Cervical Cord Neurodegeneration in Traumatic and Non-Traumatic Spinal Cord Injury

This study aimed to compare macrostructural and microstructural neurodegenerative changes remote from a cervical spinal cord injury in traumatic spinal cord injury (tSCI) and degenerative cervical myelopathy (DCM) patients using quantitative magnetic resonance imaging (MRI). Twenty-nine tSCI patients, 20 mild/moderate DCM patients, and 22 healthy controls underwent a high-resolution MRI protocol at the cervical cord (C2/C3). High-resolution T2*-weighted and diffusion-weighted scans provided data to calculate tissue-specific cross-sectional areas of the spinal cord and tract-specific diffusion indices of cord white matter, respectively. Regression analysis determined associations between neurodegeneration and clinical impairment. tSCI patients showed more impairment in upper limb strength and manual dexterity when compared with DCM patients. While macrostructural MRI measures revealed a similar extent of remote cord atrophy at cervical level, microstructural measures (diffusion indices) were able to distinguish more pronounced tract-specific neurodegeneration in tSCI patients when compared with DCM patients. Tract-specific neurodegeneration was associated with upper limb impairment. Despite clinical differences between severely impaired tSCI compared with mildly affected DCM patient, extensive cord atrophy is present remotely from the focal spinal cord injury. Diffusion indices revealed greater tract-specific alterations in tSCI patients. Therefore, diffusion indices are more sensitive than macrostructural MRI measures as these are able to distinguish between traumatic and non-traumatic spinal cord injury. Neuroimaging biomarkers of cervical cord integrity hold potential as predictors of recovery and might be suitable biomarkers for interventional trials both in traumatic and non-traumatic SCI.


Introduction
T raumatic spinal cord injury (tSCI) and non-traumatic degenerative cervical myelopathy (DCM) are conditions that arise from focal cervical damage. 1,2 The most obvious difference between a traumatic and non-traumatic cervical myelopathy lies in the time profile of neural changes (acute onset in tSCI vs. slowly developing symptoms in DCM). [3][4][5][6] Due to progressive impairment of gait and the increasing risk of falls, DCM patients can develop a central cord syndrome, which per definition is a tSCI. 7 Experimental evidence suggests that tSCI and DCM share several aspects of myelopathy with a combination of alpha-motoneuron damage (lesion of the central gray), 8 demyelination, [9][10][11][12] and axonal damage of long projecting spinal nerve fiber tracts (white matter damage), 13,14 as well as edema and ischemic changes. 11,15 Both etiologies present with varying degrees of upper limb im-pairment [16][17][18] that can be assessed by comprehensive clinical protocols sensitive to sensorimotor functions (e.g., Graded Redefined Assessment of Strength, Sensibility, and Prehension [GRASSP]). 19,20 Although such advanced clinical assessment allows quantifying the degree of impairment, it cannot disclose the underlying pathophysiology that occurs at the microstructural level.
Quantitative magnetic resonance imaging (qMRI) shows potential to detect such specific (micro-) structural changes in the spinal cord, both in tSCI [21][22][23][24] and in DCM patients. [25][26][27] To compare the magnitude of injury-induced neurodegenerative changes in both etiologies, we applied high-resolution T2*-weighted MRI and diffusion tensor imaging (DTI) above the injury level. We hypothesized that tSCI patients should show a more pronounced pattern of neurodegenerative changes compared with DCM patients, in which the disease slowly develops over time.

Participants and study design
Patients with tSCI at cervical level (n = 29; American Spinal Injury Association Impairment Scale [AIS] A-D; mean age 47.4standard deviation [SD] 19.8 years, five female) and patients with mild and moderate DCM (n = 20, AIS D; mean age 52.0 -14.5 years; six female) were enrolled in this study at the University Hospital Balgrist Zurich between July 2010 and July 2015. DCM patients were recruited at >1 year after onset of symptoms and tSCI patients were recruited at least 2 months after injury (Table 1). All tSCI patients underwent decompressive surgery before study enrollment while all DCM patients were in pre-operative phase.
The exclusion criteria for both groups were pregnancy, head or brain lesions associated with spinal cord injury, pre-existing neurological and medical disorders leading to functional impairments, mental disorder, or contraindications to MRI, and age <18 and >70 years.
The tSCI patients were divided in two subgroups based on the severity of impairment: AIS A&B group (i.e., motor complete) including 10 tSCI patients and AIS C&D group (i.e., motor incomplete) including 19 tSCI patients to better account for the severity of tSCI. A subset of subjects (17 tSCI patients and 20 DCM patients) included in the present study have been previously presented showing cord tissue specific changes induced by spinal cord injury in tSCI 23 or cord myelopathy in DCM. 27 Additionally, 22 healthy controls (mean age 41.1 -11.4 years, eight female) were enrolled to confirm the group difference between patients and healthy controls. 23,27 All patients underwent comprehensive clinical protocols to assess neurologic and functional impairment. These included the International Standards for Neurological Classification of Spinal Cord Injury (ISNCSCI) protocol for motor score, light-touch, and pinprick score and completeness of injury; GRASSP (maximum 232 points) as ancillary outcome measures dedicated for the assessment of upper limb function 19 and the Spinal Cord Independence Measure (SCIM). Additionally, all DCM patients were assessed using the modified Japanese Orthopedic Association (mJOA) scale (maximum 18 points). The outcome measures such as the ISNCSCI protocol for the upper extremity motor score (UEMS) (e.g. pyramidal dysfunction), light touch, pinprick, SCIM, and the GRASSP protocol were applied in both SCI and DCM patients to enable the comparison between these two etiologies.
The local ethics committee of Zurich, Kantonale Ethikkommission Zürich, approved the study (KEK-ZH-Nr. -2012-0343), and the study protocols were in accordance with the Declaration of Helsinki. Informed written consent was obtained from each subject before participation.

MRI measurements
Participants were positioned head-first supine and acquisitions were conducted on a 3T MRI system (SkyraFit Siemens Healthcare, Erlangen, Germany). Radio Frequency (RF) excitation was performed using the body coil and detection was achieved using a combination of 12-channel head-coil, four-channel neck-coil, and 24-channel spine matrix. Subjects were stabilized with an MRIcompatible stifneck (Laerdal Medicals, Stavanger, Norway) to minimize motion artefact effects. As a result of motion artefacts, four patients (three tSCI patients and one DCM patients) and one control were excluded from microstructural assessment. All participants underwent a protocol consisting of a T2*weighted three-dimensional (3D) multi-echo sequence (multiple echo data image combination; MEDIC) and a diffusion-weighted imaging (DWI) sequence based on the reduced-field of view (FOV) single-shot spin-echo echo planar imaging above the injury and stenosis level. Macrostructural cord neurodegeneration was assessed by determining gray and white matter in the cross-sectional area (SCA) of the cervical cord using T2*-weighted MRI. The T2*weighted images resulted in five high-resolution axial 3D volumes of the cervical cord with a resolution of 0.25 · 0.25 · 2.50 mm 3 within 2.8 min acquisition time per volume. MRI parameters were as follows: FOV = 162 · 192 mm 2 , matrix size = 648 · 768, repetition time (TR) = 44 msec, echo time (TE) = 19 msec, flip angle a = 11°, and read-out bandwidth = 260 Hz per pixel. To quantify microstructural changes of the spinal cord at the identical level, a high-resolution DWI scan was applied with cardiac-gating (based on finger pulse oximetry) resulting in 30 diffusion-weighted images (b = 500 sec/mm 2 ) and six b0-weighted images. The DWI sequence parameters were as follows: slice thickness = 5 mm with 10% interslice gap, 10 slices perpendicularly oriented to the spine, 5/8 Partial-Fourier Imaging in phase-encoding direction, phase oversampling = 50%, and a cardiac trigger delay = 200 msec, acquisition matrix = 176 · 40, FOV = 133 · 30 mm 2 , in-plane resolutions = 0.8 · 0.8 mm 2 , TE = 73 msec, and TR = 350 msec. The triggered DWI data were acquired in blocks of two slices per cardiac cycle. The minimal time between successive triggers was 1800 msec. Each DWI dataset was acquired with four averages resulting in 144 images within a nominal total acquisition time of 6.2 min.

Data processing
Cross-sectional spinal cord area measurement. The serial longitudinal registration in SPM12 (Wellcome Trust Centre for Neuroimaging, University College London, UK) was applied to all T2*-weighted images to average the images accounting for intra-participant motion. Jim 6.0 software (Xinapse Systems, Aldwincle, UK) was used to merge the adjacent partitions resulting in 10 contiguous slices (to increase signal to noise ratio [SNR]) and to semi-automatically segment the cross-sectional cervical cord area using an active-surface model after setting a marker in the center of the cord in each of the 10 contiguous slices. 28 The gray matter and white matter cross-sectional areas were manually segmented. The mean inter-observer and intraobserver reliability for these measures were shown to be in the range of previously reported results (less than 7%). 27,29 Diffusion tensor imaging (DTI) measurement. Processing of DWI data was carried out with a modified version of the Matlabbased ACID toolbox within SPM12 optimized for the spinal cord. 30 First, we reduced the in-plane FOV to 24 · 24 mm 2 to include only spinal cord tissue. Next, diffusion weighted images were slice-wise linearly registered with 3 degrees of freedom-like translation in the frequency-and phase-encoding direction, scaling in the phaseencoding direction to correct for intra-participant motion and eddycurrent artefacts. 31 A diffusion tensor model was fitted to the DWI data by applying a robust tensor fitting algorithm that accounts for outlier volumes due to motion and physiologic artefacts 32 and resulted fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD) DTI index maps. The DTI maps were spatially normalized to a self-constructed mean diffusivity template residing in the spinal Montreal Neurological Institute space. 33 To further refine the accuracy of the registration, a manual slice-by-slice registration (in-plane translation and scaling) was performed. Finally, all DTI index maps were smoothed with a full width at half-maximum (FWHM) Gaussian kernel with 0.5 · 0.5 · 5 mm 3 . All images were visually inspected for artefacts, and the analysis was conducted on three slices from each modality at the same level.

Statistical analysis
Statistical analysis of macrostructural MRI data, demographics and clinical outcome data was performed with Stata 15 (Stata-Corp LP, College Station, TX). The mean age was not statistically different between tSCI and DCM patients (Mann-Whitney U test: Z = 1.06, p = 0.29).
First, we assessed the morphometric differences in cord area, gray matter area, and white matter area between tSCI subgroups and DCM patients by means of analysis of covariance, adjusted for age. For assessing microstructural differences between patient's groups, we used voxel-based analysis of the different DTI indexes (FA, AD, RD) in SPM12, adjusted for age. All statistical parametric maps were initially thresholded with a cluster-defining threshold of p < 0.01 (uncorrected) and clusters surpassing a cluster threshold of p < 0.05 (family-wise error corrected) are reported. Next, we used linear regression analysis to investigate the relationship between cord macrostructural and microstructural changes and clinical outcome, adjusted for age. The level of significance was set to p < 0.05.
Data availability statement. Anonymized grouped data, study protocols, and processing pipelines will be shared by request from a qualified investigator.  (Table 1).

Cross-sectional spinal cord area
We first confirmed findings from previous reports that total cross-sectional spinal cord area, gray matter area, and white matter area are decreased in tSCI patients ( p < 0.001) and in DCM patients ( p < 0.001 ) when compared with the healthy controls. 23

Discussion
This study shows extensive cord pathology above a traumatic and non-traumatic cervical spinal cord injury. While macrostructural MRI measures revealed a similar extent of remote cord atrophy, microstructural qMRI measures were able to distinguish more pronounced tract-specific neurodegeneration in tSCI patients. The discrepancy between different clinical presentation and extensive cord pathology in tSCI and DCM patients may be suggestive of compensatory mechanisms owing to the slowly progressing disease in DCM, in contrast to the blunt and abrupt neuronal damage in tSCI patients. Our findings suggest that measures of cord atrophy are insensitive to reveal disease-specific changes while advanced qMRI measures are sensitive to the underlying disease process, as it can detect tract-specific changes.

FIG. 2.
Box plots of averaged cross-sectional spinal cord, gray matter, and white matter area in spinal cord injury (SCI) and DCM patients (A-C) Smaller spinal cord, gray matter, and white matter area is observed in severely impaired SCI with AIS A&B grade compared with those in SCI with AIS C&D grade. However, there is no significant difference in cord atrophy comparing SCI and DCM groups. HC, healthy controls; AIS, American Spinal Injury Association Impairment Scale; DCM, degenerative cervical myelopathy. (*p < 0.01; **p < 0.01; ***p < 0.001. Color image is available online.
Cervical cord atrophy remote from the injury site was previously reported in tSCI 21,23 and DCM patients 27,34 when compared with the healthy controls. Here, we also confirmed that the spinal cord atrophy above a cervical injury is different in both patient groups when compared with the healthy controls. However, the remote cord atrophy above the injury level is remarkably similar when tSCI patients are compared with mild DCM patients.
At the microstructural level, previous studies showed that neurodegenerative changes in remote cord regions are evident in tSCI 23,35 and in DCM patients. 25,27,34 In this study we show that albeit similar macrostructural cord changes, the microstructural integrity of the cord is more disturbed in tSCI when compared with DCM above the level of injury. In particular, measures of AD (indicating axonal degeneration) in the dorsal column and lateral corticospinal tract were reduced in tSCI compared with DCM. In addition, FA (indicating axonal count and myelin content) 36  whereas AD values changed differently in tSCI patients when compared with DCM. 20,24 This means that AD remote from the level of stenosis in DCM patients is increased, 27 while it is decreased in tSCI patients. Increased AD in DCM patients may partially be due to elevated fiber tract density driven by compression and loss of surrounding cord structure, 37 whereas in tSCI, decreased AD may be due to both axonal loss and demyelination. 23,38 Despite the differences in etiology, the pathophysiology underlying remote cord atrophy in both tSCI and DCM patients may be driven by similar neurodegenerative mechanisms that are revealed by DTI measures. Pre-clinical studies have highlighted that a range of common primary injury mechanisms are involved in both tSCI 4 and DCM patients, 15 which include apoptosis of cells, inflammation, and vascular changes resulting in cell death at the focal injury site. 5,15,39,40 Secondary injury-induced changes evolve over time and include anterograde and retrograde axonal degeneration of spinal pathyways, 5,14,41,42 remodulation of neuronal spinal circuits, 43 dysregulation of growth factors, 39 shrinkage of the neuron soma size 44  Interestingly, remote neurodegenerative changes (i.e., atrophy) within the cervical gray matter in both tSCI and DCM groups are associated with upper limb motor function and strength, sensibility, and prehension of the upper limbs (i.e., GRASSP). Microstructural tract-specific changes (FA) above the level of injury also were related to measures of functional independence (i.e., SCIM) and upper limb function. These correlations, though, are mostly driven by SCI patients. Our findings are in line with previous reports showing that MRI derived measures of cord macrostructure and microstructure in the cervical cord are associated with clinical impairments. 21,23,27 These clinicopathologic associations suggest that remote reorganizational changes, such as remodulation of intraspinal circuits, 43 contribute to the level of upper limb function in traumatic and non-traumatic SCI. Demonstrating a link between microstructure and function by means of DTI and advanced clinical measures of upper limb function (e.g., GRASSP) points towards the applicability of such advanced qMRI measures over conventional MRI methods in clinical routines. Thus, spinal cord DTI can complement conventional MRI, with the potential to enhance current diagnosis and more importantly, predict outcome in tSCI and DCM patients. In particular, FA was found to show the strongest correlation with clinical scores (ISNCSCI scores in tSCI, mJOA and Nurick scores in DCM), where lower FA values were associated with higher impairment. Neuroimaging biomarkers sensitive to sensorimotor functions could therefore be used for the prediction of upper limb recovery and stratification for interventional trials.
This study has some limitations. DCM patients were on average 5 years older than tSCI patients. Therefore, age was considered as a covariate of no interest in all statistical analyses. Consequently, voxel-based analysis of DTI indexes in the spinal cord are still under development for the spatial normalization of the spinal cord images into a common space; an automated post-processing pipeline is a work in progress. To increase the reliability of our analysis, we therefore manually corrected the spatial normalization to the template.

Conclusion
Despite clinical differences in traumatic and non-traumatic SCI patients, cord atrophy rostral to the level of the cervical injury is similar. However, measures of cord atrophy represent an accumulation of pathophysiological changes, and as such are insensitive to reveal disease specific changes. On the contrary, advanced qMRI measures can detect tract-specific changes that are clinically eloquent. Thus, DTI of the cervical cord might be a suitable biomarker for outcome prediction and to monitor treatment effects in interventional trials in both traumatic and non-traumatic SCI.