Alterations of Microstructure and Sodium Homeostasis in Fast Amyotrophic Lateral Sclerosis Progressors: A Brain DTI and Sodium MRI Study

ABSTRACT


INTRODUCTION
Amyotrophic lateral sclerosis (ALS) is a relentlessly progressive neurodegenerative disorder leading to paralysis and ultimately death. As a heterogeneous condition, ALS is characterized by variable clinical presentations and progressions of symptoms depending on various factors such as age at disease onset, the site of onset, genetic factors, and the presence of non-motor symptoms, especially cognitive impairment. 1-4 ALS outcome drastically varies, with a median survival time from onset ranging from 24 months (North Europe) to 48 months (South Asia). 5 For 1.1% of ALS cases, the median survival time from onset is 18 months and can go up to 10 years in 5 to 13.3% of cases, demonstrating the heterogeneity of the disease. 3,6 While disability is commonly scored by the ALS functional rating scale (ALSFRS-R), the ALSFRS disease progression rate is also considered as an important marker of the disease to predict disability progression and patient survival. [7][8][9] Conventional MRI (e.g. T2 * , fluid-attenuated inversion recovery and proton density-weighted imaging) lacks sensitivity and specificity to detect abnormalities in ALS and is mainly used to exclude ALS-mimics. 10 Although conventional MRI could detect abnormal signal in ALS such as hyperintensity in the white matter along the corticospinal tract (CST), they are rare, nonspecific, and their exploration is not recommended for diagnosis. [10][11][12][13] In contrast, non-conventional MRI has gradually characterized features of neurodegeneration in ALS. 14 To a large extent, diffusion tensor imaging (DTI) studies have reported microstructure alterations in upper-motor neurons and extra-motor white matter tracts. 15 Notably, DTI has shown increased burdens of white matter pathology, concordant with neuropathological staging and correlating with disease aggressiveness. 14, [16][17][18] Recently, a sodium MRI study provided the first evidence of increased total sodium concentration (TSC) located in the CST of ALS patients, reflecting sodium homeostasis disturbance involved in metabolic failure contributing to the neurodegenerative process. 19 Mitochondrial dysfunction can mediate cell death by reducing ATP production and impairing sodium and calcium homeostasis. If ATP availability becomes insufficient to allow ion pumps to maintain the appropriate ion gradients, changes in electrical properties and excitability of motor neurons occur. Thus, investigating sodium concentration disturbances with sodium MRI could provide relevant functional information on neuron energetic status and cell viability while DTI explores efficiently microstructural disorders. The combination of sodium and diffusion imaging could therefore enable the exploration of complementary processes leading to neuronal injury. Besides, one may assume that brain regions presenting combined sodium homeostasis and microstructural alterations depends on disease aggressiveness. The present study aimed at investigating the topography of brain regions showing combined microstructural and sodium homeostasis alterations in ALS subgroups according to their disease progression rates.

Ethics and Institutional Review Board Approval
This prospective study was approved by the local ethics committee, and written informed consent was obtained from all participants.

Study Participants and study procedures
Twenty-nine patients with ALS (9 females, mean age ± SD: 54 ± 10 years old, mean disease duration ± SD: 1.6 ± 1.2 years) were recruited from the ALS reference center of our university hospital and 24 age and sexmatched healthy controls (HC) with no history of neurologic or neuropsychiatric disorder (11 females, age 51 ± 11 years old). The inclusion criteria were a diagnosis of ALS according to the revised El Escorial criteria. 20 The exclusion criteria were no current or past history of neurologic disease other than ALS, and no frontotemporal dementia, respiratory insufficiency, or substantial bulbar impairment incompatible with an MRI examination. Patients were clinically assessed immediately after the MRI and scored on the revised ALS Functional Rating Scale (ALSFRS-R). 21 Patients were clinically differentiated into fast and slow progressors according to their ALFSRS-R rate of progression, defined as ([48 -ALSFRS-R] / disease duration). A threshold of 0.5 ALSFRS-R per month was set to differentiate fast from slow progressors. 22

Data processing
Anatomical. T1w images were normalized to the Montreal National Institute 152 template (MNI152) using

SyN-ANTS non-linear registration. 25
Diffusion tensor imaging. Diffusion images were denoised using a Local Principal Component Analysis method that reduces signal fluctuations solely rooted in thermal noise. 26 Images were further corrected for eddy currents and head motion using affine registration to the associated non-diffusion-weighted images. 27 Fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD) maps were computed by fitting a tensor model. 27 FA images were aligned to the FMRIB58_FA target which is in MNI152 standard space using a non-linear registration. 27 Aligned FA images were averaged, then a "thinning" (nonmaximum-suppression perpendicular to the local tract structure) was applied to create a skeletonized mean FA image. The resulting image was thresholded (FA = 0.2) to suppress areas of low mean FA and/or high intersubject variability. 28 For each subject's FA image, the maximum FA value perpendicular to each voxel of the skeleton was projected onto the mean FA skeleton. Similarly, skeletonized MD, AD and RD images were generated in the MNI152 space using TBSS-FSL tools, prior to voxel-wise analysis.
Sodium imaging. Sodium images were reconstructed offline, denoised, and then normalized relative to the reference tube signals to compute quantitative TSC maps of the whole brain. 19,24 TSC maps were rigidly aligned to their corresponding T1w image. Linear and non-linear transformations were concatenated then used to bring TSC maps into the MNI152 standard space and spatially normalized TSC maps were smoothed with a Gaussian kernel (8 × 8 × 8 mm) prior to voxel-wise analysis.

Statistical analysis
Statistical analysis was performed using FSL (FMRIB Software Library v6.0) 27 and Statistical Package for Social Science (SPSS v23, IBM corp.).
Group comparisons. Differences for age, disease duration (DD) and the ALSFRS-R score between groups were assessed using Student's t-test or the Kruskal-Wallis test, when applicable. Differences for gender between groups were assessed using the Chi-squared test.
Voxel-wise analysis. Differences in diffusion (FA, MD, RD, AD) and sodium (TSC) maps between groups (ALS patients vs HC; fast vs HC; slow vs HC; fast vs slow) were assessed using permutation inference statistics (5000 permutations), combined with t-testing. Threshold-free cluster enhancement (TFCE) with a significance interval of p-values < 0.05 was used to correct for multiple comparisons (i.e. family-wise error correction). 28 Common regions with significant group differences in both diffusion and TSC maps were identified from the Johns Hopkins University WM tractography atlas and labels, and the Harvard-Oxford structural atlas and sorted by overlap with the corresponding tracts, cortical and subcortical regions.

RESULTS
Demographical and clinical measures of our population are reported in Table 1. Figure 1 shows an example of FA and TSC images in a HC, fast and slow progressor patients. There were no significant differences in age, nor in gender between ALS, fast and slow progressor patients and HC (all p-values < 0.05). There was no significant difference in disease duration, or in ALSFRS-R between fast and slow progressors. Disease progression rate was 1.54 ± 0.93 (mean ± SD) ALSFRS-R per month for fast progressors and 0.27 ± 0.09 (mean ± SD) ALSFRS-R per month for slow progressors patients.

ALS vs HC
Statistical maps resulting from voxel-wise analysis and tract-based spatial statistics comparing ALS patients to HC for TSC, FA and MD are presented in Figure 2.
TSC. ALS patients showed significantly higher TSC compared to HC mainly at the level of the body and genu of the corpus callosum, CSTs, bilateral corona radiata and thalamic radiation for white matter, and middle frontal, precentral, postcentral and cingulate gyri and anterior division for grey matter. These clusters had a TSC of 58.15 ± 4.54 mM (mean ± SD) in ALS patients and 53.41 ± 3.22 in HC. No clusters of significantly lower TSC in ALS compared to HC were found.
DTI. ALS patients showed significantly lower FA compared to HC mainly at the level of the bilateral corona radiata, body of the corpus callosum, forceps minor, genu of corpus callosum and CSTs. ALS patients showed significantly higher MD compared to HC mainly at the level of the bilateral corona radiata, body of the corpus callosum, CSTs, internal and external capsule and longitudinal fasciculus. No clusters of significantly higher FA or lower MD in ALS compared to HC were found. Figure 2, compared to HC, ALS patients showed significantly higher TSC and lower FA, higher TSC and higher MD, mainly at the level of the corpus callosum, CSTs and bilateral corona radiata. No clusters of significantly higher FA and lower TSC or lower TSC and lower MD in ALS compared to HC were found. A complete list of the significant clusters emerging from the voxel-wise analysis is reported in Supplementary Table 1.

Fast versus HC
Statistical maps resulting from voxel-wise analysis comparing fast ALS progressors to HC for TSC, FA and MD are presented in Figure 3.

TSC.
Fast progressors showed significantly higher TSC compared to HC mainly at the level of the body and genu of the corpus callosum, thalamic radiation, bilateral corona radiata, forceps minor and CSTs for white matter, and precentral, postcentral, cingulate, precingulate, middle frontal, superior frontal gyri, thalamus and caudate for grey and deep grey matter ( Figure 3). These clusters had a TSC of 59.23 ± 5.03mM (mean ± SD) in fast progressors and 53.12 ± 3.12 in HC. No clusters of significantly lower TSC in fast progressors compared to HC were found.

DTI.
Fast progressors showed significantly lower FA compared to HC mainly at the level of the bilateral corona radiata, body and genu of the corpus callosum, forceps minor, external capsule, uncinate fasciculus and CSTs ( Figure 3). Fast progressors showed significantly higher MD compared to HC mainly at the level of the bilateral corona radiata, body and genu of the corpus callosum, forceps minor, CSTs, internal capsule, longitudinal fasciculus, fronto-occipital fasciculus, thalamic radiation and external capsule ( Figure 3). No clusters of significantly higher FA or lower MD in fast progressors compared to HC were found. Figure 3, compared to HC, fast progressors showed significantly higher TSC and lower FA, higher TSC and higher MD, mainly at the level of the corona radiata, body and genu of the corpus callosum, forceps minor and CSTs. No clusters of significantly higher FA and lower TSC or lower TSC and lower MD in fast progressors compared to HC were found. A complete list of the significant clusters emerging from the voxel-wise analysis is reported in Supplementary Table 1.

Slow versus HC
Only FA showed significantly lower values in slow progressors compared to HC mainly at the level of the bilateral superior corona radiata, CSTs, body of the corpus callosum and thalamic radiation ( Figure 4). A complete list of the significant clusters emerging from TBSS analysis is reported in Supplementary Table 1.

Fast versus slow
No significant differences in TSC and DTI metrics were found between fast and slow progressors.
Results from TBSS analysis for RD and AD are reported in Supplementary Table 1

DISCUSSION
The present study highlighted brain regions with combined microstructural and sodium homeostasis disturbances corresponding to clinically relevant regions involved in ALS, namely, the CST and the corpus callosum. 29 We opted for a whole brain voxel-to-voxel analyses to highlight the focal tissue involvement that would be masked by a global approach such as regions of interest. Our results are in accordance with DTI studies that confirmed the impairment of the CST (subcortical to brainstem) as a main hallmark in ALS, even in patients with no upper motor neuron signs at the time of MRI but who developed pyramidal symptoms later. 16,30,31 Furthermore, callosal impairment has also been stressed by several studies, especially the motorrelated regions of the corpus callosum. 10,32 A recent meta-analysis DTI study analyzing 14 studies with 396 ALS patients exhibited two clusters of brain microstructural impairment. 33 The first cluster was located in the left corona radiata, extending to the body and splenium of the corpus callosum, left superior longitudinal fasciculus, posterior limb of the internal capsule, right corona radiata, and bilateral cingulate gyrus. The second cluster was located in the right corticospinal tract that extended to the right cerebral peduncle. Interestingly, these two clusters were found in our study to be the site of microstructural impairment but also sodium homeostasis disturbances, a marker of neurodegeneration related to mitochondrial dysfunction and energy failure. [19][20][21][22][23][24][25][26][27][28][29][30][31][32][33][34] Considering that heterogeneous disease progression rates impact prognosis and might affect the responsiveness to future treatments, recent efforts have tried to study patient stratification. 35,36 Stratifying patients by disease progression, we characterized widespread combined microstructural and ionic alterations in fast progressors while slow progressors only showed restricted microstructure damage. These results are of importance as they reflect diverse pathophysiological processes in patients with no difference in age or disease duration neither disability scale (ALSFRS-R), but who experienced different disease progression rates. Few studies have investigated white matter and grey matter alterations in fast and slow progressors. 16,17,30 A DTI study reported that lower motor neuron ALS fast progressors had a substantial impairment in the CST, frontal and prefrontal brain regions compared to HC, while slow progressors showed less severe alterations. 17 In addition, high disease aggressiveness patients showed a distinct pattern of supratentorial white matter density decreases relative to those with low aggressiveness but no significant differences in grey matter density suggesting axonal loss. 30 In our study, we found sodium alterations which reflect mitochondrial dysfunction and subsequent energy failure, both of which are key factors in the induction of pathological processes in ALS. 37,38 In vitro experiments demonstrated that axonal degeneration caused by experimental anoxia within the brain is a Ca2 + dependent process that can be triggered by a sustained Na+ influx driving reverse Na + -Ca2 + exchange and importing damaging levels of Ca2+ within the axons. 39 This early work suggested that ATP depletion and consequent Na + -K + -ATPase failure might result in breakdown of ionic gradients because Na + ions enter the axon via persistently activated Na+ channels. An additional study reported that axons may degenerate because nitric oxide (NO) can inhibit mitochondrial respiration resulting in energy failure and intra-axonal accumulation of sodium. Interestingly, axons could be protected from NO-mediated damage using Na + channel blockers. 40

Limitations
The cross-sectional design of the study did not allow us to assess the course of the disease between fast and slow progressors and investigate whether fast progressors are an ALS phenotype, as suggested in some studies, 16,17 or a mal-adaptative condition. In the present study, fast and slow progressors were differentiated using a threshold of 0.47 for disease progression rate. This choice was based on the results of a previous study, 22 which found that this threshold was a significant predictor of survival in ALS. Nevertheless, as no consensus is available, this choice may be open to discussion. Another limitation is related to the restricted number of patients which prevented better staging between subgroups and might explain the lack of significant difference between fast and slow progressors. Finally, a neuropsychological assessment would have helped to explain if clusters found in the frontotemporal lobe of fast progressors were due to cognitive deficits. Future multi-centric and longitudinal imaging studies will be of interest to identify early markers of neurodegeneration and predict the course of the disease of individual patients. 41

Conclusion
The present brain DTI and sodium MRI study evidenced combined microstructural and sodium homeostasis alterations in ALS. These alterations were in accordance with disease aggressiveness. Fast progressors showed a more widespread brain tissues damage than slow progressors when compared to healthy controls. Our study highlights the pertinence of a multinuclear MRI approach to stratify patients according to their disease aggressiveness.