OUP user menu

Sodium accumulation is associated with disability and a progressive course in multiple sclerosis

David Paling, Bhavana S. Solanky, Frank Riemer, Daniel J. Tozer, Claudia A. M. Wheeler-Kingshott, Raju Kapoor, Xavier Golay, David H. Miller
DOI: http://dx.doi.org/10.1093/brain/awt149 2305-2317 First published online: 25 June 2013

Summary

Neuroaxonal loss is a major substrate of irreversible disability in multiple sclerosis, however, its cause is not understood. In multiple sclerosis there may be intracellular sodium accumulation due to neuroaxonal metabolic dysfunction, and increased extracellular sodium due to expansion of the extracellular space secondary to neuroaxonal loss. Sodium magnetic resonance imaging measures total sodium concentration in the brain, and could investigate this neuroaxonal dysfunction and loss in vivo. Sodium magnetic resonance imaging has been examined in small cohorts with relapsing-remitting multiple sclerosis, but has not been investigated in patients with a progressive course and high levels of disability. We performed sodium magnetic resonance imaging in 27 healthy control subjects, 27 patients with relapsing-remitting, 23 with secondary-progressive and 20 with primary-progressive multiple sclerosis. Cortical sodium concentrations were significantly higher in all subgroups of multiple sclerosis compared with controls, and deep grey and normal appearing white matter sodium concentrations were higher in primary and secondary-progressive multiple sclerosis. Sodium concentrations were higher in secondary-progressive compared with relapsing-remitting multiple sclerosis in cortical grey matter (41.3 ± 4.2 mM versus 38.5 ± 2.8 mM, P = 0.008), normal appearing white matter (36.1 ± 3.5 mM versus 33.6 ± 2.5 mM, P = 0.018) and deep grey matter (38.1 ± 3.1 mM versus 35.7 ± 2.4 mM, P = 0.02). Higher sodium concentrations were seen in T1 isointense (44.6 ± 7.2 mM) and T1 hypointense lesions (46.8 ± 8.3 mM) compared with normal appearing white matter (34.9 ± 3.3 mM, P < 0.001 for both comparisons). Higher sodium concentration was observed in T1 hypointense lesions in secondary-progressive (49.0 ± 7.0 mM) and primary-progressive (49.3 ± 8.0 mM) compared with relapsing-remitting multiple sclerosis (43.0 ± 8.5 mM, P = 0.029 for both comparisons). Independent association was seen of deep grey matter sodium concentration with expanded disability status score (coefficient = 0.24, P = 0.003) and timed 25 ft walk speed (coefficient = −0.24, P = 0.01), and of T1 lesion sodium concentration with the z-scores of the nine hole peg test (coefficient = −0.12, P < 0.001) and paced auditory serial addition test (coefficient = −0.081, P < 0.001). Sodium concentration is increased within lesions, normal appearing white matter and cortical and deep grey matter in multiple sclerosis, with higher concentrations seen in secondary-progressive multiple sclerosis and in patients with greater disability. Increased total sodium concentration is likely to reflect neuroaxonal pathophysiology leading to clinical progression and increased disability.

  • multiple sclerosis
  • MRI
  • disease progression
  • disability
  • Na+ channel

Introduction

Multiple sclerosis is a disabling neuroinflammatory disease of the CNS (Compston and Coles, 2008). In ∼85% of patients presentation is with a relapse (Scalfari et al., 2010), where focal lesions of lymphocytic infiltration, inflammatory demyelination, axonal transection and oedema transiently disrupt axonal conduction (Youl et al., 1991; Felts et al., 1997), followed by remission due to resolution of inflammation, remyelination, electrophysiological adaptation and neural plasticity (Craner et al., 2004a, b; Black et al., 2007; Trapp and Stys, 2009; Sumowski et al., 2010). With time the majority of patients with relapse onset multiple sclerosis develop secondary-progressive disease, where disability gradually increases independent of relapses, and the remaining 15% of patients have primary-progressive multiple sclerosis where gradual increase in disability is seen from disease outset without relapses (Kremenchutzky et al., 2006). Histopathological studies, and studies using putative imaging measures of neuroaxonal damage, have identified neuroaxonal loss as the likely major substrate of this irreversible progressive disability (De Stefano et al., 1998; Bjartmar et al., 2000; Fisniku et al., 2008; Frischer et al., 2009; Tallantyre et al., 2009, 2010). The mechanisms leading to progressive neuroaxonal loss are not well understood (Reynolds et al., 2011).

In the brain, sodium concentration is maintained at a markedly different concentration in the larger intracellular compartment (∼85% volume, ∼12.5 mM) compared with the smaller extracellular compartment (∼140 mM) by the energy dependent Na/K/ATPase pump, with total sodium concentration being a weighted average of the two (Ames, 2000; Syková and Nicholson, 2008; Fleysher et al., 2013). Experimental and histopathology studies suggest that neuroaxonal dysfunction and loss in multiple sclerosis may be caused by an inability of axons to maintain this electrochemical gradient due to mitochondrial dysfunction (Mahad et al., 2008, 2009) and electrophysiological adaptation to demyelination (Craner et al., 2004a, b). This would lead to increased total sodium concentration due to intracellular sodium accumulation (Trapp and Stys, 2009). Intracellular sodium accumulation may cause neuroaxonal dysfunction by impairing action potential propagation (Lo et al., 2003), and cause neuroaxonal death by causing reverse activity of the sodium/calcium antiporter, leading to lethal calcium influx (Lassmann et al., 2012). Neuroaxonal loss per se may also be associated with an expanded extracellular space (Périer and Gregoire, 1965), thus increasing the amount of extracellular fluid and hence total sodium concentration. The total sodium concentration in the brain could therefore provide an indication of neuroaxonal dysfunction and loss in vivo.

23Na is a MRI visible nucleus, and sodium MRI can potentially quantify the extent and regional distribution of increases in sodium concentration due to intracellular sodium accumulation and/or increased proportion of extracellular fluid in vivo. 23Na is more difficult to image than hydrogen (1H) due to lower concentration, lower gyromagnetic ratio and shorter T2 (Nielles-Vallespin et al., 2007; Ouwerkerk, 2007). However, the advent of higher field MRI systems (≥3.0 T), and advances in scanner and sequence design have facilitated development of clinically feasible scan protocols.

There have been two studies in multiple sclerosis using sodium MRI, both were of small cohorts, all of whom had relapsing-remitting disease, and most of whom had minimal disability [expanded disability status scale (EDSS) < 3]. The first study, of 17 patients, reported increased sodium concentration in lesions and normal appearing white and grey matter, with modest correlations seen between sodium concentrations and EDSS (Inglese et al., 2010). The second study, of 26 patients, found a more widespread increase in sodium concentration in normal appearing white and grey matter in patients with longer disease duration (Zaaraoui et al., 2012). Sodium imaging findings in primary and secondary-progressive multiple sclerosis, where disability is higher and neuroaxonal damage and loss more extensive, have not been reported.

We now report a study in which sodium imaging was undertaken in a large cohort of 70 patients with multiple sclerosis that included all three major clinical subgroups (relapsing-remitting, secondary, and primary-progressive), and who exhibited the typically wide range of disability that characterizes these subgroups. Our study is the first to comprehensively investigate brain sodium concentrations across the full clinical spectrum of multiple sclerosis, including many subjects with high levels of disability (43 had a progressive form of multiple sclerosis with a median EDSS of 6). The study was undertaken to explore our hypothesis that increases in tissue sodium concentration, reflecting neuroaxonal metabolic dysfunction and loss, would be most evident in patients with progressive forms of multiple sclerosis and higher levels of disability.

Materials and methods

Patients and samples

The study was approved by the research ethics committee of University College London, UK. Informed consent was obtained from participants. Patients were recruited from clinics, and control subjects were recruited from a database of volunteers interested in research. Inclusion criteria were diagnosis of clinically definite multiple sclerosis (Lublin and Reingold, 1996; Polman et al., 2011), or no neurological disease in the healthy control subjects, and age between 18 and 65 years. Exclusion criteria were presence of other neurological conditions, cardiovascular or respiratory disease, contraindication to MRI, pregnancy or breastfeeding. Subjects had medical and neurological history and examination on the day of their MRI scan. Patients with multiple sclerosis were assessed using the EDSS score (Kurtzke, 1983), and multiple sclerosis functional composite score (MSFC) (Fischer et al., 1999), consisting of a timed 25 ft walk, nine hole peg test of both hands, and 3 s paced auditory serial addition test B (PASAT). Z-scores of the nine-hole peg test and PASAT, and MSFC scores were calculated using the MFSC Administration and Scoring Manual, from National Multiple Sclerosis Society Task Force Database normative values (Fischer et al., 1999). All subjects were at least 3 months from their last clinical relapse and none was receiving steroids. Scans were performed between September 2011 and October 2012.

Magnetic resonance imaging acquisition

A 3.0 T MRI scanner (Achieva, Phillips Healthcare Systems) was used for all imaging protocols. 23Na images were obtained using a single tuned 23Na coil (RAPID biomed) using a ramp sampled radial ultra short echo time sequence, yielding an effective echo time of 0.27 ms, repetition time 120 ms, field of view 240 × 240 × 200 mm3, nominal voxel size 4 × 4 × 4 mm3, 50 axial slices, acquisition time 23 min. A phase-encoded ‘stack of stars’ with 3D volume excitation, Cartesian slice selection and 2D cylindrical stack of radial readouts were used to fill k-space (Glover and Pauly 1992; Riemer et al., 2012). Calibration phantoms of 66 mM and 33 mM sodium concentration were fixed to the subject’s head during the scan. A 1H T2-weighted dual echo scan (echo time 11/85 ms, repetition time 3875 ms, field of view 240 × 240 × 144 mm3, voxel size 1 × 1 × 4 mm3, 36 axial slices, acquisition time 9 min) was then acquired while the subject was still placed in the sodium coil, using the scanner body coil. These data were used in subsequent registration steps as detailed below.

The 23Na coil was changed for a 32 channel receive 1H coil (Phillips Healthcare Systems), and three further 1H structural scans were performed: (i) a 2D fast spin multi echo scan with parameters: echo time1 16 ms, Δecho time 16 ms, seven echoes, repetition time 5230 ms, field of view 240 × 180 × 144 mm3, voxel size 1 × 1 × 2 mm3, 72 axial slices, acquisition time 6 min 48 s. The fourth echo corresponding to a echo time of 64 ms was used for T2 hyperintense lesion identification as a T2-weighted scan; (ii) a 2D T1 spin echo scan with parameters: echo time 10 ms, repetition time 650 ms, field of view 240 × 180 × 144 mm3, voxel size 1 × 1 × 2 mm3, 72 axial slices, acquisition time 7 min 22 s. This was used for identification of T1 hypointense lesions; and (iii) a 3D T1-weighted turbo gradient echo structural scan with parameters: echo time 3.1 ms, repetition time 6.9 ms, inversion time 824 ms, field of view 256 × 256 × 180 mm3, voxel size 1 × 1 × 1 mm3, 180 sagittal slices, acquisition time 6 min 31 s. This sequence was used for tissue class segmentation.

Image post-processing and registration

Image post-processing was performed off-line on a LINUX workstation using JIM 6.0 (Xinapse systems, http://www.xinapse.com). Total sodium concentration was quantified on a voxel-by-bowel basis using a linear calibration method, where signal intensity of each voxel was scaled with signal intensity measured from calibration phantoms, divided by a correction factor of 0.98 to account for their T1 relaxation rate and experimental repetition time (in this study: T1 = 30 ms, repetition time = 120 ms) (Christensen et al., 1996; Nielles-Vallespin et al., 2007; Inglese et al., 2010).

Using the 3D T1-weighted gradient echo sequence, white matter lesions were outlined using the region of interest toolkit of JIM 6.0 and filled with simulated normal appearing white matter signal intensities to improve segmentation accuracy (Chard et al., 2010), neck slices were removed (Popescu et al., 2012), and images were skull stripped using the BET toolkit of FSL 4.1.3. These images were then segmented into white and grey matter and CSF volume fraction maps using the FAST algorithm (version 4.1 of FSL 4.1.3) and deep grey matter masks were created using the FIRST algorithm (version 1.2 of FSL 4.1.3) (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki) (Patenaude et al., 2011). The FAST algorithm has an advantage over other segmentation methods in that it outputs volume fraction maps, which were used for the subsequent CSF partial volume correction. Grey, white matter and CSF volumes were calculated using the partial volume estimate method (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FAST#Tissue_Volume_Quantification) (Klauschen et al., 2009), and fractional volumes were calculated by dividing grey or white matter volume by total intracranial volume.

Using an affine transformation within the NiftiReg toolkit (University College London, www.sourceforge.net/projects/niftyreg) (Ourselin et al., 2001, 2002) the following three image registration steps were performed: (i) sodium image to T2-weighted image obtained with the body coil; (ii) T2-weighted image obtained with the body coil to the first echo of the 2D T2-weighted image obtained with the 32 channel coil; and (iii) 2D T2-weighted image obtained with the 32 channel coil to 3D T1 gradient echo image. The matrices from these registration steps were multiplied together, and the inverse used to transform white, grey and CSF volume fraction maps to native sodium space.

Correction of the partial volume effect of cerebrospinal fluid

Because sodium concentration is higher in CSF than brain tissue (140–150 mM versus 30–40 mM; Harrington et al., 2010) small amounts of CSF partial volume may cause artefactual increase in apparent tissue sodium concentration. To correct for this effect we implemented a voxel-by-voxel modified partition-based correction method to remove the contribution of CSF sodium from sodium concentration maps in native sodium space (Soret et al., 2007). Tissue volume fraction maps were created by addition of grey matter and white matter volume fraction maps transformed to native sodium space. CSF sodium concentration was calculated on a subject-by-subject basis by using the CSF volume fraction map transformed to native sodium space to identify voxels containing 100% CSF, and calculating average sodium concentration in those voxels. Tissue sodium concentration was obtained by the equation: Embedded Image Where C = sodium concentration and VF = Volume Fraction

Only voxels with at least 20% tissue volume fraction were used, and to reject unrealistically high sodium concentrations, a threshold was set of 2 standard deviations (SD) above the mean of the sodium concentration in voxels with at least 95% tissue volume.

Lesions, normal appearing white matter, and cortical and deep grey matter mask creation

Two-dimensional T2-weighted spin echo and T1-weighted spin echo sequences obtained with the 32-channel coil were registered to the 3D T1-weighted turbo gradient echo sequence using the Niftireg toolkit. Lesions were outlined on these registered images using a semi-automated segmentation technique based upon local thresholding, utilizing the region of interest toolkit in Jim 6.0, and were classified as T1 hypointense (hyperintense on the T2-weighted scan and equal or lesser signal intensity than grey matter on the T1-weighted scan) (Van Waesberghe et al., 1999) and T1 isointense (hyperintense on the T2-weighted scan and isointense on the T1-weighted scan) lesions. Lesion volume was obtained by multiplying lesion area by slice thickness for each lesion class.

Grey and white matter volume fraction maps were binarized with lower threshold of 80% to create white and grey matter masks, and masks of caudate, putamen, pallidum and thalamus from the FAST algorithm were combined to produce a mask of the deep grey matter. Cortical grey matter mask was obtained by subtraction of deep grey matter mask from the grey matter mask, and normal appearing white matter mask was obtained by subtraction of the T2 lesion mask from the white matter mask.

Statistical analysis

SPSS 21 was used for the majority of the statistical analysis (IBM), and STATA 10 (Stata Corporation) was used for regression and bootstrapping.

Correlations between T1 and T2 lesion volume, and EDSS and MSFC were assessed using Spearman’s rank correlation coefficient. Associations between white and grey matter fractions with age and gender were assessed using linear regression. Differences between white and grey matter fractions between controls and multiple sclerosis subgroups were assessed using linear regression with age and gender covariates on a subject type indicator. Linear regression was used to investigate associations between MSFC and EDSS with white and grey matter fractions with covariates of disease duration and gender.

ANOVA for repeated measures with Bonferonni correction was used to compare sodium concentrations between cortical grey matter, white matter and deep grey matter within controls. Analysis of variance with Bonferonni correction was used to compare sodium concentrations in cortical grey matter, normal appearing white matter and deep grey matter between controls and multiple sclerosis subgroups, and sodium concentrations in T1 hypointense and isointense lesion between multiple sclerosis subgroups. This step was repeated in an ANCOVA analysis with Bonferonni correction with subject age as an additional covariate, to ensure that any differences seen were not due to age-related effects.

Univariate correlations between cortical grey matter, normal appearing white matter, deep grey matter, T1 hypointense and T2 hyperintense lesions sodium concentrations and disability scales were made using Spearman’s rank correlation coefficient for EDSS, and using Pearson’s correlation coefficient for the MSFC, 25 ft walk speed, and z-scores of the nine-hole peg test and PASAT.

Following univariate correlations we used stepwise linear regression modelling to identify sodium concentrations independently associated with EDSS, MSFC and its subscores. EDSS, MSFC, 25 ft walk speed or z-score of the PASAT or nine-hole peg test were the dependent variables. Disease duration was entered into the regression model, and the different tissue type sodium concentrations were added as independent variables to be tested in the stepwise analysis.

Since the EDSS is not normally distributed, confidence intervals and P-values obtained from regression against EDSS were confirmed using non-parametric bias-corrected and accelerated bootstrap estimates with 1000 replicates. This procedure has the advantage over standard non-parametric methods that it allows associations with EDSS to be adjusted for potential confounders in a standard regression context (Carpenter and Bithell, 2000).

A P-value of <0.05 was defined as indicating statistical significance. Unless otherwise stated results are presented as mean ± SD.

Results

Demographics

A total of 97 subjects were studied; 27 healthy control subjects, 27 patients with relapsing-remitting, 23 patients with secondary-progressive, and 20 patients with primary-progressive multiple sclerosis. Characteristics of the patient groups are shown in Table 1. Twenty-nine patients received disease-modifying therapies. In patients with relapsing-remitting multiple sclerosis, nine received beta-interferon 1a, two beta-interferon 1b, three glatiramer acetate, four natalizumab, and one fingolimod. In patients with secondary-progressive multiple sclerosis, five received beta-interferon 1a, two beta-interferon 1b, two glatiramer acetate and one patient received natalizumab. No patients with primary-progressive multiple sclerosis received disease modifying therapy.

View this table:
Table 1

Demographics and structural imaging features in healthy controls and patients with multiple sclerosis

Healthy control subjectsAll patients with multiple sclerosisRelapsing-remitting multiple sclerosisSecondary-progressive multiple sclerosisPrimary-progressive multiple sclerosis
Patient demographics
    Gender F:M16:1141:2916:1112:1113:7
    Age in years42.9 ± 11.345.8 ± 11.237.9 ± 7.950.7 ± 10.350.9 ± 10.5
    Disease duration in years14.0 ± 9.410.2 ± 6.120.5 ± 11.211.5 ± 6.9
    Median EDSS (range)5.5 (0–8.5)2.0 (0–6.0)6.0 (3.5–7.5)6.0 (3.5–8.5)
Structural imaging features
    White matter fraction36.3 ± 1.5%34.8 ± 1.6%**35.2 ± 1.6%34.3 ± 1.7%**35.1 ± 1.6%**
    Grey matter fraction40 ± 1.6%39.3 ± 1.7%39.8 ± 1.7%39.0 ± 1.7%39.2 ± 1.7%
    T2 lesion volume15.3 ± 16.5 ml8.7 ± 11.6 ml20.5 ± 14.4 ml##17.6 ± 21.6 ml
    T1 lesion volume9.8 ± 12.6 ml4.5 ± 6.7 ml13.4 ± 10.7 ml##12.3 ± 17.9 ml
  • Unless otherwise specified data shown are mean ± standard deviation.

  • **P < 0.01 = difference from healthy controls, ##P < 0.01 difference from relapsing-remitting multiple sclerosis.

Structural imaging features

Lesion volumes and tissue type fractions are shown in Table 1. T1 and T2 lesion volumes were significantly higher in patients with secondary-progressive compared with relapsing-remitting multiple sclerosis. Significant correlations were seen of T1 hypointense lesion volume with EDSS (r = 0.38, P = 0.001) and MSFC (r = −0.54, P < 0.001), and of T2 hyperintense lesion volume with EDSS (r = 0.35, P = 0.003) and MSFC (r = −0.501, P < 0.001). No association was seen between disease duration and lesion volumes.

In controls grey matter fraction was reduced with increasing age (coefficient = −0.1%, P < 0.001) (i.e. a reduction of 0.1% grey matter fraction per increase in year) and was higher in females than males (mean difference +1.4%, P = 0.008). White matter fraction was reduced with increasing age (coefficient −0.1%, P = 0.013) but there was no significant difference between males and females.

Allowing for age and gender, white matter fraction was significantly reduced in all patients with multiple sclerosis, and in secondary and primary-progressive multiple sclerosis subgroups as compared to controls. Differences between grey matter fractions were not significant, though tended to be lower in multiple sclerosis. Grey matter fraction was significantly correlated with EDSS (coefficient = −0.40, P = 0.009) and MSFC (coefficient = 1.07, P = 0.014). No significant associations were seen between white matter fraction and EDSS or MSFC.

Grey and white matter sodium concentrations in control subjects

Sodium concentrations are shown in Table 2. Sodium concentration was higher in the cortical grey matter than in the white matter (36.0 ± 2.7 mM versus 32.1 ± 2.2 mM), and was slightly higher in cortical as compared to deep grey matter (36.0 ± 2.7 mM versus 35.1 ± 2.2 mM, P < 0.001 for both comparisons). No significant association was seen between age or gender and white or grey matter sodium concentration.

View this table:
Table 2

Sodium concentrations in different tissue types in healthy controls and patients with multiple sclerosis

Healthy control subjectsAll patients with multiple sclerosisRelapsing-remitting multiple sclerosisSecondary-progressive multiple sclerosisPrimary-progressive multiple sclerosis
Cortical grey matter36.0 ± 2.7 mM40.0 ± 3.9 mM**38.5 ± 2.8 mM**41.3 ± 4.2 mM**,##40.5 ± 4.4 mM**
Normal appearing white matter32.1 ± 2.2 mM34.9 ± 3.3 mM**33.6 ± 2.5 mM36.1 ± 3.5 mM**,##35.2 ± 3.3 mM**
Deep grey matter35.1 ± 2.2 mM37.1 ± 3.1 mM**35.7 ± 2.4 mM38.1 ± 3.1 mM**,##37.8 ± 3.2 mM**
All lesions45.8 ± 7.7 mM42.1 ± 7.3 mM47.8 ± 6.1 mM48.1 ± 8.4 mM
T1 hypointense lesions46.8 ± 8.3 mM43.0 ± 8.5 mM49.0 ± 7.0 mM#49.3 ± 8.0 mM#
T1 isointense lesions44.6 ± 7.2 mM41.8 ± 6.5 mM46.4 ± 5.2 mM46.0 ± 9.2 mM
  • Data presented are mean tissue sodium concentration in mM ± standard deviation.

  • **P < 0.01 difference from healthy controls, #P < 0.05 difference from relapsing-remitting multiple sclerosis, ##P < 0.01difference from relapsing-remitting multiple sclerosis.

Grey and white matter sodium concentrations in patients with multiple sclerosis

Sodium concentration was higher in patients with multiple sclerosis as compared to controls in cortical grey matter (40.0 ± 3.9 mM versus 36.0 ± 2.7 mM), normal appearing white matter (34.9 ± 3.3 mM versus 32.1 ± 2.2 mM), and deep grey matter (37.1 ± 3.1 mM versus 35.1 ± 2.2 mM, P < 0.001 for all comparisons) (Fig. 1). When subgroups were examined with Bonferonni correction, cortical grey matter sodium concentration was significantly higher than controls in all subgroups of multiple sclerosis (Fig. 2A), and deep grey matter and normal appearing white matter sodium concentration was significantly higher than controls in primary and secondary-progressive multiple sclerosis (Fig. 2B and C). Sodium concentration was significantly higher in patients with secondary-progressive compared to relapsing-remitting multiple sclerosis in cortical grey matter (41.3 ± 4.2 mM versus 38.5 ± 2.8 mM, P = 0.008), normal appearing white matter (36.1 ± 3.5 mM versus 33.6 ± 2.5 mM, P = 0.018) and deep grey matter (38.1 ± 3.1 mM versus 35.7 mM ± 2.4 mM, P = 0.02) (Fig. 2). The differences between other multiple sclerosis subgroups were not statistically significant. The addition of age as a covariate did not alter the significance of the associations seen.

Figure 1

Raw sodium images in sodium space (top), tissue sodium maps with CSF partial volume correction (middle) and T2-weighted images (bottom) registered to the T1 volumetric scan in controls and patients with multiple sclerosis. Increased sodium is seen in the lesions in the patient with relapsing-remitting multiple sclerosis (B) and more extensive increases in sodium in lesions and normal appearing white matter is seen in more disabled patients with secondary-progressive (C) and primary-progressive multiple sclerosis (D). Note that CSF partial volume correction leads to an apparent reduction in sodium concentration in the ventricles and sulci, and due to interpolation effects during registration, tissue sodium concentration appears artefactually reduced at the cortical grey matter / CSF boundary. RRMS = relapsing-remitting multiple sclerosis; SPMS = secondary-progressive multiple sclerosis; PPMS = primary-progressive multiple sclerosis.

Figure 2

Box plots of regional sodium concentrations in cortical grey matter (A), normal appearing white matter (B) and deep grey matter (C) in controls and multiple sclerosis subgroups. RRMS = relapsing-remitting multiple sclerosis; SPMS = secondary-progressive multiple sclerosis; PPMS = primary-progressive multiple sclerosis.

Lesion sodium concentrations

Sodium concentration was significantly higher in all lesions (45.8 ± 7.7 mM), T1 hypointense (46.8 ± 8.3 mM) and T1 isointense lesions (44.6 ± 7.2 mM) compared with normal appearing white matter (34.9 ± 3.3 mM), and was significantly higher in T1 hypointense compared with T1 isointense lesions (P < 0.001 for all comparisons) (Figs 3 and 4).

Figure 3

Box plots of sodium concentrations in T1 hypointense lesions (A) and T1 isointense lesions (B) in multiple sclerosis subgroups. RRMS = relapsing-remitting multiple sclerosis; SPMS = secondary-progressive multiple sclerosis; PPMS = primary-progressive multiple sclerosis.

Figure 4

Tissue sodium concentration maps with CSF partial volume correction, T2-weighted images and T1-weighted images in two patients with multiple sclerosis. Increased sodium concentration is seen in a T1 isointense lesion (A) and higher sodium concentration is seen in a T1 hypointense lesion (B).

Sodium concentration in T1 hypointense lesions was significantly higher in patients with secondary-progressive (49.0 ± 7.0 mM) and primary-progressive (49.3 ± 8.0 mM) compared with relapsing-remitting multiple sclerosis (43.0 ± 8.5 mM, P = 0.029 for both comparisons) (Fig. 3A). A trend was seen towards an increase in sodium concentration in T1 isointense lesions in secondary-progressive compared with relapsing-remitting multiple sclerosis but this was not significant (P = 0.067) (Fig. 3B). The addition of age as a covariate did not alter the significance of the changes seen.

Univariate correlations between sodium concentrations and clinical disability scores

Univariate correlations between EDSS and MSFC scores and sodium concentrations are shown in Table 3. EDSS was significantly correlated with sodium concentration in deep grey matter (r = 0.274, P = 0.023), and with T1 isointense lesion sodium concentration (r = 0.238, P = 0.049), and MSFC was significantly correlated with all tissue type sodium concentrations, with the strongest correlation being between T1 hypointense lesion sodium concentration (r = −0.346, P = 0.004).

View this table:
Table 3

Univariate correlations between regional sodium concentrations and EDSS and MSFC scores

Embedded Image
  • Significant correlations are shown in bold, with the colour indicating the relative strength of the correlation coefficient from weaker (green) to stronger (red).

  • r = Spearman’s (for EDSS) or Pearson’s correlation coefficient (for MSFC, z 9HPT, z PASAT and 25 ft walk speed). P = two sided P-value. z 9HPT = z score of the 9 hole peg test, z PASAT = z score of the paced auditory serial addition test.

Significant correlations were seen with all tissue type sodium concentrations with the z-scores of the PASAT and nine-hole peg test, with the strongest correlation being with T1 hypointense lesion sodium concentration. Significant correlation with 25 ft walk speed was seen with deep grey matter sodium concentration only (r = −0.286, P = 0.017). Univariate correlations between EDSS and MSFC scores and sodium concentrations within multiple sclerosis subgroups are shown in Supplementary Table 1.

Multivariate analysis of regional sodium concentrations with clinical disability scores

Stepwise linear regression analysis was performed to determine independent associations between EDSS and MSFC scores, and sodium concentrations in cortical and deep grey matter, normal appearing white matter, and T1 hypointense and isointense lesions (Table 4). Independent association was seen between deep grey matter sodium concentration and EDSS (coefficient = 0.24, P = 0.003) and 25 ft walk speed (coefficient = −0.24, P = 0.01). Independent association was also seen between T1 lesion sodium concentration and MSFC (coefficient = −0.35, P = 0.003), z-score of the nine hole peg test (coefficient = −0.12, P < 0.001) and z-score of the PASAT (coefficient = −0.081, P < 0.001).

View this table:
Table 4

Outcome of the stepwise multivariate regression between clinical rating scores and all sodium concentrations for all patients with multiple sclerosis

Clinical rating scoreRegional sodium concentrationCoefficientP-value
EDSSDeep grey matter0.240.003
MSFCT1 hypointense lesions−0.350.003
z 9HPTT1 hypointense lesions−0.12<0.001
z PASATT1 hypointense lesions−0.081<0.001
25 ft walk speedDeep grey matter−0.240.01
  • All independently significant associations are shown.

  • z 9HPT = z score of the 9 hole peg test; z PASAT = z score of the 3 s paced auditory serial addition test B.

Partial volume correction

An upper threshold exclusion was applied during CSF partial volume correction to remove voxels with sodium concentrations >2 SD above the mean concentration of voxels containing at least 95% tissue volume. In total 1.64% of total voxels were excluded from masks, with a greater proportion of voxels excluded from the cortical grey matter (3.04%) than white matter (1.1%) or deep grey matter masks (0.82%, P < 0.001 for both comparisons). No significant difference was seen between the proportion of voxels excluded between patients and controls in cortical grey matter (3.01% versus 3.11%, P = 0.8), white matter (1.23% versus 1.04%, P = 0.6), or deep grey matter (0.77% versus 0.94%, P = 0.9).

Discussion

There are four main findings of our study. First, it shows that significant and widespread increases in sodium concentration are seen in primary and secondary-progressive multiple sclerosis, affecting the cortical and deep grey matter, normal appearing white matter and lesions. Second, sodium concentrations in cortical and deep grey matter, normal appearing white matter and T1 hypointense lesions are significantly higher in secondary-progressive compared with relapsing-remitting multiple sclerosis. Third, previous findings of an increase in sodium in lesions and cortical grey matter in relapsing-remitting multiple sclerosis are confirmed (Inglese et al., 2010; Zaaraoui et al., 2012). Finally, there are significant correlations of sodium concentration with clinical measures of disability and impairment: specifically, increases in EDSS and slower walking speed were independently associated with increase in deep grey matter sodium concentration, and a worse score on measures of cognition (PASAT) and upper limb function (nine-hole peg test) were independently associated with increased sodium concentration within T1 hypointense lesions.

Potential mechanisms for increase sodium concentration in multiple sclerosis—increased extracellular volume

As sodium concentration is a weighted average of intra- and extracellular sodium concentrations, increase in the volume of the extracellular fluid, which has higher sodium concentration, will increase total sodium concentration.

Lesions occur within the cortex in progressive multiple sclerosis (Kutzelnigg et al., 2005; Reynolds et al., 2011; Choi et al., 2012), which can be extensive (et al., 2003). Despite their extent, these lesions are poorly visualized on the T1 and T2-weighted MRI sequences used in this study (Geurts et al., 2005), so would have been included in the cortical grey matter masks. Considerable neuroaxonal loss is seen in these lesions in patients with progressive disease (Peterson et al., 2001; Vercellino et al., 2005), and this could lead to increased extracellular fluid volume, and may account for the increased cortical sodium concentrations seen. The greater extent of cortical lesions, and greater amount of neuroaxonal loss in lesions in secondary-progressive compared with relapsing-remitting multiple sclerosis, (Lucchinetti et al., 2011) may also account for the higher cortical sodium concentrations seen in secondary-progressive compared with relapsing-remitting multiple sclerosis.

Increased sodium concentration seen in normal appearing white matter and deep grey matter in progressive multiple sclerosis may also be due to neuroaxonal loss. Neuroaxonal loss in normal appearing white matter is more extensive in progressive as compared to relapsing-remitting multiple sclerosis (Kutzelnigg et al., 2005; Frischer et al., 2009; Tallantyre et al., 2010), which can also be inferred in our study from the reduction in white matter volume seen in progressive multiple sclerosis. Neuroaxonal loss is also seen in deep grey matter (Evangelou et al., 2000; Cifelli et al., 2002; Geurts et al., 2006, Vercellino et al., 2009; Schoonheim et al., 2012; Shiee et al., 2012), and associations seen between thalamic neuronal loss and diffuse abnormalities in relevant projections through the normal appearing white matter suggest that it may occur secondary to trans-synaptic degeneration from distal axonal damage (Geurts et al., 2006; Kolasinski et al., 2012). Normal appearing white matter neuroaxonal loss in progressive multiple sclerosis may lead to increased extracellular fluid and increased sodium concentration in normal appearing white matter, and may also lead to neuroaxonal loss and increased sodium concentration in deep grey matter through trans-synaptic degeneration.

Finally, in lesions the increased sodium concentration is likely to be caused, at least in part, by expansion in the extracellular space secondary to neuroaxonal loss (Périer and Grégoire, 1965; Van Waesberghe et al., 1999; Bitsch et al., 2001; Schmierer et al., 2004; Moll et al., 2011). Greater neuroaxonal loss in T1 hypointense lesions (Van Walderveen et al., 1998; Bitsch et al., 2001; Schmierer et al., 2004), particularly as seen in primary and secondary-progressive multiple sclerosis (Scanderbeg et al., 2000; Tallantyre et al., 2009), could also account for the higher sodium concentrations seen in these lesion and patient types.

Potential mechanisms for increased sodium concentration in multiple sclerosis—increased intracellular sodium concentration

Sodium is maintained at a lower concentration in the intracellular space by the energy dependent Na/K/ATPase (Ames, 2000). Mitochondrial neuronal dysfunction, as seen in multiple sclerosis, may lead to inadequate ATP to power Na/K/ATPase (Lu et al., 2000; Dutta et al., 2006; Mahad et al., 2008, 2009; Aboul-Enein et al., 2010; Campbell et al., 2011; Nikić et al., 2011; Fischer et al., 2012). This may be compounded by upregulation and spread of voltage-gated sodium channels to areas of demyelinated axolemma (Craner et al., 2004a, b; Black et al., 2007), allowing restoration of conduction across demyelinated sections, at the expense of greater sodium influx per action potential transmitted (Waxman, 2008). A combination of increased influx and reduced mitochondrial energy to export sodium could lead to intracellular neuroaxonal sodium accumulation (Trapp and Stys, 2009).

Although cortical lesions are seen in patients with relapsing-remitting multiple sclerosis, in contrast to progressive multiple sclerosis, neuroaxonal loss occurs rarely, but demyelination is seen (Lucchinetti et al., 2011). Upregulation and redistribution of sodium channels in demyelinated lesions, coupled with mitochondrial dysfunction as reported in the cortex (Campbell et al., 2011) may lead to intracellular sodium accumulation and consequent increase in cortical sodium concentration. The visualization of swollen axons within cortical lesions provides additional indirect evidence of this intra-axonal sodium accumulation (Lucchinetti et al., 2011).

In normal appearing white matter, although neuroaxonal loss and consequent increase in extracellular fluid volume may be a significant cause of the increase in sodium concentration seen in progressive multiple sclerosis, we found significant univariate correlations between MSFC and normal appearing white matter sodium concentration, but not with white matter fraction. This suggests normal appearing white matter sodium may be sensitive to an aspect of pathophysiology related to clinical function other than tissue loss per se. Mitochondrial dysfunction is reported in normal appearing white matter (Dutta et al., 2006; Aboul-Enein et al., 2010) and could lead to intracellular sodium accumulation and consequent neuroaxonal dysfunction (Lo et al., 2003).

Finally, putative estimation of the of extracellular fluid volume within lesions suggests that lesional sodium concentration is greater than would be predicted by increases in extracellular volume alone, and may additionally reflect intracellular sodium accumulation (see Supplementary material for full details of the method and limitations). The combination of sodium channel upregulation and redistribution in demyelinated lesions (Craner et al., 2004a, b; Black et al., 2007), combined with mitochondrial dysfunction as seen within lesions (Lu et al., 2000; Mahad et al., 2008) could lead to intracellular sodium accumulation.

Intracellular sodium accumulation could lead to reverse activity of the sodium/calcium antiporter, leading to intracellular calcium import, activation of calcium dependent proteases, and neuroaxonal death and loss (Waxman 2008; Nikić et al., 2011; Lassmann et al., 2012). This pathophysiological process may not be inevitable, however, and spontaneously reverses in a number of neurons under physiological conditions (Nikić et al., 2011). It may also be a target for therapeutic intervention, with studies in experimental models of multiple sclerosis showing axonal survival is enhanced by protection of mitochondria from oxidative injury by scavengers of reactive oxygen species (Nikić et al., 2011), and by blockade of voltage-gated sodium channels (Kapoor et al., 2003; Lo et al., 2003) or acid sensing ion channels (Vergo et al., 2011). Sodium MRI may be a potential tool for detecting a neuroprotective effect of these and similar experimental therapies in multiple sclerosis in vivo.

Associations between tissue sodium concentrations and clinical impairment and disability

We found significant independent association of EDSS and ambulation with deep grey matter sodium concentration, and associations with EDSS have also been reported between deep grey matter atrophy (Audoin et al., 2006; Rocca et al., 2010) and microstructural damage (Mesaros et al., 2011). The deep grey matter is an area of extensive connectivity, which transmits information between subcortical and cortical areas (Kreitzer and Malenka, 2008), and the associative processing that this connectivity facilitates may be particularly important for the integrative control of the volitional, sensory and subcortical inputs required for successful ambulation (Takakusaki et al., 2008). Increased sodium concentration in deep grey matter may reflect neuroaxonal dysfunction or loss in the deep grey matter, which could directly disrupt this processing, or could be a marker of trans-synaptic degeneration from axonal damage in clinically eloquent white matter tracts (Kolasinski et al., 2012).

Increased T1 hypointense lesion sodium concentration was significantly associated with worse performance in the MSFC, PASAT and nine-hole peg test scores, which could suggest that the extent of neuroaxonal dysfunction or loss within lesions is associated with clinical impairment, a finding that would concur with histopathological studies (Bjartmar et al., 2000; Tallantyre et al., 2009; Schirmer et al., 2011). Performance in the PASAT and nine-hole peg test requires co-ordination of neural activity across cortical and cerebellar regions through white matter tracts (Jasperse et al., 2007; Forn et al., 2012), and a greater extent of neuroaxonal dysfunction or loss in lesions within these tracts—as expected in T1 hypointense lesions with the highest sodium concentration—could impair these tasks by disrupting this connectivity (Lowe et al., 2006).

Structural magnetic resonance imaging measures

We found significant association between worse performance in the EDSS and MSFC scores with reduction in grey matter fraction, which accords with previous studies (Fisniku et al., 2008). Whilst a trend towards lower grey matter fraction in multiple sclerosis was seen, this was not significant, whereas other studies have shown significant reduction in grey matter fraction in multiple sclerosis (De Stefano et al., 2003; Fisniku et al., 2008). The lack of significance in our study may be a function of the relatively large standard deviation with respect to mean difference seen. This may reflect the range of disabilities sampled in the patient cohorts, and a potentially greater variability in grey matter volume estimates seen using the FSL as compared to the SPM software package, which was used in the other studies (Klauschen et al., 2009).

Limitations and future directions

There are several limitations of our study that could be addressed in future research. Mean age was higher in patients with primary and secondary-progressive multiple sclerosis compared to relapsing-remitting multiple sclerosis and controls. Although this difference might have biased our findings, there was no association between age and tissue sodium concentration in the control group with ages ranging from 22 to 65 years, and the addition of age as a covariate in the statistical analysis did not alter the significance of differences seen between sodium concentrations in controls and multiple sclerosis clinical subgroups. For these reasons, we think it unlikely the differences in age of the multiple sclerosis subgroups affected our results, but further studies could include exactly age-matched controls to confirm this.

A limitation of sodium MRI is that due to inherently low signal, voxel sizes need to be larger, thereby potentially inducing partial volume effects. The main concern in this regard is inclusion of voxels that partially contain CSF, where there is a much higher sodium concentration than in brain tissues. We reduced the impact of this on our results by using a voxel-by-voxel modified partition-based correction. As part of this method a threshold is set to reject aberrant voxels, which could potentially bias our results. A greater proportion of voxels was excluded from the cortical grey matter than white or deep grey matter masks, which is likely due to the complex architecture and adjacency to CSF of the cortex, which causes more partial volume effects. Importantly, there were no significant differences in the proportion of voxels excluded between patients and controls for each tissue class, making it improbable that these corrections accounted for differences seen in sodium concentrations between patients and controls. No correction was made for partial volume effects from different tissue types (i.e. white versus grey matter), but these differences are much smaller than that between brain tissues and CSF and unlikely to have affected our study findings. Future studies at higher field strength may be able to use the additional signal to reduce voxel size, and consequently further reduce partial volume effects (Fleysher et al., 2013).

In addition to altering concentrations of sodium, multiple sclerosis pathology may also affect its relaxivity. Although this would be unlikely to affect our results since we used an ultra-short echo time sequence, MRI techniques able to quantify sodium relaxivity have been described (Madelin et al., 2012), and their use in future studies may give additional information on multiple sclerosis pathophysiology.

Sodium MRI studies in multiple sclerosis have not been able to directly distinguish between intracellular sodium accumulation and increase in extracellular volume. Future studies using triple quantum filtering techniques thought to isolate the signal from intracellular sodium (Fleysher et al., 2013), or combination studies with other MRI modalities sensitive to neuroaxonal loss (Paling et al., 2011), may enable the relative contribution of intracellular and extracellular sodium concentrations to be directly ascertained. It would be of particular interest to elucidate to what extent increases in tissue sodium concentration seen in relapsing-remitting multiple sclerosis were due to intracellular sodium accumulation, as this may represent neuroaxonal metabolic dysfunction rather than loss, and thus indicate a window for neuroprotective intervention, and a potentially responsive method for detecting such effect.

Finally, although we found significantly higher regional sodium concentrations in association with disability and a progressive disease course, our study had a cross sectional design. Further longitudinal studies are needed to elucidate the evolution of sodium accumulation at different stages of the disease (Filippi et al., 2012) and could clarify whether tissue sodium concentrations help predict the future clinical course of multiple sclerosis and potentially identify patients most likely to benefit from therapeutic neuroprotective interventions before the onset of irreversible tissue damage.

Funding

This work was funded by the National Institute for Health Research University College London Hospitals Biomedical Research Centre [grant code 6DFB].

Supplementary material

Supplementary material is available at Brain online.

Acknowledgements

The NMR unit where this work was performed is supported by grants from the Multiple Sclerosis Society of Great Britain and Northern Ireland, Philips Healthcare, and supported by the National Institute for Health Research University College London Hospitals Biomedical Research Centre. The authors would also like to thank Dr Daniel Altmann for statistical advice and support, Dr Felicity Kay for help with proof reading the manuscript, and all the patients and controls studied for their help.

Abbreviations
EDSS
Expanded Disability Status Score
MSFC
Multiple Sclerosis Functional Composite score
PASAT
Paced Auditory Serial Addition Test

References

View Abstract