556301
research-article2014
MSJ0010.1177/1352458514556301Multiple Sclerosis JournalC Wang, D Paling et al.
MULTIPLE SCLEROSIS MSJ JOURNAL
Original Research Paper
Axonal conduction in multiple sclerosis: A combined magnetic resonance imaging and electrophysiological study of the medial longitudinal fasciculus
Multiple Sclerosis Journal 2015, Vol. 21(7) 905–915 DOI: 10.1177/ 1352458514556301 © The Author(s), 2014. Reprints and permissions: http://www.sagepub.co.uk/ journalsPermissions.nav
Chenyu Wang, David Paling, Luke Chen, Sean N Hatton, Jim Lagopoulos, Swee T Aw, Matthew C Kiernan and Michael H Barnett
Abstract Objective: The objective of this paper is to inform the pathophysiology of medial longitudinal fasciculus (MLF) axonal dysfunction in patients with internuclear ophthalmoplegia (INO) due to multiple sclerosis (MS), and develop a composite structural-functional biomarker of axonal and myelin integrity in this tract. Methods: Eighteen patients with definite MS and clinically suspected INO underwent electrical vestibular stimulation and search-coil eye movement recording. Components of the electrically evoked vestibulo-ocular reflex (eVOR) were analyzed to probe the latency and fidelity of MLF axonal conduction. The MLF and T2-visible brainstem lesions were defined by high-resolution MRI. White matter integrity was determined by diffusion-weighted imaging metrics. Results: eVOR onset latency was positively correlated with MLF lesion length (left: r = 0.66, p = 0.004; right: r = 0.75, p = 0.001). The mean conduction velocity (±SD) within MLF lesions was estimated at 2.72 (±0.87) m/s. eVOR onset latency correlated with normalized axial diffusivity (r = 0.66, p < 0.001) and fractional anisotropy (r = 0.44, p = 0.02) after exclusion of cases with ipsilateral vestibular root entry zone lesions. Conclusions: Axonal conduction velocity through lesions involving the MLF was reduced below levels predicted for natively myelinated and remyelinated axons. Composite in vivo biomarkers enable delineation of axonal from myelin processes and may provide a crucial role in assessing efficacy of novel reparative therapies in MS.
Keywords: Multiple sclerosis, magnetic resonance imaging, electrophysiology, medial longitudinal fasciculus, demyelination, remyelination, axonal conduction Date received: 16 March 2014; revised: 23 June 2014; 18 August 2014; accepted: 17 September 2014 Introduction Multiple sclerosis (MS) is a progressive central nervous system (CNS) inflammatory disorder characterized by the accrual of white and gray matter lesions that exhibit varying degrees of inflammatory cell infiltration, demyelination, gliosis and neuroaxonal loss.1 While the clinical manifestations of MS are protean, reflecting the fact that pathology can affect any site within the CNS, the etiology of most symptoms can be traced to quantitative and qualitative disruption of neuroaxonal conduction in clinically eloquent white matter tracts.2,3
Experimental studies have indicated two mechanisms by which conduction can be restored that can be separated by lesional axonal conduction velocity. The first is via upregulation and spread of voltage-gated sodium channels to areas of demyelinated axolemma,4,5 leading to restoration of axonal conduction, albeit at far lower velocity than myelinated fibers.6 The second is remyelination of axons, a common feature of MS lesions,7,8 which leads to restoration of secure conduction at near-normal velocities,9 clinical remission and long-term neuroprotection.10–12 Distinguishing these mechanisms in vivo may inform
Correspondence to: Michael H Barnett Sydney Neuroimaging Analysis Centre; Brain and Mind Research Institute, University of Sydney, Level 4, 94 Mallett St Camperdown, Sydney, Australia. michael@sydneyneurology. com.au C.W. and D.P. contributed equally to this work. Chenyu Wang Sean N Hatton Jim Lagopoulos Sydney Neuroimaging Analysis Centre, Sydney, Australia/Brain and Mind Research Institute, University of Sydney, Sydney, Australia Matthew C Kiernan Brain and Mind Research Institute, University of Sydney, Sydney, Australia David Paling Royal Hallamshire Hospital, Sheffield, UK and Department of Neuroscience, University of Sheffield, Sheffield, UK Luke Chen Swee T Aw Central Clinical School, University of Sydney, Sydney, Australia/Royal Prince Alfred Hospital, Sydney, Australia Michael H Barnett Sydney Neuroimaging Analysis Centre, Sydney, Australia/Brain and Mind Research Institute, University of Sydney, Sydney, Australia/ Royal Prince Alfred Hospital, Sydney, Australia
http://msj.sagepub.com 905
Multiple Sclerosis Journal 21(7) MS pathogenesis and provide a method for developing and testing therapies to enhance remyelination. The medial longitudinal fasciculus (MLF) is a highly myelinated tract that connects the oculomotor nuclei in the midbrain to the abducens nuclei in the pons, and is essential for coordination of conjugate horizontal eye movements.13 Disruption of this tract can cause internuclear ophthalmoplegia (INO), characterized by slowing or limitation of ipsilateral adducting eye relative to abducting eye saccades. INO is common in MS and can lead to a break in binocular fusion,14 diplopia, loss of depth perception, and reduction in quality of life.15–17 Since vestibular afferents travel through the MLF,18,19 quantitative analysis of the vestibular-ocular reflex (VOR) can be used to probe alteration in the latency and fidelity of MLF axonal conduction in vivo.20,21 In particular, electrical vestibular stimulation (EVS) provides an objective, precise and reproducible method of measuring the VOR.22,23,25 EVS assesses global vestibular function by stimulating both semicircular canal and otolith afferents to generate four parameters: evoked vestibulo-ocular reflex (eVOR) onset latency, tonic eVOR, phasic eVOR initiation and phasic eVOR cessation. In turn, correlation of eVOR metrics with high-resolution structural imaging of the brainstem in patients with INO may provide an opportunity to study the functional impact of MS lesions within the MLF.24 Figure 1 schematically illustrates the excitatory and inhibitory horizontal semicircular canal pathways and the three-neuron reflex arc.25 Diffusion tensor imaging (DTI) is an established application of MRI that is sensitive to the microstructural organization of white matter tracts. Disruption of white matter tract organization is reflected by a reduction in fractional anisotropy (FA).26 However, both demyelination and discrete loss of axons can result in reduced FA. Complementary DTI metrics that help to distinguish between these two processes include quantitative measures of axial diffusivity (AD) and radial diffusivity (RD) that model water diffusion along or across the axis of the white matter tract and reflect the integrity of axons and myelin, respectively.27 A previous study has shown that infrared oculography measurements of binocular saccadic dysconjugacy28 are associated with increased mean diffusivity (MD) and decreased FA in the MLF.29 While this study highlighted the potential of combined analysis of functional and imaging parameters, axonal conduction velocity in the MLF was not measured since saccadic dysconjugacy is sensitive to both axonal conduction delay and failure. As such,
Figure 1. Schematic illustration of the excitatory (solid red line) and inhibitory (dashed red line) horizontal semicircular canal pathways and the 3-neuron reflex arc comprising 1: vestibular nerve; 2: vestibulo-ocular secondary neuron; 3: abducens motorneuron. LR: lateral rectus muscle; MR: medial rectus muscle; ON: oculomotor nucleus; AN: abducens nucleus; VN: vestibular nucleus; MLF: medial longitudinal fasciculus. Equivalent threeneuron reflex arcs also exist for the vertical semicircular canal pathways.25
the present study sought to determine the pathophysiology of MLF dysfunction in MS by correlating eVOR parameters with lesion length and DTI metrics in patients with clinically suspected INO, and ultimately to define a composite in vivo biomarker of axonal and myelin integrity. Materials and methods Participants The study was approved by the Sydney Local Health District and University of Sydney ethics committees. Written informed consent was obtained from all participants. MRI and electrical vestibular stimulation were recorded in 18 patients with clinically definite MS and clinically suspected INO recruited from the MS Clinic, Brain and Mind Research Institute, University of Sydney (Table 1). Disability was
906 http://msj.sagepub.com
C Wang, D Paling et al. Table 1. Patient data. Patient
Age/Sex
Disease type/ duration (years)
EDSS
Current treatment
INO
Right MLF Latencya (s)
Left MLF Latencya (s)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
44/F 28/F 53/F 64/M 61/F 48/F 33/M 60/M 30/F 46/F 40/M 39/M 66/M 57/M 50/F 42/F 44/F 49/M
RRMS/7 RRMS/1 RRMS/4 SPMS/17 PPMS/23 RRMS/12 RRMS/12 RRMS/12 RRMS/1 RRMS/12 RRMS/3 SPMS/25 RRMS/25 RRMS/34 RRMS/13 RRMS/16 RRMS/7 RRMS/3
2.0 1.0 3.0 5.5 2.5 6.0 6.0 2.5 1.0 1.0 2.5 4.0 3.0 2.5 3.0 4.0 5.5 2.5
β-IF 1b GA GA GA Nil Natalizumab Fingolimod Nil GA β-IF 1a Fingolimod Nil Nil β-IF 1b β-IF 1a Fingolimod GA Fingolimod
None Right Right Right Left Bilateral Bilateral Bilateral Bilateral Bilateral Bilateral Bilateral Bilateral Bilateral Bilateral Bilateral Bilateral Bilateral
8.8 12.7 12.4 11.6 8.8 14.2 12.2 13.4 11.2 11.8 14.5 12.6 12.8 14.8 13.2 11.8 17.2 14.0
8.8 8.8 8.8 9.2 11.8 12.8 12 13 11.4 12.2 14.1 13.4 12.4 15.4 12 11.6 17.4 13.8
EDSS: Expanded Disability Status Scale; F: female; M: male; GA: glatiramer acetate; IF: interferon; PPMS: primary progressive multiple sclerosis; RRMS: relapsing–remitting multiple sclerosis; SPMS: secondary progressive multiple sclerosis; eVOR: evoked vestibulo-ocular reflex; INO: internuclear ophthalmoplegia. aeVOR onset latency.
assessed using the Expanded Disability Status Scale (EDSS).
MRI acquisition MRI scans were acquired at 3 Tesla (GE MR750 Discovery, General Electric Healthcare, Fairfield, CT, USA) with an eight-channel phased array head coil and the following parameters: axial brainstem highresolution T2-weighted spin echo images with two acquisitions, brainstem diffusion-weighted images (DWI) using an echo planar imaging sequence (four images without gradient loading (b0 s/mm2) prior to the acquisition of 61 images (each containing 36 slices) with a uniform gradient loading (b = 1000s/ mm2), and whole-brain volumetric sagittal T1weighted images (Table 2). MRI analysis Brainstem T2 hyperintense lesions were outlined on the T2-weighted images by a trained neuroimaging analyst (CW) using a semi-automated technique based on local thresholding using the region of interest toolkit of JIM 6.0 (Xinapse Systems, Northants, UK; www.xinapse.com). For each case, the path of the MLF was derived from a neuroanatomical atlas,30 and
coregistered with the lesion-marked T2-weighted image. Overlap between outlined lesions and the atlasderived MLF mask was used to calculate the “MLF lesion length” in the imaging slice direction. Lesions at the root entry zone of the ipsilateral vestibular nerve, which could potentially impact (prolong) eVOR latency measurements and impair phasic responses, were identified on T2-weighted sequences (Figure 2). All DWI was initially analyzed using the FMRIB Software Library (FSL v5.0; www.fmrib.ox.ac.uk/ fsl).31 First, voxel-wise maps of FA, AD, RD and MD (Figure 3) were derived using the FMRIB Diffusion Toolbox (FDT). Brainstem DTI and T2 images were then coregistered to the three-dimensional (3D) T1 images using the MrDiffusion toolbox (http://white. stanford.edu/newlm/index.php/MrDiffusion) and Robust Multiresolution Affine Registration model in Slicer (http://www.slicer.org). Coregistered images were then transferred into native T2 space by applying an inverted T2-to-T1 transformation matrix using FMRIB’s Linear Image Registration tool (FLIRT). T2 lesion and MLF masks were then superimposed on the coregistered images for DTI analysis. DTI measurements were divided into three categories: average DTI matrices along the entire MLF,
http://msj.sagepub.com 907
Multiple Sclerosis Journal 21(7) Table 2. MRI sequence parameters. Parameter
T2-weighted
T1-weighted
DWI
Imaging mode Position Slice thickness (mm) Field of view (mm) Matrix (frequency × phase) Repetition time (ms) Echo time (ms) Flip angle (degrees) Number of averages Bandwidth (Hz/Pxl)
Axial Brainstem 2 220×220 512×512 6178 110 90 2 122
Sagittal Whole brain 1 256×256 256×256 7.1 2.7 12 1 325
Axial Brainstem 2 240×240 96×96 6000 83.2 90 1 1953
MRI: magnetic resonance imaging; DWI: diffusion-weighted imaging.
Figure 2. Lesion at the root entry zone of the left vestibular nerve in a patient with a separate ipsilateral lesion involving the MLF and ipsilateral INO. MLF: medial longitudinal fasciculus; INO: internuclear ophthalmoplegia.
average DTI matrices in the proportion of MLF containing MS lesion as defined on T2-weighted sequences (“lesional-MLF”), and average DTI matrices in the remainder of MLF (“non-lesional MLF”). Analysis of MLF tract-based metrics was performed in MATLAB (http://www.mathworks.com.au) using an in-house-developed toolkit. A normal MLF model was constructed using two independent techniques to correct for varying DTI metrics along the MLF due to partial volume effects through adjacent structures (for details see supplementary data). DTI metrics derived from the first of these models were subtracted from those measured directly from patients with INO, and normalized DTI measurements (ΔRD, ΔAD, ΔFA and ΔMD) were averaged across all slices through the MLF for correlation with eVOR results.
Electrical vestibular stimulation, eye movement recording During recording the patient was supine with the head secured to prevent motion artifact while viewing a 2 mm light-emitting diode (LED) located 600 mm away in semidarkness. EVS consisted of bilateral, bipolar 100 ms direct-current pulses delivered at 0.9~7.5 mA steps via 4 cm2 SureFit trans-mastoid surface electrodes (ConMed, Utica, NY, USA). Sixty pulse repetitions were delivered at 1 per second by a DS5 isolated bipolar constant current stimulator (Digitimer Ltd, Welwyn Garden City, UK) controlled by in-house automated software developed in LabView 7.1 (National Instruments, TX, USA). EVS was tested in left-cathode/right-anode or right-cathode/left-anode configuration. Binocular, 3D eye positions in three orthogonal axes (torsional, horizontal and vertical) were measured using dual search coils23 (Skalar, Delft, The Netherlands) after precalibration for gains and offsets. Search coil signals recovered by preamplification and phase detection (CNC Engineering, Seattle, WA, USA) were sampled together with current-switch signal at 5-kHz and 24-bit resolution. The eye movement recording system had a resolution of mean +2 SD).32 The MLF with VDI-defined INO was designated as “INO-MLF.” Electrically eVOR analysis Data were analyzed offline using Labview software (National Instruments, Austin, TX, USA). After excluding trials with blinks, the remaining 30~40 runs were averaged and filtered with a low-pass filter with bandwidth of DC-140 Hz. The eVOR onset latency
908 http://msj.sagepub.com
C Wang, D Paling et al.
Figure 3. Comparison of axial T1 and T2-weighted images with DTI metrics acquired through the mid-pons in a healthy adult (Column (a)) and a patient with unilateral INO (Columns (b) and (c)). Within the lesion, there is a clear decrease in FA and AD signal intensities concomitant with an increase in RD signal. DTI: diffusion tensor imaging; INO: internuclear ophthalmoplegia; FA: fractional anisotropy; AD: axial diffusivity.
http://msj.sagepub.com 909
Multiple Sclerosis Journal 21(7) was calculated automatically, and defined as the interval between EVS onset and the point when torsional eye velocity first exceeded one standard deviation (SD) of the baseline noise. As in previous analyses,22 torsional eye movement was chosen because it was the largest of the torsional, vertical and horizontal responses to EVS, all of which have identical latencies. The phasic eVOR to 7.5 mA current intensity was defined as the mean peak torsional eye acceleration at initiation and cessation.23 Mathematical modeling of lesional axonal conduction velocity Aggregating our experimental data, we found a mean eVOR onset latency of 8.9 ms (range 8.8~9.2 ms) in the unaffected MLF (versional dysconjugacy index ≤1.2) determined by search coil recording of unilateral INO patients, which is similar to previously reported results in healthy controls.22 Latency delay due to the presence of a lesion was therefore calculated by subtracting 8.9 ms from the measured latency. Assuming normal axonal transmission of 32 m/s in ascending MLF axons,33 and assuming that impaired axonal conduction is attributable to demyelination in the lesional MLF, axonal conduction velocity through the lesion can be calculated using the formula:
∆τ =
l l − v1 v 2
where Δτ is the latency increases due to lesion presence, derived by subtracting 8.9 ms from measured latency, l is the length of the lesion along the MLF, ν1 is the conduction speed of demyelinated MLF, and ν2 is the average conduction speed in the normal MLF (32 m/ms). Statistics Statistical analyses were performed using the Statistical Package for the Social Sciences (SPSS 22.0 for Mac). Pearson partial correlations (r) explored the linear correlation between eVOR properties and MRI metrics. As there was an a priori hypothesis that white matter disruption would result in more abnormal eVOR (e.g. as the case shown in Figure 3), significance was set at p < 0.05 (one tailed). Results Patient characteristics Eighteen patients (10 females, eight males; mean age 47.7 years) with MS and clinically suspected INO were
Figure 4. Lesion distribution along the MLF in 18 MS patients with INO. MLF: medial longitudinal fasciculus; MS: multiple sclerosis; INO: internuclear ophthalmoplegia.
recruited (Table 1). The mean duration of MS was 12.6 years ± 9.4 SD and mean EDSS 3.2 ± 1.6 SD. Seventyeight percent of patients were on immunotherapy at the time of testing. Search coil recording confirmed INO in 17 patients. In four patients, the INO was unilateral: In two it was on the right and in two it was on the left.
Lesions in the MLF All patients with INO had a corresponding lesion on the T2-weighted MRI. The maximum diameter of lesions involving the MLF was generally much greater than the caliber of the MLF itself, and in all instances involved the entire outlined tract and part of the adjacent brainstem in individual axial T2 slices (Figure 3). In no case was an INO detected by search coil recording in the absence of a visible lesion on MRI. Lesions occupied a mean of 30% ± 11.7% SD of the length of the MLF. The distribution of lesions along the MLF in all patients is shown in Figure 4. In a single patient with clinically suspected INO at screening, the clinical examination had normalized at time of search coil recording three weeks later, the VDI was normal and no T2visible pathology was identified on MRI. In five
910 http://msj.sagepub.com
C Wang, D Paling et al. Table 3. MLF eVOR and direct MLF DTI measurements.
Latency (ms) Peak acceleration (initiation) (degrees/s2) Peak acceleration (cessation) (degrees/s2) MLF mean FA MLF mean AD (×10−3mm^2/s) MLF mean RD (×10−4mm^2/s) MLF mean MD (×10−4mm^2/s)
(INO) MLF ± SD
(non-INO) MLF ± SD
13.12±1.563 703.52±241.647 781.48±357.015 0.41±0.035 1.33±0.095 6.85±0.711 9.00±0.751
8.87±0.163 1317.00±544.250 1722.40±627.944 0.42±0.048 1.25±0.034 6.34±0.569 8.41±0.400
MLF: medial longitudinal fasciculus; eVOR: evoked vestibulo-ocular reflex; INO: internuclear ophthalmoplegia; DTI: diffusion tensor imaging; AD: axial diffusivity; FA: fractional anisotropy; RD: radial diffusivity; MD: mean diffusivity; SD: standard deviation.
patients, lesions were identified at the root entry zone: three bilateral and two unilateral (both with right-sided lesions).
Electrically eVOR The eVOR onset latency (Table 3) was prolonged (p < 0.001) on the side with INO-MLF (13.12 ±1.6 ms, mean ±SD) compared to the side without INO-MLF (8.87 ± 0.16 ms). Similarly peak phasic acceleration (703 ± 241 degrees/s2) and deceleration (781 ± 357 degrees/s2) were reduced (acceleration: p = 0.064; deceleration: p = 0.027) on the INO-MLF side, compared to the normal side (acceleration: 1317 ± 544 degrees/s2, deceleration: 1722 ± 627 degrees/s2). There was a negative correlation between eVOR latency and peak phasic acceleration (r(34) = −0.61, p < 0.001) and peak phasic deceleration (r(34) = −0.68, p < 0.001). Correlation between eVOR and T2 lesion length along the MLF The eVOR onset latency was correlated with MLF lesion length (Figure 5). Greater left MLF lesion length was associated with prolonged eVOR onset latency (r(15) = 0.66, p = 0.004), and decrease in peak acceleration initiation (r(14)= −0.47, p = 0.043) and cessation (r(14) = −0.58, p = 0.014). Similarly, greater right MLF lesion length was associated with prolonged latency (r(15) = 0.75, p = 0.001) and decrease in peak acceleration cessation (r(15) = −0.64, p = 0.005), but not with decrease in peak acceleration initiation (r(15) = −0.26, p = 0.17). The association of eVOR onset latency with MLF lesion length remained significant after exclusion of the eight of 36 INO-MLFs with an associated T2visible ipsilateral vestibular root entry zone lesion (left: r(12) = 0.51, p = 0.045; right: r(10) = 0.75, p = 0.006), but the associations with peak acceleration initiation and cessation were no longer significant, with the
Figure 5. Scatter plot of lesion length along MLF vs. eVOR onset latency measurement. MLF: medial longitudinal fasciculus; eVOR: evoked vestibulo-ocular reflex.
exception of right cessation (left initiation: r(11) = −0.18, p = 0.298; right initiation: r(10) = −0.29, p = 0.209, left cessation: r(11) = −0.49, p = 0.064; right cessation: r(10) = −0.55, p = 0.049).
Correlation between eVOR and normalized DTI metrics Whole MLF metrics. Increased AD (p = 0.002), but not MD (p = 0.071) or RD (p = 0.111), were observed in INO-MLF when compared with the non-INO-MLF (Table 3). FA was reduced in INO-MLF but did not reach statistical significance (p = 0.630). Of the 22 INO-MLFs that were not associated with ipsilateral vestibular root entry zone lesions, eVOR onset latency correlated with ΔAD (r(22) = 0.66, p < 0.001), ΔFA (r(22) = 0.44, p = 0.021) and ΔMD (r(22) = 0.51, p = 0.008). However, there were
http://msj.sagepub.com 911
Multiple Sclerosis Journal 21(7) Table 4. Correlation of normalized DTI metrics along the entire MLF with eVOR onset latency. Excluding INO-MLFs with associated root entry zone lesions ΔRD ΔAD Latencya (n = 22) 0.06 (0.392) 0.66 (
research-article2014
MSJ0010.1177/1352458514556301Multiple Sclerosis JournalC Wang, D Paling et al.
MULTIPLE SCLEROSIS MSJ JOURNAL
Original Research Paper
Axonal conduction in multiple sclerosis: A combined magnetic resonance imaging and electrophysiological study of the medial longitudinal fasciculus
Multiple Sclerosis Journal 2015, Vol. 21(7) 905–915 DOI: 10.1177/ 1352458514556301 © The Author(s), 2014. Reprints and permissions: http://www.sagepub.co.uk/ journalsPermissions.nav
Chenyu Wang, David Paling, Luke Chen, Sean N Hatton, Jim Lagopoulos, Swee T Aw, Matthew C Kiernan and Michael H Barnett
Abstract Objective: The objective of this paper is to inform the pathophysiology of medial longitudinal fasciculus (MLF) axonal dysfunction in patients with internuclear ophthalmoplegia (INO) due to multiple sclerosis (MS), and develop a composite structural-functional biomarker of axonal and myelin integrity in this tract. Methods: Eighteen patients with definite MS and clinically suspected INO underwent electrical vestibular stimulation and search-coil eye movement recording. Components of the electrically evoked vestibulo-ocular reflex (eVOR) were analyzed to probe the latency and fidelity of MLF axonal conduction. The MLF and T2-visible brainstem lesions were defined by high-resolution MRI. White matter integrity was determined by diffusion-weighted imaging metrics. Results: eVOR onset latency was positively correlated with MLF lesion length (left: r = 0.66, p = 0.004; right: r = 0.75, p = 0.001). The mean conduction velocity (±SD) within MLF lesions was estimated at 2.72 (±0.87) m/s. eVOR onset latency correlated with normalized axial diffusivity (r = 0.66, p < 0.001) and fractional anisotropy (r = 0.44, p = 0.02) after exclusion of cases with ipsilateral vestibular root entry zone lesions. Conclusions: Axonal conduction velocity through lesions involving the MLF was reduced below levels predicted for natively myelinated and remyelinated axons. Composite in vivo biomarkers enable delineation of axonal from myelin processes and may provide a crucial role in assessing efficacy of novel reparative therapies in MS.
Keywords: Multiple sclerosis, magnetic resonance imaging, electrophysiology, medial longitudinal fasciculus, demyelination, remyelination, axonal conduction Date received: 16 March 2014; revised: 23 June 2014; 18 August 2014; accepted: 17 September 2014 Introduction Multiple sclerosis (MS) is a progressive central nervous system (CNS) inflammatory disorder characterized by the accrual of white and gray matter lesions that exhibit varying degrees of inflammatory cell infiltration, demyelination, gliosis and neuroaxonal loss.1 While the clinical manifestations of MS are protean, reflecting the fact that pathology can affect any site within the CNS, the etiology of most symptoms can be traced to quantitative and qualitative disruption of neuroaxonal conduction in clinically eloquent white matter tracts.2,3
Experimental studies have indicated two mechanisms by which conduction can be restored that can be separated by lesional axonal conduction velocity. The first is via upregulation and spread of voltage-gated sodium channels to areas of demyelinated axolemma,4,5 leading to restoration of axonal conduction, albeit at far lower velocity than myelinated fibers.6 The second is remyelination of axons, a common feature of MS lesions,7,8 which leads to restoration of secure conduction at near-normal velocities,9 clinical remission and long-term neuroprotection.10–12 Distinguishing these mechanisms in vivo may inform
Correspondence to: Michael H Barnett Sydney Neuroimaging Analysis Centre; Brain and Mind Research Institute, University of Sydney, Level 4, 94 Mallett St Camperdown, Sydney, Australia. michael@sydneyneurology. com.au C.W. and D.P. contributed equally to this work. Chenyu Wang Sean N Hatton Jim Lagopoulos Sydney Neuroimaging Analysis Centre, Sydney, Australia/Brain and Mind Research Institute, University of Sydney, Sydney, Australia Matthew C Kiernan Brain and Mind Research Institute, University of Sydney, Sydney, Australia David Paling Royal Hallamshire Hospital, Sheffield, UK and Department of Neuroscience, University of Sheffield, Sheffield, UK Luke Chen Swee T Aw Central Clinical School, University of Sydney, Sydney, Australia/Royal Prince Alfred Hospital, Sydney, Australia Michael H Barnett Sydney Neuroimaging Analysis Centre, Sydney, Australia/Brain and Mind Research Institute, University of Sydney, Sydney, Australia/ Royal Prince Alfred Hospital, Sydney, Australia
http://msj.sagepub.com 905
Multiple Sclerosis Journal 21(7) MS pathogenesis and provide a method for developing and testing therapies to enhance remyelination. The medial longitudinal fasciculus (MLF) is a highly myelinated tract that connects the oculomotor nuclei in the midbrain to the abducens nuclei in the pons, and is essential for coordination of conjugate horizontal eye movements.13 Disruption of this tract can cause internuclear ophthalmoplegia (INO), characterized by slowing or limitation of ipsilateral adducting eye relative to abducting eye saccades. INO is common in MS and can lead to a break in binocular fusion,14 diplopia, loss of depth perception, and reduction in quality of life.15–17 Since vestibular afferents travel through the MLF,18,19 quantitative analysis of the vestibular-ocular reflex (VOR) can be used to probe alteration in the latency and fidelity of MLF axonal conduction in vivo.20,21 In particular, electrical vestibular stimulation (EVS) provides an objective, precise and reproducible method of measuring the VOR.22,23,25 EVS assesses global vestibular function by stimulating both semicircular canal and otolith afferents to generate four parameters: evoked vestibulo-ocular reflex (eVOR) onset latency, tonic eVOR, phasic eVOR initiation and phasic eVOR cessation. In turn, correlation of eVOR metrics with high-resolution structural imaging of the brainstem in patients with INO may provide an opportunity to study the functional impact of MS lesions within the MLF.24 Figure 1 schematically illustrates the excitatory and inhibitory horizontal semicircular canal pathways and the three-neuron reflex arc.25 Diffusion tensor imaging (DTI) is an established application of MRI that is sensitive to the microstructural organization of white matter tracts. Disruption of white matter tract organization is reflected by a reduction in fractional anisotropy (FA).26 However, both demyelination and discrete loss of axons can result in reduced FA. Complementary DTI metrics that help to distinguish between these two processes include quantitative measures of axial diffusivity (AD) and radial diffusivity (RD) that model water diffusion along or across the axis of the white matter tract and reflect the integrity of axons and myelin, respectively.27 A previous study has shown that infrared oculography measurements of binocular saccadic dysconjugacy28 are associated with increased mean diffusivity (MD) and decreased FA in the MLF.29 While this study highlighted the potential of combined analysis of functional and imaging parameters, axonal conduction velocity in the MLF was not measured since saccadic dysconjugacy is sensitive to both axonal conduction delay and failure. As such,
Figure 1. Schematic illustration of the excitatory (solid red line) and inhibitory (dashed red line) horizontal semicircular canal pathways and the 3-neuron reflex arc comprising 1: vestibular nerve; 2: vestibulo-ocular secondary neuron; 3: abducens motorneuron. LR: lateral rectus muscle; MR: medial rectus muscle; ON: oculomotor nucleus; AN: abducens nucleus; VN: vestibular nucleus; MLF: medial longitudinal fasciculus. Equivalent threeneuron reflex arcs also exist for the vertical semicircular canal pathways.25
the present study sought to determine the pathophysiology of MLF dysfunction in MS by correlating eVOR parameters with lesion length and DTI metrics in patients with clinically suspected INO, and ultimately to define a composite in vivo biomarker of axonal and myelin integrity. Materials and methods Participants The study was approved by the Sydney Local Health District and University of Sydney ethics committees. Written informed consent was obtained from all participants. MRI and electrical vestibular stimulation were recorded in 18 patients with clinically definite MS and clinically suspected INO recruited from the MS Clinic, Brain and Mind Research Institute, University of Sydney (Table 1). Disability was
906 http://msj.sagepub.com
C Wang, D Paling et al. Table 1. Patient data. Patient
Age/Sex
Disease type/ duration (years)
EDSS
Current treatment
INO
Right MLF Latencya (s)
Left MLF Latencya (s)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
44/F 28/F 53/F 64/M 61/F 48/F 33/M 60/M 30/F 46/F 40/M 39/M 66/M 57/M 50/F 42/F 44/F 49/M
RRMS/7 RRMS/1 RRMS/4 SPMS/17 PPMS/23 RRMS/12 RRMS/12 RRMS/12 RRMS/1 RRMS/12 RRMS/3 SPMS/25 RRMS/25 RRMS/34 RRMS/13 RRMS/16 RRMS/7 RRMS/3
2.0 1.0 3.0 5.5 2.5 6.0 6.0 2.5 1.0 1.0 2.5 4.0 3.0 2.5 3.0 4.0 5.5 2.5
β-IF 1b GA GA GA Nil Natalizumab Fingolimod Nil GA β-IF 1a Fingolimod Nil Nil β-IF 1b β-IF 1a Fingolimod GA Fingolimod
None Right Right Right Left Bilateral Bilateral Bilateral Bilateral Bilateral Bilateral Bilateral Bilateral Bilateral Bilateral Bilateral Bilateral Bilateral
8.8 12.7 12.4 11.6 8.8 14.2 12.2 13.4 11.2 11.8 14.5 12.6 12.8 14.8 13.2 11.8 17.2 14.0
8.8 8.8 8.8 9.2 11.8 12.8 12 13 11.4 12.2 14.1 13.4 12.4 15.4 12 11.6 17.4 13.8
EDSS: Expanded Disability Status Scale; F: female; M: male; GA: glatiramer acetate; IF: interferon; PPMS: primary progressive multiple sclerosis; RRMS: relapsing–remitting multiple sclerosis; SPMS: secondary progressive multiple sclerosis; eVOR: evoked vestibulo-ocular reflex; INO: internuclear ophthalmoplegia. aeVOR onset latency.
assessed using the Expanded Disability Status Scale (EDSS).
MRI acquisition MRI scans were acquired at 3 Tesla (GE MR750 Discovery, General Electric Healthcare, Fairfield, CT, USA) with an eight-channel phased array head coil and the following parameters: axial brainstem highresolution T2-weighted spin echo images with two acquisitions, brainstem diffusion-weighted images (DWI) using an echo planar imaging sequence (four images without gradient loading (b0 s/mm2) prior to the acquisition of 61 images (each containing 36 slices) with a uniform gradient loading (b = 1000s/ mm2), and whole-brain volumetric sagittal T1weighted images (Table 2). MRI analysis Brainstem T2 hyperintense lesions were outlined on the T2-weighted images by a trained neuroimaging analyst (CW) using a semi-automated technique based on local thresholding using the region of interest toolkit of JIM 6.0 (Xinapse Systems, Northants, UK; www.xinapse.com). For each case, the path of the MLF was derived from a neuroanatomical atlas,30 and
coregistered with the lesion-marked T2-weighted image. Overlap between outlined lesions and the atlasderived MLF mask was used to calculate the “MLF lesion length” in the imaging slice direction. Lesions at the root entry zone of the ipsilateral vestibular nerve, which could potentially impact (prolong) eVOR latency measurements and impair phasic responses, were identified on T2-weighted sequences (Figure 2). All DWI was initially analyzed using the FMRIB Software Library (FSL v5.0; www.fmrib.ox.ac.uk/ fsl).31 First, voxel-wise maps of FA, AD, RD and MD (Figure 3) were derived using the FMRIB Diffusion Toolbox (FDT). Brainstem DTI and T2 images were then coregistered to the three-dimensional (3D) T1 images using the MrDiffusion toolbox (http://white. stanford.edu/newlm/index.php/MrDiffusion) and Robust Multiresolution Affine Registration model in Slicer (http://www.slicer.org). Coregistered images were then transferred into native T2 space by applying an inverted T2-to-T1 transformation matrix using FMRIB’s Linear Image Registration tool (FLIRT). T2 lesion and MLF masks were then superimposed on the coregistered images for DTI analysis. DTI measurements were divided into three categories: average DTI matrices along the entire MLF,
http://msj.sagepub.com 907
Multiple Sclerosis Journal 21(7) Table 2. MRI sequence parameters. Parameter
T2-weighted
T1-weighted
DWI
Imaging mode Position Slice thickness (mm) Field of view (mm) Matrix (frequency × phase) Repetition time (ms) Echo time (ms) Flip angle (degrees) Number of averages Bandwidth (Hz/Pxl)
Axial Brainstem 2 220×220 512×512 6178 110 90 2 122
Sagittal Whole brain 1 256×256 256×256 7.1 2.7 12 1 325
Axial Brainstem 2 240×240 96×96 6000 83.2 90 1 1953
MRI: magnetic resonance imaging; DWI: diffusion-weighted imaging.
Figure 2. Lesion at the root entry zone of the left vestibular nerve in a patient with a separate ipsilateral lesion involving the MLF and ipsilateral INO. MLF: medial longitudinal fasciculus; INO: internuclear ophthalmoplegia.
average DTI matrices in the proportion of MLF containing MS lesion as defined on T2-weighted sequences (“lesional-MLF”), and average DTI matrices in the remainder of MLF (“non-lesional MLF”). Analysis of MLF tract-based metrics was performed in MATLAB (http://www.mathworks.com.au) using an in-house-developed toolkit. A normal MLF model was constructed using two independent techniques to correct for varying DTI metrics along the MLF due to partial volume effects through adjacent structures (for details see supplementary data). DTI metrics derived from the first of these models were subtracted from those measured directly from patients with INO, and normalized DTI measurements (ΔRD, ΔAD, ΔFA and ΔMD) were averaged across all slices through the MLF for correlation with eVOR results.
Electrical vestibular stimulation, eye movement recording During recording the patient was supine with the head secured to prevent motion artifact while viewing a 2 mm light-emitting diode (LED) located 600 mm away in semidarkness. EVS consisted of bilateral, bipolar 100 ms direct-current pulses delivered at 0.9~7.5 mA steps via 4 cm2 SureFit trans-mastoid surface electrodes (ConMed, Utica, NY, USA). Sixty pulse repetitions were delivered at 1 per second by a DS5 isolated bipolar constant current stimulator (Digitimer Ltd, Welwyn Garden City, UK) controlled by in-house automated software developed in LabView 7.1 (National Instruments, TX, USA). EVS was tested in left-cathode/right-anode or right-cathode/left-anode configuration. Binocular, 3D eye positions in three orthogonal axes (torsional, horizontal and vertical) were measured using dual search coils23 (Skalar, Delft, The Netherlands) after precalibration for gains and offsets. Search coil signals recovered by preamplification and phase detection (CNC Engineering, Seattle, WA, USA) were sampled together with current-switch signal at 5-kHz and 24-bit resolution. The eye movement recording system had a resolution of mean +2 SD).32 The MLF with VDI-defined INO was designated as “INO-MLF.” Electrically eVOR analysis Data were analyzed offline using Labview software (National Instruments, Austin, TX, USA). After excluding trials with blinks, the remaining 30~40 runs were averaged and filtered with a low-pass filter with bandwidth of DC-140 Hz. The eVOR onset latency
908 http://msj.sagepub.com
C Wang, D Paling et al.
Figure 3. Comparison of axial T1 and T2-weighted images with DTI metrics acquired through the mid-pons in a healthy adult (Column (a)) and a patient with unilateral INO (Columns (b) and (c)). Within the lesion, there is a clear decrease in FA and AD signal intensities concomitant with an increase in RD signal. DTI: diffusion tensor imaging; INO: internuclear ophthalmoplegia; FA: fractional anisotropy; AD: axial diffusivity.
http://msj.sagepub.com 909
Multiple Sclerosis Journal 21(7) was calculated automatically, and defined as the interval between EVS onset and the point when torsional eye velocity first exceeded one standard deviation (SD) of the baseline noise. As in previous analyses,22 torsional eye movement was chosen because it was the largest of the torsional, vertical and horizontal responses to EVS, all of which have identical latencies. The phasic eVOR to 7.5 mA current intensity was defined as the mean peak torsional eye acceleration at initiation and cessation.23 Mathematical modeling of lesional axonal conduction velocity Aggregating our experimental data, we found a mean eVOR onset latency of 8.9 ms (range 8.8~9.2 ms) in the unaffected MLF (versional dysconjugacy index ≤1.2) determined by search coil recording of unilateral INO patients, which is similar to previously reported results in healthy controls.22 Latency delay due to the presence of a lesion was therefore calculated by subtracting 8.9 ms from the measured latency. Assuming normal axonal transmission of 32 m/s in ascending MLF axons,33 and assuming that impaired axonal conduction is attributable to demyelination in the lesional MLF, axonal conduction velocity through the lesion can be calculated using the formula:
∆τ =
l l − v1 v 2
where Δτ is the latency increases due to lesion presence, derived by subtracting 8.9 ms from measured latency, l is the length of the lesion along the MLF, ν1 is the conduction speed of demyelinated MLF, and ν2 is the average conduction speed in the normal MLF (32 m/ms). Statistics Statistical analyses were performed using the Statistical Package for the Social Sciences (SPSS 22.0 for Mac). Pearson partial correlations (r) explored the linear correlation between eVOR properties and MRI metrics. As there was an a priori hypothesis that white matter disruption would result in more abnormal eVOR (e.g. as the case shown in Figure 3), significance was set at p < 0.05 (one tailed). Results Patient characteristics Eighteen patients (10 females, eight males; mean age 47.7 years) with MS and clinically suspected INO were
Figure 4. Lesion distribution along the MLF in 18 MS patients with INO. MLF: medial longitudinal fasciculus; MS: multiple sclerosis; INO: internuclear ophthalmoplegia.
recruited (Table 1). The mean duration of MS was 12.6 years ± 9.4 SD and mean EDSS 3.2 ± 1.6 SD. Seventyeight percent of patients were on immunotherapy at the time of testing. Search coil recording confirmed INO in 17 patients. In four patients, the INO was unilateral: In two it was on the right and in two it was on the left.
Lesions in the MLF All patients with INO had a corresponding lesion on the T2-weighted MRI. The maximum diameter of lesions involving the MLF was generally much greater than the caliber of the MLF itself, and in all instances involved the entire outlined tract and part of the adjacent brainstem in individual axial T2 slices (Figure 3). In no case was an INO detected by search coil recording in the absence of a visible lesion on MRI. Lesions occupied a mean of 30% ± 11.7% SD of the length of the MLF. The distribution of lesions along the MLF in all patients is shown in Figure 4. In a single patient with clinically suspected INO at screening, the clinical examination had normalized at time of search coil recording three weeks later, the VDI was normal and no T2visible pathology was identified on MRI. In five
910 http://msj.sagepub.com
C Wang, D Paling et al. Table 3. MLF eVOR and direct MLF DTI measurements.
Latency (ms) Peak acceleration (initiation) (degrees/s2) Peak acceleration (cessation) (degrees/s2) MLF mean FA MLF mean AD (×10−3mm^2/s) MLF mean RD (×10−4mm^2/s) MLF mean MD (×10−4mm^2/s)
(INO) MLF ± SD
(non-INO) MLF ± SD
13.12±1.563 703.52±241.647 781.48±357.015 0.41±0.035 1.33±0.095 6.85±0.711 9.00±0.751
8.87±0.163 1317.00±544.250 1722.40±627.944 0.42±0.048 1.25±0.034 6.34±0.569 8.41±0.400
MLF: medial longitudinal fasciculus; eVOR: evoked vestibulo-ocular reflex; INO: internuclear ophthalmoplegia; DTI: diffusion tensor imaging; AD: axial diffusivity; FA: fractional anisotropy; RD: radial diffusivity; MD: mean diffusivity; SD: standard deviation.
patients, lesions were identified at the root entry zone: three bilateral and two unilateral (both with right-sided lesions).
Electrically eVOR The eVOR onset latency (Table 3) was prolonged (p < 0.001) on the side with INO-MLF (13.12 ±1.6 ms, mean ±SD) compared to the side without INO-MLF (8.87 ± 0.16 ms). Similarly peak phasic acceleration (703 ± 241 degrees/s2) and deceleration (781 ± 357 degrees/s2) were reduced (acceleration: p = 0.064; deceleration: p = 0.027) on the INO-MLF side, compared to the normal side (acceleration: 1317 ± 544 degrees/s2, deceleration: 1722 ± 627 degrees/s2). There was a negative correlation between eVOR latency and peak phasic acceleration (r(34) = −0.61, p < 0.001) and peak phasic deceleration (r(34) = −0.68, p < 0.001). Correlation between eVOR and T2 lesion length along the MLF The eVOR onset latency was correlated with MLF lesion length (Figure 5). Greater left MLF lesion length was associated with prolonged eVOR onset latency (r(15) = 0.66, p = 0.004), and decrease in peak acceleration initiation (r(14)= −0.47, p = 0.043) and cessation (r(14) = −0.58, p = 0.014). Similarly, greater right MLF lesion length was associated with prolonged latency (r(15) = 0.75, p = 0.001) and decrease in peak acceleration cessation (r(15) = −0.64, p = 0.005), but not with decrease in peak acceleration initiation (r(15) = −0.26, p = 0.17). The association of eVOR onset latency with MLF lesion length remained significant after exclusion of the eight of 36 INO-MLFs with an associated T2visible ipsilateral vestibular root entry zone lesion (left: r(12) = 0.51, p = 0.045; right: r(10) = 0.75, p = 0.006), but the associations with peak acceleration initiation and cessation were no longer significant, with the
Figure 5. Scatter plot of lesion length along MLF vs. eVOR onset latency measurement. MLF: medial longitudinal fasciculus; eVOR: evoked vestibulo-ocular reflex.
exception of right cessation (left initiation: r(11) = −0.18, p = 0.298; right initiation: r(10) = −0.29, p = 0.209, left cessation: r(11) = −0.49, p = 0.064; right cessation: r(10) = −0.55, p = 0.049).
Correlation between eVOR and normalized DTI metrics Whole MLF metrics. Increased AD (p = 0.002), but not MD (p = 0.071) or RD (p = 0.111), were observed in INO-MLF when compared with the non-INO-MLF (Table 3). FA was reduced in INO-MLF but did not reach statistical significance (p = 0.630). Of the 22 INO-MLFs that were not associated with ipsilateral vestibular root entry zone lesions, eVOR onset latency correlated with ΔAD (r(22) = 0.66, p < 0.001), ΔFA (r(22) = 0.44, p = 0.021) and ΔMD (r(22) = 0.51, p = 0.008). However, there were
http://msj.sagepub.com 911
Multiple Sclerosis Journal 21(7) Table 4. Correlation of normalized DTI metrics along the entire MLF with eVOR onset latency. Excluding INO-MLFs with associated root entry zone lesions ΔRD ΔAD Latencya (n = 22) 0.06 (0.392) 0.66 (
