Characterisation of multiple sclerosis neuroinflammation and neurodegeneration with relaxation and diffusion basis spectrum imaging

Abstract

Background:

Advanced magnetic resonance imaging (MRI) methods can provide more specific information about various microstructural tissue changes in multiple sclerosis (MS) brain. Quantitative measurement of T₁ and T₂ relaxation, and diffusion basis spectrum imaging (DBSI) yield metrics related to the pathology of neuroinflammation and neurodegeneration that occurs across the spectrum of MS.

Objective:

To use relaxation and DBSI MRI metrics to describe measures of neuroinflammation, myelin and axons in different MS subtypes.

Methods:

103 participants (20 clinically isolated syndrome (CIS), 33 relapsing remitting (RRMS), 30 secondary progressive, 20 primary progressive) underwent quantitative T₁, T₂, DBSI and conventional 3T MRI. Whole brain, normal appearing white matter, lesion and corpus callosum MRI metrics were compared across MS subtypes.

Results:

A gradation of MRI metric values was seen from CIS to RRMS to progressive MS. RRMS demonstrated large oedema related differences while progressive MS had the most extensive abnormalities in myelin and axonal measures.

Conclusion:

Relaxation and DBSI-derived MRI measures show differences between MS subtypes related to the severity and composition of underlying tissue damage. RRMS showed oedema, demyelination, and axonal loss compared to CIS. Progressive MS had even more evidence of increased oedema, demyelination and axonal loss compared to CIS and RRMS.

Introduction

MRI has become an indispensable tool for diagnosing and monitoring multiple sclerosis (MS). Although conventional MRI is able to detect obvious areas of damage in the central nervous system, it does not have the specificity to differentiate between the various underlying tissue pathologies such as demyelination, axonal loss and gliosis.^1-3 A number of quantitative and advanced MRI techniques exist that are thought to be more sensitive and specific to changes in tissue microstructure, and may give more information about the actual tissue damage in MS. There is an extensive literature on metrics derived from diffusion tensor imaging (DTI) and magnetization transfer imaging, which can quantify abnormalities in MS lesions and normal appearing white matter (NAWM), and detect differences between MS subtypes with high sensitivity, but with low pathological specificity.^4-9

Other advanced MRI approaches may provide more specific information about tissue components, but have not, as of yet, been as widely applied in large studies or MS subtype analysis. Quantitative measurement of T₂ relaxation¹⁰ can determine the myelin water fraction (MWF), a histopathologically validated marker for myelin^11-14 which reflects the fraction of water between myelin layers. T₂ relaxation measurement can also provide the geometric mean T₂ (GMT₂), reflecting T₂ of intra and extracellular water which is related to tissue structure. Assessment of T₁ relaxation reflects overall water mobility and is closely related to water content (WC), which can be determined using metrics derived from T₁ and T₂ experiments.¹² Diffusion basis spectrum imaging (DBSI) employs a more sophisticated model than DTI to define anisotropic diffusion tensors (myelinated and unmyelinated axons) and isotropic diffusion tensors (cells and extracellular space) to quantify axonal injury, myelination, inflammation and oedema.¹³ Metrics derived from DBSI analysis are proposed to more accurately describe MS pathology than conventional DTI¹⁴ and include axial diffusivity (AD, axonal integrity),^13,15 radial diffusivity (RD, modulated by myelin),^13,15,16 fibre fraction (FF, density of axons),¹⁶ isotropic restricted diffusion fraction (RF, changes in cellularity due to inflammation)^13,15-17 and isotropic non-restricted diffusion fraction (Non-RF, increases with vasogenic oedema).¹⁷

The aforementioned more specific MRI approaches have been used to characterise MS pathology, reporting reduced MWF,¹⁸ increased T₁¹⁹ and T₂²⁰ and abnormalities in DBSI-related metrics^21-23 in lesions and NAWM. However, the majority of MWF reports are from relapsing remitting MS (RRMS) subjects and a comprehensive comparison between the different MS subtypes has not been reported. Furthermore, while DBSI has been applied to MS, the literature is very small.^16,21-23 There are no studies examining DBSI and MWF in the same cohort.

The purpose of this study was to use quantitative metrics derived from relaxation and DBSI MRI experiments to describe measures of neuroinflammation, myelin, and axons in different MS subtypes. Our analysis quantified measures from whole brain, NAWM, lesions and the corpus callosum (CC). The CC was chosen as a specific region of interest (ROI) as damage to this structure is often detected in MS.²⁴ Also, the smaller CC ROI size and more uniform tissue composition compared to NAWM and whole brain may make assessment of the CC more sensitive to tissue damage detection and differentiation between MS subtypes.

The expectation was that, in comparison to participants with clinically isolated syndrome (CIS), MR metrics from progressive MS will show larger differences than the relapsing remitting form of the disease. This is the largest MS-focused DBSI study to date, and we present our findings in comparison to literature established measures reflecting disease severity of lesion volume, normalized brain volume and cortical thickness.¹⁶

Methods

Participants

103 participants (20 CIS, 33 RRMS, 30 secondary progressive (SPMS), 20 primary progressive (PPMS)) were enrolled in the study (demographics in Table 1). The study was approved by the clinical research ethics board at our institution and all participants gave written informed consent.

Table 1:

Participant demographics. Mean (range) for age and disease duration, and median (range) for EDSS are listed.

Subtype	N	Sex	Age (y)	EDSS	Disease Duration (y)
CIS	20	15F/5M	36 (21-60)	1.0 (0-4.0)	2 (0.2-8)
RRMS	33	21F/12M	43 (24-59)	2.0 (0-4.0)	13 (0.5-41)
SPMS	30	18F/12M	57 (48-69)	5.5 (2.0-8.0)	22 (8-34)
PPMS	20	7F/13M	60 (45-70)	5.0 (2.0-6.5)	13 (1-36)

CIS=clinically isolated syndrome; RRMS=relapsing remitting multiple sclerosis; SPMS=secondary progressive multiple sclerosis; PPMS=primary progressive multiple sclerosis; N=number of participants; EDSS=expanded disability status scale

MR Experiments

MRI data were acquired on a Philips Achieva 3.0 T scanner (Philips Medical Systems, Best, The Netherlands). Scanning sequences included 48-echo GRAdient and Spin-Echo (GRASE) T₂ relaxation (TR=1073ms, TE=8ms, acquired voxel size=1x1x5mm³ with 20 slices, reconstructed voxel size=1x1x2.5mm³ with 40 slices, SENSE factor=2),²⁵ inversion recovery T₁ relaxation (TIs=150, 400, 750, 1200, 2100ms, TR=3000ms, TE=3.2ms, voxel size=1x1x2.5mm³, 40 slices), DBSI (99 directions, range of b values=0-1500, TR=4798ms, TE=79ms, voxel size= 2x2x2mm³, 40 slices),¹³ structural proton-density (PD)/ T₂-weighted (TR=2900ms, TE=8.42/80ms, voxel size=1x1x3mm³, 54 slices), and 3D T₁-MPRAGE (TR=3000ms, TE=3.5ms, TI=926, voxel size=1x1x1mm³, 150 slices).

Data Analysis

Voxel-wise T₂ distributions were calculated using a modified Extended Phase Graph algorithm combined with regularized non-negative least squares and flip angle optimization (analysis program available from: https://mriresearch.med.ubc.ca/news-projects/myelin-water-fraction/) developed at the University of British Columbia.^26,27 MWF was defined as the fraction of signal with T₂<40ms. GMT₂ of the intra/extracellular water pool was calculated for T₂ values between 40 and 200ms. T₁ was fit to a single exponential using in-house software. WC was calculated using the reference method.²⁸ DBSI data was analysed with in-house software (Matlab, The MathWorks, Inc., Natick, MA) to calculate FF, AD, RD, RF and Non-RF maps.¹³ MWF, GMT₂, T₁, WC, DBSI-derived metric maps and 3DT₁ images were registered to PD images using FLIRT (FSL toolbox).²⁹

Regions of interest

Whole brain masks were created using brain extraction from the advanced normalization tools (ANTs). NAWM masks were created using FAST³⁰ on the registered 3DT₁. Lesions were identified on the PD/T₂ images by an experienced radiologist and labeled with one or more seed points. Lesions were then automatically segmented using the seed points.³¹ The JHU atlas CC ROI was registered to the PD images using FLIRT²⁹ to obtain a total CC mask. Masks were overlaid onto registered MWF, GMT₂, T₁, WC and DBSI-derived metric maps to obtain mean measurements.

Volumetrics

Normalised brain volume (NBV) was determined by a combination of the FSL tools FAST and FIRST.^29,30 Cortical thickness (CTh) was calculated using ANTs.³² Lesion volume (LV) was defined as the summation of all the voxels within the lesion mask. Due to large differences between subtypes, the log of the LV was used in the comparison analysis.

Statistics

Differences in MR measurements between CIS, RRMS, SPMS and PPMS were calculated using an ANCOVA with age as a covariate. Subsequent pair-wise comparisons were done with a post-hoc Bonferroni correction. Adjusted p-values <0.05 were considered significant.

Results

Example images of each MRI metric for each participant subtype are shown in Figure 1. Boxplots for different regions divided by MS subtype are displayed in Figure 2 with significant differences between subtypes indicated. Mean metric values for different MS subtypes from whole brain, NAWM, lesions and corpus callosum are included as supplemental material (Supplemental Tables 1-4). Volumetric measurements for each subtype are shown in Figure 3 and a table of volume metrics is included as supplemental material (Supplemental Table 5). In the figures and tables, advanced imaging results are categorized based on metric interpretation into neuroinflammation (GMT₂, T₁, WC, RF, Non-RF), myelin (MWF, RD), and axons (FF, AD).

Figure 1:

Example images of the MR measures (proton density (PD), geometric mean T₂ (GMT₂), T₁, water content (WC), restricted fraction (RF), non-restricted fraction (Non-RF), myelin water fraction (MWF), radial diffusivity (RD), fibre fraction (FF), axial diffusivity (AD)) and for each subtype (clinically isolated syndrome (CIS), relapsing-remitting MS (RRMS), secondary progressive MS (SPMS) and primary progressive MS (PPMS)).

Figure 2:

Boxplots of each measurement over 4 regions of interest (whole brain (WB), normal appearing white matter (NAWM), lesions and corpus callosum (CC)) for the different subtypes (clinically isolated syndrome (CIS), relapsing-remitting MS (RRMS), secondary progressive MS (SPMS) and primary progressive MS (PPMS)). Lines connecting boxplots indicate significant differences with p-values of *<0.05 **<0.005 ***<0.0005.

Figure 3:

Boxplots of volumetric measurements for the different subtypes (clinically isolated syndrome (CIS), relapsing-remitting MS (RRMS), secondary progressive MS (SPMS) and primary progressive MS (PPMS)). Lines connecting boxplots indicate significant differences with p-values of *<0.05 **<0.005 ***<0.0005.

In general, a gradation of values was seen from CIS to RRMS to progressive MS (SPMS and PPMS) with numerous significant differences between CIS and progressive MS and fewer differences between CIS and RRMS, and RRMS and progressive MS.

CIS vs. RRMS:

Relative to CIS, the neuroinflammation-related marker Non-RF was increased in RRMS NAWM (+11%) and corpus callosum (+33%) and trending in RRMS WB (+2,4%, p=0.094), and water mobility in RRMS lesions was also higher with increased GMT₂ (+8.0%) and trending T₁ (+6.6%, p=0.093). Water content showed an increasing trend in RRMS CC (+1.6%, p=0.095) compared to CIS. RRMS lesions demonstrated greater myelin damage, with reduced MWF (−14.0%) relative to CIS lesions, and a trend for increased RD (+14.0%, p=0.08). RRMS CC MWF also showed a trend-level reduction in MWF (−14.6%, p=0.06) relative to CIS. Evidence of axonal damage was characterized by reduced FF in NAWM (−4.2%) and the CC (−7.8%), and increased AD in RRMS lesions (+8.5%) compared to CIS. Volumetric analysis revealed that, relative to CIS, RRMS showed reduced NBV (−3.3%), CTh (−9.7%) and LV (+900%).

CIS vs. progressive MS:

Numerous neuroinflammation, myelin and axonal-related differences were found between CIS and progressive forms of MS, primarily SPMS. GMT₂, reflecting water mobility, was increased in SPMS whole brain (+2.0%), SPMS CC (+4.8%), SPMS lesions (12.3%) and PPMS lesions (14.8%), relative to CIS. T₁, also related to water mobility, was higher in SPMS CC (+7.2%), SPMS lesions (9.9%) and PPMS lesions (12.9%). Water content was increased in SPMS CC (+2.2%) and trending in SPMS NAWM (+1.9%, p=0.06) and SPMS whole brain (+1.3%, p=0.052), compared to CIS. DBSI-derived RF, a measure of cellularity due to inflammation, was decreased in SPMS NAWM (−24%), SPMS lesions (−31%) and PPMS lesions (−40.6%) and trending in SPMS whole brain (−8.4%, p=0.06). Non-RF, linked to oedema, was higher in SPMS whole brain (+3.2%), SPMS NAWM (+16%), SPMS CC (+57%) and PPMS CC (+49%) and trending in PPMS NAWM (+14.2%, p=0.051).

MWF, a measure of myelin, was lower in SPMS whole brain (−13%), SPMS NAWM (−13%), SPMS CC (−24%), PPMS CC (−18%), SPMS lesions (−18%) and PPMS lesions (−18%), relative to matched CIS tissue. RD, the other myelin-related metric, was increased in SPMS lesions (+21%), PPMS lesions (23%) and trending in SPMS NAWM (+6.8%, p=0.09). FF, reflecting the apparent density of axons, was lower in SPMS CC (−14%) and PPMS CC (−12%), while AD, linked to residual axonal integrity, was higher in PPMS lesions (+11%) and trending in SPMS lesions (+8.6%, p=0.07).

Volumetric analysis revealed that SPMS showed reduced NBV (−6.8%) and CTh (−20%), which was also observed in PPMS CTh (−7.1%) and trended for NBV (−3.5%, p=0.06), relative to CIS. Lesion volume was increased by more than 100% in SPMS and PPMS compared to CIS.

RRMS vs. progressive MS:

Relative to RRMS, SPMS showed abnormalities in several neuroinflammation related measures with reduced NAWM RF (−17%), and trends for increased CC GMT₂ (+3.0%, p=0.09) and CC T₁ (+3.8%, p=0.07). Volumetric analysis found that relative to RRMS, SPMS showed reduced NBV (−3.6%) and trended towards a reduced CTh (−11.2%, p=0.07).

Secondary vs. primary progressive MS:

Only one trend towards significance was found, namely an increase in NAWM T₁ in SPMS compared to PPMS (+2.6%, p=0.09).

Discussion

Advanced MRI approaches with improved specificity for different aspects of tissue microstructure may help in differentiating damage due to neuroinflammation, more prominent in RRMS, from neurodegeneration, believed to be a larger contributor in progressive forms of the disease.³³ Some RRMS treatments exhibit high efficacy in reducing the risk of MS relapses and new MRI lesions, but existing treatments for reducing confirmed disability progression in inactive progressive MS have demonstrated underwhelming efficacy. Biomarkers of neurodegeneration will aid in the discovery and validation of future MS therapies needed to halt disease progression. Volumetric analysis can provide some information about tissue changes, and our findings were as expected¹⁶ with more atrophy, smaller cortical thickness and higher lesion volume in the later stages of disease. The ranking of subtypes from largest to smallest brain volume and cortical thickness was CIS → RRMS → progressive MS, with similar NBV and CTh for SPMS and PPMS. However, volume measures lack specificity to the underlying tissue loss responsible for reductions in tissue size. Development of in vivo biomarkers related to neurodegenerative process, like myelin and axon damage, would assist in the assessment of potential treatments. Our study provides evidence that non-invasive measurement of inflammation, myelin and axons can detect different degrees of damage across MS subtypes. In this study, myelin water fraction, a marker of myelin content, was able to discern the largest number of subtype tissue differences,⁷ followed closely by the neuroinflammation-related marker Non-RF,⁶ although these metrics provided complementary, rather than overlapping information.

Neuroinflammation, oedema and water content

We investigated a number of neuroinflammatory-related MR metrics including GMT₂, T₁, water content, and DBSI-derived isotropic restricted fraction (inflammatory cells) and isotropic non-restricted fraction (oedema). Relative to CIS, abnormal levels of all markers were observed to some extent for the tissue and MS subtypes examined. Non-RF provided MS subtype differentiation for NAWM and the corpus callosum, suggesting a global, increasing level of oedema from CIS to RRMS to SPMS/PPMS. While we did not observe subtype differences in lesions, Non-RF has been found to increase in MS lesions¹⁶ related to increased oedema. In normal-appearing CC, Sun et al.³⁴ measured a significantly higher Non-RF in SPMS than RRMS (p<0.01), which, although also higher in SPMS, we did not find significant (p=0.12). GMT₂, reflecting general mobility of water in intra and extracellular spaces, was sensitive to differences in lesion pathology between MS subtypes, with the highest values in progressive MS. This work is supportive of previous work in RRMS and SPMS reporting increased GMT₂ in MS NAWM and lesions, but differences between subtypes were not described.²⁰

T₁ relaxation highlighted differences in NAWM between CIS and SPMS, but not RRMS, a finding supported by Scanderberg et al., who reported significantly higher T₁ in SPMS compared to healthy controls whereas RRMS was not different.³⁵ In brain, a model of water exchange suggests that 1/T₁ is linearly proportional to 1/WC. This model was confirmed in animal tissue^36-39 as well as in MS.⁴⁰ Our measure of WC agreed with T₁ in highlighting differences between CIS and SPMS WB, NAWM and CC. Interestingly, lesion T₁ increased from CIS to RRMS to progressive MS, but this pattern was not observed for WC, suggesting that an upper limit of water content may be reached for lesions, but T₁-related increases may continue with MS progression.

Finally, RF highlighted differences between subtype NAWM with a gradation in RF from CIS to RRMS to progressive MS. Previous DBSI work in preclinical models^13,15,17 and autopsied spinal cord¹⁶ found RF correlated significantly and positively with increased cellularity. Shirani et al. found that RF within white matter lesions was one of the most important DBSI metrics for separating subtypes of MS.²² Their results suggested that RRMS WM lesions were more similar to PPMS than SPMS. Our current study showed significant differences in lesion RF between CIS and both progressive MS subtypes, possibly related to the inflammatory nature of early lesions compared to chronic lesions.⁴¹ Also, unlike a previous post-mortem study which found increased inflammation in the NAWM of progressive MS,⁴² we found a lower restricted fraction in SPMS compared to CIS as well as RRMS. Differences in the number of study participants, range of EDSS and age from previous studies may contribute to these discrepancies.

Myelin

Myelin related metrics examined were MWF and DBSI-determined radial diffusivity. MWF was particularly sensitive to subtype differences in lesion and corpus callosum myelin, and overall was able to detect more differences than RD. MWF and RD both demonstrated differences in CIS vs. SPMS NAWM, and CIS vs. progressive MS lesions. These observations are supported by previous studies which report reduced MWF in MS NAWM and lesions^18,43,44 as well as reduced NAWM myelin content detected in histopathological studies.^3,33 Our findings are also in line with a study by Oh et al who reported SPMS and RRMS with >5 year disease duration showed lower NAWM MWF compared to CIS and RRMS with <5 years disease duration.⁴³ We were unable to detect any MWF difference between RRMS and SPMS, which is in contrast to work by Dayan et al.⁴⁵ who reported lower whole brain MWF in SPMS (n=23) vs. RRMS (n=134); this discrepancy may be due to the much larger RRMS study population in the previous study. Although only a trend for subtype differences were found for CC RD, work by Shirani et al. using a nonparametric decision tree-based regression and classification method suggests that RD (as well as FF and Non-RF) in CC NAWM may be important in distinguishing MS subtypes.²² Lesions in MS participants showed lower myelin content than lesions in CIS participants. A histology study comparing lesion remyelination in different MS subtypes showed that extensive remyelination occurred in a subset of lesions from each MS subtype, however RRMS had the largest fraction of these remyelinated lesions.⁴⁶ The greater amount of remyelination in RRMS may indicate a greater capacity for repair in earlier stages of MS.

The relationship between both proposed myelin markers and myelin content are supported by pathological studies. MWF is highly correlated with histological staining of myelin with Luxol fast blue and toluidine blue,^11,46-48 providing strong evidence that MWF is a measure for myelin content. Correlations between increases in DBSI-derived RD and demyelination was found in preclinical studies,^13,15,17 post-mortem spinal cord¹⁶ and a biopsied MS lesion,²³ also supporting the relationship between RD and myelin.

Axons

Axonal descriptors stem from DBSI modelling which yielded fiber fraction and axial diffusivity. Fiber fraction was particularly sensitive to subtype differences in the corpus callosum, highlighting this measure for detection of subtle variations in regions related to axonal density. This observation is in alignment with studies which show that more progressive MS has greater axonal loss.^33,49 Shirani et al. showed that FF (along with Non-RF and RD) in CC NAWM helped differentiate subtypes using a nonparametric decision tree-based regression and classification method.²² Using this same classification method, they also showed that FF in lesions was important in differentiating MS subtypes, particularly PPMS and RRMS;²² our current study showed no difference in lesion FF between subtypes. Differences in AD were however demonstrated between CIS and PPMS and CIS and RRMS lesions, as well as a trend between CIS and SPMS lesions. Lesions from progressive MS participants showed greater axonal abnormality than RRMS participant lesions. The interpretation of both FF and AD as axonal-related measures have some histological validation support, with previous DBSI studies reporting decreased AD with acute axonal injury⁵⁰ and a significant positive correlation between FF and axonal density.^17,23

Limitations

The largest shortcoming of this study was the lack of a healthy control group with which to compare the MS participants to. Our findings are instead presented in the context of the gradation across the MS spectrum with CIS, the earliest stage of MS, serving as the baseline. While this is the largest DBSI study to date, and one of the largest T₂-based myelin water imaging reports in MS, the smaller number of participants with CIS and PPMS may have made it more difficult to detect differences in those subtypes. The different subtypes will also differ in age, disease duration and EDSS by their nature. Because age, disease duration and EDSS are intrinsically linked to the different MS phenotypes, it is very difficult to separate their effects from those of the disease course. Although we have not found significant effects of age nor sex on myelin water fraction in healthy controls, as seen in a recently published study of 100 healthy brains,⁵¹ we did account for the effect of age in our reported comparisons between subtypes.

Our inclusion of the CC ROI was intended for comparison of values from a significant white matter structure known to be affected in MS.²⁴ In the CC ROI, lesions were not excluded as we wanted to compare measurements taken along the whole structure regardless of the presence of lesions or any other subtle abnormalities. The integrity of the whole structure including both normal appearing and lesional tissue should impact CC function which will be tied to clinical performance. The fraction of CC voxels that contain lesion tends to be small and therefore the impact on the mean MR measurement should be minimal.

Finally, it is also worth noting that while the quantitative MR measures reported are classified based on metric assignments to neuroinflammation, myelin and axons, these metrics may be influenced by numerous factors which can confound results and interpretation. Brain tissue microstructure and alterations which occur from MS pathology are highly complex; a multi-modality approach like the one used here, where several different quantitative MR methods are used to assess the same tissue component (e.g. myelin with both T₂ relaxation and diffusion), is recommended to increase confidence in results and highlight discrepancies due to confounding factors.

In summary, advanced imaging can show differences between MS subtypes related to underlying tissue damage. This is the first study to use both T₂ relaxation and DBSI in MS, and the largest study human in vivo DBSI study to date. Compared to CIS, NAWM in RRMS showed demyelination, oedema and axonal loss. In progressive MS, there was even more evidence of increased oedema, demyelination and axonal loss compared to CIS and RRMS. A multi-modality approach using T₁, T₂ and diffusion basis spectrum imaging can provide information about inflammatory and neurodegenerative processes across the spectrum of MS which may be useful for assessing therapies targeting specific types of damage occurring in the disease.