Perilesional neurodegenerative injury in multiple sclerosis: Relation to focal lesions and impact on disability

Abstract

Background:

Axonal injury is the primary source of irreversible neurological decline in persons with multiple sclerosis (pwMS). Identifying and quantifying myelin and axonal loss in lesional and perilesional tissue in vivo is fundamental for a better understanding of multiple sclerosis (MS) outcomes and patient impairment. Using advanced magnetic resonance imaging (MRI) methods, consisting of selective inversion recovery quantitative magnetization transfer imaging (SIR-qMT) and multi-compartment diffusion MRI with the spherical mean technique (SMT), we conducted a cross-sectional pilot study to assess myelin and axonal damage in the normal appearing white matter (NAWM) surrounding chronic black holes (cBHs) and how this pathology correlates with disability in vivo. We hypothesized that lesional axonal transection propagates tissue injury in the surrounding NAWM and that the degree of this injury is related to patient disability.

Methods:

Eighteen pwMS underwent a 3.0 Tesla conventional clinical MRI, inclusive of T1 and T2 weighted protocols, as well as SIR-qMT and SMT. Regions of interests (ROIs) were manually delineated in cBHs, NAWM neighboring cBHs (perilesional NAWM), distant ipsilateral NAWM and contra-lateral distant NAWM. SIR-qMT-derived macromolecular-to-free pool size ratio (PSR) and SMT-derived apparent axonal volume fraction (V_ax) were extracted to infer on myelin and axonal content, respectively. Group differences were assessed using mixed-effects regression models and correlation analyses were obtained by bootstrapping 95% confidence interval.

Results:

In comparison to perilesional NAWM, both PSR and V_ax values were reduced in cBHs (p < 0.0001) and increased in distant contra-lateral NAWM ROIs (p < 0.001 for PSR and p < 0.0001 for V_ax) but not ipsilateral NAWM (p = 0.176 for PSR and p = 0.549 for V_ax). V_ax values measured in cBHs correlated with those in perilesional NAWM (Pearson rho = 0.63, p < 0.001). No statistically relevant associations were seen between PSR/V_ax values and clinical and/or MRI metrics of the disease with the exception of cBH PSR values, which correlated with the Expanded Disability Status Scale (Pearson rho = −0.63, p = 0.03).

Conclusions:

Our results show that myelin and axonal content, detected by PSR and V_ax, are reduced in perilesional NAWM, as a function of the degree of focal cBH axonal injury. This finding is indicative of an ongoing anterograde/retrograde degeneration and suggests that treatment prevention of cBH development is a key factor for preserving NAWM integrity in surrounding tissue. It also suggests that measuring changes in perilesional areas over time may be a useful measure of outcome for proof-of-concept clinical trials on neuroprotection and repair. PSR and V_ax largely foiled to capture associations with clinical and MRI characteristics, likely as a result of the small sample size and cross-sectional design, however, longitudinal assessment of a larger cohort may unravel the impact of this pathology on disease progression.

1. Introduction

Axonal injury is the main source of irreversible neurological decline in persons with multiple sclerosis (pwMS) (Mahad et al., 2015). Histopathological evidence shows that axonal pathology within lesions propagates damage in neighboring normal-appearing white matter (NAWM) through anterograde and retrograde degeneration (Dziedzic et al., 2010; Evangelou et al., 2000; Lovas et al., 2000). Identifying and quantifying myelin and axonal loss in lesional and perilesional tissue in vivo is fundamental for a better understanding of patient impairment and multiple sclerosis (MS) outcomes.

Evidence of neurodegeneration characterized by axonal and myelin injury in NAWM has been detected by several imaging studies using quantitative techniques in vivo. These methods indirectly infer on the integrity of myelin and axons. In particular, the findings of (i) a reduction in the N-acetyl-aspartate, a marker of axonal integrity quantified using magnetic resonance spectroscopy and (ii) changes in diffusion tensor imaging derived axial diffusivity in NAWM tracts provide evidence that axonal injury does occur outside focal lesions. Similarly, changes in (i) diffusion tensor imaging derived radial diffusivity, (ii) myelin water imaging derived myelin water fraction and in the (iii) magnetization transfer ratiomeasured outside lesions indirectly infer on myelin injury occurring in seemingly healthy tissue (Fu et al., 1998; Santos et al., 2002; Ciccarelli et al., 2003; Laule et al., 2004; Giacomini et al., 2009; Kolind et al., 2012; Liu et al., 2015; Brown et al., 2017; Kolasa et al., 2019). However, the relation between lesional and non-lesioned tissue damage is less clear (De Stefano et al., 2002; Ramió-Torrentà et al., 2006)

We used two advanced imaging techniques to indirectly infer on myelin and axonal integrity: (i) selective inversion recovery quantitative magnetization transfer (SIR-qMT), from which we derived the macromolecular-to-free pool size ratio (PSR), and (ii) multi-compartment microscopic diffusion magnetic resonance imaging (MRI) with the spherical mean technique (SMT), from which we obtained the axonal volume fraction (V_ax). Previous studies from our group have shown that PSR and V_ax hold promise as sensitive and more specific biometrics of myelin and axonal integrity, respectively (Janve et al., 2013; Kaden et al., 2016; Bagnato et al., 2018; West et al., 2018). We measured PSR and V_ax values in chronic black holes (cBHs), neighboring (perilesional) NAWM, distant ipsilateral NAWM and contra-lateral (distant) NAWM in pwMS and assessed their impacts on patients’ clinical impairment and radiological disease progression.

We hypothesized that lesional axonal transection propagates tissue injury in the surrounding NAWM and that the degree of this pathology correlates with disability. Therefore, we expected that PSR and V_ax values differ between perilesional and distant NAWM and that perilesional NAWM damage correlates with damage in individual cBHs as well as clinical and radiological measures of disease.

2. Matherials and methods

2.1. Standard protocol approvals, registrations, and patient consents

The study was approved by the local Institutional Review Board. A signed consent form was obtained from each subject prior to enrollment and the study followed all the Health Insurance Portability and Accountability Acts of 1996 (HIPAA) regulations for the U.S. studies.

2.2. Study design and cohort

This study was designed as a cross-sectional pilot to assess the feasibility of a full-scale investigation and inform power calculations. Eighteen pwMS (Table 1) were consecutively enrolled from the Vanderbilt MS Center (Nashville, TN) over a time span of six months, between February and August 2017. The presence of any of the following in the six months preceding enrollment was considered an exclusion criterion: clinical relapses or steroids treatment, changes in Expanded Disability Status Scale (EDSS) (Kurtzke, 1983), and active lesions on clinical scans. All study participants underwent a brain MRI and a neurologic visit to rate disability using the EDSS and the Timed 25-Foot Walking (T25-FW) (Cutter et al., 1999).

Table 1.

Demographic and Clinical Features of the Study Cohort (n=18)

Age	45.5 ± 14.6
Sex (male/female)	6/12
Ethnicity (African American/Caucasian)	2/16
MS type (CIS/RRMS/SPMS)	4/11/3
Disease duration (years)	11.5 ± 10.6
EDSS score*	1 (0 – 6.5)
T25-FW (seconds)	10:37 ± 18:12

CIS = clinically isolated syndrome; EDSS = expanded disability status scale; RRMS = relapsing remitting multiple sclerosis; SPMS = secondary progressive multiple sclerosis; T25-FW = timed 25-foot walk.

^*median (range)

2.3. Images acquisition and post-processing

A whole-body 3.0 Tesla dStream MRI scanner (Philips Healthcare, Best, The Netherlands) equipped with a 32-channel head coil (NOVA Medical, Washington, MA) was used for imaging. SIR-qMT and SMT, as well as T1-weighted (T1-w) and T2-w turbo spin echo (TSE) and fluid attenuated inversion recovery (FLAIR) clinical scans, were acquired using pulse sequence parameters detailed in Table 2. There was no interslice gap for any of the sequences. SIR-qMT had between 16 and 25 slices, depending on lesion distribution. SMT had a coverage of 50 slices and clinical scans covered the entire brain as conventionally done.

Table 2.

Pulse Sequences Parameters

SIR-qMT	SMT	T1-w / T2-w TSE / FLAIR
Scan time = approx. 1 min. / slice	Scan time = approx. 15 min	Scan time: approx.. 30 min (total)
Single-slice with TSE readout	SE EPI readout	T1-w TR= 500 ms
TE = 80 ms	TE = 74 ms	T1-w TE = 14 ms
TI = 10, 10, 278, 1007 ms	TR = 13.5 s	T2-w TR = 4000ms
TD = 684, 4171, 2730, 10 ms	2 b shells (1000 and 2500 s/mm2	T2-wTE = 90 ms
TSE factor = 26	Diffusion directions = 90	FLAIR TR = 9000 ms
Echo spacing = 5.9 ms	(45 on each b shell)	FLAIR TE = 114 ms
SENSE factor = 2.2	Multiband factor = 2	Resolution = 0.4 x 0.4 x 2
Resolution = 2 × 2 × 4 mm3	Resolution = 2 × 2 × 2 mm3	mm3

b = diffusion weighting factor; EPI = echo planar imaging; FLAIR = Fluid-attenuated inversion recovery; FFE= fast field echo; SENSE = Sensitivity Encoding; SIR-qMT = selective inversion recovery quantitative magnetization transfer imaging, SMT = multi-compartment microscopic diffusion MRI with spherical mean technique; TD = pre-delay time from end of turbo spin echo (TSE) readout until next inversion pulse; TE = echo time; TI = inversion time; TR = repetition time.

The PSR parametric maps were computed based on an optimized protocol (Dortch et al., 2018) of the SIR-qMT method (Gochberg and Gore, 2003; Dortch et al., 2011) using turbo spin-echo readouts. By fixing the MT exchange rate k_mf = 12.5 s⁻¹ (Dortch et al., 2018), four independent parameters, i.e., PSR, R_1f (the spin-lattice relaxation rate of free water pool), S_f (the inversion coefficient of the free water pool), and M_f (the magnetization of the free water pool before inversion) can be fit using a non-linear least-square fitting method (Xu et al., 2014; Dortch et al., 2018).

The V_ax maps were created using multi-compartment microscopic diffusion imaging based on the SMT (Kaden et al., 2016; Bagnato et al., 2019). Briefly, diffusion-weighted signals of the same b values were averaged over all gradient directions (i.e., spherical mean) and then a two-compartment model was fit to the signals to extract two independent parameters: V_ax, the apparent intra-axonal volume fraction, and D_ax, the apparent axial intra-axon diffusivity. We focused on V_ax as previous studies from our group have found that it is a sensitive indicator of axonal integrity (Bagnato et al., 2019).

All diffusion-weighted images were corrected for both eddy current and susceptibility-induced distortions using the topup and eddy toolbox in FSL (https://fsl.fmrib.ox.ac.uk/fsl/). All parametric maps, i.e. PSR and V_ax, T2-w FLAIR and T1-w SE images were co-registered to a high-resolution T2w TSE anatomical image. A linear registration approach based on mutual information cost function and six degrees of freedom was used in FSL FLIRT. Co-registration accuracy was assessed through a visual inspection of parametric maps overlaid on anatomical images.

2.4. Image analysis

2.4.1. PSR and V_ax measurements of cBHs, perilesional NAWM and distant NAWM

Regions of interest (ROIs) were manually drawn to capture (i) cBH lesions, (ii) areas adjacent to each cBH, referred to as perilesional NAWM hereafter, (iii) distant ipsilateral areas, which were at least 30 mm from perilesional NAWM ROIs and (iv) NAWM contralateral to each perilesional NAWM, referred to as distant NAWM hereafter. Care was taken to avoid areas of distant ipsilateral and contralateral NAWM which were in dose proximity to other T2-lesions or cBHs. Data from five out of 18 patients (four persons with relapsing remitting MS and one with secondary progressive MS) were excluded from final analysis. The reasons for exclusion were as follows: one patient had no cBHs, one patient had severe artefacts on the PSR image that prevented the accurate assessment of the anatomical location of NAWM ROIs and three patients had bilateral or midline cBHs, which precluded creating contralateral distant NAWM ROIs.

We illustrate example ROI placements in Fig. 1. cBH ROIs (n = 484) were manually drawn to include the entire BH area on individual slices. For each cBH, perilesional, distant ipsilateral and distant contralateral NAWM, ROIs were drawn, making a total of 1936 individual ROIs. The size of the perilesional, ipsilateral and distant NAWM ROIs were identical to each other, however, the size of these ROIs differed between lesions depending on the anatomical location and adjacent parenchymal tissues (Riva et al., 2009). The size of the ROIs ranged from 8 to 168 voxels [mean ± standard deviation (SD) = 46.1 ± 23.7 voxels, median= 40 voxels]. The number of ROIs per case varied depending on cBH lesion burden and ranged from 6 to 55 (mean ± SD = 29.2 ± 15.5, median = 27).

Fig. 1. Region of interest (ROI) placement.

ROIs were drawn around cBHs (red), perilesional NAWM (blue), ipsilateral NAWM (orange) and distant NAWM (green) on (A) T1-weighted, (B) T2-weighted FLAIR, (C) V_ax and (D) PSR parametric maps. The corresponding V_ax and PSR signal intensity values are visible on the left (V_ax) and right (PSR). The size of each of the corresponding perilesional, ipsilateral and distant NAWM ROIs were identical. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

2.4.2. Lesion volume computations

The entire white matter lesion burden on T2-w and on T1-w TSE sequences was manually segmented using graphic tools available in MIPAV (http://mipav.cit.nih.gov/). Thereafter, we used MATLAB (Mathworks, Natick, MA) to generate a subtraction mask that removed all cBHs (defined as lesions, which were simultaneously hypointense on T1-w TSE and hyperintense on T2-w images) (Bagnato et al., 2003) from the masks of the white matter lesions measured on T2-w TSE only. This subtraction mask, hereby defined as T2-lesion mask, was obtained to ensure that cBHs were not computed on the T2-lesion burden. cBHs and T2-lesion masks were used to compute lesion volumes.

2.4.3. Brain atrophy assessment

Brain parenchymal fraction (BPF) was computed as a surrogate measure of brain atrophy (Bermel and Bakshi, 2006). For each subject, we used the T1-w TSE and T2-w FLAIR, which were corrected for inhomogeneity using the N4 algorithm (Tustison et al., 2010), before each of the images was segmented using a publicly available tissue probability atlas (https://www.nitrc.org/projects/imc_brain_atlas/). The atlas T1-w image was then registered to each of the bias-corrected T1-w images using rigid, affine and deformable registration, performed using the normalized cross-correlation metric (patch radius: 4 × 4 × 4 voxels) of the greedy registration package (https://github.com/pyushkevich/greedy) (Xie et al., 2018). The resulting transformations were applied to the atlas-based tissue probability maps before being applied to the subject’s individual T1-w image space. Voxel-level tissue classification was then performed using an expectation maximization algorithm (https://www.nitrc.org/projects/abc/) (Leemput et al., 1999). Finally, for each subject, a skull-strip mask was created from the combined tissue labels and used to bootstrap the registration. The process of subject-to-atlas registration was repeated with the use of the skull-strip mask in order to constrain the registration, while tissue segmentation was performed based on the results of the bootstrapped registration. Lesions were manually filled and the BPF was calculated as the ratio of the total volume of gray matter, white matter, cerebellum, brainstem and lesions to the total volume of gray matter, white matter, cerebellum, brainstem, lesions and cerebrospinal fluid (Rudick et al., 1999).

2.5. Statistical analysis

Data are expressed as mean ± SD for demographic data and mean ± standard error (SE) for PSR and V_ax. For each non-categorical clinical measure, the mean (or, if appropriate, the median) was used to divide patients into two groups, which were subsequently compared. Throughout the entire statistical analysis, we assumed that MRI measures were correlated within the same subject. On the basis of this assumption: (i) differences in PSR and V_ax values between different tissue types (using perilesional NAWM ROIs as the reference) and (ii) differences in PSR/V_ax and clinical measures (sex, ethnicity, MS type, disease duration, EDSS score and T25-FW) in all different tissue types were assessed using a mixed-effects model with subject as a “random effect”. Additionally, Pearson correlation analysis with its bootstrapping 95% confidence interval (CI) estimated correlations between variables.

All statistical tests were two-sided and a p value ≤ 0.05 was considered statistically significant. Statistical analyses were performed using statistical software R (version 3.4.0, R Foundation, Vienna, Austria) and SAS (version 9.4).

3. Results

3.1. Differences in PSR and V_ax values between cBHs, perilesional and ipsilateral and contralateral distant NAWM

In comparison to the reference group, i.e. perilesional NAWM, both PSR (Fig. 2a) and V_ax (Fig. 2b) values were significantly different in cBHs and distant NAWM. Specifically, compared to perilesional NAWM (0.109 ± 0.002), PSR values were reduced in cBHs (0.062 ± 0.002, p < 0.0001) and increased in distant NAWM (0.120 ± 0.002, p < 0.001). Although a trend toward higher values was seen in distant ipsilateral NAWM (mean = 0.130 ± 0.002, p = 0.176), group differences did not reach statistical significance. Similarly, compared to perilesional NAWM (0.398 ± 0.027), V_ax values were reduced in cBHs (0.280 ± 0.002, p < 0.0001), increased in distant NAWM (0.476 ± 0.002, p < 0.0001) and did not differ significantly in distant ipsilateral NAWM (0.432 ± 0.01, p = 0.549).

Fig. 2. PSR and V_ax values in the four tissues studied: cBHs, perilesional (reference group), ipsilateral NAWM and distant NAWM.

(A) PSR values in cBH, perilesional, ipsilateral and distant NAWM; (B) V_ax values in cBH, perilesional, ipsilateral and distant. Box plots represent mean and interquartile range, whiskers represent 95% confidence interval. Asterisks over each bar represent significance level (** = p < 0.01, *** = p < 0.001).

In comparison to perilesional NAWM, the mean percentage change in PSR values in the different tissues was: −43.30 for cBHs, 15.7 for ipsilateral NAWM and 10.19 for distant NAWM. The percentage change in V_ax values in the different tissues in comparison to perilesional NAWM was: −29.77 for cBHs, 11.3 for perilesional NAWM and 19.63 for distant NAWM.

Due to the lack of significant differences between perilesional and ipsilateral NAWM, we excluded the latter from subsequent analyses and focused only on the three tissue types: cBHs, perilesional NAWM and distant (contralateral) NAWM.

3.2. Associations between PSR and V_ax values

None of the PSR and V_ax values were significantly correlated with each other within the cBHs, perilesional and distant NAWM. However, V_ax values measured in cBHs correlated with those in perilesional NAWM (Pearson rho= 0.63 and 95% CI = 0.29 – 0.9, p < 0.001) (Fig. 3).

Fig. 3. Association between cBHs and perilesional V_ax..

V_ax values measured in cBH correlated with those in perilesional NAWM (Pearson rho = 0.63 and 95% CI = 0.29 – 0.9, p < 0.001).

3.3. Associations between MRI and clinical metrics

There were no significant associations between any of the lesional and non-lesional PSR and V_ax measures and other clinical or radiological metrics of disease with the exception of cBHs PSR values, which correlated with the EDSS score (Pearson rho= −0.63, and 95% CI= −0.88, −0.12, p = 0.03) (Table 3).

Table 3.

Association between PSR and with other measures of the disease

	Years of disease	EDSS	T25-FW	Lesion volume		BPF
				T2-Lesion	cBH
cBH
PSR	−0.53	−0.63 (p=0.03)	−0.55	−0.38	−0.33	−0.15
Vax	0.12	0.35	−0.01	−0.24	−0.06	−0.15
Perilesional NAWM
PSR	−0.28	−0.20	−0.23	−0.36	−0.20	−0.45
Vax	0.19	0.35	−0.05	−0.16	0.00	−0.24
Distant NAWM
PSR	−0.47	−0.33	−0.40	−0.48	−0.29	−0.17
Vax	0.32	0.39	0.01	0.14	0.14	−0.19

Values are represented as Pearson correlation coefficient (r) and statistically significant correlation values with the corresponding p value are indicated in bold.

BPF = brain parenchymal fraction; cBH = chronic black hole; EDSS: Expanded Disability Status Scale; NAWM = Normal Appearing White Matter; PSR = macromolecular-to-free pool size ratio; T25-FW = timed 25-foot walk test; V_ax = apparent axonal volume fraction.

Exploratory analyses were performed to look for differences in cBH, perilesional and distant NAWM PSR and V_ax values in patients at different stages of the disease in terms of MS phenotype and disease duration. We found a decreasing trend for PSR values in the distant NAWM with longer disease duration (over 12 years, p = 0.07) (Fig. 4).

Fig. 4. Distribution of PSR values in distant NAWM in patients with different disease durations.

Patients were split into two groups based on the median disease duration: ≤12 years (n = 6) and >12 years (n = 7). Y-axis represents PSR values derived from distant NAWM. Box plot represents mean and interquartile range, whiskers represent 95% confidence intervals.

4. Discussion

The results of our investigations show that compared to perilesional NAWM, indices of myelin and axonal integrity, i.e., PSR and V_ax respectively, are significantly decreased in cBHs but increased in distant contralateral NAWM. Moreover, the degree of axonal injury in the NAWM surrounding cBHs correlates with that within the lesion. We largely failed to find significant associations between this pathology and clinical measures but we did find evidence for a significant negative correlation between cBH PSR values and EDSS. Altogether the findings provide an in vivo demonstration of the well-known pathological notion that lesional axonal transection affects the neighboring, normal appearing tissue via retrograde and anterograde degeneration.

While our findings parallel an established pathological concept (Trapp et al., 1998; Evangelou et al., 2000; Moll et al., 2011), the novel aspects of our study compared with previous literature investigating this topic (Fu et al., 1998; Ciccarelli et al., 2003; Liu et al., 2015) are the detailed analytical approach and the use of innovative advanced imaging methods. In comparison with other previously reported techniques, SIR-qMT-derived PSR and SMT-derived Vax offer crucial advantages with regards to increased pathological specificity. Being sensitive only to the bound pool of macromolecules, PSR provides a more accurate and specific measure of myelin content compared to other techniques, such as MTR (2013; Bagnato et al., 2018; Dortch et al., 2018). Similarly, SMT overcomes the problem of fiber orientation dependence, thus increasing signal-to-noise ratio and pathological specificity compared to other diffusion-based techniques (Kaden et al., 2016; Bagnato et al., 2019).

4.1. Myelin and axonal damage in cBHs, neighboring tissue and distant NAWM

cBHs are characterized by permanent axonal damage resulting from acute demyelinating events and some chronic active inflammation (Trapp et al., 1998; van Walderveen et al., 1998). Previous studies found that between 70 and 80% of axons within chronic lesions are lost (Tallantyre et al., 2009; Frischer et al., 2009; van Walderveen et al., 1998) leading to perilesional damage in the form of retrograde and anterograde neurodegeneration. In this study, we show that both demyelination and neurodegeneration extend into the NAWM surrounding cBHs, which can be detected using PSR and V_ax. Moreover, the significant correlation between cBH and perilesional NAWM V_ax values suggests that the extent of remote neurodegeneration, but not demyelination, is a function of the level of axonal loss within cBHs.

Our findings have several important implications for our understanding of MS and its effective treatment. First, our data suggest that preventing cBH development is key for preserving the axonal integrity of the surrounding NAWM, irrespective of demyelination. This highlights the importance of effectively treating active lesions, irrespective of associated symptoms, at the time when the blood brain barrier breakdown occurs, to minimize the likelihood of conversion into cBHs. While abating inflammation is not conceptualized as a neuroprotective intervention, preventing its downstream consequences is. The latter is even more important if one considers that once formed, cBHs are unlikely to be affected by current disease modifying agents. Previous studies have indeed shown that once a cBH has formed, currently available disease modifying agents are unlikely to act upon it. For example, cBHs persist for the same amount of time irrespective of whether they originated during natural history or with treatment (interferons) (Bagnato et al., 2005). As such, effectively preventing their development is critical in order to prevent injury propagation in neighboring areas.

Secondly, our findings suggest that perilesional NAWM areas are considerably more vulnerable to disease pathologies, compared to distant ones. Changes in these specific areas could therefore provide a biometric of outcome measures in clinical trials on neurodegeneration and repair; i.e., perilesional damage is likely to be more informative and measurable than injury in whole NAWM or normal appearing gray matter structures.

4.2. Associations between PSR/V_ax and clinical or other radiological metrics of disease

Our study failed to show significant correlations between lesional and non-lesional PSR and V_ax values and clinical or radiological measures of the disease, with the exception of a moderate negative correlation between myelin integrity within cBHs and EDSS, which we also reported previously (Bagnato et al., 2020). Our data also showed a trend of decreasing PSR values in distant NAWM, with longer disease duration reflecting a widespread reduction in myelin integrity over time.

Two possible considerations, that are not mutually exclusive, may explain our findings. First, the cohort of patients included in this study consisted mainly of minimally disabled patients, with only three patients in the progressive stage of the disease. As demyelination is thought to predominate in the earlier disease stage with neurodegeneration becoming more prominent later on, the characteristics of our cohort likely explain our lack of significant associations, especially with V_ax. With a larger cohort, we expect to have been able to show the anticipated associations between PSR and V_ax and clinical and MRI metrics. Alternatively, it might also be possible that V_ax values might lack the sensitivity to unearth clinical associations in a cross-sectional setting.

Similarly, we did not find significant correlations between PSR and V_ax, and between these two metrics and lesion load. A large body of evidence provides support for axonal transection being a direct result of active demyelination (Evangelou et al., 2000; Kutzelnigg and Lassmann, 2005); alternatively, evidence exists that the two processes are, at least partially, independent (Bitsch et al., 2000; De Stefano et al., 2002). While our sample size might have led to our negative findings, this study appears to provide supporting evidence for the latter theory. This points to a more complex etiology of MS pathology and suggests that a multi-faceted approach to treatment, which targets both demyelination and axonal loss, is likely needed to successfully treat MS pathology.

4.3. Study limitations and conclusions

Our study suffered from a few limitations which must be discussed in order to derive accurate conclusions. First and foremost, being a pilot investigation, our sample of pwMS was relatively small and consisted mainly of minimally disabled persons with relapsing remitting MS, which likely contributed to our lack of significant associations in this study. Moreover, we only studied cBHs and as such did not capture any potential changes in PSR and V_ax associated with blood-brain barrier breakdown and with T2-lesions. This, however, this was outside the scope of our current study.

Despite these limitations, our data provide an elegant in vivo demonstration for retrograde and anterograde trans-synaptic degeneration in the perilesional area surrounding cBHs, that can be detected using PSR and V_ax. This has important therapeutic implications advocating for cBHs prevention, which in turn, can help preserve axonal integrity in the surrounding NAWM. It also suggests that perilesional areas are those where occult NAWM can be more easily detected and measuring changes in these regions across time may be a useful measure of outcome for proof-of-concept clinical trials on neuroprotection and repair. Whilst we failed to report significant associations between PSR and Vax and many clinical metrics in this exploratory, pilot study, it is likely that longitudinal studies will be more informative than a cross-sectional analysis, as presented here.

Abbreviations:

BPF

brain parenchymal fraction

cBH

chronic black hole

CI

confidence interval

EDSS

expanded disability status scale

FLAIR

fluid attenuation inversion imaging

GM

gray matter

MRI

magnetic resonance imaging

NAWM

normal-appearing white matter

PSR

macromolecular-to-free pool size ratio

pwMS

persons with multiple sclerosis

ROI

region of interest

SD

standard deviation

SE

standard error

SIR-qMT

selective inversion recovery quantitative magnetization transfer

SMT

multi-compartment diffusion

MRI

with the spherical mean technique

TSE

turbo spin echo

T1-w

T1-weighted

T2-w

T2-weighted

T25FW

timed 25-foot walking

V_ax

axonal volume fraction