Longitudinal ultra-high field MRI of brain lesions in neuromyelitis optica spectrum disorders

Abstract

Background

In neuromyelitis optica spectrum disorder (NMOSD), clinical disability in NMOSD patients is relapse-related and progressive phase is rare. This observation raises the question whether there is any radiographic disease activity. The aim of present study was to determine the longitudinal changes in cerebral lesion number, lesion size, lesion-to-venule relationship, and morphological patterns of lesions in NMOSD using multiparametric 7T MR imaging. We also aimed to assess brain volume changes in NMOSD.

Methods

A cohort of 22 patients with NMOSD underwent high-resolution 3D-susceptibility weighted imaging (SWI) and 2D-gradient-echo (GRE-T2*) weighted imaging on 7T MRI of brain at baseline and after ~2.8 years of follow-up. Morphologic imaging characteristics, and signal intensity patterns of lesions were recorded at both time points. Lesions were classified as “iron-laden” if they demonstrated hypointense signal on GRE-T2* images and/or SWI as well as hyperintense signal on quantitative susceptibility mapping (QSM). Lesions were considered “non-iron-laden” if they were hyperintense on GRE-T2*/SWI and isointense or hyperintense on QSM. Additionally, fractional brain parenchymal volume (fBPV) was computed at both time points.

Results

A total of 169 lesions were observed at baseline. At follow-up, 6 new lesions were found in 5 patients. In one patient, a single lesion could not be detected on the follow-up scan. No appreciable change in lesion size and vessel-lesion relationship was observed at follow up. All lesions demonstrated hyperintense signal intensity on GRE-T2* weighted images and isointense signal on QSM at both time points. Therefore, these lesions were considered as non-associated with iron pathology. Additionally, no significant change in brain volume was observed: fBPV 0.78 ± 0.06 at baseline vs. 0.77 ± 0.05 at follow up, p>0.05.

Conclusion

Cerebral lesions in NMOSD patients remain ‘inert’ and do not show any substantial variations in morphological characteristics during a 2–3-year follow-up period.

1. Introduction

Neuromyelitis Optica spectrum disorders (NMOSD) is a relapsing inflammatory disorder of central nervous system (CNS) characterized by attacks of optic neuritis, longitudinally extensive transverse myelitis (Wingerchuk et al., 2007; Tackley et al., 2014) and several other less common, but well-described clinical syndromes (Wingerchuk et al., 2017). Antibody targeting astrocytic aquaporin-4 (AQP4) water channels is a highly specific marker of NMOSD, but at least 10–20% of NMOSD patients are AQP4 antibody (AQP4-ab) seronegative (Jiao et al., 2013). Aquaporin-4 seropositive NMOSD is considered as a primary astrocytopathy since aquaporin-4 (AQP4) water channels in CNS are mainly concentrated on the foot processes of astrocytes (Fujihara K, 2011; Lucchinetti et al., 2014; Lennon et al., 2004). MR imaging is used to support the clinical diagnosis, characterize disease activity, monitor treatment response and advance understanding of the mechanisms of tissue damage in NMOSD (Sinnecker et al., 2012; Cabrera-Gomez JA, 2012; Iyer et al., 2014). Typical brain MR imaging findings in NMOSD include diencephalic and brainstem lesions (Pittock et al., 2006; Kremer et al., 2014), as well as periependymal and subcortical white matter lesions that are usually different from multiple sclerosis (MS) lesions (Huh et al., 2014; Kim et al., 2015).

Ultra-high field (7T) MR imaging has the potential to shed light on patho.genetic and pathophysiologic aspects of NMOSD by virtue of markedly improved venous and iron contrast within different tissue compartments (Sinnecker et al., 2015). Using 7T MR imaging, several studies have shown that a vast majority of MS lesions contain a central venule (Ge et al., 2008; Sinnecker et al., 2012, 2019). In contrast, few NMOSD lesions have central vein sign (Kister et al., 2013; Sinnecker et al., 2012). Additionally, significantly higher degree of abnormal iron deposition has been observed in deep gray-matter regions in MS compared to those of NMOSD patients (Chen et al., 2012). Moreover, NMOSD lesions are not associated with abnormal iron accumulation, while most MS patients will have at least some iron-containing lesions (Chawla et al., 2016; Sinnecker et al., 2016; Schumacher et al., 2016). Collectively, these studies suggest that iron-related pathology may differentiate in these two diseases. Perhaps, absence of pathologic iron deposition in NMOSD is relevant to the observation that progressive phase is very rare in NMOSD, unlike in MS (Wingerchuk et al., 2007b).

In a longitudinal multiparametric 7T MRI study of MS lesions over a mean period of 2.4 years, we observed morphological changes in some MS lesions, and with patients developing new, mostly iron-laden lesions during follow-up (Chawla et al., 2018). It would be instructive to compare these findings with progression of NMOSD cerebral lesions on ultra-high field MRI in order to better understand the pathophysiologic processes that drive NMOSD disease. Therefore, our objective in the present study was to investigate the longitudinal changes in lesion size, lesion-to-venule relationship, and morphological patterns of lesions using multiparametric 7T MR imaging in a cohort of patients with NMOSD.

2. Materials and methods

2.1. Subjects

This retrospective, longitudinal study was conducted at two academic MS referral centers - New York University Medical Center, New York, and Charité - Universitätsmedizin, Berlin. Both sites received approval from local Institutional Review Boards. Written informed consent was obtained from all patients prior to study entry. All patients met upon the international consensus diagnostic criteria (Wingerchuk et al., 2015) for NMOSD. Twenty-two NMOSD patients (age 47.8 ± 12.9 years; 21 females and one male; mean disease duration 10.3 ± 7.7years, range = 0.5–26.0years) were enrolled in this study. Nearly all patients (21/22) were seropositive for AQP4-Ab as evaluated by the similar cell-based assay. All patients underwent MR imaging at baseline and at a follow-up period. The time interval between the two MR imaging sessions ranged from 1.3 to 4.3 years (mean = 2.8 ± 0.80 years). None of the patients had clinical attacks involving brain stem or cerebellum during follow up. However, 8/22 (36%) patients had one and 3/22 (14%) patients had 2 clinical attacks involving optic nerve or spinal cord during the follow-up period. Additionally, no patient had any brain manifestations during the course of disease. The demographic and clinical characteristics of patients at baseline are summarized in Table 1.

Table 1.

Demographic and Clinical Characteristics of NMOSD Patients at baseline.

ID	Characteristics
1.	Total Number of Patients	22
2.	Gender	Male=1; Females=21
3.	Mean Age at Baseline MRI	47.8 ± 12.9years (29–75years)
4.	Disease Duration	10.3 ± 7.7years (0.5–26.0years)
5.	AQP4 Ab Positive	21 (95%)
6.	Mean Number of Optic Neuritis	2.4 ± 1.9 (range, 0–6)
7.	Mean Number of Transverse Myelitis	2.2 ± 1.6 (range, 0–4)
8.	Disability at Follow-up
	Ambulatory	17 (77%)
	Assistive Device	4 (18%)
	Wheelchair	1 (5%)
9.	Treatment at Follow-up
	Oral Immunosuppressants	18 (82%)
	Rituximab	8 (36%)
10.	Co-morbid Diseases
	Autoimmune Disease	5 (23%)
	Vascular Risk Factors	6 (27%)
11.	Mean Duration between MRI scans	2.8 ± 0.8years (1.3–4.3years)
12.	Number of Relapses involving Optic Nerve or	Spinal Cord between MRI Scans
	0	11 (50%)
	1	8 (36%)
	2	3 (14%)
13.	Number of Relapses involving Cerebrum, Brain Stem or Cerebellum between MRI Scans	None (0%)

2.2. Ultra-high field MRI and data post-processing

All patients underwent ultra-high-field MR imaging using identical whole-body 7T human MR systems (MAGNETOM, Siemens Medical Solution, Erlangen, Germany) equipped with 24-channel phased array coil (Nova. Medical, Inc., MA) both at baseline and follow-up period at both sites. The imaging protocol included a high-resolution axial 2D-gradient-echo (GRE-T2*) weighted imaging, high-resolution axial 3D-susceptibility-weighted imaging (SWI), T2-weighted fluid-attenuated inversion recovery (FLAIR), and sagittal T1-weighted 3D-MPRAGE sequences using sequence parameters reported previously (Chawla et al., 2016). In brief, the acquisition parameters for GRE-T2*weighted imaging were: TR/TE = 580/25 ms, flip angle = 35°, slice thickness = 2 mm, FOV = 240 × 240 mm², voxel size = 0.2 × 0.2mm², for sagittal 3D-SPACE-FLAIR TR/TE/ TI = 8000/380/2100 ms; isotropic voxel size = 1.0 × 1.0 × 1.0mm³ and for sagittal T1-weighted 3D-MPRAGE, TR/TE/TI =2000/2.92/1100 ms, isotropic voxel size = 1.0 × 1.0 × 1.0mm³

High-resolution, flow-compensated 3D-SWI images were acquired with the following parameters: TR/TE = 27/18 ms, flip angle=18°, slice thickness = 2 mm, FOV = 240 × 240mm², base resolution = 1024, voxel size = 0.2 × 0.2 × 2mm³, bandwidth = 110 Hz/px, acquisition time = 7:49 min, and iPAT factor = 2.

Identical imaging parameters were used in both sites for all the sequences. However, for 3D-SWI sequence, a voxel size of 0.2 × 0.2 × 2mm³ was used at NYU site whereas a voxel size of 0.5 × 0.5 × 2mm³ was used at the German site. To avoid, susceptibility artifacts from air-tissue interfaces, only supratentorial brain regions were covered while acquiring 2D-GRE-T2* and 3D-SWI. None of the patients received intravenous contrast agents. None of the patients received intravenous contrast agent at both time points.

The source magnitude and phase images from each SWI scan were obtained and were used to generate SWI venograms. All phase images were reconstructed and corrected for field inhomogeneities with Hanning high-pass filter of size 96 × 96 using SPIN image processing software (signal processing in MR http://www.mrc.wayne.edu). The original magnitude image was multiplied by the phase mask four times in order to enhance the visibility of lesions and venous structures. Finally, SWI venograms were created by performing minimum intensity projection (mIP) over 2 contiguous slices (4 mm thick).

Susceptibility weighted imaging and mapping (SWIM) algorithm developed by Haacke’s group (Haacke et al., 2010; Liu et al., 2013) was used to reconstruct quantitative susceptibility mapping (QSM) maps from high-resolution 3D-SWI data. The SWIM allows the following post-processing steps: skull stripping to remove the artifacts caused by skull and brain tissue interface using the brain extraction tool, followed by phase unwrapping using a Laplacian operator. To remove background field inhomogeneity, a variable high-pass filter of 32 pixels size was applied and, finally, inverse filtering was performed to generate QSM maps.

2.3. Data analysis

All MR images were analyzed using Image J software (imagej.nih.gov/ij/). Supratentorial lesions were analyzed side-by-side on axial GRE-T2*, 3D-SWI and QSM images. For each lesion, the following morphological imaging characteristics were recorded: 1) lesion count, 2) largest cross-sectional diameter, 3) location, 4) the presence of one or multiple central intralesional venules, and 5) signal intensity pattern within the lesions. Based upon the largest cross-sectional diameter, lesions were classified as small (<5 mm) or large (>5 mm). Lesions were classified as “iron-laden” if they demonstrated hypointense signal on T2*-weighted GRE images and/or SWI as well as hyperintense signals on QSM. Lesions were considered “non-iron-laden” if they were hyperintense on GRE-T2*/SWI and isointense or hyperintense on QSM. All MR images (GRE-T2, SWI and QSM) were analyzed and interpreted at both time points by two investigators (IK and SC) jointly. These investigators worked together in achieving the final consensus. In event of disagreement in evaluating data, a senior author (YG, with over 22 years of experience in the field of advanced MR imaging in studying MS/NMOSD diseases) was consulted.

Additionally, the brain extraction tool (BET) (Smith SM, 2002) and the automated segmentation tool from the FMRIB Software Library (Zhang et al., 2001) was used to segment gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) regions and to measure corresponding tissue volumes using T1-weighted 3D-MPRAGE images. To assess brain atrophy from patients at both time points, the fractional brain parenchymal volume (fBPV) was computed as the ratio of brain parenchymal volume (GM + WM) to the total intracranial volume (GM + WM + CSF) as described previously (Marshall et al., 2016).

2.4. Statistical analyses

Mann-Whitney U tests were performed to look for differences in fBPV and cross-sectional diameters of NMOSD lesions between baseline and follow-up period. A probabilistic (P) value of less than 0.05 was considered significant. All statistical analyses were performed using a statistical package, SPSS for Windows (v. 18.0; Chicago, IL).

3. Results

A total of 169 cerebral lesions were observed in 22 patients at baseline (mean = 7.68; median = 8/patient, range = 0–50 lesions in patients). At follow-up, 6 new lesions were found in 5 patients. Representative images from a patient demonstrating two new lesions at follow-up (not visible at baseline) are shown in Fig. 1. In one patient, a single lesion could not be detected on the follow-up scan. Altogether, a total of 174 lesions (mean = 7.9; median = 8/patient) were observed after a mean follow-up period of 2.8 years, ~3.0% increase from baseline. Distribution of lesions by size at baseline was as follows: 126 (74.6%) were <5 mm in largest diameter, and 43 (25.4%) were >5 mm in diameter. The mean cross-sectional diameters for NMOSD lesions were 3.64 ± 1.51 mm at baseline and 3.67 ± 1.55 mm at follow-up. We did not observe a significant (p>0.05) change in size of pre-existing lesions during follow up (Fig. 2).

Fig. 1.

An axial GRE-T2* weighted image at follow-up period shows two hyperintense lesions (NL1 and NL2; white arrows) with isointense signal intensity on corresponding QSM maps consistent with non-iron related pathology. These lesions are not visible at baseline on axial GRE-T2* weighted image.

Fig. 2.

Axial GRE-T2* weighted images show two hyperintense lesions (L1 and L2; white arrows) at baseline. These lesions showed similar morphological characteristics and signal intensity pattern on GRE-T2* weighted images at follow-up suggesting these lesions remained inert during a period of 2.62 years.

Distribution of lesions by location was as follows: subcortical white matter – 128 (75.7%), juxtacortical white matter – 16 (9.5%), periventricular white matter – 10 (5.9%), corpus callosum – 10 (5.9%) and basal ganglia – 5 (3%). The majority of the new lesions (3/6, 50.0%) were located in the subcortical white matter region. Central venule was identified in 19/169 (11.2%) of lesions at baseline and was still visible in 16 out of these 19 lesions at the follow up (84.2%). None of the six new lesions contained a central venule and none of the lesions without a central venule at baseline contained a central venule at follow up.

At baseline, all lesions demonstrated hyperintense signal intensity relative to the surrounding brain parenchyma on GRE-T2* weighted images. All these lesions showed isointense signal on QSM and hence were inconspicuous on these images. By definition, as described in Method section, none of the lesions were associated with abnormal iron deposition. There was no appreciable change in the signal intensity pattern within the lesions and all of them remained non-iron laden at follow-up. Moreover, none of the 6 new NMOSD lesions had evidence of iron-related signal alterations.

No significant decrease in brain volume was observed during the observation period: fBPV 0.78 ± 0.06 at baseline vs. 0.77 ± 0.05 at last follow up, p>0.05.

4. Discussion

In this pilot longitudinal study, morphological and structural features of lesions from NMOSD patients were followed over an average period of 2.8 years using ultra-high field GRE-T2*, SWI, and QSM. Several important observations were made. Firstly, there was no appreciable change in lesion size and lesion-vessel relationship throughout the duration of follow-up in these NMOSD patients, and only a modest increase in lesion number (3% increase). Secondly, none of the lesions were associated with iron-related pathology at baseline or at follow-up. Thirdly, new lesions were also not associated with iron-related pathology. Collectively, these findings suggest that cerebral lesions seen in NMOSD appear to be ‘inert’ on multi-parametric high-field MRI. The observed lesion characteristics are consistent with demyelination, edema, micro-necrosis, and gliosis (Chawla et al., 2016). Alternatively, these lesions might be due to non-NMOSD pathology, e.g. microvascular disease and age-related changes.

In contrast to NMOSD lesions herein described, MS patients exhibited four distinct morphological patterns of MS lesions based upon signal intensity on GRE-T2*/SWI and QSM images (Chawla et al., 2018). The majority of MS lesions were associated with non-iron related pathology – similar to NMOSD lesions, however, a minority of MS lesions contained iron in either nodular or ring-like form. Our findings were concordant with previous reports (Haacke et al., 2009; Bagnato et al., 2011; Mehta et al., 2013; Sinnecker et al., 2016) that have documented abnormal iron accumulation in both acute and chronic active phases of lesion development in MS. Taken together, these studies indicate that dysregulation of iron metabolism is a feature of MS, but not of NMOSD.

Longitudinal studies (Dal-Bianco et al., 2017; Bian et al., 2013; Zhang et al., 2016) tracking the evolution of MS lesions using high-field MRI have reported temporal variations in morphological features and signal intensity pattern in- only a few lesions. Similarly, in our recent longitudinal study in MS (Chawla et al., 2018), a vast majority of iron and non-iron laden lesions at baseline did not show any obvious change in the morphological characteristics during a follow-up course of ~2.4 years. However, a few non-iron laden lesions baseline exhibited iron accumulation at follow-up possibly suggesting a dynamic phenotype alteration of M2-polarized macrophages to iron-enriched M1-polarized population of cells in these lesions (Martinez et al., 2009).

We did not find significant changes in whole-brain atrophy between the two time points. The latter finding is in agreement with prior studies (Liu et al., 2018; Matthews et al., 2015) that have reported no significant changes in whole-brain or thalamic atrophy in NMOSD patients during a follow-up period of one year. Taken together with the results from the prior studies, our finding suggests very limited radiographic evidence of brain neurodegeneration over a few-year period in treated NMOSD patients. On the other hand, investigators using proton MR spectroscopy (Pichiecchio et al., 2012; Bichuetti et al., 2008; Aboul-Enein et al., 2010), magnetization transfer imaging and diffusion imaging techniques (Filippi et al., 1999; Pichiecchio et al., 2012; Finke et al., 2016) have reported that NMOSD is associated with decreased metabolic activities and subtle alterations in the tissue microstructures in the short-term. These cross-sectional studies support the notion that pathophysiologic changes underlying radiographic evolution of NMOSD lesions in the brain are associated with a slow process. The absence of overt changes in morphological characteristics during follow up confirms the clinical impression that disability in NMOSD is almost entirely due to relapses and that NMOSD very seldom follows a secondary progressive course (Wingerchuk et al., 2007, 2006).

There were some limitations of our study. Sample size was relatively small due to both relative rarity of NMOSD and technical difficulties associated with ultra-high field MR imaging, which precluded us from enolling more patients. All of our patients exhibited similar kind of ‘non-specific’, mostly small and subcortical NMOSD lesions. A number of other kinds of ‘NMOSD -specific’ lesions have been described in the literature (Cabrera-Gomez JA, 2012), but were not observed in our series. Perhaps, this is due to the fact that none of our patients had history of cerebellar, diencephalic or brainstem relapses of the kind seen in NMOSD as reported previously (Wingerchuk et al., 2015). Our study was also limited by the fact that we used a relatively short follow-up duration of ~2.8 years. Future studies with larger sample sizes and longer follow-up periods are warranted to confirm our results. Another shortcoming of the current study was that R2* or T2* mapping was not performed due to time constraints; these sequences may provide additional quantitative information for characterization of NMOSD lesions. Finally, due to technical limitation, ultra-high field imaging is presently limited to cerebrum and does not cover posterior fossa regions.

In conclusion, ultra-high field (7T) GRE-T2*, SWI and QSM longitudinal study of 22 patients with NMOSD have shown that during the course of 2.8-year follow-up, the lesions did not demonstrate any variation in the morphological characteristics of the signal intensity pattern, suggesting that asymptomatic cerebral NMOSD lesions, once formed, tend to remain relatively inert. The absence of iron pathology or brain atrophy may be relevant to the observation of no clinically-overt progressive phase in NMOSD.