Magnetic Resonance Imaging and Spectroscopy in Late-Onset GM2-Gangliosidosis

Abstract

OBJECTIVE:

Our study aimed to quantify structural changes in relation to metabolic abnormalities in the cerebellum, thalamus, and parietal cortex of patients with late-onset GM2-Gangliosidosis (LOGG), which encompasses late-onset Tay-Sachs disease (LOTS) and Sandhoff disease (LOSD).

METHODS:

We enrolled 10 patients with LOGG (7 LOTS, 3 LOSD) who underwent a neurological assessment battery and 7 age-matched controls. Structural MRI and MRS were performed on a 3T scanner. Structural volumes were obtained from FreeSurfer and normalized by total intracranial volume. Quantified metabolites included N-acetylaspartate (NAA), choline (Cho), myo-inositol (mI), creatine (Cr) and combined glutamate-glutamine (Glx). Metabolic concentrations were corrected for partial volume effects.

RESULTS:

Structural analyses revealed significant cerebellar atrophy in the LOGG cohort, which was primarily driven by LOTS patients. NAA was lower and mI higher in LOGG, but this was also significantly driven by the LOTS patients. Clinical ataxia deficits (via the Scale for the Assessment and Rating of Ataxia) were associated with neuronal injury (via NAA), neuroinflammation (via mI), and volumetric atrophy in the cerebellum.

INTERPRETATION:

The decrease in NAA in the cerebellum suggests that, in addition to cerebellar atrophy, there is ongoing impaired neuronal function and/or loss, while an increase in mI indicates possible neuroinflammation in LOGG (more so within the LOTS subvariant). Quantifying cerebellar atrophy in relation to neurometabolic differences in LOGG may lead to improvements in assessing disease severity, progression, and pharmacological efficacy. Lastly, future neuroimaging studies in LOGG are required to contrast LOTS and LOSD more accurately.

1. Introduction

GM2-Gangliosidosis is a group of rare, neurodegenerative lysosomal storage disorders¹ that carry an estimated incidence rate of 1 in 222,000-422,000 live births.² Tay-Sachs disease and Sandhoff disease, the subvariants of GM2-gangliosidosis, are caused by mutations in the HEXA or HEXB gene, respectively. These mutations lead to dysfunctional hexosaminidase enzyme activity, thus creating an accumulation of gangliosides in cells.³ These subvariants are further defined by age of onset, ranging from classic infantile-onset, a rapidly progressive disease resulting in early death in childhood (type I), to adult-onset (type III), a chronic disease course that appears at approximately 21 years of age.³ The latter, known as late-onset GM2-gangliosidosis (LOGG), manifests more heterogeneously with slowly progressive ataxia, muscular weakness, cognitive dysfunction, and/or neuropsychiatric symptoms.^3-6

Structural magnetic resonance imaging (sMRI) of the brain in LOGG, often presented as case studies,^4,7-12 demonstrates severe cerebellar atrophy, particularly in the vermis.^3,4,13 However, beyond cerebellar atrophy, which does not always predict clinical dysfunction³ or disease duration,¹⁴ other aberrations remain lesser-known. These findings include thalamic hypodensities and lesions,^12,15 cerebral atrophy,^5,16 mild midbrain and brainstem atrophy,¹¹ ventricular enlargements,^4,9 white matter lesions,¹⁵ normal-appearing supratentorial structures in patients with cerebellar atrophy,¹⁴ and a lack of structural abnormalities.^15,16 Discrepancies in sMRI findings may relate to disease subvariants; Late-Onset Tay-Sachs (LOTS) patients tend to express more severe cerebellar atrophy,^4,8,10,13 while Late-Onset Sandhoff Disease (LOSD) patients have demonstrated a broad range of cerebellar atrophy^15,16 and are more likely to present with unremarkable brain imaging, especially at the time of diagnosis.^15,17 Furthermore, the extent of cerebellar atrophy in LOGG has yet to be quantified,¹⁸ and its relationship to the underlying metabolic pathophysiology has not been evaluated.

Proton Magnetic Resonance Spectroscopy (¹H-MRS), which allows for noninvasive metabolic measurements, has the unique ability to provide information about neuropathophysiology in disorders such as LOGG. N-acetylaspartate (NAA) is a marker for neuronal health and viability.^19,20 When choline (Cho), a membrane marker, decreases,²¹ it is associated with hypo-myelinating processes.²² Creatine + Phosphocreatine (Cr) are associated with energy metabolism.²¹ Increased glutamate-glutamine (Glx) is indicative of excitotoxicity, while decreased Glx is a marker of neuronal loss.²³ Myo-inositol (mI) is a likely neuroinflammatory/glial marker.²⁴ Previous studies in neurodegenerative diseases have shown elevated mI/Cr, which could be attributed to neuroinflammation in patients with amyotrophic lateral sclerosis (ALS)²⁵ and Alzheimer’s Disease.²³ Furthermore, a direct association was reported between mI, measured by in vivo 1H-MRS, and glial activation assessed with [11C]-PBR28 uptake in ALS – both presumed markers for neuroinflammation.²⁶ However, ¹H-MRS so far has only been systematically applied to infantile-onset GM2-gangliosidosis.^27-29 MRS reports of LOGG patients^12,30,31 are sparse, with case reports limited to cerebellar assessments, leaving supratentorial biochemistry largely unexplored. Understanding the relationship between MRS and sMRI may provide greater insight into disease pathology, particularly in metabolic diseases like LOGG. MRS cerebellar abnormalities in LOGG, such as lower NAA/Cr and increased mI/Cr ratios, have been consistent with sMRI findings compared to controls.¹³ Yet, MRS has occasionally revealed more about underlying neuronal mechanisms than sMRI imaging alone. Pretegiani et al. detailed a case report of LOSD presenting with pendular nystagmus, palatal tremor and progressive clinical ataxia where imaging revealed no evidence of brainstem or cerebellar atrophy, but MRS revealed low NAA/Cr and Cho/Cr ratios in both the pons and cerebellar white matter.³⁰ Another study demonstrated decreased NAA levels compared to controls in normal-appearing white matter and the thalamus.³¹ LOGG patients who show progressive cerebellar atrophy yet normal-appearing supratentorial structures on sMRI,¹⁴ have shown decreased NAA in these supratentorial regions via MRS.³¹ Thus, MRS metabolic findings may precede structural aberrations.

In this study, we utilized whole-brain exploratory sMRI to examine global structural aberrations and single-voxel MRS to investigate underlying metabolic abnormalities in the cerebellum, parietal cortex, and thalamus. Specifically, with regards to structural metrics, we investigated changes in subcortical volume, cortical thickness, and cortical surface area in the whole brain to identify the regions most affected by LOGG. Furthermore, metabolic metrics included NAA, Glx, mI, Cho, and Cr as biomarkers of neurodegeneration, neuroinflammation, and energy metabolism. Based on prior literature and our analyses, it was determined that the cerebellum was a key region of interest. Therefore, we then examined the relationship between structural and metabolic metrics within the cerebellum and how each of these metrics may be associated with clinical presentations of ataxia, non-motor dysfunction, disease severity, and age of disease onset. Utilizing both quantifiable neuroimaging metrics and quantifiable clinical scales may lead to biomarkers that more accurately reflect disease severity and are more sensitive to the heterogenous disease burden of LOGG. Our secondary aim was to identify at-risk brain regions³² in LOGG by solidifying less consistent structural^10,11,14,15 and supratentorial neurometabolic³¹ findings, which may serve as future therapeutic targets. Finally, we accounted for potential differences^3,6,33 in disease presentation of LOTS versus LOSD, thereby providing new pathophysiological and neuroimaging insights into these related³⁴ but distinct disorders that are not always distinguished in studies.

2. Methods

2.1. Standard Protocol Approvals, Registrations, and Patient Consents:

Ethical approval for the study was obtained through the Massachusetts General Hospital Institutional (MGH) Review Board. All participants provided written informed consent.

2.2. Participants:

The targeted population for this study were patients with a genetically confirmed biochemical diagnosis of LOGG and age-matched healthy controls. The diagnosis was defined by (a) absent to near-absent beta-hexosaminidase enzymatic activity in the serum or white blood cells or (b) mutation analysis of the HEXA and HEXB genes to distinguish pseudo deficiency alleles from disease-causing alleles. All participants were 18 years of age or older and willing to undergo MRI. Patients were recruited through either the MGH Leukodystrophy Clinic or the National Tay-Sachs and Allied Diseases Association (NTSAD). Participants were deemed as healthy controls through a self-reported history of no known neurological or psychiatric illness. Imaging consisted of two visits at six-month intervals for patients and one visit for age-matched controls. All LOGG patients were examined by a neurologist specializing in ataxia (C.D.S.) and disease severity was assessed with the following scales:⁶ the LOTS Severity Scale,³³ the Brief Ataxia Rating Scale (BARS),³⁵ the Friedreich Ataxia Rating Scale (FARS),³⁶ and the Scale for the Assessment and Rating of Ataxia (SARA).³⁷ The BARS, FARS and SARA scores were consistently adjusted for clinical weakness (i.e. if limbs were excessively weak and unable to be assessed a score of 0 was given, as opposed to a maximum score) so that the clinical extent of ataxia was recorded, as per Stephen et al. (2020).⁶ These adjustments were conducted to provide a more accurate assessment of clinical ataxia severity and to avoid inappropriately high scoring owing to weakness, which was particularly prevalent in the lower extremities and a frequently precluded assessment of lower extremity dysmetria. Additional non-motor features were assessed, including the Epworth Sleepiness Scale to quantify sleep quality³⁸ and the Cerebellar Cognitive Affective Syndrome Scale (CCAS) to assess cerebellar-related cognitive/neuropsychiatric deficits.³⁹

2.3. MR Imaging and Spectroscopy:

MR imaging and ¹H-MR spectroscopy were performed on a 3T Magnetom Tim Trio scanner (Siemens Healthcare, Erlangen, Germany) using a 32-channel receive coil at the Athinoula A. Martinos Center for Biomedical Imaging at MGH. A 3D T1-weighted multi-echo Magnetization Prepared Rapid Acquisition of Gradient Echo (MPRAGE) image was acquired with the following parameters: repetition time (TR)=2530ms, inversion time=1200ms, echo time (TE)=1.64ms, flip angle=7°, field of view=256×256 mm², number of slices=176, voxel size=1 × 1 × 1 mm³.

Three ¹H-MRS volumes of interest (VOIs) were placed (Figure 1), based upon the MPRAGE images, in the cerebellar vermis (2 × 2 × 2 cm³), the parietal cortex at midline (2 × 2 × 2 cm³), and the left thalamus (2 × 2 × 1.5 cm³). The thalamus was selected based on early animal model research³² and a previous human study,³¹ while the parietal cortex was selected as a supratentorial example voxel³¹ to explore these potential at-risk regions. These two regions allowed for deep gray nuclei and cortex comparisons. A point-resolved spectroscopy (PRESS) sequence (TE/TR=30ms/1700ms, number of averages (NA)=128, vector size=1024, bandwidth=1200Hz) with water suppression enhanced through T1 effects (WET) was used to measure brain metabolites. Prior to data acquisition, the VOIs underwent an automatic shim routine based on gradient double acquisition (GRE-shim) using first and second order shims followed by first order manual shimming. To estimate absolute concentrations, a water unsuppressed spectrum (NA=4) was also acquired in the same VOI. Additionally, to further evaluate the consistency of voxel coverage across participants, a probabilistic map of voxel placement in standard space was created (Figure 1). To this end, we used FSL FNIRT to calculate a nonlinear transformation between each participant’s native T1 volume and the MNI152 template⁴⁰ and then applied the resulting transformation to the MRS voxel. In addition, MRS masks in MNI space were used to calculate the voxel centroid for each participant using Python scripts (https://github.com/nwd2918/MRS-voxel-plot).

Figure 1. MRS voxel placement and MR Spectra.

Voxel overlap density maps of the (A) cerebellum (B) parietal cortex and (C) left thalamus for all participants. Individual MRS voxels were transformed to the MNI space and then combined to show the overlap between participants. Corresponding MR spectra, average (solid) and standard deviation (shaded area) from LOGG patients are shown below each map.

2.4. Data Processing:

Each T1-weighted multi-echo MPRAGE image was reconstructed and processed with FreeSurfer version 7.1.0 (May 2020)⁴¹ to facilitate volumetric and surface-based structural analyses. For whole-brain volumetric subcortical segmentation processing, a new algorithm, Sequence Adaptive Multimodal Segmentation (SAMSEG) was run alongside recon-all, the standard reconstructive processing pipeline. Surface-based processing consisted of cortical thickness and cortical surface area parcellation measurements for each participant’s left and right hemispheres (APARC). The output files for each participant contained quantitative total voxel calculations for both the segmentations and parcellations, which were automatically assigned based upon the default Desikan-Killiany atlas.⁴² For subcortical volumetric analyses, 37 of 41 regions-of-interest (ROIs) volume segmentation labels from the Desikan-Killiany atlas were selected. All 35 surface parcellation labels based on the atlas were used in the surface-based analyses, and thus no labels were excluded. In addition, “total cerebellum volume” was calculated as the sum of the cerebellar cortex and cerebellar white matter across both hemispheres. Subcortical volume and cortical surface-based (thickness) outputs were normalized by estimated total intracranial volume (eTIV) to normalize for total head size, accounting for sex differences.⁴³

MR spectra were processed and fitted using a simulated basis-set with LCModel v6.3.⁴⁴ Quantified metabolites included NAA, Cho, mI, Cr, and Glx. These metabolites were selected as they are typically robust measures on 3 Tesla with low CRLB, a benchmark of accuracy of the fit. (Please see results section 3.3.) The signal-to-noise ratio (SNR), estimated by LCModel and defined as peak height of NAA divided by the root mean square of the noise of the LCModel fit, was greater than 7 for all spectra (mean±SD=22±8, ≥5 is considered acceptable). Absolute metabolite concentrations were reported relative to the water-unsuppressed spectra and concentrations were corrected for partial volume effects as follows: each MRS voxel was first registered to the T1-anatomical space and segmented using Gannet toolkit v3.1⁴⁵ and SPM12,⁴⁶ and then segmented tissue fractions were corrected for metabolite concentrations derived from LCModel to account for cerebrospinal fluid (CSF) content.⁴⁷

2.5. Statistical Analyses:

To assess potential disease progression, first and second visits were analyzed with paired, two-tailed T-tests for volumetric, surface-based, and metabolic data (p<0.05). The within-group, between-visit analyses revealed non-significant results. Thus, visit 2 LOGG data were selected for between-group comparisons given the degenerative nature of the disease, albeit non-significant within six months. Additionally, while eTIV normalization accounted for volumetric sex differences, Pearson Chi-Square tests and Wilcoxon Rank Sum tests were conducted in R Studio to assess between-group sex and age differences, respectively.

For between-group comparisons, ANCOVAs were conducted in MATLAB to compare volumetric, surface-based, and metabolic measures in these groups, controlled for age: i) all LOGG patients (10) vs. controls (7) and ii) only LOTS patients (7) vs. controls (7). While LOTS patients (7) vs. LOSD patients (3) was of interest, the small sample size of the incredibly rare LOSD cohort would have made for poor statistical inference. 95% confidence intervals were also recorded across all measures. (Note that confidence intervals are expressed as fractions for volumetric results due to eTIV normalization.) Results were corrected for multiple comparisons using the false discovery rate (FDR) method (p<0.05). However, metabolite concentrations in the parietal cortex and thalamus were considered secondary outcome markers to generate hypotheses for future studies, so FDR correction was not applied, given the exploratory approach. These regions are either minimally studied or thought to be at-risk in LOGG patients.^31,32

Cerebellar results were considered our primary outcome markers. Thus, the associations between total cerebellar volume, cerebellar metabolites NAA and mI, and clinical metrics were assessed using the partial correlation coefficient in R Studio, with age as a regressor of no interest. Since age of the participant was the sum of age at disease onset and disease duration (i.e., interrelated), only the association between age at disease onset and cerebellar markers were assessed, controlling for age in the partial correlation framework. Due to known 4^th ventricular enlargement^4,9 in LOGG, specifically LOTS, 4^th ventricle volume was also correlated with the ataxia scales and disease severity. All associations were corrected for multiple comparisons using the FDR method.

2.6. Data Availability Statement:

Due to the extremely rare nature of LOGG, the Institutional Review Board Detailed Protocol states that study data, including de-identifiable imaging and clinical data, cannot be shared externally without IRB approval and a Data User Agreement in place.

3. Results

3.1. Participants:

Seven of the ten LOGG patients (LOTS=7, LOSD=3) were age-matched to a healthy control (Table 1). Age-matching was not dependent on disease subvariant. It was noted that the 3 LOSD were older than the 7 LOTS patients. Results of the Wilcoxon Rank Sum Test indicated that between all LOGG patients and healthy controls, neither age (W=36, p= 0.961) nor sex (χ²= 0.014, p= 0.906) were significantly different. This was also the case between LOTS patients and healthy controls for age (W=31, p= 0.442) and sex (χ²= 0.311, p= 0.577) differences. Detailed clinical assessments of the 10 patients are published in Stephen et al. (2020). Four of ten patients also underwent a mutation analysis and results were as follows: Patient 1: heterozygous for HEXA:c.805G>A (p.Gly269Ser) and HEXA:c.1274_1277dupTATC (p.Tyr427Ilefs*5) mutations, Patient 2: heterozygous for HEXA:c.1274_1277dupTATC and HEXA:c.805G>A (p.Gly269Ser) mutations, Patient 3: heterozygous for the HEXB:298delC mutation and a HEXB:p.Gly473Ser variant of unknown significance, and Patient 4: heterozygous for the HEXA:c.1274_1277dupTATC and HEXA:c.805G>A (p.Gly269Ser) mutations.

Table 1:

Participant demographics and clinical assessment scores reported as mean ± standard deviation along with ranges for age at assessment, age at disease onset, and disease duration. Patient information is from visit 2.

Demographic & Clinical Assessment Information	Participant Group
	LOGG (N=10)	LOTS (N=7)	LOSD (N=3)	CTRLS (N=7)
Demographic Details
Sex (Male: Female) (N)	6:4	5:2	1:2	4:3
Age (Years)	42.0 ± 12.6 (22 - 63)	37.6 ± 11.2 (22 - 58)	53.3 ± 10.5 (42 - 63)	42.7 ± 14.3 (23 - 62)
Age at Disease Onset (Years)	19.4 ± 9.7 (8 - 36)	16.7 ± 9.3 (8 - 30)	25.7 ± 9.3 (18 - 36)	- -
Disease Duration (Years)	22.6 ±11.6 (6 - 40)	20.9 ± 9.0 (11-32)	26.7 ± 18.1 (6 - 40)	- -
Clinical Assessment Scores
FARS	25.2 ± 14.0	27.6 ± 13.0	19.7 ± 17.5	-
BARS	5.1 ± 4.1	6.2 ± 4.1	2.3 ± 2.8	-
SARA	8.4± 4.3	9.9 ± 3.7	5.0 ± 4.3	-
LOTS Severity Scale	6.9 ± 3.1	7.4 ± 3.3	5.7 ± 2.9	-
CCAS	100.2 ± 10.8	98.7 ± 11.7	103.7 ± 9.6	-
Epworth Sleepiness	8.7 ± 5.5	7.4 ± 3.7	11.7 ± 8.6	-

Abbreviations: LOGG=Late-Onset GM2-Gangliosidosis; LOTS=Late-Onset Tay-Sachs Disease; LOSD=Late-Onset Sandhoff Disease; CTRLS=Healthy Controls; FARS=Friedreich Ataxia Rating Scale; BARS=Brief Ataxia Rating Scale; SARA=Scale for the Assessment and Rating of Ataxia; CCAS=Cerebellar Cognitive Affective Syndrome Scale

3.2. sMRI Results:

3.2.1. Subcortical volumes

No manual edits to the labels were performed. No participants were discarded from the statistical analyses based on image quality. Mean total brain volume (eTIV) was noted as 1,494,276 mm ³ in LOGG patients, 1,514,696 mm ³ in LOTS patients, 1,446,628 mm³ in LOSD patients, and 1,521,744 mm³ in controls. Total brain volume differences were insignificant for both LOGG (p=0.54) and LOTS (p=0.80) patients when compared to controls.

Upon visual inspection of the MR images (Figure 2), cerebellar atrophy appeared to be predominantly present in the 7 LOTS patients and minimal in the 3 LOSD patients. Subcortical volumes reduced between visits but were non-significant, particularly in the cerebellum (p=0.14).

Figure 2. T1-weighted MR images showing cerebellar atrophy.

Defaced coronal and sagittal MRI slices (MNI space: x=0) of (A) patient presenting with late-onset Tay-Sachs disease showing profound cerebellar atrophy, (B) patient presenting with late-onset Sandhoff disease, and (C) healthy control.

Mean (non-normalized) volume and statistical results in the cerebellum for each cohort can be seen in Table 2, alongside the statistical results. The 7 LOTS patients showed more severe total cerebellar atrophy when separately compared to controls than when included in the collective LOGG disease category. Specifically, gray matter atrophy was primarily driven by LOTS patients (Table 2). LOTS patients also showed an enlarged 4^th ventricle (p=2.20 × 10⁻⁴, 95% CI [0.0009, 0.0019]) when compared to controls. No other subcortical regions revealed statistically significant differences.

Table 2:

Percent change, p-values, and 95% confidence intervals for ANCOVA between-group comparisons using eTIV-normalized cerebellar data. Raw mean volumes with standard deviations (non-normalized via eTIV division) for cerebellar regions are also reported. LOGG patients exhibited lower volumes than controls, but LOTS patients had lower volumes than any other cohort, which was most prominent in the cerebellar cortex. Cerebellar white matter atrophy was similar between disease subvariants. All p-values are un-adjusted but are corrected for multiple comparisons and noted in bold.

Cerebellum Volumetric Values	Total Cerebellum Volume	Left Cerebellum White Matter	Left Cerebellum Cortex (Gray Matter)	Right Cerebellum White Matter	Right Cerebellum Cortex (Gray Matter)
% Change p-values [95% CI]
10 LOGG vs. 7 CTRL	−33.1% 0.0016* [0.0125, 0.0428]	−20.0% 0.0053* [0.0006, 0.0028]	−36.9% 0.0023* [0.0051, 0.0191]	−19.8% 0.0041* [0.0007, 0.0026]	−36.3% 0.0020* [0.0054, 0.0191]
7 LOTS vs. 7 CTRL	−42.9% 7.38×10−6* [0.0276, 0.0474]	−21.5% 0.0088 [0.0007, 0.0032]	−49.4% 1.25×10−6* [0.0131, 0.0206]	−20.6% 0.0098 [0.0006, 0.0030]	−47.9% 3.16×10−6* [0.0127, 0.0209]
Volume (mm3) (mean± SD)
LOGG Patients	90,039 ±24,254	11,484 ± 1,610	33,070 ±11,412	11,150 ±1,511	34,335 ± 11,119
LOTS Patients	76,929 ±14,374	11,280 ±1,902	26,507 ±5,160	11,039 ±1,830	28,103 ±5,751
LOSD Patients	120,630 ± 4,705	11,960 ± 564	48,385 ± 1,888	11,409 ± 295	48,875 ± 2,010
Controls	134,617 ± 18,800	14,370 ± 2,419	52,442 ± 7,009	13,904 ± 2,351	53,901± 7,547

^*Denotes regions significant after FDR multiple comparisons corrections

3.2.2. Cortical thickness and surface area

LOTS patients showed a trend towards greater mean cortical thickness in the right (p=0.06) and left (p=0.12) hemispheres with an average thickness across the cortex of approximately 2.51 mm³ in LOTS compared to 2.42 mm³ in controls and 2.45 mm³ in LOSD patients. LOTS patients expressed significantly greater thickness after FDR correction only in the left fusiform area compared to controls (p=2.07×10⁻⁴, 95% CI [0.1467, 0.3424]). The final structural metric collected was surface area of the cortex for which there were no significant between-group differences.

3.3. ¹H MRS Results:

3.3.1. MRS data quality

The Cramér-Rao lower bounds (CRLB) for the metabolites under investigation were all below 13% (≤ 15-20% is considered acceptable).⁴⁸ No data were excluded due to low SNR or high CRLB. One MRS dataset in the parietal cortex of one LOGG patient and one dataset in the thalamus of another LOGG patient needed to be excluded as no water reference spectrum was acquired. Another MRS dataset in the thalamus of a healthy control needed to be excluded as upon inspection, the thalamus had not been sufficiently included in the VOI.

3.3.2. Cerebellum metabolites

Cerebellar metabolite concentrations were considered our primary outcome biomarkers. None of the metabolites showed a significant change over the study course of 6 months (p>0.09 across analyses taken together). In accordance with the structural results, only the findings of the second visit are described.

Group analyses between all LOGG patients (10) and controls (7) (Table 3) revealed no significant metabolic differences after FDR correction (p<0.05). However, noteworthy differences in metabolites between LOGG patients and controls were significantly driven by the LOTS population. Group analyses between the 7 LOTS patients and controls after FDR correction (p<0.05) revealed elevated mI, and decreased NAA, Glx, and Cr (Table 3). Only Cho values seemed to not be driven by LOTS.

Table 3:

Percent change, p-values, and 95% confidence intervals for ANCOVA group comparisons. Mean cerebellar metabolic values with standard deviations are also reported. LOTS patients drove most of the significant metabolic differences. All p-values are un-adjusted but are corrected for multiple comparisons and noted in bold.

Cerebellum Metabolic Values	NAA	mI	Glx	Cr	Cho
% Change p-values [95% CI]
10 LOGG vs. 7 CTRL	−20.0% 0.015 [0.3657, 2.4934]	26.7% 0.043 [0.1005, 3.0279]	−16.9% 0.065 [−0.0570, 4.2779]	−5.6% 0.14 [−0.1778, 1.1582]	−5.8% 0.071 [−0.0064, 0.2423]
7 LOTS vs. 7 CTRL	−28.2% 0.0003* [1.093, 2.923]	38.7% 0.0064* [0.8773, 3.6513]	−25.6% 0.0018* [1.4071, 4.9871]	−10.0% 0.0065* [0.3354, 1.4062]	−6.3% 0.12 [−0.0257, 0.2759]
Concentration (mean± SD)
LOGG Patients	5.71±1.24	7.40±1.67	10.41±2.59	8.20±0.71	1.96±0.10
LOTS Patients	5.13±0.99	8.10±1.48	9.32±2.04	7.82±0.41	1.95±0.12
LOSD Patients	7.06±0.26	5.77±0.58	12.94±2.00	9.09±0.27	1.98±0.07
Controls	7.14±0.50	5.84±0.81	12.52±0.74	8.69±0.51	2.08±0.14

^*Denotes regions significant after FDR multiple comparisons corrections

3.3.3. Parietal cortex and left thalamus metabolites

Additionally, we performed MRS in the parietal cortex and thalamus. In the parietal cortex, there was a 10.3% decrease in NAA in LOGG patients (8.87±1.14) compared to controls (9.89±0.55), (p=0.042, 95% CI [0.0118, 2.0333]), which was not driven by the LOTS cohort (8.78±1.43), (p=0.12, 95% CI [−0.1672, 2.3923]). No significant differences in the thalamus were found.

3.4. sMRI-MRS Interaction:

Figure 3 presents the interaction between sMRI normalized total cerebellar volumes and MRS cerebellar metabolic concentrations. Across volumes and metabolites, there was a clear separation between LOTS patients and controls. These differences were more prominent for normalized volumetric values as compared to metabolite concentrations. Additionally, LOSD patients were observed to fall within the control cluster, highlighting a distinction among the LOGG subvariants. Statistical results of the relationship between sMRI and MRS amongst the LOGG patients can be viewed in Table 4.

Figure 3. Association between normalized cerebellum total volumes and cerebellar metabolite concentrations for patients and controls.

Figure 3 is intended to help visualize the degree of association between the two variables and separation of groups. The narrower the ellipse the higher the correlation between the variables. Correlative statistics were conducted only for LOGG patients between normalized cerebellar volume and mI (R = −0.57, p = 0.106), NAA (R = 0.78, p = 0.013), Cho (R = 0.62, p = 0.19), Cr (R = 0.81, p = 0.008), and Glx (R = 0.70, p = 0.034) (Table 4).

Table 4:

Partial Correlation Matrix of cerebellar results amongst the LOGG patient population with age regressed as a variable of no interest. All p-values are un-adjusted but are corrected for multiple comparisons and noted in bold.

Cerebellum Correlations	mI	Cho	Cr	NAA	Glx	Total Volume	SARA
mI	-	p = 0.92 R = 0.04	p = 0.27 R = −0.415	p = 0.015* R = −0.77	p = 0.12 R = −0.57	p = 0.106 R = −0.57	p = 0.036 R = 0.70
Cho		-	p = 0.164 R = 0.16	p = 0.21 R = 0.46	p = 0.172 R = 0.50	p = 0.19 R = 0.62	p = 0.28 R = −0.40
Cr			-	p = 0.043 R = 0.68	p = 0.057 R = 0.65	p = 0.008* R = 0.81	p = 0.175 R = −0.495
NAA				-	p = 0.005* R = 0.84	p = 0.013* R = 0.78	p = 0.0036* R = −0.85
Glx					-	p = 0.034 R = 0.70	p = 0.057 R = −0.65
Vol						-	p = 0.028* R = −0.72

^*Denotes regions significant after FDR multiple comparisons corrections

3.5. Neuroimaging and Clinical Metrics:

Amongst all LOGG patients, there was a significant association between the SARA scores and total cerebellar volumes (R= −0.72, p=0.028), which was stronger when analyzing only the LOTS cohort (R= −0.91, p=0.011). In the LOTS cohort, total cerebellar volumes were also significantly associated with the FARS (R= −0.91, p=0.012) and BARS (R= −0.82, p=0.047) ataxia scales. No other associations between clinical metrics and structural findings throughout the brain, cortically or subcortically, were significant after FDR correction.

For the associations between clinical and metabolic markers in LOGG, we only focused on cerebellar NAA and mI, as these two were the most robust metabolic markers of LOGG as shown in the between-group analyses. Amongst LOGG patients, NAA in the cerebellum negatively correlated with the total scores of FARS (R= −0.75, p=0.020), BARS (R= −0.74, p=0.022) and SARA (R= −0.85, p=0.0036), but not the Disease Severity Scale (R= −0.46, p=0.21). Cerebellar mI showed a trending association with the SARA after FDR correction (R=0.70, p=0.036) but a significant association with age at disease onset (R=−.87, p= 0.0025). When assessing LOTS patients independently, only NAA was significantly associated with the SARA (R= −0.82, p=0.048). However, mI was still significantly correlated to age at disease onset in LOTS patients (R=−0.91, p=0.0113). Table 4 shows the relationship among metabolite levels, the sMRI cerebellar volumes, and the SARA scores in the LOGG cohort, with age used as a regressor of no interest. Figure 4 presents the associations between cerebellar metabolites NAA and mI and cerebellar volume with clinical ataxia deficits (SARA scores).

Figure 4. Associations between clinical ataxia severity (SARA score) compared to cerebellar metabolic markers and cerebellar volume.

LOTS patients are illustrated as blue circles and LOSD patients as orange squares. Plots comparing clinical ataxia severity (SARA score) with cerebellar metabolites A) mI and B) NAA, and C) total cerebellar volume are shown. Values plotted, not regressed for age, are shown alongside shaded 95% confidence intervals.

4. Discussion

This study revealed that the brains of those affected by LOGG show widespread neuronal dysfunction, inflammation, and altered energy metabolism within regions of cerebellar atrophy. While clinical ataxia was most clearly associated with total cerebellar volumes and NAA levels, it may be the result of complex sphingolipids affecting a larger network of neuronal dysfunction and causing inflammatory changes.⁴⁹ Interestingly, our data suggest that solitary β-hexosaminidase A enzyme deficiency (LOTS) was more severely affected than the combined deficiency of β-hexosaminidase A and B enzymes (LOSD), however this difference may have been related to differences in genetic mutation severity, as our LOSD cohort did not have the cerebellar phenotype.⁶ The slower neurodegenerative nature of LOGG is supported by the lack of significant changes between six-month visits. In comparison, infantile-onset GM2-Gangliosidosis reveals significant increases in brain volume, and more specifically, in the ventricles, cerebral cortex, and cerebellar cortex at one-year follow up. Juvenile-onset shows global brain atrophy as well as atrophy in the corpus callosum, caudate, and basal ganglia in combination with enlarged ventricles.¹⁸ Although there have been no known natural history MRS studies in infantile-onset, previous studies demonstrated similar differences in NAA and mI compared to controls.²⁹

Significant cerebellar atrophy, especially in the cerebellar cortex (gray matter), was seen across the LOGG cohort but primarily driven by LOTS patients. This finding, together with that of enlarged 4^th ventricles, replicates prior studies, albeit mostly case studies,^4,9,10,11,14 that have shown more severe cerebellar atrophy in LOTS patients compared to other studies that assessed LOSD patients.^15,17,30 The LOSD patients in this particular study did not have evidence of global cerebellar involvement and instead clinically presented with a lower motor neuron neuromuscular phenotype.⁶ A notable, similar reduction in cerebellar white matter volume was observed in both LOGG and LOTS. In analyzing total brain volume, LOGG patients had less, though insignificantly less, mean total volume compared to controls than LOTS patients, which may have been driven by the addition of LOSD patients to the LOGG analysis. Previous LOSD¹⁶ and LOGG studies⁵ have highlighted patients who visually presented with cortical atrophy.

LOGG patients with normal supratentorial structures on a conventional brain MRI¹⁴ have shown lower concentrations of NAA in the cerebellum and other supratentorial brain regions,³¹ which may be indicative of metabolic alterations preceding structural changes. Yet, the relationship between structural atrophy and underlying metabolic energy demands is unexplored, especially amongst regions like the cerebellum that require considerable metabolic energy.⁵⁰ Elevated mI and decreased NAA in the cerebellum of LOGG compared to controls was driven solely and significantly by the LOTS cohort. Gene expression profiles reported in GM2-Gangilosidosis suggest that neuroinflammation plays a major role in neuronal atrophy.⁵¹ Therefore, our observation of elevated mI and decreased NAA may be the result of glial activation due to accumulation of gangliosides in neurons.⁵² Differences in Glx and Cr in the cerebellum of LOGG patients compared to controls was also significantly governed by the LOTS patients. Decreased levels of Cr may reflect altered energy metabolism and/or neuronal loss.⁵³ Notably, Cr was highly correlated to cerebellar atrophy and not to other metabolites, which emphasizes the cerebellum’s high energy consumption.⁵⁰ A decrease in Glx in conjunction with a decrease in NAA is not surprising as most neurodegenerative diseases show decreased Glx or Glu, which reflects either neuronal loss or a decreased production of Glu by degenerating neurons.^54,55 No significant change in Cho levels were observed in our patients. There have been conflicting previous reports of Cho changes in LOGG^30,31 related to demyelination and hypomyelination.^22,29 The lack of significant differences in our study population may be due to various disease stages. Unlike in the cerebellum, the metabolites in the secondary regions of interest, such as decreased NAA in the parietal cortex, lacked a distinction between disease subvariants.

Cerebellar neuroimaging measures were associated with clinical ataxia metrics, a relationship that has not been previously investigated. Cerebellar metabolites have previously only been correlated with mobility and disease duration.³¹ Yet, NAA was the most prominent metabolic marker of clinical ataxia and was associated with all three ataxia measures: the SARA (which focuses mainly on gait and posture), the FARS (which also includes neuromuscular features), and the BARS (which is broader in scope, assessing all 5 cardinal cerebellar features: eye movements, speech, arm/leg appendicular dysmetria and gait) amongst both the LOGG and LOTS cohorts. Elevated cerebellar mI was associated mildly with clinical deficits on the SARA scores but significantly with earlier disease onset, which differs from previous findings.³¹ Amongst the LOGG cohort, total cerebellar volume was significantly associated with only the SARA, but when assessing just the LOTS patients, all three clinical ataxia assessments were significant. Overall, clinical ataxic deficits seemed to be driven by a combination of neuroinflammation (mI), neuronal injury (NAA), and volumetric atrophy in the cerebellum. Additionally, while some metabolites like creatine, a marker for energy metabolism, were not significant, there were strong correlations with neuronal injury as seen in Table 4. We therefore recommend that in order to obtain holistic structural and neurometabolic effects of LOGG on the brain, future investigations study sMRI and MRS in tandem.

Our secondary aim was to explore and quantify other neural aberrations in LOGG beyond the cerebellum. Although previous neuroimaging studies in LOGG patients have found inconsistent abnormalities in the cerebral cortex,^5,16 thalamus,^12,15 and mesencephalon,¹¹ our structural results did not support any of those findings as potential at-risk regions in LOGG. These previous studies only noted visual abnormalities on conventional imaging, but our quantification of changes showed little difference in size. However, at the cortical level, LOTS patients tended to have greater mean cortical thickness across hemispheres, which was significantly greater in the left fusiform area. There have been no known reports of impaired facial recognition in LOGG, however, the fusiform area has been implicated in abnormal cerebellar networks in other disorders.^56,57 This is a previously unreported observation that may be a result of ganglioside accumulation in neurons⁵², but future studies with greater statistical power should further evaluate this discovery since the fusiform area borders the cerebellum.

With regards to MRS, while the thalamus was considered a potential at-risk region based on animal models³² and a previous human study,³¹ no significant differences in the thalamus were found in our patient population. The parietal cortex was used to measure supratentorial metabolic differences, which have been sparsely studied.³¹ Our results showed significantly decreased NAA, not exclusively driven by the LOTS patients, indicating that both LOTS and LOSD may be at risk for supratentorial neuronal injury or loss. However, the effect size was much smaller compared to the cerebellar metabolic differences.

Our final aim was to delineate the heterogeneity of LOGG by contrasting LOTS and LOSD patients. Studies often look at LOGG as a collective group of disorders due to their similarities in disease classification³⁴ and difficulties recruiting subvariants of sufficient sizes to appropriately characterize these ultra-rare diseases.³ However, this grouping may contribute to poor specificity of structural abnormalities, since limited research suggests unique neuroanatomical vulnerabilities for each of the subvariants.^{3,4,8-11,13,15-17,30} While LOTS patients are characterized by severe cerebellar atrophy,^4,8,11,13 and enlarged ventricles,^4,9 LOSD patients have shown more heterogeneous presentations^15,17,30 with multiple phenotypes, which may give rise to the varying degrees of cerebellar atrophy.^6,15 LOSD, which is rarer than LOTS, has understandably undergone fewer neuroimaging studies compared to LOTS. Our aim considers stark differences in visual neuroanatomy, which showed minimal cerebellar atrophy and more typical cerebellar metabolic values among our 3 LOSD patients. Additionally, plotted cerebellar structural and neurometabolic values showed clear clustering between these subvariants (Figure 3). We infer that LOSD patients therefore played a role in the statistical differences observed between all LOGG patients (LOTS + LOSD) and only the LOTS patients, especially with regards to the cerebellum. This inference supports the notion that LOTS has important structural distinctions, and a combined LOGG group may be obscuring important clinical phenotypes of both brain structure and clinical presentation.

The small sample size of LOSD patients, albeit a comparatively rarer disease than LOTS, limited our ability to truly characterize imaging differences among these two LOGG subvariants. Additionally, all three LOSD patients were older and presented with a lower motor neuron phenotype, rather than a cerebellar phenotype of the disease,⁶ which may have driven cerebellar structural, metabolic, and clinical ataxia differences. In our study mutation burdens could not be fully addressed, as only a small number of the patients had undergone genetic testing. The majority of the ten participants were diagnosed prior to the modern genomic era by biochemical testing alone. Therefore, interpreted differences between LOTS and LOSD could be related to individual mutations rather than the disease subvariants themselves. With regards to cerebellar atrophy, the cerebellar vermis, which has previously shown significant visual atrophy,¹⁴ is not an official automated segmentation of FreeSurfer. Thus, its degree of atrophy should be investigated in future studies. Supratentorial regions should also continue to be further explored in more detail. Healthy controls, who were not expected to have any changes at a six-month follow up, only completed one imaging visit. This limitation, in addition to lacking clinical assessments from healthy controls, prevented us from conducting a true cross-sectional analysis across visits. However, our results, combined with previously discussed differences between LOTS and LOSD,⁶ point us towards the need to evaluate these diseases independently in the future due to differences in LOSD phenotype expression. Finally, as the first natural history neuroimaging study to examine MRS and quantified sMRI in LOGG, there are many unanswered questions as to whether neuroimaging predicts disease progression. Currently, patients are only being diagnosed in symptomatic phases of the disease course, and we do not know much about the preclinical stages of disease. Although we hypothesize that our techniques carry predictive power, longer studies, extending beyond a six-month period, are required to assess whether structural or metabolic brain changes precede clinical dysfunction.

Quantification of cerebellar atrophy in LOGG will lead to imaging biomarkers that may improve assessments of disease severity, progression, and pharmacological efficacy while alluding to potential differences in disease burden between LOTS and LOSD. We also highlight a unique relationship between neuronal dysfunction, neuronal loss, neuroinflammation, altered energy metabolism, and atrophy in the cerebellum by utilizing sMRI and MRS in tandem to study LOGG. These biomarkers have shown to accurately reflect quantified clinical ataxia metrics, which previous studies have not examined. Metabolites also further correlated with age of disease onset. Rare diseases, such as LOGG, will directly benefit from imaging biomarkers since disease characteristics will be more appropriately defined and interventions will be more efficiently explored.

5. Conclusion

As part of the first natural history neuroimaging study in LOGG to quantify structural aberrations and MRS in relation to clinical ataxia, our findings indicate that in our cohort, LOTS patients presented with severe cerebellar atrophy, while this was not seen in our LOSD patients, who were of the neuromuscular phenotype. Furthermore, neuroinflammation, as measured by mI, and neurodegeneration, as measured by NAA appear within regions of cerebellar atrophy. This combination of neuroinflammation, neuronal injury, and volumetric atrophy in the cerebellum seems to drive clinical ataxia in LOGG. However, differences in findings between disease subcategories indicate that future studies should evaluate LOTS and LOSD independently. Future research will also need to investigate whether imaging could represent a biomarker of disease progression, requiring a longer period of observation than six months. Finally, by using both sMRI and MRS imaging techniques in studying metabolic diseases such as LOGG, more accurate imaging biomarkers may come to light, potentially improving disease severity assessments as well as therapeutic interventions.