Characterization of Central Manifestations in Patients with Niemann-Pick Disease Type C

Abstract

Purpose:

Niemann-Pick Disease Type C (NPC) is a rare lysosomal storage disease characterized by progressive neurodegeneration and neuropsychiatric symptoms. This study investigated pathophysiological mechanisms underlying motor deficits, particularly speech production, and cognitive impairment.

Methods:

We prospectively phenotyped 8 adults with NPC and age-sex-matched healthy controls using a comprehensive assessment battery, encompassing clinical presentation, plasma biomarkers, hand-motor skills, speech production, cognitive tasks, and (micro-)structural and functional CNS properties through MRI.

Results:

NPC patients demonstrated deficits in fine-motor skills, speech production timing and coordination, and cognitive performance. MRI revealed reduced cortical thickness and volume in cerebellar subdivisions (lobule VI and crus I), cortical (frontal, temporal, and cingulate gyri), and subcortical (thalamus and basal ganglia) regions, and increased choroid plexus volumes in NPC. White matter fractional anisotropy was reduced in specific pathways (intracerebellar input and Purkinje tracts), while DTI graph theory analysis identified altered structural connectivity. NPC patients exhibited altered activity in sensorimotor and cognitive processing hubs during resting-state and speech production. Canonical component analysis highlighted the role of cerebellar-cerebral circuitry in NPC and its integration with behavioral performance and disease severity.

Conclusion:

This deep phenotyping approach offers a comprehensive systems neuroscience understanding of NPC motor and cognitive impairments, identifying potential CNS biomarkers.

Introduction

Niemann-Pick Type C Disease (NPC, OMIM #257220) is a rare inherited lysosomal storage disorder (LSD) characterized by progressive neurodegeneration and neuropsychiatric symptoms due to intralysosomal lipid accumulation in neurons¹. The prevalence of NPC is estimated at 3.9 cases per million in the United States, a rate potentially underestimated due to delayed diagnosis or misdiagnosis². Approximately 95% of pathogenic variants occur in the NPC1 gene, whereas 5% occur in the NPC2 gene³. Despite an increasing comprehension of the molecular and genetic factors of NPC, currently no Food and Drug Administration approved treatments are available, although multiple therapeutic candidates are in development¹.

Prior studies have demonstrated that motor (i.e., speech production, swallowing, fine motor skills) and cognitive (i.e., working memory) impairments substantially reduce quality of life for NPC patients and caregivers^{2, 4}. Clinical scales such as the NPC-Clinical Severity Scale (NPC-CSS), NPC Suspicion Index (NPC-SI), and NPC Clinical Database Score Sheet (NPC-CDB) are frequently employed to assess these domains for disease status and treatment effect⁴. However, the neurobiological causes of these symptoms, and relatedly, variable symptom severity, remain largely unknown.

Diagnosis and characterization of disease severity in NPC is aided by quantification of plasma biomarkers such as cholestane-3β,5α,6β-triol (C-triol), 7-ketocholesterol (7-KC), and lysoSM-509 (N-palmitoyl-O-phosphocholineserine [PPCS])³. Yet, peripheral biomarkers do not provide a direct assessment of CNS alterations that may underlie motor, cognitive, or psychiatric symptoms^{2, 4}. Indeed, recent work suggests that quantifying neurofilament light chain (NfL) expression in cerebral spinal fluid may be important for monitoring patients and detecting treatment effect⁵. Earlier magnetic resonance imaging (MRI) studies have defined white and gray matter atrophy in NPC patients^{6, 7} and have provided initial evidence demonstrating possible protective effects of Miglustat, a treatment used off-label in NPC^{6, 8}. Though informative, these studies in NPC are limited by factors such as retrospective design and non-standardized data acquisition methods.

Identifying objective CNS biomarkers that closely align with functional status of patients are currently critically needed in NPC. Addressing this gap will improve the neurobiological understanding of the neuropsychiatric spectrum observed in NPC as well as provide a foundation for enhancing clinical trial readiness and thereby aid the development and validation of innovative treatment approaches. Thus, in the present study, we prospectively applied a cross-sectional design to probe two prevalent symptom domains that diminish quality of life in NPC – motor deficits, particularly in regard to speech production, and cognitive impairment². We aimed to uncover pathophysiological mechanisms at the systems neuroscience level underlying motor and cognitive symptoms in NPC by employing a deep phenotype approach in adult male and female NPC patients, combining structural and functional neuroimaging with clinical assessments, plasma biomarker quantification, and behavioral tasks assessing speech, fine-motor skills, physical abilities, and cognition.

Materials and Methods

Patient Recruitment

This study, approved by the Boston Children’s Hospital (BCH; Supplemental Table 1) and Massachusetts General Brigham (MGB) Institutional Review Boards, adhered to Helsinki criteria for human subject research. Participants were recruited from the BCH Lysosomal Storage Disorder Program and National Niemann Pick Disease Foundation. Age and sex-matched healthy controls were recruited locally. All participants provided written informed consent, and the UCSD Brief Assessment of Capacity to Consent⁹ was administered to assess decisional impairments in patients. In cases where patients were unable to provide informed consent, their legally authorized representative granted consent, and patient assent was also obtained.

Clinical Instruments

Multiple clinical instruments were used to assess the disease status of each NPC patient and consisted of the NPC-SI¹⁰, the NPC-CDB¹¹, and the Scale for Assessment and Rating of Ataxia (SARA)¹². All participants completed the Neuro-Quality of Life (QoL)¹³, Patient-Reported Outcome Measurements Information System (PROMIS), and Scale for the Assessment of Positive Symptoms (SAPS)¹⁴ questionnaires. Legal guardians completed questionnaires for severely impaired NPC patients.

NPC Biomarker Assessment

Plasma and serum were extracted from 10 mL blood and frozen at −80°C. Plasma biomarkers (i.e., C-triol, 7-KC, and PPCS) were assessed for each patient using liquid chromatography-tandem mass spectrometry (Mayo Clinic BioPharma Diagnostics, Rochester, MN).

Motor and Cognitive Test Battery

National Institutes of Health (NIH) Toolbox Motor Batteries were used to evaluate upper-extremity motor functioning and consisted of the 9-hole pegboard dexterity test and hand grip dynamometry¹⁵. Lower-limb function assessments were omitted due to non-ambulatory patients. Neurocognitive evaluation included the Delis–Kaplan Executive Function System (D-KEFS) Verbal Fluency and Trail Making Tests¹⁶ and the Wechsler Abbreviated Scale of Intelligence II (WASI-II) Vocabulary, Matrix Reasoning, and Digit Span Tests¹⁷.

Speech Acquisition and Analysis

Speech samples were recorded using Redenlab Online^™ on an iPad (Apple Inc). Tasks included (i) 30 single-syllable consonant-vowel-consonant words (e.g., box, say, big) as written visual stimuli, (ii) a 10-second sustained vowel /a:/, and (iii) 10 seconds of rapid syllable repetition (/pata/)¹⁸, with the latter two tasks repeated twice to minimize unfamiliarity impact. Tasks spanned a continuum of cognitive motor complexity, from pure motor elements (e.g., vowel) to contemporaneous language production (e.g., single-syllable word production)¹⁹.

Voice quality features for acoustic analysis were derived from the sustained vowel and included syllables per second, variability of syllables per second, syllable duration, variability of loudness (intensity CoV; coefficient of variation), harmonics to noise ratio (HNR), fundamental frequency variability (f0 CoV), and cepstral peak prominence smoothed (CPPS)¹⁹. Timing features included syllabic rate, silence length, and their variability²⁰.

Neuroimaging Acquisition

MRI data were acquired on a Siemens 3T Prisma using a 64-channel head coil at BCH. Scan protocols and parameters followed the Human Connectome Project²¹. Anatomical MRI used multi-echo magnetization-prepared rapid acquisition with gradient echo (MEMPRAGE) to assess morphological properties²¹. T2 turbo spin echo data were collected using the following parameters: 1 mm isotropic, TR/TE=7000ms/394ms, CAIPIRINHA = 2, 160 sagittal slices. Scan time: 3:53 min. Diffusion tensor imaging (DTI) was acquired to assess white matter (WM) integrity and interregional structural connectivity²¹. Resting-state functional MRI (rsfMRI) was employed to estimate intrinsic functional connectivity²¹.

Task fMRI data were obtained using a sparse-sampling fMRI protocol²² to estimate functional responses during speech production. T2*-weighted echo-planar images were acquired using a multi-band sequence with 2 mm³ resolution, TR/TE = 800ms/37ms, 72 slices, GRAPPA = 2, multiband factor = 6. A temporal delay of 5200 ms (silent period) was incorporated to sample the estimated peak blood-oxygen-level-dependent (BOLD) response elicited by the speech production. 30-word conditions and 30 baseline (‘^^^’) conditions, with participants remaining silent during baseline, were presented in a pseudorandomized manner, and consisted of words identical to those presented in the speech production task that was performed outside of the MRI scanner. Scan time: 6:00 min.

Neuroimaging Analysis

Anatomical MRI:

MEMPRAGE images were processed with the FreeSurfer (http://surfer.nmr.mgh.harvard.edu) recon-all pipeline to extract the morphometric properties of cortical and subcortical brain structures, implementing the Desikan-Killiany atlas²³ and the Automatic Segmentation of Subcortical Structures (ASEG) atlas, respectively. Cerebellum analysis employed the spatially unbiased infratentorial template (SUIT) Toolbox²⁴ in Statistical Parametric Mapping (SPM) 12 (http://www.fil.ion.ucl.ac.uk/spm/), with lobular volumes extracted from native-space images using SUIT Toolbox scripts. ITK-SNAP (version 3.8)²⁵ was used for manual segmentation and volume calculation of the choroid plexus (CP) in lateral ventricles on T1-weighted images and cross-checked with T2-weighted turbo spin echo images.

DTI:

DTI data was processed using 3D Slicer with default parameters (http://dmri.slicer.org)²⁶. WM tracts were automatically identified using the whitematteranalysis package advanced tractography parcellation strategy, and group-level comparisons of fractional anisotropy (FA) values were conducted. Structural connectivity matrices were generated for each subject using the automated anatomical labeling template (AAL), resulting in 116 network nodes. DTI graph theoretical metrics were computed for binarized connectivity matrices describing pairwise connectivity between nodes using the Brain Connectivity Toolbox²⁷. Specifically, node degree, betweenness centrality, and clustering coefficient were calculated.

Speech production fMRI:

Preprocessing utilized FMRIB’s Software Library (FSL) version 6.0.5. A general linear modeling (GLM) approach incorporated an explanatory variable matching the pseudorandomized presentation of three-letter words or baseline condition. Between-group analysis was performed using a fixed effects model. Analysis at the region-of-interest (ROI) level was also performed and involved averaging parameter estimates across all voxels within ROIs defined by the Harvard-Oxford and Desikan-Killiany-Tourville atlases²³.

rsfMRI:

A whole-brain ROI resting-state functional connectivity analysis was performed using the CONN-fMRI fc toolbox v18.b (https://web.conn-toolbox.org) in conjunction with SPM 12 using default parameters.

Statistical Analysis

To compare clinical and behavioral endpoints between NPC patients and HCs, non-parametric Mann-Whitney U tests (two-tailed, α = 0.05) were performed. Neuroimaging endpoints were compared between NPC patients and HCs by means of multivariate analyses of covariance (MANCOVAs) with age, sex, and total intracranial volume as covariates. P-values were corrected using the Bonferroni method. Pearson correlation analyses were performed to determine associations among study measures quantified. A sparse canonical correlation analysis (CCA) was performed on the data of NPC patients and HCs combined to investigate the multivariate relationships between neuroimaging measures and clinical and behavioral endpoints²⁸. All pairs of canonical variables (CVs) with correlations larger than 0.5 were reported in the results. To understand the potential clinical interpretation of CVs corresponding to neuroimaging endpoints, Pearson correlations between scores of these CVs and six relevant biomarkers and clinical scores (C-triol, 7-KC, and PPCS levels, and NPC-SI, NPC-CDB, and SARA scores) among NPC patients were reported. The sparse CCA was conducted using PMA package in R.

Results

Study Participant Demographics and Clinical Characteristics

Three females and five males with NPC (age: mean ± standard deviation (SD): 36.00 ± 14.27 years) and three female and five male HCs (36.13 ± 14.23 years) completed the study protocol (Supplemental Table 2). Genetically confirmed, adult patients with NPC presented with variable clinical severity as revealed by the NPC-SI, NPC-CDB, and SARA as well as plasma biomarker expression of C-triol, 7-KC, and PPCS. The mean ng/mL ± SD expression of biomarkers in our NPC cohort was 54.69 ± 12.62 (130.06 ± 30.03 nmol/L) for C-triol, 108.20 ± 56.09 (270.14 ± 140.01 nmol/L) 7-KC, and 10.20 ± 5.09 (20.05 ± 10.00 nmol/L) for PPCS. Overall, higher scores on the NPC-SI were associated with higher expression of C-triol (r = 0.75, p = 0.03), 7-KC (r = 0.80, p = 0.02), and PPCS (r = 0.63, p = 0.09). Significant associations were not detected between the NPC-CDB or SARA with any plasma biomarker.

Clinical Questionnaire Responses

Responses on the NeuroQoL-swallowing and PROMIS-physical function questionnaires significantly differed between patients with NPC and HCs (Supplemental Table 3). On the SAPS, only NPC 002 and 006 experienced hallucinations and delusions.

Motor and Cognitive Impairments in NPC

Patients with NPC demonstrated significant hand-motor and cognitive deficits across all tasks, with greater deficits appearing in specific neurocognitive domains, for example, on the trail number (numerical processing), trail vision (attention), and fluid switch response (cognitive flexibility) tasks (Figure 1). NPC 006 was unable to perform the trail letter, trail switching, trail motor, and 9-hole pegboard dexterity tests.

Figure 1: Motor Deficits and Cognitive Impairment in NPC.

(A) The average age-corrected standard scores ± SEM for the 9-hole pegboard and grip strength tasks are reported. For both tasks – and for both dominant and non-dominant hands – NPC patients scored significantly lower than HCs. * Indicates significant group-differences determined by Mann-Whitney U tests at p < 0.05. (B) Significant differences in motor (trail motor) and cognitive (trail number, letter, switching, vision, and fluency switch response) functioning were observed between NPC patients and HCs, as assessed via the Delis-Kaplan Executive Function System (D-KEFS). NPC patients scored significantly lower than HCs for all measures. Values are mean scaled scores ± SEM. * Indicates significant group-differences determined by Mann-Whitney U tests at p < 0.05. (C) NPC patients scored significantly lower on the WASI-II digit span task than HCs. Values are mean scaled total scores ± SEM. * Indicates significant group-differences determined by Mann-Whitney U tests at p < 0.05. (D) WASI-II FSIQ-2 scores were calculated using Vocabulary and Matrix Reasoning assessments. NPC patients had significantly lower WASI-II FSIQ-2 scores than HCs, indicative of decreased cognitive ability in the NPC patient sample. Values are mean scores ± SEM. * Indicates significant group-differences determined by Mann-Whitney U tests at p < 0.05.

Reduced Cerebellar and Subcortical Volumes, Cortical Thinning, and Increased Choroid Plexus Volumes in NPC

Cerebellar atrophy in patients with NPC compared to HCs was particularly evident within lobule VI and crus I subdivisions (Figure 2; Supplemental Table 4). Regional cortical thinning was mostly found in areas of the frontal and temporal cortex in patients with NPC compared to HCs. Analysis of subcortical volumes revealed significant reductions of the bilateral thalamus and multiple striatal sub-divisions for NPC patients relative to HCs (Figure 2). CP volume was greater in patients with NPC compared to HCs (controlled for ventricular and supratentorial volume, F(1,13) = 5.23, p = 0.04; controlled for ventricular and supratentorial volume, age, and sex, significant at F(1,11) = 4.95, p = 0.05) (Figure 2). CP volume was significantly associated with thalamic volumes (left: r = −0.50, p = 0.05; right: r = −0.53, p = 0.03) and plasma PPCS (r = 0.71, p = 0.05).

Figure 2. Reductions in cerebellar, cortical, and subcortical morphological properties in NPC.

(A) Atlas-based segmentation of cerebellar subdivisions and cortical gyri. Cerebellar subdivisions are depicted on the spatially unbiased infratentorial template (SUIT) flatmap. Cortical regions showing significant group-level differences between NPC patients and HCs are highlighted. Individual CNS regions were defined using the Desikan-Killiany atlas²³. (B) MANCOVA analysis with age and sex as covariates revealed significant reductions in volume in the left lobule VI and both left and right crus I of the cerebellum in NPC patients relative to HCs. Values are mean volumes ± SEM. * Indicates significant at Bonferroni corrected p < 0.05. (C) Reduced cortical thickness was found for NPC patients in several cortical areas, including the bilateral posterior cingulate, bilateral pars triangularis, and left rostral middle frontal gyrus. Analysis consisted of MANCOVA with age, sex, and intracranial volumes as covariates. Values are mean thickness ± SEM. * Indicates significant at Bonferroni corrected p < 0.05. (D) Significant subcortical volume reductions were found bilaterally in the thalamus, putamen, nucleus accumbens, and caudate for MANCOVA analysis with age and sex as covariates. Values are mean volumes ± SEM. * Indicates significant at Bonferroni corrected p < 0.05. (E) CP volume was greater in NPC patients relative to HCs. The ANCOVA analysis was controlled for ventricular and supratentorial volume. Values are mean volumes ± SEM.

Reduced Fractional Anisotropy and Alterations in Nodal Network Properties in NPC

Compared with HCs, significant reductions in FA values were observed in WM tracts of NPC patients, specifically within the intracerebellar input and Purkinje tracts as well as cingulum bundles (Figure 3). Interestingly, while group differences in cerebellar tracts were more robust within the right cerebellar hemisphere, the cingulum bundle showed more significant reductions in FA in the left cerebral hemisphere in the NPC cohort. Graph theory analysis of DTI data revealed reduced nodal structural connectivity of several speech-motor (i.e., pars triangularis), motor (i.e., anteroventral thalamus and supplementary motor cortex), and multimodal (i.e., superior frontal gyrus, superior parietal gyrus, and Rolandic operculum) regions, as indicated by generally reduced node degree and betweenness centrality, but higher clustering coefficient (Figure 3).

Figure 3. Alterations in white matter microstructure and structural network connectivity in patients with NPC.

(A) Significant group-level differences in FA were particularly observed in the right intracerebellar input and Purkinje tracts as well as the left and right cingulum bundles (see yellow highlighted tracts). (B) DTI-based graph theoretical MANOVA analysis of nodal properties revealed reduced nodal degree in the left pars triangularis, left calcarine, right inferior parietal gyrus, and right anteroventral thalamus in NPC patients relative to HCs. Overall reduced betweenness centrality was found in the left pars triangularis, right superior and inferior parietal gyri, and cerebellar vermis III. However, increased betweenness centrality was found in the right Rolandic operculum in NPC patients. Higher clustering coefficient was found in the left lingual gyrus, right superior frontal gyrus, right supplementary motor area, and cerebellar vermis III of NPC patients. Changes in node degree and betweenness centrality indicate differences in connectivity and efficiency of information flow, respectively, while altered clustering coefficient suggests differences in local segregation. * Indicates significant at Bonferroni corrected p < 0.05.

Reduced CNS Activity During Speech Production in NPC

NPC 004 demonstrated a strong BOLD fMRI response during production of 30 three-letter words, whereas functional activity was comparatively lower in NPC 005 who had greater extent of neurodegeneration (Figure 4). In accordance with fMRI data, recordings collected during syllable repetition demonstrated more intact speech production for NPC 004 relative to NPC 005. In accord with single patient finding, group-level comparison of speech production fMRI responses revealed significantly reduced BOLD activity in NPC patients relative to HCs within key hubs of the speech production (supplementary motor cortex) as well as the auditory (Heschl’s gyrus) processing circuits (Figure 4; Supplemental Figure 1). Analysis of speech production behavioral metrics revealed speech-motor timing and coordination difficulties (longer syllable duration [z = −2.52, p = 0.01] and fewer syllables per second [z = −2.52, p = 0.01]) experienced by patients with NPC. All patients were able to perform the speech production tasks inside and outside of the MRI. Finally, a cross-correlation analysis encompassing functional responses within the speech-production network and behavioral speech production metrics further identified brain-behavior relationships (Figure 4; Supplemental Figure 2).

Figure 4. Speech Production Functional MRI and Acoustic Metrics Results.

(A) Study participants generated three-letter words (e.g., box, say, big) during acquisition of fMRI data. Single-subject fMRI responses during speech showed robust CNS activity, but with more activity for NPC 004 vs. NPC 005. BOLD activation maps were superimposed on subject-specific anatomical images. Statistical maps were cluster-size corrected at p < 0.05. Analysis of speech recordings made outside of the MRI scanner and during syllable repetition revealed more intact speech for NPC 004 vs. NPC 005. (B) Regions showing significant reductions in BOLD activity during speech production in NPC patients compared to HCs are displayed on the MNI template brain. (C) Group-level MANOVA comparison of parameter estimates extracted from atlas-defined ROIs further revealed significant reductions (NPC < HC) in BOLD responses during speech production in speech-motor network hubs. * Indicates significant at Bonferroni corrected p < 0.05. (D) Cross-correlation analysis of NPC patients and HCs combined identified significant associations between voice quality features and BOLD activity during speech production. For example, fundamental frequency variability (f0 CoV) was negatively correlated with the BOLD responses in the right supplementary motor cortex (r = −0.50, p = 0.05), right dorsal cingulate motor area (right: r = −0.64, p = 0.01), left Heschl’s gyrus (r = −0.59, p = 0.02), and left ventral premotor cortex (r = −0.56, p = 0.03), while syllables per second was significantly correlated with activity within the left supplementary motor cortex (r = 0.59, p = 0.02), left anterior central operculum (r = 0.68, p = 0.01), bilateral anterior cingulate (left: r = 0.64, p = 0.01; right: r = 0.60, p = 0.01), left dorsal cingulate motor area (r = 0.51, p = 0.05), right Heschl’s gyrus (r = 0.55, p = 0.03), and left anterior insula (r = 0.57, p = 0.02).

Altered Resting-State Network (RSN) Functional Connectivity in NPC

Seed-based RSN analysis revealed altered coupling between the posterior cingulate cortex and right cerebellar crus I in NPC patients as compared to HCs, as the NPC cohort showed a positive functional association between regions and HCs presented with an anti-correlation (NPC: r = 0.16; HC: r = −0.24). ROI-to-ROI RSN analysis showed a trend of increased connectivity identified between the cerebellum and hubs within the Default Mode Network (DMN) (e.g., posterior cingulate cortex) in NPC patients compared to HCs, whereas a trend in diminished functional connectivity was observed between the medial prefrontal cortex and cerebellar lobule VIII (significant at uncorrected p < 0.001) (Supplemental Figure 3).

Multivariate Associations between Different Modalities

Sparse CCA linked neuroimaging measures to three categories of clinical and behavior endpoints: acoustic metrics, motor assessment and neurocognitive performance (Figure 5). In the correlation between neuroimaging and acoustic metrics, six CV pairs emerged, predominantly comprising rsfMRI, speech production fMRI, and DTI graph theoretical endpoints. Morphological measures contributed exclusively to CV 5 (variability in syllables per second). Two DTI FA measures were selected: the right intracerebellar parallel tract for CV 3 (syllable duration) and the left intracerebellar parallel tract for CV 5 (variability of syllables per second). When correlating neuroimaging endpoints to motor assessment, five CV pairs emerged, all exclusively involving rsfMRI connectivities. For neurocognitive scores, CCA identified seven CV pairs. Most neuroimaging CVs were loaded by rsfMRI measures, with CV 2 (trail vision) and CV 5 (fluency switch) also exhibiting non-zero loadings for the DTI clustering coefficient of the left lingual gyrus, and CV 3 (full scale IQ) displaying a non-zero loading for the DTI clustering coefficient of the right superior frontal gyrus.

Figure 5. Canonical Correlation Analysis Links Neuroimaging Endpoints and Clinical and Behavioral Measures.

(A) The loading matrices of CV pairs linking neuroimaging endpoints to acoustic metrics, motor assessment and neurocognitive metrics in the sparse CCA are illustrated. CCA of acoustic metrics identified six CV pairs, CCA of motor assessment endpoints revealed five pairs of CVs, and CCA of neurocognitive metrics identified seven CV pairs. (B) The neuroimaging CVs of the acoustic metrics consisted of individual measures from the rsfMRI, speech production fMRI, and DTI graph theoretical analysis. The morphological MRI measures (i.e., only choroid plexus volume) were included by CV 5, corresponding to the variance in syllable duration. Only two DTI measures were included and consisted of the right intracerebellar parallel tract for CV 3 (syllable duration) and the left intracerebellar parallel tract for CV 5 (variability of syllables per second). The neuroimaging CVs associated with motor assessment endpoints were exclusively loaded by rsfMRI connectivity measures. The CCA of neurocognitive metrics identified non-zero loadings only for rsfMRI endpoints. (C) The Pearson correlations between neuroimaging CVs and the seven NPC-related biomarkers are displayed for CVs derived from the CCA of acoustic metrics, motor assessment, and neurocognitive score. The neuroimaging CVs of acoustic metrics demonstrated a significant correlation for CV 3 with SARA total score, linking the voice quality feature of fundamental frequency variability (f0 CoV) to SARA scores. When considering CVs derived from motor assessment, significant correlations were found between CV 4 (trail motor) and expression levels of C-triol and 7-KC. For the neurocognitive CCA, both CV 1 and CV 7 (fluency switch and trail switch, respectively) are significantly correlated with C-triol and 7-KC while CV 2 (fluency switch) are significantly correlated with the NPC-SI.

Correlations between Neuroimaging CVs and NPC Biomarkers

Pearson correlation analysis between neuroimaging CVs and six NPC-related biomarkers revealed additional brain-behavior relationships (Figure 5). For the neuroimaging CVs derived from the CCA of acoustic metrics, the significant correlation was attained between CV 6 (f0 CoV) and SARA scores. For motor assessments, significant correlations were detected for CV 4 (trail motor) with C-triol and 7-KC biomarkers. For the neurocognitive CCA, neuroimaging CV 1, CV 2, CV 4, and CV 7 are significantly correlated with C-triol levels. CV 1, CV 2, CV 4, and CV 7 correspond to fluency switch, trail vision, trail letter, and trail switching scores, respectively, and are measures of attention, processing speed, and cognitive flexibility.

Discussion

In NPC, aberrant lysosomal functioning initiates a range of central manifestations including neurodegeneration, impairments in motor and cognitive performance, or psychosis²⁹. To date, identifying biological factors or signatures in NPC that explain or associate with the severity of neuropsychiatric presentation has been challenging as the expression of plasma biomarkers (e.g., PPCS) or the degree of neurodegeneration (e.g., hippocampal volume loss) do not consistently track with symptom severity or behavioral phenotypes³⁰. Currently, there is a critical unmet need to identify objective CNS biomarkers that (i) elucidate neurobiological causes of neuropsychiatric symptoms, particularly in regard to motor and cognitive deficits, (ii) have the sensitivity to detect early treatment effects, (ii) can help understand treatment resistance, and (iii) can be utilized to stage patients with NPC. To this end, an initial systematic and multidisciplinary evaluation of adult patients with NPC with variable disease severity was performed with a focus on unfurling the neurobiological underpinnings in two core symptom domains - (speech-)motor deficits and cognitive impairments.

Variability of NPC disease severity was revealed by several clinical tools as well as plasma biomarkers, where only higher scores on the NPC-SI associated with elevated expression of circulating biomarkers (i.e., C-triol and 7-KC). Overall, NPC patients presented with consistent deficits in the physical domain, as indicated by the Neuro-QoL-swallowing and PROMIS-physical function. This finding extended to hand-motor functioning as well as cognitive functioning, as NPC patients showed worse performance than HCs on several measures (e.g., D-KEFS).

From a morphological perspective, our study confirms the presence of gray and white matter loss distributed throughout the cerebellum, cortex, thalamus, and striatum in NPC patients, but provides more specificity regarding which neuronal substrates are affected^6–8. Within the cerebellar gray matter, we identified focal volume reduction in lobule VI and crus I subdivisions, which are involved in higher-level processes such as, motor processing, executive functioning, working memory, and affective processing³¹. Similar atrophy was observed within fronto-temporal areas implicated in attention, working memory and language-related processing³². Subcortically, reduced volumes were found in structures relevant to sensorimotor control, executive functioning, and affective processing³³, consisting of the bilateral thalamus as well as the dorsal (caudate and putamen) and ventral (nucleus accumbens) striatum. Additionally, we found evidence of choroid plexus enlargement in patients with NPC, which inversely correlated with bilateral thalamic volumes and associated positively with levels of PPCS³⁴. The choroid plexus plays an important role in cerebrospinal fluid production and immunosurveillance of the CNS³⁵. Investigations in other conditions, including psychosis, neuropsychiatric lupus, and multiple sclerosis, have linked choroid plexus enlargement to increased levels of neuroinflammation or cognitive dysfunction^{35, 36}. Although our study cannot confirm neuroinflammation, prior NPC investigations using 18 kDa translocator protein (TSPO) positron emission tomography (PET) have indicated neuroinflammatory mechanisms³⁷.

Assessment of WM using DTI revealed reduced FA of the intracerebellar input and Purkinje fibers, a finding that is in line with preclinical observations where Purkinje cell neurodegeneration has been noted in mouse models of NPC³⁸. Additionally, we saw reduced FA within the cingulum bundles, which project between fronto-temporal and cingulate cortices, regions that have correspondingly shown reductions in cortical thickness in NPC patients relative to HCs. Findings related to FA are particularly intriguing considering evidence that available therapies (e.g., Miglustat) can impact CNS microstructure, and therefore, FA could be representative of a sensitive pharmacodynamic biomarker⁸. Graph theoretical analysis of DTI data suggested structural network differences of nodal properties, as indicated by reductions in structural connectivity strength (i.e., node degree) and efficiency of information flow (i.e., betweenness centrality), but increased local segregation (i.e., clustering coefficient). Together, these findings point to altered network connectivity of main hubs involved in speech-motor production and processing, comprising the pars triangularis, inferior parietal lobe, and supplementary motor cortex.

The fMRI speech-production paradigm (i.e., three-letter words phonation) demonstrated reduced BOLD responses in speech-motor and auditory or language processing areas in NPC patients. In line with morphological findings, we saw less activity specifically in cerebellar lobule VI. Several robust decreases in BOLD activity during speech production were localized to motor processing regions. These regions include the supplementary motor cortex, a region involved in speech motor control as an initiation and timing interface, dorsal cingulate motor area, involved in simple speech tasks³⁹, and ventral premotor cortex, an area implicated in speech production and perception²². In accord with reductions in BOLD responses, analysis of acoustic metrics derived from speech recordings during syllable repetition outside of the MRI environment demonstrated longer syllable duration and fewer syllables per second, indicating NPC patients take longer to say sound units and have slower speech, respectively. This brain-behavior relationship was further confirmed by associations found between the observed reductions in functional responses of the speech-production paradigm and the performance on behavioral speech-production metrics.

In patients with NPC, rsfMRI analysis revealed altered functional coupling between crus I and the posterior cingulate cortex, a key hub of the DMN. Specifically, these regions showed inverse correlations within HCs, whereas NPC patients presented with a loss of the normal pattern of anti-correlation (i.e., positive associations), indicating faulty decoupling of the DMN with crus I. Although not significant, the ROI-to-ROI resting-state connectivity analysis revealed a trend of increased coupling of the cerebellum and important hubs within the DMN (i.e., posterior cingulate cortex and lateral parietal lobe). However, a trend of reduced coupling between another important node of the DMN, the medial prefrontal cortex, and the cerebellar lobule VIII was found.

Our sparse CCA further characterized brain-behavior relationships based on acoustic metrics, motor assessments, and neurocognitive performance by linking these endpoints with findings of neuroimaging modalities. Acoustic metrics were paired with almost all measures, including microstructural findings, nodal network properties, speech production fMRI, and rsfMRI connectivities, whereas CCA of motor and neurocognitive performances were exclusively paired with rsfMRI measures. Pearson correlation analysis revealed that motor performance and cognitive flexibility associated with expression levels of C-triol and 7-KC biomarkers, but not PPCS, and with NPC-SI scores.

The current set of findings provide new information related to the brain-behavior relationships associated with speech-motor, hand motor, and cognitive impairments in adult patients with NPC. In addition, this investigation describes a framework for further investigation of the neurobiological underpinnings of NPC in future natural history studies involving a broader cohort of NPC patients, including pediatric populations or in the context of a clinical trial. Nonetheless, multiple study limitations should be considered. Firstly, this initial phenotyping investigation involved a small sample size. Within the NPC cohort, heterogeneity across patients may exist in terms of severity of neurodegeneration, neuropsychiatric symptoms, or overall disease burden and introduce study variability, potentially reducing sensitivity to detect further significant differences. Despite these limitations, in this well-controlled and prospective study, we observed altered (micro-)structural and functional properties in CNS regions associated with motor and cognitive impairments. Nonetheless, it is important to consider that NPC is an ultra-rare disease and larger cohorts may be feasible in future multi-center investigations. Longitudinal tracking of brain changes from disease onset through progression of NPC will additionally be needed to characterize changes in gray and white matter and their associations with clinical and behavioral endpoints over time. Moreover, we hypothesize that in future NPC studies involving larger sample sizes and longitudinal patient evaluation, a deep phenotyping approach that incorporates neuroimaging biomarkers should also evaluate blood-based biomarkers of CNS disease such as NfL, glial fibrillary acidic protein, phosphorylated tau, among others⁵ In conclusion, this investigation takes a step towards better understanding of structural and functional CNS alterations in NPC, particularly in the context of speech production deficits and cortico-cerebellar circuitry. This study also indications the potential role of the protective CNS barriers (i.e., choroid plexus) in NPC pathophysiology. Finally, a deep phenotyping framework is described, which upon further implementation, may provide a novel and robust CNS biomarkers in NPC that can be used to examine patients and enhance therapeutic evaluation.

Data Availability

The data used in our current study are available upon request to the corresponding author and following implementation of a formal data sharing agreement.