Evolution of structural neuroimaging biomarkers in a series of adult patients with Niemann-Pick type C under treatment

Neurometabolic disorders encompass a very wide spectrum of diseases, of heterogeneous clinical and radiological presentations. While their diagnosis remains difficult, treatments are emerging, emphasizing the need of non-invasive techniques to evaluate their efficacy. Studying neurometabolic diseases also contributes to the better understanding of neurodegenerative disorders related to metabolic dysfunction such as Parkinson disease [1] or Huntington disease [2]. Niemann-Pick type C disease (NPC) is a genetic lysosomal lipid storage disorder of autosomal recessive inheritance. It results from the mutation of either NPC1 (95%) or NPC2 (5%). The resulting impairment of the processing and utilization of endocytosed cholesterol leads to the accumulation of free cholesterol and glycosphingolipids in many tissues, including the brain. Its incidence is difficult to evaluate precisely, and lies around 1 in 120,000 births, but is probably underestimated, particularly in the adult population [3].

Diverse clinical presentations exist, ranging from a very severe perinatal form to a neurodegenerative adult-onset form. Neurological symptoms include mental retardation, cerebellar ataxia with dysarthria and dysphagia, dementia and vertical supranuclear gaze palsy (highly evocative). Psychiatric disorders, epilepsy and dystonia may also occur. In adults, isolated ataxic and/or psychiatric forms have been described, underlining the complexity of the diagnosis. Visceral involvement (liver, spleen, sometimes lungs) is also rather frequent [3, 4]. The diagnosis of NPC was historically made using fillipin staining but is now currently based on elevated plasma oxysterols (cholesterol oxidation products) [5], and confirmed by the molecular analysis of NPC1 and NPC2 genes. A substrate reduction therapy exists for NPC patients using miglustat, an inhibitor of glucosylceramide synthase, hence preventing glycosphingolipids accumulation. Miglustat has shown a partial clinical efficiency [6, 7].

Conventional cerebral magnetic resonance imaging (MRI) can be normal in NPC patients or shows unspecific cortical or cerebellar atrophy [8, 9]. White matter abnormalities may be present on T2-weighted images, especially in the severe infantile form [3, 10]. Cerebral volumetry [8, 9, 11] and diffusion tensor imaging (DTI) studies have been performed in NPC [9, 11, 12], but rarely in combination [9, 12] and mainly in children, and few neuroimaging studies have assessed the benefit of treatment in NPC [8, 12, 13]. Accordingly, we performed a retrospective study using both volumetry and DTI on a large cohort of adult patients with NPC at baseline, compared to controls, and under treatment.

At baseline, volumetric data were available for 13 NPC patients while DTI metrics were available for 8 NPC patients. Their clinical and demographic characteristics are shown on Table 1. Eleven NPC patients were then followed over a mean of 5 years (1–9 years) with MRI available at different time-points, as one patient had just been diagnosed and another patient was lost to follow up. These 11 patients were treated with miglustat, but two interrupted their treatment after 6 months, the first patient because of a severe weight loss and the second patient due to major digestive and psychiatric side effects. Follow-up time extended up to 9 years. The 2 patients who interrupted treatment had baseline MRI, and radiological follow-up at 1.5 and 4.5 years. Among all patients with MRI available for follow-up, 3 worsened clinically: one with sustained miglustat therapy and the 2 patients who interrupted treatment. All of the others had a global stabilization-i.e., loss or gain in overall score ≤1 (Table 2).

For volumetric analyses at baseline, the control group (n = 13) had a mean age of 35 (±12) years and a sex-ratio comparable to patients (M/F: 7/6). Eight NPC patients had radiological follow-up at 1 (n = 6), 1.5 (n = 3), 2 (n = 6), 3 (n = 3), 4 (n = 4), 4.5 (n = 3), 5 (n = 4) or 9 (n = 3) years. Analyses with VBM and ROI methods both showed global atrophy and decreased volume in the basal ganglia of NPC patients compared to controls (Fig. 1, Additional file 1: Table S1). The mean volume (% of total intracranial volume ± SD [min-max]) of superior cerebral peduncles was smaller in the patients group compared to controls – 0.01% ±0.002 [0.01–0.02] vs. 0.02% ± 0.002 [0.01–0.02], p = 0.003 – as well as the anterior corpus callosum – 0.04% ± 0.01 [0.02–0.07] vs. 0.06% ±0.01 [0.04–0.09], p = 0.006. When looking at the evolution of the volume of regions of interest in patients treated with miglustat, a significant atrophy was found over time in the thalamus, pallidum, and amygdala (data not shown). There was a tendency to slower atrophy in caudate nucleus, putamen and corpus callosum in treated patients (n = 3) compared to patients who withdrew treatments (n = 2) but that was not significant (Fig. 2a and b). We found no significant correlations between volume changes and clinical scores. There was no significant difference either in volume changes between patients who worsened clinically and patients who stabilized.

For DTI analyses, the control group (n = 8) had a mean age of 34 years (±14), and a sex-ratio comparable to patients (M/F: 5/3). Four NPC patients had radiological follow-up at 1, 2, 3, 4 (n = 4), 4.5 (n = 2), 5 (n = 4) and 9 (n = 3) years. At baseline, TBSS analyses found significantly decreased FA in the posterior corona radiata and left forceps major (Fig. 3). ROI analyses found a significantly decreased FA in the corpus callosum, anterior corona radiata and internal capsule, posterior thalamic radiations, cerebellar peduncles, cingulate gyrus, and longitudinal and fronto-occipital fasciculii (Table 3). RD was globally increased in NPC patients compared to patients (0.00062 ± 0.00003 [0.00064–0.00065] vs. 0.00075 ± 0.00007 [0.00070–0.00081, p < 0.001). After 2 years of treatment with miglustat, TBSS analyses showed significantly increased FA in the corpus callosum, forceps minor and superior region of corona radiata, as well as significantly decreased RD in the corpus callosum (Fig. 4). ROI analyses also showed increased FA after 2 years of treatment in the right hippocampal cingulum (0.321 ± 0.170 vs 0.308 ± 0.546, p = 0.029) and in the anterior corona radiata (0.431 ± 0.178 vs 0.349 ± 0.180, p = 0.029). However, such difference was no longer observed at further time points (years 3, 4, 5 and 9) compared to baseline MRI. FA in the corona radiata tended to worsen more rapidly in non-responding or untreated patients (Fig. 5a). Moreover, the overall score and the dysphagia score seemed more likely to stabilize if the change in FA was minor (Fig. 5b).

The novelty of our work lies in the following points: (i) this study is the largest performed in adult NPC patients using two different neuroimaging modalities; (ii) the whole brain was analyzed and not only regions of interest; (iii) most patients had follow-up scans under treatment. As suggested by visual inspection from brain MRI of NPC patients [4, 8‐10], our volumetric analyses showed that NPC patients display cerebral atrophy compared to healthy controls, especially the corpus callosum and the cerebral peduncles. Treatment with miglustat failed to prevent further atrophy of the basal ganglia over time. However, it is unclear how volume changes reflect disease severity and/or therapeutic efficacy in NPC since no correlation between clinical and MRI metrics were found in our cohort. Likewise, changes in brain volumes and clinical expression of neurological conditions are not always related as shown in neurodegenerative disorders such as Alzheimer disease [18]. Thus, it remains to be established whether brain volumetry can generate clinically relevant biomarkers in NPC, especially in longitudinal research studies. On the other hand, DTI analyses showed decreased FA in NPC patients compared to controls, especially in the corpus callosum, as previously shown [12, 19], but also the internal capsule, the corona radiata and the cingulate gyrus-with a possible correlation with clinical manifestations for the two latter. The greater changes observed in RD suggest a preferential alteration of the myelin sheath in NPC patients. Overall, the DTI changes identified in NPC patients in key white matter regions may directly reflect cerebral metabolic alterations. These data are coherent with recent work in NPC dedicated to white matter analysis [20]. Of note, treatment with miglustat led to an early, although transient, improvement of DTI metrics.

This study has some limitations. First, due to their acquisition in a clinical setting, some data did not pass quality control requirements and were discarded. Second, combining data from different field strengths can bias the outcome of semi-automated methods such as VBM analyses. Still, data acquisition in clinical routine can be implemented in a multicentric setting with a more systematic radiological follow-up. Third, the sample size of the study is relatively small, a problem inherent to any rare neurometabolic disease, although greater than previously reported in adults with NPC [8, 9, 12]. Last, it can be difficult to establish the relevance of neuroimaging parameters in NPC as (i) there is a lack of natural history data prior to miglustat, (ii) a vast majority of patients are now treated with miglustat, preventing us from assembling an untreated control group and, (iii) the clinical impact of miglustat is relatively weak [6, 7], so that a strong biomarker effect is less likely.

Overall this work suggests that DTI can provide relevant biomarkers of disease progression in NPC, including in a clinical setting, but should be confirmed in prospective studies, especially if drugs with a greater effect on disease progression become available.