A Differential Proteomics Study of Cerebrospinal Fluid from Individuals with Niemann-Pick Disease, Type C1

Abstract

Niemann-Pick, type C1 (NPC1) is a fatal, neurodegenerative disease, which belongs to the family of lysosomal diseases. In NPC1, endo/lysosomal accumulation of unesterified cholesterol and sphingolipids arises from improper intracellular trafficking resulting in multi-organ dysfunction. With the proximity between the brain and cerebrospinal fluid (CSF), performing differential proteomics provides a means to shed light to changes occurring in the brain. In this study, CSF samples obtained from NPC1 individuals and unaffected controls were used for protein biomarker identification. A subset of these individuals with NPC1 are being treated with miglustat, a glycosphingolipid synthesis inhibitor. Of the 300 identified proteins, 71 proteins were altered in individuals with NPC1 compared to controls including cathepsin D, and members of the complement family. Included are a report of 10 potential markers for monitoring therapeutic treatment. We observed that pro-neuropeptide Y (NPY) was significantly increased in NPC1 individuals relative to healthy controls; however, individuals treated with miglustat displayed levels comparable to healthy controls. In further investigation, NPY levels in a NPC1 mouse model corroborated our findings. We posit that NPY could be a potential therapeutic target for NPC1 due to its multiple roles in the central nervous system such as attenuating neuroinflammation and reducing excitotoxicity.

Introduction

Niemann-Pick, type C (NPC) is an autosomal recessive, neurodegenerative disease caused by mutations in either NPC1 or NPC2 ^[1], which encode proteins that are involved in cholesterol and lipid trafficking through the lysosome ^[2]. As a result of biallelic pathological variants in either NPC1 or NPC2, an accumulation of unesterified cholesterol and sphingolipids occurs in the late endosomal-lysosomal system and is a biochemical hallmark of the disease ^[3]. Further, this lipid accumulation results in a number of downstream events and clinical phenotypes including hepatosplenomegaly, vertical supranuclear gaze palsy, cognitive decline and progressive ataxia due to cerebellar degeneration, among others ^[3,4]. To date, there is no FDA approved therapy for NPC, although a glycosylceramide synthase inhibitor, miglustat (Zavesca^®), is an approved therapy in Europe, Japan, and other countries to ameliorate neurological manifestations ^[5,6]. In the United States, miglustat is available off-label ^[7]. To address the gap in therapeutic opportunities several other potential therapeutic drugs have been studied, including, 2-hydroxypropyl-beta-cyclodextrin ^[8–10]; arimoclomol ^[11], histone deacetylase inhibitors ^[12] and N-acetyl-L-leucine (NALL), an acetylated derivative of the essential amino acid leucine ^[13].

Over the past decade, proteomic studies in NPC1 have been carried out utilizing either animal models or biological samples obtained from patients. Many biomarkers have been reported and shed light on the pathophysiology and downstream alterations that occur in NPC1. For example, Sleat et al. purified lysosomal proteins containing mannose 6-phosphate and found prosaposin, which is the precursor of saposins A, B, C, and D that have functions in the degradation of glycosphingolipids, and β-hexosaminidase that is important in degradation of ganglioside GM2 were both increased in the NPC1 mouse brains ^[14]. Recently, our group performed mass spectrometry-based proteomics studies on the liver, spleen, cerebral cortex and cerebellum of late stage Npc1^−/− mice and found alterations across these different tissues including, RAC-alpha serine/threonine-protein kinase (AKT), which stimulates cell growth and proliferation ^[15], was downregulated in both spleen and liver ^[16]; lysosome membrane protein 2 (LIMP2), lysosome-associated membrane glycoprotein 1 (LAMP1) and Ras-related protein Rab-7a (RAB7A) were all upregulated in the liver, cerebellum and cerebral cortex ^[17]. Glutathione metabolism and associations with oxidative stress were also found to be altered in the Npc1^−/− cerebellar tissue compared to control mice ^[18]. Using fibroblasts from individuals with NPC1, procathepsin D, the precursor of the lysosomal aspartyl protease of cathepsin D (CTSD) ^[19], was found to be increased-which coincides with reports of reduced CTSD activity ^[20]. In NPC1 fibroblasts containing the missense variant, I1061T, (the most common pathological variant in NPC1 ^[21]), superoxide dismutase [Mn], mitochondrial (SOD2) was found to be downregulated compared to control cells ^[22]. Furthermore, some protein targets have been investigated using ELISA to measure NPC1 CSF levels of superoxide dismutase [Cu-Zn] (SOD1), an oxidative stress protein, which was increased relative to healthy controls ^[18]. In addition, the concentration of fatty acid binding protein 3 (FABP3) was also higher in individuals with NPC1 CSF samples compared to controls ^[18]. However, to date, there is no comprehensive proteomics study in CSF in individuals with NPC1.

In the current study, we used isobaric tagging combined with mass spectrometry to evaluate the altered proteome in CSF from individuals with NPC1. Our results show that many innate immune complement components and fibrinogen are downregulated in individuals with NPC1. On the contrary, neuromodulin (GAP43), antithrombin III (AT III, SERPINC1), CTSD and NPY are all upregulated. Additionally, the NPC1 mouse model was used to validate our findings of altered levels of NPY in NPC1. Further, the altered CSF proteome of NPC1 individuals treated with miglustat was compared where 10 potential biomarkers for monitoring treatment were identified.

Materials and methods

Reagents and chemicals

All reagents and chemicals were obtained and used without additional purification unless otherwise noted. Solvents were purchased from Sigma-Aldrich (MO, USA) or VWR and of LC-MS grade when available.

CSF samples collection and preparation

CSF samples were collected via lumbar puncture in accordance with NICHD IRB approved protocols. Written informed consent was obtained from the participant, parent, or legal guardian. Assent was obtained when appropriate. Consent was witnessed and became part of the participant’s medical record at the NIH Clinical Center. All NPC1 individuals have undergone genetic sequence confirmation of NPC1 disease. Of the 22 individuals with NPC1, 10 participants were currently taking off-label miglustat at the time of CSF collection. Control CSF (N = 12) was obtained from unaffected adults (Biochemed Services, Winchester, VA). Following collection, aliquots of CSF were immediately frozen at −80 °C until use. A pooled reference sample was generated using equal CSF volumes. The pooled reference was used to normalize protein abundance across multiple iTRAQ kits given the limited labeling of 8 samples per kit. The age ranges and sex for NPC1 and adult control specimens are summarized in Table S1.

CSF samples were processed individually for NPC1 individuals, controls, and the pooled reference. To reveal more of the low-abundance proteins in CSF, depletion of human serum albumin is commonly performed ^[23] and was included in this study. In each case, CSF was diluted with Agilent concentrated loading buffer (4X) by mixing 125 μL of buffer with 375 μL of CSF and centrifuged in a 0.2 μm filter column for 1 minute at 14,000 rpm. The resulting sample was then injected onto a human serum albumin immunoaffinity column (Agilent Technologies, Santa Clara, CA) using the manufacturer’s guidelines. The unbound, non-albumin containing fraction was collected and used for downstream quantitative proteomic analysis. Depletion was monitored via UV-Visible absorption at 280 nm. The albumin-depleted samples were then concentrated and desalted using 5 kDa molecular weight cut-off filters and re-suspended in an iTRAQ appropriate buffer (100 mM triethylammonium bicarbonate, 10% acetonitrile). Protein concentration was determined using the Bradford method and 30 μg of total protein from each sample was used for proteomic analysis.

Proteolytic digestion, iTRAQ labeling, and fractionation

The concentrated, desalted, albumin-depleted proteome (30 μg) for each patient was diluted to a total volume of 20 μL (TEAB/ACN buffer) then reduced, alkylated and trypsin digested as suggested by the manufacturer. In addition, recombinant green fluorescent protein (GFP, 50 fmol per 1 μg total protein), used as a digestion standard, was added to each sample prior to digestion and pre-digested alcohol dehydrogenase (50 fmol per 1 μg total protein; Waters Corporation, Milford, MA) as a post-digestion internal standard was added prior to iTRAQ labeling. Each sample was labeled with an appropriate iTRAQ label according to the manufacturer’s protocol and the labels were randomly assigned for each cohort/experiment to ensure no experimental bias introduction for a particular tag or participants group. For each cohort experiment, four individuals with NPC1, three controls and the pooled reference sample were analyzed. Following the labeling reactions, each tube was quenched by the addition of 50 μL of water. Digestion verification and labeling was confirmed by removing 0.5 μL from each sample at the completion of the appropriate step and analyzed via MALDI-MS and MS/MS. The combined iTRAQ labeled samples were fractionated into ten fractions using strong cation exchange (Sciex/Life Technologies ICAT SCX Kit) then desalted with Omix C18 tips (Varian/Agilent, Santa Clara, CA). The desalted samples were vacuum concentrated to remove organic solvents and re-suspended in 40 μL of 0.1% TFA prior to LC-MS/MS analysis.

Liquid chromatography-mass spectrometry analysis (LC-MS)

Mass spectrometry analysis of each of the SCX fractions were analyzed using an Agilent 6550 iFunnel Q-TOF (Agilent Technologies, Santa Clara, CA) which was interfaced to an Agilent 1260 HPLC Chipcube. Solvent A was 0.1% formic acid and solvent B was acetonitrile containing 0.1% formic acid. Peptides were loaded onto a High Performance Chip (G4240-62030, Agilent Technologies Inc., Santa Clara, CA, USA) in 0.1% formic acid and trapped on the 360 nL enrichment column using a sample flush volume of 5 μL. Peptide separation was performed using the following gradient: 3% B at 0 min, 3-25% B at 105 min, 40% B at 120 min, 90% B at 125 min, hold at 90% B until 130 min. Source parameters were: gas temp (225°C), drying gas (13 L/minute), VCap (1800V) and Oct 1 RF Vpp (750V). Acquisition parameters for mass spectrometry analysis were as follows: 8 MS scans/second (m/z 300-1700), 3 MS/MS scans/second (m/z 50-1700) and maximum precursor selection of the top 20 ions of charge states ≥+2 with greater than 5000 counts triggered MS/MS at a normalized collision energy determined for each peptide using slope of 4 and offset of 2 for +2 charge state species and slope of 4.6 and offset of −4.8 for species with charge state ≥ +3. Low energy CID was used to fragment the selected precursors. Noting, only biological replicates of different cohorts were performed in this study.

Raw data (.d file format) was converted to a .mgf file for each fraction using the Mass Hunter Qualitative Browser (v7.1) and then merged using Proteowizard to generate a single .mgf file for each cohort. Peak areas were extracted, and in-spectrum normalization performed as previously described ^[24] with a requirement of reporter ion intensity sum >1600, using the R software program. The resulting .mgf file was submitted for protein identification in Mascot by searching against the Homosapien Swiss Protein database as well as in inhouse library containing the sequences for GFP (Aequorea victoria) and alcohol dehydrogenase (Saccharomyces cerevisiae) with fixed modifications iTRAQ 8-plex (K and N-term) and variable modifications oxidation (M) and MMTS (C). Mass error tolerances were 10 ppm for MS and 0.1 Da for MS/MS spectra. Protein identification results were exported as .DAT files and imported in Scaffold v4.8.4 (Proteome Software, Portland, OR). Protein identifications were considered at the 95% confidence interval with a minimum of 2 unique peptide assignments, each having a confidence interval minimum of 99%. Data from all cohorts were normalized across experiments via the pooled reference sample in Scaffold. Relative quantification was performed across all groups (NPC1-miglustat vs control, NPC1+miglustat vs NPC1-miglustat and NPC1+miglustat vs control) using an unpaired T-test where the p-value was adjusted using the Benjamini-Hochberg procedure to decrease the false discovery rate (FDR) with the cutoff set to 0.05. Further, evaluation of the fold change (FC) for the internal standards across all cohorts and biological replicates was evaluated and it was determined that no additional normalization was needed.

Cerebellar tissue lysates and Western blot analysis

NPC1 mice were obtained from Jackson laboratories (Npc1tm(I1061T)^Dso_; strain: 027704) ^[25] and heterozygous breeding pairs were maintained using institutional IACUC-approved protocols (UIC #21-118). Genotype was determined by polymerase chain reaction. Cerebellar tissue lysates from individual animals were prepared by homogenizing and lysing in RIPA buffer (1% NP-40, 150 mM sodium chloride, 0.5% sodium deoxycholate, 0.1% SDS, 50 mM Tris pH 8.0). Lysate concentrations were determined by the Pierce BCA assay using the manufacturer’s protocol. Lysates (30 μg) were subjected to separation by electrophoresis and Western blot for evaluation of NPY expression. The samples were thermally denatured in Laemmli buffer containing 4% SDS at 95 °C for 10 min then loaded onto a 4-12% SDS-PAGE gel (Novex Life Technologies, CA, USA). Separated proteins on the gel were transferred to a PVDF membrane using a semi-wet transfer method at 20 V for 60 min. Blocking was carried out in 5% (wt/v) BSA with 0.1% (v/v) Tween-20 at room temperature for 1 hour. Rabbit anti-NPY (1:1000 dilution, Cell Signaling, D7Y5A) was added to the blocking buffer and incubated at 4 °C overnight, followed by secondary antibody (1:2000 dilution, R & D Systems, HAF008) incubated at room temperature for 1 hour. The antibody against NPY recognizes both mature NPY (4 kDa) and precursor (11 kDa). Protein bands were visualized with SuperSignal West Pico PLUS Chemiluminescent Substrate (ThermoScientific, 34580) and imaged on an Azure imaging system (Azure Biosystems, CA, USA). Densitometry analyses were performed using ImageJ (https://imagej.nih.gov/ij/) ^[26]. GraphPad Prism (v9.3.1) was used to determine statistical significance between groups using an unpaired T-test with two-stage set-up method of Benjamini, Krieger and Yekutieli; the significance was considered when p-value < 0.05.

Results

Protein identification and quantification in NPC1 CSF samples

To identify protein biomarkers in the individuals with NPC1, CSF samples obtained from participants and healthy controls were depleted of human serum albumin (HSA), digested with trypsin, and labeled with iTRAQ8plex reagents followed by fractionation and LC-MS analysis. The experimental workflow is shown in Figure 1. A total of 300 proteins were identified in this study (Table S2), where 71 proteins were observed to be altered (FDR < 0.05, unpaired T-test (p < 0.01366) with Benjamini-Hochberg correction) when comparing between individuals with NPC1 and healthy controls (Figure 2, Table S3). Included in the altered proteome, were proteins associated with the complement system, which is part of innate immune defense ^[27], including complement factor I (CFI), complement factor B (CFB), complements C2 (C2), C3, C9, complement factor H (CFH), complement C1r subcomponent (C1R), and complement C1q subcomponent subunit C (C1QC)-all of which were observed to be downregulated indicating reduced innate immune response in the individuals with NPC1. Three polypeptide chains fibrinogen (essential to blood clot formation) ^[28], fibrinogen alpha, beta, and gamma chain (FGA, FGB, FGG) were also downregulated in individuals with NPC1 compared with controls. On the other hand, GAP43 (important in axon growth) ^[29], AT III (inhibits thrombin) ^[30], CTSD (lysosomal proteolytic enzyme) ^[31], and NPY (neurotrophic factor and an inhibitor of excitotoxicity and neuroinflammation in the brain) ^[32], were all increased in the CSF obtained from individuals with NPC1.

Figure 1. Workflow for the analysis of human cerebrospinal fluid.

Cerebrospinal fluid (CSF) samples were depleted of human serum albumin (HSA) to maximize the measured proteome coverage. The depleted CSF was reduced, alkylated and trypsin digested followed by iTRAQ8plex labeling. The labeled samples were combined and fractionated by strong cation exchange and analyzed by Q-TOF LC-MS and database searching. Schematic was created with BioRender.com.

Figure 2. Volcano plot of the measured cerebrospinal fluid proteome.

Volcano plot depicting the distribution of altered proteins (unpaired T-test (p < 0.01366) with Benjamini-Hochberg correction, circled dots above the grey line) in untreated individuals with NPC1. Complement proteins that were downregulated are annotated in blue, fibrinogen subunits annotated in green and other upregulated proteins of interest (GAP43, AT III, CTSD, and NPY) are annotated in red.

To further investigate the altered 71 proteins in individuals with NPC1 CSF without miglustat treatment, pathway analysis was performed using Ingenuity Pathway Analysis software (IPA, Qiagen, Redwood City, CA) to elucidate biological relevance. Figure 3 shows the major processes affiliated with altered CSF proteins in NPC1 including the complement system, glycolysis I, acute phase response signaling (C1R, complement C5, CFB, alpha-1-antichymotrypsin (SERPINA3), transthyretin (TTR)), FXR/RXR activation, sucrose degradation V, LXR/RXR activation, atherosclerosis signaling, maturity onset diabetes of young signaling, gluconeogenesis I, neuroprotective role of THOP1 in Alzheimer disease, among others. The complement system is the most significant altered IPA pathway in accordance with the downregulated complement proteins shown in Figure 2.

Figure 3. Pathway analysis of the altered cerebrospinal fluid proteome.

Top ranked (top 15) conical pathways of the differential proteins in the cerebrospinal fluid of individuals with NPC1. Significance was determined using a Right-tailed Fisher’s exact test to determine probability of pathways enriched when searched against the Ingenuity Pathway Knowledge Base.

Next, we sought to individually investigate altered proteins identified in our discovery study, such as NPY. It is well-know that the NPY precursor (11 kDa) can be cleaved into three peptides, an N-terminal signaling peptide, mature NPY (4 kDa), and a C-flanking peptide of NPY (CPON) that can be co-released with mature NPY from dense core vesicles ^[33]. In the present study, two unique peptides of the CPON region were detected by MS; however, it is not definitive that if these peptides measured by MS were from NPY precursor after trypsin digestion or the cleaved form of the CPON. Since mature NPY plays critical roles in the brain, including neurogenesis and neuroprotection, and has to been reported to be altered in many neurodegenerative diseases ^[32]; we were particularly interested in investigating more about NPY. Moreover, the alteration of NPY has not been reported in NPC1 previously. Thus, to investigate if this change also occurs in mouse models, of NPC1, we used the Npc1^{I1061T/ I1061T} mouse model, which has the most common pathological variant in NPC1 disease ^[25] and corresponds to the classic juvenile (onset at age 5-16) phenotype ^[34] in humans. In the Npc1^{I1061T/ I1061T} mouse model, the onset of a visible resting tremor starts at approximately 8 weeks of age ^[25], with an 8-week-old mouse considered as juvenile period (before adulthood) ^[35]. The cerebellum is the most affected brain area in NPC1 ^[1], thus cerebellar tissues were collected from 4-, 9- and 15-week-old mice and used to perform Western blot analysis. These time points correspond to a pre-symptomatic, symptomatic, and near-terminal time point in this mouse model. With this experiment, we aimed to evaluate the expression of NPY across disease progression. Western blot analysis showed an increase in expression of mature NPY (4 kDa) in 15-week Npc1^{I1061T/ I1061T} mice, but not in 4- or 9-week animals, compared to age-matched controls (Figure 4). This finding suggests that both mature NPY (measured by Western blot) and CPON (detected by MS) were increased in NPC1 and may be used as a marker for monitoring the late stages of the disease progression and can be a potential target for future therapeutic development.

Figure 4. Western blot analysis of NPY expression in a mouse model of NPC1.

A) Protein lysates from the cerebellum of 4-, 9-, and 15-week control and Npc1^{I1061T/ I1061T} mice (N = 4 per genotype) were subject to electrophoresis and Western blot analysis of Pro-Neuropeptide Y (NPY). B) Data analysis revealed increased expression of NPY only at 15-weeks, a later timepoint of disease. Data is reported as relative to α-tubulin. Statistical significance was determined using an unpaired T-test with two-stage set-up method of Benjamini, Krieger and Yekutieli within GraphPad Prism v9.3.1. * p-value < 0.05

Miglustat treatment reverses the levels of some proteins in individuals with NPC1

Approximately half of individuals with NPC1 seen in the clinic are treated with miglustat, therefore a subset of participant specimens included in this study were collected during miglustat treatment. Here, we evaluated the quantitative proteome differences in the CSF between NPC1 individuals treated with miglustat versus those that are untreated and found 52 proteins (FDR < 0.05, unpaired T-test (p < 0.00891) with Benjamini-Hochberg correction) were different between the groups (Figure 5A, Table S4). However, when comparing NPC1 individuals treated with miglustat versus healthy controls, 96 altered proteins were observed (FDR < 0.05, unpaired T-test (p < 0.01516) with Benjamini-Hochberg correction) (Figure 5B, Table S5). Interestingly, in NPC1 individuals treated with miglustat, the complement protein, C1R (p < 0.0001, Log₂FC = 0.4) was increased compared to untreated NPC1 individuals (Figure 5A), and the expression level was comparable (p = 0.28, Log₂FC = 0.1) with healthy controls (Figure 5B). Similarly, CFH was increased in NPC1 individuals treated with miglustat when compared to either NPC1 individuals without treatment (p < 0.0001, Log₂FC = 0.46) (Figure 5A) or healthy controls (p = 0.001, Log₂FC = 0.27) (Figure 5B). These results indicate miglustat treatment normalizes these complement proteins in NPC1 and these may be biomarkers for monitoring therapeutic treatments. Two subunits of the fibrinogen, FGA (p = 0.00021, Log₂FC = 0.28) and FGB (p < 0.0001, Log₂FC = 0.35), were increased in NPC1 individuals treated with miglustat compared to individuals without treatment (Figure 5A); however, they were decreased (FGA: p = 0.0022nLog₂FC = −0.49; FGB: p = 0.0024, Log₂FC = −0.48) when comparing to healthy controls (Figure 5B), indicating that miglustat increases the CSF levels of FGA and FGB to some point but not enough to normalize them to control levels. Additionally, the levels of C3 (p = 0.0026, Log₂FC = −0.08) was even lower in NPC1 individuals treated with miglustat (Figure 5A), noting that C3 (p < 0.0001, Log₂FC = −0.2) levels were already reduced in untreated NPC1 individuals compared to controls (Figure 2). This result indicates that C3 will not be a suitable marker for monitoring therapeutic treatments.

Figure 5. Volcano plot showing the differential NPC1 cerebrospinal fluid proteome.

A) Comparison of individuals with NPC1 therapeutically treated with and without miglustat, an inhibitor of glycosylceramide synthesis. B) Comparison of individuals with NPC1 treated with miglustat relative to healthy controls. Complement proteins are annotated in blue, fibrinogen subunits annotated in green and other proteins of interest (GAP43, AT III, CTSD, and NPY) are annotated in red. C) The changing trend of NPY in comparison of untreated individuals (NPC1-mig) and controls, NPC1 individuals treated with miglustat (NPC1+mig) and untreated individuals (NPC1-mig), and NPC1 individuals treated with miglustat compared to controls (NPC1+mig and controls). The p-value corresponding to each comparison is indicated above or below the bar graph. Significant p-value is denoted in red text.

Both AT III (p = 0.11, Log₂FC = −0.09) and NPY (p = 0.16, Log₂ FC = −0.3) were slightly decreased in NPC1 individuals treated with miglustat, although not statistically significant (Figure 5A, Table S4). However, the changes in expression of these two proteins (AT III, p = 0.23, Log₂FC = 0.08; NPY, p = 0.068, Log₂FC = 0.39) were also no longer different when comparing NPC1 individuals treated with miglustat versus controls (Figure 5B, Table S5). The changing trend of NPY among all three comparison groups is shown in Figure 5C. These findings suggest that miglustat can reduce AT III and NPY expression at some level. On the other hand, CTSD (p = 0.038, Log₂FC = 0.15) and GAP43 (p = 0.96, Log₂FC = 0) still had comparable levels between NPC1 individuals treated with miglustat versus untreated NPC1 individuals (Figure 5A, Table S4), which is consistent when evaluating significant changes between NPC1 individuals treated with miglustat to healthy controls (CTSD, p < 0.0001, Log₂FC = 0.44; GAP43 p < 0.0001, Log₂FC = 0.86) (Figure 5B, Table S5).

Potential biomarkers for monitoring miglustat therapeutic efficiency

Miglustat, an inhibitor of glucosylceramide synthase ^[36], which is approved in the US for another lysosomal storage disorder, Gaucher disease, is available off-label and has been investigated in several clinical trials ^[37–39]. Treatment of miglustat improved or stabled disease in 70.5% of NPC1 participants in a recent report of clinical study, though with some confounding issues including chronic diarrhea and seizures ^[40]. While miglustat has been investigated for several years, the comprehensive differential proteome of CSF has not been evaluated rather, select proteins have been pursued in NPC1 individuals treated with miglustat. Using our dataset to identify potential biomarkers of therapeutic effectiveness of miglustat, we overlapped the list of altered proteins between untreated NPC1 individuals and controls, and the list of altered proteins between NPC1 individuals treated with miglustat and untreated individuals (Figure 6A, Table 1). Seventeen proteins were in common between these two comparison groups as listed in Table 1, implying potential candidates for monitoring therapeutic effectiveness. To test the specificity of whether these common proteins can characterize differences among three cohorts, we performed an unbiased clustering based on t-distributed stochastic neighbor embedding (t-SNE) analysis as shown in Figure 6B. The t-SNE analysis shows that these 17 proteins are able to distinctly cluster into three groups within the current study, indicating these proteins have the potential as candidates for monitoring therapeutic effectiveness with further investigation. To further narrow down this list, we include 10 out of the 17 proteins (bolded in Table 1) that could be considered potential biomarkers for monitoring therapeutic treatment due to their opposite direction of movement in these two comparison groups. For example, FGA, FGB, CFH, C1R, haptoglobin (HP), osteopontin, (SPP1), IgGFc-binding protein (FCGBP), galectin-3-binding protein (LGALS3BP), apolipoprotein L1 (APOL1), and prolow-density lipoprotein receptor-related protein 1 (LRP1) were downregulated (FDR < 0.05, unpaired T-test (p < 0.01366) with Benjamini-Hochberg correction) in individuals with NPC1 CSF compared to controls, however, they were upregulated (FDR < 0.05, unpaired T-test (p < 0.00891) with Benjamini-Hochberg correction) in individuals with NPC1 treated with miglustat compared to untreated NPC1 individuals. Further studies are needed to investigate these proteins in larger cohorts and to determine if these changes are specific to miglustat treatment or could be used to monitor other therapeutic treatments of NPC1.

Figure 6. Altered cerebrospinal fluid proteins observed between comparisons of untreated NPC1 individuals to healthy controls and to NPC1 individuals with miglustat treatment.

A) Venn diagram analysis showing the overlap of altered proteins observed in untreated NPC1 individuals (relative to healthy controls) and individuals with NPC1 treated with miglustat (relative to untreated NPC1 individuals). A total of 17 proteins are common between both comparisons, regardless of treatment status. B) The t-distributed stochastic neighbor embedding (t-SNE) plot using these 17 proteins indicating the specificity of these proteins as potential biomarkers for monitoring therapeutic effectiveness. Courtesy: Oliveros, J.C. (2007-2015) Venny. An interactive tool for comparing lists with Venn's diagrams. Publicly available at https://bioinfogpcnbcsices/tools/venny/index.html.

Table 1.

Overlap of differential proteins.

Protein name	Protein Abbreviation	Untreated NPC1 vs. Control		Miglustat Treated vs. Untreated NPC1
		Log2FC	p-value	Log2FC	p-value
Fibrinogen alpha chain	FGA	−0.78	<0.0001	0.28	0.00021
Fibrinogen beta chain	FGB	−0.83	<0.0001	0.35	<0.0001
Cluster of haptoglobin	HP	−1.16	<0.0001	0.56	0.00011
Complement C3	C3	−0.2	<0.0001	−0.08	0.0026
Osteopontin	SPP1	−0.49	<0.0001	0.22	0.0018
IgGFc-binding protein	FCGBP	−0.57	<0.0001	0.46	0.00034
Secretogranin-1	CHGB	0.23	<0.0001	0.23	<0.0001
Complement factor H	CFH	−0.2	<0.0001	0.46	<0.0001
Galectin-3-binding protein	LGALS3BP	−0.22	<0.0001	0.37	<0.0001
Chitinase-3-like protein 1	CHI3L1	−0.34	0.00013	−0.24	0.00063
Complement C1r subcomponent	C1R	−0.29	0.00028	0.4	<0.0001
Ectonucleotide pyrophosphatase/phosphodiesterase family member 2	ENPP2	0.19	0.00062	0.18	0.0024
Apolipoprotein L1	APOL1	−0.64	0.0029	0.62	0.0017
Keratin, type I cytoskeletal 9	KRT9	−0.17	0.003	−0.29	<0.0001
Keratin, type II cytoskeletal 1	KRT1	−0.14	0.0034	−0.17	<0.0001
Prolow-density lipoprotein receptor-related protein 1	LRP1	−0.53	0.0084	0.62	0.0007
Secretogranin-3	SCG3	0.19	0.0085	0.22	0.0024

Discussion

The present study is a comprehensive investigation comparing the CSF proteomes between 1) untreated individuals with NPC1 and healthy controls, 2) individuals with NPC1 treated with miglustat and untreated NPC1 individuals and, 3) individuals with NPC1 treated with miglustat versus healthy controls. We observed 71, 52, and 96 (FDR < 0.05) altered proteins, respectively, for each of the comparisons. Among these altered proteins, the increased level of NPY was further validated by Western blot using a NPC1 mouse model.

Differential proteins that are reported in the present study are involved in many pathways, including the complement system. The complement system serves as a first line of defense against altered host cells and foreign intruders ^[27] and is known to be dysregulated in NPC1 disease ^[41–44]. In the brain, the complement system plays important roles in synaptic pruning and neuroinflammation, and most of complement components increase with age and further increase in some neurodegenerative diseases, such as Alzheimer disease ^[45]. Considering individuals with NPC1 are prone to be young children and CSF samples of healthy controls used in present study are from adults, lower expression level of many complement components in NPC1 participants are expected, however, it is unclear whether this difference of expression level between young children and adults is significant. Prior work has reported that complement C3 transcripts are increased in the 7-week-old mice and postmortem individuals with NPC1 brain tissues ^[41]. Similar observations were reported by Lopez et al. that the mRNA levels of C3, C1qa, b, and c are elevated in the cerebellum of 3- and 7-week Npc1^−/− mice ^[42]. Interestingly, a recent study stained liver tissue and showed that C3 is significantly higher at 3- and 8-week-old Npc1^−/− mice, but decreased to a comparable level as control mice at 10-weeks of age ^[44]. However, these results contradict our current study that C3, C1QC, and other complement components were decreased in the CSF of individuals with NPC1. This discrepancy may be due to previous studies using mouse models as a proxy for human disease and tissue versus CSF as the sample source. Secondly, the expression level of C3 decreases with age ^[44,46], and the previous studies used younger mice with higher levels of C3. Thirdly, the previous study was comparing mRNA levels of C3, while the current study measured protein levels. It is well known that under many circumstances the amount of mRNA is not always correlated with the amount of the protein due to post-transcriptional, translational and degradation regulation ^[47]. In addition, tissue specificity may result in different expression level of the same protein ^[48], since the previous study used brain and liver tissues, but the current study used CSF that was obtained from individuals with NPC1 which may represent extracellular level that are not captured in the same way. Finally, it is well known that complement components are secreted proteins ^[27], and with disease progression there may be less secretion into the CSF while they accumulate in some of the brain areas. This phenomenon can additionally explain the increased C3 in postmortem individuals with NPC1 brains reported by Cologna, et al ^[41]. Interestingly, complement components, namely C1Q and C3, are not involved in neuronal death in a NPC1 mouse model, supporting that neuroinflammation, a well-known feature ^[41,49], is a secondary response of neurodegeneration in NPC1 but not the cause of neurodegeneration ^[42]. This could possibly explain why the complement members are not increased in NPC1 disease the same way as in other neurodegenerative diseases ^[45,50].

Alteration of fibrinogen to our knowledge has not been reported in NPC1 disease. Fibrinogen plays an important role in blood clotting, which is initiated by the cleavage of thrombin, releasing fibrinopeptide A and B ^[28]. In the current study, all three polypeptide chains of fibrinogen (i.e., FGA, FGB, FGG) were detected and found to be decreased in the CSF of individuals with NPC1 compared to controls (Figure 2, Table S3). Consistently, the defective platelet function in the blood of Npc1^−/− mice with impaired thrombin-stimulated platelet aggregation and prolonged bleeding time was reported in a recent study ^[51]. Although there is no direct correlation of fibrinogen levels between blood and CSF ^[52], downregulated fibrinogen in CSF indicates that fibrinogen may also be reduced in the blood since impaired blood clotting has been reported in Npc1^−/− mouse model ^[51]. Notably, AT III, a thrombin inhibitor that inhibits coagulation of the blood ^[30], was upregulated (p < 0.00052, Log₂FC = 0.17) in individuals with NPC1 in the current study (Figure 2, Table S3). Together, the current data is consistent with the notion of impaired blood clotting that was observed in the Npc1^−/− mouse ^[51].

Other upregulated proteins in individuals with NPC1 reported in this study including GAP43, which is critical to initiate axon growth in the CNS ^[29], and CTSD (Figure 2, Table S3), an aspartic protease that degrades proteins in the lysosome ^[31], which was reported in an early study to be significantly increased in individuals with NPC1 serum samples ^[53]. The upregulated GAP43 and CTSD indicate the complimentary regulation that counterparts the neurodegeneration in NPC1. However, the levels of GAP43 (p = 0.96, Log₂FC = 0) and CTSD (p = 0.038, Log₂FC = 0.15) were comparable between individuals with NPC1 treated with miglustat and untreated NPC1 individuals (Figure 5A, Table S4), indicating miglustat does not affect the CSF levels of either GAP43 or CTSD and an additional therapeutic treatment may be needed to further reverse these altered proteins in individuals with NPC1.

Calbindin D (CALB1) is a classical Purkinje cell marker and has been reported in a previous study as a disease progression marker, being significantly elevated in the 3-week-old feline NPC1 model and continues to increase to the end-stage of disease ^[54]. However, biweekly administration of 2-hydroxypropyl-β-cyclodextrin (HPβCD) intrathecally normalized CALB1 concentration to control levels ^[54]. Similarly, an increase level of CALB1 in CSF from individuals with NPC1 compared to healthy controls were noted ^[54]. Furthermore, individuals with NPC1 treated with off-label miglustat showed a reduction in CALB1 expression compared to pretreatment ^[54]. In the current study, CALB1 was slightly increased in the CSF (p < 0.04, Log₂FC = 0.33) in individuals with NPC1 (Table S3), however, it is not below the cut-off of FDR < 0.05 (adjusted p-value < 0.01366). This observation may be due to small sample sizes or a difference between the detection methods used in the two studies. Moreover, reporter ion underestimation may explain this observation ^[55]. Another challenge associated with this finding is that CALB1 expression levels reduce with age ^[56], the expression of CALB1 can be very different if individuals with NPC1 selected were from one certain age group versus a wide range of ages which had to be used for the control group.

From our work, we propose NPY as a disease marker of NPC1. It is worth noting that the CSF represents the average protein level from the whole brain and not specifically reflects a certain brain area, for instance, an early study reported that the number of NPY-positive cells in the cerebral cortex was not altered in the Npc1^−/− mice compared to control mice, although the expression level of NPY was not assessed ^[57]. However, in the present study, NPY was observed to be increased in the cerebellar tissue of the Npc1^{I1061T/ I1061T} mice (Figure 4). NPY, a 36 amino acid peptide, has various roles in the brain including trophic support, stimulating autophagy, decreasing excitotoxicity, and attenuating neuroinflammation ^[32,58]. In addition, NPY receptors belong to the class A or rhodopsin-like G-protein coupled receptors (GPCR) and regulate calcium signaling ^[59]. Many studies have reported NPY levels are altered in various neurodegenerative diseases, including Alzheimer, Parkinson, Huntington, and Machado-Joseph diseases ^[32]. Similarly, NPY was slightly decreased (p = 0.16, Log₂FC = −0.3) in individuals with NPC1 treated with miglustat compared to untreated NPC1 individuals (Figure 5, Table S4). This supports the hypothesis that upregulated NPY in individuals with NPC1 may be a self-regulatory mechanism to combat increased neuroinflammation. Neuroinflammation markers in CSF from individuals with NPC1 were previously reported, including interleukin (IL) 3, IL -10, and -13 ^[41], which were slightly decreased in individuals with NPC1 who were treated with miglustat, similar to the observation for NPY in the current study. These data suggest that miglustat treatment in individuals with NPC1 reduces the neuroinflammation at some level. Alternatively, noting that individuals with NPC1 have seizures ^[60], which is due to neuronal hyperexcitability, upregulation of NPY in individuals with NPC1 could potentially be a complementary regulatory mechanism to decrease neuronal hyperexcitability ^[61]. It is reasonable to consider the role of NPY in the nitric oxide (NO) pathway which has been noted to be altered in the current study (Figure 3). Nitric oxide is the mediator of NPY in cell proliferation in the brain and APOE induces NO production ^[62–65]. Since APOE was increased (p = 0.00037, Log₂FC = 0.25) in CSF of individuals with NPC1 (Table S3) in this study, correlating with upregulated NPY in NPC1 disease, NPY could be an interesting target for further investigation.

We report 17 differential proteins that overlap between NPC1 individuals compared to controls and between individuals treated with miglustat and untreated NPC1 individuals, we focused on the 10 proteins that moved in different directions in Table 1 that can be potential biomarkers for monitoring therapeutic treatment of miglustat. Beyond the complement system, fibrinogen, decreased in NPC1 and has been reported to be increased in other neurodegenerative diseases, including Alzheimer ^[66] and multiple sclerosis ^[67] involving abnormal fibrin deposition in the brain and further triggering neuroinflammatory response ^[68]. Another potential biomarker for monitoring therapeutic treatment is osteopontin (SPP1), which has been reported to be upregulated in response of neuroinflammation and injury in multiple sclerosis and Parkinson disease ^[69]. Besides, SPP1 overexpression promotes multiple cancer cell proliferation, angiogenesis, and metastasis in lung, liver and breast cancer ^[70]. Although keratins 1 and 9 listed in Table 1 may indicate potential contamination during sample collection or preparation, these two proteins were reported to be present in the CSF relating to other diseases ^[71,72]. Additionally, keratins 1 and 9 were not considered as potential therapeutic biomarkers in the present study due to altered levels being consistent in direction. NPY is the most intriguing protein in this study due to many of its important roles in the brain as discussed earlier. Furthermore, we suggest that the other differential proteins should be further investigated in either human or animal models with different techniques to broaden the understanding of the pathophysiology of NPC1. It would also be essential to validate in a larger cohort of these 10 biomarkers that may respond to therapeutic intervention and determine if they correlate with clinical aspects of NPC1.

Significance of the study:

Herein we have performed a discovery-based differential proteomic analysis of CSF from individuals with NPC1. This study revealed altered proteins that had not been previously shown, fibrinogen and NPY and we discuss their potential use as biomarkers. Additionally, proteins involved in the complement cascade and lysosomal enzymes were found to be secreted into the CSF, keeping in accordance with prior studies. Further, it was observed that NPC1 individuals therapeutically treated with miglustat displayed differing secretory proteomes compared to those who have not undergone treatment. Thus, our data provides unique information about NPC1 disease, changes associated with miglustat therapy and thus, provides candidate biomarkers to consider for monitoring in future therapeutic trials.

Abbreviations:

APOL1

Apolipoprotein L1

AT III

Antithrombin III

C1QC

Complement C1q subcomponent subunit C

C1R

Complement C1r subcomponent

C2

Complement 2

C3

Complement 3

C9

Complement 9

CALB1

Calbindin D

CFB

Complement factor B

CFH

Complement factor H

CFI

Complement factor I

CSF

Cerebrospinal fluid

CTSD

Cathepsin D

FABP3

Fatty acid binding protein 3

FABP7

Fatty acid binding protein 7

FCGBP

IgGFc-binding protein

FGA

Fibrinogen alpha chain

FGB

Fibrinogen beta chain

FGG

Fibrinogen gamma chain

GAP43

Neuromodulin

HAS

Human albumin serum

HP

Haptoglobin

LGALS3BP

Galectin-3-binding protein

LRP1

Prolow-density lipoprotein receptor-related protein 1

NO

Nitric oxide

NPC1

Niemann-Pick, type C1

NPY

Pro-neuropeptide Y

SPP1

Osteopontin