Plasma Signature of Neurological Disease in the Monogenetic Disorder Niemann-Pick Type C*

Background: Neurodegenerative diseases are associated with inflammation in the brain and other organs. Results: In murine Niemann-Pick Type C, cathepsin S was elevated and a plasma marker of liver inflammation. Lysozyme was an inflammatory plasma marker derived from both liver and brain. Conclusion: Their dual analysis suggested four distinct severity states of neurodegeneration. Significance: Cerebral inflammatory disease may be detectable in plasma. Early diagnosis of neurological disorders would greatly improve their management and treatment. A major hurdle is that inflammatory products of cerebral disease are not easily detected in blood. Inflammation in multiple organs and heterogeneity in disease present additional challenges in distinguishing the extent to which a blood-based marker reflects disease in brain or other afflicted organs. Murine models of the monogenetic disorder Niemann-Pick Type C present aggressive forms of cerebral and liver inflammatory disease. Microarray analyses previously revealed age-dependent changes in innate immunity transcripts in the mouse brain. We have now validated four putative secretory inflammatory markers that are also elevated in mouse liver. We include limited, first time analysis of human Niemann-Pick Type C liver and cerebellum. Furthermore, we utilized 2-hydroxypropyl-β-cyclodextrin (HPβCD, an emerging therapeutic) administered intraperitoneally in mice, which abrogates inflammatory pathology in the liver but has limited effect on the brain. By analyzing the corresponding effects on inflammatory plasma proteins, we identified cathepsin S as a lead indicator of liver disease. In contrast, lysozyme was a marker of both brain and liver disease. 2-Hydroxypropyl-β-cyclodextrin had no effect on transcripts of neuron-specific 24-hydroxylase, and its product 24(S)-hydroxycholesterol was not a useful indicator in mouse plasma. Our data suggest that dual analysis of levels of the inflammatory markers lysozyme and cathepsin S may enable detection of multiple distinct states of neurodegeneration in plasma.

Inflammatory proteins, especially those of innate immunity, are under investigation as biomarkers to monitor disease onset and progression in a wide range of neurodegenerative and metabolic disorders (1)(2)(3). Heterogeneity in the progression of these diseases underscores the critical need for biomark-ers. This is particularly so for inherited lysosomal disorders because they are rare, which increases the challenges of detection and treatment. Multiple organs may be affected, raising the question of whether markers reflect change in one or more organ systems. Plasma markers for neurological disease have been particularly elusive in both rare and more prevalent neurodegenerative disorders (such as Alzheimer and Parkinson diseases).
Niemann-Pick Type C (NPC) 2 is an autosomal recessive neurodegenerative, lysosomal disorder caused by defects in function of either the NPC1 or NPC2 gene, although in ϳ95% of patients, disease is caused by a defect in NPC1 (4). Progressive neurodegeneration is a prominent feature. In addition, NPC is also recognized as a significant cause of liver disease in early life (5)(6)(7). A mouse model, BALB/c Npc1 Ϫ/Ϫ also known as Npc1 nih , where the Npc1 gene is truncated (8) enables the study of aggressive forms of brain and liver disease. Furthermore, because terminal stage disease manifests in less than 90 days, it provides a relatively short model to monitor both neurodegenerative and liver disease.
Multiple inflammatory, innate immune changes have been reported by transcriptional and protein analyses in the liver, spleen, and brain of NPC animals (9 -12). At the cellular level, there is prominent accumulation of foamy macrophages in liver (9,10,13) and activation of microglia in brain (14). Impaired development and reduced natural killer T cells in spleen and thymus have been found in NPC null mice (15,16). In addition, expression arrays suggest transcriptional changes in NPC cells grown in in vitro cultures (17,18).
We investigated conserved transcriptional changes seen in the brain throughout the life span of the Npc1 nih mouse by examining animals at six different ages from weaning to late neurodegeneration (19). These analyses revealed innate immunity trends that could not be obtained from isolated (or a few) time points. We compared them with changes in the liver to identify age-dependent elevation of eight genes of lysosomal innate immunity proteins in the brain and liver, and results * This work was supported, in whole or in part, by National Institutes of Health suggested that they may be potentially suitable as biomarkers for disease in both organs and secreted into plasma. The top candidate, lysozyme, was validated in plasma of Npc1 nih and Npc1 nmf164 (Npc1 nmf ; a BALB/c strain with a point mutation (D1005G) in the NPC1 protein). Our analyses also revealed that neutrophils accumulate in the NPC liver, suggesting a new cellular component that contributes to inflammatory damage there. In independent studies, Cluzeau et al. (20) correlated age-dependent gene expression in mouse liver to identify two plasma markers validated in mice and humans, but their link to molecular changes in the brain was not investigated.
Our interest is also to understand how potential biomarkers and inflammatory changes will serve to assess therapies and their differential effects on disease in brain. To do this, we expanded validation of candidate genes using multiple members of the cathepsin family in brain and liver of murine models. We also extended findings in mice in a limited, first time molecular analysis of human cerebellum and liver. Furthermore, we monitored changes in cathepsins as well as previously identified lysozyme in mice treated with 2-hydroxypropyl-␤-cyclodextrin (HP␤CD; commonly known as cyclodextrin), an emerging therapeutic known to improve disease outcomes in mice (21)(22)(23)(24) and being expanded for use in humans. Cathepsins are cysteine and aspartic proteases that are secreted into the body fluid including blood, and several cathepsins have been identified as blood-based markers for several cancers and inflammatory diseases (25)(26)(27). However, use of cathepsins as plasma biomarkers in neurodegenerative lysosomal disorders has been poorly explored. Lysozyme transcripts are the most highly elevated in the brain, and their elevation in mouse plasma has been reported (19), but how the contribution from the liver could be distinguished from that in the brain remained unknown.

EXPERIMENTAL PROCEDURES
Materials-All fine chemicals were obtained from Sigma unless otherwise indicated. For immunohistochemistry, rat anti-mouse Ly-6G (clone 1A8, BioXcell) was used to detect neutrophils, and monoclonal anti-calbindin (C9848, Sigma) antibody was used for Purkinje neurons. Rabbit anti-cathepsin S (CTSS) (H-50) antibody was from Santa Cruz Biotechnology (Dallas, TX). Antibodies to lysozyme (28) were a kind gift from Professor Tomas Ganz (University of California at Los Angeles). Oligonucleotides for quantitative PCR (qPCR) were purchased from Invitrogen.
Production of Npc1 nih and Npc1 nmf164 Mutant Mice-A breeding pair of Npc1 nih (BALB/c Nctr-Npc1 m1N /J) mice was purchased from The Jackson Laboratory (Bar Harbor, ME). This strain is a widely used NPC BALB/c strain (8) carrying a truncation and premature translation of NPC1 protein that was originally established by Peter Pentchev at the National Institutes of Health (Bethesda, MD). Npc1 nmf164 is a BALB/c strain derived from the recently described Npc1 nmf164 in C57BL/6J (29) that contains an ethylnitrosourea-induced point mutation in the Npc1 gene. The mutation is a single nucleotide change (A to G at cDNA bp 3163) causing an aspartate to glycine change at position 1005 (D1005G) that results in slower disease progression due to partial loss of NPC1 function. The mutation was transferred from C57BL/6J to the BALB/c strain by Robert P.
Erickson, University of Arizona Health Sciences Center (Tucson, AZ). Homozygous mutants of both strains (Npc1 Ϫ/Ϫ ) along with wild type littermates (Npc1 ϩ/ϩ ) were generated by crossing heterozygous mutant (Npc1 ϩ/Ϫ ) males and females in house. Npc1 nih mouse pups were genotyped according to published protocols (8), and Npc1 nmf164 mice were genotyped based on PCR followed by digestion with BstEII (29). In this study, unless otherwise indicated, Npc1 nih mice were used.
RNA Extraction-In mice, formalin-fixed paraffin-embedded tissue was sectioned (4 -5 m), and total RNA was isolated using an RNeasy FFPE kit (Qiagen, Germantown, MD), which included treatment with DNase. Frozen human livers and cerebella from four NPC patients and four age-, gender-, and ethnicity-matched controls were obtained from the NICHD Brain and Tissue Bank for Developmental Disorders (University of Maryland, Baltimore, MD) as approved by the Institutional Review Board of the University of Notre Dame (FWA 00002462). Total RNA was isolated using an RNeasy kit (Qiagen). Eluted RNA was further digested with RNase-free DNase I and repurified using an RNeasy column. The quality of RNA was checked using a Bioanalyzer chip (Agilent Technologies, Santa Clara, CA), and the quantity was determined using a Nanodrop 2000 (Thermo Fisher Scientific, Waltham, MA).
Quantitative PCR-qPCR was performed using a Power SYBR Green RNA-to-C T 1-Step kit and an ABI Prism 7500 Fast Real-time PCR System (Applied Biosystems). The reaction was carried out in 20 l using 100 nM primers and 5-100 ng of total RNA as template. The thermal cycling parameters were as follows: step 1, 48°C for 30 min; step 2, 95°C for 10 min; step 3, 95°C for 15 s; step 4, 60°C for 15 s. Steps 3 and 4 were repeated for 40 cycles followed by melt curve analysis. The nucleotide sequences of gene-specific primers and their sources are listed in Table 1. Specific amplification was validated by analysis of template titration, melt curves, and agarose gel electrophoresis. In both mouse and human tissues, the mRNA levels were normalized to the housekeeping gene glyceraldehyde 3-phosphate dehydrogenase (Gapdh). -Fold change was calculated using a relative standard curve method after correcting for PCR efficiency. In mice, -fold change in expression levels of different genes in Npc1 Ϫ/Ϫ was calculated relative to average levels of expression in Npc1 ϩ/Ϫ mice. In human tissues, -fold change in transcript expression in NPC liver and cerebellum was expressed relative to average expression in age-matched controls.
Lysozyme Activity Assay-Lysozyme activity in plasma was measured using a fluorescence-based lysozyme assay kit (Enz-Check, Invitrogen) as describer earlier (19). Plasma corresponding to 25 g of protein from female and male Npc1 nih mice was used in a 100-l reaction volume. The reaction was carried out at 37°C for 24 h. Fluorescence was read using excitation/emission of 494/518 nm in a multiwell plate reader (Spectramax M2, Molecular Devices). The values obtained were normalized by dividing them by the mean value of lysozyme obtained among untreated Npc1 ϩ/Ϫ mice. Purified chicken egg white lysozyme was used as a positive control.
Cathepsin S ELISA-Total plasma cathepsin S was determined using an ELISA Duo Set kit (DY1183) from R&D Systems (Minneapolis, MN) according to the manufacturer's instructions. Plasma of Npc1 ϩ/ϩ and Npc1 ϩ/Ϫ mice of both Npc1 nih and Npc1 nmf164 strains was diluted to 1:10, whereas that of Npc1 Ϫ/Ϫ mice of both strains was diluted to 1:20. All measurements were done in triplicate wells. For normalization, the raw absorbance values were divided by the average absorbance of Npc1 ϩ/Ϫ mice of each strain of a given age group.
24(S)-Hydroxycholesterol (24(S)-HC) ELISA-Plasma 24(S)-HC concentration was determined using an ELISA kit from Enzo Life Sciences (Farmingdale, NY) according to the manufacturer's instructions. Plasma was diluted to 1:1000 in the supplied buffer, and measurements were done in triplicate wells. Pure 24(S)-HC (supplied with the kit) was used to prepare the standard curve. 24(S)-HC concentration was normalized to plasma protein content.
Organ Harvest and Immunohistochemistry-Mice were sacrificed by asphyxiation using CO 2 . The circulatory bed was washed with PBS (pH 7.4) and subsequently perfused with 10% neutral buffered formalin (ϳ4% formaldehyde). The organs (brain and liver) were surgically harvested and stored in 4% formaldehyde at room temperature until transfer to paraffin. Paraffin-embedded tissue sections (3-4 m) were dewaxed in xylene and alcohol. For Ly-6G and calbindin staining, antigen retrieval was done by preincubating deparaffinized samples with 0.05% proteinase K (Dako, Germany) in 50 mM Tris-HCl (pH 7.5) for 8 min at room temperature. CTSS and lysozyme were retrieved by boiling the sections in acidic conditions for 30 min. Sections were incubated with anti-Ly-6G (20 g/ml), anticalbindin (1:1000), anti-CTSS (20 g/ml), or anti-lysozyme (1:20) overnight at 4°C. Reagents were prepared according to the manufacturer's instructions (Vector Laboratories). The staining protocol used was described previously (19). The secondary antibody for neutrophil staining was biotinylated rabbit anti-rat IgG (mouse absorbed, Vector Laboratories) and for Purkinje neurons was biotinylated horse anti-mouse IgG (Vector Laboratories).
For fluorescence microscopy, FITC-conjugated IgG (MP Biomedicals, Solon, OH) was the secondary antibody. Sections stained only with secondary antibodies served as controls. Bright field images were acquired on a Nikon Olympus microscope using a Nikon digital DS-Fi1-U2 camera controlled by NIS-Elements F3.0 Nikon software (all from Nikon Instruments Inc., Tokyo, Japan). Images were visualized with an A10 PL 10ϫ/0.25, DPlan Apo 40ϫ/1.00 oil immersion, or DPlan Apo 100ϫ/1.30 oil immersion objective lens (Nikon). Fluorescence microscopy and digital image collection were performed using an Olympus IX inverted fluorescence microscope and a Photometrix cooled charge-coupled device camera (CH350/ LCCD) driven by DeltaVision software from Applied Precision (Seattle, WA). DeltaVision software (softWoRx) was used to deconvolve these images. Images were visualized with a 40ϫ oil immersion objective lens and are single optical sections. NIH ImageJ software was used to process and quantify the fluorescence intensity of CTSS and lysozyme.
Drug Injections and Blood Withdrawal-Starting at P21and once a week thereafter, Npc1 nih and Npc1 nmf164 mice were injected intraperitoneally with 20% HP␤CD (4000 mg/kg) prepared in 0.2% DMSO and 0.9% saline. Control mice received 0.2% DMSO in 0.9% saline. Blood was collected in EDTA tubes (BD Biosciences) either via cheek bleed or terminal heart bleed This study R, 5Ј-GTTGTCATACCAGGAAATGAGC-3Ј from mice. Plasma was separated by centrifugation at 2500 rpm for 15 min and stored at Ϫ70°C until used.
Statistical Tests-Student's t test was carried out to determine the statistical significance of the data. p Ͻ 0.05 was considered significant.

Validation of Cathepsins B (Ctsb), D (Ctsd), and S (Ctss) in
Liver and Brain of NPC Mice and NPC Patients-Of 12 potential biomarker genes identified in our previous study (19), three belonged to the cathepsin family. These were Ctsb, Ctsd, and Ctss. Although there is no information about cathepsin S in NPC disease, cathepsins B and D have been reported to be overexpressed in the cerebellar neurons in Npc1 Ϫ/Ϫ mouse brain and have been linked to increased neurodegeneration (30 -32), suggesting that the family may be suitable for further investigation.
Disease progression as a function of age in Npc1 Ϫ/Ϫ mice is shown schematically in Fig. 1A. Our microarray data suggested that the -fold up-regulation of Ctsb, Ctsd, and Ctss was 1.5, 3.2, . D, panels i-iii, expression analysis of CTSS, CTSD, and CTSB in liver and cerebellum of human NPC patients. Total RNA was isolated from frozen livers and cerebella from four NPC and four control subjects. Expression levels of cathepsins were determined by qPCR. -Fold change is relative to the average value of control subjects. Change above 1 (shown by dotted line) represents the extent of overexpression. For both mouse and human qPCR studies, Gapdh was used as an internal control. *, p Ͻ 0.005. and 6.2, respectively, in the liver of late stage Npc1 Ϫ/Ϫ mice compared with age-matched control mice (Table 2). Similarly, transcript levels of Ctsb, Ctsd, and Ctss were 2.8-, 1.9-, and 2.7-fold higher in the brain of Npc1 Ϫ/Ϫ mice compared with controls ( Table 2). To validate the microarray data, we performed qPCR to determine transcript increases for Ctsb, Ctsd, and Ctss in liver and brain of Npc1 Ϫ/Ϫ mice at a late symptomatic stage (Fig. 1). As shown in Fig. 1B (panels i-iii), in the liver, the -fold change was 4.4 for Ctsb, 11.7 for Ctsd, and 60. 6 for Ctss in Npc1 Ϫ/Ϫ compared with age-matched control mice. In the brain, the -fold increase of Ctsb was 1.5, that of Ctsd was 3.7, and that of Ctss was 3.4 ( Fig. 1C, panels i-iii).
The -fold change detected by qPCR was not the same as seen in the microarrays. Many factors such as mRNA extraction and stability, hybridization efficiency, and differences in the efficiency of cDNA synthesis may contribute to this discrepancy. Although microarrays are useful in obtaining trends of change, qPCR provides the quantitative confirmatory data.
Cluzeau et al. (20) have reported that plasma cathepsin D is elevated in NPC patients. However, information on levels of cathepsins D, S, and B in human organs is not available. Therefore, we obtained frozen livers and cerebella from four NPC and four control subjects matched for age, gender, and ethnicity. As shown in Fig. 1D, panel i, we detected increased transcripts of CTSS in liver (1.4-, 1.8-, and 2.6-fold) as well as in cerebellum (1.4-, 1.7-, and 2.8-fold) of three NPC patients. In the fourth NPC patient, CTSS transcript was unchanged in liver but decreased in cerebellum compared with controls ( Fig. 1D, panel i). In contrast, CTSD expression was not increased in liver, but increases (1.3-and 7.6-fold) were seen in two NPC cerebella (Fig. 1D, panel ii). CTSB showed elevation (5.2-fold) in one of four liver samples and in the cerebellum of all four NPC patients (1.3-, 1.3-, 1.5-, and 1.7-fold) (Fig. 1D, panel iii). Because the sample size is small, the data do not rule out CTSD or CTSB as potential markers. Nevertheless, because there was an increase in CTSS in three of four patient samples for both organs and CTSS showed the greatest change in the mouse liver (60.6-fold), we investigated it as the lead marker of interest in subsequent work.
We further examined plasma from Npc1 nmf mouse. Previous studies suggested that Npc1 nmf mice in the C57BL/6J background have a life span of ϳ112 days and develop progressive disease (29). BALB/c Npc1 nmf have a comparable life span (ϳ120 -125 days) and exhibited weight loss from 85 to 90 days (19). As shown in Fig. 2C, plasma CTSS levels were indeed elevated ϳ1.4 -1.6-fold (p Ͻ 0.05) in both early (ϳ75 days) and later (100 days) symptomatic stages. Remarkably, HP␤CD reduced levels of CTSS at late stages (80 -114 days) in Npc1 nih (Fig. 2E) and mild to moderately symptomatic Npc1 nmf mice to those seen in healthy controls (Fig. 2, E and F). These findings were surprising because HP␤CD-treated Npc1 nih and Npc1 nmf mice manifest disease at 100 days of age.
Effect of HP␤CD Treatment on Ctss, Ctsd, Ctsb, and Other Inflammatory Marker Expression in Liver and the Pathologies of the Organ in Npc1 nih Mice-To investigate whether cathepsin levels in the plasma of Npc1 Ϫ/Ϫ mice reflect disease status of the liver and its response to HP␤CD, we studied the effect of treatment on (i) the expression levels of Ctss, Ctsd, and Ctsb and (ii) liver pathology. After HP␤CD treatment, the expression of Ctss in the liver of late stage Npc1 Ϫ/Ϫ mice was markedly reduced and equivalent to that in control mice (Fig. 3A). Similar trends were also observed in the expression of Ctsd and Ctsb (Fig. 3, B and C).
We undertook analysis of additional inflammatory markers and histology. We studied the expression of two inflammatory genes, Cd68 (macrophage marker) and Itgax (a marker of activated macrophage, granulocytes, dendritic cells, etc.; also known as Cd11c). qPCR analysis showed that Cd68 was up-regulated by ϳ88-fold (Fig. 3D) and Itgax was up-regulated by ϳ400-fold (Fig. 3E) in Npc1 Ϫ/Ϫ mice at late stages (70 -83 days) in the liver. These were reduced to normal levels after HP␤CD treatment (Fig. 3, D and E), suggesting amelioration of liver inflammation. A third marker, Cola1 (procollagen type a1), shown to be up-regulated during liver fibrosis (10) was increased (ϳ1.3-fold) in Npc1 Ϫ/Ϫ . HP␤CD induced anomalous reduction in its expression (Fig. 3F), suggesting that although HP␤CD treatment reduced inflammation it may also adversely change levels of important molecular determinants of the liver.
To study the expression of CTSS protein and its localization in the liver, sections were subjected to immunohistochemistry using anti-CTSS antibodies. The liver of Npc1 ϩ/Ϫ mice (age, 80 days) showed healthy hepatocyte architecture (Fig. 4A). In contrast, numerous large foamy macrophages containing high levels of CTSS were seen in Npc1 Ϫ/Ϫ mice of the same age (Fig. 4B, blue arrows). Saline had no effect (Fig. 4C), but HP␤CD treatment eliminated accumulation of foamy macrophages and dramatically reduced CTSS accumulation in Npc1 Ϫ/Ϫ (Fig. 4D).
We reported previously that giant foci of neutrophils accumulate in liver of Npc1 Ϫ/Ϫ mice, suggesting that they contribute to the inflammatory response (19). Therefore, we also examined the effects of HP␤CD on neutrophil accumulation in the liver. Immunohistochemical analyses did not show neutrophil infiltration in healthy mice at 54 and 80 days (Fig. 4, E1 and  E2). In contrast, in diseased mice, clusters of neutrophils were clearly seen (Fig. 4, F1, Ly-6G-positive cells in brown, shown by blue arrows). The number of neutrophils increased in the liver of diseased mice at 80 days (Fig. 4, F2). Administration of saline had no effect (Fig. 4, G1 and G2), but HP␤CD treatment reduced neutrophil clusters at both 50 (Fig. 4, H1) and 80 days (Fig. 4, H2). However, there were differences in hepatocyte architecture (decreased cytoplasmic staining with slightly irregular plasma membrane) in animals treated with HP␤CD, suggesting that treatment does not completely restore all aspects of liver health (at least as judged by histochemistry).
Our previous microarray studies also reported the up-regulation of lysozyme transcripts in the liver of diseased animals.
As shown in Fig. 4I, we detected ϳ85-fold up-regulation of Lysozyme1 (Lyz1) gene in the Npc1 Ϫ/Ϫ liver at late stages of disease compared with age-matched control mice. HP␤CD treatment abrogated Lyz1 overexpression in the liver of Npc1 Ϫ/Ϫ mice, bringing transcript levels back to those seen in healthy mice (Fig. 4I). To confirm that lysozyme protein levels were also elevated and determine the site(s) of concentration, we undertook immunohistochemical analyses of liver sections using antibodies to mouse lysozyme (see "Experimental Procedures"). Large foamy macrophages contained high levels of lysozyme in the liver of Npc1 Ϫ/Ϫ mice (age 80 days). They were absent in healthy animals (Fig. 4, J and K). HP␤CD treatment (Fig. 4M) but not saline (Fig. 4L) largely eliminated the macrophages containing lysozyme from the liver. A low, basal level of lysozyme expression was seen in hepatocytes but at levels comparable with that in healthy mice (Fig. 4M).
Together, these data provide new markers (such as CTSS and lysozyme) to confirm prior findings that HP␤CD treatment improves inflammation in the liver. Furthermore, they show that HP␤CD treatment returns CTSS levels in the liver of diseased animals to that seen in healthy counterparts (Figs. 3 and 4), which is analogous to effects seen with this marker in plasma (Fig. 2).
Effect of HP␤CD Treatment on Ctss, Ctsd, and Ctsb Expression in Brain and the Pathologies of the Organ in Npc1 nih Mice-In the brain, HP␤CD treatment resulted in partial reduction of Ctss and Ctsd (Fig. 5, A and B), whereas Ctsb expression returned to normal levels (Fig. 5C). Activation of microglia and astrocytes in the brain has been associated with neuroinflammation and neurodegeneration in NPC disease (14,33,34). qPCR analyses of associated markers revealed ϳ24-fold up-regulation of Cd68 in Npc1 Ϫ/Ϫ mice (70 -83 days) compared with age-matched Npc1 ϩ/Ϫ mice (Fig. 5D). The expression of Cd68 in Npc1 Ϫ/Ϫ mice treated with HP␤CD was reduced significantly, but there was still ϳ6-fold higher transcript relative to Npc1 ϩ/Ϫ mice (Fig. 5D). Furthermore, mutant animals showed an ϳ44-fold increase in Mip-1␣ transcript, which was reduced to ϳ27-fold upon treatment with HP␤CD (Fig. 5E). Untreated Npc1 Ϫ/Ϫ mice at late stages showed ϳ9-fold higher Gfap transcript compared with age-matched controls (Fig. 5F) that was slightly reduced in HP␤CD-treated animals (Fig. 5F). These results are consistent with the data with Ctsb, Ctsd, and Ctss as markers and prior reports that HP␤CD partially alleviates neuroinflammation in Npc1 Ϫ/Ϫ mice (23, 35-37).
Immunohistochemical analyses of brain sections showed enhanced labeling of CTSS in the pyramidal neurons of hippocampus of Npc1 Ϫ/Ϫ mice (Fig. 5, G-J), recognized by their characteristic morphology. Quantitative analysis using ImageJ revealed an 1.7-fold (p Ͻ 0.05) increase in CTSS in Npc1 Ϫ/Ϫ mice. HP␤CD treatment slightly reduced CTSS levels (ϳ20%), whereas saline had no effect (Fig. 5K). CTSS was expressed at low levels in the rest of the brain (not shown), but there were no differences between Npc1 ϩ/Ϫ and Npc1 Ϫ/Ϫ mice elsewhere including the cerebellum (data not shown).
Loss of Purkinje neurons in the cerebellum is a characteristic feature of NPC disease and has been used as a benchmark to study brain pathology (24,34,38,39). To study the effect of HP␤CD on Purkinje neuron death, immunohistochemical staining of sagittal sections of brain of Npc1 Ϫ/Ϫ mice was carried out using anti-calbindin (a marker of Purkinje neurons) antibody. Mouse cerebellum is composed of 10 (I-X) different lobules. All cerebellar sections were examined; however, in Fig. 6, micrographs corresponding to IX lobule are shown as representative images. Numerous Purkinje neurons (Fig. 6A, stained in brown, shown by arrows) were clearly seen in the cerebellar section of Npc1 ϩ/Ϫ mouse at 80 days, whereas in Npc1 Ϫ/Ϫ mice of the same age, the numbers were markedly reduced (five to eight/lobule) (Fig. 6B). This was unchanged upon treatment with saline (Fig. 6C). However, treatment with HP␤CD preserved additional Purkinje neurons in the IX lobule of the cerebellum (Fig. 6D) although at lower levels than in Npc1 ϩ/Ϫ animals. The intensity of calbindin-positive neurons in HP␤CD-treated Npc1 Ϫ/Ϫ mice were reduced compared with Npc1 ϩ/Ϫ animals (Fig. 6, compare intensity of brown staining seen in A1 and A2 with that in D1 and D2). Purkinje neurons were also seen in the X lobule (data not shown). Even after HP␤CD treatment, Purkinje neurons were barely seen in the rest of the cerebellar regions of Npc1 Ϫ/Ϫ mice  MARCH 21, 2014 • VOLUME 289 • NUMBER 12

JOURNAL OF BIOLOGICAL CHEMISTRY 8057
(data not shown). A semiquantitative analysis of Purkinje neurons in a cerebellar section showed they were significantly reduced in number in Npc1 Ϫ/Ϫ (39 Ϯ 11) mice as compared with Npc1 ϩ/Ϫ mice (473 Ϯ 75). Saline treatment showed no effect (40 Ϯ 8), whereas HP␤CD treatment showed an increased number of Purkinje neurons (100 Ϯ 16) (Fig. 6E). The data are consistent with prior results (22, 24, 40) that HP␤CD treatment may to a small degree resolve neuroinflammation and inhibit loss of Purkinje neurons. Furthermore, the data are consistent with our marker analysis for inflammatory proteins including the cathepsins.

Characterization of Lysozyme Levels in Plasma and Brain in HP␤CD-treated Npc1 nih Mice at Terminal Stages of Disease: Localization of Lysozyme Elevation in the Cerebellum and Development of a Composite Scale to Distinguish among Four
Distinct States of Cerebral and Liver Disease-Our prior study (19) identified lysozyme transcripts as the most highly elevated in the brain of Npc1 nih mice. We further validated the elevation of lysozyme in the plasma of Npc1 nih as well as a second mouse model, Npc1 nmf . Additionally, we showed that plasma lysozyme levels elevated in asymptomatic Npc1 nmf mice (age, ϳ50 days) were reduced by HP␤CD treatment and rendered comparable with untreated wild type animals (19).
To examine time points of advanced disease, we returned to the shorter Npc1 nih model. This model typically manifests phenotypic symptoms (weight loss, gait, tremor, etc.) from ϳ50 to 55 days and survives up to ϳ80 -84 days (Fig. 7A). Previous studies have shown that weekly injections of HP␤CD to Npc1 nih reduce disease and extend survival (23,24). Therefore, we treated Npc1 Ϫ/Ϫ mice with HP␤CD or vehicle control (0.2% DMSO in 0.9% saline) by once a week drug injections (4000 mg/kg) starting at age P21. Npc1 nih mice treated with this regime showed delayed onset of symptoms (from ϳ49 -56 to ϳ70 -80 days) and survived ϳ105-112 days (diagrammatically represented in Fig. 7A). As reported earlier, the plasma lysozyme activity of vehicle-treated Npc1 Ϫ/Ϫ mice was elevated on average 2-fold in early symptomatic animals (50 -60 days) (Fig. 7B) and remained elevated at the late symptomatic stage (70 days) compared with age-matched controls (Fig. 7C). Furthermore, HP␤CD treatment at 50 -60 days reduced plasma lysozyme activity levels seen in wild type mice (Fig. 7B). At 80ϩ days, HP␤CD-treated animals showed a reduction of plasma lysozyme compared with mock-treated animals but nonetheless displayed a 1.5-fold increase compared with normal animals (Fig. 7C). This persistent elevation of plasma lysozyme could not be derived from the liver because as shown previously HP␤CD treatment restored lysozyme transcript and protein to normal levels in the liver (Figs. 3 and 4). Therefore, we examined the brain where qPCR revealed an ϳ81-fold up-regulation of Lyz1 in Npc1 Ϫ/Ϫ mice at late stages of disease (Fig. 7D). Moreover, after HP␤CD treatment, Lyz1 expression remained elevated by ϳ28-fold (Fig. 7D).
In immunolocalization studies by fluorescence microscopy, low levels of lysozyme were detected throughout the brains of normal and diseased mice except in the cerebellum where there was a marked increase in the mutant animals (Fig. 8, B-K). Numerous cells highly positive for lysozyme were seen in the cerebellar white matter of Npc1 Ϫ/Ϫ mice (80 days) but not in Npc1 ϩ/Ϫ mice (Fig. 8, B and C). HP␤CD treatment had no significant effect on the lysozyme levels in the cerebellum (Fig. 8, E and F). Lysozyme was also elevated in the molecular layer of the cerebellum (Fig. 8, G-K). In this region, healthy mice showed minimal levels of lysozyme staining (Fig. 8G), whereas elevated fibrillar staining was seen in the Npc1 Ϫ/Ϫ mice (Fig. 8H). Saline treatment had no effect, but HP␤CD treatment resulted in a minor reduction (ϳ25%; Fig. 8, I-K). Although CTSS was increased in the hippocampus of mutant mice (Fig. 5, G-J) lysozyme was unchanged here (Fig. 8, L-P) and as indicated earlier in the rest of the brain (data not shown). In several instances, -fold changes in plasma levels of both cathepsin and lysozyme were considerably lower than their transcript levels in brain and liver but more in keeping with changes seen by immunohistochemistry, which is to be expected because the latter is a readout of protein levels in tissue. Taken together, these data suggest that HP␤CD given postweaning into the body cavity can deplete lysozyme in the liver. It may also reduce to a small extent lysozyme in the brain, but significant levels persist. Remarkably, lysozyme elevation in the brain of diseased animals appears concentrated in the cerebellum whose function is prominently compromised in NPC (summarized in Fig. 8Q).
Nonetheless, prior to HP␤CD treatment, plasma levels of lysozyme likely reflect inflammation in the brain as well as the liver in mice and humans (Figs. 4, 7, 8, and 9A). To estimate the contribution from the inflamed liver, we needed a second marker whose levels in plasma solely reflect that of the liver (such as CTSS). Thus, we considered that lysozyme along with CTSS may contribute to a composite, quartile scale for both cerebral disease and inflamed liver (Fig. 9B, panels i-iii). The first two quartiles reflect elevated lysozyme and thus potential contribution from cerebral disease (Fig. 9B). However, simultaneous elevation of CTSS in the second quartile suggests liver inflammation that can also contribute lysozyme to the plasma. Accordingly, HP␤CD, which abrogates CTSS, decreases lysozyme by ϳ50%, reflecting that one-half of lysozyme activity is contributed by the liver and the other half is contributed by the brain. Thus, for the same lysozyme levels, quartile 2 is expected to reflect more moderate levels of cerebral disease compared with quartile 1 (Fig. 9B). Quartile 4 reflects low disease, whereas quartile 3 is indicative of just liver disease (Fig. 9B).
Comparison with Oxysterol Markers-Oxysterol species are emerging as markers of NPC disease (41,42). Plasma oxysterols (7-ketocholetserol and 3␤,5␣,6␤-triol) generated by non-enzymatic pathways are largely produced by the liver (43,44) and thus are likely to be more useful to understand liver pathology rather than brain pathology. However, 24(S)-HC is derived from cholesterol by an enzyme, 24-hydroxylase, which is primarily expressed in the neurons of the central nervous system (45,46). As shown in Fig. 10A, we detected an ϳ25% reduction in the expression of Cyp46a1 (24-hydroxylase) gene in the brain of Npc1 Ϫ/Ϫ mice at late stage disease. This is consistent with a slight reduction of 24(S)-HC reported in NPC patients (42), but our analysis of human cerebellum in four patients suggested variability in transcript levels (Fig. 10B). Unexpectedly, the levels of 24(S)-HC in the plasma of Npc1 Ϫ/Ϫ mice were elevated and remained largely unaffected after HP␤CD treatment (Fig. 10C). Thus, plasma 24(S)-HC may not assess neuropathology in mouse models.

DISCUSSION
Oxysterols are emerging as sensitive blood-based biomarkers for NPC (42). However, they are largely products of the liver, not the brain. In addition, the disease is heterogeneous with respect to both neurological and metabolic symptoms as well as age of onset, which strongly argues for the need for multiple markers.
Elevation of several cathepsins including CTSB, CTSD, and CTSS has been implicated in neurodegenerative diseases (47). The level and activity of CTSB and CTSD are elevated in the hippocampal, cerebellar, and cortical neurons (30,31,34) of Npc1 Ϫ/Ϫ mice. By immunohistochemistry, CTSS can be detected in almost all regions of the brain. However, CTSS was elevated only in hippocampal neurons of Npc1 Ϫ/Ϫ mice compared with healthy counterparts. It is possible that the hippocampal neurons can tolerate a minor elevation of these proteases and remain resistant to degeneration. Increased cytosolic levels of CTSB and CTSD have been shown to activate the autophagic pathways, thereby leading to neuronal death in Npc1 Ϫ/Ϫ cells or mice (31,32). CTSS may do the same. Addi-tionally, activated microglia can release CTSB and CTSD that along with CTSS can induce neuronal death through digestion of extracellular matrix (48).
But importantly, CTSS detected in plasma of NPC mouse models does not reflect cerebral disease but is derived largely from the liver. Our studies suggest that among the cathepsins, cathepsin S appears to be the best candidate biomarker for liver disease. Although transcript analysis in mouse organs suggests that Ctss increases gradually, direct measurements in plasma revealed high levels from the outset. The marked elevation of Ctss in the liver and its concomitant responsiveness to HP␤CD treatment in plasma and liver suggest that it may be a preferred marker of early liver disease. This is of value because although neurodegeneration is a prominent feature and linked to fatal disease, NPC is recognized as a significant cause of liver disease in early life (5)(6)(7). A history of neonatal jaundice or persisting hepatosplenomegaly is common among patients with early and late infantile onset disease. NPC is the second most common cause of neonatal cholestasis resulting in liver failure and death of ϳ10% patients (49,50). Thus, along with oxysterols, plasma CTSS may also help in the diagnosis of NPC, particularly in a newborn child or infants manifesting cholestatic jaundice along with hepatomegaly or splenomegaly. The Purkinje cell layer in the cerebellum contains two types of cells, Purkinje neurons and Bergmann glial cells. At advanced disease states, the Purkinje neurons are largely lost in Npc1 Ϫ/Ϫ mice. This suggests that increased lysozyme in the Purkinje cell layer and molecular layer is due to its expression and secretion by Bergmann glial cells. Activated microglia and Bergmann glial cells may secrete a higher level of lysozyme that may also play a role in the loss of Purkinje neurons in Npc1 Ϫ/Ϫ mice (through mechanisms that remain undefined). Lysozyme at higher concentration has been shown to be amyloidogenic (51), and exposure of cultured rat neurons to oligomers of hen egg white lysozyme has been found to induce hyperphosphorylation of tau (52). In fact, neurons expressing lysozyme have been shown to have increased hyperphosphorylated tau in the mucopolysaccharidosis type IIIB mouse brain (28). Therefore, it is plausible that overexpression of lysozyme may allow it to reach a critical concentration at which it either oligomerizes or aggregates and serves as a template for the aggregation of tau and its phosphorylation in the cerebellum. Importantly, cerebellar ataxia is a major clinical symptom of NPC.
Prior studies have suggested that macrophage activation and accumulation in the liver are responsive to HP␤CD treatment (22,23). We confirm that with two new markers, CTSS and lysozyme, and also show that neutrophil accumulation is reduced, suggesting that both types of inflammatory cells respond to lipid accumulation. One possibility is that anoma-  gene in NPC mouse brain. The Cyp46a1 RNA level was determined by qPCR with total RNA extracted from the brain of Npc1 ϩ/Ϫ (ϩ/Ϫ; n ϭ 6), Npc1 Ϫ/Ϫ (Ϫ/Ϫ; n ϭ 6), and HP␤CD-treated Npc1 Ϫ/Ϫ (Ϫ/Ϫ; n ϭ 6) mice. -Fold increase is expressed relative to transcript levels in Npc1 ϩ/Ϫ mice. B, expression of CYP46A1 gene in the cerebella of four NPC and four control subjects. Expression levels were determined by qPCR. -Fold change is relative to the average value of control subjects. Change above and below 1 (shown by the dotted line) represents the extent of up-and down-regulation, respectively. For both mouse and human qPCR studies, Gapdh was used as an internal control. Horizontal bars show the median values. C, 24(S)-HC levels in Npc1 nih Npc1 Ϫ/Ϫ (Ϫ/Ϫ) mice treated with saline or HP␤CD compared with Npc1 ϩ/Ϫ (ϩ/Ϫ). The plasma concentration of 24(S)-HC was determined by ELISA (see "Experimental Procedures"). lous neutrophil migration occurs in response to changes in lipid gradients to inflict inflammatory damage, which is then removed by macrophage action. At late stages of disease, reduction of inflammatory proteins lysozyme and CTSS in plasma closely corresponds to reduction of inflammation in the liver. However, the liver is not completely "normal." The observed reduction of collagen in liver can be correlated with compromised cellular organization, suggesting that high levels of HP␤CD in circulation may also have adverse effects on the liver. Nonetheless, the dramatic reduction in inflammation may outweigh adverse effects, resulting in a net benefit. Further studies are required to establish improved liver function.
HP␤CD injections have been shown previously to slightly but detectably improve brain pathology and levels of inflammatory markers (21-24, 35, 37). Our data are consistent with these findings both with respect to organ pathologies and marker analysis. Nonetheless, the improvement in liver pathology after HP␤CD treatment far exceeds that in the brain. In the initial microarray analysis of the age-dependent increase in transcripts, lysozyme was the topmost transcript hit. Because both brain pathology and plasma lysozyme levels are relatively refractory to intraperitoneal HP␤CD injections, it is likely they are linked. Indeed, HP␤CD-treated animals, although rescued in liver pathology, nonetheless die of cerebral disease.
How loss of NPC protein function leads to neuroinflammation is poorly understood. One possibility is that lysosomal functions are compromised due to harmful accumulation of cholesterol and other lipids. In response, cellular systems may compensate the functional loss by overexpressing lysosomal proteins such as cathepsins and lysozyme. This may be a general phenomenon as neuroinflammation is the hallmark of almost all lysosomal storage diseases (2,53). Malfunctioning of the lysosomal system may hamper phagocytosis, rapid membrane synthesis, and recycling in macrophages and microglial cells, which in turn may lead to their activation and subsequent overexpression of markers of neuroinflammation.
Inflammatory proteins corresponding to members of chemokines and cytokines family have been explored in cerebrospinal fluid of NPC patients; however, further investigation is required to establish their usefulness as biomarkers (54). Oxysterols largely reflect liver function, although 24(S)-HC has been proposed as a marker for neuronal disease in humans because it is produced in the brain. However, the Npc1 Ϫ/Ϫ mouse model failed to provide insights into the utility of this marker for human disease.
Rather, our data show that plasma lysozyme is derived from the brain and overexpressed in the cerebellum. This is important because cerebellar ataxia is a major symptom of NPC. Lysozyme in conjunction with CTSS may be used to distinguish distinct states of brain and liver disease that have hitherto not been possible but would be very helpful to monitor the progression and management of human disease. In this regard, mouse models may be particularly helpful in dissecting the differential response of major disease organs to emerging therapeutics in both preclinical and clinical studies.