Cerebellar Alterations and Gait Defects as Therapeutic Outcome Measures for Enzyme Replacement Therapy in α-Mannosidosis

Abstract

α-Mannosidosis is a rare lysosomal storage disease with accumulation of undegraded mannosyl-linked oligosaccharides in cells throughout the body, most notably in the CNS. This leads to a broad spectrum of neurological manifestations, including progressive intellectual impairment, disturbed motor functions and cerebellar atrophy. To develop therapeutic outcome measures for enzyme replacement therapy (ERT) that could be used for human patients, a gene knockout model of α-mannosidosis in mice was analyzed for CNS pathology and motor deficits. In the cerebellar molecular layer, α-mannosidosis mice display clusters of activated Bergman glia, infiltration of phagocytic macrophages and accumulation of free cholesterol and gangliosides (GM1), notably in regions lacking Purkinje cells. α-mannosidosis brain lysates also displayed increased expression of Lamp1 and hyperglycosylation of the cholesterol binding protein NPC2. Detailed assessment of motor function revealed age-dependent gait defects in the mice that resemble the disturbed motor function in human patients. Short-term ERT partially reversed the observed cerebellar pathology with fewer activated macrophages and astrocytes but unchanged levels of hyperglycosylated NPC2, gangliosides and cholesterol. The present study demonstrates cerebellar alterations in α-mannosidosis mice that relate to the motor deficits and pathological changes seen in human patients and can be used as therapeutic outcome measures.

INTRODUCTION

Lysosomal storage disorders (LSDs) are a group of individually rare human genetic diseases characterized by defects in lysosomal hydrolysis of macromolecules (e.g. lipids and glycoproteins), leading to accumulation of undegraded material in the lysosome (1). α-Mannosidosis is a progressive LSD caused by deficiency of the lysosomal hydrolase α-mannosidase (LAMAN) (2, 3), and accumulation of polymannose oligosaccharides in the endosomal/lysosomal system. Abnormalities in α-mannosidosis patients include intellectual and psychiatric disability and prominent motor defects (4). These motor defects are due to multiple factors including orthopedic pathology, but also ataxia that results from cerebellar atrophy and demyelination (5, 6). Patients with α-mannosidosis patients present a variety of clinical symptoms. At present, 3 clinical types have been suggested: a mild and moderate form with late onset (>10 years) and slow disease progression (type I); development of ataxia at the age of 20 to 30 years (type II); and severe forms leading to an early death from CNS involvement or myopathy (type III). Most patients have the moderate form (type II) of the disease (4).

Animal models of α-mannosidosis in guinea pigs, cats, and mice have been used to study the consequences of α-mannosidase deficiency and to develop therapeutic approaches for this carbohydrate storage disorder. In particular, mice with targeted disruption of the LAMAN gene Man2b1 (7) were shown to be a valid model for α-mannosidosis (8, 9). Cerebellar pathology comparable to that in patients has been described in both guinea pigs and cats with this disease, but not yet in α-mannosidosis mice. Behavioral analyses revealed neurocognitive impairments in adult α-mannosidosis mice that mimic many aspects of human α-mannosidosis (8–10). Detailed assessment of motor system dysfunction, including sensitive methods of quantitative gait analysis, and cerebellar pathology in these mice has not been performed. At present, no therapeutic treatment other than bone marrow transplantation is available for α-mannosidosis patients but preclinical enzyme replacement therapy (ERT) in the mouse model has shown promise (10, 11). Short-term, high-dose ERT successfully decreased neuronal storage of sugars in the brain of α-mannosidosis mice and correlated with improved neuromotor abilities (10). Inasmuch as clearance of stored sugars was prominent in the hippocampus and other brain regions but not in the cerebellum, it is unclear as to how ERT contributed to the observed improvements in motor function.

To investigate whether neuromotor deficits in α-mannosidosis mice relate to the cerebellar pathology that is comparable to that seen in humans, and that can be used as outcome measures in therapy studies, we performed detailed histological and biochemical analyses on the cerebellum of α-mannosidosis mice. Sensitive methods for quantitative gait analyses are included to determine the functional consequences of the cerebellar findings. In addition, the effects of a short-term, high-dose ERT regimen on the pathological alterations were evaluated.

MATERIALS AND METHODS

Animals

α-Mannosidase knockout (KO) mice, (referred to here as “α-mannosidosis mice”) were generated by targeted disruption of the Man2b1 gene leading to a deficiency in LAMAN activity (7). KO mice and their wild-type (WT) littermates were bred on a C57Bl/6 background and maintained under standard housing conditions. Genotyping was carried out by polymerase chain reaction, as described (7). Six- to 14-month-old α-mannosidosis animals and age-matched WT animals were studied. All animal experiments were approved by local authorities.

Primary Antibodies and Reagents

Monoclonal mouse anti-glial fibrillary acidic protein (GFAP) and Filipin were purchased from Sigma-Aldrich (St. Louis, MO). Monoclonal mouse anti-Lamp1 (1D4B) and mouse anti-Lamp2 (Abl-93) were purchased from Developmental Studies Hybridoma Bank (Iowa City, Iowa). Antibodies to myelin basic protein (MBP) and calbindin were from Calbiochem (Darmstadt, Germany), to GM1 from Biomol (Hamburg, Germany), and to Actin from Sigma-Aldrich. Rat anti-CD68 and mouse anti-F4/80 were from Abd Serotec (Oxford, UK). Rabbit anti-NPC2 protein antiserum was a kind gift from Shutish Patel (12). Goat anti-mouse IgG, goat anti-rabbit IgG and Vectastain ABC kit were from Vector Laboratories (Burlingame, CA). Fluorophore conjugated secondary antibodies (Alexa Fluor 488 and 546) and DAPI were purchased from Molecular Probes (Eugene, OR). Appropriate negative controls omitting primary antibodies were included in immunohistochemistry (IHC) protocols to confirm specificity. Chemicals were purchased from Sigma-Aldrich.

Intravenous Injection of Recombinant Human LAMAN

Recombinant human LAMAN was purified from CHO cells as described (11). It was injected twice a week (4 ×) into the tail vein of adult mice in a dose of 500 U per kg body weight. Mock-injected KO and WT mice given the same volume of PBS served as controls. LAMAN activity was determined as previously described (10).

Tissue Collection

Animals were deeply anesthetized and subsequently perfused with 0.1 M phosphate buffer (PB), pH 7.4 followed by 4 % paraformaldehyde in PB (for free floating sections), or 6 % glutaraldehyde supplemented with 1 % procaine (for electron microscopy). Brains were removed and cut sagittally into halves. For free-floating sections, brains were post-fixed in 4% paraformaldehyde overnight at 4°C, and stored in 30% sucrose until sectioning. Samples for electron microscopy were processed as previously described (10).

Western Blot Analysis

For Western blots, brains were removed after PB perfusion and homogenized in 9 volumes (v/w) of ice-cold lysis buffer (10 mM Tris-HCl pH 7.4, 150 mM NaCl, 5 mM EDTA, 1 % Triton X-100 with protease inhibitors). After incubation on ice for 30 minutes followed by sonification, homogenates were cleared by centrifugation at 13,000 rpm for 15 minutes at 4°C. Equal amounts of protein were separated on 12.5 % or 15 % SDS-PAGE gels, blotted onto nitrocellulose membranes and incubated with the respective primary antibodies. Horseradish peroxidase-conjugated secondary antibodies were detected by chemiluminescence (SuperSignalWest, Pierce, Pittsburgh, PA). Actin was used as a loading control.

Immunohistochemistry

For IHC, sagittally cut, frozen brain halves were sectioned on a Leica 9000s microtome into 50-µm-thick free-floating slices. When staining for gangliosides and cholesterol, unfrozen brain halves were cut sagittally into 35-µm-thick free-floating sections using a Leica vibratome. The sections were rinsed in 0.1 M PB and subsequently blocked with 4% normal goat serum in 0.1 M PB with 0.2 % bovine serum albumin and 0.25% Triton X-100 for permeabilization at room temperature (RT) for 2 hours. Primary antibodies were diluted in blocking solution and incubated overnight at 4°C with the sections at gentle agitation. After washing 3× with washing buffer (0.1 M PB with 0.25 % TX-100), appropriate secondary antibodies (fluorophore conjugated or biotinylated) diluted in washing buffer were incubated with the sections for 2 hours at RT. After additional 3 washes with washing buffer, the sections were coverslipped with MOWIOL/DABCO (fluorophore-labelled sections) or processed for 3, 3’diaminobenzidine (DAB)-visualization. DAB-stained sections were mounted, dehydrated and coverslipped with Permount™. Some sections were counterstained with Nissl stain. DAB-stained sections with the anti-GFAP-antibody were detected with Horseradish peroxidase-labelled secondary antibody omitting Biotin/Streptavidin-step.

Histochemistry

For filipin histochemistry of brain sections, mice were perfused with 4% paraformaldehyde in 0.1 M PB and post-fixed in the same fixative overnight followed by storage in 0.1 M PB at 4°C. Filipin histochemistry was performed on 35-µm-thick vibratome sections. All steps were performed at RT. Sections were washed 2 × 10 minutes in PBS and 2 × 10 minutes in 0.2% Saponin/PBS. Slices were incubated in Filipin (0.05 mg/ml) for 20 minutes, washed 2 × 10 minutes in 0.02% Saponin/PBS and 2 × 10 minutes in PBS. Slices were mounted in Prolong antifade mounting solution (Invitrogen).

Isolation of Neutral Oligosaccharides

Neutral oligosaccharides were extracted as described (11). In brief, brains were homogenized in 9 volumes (v/w) HPLC-grade water. After sonification, proteins were precipitated by methanol and extracted by addition of chloroform/water. Supernatants were desalted by incubation with mixed-bed ion-exchange resin (AG 501-X8, Bio-Rad, Hercules, CA) and soluble material was lyophilized.

Thin Layer Chromatography of Neutral Oligosaccharides

Lyophilized oligosaccharides were resuspended in HPLC-grade water and spotted onto Silica gel thin layer chromatography plates (20 × 20 Silica gel F60, Merck, Darmstadt, Germany). Oligosaccharides were separated by n-butanol/acetic acid/water (100:50:50) development overnight followed by n-propanol/nitromethan/water (100:80:60) development for 4 hours. After drying, plates were sprayed with 0.2% (w/v) orcinol-solution (20% H₂SO₄ dissolved in water) and heated at 110°C until dark bands appeared.

Quantitative Gait Analysis

α-Mannosidosis mice 6 and 19 months of age underwent treadmill gait analysis, as described (13). Following a 30-second habituation period, 4 trials of 60-second treadmill walking were conducted, during which mice were filmed by a ventrally placed webcam. These trials were executed at different combinations of speed and slope (16 cm/s and 0°; 16 cm/s and 10°; 22 cm/s and 0°; 22 cm/s and 10°). Mice were encouraged to keep pace with the treadmill using an electric grid placed at the end of the treadmill. The number of required electrical stimulations (errors) was registered during each trial. Several gait variables including base-width (distance between contralateral paws), front/hind distance (distance between front paw print and subsequent hind paw print) and stride length (distance between 2 subsequent prints of the same paw) were extracted from these video data using an automated algorithm (14). Additionally, incongruity coefficients (ICs) for related variables were calculated as a measure of disturbed gait. These coefficients are defined as the absolute value of the difference between the z-scores of 2 related gait variables, for example:

$$\text{IC (FB, HB)}=|\text{Z (FB)}-\text{Z (HB)}|=\mathit{|}\mathit{\text{((FB-FBmean)/FBstdev)}}-\mathit{\text{((HB-HBmean)/HBstdev)}}\mathit{|}$$

With IC = incongruity coefficient; FB = front base (distance between front paws); HB = hind base (distance between hind paws); stdev = standard deviation. Dissimilar deviations from the group mean result in a higher IC based on these variables and is a sign of ill-balanced or uncoordinated gait.

RESULTS

Regional Differences in Carbohydrate Storage in Brains of α-Mannosidosis Mice

Neutral oligosaccharides were extracted from different brain regions including cortex, cerebellum, brainstem and midbrain of α-mannosidosis mice and separated by thin layer chromatography (Fig. 1). Whereas mannosyl-linked oligosaccharides were present in all brain regions, remarkably high levels of stored sugars were detected in the cerebellum of α-mannosidosis mice.

Figure 1.

Thin layer chromatography of extracted oligosaccharides from whole brain of wild-type and α-mannosidosis (KO) mice and of different brain regions of 2 KO mice. The cerebellum shows the highest amount of storage material. Oligosaccharides with one N-acetylglucosamine and 2 – 9 mannose residues are labelled as M2 – M9. Arrows mark orcinol-positive bands insensitive to lysosomal hydrolase α-mannosidase.

Storage of Gangliosides, Cholesterol and Autofluorescent Material in the Cerebellum of α-Mannosidosis Mice

Based on our thin layer chromatography analyses, the cerebellum is the site of highest primary storage. To investigate whether primary lysosomal storage secondarily affects the distribution of gangliosides as in many other LSDs (15), we performed IHC on brain sections of 12-month-old KO and WT mice using antibodies specific for GM1, GM2 and GM3 gangliosides. The secondary accumulation of gangliosides in the brain is a common hallmark of LSDs (15). Among the tested gangliosides, only GM1 was elevated in the KO brains (Fig. 2B, D–F). In WT, GM1 was barely detectable (Fig. 2A, C), but pyramidal neurons of the hippocampus of KO mice displayed slightly increased levels of GM2 and GM3 (data not shown). Vesicular GM1 accumulation was apparent throughout the brain, including the cerebellum (Fig. 2A–D, F), cerebral cortex and midbrain (Fig. 2E). In the cerebellum, GM1 accumulated exclusively within the molecular layer where it was found within the dendritic tree of Purkinje cells and in vesicular-like structures within the cytoplasm of glial and neuronal cells (Fig. 2D, F).

Figure 2.

(A–F) Vibratome sections of formalin-fixed brains from 11-month-old wild-type (WT; A, C) and α-mannosidosis (KO) (B, D–F) mice immunostained for GM1-gangliosides (brown) with Nissl counterstain (nuclei, blue). GM1-positive vesicles accumulate in the molecular layer (ml) of the KO mouse cerebellum (B, D, F; black arrows), and in neurons within the midbrain (mb, E). No vesicular GM1 accumulation was found in the cerebellar granular layer (gl) or white matter (wm). (G–J) Filipin-positive free cholesterol accumulates in the KO mouse brain specifically in the molecular layer of the cerebellum (H–J), but not in WT brain (G, I). Scale bars: A–B, G–H, 100 µm; C–F, I–J, 50 µm.

Because ganglioside storage correlates at least partially with cholesterol storage (16), we used the fluorescent dye filipin to stain free cholesterol in the KO and WT tissue (Fig. 2G–J). In KO (Fig. 2 H, J), but not in WT (Fig. 2 G, I) brains, cholesterol accumulated specifically in the cerebellar molecular layer in a pattern comparable to that of GM1. However, other brain regions with elevated GM1 levels lack stored cholesterol. IHC on brain sections of 6-month-old mice revealed less prominent storage of both lipids, indicating their progressive accumulation (not shown). Consistent with these observations, subcellular fractionation of whole brain homogenates showed a significantly decreased density of lysosomal compartments in the KO brains, suggesting an accumulation of metabolites with low density like lipids within α-mannosidosis lysosomes (Supplemental Fig. 1).

Autofluorescence

Unstained vibratome brain sections of KO and WT brains at 12 months were used to study their autofluorescent properties following excitation of different wavelengths (UV light, 350nm, 488 nm and 633 nm) (Fig. 3B, D). Autofluorescent material was detected in Purkinje cells within the molecular layer and the deep nuclei of the cerebellum in KO mice. Autofluorescence was also visible in Purkinje cells of WT mice, a finding consistent with the reported accumulation of lipofuscin in aged Purkinje cells (Fig. 3B, upper panel) (17). Autofluorescence was detected in all excitation wavelengths with highest emission after UV-excitation at 350 nm. Cholesterol was present within most of these autofluorescent structures as indicated by co-staining with Filipin (Fig. 3B, lower panel).

Figure 3.

Macrophage/microglia in α-mannosidosis (KO) mice. (A, B, D) Vibratome sections of formalin fixed brains from 12-month-old wild-type and KO mice were analyzed for activation of macrophages by either 3, 3’ diaminobenzidine (DAB) immunohistochemistry (A) or immunofluorescence (D) using anti-CD68 (A) or F4/80 (D) antibodies. Autofluorescence images (B, D) were taken with excitation wavelengths of 488 nm, 633 nm and UV light. In the KO brain anti-CD68 strongly labels activated microglia that are found predominantly in the cerebellar molecular layer (ml) where there is abundant storage of autofluorescent material (B, upper panel) (pcl = Purkinje cell layer; gl = granular layer). Cholesterol-containing autofluorescence material demonstrated by Filipin staining (B, lower panel in red) is taken up by F4/80-positive macrophages (D, green). (C) Electron micrographs from macrophages in the cerebellar molecular layer of a KO mouse display cytoplasmic (left) multilamellar (*) and (right) lipofuscin like (❋) material. Scale bars: A (upper panel): 400 µm, (lower panel) 50 µm; B (upper panel): 500 µm, (lower panel) 15 µm; D: 50 µm.

Activation of Macrophages and Astrogliosis

Activation of CNS resident macrophages and astrocytes is characteristic of LSD neuropathology. To investigate whether microglial activation also occurs in α-mannosidosis, the expression of CD68/macrosialin (18) (Fig. 3A) and F4/80 (Fig. 3D) was studied by IHC of WT and KO brains. The KO brains displayed numerous CD68-positive macrophages with plump, amoeboid shapes indicative of activation mainly in the cerebellar molecular layer whereas WT brains showed ramified microglial cells with fine processes characteristic of a non-activated state. Activated microglia were often found in a cluster-like assembly and gradually increased in number from anterior to posterior lobes (Fig. 3A). The granular cell layer, white matter tracts and deep cerebellar nuclei of KO cerebellum were mostly devoid of activated microglia. Intracellular autofluorescence was visible in F4/80-positive macrophages, suggesting phagocytic uptake of autofluorescent storage material (Fig. 3D). Electron micrographs of macrophages in the cerebellar molecular layer of KO mice displayed cytoplasmic multilamellar and lipofuscin like material indicative of lipid storage (Fig. 3C).

Because microglial activation is often accompanied by reactive astrogliosis and demyelination, we performed immunofluorescence and Western blot analyses of GFAP and MBP, which are specific for astrocytes and myelin, respectively. As in the cerebellum, WT and KO brains showed comparable MBP distribution (Fig. 4A, upper panel) and expression levels (Fig. 4C), suggesting normal myelination the KO mice. Myelin staining with Luxol fast blue also showed no differences in the overall appearance of cerebellar myelin between the 2 genotypes (not shown). In contrast, GFAP was slightly upregulated in the KO mice (Fig. 4A, lower panel) that was confirmed by Western blot analysis (Fig. 4C), suggesting an activation of astrocytes upon disruption of LAMAN activity. Upregulation was also evident for the lysosome-associated membrane proteins Lamp1 and Lamp2 (Fig. 4C).

Figure 4.

Myelin and astrocytes in the cerebellum of α-mannosidosis (KO) and wild type (WT) mice. (A, B, D–F) Vibratome sections of formalin-fixed brains from 11- to 14-month-old WT and KO mice were stained by immunofluorescence (A, D) and 3, 3’ diaminobenzidine (DAB) immunohistochemistry (B, E, F) for myelin basic protein (MBP) (A; upper panel, green; nuclein in blue) and the astrocytic marker glial fibrillary acidic protein (GFAP) (green, lower panel). MBP expression is unchanged whereas GFAP is slightly increased in the KO cerebellum. In the KO cerebellum, there is reactive Bergmann astrogliosis (B) (arrowhead). (C) Western blots from total KO brain lysates (n = 3 for each genotype) showed upregulation of GFAP and of the lysosomal membrane proteins Lamp1/2. Actin served as a loading control. (D) To assess the correlation between Purkinje cell loss (*) and reactive astrogliosis, immunofluorescence was performed with antibodies to GFAP (red) and Calbindin (green). Autofluorescent material (predominantly in microglia) was detected under UV excitation (blue, arrows). Merge shows astrogliosis in areas devoid of Purkinje cells. (E, F) DAB staining reveals minor loss of Purkinje cells in α-mannosidosis KO mice (lower panel) compared to wild-type mice (upper panel) (wm = white matter; gl = granular layer; ml = granular layer). Scale bars: A: 200 µm; B: 100 µm, D–F, 50 µm

Regional Restricted Astrogliosis and Partial Loss of Purkinje Cells

To study reactive astrocytosis in more detail, we performed GFAP IHC that revealed a regional restricted astrogliosis (arrowhead) of Bergmann glia within the molecular and the Purkinje cell layer of the cerebellum of KO mice (Fig. 4B). To determine whether the regionally restricted astrogliosis was associated with changes within the Purkinje cell layer as in other murine LSD models (19–21), sections of KO and WT mice were co-stained for calbindin, (a protein dominantly expressed in the soma and dendritic tree of Purkinje cells [22]) and GFAP (Fig. 4D). The reactive astrogliosis occurred predominantly within regions devoid of Purkinje cells, as indicated by absence of their dendritic trees. Immunolabeling of calbindin revealed a comparable distribution between the Purkinje cell layers of KO and WT mice (Fig. 4E), but KO cerebellum displayed scattered regions lacking Purkinje soma and dendritic trees (Fig. 4F), indicating partial loss of this cell type in the KO mice. Surviving Purkinje cells displayed spheroids that were close to Purkinje cell bodies (Supplemental Fig. 2).

Quantitative Gait Analysis

At the age of 6 months there was no difference between the WT and KO mice in any of the gait parameters (Table). However, at age 19 months KO mice showed an increased incongruity coefficient (IC) for left and right stride lengths (p < 0.05). Left and right stride lengths were closely related in WT mice but there was much more incongruity in the KO mice. Scatter plot and respective regression lines illustrate this difference (Fig. 5). A similar trend occurred in the relationship between front strides and hind strides (Table). These results indicate that aging KO mice show increasingly inharmonious gait dynamics and uncoordinated gait; they also displayed a higher number of errors in their treadmill walking at both ages (data not shown), as described previously (10).

Table.

Gait Parameter Values for Wild-Type and α-Mannosidosis (Knockout) Mice at ages 6 and 19 Months

	6 months		19 months
	WT	KO	WT	KO
	(n = 20)	(n = 17)	(n = 10)	(n = 11)
FB (cm)	1.57	1.56	1.67	1.76
HB (cm)	2.56	2.60	2.73	2.76
LFHD (cm)	1.36	1.46	1.64	1.56
RFHD (cm)	1.36	1.45	1.63	1.31
LFS (cm)	4.87	4.65	4.69	4.30
RFS (cm)	4.77	4.68	4.71	4.63
LHS (cm)	4.85	4.83	4.70	4.67
RHS (cm)	4.96	4.76	4.61	4.79
IC (FB-HB)	0.88	0.77	0.55	0.80 (p = 0.16)
IC (LFHD-RFHD)	0.75	0.69	0.61	0.58
IC (LS-RS)	0.27	0.31	0.21	0.92* (p < 0.05)
IC (FS-HS)	0.21	0.22	0.28	0.90 (p = 0.06)

Gait variables including base-width (distance between contralateral paws), and stride length (distance between 2 subsequent prints of the same paw) were extracted from video data using an automated algorithm (14).

Abbreviations: WT, wild type, KO, knockout, FB = Front Base (distance between front paws); HB = Hind Base (distance between hind paws; LFHD = left front/hind distance (distance between front paw print and subsequent hind paw print); RFHD = right front/hind distance; lfs = left front limb stride; rfs = right front limb stride; lhs = left hind limb stride; rhs = right hind limb stride; IC = incongruity coefficient; ls, left strides; rs, right strides; fs, front strides; hs, hind strides.

Figure 5.

Scatter plot and regression lines for left vs. right stride lengths. Wild-type (black dots) mice show strongly related values of left and right strides, all data points closely approaching the regression line. α-mannosidosis (KO) mice (white dots) show more divergence of left and right strides, and the data points are more widespread.

Enzyme Replacement Therapy

Short-term, high-dose treatment with recombinant human LAMAN resulted in partial correction of the primary storage of oligosaccharides in different brain regions (10). Here, we investigated the effect of this ERT regimen on the cerebellar phenotype. Detailed examination of the cerebellum by thin layer chromatography did not reveal any effects on oligomannose storage after 4 injections with 500 U per kg body weight LAMAN, although high amounts of LAMAN activity were present in this brain region (Supplemental Fig. 3). No obvious alterations in the size or number of storage vacuoles were observed for the Purkinje cells (Supplemental Fig. 3B). To evaluate changes in microglia activation, astrogliosis and secondary storage of lipids in response to the ERT, IHC and Western blots were performed.

Untreated KO mice displayed microglia of the amoeboid, phagocytic shape, whereas in the brains of ERT-treated mice, CD68-positive microglia were reduced slightly in number and had less of the amoeboid shape but with fine processes (Fig. 6A, first panel). Lamp1 immunoreactivity was markedly reduced in all LAMAN-treated animals (Fig. 6B, upper panel) and large Lamp1-positive inclusions in the molecular layer, probably located in microglial cells, disappeared almost completely (Fig. 6A, second panel). After ERT, hypertrophy of Bergmann glia with processes ranging deep into the molecular layer were at least partially normalized to the WT level, as demonstrated by GFAP immunostaining (Fig. 6A, third panel).

Figure 6.

Effects of enzyme replacement therapy (ERT) in α-mannosidase (KO) mice. (A) Vibratome sections of cerebella from wild-type (WT) (left column), KO (middle column) and ERT-treated KO mice (right column) were stained for CD68, Lamp1 and glial fibrillary acidic protein (GFAP). For detection of cholesterol, sections were stained with Filipin and evaluated by fluorescence (lowermost panels). Lamp1 immunoreactivity and astrogliosis of Bergmann glia were nearly comparable to those in WT mice after ERT in KO mice; slight effects of ERT were observed on microglia (CD68). No differences of cholesterol accumulation were observed after ERT. (B) Western blots of whole brain lysates (WT, KO and ERT treated KO; n = 2 each genotype), showed almost complete normalization of Lamp1-specific bands after ERT and only faint effects on the amount of the cholesterol binding protein NPC2. No effects on the molecular weight differences between WT and KO mice were detected after ERT. Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) served as loading control. Scale bars: A: 200 µm; B: 100 µm.

Finally, we evaluated the ERT effects on the secondary storage of lipids in the cerebellum and on NPC2, a soluble lysosomal protein mediating the export of free cholesterol from lysosomes. NPC2 was recently shown to be hyperglycosylated in KO mice due to impaired processing of its glycan moiety by missing LAMAN activity (23). Filipin labelling of free cholesterol did not change after high dose ERT regimen (Fig. 6A, lowermost panel) and was accompanied by persistent storage of GM1 gangliosides (data not shown). The hyperglycosylation of NPC2 also did not change in KO mice after ERT (Fig. 6B, lowermost panel), suggesting insufficient amounts of active LAMAN for glycan trimming.

DISCUSSION

Understanding the neuropathological characteristics of α-mannosidosis will be important for future ERT trials in humans. The present study characterizes cerebellar pathology as a future outcome measure to assess therapeutic efficacy. Neuropathological changes occurred especially in the molecular layer of the cerebellum, the location of highest oligosaccharide accumulation in the brain. Progressive gait disturbances in aged KO mice indicate that gait impairments are prominent features in the mouse model, as well as in human patients. Pathological changes further included secondary accumulation of GM1 ganglioside, free cholesterol and lipophilic autofluorescent material that was partially engulfed by infiltrated CD68/ F4/80-positive macrophages. Regionally restricted gliosis of Bergman glia was found in regions that were devoid of Purkinje cells, and surviving Purkinje cells showed dendritic swelling and axonal spheroids.

Pathogenesis of α-Mannosidosis

Although no detailed neuropathological reports are available for cases of human of α-mannosidosis, different animal models have been described to examine the pathogenesis of the disease (7, 24–26). Previous studies focussed on the hippocampus (10) and did not reveal significant activation of microglia and astroglia; however, the investigated cohort of mice was younger (>3 months) than those in the present study.

The clinical disease course described here is more subtle than that observed in cattle, cats and guinea pigs with prominent cerebellar atrophy and progressive motor dysfunctions (27, 28). The disease in cats is most prominent and leads to premature death (29). Despite the lack of demyelination in the mouse model and different patterns of brain ganglioside and cholesterol accumulation, overall histopathological courses of the disease are similar in all animal models. The comparison of the clinical and histopathological course of the disease across different animal models suggests that loss of Purkinje cells contributes to the severity of the disease. Accordingly, α-mannosidosis cats show the most rapid loss of this cell type (30), whereas later onset loss is obvious in guinea pigs (27), and there is only mild Purkinje cell death in mice. Thus, factors other than storage in the different species must be crucial for the survival of affected cells.

A possible explanation for this obvious discrepancy could be the difference in lysosomal oligosaccharide degradation between cats and cattle on one hand, and rodents and humans on the other. In contrast to cattle and cats, rodents and humans express a lysosomal chitobiose and a second lysosomal α-mannosidase (31) that is specific for cleavage of the α-1,6-mannosidic core of glycoproteins (3, 32). These differences in enzyme expression lead to considerably different primary storage products (33, 34), and possibly to a different pathology between species.

Lipid Storage

Ganglioside and cholesterol storage are prominent features in LSDs with neurological involvement. They have been described in LSDs with primary defects in lipid-degradation, gangliosidoses and Niemann-Pick type C, but also in LSDs unrelated to lipid-degradation, such as mucopolysaccharidoses and glycoproteinoses (15). Secondary storage of gangliosides and cholesterol is well documented in the animal models of α-mannosidosis (27, 35, 36). Interestingly, α-mannosidosis cats, guinea pigs and mice show a different pattern of accumulated gangliosides and cholesterol. In cats, GM2 storage predominates over GM3 whereas guinea pigs store mainly GM3. In both animal models, GM1 storage has not been reported and cholesterol accumulation appears after that of ganglioside storage (27). In contrast, KO mouse brains display prominent storage of GM1 but not of GM2 and GM3. Cholesterol was only found to be accumulated in the molecular layer of the cerebellum, where it coincided with GM1 storage. These data indicate that in mice, cholesterol storage is not secondary to that of GM2 and/or GM3 gangliosides as suggested for the cat and guinea pig models, but rather depends on GM1. Further developmental analyses have to clarify whether GM1 storage in the cerebellum is secondary to cholesterol or vice versa. Recent data suggest that LAMAN is involved in the trimming of lysosomal glycoproteins including the cholesterol binding enzyme NPC2 (23) and in mouse tissues deficient for α-mannosidase, NPC2 is hyperglycosylated. Increased glycosylation of NPC2 slows the cholesterol transfer rate of the protein (37), suggesting that this effect could alter the cholesterol and ganglioside metabolism in LAMAN deficient tissues. So far, NPC2 hyperglycosylation has only been shown in the mouse model of α-mannosidosis and further studies have to clarify how this protein contributes to the disturbed lipid metabolism in the other animal models for α-mannosidosis.

Astrogliosis and Inflammation

CNS inflammation (i.e. activation of astrocytes and microglia) is a common feature of neurodegenerative diseases including LSDs (38–41). Recent publications describe cerebellar specific pathological alterations in patients and in different mouse models of neuronal ceroid lipofuscinosis (19–21), including localized astrogliosis of Bergmann glia. Patches of hypertrophic Bergman glia were found in close proximity to degenerated Purkinje cells or breaches within the Purkinje layer (19, 20), alterations that resemble those found in the present study. These results support the idea that impaired function and/or degeneration of Purkinje cells leads to a restricted astrogliosis of the functional closely linked Bergmann glia. However, we cannot completely rule out the possibility of deleterious effects of the activated Bergmann glia on the surrounding Purkinje cells, leading to their impaired function. In this regard it should be noted that mouse models for Niemann-Pick C disease that are deficient for either NPC1 or NPC2, show a similar distribution of accumulated cholesterol within the cerebellum although the NPC proteins are differentially expressed in Bergmann glia and Purkinje cells (12), respectively, highlighting the crosstalk between both cell types.

Microglia activation is well documented for several LSD mouse models and patients (40–43), but the mechanism of induction of this inflammatory response is unclear. In principle, self-activation of microglia due to the accumulation of primary or secondary storage material as well as exogenous stimuli of inflammation and/or phagocytosis is possible. Macrophages are highly active in the catabolism of mannose-containing glycoproteins and, therefore, are the most prominently affected cell type in α-mannosidosis mice, at least in peripheral tissues (7). The presence of undegraded, autofluorescent material within brain microglia argues for only partial clearance of phagocytosed lipophilic material due to defects in lysosomal breakdown. Therefore, it is likely that the prominent accumulation of undegraded substrates within CNS resident macrophages triggers their self-activation. However, the focal distribution of highly activated microglia / phagocytes within the molecular layer of the KO cerebellum rather suggests an exogenous activation of microglia rather than self-activation. In mouse models of GM1 and GM2 gangliosidoses, CNS inflammation is prominent and the reduction of these stored gangliosides decreased microglia activation in vivo (40), whereas mixed brain gangliosides and GM1 alone can induce microglia ramification in vitro (44), indicating that the amount of stored gangliosides in the brain is crucial for microglia activation possibly via toll-like receptors (45, 46). We therefore speculate that in α-mannosidosis mice, GM1 storage is the main factor stimulating microglia activation but our data suggest that both exogenous stimuli as well as endogenous storage likely contribute to the inflammatory responses.

Gait Analysis

In accordance with our previous observations, the youngest group of KO mice displayed more walking errors on the treadmill device (10). Previously employed methods and measures failed to demonstrate neuromotor dysfunction in these mice (8, 9). Here also, 6-month-old KO mice did not show alterations in any of the measured gait variables or in their interrelationships. This could mean that the observed treadmill errors might not be attributed to uncoordinated gait performance proper, but rather to deficits in gross ambulatory ability, tardiness, or procedural learning defects. Treadmill performance is typically impaired in cerebellum-lesioned mice (47, 48). The observed defect in the youngest mice could represent a decrease in cerebellum-dependent gait adaptability, since treadmill walking mechanics substantially differ from normal over ground locomotion in rodents (49, 50). Blanz et al reported amelioration of treadmill performance in LAMAN KO mice after high-dose ERT in the absence of effects on Purkinje and granule cells, which might to indicate that reduction of cerebral storage, reinstates compensatory neural mechanisms (10).

Gait and coordinated movements obviously further deteriorated as the mice grew older and aged KO mice showed reduced congruence of stride lengths. Significant incongruity occurred in contralateral stride lengths, whereas concomitant tendencies appeared in front-versus-hind stride lengths. Divergence of normally concurrent gait variables was also reported in another LSD murine model (51). Notably, mice with electrolytic lesions of the cerebellum show a similar gait profile (unpublished observations during validation of the treadmill device), and α-mannosidosis guinea pigs show gait changes that might be the consequence of cerebellar dysfunction and spinal cord pathology (52). The presently described progressive degeneration of Purkinje cells following metabolic perturbation and lysosomal storage may underlie the progressive decline of coordinated movements, neuromotor adaptability and gait in KO mice.

ERT

It has previously been shown that short-term ERT with high doses of recombinant LAMAN can reduce storage and reverse the ataxic phenotype of α-mannosidosis mice (10). We used the same treatment regimen to investigate this therapeutic effect on the cerebellar neuropathology. Even though the cerebellum showed highest uptake of recombinant enzyme, storage reduction in the cerebellum was not obvious with storage vacuoles persisting in Purkinje cells. Despite the limited effect on cerebellar neurons, Lamp1 immunoreactivity (most probably located to microglia) was markedly reduced and anti-CD68 and -GFAP staining revealed a partial effect on microglia and Bergmann glia activation, respectively. These data suggest an uptake of recombinant LAMAN either into glial cells or a decrease in exogenously induced activation. Analyses of the glycan structures of recombinant LAMAN revealed a low degree of phosphorylation coinciding with cellular LAMAN uptake that has been shown to be at least partially independent of mannose 6-phosphate (10). Because glia express several carbohydrate specific receptors, we speculate that some of the recombinant enzyme was taken up preferentially by these cells leading to a partial decrease in Lamp1 expression and glial activation. The precise mechanisms of LAMAN uptake into neural cells has to be investigated in detail and will be object of future research. In contrast to the observed normalization of Lamp1 and the slight reduction of glial activation, cholesterol and GM1 accumulation and NPC2 glycosylation did not respond to ERT, suggesting that the short duration of the treatment may limit the potential effect of ERT on neuronal storage reduction. The effect of long-term ERT on neuronal storage could not be assessed since humoral immune responses following frequent injections of the recombinant enzyme precluded longer ERT treatment in the classical α-mannosidosis KO mouse model. To overcome these limitations, we are currently generating an immune-tolerant mouse model. Induction of immune-tolerance by transgenic approaches has been shown to be a valuable tool for long-term enzyme treatment in other LSDs (53, 54).

In summary, cerebellar alterations including loss of Purkinje cells, inflammation and motor deficits are common to all α-mannosidosis animal models and relate to the observed pathology in human patients. The observed cerebellar alterations in α-mannosidosis mice may underlie the progressive motor deficits that occur in these mice as they age. Short-term high-dose ERT effectively corrected some aspects of cerebellar pathology in α-mannosidosis mice including macrophage activation, astrogliosis and Lamp1 upregulation, but only long-term studies will clarify whether ERT has the potential to ameliorate completely the CNS pathology. The observed cerebellar pathology and ataxia characterized herein may be used in forthcoming ERT studies as possible therapeutic outcome measures.