Effect of high dose, repeated intra-cerebrospinal fluid injection of sulphamidase on neuropathology in mucopolysaccharidosis type IIIA mice


  • K. M. Hemsley,

    Corresponding author
    1. Lysosomal Diseases Research Unit, Department of Genetic Medicine, Children, Youth and Women’s Health Service
    2. Department of Paediatrics, University of Adelaide, North Adelaide, South Australia, Australia
      *K. M. Hemsley, Lysosomal Diseases Research Unit, Department of Genetic Medicine, Children, Youth and Women’s Health Service, 72 King William Road, North Adelaide, SA 5006, Australia. E-mail: kim.hemsley@adelaide.edu.au
    Search for more papers by this author
  • H. Beard,

    1. Lysosomal Diseases Research Unit, Department of Genetic Medicine, Children, Youth and Women’s Health Service
    Search for more papers by this author
  • B. M. King,

    1. Lysosomal Diseases Research Unit, Department of Genetic Medicine, Children, Youth and Women’s Health Service
    Search for more papers by this author
  • J. J. Hopwood

    1. Lysosomal Diseases Research Unit, Department of Genetic Medicine, Children, Youth and Women’s Health Service
    2. Department of Paediatrics, University of Adelaide, North Adelaide, South Australia, Australia
    Search for more papers by this author

*K. M. Hemsley, Lysosomal Diseases Research Unit, Department of Genetic Medicine, Children, Youth and Women’s Health Service, 72 King William Road, North Adelaide, SA 5006, Australia. E-mail: kim.hemsley@adelaide.edu.au


Mucopolysaccharidosis type IIIA (MPS IIIA) is an inherited neurodegenerative lysosomal storage disorder characterized by progressive loss of learned skills, sleep disturbance and behavioural problems. Reduced activity of sulphamidase (N-sulphoglucosamine sulphohydrolase; SGSH; EC results in intracellular accumulation of heparan sulphate (HS), with the brain as the primary site of pathology. We have used a naturally occurring MPS IIIA mouse model to determine the effectiveness of SGSH replacement through the cerebrospinal fluid (CSF) to decrease neuropathology. This is a potential therapeutic option for patients with this disorder. Mice received intra-CSF injections of recombinant human SGSH (30, 50 or 70 μg) fortnightly from 6 to 18 weeks of age, and the cumulative effect on neuropathology was examined and quantified. Anti-SGSH antibodies detected in plasma at euthanasia did not appear to impact upon the health of the mice or the experimental outcome, with significant but region-dependent and dose-dependent reductions in an HS-derived oligosaccharide observed in the brain and spinal cord using tandem mass spectrometry. SGSH infusion reduced the number of storage inclusions observed in the brain when visualized using electron microscopy, and this correlated with a significant decrease in the immunohistochemical staining of a lysosomal membrane marker. Reduced numbers of activated isolectin B4-positive microglia and glial fibrillary acidic protein-positive astrocytes were seen in many, but not all, brain regions. Significant reductions in the number of ubiquitin-positive intracellular inclusions were also observed. These outcomes show the effectiveness of this method of enzyme delivery in reducing the spectrum of neuropathological changes in murine MPS IIIA brain.

Mucopolysaccharidosis type IIIA (MPS IIIA) or Sanfilippo syndrome is a lysosomal storage disorder (LSD) characterized by severe central nervous system (CNS) degeneration and is the most common of four MPS III subtypes (A, B, C and D). Each results from a deficiency of a different lysosomal enzyme involved in heparan sulphate (HS) degradation and is associated with accumulation of incompletely degraded HS in the lysosome/late endosomal compartment, its elevation in urine and clinical disease. MPS IIIA results from a mutation in the gene encoding sulphamidase (N-sulphoglucosamine sulphohydrolase; SGSH; EC, a sulphatase that cleaves glucosamine-N-sulphate bonds at non-reducing ends of HS fragments (Hopwood & Morris 1990; Scott et al. 1995). Intracellular inclusions are observed in the brain as early as the second trimester (Ceuterick et al. 1980). Heparan sulphate accumulation is accompanied, for unknown reasons, by lipids including GM2 and GM3 gangliosides and cholesterol (McGlynn et al. 2004).

Other neuropathological changes are also evident, including deposition of protein aggregates containing phosphorylated α-synuclein, ubiquitin and amyloid (Ginsberg et al. 1999; Hamano et al. 2008; Savas et al. 2004). Astrogliosis has been shown in MPS IIIB patients (Hamano et al. 2008; Tamagawa et al. 1985) and MPS IIIA (Savas et al. 2004) and MPS IIIB (Li et al. 2002) mice. Microgliosis (Ohmi et al. 2003) is accompanied by alterations in oxidative stress markers (Hamano et al. 2008; Villani et al. 2007) and autophagy appears to be defective (Elleder et al. 1997; Ryazantsev et al. 2007; Settembre et al. 2008). Specific cell death is observed and presumably underlies the atrophy described in post-mortem studies (Hamano et al. 2008; Heldermon et al. 2007; Zafeiriou et al. 2001).

Clinical manifestation of disease generally occurs in early childhood, although diagnoses have been made in adults (Van Hove et al. 2003). Patients exhibit progressive mental retardation, hyperactivity, aggressive behaviour, insomnia, apnoea, hepatosplenomegaly and mild dysostosis multiplex (Neufeld & Muenzer 2001). At present, there is no treatment and patients progressively decline and die. There is no evidence to suggest that bone marrow or haematopoietic stem cell transplantation is clinically effective, although it does improve the quality of life in some other CNS-affecting LSD (Hopwood et al. 1993; Staba et al. 2004).

We are utilizing a naturally occurring mouse model of MPS IIIA that exhibits many of the biochemical, pathological and clinical features seen in humans (Bhaumik et al. 1999; Crawley et al. 2006; Hemsley & Hopwood 2005; Savas et al. 2004). We and others have shown that intra-cerebrospinal fluid (CSF) delivery of recombinant human lysosomal enzyme is an effective means of treating CNS disease (Dickson et al. 2007; Hemsley et al. 2007; Kakkis et al. 2004). However, examination of the effectiveness of this treatment has been limited, until now, to determining the impact on primary lysosomal storage. This study advances our understanding of the disease course and the utility of this treatment in reducing or delaying a variety of disease parameters.

Materials and methods


Male, congenic C57BL/6J MPS IIIA mice [or unaffected (−/+ or +/+) littermates] were used in this study. Ten generations of backcrossing were undertaken to establish the strain (Crawley et al. 2006). The mice were genotyped using established methods (Bhaumik et al. 1999) and were bred, housed and maintained in the Children, Youth and Women’s Health Service (CYWHS) Animal House, with all breeding and experimental procedures undertaken with the approval of the CYWHS Animal Ethics Committee (Adelaide, South Australia) with regard to the guidelines of the National Health and Medical Research Council of Australia on the use and care of experimental animals.

Surgical methods

Recombinant human SGSH (rhSGSH) was manufactured by Shire Human Genetic Therapies as previously described (Hemsley et al. 2007). Enzyme was utilized at 21 μg/μl as determined by a bicinchoninic acid (BCA) protein assay kit (Pierce 23235; Quantum Scientific, Murarrie, Australia). Mice were pretreated with Robinul (0.01 mg/mouse; Wyeth Ayerst, Baulkham Hills, Australia) and anaesthetized with ketamine (87 mg/kg; Parnell Laboratories, Mascot, Australia)/xylazine (13 mg/kg; Troy Laboratories, Smithfield, Australia) (intraperitoneal). Recombinant human SGSH or vehicle (50 mm sodium acetate and 100 mm NaCl, pH 5.0) was injected into the cerebellomedullary cistern using a 27-G dental needle attached to an infusion pump through plastic tubing. Groups of 10–15 MPS IIIA mice received one of three rhSGSH doses (30-, 50- and 70-μg rhSGSH per mouse) or vehicle fortnightly from 6 to 18 weeks of age; unaffected mice received vehicle only. All mice received warmed dextrose (4%) in NaCl (0.18%) following surgery (3 ml/100 g subcutaneous). Acetaminophen (0.16 mg/ml; GlaxoSmithKline, Boronia, Australia) was supplied in drinking water for 3 days before surgery, the day of surgery and 3 days post-injection (Messier et al. 1999).

Necropsy and sample collection

At 21 weeks of age (3 weeks after the last intra-CSF injection), mice were euthanased by CO2 asphyxiation and blood samples were taken through cardiac puncture. Some mice in each group were then fixation-perfused for light microscopy examination [4% paraformaldehyde in phosphate-buffered saline (PBS), pH 7.4] or electron microscopy studies (2% glutaraldehyde/4% paraformaldehyde in 0.1 m sodium phosphate buffer pH 7.2), and the brain/spinal cord was harvested for histological assessment. The tissues for biochemistry were obtained from PBS-perfused mice (pH 7.4), with samples taken from the cervical (cerebellomedullary cistem; CM), thoracic and lumbar levels of the spinal cord. In addition, five 2 mm hemicoronal brain slices were collected using a brain blocker (Braintree Scientific, Braintree, MA, USA). These samples were frozen for subsequent measurement of an HS-derived oligosaccharide.

Tandem mass spectrometric analysis of HS-derived glucosamine-N-sulphate [α-1,4] hexuronic acid accumulation

The relative amount of glucosamine-N-sulphate [α-1,4] hexuronic acid (HNS-UA) in MPS IIIA or unaffected mouse tissues was measured by tandem mass spectrometry as previously described (King et al. 2006). Brain and spinal cord tissues were homogenized in 0.02 m Tris/0.5 m NaCl, pH 7.4, freeze/thawed and the supernatant was collected. Samples were derivatized with 1-phenyl-3-methyl-5-pyrazolone (Sigma-Aldrich, Castle Hill, Australia) in the presence of chondroitin sulphate disaccharide (Sigma-Aldrich) as an internal standard. Chloroform-extracted reaction mixtures were applied to C18 solid phase extraction cartridges (UCT, Bristol, PA, USA), desalted and eluted with 50% CH3CN/0.025% formic acid. Samples were injected into a PE Sciex API 3000 mass spectrometer and oligosaccharides were measured through multiple-reaction monitoring in negative ion mode. The relative amount of oligosaccharide was determined by relating the peak height to that of the internal standard (Fuller et al. 2004). The protein content of supernatants was determined with a micro-BCA protein assay kit (Progen Biosciences). Tissues were processed and analysed in one batch to avoid batch–batch variation. The intra-assay coefficient of variation in the assay was determined using 10 replicates of one homogenized MPS IIIA mouse brain. The brain sample was homogenized in the same manner as the test samples, aliquoted and stored frozen. Aliquots of the homogenate (containing protein in the same concentration range as test samples) were derivatized and analysed simultaneously to obtain an intra-assay variation of 15.8%.

Electron microscopy

Fixed samples of cerebral cortex and cerebellum were processed into resin as previously described (Savas et al. 2004). Briefly, samples were orientated and sliced to 1 mm thick or less, post-fixed in 1% osmium tetroxide, dehydrated and embedded in resin. About 1 μm thick sections were stained with toluidine blue and then 80 nm ultra-thin sections were stained with 2% uranyl acetate/1% lead citrate and examined with a Phillips CM100 transmission electron microscope. Ultra-thin electron microscopy (EM) sections were examined firstly at low power and then at a higher power to identify intracellular organelles. A qualitative description was made of the contents of intracellular inclusions, i.e. amorphous/granular or lamellated/membranous structures, and the abundance of stored material. Representative photos were taken of various cell types.

Histochemistry, immunohistochemistry and quantification of neuropathology using light microscopy

Reagents for staining

Rabbit polyclonal antibodies against ubiquitin (Z458) and glial fibrillary acidic protein [(GFAP), Z334] were purchased from Dako (Glostrup, Denmark). A monoclonal anti-lysosomal integral membrane protein (LIMP-II) antibody (Mimotopes, Clayton, Australia) raised against a synthetic peptide (Mimotopes, Australia), mapping to the C-terminal region of human LIMP-II, was generously provided by Dr E. Parkinson-Lawrence (Lysosomal Diseases Research Unit, Adelaide, South Australia). Peroxidase-conjugated lectin from Bandeiraea (Griffonia) simplicifolia (isolectin B4, BSI-B4, L5391; Sigma, St Louis, MO, USA), which recognizes α-galactosyl groups, was used as a histochemical marker for activated microglia. Biotinylated secondary antibodies, donkey anti-rabbit immunoglobulin G (IgG) and donkey anti-mouse IgG were purchased from Chemicon International (Millipore, MA, USA).


All procedures and post-staining image analyses were undertaken by an experimenter blinded to genotype/treatment status. Six-micron-thick sagittal sections of fixed paraffin-embedded brain tissue were cut on a rotary microtome (Leica, Wetzlar, Germany) at a lateral depth from the midline of between ∼0.5 and 1.1 mm [based on the Mouse Brain Atlas (Paxinos & Franklin 2001)]. After being deparaffinized, the sections were pretreated with 0.5% hydrogen peroxide in methanol for 30 min to quench endogenous peroxidase. Heat-induced epitope retrieval (citrate buffer pH 6 for GFAP and DakoCytomation target retrieval solution for LIMP-II) was undertaken to reduce non-specific staining prior to treatment with blocking serum (5% normal donkey serum and 1% bovine serum albumin in PBS) for 1 h. Sections were incubated overnight at room temperature with primary antibodies (anti-ubiquitin 1:8000, anti-GFAP 1:12000 and anti-LIMP-II 1:800), diluted in blocking serum. After washing in PBS, the sections were incubated in species-specific biotinylated secondary antibody, diluted 1:800 in PBS, for 1 h. Sections were visualized using a Vectastain ABC kit (PK-6100; Vector Laboratories, Burlingame, CA, USA) and diaminobenzidine (DAB) liquid substrate system (D7304; Sigma), enhanced by the addition of cobalt chloride (CoCl2). Histochemical staining for BSI-B4 was performed by quenching endogenous peroxidase as described above, followed by enzymatic epitope retrieval with 0.05% trypsin (T7409; Sigma) in 0.1% calcium chloride, pH 7.8, and incubation overnight in BSI-B4 (1 mg/ml) diluted 1:80 in Tris buffer, with the addition of CaCl2 and MgCl2. The reactions were visualized with DAB/CoCl2, as for immunostaining. Sections were batched for each stain.

Mounted and coverslipped sections were viewed on an Olympus BX41 microscope, and digital images were collected using an Olympus CC12 camera. Several brain regions per stain were imaged and analysed for each experimental mouse group. In sagittal sections, three regions of the cerebral cortex were examined – rostral, medial and caudal cortex – each covering an area of 0.55 mm2, with the area of interest extending ∼0.8 mm ventral from the cortical surface. An area of 0.16 mm2 (i.e. under × 200 magnification) was examined in the thalamus, cuneate, cerebellar white matter and inferior colliculus. An area of 0.58 mm2 (i.e. under × 100 magnification) was examined in the hippocampus, periaqueductal grey, hypothalamus and superior colliculus. Brain regions were defined using anatomical landmarks with reference to the mouse brain atlas of Paxinos & Franklin (2001).

All parameters of imaging and calibration remained constant for each comparative region and stain. GFAP, LIMP-II and ubiquitin images were analysed using MetaMorph (version 6.1; Molecular Devices, Victoria, Australia) software. Thresholding based on the optical density (OD) of positive immunostaining was applied to the images in a consistent manner. Data are reported as either %GFAP and %LIMP-II immunoreactivity or density of immunopositive ubiquitin inclusions (number/mm2). Analysis Lifescience Research software (Olympus, Victoria, Australia) was used to examine the BSI-B4 images and the number (per mm2) and size (μm2) of activated microglia were determined.

Detection of anti-rhSGSH antibodies in blood plasma

Plasma samples taken at euthanasia were assayed using a horseradish peroxidase-based assay to detect antibodies to rhSGSH as previously described (Hemsley et al. 2007). Antibody complexes were determined at OD 414 nm and antibody titres expressed as the dilution for an absorbance greater than two standard deviations above the blank.


Data are expressed as mean ± standard error of the mean. Our experience with the measurement of HNS-UA in MPS IIIA mouse brain samples using tandem mass spectrometry indicates that these data exhibit a normal distribution; therefore, we undertook parametric statistical analysis (anova) of these data. Significant main effects were subsequently explored by post hoc testing with a Bonferroni correction for multiple comparisons. For histological data, given the small sample sizes, normality could not be established; thus, these data were first transformed (= log [+ 1]), then analysed using anova with post hoc testing as above. In all cases, < 0.05 was considered to be statistically significant.


Response to injections

The study was conducted with a total of 61 mice that that were administered 349 injections over the course of the experiment. Five deaths were recorded: one occurred immediately following anaesthesia (one MPS 30-μg SGSH-treated mouse; 10 weeks of age); one was possibly because of damage to the brainstem during surgery (unaffected vehicle-treated mouse; 6 weeks of age); two were believed to be because of smothering by littermates during recovery (both MPS 30-μg SGSH-treated; 8 weeks of age) and one was because of an unknown cause (17 weeks of age; unaffected vehicle-treated mouse). Signs of a hypersensitivity reaction to infused protein (e.g. altered respiration, swelling, diarrhoea, rash and cyanosis) were not observed at any time.

Effect of treatment on HNS-UA levels in brain and spinal cord

At euthanasia (3 weeks after the final injection of rhSGSH, i.e. 21 weeks of age), very large increases in the relative level of HNS-UA were observed in MPS IIIA vehicle-treated mouse brain and spinal cord when compared with the levels seen in unaffected mice (Fig. 1). Statistically significant differences were observed in brain slice 1 (F4,16 = 44.79, < 0.001), slice 2 (F4,17 = 15.58, < 0.001), slice 3 (F4,16 = 36.23, < 0.001), slice 4 (F4,17 = 55.21, < 0.001) and slice 5 (F4,17 = 173.9, < 0.001). HNS-UA levels in MPS vehicle-treated mice were also statistically increased in the cervical (F4,16 = 61.52, < 0.001), thoracic (F4,17 = 21.6, < 0.001) and lumbar (F4,17 = 10.22, < 0.001) spinal cord.

Figure 1.

Effect of repeated rhSGSH injection on the relative level of HNS-UA (per mg protein) in five hemicoronal brain slices and cervical (CM), thoracic and lumbar cord. = 3–5 mice per treatment group. Each data point represents a mouse. The same symbol has been used for each mouse in a group in each tissue sample. Arrow on brain slice indicates injection site.

A decrease in the relative amount of HNS-UA was observed following treatment with all three doses of rhSGSH in brain slice 1 (F4,16 = 44.79, < 0.001), slice 4 [F4,17 = 55.21, < 0.05 (30 μg),< 0.001 (50 and 70 μg)], slice 5 (F4,17 = 173.9, < 0.001) and the cervical (F4,16 = 61.52, < 0.001) and thoracic [F4,17 = 21.6, < 0.001 (30 and 50 μg), < 0.01 (70 μg)] cord when compared with HNS-UA levels in the same regions in vehicle-treated MPS IIIA mice (Fig. 1). Only the two higher doses were able to significantly reduce HNS-UA in slice 3 [F4,16 = 36.23, < 0.05 (50 μg), < 0.01 (70 μg)]. The magnitude of the difference between vehicle-treated and enzyme-treated mouse HNS-UA levels varied between the regions sampled and was greatest in brain slices 1, 5 and the cervical spinal cord.

A dose-dependent decrease in HNS-UA levels was evident in slices 4 and 5. In slice 4, 30-μg rhSGSH was less effective than both 50- and 70-μg rhSGSH (F4,17 = 55.21, < 0.01). The highest dose of 70-μg rhSGSH was more effective than 30 μg in reducing HNS-UA in slice 5 (F4,17 = 173.9, < 0.01). Dose effects were not seen in any other region (Fig. 1). No statistically significant decrease in HNS-UA content was seen in slice 2 or lumbar cord with treatment. Despite the overall decrease in disaccharide levels in treated mice, HNS-UA remained greater in all MPS mice regardless of treatment status than in unaffected vehicle-treated mice.

Immunohistochemistry and histochemistry

Following fixation–perfusion, brain sections from three mice per group were stained using markers of lysosomal membranes, inflammatory and neurodegenerative pathology. Quantitative assessments were made of marker staining in a variety of stain-dependent brain regions. The regions were selected on a stain-by-stain basis with regard to previous studies and unpublished data (Hemsley et al. 2007; Savas et al. 2004; unpublished data) and are shown in Fig. 2. The region defined as ‘rostral cortex’ consists of the rostral secondary motor cortex, the area termed ‘medial cortex’ corresponds to primary and secondary motor cortical fields and the area referred to as ‘caudal cortex’ contains retrosplenial and secondary visual cortex. The area of the hippocampus surveyed was field CA1; the cortex of the inferior colliculus and the zonal and superficial grey layers of the superior colliculus were examined. The cerebellar white matter consisted of the white matter located within the folia; we assessed a region in the thalamus (approximately corresponding to the ventrolateral nuclei), the ventrolateral periaqueductal grey (adjacent to the second cerebellar lobule), the ventromedial hypothalamic nuclei and the cuneate nucleus.

Figure 2.

Effect of repeated rhSGSH injection on LIMP-II expression in various brain regions. Mean ± SEM. = 2–5 mice per treatment group. *, Significantly different to MPS vehicle-treated mice; ∞, significantly different to unaffected vehicle-treated mice. Photos depicting LIMP-II immunostaining in rostral cortex (a–c) and inferior colliculus (d–f) are shown. A higher power photo of LIMP-II-positive vesicles in a pyramidal neuron in the rostral cerebral cortex is shown in (g). Scale bar is 200 μm in (a–f) and 10 μm in (g). A mouse brain section illustrating the regions examined in the histological analyses is shown (h). All sections were ∼0.5–1.1 mm lateral to the midline and were determined using a mouse brain atlas. Brain regions analysed were chosen on a stain-by-stain basis.

LIMP-II immunostaining

Detection of the lysosomal/late endosomal membrane protein LIMP-II/LGP85 was used to identify changes in lysosomal membrane content with treatment in MPS IIIA mice (Fig. 2). Low levels of diffuse intracytoplasmic staining were observed in unaffected mice (Fig. 2a,d); however, in vehicle-treated MPS IIIA mice, LIMP-II staining was characterized by intensely staining intracytoplasmic vesicular structures (Fig. 2b,e) distributed throughout most brain regions. The cytoplasm of pyramidal neurons in the cortex contained particularly large numbers of stained vesicles (Fig. 2g). Immunopositive regions in vehicle-treated MPS IIIA mice included the brainstem, striatum, olfactory bulb and hypothalamus; however, these areas were not quantified. Large, statistically significant increases in staining were observed in vehicle-treated MPS IIIA mice in the rostral (F4,13 = 12.18, < 0.001), medial (F4,13 = 29.02, < 0.001) and caudal (F4,13 = 12.85, < 0.001) cerebral cortex, the superior colliculus (F4,13 = 22.41, < 0.001), inferior colliculus (F4,12 = 22.5, < 0.001), cerebellar white matter (F4,11 = 5.446, < 0.01), hippocampus (F4,13 = 33.76, < 0.001) and thalamus (F4,13 = 15.12, < 0.001) when these regions were compared with those in unaffected mice.

Enzyme-treated MPS IIIA mouse brain sections exhibited significant reductions in LIMP-II immunoreactivity when compared with those from vehicle-treated MPS IIIA mice (Fig. 2). Statistically significant reductions in immunoreactivity were observed in the rostral cortex [F4,13 = 12.18, < 0.05 (30 and 50 μg)], medial cortex [F4,13 = 29.02, < 0.05 (50 μg)] and caudal cerebral cortex [F4,13 = 12.85, < 0.05 (30 μg), < 0.01 (50 μg)], the superior colliculus [F4,13 = 22.41, < 0.001 (30 and 50 μg)], inferior colliculus [F4,12 = 22.5, < 0.001 (30 and 50 μg)] and the thalamus [F4,13 = 15.12, < 0.05 (30 and 50 μg)]. In some cases, the level of immunoreactivity observed following 30- or 50-μg rhSGSH was no longer different to that seen in unaffected mice. Dose-dependent effects of therapy on LIMP-II expression were not readily seen at the rhSGSH concentrations used here. Statistical analysis was only undertaken where group sizes were = 3 or greater, and as batch-stained tissue sections were only available from two mice from the 70-μg rhSGSH group, statistical evaluations of the data from this treatment group were unable to be undertaken; however, the trends are similar to those seen with other enzyme doses. No significant reduction in staining was seen in the cerebellar white matter and hippocampus following enzyme treatment, although a trend towards reduction was seen in all rhSGSH groups.

BSI-B4 staining

BSI-B4, which labels activated microglia and vascular endothelium in rodent tissue (Streit & Kreutzberg 1987), was used to identify microglial activation (Fig. 3). In contrast to resting/ramified microglia, which are pale and have fine processes and small soma, activated microglia stain intensely and have rounded soma with thick, shortened processes. In unaffected mice, very pale-staining microglia with a ramified morphology and vascular endothelium were observed (Fig. 3a,d,g), with intensely stained activated microglia only very rarely seen. Because of threshold parameters, only intensely stained and therefore activated microglia were quantified. Compared with vehicle-treated unaffected mice, vehicle-treated MPS IIIA mice exhibited significant numbers of activated microglia in the rostral (F4,14 = 55.13, < 0.001) (Fig. 3b), medial (F4,14 = 401.2, < 0.001) and caudal (F4,14 = 49.92, < 0.001) cerebral cortical regions, the superior (F4,14 = 26.4, < 0.001) and inferior colliculi (F4,12 = 7.985, < 0.01) (Fig. 3e), the white matter of the cerebellum (F4,12 = 29.07, < 0.01), the thalamus (F4,14 = 86.64, < 0.001) (Fig. 3h) and the hippocampus (F4,14 = 49.48, < 0.001). Other areas such as the olfactory bulb and striatum also contained BSI-B4-positive cells, but quantification was not undertaken.

Figure 3.

Effect of repeated rhSGSH injection on the number of isolectin B4-stained activated microglia in various brain regions. Mean ± SEM. = 2–5 mice per treatment group. *, Significantly different to MPS vehicle-treated mice; ∞, significantly different to unaffected vehicle-treated mice. Photos depicting lectin staining in rostral cortex (a–c), inferior colliculus (d–f) and thalamus (g–i) are shown. Scale bar is 200 μm.

Following administration of rhSGSH, significantly fewer activated microglia were seen in the more superficial regions of the brain, e.g. rostral cerebral cortex [F4,14 = 55.13, < 0.01 (50 μg), < 0.05 (70 μg)] (Fig. 3c), superior colliculus [F4,14 = 26.4, < 0.05 (30 μg), < 0.01 (50 μg), < 0.001 (70 μg)] and inferior colliculus [F4,12 = 7.985, < 0.05 (30 and 70 μg)] (Fig. 3f), but there was little or no change in deeper regions, e.g. thalamus (Fig. 3i) and hippocampus. The number of BSI-B4-positive cells observed in the brain in mice treated with different doses of rhSGSH was not obviously different. Microglial cell size was also reduced in the cerebral cortex, particularly rostrally and caudally, but not to statistically significant levels (data not shown).

GFAP immunostaining

GFAP staining appeared more intense in all MPS IIIA mice in most brain regions, regardless of treatment status when compared with the paler staining seen in unaffected mice (Fig. 4a,d). MPS IIIA vehicle-treated mouse brain showed significantly increased expression of GFAP in several regions, including the rostral cortex (Fig. 4b) (F4,14 = 4.658, < 0.01), medial cortex (F4,14 = 66.02, < 0.001), caudal cortex (F4,14 = 19.51, < 0.001), the superior colliculus (F4,13 = 8.496, < 0.001), inferior colliculus (F4,13 = 19.36, < 0.001; Fig. 4e), thalamus (F4,14 = 155.0, < 0.001) and cuneate (F4,11 = 11.68, < 0.01) when compared with unaffected vehicle-treated mouse brain. Astrogliosis was not evident in MPS IIIA hippocampus or cerebellar white matter. Only the inferior colliculus (Fig. 4f) showed a statistically significant non-dose-dependent decrease in GFAP expression with enzyme treatment [F4,13 = 19.36, < 0.01 (30 and 70 μg)], with smaller non-significant reductions also seen in the rostral (Fig. 4c) and caudal cerebral cortex and the superior colliculus.

Figure 4.

Effect of repeated rhSGSH injection on the expression of GFAP in various brain regions. Mean ± SEM. = 2–5 mice per treatment group. *, Significantly different to MPS vehicle-treated mice; ∞, significantly different to unaffected vehicle-treated mice. Photos depicting GFAP staining in rostral cortex (a–c) and inferior colliculus (d–f) are shown. Scale bar is 200 μm.

Ubiquitin immunostaining

Ubiquitin-positive inclusions were rarely observed in unaffected vehicle-treated mice (Fig. 5a). In contrast, many regions of MPS IIIA vehicle-treated mouse brain exhibited large numbers of rounded intracytoplasmic ubiquitin-positive inclusions with variable staining intensity and size (Fig. 5). Significant numbers of inclusions were seen in the superior (F4,14 = 88.39, < 0.001) and inferior colliculi (F4,13 = 23.53, < 0.001) with slightly lower numbers in the hypothalamus (F4,14 = 134.3, < 0.001) and periaqueductal grey (Fig. 5b) (F4,14 = 427.5, < 0.001) compared with vehicle-treated unaffected mice. The cuneate exhibited very large round inclusions, often with non-homogeneous staining, along with many smaller inclusions; however, this region was not quantified. In mice treated with rhSGSH, the number of inclusions in the inferior colliculus [F4,13 = 23.53, < 0.001 (30 μg), < 0.001 (50 μg), < 0.01 (70 μg)] and superior colliculus [F4,14 = 88.39, < 0.001 (30, 50 and 70 μg)] was significantly lower than in vehicle-treated mice. Fewer lesions were also seen in the hypothalamus [F4,14 = 134.3, < 0.05 (30 μg), < 0.001 (50 μg)] and periaqueductal grey (Fig. 5c) [F4,14 = 427.5, < 0.05 (30 μg), < 0.001 (50 μg), < 0.01 (70 μg)].

Figure 5.

Effect of repeated rhSGSH injection on ubiquitin-positive lesion formation in various brain regions. Mean ± SEM. = 3–5 mice per treatment group. *, Significantly different to MPS vehicle-treated mice; ∞, significantly different to unaffected vehicle-treated mice. Photos depicting ubiquitin-positive lesions in periaqueductal grey (a–c) are shown. Scale bar is 200 μm.

Electron microscopy

A qualitative examination of the molecular and Purkinje cell layers of the cerebellum and medial cerebral cortex immediately superior to the hippocampus was undertaken on tissues from three mice per experimental group.

Cerebral cortex

Initially, toluidine blue-stained 1 μm semithin sections were analysed by light microscopy, showing significant differences between MPS IIIA and unaffected vehicle-treated mice at 21 weeks of age. Pial cell cytoplasm contained large numbers of foamy vacuoles in vehicle-treated MPS IIIA mice, a feature absent in unaffected mice. No readily discernable difference in the number of inclusions in this cell type was observed in enzyme-treated mouse brain sections compared with those from vehicle-treated MPS IIIA mice. This observation was confirmed at the EM level.

When ultra-thin sections were examined at low power (× 800), mixed intraneuronal (membranous, amorphous and combinations of the two) and foamy, amorphous perineuronal microglial inclusions were the most obvious pathological feature. Perivascular, free and perineuronal microglia in mice from all treatment groups contained foamy inclusions. However, engorged perineuronal, and more occasionally, free microglia were commonly observed in the vehicle-treated affected mice, with very large numbers of ‘foamy’ vacuoles detected. Pericytes often contained this foamy material; however, fewer of these cells were observed.

The number of microglia in this region that contained amorphous inclusions appeared to decrease with treatment, with increasing numbers of both free and perineuronal microglia devoid of inclusions, observed. There was no readily discernable effect of enzyme dose at this site in the cerebral cortex. While much of the material in the microglial inclusions was amorphous in nature, some membranous structures were also observed. None of these structures was seen in unaffected mice. Fewer storage vesicles were observed in microglia in both superficial and deeper regions of the cerebral cortex in enzyme-treated mice.

In vehicle-treated mice, endothelial cells were observed to occasionally exhibit storage inclusions, with amorphous contents. Affected cells were more readily observed in the larger vessels. There was no obvious effect of enzyme delivery on this cell type, with endothelial cells in cortical sections from mice treated with all doses of enzyme exhibiting comparable numbers of lesions with similar contents.

Ultra-thin sections also showed pyramidal neurons containing large numbers of amorphous and/or mixed inclusions in vehicle-treated MPS IIIA mice (Fig. 6b,f). Additionally, very large numbers of multimembranous structures were also seen in many of these neurons, particularly in the deeper cortical layers. In enzyme-treated mice, fewer neurons appeared to contain multimembranous material, and the number of mixed and amorphous inclusions was decreased (Fig. 6c,d,g,h). Lesions were not seen in unaffected mice (Fig. 6a,e).

Figure 6.

Pyramidal neurons in the upper (∼layer 3; a–d) and lower (∼layer 5; e–h) cerebral cortex in unaffected (a and e), MPS vehicle-treated (b and f) and enzyme-treated MPS IIIA (c and g, 30 μg; d and h, 70 μg) mice. Scale bar is 10 μm in (e) and 5 μm in all other figures.


The most obvious cell type exhibiting pathology in the cerebellum was the Purkinje cell. In EM sections taken from vehicle-treated MPS IIIA mice, these cells were observed to contain many medium-sized inclusions and often several large ones (Fig. 7b). No inclusions were seen in unaffected mice (Fig. 7a). In enzyme-treated MPS IIIA mice, there appeared to be fewer inclusions in these cells (Fig. 7c,d). The dendrites extending into the outer molecular layer were found to contain significant numbers of multimembranous structures and also some mixed inclusions (Fig. 7e–h) in both enzyme-treated and vehicle-treated MPS IIIA mice but not unaffected mice.

Figure 7.

Purkinje neurons in cerebellum in unaffected (a), MPS vehicle-treated (b) and enzyme-treated MPS IIIA (c, 30 μg and d, 70 μg) mice. The dendrites of the neurons extending into the molecular layer could be seen to contain numerous multimembranous structures. MPS vehicle-treated mice (e and f) and MPS enzyme-treated mice (g and h, 30 μg) are shown. Scale bar is 10 μm in (a–c and g) and 5 μm in all other figures.

Development of antibodies to rhSGSH following intra-CSF delivery of enzyme

The antibody titre for each mouse plasma sample is shown in Table 1. Of the 29 plasma samples taken from MPS IIIA mice receiving enzyme at euthanasia, 22 exhibited significant anti-rhSGSH antibody titres. Of those mice not exhibiting antibodies, three received 30-μg rhSGSH and two received each of 50- and 70-μg rhSGSH.

Table 1.  Anti-rhSGSH antibody titres for plasma samples taken at euthanasia
UnaffectedMPS vehicleMPS 30 μgMPS 50 μgMPS 70 μg
  1. Data are the dilution at which OD > 2SD above the blank value. Each value is an individual mouse.

<25<25<25320012 800
<25<2512 800640012 800
<25<2512 800640012 800
<25<2525 60012 80012 800
<25<2525 60012 80025 600
<25<2551 20012 80025 600
505051 20025 60051 200
505051 20051 200No sample
No sample200 


There is no safe and effective therapy for many LSDs that affect the brain, therefore patients progressively decline and die. We have investigated the effectiveness of a potential therapy option – replacing lysosomal enzyme through CSF delivery – in ameliorating primary lysosomal storage, in addition to other more complex pathological changes. High doses of CSF-delivered rhSGSH significantly reduced the relative level of an HS-derived disaccharide marker, HNS-UA, and we have shown for the first time that this treatment strategy also diminishes many of the other neuropathological ramifications of lysosomal storage in MPS IIIA mouse brain.

Effect of rhSGSH on HS storage

Our previous studies in MPS IIIA mice (Hemsley et al. 2007), and those of our colleagues using a canine model of MPS I (Dickson et al. 2007; Kakkis et al. 2004), provided evidence to suggest that this methodology is effective at reducing primary lysosomal storage in both conditions and is capable of improving clinical function in MPS IIIA (Hemsley et al. 2007). The dose of rhSGSH administered in this study was higher than that used previously in mice (5–20 μg; Hemsley et al. 2007), and, as a consequence, the effect of treatment on the relative level of HNS-UA was greater on average, particularly in the middle of the brain (hemicoronal slices 2 and 3). These slices contain the lateral ventricle (LV), medial cortex, hippocampus and thalamus. There was an average reduction of 34% and 44% in slices 2 and 3 (respectively) with 70-μg rhSGSH compared with HNS-UA levels recorded in vehicle-treated affected mice. Fortnightly delivery of 20-μg rhSGSH permitted only an average of 15% and 12% reduction in the respective slices.

The mice in the original study were euthanased 24 h after the injection at 18 weeks of age, whereas those in the present study were killed at 3 weeks after the last injection at 21 weeks of age. Unpublished observations (K.M.H.) indicate the ability of intra-CSF-delivered rhSGSH to significantly reduce HNS-UA in brain within 24 h. Therefore, the reduction in HNS-UA seen here is potentially an underestimation of the effectiveness of rhSGSH to catabolize HS-derived oligosaccharides because the 3-week interval between the last injection and euthanasia is likely to have resulted in reaccumulation of HS.

Supporting these biochemical observations is the quantitative immunohistochemistry using a lysosomal membrane marker. We observed very large reductions in LIMP-II immunoreactivity with treatment, particularly in areas close to the surface of the brain (therefore exposed to CSF) and in regions located in the rostral and caudal aspects of the brain, i.e. superior and inferior colliculi, and rostral and caudal cerebral cortex. Less of an effect on LIMP-II immunostaining was observed in deeper and more centrally located brain regions (e.g. thalamus and hippocampus). Although the reductions in HNS-UA and LIMP-II staining were large and statistically significant in some regions, complete removal of lysosomal storage was not seen.

Semi-quantification of lysosomal vacuolation observed at the EM level was not undertaken in present study as comparison of very small areas (1 mm2) of brain tissue taken from different mice is potentially subject to considerable error, even when tissue samples are dissected using a mouse brain blocker, as they were here. When a qualitative assessment of lysosomal storage inclusions was made, storage vesicles were still present; however, there was an apparent decrease in the accumulation of foamy macrophages in the cortical neuropil and fewer membranous structures in cortical pyramidal and cerebellar Purkinje neurons. The sample of medial cortex examined at the EM level is likely to, on the basis of the mass spectrometry data for slice 3 and the medial cortex immunohistochemical data, exhibit less reduction in storage and other pathology than, say, the rostral or caudal cortical regions.

Furthermore, there was no easily discernable effect of treatment on pial cell storage pathology, which is curious, given the ready access this cell type has to the enzyme, and our previous observations in mice treated with lower doses of rhSGSH (Hemsley et al. 2007). One explanation for these observations is the different interval between the last injection and euthanasia in the two studies (24 h in the previous study and 3 weeks in this study). The time taken for reaccumulation of storage material is as yet unknown; however, the 3-week interval in present study is likely to have seen a reaccumulation of some lysosomal storage pathology. Similar observations were also made in MPS I dogs when treatment–euthanasia intervals were increased (Dickson et al. 2007; Kakkis et al. 2004).

These observations highlight the importance of developing sensitive, relevant biomarkers by which to monitor the time–course of removal and return of various aspects of pathology within the brain, to enable treatment frequency to be optimized in patients.

Effect of rhSGSH on other neuropathology

Microgliosis and astrogliosis are evident early in the course of disease in murine MPS IIIA (by 8 weeks of age, the earliest time examined; unpublished data) and MPS IIIB (4+ weeks of age; Li et al. 2002; Ohmi et al. 2003). The brain regions involved and the number of activated glia per region increases as the disorder progresses. Recombinant human SGSH treatment was effective at ameliorating some of the inflammatory changes. We observed reductions in microgliosis and astrogliosis in region-dependent areas of brain. As observed with LIMP-II, surface areas at the more rostral and caudal aspects of the brain (superior and inferior colliculi and rostral and caudal cerebral cortex) exhibited the greatest reduction in inflammatory pathology. Interestingly, although LIMP-II immunostaining was decreased in the thalamus, neither microglial nor astrocyte activation was reduced in this region, potentially suggesting that the level of lysosomal storage reduction in this area following enzyme delivery was insufficient to impact upon the glial activation state. Microglial activation has been referred to as a ‘double-edged sword’ (Stoll et al. 2002) as the cells are involved in both neuroprotective and neurodegenerative processes. In the case of an acute insult, the former may outweigh the latter, but where the activating factor (in the case of MPS IIIA, this may be HS, secondarily stored lipids or another unknown factor) is supplied continuously, neurotoxicity may prevail.

It will be important to determine the exact nature of the trigger and the ‘threshold’ of pathology required for glial activation to administer therapies most effectively. The time–course and nature of the relationship between primary HS and secondary ganglioside storage and neuroinflammatory changes has not yet been established for any LSD. However, we hypothesize that relatively low levels of HS (and/or gangliosides) may be sufficient to initiate a cascade of events resulting in activation of microglia. Ohmi et al. (2003) showed activated microglia in MPS IIIB mouse brain between 2 weeks and 1 month of age, a time at which low but significant amounts of HNS-UA are seen in MPS IIIA brain (Crawley et al. 2006; King et al. 2006). Microglia are one of the first cell types to exhibit lysosomal inclusions and the material presumably originates from both endogenous sources and phagocytosis.

Secondarily stored gangliosides may play a key role in microglial activation. In the LSD Tay-Sachs and Sandhoff diseases, where gangliosides are the primary lysosomal storage product, mice exhibit elevated numbers of macrophage/microglial cells in the brain and increased expression of inflammatory molecule genes (Jeyakumar et al. 2004; Myerowitz et al. 2002; Wada et al. 2000). Deletion of the macrophage inflammatory protein 1α gene reduces microglial infiltration and neuronal apoptosis and improves neurological function in Sandhoff mice (Wu & Proia 2004). Improved clinical parameters and lifespan are also observed in Sandhoff mice when anti-inflammatory or antioxidant medications are administered (Jeyakumar et al. 2004).

The RhSGSH treatment also prevented the accumulation of intracellular ubiquitin-positive inclusions, shown to be present in certain brain regions by ∼7 weeks of age (Savas et al. 2004). The ubiquitin-tagged proteins have not yet been positively identified but are hypothesized to accumulate following autophagic blockade (Settembre et al. 2008). Treatment appeared to reduce the number of lesions in the hypothalamus, located adjacent to the ventral surface of the brain, and the periaqueductal grey. Based on our observation that at ∼6 weeks of age (when the therapy began), lesions are generally only found in the cuneate nucleus (Savas et al. 2004); we believe that treatment has prevented the development of lesions rather than enabling their removal. Lesions in the cuneate seem to be largely impervious to treatment, despite their proximity to the injection site, suggesting their irreversibility (Hemsley et al. 2007; Savas et al. 2004).

Antibody response to rhSGSH delivery

It is possible that the therapeutic effect of sulphamidase administration was submaximal in the treated mice as a result of the antibodies generated towards rhSGSH. In a recent study by Matzner et al. (2008), antibodies were detected in mice receiving multiple intravenous injections of recombinant human arylsulphatase A (rhASA). Reduced sulphatide clearance in tissues was observed in antibody-generating mice. The antibodies did not inhibit enzyme activity but were shown to reduce rhASA stability and thus half-life, decrease mannose-6-phosphate receptor-mediated cellular uptake and modify intracellular targeting of enzyme.

Whether the anti-rhSGSH antibodies generated in our mice had similar effects is unknown; however, whether they reduce the clinical efficacy of this treatment strategy could potentially be established by inducing immune tolerance to rhSGSH prior to intra-CSF enzyme delivery. The time–course of antibody generation in the mice is not known; however, in a similar study in MPS IIIA Huntaway dogs (Hemsley, K. M., Norman, E. J., Crawley, A. C., Auclair, D., King, B., Fuller, M., Lang, D. L., Dean, C. J., Jolly, R. D. & Hopwood, J. J., submitted), significant anti-rhSGSH antibody titres were observed in both CSF and plasma after two to three intra-CSF injections.

Feasibility of intra-CSF injections

The delivery of compounds into CSF for treatment of brain disease is clinically acceptable and feasible. CSF is made and released within the lateral ventricle (LV) and third and fourth ventricles. The injection site used in all the studies to date, the cerebellomedullary cistern, is downstream of the ventricles, and this might explain the difficulty in delivering enzyme back into the centre of the brain (i.e. middle of slices 2 and 3). We predict that for ongoing clinical efficacy, a large proportion of the brain must exhibit a significant and maintained reduction in neuropathology; therefore, other CSF injection routes such as LV delivery should be considered as they might provide even greater enzyme penetration and distribution.

An adenovirus expressing β-glucuronidase was injected into the MPS VII mouse LV or cerebellomedullary cistern by Ghodsi et al. (1998) and, of the two routes, virus injected into the LV achieved the greatest distribution. While repeated intra-LV injections are invasive and not feasible in humans, surgical implantation of a port or pump device into the ventricle is practical (e.g. Medtronic SynchroMed Infusion system; Gilmartin et al. 2000) and would allow continuous enzyme delivery. These devices have been used in several other disorders (e.g. spasticity in cerebral palsy, chronic pain and spinal symptoms in multiple sclerosis; Boviatsis et al. 2005; Rudich et al. 2004; Scheinberg et al. 2001). Finally, long-term administration of recombinant enzymes is safe, despite patients developing antibodies to intravenously infused protein as was observed in the mice in the present study. While, as previously discussed, this could affect the efficacy of therapy and pose a safety issue, not all patients develop antibodies and in those who do, immune tolerization often occurs (Kakavanos et al. 2003).

In summary, intra-CSF delivery of rhSGSH is effective at reducing lysosomal storage and other neuropathology in MPS IIIA. Indicators of activated microglia and astroglia may provide biomarkers of disease (and pathology amelioration) in this and other related conditions. The development of strategies permitting ongoing treatment without the need for repeated intra-CSF injections would feasibly allow patients to be treated in the medium to long term. However, higher doses of rhSGSH and/or alternative routes of delivery into CSF should be considered to provide maximal delivery of therapeutic enzyme to both superficial and deeper brain regions.

Conflict of interest

The authors received funding for this project from the Australian National Health & Medical Research Council (J. J. H. and K. M. H.), the Sanfilippo Children’s Research Foundation (J.J.H. and K.M.H.) and Shire Human Genetic Therapies (J.J.H.). Shire Human Genetic Therapies provided the rhSGSH. The funding sources did not have any role in study design, data collection, data analysis, interpretation of data, writing of the report or in the decision to submit the paper for publication.


An international patent is held by J.J.H. and others for mammalian sulphamidase and genetic sequences encoding it for use in the investigation, diagnosis and treatment of subjects suspected of suffering from sulphamidase deficiency (US Patent no. 5863782).


We gratefully acknowledge the advice and support of Dr Allison Crawley and the assistance of Amanda Luck, Tina Rozaklis, Suzie Brodie and Hanan Hannouche. We thank Dr Emma Parkinson-Lawrence for the generous provision of the LIMP-II antibody, we are grateful to Lynn Scarman and Leslie Jenkins-White for caring for the mice, acknowledge Lyn Waterhouse (Adelaide Microscopy) for processing EM samples and thank Drs Mike Concino and Satyajit Ray at Shire Human Genetic Therapies for preparing the rhSGSH. We also thank the Linton Family and the Sanfilippo Children’s Research Foundation for generously providing funding to assist this study.