Central nervous system pathology in preclinical MPS IIIB dogs reveals progressive changes in clinically relevant brain regions

Mucopolysaccharidosis type IIIB (MPS IIIB; Sanfilippo syndrome B) is an autosomal recessive lysosomal storage disorder caused by the deficiency of alpha-N-acetylglucosaminidase activity, leading to increased levels of nondegraded heparan sulfate (HS). A mouse model has been useful to evaluate novel treatments for MPS IIIB, but has limitations. In this study, we evaluated the naturally occurring canine model of MPS IIIB for the onset and progression of biochemical and neuropathological changes during the preclinical stages (onset approximately 24–30 months of age) of canine MPS IIIB disease. Even by 1 month of age, MPS IIIB dogs had elevated HS levels in brain and cerebrospinal fluid. Analysis of histopathology of several disease-relevant regions of the forebrain demonstrated progressive lysosomal storage and microglial activation despite a lack of cerebrocortical atrophy in the oldest animals studied. More pronounced histopathology changes were detected in the cerebellum, where progressive lysosomal storage, astrocytosis and microglial activation were observed. Microglial activation was particularly prominent in cerebellar white matter and within the deep cerebellar nuclei, where neuron loss also occurred. The findings in this study will form the basis of future assessments of therapeutic efficacy in this large animal disease model.


Results
The accumulation over time of non-degraded HS polysaccharide and oligosaccharide chains was quantified in MPS IIIB dogs at 1, 2, 3, 6, 9, and 12 months of age. Total HS, as defined by the sum of the internal disaccharides of HS, and the MPS IIIB-specific non-reducing end of HS (the substrate of NAGLU enzymatic activity) were measured in brain tissue (cerebral grey matter), CSF, plasma, and urine ( Fig. 1, Supplementary Figure 1, Individual dog IDs presented in Table 1). Unaffected dogs had low levels of HS and HS-NRE in brain tissue, CSF, plasma, and urine at all time points tested (Fig. 1, Supplementary Figure 1). In contrast, levels of total HS and HS-NRE in brain, CSF, plasma and urine of MPS IIIB dogs were markedly increased over levels in unaffected dogs (Fig. 1, Supplementary Figure 1). In the cortex of MPS IIIB dogs, high levels of HS and HS-NRE were detected at 1 month of age (the earliest time point tested), and levels of total HS increased over time and began to plateau by about 12 months of age (Fig. 1A). In the CSF of MPS IIIB dogs, levels of total HS and HS-NRE decreased over time, reaching a steady state of approximately 17-fold over levels in unaffected dogs by approximately 6 months of age (Fig. 1B).
Increased sulfation of HS occurs in mouse models of MPS IIIB and other MPS disorders 18 . To assess whether the composition of the HS changes over time in MPS IIIB dogs, we quantified the extent of the distinct sulfation modifications (2-O, 6-O, and N-sulfation) as well as the abundance of N-acetylated glucosamine. We found that the abundance of 2-O sulfated and N-sulfated HS subtly increased while the abundance of N-acetylated HS Scientific Reports | (2020) 10:20365 | https://doi.org/10.1038/s41598-020-77032-y www.nature.com/scientificreports/ subtly decreased over time in brain tissue of MPS IIIB dogs (Supplementary Figure 2A). The composition of HS in CSF remained consistent over the time period measured (Supplementary Figure 2B). In summary, MPS IIIB dogs have high levels of HS and HS-NRE in several tissues including brain and CSF. These levels are consistently higher in MPS IIIB dogs than unaffected dogs, even at 1 month of age, the earliest time point evaluated. These results demonstrate a way to directly measure the accumulation of NAGLU substrates (HS and HS-NRE) by LCMS in the brain and in fluids which is easily monitored in the clinic (CSF, plasma, and urine) and provides a tool to measure the primary biochemical activity of NAGLU and its correction with therapy.

MPS IIIB disease forebrain pathology
To understand the effects of HS accumulation on brain pathology, changes in brain architecture were evaluated by histology.
Increased lysosomal storage in the forebrain of MPS IIIB dogs. Lysosomal membrane protein-1 (LAMP1) has been used as a surrogate marker of lysosomal storage burden and has been shown to indicate the size of the lysosomal compartment, despite only being a proxy measure 10 . Tissue from the forebrain of MPS IIIB dogs 15 months of age showed increased LAMP1 staining over the corresponding unaffected controls (Fig. 2). Due to the diffuse background staining within the neuropil of the oldest available MPS IIIB dogs (data not shown), increased LAMP1 staining was most evident in 15-month-old MPS IIIB dogs compared to age-matched and CSF (B) from MPS IIIB (circles, grey bars) and unaffected dogs (triangles, empty bars) at different ages from 1 to 12 months old. Levels in cerebral grey matter were normalized to wet brain mass. Bars illustrate the mean (horizontal line) and standard deviation of data for time points at which samples from more than two dogs were analyzed. Data points with a horizontal line through them but no bar represent more than one sample was analyzed but the variability was extremely small. Data points with no horizontal line indicate a sample from a single animal was analyzed. See Supplementary Table 1 for raw data.  (Fig. 2). A dense accumulation of punctate LAMP1 immunoreactivity was present to different extents in the cytoplasm of individual cells. These cells predominantly had neuronal morphology, although other smaller cells, presumed to be glia, also contained intense LAMP1 immunoreactivity. This staining of individual cells also revealed a series of different regional patterns of LAMP1 immunostaining within the forebrain of 15-month-old MPS IIIB dogs.
In the cortex of 15-month-old MPS IIIB dogs, immunostaining for LAMP1 varied in intensity across the cortical laminae, and at different levels of the brain (Fig. 2). Rostrally, LAMP1 immunostaining was darker within cells in both superficial (II/III) and deeper laminae (VI), but was less intense in the other laminae. In more caudal regions of cortex, this distinction was less apparent with a more even, pan-laminar staining of cells (Fig. 2). Within the white matter, there was also a stark contrast in the distribution of LAMP1 staining; the white matter of the corona radiata that lies directly under cortical grey matter had more intense LAMP1 immunoreactivity than the adjacent white matter of the corpus callosum, where much less LAMP1 staining was evident.
Hippocampal neurons displayed subfield-specific LAMP1 staining in both unaffected and MPS IIIB dogs. This immunoreactivity was much darker in pyramidal neurons of CA3 rather than CA1 of MPS IIIB dogs and also within granule neurons of the dentate gyrus (Fig. 3). There was also staining of LAMP1 with what appeared to be glia scattered through the hippocampal neuropil of 15-month-old MPS IIIB dogs that was not present in unaffected control dogs (Fig. 3). MPS IIIB dogs also had markedly more LAMP1 immunoreactivity throughout the thalamus and hypothalamus, with much more staining within morphologically identified glia, as well as more numerous large, darkly LAMP1-stained neurons in both structures (Fig. 4). In the hypothalamus, the neurons in the paraventricular nucleus seemed to have a particularly increased intensity of LAMP1 staining, as did the nearby substantia nigra. Although unaffected control dogs displayed strong LAMP1 staining in neurons within the substantia nigra pars compacta, these nigral neurons appeared to be enlarged in MPS IIIB dogs. This morphological appearance is indicative of pronounced storage accumulation, consistent with more intense LAMP1 immunostaining in these affected dogs. In addition, other cell types (likely microglia) were also stained in this brain region of MPS IIIB dogs. In the thalamus, the dorsolateral geniculate nucleus had robust LAMP1 staining in both MPS IIIB and unaffected dogs, though staining in neurons of MPS IIIB animals was much darker, as was background staining of the neuropil.

MPS IIIB dogs do not display cortical atrophy.
Previous data from MPS IIIB mice revealed no evidence of significant cortical atrophy 9 . Given that mice may not live long enough to develop the full extent of cortical pathology, we evaluated the presence of cortical atrophy in MPS IIIB dogs by measuring cortical thickness in the most severely affected MPS IIIB dogs ≥ 20 months of age. Measurements were obtained from the somatosensory cortex, a representative region of the sensory cortex that is most severely affected in large animal  The thalamus (Th) of MPS IIIB dogs generally exhibited darker LAMP1 immunostaining; the dorsolateral geniculate nucleus (dGN) appeared to contain especially intense immunostaining for this marker. This was largely due to intense LAMP1 immunoreactivity within the neuronal cytoplasm (dGN-20×) but also staining within the neuropil, presumably within dendritic processes (dGN-20×). Numbers next to ages indicate animal ID number.
Scientific Reports | (2020) 10:20365 | https://doi.org/10.1038/s41598-020-77032-y www.nature.com/scientificreports/ ity in canine tissue. Instead, immunostaining was performed with Iba1, a microglial marker that is present in quiescent cells and is increased with microglial activation 21 . Marked differences in microglial activation and morphology were observed between unaffected and MPS IIIB dogs. Such changes were evident by 3 months of age and became more pronounced over time (Fig. 5).
Surveying the distribution of Iba1 immunoreactivity across brain regions of MPS IIIB dogs revealed no particular focal area of microglial activation. For example, within the cortical mantle, Iba1 immunoreactivity was uniformly increased across the depth of all cortical regions of MPS IIIB dogs, with no greater increase in any individual lamina (Fig. 5, upper panels). Similar but less pronounced changes were evident within the underlying white matter (Fig. 5, upper panels). A notable exception was the presence of localized clusters of intensely stained Iba1-positive microglia in white and grey matter of MPS IIIB dogs of all ages (Fig. 6). Similar microglial clusters were also observed in unaffected dogs, but to a lesser extent. The distribution of these clusters did not appear to favor any specific brain region, but occurred with a similar frequency at all rostro-caudal levels of the brain in both cortical and subcortical structures. Equally, there was no evidence that the clusters associated with blood vessels and the frequency of the clusters did not appear to correlate with disease progression.
There were also clear differences in microglial morphology between unaffected and MPS IIIB dogs, which became more pronounced in MPS IIIB dogs with age. In unaffected animals, the many thin ramified Iba1-positive processes characteristic of quiescent or resting microglia gave the background of the neuropil a "lacy" appearance ( Fig. 5, cortical layer inset), an appearance that was less apparent in MPS IIIB dogs. Morphological changes in microglia in MPS IIIB dogs reflected their activated state, being not only more intensely stained than cells in unaffected controls, but also with fewer thickened processes and an enlarged cell soma, typical of activated microglia.
Iba1-positive morphologically active microglia were particularly evident in the cortical mantle of MPS IIIB dogs from 3 months of age and were distributed uniformly throughout the cortex (Fig. 5, and with age morphological differences between Iba1-positive microglia became more pronounced. Twenty-month-old MPS IIIB dogs had increased incidence of distended microglial cell bodies in the cortical neuropil (Fig. 5). Measurements of the cross-sectional area of Iba1-positive microglial soma confirmed that these microglia were significantly larger in MPS IIIB dogs than their wild-type counterparts at 20 months of age or older (Supplemental Figure 2). There was also a characteristic increase in microglial activation in the white matter in MPS IIIB dogs that was not seen in unaffected controls (Fig. 5). Although the morphological changes in white matter microglia were Increased lysosomal storage in the cerebellum of MPS IIIB dogs. Lysosomal storage burden was greatly increased in several areas of the cerebellum in the schipperke MPS IIIB dogs 17 . In our study, immunostaining of LAMP1, a surrogate marker for storage burden, was increased in MPS IIIB dogs at all ages examined, as compared to age-matched unaffected dogs, although there was significant inter-sample variability in the immunostaining intensity (Fig. 7). In unaffected dogs, LAMP1 staining was very faint, and only appreciable within the characteristically large soma of Purkinje neurons, whose morphology was readily apparent. In contrast, there was dramatically more LAMP1 staining in MPS IIIB dogs. While this increase was evident at all ages examined, the difference between the diseased state and unaffected controls was more obvious by 9 months of age. In these MPS IIIB dogs, LAMP1 not only intensely stained Purkinje neurons and their projections within the molecular layer, but also stained the granule cell layer and DCN, becoming especially pronounced at 15 months of age (Fig. 7). In the oldest MPS IIIB dogs, the pattern of LAMP1 staining changed again, with far fewer stained Purkinje neurons, markedly less staining in the molecular layer, and fewer, more intensely stained cells in the granule cell layer. This change in distribution could reflect the loss of Purkinje neurons and granule cells in MPS IIIB dogs, with more LAMP1 staining in microglia (Fig. 7). These findings suggest that storage burden, as measured by LAMP1 immunostaining, reaches a peak in brains of MPS IIIB dogs before tapering off in later stages of the disease, likely due to cell death of various cell types expressing LAMP1.
Progressive gliosis in the MPS IIIB cerebellum. The activation of both astrocytes and microglia occurs as a response to cell damage in LSDs such as MPS IIIB 9,10,20 . Using GFAP as a marker, we assessed the extent of astrogliosis in canine MPS IIIB disease.
Progressive astrogliosis in the cerebellum of MPS IIIB dogs. Across all age groups, GFAP immunostaining was consistently elevated in MPS IIIB dogs compared to the unaffected controls. However, the high inter-animal variability precluded quantitative analyses. However, even unaffected animals at all ages displayed a substantial number of GFAP-positive Bergman glia within most folia, although these cells displayed relatively thin processes (Fig. 8, Bergman glia). This was accompanied in unaffected dogs by relatively little staining of fibrous astrocytes within the cerebellar white matter (Fig. 8, white matter). In contrast, the pattern and intensity of GFAP immunostaining was markedly different in MPS IIIB dogs at all ages examined (Fig. 8, grey), and became more complex with disease progression. Bergman glia in MPS IIIB dogs displayed much thicker processes than in corresponding age-matched unaffected controls (Fig. 8), and a distinctive band of GFAP immunostaining was evident at the interface between granule and molecular layers, marking where Purkinje cells normally sit (Fig. 8, Purkinje layer). This reached its peak intensity in MPS IIIB dogs by 15 months of age, before declining in older MPS IIIB dogs (Fig. 8). This was accompanied by progressively more GFAP immunostaining with increased age within both the granule cell layer, and perhaps most notably within the white matter and   (Fig. 8). Increased GFAP immunostaining within the white matter was already visible in MPS IIIB dogs from 9 months of age, revealing a distinctive outlining of astrocyte processes around blood vessels within the cerebellar white matter (Fig. 8). These results suggest progressive changes in astrocyte activation in MPS IIIB dogs, specifically around the Purkinje cell layer and cerebellar white matter.
Microgliosis is most prominent in the DCN in cerebellum of MPS IIIB dogs. Iba1 immunoreactivity was much less variable than GFAP staining and revealed changes in morphology and increased Iba1 immunostaining in MPS IIIB dogs compared to the unaffected controls from 9 months of age, and were more pronounced at later stages ( Fig. 9). Compared to the lacy appearance of the neuropil in unaffected dogs, Iba1positive microglia in MPS IIIB dogs displayed fewer, thicker processes and had enlarged cell soma. A distinctive appearance of very large microglia and/or clusters of intensely stained Iba1-positive cells were seen sporadically within both granule and molecular layers of MPS IIIB dogs from 9 months of age. Compared to astrocytosis, microglial activation was less pronounced within the white matter of MPS IIIB dogs, but was still markedly higher than in unaffected dogs, with localised activation evident in the DCN (Fig. 9). Quantification of Iba1 immunoreactivity via thresholding image analysis revealed some inter-animal variability. In particular, one of the 9-month-old MPS IIIB brains had markedly increased Iba1 immunostaining than any other brain of the same age. However, there was a significant increase in Iba1 immunoreactivity in white matter of MPS IIIB dogs at > 20 months of age (Fig. 9D), with a trend starting at 15 months of age. A similar trend in relative Iba1 immunoreactivity was seen in the DCN of MPS IIIB dogs, but not in their cerebellar cortices in either the molecular or granular layers. These results indicate, similarly to the atrophy, much of the cerebellar microgliosis occurs in the white matter. While there were obvious changes in microglial morphology in the cerebellar cortical layers of MPS IIIB dogs, significant changes in Iba1 immunoreactivity were detected only in the white matter and the DCN.

Loss of neurons and white matter atrophy in the cerebellum of MPS IIIB dogs. Purkinje cells
and neurons in the DCN provide the major efferent output from the cerebellum, indicating their importance for cerebellar functions. Loss of these cells was analyzed by Nissl staining and the sections were used to evaluate white matter volume loss and cerebellar cortical thickness (Fig. 10). The thickness of cerebellar molecular and granular layers was measured in the posterior lobe of the cerebellum both separately and together, but revealed no significant change in the thickness of any cerebellar cortical layer in MPS IIIB dogs at any age (data not shown). Cerebellar white matter volume changes were apparent in MPS IIIB dogs from 15 months of age, but became statistically significant at > 20 months of age when the volume was decreased by more than 50% (Fig. 10A). There was an increase in white matter volume in unaffected dogs over time, potentially due to ongoing myelination during postnatal maturation, which was not evident in MPS IIIB dogs, resulted in a significantly smaller volume in the oldest MPS IIIB dogs (Fig. 10A). A marked decrease in Purkinje cells (15-20%) was www.nature.com/scientificreports/ observed in the posterior lobe of the cerebellum of MPS IIIB dogs at all ages when compared to age-matched unaffected controls. Though this did not reach statistical significance, the trend in Purkinje cell loss with disease progression was evident (Fig. 10B). Optical fractionator counts of Nissl-stained DCN neurons showed a pronounced loss of these cells in MPS IIIB dogs by 9 months of age, which then remained stable at later stages of disease (Fig. 10C). Compared to unaffected controls, the loss of DCN neurons in MPS IIIB dogs became statistically significant after 15 months of age. However, compared to 3-month-old MPS IIIB dogs, the loss of DCN neurons was statistically significant from 9 months onwards (Fig. 10C). Taken together, these data indicate that cerebellar atrophy occurs primarily in the white matter, and that there is progressive loss of DCN neurons that has not been reported previously.

Discussion
In this study, we assessed the development of disease pathology in the canine model of MPS IIIB before they display any signs, by evaluating biochemical and histopathological measures of lysosomal storage accumulation, neurodegeneration, and atrophy.
The changes in levels of both HS and MPS IIIB-specific HS-NRE were evaluated in cortical grey matter, CSF, plasma, and urine from MPS IIIB dogs. Even by 1 month of age, HS and HS-NRE levels in tissues and body fluids from MPS IIIB dogs were higher than the levels in the corresponding samples from unaffected dogs. However, the kinetics of HS and HS-NRE accumulation differ across the four tissues or body fluids. While levels of HS and HS-NRE in cortical grey matter of MPS IIIB dogs increase over time, levels of HS and HS-NRE in CSF and urine of MPS IIIB dogs were higher at 1 month of age than at subsequent time points. Despite this early decrease, levels of HS and HS-NRE in CSF and urine of MPS IIIB dogs at these later time points remain elevated in comparison to the levels present in unaffected dogs of any time point. High levels of urinary GAGs have been reported previously in healthy newborn humans and have been suggested to be associated with increased turnover of extracellular matrix during development [22][23][24] . Elevated levels of HS and HS-NRE in the CSF and urine of 1-month old MPS IIIB dogs could be indicative of the ongoing brain growth and increased turnover of extracellular matrix that occurs during this time period 25 . Further studies that include additional WT and MPS IIIB dogs at younger ages will be needed to fully understand the mechanism driving the early elevated levels of HS and HS-NRE in MPS IIIB dogs. Any future studies of potential therapies that initiate prior to HS and HS-NRE levels reaching steady state in the CSF of MPS IIIB dogs will need to distinguish between the developmentally associated decrease in HS levels, and that due to any therapeutic benefit at these early time points.
Disease-related pathology illustrated by increased intensity of LAMP1 immunostaining was observed in multiple brain regions of preclinical MPS IIIB dogs and was accompanied by a progressive increase in microgliosis (cortex and cerebellum) and astrogliosis (cerebellum). The distribution of these changes is similar to that of histological changes seen in the brain of an MPS IIB patient 26 , suggesting that the dog model recapitulates this aspect of the human disease.
Pathological evaluation of the cerebellum confirmed it to be the most severely affected brain region in preclinical MPS IIIB dogs, displaying neuropathological changes that are evident relatively early in disease progression. One of the most prominent and previously unreported findings within the cerebellum was a substantial 50% loss of DCN neurons. This loss occurred relatively early in disease progression (present by 9 months of age) and did not appear to progress further, indicating that the majority of DCN neurodegeneration occurs in early stages of the disease. As such, it may be important to look in patients for symptoms that could potentially be related to this DCN loss. There was also significant cerebellar white matter atrophy in preclinical MPS IIIB dogs, which had www.nature.com/scientificreports/ almost half the volume compared to unaffected dogs. This may be due to axon loss or a demyelination process resulting from a degeneration or dysfunction of oligodendrocytes or defective myelination during post-natal development. However, we observed an increase in white matter volume in unaffected dogs suggestive of ongoing myelination during this time in development. The degree of Purkinje cell loss in these MPS IIIB dog samples was relatively small and not significantly different from controls, as was previously reported in older animals 17 . Quantitative analysis revealed clear signs of increased storage burden in cerebellar cortical layers of MPS IIIB dogs in addition to DCN and cerebellar white matter. Taken together these observations indicate that the DCN are the primary pathological target within the MPS IIIB cerebellum, exhibiting pronounced neurodegeneration and atrophy of surrounding white matter tissue. The time-dependent accumulation of HS in the forebrain and progressive increase in LAMP1 staining in both the forebrain and cerebellum in MPS IIIB dogs indicate that MPS IIIB disease pathology clearly occurs prior to the onset of clinical signs in animals. Such signs only occur subsequent to 20 months of age, with all but the most severely affected dogs examined in this study being younger than this. While disease-related pathology is present throughout the brain, the cerebellum is more dramatically affected than other regions in preclinical MPS IIIB dogs. Although the loss of Purkinje neurons has been documented previously in the schipperke MPS IIIB dogs 17 , our current findings provide the first evidence for pronounced pathology within deeper structures of the cerebellum that include the white matter and DCN.
While human MPS IIIB disease is thought to primarily affect cortical brain regions, the extent to which the cerebellum is affected and plays a role in the disease symptoms has not been thoroughly investigated. Interestingly, early stage MPS IIIB disease shares many symptoms with autism despite the different etiologies and MPS IIIB is often misdiagnosed as autism 27 . Similar symptoms include motor dysfunction, aggressiveness, hyperactivity, and language and communication deficits 27 . In addition to changes in brainstem volume and cortical regions, autism is associated with cerebellar atrophy and dysfunction, which may cause motor dysfunction as well as some non-motor symptoms such as deficits in language, communication, social interactions, stereotyped behavior, and cognitive dysfunction 28,29 . Similar symptoms are observed in MPS IIIB patients, suggesting that the cerebellar pathology seen in the MPS IIIB dogs may have relevance for early stages of human MPS IIIB disease. Later stages of the human MPS IIIB disease, which ultimately result in a vegetative state, are likely to involve more widespread neurodegeneration.
Our study reveals the dynamics of the onset and progression of neuropathology in this large animal model of MPS IIIB disease, before the onset of disease signs. We describe a series of markers that can be used for assessing therapeutic benefit in future studies, focusing on well-defined neuropathological and biochemical endpoints including the direct measure of NAGLU substrate accumulation (HS and HS-NRE), Iba1 as a marker of microglial activation, and LAMP1 as a marker of lysosomal enlargement.

Materials and methods
Dogs and tissue extraction. The outbred MPS IIIB canine colony was founded based on a naturallyoccurring mutation in schipperke dogs 17 . Animals were genotyped by 7 days old with polymerase chain reaction at the canine NAGLU locus, and enzyme deficiency was confirmed by testing their serum enzyme levels 30 . The MPS IIIB dogs and their unaffected counterparts were produced and maintained at Iowa State University's Department of Animal Science according to guidelines of the USDA and NIH, and under approved protocols of the ISU IACUC. Animals were kept with their dams until weaning, and at 4 months segregated by sex. After weaning, dogs were fed ad libitum. CSF was aseptically collected under full anaesthesia by puncture of the cisterna magna after which dogs were euthanized by intravenous administration of sodium pentobarbital at approximately 1, 2, 3, 6, 9, 12, 15, and 20 months of age, after which encephalectomy was performed by Iowa State University Department of Animal Science staff. The numbers of individual dogs analysed at each age is summarized in Table 1 (SensiPro) and Table 2 (histology). Brains were cut sagittally into hemispheres at the level of the longitudinal fissure, with one half immersion fixed in 10% neutral buffered formalin and the other half frozen and stored at − 80 °C. Tissue samples for HS analysis were derived from frozen hemispheres, which were placed at − 20 °C overnight, with tissue samples dissected directly from these − 20 °C frozen brains. Cortical grey matter was taken from the medial aspect of the precruciate gyrus. Urine and blood samples were collected pre-mortem (serum) or immediately post-mortem (urine).
Quantification of total and disease-specific heparan sulfate. Total HS and MPS IIIB-specific HS-NRE in homogenized brain tissue (cerebral grey matter), CSF, serum and urine from approximately 1, 2, 3, 6, 9, and 12-month-old unaffected or MPS IIIB dogs (exact ages of dogs for whom brain tissue, CSF, plasma, and urine were analyzed varied by sample availability) were quantified using the HS-NRE liquid chromatographymass spectrometry (LCMS) assay (SensiPro) as described previously 2,10,31 . Briefly, HS from the samples was purified by anion exchange chromatography and digested with heparan lyases (IBEX Technologies). The HS fragments were tagged with isotropically labelled aniline by reductive amination and quantified by LCMS. Total HS (internal disaccharides) and the MPS IIIB-specific HS-NRE (N-acetylglucosamine-2-sulfated iduronic acid-N-sulfated glucosamine [GlcNAc-IdoA2S-GlcNS], referred to as A0I2S0) are expressed as pmol/mg brain wet weight, pmol/0.6 μl CSF or plasma, or pmol/nmol creatinine in urine.
Histological preparation of brain tissue. Brains were stored in formalin for extended periods before histology was performed. Before beginning histology, brain hemispheres were submerged in cryoprotectant, 30% sucrose in Tris-buffered saline (TBS: 50 mM Tris, 150 mM NaCl, pH 7.6) for at least two weeks. The forebrain was then separated from the cerebellum and cut perpendicular to the long axis of the brain into three rostrocaudal blocks that were sectioned individually. Blocks were coated with Lipshaw M-1 cryomatrix resin Nissl staining. In order to visualize tissue cytoarchitecture, regularly spaced sections were Nissl stained with cresyl violet, as described previously 32 . Briefly, every forty-eighth 50 µm section through each block was mounted onto microscope slides previously double coated with gelatin-chrome allum (ThermoFisher Scientific). After air drying for at least 24 h, sections were incubated in 0.025% cresyl violet (Merck, Darmstadt, Germany) and 0.05% acetic acid (VWR, Dorset, UK) at 56 °C for 30 min. Sections were then differentiated through a series of graded concentrations of industrial methylated spirits (IMS, VWR) and cleared in xylene (VWR) for at least 24 h before coverslipping with DPX (VWR), a xylene-based mountant.
Immunohistochemical staining. To assess lysosomal storage and microglial activation in the forebrain and cerebellum, regularly spaced sections were selected, and immunohistochemically stained, using standard protocols 33 , for the markers LAMP1 (lysosomal membrane protein 1, lysosomes) and Iba1 (ionized calcium binding adaptor molecule 1, a calcium binding protein specific to microglia and macrophages), in addition to GFAP (glial fibrillary associated protein, astrocyte activation) in the cerebellum.

Cortical thickness.
To assess potential differences in cortical thickness, Nissl-stained sections were measured perpendicularly from the pial surface to white matter border, and this thickness was analyzed with StereoInvestigator software (MBF Bioscience Inc., Williston, VT, USA). These measurements were made in the primary somatosensory cortex, the rostral suprasylvian gyrus; this region was identified according to neuroanatomical landmarks 34 . Three sections containing this region were selected from each brain, and slides were blinded to age www.nature.com/scientificreports/ and treatment before analysis started. Ten measurements of cortical thickness were made per section, and these measurements were made in areas as far from gyrus curvature as possible to minimize variability. The mean regional cortical thickness was then calculated for each brain.

Stereological counts of neuron number.
Optical fractionator counts of the number of Nissl-stained DCN neurons were made using an appropriate grid size and dissector frame. Only neurons with a clearly identified morphology that fell within the dissector frame were counted. For all optical fractionator estimates, the mean coefficient of error of individual estimates was calculated according to the method of Gundersen and Jensen 35 and was less than 0.08 in all these analyses. The number of Purkinje neurons was counted manually and expressed as the number per unit length within the same folium in each brain.
Measurements of microglial area. The mean size of Iba1-stained microglia in the forebrain was obtained by measuring the cross-sectional area of at least 100 morphologically identified Iba1-stained microglia in each animal using StereoInvestigator software, and the mean microglial area determined ± standard area of the mean (SEM).
Thresholding image analysis. The expression of GFAP, Iba1, and LAMP1 was measured in the cerebellum by semi-quantitative image thresholding analysis of each antigen. From three consecutive immunostained sections, 10 non-overlapping images were captured from each section for each of the areas examined. All images were captured at 40X objective by a live video camera (JVC, 3CCD, KY-F55B) mounted on Zeiss Axioplan universal microscope (Germany). The lamp intensity, video camera setup, and calibration were kept constant throughout the image capturing process. Image-Pro Plus 5.0 software (MediaCybernetics, Chicago, IL, USA) was used to analyze the images with an appropriate threshold selected to distinguish the foreground immunostaining above the background. Macros were recorded to transfer the data to a spreadsheet for statistical analysis. The data obtained from the thresholding analysis was plotted graphically as a mean percentage area of immunoreactivity per treatment group ± SEM.
Statistical analysis. All statistical analyses were performed blinded to genotype and age, which was only revealed once all morphological analyses were completed. A Kolmogorov-Smirnov test, a nonparametric test which assesses the empirical distribution functions of two samples, was applied to measured microglial cell body areas (GraphPad Prism, La Jolla, CA). A t-test was used to determine whether there was a statistically significant difference between genotype groups in cortical thickness of somatosensory cortex. Results were considered statistically significant if they produced a p value less than or equal to 0.05.