The GM2 gangliosidoses are progressive neurodegenerative disorders due to defects in the lysosomal β-N-acetylhexosaminidase system. Accumulation of β-hexosaminidases A and B substrates is presumed to cause this fatal condition. An authentic mouse model of Sandhoff disease (SD) with pathological characteristics resembling those noted in infantile GM2 gangliosidosis has been described. We have shown that expression of β-hexosaminidase by intracranial delivery of recombinant adeno-associated viral vectors to young adult SD mice can prevent many features of the disease and extends lifespan. To investigate the nature of the neurological injury in GM2 gangliosidosis and the extent of its reversibility, we have examined the evolution of disease in the SD mouse; we have moreover explored the effects of gene transfer delivered at key times during the course of the illness. Here we report greatly increased survival only when the therapeutic genes are expressed either before the disease is apparent or during its early manifestations. However, irrespective of when treatment was administered, widespread and abundant expression of β-hexosaminidase with consequent clearance of glycoconjugates, α-synuclein and ubiquitinated proteins, and abrogation of inflammatory responses and neuronal loss was observed. We also show that defects in myelination occur in early life and cannot be easily resolved when treatment is given to the adult brain. These results indicate that there is a limited temporal opportunity in which function and survival can be improved—but regardless of resolution of the cardinal pathological features of GM2 gangliosidosis, a point is reached when functional deterioration and death cannot be prevented.
The GM2 (monosialoganglioside 2) gangliosidoses, Tay–Sachs disease (TSD; OMIM 272800), Sandhoff disease (SD; OMIM 268800) (1,2) and GM2 activator protein deficiency (OMIM 272750) (3), are a heterogeneous group of rare neurodegenerative lysosomal storage diseases (LSDs), transmitted in an autosomal recessive fashion. They result from defects in the enzyme, β-N-acetylhexosaminidase (EC 22.214.171.124) and GM2 activator protein. Absence or malfunction of these proteins leads to lysosomal accumulation of their substrates (glycosphingolipids, glycoproteins and glycosaminoglycans) with devastating effects on the nervous system (4,5). There are three isoforms of β-N-acetylhexosaminidase: β-hexosaminidase A (HEX A), a heterodimer of α and β subunits; β-hexosaminidase B (HEX B), a β subunit homodimer; and β-hexosaminidase S (HEX S), an α subunit homodimer (6–8). Mutations in either of the genes (HEXA and HEXB) encoding the subunits or GM2 activator protein (GM2A) can cause disease which has a very similar clinical course. The infantile form of GM2 gangliosidosis is the most common and invariably fatal, but attenuated adult forms also occur (9).
The SD mouse model was generated by targeted disruption of the Hexb gene (10). The mice appear indistinguishable from their littermates for the first 3 months of life, but thereafter a fulminant neurodegenerative disease develops, with death occurring at 4–5 months of age. Stereotypic disease manifestations, which include spasticity, muscle weakness, rigidity, tremor and ataxia, closely resemble those seen in the acute form of human GM2 gangliosidosis. The SD mouse has become invaluable for testing prospective therapies, and to interrogate underlying pathophysiological mechanisms of disease.
The genetic basis of GM2 gangliosidosis is well established, but the molecular events leading from disease-causing mutation to the clinical manifestations remain obscure. Post-mortem examination of brains from patients suffering from GM2 gangliosidosis has revealed marked variation in neuronal cell density and the extent of gliosis (11).
Experimental evidence suggests that abnormal accumulation of sphingolipids (SLs) disturbs endosomal transport and sorting (12). Defective degradation of SLs is often accompanied by excessive formation of the cognate deacylated molecule, to which additional biological effects are often attributed (13). In human GM2 gangliosidosis, lyso-GM2 is found in excess and although the contribution of this species to the development of the disease is unclear, it is presumed to be damaging to neural tissue (14).
One of the pathological hallmarks of neurodegenerative disease in GM2 gangliosidosis is the development of an inflammatory response. Wada and colleagues reported that activation of microglia preceded acute neurodegeneration (15). However, the question remains as to whether this response is a primary or secondary contributor to the pathogenesis.
Phospholipid (16) and myelin lipid mass are reduced in human and animals models of GM2 gangliosidosis (11,17), but again it is unclear whether the observed reduction is a consequence or a cause of neurodegeneration.
Successful use of recombinant adeno-associated vectors (rAAV) encoding β-hexosaminidase in the treatment of severe GM2 gangliosidoses in mice (18,19) and more recently cats (20), clearly mandates translation of this approach to patients with Tay–Sachs and SDs. However, since patients at different stages of evolution of the diseases will ultimately be potential candidates for this intervention, it is imperative to determine the extent to which the disease can be arrested or reversed and how much functional recovery might be accrued by gene transfer carried out at different points in the course of the illness. Here we present data from experimental studies in which these questions are addressed using rAAV vectors to deliver complementing β-hexosaminidase in the brain of SD mice.
Timing of rAAV injections determines extent of preservation of function and survival
We compared therapeutic outcomes of intracranial administration of rAAV2/1α + rAAV2/1β in four cohorts of SD mice. Mice were injected bilaterally into the striatum and cerebellum at 4, 8, 10 or 12 weeks (w) of age. Controls were untreated SD [SD (UT)] and heterozygous or wild-type (normal controls) animals. All mice in the study were checked and weighed daily and killed when they reached their humane end point.
Survival was analysed with the log-rank (Mantel–Cox) test. Intracranial rAAV-mediated gene delivery of β-hexosaminidase had a marked effect on the lifespan of this acute mouse model of GM2 gangliosidosis (P < 0.0001), but critically, it depended on age at the time of intervention as shown by the Kaplan–Meier (Fig. 1A) and scatter plots (Fig. 1B). Survival of the different age-treated groups was examined by one-way analysis of variance (ANOVA) and adjusted for multiple post hoc comparisons by the Bonferroni's method. Median survival was 730 days for untreated normal controls (n = 22), and 131 days for mutant animals [SD (UT)] (n = 37). In contrast, median survival of groups of SD mice treated at four different ages was: 4w (615 days, n = 22), 8w (233 days; n = 5), 10w (292 days; n = 5) and 12w (126 days; n = 12). Greatest longevity was attained by the group treated at the youngest age (4w) (P < 0.0001). Importantly, life was also extended when the treatment was administered at 8w (P < 0.001) and 10w (P < 0.0001), but not at 12w (P > 0.05). Although, no difference in survival between SD mice injected at 8w and 10w was noted (P > 0.05), the evolution of neurological disease signs differed between the two groups. While animals treated at 8w exhibited a delay in disease onset with signs that developed slowly over time, those injected at 10w had disease onset that resembled that of SD (UT). Hence, to secure long-term functional rescue in a disease with such devastating effects on the entire nervous system, treatment should be given as early as possible. Notwithstanding, that benefit can be accrued even when treatment is substantially delayed was revealing. As predicted, a point occurs when the putative treatment improves neither function nor survival of the recipient (Supplementary Material, Videos S1–S5).
Abundant and widespread expression of human β-hexosaminidase after intracranial injection of viral vectors
In situ β-hexosaminidase activity was detected by histochemical staining and in transduced cells by in situ hybridization (ISH) with an antisense RNA probe against the woodchuck hepatitis virus post-transcriptional regulatory element (WPRE)-bovine growth hormone polyadenylation signal (BGHpA) viral sequence. Analysis of the brain and spinal cord of three animals was performed for each experimental group.
Injections into the striatum and cerebellum induced widespread distribution of virus and enzyme in all age groups tested. Consistent with our previous studies, staining for β-hexosaminidase and viral RNA was most intense at the injection site (Fig. 2A–I), and interconnected areas such as the substantia nigra (SN)—presumably labelled by retrograde transport as shown in a 12w injected mouse (Fig. 2J–L). Viral dispersion was also noticeable along white matter tracts such as corpus callosum (Fig. 2C, F and I), and choroid plexus by leakage and/or diffusion of infusate into the ventricular system (Fig. 2F and I). The overall pattern of β-hexosaminidase was similar between all age groups, but the most abundant staining was consistently seen in the 12w injected mice (Fig. 2G and H).
Periodic acid-Schiff reagent stain, α-synuclein and ubiquitin inclusions are cleared by expression of β-hexosaminidase
The brains and spinal cords of three animals from each age group were stained with periodic acid-Schiff reagent (PAS) to detect stored glycoconjugates, a hallmark of GM2 gangliosidoses. α-Synuclein aggregates were revealed by immunohistochemistry (IHC) with two antibodies, one monoclonal and the other polyclonal. Ubiquitin-containing inclusions were also detected by IHC.
Pathological amounts of GM2 ganglioside and related glycoconjugates, which have been reported elevated as early as in fetal life, appear to increase linearly over time, and by age 4w are easily visible by PAS staining (Supplementary Material, Fig. S1). All SD-treated animals analysed at their humane end point, including those injected at 12w, showed undetectable or, if at all present, small pockets of PAS-positive material throughout the neuraxis (Fig. 3). This finding shows that expression of β-hexosaminidase rapidly clears the stored material and restores apparent normal histology.
On the basis of reports that aggregates of the pre-synaptic protein α-synuclein are present in the neuronal soma of SD (UT) mice, we investigated whether gene transfer can clear and/or prevent its accumulation. The two antibodies used for IHC revealed an identical pattern of α-synuclein staining in SD (UT) animals that increased in intensity over time (data not shown). However, when brain and spinal cord tissue extracts from animals of different ages were analysed by western blots with the same antibodies, the amount of monomeric α-synuclein in SD (UT) samples was decreased when compared with normal controls (Supplementary Material, Fig. S2). IHC-staining with the pre-synaptic protein, synaptophysin, showed a normal pattern of staining in SD (UT) mice, and by western blotting no discernible differences were observed between SD (UT) and normal controls (data not shown). As shown in Figure 4A–D, gene transfer is highly efficient at normalizing the levels of α-synuclein in all age groups tested.
Inclusions-containing ubiquitinated proteins have been detected in several glycosphingolipidoses, and to test whether gene transfer can reverse this abnormality, we first established the spatial distribution of ubiquitinated protein inclusions in SD (UT) mice. In contrast to the staining of most neurones by PAS and with antibodies against α-synuclein, ubiquitin aggregates were detected only at a few anatomical sites in SD (UT) mice; staining was weak in thalamic nuclei, deep cerebellar nuclei, pons and medulla, but ubiquitin aggregates were prominent in the grey matter of the spinal cord (Fig. 4I). In animals examined at their humane end points and treated at ages 4w, 8w and 10w, this pathological feature had resolved completely (Fig. 4K), while those injected at 12w had partial clearance—the intensity of staining was reduced compared with that in SD (UT) mice, but the number of cells containing inclusions was similar (Fig. 4J). In contrast, glycoconjugate accumulation was cleared by treatment (Fig. 4N and P), compared with SD (UT) (Fig. 4M) and normal controls (Fig. 4P).
Gene transfer ameliorates neuroinflammation in SD mice
Innate inflammatory responses are characteristic of most neurodegenerative disorders and a widely held view is that, rather than containing the cytological injury, proliferation and activation of microglia and astroglia may accelerate the disease process.
We investigated possible regional differences in microglia and astroglia proliferation/activation in mouse models of SD, and demyelinating, Krabbe disease (KD; twitcher mouse). We compared the inflammatory response in SD (UT) and twitcher, at their humane end point of 4 months and 38 days of age, respectively, to age-matched normal controls. Abundance of mRNA encoding chemokines Mip-1α and Rantes and microglial and astroglial markers, Cd68 and gfap, respectively, was determined by real-time PCR (relative to β-actin). Spinal cord and brain were analysed separately. The brain was dissected into olfactory bulb, cerebrum (neocortex, hippocampus, thalamus and hypothalamus), and cerebellum and brain stem (superior and inferior colliculus, midbrain, pons and medulla). Twitcher and SD (UT) mice differed in their inflammatory response. Particularly, striking was the difference in Rantes expression between the two models; while Rantes is highly expressed in twitcher mice, the levels in SD (UT) tissue are similar to those noted in normal controls (Fig. 5A and B). In both models of disease, up-regulation of markers of gliosis is more prominent in hindbrain and spinal cord than in forebrain, coincident with white matter-rich regions (Fig. 5A and B).
SD (UT) mice develop a stereotypic disease with onset at ∼13w age and tremor as the first noticeable sign of neurodegeneration. To correlate expression of markers of gliosis with disease progression, we measured mRNA levels of Cd68 and Gfap (relative to β-actin) by real-time PCR at different time points in spinal cord of SD (UT) mice. In agreement with previous reports, gene expression was elevated at 1 and 2 months of age, with the greatest increase occurring at age 13w, concomitant with the beginning of the symptomatic phase. Thereafter, the levels remained elevated until the animals reached the humane end point (Fig. 5C and D).
We next examined effectiveness of gene transfer at reducing neuroinflammation when treatment is delayed. Because detailed anatomical differences are best revealed by IHC, staining patterns of Cd68 in brain and spinal cord sections of 4w and 12w injected SD animals were compared with mutant and normal controls. SD mice injected at 4w (Fig. 6G–I) had stained microglia numbers similar to normal control mice (Fig. 6J–L) and with morphology consistent with the resting state with respect to ramified spindly processes emanating from a central cell body (Fig. 6O), as opposed to the amoeboid phagocytic state of microglia in SD (UT) (Fig. 6A–C, M and P). The 12w injected mice also harboured microglia numbers and morphology similar to normal controls, except for a few scattered cells in the VPM/VPL (ventroposterior medial and lateral) nuclei in the thalamus, and white matter of the spinal cord (Fig. 6D–F).
Neurodegeneration and death is prevented by treatment
To correlate the emergence and progression of disease signs to particular anatomical regions, we studied the temporal and spatial pattern of neurodegeneration in brain and spinal cord of mutant Sandhoff mice using the chemical-development-silver method of Gallyas (21). This method detects components of neurons undergoing degeneration, such as lysosomes, axons and terminals that bind silver ions with high affinity.
SD (UT) and age-matched normal control littermates were studied at 2, 5, 8, 10w, 12w and 17w (4 months) age. Metallic grain deposition increased with age (Fig. 7A–D). It was readily observed as early as age 2w in mossy fibres of the hippocampus (Fig. 7A), and neuronal processes in deep cerebella nuclei. In contrast to ramified microglia that occupied surrounding structures, amoeboid microgliosis was associated with silver staining at these intensely stained sites (Fig. 6A and M). By age 5w, labelling became apparent in brain sensory pathways, including olfactory and optic tracts, auditory lateral lemniscus, lamina 1 of spinal cord and gracile nucleus. Notably, there was no obvious staining of large white matter tracts (Supplementary Material, Fig. S3). At 12w, silver deposition was stronger in grey matter than in white matter; and by the end stage of the disease at 17w, the entire brain and spinal cord became heavily stained, with many abnormally swollen bulbous axons (axonal spheroids), typical of advanced axonal degeneration (Fig. 7H). A conspicuous exception was the lack of silver deposition in some cranial nerves, such as the facial nerve (Supplementary Material, Fig. S4U–X).
Consistent with the progressive nature of GM2 gangliosidosis, we found increased metal deposition with advancing age. However, we also learned that neurodegeneration starts early, by age 2w is already detectable, months before disease onset becomes apparent (at ∼13w age), and that the time course of degeneration differs between anatomical regions. In addition, while some structures are particularly vulnerable, others appear to escape degeneration altogether.
We compared the effect of gene transfer given at age 5w and 12w on amelioration of neurodegeneration. Two animals per group were sacrificed at age 17w and processed for silver staining. Whereas mutant mice injected at 12w reached their humane end point at 17w and developed stereotypic signs of disease, those infused at 5w had an apparent normal phenotype. Their brains and spinal cords were removed and examined in detail. As illustrated in Figure 7, the striking observation was that even when given late in the development of the disease, treatment could prevent and/or reverse neurodegeneration. Prominent grain deposition remained in lamina 1 of spinal cord and gracile nuclei in both age groups. In mice injected at the latter time point, the internal capsule and white matter of the cerebellum had reduced but significant residual staining (Supplementary Material, Fig. S4).
We have previously reported the absence of gross neuronal loss in murine GM2 gangliosidosis, one exception being the VPM/VPL nuclei of the thalamus, which lose neurones progressively. We have recently reported a combined VPM/VPL neuronal loss, compared with age-matched normal controls, of ∼15, 37 and 50% by the ages of 2, 3 and 4 months, respectively. We have also showed that gene transfer when given at age 4–5w could prevent neuronal loss (22).
Using the VPM/VPL nuclei as a paradigm of neuronal rescue by gene transfer, we studied lifelong preservation of neuronal density in long surviving animals, and prevention of further loss when treatment is delayed. NeuN-stained cells in the VPM/VPL of SD mice, injected at 4w, 8w, 10w and 12w and killed at their humane end points were counted (Fig. 8A–F). Controls were SD (UT) mice killed at the humane end point of ∼4 months, and normal control mice at ages 4 months–2 years. Mean (±SEM) neuronal numbers were: normal controls, 895 ± 13; SD (UT), 533 ± 72; injected at 4w: 1027 ± 32, 8w: 806 ± 33, 10w: 725 ± 74 and 12w: 464 ± 14. In this study, neuronal numbers had fallen by 41% in SD (UT) at the humane end point (P < 0.01, Bonferroni post hoc test), relative to normal controls. In contrast to animals injected at 4w (age: 337–730 days) whose cell density was similar to normal controls, those infused at 12w did not significantly differ from SD (UT) (P > 0.05, Bonferroni post hoc test). Although the Bonferroni's post hoc test for the 8w and 10w treatment groups compared with normal controls returned P > 0.05, there was a downward trend in cell numbers; a 10 and 19% decline in neuronal cell counts was noted, respectively. Taken together, the results indicate that rescue by gene transfer was lifelong and that delayed treatment could still prevent further deterioration. However, when treatment is given just before the onset of signs of disease (12w), cell death is not preventable (Fig. 8G).
Myelin defects are present early in the course of SD and persist after treatment
We concluded from the results described above that delayed treatment had a marked ameliorating effect on classical pathological biomarkers of disease in the mouse model of GM2 gangliosidosis. However, despite these improvements, lifespan and neurological function was not rescued when treatment was administered late in the course of the disease.
In a quest for other pathological changes that might explain why the time of intervention is critical for preservation of function and to increase survival, we first investigated the integrity of myelin in SD (UT) mice. Using the classic demyelinating twitcher mouse as a control, we studied the pattern of myelin gene expression in SD (UT) (n = 3) and twitcher (n = 3) mice at their humane end point of ∼4 months and 38 days of age, respectively. Abundance of mRNA of the myelin markers mag, plp and cgt (relative to β-actin) was analysed by real-time PCR. Spinal cord and dissected brain were measured separately. Moribund SD (UT) and twitcher mice showed a consistent pattern of reduced expression of myelin markers in all brain regions and spinal cord, when compared with normal controls (Fig. 9A and B). To explore whether the derailment of myelin integrity was limited to the terminal phase of the disease, expression of cgt mRNA was quantified in SD (UT) spinal cord at ages 5–18w. Unexpectedly, cgt gene expression was about half that of normal controls at all ages tested (Fig. 9C).
We next examined if myelin composition was altered by comparing myelin protein amounts of non-compacted cnpase (2′,3′-cyclic nucleotide 3′-phosphodiesterase; relative to tyrosine hydroxylase (th)) (Fig. 9D), and compacted, plp and mbp (proteolipid protein and myelin basic protein; relative to β-actin) (Fig. 10A and C) in cerebrum extracts from SD (UT) and normal controls. Densitometry measurements of cnpase, plp and mbp protein species in westerns blots showed that protein amounts are about half those of normal controls (Fig. 10B and D), hence confirming differences in myelin composition at the level of polypeptide translation as well as gene transcription. Although, as pointed out earlier, gross neuronal cell death that could account for the observed myelin reduction was not evident, to confirm that abundance of a neuron-specific marker was unchanged, β-tubulin III content was determined by western blotting (Fig. 10E). Densitometry measurements revealed no significant difference in β-tubulin III expression (relative to β-actin) between SD (UT) and brain tissue from normal control mice (Fig. 10F). These results suggest that early reduction in myelin protein composition occurs throughout the entire neuraxis in the absence of gross neuronal density loss.
We then tested whether myelin protein composition was preserved after treatment was given to mutant animals at the ages of 4w, 8w and 12w. Western blot analysis of plp, mbp, β-tubulin III, β-actin and synaptophysin protein content was performed on cerebrum extracts from three different animals at their humane end points: 425–539, 384–532 and 127–140 days for 4w, 8w and 12w injected SD, respectively (Fig. 11A, C and E). As controls, age-matched wild-type mice were killed at various time points (112–547 days). Densitometry measurements of plp and mbp species (relative to β-actin) showed that animals treated at 12w had a consistent protein loss of ∼60% compared with normal controls (P < 0.05, Bonferroni post hoc test) (Fig. 11B and D). Animals infused at the earlier time points of 4w and 8w had myelin protein content that was also reduced compared with age-matched normal controls, but there was considerable variation between animals within the same treatment group. To ascertain that reduction in myelin proteins was not caused by neuronal loss after surgery, we compared the relative amounts of neuron-specific proteins synaptophysin and β-tubulin III with normal controls, and found no significant differences between the groups (P > 0.05, Bonferroni post hoc test) (Fig. 11E and F). Although a larger cohort of animals needs to be studied to confirm these findings, the data suggest that development of an abnormal composition of myelin protein is an early event that is relatively refractory to correction by gene transfer.
The GM2 gangliosidoses Tay–Sachs and SDs are prototypical neurodegenerative lysosomal storage disorders caused by β-hexosaminidase deficiency; as a consequence, lysosomal degradation of glycosphingolipids and related glycoconjugates is impaired and accumulates progressively in lysosomes. It is believed that this prime pathology leads to neuronal dysfunction and disease manifestations. In the most severe infantile form, symptoms start a few months after birth with characteristic regression of achieved milestones, motor defects, seizures, together with visual and hearing loss.
Mice lacking functional β-hexosaminidase display several signs of disease that mimic the human condition. We have previously shown that many traits of the mouse phenotype, including motor function, histopathological features and premature death, can be prevented and/or reversed by expression of human β-hexosaminidase mediated by gene transfer, when delivered to the young adult mouse brain (18,19). This would suggest that absence of HEX A and B during brain development does not compromise brain function irreversibly. This notion is moreover, supported by findings in two inducible models of Tay–Sachs-related diseases; when transgenic Hexb is silenced at 5w age, disease progresses stereotypically as in the germline hexb knockout mouse, indicating the absence of developmental events modifying the course of the disease (23).
Often, the child with GM2 gangliosidosis is born to parents unaware of their carrier status; without widespread screening programmes in the general neonatal population, months and even years may pass before the first manifestations are recognized and diagnosed. As a result, patients present with diverse signs of disease and with varying degrees of disability: were gene therapy or other putative treatments with strong effects to become available, appropriate timing of the intervention is likely to be problematic. With a potential treatment now in the horizon, it is critically important to understand the nature of the neuropathological features, their evolution and extent to which they can be reversed in order to evaluate what benefits might be accrued from gene transfer delivered at differing stages of the disease. With these goals in mind, we co-infused monocistronic rAAV vectors expressing both subunits of human β-hexosaminidase, α and β, into the brains of SD mice at 4w, 8w, 10w and 12w age.
The single-stranded DNA genome of rAAV requires conversion to the biologically active double-stranded form through annealing of complementary plus and minus molecules in the cell nucleus, which occurs gradually over a period of ∼6w (24). The implication for our timed interventions is that the putative therapeutic effect will lag behind the injection time, and thus, we consider that in the SD mouse model the 4w, 8–10w and 12w injections broadly represent interventions at the asymptomatic, early symptomatic and late symptomatic phases of disease, respectively. Results here presented show that optimal function and survival is achieved when treatment is given during the asymptomatic phase, before signs of disease are evident. Importantly, the lifespan of animals injected at the early symptomatic phase was also markedly increased, and although eventually tremor and motor deficits became apparent, the animals remained capable of moving with ease around the cage, long after the onset of disease signs. In stark contrast, animals injected at the late phase of the illness showed progression of disease which was indistinguishable from that of untreated mutant mice.
Cabrera-Salazar and colleagues reported similar results for the neurodegenerative LSD late infantile neuronal ceroid lipofuscinosis (cLINCL). In the cLINCL mouse, early intervention was essential for enhanced therapeutic benefit, and motor function had limited recovery when treatment was started after disease onset (25). Brooks and colleagues described recovery of behavioural function after treatment was carried out once functional deficits were established in murine mucopolysaccharidosis type VII; the authors point out that restoration of function was possible because, as their data showed, neuronal impairment was not beyond repair (26). Similarly, Heldermon and colleagues have recently reported improved outcomes in the mouse model of Sanfilippo B when treatment is given neonatally (27).
The general consensus is that although the adult brain can show considerable structural plasticity for most inborn errors of metabolism with neurodegenerative effects, once a skill is lost, recovery is unlikely. Clearly, to secure long-term functional rescue of GM2 gangliosidosis, treatment should be given as early as possible. Nevertheless, it was important to discover not only that therapeutic benefit can be accrued even when rAAV infusions are delivered when the disease is clinically established but also that a point is reached when treatment does not translate into improved function or survival.
Our data demonstrate that viral transduction, transgene expression and bio-distribution are not hampered by the advancing process of neurological disease. The pattern of β-hexosaminidase staining was similar between all age groups tested, but the most abundant staining was consistently seen in the 12w injected mice. We attribute the difference to the short-time elapsed between viral infusion and killing of the animals, compared with mice injected at earlier time points. This might suggest loss of virus and/or enzyme activity over time. Alternatively, structural changes in the brain related to the pathological process itself might facilitate bio-distribution of viral particles and enzyme. The observation that virus can still be taken up and transported retrogradely when injected at age 12w is a strong indication that, in spite of severe disease already present at this age, axonal integrity and transport is not fully compromised. Concomitant with widespread β-hexosaminidase activity, stored glycoconjugates were rapidly cleared and apparent normal histology restored.
α-Synuclein is highly expressed in neurons and glia (28). It binds to a variety of proteins (29), lipid vesicles (30) and is involved in lipid metabolism (31). Progressive accumulation of α-synuclein is associated with the development of numerous neurodegenerative diseases, including Parkinson's disease, dementia with Lewy bodies, Alzheimer's disease, multiple system atrophy, multiple sclerosis and LSDs. Although the molecular mechanisms linking α-synuclein accumulation and disease manifestations are not fully understood, a known predisposing factor is increased intracellular amounts of the protein, caused by enhanced expression or reduced degradation (32). Mounting evidence indicates that faulty clearance of α-synuclein is due to impairment of one or other of its principal pathways of cellular degradation, the ubiquitin–proteosome system (33) and autophagy–lysosomal pathway (34). Recently, Suzuki et al. demonstrated histologically that the presence of α-synuclein aggregates in several of human lipidoses (35). In line with these findings, we detected extensive α-synuclein aggregation in many areas of the brain and spinal cord of SD mice. The aggregates are already evident at age 5w and increase over time, and free α-synuclein forms are depleted in these tissues. The contribution of α-synuclein pathology to the progression and disease manifestations of GM2 gangliosidosis remains to be defined, but it is conceivable that both phenomena, aggregation and reduced amounts of functional free α-synuclein, contribute to pathogenesis. Gene transfer at all ages tested was highly efficient at clearing α-synuclein accumulation from the tissues of SD mice. It has been assumed that α-synuclein accumulation is linked to glycoconjugate storage in neurones, but intriguingly Ashe and colleagues have shown that α-synuclein aggregation can be cleared by treatment with iminosugar-based glucosylceramide synthase inhibitors, while glycosphingolipids GM2 and GA2 levels related directly to the disease remain as high or are even higher than those in untreated animals (36).
Post-translational ubiquitin conjugation of proteins is a key regulator of sorting, trafficking, turnover of integral membrane proteins and the quality-control system that targets defective proteins to proteasomes or lysosomes for proteolysis (37). Inclusions-containing ubiquitinated protein aggregates have been detected in tissues of patients and animal models of (LSDs) (38). Bifsha et al. attributed ubiquitin accumulation in these diseases to reduced expression of ubiquitin C-terminal hydrolase, UCH-L1 (39). While Zhan viewed ubiquitin inclusions in LSDs as a non-specific epiphenomenon of no biological significance, Bifsha and colleagues suggest that concentrations of monoubiquitin below a critical level might undermine proteosomal activity. We report here ubiquitin inclusions localizing to only a few regions of the brain and spinal cord in the SD (UT) mouse, in contrast to neuraxis-wide staining of glycoconjugates. Gene transfer fully cleared the ubiquitin inclusions when SD mice were injected at 4w, 8w and 10 of age. Unlike the clearance of glycocongugates and α-synuclein, ubiquitin aggregates were only partially resolved when mice were injected with rAAV at age 12w. These results suggest that accumulation of ubiquitin might occur in a distinct subcellular compartment; alternatively, ubiquitinated proteins might be more resistant to degradation than glycoconjugates and α-synuclein. While the biological significance of ubiquitin inclusions in SD and other LSDs remains to be elucidated, we have demonstrated that they respond and are cleared by gene transfer expressing relevant disease proteins.
There is overwhelming evidence that an important component of all neurodegenerative diseases is the generation of an innate inflammatory response within the nervous system. Microglial and astroglial cells play a key role in the development and maintenance of this response, showing enhanced proliferation and activation. Several lines of evidence support the idea that reactive gliosis in these diseases negatively contributes to disease progression. Bone marrow transplantation in SD mice improved survival, ameliorated disease signs, and suppressed expansion of activated microglia in the absence of significant amounts of corrective enzyme or decreased amounts of ganglioside storage in the nervous system (15). Treatment of SD mice with the iminosugar-based glucosylceramide synthase inhibitor Genz-529468 increased lifespan, improved behavioural abnormalities and reduced inflammatory responses without restitution of the missing enzyme or reduction of gangliosides in brain (36). Similarly, deletion of the macrophage inflammatory protein Mip-1α in SD mice resulted in improved survival and function (40). The molecular mechanism/s underlying these observations is unknown, but the nature of the insult to the nervous system is believed to specifically modulate the inflammatory reaction. We compared the expression of markers of activated microglia and macrophage between SD and twitcher mice, classic models of neurodegenerative and demyelinating diseases, respectively. We found that in both models of disease up-regulated markers were more prominent in hindbrain and spinal cord than in forebrain, coincident with regions particularly rich in white matter, and the stronger reaction occurred in the twitcher. Particularly, striking was the high induction of the chemokine Rantes in twitcher mice, while it remained unaltered in SD mice. We attribute Rantes expression in the twitcher to the acute demyelinating disease characteristic of this model of KD, presumably caused by the toxic metabolite psychosine (41). The greatest induction of Rantes occurred in the spinal cord, and this correlates with prominent demyelination and infiltration of large perivascular multinucleated macrophages, known as globoid cells.
MIP-1α and Rantes promote inflammation by chemoattraction of specific subsets of haematopoietic cells; while Mip-1α attracts cytotoxic T cells and B lymphocytes (42), Rantes attracts monocytes, CD4+ and CD8+ lymphocytes (43) and increases the adherence of monocytes to endothelial cells. Of note, whereas Ohno reported T cell infiltration in the twitcher nervous system (44), Jeyakumar and colleagues could detect no CD4+ and only a small number of CD8+ lymphocytes in the thalamus of SD mice, in spite of a compromised blood–brain barrier (45). In the demyelination model of cuprizone intoxication, the blood–brain barrier remains intact and up-regulation of chemokines Rantes and Mip-1α coincides with the appearance of a few microglia and perturbation of oligodendrocytes, preceding oligodendrocyte apoptosis. Massive microglia and astrocyte recruitment accompanies demyelination. Absence of Mip-1α delayed demyelination, reduced microglia numbers and chemokine TNF-α in this disease model (46). Similarly, administration of Rantes antisera attenuated both macrophage infiltration and demyelination in mouse hepatitis virus-infected mice (47); and in experimental autoimmune neuritis, administration of anti-MIP-1α antibody suppressed clinical signs, and inhibited inflammation and demyelination (48). Jeyakumar found up-regulation of TNFα in SD mice with increasing age, but when they treated the animals with bone marrow transplantation or the iminosugar-based glucosylceramide synthase inhibitor NB-DNJ expression of TNFα, MHC class II and CD68 were drastically reduced.
We examined gliosis as a pathological correlate of disease progression in SD mice, paying particular attention to the last one and a half months of life, a period of precipitous pathophysiological events resulting in death. In agreement with findings by others (15,45), up-regulation of inflammatory markers was already detectable by 1 month of age, but the largest increase occurred at 13w, concomitant with the onset of overt disease; and remained elevated until the animals reached their humane end point. In spite of deep-seated inflammation, and even when given during the late symptomatic phase of the disease, gene transfer reversed these pathological features suggesting that while the inflammatory response might modify disease progression, it is an unlikely primary contributor.
Neuronal loss at the time of death in human GM2 gangliosidosis has been found to vary considerably between cases (11). Huang examined two cases and identified extensive cellular apoptosis that appeared to include all classes of cells: neurones, oligodendrocytes, astrocytes, microglia and vascular pericytes. In the SD mouse, a number of studies have reported neuronal apoptosis correlating with disease progression and ultimate demise of the animals (45,49,50). We have also described discrete and gradual loss of neurones in nuclei VPM/VPL of the thalamus and lateral vestibular nucleus in brain stem, but could not detect gross neuronal loss that could account for this phenomenon (22). It is possible that modest and diffused losses, difficult to detect, are responsible for the SD stereotypic phenotype; alternatively, neuronal dysfunction, rather than massive neuronal loss, might be the major cause of disease in mice. We studied neurones undergoing degeneration by the exquisitely specific and sensitive chemical-development-silver method of Gallyas (21), and consistent with GM2 gangliosidosis being a progressive neurodegenerative disease, metal deposition in soma and processes of neurons increased with age in most areas of the brain and spinal cord. However, we discovered that degeneration starts at a very early age, and its time course differs between anatomical regions. Sensory pathways are among the first sites to show signs of degeneration, followed by staining of large white matter tracts. It is especially notable that some cranial nerves appear to escape degeneration altogether. Furthermore, the appearance of amoeboid microglia coincided with regions of greatest deposition of silver. In line with our histological findings of early signs of degeneration in the hippocampus and cerebellum, Gulinello reported early motor coordination and memory deficits in the SD mouse (51), and Hu using magnetic resonance imaging found significant alterations starting at 6–7w age (52).
When we treated animals by gene transfer at age 4w (asymptomatic) and 12w (late symptomatic), killed them at the humane end point of the latter group and silver stained them, we saw a remarkable clearance of labelling compared with age-matched untreated controls. We concluded that even when given late in the development of the disease, the treatment could prevent and possibly reverse neurodegeneration. Although we are unable to follow the fate of individual neurones, it is impossible to be certain that reversion of degeneration as detected by silver deposition has occurred; however, it is highly plausible. We interpret the remaining silver deposition mostly as a failure to effectively provide enough quantities of enzyme at these sites in time. Alternatively, as proposed by Cabrera-Salazar et al. who treated the mouse model of late infantile Batten disease, neuronal degeneration may only be halted or slowed rather than reversed (25).
The VPM/VPL nuclei of the thalamus are central to the relay of sensory information from the hindbrain and spinal cord to the sensorimotor cortex. We have recently reported gradual neuronal density loss with age and prevention when gene transfer is given at 4–5w age (22). Using the VPM/VPL nuclei as a paradigm of neuronal rescue, we studied lifelong preservation of neuronal density in long surviving animals, and prevention of further loss when treatment is delayed. We established that when treatment is given during the asymptomatic phase of the disease neuronal preservation last for the entire life of the animal, and importantly, if the injections are delivered during the early symptomatic phase further lost is prevented, but if given during the late phase of disease, neuronal loss proceeds unabated by treatment.
In his seminal neuropathological investigations, Bernard Sachs already recognized that abnormalities of the white matter occur in GM2 gangliosidosis, and since then numerous case studies have reported varying degrees of demyelination (11,53). Haberland and Brunngraber described one such case as failure to myelinate, with increased ganglioside in cerebral white matter and significant decrease in cerebrosides and glycosaminoglycans, together with derangement in glycoprotein structure. Comparative studies of the of lipid content in mouse, cat and human samples of GM2 gangliosidosis at the end stage of the disease by Baek et al. demonstrated significant reduction of myelin-enriched lipids, cerebrosides and sulphates compared with the brain of suitable control subjects (17). Recently, the cerebral white matter of the sheep model of TSD at 8 months of age has been analysed histologically and shown to have decreased myelin-specific staining (54). Opinion has been divided as to whether myelin defects are merely a consequence of primary neuronal disease or constitute primary changes of maturation in myelin development.
Using the demyelinating twitcher mouse as control, we investigated myelin integrity in SD mice. Oligodendrocyte-specific genes were down-regulated in all regions of the brain and as early as age 5w, both at the level of transcription and translation. Expression of cgt [uridine diphosphate (UDP)-galactose:ceramide transferase], the key enzyme in galactolipid biosynthesis, is down-regulated in all areas of the brain and from a young age. Thus, reduced cerebroside and sulphatide in murine GM2 gangliosidosis can be explained, at least partly, by a reduction in this essential protein. Taken together, our findings cannot easily be accounted for as a consequence of neuronal loss or axonal degeneration. Kroll and colleagues examined the feline model of SD by magnetic resonance imaging and their findings are consistent with delayed myelination. Notably, when these authors analysed brain tissue with the myelin-specific stain, luxol fast blue, the intensity of staining was reduced, and yet myelin ultra-structure appeared normal (55). The crucial question immediately posed is whether myelin composition can be preserved by gene transfer and if so, to what extent would it be effective when carried out at different stages during the evolution of disease. Unexpectedly, myelin protein content was variably reduced in animals treated during the asymptomatic and early symptomatic phases, and in animals injected during the late symptomatic phase it appeared no different from SD (UT) controls. Although a larger cohort of animals needs to be studied to confirm these results, our findings indicate that early myelin defects are refractory to treatment when this is given in adult life.
Several recent reports indicate that myelin deficits might be a common feature among LSDs with neurodegenerative features—as well as contributing an important element of the pathogenesis early in the course of the given disorder. Lending support to this notion are recent studies that include cLINCL Cln8 (56), Nieman-Pick disease type A (57) and fucosidosis (58).
We concur with the views of Haberland and colleagues that demyelination is a central part of the development of neurological injury in SD disease, and that this is likely caused by more than one process; the result of abnormal myelinogenesis and secondary myelin degeneration as a contributory factor (11). With the availability of different animal models of GM2 gangliosidosis—and effective means for long-term therapeutic gene transfer combined with a greater understanding of myelin biology—the tools to explore these fundamental questions that have long intrigued investigators in the field are now available.
We have investigated the nature and progression of the neurological injury in GM2 gangliosidosis—testing whether it can be prevented, halted or reversed using gene transfer delivered at key times during the evolution of the disease. We have moreover studied the salutary effects on survival and well-being of the animals. The results show that there is a restricted temporal opportunity in which function and survival can be improved—but regardless of a complete resolution of the cardinal pathological features of GM2 gangliosidosis, a point is reached when functional deterioration and death cannot be prevented.
MATERIALS AND METHODS
The SD knockout mouse (B6;129SHexbtm1Rlp) (10) and natural mutant KD twitcher mouse (C57BL/6J Galctwi) (59) were obtained from The Jackson Laboratory (Bar Harbor, ME, USA). The twitcher mouse, a classic model of demyelination, has a mutation in the galactosylceramidase (galc) gene and is an authentic model of infantile KD (OMIM 606890). The animals appear normal at birth but develop an acute disease, and die at 40 days of age. Both strains are maintained by heterozygous matings. Studies were conducted by using protocols approved under license by the U.K. Home Office (Animals Scientific Procedures Act, 1986). Hexb and Galctwi genotyping was determined by PCR as described elsewhere (18,60). Mice had access to food and water ad libitum and were provided with nutritional supplements (Transgel; Charles River Laboratories) on the cage surface. The approved humane end point applied to mice throughout this study was defined as the loss, between 10 and 20%, of pre-symptomatic body weight for SD, and 20% from maximum achieved weight for twitcher mice. Animals were killed at any time if they developed clinical signs such as visceral enlargement, tumours and self-inflicted injuries or when they reached the age of 2 years.
Vector construction and production
Human cDNAs coding for β-hexosaminidase α and β subunits were fused at their 3′-ends to human immunodeficiency virus protein transduction domain, WPRE and BGHpA, as described (18). rAAV viruses were produced by triple plasmid co-transfection of HEK 293 cells as described (18). Total dose injected was 3.2 × 1010 and 4.7 × 1010 DNAse-resistant particles) for rAAV2/1α and rAAV2/1β, respectively.
Intracranial stereotaxic infusion of rAAV vectors
SD mice were anaesthetized and placed in a stereotaxic frame (Kopf Instruments, Tujunga, CA, USA). Burr holes were drilled over target sites, and vector mix was delivered vertically by using coordinates relative to bregma; striatum, AP (anterior–posterior): −0.1 mm; ML (medial–lateral): ±2.0 mm; and DV (dorsal–ventral): −3.0 mm; and cerebellum, AP: −6.0 mm; ML: ±1.5 mm; DV: −3.0 mm. Animals received 3 μl of vector mix per site (rAAV2/1α:rAAV2/1β:20%, w/v, mannitol; 1.1:1.1:0.8), at 0.8 μl/min infusion rate. The needle was withdrawn after 5 min. Animals were given post-operative analgesia (Rimadyl, Large Animal Solution; Pfizer, Kent, UK) and placed in an incubator at 37°C to recover.
Mice were killed by CO2 asphyxiation and organs snap-frozen on dry-ice, or transcardially perfused with cold phosphate buffered saline (PBS) followed by PBS containing 4% paraformaldehyde (pH 7.4) for histochemical staining. Perfused tissue was post-fixed in the same fixative for a few hours and cryoprotected in 30% sucrose overnight at 4°C. Fifteen-micrometer coronal sections were mounted on glass slides, dried for a few hours at room temperature (RT) and stored at −80°C.
Neurodegeneration was detected on 30 μm sections by the chemical-development-silver method of Gallyas (21), using the FD NeuroSilver™ Kit II (FD NeuroTechnologies, Ellicott City, MD, USA), according to the manufactures’ recommendations.
For non-perfused tissue, sections were warmed at RT for ∼30 min and fixed in cold PBS containing 4% paraformaldehyde (pH 7.4) for 10 min before staining.
β-Hexosaminidase activity was detected with naphthol AS-BI N-acetyl-β-glucosaminide (Sigma, Poole Dorset, UK) as described elsewhere (61).
PAS (Leica, UK)-stained sections were counterstained with haematoxylin (Leica). Prior to mounting in DPX (dibutyl phthalate xylene, british drug houses), sections were dehydrated and cleared with a series of solutions of increasing concentrations of ethanol and xylene.
Immunohistological staining was performed with primary antibodies: rabbit polyclonal anti-α-synuclein [(C-20)-R; Santa Cruz Biotechnologies, Inc., 1/50], mouse monoclonal anti-α-synuclein (42/α-Synuclein; BD Transduction Laboratories™, 1/50), mouse monoclonal anti-ubiquitin (Ubi-1; MAB1510, Millipore, CA, USA; 1/250), rat monoclonal anti-Cd68 (MCA1957; Serotec, Oxford, UK; 1/50); mouse monoclonal anti-NeuN (MAB377; Chemicon International, CA, USA; 1/100); and biotinylated secondary antibodies (Vector Laboratories, UK): goat anti-rabbit (BA-1000; 1/1000), horse anti-mouse (BA-2000; 1/1000) and rabbit anti-rat (BA-400; 1/1000). Staining was based on the avidin–biotin peroxidase technique (Vectastain ABC HP Kit; Vector Laboratories, San Francisco, CA, USA), developed with 3′-diaminobenzidine and counterstained with Cresyl violet. For anti-NeuN staining, tissue was antigen retrieved by incubating slides at 95–100°C for 20 min in 10 mm trisodium citrate, pH 6.0. They were then left to cool at RT for 30 min. Slides were mounted with DPX.
ISH was performed with digoxigenin-labelled riboprobes against the WPRE-BGHpA sequence, present in both rAAVα and rAAVβ vectors. ISH-positive cells were detected by colorimetric staining with alkaline phosphatase-mediated 5-bromo-4-chloro-3-indolylphosphate-nitroblue tetrazolium reaction, essentially as described (62).
For polyacrylamide-gel electrophoresis, after reduction and denaturation in sodium dodecyl sulphate and 4% β-mercaptoethanol, tissue extracts were run in 4–15% linear gradient gels (161–1122; Bio-Rad). One to 40 μg protein samples were heated at 90°C before gel loading, except for those blotted with the anti-plp antibody. Western blots were processed with primary antibodies: mouse monoclonal anti-cnpase (11-5B; Sigma-Aldrich, MO, USA; 1/1000); rabbit polyclonal anti-th (AB152; Millipore; 1/1000); chicken polyclonal anti-plp (NB100-1608; Novus Biologicals; 1/1000); mouse monoclonal anti-mbp (SMI-94; Calbiochem, CA, USA; 1/1000); mouse monoclonal anti-β-actin (AC-74; Sigma-Aldrich; 1/5000); mouse monoclonal anti-β-tubulin III (T8660; Sigma-Aldrich; 1/1000); mouse monoclonal anti-synaptophysin (SY38; Abcam, UK; 1/1000). Horseradish peroxidase conjugated secondary antibodies were: goat anti-mouse (P0447; DakoCytomation, UK; 1/5000), goat anti-rabbit (401315; Calbiochem; 1/5000) and goat anti-chicken (ab97135; Abcam; 1/5000). Blots were developed with Amersham™ ECL™ western blotting Analysis System (RPN2109; GE Healthcare, UK) following the manufactures’ recommendations.
mRNA was extracted from 20 mg of tissue using the PNA/DNA/Protein extraction kit (23500; Norgen Biotek Corp, Canada), and first-strand cDNA was synthesized by reverse transcription of 300–500 μg of mRNA in a 20-μl total volume with random primers (205311; Qiagen) following the manufactures’ recommendations. Relative quantitation was performed by real-time PCR on Applied Biosystems 7500 Fast Real-time PCR System (Applied Biosystems). One microliter of reverse transcription reaction was mixed with 300 nmol of each primer and Power SYBR green PCR master mix (4367662; Applied Biosystems) was added to a final volume of 20 μl. Thermal cycling conditions were: 95°C for 10 min 1×, 95°C for 15 s 1× and 60°C for 1 min 40×. Primers were: Mip-1α (macrophage inflammatory protein 1α) forward (F): 5′-TCTGTACCATGACACTCTGC-3′ and reverse (R): 5′ AATTGGCGTGGAATCTTCCG 3′; Rantes (regulated on activation normal T cell expressed and secreted) (F): 5′-AGT GCT CCA ATC TTG CAG TC-3′ and (R): 5′-AGC TCA TCT CCA AAT AGT TG-3′; Gfap (glial fibrillary acidic protein) (F): 5′-AGTAACATGCAAGAGACAGAG-3′ and (R): 5′-TAGTCGTTAGCTTCGTGCTTG-3′; mag (myelin-associated glycoprotein) (F): 5′-TACAACCAGTACACCTTCTCGG-3′ and (R): 5′-ATACAACTGACCTCCACTTCCG-3′; plp (F): 5′-TTCCAGAGGCCAACATCAAG-3′ and (R): 5′-AGGAGCCATACAACAGTCAG-3′; cgt (UDP-galactose:ceramide galactosyltransferase) (F): 5′-AGTTTCCAAGACCAACGCTGC-3′ and (R): 5′-TGTTCCTGAGCACCACTTACC-3′; Cd68 (cluster of differentiation 68) primer set: Mm-Cd68-1-SG QuantiTect primer assay (QT00254051; Qiagen), β-actin primer set: Mm-Actb-2-SG QuantiTect Primer assay (QT01136772; Qiagen).
Tissue sampling and counting
Determination of relative neuronal density in VPM/VPL (ventroposterior medial and lateral) nuclei of the thalamus was achieved by sampling three NeuN-stained sections bilaterally at 180 µm intervals. Area of the structure and number of neurons present on a section were determined with ImageJ software (NIH, v1.41). The statistics were analysed with one-way ANOVA and Bonferroni multiple post hoc comparisons using GraphPad Prism v5.0 (GraphPad Software).
The Kaplan–Meier survival curve was analysed with the log-rank equivalent to the Mantel–Cox test. The statistics were analysed with one-way ANOVA and Bonferroni multiple post hoc comparisons using GraphPad Prism v5.0 (GraphPad Software). Values with P < 0.05 were considered significant. The Student's t-test was applied when comparing two samples. *P < 0.05; **P < 0.01; ***P < 0.001.
Conflict of Interest statement. None declared.
This work was supported by The National Institute of Health Research University of Cambridge Biomedical Research Centre (Metabolic theme); and an unrestricted grant from Cambridge in America.