Using autoregulatory constructs, we report the generation of inducible mouse models of Sandhoff disease. The single inducible constructs used here showed widespread expression of β-hexosaminidase in the mouse brain and rescued acute Sandhoff disease. Our inducible models displayed near-total gene silencing in the presence of doxycycline: administration of the agent induced Sandhoff disease with all its stereotypic features in the adult mouse. Moreover, expression from the transgenic constructs proved to be reversible on withdrawing the doxycycline in vivo.
The use of a single genetic construct carrying both elements (the tet-transactivator and coding sequence expressed from a tet-responsive element bearing promoter) of the tet-inducible system is not in common use for generating transgenic animals by pronuclear microinjection. This stratagem has been used to deliver autoregulatory cassettes by viral vectors 
and to generate targeted ‘knock-ins’ in stem cells utilizing the Rosa26 locus 
. Here we show the utility of this approach is feasible for creating functional inducible transgenic mice by pronuclear injection into fertilized oocytes. The obvious advantage of using a single genetic cassette is that breeding schedules are simplified and reduced numbers of animals are required when an inducible system is bred onto a knockout background.
When crossed onto a Hexb−/−
background, both the Hex and SYN inducible cassettes rescued the mouse from acute Sandhoff disease. However, there were differences in expression pattern between the two constructs. Although the Hexb
promoter used to drive the Hex inducible cassette was intended to provide systemic expression of Hexb
based on its ‘housekeeping function’, expression of β-hexosaminidase activity outside the brain was only found in the skin and skeletal muscle. This precludes assessment of the role of β-hexosaminidase activity in organs such as the liver and kidneys; however it was surprising that for each promoter, expression throughout the central nervous system was similar, since, even for the construct driven by the Hexb
promoter, expression appeared to be confined to neurons. Lack of expression from the Hex transgene may have been due to the absence of expression elements in the construct used in this study. Alternatively, this phenomenon could be explained by position effects 
Although expression of transgenic β-hexosaminidase throughout the central nervous system rescued the Hexb−/−
mouse from acute Sandhoff disease, rescue was incomplete and residual neurodegenerative disease became apparent beyond six months of age (Videos S5
). At the humane endpoint, Hexb−/−HexTg
mice had a mild tremor, and sporadic glycoconjugate storage was seen in Purkinje cells in the cerebellum as revealed by PAS staining; indeed, both strains showed abundant storage of glycoconjugate in lobe ten of the cerebellum. Storage was also detected in regions that interact with the cerebellum such as the pons and the red nucleus. Of note, no pathological storage was found in the substantia nigra at the delayed humane endpoint in these animals.
The most obvious aspect of residual neurological disease in the transgenic animals was increased limb tone (spasticity) - observed as clasping of the limbs when the mice were lifted by their tails. This disease feature was associated with a reproducible pattern of storage in the brain at the humane endpoint (Figure S4
, Table S2
). Most cerebral structures in the forebrain were free of storage material. Importantly, the motor cortex (origin of the cortico-spinal tract) showed no storage of glycolipid and the striatum (except for the globus pallidus) was also free of PAS-staining material. Accumulation of glycoconjugate was found in neurons, accompanied by CD68-positive amoeboid microglia, in the reticular nuclei in the pons and medulla and throughout the ventral spinal cord (). Although storage did occur in other centres of the brain, such as the septum (both strains) and the hypothalamus (Hexb−/−HexTg
), pathological changes in reticular nuclei of the medulla and pons and in ventral spinal interneurons are thought to contribute to spasticity through modulation of lower motor neurons 
. The ventral spinal cord in Hexb−/−HexTg
mice at the humane endpoint showed loss of PAX2-positive ventral interneuron density. Loss of PAX2-staining ventral interneurons is supported by similar results obtained by staining for the neural cell marker NeuN 
. It is noteworthy that this result was comparable to loss of ventral interneurons in a study of a spastic mouse model 
, suggesting that loss of interneurons in the ventral spinal cord might be responsible for limb spasticity in long-surviving Hexb−/−HexTg
mice. Similar loss of PAX2-positive interneuron density was also observed in the Hexb−/−
animal at its endpoint, and although limb clasping was less marked, this probably reflects increasing spastic paralysis.
Declining motor function () may contribute to the final deterioration of the animals, which reach the humane endpoint () past one year of age. Although we cannot exclude the presence of disease in peripheral organs playing a role in weight loss, other studies from this laboratory 
showed that correction in the central nervous system was sufficient to rescue mice from acute Sandhoff disease for up to two years.
Efficient silencing of transgenic Hexb
mRNA expression was seen in both inducible strains of mice when exposed to doxycycline (). In the brain, transcript disappeared within 24 hours of doxycycline exposure and half life of enzyme activity was about four days, similar to human HEXB
activity in fibroblasts that had a half life of six days 
. In contrast, recovery of transgene expression upon withdrawal of doxycycline differed greatly between the two strains. In the frontal cortex of the Hex line, Hexb
transgene recovered to expression levels comparable with pre-doxycycline exposure, within six days. The frontal cortex of the SYN line recovered its transgene expression much more slowly, such that 18 days after doxycycline withdrawal β-hexosaminidase activity had only recovered to approximately one third of pre-doxycycline exposure levels. Rapid recovery of transgene expression in the Hex line after doxycycline withdrawal means that it was not slow clearance of doxycycline that was causing sluggish recovery of transgene expression in the SYN line; this was considered because the skeleton can act as a reservoir of doxycycline 
. Incomplete recovery of transgene expression in tet-inducible systems has been reported in other models 
and this is largely ascribed to changes in chromatin configuration during silencing that influence the accessibility and expression of transgenes 
Suppression of β-hexosaminidase expression in the adult mouse induced an acute phenotype of Sandhoff disease. This was comparable in progression and phenotype to the disease seen in the germline Hexb−/−
animal, where pathogenesis takes place alongside dynamic developmental processes. During rodent brain development, ganglioside synthesis changes dramatically. Gangliosides GM3 and GD3, which are abundant in the early fetal brain, decrease in expression and give way to increasing synthesis of GM1 and GD1a as neurons differentiate 
. Additionally, studies on developing postnatal cerebral cortex show a large temporary induction of GM2 expression that correlates tightly with strict developmental patterns of dendritogenesis 
. Based on this observation, we predicted that after initiation in the adult mouse, progression of disease may take much longer than in the conventional germline model of Sandhoff disease. The fact that Sandhoff disease progressed largely unmodified when initiated in the adult mouse suggests that either developmental changes in ganglioside synthesis are insignificant compared with adult neuronal output of ganglioside, or that lysosomal disorder disease processes are resilient to fluctuations in the absolute amount of storage material. Absence of a developmental component of neurological pathogenesis has also been observed in a conditional model of Niemann-Pick type C disease 
and may be a general feature of neuronopathic lysosomal disorders.
In the present study, a high dose of doxycycline was used to completely abolish transgene expression. However, it may be possible to titrate the dose of doxycycline to give partial correction of the phenotype - as has been reported in studies of GDNF expression induced by adeno-associated viral vectors under the control of the tet system in the substantia nigra 
. By allowing partial rescue of the Sandhoff phenotype mediated by either of the transgenes, this approach might be used to generate attenuated models of GM2 gangliosidosis.
We contend that the inducible models of Sandhoff disease reported here will facilitate exploration of the pathogenesis in GM2 gangliosidosis. In mice that were allowed to develop residual elements of Sandhoff disease, regionalized storage of glycolipids and related storage molecules in the brain showed that limb hypertonicity can arise independently from glycolipid storage in the motor cortex and could originate in the hindbrain and spinal cord. Tetracycline-inducible models of acute Sandhoff disease were responsive to doxycycline and showed that the pathogenesis of acute Sandhoff disease is not dependent or significantly modified by processes that occur in the developing brain. Future research will focus on the effects of increasing transgene dosage; in addition, the inducible models will be useful for identifying early cellular events in the evolution of experimental GM2 gangliosidosis in vivo.