Previously, we had shown that a single AAV injection into the cerebral lateral ventricles of neonatal GM1 mice was sufficient to correct the neurochemical phenotype 
. In this study, we examined the therapeutic efficacy of an AAV2/1 vector encoding mouse βgal infused bilaterally into the thalamus, or thalamus and deep cerebellar nuclei of adult GM1 mice. This intraparenchymal infusion approach led to sustained high-level expression of βgal in many of the CNS structures analyzed, and dramatic reductions in GM1-ganglioside and GA1 content. Surprisingly the motor performance of AAV-treated GM1 mice declined over time at a rate similar to that of untreated GM1 mice. Nonetheless, the median survival of both groups of AAV-treated mice was significantly increased compared to untreated GM1 mice.
Our results show that thalamic delivery of an AAV2/1 vector encoding mouse βgal is an effective approach to achieve widespread distribution of therapeutic levels of this enzyme in the adult GM1 mouse brain. The distribution of βgal in the brain was uneven with anterior parts of the brain receiving less enzyme than posterior regions closer to the injection site. The simplest explanation is that enzyme diffusion limits its distribution to areas near the injection site. However we have shown that βgal, like many other lysosomal enzymes 
, can be transported by axonal retrograde transport from the site of injection 
. Thus an alternative explanation is that the βgal distribution pattern observed here is a direct result of the tropism of AAV2/1 vector for a subset of thalamic nuclei, namely dorsal and lateral nuclei. The faint ISH signal present in the cerebral cortex () could be due to axonal transport of vector-derived mRNA in thalamic neurons that project to the cerebral cortex, similar to our previous observation in the hippocampal system 
. Alternatively, it could be due to leakage of vector along the needle track with subsequent transduction of cortical cells. This could explain the intense staining observed in these cortical regions. Additional experiments will be necessary to determine whether more widespread transduction of thalamic nuclei leads to wider distribution of βgal in the brain. Others have also injected AAV vectors encoding lysosomal enzymes into the adult thalamus in mouse models of other LSDs, but it has been done in combination with several other targets, and thus it is difficult to assess the contribution of thalamic transduction to the overall pattern of enzyme distribution in the CNS 
. Recently, Kells and colleagues showed that thalamic infusion of an AAV2 vector encoding GDNF resulted in extensive distribution of this growth factor to the frontal cortex 
. These results support the notion that AAV-mediated genetic modification of thalamus may indeed be an effective approach to distribute other secreted therapeutic proteins to large regions of the brain.
The main finding from our studies is that bilateral injections of AAV vector in adult GM1 mice lead to sustained reduction of GM1-ganglioside and GA1 content to nearly normal levels in most CNS structures analyzed. The spinal cord was an exception, however, as GM1 levels were reduced to only ~50% of the levels found in age-matched untreated GM1 mice. Both groups of AAV-treated mice survived significantly longer than untreated GM1 mice, with the group receiving bilateral injection of AAV vector in the thalamus and DCN (AAV-TC group) surviving until 1 year of age. Nonetheless the motor performance of both groups of AAV-treated GM1 mice declined over time and was indistinguishable from that of age-matched untreated GM1 mice. According to our neurochemical analysis, combined bilateral injection into thalamus and DCN (AAV-TC group) is effective in restoring GM1-ganglioside level to normal and eliminating GA1 glycosphingolipid in the cerebellum, while bilateral thalamic injections (AAV-T group) lead to partial correction (, light gray bars), despite achieving enzymatic activity levels comparable to those in cerebellum of HZ mice (, light gray bars) (This apparent paradox is addressed below). The effect on lipid content in other CNS structures is comparable in both treatment groups. The difference in median survival between AAV-T and AAV-TC GM1 mice suggests that disease progression in the cerebellum is a contributing factor to the phenotype in GM1 mice. Previous studies have shown that the earliest and most profound changes in neuroinflammatory markers and cytokine production are found in cerebellum, brainstem and spinal cord 
. The unexpected finding in this study was that GM1-ganglioside and GA1 levels in the spinal cord were essentially the same in both groups of AAV-treated mice. βgal activity in the spinal cord of both groups of AAV-treated GM1 mice was comparable at all time points analyzed at approximately 50% of enzymatic activity in HZ spinal cord (same observation as in the cerebellum of AAV-T GM1 mice). This level of enzymatic activity should be sufficient to completely correct lysosomal storage in the spinal cord. This apparent paradox may be explained by transport of βgal from injected structures in the brain to a relatively small subset of cells in the spinal cord where βgal activity reaches high levels and skews the total enzymatic activity in the tissue. The fact that we do not observe normalization of GM1-ganglioside level and elimination of GA1 glycosphingolipid suggests that secondary distribution of enzyme to other cells may be inefficient. Our present observation that injection of an AAV vector into DCN does not improve significantly lysosomal storage in the spinal cord is surprising. Previous studies have shown that AAV-mediated gene transfer to DCN in mice is an effective approach to supply lysosomal enzymes and growth factors to the spinal cord 
. Moreover this target has also been utilized in several studies of AAV-mediated gene delivery to the adult brain in mouse models of other LSDs, but also in the context of multi-target injections 
, and thus difficult to discern the contribution of AAV-transduced DCN to the overall therapeutic effect. Several explanations are possible for our results: 1) Stereotaxic injections missed the intended DCN target, and transduction of other cerebellar regions is sufficient to supply the cerebellum with therapeutic levels of βgal, but inadequate for axonal transport of enzyme to the spinal cord; 2) differences in vector preparation/purification method that could affect its tropism, similar to the effect reported for other AAV serotypes 
; 3) severity of spinal cord involvement may be disease-specific, and in GM1 mice it appears to be one of the first structures in the CNS where alterations are detected 
; 4) higher levels of βgal may be necessary in the spinal cord of GM1 mice to achieve therapeutic correction compared to the amount of enzyme necessary in other mouse models of LSDs. Although we documented nearly normal GM1-ganglioside levels and complete elimination of GA1 in cerebrum (cortex, brainstem and subcortical structures) and cerebellum of AAV-TC GM1 mice at 4 months post-injection (5.5 to 6 months of age), these glycosphingolipids remained elevated in the spinal cord (2.2-fold above normal). Since motor performance was essentially indistinguishable from age matched untreated GM1 mice, disease progression in the spinal may be responsible for the progressive decline in motor performance of AAV-treated GM1 mice in the present study.
Myelin loss has been reported in patients and animal models of GM1- gangliosidosis 
. Previously we have shown that intracerebroventricular injection of an AAV2/1-CBA-βgal vector in neonatal GM1 mice restored cerebrosides, which are myelin-enriched lipids, to normal levels in the cerebrum at 3 months of age 
. The concentration of cerebrosides is directly proportional to the amount of myelin and has been used as a myelin marker in remyelination studies in mice 
. Our results further document the presence of a biochemical alteration in GM1 mouse myelin, and suggest that AAV-treatment partially corrects it (). Previous studies have shown that the timing of intervention is a critical parameter to prevent axonal degeneration in classical late infantile neuronal ceroid lipofuscinosis mice, where early treatment extends considerably the longevity of AAV-treated mice 
. We have performed additional studies in 4 week-old GM1 mice injected bilaterally in the thalamus and DCN with the same dose of AAV2/1-CBA-mβgal vector used in this study (with and without i.v. infusion of liver-specific AAV vector) to test this possibility, but did not observe any improvement in motor performance compared to age matched untreated GM1 mice (data not shown).
A potentially important observation in this study was the apparent preservation of some visual function in AAV-treated GM1 mice beyond 6 months of age, while age-matched untreated GM1 control mice did not respond to visual stimuli, as previously described 
. Also the observation of normal peak implicit times with reduced amplitudes in the averaged responses in AAV-treated mice suggests that some (but not all) cells in the retina may have retained normal function. Since retinal ganglion cells project to the lateral geniculate nucleus in the thalamus, and this region appeared to express high levels of βgal (as indicated by some of the darkest X-gal staining anywhere in the brain – See ), it is possible that some enzyme and/or AAV vector may have been transported via retrograde axonal transport to the retina. The histological evidence of βgal activity in the retinal ganglion cell layer of AAV-treated GM1 mice supports this notion, and could be the basis for preservation of some visual function in these mice. The effectiveness of transport of lysosomal enzymes and/or AAV vector from the thalamus to the eye requires additional experiments to characterize visual function in detail (electroretinograms, VEP and acuity) over time, and correlate it to GM1-ganglioside levels and histopathology in the retina.
Although it is possible that storage in other structures in the CNS, or remaining myelin deficits, may be involved in the long-term phenotype observed in AAV-treated GM1 mice, the symptoms that were displayed throughout the experiment are consistent with spinal cord involvement. Thus we hypothesize that to achieve long-term survival of GM1 mice with stable or improved motor performance, it will be necessary to devise a strategy to effectively address disease progression in the spinal cord. Several recent studies have shown that infusion of recombinant lysosomal enzymes into the cerebral spinal fluid via the lateral ventricles or intrathecally, is an effective approach to reach therapeutic levels throughout the CNS 
. Also intracerebroventricular and intrathecal infusion of AAV vectors appear to be effective in delivering therapeutic levels of lysosomal enzymes to many regions of the CNS in various mouse LSD models 
. Combination of bilateral thalamic injections with one of these modalities may be an effective way to achieve complete correction throughout the CNS. Other targets in the CNS such as the ventral tegmental area (VTA) 
and external capsule (EC) 
appear to be highly effective to achieve widespread distribution of lysosomal enzymes in the brain after a single injection of an AAV vector. Rigorous studies will be necessary to compare the effectiveness of different AAV vector engineered enzyme producing centers in the brain (thalamus, VTA, EC) to achieve global distribution of those enzymes throughout the adult GM1 mouse CNS. To our knowledge this is the first study demonstrating the potential of thalamic gene delivery to achieve global distribution of a lysosomal enzyme throughout the mouse cerebrum. Our studies suggest that the thalamus can be used as a central node in a ‘built-in’ network for widespread distribution of lysosomal enzymes, and possibly other secreted therapeutic proteins throughout the cerebrum.