Subthalamic Glutamic Acid Decarboxylase Gene Therapy: Changes in Motor Function and Cortical Metabolism

Introduction

Parkinson's disease (PD) is characterized by a progressive loss of dopaminergic neurons and decline of striatal dopamine. This neurodegenerative process leads to the disinhibition of glutamatergic neurons within the subthalamic nucleus (STN) with a concomitant increase in the activity of excitatory projections to the internal globus pallidus and the substantia nigra pars reticularis (Wichmann and DeLong, 1998; Hamani and Lozano, 2003). Increased pallidal output is responsible for many of the motor symptoms of PD (Brown et al, 2001). A variety of surgical approaches have been introduced to modulate the overactive STN output in PD, including high-frequency deep brain stimulation (Benabid et al, 1996; Lozano and Mahant, 2004) and stereotaxic lesioning (Alvarez et al, 2001; Su et al, 2003) of this structure. Deep brain stimulation represents an increasingly established and successful procedure with improved motor ratings at follow-up periods of up to 4 years (Lozano and Mahant, 2004; Rodriguez-Oroz et al, 2005). Although long-term treatment might not be limited by a major decline in stimulation effects, device-related complications can affect up to 20% of implanted leads (Oh et al, 2002), posing an additional burden on patients with advanced PD who already suffer considerable morbidity.

Recently, gene therapy has been proposed as a novel approach for the treatment of PD (Chen et al, 2005). In general, this therapeutic strategy has focused on the local delivery of genes to enhance dopamine production as well as neurotrophic agents (Bankiewicz et al, 2000; Kordower et al, 2000). Another approach for PD has been suggested using gene therapy to reduce abnormal pallidal output (During et al, 2001). We tested the effect of subthalamic delivery of adeno-associated viral vectors (AAV) encoding glutamic acid decarboxylase (GAD). This enzyme is rate-limiting for the synthesis of GABA, the major inhibitory neurotransmitter in the brain. Our aim was to impart inhibitory GABAergic properties to STN glutamatergic neurons and thereby alter the resulting activity of downstream targets, which influence movement. Direct support for the therapeutic effect of increased GABAergic transmission in the STN is provided by improvements in primate models of PD after a single injection of muscimol, a GABA agonist, into this area (Baron et al, 2002). Human studies have replicated these findings, demonstrating brief resolution of parkinsonian symptoms after a single injection of a GABA agonist into the STN (Levy et al, 2001). Based on this, we reported behavioral improvement after the delivery of AAV-GAD into the STN of a standard rodent model of PD (Luo et al, 2002). This was associated with increased evoked GABA release from the STN, a phenotypic change in STN neuronal firing, and a reduction in nigral dopamine cell death. Several of these findings have recently been independently replicated (Lee et al, 2005), with data suggesting that the degree of motor and physiological improvement may correlate with the level of STN GAD expression after gene transfer. In the current study, we examined the potential of subthalamic AAV-GAD to alter physiology and alleviate motor symptoms in a non-human primate model of parkinsonism.

Materials and methods

Results

Behavioral Findings

During the 55 weeks after AAV therapy, the animals maintained a constant weight and did not display signs of nausea, vomiting, diarrhea, fever, or infection. Throughout the study, they were cooperative during the testing sessions and were responsive to food stimuli. They also did not develop dyskinesias, hemiballism, or dystonia.

There was no difference between presurgical CRS scores for the AAV-GAD and the AAV-GFP groups (t=0.47, P=0.7, Student's t-test). Analysis of the CRS scores for weeks 0 to 55 revealed that the pattern of change in these ratings over time was different for the two AAV treatment groups (Figure 1). This is supported by RMANOVA, which revealed a significant group × time interaction (F (49, 539)=2.88, P<0.001). This analysis also showed a significant main effect of time (F (49, 539)=3.0, P<0.001) but a nonsignificant main effect of group (F (1, 11)=0.539, P>0.4). Graphical review of the fitted curves for the two groups revealed a gradual decline in CRS scores in the GAD group that continued until reaching a plateau at about week 25. By contrast, the GFP group demonstrated a decline in CRS scores over the first 10 weeks, followed by an increase to approximately 30 weeks with subsequent stabilization. In the GAD group, comparison of CRS subscale values at week 55 with presurgical values revealed significant improvements in bradykinesia (−16%, t=−2.52, P<0.05), gross motor skills (−26%, t=−2.93, P<0.05), and tremor (−36%, t=−3.87, P<0.01). These changes were not significant in the GFP group (changes of −8%, −10%, and −17% in these subscales, P>0.15). For the fine motor task, no significant difference between treatment groups was observed before MPTP administration or after subthalamic AAV delivery (P>0.05 for all time points).

OPEN IN VIEWER

Figure 1

Fitted curves illustrating the time course of CRS scores in unilateral MPTP monkeys receiving STN injection of either AAV-GAD or AAV-GFP. The data are consistent with a significant difference in the trajectories of the GAD and GFP treated animals (see text). (The timeline of the study is illustrated on top. All monkeys were scored zero at baseline before right intracarotid MPTP administration (black arrow). After neurotoxic exposure (black horizontal bar), the monkeys exhibited hemiparkinsonian signs. They were divided into two groups matched according to the CRS ratings obtained 5 days after MPTP administration. At day 5, STN-AAV surgery was performed (gray arrow), corresponding to week 0 on the x-axis. Weekly CRS ratings continued until weeks 55 during post-AAV surgical period (gray horizontal bar). At each time point, mean CRS scores for animals in the GAD and GFP groups are represented by filled and open circles, respectively.)

Positron Emission Tomography

Hemispheric comparison of the whole group (Figure 2) revealed a significant effect of MPTP administration, with relative hypermetabolism in the ipsilateral (right) globus pallidus (x=12, y=12, z=−8; Z_max=7.5; P<0.001, corrected) and in the contralateral cerebellum (x=−16, y=−4, z=−6; Z_max=6.2; P<0.001, corrected). Post hoc analysis of metabolic values in each of these regions did not reveal significant differences between the two AAV treatment groups.

OPEN IN VIEWER

Figure 2

Statistical parametric maps of FDG PET data comparing MPTP-treated hemispheres with contralateral untreated hemispheres (see text). Relative metabolic increases were identified in the (A) globus pallidus ipsilateral to MPTP injection and (B) the contralateral cerebellar hemisphere. (SPM{t}maps were superimposed on a single-subject macaque MRI T1 template. The color scale represents T-scores thresholded at 3.5, P<0.001).

Comparison of the two AAV groups conducted blind to treatment arm (Figure 3) disclosed a relative increase with GAD treatment that was localized to the ipsilateral motor cortex (gyrus centralis anterior, x=14, y=10, z=26; Z_max=3.5, P<0.001, uncorrected). No significant metabolic reductions were present in the GAD group relative to the GFP controls. No group differences were observed in the vicinity of the STN, and there were no differences in estimates of global metabolic rate. Correlational analysis (Figure 4) disclosed a significant relationship between CRS ratings at week 55 and metabolism in the right motor cortex (x=16, y=14, z=18; T=3.6; P<0.001, uncorrected).

OPEN IN VIEWER

Figure 3

(A) Statistical parametric map comparing whole-brain FDG PET data from the GAD and GFP-treated animals (see text). Significant metabolic increases were identified in the GAD-treated group in the right motor cortex (arrow). (SPM{t} maps were superimposed on a single-subject macaque MRI T1 template (sagittal slice). The color scale represents T-scores thresholded at 3.3, P<0.001). (B) Scatterplot illustrating individual metabolic values and group means (±s.d.) for this region in the GAD (squares) and GFP treated (circles) monkeys. Exclusion of GFP animals with striato-pallidal necrosis (shaded circles; see text) does not change the significance of the group differences in cortical glucose utilization (P<0.01 for comparisons between the GAD and GFP groups conducted with and without these control animals).

OPEN IN VIEWER

Figure 4

(A) Statistical parametric map of correlations between regional glucose utilization measured by FDG PET and clinical motor ratings (CRS) at 55 weeks after right subthalamic AAV surgery (see text). This analysis revealed a significant correlation between metabolism and motor ratings that was localized to the ipsilateral motor cortex (arrow). (SPM{t} maps were superimposed on a single-subject macaque MRI T1 template (axial slice). The color scale represents T-scores thresholded at 2.0, P<0.005). (B) Significant linear correlation between individual values of glucose utilization in the motor cortex and CRS ratings across the combined group of GAD (squares) and GFP (circles) treated animals.

Morphology

At 1 year after MPTP and AAV treatment, there was evidence of GAD and GFP expression in the injected STN of the monkeys (Figure 5), with the exception of one GFP control animal. Green fluorescent protein immunoreactivity was observed in the area of the needle tract, in nearby structures such as the substantia nigra and the lenticular fascicle/cerebral peduncles (Figure 5B). While some of the staining was in cell bodies, suggestive of local spread, most of this staining appeared to be in fiber tracts extending out from the STN. Aside from a single GFP-positive cortical neuron in one animal, there was no evidence of axonal transport of virus to other distant structures. Native GAD immunoreactivity within fibers projecting to the STN was very strong, which prevented quantification of STN neurons transduced with the GAD transgene. Although GFP staining was qualitatively observed within the STN of all but one of the GFP animals, great variability was observed with regard to both cell numbers and location within the STN.

OPEN IN VIEWER

Figure 5

(A) GAD immunohistochemistry of a right brain section through the STN from a monkey receiving intracarotid MPTP followed by subthalamic AAV-GAD injection. Gene expression was observed without inducing morphological alterations 1 year after AAV treatment. Inset: High magnification image of GAD positive neuron-like cells in the STN. (B) GFP immunohistochemistry of a right brain section through the STN from a monkey receiving intracarotid MPTP followed by subthalamic AAV-GFP injection. Strong expression of GFP was observed 1 year after AAV treatment. Inset: High magnification image of GFP positive astrocyte- (a) and neuron-like (b) cells in the STN. (C, D) CD68 immunostaining of a right brain section through the STN from a monkey receiving intracarotid MPTP followed by subthalamic injection of (C) AAV-GAD or (D) AAV-GFP. CD68 positive cells (inset, high magnification) were found in the areas of AAV expression (scale: A–D=1000 μm; insets=60 μm).

CD8 and CD45 immunoreactivity was virtually absent in all the animals (not shown). Mild CD68 immunoreactivity was observed in the area of injection and was most prevalent in GFP-treated monkeys (Figures 5C and 5D). Three monkeys in the GFP group exhibited an unexpected necrosis of the ipsilateral striatum and globus pallidus, associated with strong CD68 immunostaining. We did not find these animals to be influential in the metabolic differences that were noted across treatment groups (Figure 3).

Qualitative observations of brain sections stained for TH revealed unilateral loss of striatal fibers and nigral cells ipsilateral to MPTP administration. Striatal optical density confirmed asymmetrical loss of dopaminergic fibers. Comparison of lesioned and unlesioned hemispheres revealed that AAV-GFP animals had a loss of 72% in the caudate and 76.7% in the putamen, while AAV-GAD had a loss of 74% and 76.6% respectively. Similarly, stereological cell counts of TH positive nigral cells showed a profound loss of cells ipsilateral to the MPTP lesion without significant changes contralaterally. Statistical analysis did not reveal significant differences between treatment groups. This suggests that the degree of lesioning was similar between groups, with no histological evidence of recovery in either group after 1 year.

Discussion

In this study we demonstrate the feasibility of subthalamic GAD gene transfer in a non-human primate model of PD as highlighted by the detection of AAV vector product 1 year after surgery. During the course of the follow-up period, AAV-GAD proved safe, without the development of abnormal involuntary movements or any other adverse event. This therapy affected the course of MPTP-induced motor signs compared with the GFP-treated control animals, as reflected in the relative improvement of the clinical rating scores. In fact, there was a significant improvement in clinical subscores for bradykinesia, tremor, and gross motor skills that was not seen in the GFP controls, further supporting a different effect of the GAD virus on post-MPTP motor deficits compared with controls. The findings at necropsy indicated that these changes were not associated with differences in the integrity of nigrostriatal dopaminergic system as measured by TH immunoreactivity. The behavioral observations were supported by PET imaging results demonstrating that motor cortical glucose metabolism was elevated in the GAD group relative to the GFP controls. Moreover, metabolism in this region correlated significantly with contemporaneous motor ratings.

Our findings support the notion that delivery of AAV-GAD to the STN modulates the excitatory output of this structure, changing the overall neurotransmitter balance in the basal ganglia. The time course of the GAD CRS curve with a progressive improvement over several weeks is consistent with the delay in reaching peak expression found with AAV-mediated gene transfer (Herzog et al, 1997; Thomas et al, 2004). The nadir of the GAD CRS curve at week 25 suggests that either GAD transgene expression reached its peak plateau level at this time point or that the GAD gene transfer is associated with delayed secondary adaptive changes in the physiology of transduced and surrounding cells.

By contrast, the GFP control group has a different trajectory of change in CRS scores over the 1-year follow-up period. These CRS scores showed an unexpected initial decline, followed by an increase until approximately week 30. Subsequently, these values stabilized at a higher level than in the GAD group. The initial clinical improvement in the GFP group cannot be explained by recovery of the dopaminergic system because at post mortem the loss of nigral cells and dopaminergic terminals was similar for both treatment groups. Rather, this decline resulted from the substantial clinical improvement in three monkeys in the GFP control group. At necropsy, these monkeys displayed necrotic lesions in the striatum and globus pallidus, but not in the STN. The cause of these lesions is unknown, although similar areas of tissue destruction have occasionally been noted after intracarotid MPTP injection as has been reported elsewhere (Forno et al, 1995; Emborg et al, 2006). While it is surprising that this only occurred in control animals, there is no evidence to suggest that either GFP or GAD significantly altered this apparent MPTP toxicity. The equivalent number of TH positive nigral neurons and dopaminergic terminals between groups suggests that both the degree of MPTP-induced nigral lesioning as well as the stability of the lesions were no different between groups.

It is likely that the presence of striato-pallidal lesions in these three GFP animals was responsible for the absence of a significant group difference (i.e., main effect) in our data. By selectively improving outcomes in the GFP control group, the pallidal lesions behaved similarly to therapeutic pallidotomy, and worked against the primary hypothesis of relative functional benefit in the GAD treated group. Owing to sample size limitations, we could not perform a formal statistical comparison of the time course of CRS change in the GAD group with that observed in ‘true controls’, that is, the unlesioned GFP subgroup. Nonetheless, a graphical examination of the data (see figure in Supplementary Material) revealed that mean CRS scores for these control animals were consistently higher than that for the GAD group across all time points to week 55. By contrast, the three lesioned GFP monkeys, though initially exhibiting greater clinical improvement, approximated the CRS ratings of the GAD treated animals during the later stages of the study. Indeed, at week 55, the GAD group and the three lesioned GFP animals had similar degrees of clinical improvement (27% and 23% decline in CRS ratings relative to baseline), with comparatively less change (13%) in the unlesioned GFP animals. We note that excluding these animals in post hoc analysis of the imaging data did not alter the finding of relative increases in cortical metabolism in the GAD treated group (see Figure 3B).

In contrast to differences in the time course of the CRS, fine motor testing did not differ in the two AAV treatment groups. This aspect of motor performance may not have been affected by STN gene therapy. It is conceivable that fine motor improvement could be achieved by a more localized delivery of the viral vector within the STN or through the delivery of larger titers of virus. Alternatively, different mechanisms may underlie the features of motor performance captured by the two measures. Indeed, human PD patients treated with STN deep brain stimulation have been shown to have a better response to akinesia of proximal arm movement compared with fine motor finger movements of the hand (Wenzelburger et al, 2003). Consistent with this, we observed significant improvement in subscores for bradykinesia, tremor, and gross motor skills, which was not evident in the GFP control group. This suggests that these specific functions may be more sensitive to the effects of AAV-GAD.

The results of the FDG PET imaging 1 year after gene therapy show significant elevations in ipsilateral motor cortical glucose utilization in the GAD group relative to controls. We attribute this to an increase in the activity of thalamic projections to cortical motor regions, occurring with the modulation of pathological brain networks during antiparkinsonian therapy (Fukuda et al, 2001; Trošt et al, 2006; cf. Eckert and Eidelberg, 2005). This is consistent with previous findings of increased motor cortical activity in PD patients undergoing a variety of stereotaxic procedures for advanced symptoms (Eidelberg et al, 1996; Stefurak et al, 2003; Sestini et al, 2005). Metabolism in that cortical region also correlated with CRS motor ratings at 55 weeks. These findings are also in accord with a recent study of motor cortical stimulation, which resulted in improvement in CRS ratings and regional metabolism in a primate model of parkinsonism (Drouot et al, 2004).

We note that increased metabolism in the motor cortex of the GAD group could have emerged through alterations in the activity of other projections, such as those from the STN directly to this area (Hamani and Lozano, 2003). However, imparting an inhibitory GABA-ergic phenotype to glutamatergic STN-cortical projections does not necessarily change local energy requirements (Chatton et al, 2003). Moreover, this would likely change the nature of the relationship of metabolism to the motor scores. In this case, an increase in cortical inhibition subsequent to GAD therapy would be associated with a positive correlation between glucose utilization and CRS ratings, reflecting greater disability at higher rates of local metabolism. Our finding of a significant negative correlation between these measures relates increasing cortical metabolism to better performance, as observed in PD patients scanned in the rest state (Carbon and Eidelberg, 2002).

We also performed a hemispheric comparison to identify the regional effects of MPTP administration on glucose utilization in the anesthetized primate. In this acute, unilateral MPTP model, metabolic increases relative to the unlesioned hemisphere were present in the ipsilateral GP and contralateral cerebellum. Indeed, elevation in GP metabolism in MPTP primates was found using autoradiographic techniques in a similar unilateral intracarotid model (Palombo et al, 1990), and with in vivo PET imaging after chronic systemic MPTP exposure (Brownell et al, 2003). However, increased cerebellar glucose utilization was not a feature of the prior monkey studies. By contrast, the present findings of pallidal and cerebellar hypermetabolism subsequent to MPTP are consistent with FDG PET findings in PD patients (Eidelberg et al, 1994; Moeller et al, 1999). Importantly, post hoc comparison of local metabolism in these regions did not reveal significant differences between the GAD and GFP groups, indicating that the effect of the neurotoxin was similar in the two STN AAV treatment arms, in accord with the pretreatment CRS ratings and with pathological analysis.

This study was limited by several technical issues. Although the number of animals and length of follow-up here is equivalent to that of other primate gene therapy studies (Bankiewicz et al, 2000; Kordower et al, 2000), the number of animals may still be limiting, especially with the unexpected findings of necrotic lesions in the control group. A further limitation is the surgical technique, which involved MRI localization and stereotactic targeting of the STN. In humans, STN surgery is often performed with electrophysiological mapping to optimize targeting precision. This method was not performed in this study because of the extensive animal training required to perform electrophysiology and surgery in the waking condition. The far smaller size of the overall brain and the STN itself in these animals could lead to even greater variability in the site and amount of vector reaching potentially key subregions of the STN. Finally, complete information concerning the neural pathways affected by gene therapy requires a direct comparison of regional glucose utilization before and after treatment. Nonetheless, despite the limitations of the cross-sectional design that was employed, we identified a significant group difference in cortical metabolism that correlated with contemporaneous behavior descriptors. Further imaging studies incorporating a within-subject design will be needed to delineate the effects of this intervention on metabolic activity in subcortical motor regions (cf. Trošt et al, 2006).

Our observations in the primate suggest that subthalamic delivery of AAV-GAD is safe. All 13 AAV treated animals survived to the 1-year end point without adverse effects. Nonetheless, the possibility of late effects of STN gene therapy cannot be dismissed based on the results of a 1-year study of medication-naïve hemiparkinsonian non-human primates. Additionally, despite likely improvement in the function of basal ganglia—thalamocortical motor pathways after subthalamic AAV-GAD therapy, we could not assess potential treatment effects on the modulation of parallel limbic or cognitive circuits.

In conclusion, the results of these primate studies suggest that subthalamic delivery of AAV GAD is a feasible approach to the symptomatic treatment of parkinsonism. Indeed, these findings provide a useful proof-of-principle for the development of other gene therapies using AAV vectors. Further studies are warranted to optimize the delivery of viral vector, regulate gene expression, and to improve the behavioral outcome of gene therapy for PD and other disorders.