ARTICLE IN BRIEF:
In a gene therapy experiment that delivered glutamic acid decarboxylase into the subthalamic nucleus of Parkinson's disease patients, researchers observed on neuroimaging scans that new linkages were formed between the basal ganglia and the cortical premotor and motor regions. They also observed that study participants who expressed the network more had better clinical outcomes than did those who had lower levels of induced network expression.
Patients treated with gene therapy for Parkinson's disease (PD) express a distinct metabolic brain network, with new nodal connections different from those seen in patients receiving sham therapy, according to new findings published online first November 28 in Science Translational Medicine.
Gene therapy for Parkinson's disease has shown some modest promise in early clinical trials. In a phase 2 clinical trial in which glutamic acid decarboxylase (GAD) was delivered via an adeno-associated viral vector (AAV) into the subthalamic nucleus (STN), 16 participants who received gene therapy had significantly better motor control after six months than did 21 controls who underwent sham surgery, according to a 2017 report in the Journal of Clinical Investigation Insight. Motor scores on the Unified Parkinson's Disease Rating Scale improved in the control group by about four points (12.7 percent) over six months, while the treatment group improved by eight points (23.1 percent) over the same period. The improvements persisted at a one-year open-label follow-up.
But the modest functional benefits, about 25 percent, were no better than deep brain stimulation and were not considered significant enough to pursue further development. Now, however, neuroscientist David Eidelberg, MD, professor and head of the Feinstein Center for Neurosciences at the Feinstein Institute for Medical Research and professor of molecular medicine and neurology at the Donald and Barbara Zucker School of Medicine at Hofstra/Northwell on Long Island, and colleagues have identified a novel mechanism underlying the gene therapy patients' improvements that could give this treatment a renewed chance.
The trial's investigators had theorized that STN AAV2-GAD therapy might modulate the activity of the PD-related covariance pattern (PDRP), a disease-specific metabolic network that has been demonstrated to be an objective and specific indicator of progression in Parkinson's disease that correlates with functional measures. A 2017 study in Human Brain Mapping had demonstrated that medication and deep brain stimulation both suppress PDRP activity, in direct proportion to the motor benefit achieved.
But when Dr. Eidelberg decided to take another look at FDG-PET metabolic brain imaging data obtained from the trial's subjects, he found that neither gene therapy nor sham surgery had any effect on the PDRP network at all. To identify patterns in the scans, taken presurgery, at six months after surgery, and again at 12 months, he used a computational method known as ordinal trends/canonical variates (OrT/CBA), a form of supervised principal components analysis (PCA), in which scans from each subject were ordered by time point. This multivariate approach is designed to detect spatial covariance patterns (networks) for which subject expression scores consistently change across time and/or treatment states.
“Using these computational methods, we found that the PDRP pattern continued to worsen at the same rate in both the gene therapy and the sham surgery groups,” Dr. Eidelberg said. “Both groups had the same level of disease progression in PDRP that you would expect to see over time, and yet they were getting functionally better, not worse. What was happening?”
The investigators studied the sham network first and found it was associated with the expression of a distinct cerebello-limbic circuit, a different brain region from either PDRP or what they ultimately found to be involved with the gene therapy in a 2014 report in the Journal of Clinical Investigation.
They next applied the OrT/CVA computational methods on the gene therapy patients, where they found a unique network far different from either PDRP or the network activated in sham therapy patients. They called it the AAV2-GAD-related metabolic covariance pattern (GADRP). “It wasn't seen at baseline and evolved over time strictly in the gene therapy patients, a remodeling of network connections that we believe allowed the rest of the brain, in the motor system at least, to adapt and take on a different network feature than what would have otherwise occurred,” said Dr. Eidelberg.
These new connections improved linkages between regions in the basal ganglia and the cortical premotor and motor regions. They recruited nearby nonmotor pathways to connect the left superior frontal node to the left caudate nucleus, the right superior frontal node to the right supramarginal gyrus in the parietal lobe, and the left anterior putamen and globus pallidus to the ipsalateral thalamic node. Gene therapy patients also had increased metabolic activity in the premotor region and supramarginal gyrus, and decreased activity in the basal ganglia, the ventral anterior and medial dorsal thalamic nuclei, and the inferior frontal gyrus.
“We found that this network ultimately allowed the site of the gene therapy and the subthalamic nucleus to connect and get signaling to the motor cortex. Indeed, we found that people who expressed the network more had better clinical outcomes than did those who had lower levels of induced network expression,” said Dr. Eidelberg.
Study co-author Michael Kaplitt, PhD, vice-chairman for research and director of Stereotactic and Functional Neurosurgery and the Laboratory of Molecular Neurosurgery at Weill Cornell Medical College, performed the first human gene therapy surgery for PD in 2003. “What I love most about the computational method used in this project is that it is a very unbiased way to ask questions about what the brain is actually doing,” he said. “PCA is an unbiased mathematical way of looking for patterns of changes in the brain that are unique to a particular population of patients, without specifying in advance what you think those patterns are going to be.”
But does the identification of a brain network pattern that uniquely defined the gene therapy patients mean this pattern is relevant to gene-therapy treatment goals, rather than an immune reaction to the virus or something else not relevant to Parkinson's? Dr. Kaplitt thinks it does. “There was good correlation between the degree of change in this pattern and the degree of clinical improvement patients experienced,” he said, “which suggests it's meaningful biologically and medically.”
Dr. Kaplitt noted that the computational technique used in the study may help illuminate the underlying biology of experimental therapies in the nervous system. “By combining PCA with FDG-PET, we can look at what is happening in the brain of a patient we are treating and try to understand better whether or not a given new therapy is working, and how it may be working, in order to improve it or to select patients who may be the best candidates for therapy. This could really be applied to any other neurodegenerative disease where you can get a PET signal. That's really what we want to be doing when we talk about personalized medicine in neuroscience.”
Jon Stoessl, MD, professor and head of neurology and director of the Pacific Parkinson's Research Centre and National Parkinson Foundation Centre of Excellence at UBC and Vancouver Coastal Health, said it should not be surprising to see novel neural network modifications following gene therapy. “You're not actually making a lesion as with DBS; instead, you're converting what's normally an excitatory structure, the subthalamic nucleus, into one that's inhibitory by virtue of gene therapy. It would make sense that there would be some differences in neural network activation.”
But he advised caution. “First, this new pattern accounts for only about 1.9 percent of variability. It's also hard for me to visualize exactly how converting the subthalamic nucleus to a GABA-expressing structure results in all these other changes. I don't think they've fully explained It. All they're really doing is looking at the relative changes in glucose metabolism and how they interrelate between different regions; that doesn't necessarily tell you that there's a connection between one region and another. It's interesting and worth pursuing, but I wouldn't run out and have the procedure based on these results.”