Building Up to a Breakthrough

We sat down with David Eidelberg shortly after the publication of his group’s most recent paper in Science Translational Medicine (1), which we covered in this article. Here, Eidelberg discusses 20 years of research in the field of functional brain connectivity and brain network mapping – and its translation into treatments for neurodegenerative disorders.

Good science takes time

Our recent paper, applying gene therapy to treat Parkinson’s disease, comes off the back of about 20 years of prior publications; a review in Lancet Neurology from June 2018 (2) provides the backstory of how we got where we are today for those that are interested. The overall concept is that of functional brain maps – in our case often glucose metabolism. But lately we’ve been doing similar work with rest state functional magnetic resonance imaging (FMRI).

We recognized a long time ago that neurological disorders are not just the disease of one area of the brain, but rather the result of abnormal systems or networks. Such abnormal networks can be present before clinical symptoms emerge and they develop over time. By measuring them, we realized that they are stable enough to be used as signatures – characteristic and unmistakable signals of disease that would be missed, if looking at the brain region by region. Having discovered these networks, we needed to validate them; it’s one thing to have research based on 50 patients, but we needed to be certain that such networks are present in every population of a disease. Moreover, we needed to be able to quantifiably determine if there’s a disease related abnormality.

We developed algorithms for that very purpose – to characterize brain networks. But our algorithms can also be used for tracking; by measuring brain networks over time, you can see how fast the disease is progressing over time. As a result, there’s a lot of interest in network progression and also the preclinical detection of networks. We found that people with a propensity for disease – those with genotypic markers or other symptoms (for example, sleep disorders in Parkinson’s) – had abnormal expression of the networks years before they developed actual symptoms.

Serendipity – and mystery – are a scientist’s best friend

A lot of our work – isolating the networks, confirming their presence, looking at a treatment, modulating a given network has been very interesting – and we’ve written papers on those topics (3, 4). But using social network theory to try to understand what it means to have a core or peripheral involvement in the network is currently exciting. We’ve found that certain regions of the brain are hyper-connected. These “cores” are highly susceptible to modulation, whilst peripheral zones are much harder to access (in fact, deleting those peripheral nodes wouldn’t affect the network all that much). Now, we’re investigating how different genotypes in Parkinson affects different parts of the network in terms of propensity mechanisms. We call generic pathological networks in the brain “related patterns” (RPs), and we’ve been able to characterize and validate them.

Through several longitudinal studies, we’ve found that the activity of Parkinson’s disease related pattern (PDPR) goes up over time in Parkinson’s patients. When you treat someone – either by depleting dopamine with the drug levodopa or using deep brain stimulation (DBS) of the sub thalamic nucleus (STN), you’re able to knock down PDRP expression to a certain extent. With either treatment, the network isn’t completely shut down, but it is suppressed enough to have a statistically relevant effect: people with higher suppression of PDRP have better outcomes in general. And so we want to target PDPR to alleviate symptoms.

The objective of the gene therapy approach was – perhaps somewhat naively – to get the same result as DBS. By introducing a gene that inhibits glutamate (the main neurotransmitter in the STN), we hoped to drive some neurons into an inhibitory state, which would be akin to the approach taken by high frequency stimulation or even experimental ablation of the nucleus altogether. The problem? We learnt that, over time, our patients had consistent increases in their PDRP, just like the untreated group; in other words, there was no modulation of PDRP by gene therapy, in contrast to what we predicted. Yet at the same time, we were seeing improvements in clinical outcomes, so something unexpected was happening.

So the hunt began. We wanted to know if gene therapy was inducing a brand new brain network that was achieving some level of clinical response in Parkinson’s patients – independent of PDPR suppression. It would be a first-of-its-kind discovery. Looking at the rest of the brain, we saw that there was indeed a significant network with unique topography that was not present in people without Parkinson’s; crucially, it was also not present in untreated Parkinson’s patients or those receiving conventional treatments.

We applied graph theory to conduct nodal analysis of the network subspace; very small areas of the brain that you’d never be able to see visually. We saw a sequence downstream of where our gene therapy had been induced. The neurotransmitter phenotype was older than the STN, but we were observing the occurrence of plasticity downstream. There was an indirect rewiring of neural connections, communicating the STN target to the motor cortex. And that does not happen automatically. We also saw that the network continued to mature over the course of 12 months – even in those who were already doing well from a clinical perspective at six months.

Expect plenty of (big) bumps in the road to translation

I would say that the outcomes for the gene therapy treatment were good compared with sham surgery – they had a gain of 4 or so points on the Unified Parkinson’s Disease Rating Scale (UPDRS). Despite this, it was hard for people to get around the potential invasive nature of the viral vector treatment; after all, you could achieve the same results through deep brain stimulation. We had a small biotech company, but no takers for a phase III trial – and the company went out of business.

Despite the setback, I felt that our one-year data deserved to be published – hence the aforementioned paper in Science Translational Medicine. After all, if the treatment was totally useless, patients would go right back to where they started. Now, there’s been a renewal of interest in the technology – in part, because of the story on network remodeling.

But to succeed, we need to stop thinking about research as we have done for the past 40 years. Otherwise, we will end up with the same barrier at the clinical trial end; the treatments we explore are often not all that more powerful than sham effects, which is not enough to keep the translational train running. I understand that people want to see cures or levels of improvement that might not be possible. But first, we need to try to temper expectations. Second, we must apply cutting-edge methods to redefine how we determine the effect of interventions versus sham effects – something that cannot simply be defined by clinical ratings. Only then – when genuine but perhaps subtle promise can be observed – are we going to spur additional interest. Right now, it feels like the field has burnt itself out with too many ideas that were tested too soon in pursuit of too many unrealistic expectations.

The real test for us will be to use our network analysis on real patients; not everyone will have the same precise networks we’re looking at now. The step beyond that is to reconstruct a patient’s individual network to design treatments – an important move towards personalized or customized medicine.

The big picture? Research in this area could (and hopefully will) open new doors in the diagnosis, monitoring and treatment of neurodegenerative diseases – but only if all stakeholders work within a framework of reasonable expectations and rewards.