what we do

How do we take a disease association found by genomewide screen and derive true biological meaning from this signal? In our reviews of the post-GWAS era and the increasing role of "omics" in the science of neurodegeneration, we have highlighted the enormous gap between the number of loci associated with various human traits and ailments, and the number of loci for which we have any functional understanding. Specifically, as a field, we should be asking: To what gene does the variant associated with disease refer? What is the effect on that gene of interest? How does common variation affect the gene in this way? What happens to the function or expression of the gene as a consequence of the small changes in sequence? What is the consequence on a cellular or organismal level of the changes in function/ expression of the gene/protein? 

We began to ask these questions after Alice helped to identify a new genetic risk factor for frontotemporal lobar dementia as a postdoc. We eventually tied this locus to the target gene TMEM106B, identifying the GWAS causal SNP and demonstrating that variation here alters TMEM106B expression through effects on CTCF and chromatin architecture. We also showed that TMEM106B is a lysosomal protein and that alterations in expression cause multiple lysosomal abnormalities, with evidence for genetic modifier effects on C9orf72 expansion carriers, as well as degree of TDP-43 pathology in ALS brain and in cell models.

Taking the experience gained in our dissection of TMEM106B, we are interested in functionally exploring the many (>200) other loci associated by GWAS with neurodegenerative diseases. 

Our current focus is on Parkinson's disease, where we have (1) dissected the effect of common missense variants in the PD risk gene TMEM175 in cells, animals, and humans (collaborating with the Ren and Luk labs) and (2) shown that the PD risk gene GPNMB is both necessary and sufficient for cells to take up alpha-synuclein and form pathology, using CRISPR-Cas9 knockout in iPSC-derived neurons. 

This project grows out of Alice's clinical life as a PD neurologist, which provides her with a regular reminder that we have no disease-modifying therapies for any of the neurodegenerative diseases. Nothing to stop or even slow down the disease process. If we are going to make any headway, along with finding new therapeutic targets, we need to make clinical trials better. 

One way that this may be possible is through the development of robust biomarkers that can help to reduce heterogeneity in trial enrollment, stratify responses by biomarker-defined subtypes of disease, tell us whether the drug is finding its molecular target. To reduce heterogeneity in trial enrollment (read: pick the right people to test), we have been looking for biomarkers that predict the development of a clinically important phenotype that develops in only some PD patients, cognitive decline and/or dementia. We have shown that the ADNC-RS, a simple genetic predictor based on Alzheimer's Disease (AD) GWAS risk variants, can predict which PD patients will have concomitant AD pathology at autopsy, a result which has ramifications for crossover therapy. We identified epidermal growth factor (EGF) as a novel blood-based biomarker that not only correlates with poor cognitive performance in PD, but also predicts which cognitively intact patients will convert to dementia in just over 1 year. We have more recently found that plasma melanoma inhibitory activity (MIA) levels predict cognitive decline in PD, with Mendelian randomization analyses suggesting that MIA is causally involved in these processes. We have also found markers that predict whether PD patients will have a "better" vs. "worse" course, demonstrating that plasma levels of Apolipoprotein A1 (ApoA1) may modulate risk for the development of PD.  

We are experienced in moving promising biomarkers from discovery stages to replication in international cohorts of patients. We note that EGF, MIA, and ApoA1 emerged from large-scale unbiased protein screens, an approach we pioneered in biomarker discovery, and both have been successfully replicated by ourselves and others. 

We believe that by intersecting genomic screens (representing the risk with which an individual is born) and large-scale protein biomarker screens (a readout of the individual's current state, including exposures during life), we can find dual genomic x biomarker hits that are particularly promising to investigate in model systems.

Our work on GPNMB exemplifies this approach, since GPNMB is both a PD GWAS target gene and a blood biomarker associating with disease severity. Because GPNMB emerged from both genomic and biomarker screens, we investigated the effects of GPNMB loss on iPSC-derived cortical neurons (iPSC-N). To our great surprise, decreasing GPNMB expression in iPSC-N rescued pathology. 

We are currently following ApoA1 and MIA (both hits that emerged from our large-scale biomarker screens, both implicated causally by Mendelian randomization analyses). Specifically, we are manipulating levels of these genes/proteins in model systems (iPSC-N, animals) in order to understand how higher levels of ApoA1 might protect individuals from key events in alpha-synucleinopathy and how higher levels of MIA may increase risk for the development of both Lewy body and AD pathology.