First PWAS for AD, now TWAS for PD. Hot on the heels of a study that linked AD variants to changes in a protein-wide association study (Feb 2021 news), geneticists now have linked Parkinson’s genome-wide association hits to genes in transcriptome-wide association and epigenetic studies. In the February 1 JAMA Neurology, John Hardy, Nicholas Wood, and colleagues at University College London reported 11 PD risk gene candidates.

  • Marrying GWAS with transcriptomic and epigenetic data IDs 11 PD genes.
  • Four are involved in the lysosomal pathway.
  • Nine are linked to familial PD genes and regulate cell growth.
  • Most are highly expressed in glial cells, not neurons.

Five regulate gene expression and six affect splicing. Four are new PD genes. Nine of the 11 encode proteins that interact with known familial PD genes involved in cell growth and differentiation. Of the seven known PD risk genes, four are in the lysosomal pathway, which is tied to PD. Expression changes were more prevalent in glial cells than in neurons. That these loci emerged from multiple different analyses may strengthen researchers’ confidence about them. “Potential biases and flaws in any one dataset are mitigated when hits are consistent across them all,” Wood told Alzforum.

“This type of large-scale bioinformatics analysis is important to narrow down potential leads for researchers working on functional studies,” Konstantin Senkevich at Montreal’s McGill University told Alzforum.

Matching GWAS hits to changes in gene expression is crucial to understanding how variants act in disease. This strategy has taught scientists about the function of at least 17 PD and 23 Alzheimer’s disease risk gene candidates so far (Feb 2021 news; Oct 2020 news). 

In this latest work, a collaboration between the U.K. Brain Expression Consortium (UKBEC) and the International Parkinson’s Disease Genomics Consortium (IPDGC), first author Demis Kia and colleagues identified genes that increase a person’s risk for sporadic PD. They started with a PD GWAS dataset of 26,035 cases and 403,190 controls of European ancestry (Chang et al., 2017). Of the 8 million single-nucleotide polymorphisms (SNPs) analyzed in that study, the researchers searched gene-expression databases for any genes near PD variants that were either up- or downregulated in people with PD. They chose any variants that associated with PD with a P value below 5 × 10−8. Similarly, they looked for genes near those SNPs that might have alternatively spliced variants or epigenetic changes that associate with the disease.

For expression and alternatively spliced variants, Kia and colleagues queried three data sets: Braineac, the Genotype-Tissue Expression (GTEx), and CommonMind Consortium (CMC). Braineac, which is part of the U.K. Brain Expression Consortium run by Hardy, includes expression data of 515 genes from 10 brain regions of 134 healthy controls. GTEx includes expression data of 748 genes from healthy tissue of 13 areas, each with between 80 and 154 samples. CMC includes data on expression of 5,420 genes and splicing of 3,315 genes from the dorsolateral prefrontal cortex of 206 people with schizophrenia, 52 people with affective disorder, and 206 healthy controls.

For genome-wide methylation, the authors analyzed DNA from the substantiae nigrae and frontal cortices of 134 people with PD who had donated their brains to the Parkinson Disease U.K. Brain Bank. The researchers looked for CpG methylation within a 500 kb window around the 8 million SNPs, turning up 37,460 methylation sites for further analysis.

Kia and colleagues compiled data from all three expression sets and analyzed it with statistical tools called Coloc and TWAS (see image below). Short for “co-localization,” Coloc calculates the probability that a single variant changes both PD risk and gene-expression levels. TWAS predicts whether gene expression is associated with disease by correlating GWAS hits with expression quantitative trait loci (eQTL). Linking loci to gene expression, splicing, or methylation provides functional context to GWAS hits (Apr 2019 conference news). 

Go With the Flow. Gene expression (left) and splicing (right) analyses whittled thousands of genes down to 11 candidates. [Courtesy of Kia et al., JAMA Neurology, 2021.]

After Coloc analysis, nine and 42 genes from Braineac and GTEx, respectively, were expressed differently in PD than in controls and came with higher PD risk. TWAS found 137 differently expressed genes in CMC, of which 61 lay close to a known PD variant. Comparing Coloc and TWAS revealed seven overlapping hits in Braineac and 18 in GTEx. Only these five genes were common to both data sets: WDR6, CD38, GPNMB, RAB29 (also called RAB7L1), and TMEM163.

Next, the researchers searched for differently spliced genes that lay close to the GWAS SNPs. Coloc identified 25 genes from Braineac linked to PD and expression, of which 15 were alternatively spliced as indicated by expression of specific exons rather than the whole gene. Kia compared these hits to those obtained from TWAS. In CMC, TWAS spotted 129 genes that had evidence of splicing, 40 of them near PD SNPs. Lo and behold, six of those also appeared in the Coloc list—ZRANB3, PCGF3, NEK1, NUPL2, GALC, and CTSB. In total, this gave the researchers 11 genes whose expression or splicing changes in PD.

What about methylation? Of the more than 37,000 CpG sites analyzed, methylation at 134 of them, across 107 genes, was altered in the substantiae nigrae of people with PD. Of those, 116 sites in 93 genes were statistically significant, and three of those genes were also among the 11 that have altered expression in PD: GPNMB, TMEM163, and CTSB.

Of the 11 finalists, four are newcomers to PD research. They are WDR6, which suppresses cell growth; PCGF3, which remodels chromatin; ZRANB3, which stabilizes DNA replication forks; and the cell-cycle regulator NEK1.

In contrast, GALC, CTSB, GPNMB, RAB29, TMEM163, CD38, and NUPL2 had all popped up in previous PD GWAS (Murthy et al., 2017Jul 2014 news; Mar 2014 conference news). 

How do these genes act in PD pathogenesis? Among the 11, the lysosomal pathway emerged most clearly. GALC encodes a lysosomal galactosidase, CTSB a lysosomal protease, GPNMB a lysosomal regulator. All are implicated in lysosomal storage disorders (reviewed in Bellomo et al., 2020; Moloney et al., 2018). RAB29 recruits LRRK2 to lysosomes, creating a complex crucial for clearing trans-Golgi vesicles. LRRK2 is a major PD gene (Sep 2018 news; Feb 2014 news). 

Lysosomes have long been implicated in PD. Variants in the lysosomal glucocerebrosidase gene boost risk for the disease, and deficiencies in the enzyme have been linked to poor clearance of LRRK2 and α-synuclein, the major component of Lewy bodies (Sep 2017 news; Mar 2014 conference news). 

As for the other seven genes, their links to PD are tenuous. Although transmembrane protein TMEM163, which is involved in amino acid metabolism, was linked to PD in multiple previous GWAS, small studies conducted in Spain, China, and Taiwan found no association (Tejera-Parrado et al., 2019; Wang et al., 2019; Chang et al., 2019). CD38, a glycoprotein expressed on the surface of many white blood cells, regulates neuroinflammation and has been indirectly implicated in neurodegeneration (Guerreiro et al., 2020). Nucleoporin-like protein 2 (NUPL2) may be involved with exporting mRNA from the nucleus into the cytoplasm.

Three of the 11 candidates have links to other neurological diseases. CTSB came up in a recent AD GWAS (Jul 2020 news), and transcriptional profiling of microglia found GPNMB upregulation in mouse models of amyloidosis and tauopathy (Aug 2018 news). NEK1 is a rare risk variant in amyotrophic lateral sclerosis (Aug 2016 news). 

Almost All Glia. PD gene candidates are more highly expressed in human glial cells, including astrocytes (gray), microglia (light blue), and oligodendrocytes (orange and tan), than in neurons (dark blue) or endothelial cells (white). [Courtesy of Kia et al., JAMA Neurology, 2021.]

Kia and colleagues then asked which cell types express these 11 candidate genes. They pulled expression information from transcriptome data of brain cells from healthy people. Astrocytes, oligodendrocytes, and microglia throughout the brain, including in the substantia nigra, more highly expressed most of these genes than did neurons (see image above). “This shows that the immune system is playing an important role in Parkinson’s disease,” Senkevich said.

Mathias Toft, University of Oslo, Norway, was surprised by the preponderance of glial expression. His group recently compared PD GWAS hits with areas of active chromatin, a marker of expression, and found them mostly in cortical neurons (Berge-Seidl et al., 2021). “We found enrichment of PD risk variants in regulatory gene regions of neurons, indicating that neurons, not glial cells, are the primary mediators of genetic risk for PD,” he wrote to Alzforum (full comment below).

Kia and colleagues took their analysis one step further by compiling co-expression networks to determine how the 11 genes fit into molecular pathways in different areas of the brain. In the nucleus accumbens, the caudate, and putamen, the scientists found NUPL2 to be involved in protein ubiquitination and degradation. In the frontal cortex and caudate, respectively, TMEM163 and ZRANB3 coordinate synapse signaling and cell communication.

Taking a similar approach at the protein level, the scientists found that nine of the 11 proteins interacted with proteins known to cause familial PD at a rate higher than chance alone would predict. Such a network regulates ErbB and the related EGFR signaling pathways, which are crucial to cell growth and differentiation (see image below).

Guilt by Association. In a protein-protein network analysis (top), nine of 11 PD candidates (green) interact through intermediates (gray) with proteins known to cause familial PD (pink), including LRRK2, PINK1, SNCA, and GBA. Searching for functional pathways (bottom) common to the familial PD genes and the nine candidates (yellow) brought out ErbB and EGFR signaling (circled in red). [Courtesy of Kia et al., JAMA Neurology, 2021.]

This study is one of a growing number that push past GWAS hits toward gene function in disease. Wood believes that as datasets expand and statistical tools advance, researchers can dig deeper into the functional implications of GWAS loci. “Marrying and analyzing large datasets is in its infancy,” he said. “Publicly available databanks are really good quality and they are here to stay. They’re going to get larger, deeper, and include more diverse populations as -omic sciences come to fruition.”—Chelsea Weidman Burke

Comments

  1. Genome-wide association studies (GWAS) have revolutionized the search for genetic risk factors of human disease, including Parkinson’s disease. However, the translation into biological understanding of disease mechanisms has been lagging behind.

    To improve the understanding of genes and mechanisms underlying PD, here Kia and colleagues have investigated the hypothesis that genes underlying PD influence disease risk by changes in gene expression, splicing, or methylation. They performed a range of complex bioinformatics analyses by combining information from a recent GWAS with relatively large datasets on gene expression, methylation, and gene splicing in the brain.

    The study identified five genes with changes in total gene expression associated with PD risk. At least two of these, GPNMB (Murthy et al., 2017) and RAB29/RAB7L1 (Pihlstrom et al., 2015) have been highlighted as likely PD-related genes in previous studies. The number of genes with altered expression seems low, as it would be expected that many of the possible ways genetic variants influence disease risk would be through gene-expression changes. This may be related to the relatively conservative approach used by the authors. Gene-expression data was taken from two publicly available sources that are based on different technologies, GTEx (RNA-seq) and Braineac (microarray). The authors used two methods, Coloc and TWAS, to test if gene-expression regulation was. Only genes that replicated across the two methods were included.

    Our group recently published results from a study using allelic expression profiling of genes located within PD-associated loci to identify cis-regulatory variation affecting gene expression (Langmyhr et al., 2021). Using this sensitive PCR-based method, allele-specific expression was identified for the majority of tested genes. This supports the hypothesis that changes to the cis-regulation of gene expression is a major mechanism behind a large proportion of genetic associations in PD.

    An intriguing finding by Kia and his colleagues was that the identified gene-expression changes were overall more prevalent in glial cells compared to neurons. This contradicts findings from recent genetic studies. We and others have found enrichment of PD risk variants in regulatory gene regions of neurons, indicating that neurons, and not glial cells, are the primary mediators of genetic risk for PD (Berge-Seidl et al., 2021).

    As the quality and quantity of gene expression data from brain tissue increase, including from single-cell experiments, we can expect a number of studies using similar approaches as Kia and co-workers. This will undoubtedly give new and important insights into disease mechanisms of PD.

    References:

    . Increased brain expression of GPNMB is associated with genome wide significant risk for Parkinson's disease on chromosome 7p15.3. Neurogenetics. 2017 Jul;18(3):121-133. Epub 2017 Apr 8 PubMed.

    . Fine mapping and resequencing of the PARK16 locus in Parkinson's disease. J Hum Genet. 2015 Jul;60(7):357-62. Epub 2015 Apr 9 PubMed.

    . Allele-specific expression of Parkinson's disease susceptibility genes in human brain. Sci Rep. 2021 Jan 12;11(1):504. PubMed.

    . Integrative analysis identifies bHLH transcription factors as contributors to Parkinson's disease risk mechanisms. Sci Rep. 2021 Feb 10;11(1):3502. PubMed.

Make a Comment

To make a comment you must login or register.

References

News Citations

  1. PWAS x GWAS? Proteome Analysis Nets 10 New Alzheimer’s Genes
  2. Epigenomic Roadmap Points to Causal Genes
  3. Expression, Expression, Expression—Time to Get on Board with eQTLs
  4. Largest Meta-GWAS Yet Uncovers New Genetic Links to Parkinson’s
  5. Researchers Build on GWAS to Parse Genetic Players in AD and PD
  6. Does LRRK2 Sweep α-Synuclein from the Cell?
  7. LRRK2 Interactions Identify New Parkinson’s Genes, Implicate Autophagy
  8. Lysosomes Take Center Stage in Parkinson’s and Frontotemporal Dementia
  9. Protecting Neurons by Ramping Up Waste Disposal?
  10. Doubling Down on Sequencing Serves up More Alzheimer’s Genes
  11. ApoE: Common Microglial Culprit in Aging, Alzheimer’s, and Tauopathy?
  12. Genetic Studies Uncover Four New ALS Genes

Paper Citations

  1. . A meta-analysis of genome-wide association studies identifies 17 new Parkinson's disease risk loci. Nat Genet. 2017 Sep 11; PubMed.
  2. . Increased brain expression of GPNMB is associated with genome wide significant risk for Parkinson's disease on chromosome 7p15.3. Neurogenetics. 2017 Jul;18(3):121-133. Epub 2017 Apr 8 PubMed.
  3. . The vicious cycle between α-synuclein aggregation and autophagic-lysosomal dysfunction. Mov Disord. 2020 Jan;35(1):34-44. Epub 2019 Nov 15 PubMed.
  4. . The glycoprotein GPNMB is selectively elevated in the substantia nigra of Parkinson's disease patients and increases after lysosomal stress. Neurobiol Dis. 2018 Dec;120:1-11. Epub 2018 Aug 24 PubMed.
  5. . A replication study of GWAS-genetic risk variants associated with Parkinson's disease in a Spanish population. Neurosci Lett. 2019 Nov 1;712:134425. Epub 2019 Aug 17 PubMed.
  6. . Association of three candidate genetic variants in ACMSD/TMEM163, GPNMB and BCKDK /STX1B with sporadic Parkinson's disease in Han Chinese. Neurosci Lett. 2019 Jun 11;703:45-48. Epub 2019 Mar 14 PubMed.
  7. . Polymorphisms of ACMSD-TMEM163, MCCC1, and BCKDK-STX1B Are Not Associated with Parkinson's Disease in Taiwan. Parkinsons Dis. 2019;2019:3489638. Epub 2019 Jan 2 PubMed.
  8. . CD38 in Neurodegeneration and Neuroinflammation. Cells. 2020 Feb 18;9(2) PubMed.
  9. . Integrative analysis identifies bHLH transcription factors as contributors to Parkinson's disease risk mechanisms. Sci Rep. 2021 Feb 10;11(1):3502. PubMed.

External Citations

  1. Braineac
  2. GTEx
  3. CMC

Further Reading

Papers

  1. . Identification of novel risk loci, causal insights, and heritable risk for Parkinson's disease: a meta-analysis of genome-wide association studies. Lancet Neurol. 2019 Dec;18(12):1091-1102. PubMed.

Primary Papers

  1. . Identification of Candidate Parkinson Disease Genes by Integrating Genome-Wide Association Study, Expression, and Epigenetic Data Sets. JAMA Neurol. 2021 Apr 1;78(4):464-472. PubMed.