Analysis of DNA methylation associates the cystine-glutamate antiporter SLC7A11 with risk of Parkinson’s disease

Parkinson’s disease (PD) is a neurodegenerative disorder characterized by Lewy body alpha-synuclein aggregates and progressive loss of dopaminergic neurons in the substantia nigra. Its prevalence rises with age (~1% over 60; 3–4% at 80) and is expected to more than double by 2040, creating substantial societal burden. PD etiology is complex, involving genetic and environmental factors, yet specific molecular mechanisms remain unclear. DNA methylation at CpG dinucleotides, a key epigenetic mechanism regulating gene expression, may mediate environmental and genetic influences on complex disease risk. Prior PD methylation studies (e.g., Chung/Chuang et al.) reported epigenome-wide significant CpGs but these did not remain significant after adjusting for blood cell composition, highlighting the need for larger, better-controlled studies. This study aims to identify robust blood-based DNA methylation associations with PD, assess their independence from cell-type composition and medication, and evaluate potential genetic versus environmental underpinnings, particularly focusing on loci such as SLC7A11 that may link to oxidative stress pathways relevant to PD.

Previous PD methylation work reported numerous epigenome-wide significant CpGs in relatively small cohorts, but associations did not withstand adjustment for blood cell type composition differences between cases and controls. There is growing recognition that blood cell composition, medication, and environmental exposures can confound methylation-disease associations. Larger samples (on the order of 2000–3000 individuals) have been suggested to robustly replicate findings. Environmental risk factors for PD, including pesticides and potential toxins such as β-methylamino-L-alanine (BMAA), have long been hypothesized to influence neurodegeneration via oxidative stress and excitotoxicity pathways. These contexts motivated the present meta-analysis across independent cohorts with rigorous adjustment for cell-type composition, and the application of summary-data Mendelian randomization to distinguish genetic from environmental contributions to methylation and expression changes at candidate loci.

- Cohorts: Two independent case–control cohorts with whole-blood DNA methylation profiled on the Illumina 450k array. Primary analyses focused on unrelated European-ancestry subsets. - SGPD (System Genomics of Parkinson’s Disease) discovery cohort: 1638 unrelated Europeans (851 PD cases, 787 controls) for main MWAS and cell-type analyses. - PEG (Parkinson’s Environment and Gene) replication cohort: 493 Europeans (281 PD cases, 212 controls). - DNA methylation QC and normalization: Background correction and normalization using internal controls; removal of probes with low detection rates, inadequate bead counts, and known cross-reactive probes; sample-level QC (sex checks, call rate, outliers). Post-QC probe counts included 242,205 CpGs in PEG; common probes across cohorts (~229k) were used for meta-analyses. - Cell type proportion (CTP) estimation: Blood CTPs imputed using the Houseman model; logistic regression tested PD status associations with individual CTPs adjusting for sex and predicted age. Explored relationships of levodopa equivalent daily dosage (LEDD) with CTPs in 494 SGPD PD cases with LEDD data, with and without adjustment for disease duration. - Association testing (MWAS): Linear mixed model-based methods implemented in OSCA were used to test associations between methylation at individual CpG probes and PD status, adjusting for sex, predicted age, predicted smoking exposure, and predicted CTPs. - MOA (mixed linear model Omic Association) and MOMENT (a more robust, slightly less powerful method that partitions probes by association strength) were both applied within cohorts. - Meta-analysis: Inverse-variance weighted fixed-effects meta-analysis of MOA and MOMENT summary statistics across SGPD and PEG for 229,017 common CpG probes (also noted ~229,071 in methods), with Bonferroni correction (p < ~2.2×10^-7 for MOA meta-analysis threshold as specified in results). - Mendelian randomization: Summary-data-based Mendelian randomization (SMR) coupled with the HEIDI test evaluated whether significant methylation signals had genetic instruments (mQTLs) and whether methylation at significant probes influenced nearby gene expression (eQTLs), and whether any observed PD association might be attributable to shared genetic instruments versus linkage or environmental/non-genetic effects. mQTL/eQTL resources included Brain-meta mQTL and Brain-eMeta eQTL datasets. - Medication effects: Tested association of cg06690548 methylation with PD medication dose (LEDD) within SGPD cases using MOA, adjusting for covariates, to evaluate possible medication confounding. - Variance component analyses and classification: Estimated proportion of variance in PD status captured by all methylation probes (σ²) using mixed models, and developed a methylation-based classifier trained in SGPD and evaluated in PEG using logistic regression and AUC as performance metric (adjusted for covariates).

- Blood cell type proportions (CTPs) differ between PD cases and controls (SGPD): - PD cases showed more granulocytes (p < 2×10^-10), fewer B cells (p = 2.16×10^-10), fewer helper T cells (CD4+ T p = 2.97×10^-15), and fewer CD8+ T cells (p = 5.29×10^-13). Natural killer cells were reduced (p = 7.53×10^-05). Findings were consistent with prior reports. - LEDD correlations in 494 SGPD PD cases: granulocytes r = 0.17 (p = 3.98×10^-05), CD4+ T r = -0.14 (p = 2.08×10^-03), CD8+ T r = -0.15 (p = 1.08×10^-03). These associations did not persist after adjusting for disease duration; CTP differences were similar between most- and least-exposed medication subgroups, suggesting medication is unlikely to explain the CTP differences. - Within-cohort MWAS: - SGPD MOA identified two epigenome-wide significant probes: cg16001422 (PLEC) on chromosome 8 (p = 2.3×10^-8) and cg26033520 (ASCC1) on chromosome 10 (p = 1.3×10^-7). These signals were not supported by MOMENT or replicated in PEG for PLEC. - PEG cohort analyses showed no inflation (λ ~1.01–1.04) and consistency in effect directions for a subset of suggestive probes. - Meta-analysis (MOA) across SGPD and PEG (1,132 PD cases; 999 controls; 229,017 probes): Two epigenome-wide significant probes were identified (Bonferroni threshold p < 2.18×10^-7): - cg06690548 (chromosome 4) near SLC7A11: beta = 1.09, SE = 0.20, p = 6.2×10^-8 (MOMENT: beta = 0.92, SE = 0.19, p = 2.3×10^-6). - cg26033520 (chromosome 10) near ASCC1: beta = -1.47, SE = 0.26, p = 1.6×10^-8 (MOMENT: beta = -1.01, SE = 0.25, p = 7.5×10^-5). - No evidence that these probes are affected by common SNPs within the critical probe sequence; no strong signals in known PD genes (e.g., SNCA). - SMR and HEIDI analyses: - cg06690548 had significant mQTLs; SMR indicated that hypermethylation at cg06690548 is associated with downregulation of neighboring SLC7A11 expression (SMR p = 3.59×10^-3), with no evidence of heterogeneity due to distinct causal variants (HEIDI p = 0.16). - There was no evidence of a genetic association between PD and either cg06690548 methylation or SLC7A11 expression, implying the PD–methylation association is unlikely to be genetically driven and may reflect environmental exposure or disease consequence. - cg06690548 methylation was not associated with PD medication dose in SGPD cases (b = 0.50, SE = 0.37, p = 0.17). - Variance and classification: - Proportion of variance in PD status associated with all methylation probes was high: σ² ≈ 0.28 (SGPD) and 0.24 (PEG). In PEG, σ² decreased to 0.21 when SGPD epigenome-wide significant probes were modeled as fixed effects, suggesting much variance remains unexplained. - A methylation-based classifier trained in SGPD achieved AUC = 0.70 (95% CI: 0.66–0.75) in PEG, indicating moderate case–control discrimination.

This study identified robust blood-based DNA methylation associations with PD, notably at cg06690548 near SLC7A11 and cg26033520 near ASCC1. The SLC7A11 finding is biologically plausible: SLC7A11 encodes the xCT subunit of system Xc−, a cystine–glutamate antiporter that regulates intracellular cystine uptake and glutathione synthesis, as well as glutamate export. Hypermethylation at cg06690548 associates with downregulation of SLC7A11, consistent with reduced glutathione levels and increased oxidative stress—mechanisms long implicated in dopaminergic neuron vulnerability in PD. Importantly, SMR/HEIDI indicated that the PD association at cg06690548 is unlikely to be explained by shared genetic variants affecting methylation or expression, pointing toward environmental exposure or disease-related processes. One environmentally relevant hypothesis involves β-methylamino-L-alanine (BMAA), a cyanobacterial neurotoxin reported to target system Xc−, potentially leading to cystine uptake reduction, glutathione depletion, oxidative stress, and glutamate-mediated excitotoxicity. While this study does not establish causality or directly implicate BMAA in PD, it provides convergent support for glutathione/glutamate dysregulation. The ASCC1 association may relate to immune and transcriptional regulation (e.g., NF-κB pathways), aligning with evidence for neuroinflammation in PD. Differences in blood cell composition between PD cases and controls were replicated but do not appear to be driven by medication exposure; nonetheless, rigorous CTP adjustment was applied to minimize confounding. The relatively high variance in PD status associated with methylation and the moderate out-of-sample classifier performance suggest substantial, but not necessarily causal, methylation differences between cases and controls. Further work is needed to disentangle cause from consequence, explore tissue specificity (brain versus blood), and clarify environmental contributions.

The study reports two previously unrecognized blood DNA methylation associations with PD at SLC7A11 (cg06690548) and ASCC1 (cg26033520). Integrative SMR analyses indicate that the SLC7A11 association likely reflects non-genetic influences, consistent with environmentally mediated pathways impacting glutathione homeostasis and oxidative stress. These findings highlight glutamate signaling, oxidative stress, and neuroinflammation as potential etiological mechanisms in PD and nominate SLC7A11 as a plausible biological target. Future research should prioritize larger PD MWAS, replication across diverse populations, assessments in disease-relevant brain tissues, longitudinal designs to evaluate temporality and causality, and targeted investigation of environmental exposures (e.g., BMAA, pesticides) that may modulate SLC7A11-related pathways.

- Tissue source: Whole-blood methylation may not reflect brain-specific epigenetic alterations central to PD pathology. - Confounding by cell type composition and medication: Although analyses adjusted for predicted CTPs and medication effects were explored, residual confounding cannot be fully excluded; conservative CTP adjustment may have reduced power and obscured true signals. - Statistical methods: Epigenome-wide significance was reached with MOA (more powerful but potentially higher false-positive rate) and not always with MOMENT (more robust but less powerful), raising caution about False Discovery Rate versus power trade-offs. - Replication: Some discovery signals (e.g., PLEC cg16001422) did not replicate or were not supported by MOMENT. - Causality: Cross-sectional design limits inference about cause versus consequence of methylation changes; SMR suggests non-genetic mechanisms but cannot specify specific environmental exposures. - Sample and probe inconsistencies: Minor discrepancies in probe identifiers and counts across sections may reflect QC filtering differences; larger, harmonized datasets are needed to resolve borderline associations.