A Parkinson risk axis that protects neurons but ignores microglia

What the work claims

This is a primary mechanistic preprint, and its central claim has several load-bearing parts. First, a genome-wide CRISPR activation screen in retinal pigment epithelial cells (RPE1) carrying dCas9-VPR, using the hCRISPRa-v2 library and fluorescently labelled alpha-synuclein preformed fibrils, nominated BAP1 as the single gene whose activation reduced fibril uptake and which also sits in a Parkinson's disease GWAS locus. The strongest enriched pathway among screen hits was glycosaminoglycan biosynthesis, matching prior work on heparan sulfate proteoglycans (HSPGs) as fibril entry factors¹.

Second, the authors assemble human genetic evidence that BAP1 expression is causally protective: in GTEx, BAP1 is expressed highest in regions spared in Parkinson's (cortex, cerebellum) and lowest in vulnerable ones (substantia nigra, basal ganglia); two fully linked variants, rs123598 and rs3432947 (D' = 1, r squared = 1), track reduced BAP1 expression in nucleus accumbens, caudate, hippocampus and cortex; and summary-data Mendelian randomization finds the risk allele raises disease risk while lowering BAP1 expression, with no heterogeneity on the HEIDI test (P = 0.16)¹.

Third, the mechanism: BAP1, with its PR-DUB partner ASXL3, transcriptionally remodels the cell surface, lowering HSPGs and raising O-linked glycoproteins such as mucins, and this two-pronged shift reduces fibril binding. And fourth, the kicker: the effect is cell-type specific. Overexpressing BAP1 cuts fibril uptake in induced pluripotent stem cell (iPSC) dopamine neurons and reduces seeded phosphorylated-synuclein pathology after three weeks, but does nothing in iPSC microglia¹.

How it works

BAP1 is the catalytic deubiquitinase of the Polycomb Repressive-Deubiquitinase (PR-DUB) complex, which removes ubiquitin from histone H2A to steer transcription. The complex's genomic targeting is set by mutually exclusive ASXL scaffolding proteins (ASXL1, ASXL2, ASXL3). In RPE1 cells, activating BAP1 with two independent guide RNAs reduced fibril accumulation across 3 to 24 hours by flow cytometry and microscopy (3 to 7 biological replicates), while siRNA knockdown of BAP1 increased uptake (n = 6), a bidirectional effect that follows the direction of the genetic risk¹.

The effect is specific: internalization of dextran, EGF, transferrin and even Tau oligomers was statistically unchanged, and a pH-sensitive label confirmed reduced fibrils reach acidic endolysosomal compartments. Cell-surface proteomics, lectin microarrays and RNA sequencing converge on the mechanism: BAP1 reduces HSPG biosynthetic machinery and increases O-linked glycosylation carriers. Chemically stripping O-glycans with benzyl-alpha-GalNAc reversed the protection, showing both glycan shifts are required¹.

The cell-type split is explained by the scaffold. In single-cell data from human midbrain organoids with integrated microglia, ASXL3 is far more abundant in dopamine neurons than in microglia (log fold change about minus 3.3), while BAP1, ASXL1 and ASXL2 show no significant cell-type difference. Knocking down ASXL3, but not ASXL1 or ASXL2, phenocopied BAP1 loss and increased fibril uptake. In five-month-old organoids derived from a patient with SNCA triplication versus an isogenic corrected control, both BAP1 and ASXL3 are expressed higher in the resilient VTA-like dopamine neurons than in the vulnerable SN-like population, at a stringent differential-expression cutoff (FDR below 0.05, absolute log fold change above 1.5)¹.

Where a skeptic should push

The most load-bearing assumption is that BAP1 is genuinely a Parkinson's risk gene. The authors are candid that BAP1's locus was nominated in the 2017 PD GWAS but did not reach genome-wide significance in the 2019 or 2025 follow-ups. The case here rests on expression quantitative trait loci, colocalization and Mendelian randomization, which argue that lower BAP1 expression causally raises risk, but that is a different and softer claim than a replicated association. Treat the genetics as suggestive, not settled¹.

Second, the donor base is thin. The dopamine neuron and microglia experiments each use one healthy-donor iPSC line, with n = 3 to 4 biological replicates per assay. The midbrain organoid findings are re-analyses of previously published single-cell datasets from companion studies, not new perturbations performed in organoids: no one has yet activated BAP1 inside a microglia-containing assembloid and watched what happens to spread. The ASXL3 causal step was demonstrated in RPE1 cells, a convenient epithelial line that is neither neuronal nor microglial. And everything is in vitro: no animal has been protected from pathology¹.

Finally, the therapeutic framing needs a hard look. BAP1 is a bona fide tumor suppressor; germline loss drives cancer predisposition syndromes. A strategy that raises BAP1 activity system-wide inherits that safety question entirely.

Cell-type specificity as an organoid assay trap

The transferable finding for organoid science is not the gene; it is the demonstration that a genetic modifier of protein-aggregate uptake can be strong in one neural cell type and invisible in another sitting millimetres away in the same tissue. Microglia are the brain's professional phagocytes and a principal route of alpha-synuclein clearance and spread; a neuroprotective candidate that reduces neuronal uptake while leaving microglial uptake untouched would change where, not whether, pathology circulates. Any uptake or spread assay run in a neuron-only midbrain organoid is structurally blind to this failure mode. The practical consequence is that for spread-relevant readouts, the assayable unit should be a microglia-integrated assembloid with cell-type-resolved measurement, such as sorted populations or single-cell profiling, rather than a bulk fluorescence endpoint¹.

The opportunity is equally concrete. The paper's protective mechanism is a druggable surface phenotype: the HSPG-versus-O-glycan balance, measurable by lectin microarray and cell-surface proteomics. That is exactly the kind of pharmacodynamic readout an organoid screening platform can industrialize, because it reports on the drug's mechanism rather than on downstream viability. Platforms that can titrate microglial inclusion and resolve cell types in the same well would turn this one-off observation into a screenable, decision-grade assay. The maturation caveat matters too: in the organoid data, BAP1 and ASXL3 expression differs between neuron subtypes and shifts with age, so passage, culture duration and regional identity could flip the apparent efficacy of any modifier tested in an immature organoid. Treat culture age as a variable, not a constant¹.

The hype-correction: BAP1 activation as a Parkinson's therapy is a hypothesis at the far end of a long translational chain, and the safety shadow of BAP1's tumor-suppressor biology is real. What this paper credibly delivers is a mechanism and a warning about how not to measure it.

The bottom line

Established: BAP1 bidirectionally regulates alpha-synuclein fibril uptake in vitro through a two-pronged remodeling of the cell surface glycoproteome, its expression tracks the geography of Parkinson's vulnerability, and the effect is dopamine-neuron specific, plausibly because ASXL3 is absent from microglia. Hypothesis: this axis meaningfully modifies disease risk or progression in patients, and manipulating it is therapeutic. What would confirm it: BAP1 or ASXL3 perturbation inside microglia-containing midbrain assembloids, replicated across donor lines, with spread rather than just uptake as the endpoint, followed by in vivo validation. What would break it: failure of the effect to replicate in independent donor lines, or loss of the cell-type split in more mature assembloids, which would suggest the specificity is a culture artifact.