Parkinson's risk genes gate fibril uptake at the cell surface

What the work claims

This is a primary mechanistic study, still a preprint. Karpilovsky, Goiran and colleagues at the Montreal Neurological Institute ran a genome-wide CRISPR activation screen for genes that change how much fluorescently labeled alpha-synuclein preformed fibril (PFF) a cell internalizes, cross-referenced the hits against Parkinson's disease GWAS loci, and followed the single overlapping gene, BAP1, into mechanism and genetics.¹

The claims, ordered by strength. Mechanistically: BAP1, a deubiquitinase in the Polycomb repressive deubiquitinase (PR-DUB) complex, acts as a transcriptional master regulator of the cell surface glycoproteome; raising BAP1 lowers heparan sulfate proteoglycans (HSPGs) and raises O-linked, mucin-type glycosylation at the plasma membrane, and the combination reduces fibril binding and entry. Genetically: Mendelian randomization links lower BAP1 expression to higher Parkinson's risk. Cell-type specificity: the protective effect holds in iPSC-derived dopamine neurons but not in microglia, and a midbrain organoid carrying integrated microglia shows the likely reason, sharply lower ASXL3 expression in microglia; ASXL3 is itself a PD risk gene and the PR-DUB partner that directs where BAP1 acts.¹

How it works

The screen used retinal pigment epithelial cells expressing dCas9-VPR, infected with the genome-wide hCRISPRa-v2 library, sorted into the top and bottom 15 percent of fibril fluorescence, and analyzed with MAGeCK. The top enriched pathway term was glycosaminoglycan biosynthesis, matching the known role of HSPGs as fibril docking sites, and BAP1 was the only gene among the top hundred bidirectional modifiers that also sat in a PD GWAS locus. In follow-up experiments, CRISPRa activation of BAP1 cut fibril accumulation measured by flow cytometry and high-content microscopy, while siRNA knockdown increased it; the effect was specific, with no change in uptake of dextran, EGF, transferrin or Tau oligomers, and pHrodo labeling confirmed the missing fibrils are the ones that would have reached acidic endolysosomal compartments.¹

Disaccharide analysis by mass spectrometry showed a global drop in heparan sulfate chains, without any specific sulfation-pattern change and with no effect on chondroitin sulfate proteoglycans. RNA sequencing pointed the other way at the membrane too: mucins MUC3A, MUC12 and MUC17 were massively upregulated, and cell-surface biotin pulldown proteomics found MUC12 the most increased surface protein (log fold change above 4). Lectin microarrays read the shift as about 50 percent lower mannose-type lectin binding and about 50 percent higher O-glycan lectin binding, with wheat-germ-agglutinin staining redistributing away from the Golgi into surface ridges. Two perturbations settle causality in opposite ways: stripping O-glycoproteins with the protease IMPa restored fibril uptake in BAP1-activated cells back to control levels, while merely truncating O-glycan chains with benzyl-alpha-GalNAc did not. Within-culture correlation told the same story cell by cell: cells staining highest for the truncated Tn antigen took up the least fibril.¹

The genetics are Mendelian randomization with colocalization. In GTEx, BAP1 expression is highest in regions spared by Parkinson's (cortex, cerebellum) and lowest in vulnerable ones (substantia nigra, basal ganglia). Variants rs123598 and rs34332947 sit in complete linkage disequilibrium and track lower BAP1 expression across four brain regions; summary-data-based Mendelian randomization gave a negative effect consistent with higher expression being protective (SMR beta -0.114, P = 2.0 x 10^-3; the risk allele raises disease risk at P = 5.4 x 10^-4 while lowering expression at P = 7.3 x 10^-12), with no heterogeneity on the HEIDI test (P = 0.16) against a cortex eQTL resource of 2,443 samples and a PD GWAS of about 63,555 cases with 1.75 million controls.¹

The cell-type split is where the organoid work comes in. In iPSC-derived dopamine neurons, BAP1 activation reduced fibril uptake and, after a three-week seeding protocol, reduced phosphorylated alpha-synuclein puncta. In iPSC-derived microglia the same manipulation did nothing at 3, 6 or 24 hours. Single-cell RNA sequencing of midbrain organoids with integrated microglia showed ASXL3 expression in microglia running far below neuronal levels, especially below dopamine neurons (FDR = 4.9 x 10^-65, log fold change -3.3); since ASXL proteins are mutually exclusive PR-DUB partners that steer BAP1 to its genomic targets, low ASXL3 offers a concrete explanation for the cell-type split. Knockdown phenocopy matched: ASXL3, but not ASXL1 or ASXL2, increased fibril uptake. Finally, in five-month-old SNCA-triplication midbrain organoids, where the same group had previously shown loss of vulnerable substantia-nigra-like dopamine neurons and sparing of ventral-tegmental-area-like ones, both BAP1 and ASXL3 met the study's significance cutoff for higher expression in the resilient population.¹

Where a skeptic should push

The load-bearing assumption is that fibril uptake is the right proxy for what kills neurons. The paper shows less seeding (fewer pSyn puncta) after BAP1 activation in dopamine neurons, but it does not show that the SNCA-triplication organoid's selective neuronal loss is rescued by manipulating BAP1 or ASXL3. The vulnerability correlation in organoids is expression-level and associative; it becomes causal only when someone perturbs the axis in that model and changes which neurons die.¹

Second, the genetics are softer than the abstract implies. BAP1 was nominated by a 2017 PD GWAS but did not reach genome-wide significance in subsequent GWAS; the authors' own framing is that fine mapping, colocalization and Mendelian randomization revive it. That is a legitimate route, but a reviewer should hold the causal claim at "supported" rather than "established." Third, translation tension the paper understates: BAP1 is a tumor suppressor, with loss-of-function driving mesothelioma, renal and uveal cancers, so a chronic-elevation strategy aimed at the brain inherits an oncologic liability that any drug program would have to engineer around. Finally, most wet-lab experiments run at n = 3 to 4 biological replicates, and the whole mechanistic chain starts in an RPE1 cell line, a retinal epithelial line, extrapolated to dopamine neurons.¹

What assembloids change in uptake-targeted screens

For organoid-based drug discovery, the deepest lesson is not the target but the assay. This study hit a result that a single-cell-type program would have read as a contradiction: the same genetic manipulation cut fibril entry in dopamine neurons and did nothing in microglia. The resolution came from a midbrain organoid carrying integrated microglia, where single-cell sequencing showed the missing PR-DUB partner, ASXL3, is barely expressed in microglia. That is what assembloids are for in practice: not anatomical completeness, but a veto on cell-type-specific mechanisms that 2D screens silently generalize. A compound that blocks HSPG binding or remodels surface glycosylation would look efficacious in a neuron-only assay, look like a clean failure in a microglia assay, and only an assembloid tells you both are true at once. Screening programs chasing anti-spread therapies should treat cell-type composition of the model as a design variable that decides which mechanisms are even visible, and should expect neuron-selective hits to disagree across assay formats.¹

The opportunity is a new druggable surface. The fibril-entry field has fixated on HSPGs; this work adds O-linked glycosylation as a second, opposing lever that is transcriptionally gated and reversible, and demonstrates a genetic proof of concept that raising the barrier cuts seeding. An organoid platform that reports fibril uptake and pSyn seeding per cell type is exactly the tool for finding small molecules that mimic the effect without touching BAP1 itself, which is the right instinct given BAP1's day job as a tumor suppressor. The vulnerability map is the second opportunity: SNCA-triplication organoids sort dopamine neurons into a doomed substantia-nigra-like population and a spared ventral-tegmental-area-like one, and BAP1/ASXL3 expression tracks that split. Organoids are quietly becoming atlases of selective vulnerability, and target validation programs can now ask not just "does our target matter" but "does it matter in the neurons that actually die."¹

The threats deserve equal weight. Mechanistically, blocking uptake slows spread; it does nothing about aggregates already inside neurons, so the therapeutic window is prophylactic and progression-modifying at best. And the dual-use note writes itself: the same screen that found a protective axis would, run in the opposite direction, identify genes that accelerate fibril entry, including into microglia, which the paper shows sit outside the BAP1 mechanism entirely.¹

The bottom line

Established: a genome-wide screen nominating BAP1 as a modifier of alpha-synuclein fibril internalization; a two-pronged surface mechanism (less heparan sulfate, more O-linked glycosylation) with protease-rescue causality; cell-type specificity between dopamine neurons and microglia; and an ASXL3-based explanation visible only in a microglia-containing midbrain organoid. Supported but not established: that lower BAP1 expression causally raises PD risk (Mendelian randomization, not replication-grade GWAS), and that the BAP1/ASXL3 axis drives selective dopamine-neuron vulnerability rather than marking it. What would confirm it: perturbing BAP1 or ASXL3 inside SNCA-triplication organoids and shifting which dopamine-neuron population survives. What would break it: a lineage-traced demonstration that uptake is irrelevant to neuronal death in that model.¹