Extended Data Fig. 3: The effects of the LOAD risk-allele on chromatin accessibility, PICALM expression and PU.1 binding.

(a) Human brain OCR peaks in different cell types in comparison with our iMG at the PICALM locus. Note the rs10792832-flanking peak (highlighted in the vertical yellow box) identified in our iMG is only present in brain microglia but not in other cell types. The adjacent ATAC-seq read pileup plot (right) for rs10792832 shows significantly more ATAC-reads of allele A than allele G (i.e., ASoC) in brain microglia. Except for our iMG, data of all other cell types are from a snATAC-seq study²⁴ and reanalysed by using 10x Genomics Cell Ranger ARC 2.0.2. Ex = excitatory neurons, In = inhibitory neurons, Ast = astrocytes, oligo = oligodendrocytes, OPC=oligodendrocyte precursor cells, Vas = vascular cells. (b) IF staining of the CRISPR-engineered isogenic iMG (left panels) shows high and comparable purity (TREM2⁺/CD45⁺/PU.1⁺) (right panel). Each datapoint represents single-well measurement; for each condition (risk and non-risk), data are from two independent donor lines each with one clone from one experiment (n = 2). Scale bar: 50 µm. (c) The representative image of iMG stained with P2RY12+ and IBA1+ (left) and quantification of iMG purity (right). (d) The representative image of iMG stained with TMEM119+ and IBA1+ (left) and quantification of iMG purity (right). In (c) and (d), each datapoint represents a single-well measurement from one experiment; for each condition (risk and non-risk), data are from two independent donor lines (CD04 and CD09) each with 2 clones from one experiment with 3 wells of differentiations (n = 12). Scale bar, 50 µm. Two-sided unpaired Student’s t-test. ns = not significant. (e) qPCR shows reduced PICALM expression in risk-allele iMG. Each datapoint represents a single-well measurement from one experiment; for each donor line (CD04 and CD09) and each condition (risk and non-risk), data are from 2 clones, collected across 3 independent experiments each with 3 wells of differentiations for clone 1 and 2 experiments each with 3 wells of differentiations for clone 2 (n = 15). (f) qPCR result showing that the risk allele of rs10792832 does not affect PICALM expression in iAst. Each datapoint represents a single-well measurement from one experiment; for each donor line (CD04 and CD09) and each condition (risk and non-risk), data are from 2 clones, collected from 2 independent experiments each with 3 wells of differentiations (n = 12). In both (e) and (f), expression was normalized to GAPDH. Linear Mixed Model (LMM) was applied to test the fixed effect of genotype, with experimental round and clone identity as nested random factors; two-sided test. (g) PU.1 antibody validation in PMP cells from which iMG were further differentiated. (h) Schematic design of ChIP-qPCR to assay PU.1 binding. The illustration was created using BioRender. Sudwarts, A. (2025) https://BioRender.com/nux3l58. (i) The LOAD risk-allele G of rs10792832 shows reduced PU.1 binding in qPCR of ChIP products. n = 2 experiments for each line. Homozygous iMG was used to compare the allelic effect. (j) The allele A of rs10792832 exhibits higher Sanger sequencing peaks for PU.1 ChIP-seq PCR products of iMG from two iPSC lines heterozygous for rs10792832. Note the equal peak height of the two alleles for genomic PCR products (sequenced by using primers for on-target editing confirmation) of the same heterozygous samples as in the top panels. (k) Quantification of the allelic ratios of rs10792832 for Sanger sequencing peaks in (j). In all comparisons. * P < 0.05, ** P < 0.01, ***, P < 0.001; mean ± s.e.m.