Rationale
Converging evidence from GWAS, single-nucleus transcriptomics and EWCE consistently implicates microglia as the primary cell type mediating Alzheimer's genetic risk. The 29 Jansen et al. loci include canonical microglial genes (TREM2, INPP5D, CD33, MEF2C, MS4A) and the Murphy et al. pseudobulk reanalysis demonstrates that 96% of robust gene-expression changes in AD brains occur in microglia.
This PhD project addresses the next question: which specific microglial enhancers are activated in the disease-associated state, and where does IRF1 bind to drive this transition? Standard ChIP-seq cannot resolve this directly: commercially available anti-IRF1 antibodies cross-react across the IRF family (IRF1–9 share a highly conserved N-terminal DNA-binding domain), so peak calls cannot be unambiguously attributed to IRF1 alone. This motivates a Prime Editing + CETCh-seq strategy — epitope-tag a single endogenous IRF1 allele, then map with anti-FLAG.
Manhattan plot · 29 AD risk loci
Jansen 2019 meta-analysis · highlights microglial loci gwas_loci.jsonEach point is a genetic variant. Y-axis is −log10(p); magenta marks loci whose nearest gene is primarily expressed in microglia. APOE is capped at 60 for visual sanity.
Loci table
Sortable · click column headers| # | Gene | Chr | Lead SNP | OR | −log₁₀(p) | Microglial? |
|---|
26 DEGs survive FDR after pseudobulk correction, down from 23,923 in the original Mathys analysis. Twenty-five are microglia-resident.
DEG count by method
Pseudoreplication vs. pseudobulkRobust DEGs · effect sizes
Directionality per geneColour scale shows standard deviations above the mean of 10,000 bootstrap samples. Microglia are the top cell type for Alzheimer's at ζ = 4.1σ.
Aim timeline
2026-Q3 → 2030-Q1Identify IRF1-responsive microglial enhancers via ATAC-seq + H3K27ac overlay with the 29 Jansen loci.
Prime-edit a 3×FLAG tag into the endogenous IRF1 C-terminus in iPSC-derived microglia.
CETCh-seq the tagged line under baseline and DAM-induced conditions; intersect binding sites with GWAS loci.
Run stratified LD-score regression to partition AD heritability across IRF1-bound enhancers vs. control annotations.
Analysis pipelines
snRNA-seq reanalysis · GWAS annotation · Prime Editing → CETCh-seqMethods
Single-nucleus RNA-seq reanalysis
Raw counts from Mathys et al. (2019) (Synapse syn18485175) were
reprocessed via Scanpy with standard QC filters (≥ 500 genes/cell,
< 10% mitochondrial reads, doublet detection with scDblFinder) then
aggregated to per-sample pseudobulk using decoupler. Differential
expression used DESeq2 at sample level, with age, sex and pmi as covariates;
26 genes survive FDR < 0.05 across six brain regions. Method follows
Murphy et al. (2023).
GWAS meta-analysis + annotation
Summary statistics: Jansen et al. 2019 (N = 455,258, 71,880 AD cases & 383,378 controls). Lead SNPs extracted at p < 5×10⁻⁸; coordinates intersected with ENCODE cCREs and CATlas microglia ATAC peaks (bedtools intersect). Nearest-gene assignment via GENCODE v44 protein-coding transcripts.
Prime Editing design (PE)
pegRNAs designed with PRIDICT 2.0 for the IRF1 C-terminus. Insert: 3×FLAG tag (DYKDHDGDYKDHDIDYKDDDDK), preceded by a flexible GS linker. pegRNA and nicking sgRNA pairs are screened in silico for predicted efficiency ≥ 40 % and off-target CFD ≤ 0.1.
CETCh-seq protocol sketch
iPSC-derived microglia (Abud protocol) → Prime Editing co-electroporation → FACS for edited cells → clonal expansion → genotype QC (Sanger + NGS) → baseline / LPS+IFNγ-induced DAM states → ChIP with anti-FLAG beads → paired-end sequencing at ~30 M reads/sample → MACS2 peak calling → downstream motif + enrichment analysis.