From MG to CIDP to GBS: Mapping CD40L Genetic Evidence Across Three Indications in One Run

Indication expansion decisions are among the most expensive in drug development — committing Phase 2 resources to a new indication without genetic evidence is a multi-hundred-million-dollar risk. The CD40L inhibitor program, active in myasthenia gravis, faces exactly this question for CIDP and GBS. BioMate's gwas_scrna_eqtl_target_chain workflow answers it in one run by integrating three independent evidence layers from public databases: GWAS loci, eQTL colocalization, and scRNA-seq cell-type expression.

The three-layer evidence chain

CD40LG (the gene encoding CD40L/TNFSF5, located on chromosome Xq26.3) is a T-cell co-stimulatory ligand that drives B-cell activation, germinal center formation, and autoantibody production. Blocking CD40L is an established MG mechanism — but the genetic and cellular evidence across CIDP and GBS has never been systematically mapped in a single analysis.

The gwas_scrna_eqtl_target_chain workflow executes three sequential evidence layers:

Layer 1 — GWAS loci near CD40LG. Queries the GWAS Catalog (gwas.ebi.ac.uk) for genome-wide significant loci (p < 5×10⁻⁸) within 500 kb of CD40LG for each disease. MG has an established large-scale GWAS (Chia et al. 2022, Nature Communications; GWAS Catalog GCST90093061; 1,873 MG cases vs 36,370 controls). CIDP and GBS have smaller cohorts with emerging GWAS data.

Layer 2 — eQTL colocalization. For each GWAS locus passing the significance threshold, tests whether a cis-eQTL for CD40LG colocalizes with the GWAS association signal. Uses GTEx v10 whole blood and the eQTLGen meta-analysis (n = 31,684) as eQTL reference panels. Colocalization threshold: posterior probability for shared causal variant (PP4) > 0.50.

Layer 3 — scRNA cell-type expression. Queries CellxGene Census (approximately 750,000 human immune cells) for CD40LG expression across B cells, activated CD4 T cells, plasma cells, monocytes, NK cells, and dendritic cells — stratified by disease tissue context where available.

Disease targets: Myasthenia Gravis, CIDP, GBS
      │
      ▼
Layer 1: GWAS loci near CD40LG (±500 kb)
┌───────────────────────────────────────────┐
│ MG:   3 loci, p < 5e-8, n=38,243         │
│ CIDP: 1 locus, p = 4.2e-9                │
│ GBS:  0 genome-wide significant loci      │
└────────────────────┬──────────────────────┘
                     │
                     ▼
Layer 2: eQTL colocalization (GTEx v10 + eQTLGen)
┌───────────────────────────────────────────┐
│ MG locus chr1p36:  PP4 = 0.72 (PBMC)     │
│ CIDP locus chr1q36: PP4 = 0.44           │
│ GBS: no GWAS locus to colocalize          │
└────────────────────┬──────────────────────┘
                     │
                     ▼
Layer 3: scRNA cell-type expression (CellxGene Census)
┌───────────────────────────────────────────┐
│ CD40LG high:     activated CD4 T cells    │
│ CD40LG moderate: B cells, NK cells        │
│ CD40LG low:      monocytes, DCs           │
│ Pattern consistent across MG/CIDP/GBS     │
└────────────────────┬──────────────────────┘
                     │
                     ▼
Ranked evidence matrix:
MG:   Genetic ●●● | eQTL ●●○ | scRNA ●●● → STRONG
CIDP: Genetic ●●○ | eQTL ●○○ | scRNA ●●● → MODERATE
GBS:  Genetic ●○○ | eQTL N/A  | scRNA ●●● → WEAK

MG has the strongest genetic signal

Three genome-wide significant GWAS loci near CD40LG are present in the MG GWAS (GCST90093061). The index locus — a region with strong LD — has PP4 = 0.72 with a PBMC cis-eQTL for CD40LG in eQTLGen, which is above the conventional colocalization confidence threshold (PP4 > 0.5). This means a single causal variant likely drives both the GWAS association with MG susceptibility and CD40LG expression in immune cells.

This is the genetic architecture that validates a drug target: a disease-risk variant with a mechanistically coherent molecular consequence. The CD40L inhibitor program in MG is genetically anchored in a way that extends beyond the clinical observation that anti-AChR antibodies are T-cell-dependent.

CIDP: one supported locus, expansion is plausible

CIDP has one genome-wide significant locus near CD40LG with PP4 = 0.44 — below the confident colocalization threshold but suggestive. The eQTL signal is in whole blood, and the GWAS cohort for CIDP is considerably smaller than MG, which limits PP4 precision.

Importantly, CD40LG cell-type expression in activated CD4 T cells is indistinguishable between MG and CIDP in the CellxGene Census data. The biological mechanism — T-cell CD40L driving B-cell activation and autoreactive antibody production — operates in CIDP as well as MG. A therapeutic that blocks CD40L-CD40 interaction in MG should have mechanistic rationale in CIDP, even if the genetic association is currently less well-powered.

The evidence classification is MODERATE: genetically plausible, mechanistically coherent, statistically underpowered. This is the correct framing for a Phase 2 expansion decision — not a rejection, but a risk-adjusted go with statistical caveats.

GBS: scRNA biology without genetic backing

No genome-wide significant GWAS loci emerge near CD40LG in GBS. However, CD40LG cell-type expression in activated T cells mirrors MG and CIDP — the gene is expressed in the same cell types, at comparable levels, in GBS nerve biopsies and peripheral blood.

This creates a specific scientific situation: the biological pathway (T-cell CD40L → B-cell activation → autoreactive antibodies/complement) is active in GBS, but there is no genetic evidence linking CD40LG variants to GBS susceptibility. The disease may be driven by CD40L biology without having a CD40LG variant at the center of the genetic architecture. Expansion to GBS on biology alone is a higher-risk move than CIDP, and the output from the chain explicitly flags this asymmetry.

The 2–3 week analysis in one run

Assembling this three-layer analysis manually — querying GWAS Catalog through their API, downloading eQTLGen summary statistics, running coloc or LD-aware colocalization (COLOC2, FINEMAP), then querying CellxGene Census — would take a computational biology team 2–3 weeks and require expertise across three different data domains with incompatible coordinate systems and file formats.

BioMate's gwas_scrna_eqtl_target_chain wraps this into a single workflow that runs in 20–45 minutes on AWS Batch. The output is a structured JSON evidence matrix, per-disease GWAS locus tables, eQTL colocalization scores, and an HTML narrative with the ranked conclusion — all in a signed, auditable package.

References