Genome-scale mapping of variant, enhancer and gene function in primary human CD4+ T cells

By integrating targeted and genome-wide Perturb-seq across 4.1 million primary human CD4+ T cells, this study maps the functional relationships between immune disease-associated noncoding variants, their target cis-regulatory elements, and downstream gene regulatory networks to elucidate the molecular mechanisms underlying immune diseases.

The Gist

Imagine your body's immune system as a massive, bustling city. The CD4+ T cells are the city's elite security guards and emergency responders. They are constantly on patrol, ready to react to threats like viruses or bacteria.

Now, imagine that sometimes, the city's blueprint (your DNA) has tiny typos or smudges. These are called genetic variants. In people with autoimmune diseases (like Crohn's, Multiple Sclerosis, or Rheumatoid Arthritis), these smudges are often found in the "instruction manuals" that tell the security guards how to behave, rather than in the guards themselves.

The big mystery scientists have faced for years is: Which specific smudge on the blueprint causes the security guard to go rogue? And more importantly, which specific instruction does that smudge mess up?

This paper is like a massive, high-tech detective operation that finally solves this mystery on a city-wide scale. Here is how they did it, explained simply:

1. The Problem: A City Too Big to Map

Previously, scientists knew where the smudges were (on the DNA), but they didn't know what they controlled. It's like finding a typo in a massive library of books but not knowing which sentence in which book it changes.

• The Challenge: These "instruction manuals" (called Enhancers) can be miles away from the "guards" (genes) they control. They are hidden in the dark, and we didn't have a flashlight to see which manual controls which guard.

2. The Solution: The "Switch-Off" Experiment

The researchers decided to play "Pin the Tail on the Donkey" but in reverse. They used a tool called CRISPR (think of it as a pair of molecular scissors) to systematically turn off specific instruction manuals (enhancers) in the security guards' cells.

They did this in two massive waves:

• Wave 1 (The Targeted Hunt): They picked 1,032 specific instruction manuals known to be linked to 14 different immune diseases. They turned them off one by one and watched what happened to the nearby guards. This is like turning off a specific light switch in a house and seeing which lights in the living room flicker.
• Wave 2 (The City-Wide Sweep): Then, they turned off every single guard (gene) in the cell to see how the whole city reacted. This helped them map the entire chain of command.

3. The Big Discovery: Connecting the Dots

By looking at 4.1 million cells (a number so big it's like counting every grain of sand on a beach), they built a complete map.

• The "Who's Who" List: They found 626 specific pairs where turning off an instruction manual directly changed a guard's behavior.
• The "Long-Distance" Call: They proved that some instructions are written in a different neighborhood but still control guards miles away. They confirmed that these long-distance connections are real and not just random noise.
• The "Chain Reaction": They didn't just stop at the first guard. They followed the chain.
  • Example: They found a smudge linked to Crohn's disease. It turned off a specific instruction manual, which silenced a gene called TYK2. When TYK2 went silent, the whole "Fire Alarm" system (inflammatory response) in the cell went haywire.
  • Example: They solved a mystery about Multiple Sclerosis. There were two genes, DEXI and CLEC16A, right next to each other. Scientists were confused about which one the disease smudge was attacking. The experiment showed clearly: It's DEXI. The smudge turns off DEXI, which then messes up the cell's ability to transport food (amino acids), causing the disease.

4. Why This Matters: From "Where" to "How"

Before this study, doctors could say, "You have a genetic risk for this disease."
Now, thanks to this map, they can say: "You have a smudge in instruction manual #405, which silences guard #202, causing the city to overreact to harmless pollen."

The Analogy of the "Ripple Effect":
Think of the DNA as a pond.

• The Variant is a stone thrown in.
• The Enhancer is the first ripple.
• The Gene is the second ripple.
• The Disease is the wave crashing on the shore.

This paper didn't just find where the stone was thrown; it traced the ripples all the way to the shore. They showed that even though the stones are thrown in different places across the pond, many of them create waves that crash in the same spots (shared disease pathways), while some create unique, specific waves (disease-specific pathways).

The Bottom Line

This research is like giving scientists a GPS and a wiring diagram for the human immune system. It explains exactly how tiny errors in our DNA lead to big problems in our bodies. This is a huge step forward for developing new drugs that don't just treat the symptoms, but fix the specific broken wire in the circuit.

Technical Summary

1. Problem Statement

Genome-wide association studies (GWAS) have identified thousands of genetic variants associated with complex immune diseases. However, over 60% of these variants reside in non-coding cis-regulatory elements (CREs) such as enhancers and silencers. Translating these statistical associations into mechanistic understanding faces three major hurdles:

1. Linkage Uncertainty: It is difficult to reliably infer which specific gene a non-coding CRE regulates based solely on genomic sequence or proximity.
2. Cell-Type Specificity: CRE activity is highly dependent on cellular context (e.g., T cell activation state), and many disease-relevant loci are only active in specific immune states.
3. Scalability and Sensitivity: Existing methods (eQTLs, reporter assays, or standard CRISPR screens) often lack the sensitivity to detect weak enhancer effects or the scale to map genome-wide regulatory networks in primary human cells.

2. Methodology

The authors employed a multi-stage, integrative strategy combining in silico prioritization with large-scale experimental perturbation in primary human CD4+ T cells (a key context for immune disease). The study utilized three distinct but integrated screens across 4.1 million cells from three healthy donors, profiled 3 days post-TCR stimulation (CD3/CD28).

A. Study Design & Prioritization (Variant-to-CRE)

• Target Selection: The authors intersected GWAS variants from 14 immune diseases (including IBD, MS, RA, etc.) with ATAC-seq data from activated CD4+ T cells to identify 1,032 disease-relevant CREs overlapping 4,724 variants.
• Candidate Gene Prediction: To define target genes for the CRE screen, they used five orthogonal prediction strategies to maximize recall:
  1. Proximity (<100 kb).
  2. Nearest genes outside the 100 kb window.
  3. CD4+ T cell-specific Gene Regulatory Networks (GRNs).
  4. ENCODE-rE2G predictions.
  5. CD4+ T cell eQTL links.
  • Result: 4,928 candidate CRE-gene pairs were generated.

B. Experimental Screens

1. CRE Screen (TAP-seq):

   • Technique: Targeted Perturb-seq (TAP-seq) using CRISPRi to perturb the 1,032 prioritized CREs.
   • Readout: Targeted sequencing of the 4,928 candidate genes.
   • Scale: ~1.5 million cells.
   • Goal: Determine which CREs regulate which genes (CRE-to-gene).

2. Promoter Screen (Genome-wide Perturb-seq):

   • Technique: Genome-wide CRISPRi targeting all expressed coding genes (7,664 genes), including transcription factors and the target genes identified in the CRE screen.
   • Readout: Whole transcriptome sequencing.
   • Scale: ~2.6 million cells.
   • Goal: Map downstream regulatory networks (gene-to-network).

C. Data Analysis & Integration

• Validation: Used SCEPTRE for differential expression analysis to identify significant perturbations.
• Enhancer Identification: Filtered CRE-gene pairs to identify "enhancer-like" interactions by excluding proximal promoter effects and validating against chromatin accessibility (ATAC-seq), histone marks (H3K27ac, H3K4me1), and Hi-C contact frequencies.
• Network Propagation: Integrated CRE and Promoter screen data to trace "Variant $\to$ CRE $\to$ Gene $\to$ Network" cascades (Hop 1 and Hop 2 effects).
• Dimensionality Reduction: Used MOFA (Multi-Omics Factor Analysis) to decompose transcriptional responses into latent factors representing biological programs.

3. Key Contributions & Results

A. Mapping CRE-Gene Interactions

• Discovery: Identified 626 significant cis-regulatory interactions (CRE-gene pairs).
• Statistics: On average, each CRE regulates 1.6 genes (73% downregulated, 27% upregulated upon CRISPRi).
• Prediction Benchmarking:
  • In silico predictions were 13-fold more likely to be validated than random links.
  • eQTL-based links had the highest validation rate (35%), though few in number.
  • ENCODE-rE2G showed high generalizability (32% validation rate) despite being trained on K562 cells.
  • Simple distance-based approaches (100 kb window) yielded the highest absolute number of hits but low precision.

B. Defining Enhancer-Like Interactions

• The authors refined the dataset to 106 high-confidence enhancer-like interactions (84 activating, 22 repressive) involving 85 CREs and 89 genes.
• Characteristics: These interactions showed:
  • Distance-dependent decay in chromatin accessibility and H3K27ac signals.
  • Higher Hi-C contact frequencies with target promoters compared to non-significant controls.
  • Localization within the same Topologically Associating Domain (TAD).
  • Limited activity beyond 1 Mb distance.

C. Network Propagation (Variant-to-Network)

• Cascades: By integrating the screens, they traced disease-associated variants through CREs to target genes and then to downstream networks.
• Reach: The regulatory cascades connected to 56% (4,356/7,815) of all tested genes within two "hops" (Hop 1: direct target; Hop 2: downstream of target).
• Case Studies:
  • TYK2 Locus: A distal intronic CRE (associated with IBD/Crohn's) was shown to regulate TYK2 (not the host gene PDE4A). TYK2 knockdown triggered a Type I interferon program, linking noncoding variants to inflammatory pathways.
  • DEXI/CLEC16A Locus: Resolved ambiguity in Multiple Sclerosis risk loci. Three enhancer-like CREs converged on DEXI (not CLEC16A). Downstream effects implicated amino acid transport and metabolic programs.

D. Disease-Specific vs. Shared Programs

• Shared Programs: Most diseases converged on core immune programs (T cell activation, cytokine production, viral response).
• Disease Specificity: Distinct programs were identified for specific diseases:
  • Multiple Sclerosis: Enrichment for IL-8 production.
  • Celiac Disease: Convergence on tight junction regulation (genes like DLG1, ACTG1), linking dispersed noncoding variants to intestinal barrier dysfunction.

4. Significance

• Resource Creation: Provides the first genome-scale, cell-type-specific map of variant-to-CRE-to-gene-to-network relationships in primary human CD4+ T cells.
• Methodological Advancement: Demonstrates that TAP-seq offers superior sensitivity for detecting weak enhancer effects compared to standard Perturb-seq, and that integrating orthogonal prediction strategies improves CRE-gene linking.
• Mechanistic Insight: Moves beyond statistical association to provide causal mechanisms for immune diseases, resolving "ambiguous" GWAS loci (e.g., distinguishing DEXI from CLEC16A) and identifying specific therapeutic targets (e.g., TYK2 pathway).
• Framework for Future Research: Establishes a reproducible framework ("Variant-to-CRE-to-Gene-to-Network") that can be applied to other disease contexts and cell types to decode the non-coding genome.
• Clinical Relevance: The findings directly link noncoding variants to druggable pathways (e.g., JAK-STAT signaling, metabolic transport), offering new avenues for drug discovery in autoimmune diseases.