Massively parallel functional profiling identifies CCDC88C as a risk gene for ER-positive breast cancer

By employing a lentivirus-based massively parallel reporter assay to screen over 5,000 credible causal variants, this study identified 709 functional variants across 140 risk regions and pinpointed rs7153397 as a key variant influencing CCDC88C expression, thereby establishing CCDC88C as a risk gene for ER-positive breast cancer.

The Gist

The Big Picture: Finding the "Typos" in the Recipe Book

Imagine the human genome (our DNA) as a massive, ancient recipe book for building and maintaining a human body. Sometimes, this book has tiny typos or spelling errors. Most of these errors don't matter, but some specific typos can change the recipe just enough to make a person more likely to get breast cancer.

Scientists have already found 196 different locations in this recipe book where these "cancer-risk typos" exist. However, there's a huge problem: these typos are often clustered together in a messy neighborhood. It's like finding a street with 50 houses, and you know one of them has a broken lock, but you don't know which one. Furthermore, most of these risky typos aren't in the "ingredients list" (the parts of DNA that code for proteins); they are in the "instructions" (the non-coding parts) that tell the cell how much of an ingredient to make.

The Goal: This study wanted to walk down that street of 50 houses, knock on every door, and figure out exactly which specific typo is actually breaking the lock and causing the problem.

---

The Method: The "Massive Parallel" Mailbox Test

To solve this, the researchers used a high-tech tool called lentiMPRA. Let's break down what that means using an analogy.

Imagine you have 5,000 different versions of a specific instruction manual page. Some pages have a "Red" word, and others have a "Blue" word in the same spot. You want to know: Does having the "Red" word make the machine work faster or slower than the "Blue" word?

1. The Library: The scientists took 5,116 of these risky DNA snippets. For each snippet, they made two copies: one with the "Red" version (Reference) and one with the "Blue" version (Alternative).
2. The Barcodes: They attached a unique "barcode" (like a tiny, invisible sticker) to every single copy. This is crucial. It's like giving every single letter in a massive mailroom a unique tracking number so you know exactly which letter produced which result.
3. The Factory: They put all these barcoded instructions into a delivery truck (a lentivirus) and sent them into a factory (breast cancer cells called T-47D).
4. The Count: Once inside the factory, the instructions started working. The scientists then counted how many "Red" instructions were working versus how many "Blue" instructions were working by scanning the barcodes.

If the "Red" version produced 80% more activity than the "Blue" version, they knew that specific typo was a functional driver of risk.

The Results: Finding the Culprit

After testing thousands of these "typos," they found 709 specific variants that actually changed how the cell behaved. Most of these had small effects, but they were significant.

However, the team didn't just stop at a list of numbers. They wanted to find the "smoking gun"—the one variant that explained a specific risk area on chromosome 14.

The Discovery: rs7153397 and CCDC88C
They zoomed in on a specific spot on chromosome 14. They found a typo called rs7153397.

• The Mechanism: This typo sits right in a "switch" that controls a gene called CCDC88C.
• The Effect: The "Red" version of this typo flips the switch to "High," causing the cell to produce too much of the CCDC88C protein.
• The Consequence: Too much CCDC88C seems to be linked to a higher risk of getting ER-positive breast cancer (a common type of breast cancer that grows in response to estrogen).

The Twist: Usually, when we think of "too much protein," we think of bad things. But here, the data showed something interesting: In patients who already had ER-positive cancer, having higher levels of CCDC88C was actually linked to better survival rates. It's like having a fire alarm that is too sensitive (causing the fire to start more easily), but once the fire is burning, that loud alarm helps the firefighters put it out faster.

The Verification: The "Dimmer Switch" Test

To prove they were right, the scientists didn't just guess; they tested it. They used a tool called CRISPRi (a molecular dimmer switch).

• They went into the cells and specifically turned down the activity of that specific typo (rs7153397).
• Result: When they dimmed the switch, the levels of the CCDC88C protein dropped.
• Conclusion: This confirmed that this specific typo is indeed the master switch controlling that gene.

Why This Matters

1. From "Where" to "How": Before this, we knew where the risk was (the neighborhood), but not how it worked. Now we know the specific mechanism: a typo that turns up a volume knob on a specific gene.
2. Better Tools: The study showed that looking at computer predictions (epigenetics) isn't enough. You have to physically test the DNA to see what it actually does.
3. Future Treatments: By understanding that CCDC88C is the target, scientists can now look for drugs that might lower its levels in high-risk people, or use its levels to predict how well a patient will do after a diagnosis.

Summary in One Sentence

This study used a massive, automated "mailroom" experiment to sift through thousands of genetic typos, identifying one specific switch (rs7153397) that controls a gene (CCDC88C), which acts as a double-edged sword: it increases the risk of developing ER-positive breast cancer but, if the cancer does develop, higher levels of this gene might actually help patients survive longer.

Technical Summary

1. Problem Statement

Genome-wide association studies (GWAS) have identified 196 independent genetic signals associated with breast cancer risk. However, deciphering the functional basis of these associations is challenging because:

• Linkage Disequilibrium (LD): Variants are often correlated, making it difficult to pinpoint the specific causal variant.
• Non-coding Nature: Most risk variants reside in non-coding regions, likely influencing disease risk through cis-regulatory mechanisms (modulating gene expression) rather than altering protein structure.
• Validation Gap: While computational fine-mapping (e.g., by the Breast Cancer Association Consortium, BCAC) has prioritized 7,394 "credible causal variants" (CCVs), experimental validation of their regulatory activity and target genes remains limited.

2. Methodology

The authors employed a high-throughput, experimental approach to functionally validate BCAC-prioritized CCVs.

• Screening Platform: A lentivirus-based Massively Parallel Reporter Assay (lentiMPRA) was used. This method integrates reporter constructs into the host genome, better recapitulating in vivo chromatin context compared to episomal assays (like STARR-seq).
• Library Design:
  • Target: 5,116 CCVs across 195 independent risk signals.
  • Constructs: 20,878 oligonucleotides (270 bp) were designed, centered on Reference (REF) and Alternative (ALT) alleles in both forward and reverse orientations.
  • Barcoding: Each oligonucleotide was linked to unique DNA barcodes (mean 175 barcodes per oligo) to quantify transcriptional activity via RNA sequencing.
  • Controls: Included 36 positive controls (near GATA3 and PGR enhancers) and 400 negative controls (scrambled sequences).
• Cell Model: T-47D breast cancer cells (Estrogen Receptor-positive, ER+) were chosen due to their clinical relevance and availability of epigenetic data (DNase-seq, H3K27ac, TF ChIP-seq).
• Workflow:
  1. Library generation and lentiviral packaging.
  2. Infection of T-47D cells (triplicate biological replicates).
  3. DNA and RNA extraction followed by high-throughput sequencing.
  4. Analysis: Used MPRAnalyze to quantify enhancer activity ($\alpha$) and differential allelic activity (log2 fold change, FC) between REF and ALT alleles.
• Follow-up Validation:
  • CRISPR Interference (CRISPRi): Used dCas9-KRAB to repress specific regions (rs7153397 and the CCDC88C promoter) in T-47D cells to confirm target gene regulation.
  • Clinical Correlation: Analyzed CCDC88C expression and survival data from The Cancer Genome Atlas (TCGA) and the Sweden Cancerome Analysis Network-Breast (SCAN-B) cohorts.

3. Key Contributions

• Systematic Functional Screening: This study represents the first comprehensive functional validation of BCAC-defined CCVs using lentiMPRA, moving beyond computational prediction to experimental evidence of regulatory activity.
• Prioritization Framework: Established a pipeline to filter thousands of variants down to a high-confidence set of functional drivers based on allelic differential activity and epigenetic enrichment.
• Novel Gene Discovery: Identified CCDC88C as the specific target gene for the 14q32.11 risk locus, linking a specific non-coding variant to a specific gene and clinical outcome.

4. Key Results

• Global Screening Results:
  • Out of 5,116 tested CCVs, 709 variants (14.98%) across 140 risk regions showed significant differential allelic activity (FDR-adjusted $P < 0.1$).
  • 6.47% of variants showed significant enhancer activity ($\alpha$), indicating they reside in regulatory sequences.
  • Functional variants were significantly enriched for DNase I hypersensitive sites (DHS) and EP300 binding, confirming their regulatory nature.
  • There was a weak correlation between computational fine-mapping scores (PAINTOR posterior probabilities) and experimental allelic activity, highlighting the necessity of experimental validation.
• Focus on rs7153397 (14q32.11):
  • Variant Selection: rs7153397 was prioritized due to a massive log2FC of 0.85 (80% increase in activity for the ALT allele; $P = 1.5 \times 10^{-56}$).
  • Regulatory Context: The variant maps to a region with H3K27ac marks and binding sites for FOXA1, ESR1, and EP300. It is in perfect LD with the index variant rs11341843.
  • Target Gene Identification: CRISPRi targeting of the rs7153397 region significantly reduced expression of CCDC88C (Coiled-Coil Domain Containing 88C) without affecting neighboring genes, confirming it as the causal target.
• Clinical Relevance of CCDC88C:
  • Expression: CCDC88C is significantly upregulated in both ER+ and ER- breast tumors compared to normal tissue. Crucially, expression is significantly higher in ER+ tumors compared to ER- tumors.
  • Prognosis: In ER+ patients, higher CCDC88C expression is associated with improved overall survival (Hazard Ratio = 0.53 in TCGA; HR = 0.63 in SCAN-B). This association was not observed in ER- disease.

5. Significance

• Mechanistic Insight: The study elucidates a specific mechanism for breast cancer risk at the 14q32.11 locus, where a non-coding variant alters the expression of CCDC88C, a regulator of non-canonical Wnt signaling via the PI3K–AKT pathway.
• Subtype Specificity: The findings highlight that genetic risk factors can have distinct effects depending on breast cancer subtype (ER+ vs. ER-), with CCDC88C specifically influencing prognosis in ER+ disease.
• Resource for Future Research: The study provides a prioritized list of 709 functional variants and a validated methodological framework (lentiMPRA + CRISPRi) for dissecting the "dark matter" of GWAS loci.
• Therapeutic/Prognostic Potential: Identifying CCDC88C as a driver of ER+ breast cancer with prognostic value suggests it could serve as a biomarker or a potential therapeutic target for this specific patient population.

In conclusion, this paper successfully bridges the gap between statistical genetic associations and biological function, demonstrating that high-throughput functional profiling is essential for identifying causal genes and understanding the etiology of complex diseases like breast cancer.