Resolution of the D4Z4 repeat responsible for facioscapulohumeral muscular dystrophy with HiFi sequencing

This paper introduces Kivvi, a computational tool leveraging PacBio HiFi long-read sequencing to comprehensively resolve the complex D4Z4 repeat region, enabling accurate genotyping of facioscapulohumeral muscular dystrophy (FSHD) alleles, detection of methylation patterns, and large-scale population profiling that consolidates multiple diagnostic assays into a single workflow.

The Gist

Imagine your genome is a massive, complex library. Most books in this library are easy to read, but there is one specific section, the D4Z4 repeat, that is notoriously difficult. It's like a shelf filled with thousands of nearly identical copies of the same book, stacked so tightly that it's impossible to tell where one copy ends and the next begins.

This specific section is linked to Facioscapulohumeral Muscular Dystrophy (FSHD), a condition that weakens muscles in the face, shoulders, and arms. To diagnose FSHD, doctors need to count exactly how many "book copies" (repeats) are on a specific shelf and check if the "security guard" (methylation) is doing its job to keep the books closed.

Here is the problem: The old tools used to count these books were like trying to count the pages of a book by looking at a blurry photo of the whole shelf. They were slow, labor-intensive, and often gave confusing answers because the books on Chromosome 4 (the dangerous shelf) look almost identical to the books on Chromosome 10 (the safe shelf).

The New Solution: "Kivvi" and the HiFi Camera

This paper introduces a new tool called Kivvi, powered by a special type of DNA sequencing called PacBio HiFi.

Think of the old sequencing methods as taking a photo of a crowd where everyone is wearing the same gray suit; you can't tell who is who. HiFi sequencing is like giving every person in the crowd a high-definition, 4K camera that captures their unique face and even the tiny details of their clothing.

Kivvi is the software that looks at these high-definition photos and does three amazing things:

1. It Counts the Books: It can accurately count how many repeat units are in a row, even if there are 100 of them.
2. It Identifies the Shelf: It can tell if the books are on the dangerous Chromosome 4 shelf or the safe Chromosome 10 shelf.
3. It Checks the Security Guard: It can see if the "security guard" (methylation) is working. In healthy people, the guard keeps the books locked tight. In FSHD patients, the guard is asleep, and the books (specifically a gene called DUX4) start opening up and causing damage.

How It Works (The Analogy)

Imagine the D4Z4 region as a long train made of identical train cars.

• The Problem: Sometimes the train is short (1–10 cars), which is dangerous. Sometimes it's long (100+ cars), which is safe. But the cars look so similar that old tools couldn't count them accurately. Also, some trains are on the "Bad Track" (Chromosome 4) and some on the "Good Track" (Chromosome 10).
• The Kivvi Solution: Kivvi takes a picture of the entire train in one go. It looks at the front of the train to see which track it's on. It looks at the back to see if the "safety signal" (polyA signal) is intact. Then, it counts every single car in between.

What They Found

The researchers tested Kivvi on hundreds of people, including patients with FSHD and healthy individuals.

1. It's a Detective: Kivvi successfully found the "short, dangerous trains" (contracted alleles) in 100% of the FSHD cases they checked. It also correctly identified the "long, safe trains" in most healthy people.
2. Two Types of Crime: They found two ways the "security guard" fails:
   • Type 1 (FSHD1): The train is too short (1–10 cars), so the guard can't lock it properly.
   • Type 2 (FSHD2): The train is long, but the guard is broken because of a different problem (a mutation in a gene called SMCHD1). Kivvi could spot this by seeing that all the trains on the shelf had their guards asleep, not just the short ones.
3. A World Tour: They looked at DNA from people of five different ancestral backgrounds (European, African, Asian, etc.). They discovered that the "trains" look slightly different depending on where your ancestors are from. They found "hybrid trains" that mix features from different tracks, which helps scientists understand how human evolution works.

Why This Matters

Before this, diagnosing FSHD was like solving a puzzle with missing pieces, requiring multiple different tests (gel electrophoresis, methylation tests, gene sequencing) that took weeks and cost a lot of money.

Kivvi changes the game. It allows doctors to run one single test that does everything: counts the repeats, identifies the chromosome, checks the safety signal, and measures the methylation. It's like replacing a dozen different tools with one Swiss Army knife that solves the whole mystery in a single workflow.

This means faster, cheaper, and more accurate diagnoses for patients, and it opens the door for scientists to discover new genetic clues that might help treat this difficult disease in the future.

Technical Summary

1. Problem Statement

Facioscapulohumeral muscular dystrophy (FSHD) is a prevalent inherited muscular dystrophy caused by the ectopic expression of the DUX4 gene in skeletal muscle. This expression is mediated by the contraction of the D4Z4 macrosatellite repeat on chromosome 4q35 (FSHD1, ~95% of cases) or mutations in chromatin modifiers like SMCHD1 (FSHD2, ~5% of cases).

The D4Z4 locus presents significant technical challenges for genomic analysis due to:

• Structural Complexity: It consists of a variable number tandem repeat (VNTR) of 3.3 kb units, with up to 100 copies per allele.
• Sequence Homology: High homology exists between the subtelomeric regions of chromosome 4 and chromosome 10, as well as between different haplotypes (4qA, 4qB, 10qA).
• Diagnostic Limitations: Current gold-standard diagnostics (Southern blotting, Optical Genome Mapping) are low-throughput, labor-intensive, and often ambiguous. They typically require separate assays to determine repeat size, chromosome origin, haplotype type (permissive vs. non-permissive), and methylation status.
• Assembly Errors: Standard de novo assembly of long-read data often fails in this region due to the repetitive nature of the locus.

2. Methodology: The Kivvi Tool

The authors developed Kivvi, a computational tool designed to genotype D4Z4 using PacBio HiFi (High-Fidelity) whole-genome sequencing (WGS) data. The workflow involves:

• Read Extraction & Realignment: Kivvi extracts reads overlapping D4Z4 on chromosomes 4 and 10 and realigns them to a single 3.3 kb reference unit plus a 905 bp distal region containing the polyadenylation signal (PAS).
• Repeat Unit Identification: It identifies unique repeat unit variants (22 distinct types in the example sample) based on specific nucleotide changes.
• De Bruijn Graph Assembly: A graph is constructed where nodes are unique repeat units and edges represent links found within HiFi reads (typically spanning 3–5 units). This allows for the reconstruction of full alleles.
• Haplotype Phasing (Chromosome Assignment): To distinguish between chromosome 4 and 10 alleles, Kivvi utilizes Paraphase to phase a 42 kb homologous region upstream of D4Z4. Chromosome 10 haplotypes are identified by soft-clipping at the end of the homology region, resulting in shorter upstream sequences compared to chromosome 4.
• Allele Classification:
  • Distal Type: Determined by the sequence downstream of the repeat array.
    • qAIntactPolyA: Intact PAS (Permissive, potentially pathogenic).
    • qADisruptedPolyA: Disrupted PAS (Non-permissive).
    • qB: Different distal sequence lacking the third exon/PAS.
  • Size: Reported as an exact count for fully assembled alleles or a lower bound for partially assembled ones.
• Methylation Profiling: Kivvi leverages the native 5-methylcytosine (5mC) detection capability of HiFi reads to calculate methylation levels at 101 CpG sites within the repeat units.

3. Key Contributions

• Unified Workflow: Kivvi consolidates multiple diagnostic assays (sizing, chromosome assignment, haplotype typing, and methylation) into a single WGS-based workflow.
• Comprehensive Genotyping: It is the first tool to report the full set of D4Z4 information (size, chromosome, distal haplotype, and methylation) simultaneously for all alleles in a sample.
• Population-Scale Characterization: The tool was applied to 601 individuals across five ancestral populations, revealing extensive genetic diversity, including hybrid repeat units and translocation events previously difficult to characterize.

4. Results

Validation Against Orthogonal Assays

Kivvi was validated against PFGE-Southern blot (27 control samples) and Optical Genome Mapping (10 FSHD1 patients):

• Pathogenic Allele Detection: Successfully detected 100% (11/11) of contracted (1–10 repeat units) qAIntactPolyA alleles, which are the primary cause of FSHD1.
• Non-Contracted Alleles: Correctly classified 90% of non-contracted qAIntactPolyA alleles and 100% of qB alleles in high-quality datasets.
• Accuracy: For fully assembled alleles, size calls were consistent with truth data (allowing a difference of ≤3 units).
• Limitations: Performance decreased for long (>10 RU) alleles in lower-coverage datasets, often resulting in "inconclusive" lower bounds rather than exact sizes. However, these lower bounds were sufficient to rule out pathogenic sizes in many cases.

Methylation and Disease Differentiation

• FSHD1 vs. FSHD2: Kivvi successfully distinguished between the two forms of FSHD based on methylation patterns:
  • FSHD1: Shows hypomethylation only in the contracted qAIntactPolyA allele, while other alleles remain highly methylated.
  • FSHD2 (SMCHD1 mutations): Shows global hypomethylation across all D4Z4 alleles, regardless of size or type.
• Case Discovery: Identified a potential FSHD1+2 case (Sample GeneDx001) carrying a contracted 9RU qAIntactPolyA allele and a pathogenic SMCHD1 variant, exhibiting global hypomethylation.

Population Genetics and Diversity

• Allele Frequencies: In the general population, qAIntactPolyA alleles are rare on chromosome 10 (<1%) but common on chromosome 4 (~40%).
• Hybrid Units and Alleles: PCA analysis of repeat unit sequences revealed a "hybrid zone" between qAIntactPolyA-like and qADisruptedPolyA-like units.
  • Hybrid Alleles: Identified alleles with mixed unit types (e.g., transitioning from qB-like to qADisruptedPolyA-like), suggesting translocation events.
  • Ancestral Differences: African populations showed a higher prevalence of hybrid units and hybrid alleles (47.5% of qAIntactPolyA alleles had hybrid units) compared to other populations, reflecting greater genetic diversity.

5. Significance

• Clinical Impact: Kivvi offers a scalable, high-throughput alternative to Southern blotting, enabling the consolidation of FSHD diagnostics into a single NGS workflow. This is crucial for identifying complex cases (e.g., FSHD1+2) and non-penetrant carriers.
• Biological Insight: The study provides the first comprehensive view of D4Z4 sequence variation across diverse populations, uncovering hybrid repeat structures and translocation events that were previously obscured by technical limitations.
• Future Directions: The authors propose extending Kivvi to other complex VNTRs (e.g., LPA KIV-2 repeats) and improving accuracy for mosaic alleles and in-cis duplications, paving the way for a unified approach to resolving difficult genomic regions.