Low-level mosaic variants causing the pancreatic disease congenital hyperinsulinism can be detected from blood DNA

Technical Summary: Low-level mosaic variants causing congenital hyperinsulinism detected from blood DNA

Problem Statement
A significant proportion of individuals with well-defined monogenic disorders, such as congenital hyperinsulinism (CHI) and neonatal diabetes (NDM), remain without a genetic diagnosis. While next-generation sequencing (NGS) has accelerated discovery, up to 50% of cases with monogenic phenotypes are genetically unresolved. A specific challenge in these organ-specific conditions is the detection of low-level mosaic pathogenic variants (typically defined as Variant Allele Fraction [VAF] <10%). These variants arise post-zygotically and are present in only a subset of cells. In conditions like CHI, where the affected tissue (pancreas) is inaccessible for routine testing, reliance on peripheral blood leukocyte DNA makes detection difficult. Standard genomic pipelines often miss these variants due to background noise, PCR artifacts, and the limitations of standard germline variant callers. Furthermore, distinguishing true low-level mosaicism from technical artifacts or sample contamination is a critical barrier to accurate diagnosis.

Methodology
The study systematically investigated low-level mosaic variants in a large cohort of genetically unsolved patients: 1,252 individuals with CHI and 312 with NDM.

• Targeted NGS (tNGS) Screening: The authors re-analyzed existing tNGS data from peripheral blood DNA. They utilized GATK3 Mutect2, a tool designed for somatic/low-level variant detection, to screen for variants with VAF <8% in dominant CHI genes (ABCC8, GCK, GLUD1, HK1) and dominant NDM genes (ABCC8, KCNJ11, INS). Standard germline callers (GATK HaplotypeCaller) were shown to miss variants below this 8% threshold.
• Filtering and Controls: Variants were filtered against a curated list of known pathogenic/likely pathogenic variants. To rule out coincidental findings, the authors screened a control cohort of 522 genetically solved CHI cases and the NDM cohort. Sample contamination was assessed using VerifyBamID, with a threshold set at 2%.
• Orthogonal Validation: Candidate variants identified by tNGS were validated using TaqMan-based droplet digital PCR (ddPCR). This method partitions reactions into ~20,000 droplets to provide highly accurate quantification of VAFs. Fresh aliquots of leukocyte DNA were used, and in one case, buccal cell DNA was also analyzed.
• Software Comparison: The study compared the performance of GATK3 Mutect2 against the updated GATK4 Mutect2 pipeline to assess sensitivity and positive predictive value.
• Clinical Correlation: Clinical data, including age at diagnosis and hyperammonaemia status, were compared between individuals with confirmed mosaic variants and those with heterozygous dominant variants.

Key Results

• Detection Rate: The tNGS screening identified 40 candidate low-level mosaic variants in 39 individuals with CHI (median VAF 1.4%). No candidate variants were found in the NDM cohort.
• Validation Success: Of the 35 variants subjected to ddPCR validation, 26 (74%) were confirmed as true positives. The confirmed variants had a median VAF of 3.6% (range 1.1–7.8%).
• False Positives: 9 of the 35 tested variants were false positives. These predominantly had VAFs <1% and included variants found in samples with detected contamination. Notably, no ddPCR-confirmed variants had tNGS VAFs below 1%.
• Tissue Distribution: In one individual, the mosaic HK1 variant was detected in both leukocyte and buccal cell DNA, supporting the presence of mosaicism across tissues.
• Software Performance: GATK4 Mutect2 demonstrated a higher positive predictive value (90% vs. 74% for GATK3) by generating fewer false positives, but it missed 31% of the true positive variants detected by GATK3.
• Clinical Phenotype: Individuals with confirmed mosaic GLUD1 variants showed a significantly lower prevalence of hyperammonaemia (54%) compared to those with heterozygous GLUD1 variants (92%). There was a trend toward later age at diagnosis for mosaic ABCC8, GCK, and HK1 variants compared to heterozygous cases, though this did not reach statistical significance.
• Enrichment: Low-level mosaic variants were significantly enriched in the genetically unsolved CHI cohort compared to solved CHI controls and NDM patients (p <0.0005).

Significance and Claims
The paper claims that disease-causing low-level mosaic variants in dominant CHI genes can be reliably detected from routine blood-derived DNA, provided that orthogonal validation (such as ddPCR) is employed to distinguish true variants from technical artifacts.

• Diagnostic Yield: This approach identifies a "missed" cause of disease in approximately 2% of genetically unsolved CHI cases, increasing diagnostic yield without the need for invasive pancreatic tissue sampling.
• Mechanistic Insight: The findings suggest that disorders driven by pathological overproduction or overactivation (like CHI) are more sensitive to the phenotypic effects of low-level mosaicism than loss-of-function disorders (like NDM), where a larger proportion of beta-cells must be impaired to manifest disease.
• Clinical Implications: Identifying these variants allows for accurate genetic counseling regarding recurrence risk and personalized treatment. The authors caution that relaxing screening criteria to include novel missense variants without validation could lead to incorrect diagnoses and inappropriate clinical management.
• Future Directions: The study highlights the challenge of balancing sensitivity and specificity when transitioning to whole-exome or whole-genome sequencing, which typically have lower read depths. It suggests that while ddPCR is the current gold standard for validation, scalable alternatives like single-molecule molecular inversion probes may be necessary for broader implementation.