Assessing the Clinical Relevance of BRCA1 RING Domain Variants of Uncertain Significance (VUS) with Comprehensive Computational and Functional Analyses
Cancer Genetics and Therapeutics
-
Primary Categories:
- Cancer
-
Secondary Categories:
- Cancer
Introduction:
BRCA1 serves multiple chromatin integrity functions critical to tumor suppression; up to 70% of women who inherent a pathogenic variant develop breast cancer by age 70. Accordingly, genetic screens that identify these heterozygotes enable patient stratification for prophylactic interventions, surveillance to detect cancer at an early and more-treatable stage, lifestyle and fertility counselling, and, if cancer develops, targeted therapy such as synthetic lethal PARP inhibition; benefits extend to relatives who become eligible for reflex testing. Conversely, heterozygotes found with a benign variant that does not increase cancer risk have anxieties alleviated. Unfortunately, VUS – variants for which associated cancer risk is not confidently understood – are uncovered in 10-20% of screens. This finding does not afford reflex testing for at-risk relatives, and often confuses patients who are ultimately left to make clinical care decisions based on a limited risk insight. Over 80% of BRCA1 VUS are missense, with their nuanced effects and low variant frequency generally limiting strong evidence approaches for classification outlined by The American College of Medical Genetics (ACMG). However, assays that measure variant effects on protein tumor suppressive function can also provide strong evidence for classification in a rapid, high-throughput manner. While the exact mechanism of BRCA1 tumor suppression remains unknown, an essential role of N-terminal RING is supported by significant clustering of pathogenic missense variants within the domain, and its activities in genome maintenance (e.g. homologous recombination, replication fork maintenance, transcription R-loop resolution). Current evidence supports that all RING functions depend on binding BARD1 and one of nine somewhat redundant ubiquitin-conjugating enzymes (E2); thus, an in vitro mammal assay quantifying both binding activities was hypothesized to produce strong evidence for classifying RING VUS that improves upon the physiological relevance of previous work.
Methods:
Six VUS were prioritized for functional analysis using a machine learning algorithm tailored to predict pathogenicity of BRCA1 RING variants: Molecular Feature Selection Tool evaluated the accuracy of 53 diverse existing algorithms on well-established variants within the domain, with the top 9 selected for model-building on MatLab Classification Learner. The linear support vector machine was selected as it provided a generalizable solution with perfect testing accuracy, and had a continuous output range suited for detailed prioritization. VUS with the highest and lowest model-predicted probabilities of pathogenicity were selected for functional analysis, backed with supporting evidence for classification. To assay binding, HA-tagged variant RING domain constructs were cloned into pTRE2 vectors, and transfected into HEK293T cells. Following 48h of growth, cells were harvested, and lysates subjected to immunoprecipitation by anti-HA magnetic beads to isolate variant RING and bound proteins. Binding of BARD1 and a model E2, UbcH5c, was then quantified with western blotting and fluorescent imaging.
Results:
Functional results generally agreed with computational predictions: 3/3 predicted pathogenic variants demonstrated a significant decrease in BARD1 binding activity compared to 2/3 benign controls; 2/3 predicted benign variants demonstrated similar or higher activity than wildtype across 3 biological replicates. E2 binding trends followed the same pattern, but failed to reach significance. Interestingly, one variant [NM_007294.4:c.271T>C(p.Cys91Arg)] was predicted benign computationally, but BARD1 and E2 binding activities were modestly reduced compared to 3/3 benign controls. Future projects to resolve this conflicting evidence are proposed: assays on PARP inhibitor sensitivity, and the presence of megabase-long scars distinctively arising from erroneous back-up repair in BRCA1-deficient cells.
Conclusion:
Corroborating strong functional and supporting computational evidence from this project enables reclassification of 2 VUS to likely benign; only one additional supporting (or greater) line of evidence is required for reclassification of 3 VUS to likely pathogenic according to ACMG guidelines. These reclassifications will immediately benefit heterozygotes with these variants by directing them to appropriate preventative care.
BRCA1 serves multiple chromatin integrity functions critical to tumor suppression; up to 70% of women who inherent a pathogenic variant develop breast cancer by age 70. Accordingly, genetic screens that identify these heterozygotes enable patient stratification for prophylactic interventions, surveillance to detect cancer at an early and more-treatable stage, lifestyle and fertility counselling, and, if cancer develops, targeted therapy such as synthetic lethal PARP inhibition; benefits extend to relatives who become eligible for reflex testing. Conversely, heterozygotes found with a benign variant that does not increase cancer risk have anxieties alleviated. Unfortunately, VUS – variants for which associated cancer risk is not confidently understood – are uncovered in 10-20% of screens. This finding does not afford reflex testing for at-risk relatives, and often confuses patients who are ultimately left to make clinical care decisions based on a limited risk insight. Over 80% of BRCA1 VUS are missense, with their nuanced effects and low variant frequency generally limiting strong evidence approaches for classification outlined by The American College of Medical Genetics (ACMG). However, assays that measure variant effects on protein tumor suppressive function can also provide strong evidence for classification in a rapid, high-throughput manner. While the exact mechanism of BRCA1 tumor suppression remains unknown, an essential role of N-terminal RING is supported by significant clustering of pathogenic missense variants within the domain, and its activities in genome maintenance (e.g. homologous recombination, replication fork maintenance, transcription R-loop resolution). Current evidence supports that all RING functions depend on binding BARD1 and one of nine somewhat redundant ubiquitin-conjugating enzymes (E2); thus, an in vitro mammal assay quantifying both binding activities was hypothesized to produce strong evidence for classifying RING VUS that improves upon the physiological relevance of previous work.
Methods:
Six VUS were prioritized for functional analysis using a machine learning algorithm tailored to predict pathogenicity of BRCA1 RING variants: Molecular Feature Selection Tool evaluated the accuracy of 53 diverse existing algorithms on well-established variants within the domain, with the top 9 selected for model-building on MatLab Classification Learner. The linear support vector machine was selected as it provided a generalizable solution with perfect testing accuracy, and had a continuous output range suited for detailed prioritization. VUS with the highest and lowest model-predicted probabilities of pathogenicity were selected for functional analysis, backed with supporting evidence for classification. To assay binding, HA-tagged variant RING domain constructs were cloned into pTRE2 vectors, and transfected into HEK293T cells. Following 48h of growth, cells were harvested, and lysates subjected to immunoprecipitation by anti-HA magnetic beads to isolate variant RING and bound proteins. Binding of BARD1 and a model E2, UbcH5c, was then quantified with western blotting and fluorescent imaging.
Results:
Functional results generally agreed with computational predictions: 3/3 predicted pathogenic variants demonstrated a significant decrease in BARD1 binding activity compared to 2/3 benign controls; 2/3 predicted benign variants demonstrated similar or higher activity than wildtype across 3 biological replicates. E2 binding trends followed the same pattern, but failed to reach significance. Interestingly, one variant [NM_007294.4:c.271T>C(p.Cys91Arg)] was predicted benign computationally, but BARD1 and E2 binding activities were modestly reduced compared to 3/3 benign controls. Future projects to resolve this conflicting evidence are proposed: assays on PARP inhibitor sensitivity, and the presence of megabase-long scars distinctively arising from erroneous back-up repair in BRCA1-deficient cells.
Conclusion:
Corroborating strong functional and supporting computational evidence from this project enables reclassification of 2 VUS to likely benign; only one additional supporting (or greater) line of evidence is required for reclassification of 3 VUS to likely pathogenic according to ACMG guidelines. These reclassifications will immediately benefit heterozygotes with these variants by directing them to appropriate preventative care.