A Genome-First Approach to Characterize DICER1 Pathogenic Variant Prevalence, Penetrance, and Phenotype

Abstract

Importance: Genetic disorders are historically defined through phenotype-first approaches. However, risk estimates derived from phenotype-linked ascertainment may overestimate severity and penetrance. Pathogenic variants in DICER1 are associated with increased risks of rare and common neoplasms and thyroid disease in adults and children. This study explored how effectively a genome-first approach could characterize the clinical traits associated with germline DICER1 putative loss-of-function (pLOF) variants in an unselected clinical cohort.

Objective: To examine the prevalence, penetrance, and phenotypic characteristics of carriers of germline DICER1 pLOF variants via genome-first ascertainment.

Design, setting, and participants: This cohort study classifies DICER1 variants in germline exome sequence data from 92 296 participants of the Geisinger MyCode Community Health Initiative. Data for each MyCode participant were used from the start of the Geisinger electronic health record to February 1, 2018.

Main outcomes and measures: Prevalence of germline DICER1 variation; penetrance of malignant tumors and thyroid disease in carriers of germline DICER1 variation; structured, manual review of electronic health records; and DICER1 sequencing of available tumors from an associated cancer registry.

Results: A total of 92 296 adults (mean [SD] age, 59 [18] years; 98% white; 60% female) participated in the study. Germline DICER1 pLOF variants were observed in 1 in 3700 to 1 in 4600 participants, more than double the expected prevalence. Malignant tumors (primarily thyroid carcinoma) were observed in 4 of 25 participants (16%) with DICER1 pLOF variants, which is comparable (by 50 years of age) to the frequency of neoplasms in the largest registry- and clinic-based (phenotype-first) DICER1 studies published to date. DICER1 pLOF variants were significantly associated with risks of thyroidectomy (odds ratio [OR], 6.0; 95% CI, 2.2-16.3; P = .007) and thyroid cancer (OR, 9.2; 95% CI, 2.1-34.7; P = .02) compared with controls, but there was not a significant increase in the risk of goiter (OR, 1.8; 95% CI, 0.7-4.9). A female patient in her 80s who was a carrier of a germline DICER1 hotspot variant was apparently healthy on electronic health record review. The term DICER1 did not appear in any of the medical records of the 25 participants with a pLOF DICER1 variant, even in those affected with a known DICER1-associated tumor or thyroid phenotype.

Conclusions and relevance: This cohort study was able to ascertain individuals with germline DICER1 variants based on a genome-first approach rather than through a previously established DICER1-related phenotype. Use of the genome-first approach may complement more traditional approaches to syndrome delineation and may be an efficient approach for risk estimation.

Figure 1.. Association of Germline DICER1 Variants and Thyroid Phenotypes and Malignant Tumors Stratified by Pathogenicity

Odds ratios (95% CIs) for the prevalence of each phenotype for individuals from the putative loss-of-function (pLOF) and hotspot variant, predicted deleterious, variant of uncertain significance (VUS), and likely benign (LB) groups compared with noncarriers (DiscovEHR participants who were noncarriers of a DICER1 variant) were calculated using the Fisher exact test. Carriers and noncarriers were matched by sex, race, and smoking status. Numbers in parentheses after each phenotype denote counts in DICER1 carrier cases and noncarrier cases (eTable 5 in the Supplement). ^aP < .008 is considered significant with Bonferroni correction for multiple testing.

Figure 2.. Probability of Thyroid Phenotypes (Goiters, Thyrotoxicosis, Hypothyroidism, Thyroidectomy, and Thyroid Cancer) Over Time

Kaplan-Meier plots with left-truncation-bias correction in all individuals with bioinformatically predicted DICER1 missense variation using in silico prediction tool scores, including metaSVM, CADD, and REVEL, as well as variant of uncertain significance (VUS) vs controls. The metaSVM data use separate sets of controls designed for T (tolerated) and D (deleterious). NC indicates noncarrier of DICER1 variants.