Specifications of the ACMG/AMP Variant Classification Guidelines for Germline <i>DICER1</i> Variant Curation

Hatton, Jessica N.; Frone, Megan N.; Cox, Hannah C.; Crowley, Stephanie B.; Hiraki, Susan; Yokoyama, Noriko N.; Abul-Husn, Noura S.; Amatruda, James F.; Anderson, Michael J.; Bofill-De Ros, Xavier; Carr, Ann G.; Chao, Elizabeth C.; Chen, Kenneth S.; Gu, Shuo; Higgs, Cecilia; Machado, Jerry; Ritter, Deborah; Schultz, Kris Ann P.; Soper, Emily R.; Wu, Mona K.; Mester, Jessica L.; Kim, Jung; Foulkes, William D.; Witkowski, Leora; Stewart, Douglas R.

doi:https://doi.org/10.1155/2023/9537832

Human Mutation

On this page

Abstract Introduction Methods Results Discussion Conclusions Data Availability Disclosure Conflicts of Interest Authors’ Contributions Acknowledgments Supplementary Materials References Copyright Related Articles

Research Article | Open Access

Volume 2023 | Article ID 9537832 | https://doi.org/10.1155/2023/9537832

Specifications of the ACMG/AMP Variant Classification Guidelines for Germline DICER1 Variant Curation

Jessica N. Hatton,¹Megan N. Frone,¹Hannah C. Cox,²Stephanie B. Crowley,³Susan Hiraki,⁴Noriko N. Yokoyama,⁵Noura S. Abul-Husn,⁶James F. Amatruda,⁷Michael J. Anderson,³Xavier Bofill-De Ros,⁸Ann G. Carr,⁹and Elizabeth C. Chao^5,10 et al.

Academic Editor: Jian-Min Chen

Received09 Oct 2022

Revised02 Feb 2023

Accepted08 Mar 2023

Published29 Mar 2023

Abstract

Germline pathogenic variants in DICER1 predispose individuals to develop a variety of benign and malignant tumors. Accurate variant curation and classification are essential for reliable diagnosis of DICER1-related tumor predisposition and the identification of individuals who may benefit from surveillance. Since 2015, most labs have followed the American College of Medical Genetics and Genomics and the Association for Molecular Pathology (ACMG/AMP) sequence variant classification guidelines for DICER1 germline variant curation. However, these general guidelines lack gene-specific nuances and leave room for subjectivity. Consequently, a group of DICER1 experts joined ClinGen to form the DICER1 and miRNA-Processing Genes Variant Curation Expert Panel (VCEP) to create DICER1-specific ACMG/AMP guidelines for germline variant curation. The VCEP followed the FDA-approved ClinGen protocol for adapting and piloting these guidelines. A diverse set of 40 DICER1 variants were selected for piloting, including 14 known pathogenic/likely pathogenic (P/LP) variants, 12 known benign/likely benign (B/LB) variants, and 14 variants classified as variants of uncertain significance (VUS) or with conflicting interpretations in ClinVar. Clinically meaningful classifications (i.e., P, LP, LB, or B) were achieved for 82.5% (33/40) of the pilot variants, with 100% concordance among the known P/LP and known B/LB variants. Half of the VUS or conflicting variants were resolved with four variants classified as LB and three as LP. These results demonstrate that the DICER1-specific guidelines for germline variant curation effectively classify known pathogenic and benign variants while reducing the frequency of uncertain classifications. Individuals and labs curating DICER1 variants should consider adopting this classification framework to encourage consistency and improve objectivity.

1. Introduction

The DICER1 gene (NM_177438.3) is located on chromosome 14q32.13 and contains 27 exons encoding 1,922 amino acids. Germline pathogenic variation in DICER1 is associated with an increased risk for the development of tumors in childhood and adulthood (OMIM # 601200) [1–3]. The DICER1 protein is an endoribonuclease that converts a hairpin-shaped miRNA precursor (pre-miRNA) to a mature miRNA duplex by removing the terminal loop. The RNase IIIa and RNase IIIb domains of DICER1 form two catalytic cores [4], cleaving at the 3 and 5 sides of the terminal loop, respectively, which are required to generate miRNAs derived from the 3p-arm (3p miRNAs) and 5p-arm (5p miRNAs) of the pre-miRNA accordingly.

DICER1-related tumor predisposition was first described in families with pleuropulmonary blastoma, a rare pediatric lung tumor [3]. The phenotypic spectrum has since expanded to include a wide range of benign and malignant neoplasms in both children and adults such as Sertoli-Leydig cell tumors, cervical and ovarian embryonal rhabdomyosarcoma, Wilms tumor, nasal chondromesenchymal hamartoma, pituitary blastoma, pineoblastoma, thyroid lesions, and other rare sarcomas [1, 5]. Surveillance recommendations aimed at early tumor detection exist for those with DICER1-related tumor predisposition due to germline variants in DICER1 [6–8].

Germline variant classification relies on the weighing of many pieces of evidence, such as functional data, population frequency, clinical phenotype, and family segregation data. In 2015, the American College of Medical Genetics and Genomics (ACMG) and the Association for Molecular Pathology (AMP) issued a joint publication of standards and guidelines for the classification of germline sequence variants [9] as a starting point to standardize variant classification procedures. The Clinical Genome Resource (ClinGen) [10], a National Institutes of Health- (NIH-) funded resource aimed at further refining and centralizing gene and variant curation processes, has since created a number of variant curation expert panels (VCEPs) [11] that follow the Food and Drug Administration- (FDA-) recognized guidance for Public Human Genetic Variant Databases and the ClinGen Expert Panel process to tailor and pilot gene-specific modifications of the ACMG/AMP variant curation guidelines [12–15].

The ClinGen DICER1 and miRNA-Processing Gene VCEP, hereafter referred to as the DICER1 VCEP, were formed with the goal of developing such tailored germline sequence variant curation guidelines for DICER1 and eventually other miRNA-processing genes associated with inherited syndromes (https://clinicalgenome.org/affiliation/50050/). Here, we describe the process of our VCEP formation, the evidence code specification for DICER1, and pilot curation.

2. Methods

In 2019, a variety of DICER1 experts from across North America convened virtually to form the DICER1 VCEP (https://clinicalgenome.org/affiliation/50050/), following the ClinGen VCEP protocol (https://clinicalgenome.org/site/assets/files/3635/variant_curation_expert_panel_vcep_protocol_version_9-2_3.pdf). Membership included clinicians, basic scientists, laboratory geneticists, and variant scientists. Initially, 22 group members were divided into four subgroups (phenotype, penetrance, computational, and functional) to critically assess and modify a subset of the ACMG/AMP variant curation evidence codes for DICER1-specific germline variant curation. A preliminary set of specifications was defined in November 2020 using MANE transcripts NM_177438.2 and MONDO:0017288.

The specifications were piloted on 40 DICER1 variants with submissions in ClinVar. These included 14 known pathogenic or likely pathogenic (P/LP) variants, 12 known benign or likely benign (B/LB) variants, and 14 variants with conflicting interpretations or classified as variants of uncertain significance (VUS). Classifications reflect ClinVar submissions as of November 2020 except for two of the P/LP variants which were updated more recently due to a known incongruence between one laboratory’s ClinVar submissions (VUS) and internal classifications (LP) at the time of the data pull. Pilot variants were intentionally selected such that missense, nonsense, frameshift, synonymous, and intronic variants were represented, giving the opportunity to test the performance of as many evidence codes as possible. Each variant was double-curated by two of the six biocurators to ensure evidence codes were being interpreted and applied uniformly. All biocurators had prior variant curation experience through other ClinGen VCEPs and/or employment at a commercial genetic testing laboratory offering clinical genetic testing for the DICER1 gene. In addition to published cases, relevant internal case-level data stripped of personally identifiable information was obtained by VCEP members working at testing laboratories, clinics, and the pleuropulmonary blastoma/DICER1 Registry (http://www.ppbregistry.org, NCT03382158) using an organized spreadsheet guide. Variants were curated within the ClinGen Variant Curation Interface [16]. Final classifications were determined according to the original evidence code combinations [9] plus a handful of predetermined combinations supported by a Bayesian framework [17]. In cases of conflicting benign and pathogenic evidence codes, a Bayesian point system was employed to reach a final classification [18]. Evidence codes were further adapted as appropriate during the pilot, and the final specifications were approved by the ClinGen Sequence Variant Interpretation (SVI) Committee in May 2022.

Our ACMG/AMP specifications will be updated periodically; to find the most current information, please visit https://clinicalgenome.org/affiliation/50050/ or https://cspec.genome.network/cspec/ui/svi/doc/GN024.

3. Results

3.1. DICER1-Specific Variant Curation Criteria

The DICER1 VCEP specifications to the ACMG/AMP variant curation criteria are summarized in Table 1. Eight evidence codes (PM3, PM6, PP2, PP5, BP1, BP3, BP5, and BP6) were excluded due to redundancy, irrelevance with respect to DICER1, or published ClinGen guidance [19]. The remaining 20 criteria were kept with clarifications and/or gene-specific modifications to strength or scope.

3.2. Population Data (BA1, BS1, and PM2)

3.2.1. BA1 and BS1

BA1 is standalone, and BS1 is strong evidence for benign variation based on the frequency of a variant in the general population. To determine frequency cutoffs, the VCEP first calculated a realistic maximum allele frequency for a pathogenic DICER1 variant using the Whiffin-Ware equation: [20]. Disease prevalence was set to 1 in 10,600 people (1 in 21,200 alleles) based on estimates from population databases [21]. The maximum allelic contribution was set to 0.07 based on the proportion of the most common P/LP DICER1 variant from Invitae internal data. Disease penetrance was set to 0.1 (i.e., 10%) based on the lower end of published penetrance estimates for individuals aged 50-60 years [22]. The resulting frequency, 0.00003, was conservatively increased by one order of magnitude for a BS1 cutoff of 0.0003 and another order of magnitude for a BA1 cutoff of 0.003. The VCEP chose to use noncancer gnomAD subpopulations to minimize the inclusion of cases. Generally, the most recent version of gnomAD with a noncancer subpopulation should be used. However, earlier versions should be considered relevant (e.g., superior sample size). Per published guidance, continental subpopulations must have greater than 2,000 alleles tested and a minimum of five alleles present [23].

3.2.2. PM2

The PM2 criterion is intended to provide evidence of pathogenicity for variants that are absent from population databases or present only at low levels. The VCEP identified 19 P/LP or putative loss of function DICER1 variants in noncancer gnomAD at low frequencies and expects that more will inevitably be present as databases grow. For this reason, the VCEP chose to establish a PM2 cutoff rather than to require absence. Based on the data from those 19 variants, the VCEP elected to apply PM2 for variants with a frequency less than 0.000005 across noncancer gnomAD with no more than one allele in any subpopulation and at least 20x coverage for that region of the gene in gnomAD. Such conditions would allow PM2 application for 15 of the 19 variants described previously. Per ClinGen SVI recommendations, PM2 should only be applied at a supporting level (https://www.clinicalgenome.org/site/assets/files/5182/pm2_-_svi_recommendation_-_approved_sept2020.pdf).

3.3. Computational and Predictive Data (PVS1, PS1, PM1, PM4, PM5, PP3, BP4, and BP7)

3.3.1. PVS1

PVS1 provides very strong evidence of pathogenicity for null variants in a gene where the loss of function is a known mechanism of disease. This code is particularly relevant to DICER1, as most germline causative alleles are loss of function [1, 24]. The VCEP adopted previously published recommendations for the PVS1 application [25] but provided DICER1-specific details to simplify the application such as the nonsense-mediated decay cutoff and which exons, if skipped, would result in in-frame deletions. Notably, the VCEP deviated from the typical recommendation by precluding PVS1 application for start codon variants, as the p.Met1 site is not highly conserved in DICER1, and there are three possible in-frame alternate methionine residues at p.Met11, p.Met17, and p.Met24. Furthermore, internal lab data showed that, in multiple individuals, p.Met1 variants are not associated with any DICER1 phenotype. A DICER1-specific PVS1 flowchart is provided in Supplementary Figure 1.

3.3.2. PP3 and BP4

PP3 and BP4 support level evidence codes based on computational predictors. The VCEP assessed the performance of several computational predictors, including metaSVM, CADD, BayesDel, and REVEL, on 15 known P/LP and 27 known B/LB DICER1 missense variants. The best separation was attained using REVEL, a computational meta-predictor whose score reflects 13 individual computational tools [26]. Attempts were made to trichotomize REVEL score cutoffs for PP3 and BP4 in a Bayesian fashion by calculating the odds of pathogenicity for a variant above or below a chosen threshold based on the test set of variants. Such a calculation could also be used to modify the strength of the evidence code if it could be shown, for example, that variants above a particular threshold had moderate or strong odds of pathogenicity [17]. Because few confidently curated missense variants in DICER1 currently exist, the VCEP was unable to establish cutoffs through a Bayesian approach [27] and instead selected ≥0.75 and <0.50 as the PP3 and BP4 cutoffs, respectively, based on the general REVEL use guidelines [26] and a good visual separation of 15 pathogenic and 27 benign variants. PP3 and BP4 may also be applied to splicing and noncoding variants based on the concordance of two splice predictors, MaxEntScan and SpliceAI. Until sufficient data are available to determine gene-specific splice predictor thresholds, standard MaxEntScan and SpliceAI thresholds should be used. PP3 should not be used in combination with PVS1.

3.3.3. BP7

BP7 is intended for silent variants not predicted to impact splicing. BP4 must be applied as a prerequisite for BP7 consideration. For variants meeting BP4, any silent or intronic variant at +7 to -21 positions automatically qualifies for BP7. Noncoding variants outside the +7 to -21 intronic positions may have BP7 applied if the variant is the reference nucleotide in one or more primates and/or four or more mammalian species, indicating a lack of conservation of the nucleotide.

3.3.4. PM1

Variation in critical gene regions or hotspot codons is considered moderate evidence of pathogenicity under PM1. DICER1 has seven recognized hotspot codons: p.Ser1344, p.Glu1705, p.Asp1709, p.Asp1713, p.Gly1809, p.Asp1810, and p.Glu1813 [1, 24, 28]. Variation in these codons impairs the activity of the DICER1 RNase IIIb domain while leaving the IIIa cleavage domain intact. While variants in these hotspot codons are more commonly somatic in origin, they have been observed in a mosaic state and thus are still relevant for germline curation considerations [24, 29]. The VCEP decided it was appropriate to apply PM1 at a supporting level for missense variants affecting other residues within the RNase IIIb domain (p.Y1682–p.S1846).

3.3.5. PM4

Similarly, the VCEP decided that in-frame protein length changes, considered moderate evidence of pathogenicity under PM4, were more likely to be pathogenic if located in the RNase IIIb domain [30, 31]. For this reason, PM4 can be applied at full moderate strength to in-frame indels within the RNase IIIb domain (p.Y1682–p.S1846) and at a supporting level to in-frame indels outside that domain. PM4 should not be applied to indels in repeat regions of DICER1 (p.D606-p.D609; p.E1418-p.E1420; p.E1422-p.E1425).

3.3.6. PS1 and PM5

The PS1 and PM5 codes are intended for missense variants observed at an amino acid residue where the same (PS1) or a different (PM5) predicted amino acid change has been established as pathogenic. For both codes, the VCEP specified that the other variant must have reached a pathogenic classification (likely pathogenic does not suffice) by the DICER1 VCEP and that splice effects should be ruled out by RNA data or concordance with MaxEntScan and SpliceAI. For PM5, the missense variant under investigation should have an equal or worse (i.e., higher) Grantham score than the other pathogenic variant [32]. The VCEP further expanded the scope of PS1 by allowing it to apply to noncanonical intronic nucleotide substitutions where a pathogenic splice site variant has been observed before if MaxEntScan and SpliceAI both predict an equal or greater splice impact for the variant under investigation. Because PS1, PM5, and PM1 are similar evidence types, they should not be applied together. The strongest evidence code should be used for variants meeting two or more of these codes. PM1 at supporting strength may be combined with PS1 or PM5.

3.4. Functional Data (PS3 and BS3)

3.4.1. PS3 and BS3

In vivo and in vitro functional studies provide another critical piece of evidence for variant curation under PS3 and BS3. The VCEP identified various types of functional evidence applicable to DICER1 that can be applied at different strength levels. To apply PS3 at full strength, a patient-derived RNA assay must demonstrate an out-of-frame splicing impact or an in-frame splicing impact removing more than 10% (193 residues) of the protein or disrupting the RNase IIIb domain. If a variant also has PVS1_Strong applied, PS3 should be dropped to moderate application. PS3 can also be applied at a moderate level if RNA data demonstrates an in-frame splicing impact removing less than 10% of the protein and not affecting the RNase IIIb domain. Similarly, a patient-derived RNA assay demonstrating no splicing impact qualifies for BS3, though this should be observed in more than one patient to minimize the possibility of dropout. Another functional assay of utility for DICER1 variant classification is an in vitro cleavage assay which assesses the ability of a DICER1 protein to generate 3p and 5p miRNAs [33]. Evidence of impaired or retained DICER1 cleavage function through such an assay may be used to apply PS3 or BS3, respectively, at a supporting level, provided that appropriate positive and negative controls were used. A higher strength level is not appropriate at this time as these assays are low-throughput and dependent on operator experience. PS3 cannot be applied at any strength if PVS1 is applied at full strength.

3.5. Clinical Data

3.5.1. Phenotype (PS4 and PP4)

(1) PS4. The VCEP critically evaluated known DICER1-associated phenotypes; the specificity of these phenotypes for an underlying pathogenic germline DICER1 variant was also considered. PS4 was initially intended to be an evidence code for variant-level case-control studies, with a reduced-strength option for rare variants observed in multiple affected patients but lacking statistically significant case-control studies [9]. The code has since evolved into a sophisticated proband-counting code with variable strength applications where affected, unrelated probands are allotted 0, 0.5, or 1 point each based on the specificity of their phenotypes, and the point total determines the PS4 strength application [12, 14]. The VCEP kept this framework in mind when considering the DICER1 phenotypic spectrum.

A high-specificity phenotype deserving a full proband point should reflect a greater than 80% likelihood of an underlying pathogenic germline variant in the gene of interest; a moderate-specificity phenotype deserving a half proband point should reflect a 60-80% likelihood of an underlying causative germline variant [14]. Of the nearly 30 DICER1-associated phenotypes gathered from the literature [1, 5, 34] and panel members, few had published data on the frequency of underlying germline DICER1 variants in unselected patient cohorts. Studies of pleuropulmonary blastoma [24] and pituitary blastoma [29] suggest greater than 80% specificity for an underlying pathogenic germline DICER1 variant, while cystic nephroma [35] and Sertoli-Leydig cell tumors and gynandroblastoma [36] appear to fall in the 60-80% range. More recently, studies of primary intracranial sarcomas [37, 38] and multinodular goiter in young adults [39] suggest less than 60% specificity for germline DICER1 variants.

Given the lack of large, unselected studies of these neoplasms, the VCEP elected to independently survey six clinical experts from the VCEP to categorize the phenotypes as high-specificity (much more likely than not to have a germline P/LP DICER1 variant), moderate-specificity (more likely than not to have a germline P/LP DICER1 variant), and low-specificity (less likely to have a germline P/LP DICER1 variant). A consensus was reached if 5 or more of the experts agreed on the categorization. VCEP members discussed cases of disagreement and conservatively downgraded specificity. Certain phenotypes were considered so nonspecific (e.g., adult multinodular goiter, macrocephaly) that they were not deemed fit to qualify even for low specificity. The final agreed-upon designations are summarized in Table 2.

Using Table 3 as a guide, unrelated probands may be granted a full point on the basis of a high-specificity phenotype, two moderate-specificity phenotypes, a moderate-plus a low-specificity phenotype, or a moderate-specificity phenotype plus a family history of a high- or moderate-specificity phenotype in a first- or second-degree relative. If the last combination is used and that family also contributes to PP1 meiosis counting, only a half point should be counted to avoid double-counting segregation. A proband with only one moderate-specificity phenotype should be given a half point. Anything less specific is not granted any points. Points summed across unrelated probands indicate the strength application of PS4: supporting (1 to <2 points), moderate (2 to <4 points), or strong (≥4 points). PS4 should not be applied when a variant also has population data meeting BA1 or BS1 since a common variant may be present in a proband by chance. Additionally, PS4 should not be applied to a proband with another germline variant that could have reasonably contributed to the observed phenotype or whose tumor sequencing suggests sporadic tumorigenesis.

(2) PP4. Considering PS4 proband counting, many VCEPs have discarded PP4, a code focused on patient phenotype and family history, as redundant. However, it has been recognized that PP4 may be utilized as a tumor phenotype code when appropriate [40]. With few exceptions, both benign and malignant DICER1-driven neoplasms follow a distinct modified two-hit hypothesis: one loss of function variant plus one variant selectively impairing the RNase IIIb domain function [2, 24, 41, 42]. In DICER1-related tumor predisposition, the germline variant is typically the loss of function, and the somatic second hit generally occurs in one of a handful of hotspot codons. Because this pattern is a hallmark of DICER1-driven neoplasms, the VCEP determined that evidence from somatic tumor sequencing of any DICER1-associated neoplasm, regardless of specificity, should lead to PP4 application if three conditions are met. First, the variant under investigation should not itself be in a DICER1 hotspot codon. Second, in addition to the retention of the germline variant in the tumor, somatic sequencing should reveal a previously reported somatic second hit [1, 43, 44] as summarized in Supplementary Table 1. Finally, no additional non-hotspot DICER1 variants or loss of heterozygosity should be revealed, as such a finding could reflect sporadic tumorigenesis. A flowchart simplifying the PP4 application is shown in Figure 1. A single observation of such evidence is sufficient for the PP4 application. Multiple observations cannot increase the code strength, as this would be considered proband counting. The VCEP will consider whether PP4 should be strengthened in future versions once a sufficient number of variants have been curated to allow for formal odds of pathogenicity calculations.

3.5.2. Segregation Data (BS4 and PP1)

(1) PP1 and BS4. Variant segregation and lack of segregation with disease fall under PP1 and BS4, respectively. For counting PP1 meiosis, the DICER1 VCEP adopted the same cutoffs used by other VCEPs [12–14] and informed by prior work [45, 46]. Namely, PP1 may be applied at supporting strength when 3 or 4 meioses are observed across one or more families, moderate strength when 5 or 6 meioses are observed across one or more families, and strong strength when seven or more meioses are observed across two or more families. Meioses are counted between phenotype-positive individuals with high-, moderate-, or low-specificity phenotypes as outlined in Table 2. PP1 was relaxed to include low-specificity phenotypes during the pilot, which improved its performance for pathogenic variants without resulting in excessive segregation counts. However, variant segregation with a single low-specificity phenotype (e.g., Wilms tumor) across multiple individuals is not sufficient for PP1 application. PP1 should not be applied when a variant also has population data meeting BA1 or BS1 since a common variant may appear to segregate with the disease by chance. BS4 may be applied if a proband has a phenotype-positive (must be high- or moderate-specificity), genotype-negative first-, second-, or third-degree relative. Genotype-positive and phenotype-negative individuals do not count toward BS4 but may be considered for BS2 (see 3.5.4).

3.5.3. De Novo Data (PS2)

(1) PS2. The DICER1 VCEP followed SVI recommendations for de novo criteria (https://clinicalgenome.org/working-groups/sequence-variant-interpretation/). Under the recommended framework, probands with de novo germline variants contribute 0, 0.25, 0.5, 1, or 2 points toward a de novo score based on the phenotype of the proband and whether parental relationships were confirmed (e.g., trio exome, maternity/paternity testing) or unconfirmed. Under this framework, a curator may apply either of the two de novo evidence codes originally proposed in the ACMG/AMP guidelines [9]. The DICER1 VCEP elected to adopt PS2 as the sole de novo evidence code and to exclude PM6 as redundant, instead using PS2 at a lower evidence strength when maternity/paternity was unconfirmed. The proposed point combinations are summarized in Table 3, and phenotypes are organized in Table 2. Points summed across unrelated probands indicate the strength application of PS2: supporting (0.5 to <1 point), moderate (1 to <2 points), strong (2 to <4 points), or very strong (≥4 points).

3.5.4. Allelic Data (BS2 and BP2)

(1) BS2. Because pathogenic DICER1 variants have incomplete penetrance, the DICER1 VCEP initially excluded BS2, which is considered benign evidence for a variant observed in a healthy adult. However, it became apparent during the pilot that a modified version of BS2 would be needed for multiple known benign variants to comfortably reach a benign classification. Based on a conservative neoplasm penetrance estimate of 10% in individuals aged 50-60 years with germline DICER1 variants [22] and higher penetrance in females than males, the VCEP determined that an observation of 10 or more unrelated females, who have reached 50 years of age without a tumor diagnosis, should qualify for BS2_Supporting, provided that the ratio of BS2-eligible females to PS4-eligible probands is equal to or greater than 10 : 1. Similarly, since a strong evidence code can be thought of as equivalent to four supporting level codes [17], an observation of 40 or more unrelated females, who have reached 50 years of age without a tumor diagnosis, should qualify for BS2 at full strength, provided that the ratio of BS2-eligible females to PS4-eligible probands is equal to or greater than 40 : 1. In both cases, all females should come from a single source (e.g., from a single laboratory, database, clinical cohort, or publication) to eliminate the possibility of double counting. Additionally, since the homozygous loss of function variants in DICER1 are thought to be embryonic lethal [47, 48], homozygous observations can also qualify for BS2 application. The DICER1 VCEP allows BS2 to be applied at full strength if homozygosity is observed in two or more healthy individuals or in one healthy individual if homozygosity is confirmed by parental testing. BS2_Supporting may be applied if two or more observations of homozygosity are made in individuals lacking clinical information.

(2) BP2. In cases where an additional P/LP germline DICER1 variant is found in a proband, BP2 may be applied if the P/LP variant is confirmed in trans with the variant under investigation. If the P/LP variant is in cis or in an unknown phase, three such observations are required for BP2 application, and the probands must not all carry the same P/LP variant. Similar to PS1 and PM5, the cooccurring P/LP variant must be classified by the DICER1 VCEP.

3.6. Evidence Code Combinations

Initially, the VCEP followed the originally recommended evidence code combinations [9] and stated that a single supporting evidence code should not be considered conflicting evidence if a clinically meaningful classification would otherwise be reached. However, the original combinations were not flexible enough to account for some of the combinations in round 1 of the pilot (e.g., 6 supporting pathogenic codes), and limitations with regard to resolving complex conflicting evidence code combinations are apparent. For those reasons, the VCEP pivoted to a flexible, modified Bayesian point approach for all evidence code combinations [18] for the final pilot curations. In this approach, supporting, moderate, strong, and very strong evidence codes are weighted at one, two, four, and eight points, respectively, with pathogenic evidence weighted positively and benign evidence weighted negatively. A sum of the points results in the final classification as outlined in Table 4.

3.7. Pilot

The VCEP tested the proposed evidence code specifications on 40 DICER1 variants as described in the Methods. The pilot results, including the evidence codes applied, are summarized in Table 5. To improve performance, the VCEP modified PP1, BS2, and the method for evidence code combinations as described above between round 1 and round 2 of the pilot. The changes implemented between the initial and final rounds of pilot classifications led to stronger variant classifications (i.e., more pathogenic or more benign) in nine variants (22.5%), including five variants which shifted from VUS to LB or LP.

Final VCEP classifications were clinically meaningful for 82.5% (33/40) of the pilot variants. Concordance for known P/LP and known B/LB pilot variants was 100% (14 of 14 P/LP and 12 of 12 B/LB). Pilot variants with conflicting or uncertain classifications in ClinVar reached 50% (7/14) resolution, with four variants reaching LB and three reaching LP.

4. Discussion

Under the ClinGen framework, the DICER1 VCEP developed and piloted DICER1-specific sequence variant curation guidelines. These guidelines performed very well on a set of pilot variants, with more than 80% of the pilot variants receiving a clinically meaningful classification. Furthermore, the pilot demonstrated that the guidelines could be interpreted and applied consistently by curators and that internal data sharing can be effectively integrated into the curation process. The pilot variants and evidence summaries have been submitted to ClinVar as three-star submissions [49]. Additional curation details for those variants are also available on the ClinGen Evidence Repository (https://erepo.clinicalgenome.org/evrepo/).

Past challenges in curating DICER1 missense variants have been recognized and even cited as a reason to exclude DICER1 from the ACMG secondary findings list [50]. The success of our guidelines in clarifying DICER1 variant classification not only implies fewer patients will be faced with VUS results in the future but also reduces this barrier for future reconsideration of DICER1 for the ACMG secondary findings list.

The VCEP will continue to meet regularly to further variant curation progress and submit classifications for public use. Variants will be prioritized by ClinVar classification (conflicting interpretations or VUS by multiple submitters) and by request. ClinVar currently contains ~5,000 DICER1 variant entries, including ~150 with conflicting interpretations and ~860 VUS by multiple submitters. Variant interpretations will be submitted to ClinVar within 30 days of VCEP approval. The VCEP will recurate variants classified as LP or VUS every two years to assess whether additional evidence is available. Medically significant discrepancies (i.e., P/LP vs. VUS/LB/B) between a VCEP submission and a more recent ClinVar submitter will be reviewed and updated as appropriate within six months of the discrepant submission. Other discrepancies (i.e., VUS vs. LB/B) will be reviewed within two years.

Due to the characteristic signature of DICER1 somatic mutations, the DICER1 VCEP chose to use somatic tumor testing as supporting evidence (PP4) [40]. The DICER1 VCEP is the first VCEP within the ClinGen Hereditary Cancer Clinical Domain to use somatic tumor testing to inform the PP4 application, providing a model for other VCEPs.

As more is learned and published on the DICER1 gene and the phenotypic consequences of its pathogenic variation, the VCEP will reevaluate the proposed guidelines and consider updates for future versions of the guidelines. For example, the phenotypic spectrum of the disorder may expand, or the specificity of certain phenotypes may need to be adjusted. Additionally, as more DICER1 variants are curated, the VCEP can revisit the odds of pathogenicity calculations for various evidence codes such as PP4 tumor phenotype evidence or PP3 and BP4 in silico predictor cutoffs and modify the strength of the evidence codes as appropriate. Any modifications to evidence specifications will be submitted to the SVI for approval and made publicly available on the ClinGen website (https://clinicalgenome.org/affiliation/50050/) as a resource for others curating DICER1 variants.

5. Conclusions

The DICER1-specific sequence variant curation guidelines developed by the ClinGen DICER1 VCEP show promising results on a pilot set of 40 variants, with 80% reaching clinically meaningful classifications. Consistent utilization of these guidelines may reduce the number of variants of uncertain significance returned to patients undergoing DICER1 sequencing. Future refinement of these guidelines over time is expected to further improve the clinical utility of variant classification.

Data Availability

The variant classifications made during this effort have been published in ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/), and the curated evidence collected has been made publicly available through the ClinGen Evidence Repository (https://erepo.clinicalgenome.org/evrepo/). Some detailed internal patient-level data is not publicly available for ethical and privacy reasons.

Disclosure

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Conflicts of Interest

The following authors work for laboratories that offer fee-for-service testing of DICER1: MJA, ECC, HCC, SBC, SH, JM, JLM, and NNY. The following authors have made substantial contributions to the DICER1 gene, disease literature, DRS, XB-D, KSC, WDF, SG, KAS, and MKW. NSA-H is an employee and equity holder of 23andMe, serves as a scientific advisory board member for Allelica, received personal fees from Genentech, Allelica, and 23andMe, received research funding from Akcea, and was previously employed by Regeneron Pharmaceuticals.

Authors’ Contributions

Leora Witkowski and Douglas R. Stewart are co-senior authors.

Acknowledgments

The ClinGen DICER1 and miRNA-Processing Gene Variant Curation Expert Panel thanks Steven Harrison, Leslie Biesecker, Sharon Plon, and the ClinGen Sequence Variant Interpretation Group for guidance and feedback on the development of our guidelines. We also thank past member D. Ashley Hill for her contributions and the patients whose data made curation possible. ClinGen is primarily funded by the National Human Genome Research Institute (NHGRI) with cofunding from the National Cancer Institute (NCI) through the following grants: Baylor/Stanford: U24HG009649, Broad/Geisinger: U24HG006834, and UNC/Kaiser: U24HG009650. The work of DRS, JNH, CH, MNF, and JK was supported by the Intramural Research Program of the Division of Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Health. WDF is funded by the Canadian Institutes of Health Research grant FDN-148390. KAS receives funding from NIH NCI R01 CA244940 and the Pine Tree Apple Classic Fund.

Supplementary Materials

Supplementary 1. Supplementary Table 1; DICER1 hotspot variants previously reported as somatic second hits.

Supplementary 2. Supplementary Figure 1: flowchart for DICER1-specific PVS1 code application. Modified from [25] (PMID 30192042).

References

L. de Kock, M. K. Wu, and W. D. Foulkes, “Ten years of DICER1 mutations: provenance, distribution, and associated phenotypes,” Human Mutation, vol. 40, no. 11, pp. 1939–1953, 2019.
View at: Publisher Site | Google Scholar
W. D. Foulkes, J. R. Priest, and T. F. Duchaine, “DICER1: mutations, microRNAs and mechanisms,” Nature Reviews. Cancer, vol. 14, no. 10, pp. 662–672, 2014.
View at: Publisher Site | Google Scholar
D. A. Hill, J. Ivanovich, J. R. Priest et al., “DICER1 mutations in familial pleuropulmonary blastoma,” Science, vol. 325, no. 5943, p. 965, 2009.
View at: Publisher Site | Google Scholar
H. Zhang, F. A. Kolb, L. Jaskiewicz, E. Westhof, and W. Filipowicz, “Single processing center models for human dicer and bacterial RNase III,” Cell, vol. 118, no. 1, pp. 57–68, 2004.
View at: Publisher Site | Google Scholar
I. A. González, D. R. Stewart, K. A. P. Schultz, A. P. Field, D. A. Hill, and L. P. Dehner, “_DICER1_ tumor predisposition syndrome: an evolving story initiated with the pleuropulmonary blastoma,” Modern Pathology, vol. 35, no. 1, pp. 4–22, 2022.
View at: Publisher Site | Google Scholar
J. J. Bakhuizen, H. Hanson, K. van der Tuin et al., “Surveillance recommendations for DICER1 pathogenic variant carriers: a report from the SIOPE host genome working group and CanGene-CanVar clinical guideline working group,” Familial Cancer, vol. 20, no. 4, pp. 337–348, 2021.
View at: Publisher Site | Google Scholar
K. A. P. Schultz, S. P. Rednam, J. Kamihara et al., “PTEN, DICER1, FH, and their associated tumor susceptibility syndromes: clinical features, genetics, and surveillance recommendations in childhood,” Clinical Cancer Research, vol. 23, no. 12, pp. e76–e82, 2017.
View at: Publisher Site | Google Scholar
K. A. P. Schultz, G. M. Williams, J. Kamihara et al., “DICER1 and associated conditions: identification of at-risk individuals and recommended surveillance strategies,” Clinical Cancer Research, vol. 24, no. 10, pp. 2251–2261, 2018.
View at: Publisher Site | Google Scholar
S. Richards, N. Aziz, S. Bale et al., “Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology,” Genetics in Medicine, vol. 17, no. 5, pp. 405–424, 2015.
View at: Publisher Site | Google Scholar
H. L. Rehm, J. S. Berg, L. D. Brooks et al., “ClinGen--the clinical genome resource,” The New England Journal of Medicine, vol. 372, no. 23, pp. 2235–2242, 2015.
View at: Publisher Site | Google Scholar
E. A. Rivera-Muñoz, L. V. Milko, S. M. Harrison et al., “ClinGen variant curation expert panel experiences and standardized processes for disease and gene-level specification of the ACMG/AMP guidelines for sequence variant interpretation,” Human Mutation, vol. 39, no. 11, pp. 1614–1622, 2018.
View at: Publisher Site | Google Scholar
C. Fortuno, K. Lee, M. Olivier et al., “Specifications of the ACMG/AMP variant interpretation guidelines for germline TP53 variants,” Human Mutation, vol. 42, no. 3, pp. 223–236, 2021.
View at: Publisher Site | Google Scholar
K. Lee, K. Krempely, M. E. Roberts et al., “Specifications of the ACMG/AMP variant curation guidelines for the analysis of germline CDH1 sequence variants,” Human Mutation, vol. 39, no. 11, pp. 1553–1568, 2018.
View at: Publisher Site | Google Scholar
J. L. Mester, R. Ghosh, T. Pesaran et al., “Gene-specific criteria for PTEN variant curation: recommendations from the ClinGen PTEN expert panel,” Human Mutation, vol. 39, no. 11, pp. 1581–1592, 2018.
View at: Publisher Site | Google Scholar
D. Wu, X. Luo, S. Feurstein et al., “How I curate: applying American Society of Hematology-clinical genome resource myeloid malignancy variant curation expert panel rules for RUNX1 variant curation for germline predisposition to myeloid malignancies,” Haematologica, vol. 105, no. 4, pp. 870–887, 2020.
View at: Publisher Site | Google Scholar
C. G. Preston, M. W. Wright, R. Madhavrao et al., “ClinGen variant curation interface: a variant classification platform for the application of evidence criteria from ACMG/AMP guidelines,” Genome Medicine, vol. 14, no. 1, article 6, 2022.
View at: Publisher Site | Google Scholar
S. V. Tavtigian, M. S. Greenblatt, S. M. Harrison et al., “Modeling the ACMG/AMP variant classification guidelines as a Bayesian classification framework,” Genetics in Medicine, vol. 20, no. 9, pp. 1054–1060, 2018.
View at: Publisher Site | Google Scholar
S. V. Tavtigian, S. M. Harrison, K. M. Boucher, and L. G. Biesecker, “Fitting a naturally scaled point system to the ACMG/AMP variant classification guidelines,” Human Mutation, vol. 41, no. 10, pp. 1734–1737, 2020.
View at: Publisher Site | Google Scholar
L. G. Biesecker, S. M. Harrison, and ClinGen Sequence Variant Interpretation Working Group, “The ACMG/AMP reputable source criteria for the interpretation of sequence variants,” Genetics in Medicine, vol. 20, no. 12, pp. 1687-1688, 2018.
View at: Publisher Site | Google Scholar
N. Whiffin, E. Minikel, R. Walsh et al., “Using high-resolution variant frequencies to empower clinical genome interpretation,” Genetics in Medicine, vol. 19, no. 10, pp. 1151–1158, 2017.
View at: Publisher Site | Google Scholar
J. Kim, A. Field, K. A. P. Schultz, D. A. Hill, and D. R. Stewart, “The prevalence of DICER1 pathogenic variation in population databases,” International Journal of Cancer, vol. 141, no. 10, pp. 2030–2036, 2017.
View at: Publisher Site | Google Scholar
D. R. Stewart, A. F. Best, G. M. Williams et al., “Neoplasm risk among individuals with a pathogenic germline variant in DICER1,” Journal of Clinical Oncology, vol. 37, no. 8, pp. 668–676, 2019.
View at: Publisher Site | Google Scholar
R. Ghosh, S. M. Harrison, H. L. Rehm, S. E. Plon, L. G. Biesecker, and on behalf of ClinGen Sequence Variant Interpretation Working Group, “Updated recommendation for the benign stand-alone ACMG/AMP criterion,” Human Mutation, vol. 39, no. 11, pp. 1525–1530, 2018.
View at: Publisher Site | Google Scholar
M. Brenneman, A. Field, J. Yang et al., “Temporal order of RNase IIIb and loss-of-function mutations during development determines phenotype in DICER1 syndrome: a unique variant of the two-hit tumor suppression model,” F1000Research, vol. 4, p. 214, 2015.
View at: Publisher Site | Google Scholar
A. N. Abou Tayoun, T. Pesaran, M. T. DiStefano et al., “Recommendations for interpreting the loss of function PVS1 ACMG/AMP variant criterion,” Human Mutation, vol. 39, no. 11, pp. 1517–1524, 2018.
View at: Publisher Site | Google Scholar
N. M. Ioannidis, J. H. Rothstein, V. Pejaver et al., “REVEL: an ensemble method for predicting the pathogenicity of rare missense variants,” American Journal of Human Genetics, vol. 99, no. 4, pp. 877–885, 2016.
View at: Publisher Site | Google Scholar
V. Pejaver, A. B. Byrne, B.-J. Feng et al., Evidence-based calibration of computational tools for missense variant pathogenicity classification and ClinGen recommendations for clinical use of PP3/BP4 criteria, bioRxiv, 2022.
View at: Publisher Site
E. Pontén, S. Frisk, F. Taylan et al., “A complex DICER1 syndrome phenotype associated with a germline pathogenic variant affecting the RNase IIIa domain of DICER1,” Journal of Medical Genetics, vol. 59, no. 2, pp. 141–146, 2022.
View at: Publisher Site | Google Scholar
L. de Kock, N. Sabbaghian, F. Plourde et al., “Pituitary blastoma: a pathognomonic feature of germ-line DICER1 mutations,” Acta Neuropathologica, vol. 128, no. 1, pp. 111–122, 2014.
View at: Publisher Site | Google Scholar
M. Apellaniz-Ruiz, L. de Kock, N. Sabbaghian et al., “Familial multinodular goiter and Sertoli-Leydig cell tumors associated with a large intragenic in-frame DICER1 deletion,” European Journal of Endocrinology, vol. 178, no. 2, pp. K11–K19, 2018.
View at: Publisher Site | Google Scholar
M. Apellaniz-Ruiz, M. Segni, M. Kettwig et al., “Mesenchymal hamartoma of the liver and DICER1 syndrome,” The New England Journal of Medicine, vol. 380, no. 19, pp. 1834–1842, 2019.
View at: Publisher Site | Google Scholar
R. Grantham, “Amino acid difference formula to help explain protein evolution,” Science, vol. 185, no. 4154, pp. 862–864, 1974.
View at: Publisher Site | Google Scholar
M. K. Wu, G. M. Vujanic, S. Fahiminiya et al., “Anaplastic sarcomas of the kidney are characterized by _DICER1_ mutations,” Modern Pathology, vol. 31, no. 1, pp. 169–178, 2018.
View at: Publisher Site | Google Scholar
R. P. Guillerman, W. D. Foulkes, and J. R. Priest, “Imaging of DICER1 syndrome,” Pediatric Radiology, vol. 49, no. 11, pp. 1488–1505, 2019.
View at: Publisher Site | Google Scholar
L. A. Doros, C. T. Rossi, J. Yang et al., “DICER1 mutations in childhood cystic nephroma and its relationship to DICER1-renal sarcoma,” Modern Pathology, vol. 27, no. 9, pp. 1267–1280, 2014.
View at: Publisher Site | Google Scholar
K. A. P. Schultz, A. K. Harris, M. Finch et al., “DICER1 -related Sertoli-Leydig cell tumor and gynandroblastoma: clinical and genetic findings from the International Ovarian and Testicular Stromal Tumor Registry,” Gynecologic Oncology, vol. 147, no. 3, pp. 521–527, 2017.
View at: Publisher Site | Google Scholar
R. Y. Diaz Coronado, M. Mynarek, C. Koelsche et al., “Primary central nervous system sarcoma with DICER1 mutation—treatment results of a novel molecular entity in pediatric Peruvian patients,” Cancer, vol. 128, no. 4, pp. 697–707, 2022.
View at: Publisher Site | Google Scholar
C. Koelsche, M. Mynarek, D. Schrimpf et al., “Primary intracranial spindle cell sarcoma with rhabdomyosarcoma-like features share a highly distinct methylation profile and DICER1 mutations,” Acta Neuropathologica, vol. 136, no. 2, pp. 327–337, 2018.
View at: Publisher Site | Google Scholar
M. Altaraihi, T. V. O. Hansen, E. Santoni-Rugiu et al., “Prevalence of pathogenic germline DICER1 variants in young individuals thyroidectomised due to goitre - a national Danish cohort,” Frontiers in Endocrinology, vol. 12, article 727970, 2021.
View at: Publisher Site | Google Scholar
M. F. Walsh, D. I. Ritter, C. Kesserwan et al., “Integrating somatic variant data and biomarkers for germline variant classification in cancer predisposition genes,” Human Mutation, vol. 39, no. 11, pp. 1542–1552, 2018.
View at: Publisher Site | Google Scholar
K. S. Chen, S. H. Stuart, E. K. Stroup et al., “DistinctDICER1 hotspot mutations identify bilateral tumors as separate events,” JCO Precision Oncology, vol. 2, no. 2, pp. 1–9, 2018.
View at: Publisher Site | Google Scholar
A. Garcia, L. Desrosiers, S. Scollon et al., “Distinct somatic _DICER1_ hotspot mutations in three metachronous ovarian Sertoli-Leydig cell tumors in a patient with _DICER1_ syndrome,” Cancer Genetics, vol. 262-263, pp. 53–56, 2022.
View at: Publisher Site | Google Scholar
S. Gadd, V. Huff, A. L. Walz et al., “A Children’s Oncology Group and TARGET initiative exploring the genetic landscape of Wilms tumor,” Nature Genetics, vol. 49, no. 10, pp. 1487–1494, 2017.
View at: Publisher Site | Google Scholar
M. K. Wu, N. Sabbaghian, B. Xu et al., “Biallelic DICER1 mutations occur in Wilms tumours,” The Journal of Pathology, vol. 230, no. 2, pp. 154–164, 2013.
View at: Publisher Site | Google Scholar
G. P. Jarvik and B. L. Browning, “Consideration of cosegregation in the pathogenicity classification of genomic variants,” American Journal of Human Genetics, vol. 98, no. 6, pp. 1077–1081, 2016.
View at: Publisher Site | Google Scholar
D. Thompson, D. F. Easton, and D. E. Goldgar, “A full-likelihood method for the evaluation of causality of sequence variants from family data,” American Journal of Human Genetics, vol. 73, no. 3, pp. 652–655, 2003.
View at: Publisher Site | Google Scholar
E. Bernstein, S. Y. Kim, M. A. Carmell et al., “Dicer is essential for mouse development,” Nature Genetics, vol. 35, no. 3, pp. 215–217, 2003.
View at: Publisher Site | Google Scholar
V. Teijeiro, D. Yang, S. Majumdar et al., “DICER1 is essential for self-renewal of human embryonic stem cells,” Stem Cell Reports, vol. 11, no. 3, pp. 616–625, 2018.
View at: Publisher Site | Google Scholar
M. J. Landrum, J. M. Lee, M. Benson et al., “ClinVar: improving access to variant interpretations and supporting evidence,” Nucleic Acids Research, vol. 46, no. D1, pp. D1062–d1067, 2018.
View at: Publisher Site | Google Scholar
D. T. Miller, K. Lee, W. K. Chung et al., “ACMG SF v3.0 list for reporting of secondary findings in clinical exome and genome sequencing: a policy statement of the American College of Medical Genetics and Genomics (ACMG),” Genetics in Medicine, vol. 23, no. 8, pp. 1381–1390, 2021.
View at: Publisher Site | Google Scholar

Copyright

Copyright © 2023 Jessica N. Hatton et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

PDF Download Citation

Download other formats

Order printed copies

Views

1754

Downloads

961

Citations