Re‐evaluation of missense variant classifications in NF2

Abstract Missense variants in the NF2 gene result in variable NF2 disease presentation. Clinical classification of missense variants often represents a challenge, due to lack of evidence for pathogenicity and function. This study provides a summary of NF2 missense variants, with variant classifications based on currently available evidence. NF2 missense variants were collated from pathology‐associated databases and existing literature. Association for Clinical Genomic Sciences Best Practice Guidelines (2020) were followed in the application of evidence for variant interpretation and classification. The majority of NF2 missense variants remain classified as variants of uncertain significance. However, NF2 missense variants identified in gnomAD occurred at a consistent rate across the gene, while variants compiled from pathology‐associated databases displayed differing rates of variation by exon of NF2. The highest rate of NF2 disease‐associated variants was observed in exon 7, while lower rates were observed toward the C‐terminus of the NF2 protein, merlin. Further phenotypic information associated with variants, alongside variant‐specific functional analysis, is necessary for more definitive variant interpretation. Our data identified differences in frequency of NF2 missense variants by exon between gnomAD population data and NF2 disease‐associated variants, suggesting a potential genotype‐phenotype correlation; further work is necessary to substantiate this.

The majority of pathogenic variants identified in NF2 result in truncation of the protein product, often causing loss of protein expression or creating nonfunctional proteins (Evans, 2009).
Genotype-phenotype correlations have been observed in NF2, where protein-truncating variants, such as frameshift or nonsense, result in more severe disease presentation than missense variants (Ruttledge et al., 1996;Smith et al., 2011). In cases where truncating variants result in a severe phenotype, a dominant negative action of the variant protein has been proposed (Evans, 2015). Variants in regulatory elements, such as splice sites and larger structural variants for example, ring chromosome 22, often result in variable disease presentation (Evans, 2009). Still, splice site variants positioned earlier in the NF2 transcript have been associated with more severe disease presentation (Baser et al., 2005;Kluwe et al., 1998). Investigation of missense variant genotype-phenotype correlations presents a unique challenge, as a function of an amino acid residue is not necessarily related to its position within a transcript, but rather its location within protein tertiary structures (Suckow et al., 1996).
Missense variants often represent clinical dilemmas for diagnostic services due to challenges of obtaining evidence for pathogenicity and function. Diagnostic classification of missense variants largely relies upon population frequency data and in silico predictive tools, as well as familial and functional data when available. Release of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology (ACMG-AMP) guidelines for variant interpretation (Richards et al., 2015) enabled more reproducible interpretation of variants by providing an evidence framework, facilitating more consistent clinical reporting. Subsequent revision of these guidelines has followed and the Association for Clinical Genomic Sciences (ACGS) Best Practice Guidelines for Variant Classification in Rare Disease 2020 is the framework now currently employed by the National Health Service (NHS) within the UK (ACGS best practice guidelines, 2020 https://www.acgs.uk.com/quality/ best-practice-guidelines/#VariantGuidelines. Accessed 23 August 2021). The ACGS 2020 guidelines combine the detailed guidance of Richards et al. (2015), with clarifications and developments proposed by other research groups (Tavtigian et al., 2018). Key developments in the ACGS (2020) guidelines from the ACMG-AMP include: defining variant-specific, rather than gene-specific, effects from functional studies, resolving scoring inconsistencies from combining evidence criteria, and the sub-division of pathogenic, likely pathogenic and variant of uncertain significance (VUS) classifications. Further disease-specific guidelines are currently in development through ClinGen and other curation networks, which incorporate additional disease-associated features into variant classification; for example, loss of heterozygosity (LOH) and retention of a missense variant in a tumor would be informative for NF2 variant classification. Recently proposed improvements in NF2 genetic severity scores suggest incorporation of merlin functional assays conducted in patient fibroblasts (Catasús et al., 2021), this evidence would be similarly valuable for NF2 variant interpretation.
While missense variants only account for~9% of diagnosed NF2 cases (Heineman et al., 2015), they represent >25% of observed NF2 variants in gnomAD. This disparity may be attributed to tolerability of the NF2 protein to missense variation, but might also suggest reduced phenotype severity and disease penetrance in individuals who possess missense variants. This suggestion is supported by observed phenotypic variation in familial cases of NF2, such as the c.1604T>C p.(Leu535Pro) missense variant (Heineman et al., 2015).  Figure 1 shows a flow chart detailing the order of variant compilation and numbers of variants included and excluded at each step. An extra literature mining step was conducted using LitVar to capture any missing variants (Allot et al., 2018). A total of 395 unique variants were included within the study.
A subset of variants identified in patients with a confirmed Manchester Criteria NF2 diagnosis (Table S1) or known NF2associated features, for example, unilateral VS, meningioma, ependymoma, were grouped for further analysis. A total of 97 NF2 diseaseassociated variants were included, 69 of these variants appear in public databases, 17 were identifiable in the literature, the remaining 11 were exclusive to local databases and have since been submitted to public variant databases ( Figure 1).
Variants outside the exonic regions of the primary NF2 transcript RefSeq NM_000268.4 (isoform 1) were excluded from analysis, as well as variants described as nonsense, frameshifts, insertions, deletions, indels, and synonymous.

| Variant classification tools
Evidence for clinical classification of variants was obtained and interpreted following the ACGS best practice guidelines (2020).

| Population and frequency data
Maximum credible population allele frequency was determined using the alleleFrequencyApp (cardiodb.org/allelefrequencyapp) (Whiffin et al., 2017), and was calculated to be 1.88e-07 for NF2, based on the following input parameters: monoallelic inheritance, disease incidence of 1 in 28,000 (Evans et al., 2018), allelic heterogeneity 0.01 and penetrance 0.95, accounting for the known rate of recurrent pathogenic variants and late disease onset. Strong benign evidence (BS1) was applied to any variants with an allele frequency equal to or higher than NF2 disease incidence (1/28,000). With a low maximum credible population allele frequency calculated (1.88e-07), moderate pathogenicity evidence (PM2) based on frequency data was not applied to any variant observed in gnomAD as frequency values of observed variants exceeded this value.

| Functional data
With a predicted missense constraint Z score of 2.29 in ExAC, NF2 is considered moderately intolerant of variation. However, only Z scores ≥3.09 are considered significant within the ACGS guidelines and therefore variants in NF2 are ineligible for application of evidence for missense constraint (PP2).
The DECIPHER database (Firth et al., 2009) was used to investigate possible mutational hotspots or identify regional con- Splice prediction tools were also interpreted and applied as evidence, as suggested in the ACGS 2020 guidelines. Variants that received MaxEntScan (Yeo & Burge, 2004) predictions of >15% score reduction compared to reference allele, and SpliceSiteFinder-Like (Zhang, 1998) predictions with >5% reduction, had PP3 computational evidence of pathogenicity applied in their classification.

| Clinical information
If phenotype was described, patients who fitted Manchester Criteria for NF2 disease (Table S1) (Evans et al., 1992;Smith et al., 2017) were considered to have phenotypic specificity for a disease of single etiology (PP4), applied as supporting evidence of pathogenicity.
Where possible, family history and segregation data was applied to the evidence framework.  Table S2.

| Other databases
While 395 variants were collated in total, only 97 were identified in cases with confirmed NF2-associated phenotypic features (- Table 2). All variants classified as likely pathogenic and pathogenic were identified in association with NF2 disease presentation, and were therefore assigned to both data groups in Table 2.
Seventeen NF2 missense variants had in silico computational evidence of pathogenicity (PP3) applied by splicing prediction tool scores, MaxEntScan (Yeo & Burge, 2004) and SpliceSiteFinder-Like (Zhang, 1998), in the absence of a pathogenic REVEL metascore. All seventeen of these potential splicing variants remain classified as VUS.

| Conflict with existing classifications
When all variant classifications were compared to existing ClinVar interpretations, 17 variants were in conflict with current submissions, seen in Table 3. The vast majority of these variants were downgraded in pathogenicity class.

| Rate of variation across NF2
The number of variants identified in each exon of NF2 was compared to exon size in amino acids. Missense variants recorded within gnomAD occurred at a highly consistent rate across the NF2 transcript, Figure 2. When considering all 395 NF2 variants identified in this study rates per exon differed, yet the average trendline remained consistent across the gene (Figure 2   F I G U R E 2 A comparison of rates of NF2 missense variants in gnomAD v2.1.1 (controls), all variants identified within this study, and NF2 disease-associated variants. Rates were calculated as a percentage of the number of variants in comparison to exon size in amino acids. Assumed benign variation in the gnomAD v2.1.1 (controls) data set remains consistent across the gene. In contrast, there is an increased rate of variation in a number of exons for variants identified in pathology databases rates of NF2-associated variants across exons 2 and 7 are observable in Figure 3.

| Somatic variants
From the 395 variants collated within this study, 39 had been observed exclusively in somatic samples. Many of the somatic samples were obtained from schwannoma and meningioma tumors, however, 15 of the variants were identified exclusively in non-NF2 related tumor types, such as liver, breast and lung cancers (Table S2).

| DISCUSSION
The vast majority of missense variants identified within NF2 are classified as variants of uncertain significance in accordance with the ACGS 2020 guidelines. Unfortunately, these variants remain as clinical interpetation dilemmas without sufficient evidence to ascribe or discount them as disease causing. While the VUS temperature scale provides further insight into the possible pathogenicity of a variant, many variants remain at the "cooler" end of the scale with little compelling evidence available, see Table 1 (Shimizu et al., 2002;Stokowski & Cox, 2000), it seems likely that regional constraint could be better defined for NF2. Identifying areas of regional constraint would enable the application of moderate evidence for pathogenicity that might enable the revision of a number of variants into likely pathogenic and pathogenic classifications. Exploring ways to redefine regional constraint and domain function for NF2 may prove valuable in the curation of NF2 disease-specific variant interpretation guidelines. Full details of the 395 NF2 missense variants is available in Table S2. Another consideration of ClinVar variant classifications is the age of the studies that were used to assign pathogenicity; a number of variants were submitted to ClinVar before the inception of the clinical variant interpretation guidelines suggested by Richards et al. (2015), and therefore evidence is often applied with inconsistent weighting in these earlier submissions.
When considering missense variant rates by exon size, a highly consistent rate of assumed benign variation was observed in the gnomADv2.1.1 (controls) data set ( Figure 2). In contrast, the variants collated from pathology databases for this study demonstrated  Figure 3), the sequence of exon 7 in NF2 is highly conserved across the ERM (ezrin, radixin, moesin) protein superfamily (Shimizu et al., 2002). The sequence conservation of exon 7, alongside the high rate of NF2-associated missense variants, suggests that alteration of amino acid residues in this region may disrupt critical biophysical interactions of the merlin protein. For example, the exon 7 variant c.658A>T p.(Asn220Tyr) has been reported to display reduced binding to scaffolding protein EBP50 (Stokowski & Cox, 2000); Shimizu et al. (2002) theorized that this may be due to altered residue contacts resulting in changes to subdomain orientation.
Rates of NF2-associated variants decreased toward the end of the NF2 gene, which may suggest that variants positioned later in the gene transcript are less likely to disrupt function of the protein, similar to the genotype-phenotype correlation observed in NF2 splice variants (Baser et al., 2005;Kluwe et al., 1998). Moreover, the single NF2 disease-associated variant identified in exon 17 was observed in a somatic astrocytoma sample from one individual. Astrocytomas are observed very rarely in association with NF2 (Gene Reviews-Neurofibromatosis 2, 2018. https://www.ncbi.nlm.nih.gov/books/ NBK1201/. Accessed September 02, 2021) and it is possible that this variant was acquired somatically in the tumor and is not related to NF2 disease. As the two predominant isoforms of merlin possess variant C-terminal ends (Shimizu et al., 2002), it is possible there is transcript redundancy that reduces the pathological effect of variants toward the end of the gene. As only isoform 1 of NF2 has been analyzed within this study it should be considered that some variants may confer transcript-specific effects currently unaccounted for in our interpretation.
Fifteen of the NF2 missense variants included in this study were observed exclusively in somatic samples from non-NF2 related tumors, and this is consistent with previous observations of somatic NF2 variants in multiple cancer types, such as mesothelioma, liver, and large instestine cancers (Schroeder et al., 2014).
Merlin is a known tumor suppressor, regulating multiple cell signaling pathways associated with cell proliferation and therefore tumorigenesis of multiple cancer types Trofatter et al., 1993).  (Heineman et al., 2015). Since missense variants generally lead to a milder phenotype, they are more likely to be seen as non-mosaic variants (Evans et al., 2013). interpretation guidelines by increasing the strength of the PP4 evidence class to moderate or strong, "patient phenotype or family history is highly specific for a disease with a single genetic etiology." In conclusion, most NF2 missense variants remain classified as variants of uncertain significance after application of current ACGS guidelines. Our observation of differing missense variant rates by exon of NF2, with fewer NF2-associated variants toward the Cterminus of merlin, is suggestive of a potential genotype-phenotype correlation, although further work is necessary to substantiate this.
While we provide a comprehensive list of NF2 missense variants, it is not exhaustive, and we encourage other researchers within the field to submit novel variants to public databases. This is particularly

ACKNOWLEDGMENTS
The authors wish to acknowledge the National Health Service (NHS) England funded highly specialized NF2 service, Prof Evans is an NIHR Senior Investigator. This study was supported by the Manchester

National Institute for Health Research (NIHR) Biomedical Research
Centre (IS-BRC-1215-20007). The authors have not declared a specific grant for this study from any funding agency in the public, commercial or not-for-profit sectors.

CONFLICTS OF INTEREST
The authors declare no conflicts of interest.

ETHICS STATEMENT
Ethical approval for the use of anonymised samples from the Manchester Centre for Genomic Medicine archive was obtained from the North West-Greater Manchester Central Research Ethics Committee (reference 10/H1008/74). Ethical approval for the use of deidentified data from the UAB Medical Genomics Laboratory was obtained from the UAB Institutional Review Board, project number 080926009.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions. ACGS best practice guidelines (2020):