1 INTRODUCTION

Neurofibromatosis type 1 (NF1; MIM# 162200) is one of the most common autosomal dominant disorders with a birth incidence of 1 in 2,000–3,000 (Evans et al., 2010; Lammert, Friedman, Kluwe, & Mautner, 2005; Uusitalo et al., 2015). It is characterized by a highly variable expressivity and age-dependent clinical features, including multiple café-au-lait macules (CALMs), skinfold freckling, Lisch nodules, cutaneous, plexiform and/or spinal neurofibromas, optic pathway gliomas (OPGs), neoplasms, skeletal abnormalities, and learning difficulties. The wide range of variable expression makes genetic counseling and anticipatory guidance challenging (Friedman, 2014).

NF1 results from loss-of-function pathogenic variants in the NF1 tumor suppressor gene (MIM# 613113), located on chromosome 17q11.2, a locus with one of the highest spontaneous mutation rates across single-gene human genetic disorders (Friedman, 2014; Huson, Compston, Clark, & Harper, 1989). Indeed, the extreme diversity of the NF1 mutational spectrum observed in the University of Alabama at Birmingham (UAB) cohort of approximately 8,000 unrelated NF1-affected individuals, all comprehensively analyzed with NF1 transcript analysis, includes greater than 3,000 different constitutional variants, with only 31 having a prevalence of ≥0.4% (Messiaen & Wimmer, 2012; Messiaen, in preparation). The constitutional NF1 microdeletion represents the most frequent recurrent pathogenic variant (~3–8%) and therefore was first recognized to have a genotype–phenotype association with a distinct severe form of NF1 (Kehrer-Sawatzki, Mautner, & Cooper, 2017). Nontruncating pathogenic variants affecting only a single amino acid are more likely to be associated with genotype–phenotype correlations, as they may represent gain-of-function or hypomorphic variants. Although missense variants contribute to approximately 17% of all variants found in unrelated NF1-affected individuals in the UAB cohort, very few are recurrent and therefore amenable to variant-specific genotype–phenotype analyses. Indeed, only six true nontruncating hotspots, each with a prevalence ≥0.4%, were found in the UAB cohort, i.e., missense variants at codons 844-848, 1149, 1276, 1423, and 1809 and a single amino acid deletion p.Met992del (Figure 1). To date, genotype–phenotype associations affecting pathogenic missense variants at NF1 codons 1809 and 844-848 and p.Met992del have been reported, representing approximately 1.2, 0.8, and 0.9% of unrelated probands, respectively (Koczkowska et al., 2018; Koczkowska et al., 2019; Pinna et al., 2015; Rojnueangnit et al., 2015; Upadhyaya et al., 2007). Increased efforts towards the identification of additional clinically relevant genotype–phenotype correlations are needed. Therefore, we examined with priority those individuals carrying a pathogenic variant at one of the three remaining nontruncating hotspots, i.e. at p.Met1149, p.Arg1276 and p.Lys1423.

In this cross-sectional study, we investigated the clinical spectrum of 281 individuals from 237 unrelated families identified through clinical genetic testing as being heterozygous for a missense variant at p.Met1149, p.Arg1276, or p.Lys1423.

2 METHODS

In total, 237 unrelated probands and 44 relatives, all carrying a pathogenic NF1 missense variant at either p.Met1149 (50 probands and 19 relatives), p.Arg1276 (101 probands and 18 relatives), or p.Lys1423 (86 probands and 7 relatives), were included in this study (Figure 2). Briefly, blood samples from 220 individuals (184 probands and 36 relatives) were originally sent to the Medical Genomics Laboratory at the UAB for molecular NF1 genetic testing to establish or confirm the diagnosis of NF1. Two individuals from the UAB cohort, carrying c.3445A>T (p.Met1149Leu) and c.4268A>T (p.Lys1423Met) were excluded from the genotype–phenotype studies as these specific variants are still interpreted as “variant of uncertain significance” (VUS) or “likely pathogenic”, respectively, but not (yet) “pathogenic”, according to the here applied American College of Medical Genetics and Genomics (ACMG) recommendations (Richards et al., 2015). This initial study was expanded to include an additional 63 individuals (55 probands and 8 relatives), molecularly diagnosed in collaborating institutions (please see details in the Supporting Information Methods). The collaborating institutions and their referring physicians completed the same phenotypic checklist and as they did not have prior insight into the phenotypic associations emerging from the UAB cohort, their contributions were independent and unbiased. All variants identified in this study with the confirmed origin of the variants were submitted to the LOVD and ClinVar databases.

Comprehensive NF1 molecular analysis, interpretation of variant pathogenicity and collection of clinical data (see details in Supporting Information Methods) was performed as previously described (Koczkowska et al., 2018; Koczkowska et al., 2019; Messiaen et al., 2000; Richards et al., 2015; Rojnueangnit et al., 2015). Two-tailed Fisher's exact test with p < .05 considered as statistically significant was applied. The resulting p values were adjusted for multiple comparisons using the Benjamini–Hochberg (B-H) procedure with false discovery rates (FDR) at 0.05 and 0.01 (Thissen, Steinberg, & Kuang, 2002). These statistical analyses were performed with GraphPad software. The risk ratio was calculated with the STATCALC module from the Epi Info^TM package version 7.2.3.1 (http://cdc.gov/epiinfo/index.html).

This study was approved by the UAB Institutional Review Board and all individuals participating in this study or their legal guardians signed the informed consent forms for clinical genetic testing.

3 RESULTS

4 DISCUSSION

NF1 and Noonan syndrome (NS; MIM# 163950) represent RASopathies, a group of phenotypically related conditions, caused by pathogenic variants in genes involved in the RAS/mitogen-activated protein kinase signaling transduction pathway. Due to some overlapping phenotypic features, a clinical diagnosis can be complicated. “Classic” NF1 cases presenting with Noonan-like features were classified as having neurofibromatosis-Noonan syndrome (NFNS; MIM# 601321), first described 30 years ago (Allanson, Hall, & Van Allen, 1985; Opitz & Weaver, 1985). Although few cases have been reported to carry both a pathogenic NF1 and PTPN11 (MIM# 176876) variant (Bertola et al., 2005; Nyström et al., 2009; Thiel et al., 2009), as a rule, NF1 is the genetic cause underlying NFNS. De Luca et al. (2005) identified pathogenic variants spread across the NF1 gene, including nonsense, missense, out-of-frame and small in-frame deletions, and one multiexon deletion in 16/17 unrelated NFNS individuals. Among four individuals with NFNS and PS, 2/4 carried an in-frame single amino acid deletion and 1/4 a pathogenic missense variant with 2/3 located in the GTPase-activating protein-related domain (GAP-GRD). Ben-Shachar et al. (2013) reported a higher prevalence of PS in NFNS individuals compared with the general NF1 population (9/35, 26% vs. 25/2322, 1.1%; p < .001), with NFNS cases carrying a nontruncating pathogenic variant having the highest risk (p < .001). However, this study did not provide more specific information related to location or type of nontruncating pathogenic variants and was not adjusted for multiple comparisons.

So far, only two specific NF1 pathogenic variants have been associated with a statistically significant increased prevalence of Noonan-like features compared with “classic” NF1-affected cohorts, i.e., p.Arg1809 and p.Met992del, both located outside the GRD (Koczkowska et al., 2019; Pinna et al., 2015; Rojnueangnit et al., 2015; Upadhyaya et al., 2007). Here, we report that NF1 pathogenic missense variants at p.Met1149, p.Arg1276, and p.Lys1423 also are associated with this distinct phenotype (all p < .0001, significant at FDR of 0.01 after B-H correction), with no pathogenic variants in other NS genes detected (Tables S6, S8, S10, S12, S20, and S21). A high prevalence of PS was associated with p.Arg1276, p.Lys1423 and p.Arg1809 compared to the general NF1 population as well as the UAB NF1 nonsense variant cohort (all P < 0.0001, significant at FDR of 0.01 after B-H correction, Tables 2, 3 and Table S18). These observations provide evidence that although PS overall is more frequently observed in individuals carrying one of the nontruncating pathogenic variants in the current and previous studies (48/625, 7.7%), the prevalence of PS and other cardiac/cardiovascular abnormalities is highest in the p.Arg1276- and p.Lys1423- positive individuals, both located in the GRD (at least 10-fold higher than in the “classic” NF1 population; Table 4 and Tables S31 and S32), whereas is not significantly increased in individuals carrying pathogenic variants affecting codons 844-848 and 1149. As cardiovascular abnormalities may be congenital or become symptomatic at a young age and are associated with morbidity and mortality (Friedman et al., 2002; Lin et al., 2000), all individuals carrying one of the aforementioned pathogenic NF1 missense variants should receive a detailed cardiac examination.

Watson syndrome (WS; MIM# 193520) was described in 1967 and is characterized by CALMs, PS, and intellectual disability, and is allelic to NF1 (Ben-Shachar et al., 2013; Watson, 1967). Among 24 cases with PS found in this report, one 12-year-old individual (UAB-R1256) presented solely with greater than five CALMs, PS, developmental delay, severe scoliosis, and short stature. Additionally, two children less than 4 years old (UAB-R37401FN.202 and UAB-R8917) had greater than five CALMs, PS, abnormal development with speech delay, and Noonan-like facial features as the only features. NF1 c.3827G>A (p.Arg1276Gln), identified in the largest WS family from the original study (Ben-Shachar et al., 2013; Watson, 1967) was also found in UAB-R1256. Although 40% of the p.Arg1276-positive individuals ≥9 years old (29/73) had a severe phenotype with cardiac/cardiovascular abnormalities and/or spinal neurofibromas, 26% of cases ≥9 years old (19/73) presented with only pigmentary manifestations and a variable amount of learning disabilities (Table S6). Therefore, further studies aimed at identifying potential modifying genes, unlinked to the NF1 locus, in these individuals should be considered.

Familial spinal neurofibromatosis (FSNF; MIM# 162210) is another rare subtype of NF1, characterized by the presence of multiple histopathologically confirmed neurofibromas along spinal nerve roots with few, if any, cutaneous neurofibromas, but often multiple firm subcutaneous tumors and CALMs (Korf, 2015; Pulst, Riccardi, Fain, & Korenberg, 1991; Ruggieri et al., 2015). Although pathogenic NF1 missense and splicing variants are more commonly found in these individuals (Kluwe, Tatagiba, Fünsterer, & Mautner, 2003; Messiaen, Riccardi, & Peltonen, 2003; Upadhyaya et al., 2009), no specific recurrent variants associated with symptomatic spinal neurofibromas were identified, except recently for pathogenic missense variants affecting p.Gly848 (Koczkowska et al., 2018). In the current study, we identified p.Arg1276Gln being the second specific NF1 variant associated with a high prevalence of symptomatic spinal neurofibromas (11-fold higher than in the general NF1 population; Table 4), found in 55.6% of individuals greater than 19 years (15/27; Table S10), including 4/15 with so-called spinal form of NF1 (Table S14 and Figure S5). An additional 4/6 p.Arg1276Gln-positive adults who underwent routine MRI screening had asymptomatic spinal tumors, and such individuals may possibly be eligible for treatment with MEK inhibitors. Similar to p.Gly848-positive individuals (Koczkowska et al., 2018), fewer individuals ≥19 years developed cutaneous neurofibromas compared with “classic” cohorts (p < .0001, significant at FDR of 0.01 after B-H correction; Table 2). Cutaneous neurofibromas were observed in only one-third of p.Arg1276Gln-positive cases with spinal tumors (7/19), ranging in number from 2 to 99, while the majority of NF1-affected individuals ≥30 years old with “classic” NF1 exhibit greater than 100 neurofibromas (Mautner et al., 2008). Only EUR-R32, carrying a different pathogenic variant, p.Arg1276Pro, presented with greater than 100 cutaneous and subcutaneous lesions at age of 60 years (Table S14). Among 22 individuals with spinal tumors, 16 cases had normal development in line with previous findings (Korf, 2015).

NF1 p.Arg1276 and p.Lys1423 are highly conserved residues lying in a shallow pocket in the central catalytic domain forming the RAS-binding region with the p.Arg1276 residue also called the arginine “finger,” being the most essential catalytic element for RasGAP activity (Scheffzek, Ahmadian, & Wiesmüller, 1998). Functional studies have shown that pathogenic variants affecting these residues result in a dramatic reduction of GAP activity (Poullet, Lin, Esson, & Tamanoi, 1994; Scheffzek et al., 1998; Ahmadian, Kiel, Stege, & Scheffzek, 2003). Moreover, pathogenic missense variants at residue p.Lys1423 affects the stability of the neurofibromin/RAS complex by disrupting an intramolecular salt-bridge of the GRD (Ahmadian et al., 2003).

In this study, we report an additional NF1 variant, p.Met1149Val, associated specifically with a mild form of NF1 without externally visible plexiform or symptomatic spinal neurofibromas, symptomatic OPGs or malignant neoplasms similar to the phenotypes associated with variants at p.Arg1809 and p.Met992del (Koczkowska et al., 2019; Pinna et al., 2015; Rojnueangnit et al., 2015; Upadhyaya et al., 2007). As only six probands were identified so far with either the p.Met1149Ile or p.Met1149Thr (Table S7), even larger datasets, as well as follow-up of individuals enrolled in the current study, are required to further refine a mild phenotype associated with the other substitutions at p.Met1149. Therefore, we have now shown that approximately 2.5% of all unrelated NF1 pathogenic variant-positive probands of the UAB cohort develop a mild phenotype associated with nontruncating recurrent variants at one of three residues, p.Met992, p.Arg1809, and p.Met1149. Identification of this additional genotype–phenotype correlation will benefit the genetic counseling of these individuals and allow their stratification as they may not require stringent follow-up care. Moreover, as only 75% of p.Met1149-positive individuals fulfilled the NIH diagnostic criteria when family history was excluded as a criterion (Table S9), molecular analysis significantly facilitates the NF1 diagnostic process and distinguishes Legius syndrome (MIM# 611431) from the p.Met992del, p.Arg1809, and now p.Met1149 phenotypes with clinically overlapping features.

5 CONCLUSION

Although single case reports may suggest an association of p.Arg1276 with spinal neurofibromas (Korf, Henson, & Stemmer-Rachamimov, 2005; Upadhyaya et al., 2009) or p.Lys1423 with Noonan-like features (De Luca et al., 2005) (two unrelated individuals described with each phenotype), only large datasets of postpubertal individuals, carrying the same constitutional pathogenic variant and with the phenotype recorded in a standardized way allow the establishment of clinically relevant genotype–phenotype correlations. International multicenter studies involving close collaborations between NF1 clinicians and molecular geneticists are required to unfold such associations and to enhance correct variant interpretation. Our findings demonstrate genotype–phenotype correlations at the NF1 codons p.Met1149, p.Arg1276, and p.Lys1423. Although each of the reported NF1 variants associated with a specific clinical presentation (Kehrer-Sawatzki et al., 2017; Koczkowska et al., 2018; Koczkowska et al., 2019; Pinna et al., 2015; Rojnueangnit et al., 2015; Upadhyaya et al., 2007) affects only a small percentage of NF1 individuals, together, including the current results, they affect counseling and management of approximately 10% of the NF1 population. As the number of NF1 genotype–phenotype correlations continues to accumulate, genotype-driven personalized medicine will reach a turning point in NF1 by improving the disease surveillance and stratification of NF1-affected individuals.

CONFLICT OF INTERESTS

The authors declare that there are no conflict of interests.