Accurate Classification of NF1 Gene Variants in 84 Italian Patients with Neurofibromatosis Type 1

Abstract

Neurofibromatosis type 1 (NF1) is one of the most common autosomal dominant genetic diseases. It is caused by mutations in the NF1 gene encoding for the large protein, neurofibromin. Genetic testing of NF1 is cumbersome because 50% of cases are sporadic, and there are no mutation hot spots. In addition, the most recognizable NF1 clinical features—café-au-lait (CALs) spots and axillary and/or inguinal freckling—appear early in childhood but are rather non-specific. Thus, the identification of causative variants is extremely important for early diagnosis, especially in paediatric patients. Here, we aimed to identify the underlying genetic defects in 72 index patients referred to our centre for NF1. Causative mutations were identified in 58 subjects, with 29 being novel changes. We evaluated missense and non-canonical splicing mutations with both protein and splicing prediction algorithms. The ratio of splicing mutations detected was higher than that reported in recent patients’ series and in the Human Gene Mutation Database (HGMD). After applying in silico predictive tools to 41 previously reported missense variants, we demonstrated that 46.3% of these putatively missense mutations were forecasted to alter splicing instead. Our data suggest that mutations affecting splicing can be frequently underscored if not analysed in depth. We confirm that hamartomas can be useful for diagnosing NF1 in children. Lisch nodules and cutaneous neurofibromas were more frequent in patients with frameshifting mutations. In conclusion, we demonstrated that comprehensive in silico analysis can be a highly specific method for predicting the nature of NF1 mutations and may help in assuring proper patient care.

Keywords: NF1 gene; in silico analysis; neurofibromatosis type 1; splicing; variant classification.

Figure 1

Outline of the patients investigated in this study.

Figure 2

(A) Electrophoretogram of the complementary DNA (cDNA) generated from mutation c. 1640-7G>A. Below, the sequences of wild type and mutated cDNA and genomic DNA are presented. The inserted nucleotides are in red in the cDNA sequence. In the genomic sequences, the nucleotide substitution is in red, the nucleotides inserted in the cDNA are underlined. Small letters indicate the intronic sequence. (B) Predictions computed from the Alamut Visual splicing module. In the first row, the module’s splicing algorithms are presented; in the second row, the thresholds for the splicing site recognition for each individual algorithm are presented; in the remaining rows, all the putative splicing sites recognized by at least one of the algorithms of the Alamut Visual splicing module are presented. N indicates the natural splicing site. Green numbers indicate for each algorithm the threshold or range for splicing site recognition; red numbers are used when the computed scores predict splicing alteration; black numbers are used when the splicing site is recognized (above threshold or in the range) but no splicing alteration is forecasted. wt, wild type; mut, mutation.

Figure 3

Comparison of mutations type distribution across recent literature and with Human Gene Mutation Database (HGMD). Bianchessi et al. [14], Calì et al. [13], Sabbagh et al. [6].

Figure 4

Ratio of HGMD nonsense, pathogenic missense and possibly pathogenic missense variants ≤3 nucleotides from splicing sites predicted to either impact splicing (prediction positive) or have a neutral effect on splicing (prediction negative). In the bars are the absolute numbers for each group. Over the bars, p-values for pairwise comparisons, obtained according to Tukey’s method, are represented. In the top-right corner, the global p-values for the Fisher’s exact test and for the Chi-Square trend test are presented to compare greater proportions (i.e., more than two).

Figure 5

Correlation analysis of the cumulative average score from the five Alamut Visual^® splicing predictors with the MutationAssessor scores calculated for 26 HGMD missense mutations ≤3 nucleotides from the nearest splicing site. The data points for this graph are in Table S2.