Functional Assays Combined with Pre-mRNA-Splicing Analysis Improve Variant Classification and Diagnostics for Individuals with Neurofibromatosis Type 1 and Legius Syndrome

Abstract

Neurofibromatosis type 1 (NF1) and Legius syndrome (LS) are caused by inactivating variants in NF1 and SPRED1. NF1 encodes neurofibromin (NF), a GTPase-activating protein (GAP) for RAS that interacts with the SPRED1 product, Sprouty-related protein with an EVH (Ena/Vasp homology) domain 1 (SPRED1). Obtaining a clinical and molecular diagnosis of NF1 or LS can be challenging due to the phenotypic diversity, the size and complexity of the NF1 and SPRED1 loci, and uncertainty over the effects of some NF1 and SPRED1 variants on pre-mRNA splicing and/or protein expression and function. To improve NF1 and SPRED1 variant classification and establish pathogenicity for NF1 and SPRED1 variants identified in individuals with NF1 or LS, we analyzed patient RNA by RT-PCR and performed in vitro exon trap experiments and estimated NF and SPRED1 protein expression, RAS GAP activity, and interaction. We obtained evidence to support pathogenicity according to American College of Medical Genetics guidelines for 73/114 variants tested, demonstrating the utility of functional approaches for NF1 and SPRED1 variant classification and NF and LS diagnostics.

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.

1. Introduction

Detecting and classifying genetic variants are key to diagnosing genetic disorders. Clinical and genetic investigations, in silico predictions, and population data are routinely used in genome diagnostics to detect and establish pathogenicity of identified genetic variants. Nonetheless, the number of variants of unknown clinical significance (VUS) identified can equal the number of (likely) pathogenic variants, preventing a conclusive diagnosis and hindering genetic counselling [1]. Analysis of mRNA expression, pre-mRNA splicing, and protein function can help establish variant pathogenicity but are not routinely applied by diagnostic laboratories.

Neurofibromatosis type 1 (NF1; MIM# 162200) is an autosomal dominant disorder characterized by café-au-lait macules, Lisch nodules, axillary and inguinal freckling, cutaneous neurofibromas, and a wide range of patient specific symptoms [2–4]. NF1 has an incidence of ˜1/3500 and is caused by inactivation of the NF1 tumour suppressor [5]. Due to the relatively high incidence and large size of the NF1 gene, many VUS are identified preventing a final diagnosis and preventing optimal care of NF patients.

The canonical 12 kb NF1 mRNA transcript, NM_000267.3, encodes a 2818 amino acid (320 kDa) GTPase-activating protein (GAP) called neurofibromin (NF) that acts on GTPases of the RAS family. Loss or inactivation of NF1 results in increased RAS signaling and the development of lesions characteristic for NF1 [6]. Recent studies indicate that NF dimerizes in a head-to-tail orientation and that this dimerization is important for NF activation [7–10].

Legius syndrome (LS; MIM# 611431) is an autosomal dominant disorder characterized by cafe-au-lait macules, axillary and inguinal freckling, lipomas, and macrocephaly, learning disabilities, and developmental delay [11]. LS has an incidence of ˜1/75000 and is caused by inactivation of SPRED1 that encodes the Sprouty-related protein with an EVH (Ena/Vasp homology) domain 1 (SPRED1). SPRED1 recruits the NF dimer to the plasma membrane to stimulate the GTPase activity of membrane-bound RAS [8, 12]. The functional relationship between NF and SPRED1 helps explain the phenotypic overlap between NF1 and LS. Indeed, while some amino acid substitutions impair NF RAS GAP activity to cause NF1, other changes that do not affect RAS GAP activity cause NF1 by disrupting the interaction with SPRED1 [13]. Similarly, changes to SPRED1 disrupt the interaction with NF and cause LS [14–16]. In addition to direct effects on the catalytic GAP activity and NF-SPRED1 binding, the effects of amino acid changes on the expression and stability of the NF dimer are also critical for NF function [17].

Molecular genetic analysis can establish a diagnosis of NF1 or LS: the identification of an inactivating change, such as a frameshift or nonsense variant that causes a premature stop codon and/or nonsense-mediated mRNA decay provides strong evidence to support pathogenicity. Variants that affect pre-mRNA splicing or introduce damaging changes into the NF or SPRED1 proteins are more difficult to classify. In our center, DNA-based molecular screening identified 371 NF1 or SPRED1 VUS (accounting for approximately 10% of all cases) ([18]; unpublished data). The American College of Medical Genetics and Genomics (ACMG) has provided guidelines for the interpretation of genetic variants [19]. Strong evidence for classifying NF1 and SPRED1 variants as pathogenic can be obtained by performing functional experiments (ACMG criterium PS3) [14, 16, 17, 20–23]. To provide individuals from our NF1/LS cohort with certainty regarding their affection status and follow-up and to facilitate prenatal diagnostics, we initiated functional assessment of NF1 pre-mRNA splicing and NF-SPRED1 function and implemented these tests in our diagnostic laboratory. In addition, for cases where no candidate pathogenic variant was identified by DNA-based molecular screening, we applied RNA-sequencing to help identify variants that affect the NF1 or SPRED1 transcripts [24].

We tested 114 NF1 and SPRED1 variants. The effects of 38 variants on NF1 pre-mRNA splicing and 76 variants on NF-SPRED1 function were investigated. In 11 cases, both pre-mRNA splicing and NF1-SPRED1 function were analyzed. The combination of RNA and protein studies enabled us to fully investigate the likely effects of the different variants. For some variants, mRNA-splicing analysis was required to identify the correct protein variant to test in the functional assays. For others, the demonstration of abnormal NF1 pre-mRNA splicing made the testing of NF-SPRED1 protein function redundant. The results of the functional experiments, together with clinical and genetic data, were used to (re) classify the variants, following ACMG guidelines. Our integrated approach, combining testing of both pre-mRNA splicing and protein function in a routine NF1 diagnostic testing setting, allowed (re) classification of two-thirds of the variants tested as (likely) pathogenic.

2. Materials and Methods

2.1. Editorial Policies and Ethical Considerations

Informed consent was provided by all subjects, as required by the institutional review board of the Erasmus Medical Center, and according to standard diagnostic protocols.

2.2. Patient Assessment and Selection of Variants for Testing

The Erasmus MC Department of Clinical Genetics NF1/LS cohort consists of >4900 index cases suspected of NF1 or LS based on the international clinical diagnostic criteria [25] for whom DNA has been submitted for genetic testing of NF1 and/or SPRED1. Variants were classified using the available clinical and genetic data resulting in the identification of >2267 pathogenic or likely pathogenic NF1 or SPRED1 variants and >370 VUS ([18]; unpublished data). Variants were selected for functional testing following requests received from the consultant clinical geneticist.

Splice site prediction software (Alamut Visual Plus, version 1.5.1; Sophia Genetics) was used to identify variants likely to affect pre-mRNA splicing and the assay method was determined by the availability of patient RNA and/or the complexity of the predicted/observed splice abnormalities. In cases where a nonsynonymous variant was predicted to disrupt splicing, we first analyzed the putative effects on mRNA synthesis prior to deciding whether to investigate effects on protein function. For assay validation, additional variants, either from our own cohort or from literature, were selected for comparison, as detailed in Supplementary Tables S1-S3. We tested 11 variants that have been classified as pathogenic and/or subjected to functional evaluation: NM_000267.3(NF1) p.Leu90Pro [26], p.Met992del [17, 27], p.Met1149Val [17], p.Asp1217Tyr [14], p.Arg1276Gly [28], p.Lys1423Glu [17, 23, 29], p.Asp1623Gly [17], and p.Arg1809Cys [17]; NM_152594.2(SPRED1) p.Val44Asp, p.Thr102Met [14], and p.Ser105Ala [16]. In addition, we tested 3 likely benign variants from our cohort: p.Asn1229Ser, p.Pro1232Ser, and Ile1478Val. These variants were identified in individuals for whom another pathogenic, germ-line NF1 variant, was identified (data not shown). Nomenclature for all the reported variants is according to HGVS guidelines [30].

2.3. Constructs, Antibodies, and Cell-Lines

NF1 minigene exon trap constructs and NF expression plasmids were generated using standard cloning techniques [31], Gibson assembly [32], and/or site-directed mutagenesis (see Supplementary Materials for details). All constructs were verified by sequencing of the complete insert and at least 2 independent, verified clones per variant were used to prepare separate plasmid DNA stocks for the functional experiments. Nucleotide and amino acid numbering are according to reference transcripts NM_000267.3(NF1) and NM_152594.2(SPRED1), unless specified otherwise.

Antibodies were from Cell Signaling Technology (Danvers, U.S.A.) (rabbit anti-HA; mouse anti-HA; 9B11 mouse anti-myc), Invitrogen (mouse anti-V5), Sigma-Aldrich (St. Louis, U.S.A.) (mouse and rabbit anti-FLAG), and LI-COR Biosciences (Lincoln, U.S.A.) (goat anti-rabbit 680 nm and goat anti-mouse 800 nm conjugates). Anti-FLAG affinity beads were from Sigma-Aldrich; glutathione-sepharose was from GE Healthcare (Uppsala, Sweden).

HEK 293 T and COS-7 cells were maintained in Dulbecco’s Modified Eagle Medium (DMEM) (Lonza, Verviers, Belgium) containing 10% fetal calf serum, 50 U/ml penicillin and 50 μg/ml streptomycin in a humidified 37°C, 5-10% CO₂ incubator.

2.4. Assessment of the Effects of NF1 Variants on NF1 Pre-mRNA Splicing in Patient Material

Reverse transcriptase (RT) PCR was performed on 1-2 μg total RNA as described in the Supplementary Materials.

2.5. In Vitro Assessment of the Effects of NF1 Variants on NF1 Pre-mRNA Splicing

Exon trap experiments were performed as described previously [33, 34]. See Supplementary Materials for details.

2.6. In Vitro Assessment of RAS GAP Activity

To estimate RAS GAP activity, HA-H-RAS was expressed together with either wild-type (WT) or variant NF in COS-7 or HEK 293 T cells. GTP-bound RAS was subsequently isolated using glutathione-agarose beads coated with recombinant GST-RAF-RBD [35]. See Supplementary Materials for details.

2.7. In Vitro Assessment of the NF1-SPRED Interaction

To investigate the NF-SPRED interaction, FLAG-SPRED1 (WT or variant) and NF (WT or variant) were coexpressed in HEK 293 T cells. NF-SPRED1 complexes were immunoprecipitated using anti-FLAG affinity beads (Sigma-Aldrich). See Supplementary Materials for details.

3. Results

3.1. Assessment of the Effects of NF1 Variants on NF1 Pre-mRNA Splicing

We investigated the effects of 38 NF1 variants on pre-mRNA splicing (Figure 1 and Supplementary Materials, Table S1). For 30 variants, we performed in vitro exon trapping (Figure 1(a)). Subject RNA was available for testing of 15 of these variants, allowing confirmation of the exon trap results (Figure 1(b)). For 8 additional variants, RT-PCR and Sanger sequencing of subject RNA were carried out directly, without performing exon trap experiments (Figure 1(b)). We compared the variant exon trap constructs with the corresponding wild-type (WT) NF1 exon (Figure 1(a)). The WT construct usually revealed a single, predominant product corresponding to the expected trapped exon (Figures 1(c) and 1(d)); although for a few exons, a minor, less intense, RT-PCR product corresponding to skipping of the WT exon was observed. The variant constructs showed either no difference, an abnormal splice product, or a combination of different products (Figures 1(c) and 1(d)).

[figure(s) omitted; refer to PDF]

We detected one or more abnormal NF1 splice products, either in vitro, in subject RNA, or in both for 25/38 variants (66%). Exon skipping (type I defect) [36] was observed for 20 variants; but pseudoexon incorporation (type II defects), exon truncation due to utilization of a noncanonical splice site (type III defects), and intron retention (type IV defects) were also observed (Supplementary Materials, Table S1; Figure 1(c)). In 5 cases, abnormal splicing resulted in an in-frame deletion, including the nonsynonymous c.2710A>T p.(Cys904Ser) variant. To determine whether the NF1 c.288+3 A>T, r.205_288del, p.(Arg69_Gly96del) and c.2710A>T, r.2707_2850del, p.(Cys904_Val951del) variants affected NF activity, protein function assessment was performed. In 7 cases, RNA analysis indicated that a missense change prevented canonical NF1 pre-mRNA splicing, making assessment of NF function redundant. We did not observe an effect on NF1 pre-mRNA splicing, either in vitro or in subject RNA for 13 variants. In total, analysis of NF1 pre-mRNA splicing assisted in the classification of 25/38 variants (66%) as likely pathogenic; the remaining variants were subsequently confirmed as likely benign (3 cases; data not shown), affected protein function (5 cases; see below), or remained VUS (4 cases) (Supplementary Materials, Table S1).

3.2. In Vitro Assessment of the Effects of NF1 Variants on NF RAS GAP Activity

The ability of the NF GAP-related domain (GRD; amino acids 1180-1504) to inactivate RAS can be determined by measuring the amount of active, GTP-bound RAS in the presence of the NF GRD [23]. To assess the pathogenicity of 2 variants that were not predicted to affect splicing, we determined the RAS GAP activity of the NF GRD using a pull-down assay for GTP-bound RAS (Figure 2(a)). We identified the NF1 c.3829G>C, p.(Gly1277Arg) and c.3651T>A, p.(Asp1217Glu) variants in the NF GRD in 2 individuals with NF1 and introduced both variants and the pathogenic NF1 c.3826C>G, p.(Arg1276Gly) variant into a NF V5-p.1180_1504 expression construct [23] (Figure 2(b)). The p.Arg1276Gly and p.Gly1277Arg variants lacked RAS GAP activity, as estimated from the levels of GTP-bound RAS (RAS-GTP) in the pull-down fraction (Figure 2(c)), supporting likely pathogenicity of the NF1 c.3829G>C, p.(Gly1277Arg) but not c.3651T>A, p.(Asp1217Glu) substitution.

[figure(s) omitted; refer to PDF]

Expression levels of the wild-type and variant NF V5-p.1180_1504 proteins were low (data not shown). To enhance NF GRD expression and detection, we modified the NF V5-p.1180_1504 construct by altering the sequence preceding the initiation codon to correspond to the Kozak consensus and by introducing a C-terminal V5-epitope tag. We observed robust expression of the resulting WT NF V5-p.1180_1504-V5 protein and therefore derived 12 NF1 variants identified in our NF1 cohort, including p.(Gly1277Arg) and p.(Asp1217Glu), in the NF V5-p.1180_1504-V5 expression construct (Figure 2(b)). We determined the RAS GAP activity of the expressed NF V5-p.1180_1504-V5 proteins with the pull-down assay (Figures 2(d)–2(f)). In 5 cases, the variant lacked RAS GAP activity: levels of GTP-bound RAS (RAS-GTP) were not significantly different to those in the absence of NF V5-p.1180_1504-V5 ($P>0.025$; Student’s paired $t$-test with the Bonferroni correction; shown as red bars in Figure 2(d)). In 2 cases, RAS GAP activity was significantly reduced compared to WT NF V5-p.1180_1504-V5, but significantly increased compared to the absence of NF V5-p.1180_1504-V5 ($P<0.025$; Student’s paired $t$-test with the Bonferroni correction; orange bars in Figure 2(d)), suggesting that these variants impaired RAS GAP activity but did not inactivate the GRD completely. The remaining 5 variants retained full RAS GAP activity ($P>0.025$; Student’s paired $t$-test with the Bonferroni correction; black bars in Figure 2(d)).

3.3. In Vitro Assessment of the Effects of NF1 Variants on NF P.1_2069 RAS GAP Activity

Many NF1 VUS identified in our cohort mapped outside the NF GRD. Attempts to introduce nucleotide changes into a full-length NF1 expression construct were unsuccessful. However, we were able to introduce variants into 2 expression constructs encoding the N-terminal 2069 amino acids of NF (Figure 3(a)). The only difference between these 2 WT constructs was the inclusion of sequences corresponding to a neuron-specific NF1 transcript encoding a 10 amino acid insertion (NM_000267.3(NF1) p.420insSerThrPheLysHisGlyLeuGlyThrAla; [37]. We referred to the proteins expressed from these constructs as NF p.2069myc and NF p.420ins10myc, respectively. Some initial experiments were conducted using the WT NF p.420ins10myc construct. However, most variants were derived from the WT NF p.2069myc construct as the encoded protein corresponded better with the product of the NM_000267.3 reference transcript. We did not detect significant differences in RAS GAP activity between the WT p.2069myc and p.420ins10myc proteins (Figures 3(b) and 3(c)). To further validate the assay, we investigated the correlation between NF expression levels and the estimated RAS GAP activity (Supplementary Figure S1A-C). Consistent with previous studies [17, 38], the NF signal and the estimated RAS GAP activity were dependent on the amount of NF expression construct used in the transfection experiments. Under the conditions used to compare the WT and variant NF proteins, the expression of WT NF was sufficient to increase RAS GAP activity >5-fold compared to cells not expressing any exogenous NF (Supplementary Figure S1B).

[figure(s) omitted; refer to PDF]

We introduced 69 NF1 variants in the WT NF expression constructs, including 12 previously tested in the NF V5-p.1180_1504 or NF V5-p.1180_1594-V5 constructs, and determined the RAS GAP activity of the variant proteins (Figure 3 and Supplementary Materials, Table S2). Some were as active as the corresponding WT NF protein, some had severely attenuated RAS GAP activity, and others had intermediate levels of activity. This made it difficult to assign an exact cut-off value to identify pathogenic, inactivating variants. Therefore, we devised an empirical scheme to categorize the variants (Supplementary Figure S1B). We compared the mean RAS GAP activities of the variants to WT NF. If the mean RAS GAP activity was <50% of the WT ($P<0.05$, Student’s paired $t$-test), then we considered it evidence for disruption of NF RAS GAP activity and supporting evidence for pathogenicity (ACMG criteria PS3). This was the case for 22 variants (Figure 3(b); variants shown as red bars). Of these, 17 (77%) mapped to the GRD (Figure 3(b)), confirming the importance of this region for NF RAS GAP activity. With one exception, the results with the NF p.1180_1504-V5 and V5-p.1180_1504-V5 proteins were consistent. We did not detect significant impairment of RAS GAP activity by the NF p.420ins10myc p.Thr1199Ile variant, in contrast to the reduction associated with the NF V5-p.1180_1504-V5 p.Thr1199Ile variant (compare Figures 2(d) and 3(b) and see Discussion section).

In 16 cases, RAS GAP activity was significantly reduced compared to WT NF ($P>0.05$; Student’s paired $t$-test), but was >50% of the WT value (Figure 3(b); orange bars). We did not consider this sufficient evidence to support pathogenicity. The remaining variants did not show evidence for impaired RAS GAP activity: mean activity was not significantly different to WT NF ($P>0.05$; Student’s paired $t$-test; Figure 3(b), black bars).

3.4. In Vitro Assessment of the Effects of NF1 and SPRED1 Variants on the NF-SPRED1 Interaction

Some pathogenic NF1 variants disrupt the interaction between NF and SPRED1 without affecting RAS GAP activity [13]. We used an anti-FLAG affinity matrix to coimmunoprecipitate (coIP) the WT NF p.2069myc and p.420ins10myc proteins together with coexpressed WT FLAG-SPRED1 (Figure 4(a)) and determined whether 67 NF1 (Figures 4(b) and 4(d) and Supplementary Table S2) and 5 SPRED1 variants (Figure 4(e), left; Supplementary Table S3) affected NF-SPRED1 coIP. We compared the WT and variant signals in the IP fractions (Figures 4(b) and 4(e), left) and categorized the variants using the same criteria as for the RAS GAP assay: a significant reduction ($P<0.05$, Student’s paired $t$-test) of >50% in the mean NF signal in the IP fraction (NF coIP) was evidence for an effect on the NF-SPRED1 interaction, supporting pathogenicity (Supplementary Figure S1C). We did not detect significant differences in expression or NF-SPRED1 coIP between the WT p.2069myc and p.420ins10myc proteins (Figures 4(b) and 4(c)). To validate the NF coIP assay, we determined the effect of NF expression levels on the NF signals in the IP fraction (Supplementary Materials, Figure S1D-F). The NF coIP signal correlated with the NF signal in the cell lysate and was dependent on the amount of transfected NF expression construct.

[figure(s) omitted; refer to PDF]

NF coIP was reduced >50% for 29 NF1 and 2 SPRED1 variants, including the NF1 p.Asp1217Tyr and SPRED1 p.Val44Asp variants previously shown to disrupt the NF-SPRED1 interaction [14] (Figures 4(b) and 4(e), red bars). There was a significant reduction in the NF coIP signal for an additional 11 NF1 variants, but the mean value was >50% of WT NF (Figure 4(b), orange bars) and we did not consider this sufficient evidence to support pathogenicity. The remaining variants did not differ from WT NF ($P>0.05$, Student’s paired $t$-test; Figures 4(b) and 4(e), left, black bars). We did not test 5 NF1 variants that had clearly disrupted either RAS GAP activity or NF1 pre-mRNA splicing in earlier assays (Supplementary Table S2).

3.5. In Vitro Assessment of the Effects of NF1 and SPRED1 Variants on the Expression and Stability of NF and SPRED1

Differences in RAS GAP activity and NF coIP could reflect differences in the expression and/or stability of the variant proteins. To determine the effects of NF1 and SPRED1 variants on NF-SPRED1 expression and stability, we compared WT and variant NF and SPRED1 signals in the cell lysates by immunoblotting (Figures 3(c) and 4(c)–4(e), right). In addition, we determined the effect of WT NF levels on both RAS GAP activity and NF coIP (Supplementary Figure S1). We compared the resulting titration curves to the expression, RAS GAP activity, and NF coIP of the NF1 variants, as estimated under standard assay conditions (Figure 5). Notably, some variants were expressed at levels equal to or above WT, yet were still unable to inactivate RAS or were not immunoprecipitated efficiently with SPRED1.

[figure(s) omitted; refer to PDF]

Mean NF expression was reduced by >50% for 15 NF1 variants (Figure 4(c), red bars). Of these, either RAS GAP activity, NF coIP, or both was reduced by >50% in 9 cases. Both RAS GAP activity and NF coIP were significantly reduced in 2 additional cases, but by <50% (see Discussion); in 2 cases, RAS GAP activity was reduced but by <50%, and in 2 cases, neither RAS GAP activity or NF coIP was significantly different to WT NF (compare Figure 4(c) with Figures 3(b) and 4(b) and see Supplementary Table S2).

4. Discussion

Although the diagnostic yield for NF1 and LS is >90%, the high incidence of NF1 means that many patients still lack a molecular diagnosis because either no candidate pathogenic variant or a VUS in NF1/SPRED1 is found. To complement the NF1 and SPRED1 DNA test results from our laboratory, we implemented functional assays to assess 114 NF1/SPRED1 variants in a diagnostic setting. We employed 4 functional assays: (i) analysis of subject mRNA by RT-PCR; (ii) in vitro exon trap analysis of NF1 pre-mRNA splicing; (iii) in vitro analysis of NF RAS GAP activity; and (iv) in vitro analysis of NF-SPRED1 expression and interaction. We considered the evidence sufficient for reclassification of 73/114 (64%) variants as (likely) pathogenic (class 4 and 5) (Supplementary Materials, Tables S1, S2, and S3), demonstrating the utility of functional approaches for NF1 and SPRED1 variant classification and NF1 and LS diagnostics. The results of our experiments have been submitted to the NF1 and SPRED1 Leiden Open Variation Databases (https://databases.lovd.nl/shared/genes/NF1;https://databases.lovd.nl/shared/genes/SPRED1).

In contrast to laboratories that specialize in NF1 variant detection and classification using patient RNA [21, 39], our diagnostic laboratory performs molecular screening primarily on DNA samples because direct analysis of RNA was not considered practical for routine screening in our setting [18]. The in vitro exon trap experiments provided a useful screen for identifying NF1 variants likely to affect splicing, without having to resample patients. We did not observe major discrepancies between the exon trap and RT-PCR results that would have led to a different classification for any of the variants tested, consistent with other work from our laboratory [24, 40]. Furthermore, the exon trap analysis meant that the observed in vitro effects of a potentially pathogenic variant could be communicated prior to taking a tissue sample for confirmation. The exon trap approach also assisted in resolving allele-specific patterns of pre-mRNA splicing when phasing was not possible due to a lack of informative exonic variants. For example, expression of the WT allele sometimes prevented us from determining whether the canonical NF1 transcript was also expressed from the variant allele. The exon trap experiments indicated whether a variant was likely to completely prevent canonical splicing or only have a partial effect. In 4 cases, there were minor differences between the in vitro and in vivo RNA data (Supplementary Materials, Table S1). However, we did not identify cases where a variant had a major effect on splicing in vitro but not in vivo or vice versa. Analysis of pre-mRNA splicing was also a useful screen for the functional assessments as it was not always obvious whether a variant was likely to affect splicing and/or protein function. In some cases, RNA analysis revealed abnormal NF1 splicing, making functional assessment redundant, whereas in other cases, RNA analysis indicated that functional assessment of an in-frame deletion was indicated to establish pathogenicity. In 25/38 cases (66%), functional assessment of NF1 pre-mRNA splicing provided sufficient evidence for us to classify the variant as (likely) pathogenic (class 4 and 5; ACMG criteria PS3).

Compared to the exon trapping and RT-PCR experiments, assessment of NF-SPRED1 function was labour-intensive, time-consuming, and had other limitations (see below), meaning that the findings had to be interpreted with caution and in the light of clinical and genetic evidence. Nonetheless, we obtained functional evidence to support pathogenicity for 46 NF1 and 2 SPRED1 variants, including the known pathogenic variants NF1 p.(Leu90Pro), p.(Met992del), p.(Asp1217Tyr), p.(Arg1276Gly), p.(Lys1423Glu), p.(Asp1623Gly) and p.(Arg1809Cys), and SPRED1 p.(Val44Asp) (Supplementary Materials, Tables S2 and S3). Our data showing loss of RAS GAP activity for the NF1 p.Lys1423Glu variant and disruption of the NF-SPRED1 interaction for the NF1 p.Asp1217Tyr and SPRED1 p.Val44Asp variants were consistent with previous reports [14, 17, 23].

In contrast to a recent study that analyzed full-length murine Nf1 variants [17], our expressed NF proteins lacked a segment of the C-terminal HEAT-repeat region that is involved in NF dimerization [7–10]. Nonetheless, robust, reproducible effects on RAS GAP activity, NF coIP, and/or NF expression/stability were observed, even though the variants were expressed at nonphysiological levels (Figures 3 and 4 and Supplementary Materials, Tables S2 and S3). Some differences in the estimated activity or expression might reflect variation in transfection efficiency, cell numbers, immunoblotting artefacts, or other processing errors, and it is possible that some variants that disrupted NF-SPRED1 function in our in vitro assays might retain sufficient activity in vivo to prevent NF1 or LS. With these caveats in mind, we devised an empirical scheme to categorize the variants. We considered a >50% reduction in either RAS GAP activity or NF coIP as functional evidence to support pathogenicity (Supplementary Figure S1). We did not consider a >50% reduction in expression/stability as sufficient evidence for pathogenicity unless it was concordant with significant disruption of RAS GAP activity and/or NF coIP ($P<0.05$, Student’s paired $t$-test) (Supplementary Materials, Tables S2 and S3). Variants that did not significantly reduce RAS GAP activity or NF coIP remained VUS, unless other genetic evidence was obtained to support or exclude pathogenicity, such as de novo occurrence of the variant in a sporadic case of NF1 or identification of another pathogenic variant in the same individual.

None of the variants for which we obtained evidence to support pathogenicity were identified more than once in the gnomAD (v2.1) database (https://gnomad.broadinstitute.org/) (accessed7/3/2022), and none were classified as benign or likely benign in ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/) (accessed 7/3/2022). The variants were all identified in at least one individual suspected of NF1 or LS in our cohort. The remaining variants did not show sufficient evidence for an effect on NF or SPRED1 function to support pathogenicity, even though several are described as likely pathogenic in ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/; Supplementary Materials, Table S2). In one case, we observed a discrepancy between the results of the RAS GAP assay with the NF V5-p.1180_1504-V5 GRD and NF p.420ins10myc protein. The NF1 p.Thr1199Ile variant impaired RAS GAP activity of the GRD but did not significantly affect RAS GAP activity of the NF p.420ins10 protein (compare Figures 2(d) and 3(b)). It is possible that the NF GRD and p.420ins10myc proteins have distinct sensitivities to changes in secondary structure. The extra scaffolding around the active site of the GRD provided by the p.420ins10myc protein might restrict structural changes and thereby help maintain RAS GAP activity. For this reason, and because the larger NF p.2069myc and p.420ins10myc proteins could also be used to not only investigate the effects of more variants but also interrogate the NF-SPRED1 interaction, we currently derive NF p.2069myc variants for functional assessment.

We observed good correlation between the variants affecting either RAS GAP activity or NF coIP and their location within or adjacent to the NF GRD and SPRED1 interaction domains (Figures 3(a) and 3(b) and 4(b)). Interestingly, we identified 8 NF1 variants just distal of the second, C-terminal SPRED1 interaction domain that clearly disrupted NF coIP (Figure 4(b); amino acids 1522-1809). It is possible that this region of NF is important for maintaining the correct spatial orientation of the SPRED1 interaction domains. Intriguingly, we also identified a cluster of NF1 variants, p.His1821Asp, Asp1828Asn, Asp1828Val, and Asp1828Tyr, with increased RAS GAP activity (~150%) compared to WT NF (Figure 3(b) and Supplementary Table S2). More experiments are required to investigate the significance of this finding.

In contrast to an earlier study [13], we identified variants that disrupted both RAS GAP activity and the NF-SPRED1 interaction (Supplementary Table S2; compare Figures 3(b) and 4(b)). Notably, several of these variants, NF1 p.Thr780Lys, p.Leu995Pro, p.Gly1219Arg, p.Leu1221Arg, and p.Leu1246Pro, did not reduce NF expression (Figure 4(c) and Supplementary Table S2), suggesting that they affect NF function without destabilising the protein.

Variants classified as likely benign according to ACMG criteria did not show sufficient evidence to support pathogenicity in our functional assays (Supplementary Table S2). However, we did observe differences for variants previously classified as (likely) pathogenic [17]. According to our criteria, the p. Met992del, p.Met1149Val, and p.Arg1809Cys variants did not reduce NF RAS GAP activity sufficiently to support pathogenicity. Nonetheless, NF coIP was clearly reduced for the p.Met992del and p.Arg1809Cys variants, supporting pathogenicity (the NF-SPRED1 interaction was not assessed in [17]). The absence of the C-terminal region of NF (amino acids 2070-2818) from our expressed NF variants as well as differences between murine and human NF could account for the differences between the two studies, but we note that all 3 variants have been associated with a distinct NF1 phenotype [27, 29, 41], and it is possible that other variants with (partial) retention of NF RAS GAP activity might be associated with less severe disease. Advances in the structural and functional biology of NF will help inform decisions regarding the pathogenicity of these and other variants. Furthermore, functional analysis of a larger number of known pathogenic NF1 and SPRED1 variants could help establish more accurate criteria for determining likely pathogenicity and identify correlations between specific deficits in NF function and the clinical phenotype.

We did not use the results of the functional assessment to exclude pathogenicity. We only interrogated 3 aspects of NF-SPRED1 function: RAS GAP activity, the NF-SPRED1 interaction, and expression/stability. We did not investigate other putative functions of NF or SPRED1, such as phospholipid binding [42, 43], cell invasiveness [44], or the regulation of estrogen receptor dependent transcriptional activity [45]. Furthermore, we were unable to investigate NF1 variants distal to residue 2069. Efforts to efficiently derive NF1 variants in a full-length NF1 expression construct are on-going in our laboratory.

Despite the limitations detailed above, our work enabled a molecular diagnosis to be made for individuals suspected of NF1 and LS in whom a VUS in NF1 or SPRED1 had been identified. We obtained evidence to support pathogenicity for 73/114 variants (64%) (Supplementary Materials, Tables S1, S2, and S3) and together with consideration of the clinical, population, in silico, and segregation data, functional testing helped establish likely variant pathogenicity in these cases. Implementation of functional testing in our laboratory has improved molecular diagnostics for individuals with NF1 and LS (Supplementary Figure S1) and facilitated appropriate monitoring, treatment, and prenatal diagnostic options for family planning. Our approach shows that the integration of assays for NF1/SPRED1 pre-mRNA splicing and protein function allows reclassification of a significant proportion of NF1 and SPRED1 VUS, drastically improving molecular diagnostics for individuals and families with NF1 and LS. Furthermore, our study demonstrates that diagnostic laboratories proficient in protein-based analyses such as immunoblotting and immunoprecipitation can apply similar functional approaches for variant classification, not only for NF1 and LS but also for other specific genetic conditions.

Disclosure

A preprint of this manuscript has previously been published [46] (https://yvm2020.authorea.com/doi/full/10.22541/au.165217208.81458305/v1).

Authors’ Contributions

Conceptualization was carried out by MN, JJS, TvH, and RvM; experimentation was carried out by HD, MH-W, MN, JL, MK-dH, MP, BvO, LvU, PE, and EK; clinical investigation was carried out by YvB, MvV, RO, AW, and YvI; supervision was carried out by MH-W, MN, AW, TvH, and RvM; writing, review, and editing were carried out by MN, AW, YvI, TvH, and RvM. Hannie Douben, Marianne Hoogeveen-Westerveld, Tjakko van Ham, and Rick van Minkelen contributed equally to this work.

References

[1] G. Federici, S. Soddu, "Variants of uncertain significance in the era of high-throughput genome sequencing: a lesson from breast and ovary cancers," Journal of Experimental & Clinical Cancer Research, vol. 39 no. 1,DOI: 10.1186/s13046-020-01554-6, 2020.

[2] S. Peltonen, M. Pöyhönen, "Clinical diagnosis and atypical forms of NF1," Neurofibromatosis Type 1. Molecular and Cellular Biology, pp. 17-30, DOI: 10.1007/978-3-642-32864-0_2, 2012.

[3] K. Jett, J. M. Friedman, "Clinical and genetic aspects of neurofibromatosis 1," Genetics in Medicine, vol. 12 no. 1,DOI: 10.1097/GIM.0b013e3181bf15e3, 2010.

[4] K. I. Ly, J. O. Blakeley, "The diagnosis and management of neurofibromatosis type 1," The Medical Clinics of North America, vol. 103 no. 6, pp. 1035-1054, DOI: 10.1016/j.mcna.2019.07.004, 2019.

[5] N. Ratner, S. J. Miller, "A RASopathy gene commonly mutated in cancer: the neurofibromatosis type 1 tumour suppressor," Nature Reviews Cancer, vol. 15 no. 5, pp. 290-301, DOI: 10.1038/nrc3911, 2015.

[6] K. Cichowski, T. Jacks, "NF1 tumor suppressor gene function: narrowing the GAP," Cell, vol. 104 no. 4, pp. 593-604, DOI: 10.1016/S0092-8674(01)00245-8, 2001.

[7] M. Sherekar, S.-W. Han, R. Ghirlando, S. Messing, M. Drew, D. Rabara, T. Waybright, P. Juneja, H. O'Neill, C. B. Stanley, D. Bhowmik, A. Ramanathan, S. Subramaniam, D. V. Nissley, W. Gillette, F. McCormick, D. Esposito, "Biochemical and structural analyses reveal that the tumor suppressor neurofibromin (NF1) forms a high-affinity dimer," The Journal of Biological Chemistry, vol. 295 no. 1105-1119,DOI: 10.1074/jbc.RA119.010934, 2019.

[8] C. J. Lupton, C. Bayly-Jones, L. D'Andrea, C. Huang, R. B. Schittenhelm, H. Venugopal, J. C. Whisstock, M. L. Halls, A. M. Ellisdon, "The cryo-EM structure of the human neurofibromin dimer reveals the molecular basis for neurofibromatosis type 1," Nature Structural & Molecular Biology, vol. 28 no. 12, pp. 982-988, DOI: 10.1038/s41594-021-00687-2, 2021.

[9] A. Naschberger, R. Baradaran, B. Rupp, M. Carroni, "The structure of neurofibromin isoform 2 reveals different functional states," Nature, vol. 599 no. 7884, pp. 315-319, DOI: 10.1038/s41586-021-04024-x, 2021.

[10] M. Chaker-Margot, S. Werten, T. Dunzendorfer-Matt, S. Lechner, A. Ruepp, K. Scheffzek, T. Maier, "Structural basis of activation of the tumor suppressor protein neurofibromin," Molecular Cell, vol. 82 no. 7, pp. 1288-1296.e5, DOI: 10.1016/j.molcel.2022.03.011, 2022.

[11] H. Brems, M. Chmara, M. Sahbatou, E. Denayer, K. Taniguchi, R. Kato, R. Somers, L. Messiaen, S. De Schepper, J. P. Fryns, J. Cools, P. Marynen, G. Thomas, A. Yoshimura, E. Leqius, "Germline loss-of-function mutations in SPRED 1 cause a neurofibromatosis 1 -like phenotype," Nature Genetics, vol. 39 no. 9, pp. 1120-1126, DOI: 10.1038/ng2113, 2007.

[12] I. B. Stowe, E. L. Mercado, T. R. Stowe, E. L. Bell, J. A. Oses-Prieto, H. Hernandez, A. L. Burlingame, F. McCormick, "A shared molecular mechanism underlies the human rasopathies Legius syndrome and neurofibromatosis-1," Genes & Development, vol. 26 no. 13, pp. 1421-1426, DOI: 10.1101/gad.190876.112, 2012.

[13] T. Dunzendorfer-Matt, E. L. Mercado, K. Maly, F. McCormick, K. Scheffzek, "The neurofibromin recruitment factor Spred1 binds to the GAP related domain without affecting Ras inactivation," Proceedings of the National Academy of Sciences, vol. 113 no. 27, pp. 7497-7502, DOI: 10.1073/pnas.1607298113, 2016.

[14] Y. Hirata, H. Brems, M. Suzuki, M. Kanamori, M. Okada, R. Morita, I. Llano-Rivas, T. Ose, L. Messiaen, E. Legius, A. Yoshimura, "Interaction between a domain of the negative regulator of the Ras-ERK pathway, SPRED1 protein, and the GTPase-activating protein-related domain of neurofibromin is implicated in Legius syndrome and neurofibromatosis type," The Journal of Biological Chemistry, vol. 291 no. 7, pp. 3124-3134, DOI: 10.1074/jbc.M115.703710, 2016.

[15] S. Führer, M. Tollinger, T. Dunzendorfer-Matt, "Pathogenic mutations associated with Legius syndrome modify the Spred1 surface and are involved in direct binding to the Ras inactivator neurofibromin," Journal of Molecular Biology, vol. 431 no. 19, pp. 3889-3899, DOI: 10.1016/j.jmb.2019.07.038, 2019.

[16] W. Yan, E. Markegard, S. Dharmaiah, A. Urisman, M. Drew, D. Esposito, K. Scheffzek, D. V. Nissley, F. McCormick, D. K. Simanshu, "Structural insights into the SPRED1-neurofibromin-KRAS complex and disruption of SPRED1-neurofibromin interaction by oncogenic EGFR," Cell Reports, vol. 32 no. 3, article 107909,DOI: 10.1016/j.celrep.2020.107909, 2020.

[17] A. Long, H. Liu, J. Liu, M. Daniel, D. M. Bedwell, B. Korf, R. A. Kesterson, D. Wallis, "Analysis of patient-specific NF1 variants leads to functional insights for Ras signaling that can impact personalized medicine," Human Mutation, vol. 43 no. 1, pp. 30-41, DOI: 10.1002/humu.24290, 2022.

[18] R. van Minkelen, Y. van Bever, J. N. Kromosoeto, C. J. Withagen-Hermans, A. Nieuwlaat, D. J. Halley, A. M. van den Ouweland, "A clinical and genetic overview of 18 years neurofibromatosis type 1 molecular diagnostics in the Netherlands," Clinical Genetics, vol. 85 no. 4, pp. 318-327, DOI: 10.1111/cge.12187, 2014.

[19] S. Richards, N. Aziz, S. Bale, D. Bick, S. Das, J. Gastier-Foster, W. W. Grody, M. Hegde, E. Lyon, E. Spector, K. Voelkerding, H. L. Rehm, "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, DOI: 10.1038/gim.2015.30, 2015.

[20] A. Zatkova, L. Messiaen, I. Vandenbroucke, R. Wieser, C. Fonatsch, A. R. Krainer, K. Wimmer, "Disruption of exonic splicing enhancer elements is the principal cause of exon skipping associated with seven nonsense or missense alleles of NF1," Human Mutation, vol. 24 no. 6, pp. 491-501, DOI: 10.1002/humu.20103, 2004.

[21] E. Ars, E. Serra, J. Garcia, H. Kruyer, A. Gaona, C. Lazaro, X. Estivill, "Mutations affecting mRNA splicing are the most common molecular defects in patients with neurofibromatosis type 1," Human Molecular Genetics, vol. 9 no. 2, pp. 237-247, DOI: 10.1093/hmg/9.2.237, 2000.

[22] L. M. Messiaen, T. Callens, G. Mortier, D. Beysen, I. Vandenbroucke, N. Van Roy, F. Speleman, A. D. Paepe, "Exhaustive mutation analysis of the NF1 gene allows identification of 95% of mutations and reveals a high frequency of unusual splicing defects," Human Mutation, vol. 15 no. 6, pp. 541-555, DOI: 10.1002/1098-1004(200006)15:6<541::AID-HUMU6>3.0.CO;2-N, 2000.

[23] L. Thomas, M. Richards, M. Mort, E. Dunlop, D. N. Cooper, M. Upadhyaya, "Assessment of the potential pathogenicity of missense mutations identified in the GTPase-activating protein (GAP)-related domain of the neurofibromatosis type-1 ( NF1 ) gene," Human Mutation, vol. 33 no. 12, pp. 1687-1696, DOI: 10.1002/humu.22162, 2012.

[24] J. C. W. Douben, M. Nellist, L. van Unen, P. Elfferich, E. Kasteleijn, M. Hoogeveen-Westerveld, J. Louwen, M. van Veghel-Plandsoen, W. de Valk, J. J. Saris, F. Hendriks, E. Korpershoek, E. H. Hoefsloot, M. van Vliet, Y. van Bever, I. van de Laar, E. Aten, A. M. A. Lachmeijer, W. Taal, L. van den Bersselaar, J. Schuurmans, R. Oostenbrink, R. van Minkelen, Y. van Ierland, T. J. van Ham, "High-yield identification of pathogenic NF 1 variants by skin fibroblast transcriptome screening after apparently normal diagnostic DNA testing," Human Mutation, vol. 43 no. 12, pp. 2130-2140, DOI: 10.1002/humu.24487, 2022.

[25] E. Legius, L. Messiaen, P. Wolkenstein, P. Pancza, R. A. Avery, Y. Berman, J. Blakeley, D. Babovic-Vuksanovic, K. Soares Cunha, R. Ferner, M. J. Fisher, J. M. Friedman, D. H. Gutmann, H. Kehrer-Sawatzki, B. R. Korf, V.-F. Mautner, S. Peltonen, K. A. Rauen, V. Riccardi, E. Schorry, A. Stemmer-Rachamimov, D. A. Stevenson, G. Tadini, N. J. Ullrich, D. Viskochil, K. Wimmer, K. Yohay, International Consensus Group on Neurofibromatosis Diagnostic Criteria, S. M. Huson, D. G. Evans, S. R. Plotkin, "Revised diagnostic criteria for neurofibromatosis type 1 and Legius syndrome: an international consensus recommendation," Genetics in Medicine, vol. 23 no. 8, pp. 1506-1513, DOI: 10.1038/s41436-021-01170-5, 2021.

[26] H. Xiao, L. Yuan, H. Xu, Z. Yang, F. Huang, Z. Song, Y. Yang, C. Zeng, H. Deng, "Novel and recurring disease-causing NF1 variants in two Chinese families with neurofibromatosis type 1," Journal of Molecular Neuroscience, vol. 65 no. 4, pp. 557-563, DOI: 10.1007/s12031-018-1128-9, 2018.

[27] M. Koczkowska, T. Callens, Y. Chen, A. Gomes, A. D. Hicks, A. Sharp, E. Johns, K. A. Uhas, L. Armstrong, K. A. Bosanko, D. Babovic-Vuksanovic, L. Baker, D. G. Basel, M. Bengala, J. T. Bennett, C. Chambers, L. K. Clarkson, M. Clementi, F. M. Cortés, M. Cunningham, M. D. D’Agostino, M. B. Delatycki, M. C. Digilio, L. Dosa, S. Esposito, S. Fox, M. L. Freckmann, C. Fauth, T. Giugliano, S. Giustini, A. Goetsch, Y. Goldberg, R. S. Greenwood, C. Griffis, K. W. Gripp, P. Gupta, E. Haan, R. K. Hachen, T. L. Haygarth, C. Hernández-Chico, K. Hodge, R. J. Hopkin, L. Hudgins, S. Janssens, K. Keller, G. Kelly-Mancuso, A. Kochhar, B. R. Korf, A. M. Lewis, J. Liebelt, A. Lichty, R. H. Listernick, M. J. Lyons, I. Maystadt, M. Martinez Ojeda, C. McDougall, L. K. McGregor, D. Melis, N. Mendelsohn, M. J. M. Nowaczyk, J. Ortenberg, K. Panzer, J. G. Pappas, M. E. Pierpont, G. Piluso, V. Pinna, E. K. Pivnick, D. A. Pond, C. M. Powell, C. Rogers, N. Ruhrman Shahar, S. L. Rutledge, V. Saletti, S. A. Sandaradura, C. Santoro, U. A. Schatz, A. Schreiber, D. A. Scott, E. A. Sellars, R. Sheffer, E. Siqveland, J. M. Slopis, R. Smith, A. Spalice, D. W. Stockton, H. Streff, A. Theos, G. E. Tomlinson, G. Tran, P. L. Trapane, E. Trevisson, N. J. Ullrich, J. Van den Ende, S. A. Schrier Vergano, S. E. Wallace, M. F. Wangler, D. D. Weaver, K. H. Yohay, E. Zackai, J. Zonana, V. Zurcher, K. B. M. Claes, M. Eoli, Y. Martin, K. Wimmer, A. De Luca, E. Legius, L. M. Messiaen, "Clinical spectrum of individuals with pathogenic NF 1 missense variants affecting p.Met1149, p.Arg1276, and p.Lys1423: genotype–phenotype study in neurofibromatosis type 1," Human Mutation, vol. 41 no. 1, pp. 299-315, DOI: 10.1002/humu.23929, 2020.

[28] C. Mattocks, D. Baralle, P. Tarpey, C. ffrench-Constant, M. Bobrow, J. Whittaker, "Automated comparative sequence analysis identifies mutations in 89% of NF1 patients and confirms a mutation cluster in exons 11-17 distinct from the GAP related domain," Journal of Medical Genetics, vol. 41 no. 4, article e48, 2004.

[29] M. Koczkowska, T. Callens, A. Gomes, A. Sharp, Y. Chen, A. D. Hicks, A. S. Aylsworth, A. A. Azizi, D. G. Basel, G. Bellus, L. M. Bird, M. A. Blazo, L. W. Burke, A. Cannon, F. Collins, C. DeFilippo, E. Denayer, M. C. Digilio, S. K. Dills, L. Dosa, R. S. Greenwood, C. Griffis, P. Gupta, R. K. Hachen, C. Hernández-Chico, S. Janssens, K. J. Jones, J. T. Jordan, P. Kannu, B. R. Korf, A. M. Lewis, R. H. Listernick, F. Lonardo, M. J. Mahoney, M. M. Ojeda, M. T. McDonald, C. McDougall, N. Mendelsohn, D. T. Miller, M. Mori, R. Oostenbrink, S. Perreault, M. E. Pierpont, C. Piscopo, D. A. Pond, L. M. Randolph, K. A. Rauen, S. Rednam, S. L. Rutledge, V. Saletti, G. B. Schaefer, E. K. Schorry, D. A. Scott, A. Shugar, E. Siqveland, L. J. Starr, A. Syed, P. L. Trapane, N. J. Ullrich, E. G. Wakefield, L. E. Walsh, M. F. Wangler, E. Zackai, K. B. M. Claes, K. Wimmer, R. van Minkelen, A. De Luca, Y. Martin, E. Legius, L. M. Messiaen, "Expanding the clinical phenotype of individuals with a 3-bp in-frame deletion of the NF 1 gene (c.2970_2972del): an update of genotype -phenotype correlation," Genetics in Medicine, vol. 21 no. 4, pp. 867-876, DOI: 10.1038/s41436-018-0269-0, 2019.

[30] J. T. den Dunnen, R. Dalgleish, D. R. Maglott, R. K. Hart, M. S. Greenblatt, J. McGowan-Jordan, A.-F. Roux, T. Smith, S. E. Antonarakis, P. E. M. Taschner, on behalf of the Human Genome Variation Society (HGVS), the Human Variome Project (HVP), and the Human Genome Organisation (HUGO), "HGVS recommendations for the description of sequence variants: 2016 Update," Human Mutation, vol. 37 no. 6, pp. 564-569, DOI: 10.1002/humu.22981, 2016.

[31] J. Sambrook, E. F. Fritsch, T. Maniatis, Molecular Cloning. A Laboratory Manual, 1989.

[32] D. G. Gibson, L. Young, R. Y. Chuang, J. C. Venter, C. A. Hutchison, H. O. Smith, "Enzymatic assembly of DNA molecules up to several hundred kilobases," Nature Methods, vol. 6 no. 5, pp. 343-345, DOI: 10.1038/nmeth.1318, 2009.

[33] L. G. Dufner-Almeida, S. Nanhoe, A. Zonta, M. Hosseinzadeh, R. Kom-Gortat, P. Elfferich, G. Schaaf, A. Kenter, D. Kümmel, N. Migone, S. Povey, R. Ekong, M. Nellist, "Comparison of the functional and structural characteristics of rare TSC 2 variants with clinical and genetic findings," Human Mutation, vol. 41 no. 4, pp. 759-773, DOI: 10.1002/humu.23963, 2019.

[34] D. Halim, E. Brosens, F. Muller, M. F. Wangler, A. L. Beaudet, J. R. Lupski, Z. H. C. Akdemir, M. Doukas, H. J. Stoop, B. M. de Graaf, R. W. W. Brouwer, W. F. J. van Ijcken, J. F. Oury, J. Rosenblatt, A. J. Burns, D. Tibboel, R. M. W. Hofstra, M. M. Alves, "Loss-of-function variants in _MYLK_ cause recessive megacystis microcolon intestinal hypoperistalsis syndrome," American Journal of Human Genetics, vol. 101 no. 1, pp. 123-129, DOI: 10.1016/j.ajhg.2017.05.011, 2017.

[35] M. van Triest, J. de Rooij, J. L. Bos, "Measurement of GTP-bound Ras-like GTPases by activation-specific probes," Methods in Enzymology, vol. 333, pp. 343-348, DOI: 10.1016/s0076-6879(01)33068-9, 2001.

[36] A. Anna, G. Monika, "Splicing mutations in human genetic disorders: examples, detection, and confirmation," Journal of Applied Genetics, vol. 59 no. 3, pp. 253-268, DOI: 10.1007/s13353-018-0444-7, 2018.

[37] R. T. Geist, D. H. Gutmann, "Expression of a developmentally-regulated neuron-specific isoform of the neurofibromatosis 1 (NF1) gene," Neuroscience Letters, vol. 211 no. 2, pp. 85-88, DOI: 10.1016/0304-3940(96)12730-0, 1996.

[38] D. Wallis, K. Li, H. Lui, K. Hu, M. J. Chen, J. Li, J. Kang, S. Das, B. R. Korf, R. A. Kesterson, "Neurofibromin (NF1) genetic variant structure–function analyses using a full- length mouse cDNA," Human Mutation, vol. 39 no. 6, pp. 816-821, DOI: 10.1002/humu.23421, 2018.

[39] J. M. Martinez, H. H. Breidenbach, R. Cawthon, "Long RT-PCR of the entire 8.5-kb NF1 open reading frame and mutation detection on agarose gels," Genome Research, vol. 6 no. 1, pp. 58-66, DOI: 10.1101/gr.6.1.58, 1996.

[40] J. Dekker, R. Schot, M. Bongaerts, W. G. de Valk, M. M. van Veghel-Plandsoen, K. Monfils, H. Douben, P. Elfferich, E. Kasteleijn, L. M. A. van Unen, G. Geeven, J. J. Saris, Y. van Ierland, F. W. Verheijen, M. L. T. van der Sterre, F. Sadeghi Niaraki, H. H. Huidekoper, M. Williams, M. Wilke, V. J. M. Verhoeven, M. Joosten, A. J. A. Kievit, I. M. B. H. van de Laar, L. H. Hoefsloot, M. Hoogeveen-Westerveld, M. Nellist, G. M. S. Mancini, T. J. van Ham, "RNA-sequencing improves diagnosis for neurodevelopmental disorders by identifying pathogenic non-coding variants and reinterpretation of coding variants," ,DOI: 10.1101/2022.06.05.22275956, 2022.

[41] V. Pinna, V. Lanari, P. Daniele, F. Consoli, E. Agolini, K. Margiotti, I. Bottillo, I. Torrente, A. Bruselles, C. Fusilli, A. Ficcadenti, S. Bargiacchi, E. Trevisson, M. Forzan, S. Giustini, C. Leoni, G. Zampino, M. C. Digilio, B. Dallapiccola, M. Clementi, M. Tartaglia, A. De Luca, "p.Arg1809Cys substitution in neurofibromin is associated with a distinctive NF1 phenotype without neurofibromas," European Journal of Human Genetics, vol. 23 no. 8, pp. 1068-1071, DOI: 10.1038/ejhg.2014.243, 2015.

[42] I. D'Angelo, S. Welti, F. Bonneau, K. Scheffzek, "A novel bipartite phospholipid-binding module in the neurofibromatosis type 1 protein," EMBO Reports, vol. 7 no. 2, pp. 174-179, DOI: 10.1038/sj.embor.7400602, 2006.

[43] S. Welti, S. Kuhn, I. D'Angelo, B. Brugger, D. Kaufmann, K. Scheffzek, "Structural and biochemical consequences of NF1 associated nontruncating mutations in the Sec14-PH module of neurofibromin," Human Mutation, vol. 32 no. 2, pp. 191-197, DOI: 10.1002/humu.21405, 2011.

[44] S. F. B. Fadhlullah, N. B. A. Halim, J. Y. T. Yeo, R. L. Y. Ho, P. Um, B. T. Ang, C. Tang, W. H. Ng, D. M. Virshup, I. A. W. Ho, "Pathogenic mutations in neurofibromin identifies a leucine-rich domain regulating glioma cell invasiveness," Oncogene, vol. 38 no. 27, pp. 5367-5380, DOI: 10.1038/s41388-019-0809-3, 2019.

[45] Z.-Y. Zheng, M. Anurag, J. T. Lei, J. Cao, P. Singh, J. Peng, H. Kennedy, N.-C. Nguyen, Y. Chen, P. Lavere, J. Li, X.-H. Du, B. Cakar, W. Song, B.-J. Kim, J. Shi, S. Seker, D. W. Chan, G.-Q. Zhao, X. Chen, K. C. Banks, R. B. Lanman, M. Nemati Shafaee, X. H.-F. Zhang, S. Vasaikar, B. Zhang, S. G. Hilsenbeck, W. Li, C. E. Foulds, M. J. Ellis, E. C. Chang, "Neurofibromin is an estrogen receptor- α transcriptional co-repressor in breast cancer," Cancer Cell, vol. 37 no. 3, pp. 387-402.e7, DOI: 10.1016/j.ccell.2020.02.003, 2020.

[46] J. C. W. Douben, M. Hoogeveen-Westerveld, M. Nellist, J. Louwen, M. Kroos-de Haan, M. Punt, B. van Ommeren, L. van Unen, P. Elfferich, E. Kasteleijn, Y. van Bever, M. van Vliet, R. Oostenbrink, J. J. Saris, A. Wagner, Y. van Ierland, T. van Ham, R. van Minkelen, "Functional assays combined with pre-mRNA splicing analysis improve variant classification and diagnostics for individuals with neurofibromatosis type 1 and Legius syndrome," ,DOI: 10.22541/au.165217208.81458305/v1, 2022.