Corresponding author: Andrea Zatkova, PhD, Institute of Molecular Physiology and Genetics, Slovak Academy of Sciences, Vlarska 5, 833 34 Bratislava, Slovakia. Tel: +421 9 1146 6599; E-mail: email@example.com
We performed a complex analysis of the neurofibromatosis type 1 (NF1) gene in Slovakia based on direct cDNA sequencing supplemented by multiple ligation dependent probe amplification (MLPA) analysis. All 108 patients had café-au-lait spots, 85% had axilary and/or inguinal freckling, 61% neurofibromas, 36% Lisch nodules of the iris and 31% optic pathway glioma, 5% suffered from typical skeletal disorders, and 51% of patients had family members with NF1.
In 78 of the 86 (90.7%) index patients our analysis revealed the presence of NF1 mutations, 68 of which were small changes (87.2%), including 39 (50%) novel. Among the identified mutations the most prevalent were small deletions and insertions causing frameshift (42.3%), followed by nonsense (14.1%), missense (12.8%), and typical splicing (11.5%) mutations. Type 1 NF1 deletions and intragenic deletions/duplication were identified in five cases each (6.4%). Interestingly, in five other cases nontypical splicing variants were found, whose real effect on NF1 transcript would have remained undetected if using a DNA-based method alone, thus underlying the advantage of using the cDNA-based sequencing. We show that Slovak NF1 patients have a similar repertoire of NF1 germline mutations compared to other populations, with some prevalence of small deletions/insertions and a decreased proportion of nonsense mutations.
Neurofibromatosis type 1 (OMIM# 162200) is one of the most frequent autosomal dominant disorders (1:3,000–1:4,000) and is caused by mutations in the NF1 gene (reviewed in Gutmann et al., 1997).
Clinical diagnosis of the disease is based on the presence of at least two of the seven National Institute of Health (NIH) diagnostic criteria (NIH, 1988), the most distinctive features of which are multiple benign, soft tumors called neurofibromas (NFs), and patches of skin pigmentation called “café-au-lait” spots (CALs). NF1 can affect nerves throughout the body, including those in the brain and spinal cord. Increased growth in benign diffuse NFs can be invasive and cause a distressing cosmetic defect or functional restriction of surrounding vitally important tissues. Plexiform NFs are prone to progression into malignant peripheral nerve sheath tumors (or neurofibrosarcomas). Another possible malignant tumor that may arise is pheochromocytoma. Early progressive deformation, invalid status, learning difficulties, and cognitive impairment result in severe social handicap for these patients.
The clinical phenotype is progressive and can vary, even within a family, where modified genes have been proposed as causative agents for at least part of this phenotypic variability (Harder et al., 2010; Jentarra et al., 2012). The NF1 protein, neurofibromin, is a “tumor suppressor,” an activator of Ras-GTPase (Martin et al., 1990; Yunoue et al., 2003). Impaired expression of neurofibromin leads to a persistent increased Ras activity and subsequent increased cell proliferation (DeClue et al., 1992). In addition to regulating Ras, neurofibromin also positively regulates cyclic adenosine monophosphate (cAMP) levels (Dasgupta et al., 2003). Consistent with Knudson's “two-hit” hypothesis, patients harboring NF1 germline mutation develop a range of NF1 symptoms, such as CALs, NFs, tibial dysplasia, and optic pathway tumors upon somatic mutation of the second, wild-type NF1 allele in specific cells (Xu et al., 1992; Legius et al., 1993; Shannon et al., 1994; Colman et al., 1995; Sawada et al., 1996; Rasmussen et al., 2000; Cichowski & Jacks, 2001).
The NF1 gene (17.q11.2, 280 kb genomic DNA) consists of 57 constitutive and at least three alternatively spliced exons (23a, 48a, 9br; Andersen et al., 1993; Gutman et al., 1993; Danglot et al., 1995) and codes for a transcript, which is 8454 nucleotides long (Cawthon et al., 1990; Viskochil et al., 1990; Wallace et al., 1990; Xu et al., 1990; Marchuk et al., 1991; Li et al., 1995). The majority of patients carry a heterozygous germline mutation consisting of a single substitution or small deletions or insertions. These are spread along the entire coding sequence (NM_000267.3) and frequently lead to a premature stop codon (Fahsold et al., 2000; Messiaen et al., 2000; Ars et al., 2003). In addition, 5–10% of the patients carry microdeletions or deletions of the entire NF1 gene (Rasmussen et al., 1998; Kluwe et al., 2004), including neighboring genes (Cnossen et al., 1997; Leppig et al., 1997; Lopez Correa et al., 1999; Dorschner et al., 2000). In 2% of cases the intragenic copy number of one or more exons changes through deletions or duplications (Wimmer et al., 2006). Identification of the NF1 mutation is challenging because the NF1 gene is large and complex, lacks mutational “hot-spots” and a wide spectrum of different types of mutation have been identified. Although some recurrent mutations have been described (Hoffmeyer et al., 1998; Ars et al., 2000; 2003; Fahsold et al., 2000; Messiaen et al., 2000; Wimmer et al., 2006; Griffiths et al., 2007; Valero et al., 2011), each of these is present in only a small percentage of patients, and therefore, it is necessary to analyze the entire NF1 sequence in each patient. Moreover, multiple unprocessed NF1 pseudogenes are found in the human genome and this further complicates specific primer design (reviewed in Luijten et al., 2001). In addition, approximately 50% of cases are sporadic, and thus result from de novo mutations. Messiaen et al. (2000) reported that approximately one-third of pathological NF1 mutations influence mRNA splicing and 30% of all such mutations lie outside the classical consensus splice-regulatory sequences (Messiaen & Wimmer, 2008). Many nontypical splicing mutations create new or activate cryptic splice sites, either within the exon or deep in intronic sequences. Another group of mutations include missense or stop mutations, which are present within exons and actually affect correct splicing of mRNA by abolishing regulatory sequences—so-called exonic splicing enhancers (ESE) (Messiaen et al., 1997; Zatkova et al., 2004). Nontypical splicing mutations often remain unclassified or even undetected when screening methods are based on DNA analysis alone. Therefore, in Slovakia we implemented a mutation analysis protocol based on analysis of NF1 mRNA transcripts (Messiaen & Wimmer, 2012) and herein we report mutation analysis by direct cDNA sequencing and MLPA in 124 suspected patients.
Materials and Methods
In this study, 124 suspected Slovak NF1 patients from 102 families were analyzed. After re-evaluation of NIH diagnostic criteria (NIH, 1988) in each patient, 108 (86 index cases) exhibited typical NF1 features. Clinical features of all tested patients can be found in Table S1 and these are summarized in Figure 1. All patients had CALs, 85% had axillary and/or inguinal freckling (FRC), 61% NFs, 36% Lisch nodules of the iris (LN), 31% optic glioma (OPG), 5% suffered from typical skeletal disorders (SD), and 51% had family members with NF1. The age of analyzed patients was from a few months to 67 years, with an average of 35 years and a median of 14 years. All patients gave informed consent prior to analysis.
Mutation Screening Protocol
The mutation screening protocol was essentially as described by Messiaen and Wimmer (2012). In brief, the patient's EDTA-blood sample was split into two aliquots: (i) for DNA extraction, (ii) for short-term lymphocyte culture and RNA extraction. Puromycin was added to the culture 2–4 h prior to RNA harvest to prevent nonsense mediated RNA decay (NMD). The RNA was then isolated and cDNA prepared according to standard protocols using Trisol reagent and Superscript RT (Invitrogen-Life technologies, Carlsbad, CA, USA), respectively.
The entire NF1 gene coding region was PCR amplified from total cDNA in five overlapping fragments of approximately 2 kb (covering exons 1–12a(16), 10a(12)–21(27), 19a(24)–29(38), 28(37)–38(47), and 34(43)–49(58)), using primers published in Messiaen & Wimmer (2012). All fragments were analyzed by gel electrophoresis, which enables direct identification of shorter or longer fragments, thus indicating the presence of splicing or intragene deletion/insertion mutations. Subsequently, individual PCR fragments were directly sequenced using 19 internal primers (Messiaen & Wimmer, 2012) and commercial kits (BigDye 3.1 Applied Biosystems-Life technologies, Carlsbad, CA, USA). Sequences were analyzed on ABI PRISM® 3100-Avant genetic analyzer (Applied Biosystems).
All mutations identified in cDNA were also confirmed in the patient's genomic DNA. The relevant exon with short intronic sequences was amplified and then sequenced as described above, using gDNA specific primers (available upon request).
Mutations were labelled as novel if they had not been reported in the literature and in the Human Genome Mutation Database (HGMD) (Institute of Medical Genetics, Cardiff, Wales, UK; http://www.hgmd.cf.ac.uk/ac/index.php) or in the LOVD [Leiden Open (Source) Variation Database] NF1 mutation database (http://www.LOVD.nl/NF1).
The old consensus exon number was used to describe the mutation position, followed by a new exon number in brackets, as in exon 28(37). Mutations are described as recommended by the human genome variation society (den Dunnen & Antonarakis, 2000) using NM_000267.3 as the reference sequence.
Where no mutation was found by cDNA sequencing, the patient's gDNA samples were analyzed by MLPA with NF1 microdeletion MLPA salsas P122, P081, and P082 (MRC Holland). Salsa P122 readily detects the deletions of the entire NF1 gene and P081 and P082 reveal also possible intragenic copy number changes, which escaped detection by RT-PCR. The Salsa P122 NF1 area assay was also used to distinguish between the 1.4 Mb deletions (type I, Type 1 NF1 deletion) encompassing 14 genes, with breakpoints in the NF1 low-copy repeats, and the 1.2 Mb deletions (type II), which cover 13 genes and are mediated by recombination between the JJAZ1 gene and its pseudogene (Kehrer-Sawatzki et al., 2004; Wimmer et al., 2006).
RT-PCR Test for Possible Splicing Effect of Silent Mutations
We also performed a simple RT-PCR test to verify the possible effect of silent changes c.4866G>A (p.Val1622 = , rs17880521) and c.5172G>A (p.Lys1724 = , rs17887014) on exon 28(37) splicing, as well as of c.168G>A (p.Ser56 = , rs17881168) in exon 2 splicing. Short fragments, which contained the corresponding exons were amplified from the cDNA of the patients carrying these variants and also from the controls. The size of the PCR fragments in gel electrophoresis enables distinction of products with exon skipping. The expected size of the exon 28 RT-PCR product containing exon 28 is 760 bp, while the size of that lacking exon 28 is 328 bp. Similarly, the RT-PCR product containing exon 2 is 470 bp long, and the one without exon 2 it is 326 bp.
Predictions of the Possible Impact of Amino Acid Substitution on the Structure and Function of a Human NF1 Protein
Polymorphism Phenotyping v2 (PolyPhen-2, http://genetics.bwh.harvard.edu/pph2/; Adzhubei et al., 2010) and SNAP (Screening for nonacceptable polymorphisms, http://www.rostlab.org/services/snap/submit; Bromberg & Rost, 2007) were used to predict the possible effect of all identified missense NF1 mutations on the structure and function of the human NF1 protein (P21359-2). PolyPhen-2 uses straightforward physical and comparative considerations to calculate the Naïve Bayes posterior probability that the mutation is damaging and it reports estimates of true and false positive rates. The mutation is also appraised qualitatively, as “benign,” “possibly damaging,” or “probably damaging” based on the model's false positive rate (Adzhubei et al., 2010). HumVar-trained PolyPhen-2 enabled us to distinguish mutations with drastic effects from all remaining human variations, including the abundant mildly deleterious alleles.
SNAP is a neural network-based method. Its reliability index (RI) ranges between 0 and 9, with higher reliability indexes strongly correlating with a higher accuracy of prediction. The expected accuracy at a given reliability index is the number of correctly predicted neutral or non-neutral samples in the SNAP testing set. This measure of accuracy establishes the likelihood that a given prediction is correct (Bromberg & Rost, 2007).
Predictions of the Possible Impact of Mutations on mRNA Splicing
In this study, 78 different NF1 mutations were identified by cDNA sequencing and/or MLPA in 86 index NF1 cases (90.7%) who fulfilled NIH diagnostic criteria (Figs. 2B and S1). The same mutation was found in patients in the same family. The clinical data of all cases and identified mutations is summarized in Table S1(A). The majority of the identified mutations consist of small changes (68/78, 87.2%), 39 (57.35%) of which were novel. The most prevalent of the small mutations involved were small deletions and insertions almost always causing frameshift and preliminary stop codon (42.3%), followed by nonsense (14.1%) and missense (12.8%) mutations (Fig. 2A). Interestingly, classic splicing mutations were present in only nine Slovak NF1 patients (11.5%), while nontypical splicing variants were found in another five (6.4%).
In 39 families the parents were also tested for the presence of the identified mutations and this confirmed a de novo mutation in 23 patients (59%).
All 68 small mutations were distributed throughout the entire NF1 gene. Fifteen of these were located in the exons 11–17(15–22) region (Fig. 2B), which encodes a domain rich in cysteine and serine (CSRD) (Izawa et al., 1996).
Thirteen small mutations were mapped into the Ras-GRD (Ras-GTPase activating proteins (GAP)– Related Domain) of exons 20–27a(26–35) (Martin et al., 1990; Xu et al., 1990), and this domain was also affected by three intragenic multiple exon deletions (Fig. 2B). In addition to exon 16(21) where six different NF1-causing mutations were found, multiple mutations were also observed in exons 7(9), 12a(16), and 28(37), carrying four, four, and five variants, respectively.
Two recurrent mutations were identified in this Slovak NF1 patient group. The same stop mutation c.3827G>A (p.Arg1276*) in exon 22(28) was found in two unrelated cases, P54 and P61, and the second mutation was the small deletion c.1756_59delACTA (p.Thr586Valfs*18) in exon 12a(16) observed in P87 and P118. The same nucleotide position, c.6641+1G in intron 35(44) mutated twice; once into T (c.6641+1G>T) in patient P17 and once into C (c.6641+1G>C) in P126.
Nontypical splicing mutations were found in five cases. Patient P68 showed a deletion of 75 base pairs of exon 7(9) in cDNA (r.988_1062del). The protein is predicted to lack 25 amino acids encoded in exon 7(9) (p.Ala330_Lys354del). The analysis of exon 7(9) gDNA in this patient revealed the single nucleotide c.989C>T substitution which can be classified as missense mutation p.Ala330Val. The splice site prediction analysis using HSF (Desmet et al., 2009), however, reveals that this substitution creates a cryptic donor splice site.
A similar situation was observed in patient P139, who carried the small substitution c.4268A>G leading to skipping of exon 24(32)(r.4111_4269del). In this case, the predicted protein is 53 amino acids shorter (p.Val1371_Lys1423del). This mutation was previously described as missense change p.Lys1423Arg (Han et al., 2001).
Another interesting variation occurred in P90 who has a short genomic deletion including part of intron 28(37) and exon 29(38) eliminating the acceptor splice site and causing the skipping of either exon 29(38) alone (p.Gly1737Serfs*4, r.5206_5546del329) or of both exons 29(38) and 30(39) (p.Gly1737Leufs*3, r.5206_5749del532). Both variants were observed in the patient's cDNA.
The intronic substitution c.1261–19G>A in P83 creates a novel acceptor splice-site in intron 9(11) which led to the insertion of 17 nucleotides into the transcript (r.1260insTCTTTGTTTTTCTCTAG). This gave a frameshift mutation and premature stop codon after 57 aminoacids (p.Ala422Leufs*57) (Fig. 3).
Patient P135 carried the last of the five nontypical splicing variants; substitution c.1466A>G presumably leading to the missense mutation p.Tyr489Cys. However, deletion of the distal part of exon 10b(13) leading into frameshift and premature stop (r.1466_1527del, p.Tyr489*) was observed at the cDNA level, as previously described by (Messiaen et al., 1999). The BDGP and HSF programs predicted that this substitution in codon 489 creates a novel donor splice-site, which caused the splicing effect revealed in cDNA analysis.
In addition to the frameshift mutation in exon 10b(13) (p.Ile500Asnfs*11, c.1497_98insA) that the 21-year-old patient P37a shared with his affected father (P37, 52 years old) and sister (P37b, 18 years old), this patient also carried a variant in intron 10a(12) not present in the father or sister (c.1393–32C>T, rs2905876).
PolyPhen2 and SNAP predictions of the possible effect of all missense mutations identified in our patients on the structure and function of the human NF1 protein are summarized in Table 1(A). At least one of the programs indicated a damaging or non-neutral effect for all of these. SNAP predicted p.Ile1918Thr (c.5753T>C) in exon 31(40) as a “neutral” mutation, while PolyPhen labeled it as “probably damaging” (Table 1A). This mutation was identified in the 2-year-old patient P115 who exhibited CALs and freckling.
Table 1. List of all missense (A) and silent changes (B) in the NF1 gene identified in Slovak patients. The results of PolyPhen2 and SNAP predictions for missense changes and ESE site disruption predictions for silent changes are shown. Human Splicing Finder was used to predict the effect on exon splicing regulatory elements. No sample splicing effect was observed in RT-PCR test caused by the silent changes (Figure 4), and additional disease-causing mutations were found with the silent changes in patients P63, P64, P86, and P112. The DNA numbering system is based on cDNA (NM_000267.3), with +1 corresponding to the A of the ATG
(A) NF1 missense variants in coding region
Type of the NF1 Causing Mutation
PolyPhen Predictions (HumVar)
SNAP Predictions (RI; Expected Accuracy)
Probably damaging with a score of 0.986 (sensitivity: 0.54; specificity: 0.94)
Non-neutral (8; 91%)
From mother's side many oncological diseases present
Probably damaging with a score of 0.987 (sensitivity: 0.53; specificity: 0.95)
Non-neutral (8; 91%)
p.Asp176Glu, c.528T>A, MLPA negat
Benign with a score of 0.397 (sensitivity: 0.84; specificity: 0.79) (HumDiv - possibly damaging with a score of 0.691 (sensitivity: 0.86; specificity: 0.92))
p.Ala330Val, c.989C>T (creates a cryptic donor splice site and thus deletion r.988_1062del (p.Ala330_Lys354del))
Nontypical splicing (missense)
Probably damaging with a score of 0.971 (sensitivity: 0.60; specificity: 0.93)
Non-neutral (8; 91%)
(B) NF1 silent variants
Silent NF1 Mutation
NF1 Causing Mutation
Predictions for Silent Changes (HSF)
RI* = reliability index, range 0-9, where higher reliability indexes correlated strongly with higher accuracy of prediction. Expected accuracy is the number of correctly predicted (at a given reliability index) neutral or non-neutral samples in the SNAP testing set. Both SNAP and PolyPhen2 predictions were performed using NF1protein sequence P21359-2.
c.168C>T, p.Ser56= rs17881168
c.204+2T>G in intron 2, causes skipping of exon 2: r.61_204del, p.Leu21_Met68del
new ESE site created (SRp40), new acceptor site created (score 67,79)
In four cases we observed the presence of silent NF1 variants in addition to the typical small NF1 mutation. Patient P63 carried a novel silent substitution c.5895A>C (p.Lys1965 = ) in exon 31(40) in addition to a frameshift mutation in exon 3(3). P64, P86, and P112 carried the c.168C>T variant (p.Ser56 = , rs17881168) in exon 2(2), as well as disease causing splicing mutations in intron 2(2) (P64) and intron 45(54) (P86), or a missense mutation in exon 13(18) (P112) (Table 1B).
MLPA analysis was performed in each patient where cDNA sequencing failed to identify mutations. Type 1 NF1 gene deletion was identified in five cases (6.4%) (Fig. S1, Table S1(A)), and an intragenic deletion of one to six NF1 exons was revealed in four additional patients, as indicated in Figure 2B. In addition, P138 showed duplication of almost the entire exon 45(54), which caused frameshift and shortened protein.
NF1 Cases Fulfilling NIH Diagnostic Criteria with no NF1 Mutation Identified
Our methodology failed to detect any classic NF1 mutation in eight index patients with clear NF1 phenotype (Table S1(B)). Most likely they represent a segmental or mosaic variant of the disease where it is not possible to identify the mutation without testing the affected tissue.
In the 5-year-old patient P22, we found the missense change p.Asp176Glu (c.528T>A) in NF1 exon 4b(5), which had earlier been described as a benign variant (Toliat et al., 2000; Bendova et al., 2007). This variant was also present in an unaffected mother and maternal grandmother and no other change was identified in this patient by sequencing the entire coding region of the NF1 gene and MLPA analysis. This patient has multiple CAL, freckling and NFs. Although the PolyPhen prediction of the possible effect of this mutation on NF1 function indicates that this mutation is benign, SNAP classifies it as non-neutral (Table 1A).
The silent change c.4866G>A (p.Val1622 = , rs17880521) in exon 28(37) was found in a 7-year-old patient P26 with nontypical CALs, Lisch nodules and whose mother had CALs. No classic NF1 mutation was identified. Rs17880521 had already been described as a rare polymorphism possibly affecting the ESE (Table 1B).
Nontypical Cases with NF1 Signs but no NF1 Mutations
A further 15 index cases sent to our laboratory for DNA diagnostics had clinical signs not typical for NF1 and, as our analysis shows, they did not carry typical NF1 mutations. In three of these cases, silent NF1 variants were identified (Table S1(C)). One was P42, who carries the same change in exon 28(37) as the above-mentioned patient P26. This p.Val1622 = (c.4866G>A, rs17880521) variant and also another silent change in exon 28(37) found in P35: p.Lys1724 = (c.5172G>A, rs17887014), have already been described as rare polymorphisms possibly affecting ESE (Table 1B).
The remaining silent change c.168G>A (p.Ser56 = , rs17881168) in exon 2(2) found in P30 has also been described previously. ESE-prediction programs show disruption of putative ESEs for two SR (Serine-Arginine-rich) proteins, SRp55 and SF2/ASF. The same change was identified also in P64 and P112, where NF1 was caused by a splicing mutation in intron 2 (c.204+2T>G) and missense mutation p.Leu695Pro in exon 13(18), respectively.
We also performed a simple RT-PCR test to verify the possible effect of silent changes c.4866G>A (p.Val1622 = ), c.5172G>A (p.Lys1724 = ) in exon 28(37) and c.168G>A (p.Ser56 = ) in exon 2(2) on exon splicing. As Figure 4 shows, no exon skipping was observed in the cDNA samples carrying these changes.
In this study, we performed analysis of mutations in the NF1 gene in 124 referred patients from 102 Slovak families with NF1 features. As our results demonstrate, the combination of cDNA sequencing and MLPA analysis forms a suitable diagnostic tool for identification of germline NF1 mutations, since we were able to identify mutations in 78 of 86 (90.7%) index patients who fulfilled NIH criteria. The mutation detection rate of NF1-causing mutations in families was even higher at 28 of 30 patients who had an affected parent (93.3%).
Most of the identified mutations were small changes (87.2%) where small deletions/insertions predominated (42.3%). Since the largest group of 2900 unrelated NF1 patients examined by Messiaen and Wimmer in 2012 revealed this type of mutations in 26%, the frequency in Slovakia is almost 62% higher. Intragenic copy number changes of one or more exons, such as deletions or duplications, were also quite frequent. These were found in 6.4% of Slovak patients compared to Messiaen and Wimmer's 2.5% reported in 2012. Additionally, nonsense and splicing mutations appeared less prevalent at 14.1% and 17.9% of our Slovak patients, respectively, whereas in the cohort of Messiaen and Wimmer (2012) these figures were 23% and 29%. These differences may be random, since some variations to this large cohort's results are also reported in other studies involving smaller patient groups (Lazaro et al., 1996; Cnossen et al., 1997; Rasmussen et al., 1998; Messiaen et al., 2000; Kluwe et al., 2004; Wimmer et al., 2006; Griffiths et al., 2007; Valero et al., 2011). However, the tendency towards an increased number of small deletions or insertions (42.3%) and a decreased proportion of nonsense mutations (14.1%) in our study differs also from some other cohorts with reported proportions of 18–27.5% and 23–37%, respectively (Lazaro et al., 1996; Cnossen et al., 1997; Rasmussen et al., 1998; Messiaen et al., 2000; Kluwe et al., 2004; Wimmer et al., 2006; Griffiths et al., 2007; Valero et al., 2011).
It was shown in 1998 that mutations in the NF1 GAP-related domain (GRD) (Martin et al., 1990; Xu et al., 1990) completely disable GAP activity (Klose et al., 1998), and this provides direct evidence that failure of neurofibromin GAP function is the critical element in NF1 pathogenesis. In Slovak patients, 16 NF1 mutations affect Ras-GRD and are, thus, predicted to severely decrease neurofibromin function.
Clustering of the mutation in the upstream gene segment comprising exons 11–17(15–22) has previously been observed (Fahsold et al., 2000; Mattocks et al., 2004). This region forms the CSRD domain with three cysteine pairs suggestive of ATP binding, and three potential cAMP-dependent protein kinase (PKA) recognition sites obviously phosphorylated by PKA (Izawa et al., 1996).
Based on the results of their study in neural cells, Mangoura et al. (2006) showed that neurofibromin Ras-GAP activity is regulated by PKC-dependent phosphorylation of CSRD. The authors proposed that this phosphorylation may serve to increase Ras-GAP function and to promote cell growth arrest and actin cytoskeleton-associated functions instead of cell proliferation in response to pleiotropic growth factors (Mangoura et al., 2006). In our cohort, 15 identified mutations affected a CSRD.
This study confirms the advantage of using an RNA/cDNA-based method to identify nontypical splicing variants. In patients P68, P135, and P139, the putative missense changes in exon 7 (9), exon 10b (13), and 24 (32), actually caused defective splicing. The actual effect of these substitutions would not have been revealed if DNA analysis alone had been used. Patient P139 showed skipping of exon 24, and here a single nucleotide substitution c.4268A>G in exon 24 was found which had previously been described as missense change p.Lys1423Arg (Han et al., 2001). The BDGP and HSF prediction programs showed that this substitution of adenine for guanine in codon 1423 in exon 24 weakens the donor splicing site and concurrently it also abolishes ESEs for SRp40, SF2/ASF a SF2/ASF (IgM-BRCA1). In addition, it activates the exonic splicing silencer (ESS), which may have contributed to the incorrect splicing observed in this patient.
A novel c.989C>T substitution in patient P68 created a cryptic donor splice site in exon 7(9), and thus deletion of 75 base pairs (r.988_1062del) and a shortened protein (p.Ala330_Lys354del). Using the DNA sequencing approach this novel mutation would have been described as missense change p.Ala330Val.
Patient P135 carried substitution c.1466A>G presumably leading to the missense mutation p.Tyr489Cys previously described (Messiaen et al., 1999). At the cDNA level we observed deletion of the distal part of exon 10b (13) leading to frameshift and an immediate premature stop codon (r.1466_1527del, p.Tyr489*). Here, the BDGP and HSF prediction programs showed that this substitution in codon 489 creates a novel donor splicing site which affects splicing, and this was revealed in our cDNA analysis.
Similarly, the splicing effect of the intronic variant c.1261–19G>A in patient P83 was also observed. This mutation created a novel acceptor splice site in intron 9(11), leading to insertion of part of the intron into the NF1 transcript and frameshift with the premature stop codon after 57 aminoacids (r.1260_1261 insTCTTTGTTTTTCTCTAG, p.Ala422Leufs*57).
Due to the genomic deletion of 20 base pairs that eliminated the acceptor splice-site in intron 28(37) (c.5206–5_5220delTCCAGGTTGGTTCTACTGCT) in patient P90, skipping of exon 29(38) (r.5206_5546del329, p.Gly1737Serfs*4) or both exons 29(38) and 30(39) (r.5206_5749del532, p.Gly1737Leufs*3) was observed. This led to a premature stop codon in each case. Interestingly, alternative splicing of these exons has been previously described in non-NF1 blood cells. This may have been related to sample handling, or weak regulatory sequences may simply have conferred a propensity for these exon-skips (Park et al., 1998; Vandenbroucke et al., 2002; Raponi et al., 2009).
For accurate molecular diagnosis of NF1, it is important to develop protocols, which estimate the pathogenity of the unknown missense variants and silent changes sometimes found in NF1 patients lacking other pathogenic mutations.
In this study, 39 novel mutations were identified and six of these were missense. For this type of variant we recommend PolyPhen2 and SNAP predictions be utilized, as in our study, to predict possible effects of amino acid substitution on the structure and function of the human protein. Both of these methods use information on 3D protein structures; PolyPhen-2 prediction uses straightforward physical and comparative considerations, while SNAP is a neural network-based method (Bromberg & Rost, 2007; Adzhubei et al., 2010).
We also analyzed patients’ parents whenever possible, to detect possible segregation of the mutation with the disease, and to uncover de novo mutation events. This analysis was performed in 39 families, with patient de novo mutations indicated in 59%. This number is higher than the average 50% de novo cases reported (Messiaen & Wimmer, 2008), and our results could well have positive implications for future family planning.
Segregation analysis in P22 clearly demonstrated that the previously reported benign variant p.Asp176Glu (Toliat et al., 2000; Bendova et al., 2007) is not associated with the disease, since it was also present in the unaffected mother and grandmother. SNAP analysis, however, indicated a non-neutral effect of this change. Functional studies are required to understand the precise pathogenic effects of this variant.
A recent detailed clinical study was performed on 52 of our Slovak NF1 patients, and this is the subject of a separate report (Bolcekova et al., 2013). Briefly, for comparison purposes, these subjects were divided into a group of 25 patients with optic pathway glioma (OPG) and 27 without it. The 52 patients had brain MRIs performed to detect asymptomatic gliomas. Analysis of the distribution of the NF1 mutations showed that of all patients with the mutation localized in the first tertile of the NF1 gene, the majority (71%) were those with OPG (P = 0.0049). Thus, we showed that a significant number of mutations in NF1 patients exhibiting OPG (17/25, 58%) are located in the 5’ tertile of the NF1 gene – exons 1–16(1–21). In the group without OPG, seven of the 27 (27%) individuals carried a mutation located in this region. Of these, three were under 8 years of age and therefore they remain at risk of developing OPG in later life. Although the number of the studied patients is not high and needs to be enlarged, our results agree with those observed in the larger group (Sharif et al., 2011).
Several silent exon variants were observed in our cohort and these also occurred in subjects where no other possible NF1-causing variants were observed. The HSF program predicted that these variants always cause disruption to the ESE sequences, with possible subsequent effect on splicing. However, our simple PCR test recorded no evidence of exon skipping in lymphocytes.
One of the silent variants, p.Ser56 = (rs17881168) in exon 2, was present in patients with typical NF1 signs but lacking any other identified NF1 change, and it was also detected in several patients possessing different classic NF1-causing mutations. This occurred in our cohort and also in two cases from Austria (Katharina Wimmer, personal communication). This is further proof that p.Ser56 = is most likely not associated with the NF1 phenotype. Nevertheless, further studies are needed to test for possible minor splicing anomalies caused by silent variants, which can also contribute to NF1 pathology.
In addition to the frameshift mutation in exon 10b(13) (p.Ile500Asnfs*11, c.1497_98insA) that 21-year-old patient P37a shared with his affected father (P37, 52 years old) and younger sister (P37b, 18 years old), he also carried a variant in intron 10a(12) (c.1393–32C>T, rs2905876) that was absent in the father and sister. Interestingly, the son has a plexiform NF that was not present in either family member. On the other hand, he exhibits no Lisch nodules, which are present in his affected first-degree relatives. Although it is possible that this variant, previously described as benign polymorphism (rs2905876), together with the identified frameshift mutation, contributed to the differences in the severity of the disease in this patient, further study is required to confirm this hypothesis.
Our study also included seven affected sibling pairs in families P18, P6, P89, P37, P61, P48, and P135. In all but one case (P135) the mutation was inherited from the father. Comparison of their phenotypic features indicates variability of expression even despite the low age of the siblings in some cases. In family P18, the 35-year-old father carrying the stop mutation in exon 7(9) (p.Tyr333X) (Ars et al., 2003) had CALs, FRC, LN and also multiple NFs. The same phenotype is shared by his 6-year-old daughter who also developed OPG, and by his 5-year-old daughter who had OPG and skeletal dysplasia but no NFs.
A nonsense mutation in exon 16(22) (p.Arg816*) (Maynard et al., 1997) was identified in the 44-year-old father of family P48 and also in both his 6- and 16-year-old daughters. The older sister shared his phenotype; including CALs, FRC, LNs and NFs, while the younger sister had OPG and no NFs.
A mutation causing the premature stop codon in exon 22(28) (p.Arg1276*) (Heim et al., 1995) was confirmed in a 1-year-old boy and a 4-year-old girl in family P61. Their affected father was not analyzed. Both children had CALs and FRC, while a NF was also reported in the boy.
A novel frameshift mutation in exon 34(43) (p.Phe2176Cysfs*44) was identified in the 44-year-old affected father of family P89 and also in both his 4- and 7-year-old daughters. All had CALs and FRC, and while the father and younger daughter also had LNs, these were not present in the older daughter who developed OPG.
Another novel frameshift mutation in exon 34(43) (p.Arg2162Thrfs*18) was observed in two siblings in family P6 where the father was reported to have the NF1 symptoms. Since his DNA was not analyzed, he is not included in this study, but a 47-year-old female family member had CALs as her only NF1 feature and her 43-year-old brother had FRC and multiple NFs in addition to CALs.
Families P135 and P37 have already been described above. This included the description of the phenotype differences in family P37. Mutations in exon 10b(13) were found in both these families. The mother and two female twins in family P135 carried the nontypical splicing mutation p.Tyr489Cys (pTyr489*) previously reported by (Messiaen et al., 1999). The twins were born in 2012, and they already suffer from CALs and FRC while their 27-year-old mother has CALs, FRC, LNs and NFs.
In addition to affected sibling pairs, our cohort also includes the nine affected parent-and-child pairs listed in Table S1. The child had a more severe phenotype than the parent in the five P58, P47, P16, P45, and P2 pairs.
No NF1 mutation was identified in eight typical NF1 profiles. Since our detection method is not 100% sensitive and the present protocol does not include specific study of exon 1, untranslated or promoter regions, it is possible that some mutations escaped detection. We believe that NMD did not hinder us from detecting NF1 mutations because we used puromycin, and most of our identified mutations resulted in a premature stop codon. Although our sequencing primers covered the entire coding region, there may have been an intron mutation causing a multiple exon-skip that evaded detection due to the primer locations. Therefore, we plan to increase sequence coverage by using primers in the reverse direction, as recommended by Messiaen & Wimmer (2012). Mosaicism could explain the lack of identified NF1 mutations in de novo cases, so NF mutation screening could be helpful here.
In our group of 16 nontypical NF1 cases which did not fulfil NIH criteria, three patients were later differentially diagnosed with suspected schwannomatosis (P13, P13a, and P80) and this condition was histologically confirmed in patient P13 and is also predicted for his son P13a. However, the diagnosis for P80 remains unclear, since this patient had a mixed-phenotype: generalized fibroma similar to schwanomatosis and a mixed type lipoma similar to a schwanoma, so precise histological diagnosis was ambiguous. The majority of schwannomatosis occurrences are sporadic, with approximately one-third present within families. A subset of these has recently been associated with germline mutations in the tumor suppressor gene SMARCB1/INI1 encoding a component of the SWI/SNF chromatin remodelling complex (Hulsebos et al., 2007). More recently, different somatic (but not germline) NF2 mutations in separate schwannomas, together with a SMARCB1/INI1 germline mutation, were described in one family (Sestini et al., 2008). This inactivation of both NF2 and SMARCB1/INI1 appears to be the typical mechanism in schwannomas in SMARCB1/INI1 germline mutated individuals (Hadfield et al., 2010). Following histological confirmation of schwannomas, INI1 immuno-histochemistry can be performed and then screening for SMARCB1/INI1. If this proves positive, screening for NF2 mutations can be performed.
This screening protocol was used in patients P13, with resultant exclusion of NF2 and SMARCB1/INI1 mutations.
An unspecified skin hyper-pigmentation was differentially diagnosed in a further six cases lacking NF1 mutation (P50, P35, P30, P42, P24, and P28). Two patients, P50 and P35, were tested for mutation in the GNAS1 gene, that are found in McCune Albright syndrome (reviewed in Weinstein et al., 2004). Although these proved negative, however, this finding does not exclude this syndrome, which can arise also from mosaicism. Biopsy of affected tissues, such as skin hyper-pigmentations, bones and hormone-producing glands was not performed so far.
The possibility of segmental NF1 is still under consideration in four of these patients (P33, P81, P53, and P35), and here verification can be reached by testing the affected tissue.
In patient P20 only OPG was present and in patient P71 lipomatosis was indicated.
Since some families have recently been described with a very similar phenotype to NF1 with mutations in the SPRED1 gene (15q14) (Brems et al., 2007; Messiaen et al., 2009), screening of NF1 negative patients for SPRED1 mutations has already been instituted. A preliminary result shows the presence of a nonsense SPRED1 mutation in P62, thus in this patient Legius syndrome can be diagnosed.
Diagnosis of NF1 is based on clinical assessment and the presence of two or more of the seven NF1 NIH diagnostic criteria. In our sample, we could have excluded 16 cases from genetic testing based on clinical preselection in order to avoid excessive testing. However, this is not recommended, especially in younger patients, in whom not many NF1 clinical features are currently present, and therefore only genetic testing enables definitive diagnosis.
In summary, herein we present an effective mutation detection scheme for NF1 in Slovakia that enabled us to identify disease causing mutations in 90.7% of patients fulfilling NIH criteria, among which 39 mutations were novel. Although our Slovak patients have a similar general mutation distribution pattern to patients in other countries, we noted a tendency here to an increased proportion of small deletions/insertions and a decreased number of nonsense NF1 mutations.
First, we would like to thank Katharina Wimmer (Division of Human Genetics, Medical University Innsbruck, Innsbruck, Austria) and Ludwine Messiaen (Medical Genomics Laboratory, Department of Genetics, University of Alabama in Birmingham) for their help, advice and support. We also acknowledge all the paediatricians and especially the Head of the Department of Medical Genetics of the National Cancer Institute, Dr. Lucia Copakova, for enabling us to cultivate our cells in their facilities. This contribution is the result of projects implemented in “Diagnostics of socially important disorders in Slovakia, based on modern biotechnologies” ITMS 26240220058, “Creating a Competitive Centre for research and development in the field of molecular medicine” ITMS 26240220071, supported by the Research & Developmental Operational Programme funded by the ERDF, and also by VEGA AGENCY grant No. 2/0104/10.
Conflicts of Interest
All authors declare that there is no competing financial interest in relation to the work described.