Neurofibromatosis type 1 (NF1; MIM# 613113) is an autosomal dominant inherited disease, with an incidence of 1 in 3000, caused by mutations of the NF1 gene at 17q11.2. In 95% of non-founder NF1 patients, NF1 gene mutations are identified when a comprehensive NF1 mutation analysis is applied, including an RNA-based core assay supplemented with methods to identify NF1 microdeletions [1]. The proportion of patients with large deletions (termed NF1 microdeletions) that encompass the entire NF1 gene and its flanking regions among all patients with NF1 is 5–10 % [2].

NF1 is a tumour predisposition syndrome characterised by tumours of the peripheral nerve sheaths including the pathognomonic neurofibromas. Cutaneous or dermal neurofibromas (DNF) usually grow during puberty or early adulthood at the end of single peripheral nerves and form small round tumours on the skin which never become malignant. In contrast to DNF, plexiform neurofibromas (PNF) grow along large nerve trunks involving several nerve bundles and mostly represent much larger and more complex tumours than DNF. PNF are usually congenital [3], can grow continuously and may cause organ compression, neurologic impairment and motor dysfunction. At least 10% of all PNF transform into malignant peripheral nerve sheath tumours (MPNST) which are the major cause of NF1-associated mortality [4].

NF1 is associated with considerable inter- and intra-familial variability in phenotypic expression. Nevertheless, the familial aggregation of specific symptoms suggests the influence of a strong genetic component unrelated to the constitutional NF1 mutation [5, 6]. One of the phenotypic traits with the highest estimated heritability in NF1 is the number of PNF, suggesting that one or more modifier genes might influence PNF susceptibility [6]. Recently, a single nucleotide polymorphism (SNP) rs2151280, located within the non-coding RNA gene ANRIL at 9p21.3, has been identified as being associated with the number of PNF in a family-based association study [7]. ANRIL (antisense non-coding RNA in the I NK4 locus) is transcribed in the antisense orientation to the CDKN2A/ARF and CDKN2B genes (Figure 1) and is known to influence their expression [8–10]. CDKN2A/ARF and CDKN2B are three tumour suppressor genes which play a central role in cell cycle inhibition, senescence and stress-induced apoptosis [11]. Importantly, homozygous deletion or expression silencing of these genes has been observed in a subset of PNF, atypical neurofibromas (considered as premalignant tumours) and MPNSTs indicative of their role during the malignant progression of peripheral nerve sheath tumours [7, 12, 13]. However, not only the malignant progression of PNF but also their formation may be influenced by genes at 9p21.3. This conclusion has been drawn from the observed association between the number of PNF in NF1 families and SNP rs2151280 located within the ANRIL gene. The T-allele of rs2151280 has been found to be associated with a higher number of PNF [7]. These authors investigated a total of 1105 individuals (740 NF1 patients and 365 unaffected relatives from 306 French NF1 families). It is however unclear how the number of PNF was assessed in these 740 NF1 patients. Whilst PNF can be externally visible tumours, they may also present as internal asymptomatic tumours which are not detectable by physical examination. Hence, the accurate and reliable detection of all PNF in a given patient with NF1 requires whole-body magnetic resonance imaging (MRI) [14]. In this study, we analysed 29 patients with non-mosaic NF1 microdeletions. The number of PNF as well as the total PNF volume exhibited by these patients has been thoroughly analysed by whole-body MRI and volumetric analysis in a previous study [15]. The 29 patients with NF1 microdeletions analysed here presented with different numbers of PNF (ranging from 0 to 5 PNF per patient). The purpose of our study was to examine if there was an association between ANRIL SNP rs2151280 and the number or volume of PNF observed in patients with NF1 microdeletions.

In the present study, we investigated the potential association between the T-allele of SNP rs2151280 and both the number and volume of the PNF exhibited by 29 patients with non-mosaic NF1 microdeletions. These patients harboured a total of 52 PNF > 3 cm as determined previously by whole-body magnetic resonance imaging (MRI) [15]. Tumours smaller than 3 cm in diameter were not included in the analysis because these tumours cannot be evaluated by volumetric analysis. At least 13 (25%) of 52 tumours analysed were not visible by external physical examination but were detected exclusively by MRI.

We used the Cochran-Armitage test for trends to investigate a potential association between the number of PNF observed in the 29 patients with NF1 microdeletions and SNP rs2151280 . By means of this test, an association between variable (a) with two categories and variable (b) with >2 categories was investigated. In the context of this study, variable (a) represents the genotype of SNP rs2151280 observed in the 29 patients. Variable (a) has been classified into two categories: category #1 represents the genotypes homozygous for the C-allele whereas category #2 represents the CT and TT genotypes of SNP rs2151280. Variable (b) represents the number of PNF observed, with six distinct categories being classified, ranging from category #1 (no PNF), category #2 (one PNF), category #3 (two PNF) etc. up to category #6 (five PNF) (Table 1). Within each category of variable (b), the frequency of categories #1 and #2 of variable (a) was investigated. In other words, all 29 patients were classified into categories depending upon the number of PNF observed. Subsequently, the frequency of the CC genotype vs. the CT and TT genotypes was assessed for each category of patient distinguishable by PNF number. A significant association between the PNF number and the frequency of the CC vs. CT/TT genotypes was not observed (P = 0.20, exact Cochran-Armitage test for trends).

The number of PNF in the 10 patients with a CC genotype ranged from 0 to 4 tumours with a mean value of 1.2 PNF per patient. By contrast, in the 19 patients with the genotype CT or TT, the number of PNF ranged from 0 to 5 with a mean value of 2.1. However, the observed difference between these groups of patients (those with a CC genotype and those with a CT or TT genotype) did not attain statistical significance (P = 0.21, exact Mann-Whitney-Wilcoxon test). Although PNF are mostly congenital tumours and hence the age of the patients investigated is not considered to be critical, we included an adjustment for age in our comparisons. Again, the difference in the PNF number observed in both patient groups was not found to be significant (P = 0.16, with an adjustment for age).

We also investigated a putative association between the tumour volume normalized against body weight and the rs2151280 genotype (CC vs. CT and TT genotypes) in the 29 NF1 microdeletion patients. In the group of patients with the CC genotype, the mean tumour vol-ume was 5.1 ml/kg whereas the median tumour volume was 0.52 ml/kg (95% CI: 0.00 - 9.30). In the 19 patients with CT or TT genotypes, the mean and median tumour volume were 19.8 ml/kg and 2.05 ml/kg (95% CI: 0.15 - 29.80), respectively. Although both groups of patients differed considering the median tumour volume, the confidence intervals overlap to a large extend. A significant difference in tumour volume was not detected comparing both groups of patients (P = 0.52, exact Mann–Whitney–Wilcoxon test; P = 0.19 with adjustment for age). We also did not observe a significant correlation between the total tumour volume (not normalized against body weight) or the number of PNF and the age of patients (ρ = 0.22; P < 0.24; ρ = 0.11; P < 0.58). By contrast, a correlation between the total tumour volume and the number of tumours was observed (ρ = 0.87; P < 0.001).

The chromosome 9p21.3 region harbours a cluster of important growth regulatory genes (CDKN2A/ARF and CDKN2B) that are deleted or transcriptionally silenced in a wide range of tumours such as plexiform neurofibromas (PNF) [12, 22]. The proteins encoded by the CDKN2A/CDKN2B genes act as inhibitors of the CDK4/6 cyclin-dependent kinases, thereby regulating the growth suppressive activity of the RB family of proteins. By contrast, the ARF protein binds to and inhibits the oncoprotein MDM2 which activates p53 [11]. The expression of CDKN2A, ARF and CDKN2B is very low in both young and non-neoplastic cells but increases during cell aging and oncogene-induced hyperproliferation, suggesting that the coordinated expression of these genes is a means to regulate senescence and prevent oncogene-driven hyperproliferation [11]. The polycomb repressive complexes PRC1 and PRC2 have been shown to initiate and maintain the silenced state of the CDKN2A/ARF, CDKN2B gene cluster [23, 24]. PRC1 and PRC2 are recruited to these loci by the 3.8-kb non-coding RNA ANRIL in order to regulate their expression [8, 10, 25, 26].

In a family-based association study, Pasmant et al. [7] investigated a total of five tag SNPs located at 9p21.3 in 1105 individuals (740 NF1 patients and 365 unaffected individuals from 306 NF1 families) and observed a significant association between the number of PNF and one of these five SNPs, rs2151280. This SNP, located within intron 3 of the ANRIL gene, was found to be associated with the number of PNF under a dominant model, with preferential transmission of the derived T-allele to those NF1 patients possessing a higher number of PNF. By contrast, the number of dermal neurofibromas (DNF) was not found to be associated with rs2151280. Importantly, the T-allele of rs2151280 is associated with a reduced ANRIL expression level suggesting either a functional role for SNP rs2151280 or that this SNP is in linkage disequilibrium with an additional as yet unknown functional variant which influences ANRIL expression [7]. Taken together, these findings suggested that modulation of ANRIL expression mediates PNF susceptibility in patients with NF1. It is unclear how many patients with NF1 microdeletions were included in the study of Pasmant et al. [7]. However, only 5% of patients with NF1 exhibit NF1 microdeletions and familial cases are very rare.

In this study, we investigated a putative association between the number or volume of PNF and rs2151280 in 29 patients with non-mosaic NF1 microdeletions. These patients were extremely well characterized by whole-body MRI. We did not observe an association between the T-allele of rs2151280 and either PNF number or PNF volume in these patients, suggesting that this SNP does not exert a strong effect on PNF susceptibility in this group of NF1 microdeletion patients. However, we cannot rule out the possibility of a weak association that might have remained undetected owing to the small number of patients investigated. Under the assumption of an ordered categorical distribution, we estimated that it would have been necessary to analyze approximately 300 NF1 patients to detect a significant association between tumour volume and the T-allele with a power of 80% using the Mann–Whitney–Wilcoxon test (α=5%). This estimation is however based on the observations we made in the 29 patients and implies that the distribution of tumour volumes observed is representative for the whole population of NF1 microdeletion patients. Since NF1 microdeletions are rare (occurring with an estimated prevalence of 1:70,000), the whole-body MRI investigation of 300 patients with NF1 microdeletions is scarcely feasible. As deduced from the data obtained from the analysis of the 29 NF1 microdeletion patients, a strong association between the T-allele of SNP rs2151280 and the PNF load is not obvious.

Patients with NF1 microdeletions have been reported to exhibit a more severe clinical phenotype than patients with intragenic NF1 mutations, as evidenced by an increased risk of MPNSTs, severe learning disability, cognitive impairment, developmental delay and dysmorphic facial features [16, 27–30]. However, the number of PNF, as determined by whole-body MRI, was not found to differ significantly between patients with NF1 microdeletions as a group and NF1 patients lacking large NF1 deletions [15]. Nevertheless, differences in PNF development and biology may well exist between both patient groups i.e. those with NF1 microdeletions and those with intragenic NF1 mutations. The most common type of NF1 microdeletion encompasses 1.4-Mb (termed type-1 NF1 deletion) and is associated with the loss of 14 protein-coding genes inclusive of the NF1 gene [31–33]. Potentially, the loss of one or several of the genes located within the NF1 microdeletion region in addition to the deletion of the NF1 gene, may influence tumour biology or progression. A good candidate for such a modifier gene influencing tumour development is SUZ12 which is located within the 1.4-Mb NF1 microdeletion region. One allele of SUZ12 is deleted in all patients investigated in our study. The SUZ12 protein is an essential component of the polycomb repressive complex 2 (PRC2) and somatic mutations as well as deletions of SUZ12 have recently been identified in various haematological malignancies indicating an important role for chromatin modifiers (such as the polycomb repressive complexes) in tumorigenesis [34–39]. Remarkably, the polycomb repressive complexes 1 and 2 (PRC1 and PRC2) have also been shown to regulate the expression of the CDKN2A/ARF and CDKN2B genes. ANRIL directly binds to SUZ12, an essential component of PRC2 and is required for SUZ12 occupancy of the CDKN2B locus as well as for the epigenetic silencing of CDKN2B[8].

The loss of one SUZ12 allele in patients with germline NF1 microdeletions may well influence ANRIL-mediated expression regulation of the CDKN2A/CDKN2B tumour suppressor genes. Although somatic inactivation of the NF1 wild-type allele is considered to be the PNF-initiating event in NF1 patients with intragenic mutations and patients with NF1 microdeletions [40], both patient groups may differ with regard to tumour progression because of the heterozygous constitutional deletion of SUZ12 present only in patients with NF1 microdeletions. Consistent with this hypothesis, an extremely high total PNF volume (>3,000 ml) was noted significantly more frequently in patients with NF1 microdeletions than in NF1 patients without large deletions [15].

Our findings in the present study suggest that the putative modulation of ANRIL expression by the T-allele of SNP rs2151280 does not influence PNF susceptibility in patients with NF1 microdeletions. Further studies are however needed in order to investigate possible differences in PNF development or susceptibility in NF1 patients with and without NF1 microdeletions.