Non-Coding RNA and Tumor Development in Neurofibromatosis Type 1: ANRIL Rs2151280 Is Associated with Optic Glioma Development and a Mild Phenotype in Neurofibromatosis Type 1 Patients

Non-coding RNAs (ncRNAs) are known to regulate gene expression at the transcriptional and post-transcriptional levels, chromatin remodeling, and signal transduction. The identification of different species of ncRNAs, microRNAs (miRNAs), circular RNAs (circRNAs), and long ncRNAs (lncRNAs)—and in some cases, their combined regulatory function on specific target genes—may help to elucidate their role in biological processes. NcRNAs’ deregulation has an impact on the impairment of physiological programs, driving cells in cancer development. We here carried out a review of literature concerning the implication of ncRNAs on tumor development in neurofibromatosis type 1 (NF1), an inherited tumor predisposition syndrome. A number of miRNAs and a lncRNA has been implicated in NF1-associated tumors, such as malignant peripheral nerve sheath tumors (MPNSTs) and astrocytoma, as well as in the pathognomonic neurofibromas. Some authors reported that the lncRNA ANRIL was deregulated in the blood of NF1 patients with plexiform neurofibromas (PNFs), even if its role should be further elucidated. We here provided original data concerning the association of a specific genotype about ANRIL rs2151280 with the presence of optic gliomas and a mild expression of the NF1 phenotype. We also detected the LOH of ANRIL in different tumors from NF1 patients, supporting the involvement of ANRIL in some NF1-associated tumors. Our results suggest that ANRIL rs2151280 may be a potential diagnostic and prognostic marker, addressing early diagnosis of optic glioma and predicting the phenotype severity in NF1 patients.


Introduction
In recent years, an increasing number of functions for small non-coding RNAs (sncRNAs) has been identified. They are key molecules mediating fundamental cellular processes, such as transcription, post-transcriptional modifications, chromatin remodeling, and signal transduction. Dysfunction in an

Review of Literature
The review of literature was carried out by considering the main relevant original works and reviews in the field concerning ncRNAs involved in tumor development in neurofibromatosis type 1, and distinguishing their implications for pathognomonic neurofibromas and specific tumors associated with the disease (MPNST, pilocytic astrocytoma, optic glioma).

Human Subjects
The investigations were carried out following the rules of the Declaration of Helsinki of 1975, revised in 2013. This study is part of a research grant approved by the Italian Health Minister (RFINN-2008-1204836), and according to the Italian regulations, it did not require specific ethical approval before undertaking the research, because it only uses anonymous data collected during routine patient care. The results were submitted and approved by ethical committee of Regione Lombardia, a section of C. Besta Neurological Institute Foundation.
Written informed consent was given by both the patients and subjects recruited as unaffected controls. Additional informed consent was obtained from all of the subjects who may be identified on the bases of the information contained in this article.

Patients and Phenotypic Data
One hundred and four unrelated, consecutive patients with NF1, on whom we performed molecular NF1 genetic testing at our institute between January 2013 and September 2014, were included in the study. In all, the diagnoses fulfilled National Institutes of Health (NIH) diagnostic criteria, and in 100 of them a pathogenic NF1 mutation was identified. The four patients in which NF1 mutation was not found were all adults with a certain NF1 diagnosis. The lack of NF1 mutation identification is expected in about 5% of patients; in these cases, the mutation is hypothesized to localize in one of the NF1 introns [15]. The median age was 35 . Eligible patients were identified using an electronic patient database search request. All medical records were surveyed, and the following data were collected at the time of mutation analysis and re-verified for accuracy at the time of this study: data of birth; gender; age at the time of genetic testing; mode of inheritance; and NF1 signs and symptoms, including visual impairment, pain, epilepsy, cognitive impairment, plexiform and spinal neurofibromas, optic nerve glioma, and other neoplasms of the central nervous system and other organs. Two different medical severity scales [16,17] were used to assess overall disease severity and to rate the visibility features of the disease (Tables S1 and S2). All patients underwent brain and spinal MRI (magnetic resonance imaging) scans. NF1 optic pathway glioma (OPG) cases were defined as patients diagnosed with NF1 and an OPG, confirmed by MRI scan. Spinal neurofibroma NF1 cases were defined as patients with at least one neurofibroma along the spinal root.

DNA and RNA Extraction
Constitutional DNA and total RNA extracted from peripheral blood samples were conserved in ethylenediaminetetraacetic acid (EDTA). Constitutional DNA was extracted using a NucleoSpin Blood Kit (Macherey-Nagel, Duren, Germany) in accordance with the manufacturer's instructions and stored at a temperature of −20 • C; RNA was extracted using an RNeasy Mini Kit (Qiagen, Hilden, Germany) and stored at a temperature of −80 • C. DNA from tumor samples was extracted by means of Trizol reagents (Thermo Fisher, Waltham, MA, United States) following manufacturer's instructions.

Genotyping of Single Nucleotide Polymorphism Rs2151280
The genotype of the single nucleotide polymorphism (SNP) rs2151280, located within intron 3 of ANRIL, was investigated by PCR amplification of the genomic region and subsequent sequence analysis of PCR fragments, amplified from blood-derived DNA of the 104 NF1 patients. PCR was performed by GoTaq DNA Polymerase (Promega, Fitchburg, WI, United States) following manufacturer's instructions, with the primer shown in Table S3.

Sequencing Analysis
The PCR products were bi-directionally sequenced using the Terminator v3.1 Cycle Sequencing Kit (Thermo Fisher, Waltham, MA, United States), and resolved on an automated ABI-3130xl DNA genetic analyser (Thermo Fisher, Waltham, MA, United States). The output data were analysed using SeqScape software v2.5 (Thermo Fisher, Waltham, MA, United States).

Reverse Transcription (RT) and Quantitative Real-Time PCR (qPCR)
One microgram of total RNA was reverse transcribed using the iScriptTM cDNA Synthesis Kit (Bio-Rad Laboratories Inc., Berkeley, CA, United States). The specific oligonucleotides used to amplify the CDKN2A (cyclin dependent kinase inhibitor 2a), CDKN2B (cyclin dependent kinase inhibitor 2b), ARF, and ANRIL transcripts are shown in Table S3; TBP (TATA-box binding protein) was used as a housekeeping control. The qPCR assays were made using GoTaq-qPCR master mix (Promega) and run on an iQ5 Real-Time ThermalCycler (Bio-Rad Laboratories Inc., Berkeley, CA, United States).

LOH Analysis
Microsatellite analysis was performed on genomic DNA extracted from peripheral blood mononuclear cells (PBMCs) and from tumor samples, amplified by means of PCR using a carboxyfluorescein-or hexachlorofluorescein-labelled forward primer for each marker (Sigma Aldrich, St. Louis, MO, United States) and GoTaq DNA Polymerase (Promega, Fitchburg, WI, United States) following reported procedures [18]. The DNA fragments were separated by means of capillary electrophoresis (ABI 3130XL, Thermo Fisher, Waltham, MA, United States) and analyzed using Genemapper software (version 3.1, Thermo Fisher, Waltham, MA, United States). Oligonucleotide sequences are reported in Table S3.

Statistical Analysis
Descriptive and frequency statistical analyses were carried out, and comparisons were made using the SPSS 17.0 software. Chi-square or Fisher exact tests were used to examine the differences in categorical variables between groups. Genotype-phenotype correlations were studied using a multiple logistic regression model odds ratio (OR), and 95% confidence intervals were calculated when there was a significant effect. A p value (p) < 0.05 was considered as statistically significant. The resulting p values were adjusted for multiple comparisons using the Benjamin-Hochberg (B_H) procedure, with a false discovery rate of 0.2.

Non-Coding RNA and Tumor Susceptibility in Neurofibromatosis Type 1
Some studies report on lncRNA and miRNA network involvement in different molecular mechanisms of gene expression modulation, associated with tumor development and progression in NF1 patients [1].
Two miRNAs have been directly implicated in MPNST development by miRNA expression profiling studies on peripheral nerve sheath tumor samples: miR-34a and miR-214. miR-34a is a tumor suppressor downregulated in MPNSTs, as shown by in vitro studies using the cell lines MPNST-14 (NF1 mutant) and MPNST-724 (from a non-NF1 individual). The results showed that exogenous expression of p53 or miR-34a promotes apoptotic cell death, and exogenous expression of p53 in MPNST cells induces miR-34a and other miRNAs; therefore, p53 inactivation and the following loss of expression of miR-34a may significantly contribute to MPNST development. As a result, the authors propose this miRNA as a candidate therapeutic treatment target in MPNSTs [40]. As it is the most overexpressed miRNA in MPNSTs, miR-34a is an oncogene. Other studies performed on mouse neural cells have shown that TWIST1 (twist family BHLH transcription factor 1), a regulator of metastasis highly expressed in the majority of MPNSTs, induces miR-214 expression [41].
Gong et al., performing an analysis of miRNA expression by microarray and qRT-PCR, found in NF1 and non-NF1 MPNST tumor tissues, and in tumor cell lines, the downregulation of miR-204, which is located at a cancer-associated genomic region showing a high frequency of LOH in tumors. The authors demonstrated that by restoring miR-204 expression, cellular proliferation, migration, and invasion were decreased in NF1 and non-NF1 MPNST cell lines in vitro, and tumor growth and malignant progression were reduced in non-NF1MPNST cell lines in vivo. These findings support the hypothesis that miR-204 is a tumor suppressor involved in MPNST tumorigenesis and progression, and may represent a novel biomarker for diagnosis and a possible therapeutic treatment target in MPNSTs [42].
Another miRNA that plays an important role in MPNST development is miR-21, identified as an oncogene overexpressed in MPNSTs having PDCD4 (programmed cell death protein 4) as the target gene. Itani et al. demonstrated that miR-21 expression level was significantly higher in MPNSTs than that in neurofibromas (p < 0.05), applying miRNA expression profile analysis and quantitative real-time reverse transcription PCR on MPNSTs, neurofibromas, and normal nerves in clinical samples and on three MPNST cell lines. In addition, the authors demonstrated the important role of miR-21 in MPNST progression via transfection of an miR-21 inhibitor in one MPNST cell line. These cells underwent apoptosis, indicating that miR-21 and its target PDCD4 may be candidate therapeutic targets against MPNST progression [43].
NF1 patients also have a high risk of developing gliomas [45], which are among the most common types of brain cancer and arise from the supportive ("gluey") tissue of the brain-precisely, from glial cells. A glioma is another tumor in which altered miRNA expression plays an important role, through the regulation of tumorigenic processes, such as receptor tyrosine kinase signaling, suppression of differentiation, cell cycle stimulation, apoptosis inhibition, invasion, and angiogenesis. The most frequent form of glioma in the pediatric age is the pilocytic astrocytoma (PA), a low-grade neoplasm that occurs frequently in children and young adults, whose tumorigenesis has been associated with BRAF (B-Raf proto-oncogene, serine/threonine kinase) mutations and fusion, but also with miRNA action. Ho et al., investigating different PA groups by the analysis of differentially expressed miRNAs in NF1-associated tumors versus tumors with BRAF alterations, found four miRNAs differentially expressed between NF1-associated and BRAF fusion-positive tumors: in NF1 tumors, hsa-miR-650 and hsa-miR-1276 were overexpressed, whereas hsa-miR-744* and hsa-miR-187* were under-expressed [46]. In addition, Darrigo et al. revealed a subset of 30 under-expressed miRNAs in PA samples of patients with NF1, evaluating global miRNA expression in 30 micro-dissected samples, including pediatric PAs, NF1-associated PAs, and cerebral white matter by the microarray method. These different miRNA profiles of NF1-associated PAs were previously related with known deregulated pathways in cancer, such as the cell cycle and hippo pathways, and suggest a distinct tumorigenesis process associated with miRNA dynamics in this PA subgroup [47]. Table S4 summarizes the reported expression level of miRNAs analyzed in tumors developed by NF1 patients.

Non-Coding RNA and Neurofibroma Development
Little has been reported about a possible role of miRNAs on neurofibroma development. After a study aimed at comprehensively characterizing the expression pattern of 377 miRNAs in NF1-related neurofibromas, Masliah-Planchon et al. demonstrated that miR-486-3p was the most significantly upregulated miRNA in PNFs. Interestingly, PTEN, a tumor suppressor gene, is a target of miR-486-3p. Accordingly, the authors hypothesized that in PNF development, miR-486-3p may be a major onco-miR, downregulating PTEN. Aberrant expression of miRNAs involved in the RAS-MAPK pathway, such as miR-370, miR-143, miR-181a, and miR-145, has been also reported. Further studies should be performed to assess the implication of these miRNAs in neurofibroma generation [21].
As far as the long non-coding RNAs, the only one currently related to tumorigenesis in patients affected by NF1 is ANRIL, which seems to be associated with PNF development [13]. PNFs are protruding and deforming masses that arise from multiple nerves, and also involve connective tissue and skin folds, displayed by 30% of NF1 patients [48]. ANRIL is an anti-sense, long-non-coding RNA expressed by CDKN2B-AS1 in the INK4 locus, located on chromosome 9p21 (Figure 1a). ANRIL is involved in the expression regulation of three tumor suppressor genes at the CDKN2A/B locus (p16-CDKN2A, p15-CDKN2B, and p14-ARF), via a cis-acting, polycomb-mediated epigenetic mechanism. This lncRNA stabilizes the polycomb repressive complexes 1 and 2 (PRC1 and PRC2), acting as a recruiting platform for their specific subunits (CBX7 of PRC1 and SUZ12 of PRC2). The direct interaction between ANRIL and PRC1/PRC2 allows these molecular complexes to modify the histone code of the CDKN2A/B locus, maintaining the transcriptional repression [49,50]. PRC1 and PRC2 catalyse the mono-ubiquitination of H2A on K119 and the methylation of histone H3 lysine 27, respectively [51] (Figure 1b).
As far as the molecular characterization of ANRIL, both SNPs and structural alterations of this gene have been reported. Multiple SNPs in the 9p21.3 locus were associated with risk for numerous diseases, including diabetes, stroke, coronary heart disease, melanoma, and glioma [60], while structural alterations of ANRIL, such as deletions and translocations, were identified in neurofibromas [13] and melanomas [61].
Pasmant and colleagues identified a recurrent somatic alteration in PNFs, the 9p21.3 deletion (including the CDKN2A/B-ANRIL locus), in 6 out of 22 PNF tumors from 18 NF1 patients, analysed by a genome-wide array of comparative genomic hybridization. A family-based association test (FBAT) by SNP analysis of 9p21.3 region was performed, in order to assess the role of this chromosome locus in the onset and number of PNFs in NF1 patients. The FBAT study revealed a statistically significant association of the rs2151280 SNP T allele, located within the ANRIL intron 3, with the onset of a higher number of PNFs (p < 0.001) in NF1 patients, according to a dominant model. Furthermore, after CDKN2A, CDKN2B, ARF, and ANRIL expression analysis in 124 NF1 patients' peripheral blood, the authors found that the rs2151280 T allele was significantly associated with reduced ANRIL transcript levels (p < 0.001 by Kruskal-Wallis test), suggesting a functional role of this SNP in the modulation of ANRIL expression that could promote PNF susceptibility in NF1 patients [13].
Another study carried out on a cohort of 29 NF1 microdeletion patients showed that neither the PNF number nor PNF volume were correlated with the rs2151280 T allele. Differently, in these patients PNF susceptibility was associated with the loss of one allele of SUZ12 coding for the PRC2 subunit that interacts directly with ANRIL. As SUZ12 is included in the 1.4 Mb NF1 microdeletion region, and because all patients enrolled for the study have only one copy of SUZ12, these authors hypothesize that the heterozygous constitutional deletion of this gene may influence the ANRIL-mediated expression regulation of the CDKN2A/CDKN2B tumor suppressor genes [62].

The ANRIL Rs2151280 is a Susceptibility Marker for Optic Glioma Development and Mild Phenotype in NF1 Patients
Even if the implication of lnc-RNA ANRIL in PNF development has not yet been assessed, different findings indicate that this issue should be further investigated, a challenge that could provide useful diagnostic markers or pharmacological targets.
To unravel this matter, we studied the occurrence of specific rs2151280 genotypes in a cohort of 104 NF1 patients, enrolled at the C. Besta Neurologic Institute, Milan, Italy. Furthermore, we determined the CDKN2A, CDKN2B, ARF, and ANRIL expression levels in the PBMCs of NF1 patients with a PNF. Table 1 summarises the clinical and molecular features of our cohort, while clinical details are illustrated in Table S1.
Because ANRIL is considered a susceptibility locus for several cancers, including glioma [63], we also studied a possible association of the rs2151280 genotypes or the expression level of the above genes with the development of other tumors often associated with NF1, such as optic glioma (Table S1). The obtained results have been correlated with the types of tumors developed, with Riccardi and Ablon scores [16,17] (Table 1 and Tables S1 and S2).
We analyzed the frequency of rs2151280 alleles/genotypes in our cohort by comparing them with those of the healthy population, also considering the allele or genotype distribution within NF1 patient subgroups showing DNFs, PNFs, OPGs, or other tumors (e.g., astrocytomas and MPNSTs). After the constitutional DNA analysis, a significantly different genotype distribution (0.2 > p > 0.1) was observed between the OPG subgroup of NF1 patients and the European population (1000GENOMES Phase3-V1 EUR population build 14; Table 2). Interestingly, if we consider the genotype frequency within our cohort of NF1 patients, those patients with a CC genotype had a threefold higher risk of OPG development versus patients with other rs2151280 genotypes (i.e., CT; TT). The CC genotype was detected in 41.7% of patients with an OPG, and in 17.5% of patients without (p = 0.014; OR: 3.4; 95% CI, 1.1-10.1). No difference was observed for genotypes CT and TT distribution between patients with and without OPGs (CT: 41.7 vs. 63.7, respectively; TT: 16.7 vs. 18.8, respectively). No significant effect for sex or age was found in the regression analysis. OPG patients with a CC genotype show a wide spectrum of NF1 mutations, which was also true if we considered the whole OPG group, suggesting an absence of correlation between a specific NF1 mutation and OPG onset in these patients.

SAMPLES f(C) f(T) f(CC) f(CT) f(TT)
Complete casuistry (n = 104) 0. When Riccardi and Ablon scales were considered, only patients older than 19 years were included in the analysis, due to age-dependent diseases. A significant association between genotype TT and the presence of a mild disease was detected: the TT genotype was detected in 50% patients with Ablon 1, and in 15% with a higher Ablon scale (p = 0.034; OR 5.67; 95% CI, 0.90-34.08).
We also investigated the expression of ANRIL, and its target genes CDKN2A, CDKN2B, and ARF, on PBMC RNAs from 65 patients of our NF1 cohort. As the expression of CDKN2A in the peripheral blood was very low in all samples and in the unaffected controls, it has not been reported in the results shown below. The cohort was divided into four subgroups according to the type of tumor developed by each patient, and was compared to five unrelated controls. Patients who developed multiple types of tumors were included in different sub-categories (Figure 2a-c). In general, patients with a PNF showed considerable variability, as well as those with a DNF for ARF (Figure 2b). The average expression levels of CDKN2B, ARF, and ANRIL were generally lower in NF1 patients, with the exception of PNF patients, with respect to the controls ( Figure S1). As the T-allele of rs2151280 was associated with a reduced level of ANRIL expression in PBMCs, each group of patients was furtherly subdivided according to the genotype of rs2151280 SNP (i.e., homozygous CC, TT, or heterozygous CT). We found that patients with an OPG carrying the TT genotype showed reduced ANRIL expression compared to the other subgroup of genotypes (CT and CC) (Figure 3), confirming the association previously reported [12]. Interestingly, most OPG patients carrying a CC or CT genotype did not show decreased ANRIL expression in the blood. These results are consistent with the biological effect of ANRIL expression in glioma cell lines [14]. ANRIL was found to be upregulated in glioma cell lines, and its silencing led to the inhibition of cell proliferation, migration, and invasion. ANRIL downregulation has an effect on the upregulation of miR-34a that, targeting Sirt1, mediates that downregulation, leading to the inactivation of the PI3K/AKT and mTOR pathways [14]. Furthermore, the silencing or homozygous deletion of CDKN2A, ARF, and CDKN2B has been detected in a subset of PNF, atypical neurofibromas, as well as MPNSTs, indicating that these genes have a role not only in the formation of PNFs, but probably also in MPNST development [13,64]. Nevertheless, this association was not observed in patients with NF1 microdeletion [62]. These data are consistent with our results on NF1 and the INK4b/ARF/INK4a locus LOH studies ( Figure S2). The occurrence of LOH of both the NF1 and INK4b/ARF/INK4a regions has been investigated in tumor specimens of a small subset of our NF1 patients, and was reported for the first time in DNF, PNF, and astrocytoma, indicating that the genes included in the deletion interval may have a role in the tumorigenesis progression. Accordingly, the INK4b/ARF/INK4a locus is deleted or downregulated in about 40% of human cancers, mainly related to the tumor suppressive function of CDKN2A and CDKN2B [65]; conversely, ANRIL itself shows a pro-oncogenic activity [63].
Finally, the association between the rs2151280 TT genotype and mild NF1 phenotypes supports the hypothesis of a prognostic significance of this ANRIL polymorphism being associated with a decreased ANRIL expression in blood [66].

Conclusions
This review points out how research concerning the role of ncRNAs in NF1 pathogenesis is still only beginning, and aims to promote interest in this field, starting from the current knowledge, in order to open new perspectives on diagnosis and therapies. In the last decade, RNA-based technologies have allowed new approaches to set up early diagnosis, follow-up, and therapeutic strategies, which would positively impact the quality of cures and life of affected patients in the near future. From this view, is crucial to provide new insights unravelling the complex networks of ncRNAs to develop advanced technologies for treatment and diagnosis of different diseases.
In the era of precision medicine, when dealing with a cancer-prone syndrome like NF1, new technologies, such as ncRNA-based liquid biopsies, cannot be ignored, and should be developed for screening and early detection of malignant complications. NcRNA-based research may underpin new therapeutic approaches and help in unraveling the issue of the low genotype/phenotype correlation in NF1.
From this perspective, our findings suggest that ANRIL may be a potential prognostic marker in the NF1 disease. The rs2151280 CC genotype is associated with an increased risk of OPG onset in NF1 patients, and the TT genotype is a predictive marker of a mild clinical phenotype. Although OPG generally has an indolent course, compared to non-NF1 cases, and the survival rate is good, it can sometimes cause severe visual impairment. Considering that the usefulness of brain MRI screening for the early diagnosis of optic glioma in children with NF1 is still controversial, the identification of a predictive marker may provide personalized management and follow-up for NF1 patients, with definite positive repercussions for people's quality of life and for communities' health social costs.