Isoform-specific NF1 mRNA levels correlate with disease severity in Neurofibromatosis type 1

Alternative splicing, the mechanism by which eukaryotic cells generate multiple RNAs from a single transcript, maximizes genome plasticity and versatility by promoting diversification of protein function and its spatiotemporal control [1‐3]. In humans, as many as 92–94% of multiexon genes are predicted to undergo alternative splicing [4]. This process is important in the control of developmental programs and cell physiology, as well as in the pathogenesis and progression of human diseases [5]. It has been proposed that alternative splicing contributes to the clinical variability of Mendelian disorders by altering either the level of correctly spliced RNA pools or the ratio of different mRNA isoforms that result from transcript processing [6].

Neurofibromatosis 1 (NF1 [MIM: 162200]) is the most common non-chromosomal disorder affecting development and predisposing to cancer (approx. 1:2000–3000 live births) [7, 8]. It is transmitted as an autosomal dominant trait, and is caused by germline loss-of-function mutations in the NF1 gene, which encodes neurofibromin, a GTPase negatively controlling RAS signaling [9]. Major features of NF1 include café-au-lait macules, skinfold freckling, and Lisch nodules of the iris, cutaneous and/or subcutaneous neurofibromas, variable learning disability/cognitive deficit (LD/CD), skeletal defects, and an increased risk for certain malignancies [10]. The neoplastic risk is related to functional loss of the GTPase activity of neurofibromin due to somatic hits involving NF1, according to the Knudson’s model, causing cell autonomous, and sustained activation of the Ras-mitogen-activated protein kinases (MAPK) pathway.

While NF1 is characterized by complete penetrance, variable expressivity is observed, with marked clinical variability even within families [11]. Phenotypic studies of large cohorts suggest that the type of mutation in the NF1 gene generally does not correlate with the observed phenotypic variation [12, 13]. Exceptions are the constitutional NF1 microdeletions [14] and missense variants affecting codons 844–848, which are associated with a more severe presentation [15], and the missense variants at codon 1809 [16, 17] and the 3-bp in-frame deletion, c.2970_2972del (p.Met992del), which conversely, are associated with a mild, mainly cutaneous, clinical presentation [18, 19]. Overall, the genotype-phenotype correlations identified so far have been reported to explain only a small amount of the extent of clinical variability characterizing the NF1 population [15, 19], and other factors, including stocastic events (e.g., second hits), and the genetic background (variation in modifier genes) are expected to contribute to a much larger fraction of the NF1 phenotipic variability [11]. The evidence that modifier genes contribute to the phenotypic expression of NF1 is strongly supported by familial studies [11, 13, 20, 21], which show that, independently by the NF1 mutation, the grade of phenotypic concordance between members of the same family correlates with the genetic background, and that the relative contribution of the genetic background varies depending from the feature in question [11, 20]. Further evidence that genetic modifiers are major contributor to the variable expression of NF1 comes from studies on animal models showing that Nf1^+/− mice strains have differences in phenotype severity with regards to the learning and behavioral aspects of the phenotype, as well as in the susceptibility to form astrocytomas [22‐25]. Notably, by assessing Nf1 mRNA levels in these models, it was also shown that trans-acting events modulate the phenotypic expression by impacting Nf1 gene expression [26].

NF1 is a large gene and its pre-mRNA undergoes alternative splicing. Several alternative exons that do not alter the reading frame of the gene have been identified, including 9a/9br, 10a-2, and 48a [6]. Of particular interest is exon 23a, which lies within the GAP-related domain (GRD) of neurofibromin, and is predominantly retained in most tissues, but specifically skipped in central nervous system neurons in humans [27, 28]. Of note, the two neurofibromin isoforms including/lacking the short amino acid stretch encoded by this exon differ in their ability to control Ras function [27, 28]. Isoform I, which lacks exon 23a, has ten times higher Ras-GAP activity than isoform II, in which exon 23a is retained. Biological importance of this exon during development has consistently been underlined by the observation that the mouse model in which exon 23a is constitutively deleted has a learning phenotype [24]. These considerations suggest an intriguing hypothesis in which changes in the levels of protein isoforms generated via alternative transcript processing, including alternative splicing, acts as genetic modifier in NF1 patients.

In the present study, we analyzed the levels of neurofibromin isoform I and II in circulating leukocytes of a cohort of genetically and clinically characterized NF1 patients stratified according to the severity of the phenotype, and correlated their expression levels with disease severity to assess whether alternative splicing may contribute to the variable expression characterizing NF1.

Expression levels of NF1 mRNA isoforms I and II were examined in peripheral blood leukocytes of 138 NF1 patients and compared with those of 138 population-, age- and sex-matched healthy controls. The expression value of isoform II was reported either as the sum of the expression value of both isoforms I and II pools and as isoform II data obtained subtracting the expression value of isoform I from the sum of the expression value of both isoforms I and II pools. Levels of NF1 isoforms I, isoform (I + II) and isoform II are reported in Table 2.

NF1 is the result of loss-of-function mutations in the NF1 gene. In this study, more than 75% of the mutations identified lead to the introduction of a premature termination codon in the coding sequence, which is in line with previous findings [12, 31]. Because of the nonsense-mediated RNA decay mechanism, many of these mutations are expected to lead to a reduction in the level of expression of the NF1 transcript [32]. Consistently, we found that NF1 mRNA was expressed at significantly lower levels in the peripheral blood leukocytes of NF1 patients than in healthy subjects, independently from the protein isoform that was considered and from the severity of the phenotype. Further data analysis showed that the neurofibromin isoform with higher GAP activity, isoform I, was expressed at significantly lower levels in subjects with severe phenotype respect to affected subjects with mild/moderate phenotypes, independently of the age. Moreover, when patients were compared based on the presence vs absence of LD/CD, cerebral tumors and cerebrovascular disease, analyses showed that a lower expression level of isoform I was significantly associated with occurrence of LD/CD. Such specific association is of particular relevance since isoform I is predominantly expressed in central nervous system neurons [33], and the finding that in mice, constitutional homozygous deletion of exon 23a (i.e., loss of Nf1 mRNA isoform 1 expression in all tissues) is viable, do not affect development or cause cancer predisposition but results in spatial learning and memory defects [24, 25]. Consistent with the findings collected in mice, we did not observe any significant difference in the expression of the two NF1 isoforms in relation to tumor formation or vascular disease. This is in line with the consideration that cell transformation is expected to require complete loss/functional inactivation of neurofibromin, which is more likely to depend upon somatic hits affecting the wild-type allele rather than events causing aberrant transcript processing. This also applies to vasculopathy and other NF1-related features, including café-au-lait spots or tibial pseudarthrosis, in which the somatic second hit has been detected in the pathologic tissue [34‐36]. On the opposite, learning disability phenotypes are more thought to be related to neurofibromin haploinsufficiency and therefore could be more influenced by the balance between the expression of the two neurofibromin isoforms in the brain [6, 24, 25]. It is important to underline that the association between reduced isoform I expression and learing phenotype was still present when only subjects in pediatric age were considered. This is an extremely significant observation since the LD/CD phenotype has profound implications for the management of the disease, especially in early age, and the identification of predictive markers might be useful for the clinical management of these patients [37]. Inclusion/skipping of NF1 exon 23a is a tightly regulated process during development, depending on the cellular context. This alternative splicing event is under complex control with many regulatory factors involved. Like other alternative exons, also NF1 exon 23a is characterized by the presence of weak consensus sequences surrounding the exon that are not readily recognized by the splicing machinery [6]. Although not identified yet, it is possible that variation involving cis- and/or trans-acting elements controlling/participating in exon 23a retention/skipping could result in the failure of proper NF1 transcript processing, leading to an imbalance in the distribution of the type I and type II isoforms and this in turn to phenotypic consequences in NF1 patients.

It has been widely demonstrated that dysregulation of posttranscriptional regulation, including alternative splicing, results in defective neuronal differentiation and/or synaptic connections, leading to neurodevelopmental and psychiatric disorders [38, 39]. Different genetic and chemical approaches to target components of the spliceosome to correct splicing defects have been investigated in pathological conditions including cancer and neurologic disorders. Advancements in the understanding of NF1-specific defects caused by mis-regulation of alternative splicing might increase the development of specific therapeutic options in NF1 [40‐42].

The present findings provide a first evidence for a role of circuits controlling NF1 transcript processing in modulating phenotypic expressivity in NF1, and document an association between the levels of neurofibromin isoform I mRNA and the severity of phenotype and cognitive impairment. The identification of this association between specific NF1 expression pattern and phenotype variability is remarkable and deserves further exploration. Expression studies at the protein level and in relevant tissues/cell lineages are required steps to validate the present findings.