Neurofibromatosis Type 1: From Cognitive
Profile to Future Therapeutic Targets
Paul L.C. Feyen - 6323391
Supervisor: Dr. Omrani, Azar
Co-Assesor: Dr. Werkman, Taco
Friday October 19
thLiterature Thesis - MsC in Brain and Cognitive
Sciences, Neuroscience Track, University of
Content Overview
Introduction
NF1 Cognitive Phenotype
The NF1 genotype and how this relates to the phenotype
Neurofibromin is expressed as several variants encoded by the NF1
gene
The relationship between mutations of NF1 and NF1 pathogenesis
Molecular and Cellular View on NF1-related Cognitive Problems
Upstream modulators of neurofibromin
Downstream signaling pathways of neurofibromin
Animal Models of NF1: Insights on Etiology
Behavioral tasks assigned to specific brain areas highlight the parallels
of NF1 mice and human phenotype
Increased inhibition and altered dopamine system underlie NF1
cognitive phenotype
Treatment of NF1 in Human Patients
Neurodevelopmental disorders can be treated, even in adulthood
Phase 1 clinical trials highlight the complexity of NF1 pathogenesis
Modulation of ion channel function as potential therapies for NF1
Benefits of a behavioral component in the treatment of NF1
Introduction
With an incidence of 1 in 3000 individuals, Neurofibromatosis type 1 (NF1) is among the most common autosomal dominant diseases (Huson et al, 1989). NF1, also known as von Recklinghause disease, is characterized by the presence of neurofibromas, optic pathway glioma, Lisch nodules, axillary freckling and café-‐au-‐ lait macules (NIH, 1988). NF1 symptomology develops with age, but a diagnosis can be made before the age of six (Huson et al 1989). In both child and adult patients of NF1, cognitive impairments are frequently observed (North et al, 1997; Ozonoff, 1999; Hyman et al, 2005). NF1 patients have a low-‐average IQ (Kayl and Moore, 2000), and display problems in specific cognitive domains like memory, visuospatial skills, language, executive functioning, and attention (Hyman et al 2005). In children with NF1 these deficits are the most common complication affecting quality of life (Hyman et al. 2005). Overall, the cognitive problems and associated learning disabilities pose one of the most significant sources of lifetime morbidity for patients (North et al., 1997). Given the impact which cognitive deficits have on the lives of NF1 patients, this literature study aims to bring together clinical and animal model data to develop a clear view of the molecular underpinnings of cognitive deficits in NF1, in the aim of defining reliable molecular targets for future research and potentially for pharmaceutical therapy.
NF1 Cognitive Phenotype
The cognitive phenotype of NF1 is quite diverse, and is characterized by domain specific cognitive impairment rather than a global mental deficit. This is reflected in the slightly downward shift of IQ in the NF1 population which lies within the normal distribution, as well as in the fact that several other measures sensitive to global cognitive impairments do not distinguish at all between NF1 patients and unaffected siblings (Kayl and Moore 2000). This shift in IQ does however result in a two-‐fold increase for risk of mental retardation (North et al, 1997). On the other hand, more than 80% of individuals affected by NF1 show impairments in specific cognitive domains as assessed by neuropsychological evaluations (Hyman et al, 2005, Krab et al. 2008). There is also a high correlation between NF1 and attention deficit disorder (Hofman et al.1994, Hyman et al. 2005, Mautner et al. 2002). Learning disabilities are diagnosed in up to 65% of individuals affected by NF1 (Rosser & Packer 2003), and individuals with NF1 are fourfold more likely to require special education (Krab et al. 2008a).
The fourth edition of the Diagnostic and Statistical Manual of Mental Disorders (DMS-‐IV) define a specific learning difficulty as an academic achievement that is more than twice the standard deviation on an individual’s IQ, which can occur in one of 5 major categories including reading disorders, mathematical disorders, disorders of written expression, nonverbal learning disorders, and other learning disorders which lead to academic underachievement, However, rather than fitting one such description, different types of learning disabilities have been observed in
individuals with NF1 (Levine et al 2006). Early reports suggested that the cognitive profile of NF1 resembled that of children with non-‐verbal learning disorders (Ozonoff 1999). Not only do NF1 patients show consistently poor performance in tests of visuospatial functioning and spatial learning (North, 2000; Kayl and Moore, 2000), these individuals also present other components of non-‐verbal learning disorders like increased impulsiveness, poor organizational skills and a poor social cue perception (North et al, 1997; North, 2000). Yet NF1 cannot be strictly classified into one of the categories defined in DMS-‐IV. Literacy based learning impairments are also prevalent, as are mathematical ones in NF1 (Descheermaeker et al 2005, Hyman et al, 2006). Interestingly, after controlling for IQ, children with NF1 no longer show a higher incidence of math based learning problems (Kayl and Moore, 2000). On the other hand, deficits in reading and writing have repeatedly been found to be more severe in NF1 patients than predicted from IQ scores, and is associated with specific deficits in expressive and receptive language, vocabulary, visual naming, and phonologic awareness (Hyman et al 2005). This occurrence of multiple learning disorders in NF1 is indicative of a more fundamental impairment to learning (Shilyansky 2010b). Now the aim has shifted to understanding the parse behavioral symptoms as more specific and stable phenotypes, or 'endophenotypes' with a clear anatomical, molecular, and genetic connection. In this light, the hurdles to social and academic performance can be seen to stem from learning disabilities and cognitive impairment in five specific areas, namely, reading/vocabulary, attention, executive function and planning, visuospatial function, and motor deficits (Hyman et al, 2005). These in turn are likely caused by one or more general mechanism/s affecting multiple brain systems (Shilyansky 2010b).
For one, the poor performance of NF1 patients on visuospatial, vocabulary, and reading comprehension suggest a dysfunction of hippocampus-‐dependent memory (Shilyansky 2010b). Damage to hippocampus results in specific and profound cognitive defects. Such patients often have difficulty learning and retrieving new vocabulary (Verfaelli et al, 2000), and they are often reported to have severe deficits in spatial memory and spatial navigation tasks (Burgess et al, 2002; Squire et al 2004). Interestingly, these symptoms are also commonly presented in individuals with NF1. For example, patients display poor performance in tests of visuospatial functioning and spatial learning (Kelly, 2004). As such, abnormal hippocampal function in NF1 patients could contribute to their deficits in learning and memory. Though the parallels are striking and suggest role for this brain area in NF1, dysfunction in this region alone cannot explain all the impaired cognitive functions presented in NF1.
Attention deficit disorders are another cognitive hallmark of NF1 patients, and are diagnosed as a co-‐morbid condition in 30% to 50% of patients (Hofman et al, 1994; Mautner et al, 2002; Hyman et al, 2005; Pavol et al, 2006). The impact of this condition is noticeable not only in NF1 children’s academic performance but also in their social development (Barton and North, 2004). About 40% of NF1 children seen as academic underachievers show no deficit in evaluations of learning disabilities,
and attention deficit disorders may well contribute to their poor academic performance (Kayl and Moore, 2000). NF1 patients also face social problems and are reported to appear socially awkward and withdrawn. These problems with interpersonal skills are thought to stem from a decreased attention to social cues (Kayl and Moore 2000). Attention deficit disorders have been associated with fronto-‐striatal and fronto-‐cortical abnormalities, two brain systems that are also associated with normal attention processing (Fox et al 2006; Cubillo et al 2012). Thus abnormalities in these same brains areas may contribute to the difficulty with planning and organization seen in NF1 patients.
Attention, planning, and organization are components of executive functioning, and the functioning of this multi-‐component cognitive faculty is tightly reliant on frontal cortex (Stuss and Levine 2002). Of note, patients suffering lesions to this area share symptoms with NF1 patients including limitations in planning, organization, and control of attention (Powel and Voeller 2004). The severity of attention deficits is however independent of the degree of organizational and planning deficits, suggesting that NF1 affects each component of executive function independently of the other (Hyman et al 2005). Importantly though, these components of executive function all rely on frontal brain networks, and dysfunction in this brain area has even been observed by functional imaging of NF1 patients. For instance, one study showed hypoactivation of prefrontal cortex during a rhyming task (Billingsley et al., 2003). Here participants are visually presented two nonsense words and asked to assess whether the set rhymes or not. Interestingly, hypoactivation of the corticostriatal pathway has also been observed during a delayed response task. In this visuospatial working memory task subjects must fixate on the central point of a screen and remember the location of stimuli that are presented peripherally. NF1patients show an impaired working memory performance on this task that is associated with hypoactivation of corticostriatal networks (Shilyanski et al 2010a).
Impairments in visuospatial cognition occur in over 50 percent of NF1 patients (Hyman et al 2005; Zöller et al 1997). In fact these deficits are so common that NF1 children can be distinguished from the normal population by evaluating their performance on multiple visuospatial tasks (Schrimsher et al 2003). The NF1 population performance is most frequently low in the Judgment of Line Orientation task (Hyman et al 2005), in which patients must discriminate visual stimuli presented at varying degrees of rotation. The task is thought to rely on an extensive brain system comprising the prefrontal, parietal and visual cortices (Ng et al 2000). More recently spatial working memory and visuospatial processing was also associated to activity in distinct cerebellar regions (Lee et al 2005). The degree of contribution that these different areas make to the visuospatial cognition phenotype seen in NF1 may well vary from area to area, and importantly from patient to patient. Though similar brain areas appear to be involved in different components of the NF1 cognitive phenotype, the occurrence of visuospatial impairments, just like problems with planning and organization, is independent of the presence of attention disorders (Schrimsher et al 2003).
Motor deficits are another component of the NF1 cognitive phenotype, and the impairments suggest a role for the cerebellum in the disorder. Children with NF1 are frequently described as clumsy (Cnossen et al 1998, Dilts et al 1996) and tend to reach developmental milestones at a late time (Chapman et al, 1996). The specific deficits that have been observed include many fine and gross motor functioning impairment such as problems with balance and gait, and with tasks like writing that require a visuomotor integration and fine motor skills (Hyman et al, 2003). Additionally, NF1 children were found to be impaired in several motor learning paradigms that are sensitive to cerebellar dysfunction, though they also showed normal performance in other cerebellum-‐dependent tasks (Krab et al 2011). Further suspicion of the cerebellum comes from the strikingly high level of expression of neurofibromin, the protein encoded by the NF1 gene, in cerebellar purkinje cells (Gutmann et al, 1995).
Participation from an early age in remedial teaching have been associated with enhanced long term outcomes on both cognitive and quality of life measures in individuals with Nf1-‐associated learning disorders (Hyman et al, 2006). Unfortunately, remedial teaching is not always available, pressing the need for alternative treatment options. For one, attention impairments can be treated pharmaceuticaly. NF1 patients presenting with co-‐morbid Attention-‐Deficit-‐ Hyperactivity-‐Disorder (ADHD) are known to benefit from methylphenidate treatment (Mautner et al, 2002). The development of pharmaceutical treatments for the other NF1 associated cognitive symptoms may greatly benefit the quality of life of individuals for whom remedial teaching is not available or effective.
Summarizing, the range of symptoms presented by NF1 patients lends to a presentation of a highly diverse and variable disease phenotype. In the central nervous system, the effects of NF1 on cognitive functioning can be categorized as those affecting reading/vocabulary, attention, executive function and planning, visuospatial function, and motor function (Hyman et al, 2005). In turn, dysfunction in brain areas underlying these cognitive processes may well contribute to the impairments observed in NF1 patients. The overall diversity of the phenotypes and its many clinical presentations can for a large part be understood by gaining an understanding on the genetics of the pathology.
The NF1 genotype-phenotype relationship
Underlying neurofibromatosis are mutations to the Nf1 gene (Wallace et al 1990). The resulting hereditary syndrome afflicts a variety of tissues and causes the myriad of clinical features described above. Though the disease is fully penetrant, individuals show large differences in the presentation of their phenotype, even within families (Huson 1994). One major factor contributes critically to the diverse clinical manifestation of NF1 and that is genetic background. Evidence has long demonstrated a role for trait-‐specific modifier genes in modulating the phenotypic
expression of individuals with NF1 (Easton et al 1993). NF1 gene expression itself is also a highly regulated and dynamic process involving several post-‐transcriptional modifications such as the alternative splicing mediated by the gene’s various splicing sites (Cappione et al, 1997). The result is that neurofibromin's final biochemical and functional properties are regulated at multiple levels, both during transcription and translation and by its modifier-‐gene dictated environment following expression. Modulation of NF1 also occurs at different levels, though the impact of such regulation on NF1 pathogenesis is less clear. For instance, differences in allelic expression levels, suggest that epigenetic mechanisms modulate neurofibromin protein levels and thereby physiological functionality (Stephens et al 1992). More recently, the regulation of NF1 expression by microRNA has been described in cultures of primary hippompampal neurons (Paschou and Doxakis 2012).
Neurofibromin is expressed as several variants encoded by the NF1 gene
The neurofibromin protein is encoded by the NF1 gene located at 17q11.2, a gene whose mass accounts for more than 300kb of human chromosome 17 (Li et al, 1995). The length of the gene is composed of 60 exons that are expressed as alternatively spliced transcripts. Of the 8 known variants, several are preserved across species, and 6 are expressed in humans. The isoforms are distinct in their temporal and tissue expression pattern, and further differ in functional properties (Skuse and Cappione 1997).
The naming of NF1 variants is not standardized and reflects the order of discovery and species in which it was discovered. More recently, the variants have also been referred to on the basis of exon inclusion/exclusion. The naturally occurring alternative splicing of the NF1 pre-‐mRNA results in differential inclusion of three alternative exons (23a, and 48a, 9a/9br) (Danglot et al, 1995; Andersen et al, 1993; Gutman et al, 1993). Type 1 and Type 2 NF1 exclude and include exon 23a, respectively (Andersen, 1993). The sequence of NF1 Type 1 shows a high degree of homology to the mammalian GTPase activating proteins and is referred to as NF1's guanosine triphosphatase-‐activating protein (GAP) domain (Xu et al 1990a). This domain modulates the inhibitory action of neurofibromin on the intracellular signalling molecule, RAS (Xu et al 1990a), and has been implicated in NF1 pathogenesis in humans (Klose et al 1998). The inclusion of exon 23a in NF1 Type 2 adds 63 bases within the GAP domain, and the ensuing conformational change weakens the ability of NF1 to regulate RAS (Uchida et al 1992, Andersen et al, 1993). Both Type 1 and Type 2 isoforms are expressed ubiquitously, though expression is enriched in neurons, Schwann cells, oligodendrocytes, astrocytes, leukocytes and adrenal medulla (DeClue 1991, Daston 1992b, Gutman 1999).
NF1 Type 3 and Type 4 are defined by their inclusion of exon 48a and exon 23a. The Type 3 isoform, which includes exon48a, is largely expressed in cardiac and skeletal muscle (Gutmann et al, 1995). Type 4, which includes exon 23a in addition to exon
48a, presents a similar expression profile. Exon 48a encodes 18 amino acids inserted in carboxyl terminus of neurofibromin, and therefore does no effect on the GAP domain (Skuse and Cappione 1997), but a role has been suggested for this exon in the development and differentiation of heart and skeletal muscles (Gutmann et al 1993, Gutmann et al, 1995, Skuse and Cappione, 1997).
Lastly, the inclusion of alternative exon 9a/9br in NF1 results in the addition of 10 amino acids to the amino terminal region of the transcript (Danglot et al, 1995). The presence of exon 9a/9br does not affect the GAP function of neurofibromin. Its expression seems to be restricted to the central nervous system, though it shows low expression in NF1-‐associated brain tumors (Gutman et al, 1999 & Geist & Gutman, 1996). Interestingly, the presence of exon 9a/9br on a protein level is limited to neurons of the forebrain (Gutmann et al, 1999), suggesting a role for the isoform in regulating neuronal function.
The Relationship between mutations of NF1 and NF1 pathogenesis
Convention has it that as a single gene disorder, NF1 has enormous potential for elucidating gene–brain–behavior connections. This is based on the fact that such relationships are many times more difficult and problematic to establish in multigenic disorders that are associated with mechanisms and phenotypes that are far more diverse and complex (Acosta et al, 2004). What has become clear though is that 'single-‐gene' disorders can be quite complex and as explained in the following 2 subsections, that the NF1 disorder is not as 'monogenic' as once thought. What is clear is that NF1 deficits ultimately manifest in a wide variety of clinical features with large differences in individual phenotypic severity.
Early studies showed that neurofibromin expression is essential in mammals. Genetically engineered mice modified to have a constitutive deletion of the NF1 gene face embryonic lethality between embryonic days 12.5 and 13.5 due to cardiovascular defects (Lakkis et al 1999). Importantly, NF1 and its downstream targets are highly conserved among different species, including mice and humans (Bernards et al 1993, Hajra et al 1994). Accordingly, homozygous mutations are also lethal in humans, and individuals with the NF1 disorder carry heterozygous mutations (Friedman, 1999).
Because of its large size, the NF1 gene is particularly susceptible to mutations. The NF1 gene has one of the highest mutation rates described for any human gene, and as a consequence some 30 to 50% of NF1 cases represent de novo mutations (Riccardi et al 1992, Huson et al 1994). NF1 mutations affect one allele at the DNA, mRNA, or protein level and this can result in various effects on the functions of neurofibromin. 70% of Nf1 Patients have mutations that lead to truncated, non-‐ functional version of the neurofibromin protein (Shen et al 1996, Thomson et al 2002). A recent comprehensive mutation analysis showed that 27% of NF1 mutations affect premRNA splicing (Messiaen 2008). In contrast to the abundance of
data available on mutations and their implications on a molecular level, it remains difficult to identify clear phenotype-‐genotype relationship due to the broad mutational spectrum of NF1 and the heterogeneity of symptom expression (Szudek et al 2002). Despite of these difficulties, two strong relationships have appeared, though these are derived from NF1 microdeletions cases that are associated with a stronger NF1 phenotype. (Descheemaeker et al 2004; Upadhyaya et al 2007). The first is a 3bp deletion outside of the GAP that leads to notably mild clinical phenotype and may also lead to a lower frequency of learning disabilities (Upadhyaya et al 2007). The second involves a microdeletion region of 17q11 that includes RNF135 gene, a candidate gene for the overgrowth, facial dimorphism, and possibly the more severe learning disabilities of NF1 microdeletic patients (Douglas et al 2007). Clinically presented mutations have also associated GAP dysfunction with NF1 pathogenesis. In one specifically interesting case, a missense mutation of arginine1276 into proline was found in a female member of a family with multi-‐ symptomatic NF1 phenotypes that includes malignant schwannomas. The mutation that was discovered neither impairs secondary or tertiary structure of the protein, nor does it largely affect expression levels or RAS binding of neurofibromin. Instead the mutation of the arginine finger of the NF1 GAP abolishes the enzymatic activity of GAP critical for NF1 cellular and molecular function, implicating GAP activity as a critical element of NF1 pathogenesis (Klose et al 1998). Despite the fact that NF-‐1 is a single-‐gene disease, it manifests with a variable expressivity of symptoms within families, even if individuals are affected by the same NF1 mutation (Sabbagh et al 2009).
As alluded to above, a significant factor contributing to the variable expressivity of cognitive and somatic symptoms in the NF1 populations is the contribution of modifying genes. Indeed a majority of the variability of NF1 symptom expression in mouse models and NF1 patients can be accounted for by inherited modifiers (Easton et al 1993; Sabbagh et al 2009). Modifying genes are genes whose aberrant functioning as proteins are of no pathological consequence in a wild-‐type background, but exacerbate, or increase the severity of disease symptoms in the presence of mutant NF1. One study examined 750 NF1 patients from 275 families to analyze phenotype correlations that included neurofibromas, cafe au lait spots, and learning disabilities (Sabbagh et al 2009). Of the 12 traits that were investigated, 11 showed a strong genetic component with no apparent influence of the NF1 mutations that were identified in the families. In other words, the study strongly supports a model whereby the inheritance of modifier genes modulates NF1 symptom expressivity more than mutations in the NF1 gene itself. These modifiers can be thought to interact directly with the gene or to modulate expression levels or tissue expression pattern. Alternatively the modifier genes could also be responsible for regulation of NF1 mediated signaling at the molecular or cellular level. Studies have found evidence for both these and several other mechanisms being utilized by putative NF1 modifier genes.
Molecular and Cellular View on NF1-Related Cognitive
Problems
As expected from the wide range of symptoms and the variability in the severity of expression in the NF1 population, neurofibromin is a highly dynamic regulator of multiple molecular pathways across various types of tissues and species. The neurofibromin protein contains a notorious RASGAP-‐ related domain, or GAP, which spans some 350 amino acids of this rather large 2818 amino acid long protein encoded by the NF1 gene (Xu et al 1990a). The consequent GAP mediated regulation of RAS activity is believed to account for the tumor suppressor activity of Nf1 (Xu et al 1990a, Xu et al 1990b, Jacks et al 1994, Brannan et al 1994). Though the GAP is the most well researched and characterized Nf1 domain, several more functional domains have been identified in the NF1 gene and protein. Importantly among them is an adenylyl cyclase–activating domain (Tong et al., 2002). Additional domains have been assigned roles in various cellular processes that include cell adhesion, melanosome transport, promotion of dendritic spine formation, and of neurite outgrowth (Hsue et al, 2012). One NF1 domain, the CSRD or cysteine-‐and serine-‐ rich domain, is thought to be critical for the association of neurofibromin with actin (Gregory 1993). Lastly, a domain encoded by exons 28 to 33 is thought to mediate the formation of a complex between neurofibromin and caveolin-‐1 (Boyanapalli et al 2006). Caveolin proteins are found uniquely in morphologically distinct invaginations at the plasma membrane, where they function as scaffolding protein that bind several receptors, signaling molecules, and adaptor proteins (Rajendran and Simons, 2005; Williams and Lisanti 2004). The relationships between most Nf1 domains and NF1 associated cognitive impairments remain to be elucidated. However, certain domains have been well characterized. Knowledge on the molecular pathways in which they allow NF1 to participate continues to advance our understanding of NF1 pathogenesis, and paint a picture of Nf1 as an important contributor to the regulation of neuronal function.
Upstream modulators of neurofibromin
Though it is not yet fully clear which receptors are upstream of Neurofibromin, the current line of evidence has demonstrated a strong role for growth factors as modulators of Nf1 activity. For instance protein kinase C (PKC) has been observed to quickly inhibit Nf1 in response to growth factors whose pathways utilize the tyrosine kinase receptors (Bernards & Settleman, 2005). Neurofibromin loss of function can lead to a de-‐coupling of RAS signalling from extracellular triggers. This was demonstrated by the finding that sensory neurons with a heterozygous loss of function of Nf1 no longer required the extracellular growth factor BDNF to survive and mature (Vogel et al 1995). Likely underlying this cellular effect is the up-‐ regulation of the RAS pathway, which is disinhibited following loss of NF1 function (Vogel et al 1995).
Some insight on upstream modulators of NF1 has come from the study of closely related pathologies. For instance Legius Syndrome, a neurodevelopmental disorder characterized by a cognitive and somatic phenotype similar to that of NF1, results from mutations in the SPRED1 gene (Stowe et al 2012). Like NF1, Legius Syndrome is associated with deregulated RAS signaling. Interestingly, neurofibromin is a Spred1-‐interacting protein that is necessary for Spred1's inhibitory function on the RAS pathway (Stowe et al 2012). This effect is mediated by the binding of Spred1 to NF1, which induces the localization of NF1 to the plasma membrane, where NF1 can act to down-‐regulate RAS signaling (Stowe et al 2012).
Downstream signaling pathways of neurofibromin
The NF1 gene encoding neurofibromin is expressed in a wide array of cell types in the body, though protein levels are most abundant in central nervous system cells including neurons (Daston et al 1992a; Daston et al 1992b). This tissue-‐specific enrichment of NF1 is indicative of a role for the protein in central nervous system functions. Indeed three down-‐stream targets of Nf1 signalling, RAS, cAMP, and VCP, position neurofibromin as a modulator of processes involved in the regulation of neuronal morphogenesis and synaptic plasticity. These functions of NF1 help explain the role that its mutant variants may have in the cognitive impairments and learning disabilities experienced by NF1 patients.
The RAS proto-‐oncogene is a small GTP-‐binding signalling protein that cycles between the inactive GDP-‐bound state and the active GTP-‐bound state. The GAP domain of NF1 increases the endogenous GTP hydrolyzing activity of RAS, thereby promoting the GDP-‐bound state of RAS. In this way, neurofibromin acts as a negative regulator of RAS signaling (Weiss et al, 1999). This effect of NF1 on RAS inhibits critical intracellular signalling cascades like the Extracellular signal Regulated Kinase (RAS/ERK) pathway (Habib et al 2008), and the Phosphoinositide 3-‐kinase (RAS-‐PI3K) / mammalian target of rapamycin (MTOR) pathway (Tohda, 2006). RAS-‐ERK signalling modulates synaptic plasticity by regulating both presynaptic and postsynaptic mechanisms. The presynaptic RAS-‐ERK pathway has been ascribed a role in the control of neurotransmitter release, which is realized by ERK phosphorylation of synapsin-‐1 (Tyler et al 2002, Kushner et al 2005). On the post-‐synaptic side the RAS-‐ERK pathway functions as an important signal integrator, being activated by calcium influx, BDNF binding to TRKB, and by indirect cAMP-‐PKA dependent pathway. The post-‐synaptic RAS-‐ERK signalling has been reported to be involved in AMPA receptor dynamics (Zhu et al, 2002) and regulation of protein transcription and translation events including CREB, or cAMP response element-‐binding (Thomas, 2004), both of which are processes implicated in long term potentiation (LTP). LTP serves as an in vitro measure of synaptic plasticity, which in turn is the cellular mechanism believed to underlie learning and memory (Kandel, 2001). Similarly, MTOR pathway has been implicated in the protein synthesis-‐dependent phase of LTP (Tang et al 2002).
In addition to down-‐regulating the RAS pathway, NF1 has also been implicated in up-‐regulating levels of cyclic AMP (cAMP). This regulation is well reflected by the reduced cAMP levels in Nf1-‐deficient central nervous system neurons and Nf1-‐ deficient astrocytes (Tong et al, 2002; Dasgupta et al, 2003; Warrington et al 2007). NF1 mediation of the cAMP/PKA/Rho pathway is crucial for development of normal neuronal morphology (Brown et al 2012), and heterozygous loss of function mutation in Nf1 lead to cell autonomous reductions in neurite lengths, growth cone areas, and cell survival (Brown et al 2010a). Though the mechanism is not yet defined, neurofibromin functions in the cAMP pathway at the level of adenylyl cyclase (AC) activation (Dasgupta et al, 2003). Two mechanisms utilizing receptor tyrosine kinase pathway and the heterotrimeric G-‐protein pathways respectively are likely to contribute to NF1 mediated AC signalling. The former requires the GAP activity of NF1, and the latter is a Gs alpha subunit (Gαs)-‐dependent process and
requires the C-‐terminal region of neurofibromin (Hannan et al 2006). In addition to controlling neuronal cell function by regulating morphology, cAMP can also exert a more direct effect on the electrical properties of neurons by modulating the gating kinetics and surface expression of voltage-‐gated ion channels (Biel et al, 2009).
Lastly, an important NF1 interacting partner has been discovered by the study of inclusion body myopathy with Paget's disease of bone and frontotemporal dementia (IBMPFD). Like Nf1, IBMPFD is autosomal dominant and though fully penetrant, patient populations show high degree of phenotype heterogeneity.The disorder is caused by mutations in the valosin-‐containing-‐protein (VCP), a direct protein interaction partner of neurofibromin. Both NF1 and VCP/p97 are critical for dendritic spine formation, and disruption of the interaction between neurofibromin and VCP impairs dendritic spinogenesis (Wang et al 2011).
All of the examples given above highlight the broad spectrum of NF1 signalling in the regulation of neuronal function. In this way, neurofibromin is likely to affect brain function by modulating control of dendritic spinogenesis, neuron morphogenesis, synatpic plasticity, and possibly by modulating ion channel function.
Animal Models of NF1: Insights on Etiology
Animal models have given invaluable insight on the genetic, molecular, and cellular mechanisms of NF1 pathogenesis. The power of these models comes from cleverly combining genomic and pharmacological tools with appropriate behavioral and physiological assessments. NF1 mouse models and patient populations display striking genenetic and phenotypic similarities. The NF1 gene sequence, transcriptional regulation, and downstream targets are conserved across species (Bernards et al, 1993, Hajra et al, 1994), and heterozygous mutations of NF1 function impair similar cognitive functions in mice (Li et al, 2005) and humans (Hyman et al 2005). Yet the results from such studies should always be taken in context. The evolutionary divide between mice and humans, as well as contributions
of modifier-‐genes and NF1 mutation type, must be considered when assessing the translational validity of the findings.
Behavioral Tasks Assigned to Specific Brain Areas Highlight the Parallels
of NF1 mice and human phenotype
The human NF1 cognitive profiles suggest dysfunction in hippocampus, parietal/prefrontal cortex, and cerebellum. In parallel, animal models have evaluated cognitive functions using behavioral tasks thought to rely on these areas. The most commonly employed mouse model is a heterozygous null mutant mouse,
Nf1+ / -, developed by deletion of exon 30 (Jacks et al 1994). The resulting
neurofibromin proteins are unstable and are rapidly targeted for proteolysis (Jacks et al 1994), which result in non-‐functional truncated versions of the protein (Shen et al 1996). Thereby this mouse model mimics 70% of heterozygous mutations seen in NF1 patient population. The Nf1+ / -‐ mouse displays deficits in hippocampal-‐
dependent spatial learning and memory (Silva et al 1997). The hidden platform version of the Morris Water Maze Task employed in these studies evaluates a mouse's ability to learn and remember the location of a submerged platform. A pilot study on children with NF1 revealed a similar pattern of spatial learning deficits in a virtual Morris maze (Ullrich et al, 2010). Although the performance of Nf1+ / -‐ mice is
significantly lower than wild-‐type (WT) mice on the early trials, this deficit is overcome by additional training (Costa et al. 2001, 2002; Cui et al. 2008; Silva et al. 1997). Similarly, the beneficial effects of extra training have also been described in NF1-‐associated learning disabilities, in the form of remedial teaching (Fuchs and Vaughn, 2012). The Nf1+ / -‐ mouse model further parallels human NF1 in its
sensitivity to modifier genes, so that the severity of symptom expression was modulated by the genetic background of the mice (Costa 2002). This was most clearly shown by crossing mice with a heterozygous knock out of glutamate receptor (NMDAR+/-‐) with Nf1+ / -‐ mouse. Whereas NMDAR+/-‐ showed no
performance deficit, poor task performance was exacerbated in mice carrying both mutations (Silva et al 1997). Another task employed to investigate hippocampal dependent spatial learning is contextual fear conditioning (Cui et al, 2008; Brown et al 2010b). The task measures a form of spatial learning in which an aversive stimulus (an electrical shock) is associated with a particular neutral context (spatial location), resulting in the expression of fear responses to the originally neutral context. Performance on this task was found to be impaired in multiple NF1 mouse models (Cui et al, 2008; Brown et al 2010b).
Deficits in the attention domain have also been described in Nf1+ / -‐ mice (Li et al,
2005; Brown et al 2010b). These are evaluated using a lateralized reaction time task, which requires sustained attention on two spatially separated location where visual stimuli may be presented (Jentsch, 2003), and an analogue version of the task has been used in humans to reveal attention deficits in NF1 patients (Robbins 2002).
Given the vast evolutionary differences between humans and drosophila, there are limitations on the extent that human cognition can be mimicked in flies. Nf1 animal models employing Drosophila therefore investigate the effects of NF1 on cognition by utilizing olfactory associative learning paradigms (Guo et al, 2000). In this task flies are trained by exposure to electroshock paired with one odor and subsequent exposure to a second odor without electroshock. Learning and memory is evaluated by allowing flies to choose between the two odors used during training.
Finally, Nf1+ / - mice have been used to investigate the motor deficits that are present
in the NF1 patient populations (van der Vaart 2011). Neurofibromin is enriched in cerebellar purkinje cells (Gutman et al, 1995) and the cerebellum has a modulatory role in motor function (Mier, 2002). As such, this mouse study on motor deficits utilized the Erasmus Ladder task known to depend on the cerebellum. In this task mice must move along a horizontal ladder, and learn to associate motor movements with the presentation of a tone which comes just prior to raising or lowering a ladder rung in the path of the moving animal. Though Nf1 model mice show non-‐ specific motor learning deficits in a rotarod task, in which mice are repeadetly tested for endurance in running on a rotating cylinder (van der Vaart 2011, Costa et al 2001), Nf1+ / - showed no deficits on the Erasmus ladder (van der Vaart 2011).
These results, together with the finding that NF1 children do not display deficits across all cerebellum-‐associated tasks argue against a causal role of the cerebellum in NF1-‐associated motor learning problems (Krab et al 2011). For now, the role which neurofibromin plays in the cerebellum plays in NF1 phenotype expression remains unresolved. However, a recent review looking over fMRI studies of the cerebellum reports on a host of non-‐motor functions that are mediated by this brain area including attention, executive control, working memory, and learning (Strick 2009). Future studies could asses the effects of cerebellar NF1 on these cognitive functions through a cre-‐recombinase mediated restricted expression of mutant NF1 in purkinje cells.
Increased Inhibition, Decreased cAMP, and Altered Dopamine System
Underlie NF1 Cognitive Phenotype
Several tools have been applied to direct the cellular and molecular mechanism underlying cognitive deficits in Nf1 mice and drosophila. These studies employed four approaches; (1) specific deletions of neurofibromin exon, (2) restrictive expression using cre recombinase genetic tools, and (3) pharmacological intervention.
In order to investigate the role of NF1 in molecular and cellular mechanism underlying the cognitive deficits, a combination of all the tools has been used, and though a central molecular mechanism has emerged, the entirety of the NF1 pathogenesis mechanism remains elusive. Nf1 heterozygous-‐null mice show impaired hippocampal LTP (Costa et al 2002, Li et al 2005). Interestingly, heterozygous Nf1 deletions driven by either synapsin I-‐Cre or Dlx5/6-‐Cre which
yield expression in GABAergic inhibitory neurons result in spatial learning deficits, whereas heterozygous deletions of Nf1 in excitatory neurons do not alter learning (Li et al, 2008). Consequently, LTP deficit was found to result from abberant NF1 function in GABAergic inhibitory neurons (Cui et al 2008, Shyliansky et al 2010). Both LTP and the associated deficit in Morris Water Maze task can be rescued by picrotoxin, a GABA A receptor antagonist (Costa et al 2002, Li et al 2005), which normalized the heightened inhibitory transmission observed in Nf1 mice (Costa et al 2002). Importantly, mice lacking the alternatively spliced NF1 exon 23a, which inhibits the GTPase-‐activating protein (GAP) domain of Nf1, displayed impaired visuospatial learning and thereby implicate enhanced RAS signalling in Nf1 +/-
phenotype. Furthermore, the deficits of Nf1+/- mice can be rescued by genetically
reducing the level of H-‐RAS or K-‐RAS (Costa et al 2002). The link to inhibitory neurons lies in neurofibromin's regulation of ERK, a downstream target of RAS (Shilyanski et al 2010a), and the finding that ERK dependent phosphorylation of synapsin 1 is critical for GABA release (Cui et al, 2008). In parallel to these findings in mice, fMRI-‐imaging studies in humans found that the degree of hypoactivation of dorsal lateral prefrontal circuits is correlated with working memory performance in humans (Shilyansky et al 2010a), reflecting the increased inhibition observed in multiple mouse models. The activity or RAS is dependent on its insertion in the membrane, and this anchoring requires post-‐translational isoprenylation (Konstantinopoulos et al, 2007). The rate-‐limiting enzyme in the synthesis of isoprenoids is 3-‐hydroxy-‐3-‐methylglutaryl coenzyme A (MHG-‐CoA). The enzymatic activity of MHG-‐CoA can be inhibited by Statins, such as lovastatin, a commonly used drug for the treatment of hypercholesterolemia. Interestingly, lovastatin can reverse the LTP deficits of Nf1+/-‐ mice (Li et al 2005). Furthermore lovastatin reverses the performance deficits of these mice on a visuospatial memory task as well as some components of their attention deficits (Li et al 2005). Two clinical studies of statin therapy for NF1 cognitive phenotype are currently underway.
Further interesting clues on NF1 cognitive phenotype have come from an NF1 mouse model which carries both a germline Nf1 heterozygous null mutation and a conditional homozygous knockout restricted to glial fibrillary acidic protein (GFAP) positive cells, or glia (Brown et al 2010b). Unlike heterozygous null mutants, these mice develop optic glioma (Bajenaru et al 2003, Zhu et al 2005); a tumor present in 15% of NF1 afflicted children (Listernick et al 1997). These Nf1OPG mice presented the same visuospatial cognitive profile as Nf1+/- mice and show impairment on
Morris Water Maze and Contextual Fear Condition Task. Additionaly, the mice were also found to have deficits in selective and non-‐selective attention (Brown et al 2010b). The attention deficits of the mice were reversed by methylphenidate (MPH) treatment (Brown et al 2010b). In a follow up study it was shown that Nf1OPG mice have reduced dopamine levels in the striatum, and that these levels can be normalized by MPH (Brown et al 2011b). Attention deficit hyperactivity disorder (ADHD) is a frequent co-‐morbid condition in NF1 (Hofman et al 1994), and a one year follow up study demonstrated the benefits of MPH treatment in this population in both learning ability and social behavior (Mautner et al, 2002).
