Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

34.1 Introduction

As discussed in preceding chapters, NF1 is a multisystem chronic genetic disease associated with a variety of symptoms. Another oft-cited hallmark of NF1 is its variability and unpredictability. Thus, patients typically only develop a subset of symptoms, and the severity of individual symptoms can vary dramatically between patients. Variable expressivity of a single-gene genetic disorder, like NF1, can be due to several different factors, including incomplete genetic penetrance, the nature of the culprit gene defect, whether or not a patient is a somatic mosaic, and whether or not a given patient carries genetic modifiers or has been exposed to different environments. In the case of NF1, all these factors are believed to play a role. Thus, although individuals carrying NF1 mutations develop at least some symptoms with complete penetrance, stochastic genetic or epigenetic events may result in partial penetrance of individual symptoms. An example is provided by plexiform neurofibromas, most of which are believed to develop congenitally. The fact that no more than about a third of patients develop these tumors may reflect the limited probability of loss of the wild-type NF1 allele in tumor progenitors during a restricted developmental period. Secondly, while genotype–phenotype correlations are uncommon in NF1, some notable exceptions exist. First, 5–10 % of patients harbor recurring 1.2–1.4 Mb microdeletions that include the NF1 gene and several flanking protein-coding and microRNA genes. Microdeletion patients often develop particularly severe symptoms, including mental retardation, dysmorphism, childhood overgrowth, and high numbers of unusually early onset neurofibromas (Kayes et al. 1994; Pasmant et al. 2010). In support of the view that the loss of contiguous genes may cause or modify these defects, patients carrying deletions that remove just the NF1 gene do not exhibit this more severe phenotype. Another example is provided by the finding that 21 unrelated NF1 patients carrying the same single amino acid deletion lacked obvious neurofibromas (Upadhyaya et al. 2007). Beyond effects attributable to the NF1 gene itself, variable expressivity can also reflect somatic mosaicism in sporadic patients who acquired their NF1 mutation post-zygotically (Bernards and Gusella 1994). The hypothesis that symptom-specific genetic modifiers play important roles in determining disease outcome received early support from a landmark study, which analyzed the presence or absence of five binary symptoms, as well as the severity of three quantitative symptoms amongst 175 patients in 48 NF1 families, including six monozygotic (MZ) twin pairs. This study found high correlation coefficients for four binary and two quantitative traits (neurofibroma and café-au-lait spot burden) for the MZ twins, lower scores for first-degree relatives, and lower values still for more distant family members (Easton et al. 1993). The hypothesis that unlinked modifier genes control specific NF1 symptoms has received additional support from subsequent work (Rieley et al. 2011; Szudek et al. 2002). Environmental factors are another potential cause for variable expressivity. In the case of NF1, the clearest example of a nongenetic factor affecting disease progression is the observation that neurofibromas increase in size and/or number during pregnancy (Roth et al. 2008).

The identity of modifier genes may provide clues to mechanisms responsible for disease symptoms, and since modifier genes perform rate-limiting functions, human modifiers represent prevalidated therapeutic targets. However, when the NF1 gene was identified in 1990, the human genome sequence and other tools to efficiently survey cohorts of differentially affected patients for potential modifiers were not available. Thus, our strategy to identify NF1 modifiers involved the identification of a highly conserved Drosophila NF1 (dNF1) ortholog, the generation of mutants, the analysis of phenotypes, structure–function studies, and genetic screens. The remainder of this chapter summarizes what has been learned so far and how this knowledge may be relevant to those interested in human NF1.

34.2 Identification and Structure of dNF1

The human NF1 gene sequence immediately suggested a function for the encoded protein as a GTPase-Activating Protein (GAP) for Ras (Xu et al. 1990). It also became apparent that, whereas neurofibromin shares a functional ~360 amino acid catalytic domain with RasGAPs from several species, it exhibited more extensive similarity over almost half of its length with the budding yeast Inhibitor of Ras Activity-1 and -2 (IRA1/2) proteins (Ballester et al. 1990). These findings prompted us to test whether NF1 orthologs existed in invertebrate species amenable to genetic analysis. No NF1 ortholog exists in the nematode, Caenorhabditis elegans. However, the dNF1 gene of the fruit fly, Drosophila melanogaster, was found to predict a protein 55 % identical and 69 % similar to human neurofibromin over its entire 2,802 amino acid length (The et al. 1997). As shown in Fig. 34.1, the IRA-related central segment of Drosophila neurofibromin is most similar to the human protein, but conserved regions also exist both up- and downstream. Reflecting the smaller size of the Drosophila genome, the dNF1 gene is more compact than its human counterpart. Thus, while human NF1 consists of 60 exons distributed over ~283 kb of DNA and harbors three other genes in its introns, dNF1 comprises 18 exons (alternate white/gray boxes in Fig. 34.1) and spans only 13 kb. At two locations, dNF1 mRNA is alternatively spliced. The position of alternatively spliced dNF1 exon 14 corresponds closely to where exon 43 is alternatively spliced in man (Thomson and Wallace 2002). However, while human exon 43 has been suggested to provide a nuclear localization signal (Vandenbroucke et al. 2004), whether alternative splicing of the (not obviously related) 30 amino acids predicted by dNF1 exon 14 has functional consequences, remains to be determined. At the exact location of alternatively spliced human exon 48a, the penultimate dNF1 coding exon 17 splices to three different final exons, yielding mRNAs predicting proteins with three distinct C-termini. Again, whether these C-terminal isoforms are functionally different remains unknown. In human NF1, inclusion of alternatively spliced exon 23a in the GAP-related domain (GRD) reduces the catalytic activity of the encoded protein (Andersen et al. 1993). However, the location of exon 23a does not correspond to a splice site in dNF1. Finally, we note that disease-associated missense mutations disproportionally affect residues conserved in Drosophila NF1. For example, the approximately 1,200 amino acid segment upstream of the GRD shares 46 % overall sequence identity between man and fly, but 35 of 47 missense mutations in this region (74 %) affect residues conserved in Drosophila.

Fig. 34.1
figure 00341

Diagram showing the degree of amino acid sequence identity between Drosophila and human neurofibromin. The alternate gray/white boxes below the graph indicate the size and location of 18 dNF1 exons. GRD: GAP-related domain. The extent of the protein segment related to S. cerevisiae IRA1 and IRA2 proteins is also indicated

Full size image

34.3 dNF1 Mutants and Phenotypes

No classical dNF1 mutations existed. In their absence, we mobilized a flanking P transposon to generate two dNF1 disrupting de novo integrations. Both dNF1 P1 and dNF1 P2 alleles behave as molecular and genetic nulls. Given that expression of oncogenic Ras in fly eyes causes severe defects (Bishop and Corces 1988), it seemed puzzling at first that homozygous loss of dNF1 did not cause any abnormal patterning. However, both mutants exhibited a 15–20 % reduction in linear dimensions during all phases of postembryonic development, and adult flies had a reduced tendency to fly away when released (The et al. 1997). Since these and other dNF1 phenotypes may be affected by the genetic background, and because the dNF1 P1 allele also deleted adjacent E(spl) complex genes, we made additional alleles by chemical mutagenesis of isogenized flies. Among three new alleles, dNF1 E4 is a C1045Y missense mutant, and dNF1 E1 and dNF1 E2 are Q370* and Q1062* nonsense mutations (Walker et al. 2006).

Beyond invoking Ras-independent functions, at least two other factors may explain why loss of dNF1 and expression of constitutively active Ras do not cause similar defects. First, the Drosophila genome contains five RasGAP genes (Gap1, Vap, dNF1, CG1657, CG42684) and two plexin-related potential RasGAPs. The relatively subtle dNF1 phenotypes may also reflect the apparent restriction of dNF1 expression to the nervous system (Walker et al. 2006). While loss of dNF1 does not obviously affect viability, fertility, or patterning, several macroscopic, behavioral, and biochemical phenotypes have been identified (Fig. 34.2). Interestingly, the initially identified dNF1 defects were not sensitive to genetic manipulations that affect Ras signaling strength. Rather, they were restored by increasing, and enhanced or mimicked by reducing, signaling through the cAMP-dependent Protein Kinase A (PKA) pathway (Guo et al. 1997; The et al. 1997). The suggested link between dNF1 and adenylyl cyclase (AC)/PKA signaling has motivated much subsequent research, and while there is no doubt that loss of dNF1 somehow affects AC/PKA signaling, we and others have reached conflicting conclusions as to whether dNF1 affects this pathway directly or indirectly. Before discussing what might explain this discrepancy, the following sections first briefly describe the dNF1 phenotypes illustrated in Fig. 34.2.

Fig. 34.2
figure 00342

Larval and adult homozygous null dNF1 phenotypes. dNF1 larvae lack a neuropeptide-induced rectifying K+-current at the body wall neuromuscular junction. Larvae, pupae, and adults are 15–20 % smaller than isogenic wild-type controls, and larval and adult dNF1 brain phospho-ERK levels are approximately threefold higher than in controls. Adult dNF1 flies lack normal day/night rhythmic locomotor behavior and exhibit a reduced olfactory associative learning/short-term memory performance, as well as deficits in middle-term and long-term memory. The diagram depicting the adult brain shows the location of the mushroom bodies (MBs), considered the insect equivalent of the vertebrate hippocampus. MBs play essential roles in olfactory learning and preferentially express several AC/PKA pathway proteins. Although dNF1 is expressed in MBs, whether its learning/memory-related functions involve cells within or outside of MBs remains controversial (Buchanan and Davis 2010; Gouzi et al. 2011)

Full size image

34.3.1 Postembryonic Growth Deficiency

dNF1 embryos are of normal size, but mutants are 15–20 % smaller than controls during all subsequent larval, pupal, and adult stages. By measuring adult wing cell densities, this defect was shown to reflect a reduction in cell size rather than in cell number. However, smaller mutant eyes consist of fewer normal-sized ommatidia, indicating that different tissues respond differently to loss of dNF1. Mosaic analysis provided the first indication that reduced wing growth involves a non-cell-autonomous mechanism (The et al. 1997). This conclusion is further supported by evidence that dNF1 expression is largely restricted to the nervous system (Walker et al. 2006) and by findings that neuronal re-expression of either fly or human NF1 sufficed to restore dNF1 growth (Tong et al. 2002; Walker et al. 2006). The dNF1 growth defect was not suppressed by heterozygous loss of Ras1 or the Ras exchange factor Son-of-sevenless and not enhanced in mutants carrying a gain-of-function Raf allele. Rather, hypomorphic PKA catalytic subunit mutations phenocopied the dNF1 size defect and expression of a constitutively active PKA transgene throughout development partially restored dNF1 growth (The et al. 1997).

Organism size is a function of growth rate and duration. In insects, whose growth occurs mostly during larval development, the former is controlled by insulin-like peptides and the latter by a hormonal cascade that culminates in the timed release of the molting hormone, ecdysone. No conclusive evidence implicating dNF1 in either pathway has yet been found (Walker et al. 2006). Among the various dNF1 phenotypes, the growth defect is the only one readily amenable to genetic analysis. Another reason for focusing on this defect is that reduced growth is a common symptom of human NF1 and related RASopathies (Szudek et al. 2000).

34.3.2 Behavioral Deficits

dNF1 mutants exhibit subtle behavioral defects. When flies are released, ~15 % do not fly away, even after repeated prodding (The et al. 1997). When adult flies are tapped down in their vials, mutants take longer to climb back up (Tong et al. 2007). No anatomical basis for these defects has been identified; mutants responded normally to visual or olfactory cues (The et al. 1997) and display normal locomotor activity (Williams et al. 2001). Like the reduced growth phenotype, the abnormal climbing response has been attributed to an AC/PKA signaling defect (Tong et al. 2007).

34.3.3 Neuropeptide-Stimulated K+-Current Defect

Loss of dNF1 does not affect sevenless-mediated Ras signals controlling photoreceptor cell development or torso/Ras signaling in the embryo (The et al. 1997). Hence, to analyze whether dNF1 affects other Ras pathways, we collaborated with Dr. Yi Zhong, who had reported that a rectifying K+ current elicited in Drosophila larval body wall neuromuscular junction preparations by the mammalian neuropeptide PACAP38 required intact Ras and rutabaga AC pathways (Zhong 1995). Electrophysiological analysis showed that dNF1 mutants lack the PACAP38-elicited current, which was restored by manipulating AC rather than Ras signaling (Guo et al. 1997). An unanswered question remains as to how mammalian PACAP38 stimulates a response in the absence of an obvious Drosophila PACAP receptor ortholog.

34.3.4 Circadian Rhythm Defect

The circadian clock controls various rhythmic behaviors, including daily changes in locomotor activity. When analyzed in constant darkness, dNF1 mutants showed normal cycling of the period and timeless clock genes but were either completely (NF1 P1) or nearly completely (NF1 P2) arrhythmic. Normal rhythmic behavior was restored upon elav-GAL4-driven pan-neuronal expression of UAS-dNF1, but not by restricted expression in central brain lateral neurons, the site of the circadian clock. Antibody staining revealed a circadian oscillation in phospho-ERK near nerve terminals containing Pigment Dispersing Factor, a secreted output of clock cells, and dNF1 arhythmicity was restored by mutations that attenuate Ras/ERK signaling. Thus, excess Ras/ERK signaling in dNF1-deficient neurons affects a circadian clock output pathway and circadian locomotor activity (Williams et al. 2001).

34.3.5 Associative Olfactory Learning and Short/Middle-Term Memory Defects

Drosophila is widely used to study molecular and cellular circuits responsible for learning and memory (Davis 2005). Because analysis of the neuromuscular signaling defect implicated dNF1 in a pathway involving the rutabaga AC, a well-known learning mutant, and because NF1 is associated with learning deficits in both man and mouse (Hyman et al. 2005; Silva et al. 1997), the Zhong group analyzed dNF1 mutants in a Pavlovian associative learning assay that uses electrical shocks to teach adult flies to choose between two equally aversive odors. In this test, NF1 P1 and NF1 P2 mutants both exhibited a significantly reduced learning/short-term memory performance. Arguing that developmental defects do not explain this deficiency, adult expression of a hsp70-dNF1 transgene suppressed the learning deficit. Further work showed that beyond short-term (3 min) memory, 3- and 8-h middle-term memory retention was also affected in flies lacking dNF1. Again implicating abnormal AC/PKA signaling, both learning and memory defects were suppressed by increased PKA expression (Guo et al. 2000).

34.3.6 Long-Term Memory Defect

Excess Ras/ERK signaling underlies the reduced performance of Nf1 +/− mice in the Morris water maze, which assesses spatial learning ability (Costa et al. 2002; Cui et al. 2008). By contrast, defective AC signaling appears responsible for the dNF1 olfactory learning deficit (Guo et al. 2000). Unlike the short-term olfactory learning assay, performance in the water maze requires protein synthesis-dependent long-term memory (LTM). Arguing that this difference may explain the different conclusions reached, dNF1 null mutants were found to have an LTM defect that, unlike the short-term learning and memory deficits, appeared Ras dependent (Ho et al. 2007). More recent work by others further supports roles for dNF1 in short-term (3 min), middle-term (3 h), and 24-h LTM, and specifically in memory acquisition but not decay (Buchanan and Davis 2010).

34.3.7 Mitochondrial Dysfunction

dNF1 mutants have reduced life spans and increased vulnerability to heat or oxidative stress, in association with reduced mitochondrial respiration and elevated reactive oxygen species (ROS) production. By contrast, dNF1 overexpression increased life span and improved reproductive fitness, increased resistance to heat and oxidative stress, accompanied by increased mitochondrial respiration and a 60 % reduction in ROS production. Like other dNF1 defects, these phenotypes were restored by pharmacological or genetic manipulations that increase cAMP/PKA signaling. Counteracting ROS with catalytic antioxidants also restored normal life span to homozygous dNF1 null mutants (Tong et al. 2007).

34.4 Does dNF1 Affect AC/PKA Signaling Directly or Indirectly?

As summarized above, genetic evidence strongly supports a link between dNF1 and AC/PKA signaling. A functional link is further supported by biochemical evidence, including the observation that GTPγS-stimulated but not basal AC activity is reduced in dNF1 P1 and dNF1 P2 brain membrane preparations (Guo et al. 2000). Moreover, GTPγS-stimulated AC activity and cAMP levels were also found to be reduced in day E12.5 Nf1 −/− murine brain extracts (Tong et al. 2002). Thus, while defective AC signaling appears to be an evolutionarily conserved, albeit recessive, NF1 phenotype, these results do not reveal how neurofibromin affects AC activity. Consistent with the view that non-cell-autonomous mechanisms may be involved, GTPγS-stimulated AC activity was not only reduced in Drosophila brain but also in abdominal tissue (Guo et al. 2000), where dNF1 may not be expressed (Walker et al. 2006). Also compatible with a non-cell-autonomous mechanism, the dNF1/PKA-mediated effect on mitochondrial respiration was detected in experiments with mitochondria isolated from whole dNF1 mutant flies (Tong et al. 2007; Walker and Bernards 2007).

In experiments aimed at identifying how dNF1 affects AC/PKA signaling, we and others have reached conflicting conclusions. Thus, work by the Zhong group led them to postulate the existence of two dNF1-dependent AC pathways, beyond the canonical Gαs-dependent pathway. The first NF1-requiring pathway, stimulated by serotonin and histamine, depended upon Gαs, whereas a novel AC pathway involved the EGF receptor, NF1, and Ras, but not Gαs. Using transgenic flies expressing human NF1 deletion and point mutants, the authors reported that NF1-mediated Ras regulation was essential for the novel EGF receptor-stimulated AC pathway, but not for NF1/Gαs-dependent neurotransmitter-stimulated AC activity. Moreover, providing the clearest evidence that NF1 may have Ras-independent functions, a C-terminal segment of human neurofibromin that did not include the GRD was capable of rescuing the AC-dependent organismal growth (Hannan et al. 2006) and learning defects (Ho et al. 2007).

Our conflicting conclusion that excess neuronal Ras/ERK activity is the proximal cause of dNF1 growth and learning phenotypes is based on the results of two studies. The first focused on the dNF1 growth defect. In structure/function studies similar to those performed with human NF1 transgenes (Hannan et al. 2006), we found that large dNF1 segments other than the GRD were dispensable for growth regulation, that several GAP deficient dNF1 point mutants did not restore growth, and that expression of a truncated protein representing just the dNF1 GRD was sufficient to suppress the growth defect. Moreover, the growth detect was also suppressed by neuronal expression of a Drosophila p120RasGAP ortholog. To investigate why in earlier studies, loss of Ras1 or other canonical Ras pathway components did not modify the dNF1 growth defect, we tested a comprehensive set of Ras pathway single and double mutants for their ability to restore dNF1 growth. As a molecular correlate, we also analyzed the ability of these mutants to restore the elevated larval and adult brain phospho-ERK level. None of the single mutants modified either phenotype, arguing that the tested Ras pathway components are either not involved, or not rate limiting in the pathway leading to ERK activation in dNF1 brain. Supporting the latter conclusion, some double mutants that did restore normal ERK activity also rescued the growth deficiency at least partially (Walker et al. 2006). Thus, the proximal cause of the non-cell-autonomous dNF1 organismal growth defect is excessive neuronal Ras/ERK signaling (Walker et al. 2006).

34.4.1 Drosophila Alk/Ras Signaling May Be Responsible for dNF1 Growth and Learning Defects

Further evidence implicating abnormal Ras/ERK signaling as the primary cause of dNF1 defects comes from work that began with the observation that overexpression of either the Drosophila Alk (dAlk) receptor tyrosine kinase or of its activating ligand jelly belly (jeb) phenocopied both dNF1 growth and olfactory learning phenotypes. This result was not unexpected, since previous work had identified dAlk as an activator of Ras/ERK signaling in vivo (Loren et al. 2001). However, attenuating dAlk expression or activity, either through the use of dAlk mutant alleles, dAlk shRNA constructs, expression of a dominant-negative dAlk transgene, or through pharmacological inhibition, rescued dNF1 growth, olfactory learning, and brain ERK over-activation phenotypes. In support of the hypothesis that dAlk is a rate-limiting activator of dNF1-regulated Ras/ERK signals that promote organismal growth and that limit olfactory learning, dNF1 and dAlk expression overlaps extensively in both larval and adult brain. Moreover, dAlk-GAL4 driven neuronal UAS-dNF1 expression sufficed to restore all tested defects (Gouzi et al. 2011). Finally, although a role for dAlk as a negative regulator of learning appears unusual, recent work indicates that murine Alk may have a similar role (Weiss et al. 2012).

Whether excess Alk/NF1/Ras signaling contributes to human NF1 defects is an obvious next question. Intriguingly, loss of NF1 expression is common and associated with a worse prognosis in neuroblastoma (Holzel et al. 2010), a significant proportion of which shows ALK amplification or gain-of-function ALK mutations [reviewed by Azarova et al. (2011)]. ALK overexpression and NF1 loss have similarly been implicated in glioblastoma (Powers et al. 2002; Verhaak et al. 2010). Finally, mammalian ALK is activated, either directly (Stoica et al. 2001, 2002) or indirectly (Perez-Pinera et al. 2007) by two related secreted ligands, pleiotrophin and midkine. Suggesting the intriguing possibility that inappropriate ALK signaling may contribute to NF1 tumorigenesis, midkine is overexpressed in human NF1 and has been shown to act as a mitogen for NF1 tumor cells (Mashour et al. 2001, 2004).

34.5 Conclusions

Our results have led us to conclude that excess neuronal Ras/ERK signaling is the root cause of most, if not all, dNF1 defects. However, several important questions remain. Perhaps the most vexing unresolved issue is how dNF1 affects AC/PKA signaling. One possible mechanism is that loss of dNF1 precipitates one or more neuroendocrine or neurotransmitter signaling defects. Since hormones and neurotransmitters often signal through AC-coupled receptors, such defects might be restored by increased AC/PKA signaling. We note that in this model, defective AC/PKA signaling does not necessarily involve dNF1-requiring neurons. However, although no results to date formally exclude the possibility that dNF1 affects AC/PKA signaling cell autonomously, our working hypothesis remains that cross talk between distinct neuronal populations may be involved in generating the various Ras/ERK- and AC/PKA-dependent dNF1 phenotypes. Ongoing genetic screens for dominant genetic modifiers of dNF1 phenotypes, beyond dAlk and jeb, experiments to identify the site of action of modifiers, and studies to pinpoint the cells in which increased AC/PKA signaling restores dNF1 defects, may eventually reveal the neuronal circuits via which dNF1 exerts its various functions, and provide further clues to molecular and cellular pathways responsible for human NF1 symptoms.