The study of cancer predisposition syndromes presents unique opportunities to gain insights into the genetic events associated with tumor pathogenesis. Individuals with two inherited cancer syndromes, neurofibromatosis 1 (NF1) and neurofibromatosis 2 (NF2), develop both benign and malignant tumors. The corresponding genes mutated in these two disorders encode tumor suppressor proteins, termed neurofibromin (NF1) and merlin (NF2), which function in novel ways to regulate cell growth and differentiation. Neurofibromin inhibits cell proliferation, at least in part, by modulating mitogenic pathway signaling through inactivation of p21-ras. In contrast, merlin may act as a membrane-associated molecular switch that regulates cell–cell and cell–matrix signals transduced by cell surface receptors. Significant progress in our understanding of the genetics and biology of NF1 and NF2 has elucidated the roles of these genes in tumor initiation and progression.
Received 20 December 2000; Accepted 15 January 2001.
The study of cancer predisposition syndromes affords scientists and clinicians valuable insights into cancer biology, developmental biology and general cell biology. Individuals with inherited cancer syndromes are at significant risk of developing both benign and malignant tumors as a result of starting life with a germline (inherited) mutation in one of the two copies of a specific tumor suppressor gene. Since affected individuals are heterozygous for a loss of function mutation in one copy of a tumor suppressor gene, only one additional genetic alteration (loss of the wild-type allele) is needed to facilitate tumor development. This two-step process of tumor suppressor gene inactivation was coined the ‘two-hit hypothesis’ by Alfred Knudson in his classic monograph on retinoblastoma (Fig. 1) (1). In this fashion, a cell that undergoes inactivation of both copies of a specific tumor suppressor gene has an increased growth advantage relative to cells with wild-type tumor suppressor gene function.
The proteins encoded by tumor suppressor genes mutated in cancer predisposition syndromes are growth regulators critical for the maintenance of orderly cell growth and differentiation. A simplified view of tumor suppressor gene function envisions three main intracellular compartments important for regulating cell proliferation and survival, including: (i) the cell membrane where cues from the environment are transduced to the interior of the cell; (ii) the cytoplasm, where these extracellular signals are transmitted to the nucleus; and (iii) the nucleus, where cell cycle regulation dictates whether a cell will initiate DNA synthesis and undergo mitosis. Dysregulated function of proteins in any of these three compartments can result in an increased growth advantage and can, by itself, predispose the cell to transformation or do so in concert with other genetic changes. The identification of tumor suppressors mutated in specific inherited cancer syndromes allows us to pinpoint critical signal transduction pathways, cell cycle regulatory events and extracellular cues that instruct a particular cell type to proliferate or differentiate. The elucidation of the cellular processes affected by dysfunction of these molecules may serve to identify potential targets for future therapeutic interventions.
Neurofibromatosis 1 (NF1) and neurofibromatosis 2 (NF2) are genetically distinct clinical syndromes in which affected individuals develop both benign and malignant tumors that predominantly affect the nervous system. Since many of the features of NF1 and NF2 are not related to tumor formation (café-au-lait macules in NF1 and cataracts in NF2) and the vast majority of the tumors that develop in both disorders behave in a clinically ‘benign’ fashion (slow growing and rarely malignant), it could be argued that NF1 and NF2 are not true cancer predisposition syndromes, but rather represent ‘tumor suppressor gene’ disorders. The classification of NF1 as a ‘phakomatosis’ (disorder associated with dysplastic or hamartomatous growths) further underscores the absence of a cancer phenotype. Despite the difficulties with assigning an accurate label, the fact that affected individuals are prone to tumor development suggests that the NF1 and NF2 genes are important growth regulators for a number of specific cell types. The identification of the NF1 and NF2 genes and their encoded proteins has not only provided significant insights into the pathogenesis of NF1- and NF2-associated tumors, but has also shed light on the cellular and genetic events critical for sporadic tumor formation and progression.
NF1 is the most common cancer predisposition syndrome affecting the nervous system with an incidence of 1 in 3000 worldwide (2,3). Typically, individuals with NF1 present early in life with pigmentary abnormalities, such as café-au-lait macules, skinfold freckling and iris hamartomas (Lisch nodules). In addition to these features, children with NF1 can present within the first 6 years of life with low-grade glial tumors involving the optic pathway (optic pathway gliomas or astrocytomas) (4). Although typically non-progressive, some of these astrocytomas can continue to grow and result in visual loss, hypothalamic dysfunction or other neurologic symptoms (5). Most adolescents and adults with NF1 also develop benign tumors composed of Schwann cells and fibroblasts (dermal and plexiform neurofibromas) and rare individuals will manifest with an aggressive Schwann cell tumor, termed a malignant peripheral nerve sheath tumor (MPNST). Lastly, two uncommon tumors, pheochromocytomas (adrenal medullary cancers) and leukemias, are over-represented in individuals with NF1.
Early insights into the pathogenesis of NF1-associated tumors began with the identification of the NF1 gene in 1990 (6–8). The NF1 gene product, neurofibromin, is a large cytoplasmic protein with a small central region that demonstrates sequence similarity with the GTPase activating protein (GAP) family of proteins involved in the downregulation of ras activity (Fig. 2) (9–12). Since ras activation is associated with increased cell proliferation, it is reasonable to assume that loss of neurofibromin GAP function in NF1-associated tumors would result in increased levels of activated ras sufficient for tumor formation. Increased ras activity associated with neurofibromin loss has been demonstrated in many NF1-associated tumors, including astrocytomas (13), leukemias (14), neurofibromas (15) and MPNSTs (16–18). Moreover, neurofibromin rasGAP function appears to account for the increased proliferation of neurofibromin-deficient (Nf1–/–) mouse hematopoietic cells. Whereas the increase in both proliferation and ras pathway activity could be reverted to normal in Nf1–/– cells by introducing a wild-type neurofibromin GAP domain, no effect was observed with a related rasGAP molecule, p120-GAP (19). These results argue that the consequence of neurofibromin loss on ras regulation is specific to neurofibromin and is not observed with other rasGAP molecules. In this regard, the growth suppressor function of neurofibromin in many cell types probably results from its ability to regulate ras activity.
In an effort to develop tractable models for NF1-associated cancers, much effort has been placed on generating mice with a targeted disruption of the Nf1 gene. Although mice homozygous for an Nf1 mutation die of a cardiac vessel defect during embryogenesis, Nf1+/– heterozygotes are viable, but succumb to leukemias and pheochromocytomas by 15 months of age (20,21). Unfortunately, no astrocytomas, neurofibromas or MPNSTs develop in these Nf1+/– mice. Recently, sophisticated approaches have been taken to develop more clinically accurate mouse models for NF1, including mice chimeric for Nf1 loss, mice with tissue-specific (conditional) Nf1 disruption and mice with multiple genetic alterations. One mouse model for NF1 was generated by varying the amount of null (Nf1–/–) embryonic stem cells in the developing blastocyst. Chimeric mice derived in this fashion contain tissues that lack Nf1 expression proportional to the contribution of Nf1–/– ES cells in the developing animal. Some of these chimeric animals develop tumors that histologically resemble human plexiform neurofibromas (22), suggesting that loss of Nf1 may be sufficient for the development of these tumors. In addition, several groups have focused on developing conditional, tissue-specific Nf1 knockout mice by targeting Nf1 inactivation in specific cells (e.g. Schwann cells, astrocytes and neurons). Preliminary observations suggest that Nf1 inactivation has profound tissue-specific effects on cell growth and differentiation. The analysis of these mice should yield valuable information about the consequences of neurofibromin loss in the specific tissues affected in individuals with NF1.
In an effort to determine what additional genetic events might cooperate with neurofibromin loss to promote tumor progression, several groups have focused on the role of p53, based on the observation that loss of NF1 is associated with p53 mutations in human MPNSTs (23). Mice heterozygous for both Nf1 and p53 mutations develop high-grade sarcomas, upon loss of both neurofibromin and p53 expression, which are histologically similar to human MPNSTs (24,25). When maintained on specific genetic backgrounds, these mice are also more prone to high-grade astrocytic tumors (25) in contrast to the slow-growing gliomas seen in children with NF1, which lack p53 mutations (26,27). These results argue that the spectrum of tumor formation in Nf1 mutant mice is dependent on the presence of other ‘modifying’ genes. The identification of these ‘modifying’ genes may explain why individuals within a single family that harbor the identical germline NF1 mutation develop different tumors.
Another intriguing feature of NF1 is the development of non-tumor abnormalities, including increased brain astrogliosis (28). It is likely that this increase in brain astrocytes reflects the effect of reduced, but not absent, neurofibromin expression and function. Neurofibromin is expressed in brain astrocytes both in vivo and in vitro (Fig. 3A). Studies from our laboratory and others have demonstrated that cells derived from Nf1+/– mice exhibit increased cell proliferation, elevated ras pathway activity and abnormalities in actin cytoskeleton-associated processes (Fig. 3B and C) (29–31). Moreover, heterozygosity for a different GAP molecule that downregulates ras (p120-GAP) is not sufficient to result in increased astrocyte number (Fig. 3D), again supporting the notion that ras regulation by neurofibromin is unique. This heterozygote effect probably represents an important feature of certain cancer predisposition syndromes, such that haploinsufficiency for specific tumor suppressor genes may be sufficient to confer a preneoplastic growth advantage.
Although neurofibromin tumor suppressor function has been attributed to its rasGAP activity, this portion of the molecule represents only 10% of the coding region, raising the possibility that some of the clinical features of NF1 may be attributable to other (non-rasGAP) functions of neurofibromin. Elegant studies in Drosophila have suggested that neurofibromin might function to regulate protein kinase A (PKA) signaling (32,33). Support for this notion also derives from experiments in which loss of neurofibromin influences the response of Schwann cells to cyclic AMP (34). Although a direct link between neurofibromin deficiency, PKA signaling and cAMP levels has not been demonstrated in mammalian cells, these studies suggest that mechanism of neurofibromin growth regulation may extend beyond its rasGAP activity.
Other insights into possible non-rasGAP functions for neurofibromin have derived from studies on NF1 alternative splicing. Three alternatively spliced exons (9a, 23a and 48a) have been reported. Whereas exon 48a neurofibromin is specifically expressed in muscle tissues (35,36), neurofibromin containing exon 9a is restricted to neurons in the forebrain with expression being both regionally and developmentally regulated (37,38). This unique 10-residue exon in the N-terminus of the protein may confer novel properties on neurofibromin relevant to neuronal function. Future studies aimed at generating mice with conditional disruption of these alternatively spliced exons will likely yield valuable insights into their functional significance.
NF2 is a disorder clinically distinct from NF1, in which affected individuals develop schwannomas, meningiomas and ependymomas (39). Although classified as cancers, these tumors are also typically slow growing and non-malignant. The NF2 gene was identified by positional cloning and found to bear sequence similarity to a family of proteins that link the actin cytoskeleton to cell surface glycoproteins, collectively termed Protein 4.1 molecules (40,41). Among the Protein 4.1 molecules, the NF2 gene product, merlin or schwannomin, is most closely related to ezrin, radixin and moesin (ERM proteins) (42). Merlin contains an N-terminal domain (residues 1–302), which is hypothesized to mediate binding to cell surface glycoproteins, an α-helical domain (residues 303–478) and a unique C-terminus (residues 479–595) that lacks the conventional actin binding region conserved amongst other ERM proteins (Fig. 4A). ERM proteins have been shown to be involved in cellular remodeling, but no direct growth regulatory properties have been demonstrated. There is some suggestion that ezrin might modulate programmed cell death (apoptosis) through binding to PI3-kinase (43), however no evidence exists that merlin functions in a similar manner.
In the absence of an obvious catalytic domain, insights into merlin’s possible mechanism of action have derived from several complementary approaches. The Nf2 knockout mouse dies during early embryonic development of a failure to initiate gastrulation as a result of an absence of organized extraembryonic ectoderm (44). The underlying defect in these Nf2–/– embryos does not appear to be related to cell proliferation abnormalities in the embryo itself, but rather a failure to produce the extra-embryonic structures required to generate a mesoderm-inducing signal from the embryo proper. The inability of the developing merlin-deficient embryo to generate or respond to important differentiation cues suggests a role for merlin in cell–cell signaling events critical for extra-embryonic tissue formation. In conditional Nf2 knockout mice, where merlin expression was specifically disrupted in myelin P0-expressing cells, schwannomas develop in association with peripheral nerves that histologically resemble human schwannomas (45). These results argue that loss of merlin function is sufficient for schwannoma formation in vivo.
Loss of Nf2 in Drosophila leads to embryonic lethality whereas regional loss of Drosophila merlin (D-merlin) results in a 2–3-fold increase in proliferation of morphologically normal tissue without evidence of malignant transformation (46). The Nf2–/–Drosophila phenotype is similar to that observed with the loss of another Protein 4.1 protein termed expanded. Merlin and expanded interact in vivo and the combined loss of D-merlin and expanded results in defects in both cell proliferation and differentiation (47). These results suggest a model in which D-merlin and expanded might coordinately regulate cell proliferation and differentiation by forming an active heterodimer.
Another approach to determining how merlin functions involves the identification of molecules that interact with merlin and potentially transduce its growth suppressive signal. Several candidate proteins have been identified, including CD44 (48–50), actin (51), βII-spectrin (52), SCHIP-1 (53), HRS (54), NHE-RF (55–57) and β1-integrin (58) (Fig. 4C). Merlin binds actin weakly in vitro through an alternative actin-binding domain in the N-terminus that is conserved in all ERM proteins (51). In addition, merlin interacts with an actin-binding protein, βII-spectrin, and may indirectly associate with the actin cytoskeleton in vivo by virtue of this interaction (52). The functional significance of merlin’s association with the actin cytoskeleton is supported by several findings. First, merlin-deficient tumor cells derived from the Nf2+/– mouse are highly metastatic and motile (59). Second, schwannoma cells from NF2 patient tumors have dramatic alterations in the actin cytoskeleton and display abnormalities in cell spreading (60). Lastly, inducible expression of wild-type, but not missense mutant, merlin in rat schwannoma cells reduces cell motility as well as disrupts the actin cytoskeleton during cell spreading (61). These results suggest that merlin may play an important role in regulating both actin cytoskeleton-mediated processes (e.g. cell motility) and cell proliferation, raising the possibility that merlin operates in a signaling pathway that links these two processes.
The association of merlin with CD44 and β1-integrin also raises the possibility that merlin might function as a molecular switch. CD44 is a transmembrane hyaluronic acid receptor implicated in cell–cell and cell–matrix adhesion, cell motility and metastasis (62,63). In a rat schwannoma model, merlin-induced growth inhibition was dependent upon merlin interaction with the cytoplasmic tail of CD44 in the absence of other ERM proteins (49). However, in the growth permissive state, merlin forms a complex with ERM proteins and CD44. This preferential association of merlin with CD44 may specify whether cell proliferation (and perhaps other merlin-associated functions) or growth arrest occurs (50). The ability of merlin to associate with CD44 is potentially influenced by several factors, including the phosphorylation status of merlin (64), merlin intramolecular complex formation (65) and merlin’s interactions with other binding partners (56). We have shown that merlin’s ability to function as a growth regulator is related to its ability to form two intramolecular associations (65–67) (Fig. 4B). In this fashion, merlin may cycle between ‘open’ (impaired self-association) and ‘closed’ (productive self-association) conformations in vivo that differentially determine whether it forms hetero-oligomers with ERM proteins or other molecules to transduce merlin’s growth regulatory signal. This model of merlin folding is supported by recent crystallographic data for moesin (68).
Merlin also associates with HRS (hepatocyte growth factor-regulated tyrosine kinase substrate; alternatively called HGS) (54). Merlin binds to HRS through residues in the C-terminal domain not shared amongst ERM proteins, arguing that the HRS interaction is specific to merlin. HRS has been implicated in a signaling pathway initiated by hepatocyte growth factor (HGF) binding to the c-met receptor (69). Additionally, HRS may be involved in regulating growth factor receptor degradation (70), endosomal trafficking (71), the STAT pathway (72) and Smad signaling (73). As HGF is one of the most potent stimuli for Schwann cell proliferation and motility (74), HRS represents an attractive candidate for a merlin effector protein. We have demonstrated that regulated overexpression of HRS has the same effects on growth suppression and actin cytoskeleton-mediated processes as merlin overexpression (75). Furthermore, the association between merlin and HRS appears to be regulated by merlin folding, suggesting that merlin’s ability to cycle between ‘open’ and ‘closed’ states may integrate CD44 and HGF signaling pathways relevant to growth regulation (Fig. 4D).
Inactivation of the NF2 gene and loss of merlin expression has been reported for all NF2-associated tumors. In addition, loss of merlin is seen in nearly all sporadic schwannomas and 30–70% of sporadic meningiomas, suggesting that the NF2 gene is a critical tumor suppressor in these sporadic tumors (76–81). In sporadic schwannomas, loss of merlin is also observed in the early ‘tumorlet’ stage of schwannoma formation (82). Similarly, we have demonstrated that merlin loss is observed in a high percentage of sporadic low-grade meningiomas (83). Collectively, these results argue that NF2 loss is an early initiating event in schwannoma and meningioma tumorigenesis and that merlin may function as a ‘gatekeeper’ for Schwann cells and leptomeningeal cells in the same fashion as the adenoma polyposis coli gene functions in colon cancer. An improved understanding of how merlin regulates proliferation in these cells may yield critical insights into the early events of sporadic tumor formation.
Although much attention has been given to the idea that merlin is a member of the Protein 4.1 family and structurally related to the ERM proteins, it is clear that merlin, unlike traditional ERM molecules, operates as a tumor suppressor. This unique feature of merlin suggests that it might belong to a subfamily of Protein 4.1 molecules that function as tumor suppressors. RT–PCR cloning of transcripts differentially expressed in lung cancer identified a gene located on chromosome 18p11.3, termed differentially expressed in adenocarcinoma of the lung (DAL-1), which is deleted in a high proportion of lung and breast cancers (84) (Fig. 5A). Additional studies demonstrated that DAL-1 inactivation is frequently observed in sporadic meningiomas, but not schwannomas, and that DAL-1 loss is also an early event in sporadic meningioma pathogenesis (83,85) (Fig. 5B and C). DAL-1 represents an expressed splice variant of a larger Protein 4.1 molecule (Protein 4.1B), originally termed KIAA0987, but both DAL-1 (and KIAA0987) function in vitro as tumor suppressors and impair actin cytoskeleton-mediated processes as have been described for merlin. However, DAL-1 interacts with a different set of potential effector proteins (86), arguing that DAL-1 and merlin may have distinct mechanisms of action relevant to tumor formation.
Inherited cancer syndromes represent nature’s ultimate mutagenesis screen for critical growth regulators (tumor suppressor genes). The fact that individuals with these cancer predisposition syndromes develop specific tumors points to the unique roles of each of the implicated genes in growth regulation. Although the mechanisms of action differ between these tumor suppressors, they all share the property of being initiating events in the formation of specific tumors. By studying disorders like NF1 and NF2, we derive important insights not only into the neurofibromatoses, but also into the pathogenesis of sporadic cancers. Since functional inactivation of these tumor suppressor gene products results in tumor initiation, we learn about the cellular processes critical for the regulation of growth in normal cells. This information can be used to identify new families of potential tumor suppressors, as in the case of merlin and DAL-1, as well as to focus our attention on key intracellular events that might serve as targets for future rational cancer drug design.
The study of NF1 has also suggested that reduced tumor suppressor gene expression is insufficient for tumor formation, but may be enough to confer a growth advantage to specific cells, such as astrocytes (29,30) and myeloid cells (31). This observation raises the possibility that some tumors may result not from homozygous inactivation of one or two tumor suppressor genes, but from the combined effect of heterozygosity of multiple tumor suppressor genes. In this fashion, it is possible that sporadic cancers form from the accumulation of a limited number of single genetic ‘hits’.
One of the perplexing issues in NF1 and NF2 has been the degree of phenotypic variability in individuals from the same family with the identical genetic mutation. Some of this variation may reflect stochastic events, such as the timing of the second hit, but might also result from the effects of ‘modifying’ genes that cooperate with NF1 or NF2 inactivation to predispose to specific tumors. The identification of these ‘modifier’ genes will have an enormous impact on our ability to determine the risk of cancer and to estimate the probability of specific tumor formation in individuals affected with NF1 and NF2.
I would like to thank Dr Larry Sherman as well as the members of my laboratory for their critical reading of this manuscript. This work is funded by grants from the National Institutes of Health (NS36339 and NS34848).
Department of Neurology, Box 8111, 660 South Euclid Avenue, St Louis, MO 63110, USA. Tel: +1 314 362 7149; Fax: +1 314 362 9462; Email: firstname.lastname@example.org
- signal transduction
- cell proliferation
- cell growth
- genes, neurofibromatosis 2
- tumor suppressor genes
- tissue membrane
- neurofibromatosis 1
- neurofibromatosis 2
- neurofibromatosis type 1 protein
- neurofibromin 2
- receptors, cell surface
- tumor suppressor proteins
- tumor initiation