The family described here presents a very mild course of NF2 with slowly growing and late-onset vestibular schwannomas, no marked hearing deficit before 40 to 50 years of age, and few other tumours. NF2 is arbitrarily subdivided into mild, intermediate and severe forms, and in the mild form NF2 presents at age > 30 years59. Compared to the severe form, the milder disease associates with later mean age of onset of hearing loss (33 years vs. 22 years) and mean age at death (61 years vs. 42 years)141. The course of NF2 in the members of this family demonstrated remarkable clinical uniformity, which has been noted in NF2 families54,55,142. However, reports of families with both mild and severe phenotypes17,98,141,171, and monozygotic twins with a non-identical disease18 suggest that epigenetic, stochastic or environmental factors may modify the phenotype. However, in this large NF2 pedigree with eleven affected and six as yet manifestation-free individuals, it is unlikely that the mild phenotype would be due to genetic factors other than NF2. Furthermore, there is no evidence of modifying genes or environmental factors in NF2.
In general, NF2 provides an ideal model for the genotype-phenotype analysis of a tumour suppressor gene. Since somatic inactivation of the NF2 gene is typically a large loss in 22q, the inherited germ-line NF2 mutation should primarily dictate the clinical outcome. The penetrance of NF2 is virtually complete, and limited variation of the phenotype is seen within a single family. The family in this study is one of the few large and well-characterized NF2 pedigrees and thus especially suitable for genotype-phenotype correlation. This mild NF2 disease was caused by a splice donor site mutation at intron 15 in the NF2 gene. In accordance, mild NF2 has been associated with exon 15 splice donor mutations and skipping of exon 1597,162. In these families the disease course is similar, although clinical variability frequently associates with splice site mutations99. The mutation gave rise to the expression of two transcripts in fibroblasts, a novel mutant transcript with exon 16 and 17, and the overexpression of isoform III. Also previously, overexpression of isoform III has been seen to occur with intron 15 splice donor mutations (1737 AGgt to ATgt and 1737+12 ins 220 bp), but no novel transcripts were detected88,97. This might be due to different splicing mechanisms in the tissues analyzed (fibroblasts vs. lymphoblasts and schwannoma), or due to different transcriptional consequences.
The expression of the aberrant transcripts and merlin in the patient fibroblasts at levels comparable to wild-type protein and in a schwannoma shown in this study, agrees with the concept that certain mutations, theoretically missense or in-frame changes, may result in protein products with partially retained tumour suppressor function69. This is supported by the presence of truncated and full-length merlin in a subset of schwannomas71,75,85. This highlights the importance of studying the mutational consequences at the protein level, which may aid in the prediction of the disease course and reveal domains important for the tumour suppressor function.
In the presented family, both transcripts from the mutant allele encode the C-terminus of isoform III. The decreased proliferation rate of tumours in association with the C-terminus of isoform III suggests that this isoform has partially retained the tumour suppressor function. So far, there is no information on the function of isoform III. Interestingly, transgenic mice expressing merlin truncated at the C-terminus demonstrated no tumorigenesis, and also the growth suppressive functions of Drosophila merlin are contained within the N-terminal domain65,107.
Multiple schwannomas have been suggested to represent a third form of NF, schwannomatosis, although some authors have considered it to be an attenuated form of NF258,147. In agreement with previous reports on sporadic schwannomatosis58,120, 139,215, the patients in this study had multiple schwannomas but no other manifestations of NF2 including meningiomas and ocular abnormalities, which are found in >50% and >90% of NF2 cases, respectively54,131,141,150. The favourable prognosis, presentation in middle-age and no intracranial tumours has been also recognized by others58,82,89,120,139,215. This does not distinctively rule out NF2 because, in NF2, tumours may initially appear elsewhere than on the vestibular nerves, and in late-onset families, members with an NF2 gene mutation without tumour manifestations are occasionally found in their seventies55,57,162. Although the clinical aspects in this study were not supportive of classical NF2, the schwannomas displayed histological features associated with NF2: more lobular "grape-like" growth pattern, and more Verocay bodies, uncommon in sporadic schwannomas186. Patients with two schwannomas (three of the nine cases in our series) may represent chance occurrence against the annual incidence of 0.3 - 0.4/100 000 of symptomatic spinal schwannomas177 (the incidence of occult spinal or peripheral schwannomas is unknown). None of our patients had a positive family history, although a subset of schwannomatosis is known to be familial58,89,215 as also shown by Antinheimo in a population-based study in Finland. Schwannomatosis families have been shown to associate with the NF2 gene in genetic linkage analyses58.
Recently, the molecular analysis of tumours from schwannomatosis patients suggested four alternative molecular backgrounds in schwannomatosis89. (1) The disorder may not be due to the inactivation of the NF2 gene, as three tumours of one individual with a highly aggressive disease did not display any NF2 gene abnormality, seen in > 90% of schwannomas88. This is supported by the detection of schwannomas with no alterations in the NF2 gene but deletions elsewere on 22q36. (2) It may be due to somatic mosaicism, as multiple tumours at anatomically different locations displayed the same NF2 mutations, and in one case the mutation was detected at an extremely low level in the lymphocytic lineage89. Indeed, somatic mosaicism has been recently proposed to be a fairly common (15% of cases) phenomenon in NF260. (3) Multiple tumours may be a noncontinuous spread of a single tumour, when tumours with the same NF2 mutations are restricted to a confined anatomical region. This has not been shown for schwannomas, but may explain multiple meningiomas in sporadic cases187,206,220. (4) Schwannomatosis may be due to an inherited predisposition to schwannomas, because in some familial cases tumors display different somatic mutations but LOH at always the same allele89.
No germ-line NF2 gene mutations were found in the present series, in line with other reports of sporadic or familial schwannomatosis58,89. Thus, classical NF2 mutations are unlikely, although the current mutation detection methods miss one-third of the mutations142. Occasional NF2 mutations have been identified in schwannomatosis82,147, but because some of the cases have later turned to be NF2, there is a high probability of late-onset classical NF2 in these cases97-99. Somatic mosaicism, shown to occur occasionally in sporadic schwannomatosis89, could not be excluded in this study because mutation search by exon sequencing does not detect mutations at a very low level. Also, lower mutation detection rates in mild than in severe NF2142 suggest that other mutational mechanisms not detectable by the current methods, such as large deletions of the NF2 gene, may be involved in schwannomatosis. Altogether, sporadic schwannomatosis seems to be distinct from the classic NF2, with a more favourable prognosis.
Based on the current knowledge and our observations, slight modifications can be suggested to the clinical diagnostic criteria for schwannomatosis, as proposed by Jacoby et al.89 to strengthen the discrimination to mild NF2, segmental NF2 and chance tumour occurrence. Furthermore, evidence of any intracranial tumours, especially meningiomas at young age, or other NF2 manifestations should raise suspicion of classic NF2.
The modified criteria of schwannomatosis:
1. three or more pathologically diagnosed schwannomas without an anatomically limited distribution and
2. lack of radiographic evidence of vestibular schwannoma, at age >18 years, and meningioma at age < 30 years
The NF2 tumour suppressor protein, merlin, shares structural characteristics with the ERM family of membrane-organizing proteins. Thus far, limited information has been available on its cellular distribution and functions. This study shows that merlin has a subcellular distribution beneath the cell membrane and at specialized cellular protrusions, similar to that of ERM-family members6,62,140,214. The observed distribution is in accordance with the preferential membrane-bound localization of the transfected merlin at the dorsal surface of COS-1 cells46 and at the ruffling edges in primary meningioma cells66. Furthermore, our experiments demonstrated that merlin in most subcellular regions colocalized with ezrin. The result is not surprising as ezrin, radixin and moesin are able to dimerize with each other7,63,143. Furthermore, an interaction between merlin and ezrin and the ability of head-to-tail dimerization in vitro and in vivo has been demonstrated67. The apparent discrepancy in staining results with Gonzalez-Agosti et al.66, who found a non-overlapping distribution for merlin, ezrin and moesin in a primary meningioma cell line, probably reflects cell type or stimulation specific variation in ERM protein distribution. We also observed a difference in the distribution in response to different growth substrata. When COS-1 cells were grown under conditions which result in poor development of the F-actin network, merlin appeared to redistribute ezrin from sites where it is located in untransfected cells. This is of interest as ERM proteins demonstrate interchangeability and functional redundancy78,194. For instance, overexpressed radixin displaces moesin from microvilli and filopodia78. In this respect, merlin shows functional similarity to other members of the ERM family.
The association of merlin with the cytoskeleton was indicated by its resistance to detergent extraction and by coclustering with F-actin after cytochalasin B treatment. The partial detergent-insolubility was alike the behavior of ezrin and in line with previous reports5,46. The presence of merlin in the soluble and cytoskeleton-associated fraction, like other ERM proteins, suggests that also merlin undergoes a change in conformation that could coincide with the translocation between the soluble pool and the membrane-skeleton. The ERM proteins are in their phosphorylated form when associated with the membrane-skeleton and revert to the soluble pool following dephosphorylation33,122,200,204. Intrestingly, merlin has been shown to change rapidly between phosphorylated and unphosphorylated state in response to growth inhibition stimuli by serum starvation and loss of cell adhesion179.The association between merlin and F-actin-containing cytoskeleton is evident, but whether a direct interaction exists between merlin and actin remains to be determined. The C-terminal F-actin-binding site of ERM proteins (ezrin residues 552-585) is only partially retained in merlin, but the villin/ERM actin-binding motif at this site is conserved201,202. Evidence against a funtional actin-binding site is suggested when C-domain expressed as GST-fusion proteins or by in vitro transcription/translation does not bind polymerized actin217(Turunen, unpublished). Recently, we have detected that merlin interacts with globular b-actin and filamentous a-actin by the N-domain residues 1-339 (Zhao, Sainio, Grönholm, Carpén, unpublished), in line with Xu et al., who detected an actin-binding domain at residues 178-367217. It is tempting to propose that the homologous RKKK cluster (residues 299-328 in merlin), critical for the actin-binding of ERM proteins, is the likely actin binding site also in merlin124,156. Furthermore, merlin (residues 519-590 of isoform II) binds also to another component of the cytoskeleton, bII-spectrin172.
The ERM family members bind to the cytoplasmic domain of transmembrane glycoprotein CD4479,198, and in this study such an association was also shown for merlin. Furthermore, we have recently verified the direct interaction between CD44 and the N-terminal domain of merlin by the yeast two-hybrid method (Sainio, Grönholm, unpublished). The interaction of merlin and CD44 and merlin-dependent redistribution of CD44 suggests that merlin could regulate the adhesive functions of transmembrane molecules by regulating their distribution. The aberrant expression of CD44 is associated with tumour progression and metastasis158. The intracellular interactions of the cytoplasmic domain of CD44 are known to regulate the ultrastructural localization of CD44 on the cell surface, and to affect cell motility and cell-matrix adhesion205. Thus, the tumour suppressor mechanism of merlin could be comparable to that of the APC protein, which in association with the E-cadherin/catenin complex, is associated with colon carcinogenesis19. This would provide an explanation to the impairment of cell adhesion subsequent to antisense oligonucleotide inhibition of merlin expression or expression of mutant merlin83,103. In an analogous model, transfection of ezrin to thymoma cells leads to clustering of ICAM-2 and an increase in ICAM-2-dependent adhesion77. Another mechanism may be that merlin and ERM proteins via their interaction with molecules containing PDZ domains (Na+-H+ exchanger regulatory cofactor, hNE-RF or EBP50) regulate the outside-in signalling events by clustering transmembrane proteins and key components of downstream transduction pathways34,136,152.
The membrane-cytoskeleton linker mechanism of merlin is thus similar to that of ERM proteins (Figure 3). Interestingly, NF1 tumour suppressor protein, neurofibromin, has recently been shown to associate with epithelial cell adhesion site components, CK 14, desmoplakin and b4-integrin in differentiating keratinocytes104. Thus, neurofibromin may also have a role in connecting cytoskeletal components and cell adhesion molecules.
Figure 3.Membrane-cytoskeleton linker model for merlin. In the cortical cytoskeleton, merlin and ERM proteins form head-to-tail oligomers. They interact with cytoplasmic tails of transmembrane adhesion molecules (CD44) with the N-terminal FERM domains. An F-actin binding site in FERM-domain anchors actin microfilaments.
A major function of ERM proteins is the regulation of dynamic cell surface processes. In present experiments, transfected merlin induced morphological changes which do not occur with wild-type ezrin or radixin78,124. However, in non-adherent thymoma cells, transfected wild-type ezrin induces cells to form uropod-like extensions77. Transfected cells contained membrane blebs and their cell body was significantly elongated. Such changes in cell length must be associated with reorganization of the cytoskeleton, apparently regulated by merlin. The morphogenic effect of merlin is also seen in NIH-3T3 transfectants116. In further characterization of the merlin domains responsible for the morphogenic effects, such as long cell surface projections or blebbing (Zhao, Sainio, Carpén, unpublished), revealed striking similarities with the morphogenic effects induced by expressed domains of ezrin124,125.