Article Text

Download PDFPDF

Mosaicism in sporadic neurofibromatosis type 1: variations on a theme common to other hereditary cancer syndromes?
  1. H Kehrer-Sawatzki1,
  2. D N Cooper2
  1. 1
    Institute of Human Genetics, University of Ulm, Ulm, Germany
  2. 2
    Institute of Medical Genetics, School of Medicine, Cardiff University, Cardiff, UK
  1. Dr H Kehrer-Sawatzki, Institute of Human Genetics, University of Ulm, Albert-Einstein-Allee 11, 89081 Ulm, Germany; hildegard.kehrer-sawatzki{at}uni-ulm.de

Abstract

Mosaicism constitutes a frequent complication of the genotype–phenotype relationship in genetic disease and is an important consideration for the estimation of transmission risk. Mosaicism has been identified in several hereditary cancer syndromes including retinoblastoma, familial adenomatous polyposis coli, von Hippel–Lindau disease and neurofibromatosis type 2. Recent data support the postulate that the frequency of mosaicism is increased in cancer predisposition syndromes characterised by high new mutation rates. Since the new mutation rate is very high in neurofibromatosis type 1 (NF1), mosaicism might reasonably be expected to be frequent among sporadic cases but this remains to be formally demonstrated. Here we summarise current knowledge of mosaicism in NF1, focusing on the types of mutations identified as well as their inferred developmental timing and representation in different cell types, and assess the potential impact of high frequency mosaicism on mutation screening in patients with apparent de novo NF1.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Neurofibromatosis type 1 (NF1, MIM 162200), a condition characterised by clinical symptoms affecting primarily cells of neural crest origin, affects ∼1 in 3000–4000 individuals worldwide.13 Among the clinical features pathognomonic for NF1 are café-au-lait macules (CALM), axillary and inguinal freckling, Lisch nodules and the eponymous neurofibromas—benign tumours arising as a consequence of the biallelic inactivation of the NF1 gene in Schwann cells.46 The majority of neurofibromas are of a dermal subtype; they increase in size and number with age but never become malignant. However, ∼30% of NF1 patients suffer from congenital plexiform neurofibromas7 8 which have the potential to undergo transformation into malignant peripheral nerve sheath tumours (MPNSTs). Indeed, some 4–5% of plexiform neurofibromas become malignant.911 Other classical NF1 associated clinical features include pheochromocytoma, optic gliomas, and specific bone lesions of the sphenoid wing, vertebrae and tibia. NF1 associated osseous lesions may vary in terms of their severity and are often progressive.12 Learning disability and sustained attention difficulties constitute a problem for a substantial proportion of children with NF1.13 14

NF1 is caused by mutations in the NF1 gene on 17q11.2 which encodes neurofibromin, a negative regulator of Ras GTPases.15 Since the majority of NF1 gene lesions prevent the expression of intact neurofibromin, the functional disruption of neurofibromin constitutes the underlying cause of NF1. NF1 is a classic tumour suppressor gene, and its biallelic inactivation has been confirmed in the cells of associated tumours, CALM and bone lesions.6 1625 The NF1 gene also exhibits a very high new mutation rate, estimated at between 1.4–2.6×10−426 and 3–5×10−5,27 much higher (10–100 fold) than is typical for human disease gene loci.28 The size and complexity of the NF1 gene are insufficient on their own to account for this unusually high new mutation rate. Some 30–50% of NF1 patients come to clinical attention as sporadic patients as a consequence of new mutations, and it has been suggested that a high proportion of these new mutations are actually somatic.29 As yet, however, few data are available to support this assertion.

In what follows, we shall review current knowledge of mosaicism in NF1. A mosaic may generally be defined as an individual who is composed of two or more genetically different cell populations (or lineages) derived from the same zygote. In the specific context of NF1, mosaicism implies that, in addition to the normal cells, cells harbouring an NF1 gene mutation may also be found in the soma of a given patient. In the literature, mosaicism is often described as being either germline, somatic (involving somatic cells but not the germline), or gonosomal (involving both somatic cells and the germline). Since information on the involvement of the germline is often lacking in individuals displaying evidence of somatic mosaicism, we have here declined to make these distinctions except where appropriate and have instead simply used the term “mosaicism”. We have attempted to place particular emphasis on the types of mutation identified in mosaic individuals, the proportions of cells bearing NF1 gene mutations in different tissues, and the inferred developmental timing of the mutational events. Finally, in the context of estimates of mosaicism frequency in other hereditary cancer syndromes, such as tuberous sclerosis complex, familial adenomatous polyposis coli (FAP), retinoblastoma, von Hippel–Lindau disease, and neurofibromatosis type 2, we shall assess the potential impact of high frequency mosaicism on mutation screening in patients with apparent de novo NF1.

MOSAICISM IN NF1

Mosaicism in NF1 can present as a mild yet generalised manifestation of the disease with typical symptoms such as neurofibromas and/or freckling and CALM that are not simply confined to a few body regions (“mosaic generalised NF1”). Clinically, it can be difficult to distinguish these mosaic patients from patients with an inherited NF1 gene mutation. Another form of mosaicism, termed “mosaic localized” or “segmental” neurofibromatosis, is characterised by stigmata of NF1 that are limited to one or a few areas of the body.26 3039 The prevalence of segmental NF1 has been estimated to be ∼0.002%,4042 about 10–20 times lower than the prevalence of generalised NF1. In segmental NF1, the affected body segment may exhibit either pigmentary changes or neurofibromas on their own, or alternatively neurofibromas and pigmentary changes together.34 Plexiform neurofibromas without other stigmata of NF1 have also been described34 43 and are probably caused by somatic mosaicism for an NF1 gene mutation.

Mosaicism that is confined to the germline would appear to be extremely uncommon in NF1 since only a few families have been reported with more than one affected child whose parents lack clinical signs of the disease.27 44 Germline mosaicism has, however, been confirmed by mutation analysis in the case of a clinically unaffected father who transmitted an identical 12 kb multi-exon deletion of the NF1 gene to two of his children.45 46 Although this lesion was identified in 10% of spermatozoa from the father, it was undetectable in his lymphocytes.

Somatic mosaicism in NF1 originates from postzygotic NF1 gene mutation. Mosaicism due to postzygotic back mutation of the NF1 gene (revertant mosaicism), as observed in several other human diseases,48 has not as yet been demonstrated in NF1. All the NF1 gene mutations that have been identified so far in patients exhibiting mosaicism are summarised in table 1.

Table 1 NF1 gene mutations identified in neurofibromatosis type 1 (NF1) patients with somatic mosaicism

TYPES OF NF1 GENE LESION IDENTIFIED IN MOSAIC NF1

Most of the mutations identified to date in NF1 patients with mosaicism are gross deletions of the NF1 gene region (Table 1). However, this finding is almost certainly subject to ascertainment bias since mosaic gross deletions are comparatively easy to identify by fluorescence in situ hybridisation (FISH) analysis (since single cells can be analysed). By contrast, mosaicism due to subtle intragenic NF1 gene mutations that are only present in a low percentage of cells is difficult to detect by classical analytical methods such as DNA sequencing. No dramatic mutation hotspot has been identified within the NF1 gene that would potentiate targeted mutation analyses to detect very low numbers of mutated cells. Furthermore, a considerable proportion of subtle intragenic lesions alter the NF1 splicing pattern, a proportion that has been estimated to be between 19–50%.4954 It follows that cDNA analysis (inter alia) is a prerequisite for efficient mutation screening in NF1.

Somatic gross NF1 gene deletions have been detected in patients with generalised mosaicism as well as in patients with segmental NF1, whereas somatic subtle lesions in the NF1 gene have been identified only in patients with segmental NF1 (table 1). This may reflect the fact that subtle somatic NF1 gene lesions in patients with segmental NF1 are generally found as a consequence of inordinate care being taken, in a research context, to identify the underlying causative lesions in such cases. In other words, the most laborious screening procedures have been applied in those cases where the segmental manifestations already suggested the presence of mosaicism. The number of mosaic patients investigated is, however, still very low (n = 26) and it remains likely that subtle somatic intragenic NF1 mutations will also be found in patients with generalised mosaicism and NF1 symptoms unrestricted to one or a few body segments. While the high proportion of gross deletions noted in mosaic cases of NF1 may well be due to ascertainment bias, we may nevertheless infer that large somatic NF1 deletions are not infrequent. In most instances, these deletions do not merely encompass the NF1 gene but also extend into its flanking regions. In those cases where the deletion boundaries have been precisely demarcated, the breakpoints have been found to be located within one of the NF1-REPs or SUZ12 sequences (table 1). It would thus appear that these segmental duplications have rendered the NF1 gene region recombinationally unstable, thereby facilitating the occurrence of somatic gross NF1 deletions.

Two types of recurrent NF1 deletion are known which differ with respect to the positions of the respective breakpoints: in type 1 deletions, breakpoints are located in NF1-REPa and c, whereas in type 2 deletions, the breakpoints map to the SUZ12 gene and its pseudogene. The highly homologous NF1-REPs and SUZ12 sequences flank the NF1 gene region and hence constitute target sequences for non-allelic homologous recombination (NAHR); together, they are responsible for more that 90% of all gross NF1 gene deletions.5559 However, type 1 deletions occur exclusively during meiosis60 and have so far only been detected as constitutional mutations, never in patients with mosaicism. By contrast, type 2 deletions arise exclusively during mitosis for reasons that are currently still unclear.59 A third type of large NF1 deletion, termed “atypical deletions”, are usually characterised by non-recurring breakpoints and are not known to have arisen through NAHR between highly homologous sequences located at the respective breakpoint sites. Nevertheless, one of the two breakpoints of these atypical deletions is frequently located within either an NF1-REP or SUZ12 sequence. In patients with mosaicism who manifest a gross NF1 gene deletion, the deletion in question has invariably been either type 2 or atypical (table 1).

EPIMUTATION AS A POTENTIAL MECHANISM CAUSING MOSAICISM IN NF1

While mosaicism is known to be caused by postzygotic NF1 gene mutation, it could also arise as a consequence of epimutations—that is, via alterations to the normal epigenetic modification pattern on the chromatin that can be inherited when cells divide and proliferate. Epimutations are important in tumorigenesis and typically occur at rates 1–2 orders of magnitude higher than somatic mutations.61 62 Although hypermethylation of the entire NF1 gene promoter region does not appear to occur during NF1 tumorigenesis,6365 it remains entirely possible that subtle site specific epimutations could still result in the somatic inactivation of the NF1 gene. Thus, Horan et al63 reported that tumour specific CpG methylation of six distinct CpG sites was present at positions −609, −429, 406, −383, −331 and −315 relative to the transcriptional start site of the NF1 gene in tumour tissue. Immediately 5′ to the transcriptional start site of the NF1 gene, mutation of the SP1 and CRE binding sites has been shown to give rise to a 70–90% decrease in reporter gene activity,66 while site specific methylation of these sites is sufficient to inhibit binding of their cognate transcription factors.67 More recently, site specific methylation in the NF1 gene promoter, involving the transcription factor binding sites for SP1 (−138), CRE (−10) and AP-2 (−208 and −55), has been noted in both dermal and plexiform neurofibromas.64 Finally, Fishbein et al65 have presented evidence for mosaicism of site specific methylation at positions −208, −138 and −55 on the NF1 gene promoter. Thus, changes in the site specific methylation of the NF1 promoter established at an early stage during development could, at least in principle, contribute to mosaicism by reducing NF1 gene expression.

TISSUE DISTRIBUTION AND CELL TYPES AFFECTED IN MOSAIC NF1 PATIENTS

Maertens et al23 recently performed the most comprehensive analysis to date of the representation of pathological somatic NF1 gene lesions in different cell types in individuals with segmental NF1. These authors investigated patients with segmental NF1 and neurofibromas alone (patient SNF1–1), pigmentary defects alone (SNF1–2) and pigmentary anomalies and neurofibromas together (SNF1–3), as summarised in table 1. Their analysis indicated that in segmental NF1, the proportion of mutated cells in non-neural crest derived tissues can often lie below the detection level of routine NF1 mutation screening techniques (<20%), a finding which is in accord with observations made in other cases of segmental NF1.33 35 68 However, by examining Schwann cells and/or melanocytes isolated from neurofibromas or CALM, respectively, Maertens et al23 successfully identified the mutations in a proportion of cells. The study of Maertens et al23 emphasised the importance of screening the relevant cell types (that is, Schwann cells and/or melanocytes) in cases of segmental NF1 in order to identify the underlying NF1 gene mutations. As summarised in table 1, in all cases of segmental NF1 so far investigated, cells harbouring the somatic NF1 gene lesion were either undetectable in peripheral blood lymphocytes or present only at a very low level (2–4%). By contrast, in patients with mosaic generalised NF1 and large type 2 deletions, the percentage of cells with the deletion is generally high in peripheral blood lymphocytes (70–100%),59 but lower in cells from other tissues such as buccal epithelia or skin fibroblasts.

Currently, it is unclear whether this phenomenon is specific to type 2 deletions or whether this trend also holds true for large NF1 gene deletions in general, including those of atypical type. As yet, an insufficient number of mosaic patients with atypical NF1 deletions have been systematically investigated in order to be able to determine whether the proportion of cells with the deletion is significantly higher in blood cells compared with other tissues, as observed in patients with type 2 deletions. Interestingly, type 2 NF1 deletions have recently been identified as somatic tumour specific events that can inactivate the NF1 gene in paediatric T cell acute lymphoblastic leukaemia and acute myeloid leukaemia in patients in the absence of neurofibromatosis.69 It may well be that type 2 NF1 deletions confer a proliferation advantage upon haematopoietic progenitor cells. This growth advantage could also be responsible for the higher proportion of deletion bearing cells in the peripheral blood lymphocytes of NF1 patients with mosaicism for this type of deletion. However, not only growth advantage but also cellular senescence could potentially influence the proportion of mutated cells in different tissues of patients with mosaicism.

TIMING OF THE MUTATIONAL EVENT IN DIFFERENT FORMS OF MOSAIC NF1

Irrespective of any selective advantage that may be conferred by type 2 deletions in haematopoietic cells, the very mild but generalised skin manifestations which characterise patients with type 2 deletions59 70 71 can provide specific information about the timing of the deletions during development. Since the skin/neural crest derives from the embryonic ectoderm whereas blood cells are mesodermal derivatives, we may infer that type 2 deletions are likely to have occurred at an early stage of embryonic development. The ectoblast/ectoderm and mesoblast/mesoderm are visible as distinct cell layers by the third week of human embryonic development.72 In cases where type 2 deletions are found in cells derived from both these embryonic germ layers (as deduced from the phenotype of the patients), then these deletions may reasonably be expected to have occurred in the epiblast before the third week. A brief overview of the origin and derivation of tissues in presomitic human embryos is given in fig 1. If the NF1 deletion has indeed occurred at this early developmental juncture, it is likely that the primordial germ cells (PGCs) will also harbour the deletion. Before gastrulation, the PGCs are located in the extreme proximal region of the epiblast adjacent to the extraembryonic ectoderm. Early in gastrulation, however, the PGCs move towards, and then through, the primitive streak into the extraembryonic mesoderm.73 Since three or more ancestral cells are assumed to be set aside to form PGCs, mosaicism may also affect this group of cells. Such a combination of somatic and germline (gonosomal) mosaicism has previously been reported several times in NF1—for example, three cases of the transmission of a type 2 NF1 deletion from a mother with minor skin manifestations to a child with severe clinical symptoms of the disease.59 70 71

Figure 1 Origin and derivation of tissues in presomitic human embryos. Modified from Luckett.136

In segmental NF1, the mutation is likely to have occurred at a somewhat later stage in development by comparison with mosaic generalised NF1 (Ruggieri and Huson34). However, this does not necessarily mean that the parent-to-offspring transmission risk is any lower in segmental NF1 as compared to mosaic generalised NF1. Gonosomal mosaicism may well also be found in segmental NF1. Indeed, Ruggieri and Huson34 described five patients with segmental NF1 who presented with a combination of CALM and freckling or localised skin hyperpigmentation as the only symptoms, and who were clinically ascertained through their children who presented with classical NF1. Notably, the skin segment displaying the pigmentary anomalies did not spatially co-localize with the gonads in these cases. Consoli et al68 reported on a patient with segmental NF1 and pigmentary anomalies in the absence of neurofibromas. The underlying mutation (R1947X) was not detectable by direct sequencing of polymerase chain reaction (PCR) products amplified from the peripheral blood, fibroblasts and keratinocytes of the patient. However, sequence analysis of cloned PCR products, amplified from each cell line, revealed the presence of the mutation in a small proportion (9–20%) of clones. Despite this low level of somatic mosaicism, the patient passed the mutation on to her daughter who displayed classical generalised manifestations of NF1. What is very difficult to explain pathogenetically is the vertical transmission of mosaic localised or segmental NF1 which has so far been observed in five families.34

Clearly, further systematic analyses are called for in cases of segmental NF1 and mosaic generalised NF1, particularly in those cases caused by somatic type 2 deletions. Such analyses will be required in order to determine the tissue distribution of cells bearing the pathological NF1 gene lesion, their relative proportions in different tissues, the possibility of clonal selection, and any developmental stage related changes in the proportions of normal versus mutated cells in different tissues. Such studies would certainly improve the accuracy of estimates of the transmission risk and hence the genetic counselling of affected families.

MOSAICISM IN OTHER HEREDITARY CANCER SYNDROMES

Mosaicism involving mutations in genes predisposing to a variety of different diseases is increasingly recognised to be a frequent complication of the genotype–phenotype relationship in molecular medicine.47 48 7482 Such mosaicism has acquired particular importance in the context of hereditary cancer syndromes—for example, tuberous sclerosis complex, familial adenomatous polyposis coli (FAP), retinoblastoma, von Hippel–Lindau disease, and neurofibromatosis type 2.8393 In common forms of hereditary cancer predisposition such as breast cancer and Lynch syndrome, mutational mosaicism has not so far been reported, but this may be due to a referral bias since detailed mutation analysis is not routinely performed in sporadic cases. The identification of mosaicism and the determination of its frequency are, however, extremely important since they are likely to exert a significant influence on the variability of clinical expression of a given condition as well as on its transmission risk.94

A high rate of mosaicism has recently been identified in sporadic patients with FAP.90 93 95 Somatic mutational mosaicism in the APC gene has been detected in 8/75 sporadic patients with FAP, suggesting a mosaicism frequency of ∼11%.90 Subsequently, in a comprehensive screen of 242 index patients with APC mutations, mosaicism was verified in 10 cases, representing ∼4% of the 242 patients or 21% of the apparently sporadic cases.93 These findings suggest that the rate of mosaicism in FAP is somewhat higher than previously suspected; indeed, it may occur in up to 21% of apparently sporadic patients. In the aforementioned study, four of the 10 mosaic patients exhibited an attenuated FAP phenotype characterised by a smaller number of adenomas (<100) and/or an advanced age at diagnosis. One of the 10 patients with mosaicism for an APC mutation was asymptomatic whereas five exhibited classical FAP.93

In NF2, not only is the new mutation rate high (49%),96 but also the prevalence of mosaicism. Methodical studies on large numbers of cases have indicated that ∼33% of sporadic patients with classical NF2 and bilateral vestibular schwannomas are mosaics.87 88 91 Among the NF2 patients with unilateral vestibular schwannomas, the proportion of mosaics is even higher, amounting to ∼60%.91

Taken together, these observations support the postulate that high frequency mosaicism is often to be found in cancer predisposition syndromes with high rates of new mutations (summarised in table 2). Although this idea has previously been alluded to (for example, Verhoef et al84) it does not appear ever to have been either explicitly stated or formally presented before. If the apparent correlation between a high frequency of mosaicism and a high new mutation rate in cancer predisposition syndromes is borne out by further work and is consequently shown to be a generalisable phenomenon, one explanation could be the existence of a direct relationship between the mutation rate in the soma and that in the germline. Presumably, many of the factors that influence the mutation rate of a specific gene are structural and hence constant between the soma and the germline—for example, exon/intron size and number,97 CpG dinucleotide frequency,98 chromosomal location, the presence of repetitive sequence elements or paralogous gene(s) in the immediate vicinity, the occurrence of mutation inducing sequence motifs within the gene, and intragenic sequence repetitivity.99 However, other factors may well differ between the germline and the soma—for example, CpG methylation status, chromatin structure, somatic mutational (microsatellite) instability, selection for cellular proliferation—leading to differences between both the somatic and germline mutational rates and the corresponding mutational spectra. Consistent with this conceptual framework, differences as well as similarities have been observed when comparing the germline and somatic mutational spectra of tumour suppressor genes.98 100102

Table 2 Cancer predisposition syndromes with high de novo mutation rates

In NF1, the germline and (tumour associated) somatic mutational patterns of intragenic NF1 micro-lesions appear to be remarkably similar in terms of mutation type, relative frequency of occurrence, and the putative underlying mechanisms of mutagenesis.99 However, if the germline mutation is a large deletion in the NF1 gene region, there is a tendency for the tumour associated somatic mutation to be an intragenic (and frequently frame shift causing) mutation.22 Differences between the somatic (tumour associated) and germline mutational spectra are also evident for a number of other tumour suppressor genes.98 A significantly higher rate of CGA→TGA nonsense mutations has also been noted among APC somatic mosaic mutations as compared to germline mutations.93 In any consideration of the somatic mutational rates and spectra for certain genes, the developmental timing of the mutation may well be important. Clearly, somatic mutations that lead to detectable mosaicism are likely to occur much earlier (probably during embryonic development) than somatic mutations associated with tumorigenesis. A good example of the time dependency of the somatic mutational spectrum has been reported in NF2: nonsense mutations appear to be more common than frameshift mutations in both classical and somatic mosaic NF2.102 By contrast, in patients with unilateral sporadic vestibular schwannomas (USVS) who do not fulfil the diagnostic criteria for NF2, frameshift mutations in the NF2 gene are more frequent than nonsense mutations. Somatic NF2 mutations giving rise to USVS are likely to occur much later than mutations that lead to somatic mosaic NF2. Importantly, in patients with USVS, the ratio of somatic frameshift to nonsense mutations was found to increase with age at diagnosis, consistent with an age related diminution in efficiency of some components of DNA repair capacity.102 Taken together, these observations indicate that the impact of different somatic mutational mechanisms may vary over time—for example, during early embryonic development versus later somatic mitotic divisions.

The action of different mutational mechanisms is also evident in terms of gene specific differences in observed mutational spectra between the germline and the soma. For example, type 2 NF1 deletions, that encompass 1.2 Mb and have breakpoints within the SUZ12 sequences, usually arise intrachromosomally during postzygotic cell divisions and are associated with mosaicism.59 By contrast, type 1 NF1 deletions span 1.4 Mb and have breakpoints within the NF1-REPs. Type 1 deletions occur exclusively during maternal meiosis I and are observed as constitutional deletions associated with a severe manifestation of the disease.57 60 Although non-allelic homologous recombination (NAHR) is responsible for both deletion types, a clear positional preference is apparent when mitotic and meiotic NAHR in the NF1 gene region are compared.59

Another consideration is that whereas endogenous mechanisms of mutagenesis may reasonably be expected to impact upon both the germline and the soma, exogenous mutagens are likely to exert a disproportionate effect on the somatic mutation rate. Finally, mutation rates, whether in the germline or the soma, are also going to vary quite widely between individuals as a consequence of the myriad of genetic differences in the general population that can affect DNA repair103 and/or apoptosis.104 Indeed, somatic mutational spectra have been found to differ between geographically separated populations implying differences in the underlying mutational mechanisms.105

The postulate that mosaicism is frequent in those hereditary cancer syndromes that exhibit high new mutation rates is as yet based upon a relatively small number of conditions (table 2). If, however, this idea comes to be supported by data from further studies, it is likely to have considerable impact on the search for the molecular correlates of disease features such as mild or attenuated phenotypes in other forms of hereditary cancer.

A HIGH RATE OF MOSAICISM IN PATIENTS WITH SPORADIC NF1: A HYPOTHESIS

If mosaicism does indeed turn out to be frequent in disorders characterised by high rates of new mutation, then it follows that mosaicism could also play an important role in sporadic NF1 as first postulated by Zlotogora.29 Four distinct possibilities could account for the origin of sporadic cases of NF1:

  1. The mutation could have been inherited from an unaffected parent with germline mosaicism for this mutation, ensuring that the patient would appear to be a sporadic case but would nevertheless be constitutionally heterozygous for the NF1 gene lesion.

  2. The mutation could have occurred de novo in one of the parent’s germ cells that was subsequently involved in fertilisation. This resulting offspring would also be constitutionally heterozygous for the NF1 gene lesion rather than mosaic.

  3. The mutation could have occurred postzygotically during very early cell divisions but then the mutated cells could have subsequently outgrown the normal cells with the result that the mosaicism in the sporadic patient would no longer be detectable.

  4. The mutation could have arisen postzygotically and hence would be detectable in different tissues of the sporadic patient who would consequently be mosaic for this mutation.

Considering the relative likelihoods of these scenarios, possibility 1 is likely to account for only a very small proportion of sporadic NF1 cases, since mosaicism confined to the germline has only rarely been observed.27 44 The frequency of mosaicism among sporadic NF1 patients as described in possibility 4 is so far unknown but may explain the following aspect of NF1: the vast majority of new mutations (specifically, subtle intragenic NF1 gene mutations) are of paternal origin105107 but the new mutation frequency does not correlate with paternal age.27 Since the mutation rate in the paternal germline increases with age,28 it follows that a considerable proportion of new mutations may not occur during paternal germline associated cell divisions, but may instead occur postzygotically.29 The predominance of new mutations occurring on the paternal chromosome may then be explained by gender specific differences in DNA modification of the NF1 gene during early embryonic development, but this has not so far been proven.

A high mosaicism frequency in NF1 may well influence the mutation detection rate, which should in principle be higher in familial cases than in sporadic patients, as observed in neurofibromatosis type 2.108 109 In principle, the identification of the NF1 mutation ought to be easier in the second generation because the mutation should invariably be present in 100% of the patient’s cells. By contrast, low level mosaicism in sporadic patients would inevitably hamper mutation detection owing to the high background number of normal cells. Consistent with this expectation, a significantly higher mutation detection efficiency has been noted in familial cases (87%) as compared to sporadic cases (51%).50 However, this difference was not overly apparent in the study of Messiaen et al51 who detected NF1 gene mutations in 29/29 familial cases and in 36/39 (92%) sporadic NF1 patients. The significance of the observed difference in mutation detection between familial and sporadic cases therefore remains unclear. A possible ascertainment bias concerning sporadic NF1 patients should also be considered since those individuals with very mild manifestations of the disease, and who lack affected children, may not come to clinical attention and will therefore tend to be under-represented in otherwise comprehensive mutation screening analyses such as that performed by Messiaen et al.51 This situation is rather different from that found in FAP where a high rate of mosaicism has been reported in sporadic patients. In stark contrast to those sporadic NF1 patients who present with mild manifestations, sporadic FAP patients with only a few polyps are often referred for germline APC and MUTYH mutation analysis in order to potentiate a differential diagnosis.

In two of the 36 sporadic patients with identified NF1 mutations studied by Messiaen et al,51 unequal ratios were noted between the normal and mutated alleles, suggestive of somatic mosaicism. However, the systematic and comprehensive screening of sporadic patients, including those NF1 patients with very mild manifestations of disease, would be required to provide a reliable estimate of the frequency of mosaicism in NF1 and to investigate its contribution to the variable expressivity of the disease. Such a study should also include the analysis of skin biopsies/neurofibroma tissue of sporadic NF1 patients in whom no NF1 gene mutation could be identified in blood lymphocytes. This strategy has already been shown to be very useful for mutation detection as demonstrated in NF119 23 110 and NF2.88 91 The absence of Lisch nodules in sporadic NF1 patients may also prove to be a good proxy indicator of mosaicism111 since the incidence of Lisch nodules in familial NF1 cases is much higher than in sporadic cases.112

It has been suggested that mosaicism is likely to be rare in the parents of NF1 patients with sporadic disease because the mosaicism status should be readily identifiable from clinical symptoms.113 However, the parents of patients with apparent de novo NF1 have not yet been systematically studied with respect to the presence of those minor signs of the disease that would on their own be insufficient to meet the full clinical diagnostic criteria. NF1 with very mild generalised manifestations, or segmental NF1 presenting only with localised pigmentary changes, may remain clinically unrecognised. Several reports have documented the inheritance of an NF1 gene mutation by a parent with mosaicism, some individuals being ascertained through their children with classical generalised NF1.34 68 70 Currently, however, no estimate can be given of the frequency of mosaicism among parents of apparently de novo NF1 patients.

CONCLUSION AND PERSPECTIVES

Several studies have addressed the molecular basis of the various recognised NF1 mosaic phenotypes and together these have served to improve our knowledge of mosaicism in NF1 very considerably. Although only a few cases have so far been investigated in any detail, mutation analysis in segmental NF1 has begun to yield important information about the different types of NF1 mutation and the mutational mechanisms that act somatically. The results so far obtained clearly indicate the need to investigate specific cell types such as Schwann cells and melanocytes. However, comprehensive studies have not so far been performed in patients with mosaic generalised NF1. Thus, it is currently unclear if and how mosaicism for large somatic NF1 gene deletions differs from mosaicism for intragenic NF1 mutations with respect to either the clinical phenotype or the tissue distribution and frequency of cells bearing the NF1 mutations. It also remains to be established whether the high proportion of blood cells harbouring an NF1 gene deletion, identified in type 2 deletion patients with mosaic generalised NF1, is the consequence of clonal selection or is instead a reflection of the early origin of mutational events during embryonic development. Clearly, further detailed mutation analysis of a variety of different cell types/tissues from patients with segmental and mosaic generalised NF1 will be required for us to understand the development of specific NF1 associated features, the tissue distribution of normal versus mutated cells, and probably also the transmission risks.

The frequency of mosaicism among patients with sporadic NF1 and very mild manifestations of the disease should also be considered. Any systematic screen for mosaicism either in sporadic cases or in segmental NF1 is, however, likely to be hampered by the fact that standard mutation detection methods are unlikely to be capable of identifying low level mosaicism (below 20% of cells). Improved mutation detection techniques should help to overcome this limitation. Thus, Hes et al93 achieved detection of 10–15% mutation bearing cells by optimising direct sequencing methods and were successful in detecting mutated APC alleles at a level as low as 2–5% of cells using denaturing gradient gel electrophoresis (DGGE) and quantitative pyrosequencing. In similar vein, denaturing high performance liquid chromatography (DHPLC) has been shown to be an important tool in the detection of low level mosaicism114 115 and could prove very useful in the investigation of mosaicism in sporadic NF1 patients. In patients who fulfil the NF1 diagnostic criteria but in whom no NF1 gene mutation has been identified in peripheral blood lymphocytes, mutation analysis of neurofibromas and/or CALM derived cells will be needed to detect mosaicism as demonstrated by Maertens et al23 for patients with segmental NF1. Another method that might improve the mutation detection rate in mosaic NF1 is the real-time quantitative allele discrimination assay by 3′ locked primers.116 This approach may help to identify low level mosaicism in apparently unaffected parents of patients with identified NF1 germline mutation. After characterisation of the mutation in the affected child, low level mosaicism in parents could in principle be detected by allele/mutation specific PCR, which has a very low detection limit (down to 1.4% of cells). A combination of improved mutation detection methodologies and improvements in the systematic design of studies, so as to include key cell types, should soon allow us to add new pieces into the jigsaw of our still patchy knowledge of mosaicism in NF1.

REFERENCES

Footnotes

  • Competing interests: None.