Article Text


GDNF as a candidate modifier in a type 1 neurofibromatosis (NF1) enteric phenotype

Statistics from

Editor—Neurofibromatosis type 1 (NF1) is a common human disorder (1/3500 live births) with neuroectodermal involvement primarily resulting in dermatological manifestations of café au lait spots, cutaneous/subcutaneous neurofibromas, and freckling of major skin folds.1 Owing to diagnostic uncertainties, especially in young patients, an international scoring system has been discussed and agreed upon.2 Half of the cases result from new mutations, while others show an autosomal dominant mode of inheritance. The encoded product, referred to as neurofibromin, is a member of the so called GTPase activating proteins (GAPs), and is an upstream downregulator of the RAS(p21)/RAF/MAPkinase3 and RAS/RAL4 signalling pathways. Although locus homogeneity is a hallmark of this condition, phenotypic heterogeneity has been exemplified by a wide spectrum of diversity ranging from malformation or malignant variants to virtually benign dermatological changes.1 In particular, and among the many causes of gastrointestinal involvement in NF1 patients, the association with intrinsic intestinal dysmotility, resulting from intestinal neuronal dysplasia type B (IND B)5 6 or aganglionic megacolon (Hirschsprung's disease, HSCR),7 has been documented and is now well established.

Interestingly, a substantial fraction of the phenotypic variability seen in NF1 patients might be governed by non-allelic, trait specific, “modifying” loci.8 Although the action of such modifying loci has been primarily shown in the number of café au lait spots or the number of cutaneous/subcutaneous neurofibromas, it can be speculated that such a genetic phenomenon might also be operative in other phenotypic traits, especially in individual or familial cases with an enteric phenotype.

The female proband from the family analysed here (fig 1A) had minor cutaneous manifestations of NF1, dysmorphic facial features (midface hypoplasia), congenital heart disease (ventricular septal defect, coarctation of the aorta), and congenital megacolon. She subsequently underwent a Duhamel abdominoperineal pull through and pathological examination of the whole colectomy specimen pointed to IND B (fig 2), because of findings of (1) abnormal submucosal plexuses showing focal hyperplasia (in terms of density and sizes), (2) occasional giant ganglia harbouring >10 neurones, and (3) nerve cell buds along afferent nerves.9 The older sister also had NF1 and congenital megacolon, while a brother was totally unaffected. NF1 was inherited from the mother and maternal grandmother, who both had a mainly cutaneous form of the condition. The father was healthy.10

Figure 1

(A) Family pedigree. A miscarriage, which occurred between II.2 and II.3, is not represented here. CAL=café au lait spots, NFs=neurofibromas. A cytogenetically balanced reciprocal t(15;16)(q26.3;q12.1) translocation is shared by I.1, II.1, II.2, and II.3,10 as indicated by asterisks. (B) NF1 and GDNF restriction. DNA from the proband and relatives was screened for the NF1 lesion with PCR primers and HphI enzyme cleavage as previously stated51 and for the GDNF mutation using the upper primer GDNFEx2F (5′-CAAATATGCCAGAGGATTATC-3′) and lower primer GDNFEx2R (5′-TATTTTGTCGTACGTTGTCTC-3′) with cycling conditions of 5′ at 94°C and 30 rounds of 20 seconds at 94°C, 20 seconds at 55°C, and 30 seconds at 72°C followed by restriction with HinfI endonuclease. Restriction products were size fractionated through a 8% polyacrylamide gel, stained with ethidium bromide, and visualised by ultraviolet transillumination. NF1 exon 16 and GDNF exon 2 amplimers are shown unrestricted and restricted for I.1 (lanes 1 and 2) and I.2 (lanes 3 and 4) and restricted only for II.1, II.2, and II.3 (lanes 5, 6, and 7). Arrowheads indicate theoretical fragment sizes in base pairs (bp). HphI restriction of the 552 bp wild type NF1 amplimer generates a 532 bp fragment, whereas restriction of the 559 bp mutated amplimer generates a 457 bp fragment. On the other hand, restriction with HinfI of the 326 bp wild type GDNF amplimer generates a 204 bp and a 122 bp fragment but leaves the (326 bp) mutated amplimer unaltered. Only II.1 and II.3, sharing both the paternally derived GDNF lesion and the maternally inherited NF1 mutation, have megacolon. Note the presence of heteroduplexes in NF1 heterozygotes (lanes 3-5 and 7).

Figure 2

Intestinal neuronal dysplasia. The colectomy specimen was fixed in Bouin's reagent for 12 hours. Transverse sections were cut every centimetre, from both normal and pathological segments, and were routinely processed for histology. Four millimetre serial paraffin embedded sections were stained with haematoxylin-eosin-saffron (HES) for standard light microscopy. Immunodetection of protein S100, a marker of Schwann cells, was used to outline nerve structures. After microwaving for antigen retrieval, paraffin embedded sections were incubated with Dako Z311 anti-S100 antibody (diluted 1/800) in a Ventana ES 320 automated immunostainer, using appropriate immunoperoxidase LSAB detection and revelation kits (Ventana, Tucson, AZ). Sections from normal and pathological specimens were analysed for the presence and density of intramural plexuses, the thickness of nerve trunks, and the features of ganglion cells. (A) Hyperplasia of nerve structures in the lamina propria and muscularis mucosae. Numerous nerve fibres are stained with anti-S100 antibody. Four submucosal plexuses (arrows) contain numerous immunonegative ganglion cells (immunoperoxidase staining). (B) Giant submucosal ganglion containing ⩾12 distinctive neurones, as shown by characteristic vesicular nuclei and abundant amphophilic cytoplasm (HES).

Although IND B may segregate as a monogenic disorder, no specific locus has been linked to this recognised Mendelian entity (MIM 601223). However, since ∼30% of IND B patients have accompanying aganglionosis, that is, HSCR,9 theRET proto-oncogene,11-13 the genes encoding endothelin receptor B (EDNRB),14-16 or its ligand endothelin 3 (EDN3)17 are candidate genes for isolated IND B18. Here, none of these genes was found to be mutated, suggesting that the NF1-IND B combination observed here was indeed an integral NF1 variant.10

However, the scarcity of this specific variant and the small family sizes has hindered the identification of modifying loci by phenotype and pedigree based analyses.8 The initial observation of an out of frame insertion within NF1 exon 16 (2424-2425insCCTTCAC, fig 1B) favoured a null lesion, that is, it did not support causal genotype-phenotype correlation. The fact that only those relatives with the NF1 mutation and a cytogenetically balanced reciprocal translocation t(15;16)(q26.3;q12.1) (fig 1A) had intrinsic intestinal dysmotility suggested that a modifier gene might have been altered at one of the translocation breakpoints.10 However, none of the breakpoint regions is known to harbour a gene involved in the development of the parasympathetic ganglion cells of the digestive tract and this family was investigated further for candidate genetic modifiers lying elsewhere in the genome.

As changes in GDNF 19-22 and the neurturin gene (NRTN)23suggested polygenic causation of anomalous neural crest cell derived structures, especially parasympathetic enteric neurones, these genes were good candidate modifiers. For this reason,GDNF and NRTNgene exons were analysed for DNA changes by SSCP and sequence analysis, as previously reported.19 23 WhileNRTN showed no sequence variation (data not shown), a missense mutation was found within GDNFexon 2, changing codon 93 from arginine (CGG) to tryptophan (TGG) (R93W, fig 1B). The R93W mutation could be regarded as potentially pathogenic, since (1) it resulted in a non-conservative amino acid change at an evolutionarily conserved residue in the direct vicinity of a putative propeptide cleavage site and transforming growth factor beta (TGFB) related cystein rich motifs,24 (2) it was found in other patients with distinct neural crest cell related anomalies such as HSCR (in addition to mutations ofRET)21 22 or the CCHS-HSCR association,19 and (3) it was conspicuously absent from a large number of control DNAs.21 25 Of interest, this is the first GDNF mutation seen in a patient with IND B without aganglionosis. The recurrence pattern of this mutation is probably explained by its relation to a CpG dinucleotide mutational hot spot.26

GDNF is probably the most powerful neurotrophin identified to date. It is the first identified member of a family of growth factors distantly related to TGFBs24 (see Bohn27 for a recent review and Ramer et al 28 for an update on neurotrophic effects), whose action is mediated by binding to a multicomponent system composed of RET receptor tyrosine kinase (RTK)29 and glycosyl-phosphatidylinositol (GPI) linked cell surface adapter proteins (GFRA1, GFRA2).30 31 Signalling through RET can trigger phosphatidylinositol-3 kinase (PI3K), leading to activation of either members of the RHO family of GTPases (RHO, RAC, and CDC42) with ensuing rearrangements of the actin cytoskeleton and axon outgrowth, or protein kinase B (PKB) mediated effects on metabolism or gene transcription. Interestingly, in both these pathways, PI3K requires functional RAS.32 In addition, stimulation of RET leads to SHC-GRB2-SOS complex formation and RAS activation, the RAF and RAL families of GTPases acting as downstream effectors33 (fig3). Although it is difficult to predict the exact outcome of the combination of mutations reported here for the subcellular signalling network, it can be speculated that certain pathways are markedly impaired while others are only mildly affected, especially since theNF1/GDNF double heterozygote infants depicted here have severe developmental alterations but only minor symptoms related to deregulated cell growth.10 Of special interest is that one might have expected functional recovery of the NF1mutation by the GDNF lesion, since NF1 disruption generates activated, GTP bound RAS, whereas low GDNF maintains RAS in the GDP bound form. A possible rationale comes from the observation that RAF1 is able to induce growth arrest and differentiation of discrete human carcinoma cell lines, with downregulation of endogenousRET 34 35 and resistance to activated exogenous RETalleles.36 The bona fide enteric phenotype observed here is indeed consistent with a feedback loop of RAS/RAF downstream effectors upon RET, or at least with a relatively complex interplay of factors influencing cell growth and differentiation. An interesting alternative hypothesis comes from the so called downstream model in which RAS is regarded as a regulator of neurofibromin, the process of RAS conversion from the GTP bound to the GDP bound form inducing neurofibromin to transmit signal through its influences on microtubule organisation.37 In that model, both low GDNF and low neurofibromin concur to disrupt the downstream effects of neurofibromin upon cell proliferation or differentiation.

Figure 3

NF1 (neurofibromin) and GDNF signalling partnership. Epistatic interaction of NF1 and GDNF is sustained by the signalling partnership of their respective products, neurofibromin and GDNF. Whereas alteration of GDNF is expected to balance the lack of inhibition of RAS owing to low neurofibromin, the effects on the PI3K dependent pathways are liable to aggravate the negative feedback loop of RAS/RAF on RET and RTK activity (adapted from van Weering and Bos33 and others).

Murine models were generated for both NF1and GDNF through disruption by homologous recombination in embryonic stem cells. Mice lackingGdnf(Gdnf -/-)38-40 had total renal agenesis, resulting from defective induction of the ureteric bud and absent enteric neurones, also consistent with additional manifestations of pyloric stenosis, duodenal dilatation, and congenital megacolon. These models are highly reminiscent ofRet deficient mice (Ret -/-), providing a functional confirmation that GDNF is a ligand of RET. Heterozygotes (Gdnf +/-) were indiscernible from wild type litter mates. However, heterozygousNf1 mutants (Nf1 +/-) do not replicate the human disorder (in particular, they do not develop obvious neurofibromas or pigmentation defects).41 42 Conversely, these mice are prone to age related tumours, in addition to malignancies reminiscent of human NF1 (especially phaeochromocytomas and myeloid leukaemia).42 Interestingly, homozygotes (Nf1 -/-) die in utero from severe cardiac malformation, especially involving the neural crest cell derived conotruncus,41 42 and show hyperplasia of the pre- and paravertebral sympathetic ganglia,41 indicating that the phenotype is both dosage sensitive, as inGDNF, and malformative rather than tumourous, eventually confirming the important role of neurofibromin during development. These models and the family presented here suggest murine Nf1 +/- xGdnf +/- intercrosses for the phenotypic analysis of double mutants as an ultimate demonstration of a modifier gene effect.43

Modifying genes are not just hypothetical and HSCR families have provided particularly fruitful material for eliciting such entities. Two hitherto anonymous loci have been indicted in polygenic inheritance of HSCR. The first such example was illustrated by a large inbred Mennonite HSCR pedigree that segregated a missense mutation inEDNRB 44 and otherwise showed linkage disequilibrium with marker alleles mapped to 21q22. This finding was highly suggestive of a HSCR genetic modifier linked to this chromosomal region, which might elsewhere account for the high HSCR prevalence among trisomy 21 patients. More recently, genome wide non-parametric linkage was performed on a panel of HSCR pedigrees which selected a particular subgroup in the sense that these were either unlinked to RET or showed positive linkage but with no sequence alteration identified at that locus. From these families, significant linkage to a locus in 9q31 was found, suggesting that this genetic region contains a gene whose variation entails a specific susceptibility to HSCR, with or without concomitant linkage toRET.45 More straightforward evidence for a HSCR modifier is exemplified by the genes encoding glial cell line derived neurotrophic factor (GDNF)20-22 and, more recently, neurturin (NRTN),23two highly homologous natural ligands of the RET tyrosine kinase receptor protein. Indeed, since GDNF andNRTN were found to be mutated in families also segregating well characterised RETalleles, it was postulated that alterations of these genes were not sufficient in themselves to cause HSCR, but that they probably contributed to the severity of the phenotype or to higher penetrance of the RET mutations.

Although molecular evidence for modification proper in the pathogenesis of human NF1 has not been provided to date, the possibility of epistatic interaction of the NF1(Nf1) gene with other discrete loci was recently illustrated both in man and in a murine model for human NF1.

Human pedigrees with hereditary non-polyposis colorectal cancer (HNPCC) have been reported in which children homozygous (doubly heterozygous) for an MLH1 mutation were shown to develop extracolonic malignancies of early onset and de novo NF1.46 47 These observations suggest that theNF1 gene is prone to common replication errors during mitosis and/or meiosis, and that MLH1 plays a particular role in monitoring these types of DNA lesions. These very important observations point to mismatch repair (MMR) genes as possible targets during NF1 tumour advancement and shed new light onto the rather unexpected microsatellite instability observed in NF1 derived tumours.48

In mice, different combinations of mutations ofNf1 and/or Trp53(homologous to TP53) were generated through intercrosses.49 50 The first remarkable observation was that, unlike their Nf1 +/- orNf1 -/- congeners, mice that were haploinsufficient for both Nf1 andTrp53 developed benign or malignant peripheral nerve sheath tumours (MPNSTs) and malignant Triton tumours reminiscent of human NF1. In addition, permanent cell lines established from some of these mice showed a pattern of gene expression consistent with immortalisation of pluripotent neural crest stem cells, whether the original tumours were of seemingly mesodermal or of conspicuous neural crest origin.50 These data provide further evidence of linkage of RAS with the cell cycle machinery, especially with the pathways that are linked with the activation of TP53, and confirm the commonality of neural crest involvement in NF1 pathogenesis. In addition, these intercrosses clearly support the hypothesis that functional protein-protein interaction is a strong substratum for epistasis or modification.

In the family presented here, especially in the two infants who are doubly heterozygous for theNF1/GDNF lesions, it is questionable whether GDNF modification accounts for wider involvement of neural crest cell derivatives, such as midface hypoplasia (through disruption of the cranial crest mesectoderm) or conotruncal heart disease (VSD and coarctation of the aorta), also observed in the proband. Whatever the case, our findings seemingly confirm the previous speculation that genes whose products interact functionally with RAS are potential NF1 modifiers.43 Once identified, these modifiers will provide tools to unravel interactions with the subcellular signalling network in NF1 patients and may lay the basis for new therapeutic approaches.

  • A family with neurofibromatosis type 1 (NF1, MIM 162200) and congenital megacolon (intestinal neuronal dysplasia, IND B) was investigated for possible genetic modifiers. A germline mutation in the NF1 gene, c.2424InsCCTTCAC, and a germline GDNF variant R93W were found in this family.

  • In this kindred, only members with both the paternally derived GDNF R93W and the maternally inherited NF1 mutation had megacolon.

  • Such epistatic interaction between NF1 and GDNF is in keeping with functional cross talk of the RET and RAS pathways in a complex subcellular signalling network.


The first two authors contributed equally to this work. We are grateful to Drs L Taine, P Vergnes, S Gallet, and J-F Chateil for their contribution to investigating this family, to I Laurendeau and M Olivi for technical assistance, and to Dr D Récan and coworkers for establishing and maintaining lymphoblastoid cell lines. This work was supported by the Association pour la Recherche contre le Cancer (ARC) and the French Ministère de l'Enseignement Supérieur, de l'Education Nationale et de la Recherche, and the Association Française contre les Myopathies (AFM).


View Abstract


  • * Present address: Service de Biochimie et Biologie Moléculaire, Hôpital d'Enfants Armand-Trousseau, Assistance Publique-Hôpitaux de Paris, 75571 Paris Cedex 12, France

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.