|Home | About | Journals | Submit | Contact Us | Français|
Neurofibromatosis type 1 (NF1) is a neurocutaneous disorder resulting in the growth of a variety of tumours, and is inherited in an autosomal dominant pattern. Gastrointestinal stromal tumours (GISTs) are mesenchymal tumours that commonly harbour oncogenic mutations in KIT or PDGFRA and are thought to arise from the interstitial cells of Cajal (ICC; the pacemaker cells of the gut).
To characterise two patients with NF1 and GISTs.
Two patients were genotyped for germline mutations in NF1. GISTs from both patients were genotyped for somatic mutations in KIT and PDGFRA. Loss of heterozygosity (LOH) of NF1 in one GIST was assessed by genotyping seven microsatellite markers spanning 2.39 Mb of the NF1 locus in the tumour and in genomic DNA. The known germline mutation in NF1 was confirmed in GIST DNA by sequencing. The copy number of the mutated NF1 allele was determined by multiplex ligand‐dependent probe amplification.
GISTs from both patients were of wild type for mutations in KIT and PDGFRA. In the GIST with adequate DNA, all seven markers were informative and showed LOH at the NF1 locus; sequencing of NF1 from that GIST showed no wild‐type sequence, suggesting that it was lost in the tumour. Multiplex ligand‐dependent probe amplification analysis showed that two copies of all NF1 exons were present.
This is the first evidence of mitotic recombination resulting in a reduction to homozygosity of a germline NF1 mutation in an NF1‐associated GIST. We hypothesise that the LOH of NF1 and lack of KIT and PDGFRA mutations are evidence of an alternative pathogenesis in NF1‐associated GISTs.
Neurofibromatosis type 1 (NF1) is a common neurocutaneous disorder secondary to mutations in the tumour suppressor gene NF1, and is inherited in an autosomal dominant pattern.1 A variety of benign and malignant gastrointestinal tumours have been reported in NF1, including gastrointestinal stromal tumours (GISTs).2,3 The GIST is the most common mesenchymal tumour of the gastrointestinal tract in the general population but is still rare,4 with an estimated incidence of 10–20 cases/million.5 The incidence of GIST in NF1 is unknown. Until 1998, a diagnosis of GIST subsumed a heterogeneous collection of tumours (in both the general and NF1 population), including true GISTs, peripheral nerve sheath tumours, true smooth muscle tumours and others.3,6
In 1998, Hirota et al7 reported that KIT expression in sporadic GISTs was accompanied by mutations in the proto‐oncogene receptor tyrosine kinase KIT. This hallmark expression of KIT (CD117, a transmembrane receptor for the growth factor stem cell factor), combined with molecular testing of mutations in KIT8 and PDGFRA,9 has greatly clarified the diagnosis and nosology of GISTs.10 It is hypothesised that GISTs arise from either the interstitial cells of Cajal or a precursor (henceforth, both possibilities abbreviated “ICC”), which are autonomic nerve‐related gastrointestinal pacemaker cells that regulate gastrointestinal motility.6 Although most sporadic GISTs are benign (low risk), malignant (high risk) GISTs are typically unresponsive to conventional cancer chemotherapy. The tyrosine kinase inhibitor imatinib mesylate (Gleevec, Novartis, East Hanover, New Jersey, USA) has produced dramatic clinical results in patients with widely metastatic chemotherapy‐resistant sporadic GISTs.11,12
Since 1998, true GISTs (tumours with positive staining for KIT) have been reported in NF113; most lack KIT or PDGFRA mutations. In this paper, we report two cases of KIT and PDGFRA mutation‐negative GISTs in patients with NF1. In one patient, we document the first evidence of loss of heterozygosity (LOH) of NF1 by mitotic recombination. The LOH of NF1 in GISTs in NF1 has recently been reported.14 The LOH by mitotic recombination is common in neurofibromas15; it has not been reported in other tissues in NF1. We consider the implications of NF1 haploinsufficiency and the subsequent LOH in the ICC and propose an alternative pathogenesis for GISTs in NF1.
Patient 1 presented at 34 years of age with a history of progressive left leg pain and weakness unresponsive to conservative therapy. NF1 had been diagnosed when he was a child. No other family members had a history of NF1 or GISTs. He denied any gastrointestinal symptoms.
In 1995, at age 27 years, he presented with acute peritonitis. At surgery he was found to have a ruptured cyst in the mid‐jejunum. The pathology report diagnosed a spindle cell schwannoma of the mid‐gut.
Magnetic resonance imaging of his leg showed a large, heterogeneous solid mass in the presacral space, originating from the left S1 nerve root. At surgery, a 20×15×11 cm mass was resected from the pelvis and identified as a high‐grade malignant peripheral nerve sheath tumour (MPNST). Frequent mitotic figures (>15/hpf) and extensive tumour necrosis were observed. Immunohistochemical staining with adequate controls showed focally positive S‐100 and CD34 staining, but negative KIT and smooth muscle actin (SMA) staining. Multiple, small adjacent masses were also noted studding the peritoneum (all <3 cm). They were not direct extensions of the major pelvic tumour. They exhibited a typical immunohistostaining pattern for GISTs, with strong, diffuse positive staining for KIT, focally positive for CD34 (seen in 70% of GISTs)10 and negative for S‐100. Few mitotic figures were seen.
Re‐examination of the ruptured jejunal spindle cell cystic tumour that was resected in 1995 and diagnosed as a schwannoma showed abundant KIT immunostaining, suggesting that this tumour was, in fact, a GIST. Rupture of the tumour probably seeded the peritoneum with slow‐growing tumour cells, which resulted in the multiple peritoneal GISTs that were observed 7 years later. Non‐NF1‐associated GISTs can present as a cystic mass16 and can disseminate throughout the peritoneum in ascitic fluid.17
The patient's pain improved after resection of the MPNST. He received a course of direct beam radiation to the MPNST. In consultation with his oncologist, he started treatment with imatinib mesylate. Side effects included fatigue and weight gain, and the drug was held once because of leucopenia secondary to concurrent radiation therapy. He was treated with imatinib mesylate for approximately 26 months.
The patient developed recurrent pain and weakness in the left lower extremity, and repeat magnetic resonance imaging and subsequent laparatomy showed extensive MPNST recurrence in the pelvis. A small bladder wall mass was a benign fibrous nodule, with no evidence of positive CD34 or KIT staining. Despite debulking of the MPNST and additional chemotherapy, the patient died about 6 months after the recurrence of the symptoms.
Patient 2 is a 62‐year‐old woman with NF1 and a history of breast cancer. A mass at the ileocecal junction was found on routine screening colonoscopy at age 56 years and the patient was referred for a laparotomy. The pathology report diagnosed the mass as polypoid fat with reactive changes. Two subserosal nodules in the jejunum were also incidentally noted, and biopsy performed on one of these (<0.5 cm nodule) showed a GIST based on positive staining for KIT and CD34 and negative staining for SMA and S‐100. No mitotic activity or necrosis was seen.
At the time of the GIST diagnosis, the patient was receiving chemotherapy for breast cancer. No GIST‐specific chemotherapy was given. A positron emission tomography scan at age 60 years did not find evidence of metabolically active disease. The patient remains asymptomatic of gastrointestinal complaints attributable to GISTs.
All investigations were performed under protocols approved by the University of Pennsylvania School of Medicine and the National Human Genome Research Institute Institutional Review Board.
The tissue was fixed in 10% neutralised formalin and embedded in paraffin wax. Immunostaining for S‐100, CD34, SMA and KIT was performed on the paraffin‐wax‐embedded tumour tissue using standard immunohistochemical protocol as follows. Sections of 5 μm thickness were cut and deparaffinised in xylene and rehydrated in graded alcohol. Endogenous peroxidase activity was blocked by 3% hydrogen peroxidase in methanol for 20 min. Antibodies to S‐100 (polyclonal, 1:4000, Dako, Carpinteria, California, USA), KIT (polyclonal, 1:75, Dako), CD34 (Myo10/8G12, 1:80, Becton‐Dickinson, Franklin Lakes, New Jersey, USA) and SMA (1A4, 1:500, Dako) were used. Primary antibodies were incubated for 1 h at room temperature. Immunohistochemical detection was performed on a DAKOCytomation Autostainer using the EnVision+HRP DAB System (DAKOCytomation). A positive control was used for each antibody. The primary antibody substituted by normal serum of the same species was used as a negative control.
The total coding region of the NF1 gene was analysed by reverse transcription‐polymerase chain reaction and in vitro transcription/translation as reported previously.18
Tumour DNA was extracted from two formalin‐fixed, paraffin‐wax‐embedded GISTs (one from each patient) according to the manufacturer's directions (PureGene, Gentra Systems, Minneapolis, Minnesota, USA). Using methods previously described,19,20 exons 9, 11, 13 and 17 of the KIT gene were amplified by polymerase chain reaction and screened for mutations using denaturing high‐performance liquid chromatography. Likewise, PDGFRA exons 12, 14 and 18 were analysed by the same approach.
Seven microsatellite markers spanning 2.39 Mb (table 11)) in, and flanking, NF1 were genotyped in genomic (blood) and GIST DNA from patient 1.21,22,23,24 (Analysis of the GIST from patient 2 was precluded by inadequate DNA.) Sequencing of exon 27a from the GIST DNA from patient 1 confirmed the presence of the mutated allele. To determine the number of copies of the mutated allele, multiplex ligand‐dependent probe amplification (MLPA) was performed, with appropriate controls, using the MLPA P081/082 kit, V.04 (MRC Holland, Amsterdam, The Netherlands).25
Patient 1: cDNA sequencing of fragment 3 (exons 19–29) showed (4537C→T) mutation in exon 27a, which predicts R1513X. Patient 2: a mutation in exon 30 of the NF1 gene was identified, 5747delC, which predicts L1920X. Both mutations were confirmed by sequencing in genomic DNA.
Analyses of exons 9, 11, 13 and 17 of the KIT gene and exons 12, 14 and 18 of the PDGFRA gene in tumour DNA from GISTs from both patients 1 and 2 did not show mutations in either gene.
All seven markers covering 2.39 Mb were informative in patient 1. A LOH was observed in the tumour for all markers analysed (table 11).). Sequencing of exon 27a on the DNA sample from the GIST from patient 1 showed the known mutation 4537C→T and no wild‐type sequence, suggesting that the wild‐type allele was lost in the tumour. MLPA analysis showed that two copies of all NF1 exons were present. With the exception of the control at chromosome 1p36, all positive controls showed appropriate signal strength. At chromosome 1p36, decreased signal strength suggested the loss of one copy.
Since 1998, the discovery of KIT expression in GISTs has improved tumour phenotyping and has strengthened the evidence of the association of GIST and NF1.13,26,27 There are no estimates on the incidence of GIST in NF1 using post‐1998 diagnostic criteria, although some studies have found cases of NF1 enriched in GIST populations.28,29 The two patients described here typify the GIST phenotype in NF1: both had multiple, low‐risk GISTs originating from the jejunum. Both tested tumours lacked KIT or PDGFRA mutations. Patient 1 originally presented with bowel perforation at age 27 years; patient 2 was diagnosed incidentally at 56 years. Table 22 summarises the differences between published reports of sporadic and NF1‐associated GISTs.10,13,26,27 These argue that NF1‐associated GISTs have a distinct pathogenesis.
This argument is predicated on: (1) a general lack of typical KIT/PDGFRA mutations in NF1‐associated GISTs13,26,27; (2) the evidence of NF1 LOH observed in this report and others14; and (3) the presence of ICC hyperplasia in the small bowel of individuals with NF1.13,26,27,28,30 Each of these is discussed.
Mutations in KIT are present in approximately 78% of sporadic GISTs, primarily in exon 11 (65%) and exon 9 (10%). An additional 8% of GISTs have mutations in PDGFRA, most commonly in exon 18.10 Only seven NF1‐associated GISTs have been reported with a mutation in KIT or PDGFRA. Takazawa et al26 reported four missense mutations not previously documented among thousands of sporadic GISTs examined to date, but the importance of these mutations is unknown. Cheng et al29 reported two patients with NF1‐associated GIST who had KIT exon 11 mutations that overlap with those observed in familial GIST kindreds. Given that individuals with a germline KIT exon 11 mutation may have skin hyperpigmentation, clinical confusion with NF1 is not inconceivable. Similarly, Yantiss et al30 reported an identical KIT exon 11 mutation observed in three separate GISTs from one patient. However, sequencing non‐tumour DNA for a germline mutation in KIT was not performed. In summary, NF1‐associated GISTs lack typical KIT and PDGFRA mutations.
This report is the first documentation (outside a neurofibroma) of mitotic recombination of NF1 leading to a LOH and subsequent tumour formation. NF1 LOH secondary to somatic NF1 mutations has been recently reported in NF1‐associated GISTs.14 Multiple mechanisms can lead to LOH, including deletion, non‐disjunctional chromosome loss with or without reduplication, gene conversion, point mutation, epigenetic inactivation and mitotic recombination.31 Gene conversion and chromosomal loss with reduplication are theoretically possible in the GIST from patient 1 in this report. However, mitotic recombination as a mechanism of LOH has been previously observed in NF1‐associated neurofibromas.15
Lastly, hyperplasia of the ICC in the myenteric plexus of individuals with NF1 and GIST is observed both in the vicinity of a GIST and in the macroscopically normal bowel.13,26,27,28,30 In familial GIST (kindreds harbouring germline KIT or PDGFRA mutations), this diffuse proliferation of the ICC throughout the bowel is common.5,32,33,34 The proliferation is polyclonal, suggesting that it is a hyperplasia rather than a neoplasia,35 and is regarded as a precursor lesion to GISTs. Presumably, additional KIT activations or cytogenetic changes prompt the neoplastic emergence of a GIST from the background hyperplasia.19 Precursor lesions in sporadic GISTs have not been well characterised,33 if they exist at all.
In NF1, ICC hyperplasia is the probable precursor lesion for GISTs.27,36 The ICC are of mesodermal and neural crest origin.37 Murine Nf1‐haploinsufficient melanocytes and mast cells (both of neural crest origin) have increased proliferation in response to Steel factor, the murine ligand for KIT.38 Thus, ICC hyperplasia in NF1 is plausibly secondary to NF1 haploinsufficiency.
We hypothesise that if ICC hyperplasia is secondary to NF1 haploinsufficiency, then subsequent LOH of NF1 leads to emergence of a GIST. The LOH of NF1 is probably a necessary, but not sufficient, prerequisite for tumour development. In our patient 1, the reduced signal strength of the MLPA chromosome 1p36 positive control may be a cytogenetic change that aided the emergence of the GIST. Loss of chromosome 1p (and other cytogenetic changes) have been observed in other NF1‐associated GISTs.14
Our hypothesis of alternate pathogenesis is compatible with the predominant small bowel distribution (especially in the jejunum) of GISTs in patients with NF1. Sporadic GISTs typically arise in the stomach, secondary to a somatic mutation in KIT or PDGFRA. The higher rate of sporadic gastric carcinoma (v small intestine cancers) in the general population suggests that the stomach is a relatively larger source of mutagens (eg, Helicobacter pylori, nitrite exposure, smoked or salted foods).39 Although gastric GISTs are observed in patients with NF1, precursor lesions (in the form of ICC hyperplasia) in the small intestine correspondingly increase the risk of tumour development. The notion of ICC hyperplasia as a precursor lesion is also compatible with the multiple and synchronous GISTs observed in patients with NF1, and may explain the younger age of onset of these tumours.
As ICC hyperplasia is implicated in dysmotility in cases of familial GISTs,40 it may plausibly underlie the gastrointestinal dysmotility commonly seen in NF1.3,41,42 Phenotyping with KIT and CD34 of the myenteric plexus hyperplasia3,41,42 and intestinal neuronal dysplasia43,44 observed in individuals with severe constipation and NF1 should clarify the role of the ICC in this important clinical problem.
GIST - gastrointestinal stromal tumour
ICC - interstitial cells of Cajal
LOH - loss of heterozygosity
MLPA - multiplex ligand‐dependent probe amplification
MPNST - malignant peripheral nerve sheath tumour
NF1 - neurofibromatosis type 1
SMA - smooth muscle actin
Funding: This research was supported in part by the Intramural Research Program of the National Human Genome Research Institute, National Institutes of Health.
Competing interests: None.
The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products or organisations imply endorsement by the US Government.