INTRODUCTION

Gastrointestinal stromal tumors (GISTs) represent the distinct mesenchymal tumors of the gastrointestinal tract with a wide morphological spectrum and various degrees of differentiation toward interstitial cells of the Cajal phenotype (1). Perhaps the most specific and relevant diagnostic criterion for GISTs is the immunohistochemically determined c-KIT proto-oncogene expression (2, 3); the latter encodes for a transmembrane tyrosine-kinase receptor (CD117) and has the stem cell factor as its ligand.

The identification of gastrointestinal autonomic nerve tumors (GANTs), originally described under the name plexosarcomas, was based on ultrastructural similarities with the myenteric plexus of the gastrointestinal tract (4, 5, 6, 7). Recent evidence indicates considerable morphologic, immunophenotypic, and molecular overlap between GANTs and GISTs, most significantly the common CD117 immunopositivity and the presence of c-KIT mutations (8); most likely tumors previously described as GANTs represent an ultrastructural subset of GISTs with peculiar neuroaxonal characteristics.

Although the data on the pathogenesis of GISTs rapidly have been growing over the last few years, the genetic changes leading to the initiation and progression of GISTs are still only partially known. Gain of function c-KIT mutations were found in extracellular, juxtamembrane, and kinase domains of the majority of GISTs, presumably contributing to the molecular initiation of these neoplasms (9, 10, 11, 12). Additionally, the karyotypic and fluorescence in situ hybridization (FISH) analysis, and the studies of DNA copy number changes by comparative genomic hybridization (CGH) and loss of heterozygosity, demonstrated losses in chromosome 14q, detected in ≤85% (13, 14, 15), and in 22q, detected in ≤77% of benign and malignant GISTs (13, 16, 17). The recurrent losses of chromosomal material strongly suggest other alterations, in addition to the c-KIT mutations, to be involved in tumorigenesis and/or in the progression of GISTs. These alterations may contribute to heterogeneous differentiation patterns and clinical behavior of GISTs. In contrast to conventional GISTs, however, the cytogenetic and molecular data on GANTs are limited (8, 15). The purpose of this study was to analyze the presence of c-KIT mutations and the pattern of deletions in chromosome 14 and 22 in cases of well-delineated GANTs to clarify to what extent they contribute to their pathogenesis.

MATERIALS AND METHODS

Five primary intra-abdominal neoplasms diagnosed as GANT were included in the study. Patient 1, a 25-year-old woman, presented with a duodenal tumor mass that measured 2 cm in diameter and showed low mitotic activity (1 mitosis/10 high-power fields [HPF]) (18). The tumor was diagnosed as “probably benign.” However, 8 years later, multiple peritoneal and liver metastases developed, and the patient died 10 years after the primary diagnosis. The clinical and histopathological data on the four other patients have been described elsewhere (15). They are summarized in Table 1.

TABLE 1 Clinical and Histological Features of GANTs from Five Patients and Outcome of the Disease

Histopathology

Histopathologic examination and immunohistochemical analyses were performed on tissue fixed in 10% buffered formalin and embedded in paraffin. Immunohistochemical studies were performed using polyclonal antibodies against CD117 (dilution 1/50; DAKO, Denmark) and S100 protein (dilution 1/300, DAKO) and monoclonal antibodies against CD34 (dilution 1/50, DAKO), α-smooth muscle actin (dilution 1/400, Sigma, Israel), and desmin (dilution 1/20, ICN-Cappel, Belgium) and subsequently detected with avidin-biotin peroxidase complex system and diaminobenzidine as the chromogen. Ultrastructural analysis was performed on 2.5% glutaraldehyde-fixed tumor fragments.

FISH Analysis

Dual-color FISH was performed on slides prepared from available cell suspension containing single cells obtained after overnight collagenase treatment of tumor specimens. The slides were fixed in a 3:1 mixture of methanol and acetic acid and stored at −20° C until they were analyzed. In the case of primary tumor from Patient 1, FISH was performed on nuclei extracted from paraffin-embedded tissue. Nuclei extraction was done as described elsewhere (19).

For the assessment of chromosome 14 loss, the yeast artificial chromosome (YAC) clones, isolated from the Center d’Etude du Polymorphisme Humain (CEPH, Paris, France) human genomic YAC libraries and corresponding to 14q12-q32, were used, as described elsewhere (15). Applied DNA probes were labeled with biotin-16-dUTP or digoxigenin-11-dUTP and used in pairs.

For the analysis of chromosome 22 loss, the following sets of probes were used: SpectrumGreen- and SpectrumOrange-labeled LSI 22q11.2/22q13.3 (DGCR/ARSA loci, Vysis Inc., Downers Grove, IL), SpectrumGreen-labeled LSI 22q11.2 (BCR locus, Vysis Inc.) together with digoxigenin-labeled cosmid G9 (corresponding to 22q12, proximal to EWSR1 region) or digoxigenin-labeled 96C10 (NF2/22q12 locus; kindly provided by Dr E. Zwarthoff, Rotterdam). Hybridization and detection were performed as previously described (20). Hybridization signals were analyzed by using an Axioplan 2 fluorescence microscope equipped with a cooled CCD camera and run by Cytogen system for capture and analysis (Imstar S.A., Paris, France). The number of hybridization signals for each specimen was assessed in 100 interphase nuclei. A specimen was interpreted as monosomic if only one signal for each respective probe was detected in >25% of the cells evaluated (more than three standard deviations above the average false-positive rate observed in control experiments on nuclei extracted from paraffin-embedded tissue).

CGH Analysis

Comparative genomic hybridization (CGH) was performed using DNA extracted from frozen tissue (all but one tumor) or from paraffin-embedded tissue (primary tumor from Patient 1), according to the method described by Kallioniemi and coworkers (21), with minor modification (22). The image analysis was performed using an epifluorescence microscope (Leica DMRB, Wetzlar, Germany) and image analysis software (QUIPS, Vysis, IL). The ratio of FITC/Lissamine fluorescence intensities obtained from 10 metaphase spreads per case were averaged, and the resulting profile was blotted next to the ideogram. Ratios above 1.2 and below 0.8 were considered to represent chromosomal gains and losses, respectively.

Molecular Studies

DNA for PCR amplification was obtained from fresh tissue or from paraffin-embedded material by standard methods. Exons 9, 11, 13 and 17 of the c-KIT gene was evaluated for the mutations by polymerase chain reaction amplification and direct sequencing of the amplification products, as described elsewhere (12, 23, 24).

RESULTS

Histological and Ultrastructural Appearance

Histologically, all tumors showed a spindle cell pattern (Fig. 1). All were strongly and diffusely immunoreactive with CD117 antibody, and all were negative for desmin and S100 protein. Three tumors were stained with CD34 antibody (from Patients 2, 3, and 5), revealing multifocal immunopositivity. In two tumors, focal α-smooth muscle actin expression was seen.

FIGURE 1
figure 1

Spindle cell pattern of colon GANT from Patient 2, which is virtually indistinguishable from classical GIST. Hematoxylin and eosin staining.

Ultrastructurally, the tumors consisted of nests and bundles of spindle-shaped cells, exhibiting elongated nuclei with peripheral and dispersed small clumps of heterochromatin. Some of these cells revealed long interdigitating cytoplasmic processes with short intermediate junctions. The presence of variable numbers of bulbous axonlike processes, containing many small empty vesicles and dense core granules of various diameters, was frequently observed. The intercellular spaces contained aberrant skeinoid fibers (extracellular collagen globules; Figs. 2 and 3).

FIGURE 2
figure 2

Electron micrograph of a group of spindle-shaped tumor cells with cellular processes. Note skeinoid fibers (arrow). Magnification, 4,600 ×.

FIGURE 3
figure 3

Part of the tumor cell showing bulbous axonlike processes with small empty vesicles (arrow) and a dense core granule (arrowhead). Magnification, 1,600 ×.

FISH and CGH Analysis

A summary of the cytogenetic, FISH, and CGH findings is provided in Table 2. By interphase FISH using chromosome 14–specific probes, the primary tumor of Patient 1 revealed normal, biallelic chromosome 14 complement in 91% of cells. However, in the recurrent specimen from this patient, the entire 14q11-q32 region was lost in 38% of nuclei. In this case, the loss of chromosome 14 in a subpopulation of recurrent malignant cells proves the secondary character of this genetic change, occurring most likely because of tumor progression. The results of cytogenetic analysis and dual-color FISH evaluation of 14q11-q32 losses in Cases 2–5 had been published before (15). As is summarized in Table 2, specimens from Patients 2 and 3 did not reveal chromosome 14 losses, whereas tumors from Patients 4 and 5 revealed deletions of 14q23-q24 and 14q11-q32 regions, in 55 and 93% of cells, respectively.

TABLE 2 Cytogenetic and Molecular Findings of Five GANTs Under Study

Dual-color FISH with the 22q region-specific probes revealed partial chromosome 22 losses in all five specimens. In Patient 1, 76 and 87% of nuclei showed 22q12–13 loss in the primary and recurrent tumor specimens, respectively, whereas the entire chromosome 22 loss was seen in 40% of cells in the recurrent sample. Again, the terminal partial 22q12–13 loss could be the primary genetic event, whereas the loss of the entire chromosome 22 sequences could be the progression-type genetic event in this tumor.

Tumor from Patient 5 showed total chromosome 22 loss, as the specimens from Patients 2 and 4 revealed partial 22q12-qter, and those from Patient 3, only 22q13.3 deletions. In particular, the probe specific for NF2/22q12 loci (cosmid 96C10) showed normal biallelic presence in the latter.

Fluorescence in situ hybridization results were further confirmed by CGH analysis (Table 2). One tumor (Case 2) exhibited several genetic imbalances, whereas three others showed only few changes. Nevertheless, the common findings in all specimens by CGH were partial losses of chromosome 22, with the smallest region of deletion encompassing the 22q13 region. Notably, it was the only change detectable by CGH analysis in tumor from Patient 3.

c-KIT Mutation Analysis

All tumors exhibited mutations of c-KIT. In Cases 2 and 5, the insertions of exon 9 were found, whereas other specimens showed mutations of exon 11 of the c-KIT gene (Table 2). In the samples of Patient 1, a single point mutation of codon 576 of exon 11 was present in both the primary and the secondary tumor, indicating that the later onset of the disease was due to the recurrence of the previous malignancy.

DISCUSSION

Electron microscopic examination is generally accepted as indispensable for the reliable diagnosis of GANTs. Ultrastructurally, the presence of dense core granules, cell processes, neurotubules, dystrophic axons, synapselike structures, and skeinoid fibers are characteristic for these tumors (2, 6, 7, 25). All our five cases fulfilled the ultrastructural criteria for GANT diagnosis. Simultaneously, all our cases carried c-KIT mutations that validated their classification within the general category of CD117-positive GISTs. The presence of c-KIT mutations in 5 of 10 examined GANTs was recently reported by Lee and coworkers (8), proving that in this aspect GANTs do not differ from classical GISTs. In the previous study, we demonstrated the loss of the part 14q in two of four primary GANTs (15). Another GANT analyzed in this study did not disclose chromosome 14 loss in the primary tumor, whereas the subpopulation of cells from the recurrent sample indicated partial 14q loss. This finding suggests that at least in this particular case, the chromosome 14 loss was a secondary event, most likely associated with tumor progression. On the other hand, either primary and recurrent tumors from this patient revealed loss of the terminal part of the long arm of chromosome 22. By CGH and FISH analysis, all other GANTs analyzed in our study also demonstrated the loss of genetic material from chromosome 22q, with the common overlapping area of loss at q13. Our results indicate that GISTs showing neurogenic differentiation are characterized by an accumulation of a number of genetic changes; it may explain their frequently observed more aggressive clinical behavior (6, 18). Furthermore, our data suggest that the loss of the 22q13 chromosomal region may be a common feature of these tumors. So far, few tumor suppressor genes have been localized in 22q. One of them is NF2, a gene inactivated in neurofibromatosis 2–associated and sporadic schwannomas, perineurial cell tumors, and meningiomas (26, 27, 28). Fukasawa and coworkers (17) identified NF2 gene mutation in two of 22 GISTs investigated. The authors concluded that NF2 contributes as a tumor suppressor in a small subset of GIST. By FISH analysis, NF2 loci were lost in four of five tumors we investigated. In the tumor from Patient 3, however, the NF2 locus was preserved, suggesting the gene or genes localized more telomeric to NF2 to be involved in the pathogenesis of GANTs. Alternatively, the loss of the 22q13 region could be related to the progression of our tumors. The percentage of aggressive cases is generally higher in intestinal than gastric GISTs (29), and all five tumors in the present study were intestinal in origin and clinically progressive. Apparently, additional molecular studies are needed to further define the critical region of 22q loss and to establish genes potentially responsible for the neurogenic differentiation of GANTs and/or their generally malignant clinical course.

In summary, we have used molecular cytogenetic methodologies and sequencing analysis to evaluate c-KIT mutations and chromosome 14q/22q losses in five ultrastructurally confirmed GANTs. c-KIT mutations and loss of 22q13-qter region were consistent in these tumors, suggesting that these alterations may be vital to the development and/or progression of GISTs with neuronal characteristics.

TABLE 3 Loss/Deletion of Chromosome 22q by FISH Analysis in Five GANT Cases