Background
About 20 to 30% of extra-osseous sarcomas are gastrointestinal stromal tumors [
1] which are the most frequent mesenchymal tumors of the digestive tract. Before 1998, their diagnosis was difficult and they were frequently mistaken for muscular or nervous tumors.
KIT[
2] and
PDGFRA[
3] are main driver genes of GISTs. Indeed, gain of function mutations of these genes are present in 85% of GISTs [
4,
5] and KIT inhibitors increase survival of patients with metastatic or localized GISTs [
6,
7]. The KIT inhibitor imatinib mesylate is more effective in patients with mutations in exon 11 of
KIT than in those with
KIT exon 9 mutations [
8]. The
KIT gene maps on chromosome 4q12 and encodes the protein KIT, a tyrosine kinase receptor, the activation of which leads to cell proliferation, differentiation, migration or increased survival [
9].
KIT mutations are early events in GIST oncogenesis [
10], and patients with germinal mutation of
KIT have a high incidence of GISTs [
11]. Cytogenetic studies of GISTs with karyotyping, fluorescence
in situ hybridization (FISH) and/or comparative genomic hybridization (CGH) arrays have shown that many of these tumors have gains or losses in chromosomes 5 and/or 8, and losses in 1p, 14q, 15q and 22q [
12‐
21].
The most frequent mutations of
KIT map in exon 11. In most GISTs, both mutated and wild-type transcripts are present [
22]. However, 5 to 15% of GISTs have loss of the
KIT WT allele from genomic tumor DNA [
23]. Expression of mutant KIT in the presence of WT KIT
in vitro has particular cellular effects, different to those associated with the expression of the mutant alone [
24,
25]. Patients with loss of WT
KIT have a worse prognosis than those with GISTs containing both WT and mutant alleles of
KIT[
23,
26].
We used high density whole-genome single-nucleotide polymorphism (SNP) arrays to study the mechanism of the WT allele loss in GISTs with KIT exon 11 mutations and analyze chromosome abnormalities in GISTs.
Discussion
We performed a whole genome analysis of 22 GISTs with SNP and CGH arrays, and detected numerous different chromosome alterations in most patients. Most of these alterations affected whole chromosomes or chromosome arms. These alterations consisted of gains, losses and cnLOH. We also identified the most frequently altered overlapping regions that may contain genes involved in GIST tumorigenesis. We also describe the presence of bi-allelic deletions and rare amplifications.
Chromosome alterations in GIST have previously been studied by classic cytogenetics [
12,
13,
16,
27], CGH arrays [
28] and SNP arrays [
29]. These studies demonstrated that gains and losses of whole C/CA are frequent events in GISTs. Gains frequently involve chromosomes 5 and 8, whereas most losses affect 1p, 14, 15 or 22 [
12,
13,
15‐
21]. Our work with high density SNP arrays, CGH arrays and FISH on 22 GISTs with
KIT exon 11 mutations confirms these data. Losses were detected in 14/22 (63%) tumors. The most frequently involved chromosomes/chromosome arms were 14 (n = 50%), 22 (n = 41%), 1p (n = 36%) and 15 (n = 27%), which is in accordance with previously published data [
18]. Some of these chromosomal alterations have been linked to tumor location [
28] and some studies suggest that they may have prognostic significance [
16,
18‐
20]. The tumors we studied could be classified into two groups with numerous or few chromosomes variations. The ‘numerous chromosome variations’ phenotype has been described as chromosome instability, and may be related to mutations in a driver gene [
30]. It has been reported to be associated with a poor prognosis, but our analysis failed to confirm this possibility.
An important finding of our study is the detection of frequent cnLOH in GISTs. The phenomenon of cnLOH was first described as uniparental disomy (UPD) [
31], and is responsible for parental imprinting in some inherited conditions, such as Prader-Willi syndrome [
32]. The identification of somatic cnLOH (or “acquired UPD”) in tumors was more recent [
33]. High-density SNP arrays are now available and can be used for analysis of both SNP-based genotype and DNA copy number allowing the detection of cnLOH, which was undetectable by cytogenetic methods. Such cnLOH is a frequent chromosomal alteration in human hematological malignancies, such as leukemia [
34], mantle cell lymphomas [
35] and follicular lymphoma [
36]. For cases of hematological malignancy, samples can be enriched in tumor cells by flow cytometry, making this type of analysis suitable for these diseases. Loss of a WT allele or cnLOH were previously reported in GISTs with mutations in exon 9 of
KIT[
29,
37,
38], and cnLOH was described in a few series of breast [
39], endometrial [
40], and colorectal carcinomas [
41]. However, the frequency of cnLOH in solid tumors may have been underestimated because the percentages of tumor cells in the samples analyzed were low. By contrast, the contamination of GISTs with non-tumor cells is generally low, and in the present study, more than 95% of all samples were tumor cells. Additional studies using algorithms adapted to low tumor percentage [
42] are needed to determine whether cnLOH is also frequent in other human solid tumors.
Two main types of cnLOH have been observed: whole C/CA and segmental cnLOH (telomeric or interstitial). Various mechanisms can lead to cnLOH and are probably different for these two types [
43]. Whole C/CA cnLOH may result from the loss of a chromosome/chromosome arm followed by chromosomal duplication of the paired chromosome/chromosome arm, or from a non-disjunction event during mitosis followed by loss of the supernumerary chromosome/chromosome arm. Segmental cnLOH can be secondary to a homologous somatic recombination: this seems to be the major mechanism responsible for cnLOH in hematological malignancies [
44] but still not well understood; it may be a cellular attempt to correct a deletion or repair double-strand breaks in the DNA. We show here that in GISTs, most cnLOH were whole C/CA type. Furthermore, all biGISTs with cnLOH also had losses of chromosome arms. The cnLOH in GISTs may be due to a loss of chromosome material during mitosis followed by duplication of the paired material. This hypothesis is controversial, and other studies have described cnLOH in GISTs as being mostly segmental, suggesting mitotic recombination as the major mechanism of cnLOH [
45]. Different inclusion criteria of tumors samples as well as different normalization and data-mining methods used in this study and in previously published data may in part cause these disparate observations.
The pathogenic consequences of cnLOH in cancer are multiple; it may give rise to homozygosity for a mutated tumor suppressor gene promoting tumor growth or chemotherapy resistance. Homozygosity may also change the methylation equilibrium of regulatory regions and result in epigenetic modulation of oncogenic pathways. Recently, cnLOH was observed in approximately 40% of patients with relapsed acute myeloid leukemia [
44] suggesting that it may be a mechanism for cancer progression in many cases. It has been reported that cnLOH is associated with
JAK2 or
FLT3 internal tandem duplication with oncogenic mutations in acute myeloid leukemias [
44,
46] and may also contribute to inactivating tumor suppressor genes in colorectal cancer, for example
hMLH1[
47]. Homozygosity of
JAK2 V617F and of
FLT3 mutations is associated with a poorer prognosis of patients with leukemia or myeloproliferative disorders [
46,
48].
One of our aims was to determine the causes of the
KIT mutation “homozygosity” observed in some GISTs
. We detected cnLOH in chromosome 4q in 9/11 GISTs with
KIT WT allele loss (81% of cases), whereas only one case showed LOH associated with 4q monosomy; the other case had a gain of 4q with loss of the
KIT WT allele
. Thus,
KIT mutation homozygosity is due to cnLOH in most GISTs. GISTs with homozygous mutations of
KIT have been reported to have a poor prognosis in a small number of independent series [
23,
26], although, in our series, there was no correlation with survival. However, our series was designed for molecular analyses: we specifically enriched the population with GISTs with a loss of WT
KIT, and those with mutated
KIT/WT
KIT were selected mainly to match the type of
KIT exon 11 and are thus not representative of GISTs in general. Large unbiased series are required to correlate cnLOH, gain of 4q and/or “homozygous”
KIT mutations with clinical characteristics and survival. The absence of matched blood samples may be responsible for the false positive detection of cnLOH that coincide with inherited regions of homozygosity. However, the risk of false positives is mainly associated with small segments of partial cnLOH; the risk is therefore relatively small in our series of GISTs as almost all the cnLOH detected were large.
In addition to chromosome instability, we detected numerous genomic alterations including bi-allelic deletions and amplifications. By using quantitative real-time PCR for sequences in the 9p21 region, we confirmed, with precision, the sites of the breakpoints detected with SNP arrays. We reported significant complete deletions of 48 regions involving numerous different genes; this thus implicates some of these genes in carcinogenesis, particularly those with functions in the cycle cell, DNA repair or apoptosis. Some of these genes with bi-allelic deletions map in common overlapping regions frequently altered in our series (
CDKN2A,
CDKN2B,
TUSC1), providing further arguments in favor of them having a role in GIST oncogenesis. Deletion of
CDKN2A in GISTs has already been reported, and is associated with a poor prognosis [
13]. Functional studies are necessary to determine whether the abnormalities of some of these genes may be responsible for, rather a consequence of, the substantial chromosomal instability of GISTs.
Materials and methods
Patients
KIT-positive GISTs carring mutations in exon 11 of KIT were selected from the Ambroise Paré hospital tissue bank database. All patients with loss of the WT KIT allele were included (25 samples) if there was sufficient frozen or paraffin-embedded material. Eleven tumors with heterozygous KIT exon 11 mutations were used as controls.
All GIST patients included in this genotyping analysis had previously participated in the MolecGIST study [
5]. Participants in MolecGIST provided their verbal informed consent to participate in this study after reading an information note. MolecGIST was approved by the appropriate French ethics committee: “Comité pour la Protection des Personnes se prêtant à des Recherches Biomédicales” (CPPRB, Committee for the Protection of Persons suitable for Biomedical Research) Saint Germain en Laye #06029, April 24th 2006.
All relevant information about the patients and tumor samples is given in Table
3. The median age of the 22 patients was 57 years (range [39 to 85]), and there were 12 men and 10 women. GISTs were localized in stomach (n = 10), duodenum (n = 4), small intestine (n = 6), rectum (n = 1) and mesentery (n = 1). Most samples analyzed were obtained from primary tumors, except for the samples from patients #32 s2 and #2P2 (metastases), and #1C2, #18C3 (intra-abdominal relapse). One sample was obtained after treatment with imatinib (#18C3). The types of exon 11 mutations are described in Table
3. According to the Fletcher classification estimating malignancy potential, 13 tumors were at high risk, two at intermediate risk, and three at low risk; for four samples the relevant information was not available. The mitotic count was higher than 10/50 HPF for 12 samples, between 5 and 10/50 HPF for one sample and less than 5/50 HPF for eight samples; the mitotic count was not known for one sample.
Genomic DNA was extracted from either frozen or paraffin-embedded fragments of GIST as previously described [
22]. Histological control with hematoxylin & eosin staining was performed on each sample before and after cutting the slides for DNA extraction. Tissue samples were macrodissected, and at least 90% of the cells in the samples used for DNA extraction were tumor cells. DNA samples were analyzed with a spectrophotometer (ND-100, Nanodrop®) and by electrophoresis, and only samples with a molecular weight higher than 2,5Kb were selected for SNP arrays.
Identification of KIT mutation
The method for identification of
KIT mutation has been described previously [
49]. Relative amounts of WT and mutated alleles of
KIT in patients with deletions or insertions were determined by analysis of fluorescent PCR products, and loss of WT
KIT was defined as a [mutated/wild type] ratio higher than 1.5 [
23]. Patients with a single nucleotide substitution mutation were considered as homozygous, if the WT nucleotide peak was at least three times lower than that the mutant peak in both forward and reverse Sanger sequences.
Whole genome analysis
GIST DNA samples were hybridized on Human370CNV-QUAD SNP arrays (Infinium Ilumina®) according to the manufacturer’s instructions by IntegraGen (Evry, France). This array contains over 370000 probes distributed throughout the genome with a median coverage of one probe every 5000 bases. However, the array does not include chromosome 13p, 14p, 15p, 21p or 22p markers. Chromosomes Y and X were only used to verify the sex of the patient. All genome positions were based upon NCBI36/hg18 from UCSC Genome Bioinformatics.
The fluorescence intensity data extracted using Illumina’s BeadScan software was analyzed twice by two different methods. In the first approach, the intensity data were normalized as described by Illumina Inc [
50] with IntegraGen commercial platform assistance. SNP genotyping data was plotted in Illumina Genome Viewer and Chromosome Browser of Illumina’s BeadStudio3.0 (Illumina Inc., San Diego, CA, USA) that chromosome aberrations were visualized and identified in respect to their localization. Regions with B allele Freq values and LOH Score suggestive of LOH without the modification in Log R values were considered as cnLOH segments. However, normalization and data-mining of Illumina platform was originally designed for genotyping of normal genomes. In the second analysis, the genotyping data were normalized and analyzed with tQN8 and GAP methods [
51,
52] by the biostatistics platform of la Ligue contre le Cancer. tQN8 normalization strategy was showed to improve the quality of Illumina arrays data of cancer genomes when used for LOH and copy number variations studies [
51]. GAP method was also developed for complex cancer genomes to analyze segmental copy numbers, genotype status and overall genomic ploidy of tumors [
52]. Additionally, CNA and AD were visually investigated on BeadStudio software (version3) by two independent researchers. Chromosome aberrations containing 30 SNP or more and/or longer than 100 kb were considered as true. Chromosome abnormalities were compared between tumors to detect the most frequently altered common overlapping regions (present in >20% of tumors). For tumors with ploidy higher than two, gains and losses were defined according to their ploidy level in this analysis. For tumors with chromosome numbers near 46, 69 and 92, the presence of more than four, five and six copies, respectively, were considered to be amplifications.
IntegraChipTM (CIT-CGH Homo sapiens BAC) arrays were hybridized with tumor DNA from 10 patients by the IntegraGene commercial platform according to the manufacturer’s instructions.
The results for 22 of the 36 samples hybridized on SNP arrays were of sufficient quality for analysis. Only 1 of the 13 arrays obtained with DNA extracted from paraffin-embedded tissue was of acceptable quality, whereas 21/23 samples from frozen tissue were satisfactory. The median call rate of the 22 samples was 98% (range [0.93 to 0.998]).
Quantitative Multiplex PCR of Short Fluorescent Fragments (QMPSF) analysis
QMPSF is a sensitive method involving the simultaneous amplification of short genomic fragments using dye-labeled primers under quantitative conditions, and was used for the detection of 9p21 genomic deletions. The assay employed ten primer pairs that cover a 2.8 Mb region and five genes (Telomere > MIR31/MTAP/CDKN2A/CDKN2B/DMRTA1 > Centromere) located in the 9p21 locus and generates PCR fragments of 150 to 250 base pairs. PCR conditions for QMPSF analyses of
CDKN2A and
CDKN2B loci have been described previously [
53]. Briefly, 100 ng of genomic DNA was used in a final volume of 25 uL with 0.16 mmol/L of each deoxynucleoside triphosphate, 1.5 mmol/L MgCl
2, 1 unit of thermoprime Plus DNA polymerase (ABgene), 5% dimethyl sulfoxide and 0.5 to 1.6 mmol/L of each primer, one primer of each pair carrying a 6-carboxyfluorescein label. After an initial denaturation for 3 min at 94°C, 20 cycles were performed consisting of denaturation at 94°C for 15 s, annealing at 90°C for 15 s (ramping 3°C/s) and extension at 70°C for 15 s (ramping 3°C/s), followed by a final extension step for 5 min at 70°C. PCR products were analyzed on a sequencing platform used in the fragment analysis mode in which both peak heights and areas are proportional to the quantity of template present for each target sequence.
FISH
Imprints of GISTs cells were obtained with briefly defrosted tumor samples. For each sample, one slide was stained with Giemsa for cytological control. Two centromere probes, one for each chromosomes 4 and 18 (CEN4, CEN18) (Kreatech Biotechnology) were used and six BAC probes – RP11-93 L18, RP11-983 F2 (chr17), RP11-959 K5, RP11-642 F17 (chr22), RP11-586A2 (chr4) and RP-11-121G9 (chr7) – were produced: bacteria carrying a BAC vector were grown overnight on solid agar medium, and then cultured overnight in LB medium. BAC DNA was extracted using NucleoBond PC 500 or NucleoBand Xtra BAC Kits (Macherey-Nagel) as recommended by producer. Aliquots of 1 μg of DNA were labeled by nick-translation according to the kit manufacturer’s instructions (Nick Translation Kit, Abott). The labeled DNA (the probe) was precipitated by incubation overnight at −20°C in the presence of human cot DNA, sodium acetate and ethanol, and then resuspended in hybridization buffer (LSI/WCP Hybridization Buffer, Abott). The probes were used at a final concentration of 40–50 ng/μL. Commercial probes were applied according to the manufacturer’s recommendations (Kreatech Biotechnology). Co-denaturation of the probes and the tumor section were performed to create single-stranded DNA. Fluorescence signals were analyzed using a Leica DM4000B microscope equipped with appropriate filters and a 22 DFC300FX camera under the control of LAS V4.0 software (Leica). The fluorescence signals in at least 10 nuclei were counted.
Statistical analysis
Survival curves were estimated using the Kaplan-Meier method, and differences between curves were assessed using the log-rank test. The threshold for significance was set at a p-value of 5%. The analysis was conducted using R software (2.14.1). Progression-free survival (PFS) is defined as the time interval between surgery and first evidence of disease progression or relapse. Overall survival (OS) is defined as the time interval between GIST diagnosis and death or last news. T-tests were used as appropriate.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.
The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
Conception and design: NL, ZHR, JFE. Development of the methodology, Acquisition of data: NL, ZHR, JBB, SBA, FJ, JTN, FP, EM, FC, JFE; Analysis and interpretation of data: NL, ZHR, JBB, SBA, FJ, JTN, FP, EM, AB, FC, JFE; Writing and revision of the manuscript: NL, ZHR, JFE; Study supervision: JEF; All authors have read and approved the final manuscript.