Novel Clinically Relevant Genes in Gastrointestinal Stromal Tumors Identified by Exome Sequencing

Gastrointestinal stromal tumors (GIST) are the most common mesenchymal tumors of the gastrointestinal tract (1). They are thought to arise from Cajal cells or their precursors and characteristically harbor gain of function mutations in KIT leading to constitutive activation of the KIT receptor (2).

An alternative mutation in a related tyrosine kinase, PDGFRA, is found in 35% of KIT mutation negative GIST (3, 4). Imatinib mesylate is a molecularly targeted drug known to inhibit both KIT and PDGFRA receptors and used in the treatment of recurrent and metastatic GIST or GIST with high risk of progression (5). Effectiveness of imatinib mesylate depends on the mutational status of KIT and PDGFRA (3). In contrast to metastatic GIST, radical surgery seems to be the best treatment option for localized tumors (6). The recurrence rate after radical surgery seems to depend mainly on tumor localization, size, and mitotic activity and ranges between 5% and 75% with a poor clinical outcome in relapsed patients (7, 8). Current classifications take into account tumor size, mitotic rate, and tumor location but do not include mutational data nor protein expressions of tumor cells (9–11).

The discovery of KIT and PDGFRA activating mutations in the majority of GIST represents a significant progress in understanding their biological behavior. Although further molecular mechanisms underlying the development and progression of GIST are not fully understood, they are necessary due to the wide range of clinical behavior. Chromosomal gains and losses resulting in altered gene dosage are known to be recurrent in GIST and believed to have a role in the molecular pathogenesis of these tumors (12–19). Nevertheless, the target genes remain to be identified within these regions.

The aim of our study was the identification of putative prognostic markers within the regions with recurrent gains or losses. We measured copy number variations (CNVs) and defined exact chromosomal breakpoints for the most common alterations in GIST. CNVs were screened in a large single-center cohort of GIST and correlated with clinical outcome. Exome sequencing identified mutated candidate genes in the regions of interest, and protein products of identified target genes were investigated immunohistochemically in our large single-center cohort of GIST.

We studied a total of 174 cases of GIST treated at the Medical University of Vienna between August 1992 and February 2011 in this retrospective observational study. Clinical data and follow-up were available for 145 of 174 cases. All cases were restaged according to UICC TNM classification of malignant tumors 7th edition and risk was evaluated according to Fletcher and Miettinen (9–11). As this cohort of patients has been used in several previous studies, clinical data in correlation with known risk factors have been reported previously, as well as the results of the sequence analysis about mutations of KIT (exons 9, 11, 13, and 17) and PDGFRA (exons 12 and 18; refs. 20–22). Tissue microarrays were established for immunohistochemical screening in all 145 cases. FISH was conducted on tissue microarrays of paraffin-embedded tissue in 125 of 145 cases. Apart from the 145 cases with clinical follow-up data and paraffin-embedded tissue available, further 29 GIST cases were identified with fresh-frozen tumor material suitable for microarray analysis and next generation sequencing. Tumor purity was at least 90% in samples used in all 29 cases. Institutional review board approval was obtained.

High-quality genomic DNA was obtained from 29 fresh-frozen tumor samples using the DNeasy Blood & Tissue Kit (Qiagen) and subjected to microarray analysis using the commercially available Affymetrix Genome-Wide Human SNP Array 6.0 (Affymetrix Inc.) following the protocols provided by the manufacturer. Data analysis was conducted with Affymetrix Genotyping Console 3.0.1 using the Birdseed Algorithm and Affymetrix Chromosome Analysis Suite 1.01 at a resolution of 500 kb. Genome annotations applied in data analysis referred to the human reference assembly GRCh37/hg19 as provided by the Affymetrix annotation file release na31. The reference model file used for data normalization with GTC was generated from 39 healthy control individuals. CNVs showing an overlap greater than 80% with benign CNVs of the Database of Genomic Variants (http://projects.tcag.ca/variation/) were excluded from our analysis.

Somatic CNV status of matching tumor-blood/normal tissue samples was used to confirm gains and losses using TaqMan real-time PCR or FISH.

Tissue microarrays from 125 GIST paraffin-embedded samples were used for FISH. Screening for CNVs was conducted using commercially available probes for TP73 (chromosomal band 1p36), ABL2 (1q25), CCND1 (11q13), DLEU (13q14), IGH (14q32), SNRPN (15q11), and CLTCL1 (22q11.2; Abbott and Kreatech). FISH procedures and analyses were conducted according to standard protocols.

Tissue microarrays containing 145 GIST samples were used for evaluation of protein expression. Supplementary Table S2 summarizes the antibodies used.

Immunohistochemical analysis was conducted using a Benchmark Ultra Immunostainer (Ventana), except for expression of RB, where a DAKO autostainer (DAKO) was used. A specimen was considered to be positive if the vast majority of cells (>80%) showed distinct staining. The number of cases differed between antibodies due to the use of tissue microarrays, and for investigation of FLT4 and AP1B1 the blocks had to be recut, resulting in the loss of several spots.

Thirteen of the 29 GIST of the microarray study were subjected to high-throughput sequencing. In all these cases, matched control DNA obtained from the peripheral blood or normal gastric tissue of patients were sequenced in parallel. Exome enrichment was conducted using the TruSeq sample prep (Illumina). Sequencing was conducted on a HiSeq 2000 (Illumina). All datasets were aligned with BWA v 0.6.1-r104 against the human g1k v37 genome using standard parameters. We conducted INDEL realignment with GATK v1.4, removed PCR duplicates using Picard v1.56 and recalibrated per-base quality values using GATK. Variants were then called and filtered using 2 independent pipelines based on GATK's UnifiedGenotyper (here SNPs and INDELs were filtered separately) and SAMTOOLS, respectively. The variant sets were then merged and we selected all variants from the matched normal-tumor pairs that were either unique in the tumor sample or changed their predicted genotype from being heterozygous in the normal sample to being homozygous in the cancer sample. We annotated these candidate variants using SnpEff 2.0.4 RC1 and predicted the possible effect of a variant (e.g., silent/nonsilent/nonsense mutation) using the NCBI Reference Sequence (RefSeq) gene annotations. We further determined whether a variant was already contained in the dbSNP TSI (Toscani in Italia) dataset. Candidate variants were further annotated using polyphen2 and finally manually filtered and inspected using IGV. Genes were compared with PubMed, Gene database, and the CancerGenes database for gene selection and priorization. Variants in candidate genes were confirmed by Sanger sequencing in all tested cases.

Categorical data were described with absolute and relative frequencies. Corresponding 95% confidence intervals (95% CI) were calculated according to the method of Wilson. Group differences were tested by χ² test or by Fisher exact test in the case of sparse data. Continuous data are described with median, minimum, and maximum. Differences between 2 groups for continuous and ordinal data were tested by Mann–Whitney U test.

Disease-free survival (DFS) was defined from the day of surgery until first evidence of progression of disease if complete surgical resection was possible. Patients showing advanced disease at time of initial diagnosis were excluded from analysis of DFS. Overall survival (OS) was defined as the time from primary surgery to patients' death. Survival data were graphically presented by Kaplan–Meier graphs and/or described by survival at predefined time points. Group differences are tested by the log-rank test. Cox regression models were used to estimate the effect of the expression of proteins adjusted for patients' age (<60 vs. ≥60) and risk according to Fletcher's score. Group differences are quantified with HR and corresponding 95% CIs. Proportional hazards assumptions were graphically checked by log-minus-log plots in stratified Cox regression models for every variable in the Cox model separately.

A two-tailed P-value of ≤ 0.05 was considered as significant. SPSS 20.0 (IBM) was used for all calculations. For exploratory purposes, all P values together were also adjusted to result in a false discovery rate (FDR) of 0.05 as described by Benjamini and Hochberg (23). Significant results after FDR-adjustment are marked by “^a”.

Table 1 summarizes the clinical data of the 145 patients with available clinical and follow up data. In brief, 94 GIST (64.8%) were of gastric localization (64.8%), 92 (63.4%) showed a KIT mutation, and 21 (14.5%) a PDGFRA mutation. All patients underwent initial surgical treatment; 25 patients received adjuvant imatinib mesylate. Median observation time was 37 ± 3 (SE) months. Although 7 patients showed advanced stage disease already at time of diagnosis, and no complete surgical removal of the GIST was possible, 24 patients experienced recurrent disease, and 12 died.

The microarray technology was conducted to define recurrent chromosomal regions and to determine exact breakpoints in 29 GIST tumors. Although the microarray platform used allows for the detection of CNVs down to 50 kb or smaller, we aimed to assess the minimal region of overlap of recurrent genomic regions that are known to be much larger. Therefore, we used a resolution of 500 kb and found 114 CNVs (43 gains and 71 losses) in the 29 tumor samples altogether (median number of CNVs/sample). We compared the CNVs between the 29 cases and defined rearrangements as recurrent when observed in at least 3 cases. Eighty-five of the 114 CNVs (75%) were located within recurrent rearrangements and included losses of chromosomal arms 1p (n = 13), 3p (n = 4), 13q (n = 5), 14q (n = 17), 15q (n = 7), and 22q (n = 11) as well as gains of chromosomal arms 1q (n = 3), 4q (n = 6), 5q (n = 8), 7p (n = 3), 11q (n = 4), and 12p (n = 3; Table 2).

CNVs were confirmed in all samples by FISH and TaqMan assays (1p: Hs05732825_cn; 3p: Hs01499513_cn; 5q: Hs02944177_cn; 11q: Hs06329804_cn; 13q:Hs07026395_cn; 14q:Hs02834878_cn; 15q: Hs05354966_cn; 22q: Hs01356996_cn).

A detailed list of CNVs is provided in Table 2. In conclusion, we found that the majority of chromosomal imbalances were recurrent, indicating that candidate genes relevant for GIST pathogenesis are located in these regions.

In a next step, we investigated if recurrent CNVs are of prognostic relevance. Seven of the 12 regions (1p, 1q, 11q, 13q, 14q, 15q, and 22q) were investigated on tissue microarrays containing 125 GIST tumors. About the remaining 5 regions, commercially available probes did not provide reliable signals on the tissue microarrays.

Results about frequencies of losses and gains are summarized in Supplementary Table S1 and were comparable to DNA microarrays. CNV status of investigated regions was correlated with clinical risk factor, tumor localization, DFS, OS, and KIT and PDGFRA mutations status.

FISH for chromosome 1p was successful in 119 samples. Loss of 1p (Fig. 1) was seen in 11 samples (9.2%; 95% CI, 5.2–15.8%), in 5 additional cases, a chromosomal imbalance between 1p and 1q was observed (5× 1p/1q+). When investigating the 16 cases (13.4%; 95% CI, 8.5–20.7%) with relative imbalance of 1p against all other cases, no association with clinical risk factor was seen either, but these cases showed a significantly shorter DFS (5 years DFS rate: 40.4% ± 17% SE vs. 84.6% ± 4.4% SE, P = 0.012, log-rank test) but not OS (Fig. 2). A significant association of relative loss of 1p with localization of the GIST was found (P = 0.001^a, χ² test). So of 75 gastric GIST, only 4 (5.3%; 95% CI, 5.2–33.4%) showed relative loss of 1p, but 9 of 28 (32.1%; 95% CI, 17.9–50.7%) of small intestinal GIST.

FISH for 15q was successful in only 73 samples, and a significant association of 15q deletion with tumor localization was seen (P < 0.001^a, Fisher exact test), because only 2 of 46 gastric (4.4%; 95% CI, 1.2–14.5%) GIST showed 15q deletion, but 8 of 17 (47.1%; 95% CI, 26.2–69%) small intestinal GIST.

FISH for 22q was successful in only 75 GIST, and 22q deletion correlated with gastric localization [P = 0.01, χ² test; 7 of 48 (14.6%: 95% CI, 7.3–27.2%) gastric GIST, vs. 8 of 18 (44.4%: 95% CI, 24.6–66.3%) small intestinal GIST]. Deletion of 22q correlated with the presence of KIT mutations. Although in GIST without KIT mutation (n = 28), only 3 cases (10.7%; 95% CI, 3.7–27.2%) showed a 22q deletion, there were 15 of 47 (32%; 95% CI, 20.4–46.2%) GIST with KIT mutation (P = 0.038, χ² test).

In a next step, we compared the recurrent CNVs that were observed in at least 3 cases (i.e., >10% of total cases) and defined minimal regions of overlap shared by all aberrant cases. The results are listed in Table 2. The size of the minimal regions of interest and consecutively the number of genes were too large to identify relevant candidates for the majority. Interestingly, the most frequent gain comprised a small region on chromosome 4 that included KIT and PDGFRA. Furthermore, RB1, which was recently reported to be a strong predictor of clinical outcome in GIST, was located in the deleted small region of overlap on chromosome 13. We concluded that our strategy identified successfully the 2 most prominent genes in GIST, that is KIT and PDGFRA, within our regions of interest. Nevertheless, the candidate remained elusive for the other regions.

Exome sequencing was conducted to identify putative novel candidate genes within the defined regions of interest as outlined in Table 2. We compared tumors with the matched constitutional genomes of 13 patients. The median overall number of variants/exome was 50,625 (range 38,365–58,258) per exome (Supplementary Table S3), the median number of variants (SNPs, insertions, deletions) present in tumor minus paired blood sample was 3,404 (range 1,641–13,602; Supplementary Table S4). Intergenic and intronic variants were excluded and annotation by snpEff reduced the average number from 3,404 to 1,764 (range 1,012–7,300). These variants included the switch of heterozygous mutations to homozygous mutations. Variants were considered only when they were stop mutations, splice site mutations, frame shift mutations, or missense mutations with potential pathogenic at in silico prediction. In a next step we looked for genes with recurrent variants in at least 3 patients. Recurrent variants were found in 37 genes and listed in Table 3. Sanger sequencing confirmed the variants in all tested cases. Furthermore, we compared the 37 genes with the Cancer Genes Database, the Gene Database, and GIST publications. Results are listed in Table 3. This strategy reduced the number of candidate genes to 3 or less per region. As expected KIT, PDGFRA, and RB1 were among these cases. In addition, we searched for potential interactions between the 37 genes and found positive matches between DIAPH1 and AP1B as well as MAPK8IP2 and SYNE2.

Eleven candidate genes were selected for immunohistochemical evaluation (Table 4 and Supplementary Table S2). Table 4 shows the number of GIST samples investigated, the primary staining pattern, and the number of positive cases for each antibody. Figure 1 gives samples of immunostaining. Gene mutation status has been shown to be linked to protein expression in a variety of human genes, but because protein overexpression might indicate loss of function (e.g., P53; ref. 24) as well as gain of function (KIT, ALK; refs. 25, 26), protein expression was scored as positive versus negative without any hypothesis about the type of altered gene function.

Expression of RBM5, RB, UBPY/USP8, GTSE1, MAPK8IP2, APIB1, and FLT4 was not associated with risk according to Fletcher, Miettinnen, or UICC staging and grading, DFS or OS or metastases at time of diagnosis (Table 4). RAD54L2 expression was associated with higher risk according to Fletcher (median high vs. intermediate risk; P = 0.033; Mann–Whitney U test) and higher UICC staging (median pT3 vs. pT2; P = 0.032; Mann–Whitney U test). Expression of RAD45L2 was associated with shorter OS (5 years OS rate: 80.8% ± 12.2% SE vs. 93.7% ± 3.2% SE, P = 0.042, log-rank test; Fig. 2), but not with shorter DFS. High SYNE2 expression was associated with high mitotic rate according to UICC (60% vs. 26.9%; P < 0.001^a, χ² test) and with higher risk according to Fletcher and Miettinnen (P = 0.044 and 0.031, respectively; Mann–Whitney U tests) and shorter DFS (5 years DFS rate: 43.1% ± 14.1% SE vs. 87.2% ± 4.2% SE, P = 0.001^a, log-rank test) but not OS (P = 0.06, log-rank test; Fig. 2). In addition, SYNE2 expression was less common in gastric (7 of 80) than in small intestinal GIST (9 of 22; P = 0.006^a, χ² test).

A correlation of SYNE2 and MAPK8IP2 protein expression was seen (P = 0.014; χ² test), but not between DIAPH1 and AP1B1 due to the overexpression of the latter in all samples.

GTSE1 expression was more common in small intestinal (7 of 22) than gastric (5 of 69) GIST (P = 0.013, χ² test).

Kit expression correlated with lower risk according to Fletcher (median intermediate vs. high risk, P = 0.035, Mann–Whitney U test), and lower UICC staging (median pT2 vs. pT3; P = 0.007, Mann–Whitney U test). Expression of Kit was associated with longer OS (5 years OS rate: 94.5% ± 2.4% SE vs. 67.2% ± 16.5% SE, P = 0.011, log-rank test; Fig. 2), but not DFS. The presence of a KIT mutation had no influence on DFS or OS (P > 0.05, log-rank test). A strong correlation of DIAPH1 expression with risk according to Fletcher (median intermediate vs. low risk, P < 0.001^a, Mann–Whitney U test), Miettinen (median moderate vs. very low, P < 0.001^a, Mann–Whitney U test), and UICC staging (median pT3 vs. pT2, P < 0.001^a, Mann–Whitney U test) was seen. In addition, strong DIAPH1 expression was associated with high mitotic rate according to UICC (45.1% vs. 15.6%, P = 0.001^a, χ² test) and small intestinal localization (17 of 30 vs. 26 of 74 in gastric GIST; P = 0.043, χ² test). Expression of DIAPH1 was associated with shorter DFS (5 years DFS rate: 57.3% ± 10.2% SE vs. 92.8% ± 4.1% SE months, P = 0.007, log-rank test; Fig. 2), but not with shorter OS (P > 0.05, log-rank test).

In a Cox regression model adjusting for patients' age and Fletchers score, SYNE2 expression was still associated with shorter DFS (P = 0.046; HR = 2.85; 95% CI, 1.02–7.97), and expression of DIAPH1 with longer OS (P = 0.030; HR = 0.16; 95% CI, 0.03–0.84).

When also including synchronous metastases, adjuvant administration of imatinib mesylate, KIT and PDGFRA mutations into the regression models, SYNE2 expression remained an independent prognostic factor for shorter DFS (P = 0.012; HR = 3.92; 95% CI, 1.35–11.42), whereas no relevance of KIT and PDGFRA mutations or adjuvant imatinib mesylate administration was observed.

A prognostic relevance of DIAPH1 expression on OS (P = 0.042; HR = 0.16; 95% CI, 0.03–0.95) was seen after introduction of KIT and PDGFRA mutations into the regression model, but this was not evident any more (P = 0.077) after inclusion of imatinib mesylate administration and synchronous metastases. KIT and PDGFR mutations showed again no prognostic relevance.

No significant additional effect of RAD54L2 or Kit expression was seen to explain DFS and OS additional to the effect of patients' age and Fletcher score (P > 0.05).

A combination of combined SYNE2 and DAPH1 expression showed a prognostic relevance for DFS in univariate analysis (5 years DFS rate: 25% ± 19.4% SE vs. 85.7% ± 4.4% SE, P = 0.001^a, log-rank test; Fig. 2) as well as after adjustment for patients' age and Fletchers score in a multivariate Cox regression (P = 0.025; HR = 3.58; 95% CI, 1.17–10.90). The prognostic relevance was also seen (P = 0.01; HR = 4.94; 95% CI, 1.17–16.55) when introducing adjuvant imatinib mesylate, synchronous metastases, KIT and PDGFRA mutations (in addition to age and Fletcher score) into the regression model.

The combination of SYNE2 and RAD54L2 expression, which was only evident in 2 patients, was associated with shorter OS in univariate analysis (P < 0.001^a, log-rank test), but did not reach significance in multivariate analysis of OS.

The combination of expression of target genes with high UICC mitotic rate delivered no additional prognostic information (P > 0.05, Cox regression).

No protein was associated with the presence of metastases already at time of surgery (P > 0.05, χ² test). Kit protein expression was found significantly more often in mutated tumors (P < 0.001^a, χ² test; 93.5% vs. 74.1%), whereas expression of all other proteins showed no association with KIT or PDGFRA mutation. A correlation of RB 1 expression with loss at 13q was seen (P = 0.017, Fisher exact test), whereas in GIST without nuclear RB1 expression, 24 of 66 (36.4%) showed a deletion on 13q, there were only 1 of 17 (5.9%) in patients positive for RB1 expression. In summary, expression of RAD54L2, SYNE2, KIT, and DIAPH1 correlated with risk factors and survival.

In rearward stepwise Cox regression models of DFS and OS including all relevant parameters, SYNE2 expression (P = 0–018; HR = 3.8; 95% CI, 1.26–11.48) remained an independent prognostic factor for DFS and DIAPH1 expression for OS (P = 0.042; HR = 0.59; 95% CI, 0.004–0.9) in the last step of the regression models, respectively (Supplementary Table S5).

In this study, we combined several techniques to identify putative prognostic factors within genomic GIST target regions. Microarray analysis was used to define minimum regions of interest for 12 recurrent gains or losses, FISH showed prognostic significance of 1p loss. Exome sequencing identified 37 recurrent potential pathogenic variations, and 11 candidates were chosen for immunhistochemical screening and correlation with clinical data. An association with survival was found for RAD54L2, SYNE2, KIT, and DIAPH1 expressions.

KIT is a well-known proto-oncogene in human tumors. Although KIT mutations are a common event in GIST, overexpression of KIT is observed considerably more often, showing no clear association with mutation status (27, 28).

Generally, overexpression of KIT serves as primary biomarker for induction of imitinib therapy (29), whereas the exact mutation status serves only for adjustment of it (30). The general presence of KIT mutations does not seem to be associated with worse prognosis in GIST, but deletion of exon 11 is accompanied with shorter survival in untreated patients (31). In contrast, other authors found no prognostic relevance of KIT mutations after curative surgical resection of GIST (32). Consensus exists that in patients treated with imatinib mesylate, KIT exon 9 mutations or lack of KIT mutations are associated with inferior response to imatinib (3, 33). Although many data on immunohistochemical expression of KIT as diagnostic and predictive marker for GIST do exist (34), surprisingly few data on the prognostic relevance of KIT expression are available: thus, it was not an independent prognostic factor in 95 GIST (32). In our study, KIT protein expression was found significantly more often in cases with KIT mutation. Lack of KIT expression was associated with shorter OS, but because 6 of 19 IHC-negative GIST showed a KIT mutation and 5 of 19 a PDGFRA mutation at sequencing, and 3 patients received imatinib, this subject deserves further investigation in a more homogenous collective.

Chromosomal gains and losses are consistent findings in GIST. Our data are consistent with previous reports with losses of 1p, 3p, 13q, 14q, 15q, and 22q and gains of chromosomes 4 and 5 (12–19). Only one previous study of 25 GIST tumors used the same platform and provided comparable high-resolution copy number analysis (17). Regions of interest were investigated for candidate genes by gene expression analyses in that study. This approach failed to identify PDGFRA and KIT on chromosome 4 and RB1 on chromosome 13 in contrast to our study using a combination with exome sequencing. About the prognostic relevance of CNVs in GIST, loss of 1p and 15q were reported to be more common in clinically more aggressive GIST (35). Loss of 1p has also been shown to be more common in intestinal GIST, and was associated with a more aggressive clinical course in that study as well (36). In contrast, the pathway characterized by loss of 14q has been suggested to be associated with gastric tumors with stable karyotypes and a more favorable clinical outcome (36). In our study, relative loss of 1p at FISH was significantly associated with GIST of the small bowel and with more aggressive clinical course, which is in good correlation to previous data obtained by comparative genomic hybridization (36). In contrast to the findings by Gunawan and colleagues, no association of 14q deletion with gastric localization or better prognosis was seen in our study (36). Furthermore, we failed to identify loss of RB1 expression as a prognostically relevant factor. This is in contrast to recent data, where RB1 was integrated in a genomic index for mitotic checkpoints that was reported as a strong predictor of clinical outcome in GIST (19).

Apart from established candidates for 4q gain (KIT and PDGFRA) and 13q loss (RB1), we were able to provide new candidates in further recurrent regions of interest, and to define target genes in previously known areas of CNVs. RAD54L2 on chromosomal arm 3p, DIAPH1 on 5q and SYNE2 on 14q may be of biological relevance because associations with staging and survival were observed. Although all 3 genes have not been reported in GIST so far, no data at all on RAD54L2 in human malignant disease exist, and there is only one study about SYNE2, where the presence of splice variants was reported in human non–small lung cancer (37). DIAPH1 is a downstream effector of RhoA, controls actin-dependent processes such as cytokinesis, SRF transcriptional activity, and cell motility (38, 39), and might therefore play a role in human cancer progression (40). To our knowledge, no DIAPH1 mutations in human malignant disease have been described previously.

Interestingly, SYNE2 encodes for a nuclear outer membrane protein and has been shown to play a role in MAP kinase signaling pathways (MAPK1 and MAPK2) in promyelocytic leukemia protein (41). SYNE2 also interacts with MAPK8IP2, which is one of our candidate genes on chromosomes 22 (42), and a correlation of expression of these 2 proteins was also seen in our study. Interestingly, we found recently that high MAPKAPK2 expression is also a strong prognostic predictor in the GIST cohort investigated in this study (21). Because DIAPH1 inhibition has also been shown recently to block ERK− phosphorylation (43), our findings indicate that altered MAP kinase signaling seems to be of crucial importance in GIST.

In summary, using high-throughput methods we identified several new candidate genes that may be of pathogenetic and prognostic significance in GIST. Our data indicate that the MAP kinase signaling pathway seems to play an important role in GIST. So the inclusion of MAP kinase pathway parameters in future clinical studies might be of potential benefit for the patients as selective inhibitors are available.

No potential conflicts of interest were disclosed.

Conception and design: S.F. Schoppmann, B. Streubel, P. Birner

Development of methodology: S.F. Schoppmann, P. Birner

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.F. Schoppmann, U. Vinatzer, S. Liebmann-Reindl, G. Jomrich, P. Birner

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): U. Vinatzer, N. Popitsch, M. Mittlböck, B. Streubel, P. Birner

Writing, review, and/or revision of the manuscript: S.F. Schoppmann, N. Popitsch, B. Streubel, P. Birner

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): G. Jomrich, B. Streubel

Study supervision: S.F. Schoppmann, P. Birner

The authors thank Amy Bruno-Lindner for language editing.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.