Prognostic Significance of c-kit Mutation in Localized Gastrointestinal Stromal Tumors

Gastrointestinal stromal tumors (GISTs) constitute the great majority of primary mesenchymal tumors of the gastrointestinal system. The diagnosis of GIST has been facilitated considerably by the widespread application of KIT (CD117) immunohistochemistry (1). A large proportion of tumors previously ascertained as gastrointestinal smooth muscle tumors are now proven to be GISTs. The treatment of choice for localized GISTs is complete surgical excision, but relapse is common among high-risk groups. The prediction of relapse in GISTs based on clinicopathological features, however, is often difficult. Although the most important and easily applicable criteria for predicting the tumor behavior are tumor size and mitotic activity (2, 3, 4), small portions of tumors apparently lacking mitotic activity have been found to metastasize. Several studies have identified additional prognostic factors, including immunohistochemical markers of cell proliferation (e.g., Ki-67, PCNA) and DNA flow cytometry (5, 6), but none of these has shown to be consistently relevant compared with the conventional morphological criteria.

The c-kit proto-oncogene encodes a type III receptor tyrosine kinase. Specific mutations in this gene leading to ligand-independent activation of the c-kit proto-oncogene have been detected in GISTs (7). These genomic mutations encode structurally abnormal proteins with a constitutive enzymatic activity (8). Most of the c-kit mutations in GISTs occur within the juxtamembrane region, encoded by exon 11. In addition, mutations in other regions of the c-kit gene, including exons 9, 13, and 17, have been detected, although they occur at a much lower frequency than those in exon 11 (8, 9, 10, 11, 12). Although it has been reported that GISTs with c-kit exon 11 mutations were typically of a higher grade or associated with frequent relapse and poorer outcome than those lacking these mutations (13, 14), the prognostic significance of c-kit mutations in GISTs is still controversial. In this regard, we analyzed the frequency and pattern of c-kit mutations in GISTs and the association of c-kit mutations and pathological parameters in the tumors. We also determined the prognostic significance of c-kit mutations in localized GISTs.

We retrospectively analyzed the clinical records of 215 patients who had been treated for a pathological diagnosis of GIST, leiomyoma, leiomyosarcoma, or sarcoma NOS in the gastrointestinal tract at the Asan Medical Center (Seoul, Korea) between 1990 and 2001. Of these 215 cases, paraffin blocks or clinical data were not available in 71. To identify true GIST cases, the results of light microscopic examination as well as immunohistochemical analyses using antibodies against CD117 (polyclonal, A4502; DAKO), CD34 (0786; Immunotech), desmin (monoclonal, M0760; DAKO), smooth muscle actin (monoclonal, M0851; DAKO), and S-100 (polyclonal, 00-3118; Zymed) were reviewed by two pathologists. Of the 107 cases definitively diagnosed as GISTs, 21 cases were excluded from analysis because they had a disseminated disease at diagnosis or did not receive curative surgery, leaving 86 cases for the study population. Mitotic figures were counted in 50 consecutive high power fields as recommended, and histomorphological types were recorded as spindle, epithelioid, and mixed cell type. From these results, the patients were divided into very low, low, intermediate, and high-risk groups (15). The study was approved by the Institutional Review Board.

From formalin-fixed, paraffin-embedded tissue samples, we cut 10-μm thick sections from each sample. Genomic DNA was extracted from one to three sections from each sample using a DEXPAT kit (TaKaRa, Kyoto, Japan).

PCR-SSCP analysis for exon 11 was performed in all cases. Each PCR reaction was performed in a volume of 20 μl, containing 50 ng of genomic DNA, 0.2 μM of each primer (Table 1), 0.19 μM of each deoxyribonucleotide, 50 μM KCl, 1.5 μM MgCl₂, and 10 μM Tris-HCl (pH 8.3). After amplification, the PCR products were denatured and electrophoresed on 10% polyacrylamide Tris-borate EDTA gels, and the DNA fragments were stained using a Silver Stain Plus kit (Bio-Rad, Hercules, CA). The PCR products were subcloned using a TOPO TA cloning kit (K4500-01; Invitrogen, Carlsbad, CA) and sequenced by the ABI 3700 DNA analyzer (Applied Biosystems, Foster City, CA). For each patient, five to eight clones were sequenced by using the Wizard Plus SV Miniprep DNA Purification System (A1460; Promega, Madison, WI). Patient samples negative for exon 11 mutations were subsequently PCR-amplified by using primers specific for exons 9, 13, and 17 (Table 1), followed by direct sequencing. Each patient sample was sequenced at least twice.

Statistical analyses were carried out by χ² or Fisher’s exact tests in cross tables, and the relationship between c-kit mutations and clinicopathological factors was assessed by Student t test. All statistical tests were two-sided. Survival curves were drawn according to the Kaplan-Meier method and were compared by the log-rank test. Multivariate analysis was based on Cox regression analyses. P < 0.05 was considered statistically significant.

Of the 86 patients with GISTs, 47 were males and 39 were females, with a median age of 59.5 years (range, 30–83 years). The stomach and small bowel were the most common sites of primary tumors. The median tumor size was 6 cm (range, 0.4–23 cm). Morphologically, 53 patients (61.6%) had the spindle cell type, 18 (20.9%) had the mixed cell type, and 15 (17.4%) had pure epithelioid tumors. The histological grade of the tumors was very low or low risk in 31 patients (36.0%), intermediate risk in 20 patients (23.3%), and high risk in 35 patients (40.7%).

Of the 86 GISTs, 61 (70.9%) had mutations in exon 11 of the c-kit gene. Three tumors (3.5%) had mutations in exon 9, but none had mutations in exons 13 or 17. Amino acid alterations in the 61 tumors with exon 11 mutations were deletion in 33 (54.1%) tumors, substitution in 20 (32.8%), both deletion and substitution in 4 (6.6%), insertion in 1 (1.6%), and duplication in 3 (4.9%; Table 2). Most of the exon 11 mutations were located between codons 550 and 570. There were two gastric and one intestinal GISTs carrying internal tandem duplications in the 3′-end of the c-kit juxtamembrane domain (ITDs in the 3′ c-kit-JM; Fig. 1; Table 3). The frequency of ITDs in the 3′ c-kit-JM was 4.0% in gastric GISTs and was 3.1% in intestinal GISTs. All 3 patients with ITDs in the 3′ c-kit-JM were female, had spindle cell histology, and were mitotically inactive. After 24–80 months of follow-up, they are all alive without relapse. All three exon 9 mutations were insertion of six nucleotides (1530ins6), resulting in duplications of amino acid residues Ala⁵⁰²-Tyr⁵⁰³. Two originated from the stomach and one from the small bowel. Relapses were documented in both 2 cases of the high-risk group (Table 4). The exon 11 mutations were analyzed by both PCR-SSCP and DNA sequencing. We found that DNA sequencing was more sensitive than PCR-SSCP in that the mutation rate was 71% by DNA sequencing and 53% by PCR-SSCP.

Histologically, tumors with c-kit mutations showed higher mitotic counts and higher cellularity. None of the other clinicopathological parameters showed a significant association with c-kit mutations (Table 5). The c-kit mutation was found in 40 of 53 cases with the spindle cell type (75.5%), in 15 of 18 cases with the mixed cell type (83.3%), and in 9 of 15 cases with the epithelioid cell type (60.0%). The relationship between c-kit mutations and histological subtypes was not significant either in overall GISTs (P = 0.305) or in gastric GISTs (P = 0.204).

For patients free of relapse, the median follow-up was 35.7 months. Of the 29 cases with relapses, the location of relapse in c-kit positive/negative GISTs was intra-abdominal in 13/2, the liver in 11/0, and other sites in 3/0 cases. The types of relapse did not show a significant association with c-kit mutations (P = 0.367). The 5-year relapse-free survival (RFS) rate for all patients was 29%, whereas the 5-year overall survival rate was 78%. The 5-year RFS rates for patients with tumors having a c-kit mutation was 21%, which was significantly lower than the 5-year RFS rate for patients with tumors without a c-kit mutation (60%, P = 0.098; Fig. 2). The 5-year RFS rates were also significant shorter for patients with tumors having high mitotic rates (P = 0.021; Fig. 3), larger tumor size (P = 0.015; Fig. 4), epithelioid or mixed morphology (P = 0.005), or nongastric GIST (P = 0.037). In contrast, we observed no relationship between RFS rates and the type of exon 11 mutation (missense versus all others). Multivariate analyses indicated that the presence of a c-kit mutation, mitotic counts, and tumor size was independent prognostic factors for RFS (Table 6).

We have shown here that in patients with localized GIST, the presence of a mutation in the c-kit proto-oncogene is associated with a higher rate of mitosis and adversely affects RFS. These results are in line with the previous findings, showing that tumors having c-kit mutations had higher mitotic rates and that the presence of a c-kit mutation was associated with more frequent relapses and decreased survival (13, 14). Taken together with these previous findings, our results suggest that GISTs with c-kit mutations represent malignant GISTs, which could be a potential explanation of its correlation with decreased RFS.

In contrast, other investigators (16) showed a trend toward poorer RFS in patients with GISTs with c-kit mutations than in those without c-kit mutation but without a statistical significance. This observation might be because of the small number of their series of tumors analyzed (n = 48). They also observed that patients having tumors with missense mutations in exon 11 had a favorable RFS rate than those with other types of mutations (16), a finding not observed in our study. In fact, we found that patients having tumors with deletions in exon 11 had a poorer RFS rate than those with other types of mutations.

We found three tumors (3.5%) with ITDs in the 3′ c-kit-JM. All 3 were female, had spindle cell histology, and were mitotically inactive. With follow-up for 24–80 months, they are all alive without relapse. These findings are in line with the previous findings (17). Recently, it has been reported that GISTs with the c-kit 1530ins6 mutation represent a subset of tumors occurring exclusively in the intestine and usually follow a malignant course (18). Two of three GISTs with the c-kit 1530ins6 mutation in the present study were gastric GISTs. We identified an additional five cases with the c-kit 1530ins6 mutation that were not included in the present series. All these five originated from the small bowel. In line with our observation, two studies from Japanese populations revealed four gastric GISTs of 11 GISTs with the c-kit 1530ins6 mutation (10, 19). Taken together, an ethnic difference between Western and Asian populations is suggested in the distribution of GISTs with the c-kit 1530ins6 mutation.

It has been reported that c-kit mutations are common in incidental GISTs of ≤1 cm in size, suggesting that c-kit mutations per se are of little prognostic value (20). In contrast, we found that, even when our analysis included eight incidental GISTs ≤ 2 cm in size, six of which had a c-kit mutation, the presence of a c-kit mutation was still associated with poor RFS.

Previous studies on the prognostic value of a c-kit mutation had several limitations, e.g., the mutation analyses did not include all of the exons where mutations had been identified (exons 9, 11, 13, and 17), and metastatic as well as localized tumors were included in the study samples (13, 14). In contrast to these earlier studies, we analyzed a relatively large number of cases (86 tumors) exclusively of localized GISTs and sequenced all our exons of c-kit where mutations have been documented.

Studies on the prognostic value of c-kit mutations in GISTs should be interpreted with caution. The frequency of exon 11 mutations in these tumors has been reported to vary widely from 21 to 71% (13, 16, 21, 22), which may be related to different methods of mutation detection and different populations and types of tissue analyzed. Indeed, we observed that the mutation frequency in exon 11 was 71% by DNA sequencing but was only 53% by PCR-SSCP, in agreement with previous findings (12), suggesting that the studies where PCR-SSCP was used to screen mutations might have underestimated the prevalence of c-kit mutations in GISTs.

Because of the relatively short follow-up period, we could not analyze prognostic factors in relation to overall survival. In addition, we did not additionally analyze mutations in the platelet-derived growth factor α gene in tumors negative for c-kit mutations. It has been reported that the platelet-derived growth factor α mutation frequency in 127 GISTs was 3.9% (23) and that mutations in platelet-derived growth factor α and c-kit are mutually exclusive (24). Additional studies are needed to reveal the natural history of GISTs positive for platelet-derived growth factor α mutations.

The availability of the tyrosine kinase inhibitor imatinib for the treatment of metastatic GISTs has shifted the focus of research in the treatment of localized GISTs from the prediction of malignancy to the proper selection of patients for adjuvant therapy (25). Metastatic GISTs with exon 11 c-kit mutations have been shown to better respond to imatinib than tumors with other types of mutations (23). Taken together with our results, this suggests that localized GISTs with c-kit mutations may be good candidates for adjuvant therapy with imatinib after curative resection.

In conclusion, we have shown that the presence of c-kit mutation, high mitotic count, and larger tumor size are independent factors for poor prognosis in patients with localized GISTs, with a lower RFS after curative resection. These factors can be incorporated in selecting high-risk patients as candidates for adjuvant therapy.