Polyclonal Evolution of Multiple Secondary KIT Mutations in Gastrointestinal Stromal Tumors under Treatment with Imatinib Mesylate

In the majority of gastrointestinal stromal tumors (GIST), the most common mesenchymal tumor of the gastrointestinal tract, pathogenesis is related to activating oncogenic mutations in the KIT (stem cell factor receptor) gene or the platelet-derived growth factor receptor α (PDGFRα) gene (1, 2). Both genes encode type III receptor tyrosine kinases which can be inhibited by imatinib mesylate (Glivec, Novartis, Basel, Switzerland). The response rate of advanced GISTs to imatinib mesylate reaches up to 70% (3, 4) but there is increasing evidence for resistance due to several different mechanisms. First, the location of the primary mutation may influence the receptor conformation leading to inhibition of drug binding. Second, resistance under treatment may be caused by gene amplification leading to overexpression of KIT receptor or PDGFRα. Third, still unidentified alternative receptor tyrosine kinases might be activated. Fourth, another mechanism of resistance is the acquisition of new activating mutations of the KIT or PDGFRα gene under treatment leading to functional resistance against imatinib, as shown by several groups (5–9). The same mechanism is known to be the most common reason of resistance to imatinib in patients with chronic myelogenic leukemia (10, 11).

In the present study, we aimed to explore the frequency and nature of secondary KIT mutations occurring under imatinib treatment and leading to secondary resistance, and mutational status was correlated with clinical and pathologic data. For molecular analysis, we used tumor tissue specimens obtained during surgery from 32 patients under imatinib treatment. In 18 patients, tissue was also available from the primary site before treatment. Patients had been scheduled for surgical resection of progressive tumor deposits or of residual tumor after successful imatinib therapy. In addition to the common hotspots for primary mutations (exons 9, 11, 13, and 17 of KIT; exons 12 and 18 of PDGFRα), tissues were examined for secondary mutations in exons 13, 14, and 17 of KIT encoding both tyrosine kinase domains. In the majority of cases, more than one tumor manifestation (two to seven) resected under treatment were evaluated to search for different types of secondary mutations in evolving tumor subclones.

Patients. Tumor samples from 32 patients [21 men and 11 women; median age, 56.5 years (range, 34-73 years)] were studied. They all suffered from histologically confirmed irresectable or metastatic GIST (liver and/or peritoneal metastases). The primary tumors were predominantly located in the small bowel (n = 18), five tumors arose from the stomach, and three tumors from rectum. Five tumors were classified as extragastrointestinal stromal tumors (E-GISTs). In 28 of 32 cases, more than one tumor manifestation (two to seven) was resected under treatment. In 18 of the patients, tumor samples from the primary tumor were available for comparison. The preselection of cases undergoing surgery during imatinib treatment explains why the proportion of patients with secondary tumor progression is higher than in an unselected series of patients with and without surgery under imatinib treatment. Further clinicopathologic data are shown in Table 1.

Clinical response was evaluated according to the Response Evaluation Criteria in Solid Tumors criteria (12). Four weeks after the initial start of imatinib therapy, magnetic resonance tomography or computer-assisted tomography scan was used to detect early tumor progression. Thereafter, reassessment of response was done every 3 months. Details of magnetic resonance tomography and computer-assisted tomography scan assessment have been published elsewhere (13).

Histopathology and immunohistochemistry. Histomorphologic subtype was evaluated as previously described (14) and was given as spindle cell type, epithelioid cell type, and mixed type. GIST diagnosis was confirmed by immunohistochemical analysis using antibodies against CD117 (KIT receptor), CD34, bcl-2, α-actin, desmin, S-100 protein, vimentin (all DAKO, Hamburg, Germany), PDGFRα (Santa Cruz Biotechnology, Santa Cruz, CA), and Ki-67 (MIB-1, Dianova, Hamburg, Germany) as previously described (15, 16).

Sequence analysis in KIT and PDGFRα genes. For mutational analysis, tumor tissue for DNA extraction was marked on H&E-stained slides. Areas for microdissection followed by differential mutational analysis were selected according to morphology or immunohistochemical expressions patterns under treatment with imatinib (Fig. 1). Tissue slides were deparaffinized by xylene and microdissected from serial sections (10 μm) of the primary tumors, metastases, and tumor tissue under imatinib treatment. Total DNA was extracted after pretreatment with proteinase K and absorption on silica gel membranes (Qiagen, Hilden, Germany). After estimation of DNA concentration by agarose gelelectrophoresis, relevant exons were amplified with intronic primers as previously described (15–17). The PCR products were purified using Micro Spin columns (Amersham Biosciences, Freiburg, Germany). Bidirectional DNA sequencing of the entire exons and the corresponding exon-intron boundaries was done with the Big Dye Terminator Cycle Sequencing Ready Reaction Kit (Applied Biosystems, Weiterstadt, Germany). Cycle sequencing products were precipitated with 3 mol/L sodium acetate and analyzed on an ABI PRISM 310 capillary electrophoresis system (Applied Biosystems). All sequence alterations were confirmed by an independent PCR amplification and sequencing to exclude PCR artifacts. The identity of the amplicon sequences was confirmed by database search (accession no. HSU63834, National Center for Biotechnology Information database: http://www.ncbi.nlm.nih.gov).

Statistical analysis. For descriptive data analysis, Fisher's exact tests, Student's t tests, Kaplan-Meier analysis, and log-rank tests, using SPSS 12.0 for windows, were used. P < 0.05 was considered statistically significant.

Median follow-up from first diagnosis to last evaluation was 51 months (range, 4-144 months). Median duration of response to imatinib treatment (progression-free survival; see Table 1) was 19 months (range, 0-45 months). Under treatment with imatinib, 2 tumors were primarily progressive and 25 others (78.1%) developed secondary tumor progression after 4 to 45 months of response (median, 19 months). The remaining five tumors achieved stable disease (n = 1) or partial remission (n = 4). Median survival after secondary tumor progression was 12 months (range, 1-29 months). Clinical and pathologic data of all cases are given in detail in Table 1.

Histopathologic and immunohistochemical findings. Twenty-two tumors were of spindle cell type; four GISTs showed an epithelioid phenotype; the other six lesions displayed a mixed morphology. In five cases, different morphologic subtypes were found comparing primary tumor and metastases. Before treatment, in all primary tumors, KIT receptor was expressed at least weakly and was lost in some of the metastatic lesions. Furthermore, strong reexpression of KIT receptor was observed in progressive lesions (see Fig. 2).

Primary KIT mutations. The majority of cases carried activating mutations in KIT, preferentially in exon 11 encoding the juxtamembrane domain. In detail, exon 11 mutations were found in 22 cases (68.8%) including 19 deletions (from 3 to 57 bp) and 3 point mutations. The types of KIT mutation in exon 11 are depicted in Fig. 3. Exon 9 mutations were detected in 7 cases (21.8%), always leading to a duplication of alanine-502 and tyrosine-503. No primary mutations were found in exons 12 and 18 of PDGFRα and in exons 13 and 17 of KIT. In 3 tumors (9.4%), wild-type sequences were found in all examined exons of KIT and PDGFRα (wild-type GISTs; see also Table 2).

Secondary mutations in KIT under imatinib treatment. In 14 of 32 patients (43.8%), secondary KIT mutations were identified (an example of sequence analysis is shown in Fig. 4). Ten patients developed only one type of secondary KIT mutation in the tyrosine kinase domain in exon 13 (n = 4), 14 (n = 4), or 17 (n = 2). One patient (case no. 10) showed four different mutations in exon 17 and the remaining three patients developed secondary mutations in two (n = 2) or three (n = 1) exons. Overall, secondary KIT mutations in exons 13, 14, and 17 were observed in seven, six, and five tumors, respectively. Secondary KIT mutations were only found in some of the tumor manifestations resected synchronously whereas others still carried a wild-type sequence in the respective exons.

Exon 13 point mutation led always to a substitution of valine-654 to alanine (V654A). This type of mutation has never been detected in primary GISTs up to now. In exon 14, three types of point mutations were identified. In four cases, threonine-670 was substituted by isoleucine (T670I). In one case, double point mutation led to a substitution of threonine-670 to glutamate (T670E). In one case, a point mutation leading to the substitution of serine-709 by phenylalanine (S709F) was found. Exon 17 mutations always involved the region between aspartate-816 and tyrosine-823. Sequence analyses of two tumors (nos. 10 and 27) with multiple different mutations have previously been published (17).

All 14 patients with one or more secondary KIT mutations belonged to the group of patients with secondary tumor progression whereas in 11 patients of this group, no secondary mutations could be detected. One of two patients with primarily progressive disease developed a secondary mutation in exon 14. Neither the four patients with partial remission nor the patient with stable disease carried a secondary KIT mutation. Details of sequences are summarized in Table 2.

Of the seven cases with primary KIT exon 9 mutation, 2 patients (28.6%) developed resistance due to acquired KIT mutations. In the 22 cases of KIT exon 11 mutated GISTs, 12 tumors (54.5%) were found to develop new mutations. Conclusively, the frequency of secondary KIT mutations was higher in tumors with primary KIT exon 11 mutation than in those with primary KIT exon 9 mutation, although not statistically significant. None of the three cases without any KIT or PDGFRα mutation (wt GISTs) developed new mutations. Altogether, the frequency of secondarily acquired KIT mutations under imatinib therapy in our study was ∼44%.

In summary, the primary KIT mutation was always detectable in all tumor manifestations before and under treatment. All 14 patients with newly acquired KIT mutations carried the secondary mutations only in some of their metastases whereas other lesions exhibited a wild-type sequence in the same genomic region. We never detected more than one acquired mutation in the same sample, indicating that long-term imatinib therapy leads to clonal selection of resistant tumor subclones.

Correlation of mutational and clinical data. There were no significant differences among GISTs with primary mutation in KIT exon 11, KIT exon 9, and wild-type GISTs with regard to the overall survival (from first diagnosis to death or last follow up), the time interval between first diagnosis and metastasis or recurrence, and the duration of response to imatinib. Thirteen of 25 GISTs with secondary progression and one of two primarily progressive tumors harbored secondary KIT mutations. On the other hand, none of five tumors with partial remission or stable disease under imatinib treatment developed secondary mutations in exon 13, 14, or 17. In our cohort, tumors with or without secondary KIT mutations did not differ significantly with respect to overall survival and survival after secondary tumor progression. Comparing the different loci of secondary mutations (i.e., exon 13, 14, or 17), there was no correlation with the primary tumor location (i.e., stomach, small bowel, rectum, or peritoneum) or the primary mutational type. However, tumors with secondary mutations in exon 14 were characterized by a more aggressive phenotype, indicated by a significantly shorter overall survival (mean, 41.0 months; SD, 14.8 months) compared with tumors which harbor secondary mutations in exon 13 or 17 (mean, 61.5 months; SD, 25.8 months; log-rank test, P = 0.0164; Kaplan-Meier curve not shown). Furthermore, GISTs with secondary exon 14 mutations had earlier development of metastases and/or recurrence after first diagnosis (5.5 versus 19.5 months; t test, P = 0.037) and a shorter progression-free survival under treatment (14.7 versus 19.5 months; n.s.) than tumors without exon 14 mutations. In comparison, the mean progression-free survival under imatinib treatment was 19.9 and 24.4 months for tumors with exon 13 and exon 17 mutations, respectively. The mean survival after secondary progress did not significantly differ among the different mutational groups: 13.7 months (exon 13) versus 15.6 months (exon 14) versus 16.8 months (exon 17). Cases with the acquisition of more than one secondary KIT mutation had a comparable prognosis as those with only one or no secondary mutation.

The prognosis of advanced GISTs has been improved significantly by the introduction of the tyrosine kinase inhibitor imatinib mesylate (Glivec, Novartis). However, with longer duration of treatment, the number of patients developing progressive disease increases continuously. The most recent update of the European Organization for Research and Treatment of Cancer 62005 trials shows a steady decline in progression-free survival (18). Several mechanisms for secondary progression have been proposed: (a) secondary gene amplification leading to overexpression of KIT or PDGFRα receptor, (b) activation of alternative receptor tyrosine kinases, or (c) acquisition of secondary mutations of the KIT or PDGFRα gene interfering with the inhibitory effect of imatinib. The latter mechanism is also a common reason of resistance to imatinib in chronic myelogenic leukemia (10, 11).

Several reports on secondary KIT mutations provide first evidence that this mechanism might also be a reason for imatinib resistance in GISTs (5–7, 9, 17). In addition to the primary mutation, new KIT mutations occur and are preferentially located in the first or second tyrosine kinase domain. The exchange of single amino acids in these domains probably leads to a change of the three-dimensional receptor conformation, which presumably modifies the ATP-binding pocket and thus might inhibit imatinib binding to the receptor.

In our study, we analyzed 32 cases of advanced progressive GISTs for the occurrence of acquired KIT mutations in both tyrosine kinase domains (exons 13, 14, and 17). From the majority of patients, more than one tumor manifestation was examined (total of 104 samples). In all cases, tumor tissue was resected under treatment because of tumor progression or bleeding. Most patients (78.1%) had secondary progress under treatment and two patients were primarily progressive (6.3%). The others (n = 5) had a partial response or stable disease. We found secondary KIT mutations in 14 cases (43.8%), which were located in exon 13, 14, or 17 encoding the first or the second tyrosine kinase domain, respectively. In four patients, more than one secondary KIT mutation was found. The frequency of acquired KIT mutations in GISTs is comparable with the results of another very recent study (19).

Several mechanisms increasing the risk of development of secondary resistance due to acquired KIT mutations have to be discussed. First, the amount of remaining tumor cells under imatinib treatment might influence the risk of acquisition of new KIT mutations. As shown by several groups (3, 4), location of the primary KIT or PDGFRα mutation directly predicts the clinical response to imatinib. If the inhibitory effect of imatinib depending on the underlying activating mutation is low, the larger amount of remaining tumor might be leading to a higher risk of secondary mutation. However, our own results show that tumors with an underlying primary KIT mutation in exon 11 (54.5%), known to be the subgroup with better response rates than other mutational subtypes, are at higher risk to develop secondary mutations than those with an KIT exon 9 mutation (28.6%). This observation supports our hypothesis that the probability of a secondary mutation increases with duration of imatinib treatment, which most often is longer in GISTs with exon 11 mutation than in those with exon 9 mutation or wild-type GISTs. Second, development of proliferating tumor clones could be a result of changed imatinib pharmacokinetics. Decreased imatinib levels under chronic treatment could be due to increased imatinib clearance (20) or to decreased intracellular accumulation of imatinib (21). Another possibility might be a change in the serum level of KIT or KIT ligand as proposed by Bono et al. (22). Furthermore, several drugs known to influence the serum level of imatinib are frequently taken by older GIST patients (23). Reactivation of GIST cells due to decreasing imatinib level during treatment might propagate proliferation and escape mechanisms from treatment. Third, tumor cells might lose their imatinib sensitivity due to still unknown molecular mechanisms such as activation of other tyrosine kinases or amplification of KIT or PDGFRα gene.

Interestingly, each tumor nodule under progression apparently develops an individual clonal evolution. Secondary mutations were only found in some of the examined lesions. Further evidence for an individual development of different tumor residues is provided by the detection of multiple acquired mutations in the same patient. Tumor parts with different acquired mutations were found side by side in one nodule and exhibited a different immunohistochemical expression pattern (Fig. 2) or grade of regression.

In our series, GISTs with secondary exon 14 mutation showed a more aggressive behavior with earlier metastasis and shorter progression-free survival. One might conclude that this subgroup of early metastasizing tumors is predisposed to develop exon 14 mutations whereas tumors with a slower progress might acquire secondary mutations in exon 13 or 17 after a longer treatment with imatinib.

Our results indicate that secondary KIT mutations may be one mechanism leading to late resistance to imatinib treatment. In the group of secondarily progressive tumors under imatinib treatment, their occurrence per se is not an indicator of a worse prognosis in comparison with progressive cases without secondary mutations. The latter obviously develop other molecular mechanisms to escape the inhibitory efficacy of imatinib. However, the identification of acquired mutations by analyzing all available progressive lesions under treatment clearly proofs the secondary progress and is particularly important in view of new therapeutic strategies. These might include either surgical resection of individual progressive lesions or treatment with other small molecules such as SU11248 or inhibitors of protein kinase C or phosphatidyle-inositole 3-kinase. The full understanding of the underlying molecular mechanisms of oncogenesis and resistance in GISTs may also help to gain new insights and establish further comparable strategies in other cancers types.

We thank Christiane Esch, Susanne Steiner, Barbara Reddemann, Ellen Paggen, Sandra Böhler, Theresa Buhl and Gerrit Klemm for technical assistance.