c-Kit/PDGFRA Gene Status Alterations Possibly Related to Primary Imatinib Resistance in Gastrointestinal Stromal Tumors Free

The success of targeted therapies requires the presence of a tumor addicted to a single, pathogenetically crucial, expressed, and activated oncogene. Imatinib and its mechanism of action in gastrointestinal stromal tumors (GIST) have become the paradigm for these new treatment strategies, but despite its undoubted effectiveness, imatinib does not fully control the disease in all patients and a proportion of those who do show a significant drug response relapse during the course of treatment.

It seems that the mechanisms of secondary resistance are related to the selection of subclones carrying gain or loss copies of the c-Kit gene (1) or to mutant receptors in which amino acid residues (essential for the correct “fit” of the inhibitor molecule into the tyrosine kinase pocket) are changed by DNA point mutations (2–6), but the mechanisms underlying primary resistance remain only partially understood.

c-Kit and PDGFRA mutations have been found in 88% of GIST patients (7), but ∼60% to 65% of patients treated with imatinib show a partial response, and another 20% experience stable disease (8). These responses are highly dependent on the regions of c-Kit, and PDGFRA affected by mutations as primary resistance is frequently associated with a wild-type c-Kit, mutations in exon 9 or 17 of c-Kit, or a D842V mutation in PDGFRA (9–12); moreover, codon changes in the distal part of c-Kit exon 11 correlate with a significant increase in the risk of progression (13). Unlike chronic myeloid leukemia (14), the response of GISTs does not seem to be affected by multidrug protein expression (MDR; ref. 15). It has also been pointed out that careful analyses of imatinib resistance in GISTs will probably reveal other targets (8).

Taking advantage of a series of 35 patients with advanced disease who were surgically treated after receiving imatinib (27 responders and 8 progressors), we decided to investigate residual tumoral tissue by means of molecular, biochemical, and cytogenetic analyses and, when possible, to compare the results of sequencing and fluorescence in situ hybridization (FISH) analyses with those of the recovered corresponding primary tumors.

The results showed that the presence of a genetic instability could be considered as another mechanism of primary resistance, and that several GISTs are characterized by a genuine, constitutive alteration in the copy numbers of both c-Kit and PDGFRA in addition to mutually exclusive mutations.

Patients and materials. The case material consisted of surgical samples obtained from a cohort of 35 patients (15 men and 20 women with a median age of 55 years) who underwent resection after receiving imatinib treatment (400 or 800 mg/day depending on their randomized dose in the European Organization for Research and Treatment of Cancer-Soft Tissue and Bone Sarcoma Group trial; ref. 16) as described in a previous paper (17). More precisely, the series included 27 clinically responding (group A) and 8 clinically progressing patients (group B) of the previous study (3 patients who received neoadjuvant treatment were excluded; ref. 17). Briefly, the responders were defined as patients who had received imatinib treatment for at least 12 months (the median preoperative duration of imatinib was 17 months; range, 7-39) in whom stable disease, or a partial or complete response was observed, and the progressors as those on imatinib treatment (median preoperative duration 21 months, range 6-39) showing progression after response on the basis of imaging findings. In group A, the best radiological responses at the time of surgery were stable disease in 1 patient, a partial response in 18 patients, a complete response in 7 patients, and a hepatic complete response combined with abdominal partial response in 1 patient. In group B, one patient showed primary and seven developed a limited secondary progression. In particular, two patients had two lesions in progression and five patients had only one lesion in progression (an abdominal nodule or a liver nodule in two cases each and a liver recurrence of a resected GIST with bladder invasion in one). No surgical clearing was achieved in any of the cases. Imatinib treatment was stopped 24 h before the surgical procedures, in accordance with the standard of care. None of the patients had previously received chemotherapy. Written informed consent was obtained from all of the patients.

The 35 surgical specimens came from advanced and/or metastatic GISTs: 2 primary tumors (esophagus case 3 and small intestine case 21) with metastases, 29 abdominal recurrences (10 associated with liver metastases), and 4 liver metastases (Tables 1 and 2). On the basis of the clinical records, the primary tumors (32 treated elsewhere and 3 in our institution) were located in the esophagus in 1 case, stomach in 5, small intestine in 19, colon in 2, rectum in 4, and other sites in 4. The GIST diagnosis was confirmed by means of immunohistochemical analysis (on referred slides in 32 cases and institutional slides in 3) using antibodies against CD117, CD34, desmin, and actin, as previously described (18). Thirty of the tumors were of the spindle cell type, two were epithelioid, and three had a mixed morphology.

Subsequently, a request for unstained sections was forwarded, and 27 out of 32 primary tumors were recovered for molecular and cytogenetic analyses. The received blocks or unstained sections were processed for diagnostic confirmation and selected under a microscope as material for molecular analysis (microdissected when necessary) and FISH. The primary tumors were analyzed in 30 (27 + 3) of the 35 cases (shown in Tables 1 and 2, in bold).

Pathology: immunophenotyping and histologic response assessment. The immunohistochemical analyses were made using 2-μm, formalin-fixed, and paraffin-embedded tumoral sections of representative samples of the surgical specimens using antibodies against CD117, CD34, desmin, and actin as previously described (18). Moreover, all of the samples were immunoperoxidase phenotyped using a Ki-67 antibody (clone Mib-1; DAKO, Carpinteria, CA; diluted 1:200), and antigen was retrieved by pretreating the sections in citrate buffer for 15 min at 95°C in an autoclave.

The microscopic histologic response was scored on the basis of the residual viable cell component as high in the presence of acellular areas or <10% residual tumoral cells; moderate in the presence of 10% to <50% residual tumoral cells; and mild/low when the cell component was >50% to 90%.

To determine cell proliferation, the mitotic and Ki-67 labeling indices were counted under a microscope. The mitotic index per 50 high-power fields was assessed in each of the cellular cases on the basis of the proposed guidelines (ref. 19; see Tables 1 and 2); the Ki-67 labeling index was defined as low when nuclear labeling was observed in <10% of the tumoral cells, moderate when it was observed in 10% to 20%, and high when it was observed in >30%.

Molecular analysis. A total of 72 paraffin-embedded samples from 34 of the surgical specimens were analyzed for c-Kit and PDGFRA mutations; the remaining specimen (case 7) contained no viable tumoral cells (100% regression). We also analyzed samples of primary tumor specimens obtained from 30 patients before imatinib treatment (shown in Tables 1 and 2, in bold).

The tumoral samples had various response rates (for details, see Tables 1 and 2 and the corresponding legends) and were isolated by microdissection for the molecular investigations (18); the DNA was extracted following standard procedures (20). c-Kit exons 9, 11, 13, 14, and 17 were amplified using the following primer sequences and annealing temperatures (designed): exon 9 (5′-ATTTATTTTCCTAGAGTAAGCCAGGG-3′ and 5′-ATCATGACTGATATGGTAGACAGAGC-3′ at 65°C), exon 11 (5′-ATTATTAAAAGGTGATCTATTTTT-3′ and 5′-ACTGTTATGTGTACCCAAAAAG-3′ at 54°C), exon 13 (5′-CACCATCACCACTTACTTGTTGTCT-3′ and 5′-GACAGACAATAAAAGGCAGCTTGGAC-3′ at 67°C), exon 14 (5′-TCTCACCTTCTTTCTAACCTTTTC-3′ and 5′-AACCCTTATGACCCCATGAA-3′ at 54°C), and exon 17 (5′-TGAACATCATTCAAGGCGTACTTTTG-3′ and 5′-TTGAAACTAAAAATCCTTTGCAGGAC-3′ at 65°C). The primers and PCR conditions used to amplify PDGFRA exons 12, 14, and 18 have been previously described (21, 22). The sequencing was carried out using an automated 377 DNA Sequencer ABIPRISM-PE, Applied Biosystems (Foster City, CA) following standard protocols.

Biochemical analysis: protein extraction and immunoprecipitation/Western blotting. Immunoprecipitation and Western blot analyses were only done when sufficient frozen material was available. The proteins were extracted, and 0.5 mg of protein lysates were immunoprecipitated for KIT and blotted as previously described (23).

FISH analyses. Representative areas were selected from 51 tumoral samples of 30 surgical specimens (22 from group A and 8 from group B) under microscopic control and analyzed to define the allelic status of the c-Kit and PDGFRA genes. We also analyzed samples of primary tumor specimens obtained from 30 patients before imatinib treatment (shown in Tables 1 and 2, in bold).

BAC clones RP11-586A2 (c-Kit gene, 4q12) and RP11-231C18 (PDGFRA gene, 4q12) were used as the FISH probes, and CEP4 labeled with Spectrum Orange (SO); (Vysis, Downers Groove, IL) as the control probe, as previously described (23). The probe mixtures prepared for dual-color FISH were Spectrum Green (SG)-labeled cKIT + SO-labeled CEP4, and SG-labeled PDGFRA + SO-labeled CEP4. Each mixture was placed on the preselected target area under microscopic control, covered with a coverslip, and sealed with rubber cement.

The samples were denaturated and hybridized in HYBrite Thermoblock (Vysis) and mounted using 4′,6-diamidino-2-phenylindole II antifade (Vysis). The slides were observed using a Zeiss Axioscope equipped with a 100-W mercury lamp and 4′,6-diamidino-2-phenylindole, SG, and SO filters (Vysis), and the signals were counted in at least 40 nuclei per slide under 600×/1,000× magnification. The images were acquired using a cooled CHOU 4912 camera (Applied Imaging Corporation, San Jose, CA) and McProbe software (Vysis).

One hundred interphase nuclei were evaluated for each selected area, and the ratio of KIT/PDGFRA to CEP4 was calculated. The cytogenetic pattern was classified as disomic when ≤2 chromosome 4 were present in more than 90% of cells; trisomic when there were 3 chromosome 4 in more than 10% of cells; low polysomic when ≤4 chromosome 4 were present in more than 40% of cells; high polysomic when there were ≥4 chromosome 4 in ≥40% cells; and gene amplification when the gene/chromosome ratio was >2, or when ≥15 gene copies were present in ≥10% of cells (24, 25).

After morphologic and immunophenotypical confirmation of the diagnosis of GIST, 132 samples/nodules from the 35 surgical specimens were assessed for tumor response on the basis of residual cellularity, the mitotic index, and Ki-67 labeling for each nodule in each case (the details regarding viable cellularity are shown in Tables 1 and 2).

Retaining the clinical/radiological division of responding and progressing cases, the following distribution was observed.

Responding cases. Thirteen cases fell into the high histologic response group (high responders or the low cellular/low-proliferating group) showing 0% to <50% residual viable tumoral cells in addition to no mitosis, no obvious immunostaining, or Ki-67 labeling in <10% of the cells. Eight cases showed a moderate histologic response coupled with similar mitotic and Ki-67 features to those of the first group, i.e., >50% to 90% tumoral cells, no mitoses, Ki-67 immunostaining in <10% of cells (moderate responders or the cellular/low-proliferating group). Six cases showed a mild/low histologic response, but with a mitotic index of >10/50 high-power fields (range 12-50) and Ki-67 immunostaining in 20% to 30% or >30% of cells; these were considered low responders or the cellular/high proliferating group.

Progressing cases. All eight cases were low responders (cellular/high-proliferating group).

Primary mutations. A total of 27 out of 34 patients (regardless of group) showed a mutated c-Kit exon 11 (79.5%): 13 deletions (48.2%), 4 point mutations (14.8%), and 10 cases with both alterations (37%). Of the remaining seven patients, three carried a c-Kit exon 9 mutation (8.8%), and four expressed the wild-type sequence (11.7%). No PDGFRA mutations were detected. All of the mutations are listed in Tables 1 and 2.

Secondary mutations. One group A patient (case 27) showed the previously reported imatinib-acquired mutation V654A in c-Kit exon 13. This was also detected in three group B patients (cases 32, 34, and 35), whereas one group B patient showed the known T670I substitution in c-Kit exon 14 (case 28), and another patient (case 31) carried the mutation D820N in exon 17.

Before imatinib treatment (primary tumors). In all of the analyzed cases for which primary tumor samples were available, the same primary c-Kit mutation was detected in each sample before and after imatinib treatment (shown in Tables 1 and 2, in bold).

Eight cases for which cryopreserved material was available were analyzed for KIT receptor expression and phosphorylation by means of total protein extract immunoprecipitation experiments (see Tables 1 and 2 and their legends). As shown in Fig. 1A and B, a highly expressed and phosphorylated KIT (++P; +++P) was present in the samples in both the cellular/low-proliferating and cellular/high-proliferating groups, regardless of whether the patient were in group A (cases 14, 19, 20, and 25) or group B (cases 28, 29, 30, and 31) patients.

After imatinib treatment. A total of 11 out of 23 responding cases (group A) showed a normal chromosome 4 disomic pattern (ratio, ∼1), and two cases showed a disomic pattern associated with a hemizygous deletion of c-Kit/PDGFRA genes (ratio, ∼0.5); 8 cases showed a different chromosome 4 number, consisting of trisomy (2 cases), or low or high polysomy (6 cases); 2 cases were not evaluable. Seven out of eight progressing cases (group B) showed a normal disomic pattern, and one a high polysomic pattern associated with gene deletion (ratio, ∼0.5).

Group A: Subgroup with a high histologic response. In each case, one sample was examined. All of the samples carrying a c-Kit mutation (cases 1, 2, 3, 5, 8, 9, 10, and 11) showed disomy of chromosome 4; case 12 carrying wild-type c-Kit showed trisomy.

Group A: Subgroup with a moderate response. The analyses were made of four nodules in two cases, three nodules in four cases, and two nodules in one case; case 18 was not evaluable. All of the nodules showed the same cytogenetic pattern except those from case 16, which showed a disomic condition (ratio, ∼0.5) in three nodules and low polysomy with the same ratio in one. Case 21 was normal disomic; case 19 showed a disomic condition associated with deletion (ratio, ∼0.5); cases 14, 17, and 20 showed low polysomy.

Group A: Subgroup with a low response. One nodule was analyzed in two cases (22 and 27) and more than one in the remaining three cases. No material was available for case 23. Both of the nodules of case 24 showed a chromosome 4 trisomic pattern; case 25 showed high polysomy in each of the three nodules examined; and case 26 showed high polysomy in three nodules and disomy in one. Case 27 carried a secondary mutation and, like other cases in group B, showed a disomic pattern.

Group B: Subgroup with a low response. Seven of the eight cases showed a normal disomic pattern; case 29 showed high polysomy of chromosome 4 associated with a gene deletion (ratio, ∼0.5).

Before imatinib treatment (primary tumors). Thirty primary tumors were analyzed. A total of 12 of the 22 responding cases (group A) showed a chromosome 4 disomic pattern with a ratio of ∼1; one showed a disomic pattern with a ratio of ∼0.5; two showed a trisomic pattern; six showed chromosome 4 polysomy (low or high); and one was not evaluable. Four of the eight progressing cases (group B) showed a normal disomic pattern; two showed a polysomic pattern, one of which coupled with a gene deletion (ratio, ∼0.5); and two cases were not evaluable.

Group A: Subgroup with a high histologic response. Six cases showed normal disomy (cases 5, 6, 8, 9, 10, and 11), two cases trisomy (cases 4 and 12), and two low polysomy (cases 1 and 2).

Group A: Subgroup with a moderate response. Four cases showed normal disomy (cases 14, 16, 18, and 21); case 19 was disomic with a ratio of ∼0.5; two cases (cases 17 and 20) showed low polysomy; and case 15 showed high polysomy.

Group A: Subgroup with a low response. In this subgroup, two cases showed a normal disomic pattern (cases 22 and 23); case 25 showed low polysomy; one case was not evaluable (case 27).

Group B: Subgroup with a low response. Four of the eight cases showed a normal disomic pattern (cases 30, 31, 32, and 34); cases 29 and 33 showed polysomy of chromosome 4 with gene deletion (respectively, high with a ratio of ∼0.5, and low with a ratio of >0.5 and <1); two cases (cases 28 and 35) were not evaluable.

Correlation between pre- and post-imatinib cytogenetic profiles. The available material allowed us to make a pre- and post-imatinib correlation in 23 cases (see Tables 1 and 2, rows in bold). Seventeen cases presented the same cytogenetic pattern: 11 normal disomy; 1 disomy with a ratio of ∼0.5; and the remaining 5 aneusomy (the high polysomy in case 29 was associated with gene deletion). Figures 2, 3, and 4, respectively, show cases 17, 19, and 29 and the corresponding pre- and post-imatinib histologic sections.

Three cases showed low polysomy in a primary tumor and disomy in posttreatment nodules (cases 1, 2, 33); finally, one case each showed disomy before imatinib and low polysomy afterward (case 14); disomy with a ratio of ∼1 before and disomy with a ratio of ∼0.5 after imatinib (case 16); and low polysomy before and high polysomy afterward (case 25).

On the basis of published reports, histologic examinations of resected lesions after imatinib treatment in GISTs have shown a partial response in the majority of cases and a complete response in only 12% (26, 27). The partial responses are due to c-Kit genomic amplifications or c-Kit/PDGFRA secondary mutations leading to late resistance to imatinib treatment, as well as to the presence of c-Kit or PDGFRA primary resistance mechanisms (9). Mutation-based secondary resistance will only be overcome by next-generation inhibitors (28), but the use of high-dose regimens may be beneficial in cases of primary resistance (13).

We describe a further biological mechanism possibly responsible for drug resistance, which is evident in apparently still clinically responding patients who nevertheless show morphologic, biochemical, and cytogenetic evidence of imatinib resistance.

Starting with morphology, our retrospective review not only shows that the progressing group showed a low response (all cellular/high proliferation rate cases), but also that the post-imatinib morphologic changes in the responding cases covered high responders (the low cellular/low-proliferating subgroup), moderate responders (the cellular/low-proliferating subgroup), and low responders (the cellular/high-proliferating subgroup), whose morphologic and proliferation characteristics were similar to those of the progressing cases.

In molecular terms, six of the eight progressing cases (the cellular/high-proliferating subgroup) showed changes related to resistance: one with a primary resistant mutation (case 30) and five with secondary mutations. However, with the exception of patient 27 who showed a secondary mutation and (as expected) fell into the low-response subgroup, the distribution of the so-called c-Kit primary resistance mutations (9) was similar in all of the three subgroups of responding cases.

The genomic profile of c-Kit and PDGFRA was disomic in all but one of the progressing cases (case 29 with high polysomy associated with hemizygosity), but segregated into three main profiles in the responding cases: disomic in all but one high responder subgroup, and mainly aneusomic in the moderate and low responder subgroups, with evidence of numerical variations in of both c-Kit and PDGFRA in 9 of the 12 cases.

It is worth noting that, in terms of morphology/proliferation markers and KIT expression/activation, the moderate responding (cellular/low proliferating) subgroup reflects the recently described pattern observed in GIST cell lines (29) and the mouse model carrying KIT^V558Δ/+(30), which have been proved to be responsive to imatinib and in which one of the main effect of imatinib is suppressed cell growth. Assuming that the tumoral cells from the GIST cell lines, the mouse model, and our samples all depend on activated KIT receptors, our data seem to fit preclinical models. However, at odds with preclinical experimental findings indicating the switching off of activated KIT receptor, signs of KIT activation were observed in the cases in our moderate and low-response subgroups regardless of their quiescent or proliferating status, thus not only indicating that the tumors are still dependent on KIT activation, but also that both cellular subgroups are likely to be less sensitive to imatinib. This is not entirely unexpected because the efficiency of imatinib not only depends on the inhibition of mutation-activated KIT but also on the blockade of other mechanisms supporting tumoral growth, such as the gain or loss of c-Kit/PDGFRA genes or the activation of signal transduction pathways independently of the receptors. Consequently, crossover the recommended imatinib dose could be more effective in cases carrying c-Kit/PDGFRA alterations, although further studies of a larger series with more homogeneous disease presentations and treatments could better clarify this point. In any event, we cannot exclude the possibility that the KIT activation may have been related to KIT reactivation due to imatinib discontinuation before surgery.

In each individual case, the same genetic deregulation profile was observed in all of the tumoral nodules regardless of the morphologic response pattern, thus suggesting that these alterations may represent an early event. This hypothesis, which is further supported by very preliminary data showing the same profile in pre- and post-imatinib tumoral samples from five patients, prompted us to recover as many of the primary tumors as possible to compare the pre- and post-imatinib molecular/cytogenetic profiles.

No changes were observed at the molecular level, but comparison of the pre- and post-imatinib c-Kit/PDGFRA gene profiles of 23 cases (shown in bold row in Tables 1 and 2) showed the normal disomic pattern in 11 paired samples and the same abnormal genetic pattern in six cases (cases 12, 15, 17, 19, 20, and 29), which indicates that the cytogenetic aberrations were already present at the time of disease onset and suggests that they may represent a mechanism of primary resistance. Furthermore, cases 14, 16, and 25 showed a gain/loss in the c-Kit gene copy number in the post-imatinib samples, a change that has previously been reported by Debiec-Richter (1) and which is consistent with a possible nonmutation-based mechanism of secondary resistance. Finally, cases 1, 2, and 33 changed from low polysomy to disomy which, to the best of our knowledge, has not been reported before and for which we do not yet have an explanation. Cumulatively, a gain/loss of c-Kit/PDGFRA gene copy number occurred in about 1/3 of the cases (12 of the 27 evaluable primary tumors), which suggests that it is not a chance finding but indicates a genuine and constitutive alteration that, in addition to mutually exclusive c-Kit/PDGFRA mutations, characterizes a number of GISTs.

Interestingly, unlike c-Kit and PDGFRA mutations, genomic alterations always seem to involve both genes and characteristically show the same genetic profile. Furthermore, it is worth noting that the cases characterized by secondary mutations always segregate with the disomic pattern. Accordingly, in an analysis of a group of 33 patients with secondary resistance, most (22 cases) of whom were characterized by secondary mutations, only two cases were found to have low levels of c-Kit amplification (11). Our results seem to justify investigating whether GISTs carrying gene copy number abnormalities are not inclined to develop secondary mutations, and whether nonmutation-based secondary resistance is part of the primary tumor make-up.

We did not find any correlations between the cytogenetic alterations and outcomes in our group of responding patients, possibly because of the retrospective nature of our analysis and the heterogeneity of the patients in terms of the preoperative duration of imatinib treatment, doses, tumor phase, surgical clearing, etc. Further studies are therefore essential to clarify the clinical role of these cytogenetic alterations. Moreover, the radiological evaluation also did not always correlate with the biological response and, thus, with the molecular profile of the tumors.

In conclusion, our results underline the remarkable cytogenetic heterogeneity of the tumors in the responding patients possibly related to a mechanism of late primary resistance. If these data are confirmed, assessments of GIST cytogenetic make-up and mutational analyses can be expected to become a significant part of GIST management.

We thank the following pathologists who kindly contributed case material of preimatinib tumors: C. De Gaetani (Azienda ospedaliera-Policlinico, Modena), M. Albrizio (Azienda ospedaliera “Di Venere-Giovanni XXIII”, Bari), M. Fabio (Ospedale Sant'Orsola, Ordine Ospedaliero Fatebenefratelli, Brascia), E. D'Amore (Unità Locale Socio Sanitaria, Vicenza), G. De Dominicis (Ospedale Cardarelli, Napoli), M. Assi (Ospedale Civile, Legnano), F. Stella (Ospedale Civile, Voghera), M. Barberis [Casa di Cura “Santa Maria”, Castellanza (Varese)], M. Cornaggia Medici [Ospedale Casa di Cura “San Carlo”, Paderno Dugnano (Milan)]; E. Schiaffino (Azienda Ospedaliera “San Carlo Borromeo,” Milano), M. Filotico (Studi Diagnostici Specializzati, Lecce), C. Patriarca [Presidio ospedaliero, Vizzolo Predabissi, Melegnano (Milan)]; F. Giangaspero (Ospedale M. Bufalini, Cesena), A. Parma (Ospedale Riuniti, Bergamo), F. Menestrina (Policlinico G.B. Rossi, Verona), R. Ribacchi (Azienda ospedaliera “R. Silvestrini”, Perugia), F. Crivelli [Ospedale S. Antonio Abate, Gallarate (Varese)], M. Cazzaniga [Ospedale Civile di Legnano, Po di Magenta (Magenta)], E. Padolecchia (Casa di Cura “Poliambulanza”, Brescia), G. Baruzzi (Ospedale Maggiore, Bologna); R. Colombari [Ospedale Cazzavillan, Arzignano (Vicenza)], V. Mambelli (Ospedale C.G. Mazzoni, Ascoli Piceno), R.M. Bona [Ospedale Miulli, Acquaviva delle Fonti (Bari)], M. Verroli (Presidio ospedaliero di Eboli, Salerno).