Fax: (301) 480-5142
A prospective analysis of imatinib-induced c-kit modulation in ovarian cancer†
A phase II clinical study with proteomic profiling
Article first published online: 8 JUN 2007
Copyright © 2007 American Cancer Society
Volume 110, Issue 2, pages 309–317, 15 July 2007
How to Cite
Posadas, E. M., Kwitkowski, V., Kotz, H. L., Espina, V., Minasian, L., Tchabo, N., Premkumar, A., Hussain, M. M., Chang, R., Steinberg, S. M. and Kohn, E. C. (2007), A prospective analysis of imatinib-induced c-kit modulation in ovarian cancer. Cancer, 110: 309–317. doi: 10.1002/cncr.22757
This article is a US Government work and, as such, is in the public domain in the United States of America.
- Issue published online: 29 JUN 2007
- Article first published online: 8 JUN 2007
- Manuscript Accepted: 3 APR 2007
- Manuscript Revised: 22 MAR 2007
- Manuscript Received: 9 MAR 2007
- Intramural Research Program of the Center for Cancer Research
- National Cancer Institute
- National Institutes of Health
- ovarian cancer;
- protein arrays;
- clinical trial
c-Kit and platelet-derived growth factor receptor (PDGFR) are potential molecular targets in epithelial ovarian cancer (EOC). Imatinib inhibits the kinase domain and subsequent downstream signaling of these receptor tyrosine kinases. The objective of this study was to investigate biochemical and biologic effects of imatinib on EOC.
Patients with recurrent EOC who had received no more than 4 prior regimens and who had good end-organ function were eligible. Imatinib was administered orally at a dose of 400 mg twice daily in continuous, 28-day cycles with reassessment imaging studies obtained every other cycle. Tumor core biopsies were obtained prior to and at 4 weeks into therapy; microdissected tumor and stroma were subjected to protein lysate array analysis. Blood samples were obtained monthly for cytokine measurements.
Twenty-three patients were enrolled, including 16 patients who received imatinib 600 mg daily because of gastrointestinal (GI) toxicity and fluid accumulation at the starting dose. The median time to disease progression was 2 months (range, 2–14 months). Common grade 3 toxicities included edema/ascites/pleural effusions in 11 patients (48%), GI complaints in 8 patients (35%), fatigue in 3 patients (13%), and grade 2 and 3 cytopenias in 10 patients and 3 patients (43% and 13%), respectively. Increased circulating levels of interleukin 6 were associated with grade ≥2 fluid collection (P = .02). A statistically significant trend was observed between pretreatment phosphorylated-kit levels in microdissected tumor and stroma and GI toxicity (P < .01), between tumor levels of epidermal growth factor receptor (EGFR) and PDGFR with grade of fatigue (P ≤ .005), and EGFR and phosphorylated-AKT levels with grade of ascites and edema (P ≤ .01).
The results of this study indicated imatinib had minimal activity as a single agent in EOC. Its ability to modulate its molecular targets suggests that it may be considered in combinatorial therapy. Cancer 2007. Published 2007 by the American Cancer Society.
Ovarian cancer is the leading cause of death from gynecologic malignancies in American women.1 Despite improvements in the treatment of advanced-stage disease with aggressive surgical management and adjuvant chemotherapy, >75% of patients will develop recurrent disease and become resistant to chemotherapy.2 Responses to subsequent treatment may be short-lived and can be associated with considerable toxicity. This underscores the drive for new therapeutics for ovarian cancer. Advances in molecular biology have brought a new class of therapeutic agents designed to inhibit the molecular signaling that is vital to cancer progression and survival. Protein tyrosine kinases are initial points for outside-in signaling that drives cellular vitality. They often are dysregulated, mutated, or overexpressed in human malignancies.3–5 Imatinib mesylate was the first in this class to demonstrate clinical benefit6–8. It was designed originally to inhibit the activity of bcr-abl, the fusion gene product that transforms and drives chronic myelogenous leukemia8. Additional preclinical characterization of this agent showed that the activity also encompassed other type III receptor tyrosine kinases, including c-kit9–11 and platelet-derived growth factor β (PDGFR-β).12
Both c-kit and PDGFR-β have been implicated as molecular targets for epithelial ovarian cancer.9, 13–16 Both markers are expressed in the tumor and also are found in the supporting stroma.17–20 This led to our hypothesis that imatinib would alter the growth of ovarian cancer by targeting its microenvironment and its growth machinery. We proposed that the treatment of women who had epithelial ovarian cancer with imatinib would result in perturbation of c-kit and PDGFR-β activity and/or quantity with secondary effects of altering the activation of downstream signaling targets. This Phase II trial was designed to explore the clinical activity and toxicity of imatinib in patients with ovarian cancer and to profile changes in biochemical signaling events in tumor and stroma using tissue lysate array proteomics.21
MATERIALS AND METHODS
This study was approved by the Institutional Review Board of the National Cancer Institute (Bethesda, Md), and written informed consent was obtained before enrollment. Eligible patients had recurrent epithelial ovarian cancer for which they had: received ≤4 prior regimens; had an Eastern Cooperative Oncology Group performance status of 0, 1, or 2; and had good end-organ function (leukocyte count ≥3000/mm3, absolute neutrophil count >1500/mm3, hemoglobin ≥9.0 g/dL, platelets ≥100,000/mm3, serum creatinine ≤1.5 mg/dL, transaminases ≤2.5 times the upper limit of normal, and bilirubin ≤1.5 mg/dL). Measurable disease by physical examination, noninvasive radiographic imaging, or surgical evaluation was required, including 1 lesion that would be subjected to core-needle biopsy. No chemotherapy or radiation was allowed within 4 weeks of study entry (6 weeks for carboplatin) and recovery of pre-existent clinical toxicity to grade ≤1. Ineligible patients had brain metastases, cardiac dysrhythmias that required treatment, active infection, a prior invasive cancer, or concomitant use of medication with potential interaction with imatinib. Patients who had rapidly progressive disease with potential for an imminent medical catastrophe (eg, bowel obstruction or fistula) were counseled to consider chemotherapy.
On-study and Cycle Evaluations
Patients were assessed before starting and monthly with a history and physical examination, including pelvic examination by a gynecologic oncologist, history of prescription and nonprescription drugs, and assessment of adverse events. Pretreatment studies included a complete blood count (CBC); platelet count; electrolytes; renal, mineral, and liver function panels; urinalysis; prothrombin time; CA-125; electrocardiogram; and chest radiograph within 4 days of enrollment; laboratory studies were repeated each cycle with CBC and transaminases checked weekly during the first cycle. Imaging of measurable disease was performed within 14 days of initiation and then at least every other cycle.
Treatment Plan and Dose Modifications
The initial imatinib dose was 400 mg twice daily on continuous, 28-day cycles starting the day after the initial research biopsy. Adverse-event diaries were kept by the patient and were reviewed each cycle; Adverse events were graded according to Common Toxicity Criteria (CTC) version 2.0. Treatment was interrupted for patients who experienced CTC grade 3/4 toxicity with a dose reduction if treatment was reinitiated. The dose was reduced (200 mg daily) for resolved grade 3/4 or progressive or recurrent grade 2 toxicity. Patients were not eligible to resume imatinib if the time to resolution to grade 0/1 toxicity was >4 weeks. Dose reduction to 400 mg daily was allowed. Patients who could not tolerate 400 mg daily were removed from study.
Response Assessment and Clinical Endpoints
Tumor status was characterized by using Radiographic Evaluation Criteria in Solid Tumors (RECIST) criteria.22 CA-125 was measured but was not used to monitor disease response, because the effects of imatinib on CA-125 production and secretion are unknown. Development of a new effusion, pleural or peritoneal, was not considered progressive disease, because this is a known complication of imatinib therapy.11, 23
Analysis of plasma and ascites cytokine concentrations
Blood was sampled in heparinized tubes on-study and at monthly clinic visits. Plasma was separated, aliquoted, and stored at −70°C within 4 hours. Ascites and pleural fluids that were collected for research were spun free of cells, aliquoted, and frozen within 2 hours. Concentrations of vascular endothelial growth factor (VEGF), platelet-derived growth factor AB (PDGF-AB), PDGF-BB, interleukin 6 (IL-6), IL-8, and IL-1β were measured by using commercial enzyme-linked immunosorbent assay Quantikine immunoassay kits (R&D Systems, Minneapolis, Minn).
Tissue lysate array proteomics
Patients underwent percutaneous 16-gauge to 18-gauge tumor biopsy under ultrasound or CT guidance prior to treatment and again at 4 weeks. A single attempt at biopsy was allowed; if it was unsuccessful, then patients proceeded with treatment, but no further biopsies were attempted. Core-needle biopsies were frozen immediately in optimum cutting temperature medium and were stored in liquid nitrogen until use. Frozen sections (6 μm) for laser-capture microdissection on uncharged glass slides were dehydrated, fixed, and stained using hematoxylin, and from 20,000 to 30,000 cells were captured.24 Protein was extracted using approximately 15 μL Tissue Protein Extraction Reagent buffer (Pierce Biochemicals, Rockford, Ill) diluted 1:1 with sample buffer, as described previously.25 Captured cells or lysates were stored at −80°C until protein extraction and printing of the tissue lysate array (TLA) onto glass-backed nitrocellulose membranes (Schleicher & Scheuell Bioscience, Keene, NH) in a triplicate, 5-point, serial 1:1 dilution curve using an Affymetrix 417 arrayer (Santa Clara, Calif). Controls included lysates of epidermal growth factor-stimulated HeLa cells and Fas-ligand treated Jurkat cells.
Total protein loading was quantified using colloidal gold stain. Signals of interest were detected with a titer-optimized antibody against total or phospho (p)-specific protein followed by catalyzed signal amplification system (DakoCytomation Catalyzed Signal Amplification System; DAKO Cytomation, Carpinteria, Calif). Stained arrays were digitized and spot intensities quantified using ImageQuant version 5.2 (Molecular Dynamics, Sunnyvale, Calif). Signal intensities were normalized to total protein signal to account for variations in protein loading.
The trial was conducted as a single-arm, 2-stage, Phase II clinical trial using an optimal design that evaluated the primary objective of clinical outcome, which was defined as objective response or disease stabilization for ≥6 months. The trial was designed to rule out a response/stabilization probability of 20% (P0 = .20) in favor of activity of 40% (P1 = .40) with an α = .10 and β = .10. In the first stage, if from 0 to 3 of 17 patients attained the targeted outcome, then accrual would cease, and the conclusion would be made that imatinib is not sufficiently active. If ≥4 of 17 patients attained the targeted outcome, then accrual would continue to a total of 37 patients with ≥11 patients required to consider imatinib active in epithelial ovarian cancer.
Secondary objectives include an evaluation of the effects of imatinib on PDGFR and c-kit and their downstream signal transduction pathways in tumor tissue of epithelial ovarian cancer patients, to correlate signaling events with outcome and toxicity, and to evaluate the effect of imatinib on circulating pro-angiogenic cytokine production. The biologic endpoints were deemed exploratory and there was no fixed requirement for sample size. The proteomics data were analyzed as standardized values:
Each tumor lysate assay parameter was tested for a trend according to grade of toxicity using an exact Jonchkeere-Terpstra trend test.26 For these evaluations, no formal correction for multiple comparisons was performed because of the exploratory nature of the study, the large number of parameters evaluated, and their varying degrees of independence from one another. However, a P value <.01 may be considered sufficiently small to interpret the result as statistically significant, whereas .01 <P< .05 would suggest a trend. Relative changes from baseline for a set of 6 cytokine concentrations were compared between patients with or without effusion by using an exact Wilcoxon rank-sum test. Two-tailed P values for these comparisons are presented after an adjustment using the Hochberg procedure.27
Patients and Clinical Course
Twenty-three patients were enrolled (Table 1). Seven patients started imatinib at 400 mg twice daily but, because of grade 3 toxicity in 6 of 7 patients, either the dose was reduced to 600 mg daily, or the drug was discontinued before completion of the first month of therapy. The protocol was amended to reduce the starting dose to 600 mg daily, and accrual was reinitiated.
|Patient characteristic||No. of patients|
|Median age (range), y||53 (36–68)|
|No. of prior therapies received|
Table 2 outlines the clinical effects observed. The targeted outcome was objective response or disease stabilization that lasted ≥6 months. There were no objective responses, and 2 patients had disease stabilization that lasted 8 months and 14 months, respectively. A slowing in clinical progression was observed in patients who were on treatment for at least 4 months compared with their clinical behavior prior to the initiation of imatinib therapy.
|Outcomes||No. of patients|
|Withdrawal caused by drug intolerance||6|
|Progression within or at 8 wks||7|
|Time to progression ≥6 mo (range)||2 (8-14 mo)|
|Median no. of cycles (range)||2 (<1–14)|
Grade 3 toxicity at an imatinib dose of 400 mg twice daily included new onset ascites or pleural effusion, leukopenia, anemia requiring transfusion, and gastrointestinal symptoms (Table 3). Grade 3 events that were observed at the lower dose included dyspepsia, peripheral edema, hypoalbuminemia associated with fluid retention, hypotension, and fatigue. Fluid accumulation, gastrointestinal toxicity, fatigue, and anemia were the most frequently observed events and resulted in study withdrawal by 7 patients. Five of those patients received 400 mg twice daily, and 2 patients received 600 mg daily.
|Toxicity||Grade 1||Grade 2||Grade 3|
|Abdominal pain, possibly drug related||2||3||1|
Significant fluid retention was observed in 19 of 22 evaluable patients (86%). Six patients had ascites (2 exacerbation; 4 de novo). Cytology was malignant in 3 of 4 sampled effusions. Six patients developed pleural effusions, and others had peripheral or facial edema. When imatinib therapy was discontinued, fluid accumulation ceased within days and, if removed therapeutically, did not recur without reinstitution of imatinib. This also occurred with malignant effusions. Ascites and pleural effusions developed between 2 months and 6 months into therapy.
Analysis of ascites and plasma cytokine concentrations
Plasma sampling for proangiogenic cytokine measurement was planned prospectively, prior to reports that circulating VEGF varied in response to small-molecule kinase inhibitors.28–31 The absolute ranges of measured cytokine concentrations in plasma and ascites were variable (Table 4). The relation to drug exposure was studied by assessing the maximal change from baseline. After adjustment for multiple comparisons, the maximal relative change in IL-6 concentration was significant between patients who developed ascites and/or pleural effusion and those who did not (Table 5) (P = .02). Circulating cytokine concentration dropped rapidly upon drug cessation, consistent with the clinical picture of ascites resolution, in 1 patient who was sampled after imatinib discontinuation.
|Variable||Absolute concentration, pg/mL|
|Edema||Median % change from baseline (range)|
|None or grade 1||5.2 (−81–420)||−4 (−85–316)||42 (−100–1513)||−44 (−99–439)||85 (−31–756)|
|Grade ≥2||120 (−26–738)||206 (−31–2741)||347 (−58–3577)||243 (118–972)||142 (−56–1193)|
|Hochberg adjusted P||NS||NS||NS||.02||NS|
Proteomic profiling and pharmacoproteomics
TLA was used to assess multiple signaling endpoints in microdissected tumor and stroma samples. Results were obtained from up to 14 of the tumor samples and stroma from up to 10 of the samples (Fig. 1). Reasons for an incomplete translational cohort included: no matched biopsy (14 samples), inflammatory infiltration preventing clean microdissection (4 samples), and inadequate amount of stroma (13 samples). Samples that had negative results (no measurable signal) were censored from statistical analysis. The limited clinical outcome observed precluded analysis of the relation between biochemical parameters and response. Target inhibition of c-kit was observed in 3 of 6 evaluable patients using the TLA (from −53% to 259%) (Fig. 1). Biochemical inhibition in tumor was observed for p-extracellular signal-regulated kinase (from −80% to 301%). EGFR phosphorylation, which was included as a presumptive negative biochemical control, was inhibited minimally (from −6% to 651%). Table 6 reports the parameters that were associated most strongly with toxicity, for which the P value for the association was ≤.01. Recurring patterns were addressed, because this was a selected subgroup of the potentially hundreds of comparisons, and the sample sizes were small. Pretreatment p-c-kit and EGFR potentially were linked to nausea and vomiting (P ≤ .01), whereas posttreatment EGFR (P = .001) and PDGFR (P = .004) were linked significantly to fatigue in this study.
|Variable||Toxicity grade||Mean||No.||SEM||Exact P2|
We hypothesized that imatinib would alter the local tumor microenvironment through its inhibition of c-kit and PDGFR kinases, resulting in cytostasis of tumor and/or tumor reduction. Kit, abl, and PDGFR kinases have been documented in tumor stroma, and it has been demonstrated that they are present in ovarian cancer.9, 14, 15, 32, 33 These biochemical targets were not anticipated as dominant pathways in ovarian cancer. However, their effects on angiogenesis and stroma are well known, which led to our hypothesis that therapy against the tumor and its microenvironment may be an effective molecular intervention in ovarian cancer. This Phase II trial included acquisition of pretreatment and on-treatment tumor biopsies for measuring drug-mediated target regulation. Twenty-three patients with moderately pretreated ovarian cancer were treated, for whom 15 paired biopsies were used for biochemical analysis. The 600 mg daily dose level was reasonably tolerable, and we observed 2 patients who had disease stabilization for 8 months and 14 months, respectively. No correlation between changes in biochemical parameters and outcome were observed for those 2 patients. Target inhibition of c-kit was observed in 50% of evaluable patients, whereas 1 of 8 patients had a reduction of p-EGFR (−6%). Pharmacodynamic associations were observed between the extent of on-study biopsy p-c-kit and EGFR magnitude with gastrointestinal events and between on-study biopsy EGFR and PDGFR values and fatigue.
This trial was developed to examine the ability of imatinib to remit disease or to stabilize disease for at least 6 months. The limitation of prior treatment regimens was included because of concern regarding possible pre-existing bone marrow toxicity. The 2 patients who had prolonged disease stabilization had received 1 and 2 prior therapies, respectively. A recently reported Phase II imatinib trial (600 mg daily) was limited to patients who had received ≤4 prior regimens and who were platinum-resistant; any number of prior platinum/taxane-containing regimens counted as a single exposure in that trial. The patients had received a median of 4 prior therapies, and 3 of 12 patients who were evaluable for response had disease stabilization (from 6.4 months to ≥8 months23). Those 3 patients had received from 2 to 4 prior regimens, and 2 patients had experienced no disease-free intervals. Patients in that study were required to have c-kit- and/or PDGFR/abl-expressing tumors, as evidenced by original tumor immunohistochemistry. Despite the refinement in population, the overall outcome differed negligibly from our larger study group. Current studies suggest that rates of expression of these molecular markers are more consistent with our study population.34 The Gynecologic Oncology Group study was limited to ≤2 prior regimens: That study is in its second accrual phase.
Fluid-accumulation toxicities were problematic, causing 6 patients to withdraw from our study and 1 patient with stable disease at 5 months but with persistent and symptomatic, drug-induced ascites withdrew despite a dose reduction ultimately to 400 mg daily. Edema and effusions are known adverse events for patients who receive imatinib.35, 36 In addition to peripheral and facial edema, therapy has been associated with pleural and pericardial effusions.35, 37, 38 Coleman et al. reported edema in 14% of patients and ascites in 10% of patients (grade 3 in 2 of 3 patients). New-onset ascites in ovarian cancer generally is interpreted as disease progression.2 The development of ascites was observed in 25% of patients on this study. At least 1 patient sample was negative cytologically, and the pace of reversal of ascites upon drug holiday or discontinuation was rapid. These observations argue that ascites, even if malignant, is caused and/or augmented by imatinib therapy. Ovarian cancer reportedly produced and secreted VEGF, IL-6, and IL-8 into culture medium and ascites in preclinical models,39–43 and measurable concentrations have been reported in ascites ranging up to 10-fold the concentrations observed in serum.43 Our findings demonstrate secretion of VEGF and IL-6 in ascites in the μg/mL concentration range, with only PDGFs observed in greater concentration in serum than in ascites.
The molecular targets against which imatinib are focused, abl, PDGFR, and c-kit kinases, are active targets in both stroma and tumor cells. PDGFR and c-kit signaling is important in the development and maintenance of the local vascular milieu.6, 44, 45 Our observed induction of VEGF may be a feedback response to the down-regulation of c-kit and PDGFR-mediated signaling events involved in the dynamic balance of the activated stromal microenvironment.45, 46 Recent findings in patients with chronic myelogenous leukemia have indicated that imatinib induces a decrease in VEGF levels by inhibiting VEGF gene transcription through Sp1 and Sp3 transcription factors.30 This finding of decreased VEGF levels was observed in patients with gastrointestinal stromal tumors who received imatinib.47 Our observations support those suggested by others that there is a secondary increase in VEGF and other proangiogenic cytokines with imatinib and other small-molecule signal inhibitors. This may be one explanation for the lack of activity of single-agent imatinib in ovarian cancer, a cancer with a known strong angiogenic drive.
These findings of secondary activation of pathways for which inhibition was anticipated may be some of the reasons for the lack of efficacy of imatinib in ovarian cancer and other solid tumors for which activating target kinase mutations are not dominant. Other potential arguments for the lack of single-agent activity include the insufficiency of inhibition of a single receptor tyrosine kinase to impact downstream signaling cascades,48 the possibility that there is promiscuity of the agent in vivo in an unrecognized fashion that alters the signaling balance, and that the target(s) of imatinib are neither necessary nor sufficient to alter the natural history of ovarian cancer. The proteomic endpoints tested in this trial were selected in part to address these possibilities. The lack of inhibition of EGFR phosphorylation provides some suggestion of biochemical target selectivity. The cytosolic biochemical events studied are found downstream of multiple receptor tyrosine kinases and may be limited in their imatinib-mediated down-regulation, because EGFR and possibly other kinases remain active. A similar finding was observed in our Phase II study of single-agent gefitinib in ovarian cancer.49 Our findings validate in vitro studies from Matei et al., who demonstrated that AKT activation typically occurs after therapy with imatinib mesylate.34 The power of our findings, however, rests in the ability to demonstrate this impact on molecular signaling in a prospective fashion in patient tumor samples.
These clinical and signaling results support consideration of imatinib use in ovarian cancer in a combination approach to dysregulate proangiogenic signaling, although, as a single agent, imatinib should be considered ineffective for the treatment of epithelial ovarian cancer. This would test the hypothesis that the effect of imatinib may have been masked in part by a stimulatory feedback effect on angiogenesis. Combinatorial therapy with agents such as bevacizumab to adsorb VEGF ligand or sorafenib or sunitinib, inhibitors of VEGFR kinase, may block that feedback activation. Successful combination therapy such as this may redefine a role for imatinib in ovarian cancer.
We thank Drs. Liotta, Raffeld, and A. Murgo for their expertise and helpful discussions; Mr. Frueauf and Ms. Graves for their assistance; and the nurses and clinical associates of the Medical Oncology Branch.
- 26Nonparametric Statistical Methods. 2nd ed. New York, NY: John Wiley and Sons, Inc.; 1999., .