Targeting FGFR with Dovitinib (TKI258): Preclinical and Clinical Data in Breast Cancer

Dovitinib had antitumor activity in FGFR-amplified breast cancer models

Dovitinib decreased the concentrations of pFRS2 and pERK/MAPK in a dose -dependent manner in MDA-MB-134 (FGFR1 amplified) and SUM52 (FGFR2 amplified) cell lines (Fig. 1A). Because these data indicated that dovitinib effectively inhibited FGFR signaling in vitro, we further explored whether dovitinib preferentially inhibited proliferation of FGFR1- and FGFR2-amplified breast cancer cell lines. The IC₅₀ for cell growth inhibition was 190 and 180 nmol/L in MDA-MB-134 and SUM52, respectively (Fig. 1B and Supplementary Table S1). Conversely, IC₅₀ values were more than 2,000 nmol/L in the 11 breast cancer cell lines that had neither FGFR1 nor FGFR2 amplification. We then tested the antitumor activity of dovitinib in a FGFR1-amplified in vivo model (HBCx-2 breast cancer primary xenograft, with 8 FGFR1 gene copies; ref. 13). As shown in Fig. 1C, dovitinib prevented tumor growth at the 30 mg/kg dose and caused tumor regression at the 50 mg/kg dose. On day 28, T/C calculations showed that the mean tumor volume was 24.6% and 7.2% of the vehicle control for the 30 and 50 mg/kg groups, respectively (P < 0.001 by Mann–Whitney nonparametric test). In addition, we evaluated dovitinib in an FGFR2-amplified in vivo model (HBCx-3; 10 FGFR2 copies). Similar to the results described earlier, dovitinib caused tumor regression in HBCx-3 xenografts when administered at a dose of 40 mg/kg daily until day 35 (Fig. 1D). On day 28, T/C calculations showed a mean tumor volume for the dovitinib-treated group that was 23.4% of the vehicle control group (P < 0.001). Overall, these data suggested that dovitinib had antitumor activity in FGFR-amplified breast cancer models.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Dovitinib inhibits FGFR downstream signaling and cell proliferation in FGFR1- or FGFR2-amplified breast cancer cell lines and growth of FGFR1-amplified HBCx-2 xenografts. A, effects of dovitinib on FGFR signaling were assessed by Western blot detection of pFRS2 and pERK/MAPK in FGFR1-amplified MDA-MB-124 and FGFR2-amplified SUM52 cells. B, dovitinib inhibited cell proliferation in FGFR-amplified, but not FGFR-nonamplified, breast cancer cell lines. C and D, dovitinib showed tumor growth inhibition in (C) FGFR1-amplified HBCx-2 and (D) FGFR2-amplified HBCx-3 mouse xenograft models. Error bars indicate standard error of the mean. The T/C values and P values for the Mann–Whitney nonparametric tests compared with vehicle control are shown below each graph.

Patient characteristics

On the basis of the data observed in cell lines, together with consistent results from other groups (5, 14), we conducted a phase II trial to test the hypothesis that dovitinib has antitumor activity in patients with FGFR1-amplified breast cancer. In addition to enrolling patients with FGFR1-amplified tumors, patients with nonamplified tumors were enrolled to determine if dovitinib's ability to inhibit other targets (i.e., VEGFR and PDGFR) would mediate FGFR-independent antitumor activity. A retrospective analysis was also preplanned to evaluate the activity of dovitinib in patients with FGFR2- and FGF3-amplified breast cancers. Of the 243 HER2⁻ patients who underwent molecular prescreening (i.e., tumor sample analyzed for FGFR1 assessment), 116 did not continue to the protocol-specific screening phase due to nonassessable FGFR1 status (insufficient tumor or insufficient SISH signal, n = 24) or the assigned treatment arm being closed to additional enrollment (n = 92; Fig. 2). For example, enrollment to the FGFR1⁻ arms was completed early and closed to accrual, with screening continued in order to complete accrual of the FGFR1⁺ arms. Of the 127 patients who entered the protocol-specific screening phase, 46 were excluded from study entry due to not meeting inclusion criteria, most commonly due to hematologic and biochemical laboratory values outside the requested range. The remaining 81 patients were enrolled and received treatment, with a median time of 18.3 months from first treatment to analysis cutoff. These 81 patients were divided into 4 cohorts on the basis of FGFR1 amplification and HR status: FGFR1⁺/HR⁺ (n = 23), FGFR1⁺/HR⁻ (n = 2), FGFR1⁻/HR⁺ (n = 34), and FGFR1⁻/HR⁻ (n = 22; Table 1). Notably, there were very few cases of FGFR1-amplified HR⁻ disease (n = 2), confirming there is a very low incidence of FGFR1 amplification in patients with triple-negative breast cancer, and this arm was stopped before completion of enrollment.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Study design. Patients were prescreened to determine FGFR1 status, then stratified into 4 cohorts on the basis of HR status and FGFR1 amplification. The study was designed as a Simon 2-stage with 20 patients planned for enrollment in each cohort for each stage. No arm proceeded to stage 2.

Patient characteristics for the fully enrolled cohorts (FGFR1⁺/HR⁺, FGFR1⁻/HR⁺, and FGFR1⁻/HR⁻) are reported in Table 1. Most patients had late-stage breast cancer and were heavily pretreated. For example, 70% and 78% of the FGFR1⁺/HR⁺ patients presented with 3 organs or more involved and liver metastases, respectively. In addition, 95% of all patients were previously treated with chemotherapy in the metastatic setting and all but 10 HR⁺ patients previously received therapeutic endocrine therapy. Two patients who did not receive any endocrine therapy (adjuvant, neoadjuvant, or therapeutic) were already refractory to previous chemotherapy at study entry and were not considered as candidates for endocrine therapy as per the investigator's judgment.

Dovitinib effectively increases FGF23 plasma levels, indicative of FGFR1 inhibition

FGF23 has been identified as a target gene of FGF signaling in vitro (15). Furthermore, increases in FGF23 have previously been reported as a surrogate for FGFR1 inhibition in clinical trials of dovitinib in renal cell carcinoma and melanoma (11, 16, 17). In these trials, 50% to 100% increases in plasma FGF23 from baseline were detected at cycle 1, day 15 and were correlated with reduced pERK (17). In the data presented here, plasma FGF23 levels peaked ≈100% above baseline at cycle 1, day 8 and were sustained over time, suggesting that dovitinib effectively inhibited FGFR1 signaling (Supplementary Fig. S1). The level of increase was similar among the 3 groups of patients (FGFR1⁺/HR⁺, FGFR1⁻/HR⁺, and FGFR1⁻/HR⁻), suggesting that there were no major differences in pharmacodynamic activity between the 3 groups.

Dovitinib exhibits greater antitumor effects in FGFR1-amplified as compared with nonamplified breast cancers

No complete responses or confirmed PRs were observed (overall response rate, 0%) as per the adjudicated central review (Table 2). Three patients with FGFR1⁺/HR⁺ breast cancer achieved an initial objective response, although these responses were not confirmed at a subsequent assessment. One patient showed a complete disappearance of liver lesions with persistence of nontarget bone and lymph node lesions in the first postbaseline evaluation. This would have qualified as a PR, but was not confirmed in the second evaluation (day 109) because the investigator assessed disease progression in the bone and discontinued therapy, while the liver disease remained in complete response. The second patient showed a reduction of 30.8% in target lesions in the liver, retroperitoneum, and ovaries in the second postbaseline evaluation, but was not confirmed in the 3 subsequent evaluations that showed 25.8%, 27.1%, and 28.5% reductions from baseline. The patient experienced disease progression (observed on day 274; new peritoneal lesion). The third patient had a reduction of 40.3% in target lesions. However, this was not confirmed due to clinical disease progression at day 87 (increased markers cancer antigen 15.3 and carcinoembryonic antigen, and decline in performance status). The patient refused further assessment.

Two additional FGFR1⁺/HR⁺ evaluable patients had SD for 24 weeks or longer. Therefore, 5 of 20 FGFR1⁺/HR⁺ patients (25%) achieved either an unconfirmed PR or SD 24 weeks or longer. On the contrary, only 1 patient (3%) in the FGFR1⁻/HR⁺ cohort and 2 patients (13%) in the FGFR1⁻/HR⁻ cohort achieved long-term (≥24 weeks) SD and none had tumor shrinkage greater than 30%. None of the study arms proceeded to the second stage because the predefined criteria (at least 2 confirmed objective responses) were not met.

We further explored whether higher levels of FGFR1 amplification detected by qPCR are predictive for dovitinib sensitivity in patients with HR⁺ disease. Tissue for qPCR assessment was available for 42 of 51 HR⁺ patients with measurable disease. Thirty-eight of these patients had assessable disease, and in 35 of them the reported change in tumor size was in accordance with the overall tumor response; these are included in this analysis. Thirteen and 22 of these patients were FGFR1 amplified and nonamplified by SISH, respectively. Using qPCR, with a cutoff of 6 copies for FGFR1, a total of 7 patients were identified with FGFR1 amplification. Interestingly, qPCR data suggested that 6 cases presented FGFR1 amplification by SISH but not by qPCR. There was full concordance for the remaining 22 patients with no amplification of FGFR1 by SISH (all of them nonamplified by qPCR as well). In the 7 patients amplified by qPCR (and SISH), a 20.2% reduction in the mean tumor size was observed (range, 100% reduction to 28.4% increase, with 6 of the 7 patients showing tumor shrinkage and only 1 showing tumor increase) compared with a 14.2% increase in mean tumor size for the 6 patients with FGFR1 amplification detected by SISH but not qPCR (range, 11.6% reduction to 54.0% increase). As shown in Fig. 3A, dovitinib was more effective in FGFR1-highly amplified (≥6 gene copies) breast cancer as compared with tumors with lower levels of FGFR1 amplification in this population of HR⁺ patients. In the FGFR1-amplified breast cancer group (n = 7), dovitinib induced a mean 20.2% reduction in tumor size (range, 28.4% increase to 100% reduction). Conversely, tumors with less than 6 FGFR1 copies (n = 28) had a mean 8.3% increase (range, 54.2% increase to 28.2% reduction). Overall, these data suggested that dovitinib shows more potent antitumor activity in patients with high levels of FGFR1 amplification compared with those tumors without FGFR1 amplification by qPCR.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Dovitinib induces target lesion reductions in FGF-pathway–amplified tumors. A, box plot showing the best tumor change of the target lesions on the basis of FGFR1 copy number as assessed by qPCR in HR⁺ patients with measureable disease at baseline and evaluable for this plot (n = 35). The line and + symbol within each box represents the median and mean value for that group. Whiskers above and below each box indicate the maximum and minimum values in that group, respectively. B, waterfall plot of HR⁺ FGF-pathway–amplified (black bars; FGFR1, FGFR2, or FGF3 amplification by qPCR) or FGF-pathway–nonamplified (gray bars) patients evaluable for this plot (n = 38). The asterisk (*) denotes patients whose overall lesion response is progressive disease, but the target lesion response is PR or SD. The dagger (†) denotes the FGFR1⁻/FGF3⁺ patient with 3.4 copies of FGFR1 by qPCR, the double dagger (‡) denotes the FGFR1⁻/FGFR2⁺ patients, and the section mark (§) denotes the FGFR1⁺/FGFR3⁺ patients.

FGF3 and FGFR2 amplification to complement FGFR1 amplification in selecting individuals most likely to respond to dovitinib

Because FGF3 and FGFR2 amplifications have also been reported in breast cancers (6, 18, 19), we further evaluated by qPCR whether such amplifications could define additional subsets of sensitive patients. Four patients had FGF3 amplification as measured by qPCR, with tumor reductions of 100%, 30.8%, 23.0%, and 7.5%, respectively. Interestingly, 3 of these 4 patients also had a high level of FGFR1 amplification (≥6 copies by qPCR) and the fourth presented with FGFR1-gene gain (3.4 copies of FGFR1 by qPCR and SISH negative).

FGFR2 amplification was also assessed by qPCR and detected (≥4 copies) in 2 of the HR⁺ patients (both FGFR1 nonamplified by SISH and qPCR). These 2 patients with FGFR2⁺/FGFR1⁻/HR⁺ breast cancer had tumor reductions of 28.2% and 18.5%. Two additional patients with HR⁻ disease had FGFR2 amplification by qPCR, but discontinued due to adverse events (AEs) before a postbaseline assessment.

We further analyzed the impact of amplification in FGFR1 (≥6 copies), FGFR2 (≥4 copies), or FGF3 (≥6 copies) by qPCR in the above-described group of patients with assessable and HR⁺ disease (n = 38). Ten of these 38 HR⁺ patients were defined as having FGF-pathway–amplified breast cancer (FGFR1 and/or FGFR2 and/or FGF3 amplification). Interestingly, the mean reduction in target lesions was 21.1% from baseline (range, 28.4% increase to 100% reduction) in these 10 patients. Conversely, target lesions in the remaining 28 patients who did not present with FGF-pathway–amplified breast cancer had a mean 12.0% increase from baseline (range, 54.2% increase to 15.4% reduction). Best tumor response according to the presence of gene amplification on FGF-pathway is reported in Fig. 3B. Interestingly, this analysis showed that FGF pathway amplification–negative patients were unlikely to respond to dovitinib, as 16 out of 28 patients (57.1%) without FGF-pathway amplification presented either a new lesion or tumor size increase as best response, compared with only 1 out of 10 (10.0%) FGF-pathway–amplified patients. Similar results were obtained when amplification of FGFR1 or FGFR2 was used to define FGF-pathway amplification as only 1 patient presented with FGF3 amplification but not FGFR1 or FGFR2 amplification.

Safety

All patients eventually discontinued the study, with disease progression as the most common reason reported for discontinuation (n = 47, 58.0%). AEs regardless of relationship to dovitinib were reported as the primary reason for discontinuation in 22 patients (27.2%). AEs leading to discontinuation were most commonly grade 3, the most common being asthenia/fatigue (n = 7), gastrointestinal disorders (n = 5), and investigations of laboratory abnormalities (n = 6, mainly liver function test abnormalities).

All patients (100.0%) experienced at least 1 AE regardless of relationship to study drug, most commonly vomiting (77.8%), diarrhea (76.5%), asthenia (67.9%), and nausea (67.9%; Table 3). Sixty-three patients (77.8%) experienced a grade 3/4 AE. The most common grade 3/4 AEs regardless of study-drug relationship were asthenia (23.5%), ALT increase (11.1%), diarrhea (8.6%), and vomiting, fatigue, and AST increase (7.4% each). No notable differences across the groups were observed (data not shown).

Patients were also monitored for clinical laboratory abnormalities. Overall, new or worsened grade 3 and 4 alkaline phosphatase was observed in 21.1% and 1.3% of patients, respectively, and was more common in the FGFR1⁺/HR⁺ group (43.5%). New or worsened grade 3 or 4 AST increase was observed in 16.7% of the patients (all grade 3), and new or worsened grade 3 and 4 ALT increase was also observed in 15.4% and 1.3% of patients, respectively. New or worsening grade 3 and 4 total bilirubin was observed in only 2.6% and 1.3% of patients, respectively.

The most frequently observed grade 3 or 4 hematologic clinical laboratory abnormality was lymphopenia (16.9%, all grade 3). The majority of new or worsened abnormalities in absolute neutrophils, hemoglobin, white blood cell, and platelet counts were grades 1 and 2, with few patients exhibiting a shift to grade 4. No major difference in hematologic clinical laboratory abnormalities was observed between the treatment groups.

Eight patients died on the study or during the 28-day follow-up period. In 3 cases, the fatal event was assessed by the investigator as being related to the study drug, and in the remaining 5 cases, the death was deemed as secondary to progression of underlying disease. A brief description of the 3 deaths not related to progressive disease follows. Patient #1 was a 62-year-old female with history of hypertension, diabetes mellitus, and heavily pretreated (anastrozole, tamoxifen, fulvestrant, paclitaxel, bevacizumab, and capecitabine) metastatic breast cancer (widespread metastatic liver and bone disease). After 4 weeks of therapy, the patient experienced grade 3 AST and grade 4 ALT increases, grade 4 alkaline phosphatase increase, and grade 4 total bilirubin increase. Dovitinib treatment was discontinued due to these events, and the patient died 25 days after the last dose of therapy due to acute liver toxicity (cholestatic liver damage). Patient #2, a 44-year-old female who was heavily pretreated (doxorubicin, cyclophosphamide, docetaxel, cisplatin, vinorelbine, tamoxifen, and anastrozole), had breast cancer with multiple mediastinal, pulmonary, and bone metastases at baseline. She received 8 weeks of therapy until discontinuation due to disease progression (new pleural mass and subpleural nodes). The patient started paclitaxel 12 days after study discontinuation and died of pleural hemorrhage and acute respiratory failure 15 days later (27 days after the last dose of dovitinib). Patient #3, a 62-year-old female with metastatic breast cancer (single measurable liver lesion at baseline), had previously been treated with doxorubicin, cyclophosphamide, docetaxel, bevacizumab, and capecitabine. The patient received 25 doses of dovitinib and died at day 51, 4 days after the last dose. The death was secondary to listeria infection and sepsis following treatment-emergent grade 3 neutropenia.