KDR Amplification Is Associated with VEGF-Induced Activation of the mTOR and Invasion Pathways but does not Predict Clinical Benefit to the VEGFR TKI Vandetanib

Abstract

Purpose

VEGF pathway inhibitors have been investigated as therapeutic agents in the treatment of non–small cell lung cancer (NSCLC) because of its central role in angiogenesis. These agents have improved survival in patients with advanced NSCLC, but the effects have been modest. Although VEGFR2/KDR is typically localized to the vasculature, amplification of KDR has reported to occur in 9% to 30% of the DNA from different lung cancers. We investigated the signaling pathways activated downstream of KDR and whether KDR amplification is associated with benefit in patients with NSCLC treated with the VEGFR inhibitor vandetanib.

Methods

NSCLC cell lines with or without KDR amplification were studied for the effects of VEGFR tyrosine kinase inhibitors (TKI) on cell viability and migration. Archival tumor samples collected from patients with platinum-refractory NSCLC in the phase III ZODIAC study of vandetanib plus docetaxel or placebo plus docetaxel (N = 294) were screened for KDR amplification by FISH.

Results

KDR amplification was associated with VEGF-induced activation of mTOR, p38, and invasiveness in NSCLC cell lines. However, VEGFR TKIs did not inhibit proliferation of NSCLC cell lines with KDR amplification. VEGFR inhibition decreased cell motility as well as expression of HIF1α in KDR-amplified NSCLC cells. In the ZODIAC study, KDR amplification was observed in 15% of patients and was not associated with improved progression-free survival, overall survival, or objective response rate for the vandetanib arm.

Conclusions

Preclinical studies suggest KDR activates invasion but not survival pathways in KDR-amplified NSCLC models. Patients with NSCLC whose tumor had KDR amplification were not associated with clinical benefit for vandetanib in combination with docetaxel.

Introduction

Non–small cell lung cancer (NSCLC) is the leading cause of cancer-related deaths worldwide (1), with a 5-year survival rate of only 15% for all stages combined (2). Conventional chemotherapeutic regimens have demonstrated limited efficacy. Therefore, targeted therapies designed to inhibit the VEGF pathway have been extensively evaluated. VEGF pathway inhibitors including bevacizumab and the multitargeted receptor tyrosine kinase inhibitors (TKI) vandetanib, sunitinib, and sorafenib prolong progression-free survival (PFS; refs. 3–5) and bevacizumab prolongs overall survival (OS). In the phase III ZODIAC (NCT00312377) study, the addition of vandetanib to docetaxel resulted in a statistically significant improvement in PFS (HR = 0.79, P < 0.001), but not OS in patients with NSCLC (6). Collectively, benefits from VEGFR-targeted agents have been modest in patients with NSCLC. Thus, predictive markers for identifying which patients are likely to benefit are critically needed to increase the efficacy of the agents in a subpopulation of these patients.

The progressive growth of cancers is dependent on an adequate vascular supply, and the search for tumor-derived factors that promote tumor angiogenesis lead to the discovery of VEGF (7). VEGF activates angiogenic programs in endothelial cells through binding with its receptors VEGFR-1 and VEGFR-2 or kinase insert domain receptor (KDR; refs. 8–11). Ligand binding results in VEGFR-2 dimerization and signal transduction through activation of PI3K, Akt, and MAPK pathways.

Although expression of VEGFR-2 (also known as KDR) was initially thought to primarily occur in endothelial cells, it has been detected in malignant cells, including lung cancer cells (12–15). The overexpression of KDR/VEGFR-2 on NSCLC cells assessed by IHC is associated with a poor clinical outcome (14, 15). Activation of KDR through DNA has been detected in NSCLC specimens at a relatively high frequency (9%–32%; refs. 16, 17). Recently, we have shown that NSCLC cell lines with KDR copy number gains (CNG) were associated with in vitro resistance to platinum chemotherapy, and KDR CNG was associated with shortened survival in patients treated with platinum-based adjuvant therapy but not in untreated patients (16). Gains in this region have been reported in other tumor types as well. Gene amplification at chromosome 4q12, which harbors PDGFRA, KIT, and KDR, is a common alteration in glioblastoma (18). In addition, functionally active VEGFR2 is present on ovarian cancer cells and VEGF-targeted therapies have antitumor activity in preclinical models likely through effects on both the tumor cells and the tumor-associated endothelial cells (19).

The presence of KDR CNG in cell lines and tumors from patients with NSCLC provides evidence that KDR may promote a more aggressive phenotype in NSCLC cell lines and be associated with shorter OS in early-stage patients with NSCLC treated with adjuvant therapy. Therefore, the signaling pathways activated by KDR in NSCLC were studied to test whether KDR may be a predictive marker of therapeutic benefit for VEGFR TKIs. NSCLC cell lines with and without KDR amplification and tumor specimens from patients participating in a randomized, double-blinded, multicenter, placebo-controlled phase III study (ZODIAC; NCT00312377) were available for testing the efficacy of the dual VEGFR/EGFR inhibitor vandetanib plus docetaxel versus docetaxel alone (6). We report that although KDR amplification is associated with VEGF-driven activation of mTOR, p38, and other invasion pathways, it does not predict clinical benefit to the VEGFR TKI vandetanib.

Materials and Methods

Cell lines and reagents

All NSCLC cell lines were maintained in 10% RPMI media under sterile conditions. Cediranib (AZD2171) and vandetanib (ZD6474) were obtained from AstraZeneca. Nentedanib (BIBF1120) was obtained from Boehringer Ingelheim. Imatinib, sunitinib, axitinib, and sorafenib were purchased from Selleck Chemicals. Bevacizumab was obtained from the institutional pharmacy.

Detection of HIF1α

NSCLC cell lines were serum starved for 24 hours and then pretreated with or without 1 µmol/L sunitinib or imatinib for 1 hour prior to VEGF stimulation (50 ng/mL; R&D Systems). Protein lysates were collected after 24 hours. HIF1α ELISA (R&D Systems) was performed according to the manufacturer's instructions.

Proliferation assay

Cellular proliferation was assayed using the CellTiter-Glo Luminescent Cell Viability kit (Promega) following the manufacturer's protocol. In brief, NSCLC cells were plated into 384-well plates with 1,000 cells per well. Sixteen hours after plating, the cells were treated in triplicates with sorafenib or cediranib at seven different concentrations between 1 and 10 µmol/L for 72 hours followed by 10-minute incubation with CellTiter Glo substrate. Luminescent intensity was measured using Fluostar Optima (BMG Labtech, Inc). The IC₅₀ and combination index (CI) was calculated on the basis of Chou and Talalay equation using Calcusyn software (Version 2.1, Biosoft) as previously described (20, 21).

Migration assay

A total of 700 µL of serum-free RPMI containing VEGF-B (50 ng/mL), AZD2171, sunitinib (1 µmol/L), or imatinib (1 µmol/L) was added to the lower compartment of 24-well Transwell migration inserts (8.0-µm pore size; Fisher Scientific). NSCLC cells (5 × 10⁴) in 200 µL were added to the upper chambers and incubated for 24 hours. Cells in the upper compartment were removed. Cells that migrated to the underside of the membrane were stained and counted under low magnification (40×). Migration assays using the p38 inhibitor SB203580 (1 µmol/L Biovision) or the c-Met inhibitor EMD1214063 (1 µmol/L, Merck) were conducted in a similar manner.

Reverse-phase protein array and Western blotting

NSCLC cells were treated with 4 or 40 µg/mL of bevacizumab or 1 or 10 µmol/L cediranib or vandetanib for 24 hours, and protein lysates were collected. Reverse-phase protein array (RPPA) slides were printed from lysates, and RPPA studies were performed as described previously described (22). For each cell line, a serial dilution of five concentrations was printed with 10% of the samples replicated for quality control. Immunostaining was performed with an automated autostainer (BioGenex). Each array was incubated with primary antibody. Signal was detected using a catalyzed signal amplification system (DakoCytomation California, Inc.). MicroVigene software (VigeneTech) and an R package developed at MD Anderson Cancer Center (Houston, TX; ref. 23) were used to assess spot intensity. Protein levels were quantified by the SuperCurve method (http://bioinformatics.mdanderson.org/OOMPA) as previously described (24). Data were log-transformed (base 2) and median control normalized across all proteins within a sample. All statistical analyses were performed using R packages (version 2.10.0; ref. 25). For Western blotting, cells were washed in PBS and lysed in protein lysis buffer [1% NP-40, 20 mmol/L Tris–HCl (pH 8), 137 mmol/L NaCl, 10% glycerol, 2 mmol/L EDTA, and protease inhibitor cocktail tablets; Roche]. Protein (40 µg) was used for Western blotting along with antibodies against p-MET, c-Met, VEGFR-2, p-p38, p-p70S6K, p-AKT, and AKT (1:1,000; Cell Signaling Technology). Blots were probed with antibodies against vinculin or β-actin (Sigma-Aldrich) as a loading control.

Detection of KDR CNGs in clinical specimens

Histologic sections were reviewed by a lung cancer pathologist to assess presence, quantity, quality, and histologic types of tumor tissues. At least 500 malignant cells in the tissue specimens were required to be considered adequate for analysis for copy number. To enrich for malignant cell content (>70% malignant cells), tumor tissues were microdissected from formalin-fixed, paraffin-embedded (FFPE) tissue sections for subsequent DNA extraction using the Pico Pure DNA Extraction Kit (Arcturus). KDR gene copy number was detected using methods described previously (26). For qPCR, we utilized the ABI 7300 Real-Time PCR System (Applied Biosystems) along with an endogenous (Actin) control, and all experiments were performed in triplicate. Gene copy number of greater than 4 was considered as CNG. KDR gene copy number was also examined for using FFPE 4 mm histology sections using an assay utilizing a BAC probe containing KDR sequences labeled SpectrumRed and a commercial CEP 15 probe (Abbott Molecular) labeled in SpectrumGreen or similar method. All probes were validated in normal specimens for chromosomal mapping and appropriate specificity and sensitivity with a centromeric 4q probe as an internal control.

Statistical analysis

Statistical analysis of preclinical results was performed using Student t test (two-tailed). A P value of ≤0.05 was considered statistically significant. The clinical data for PFS and OS were analyzed using a Cox proportional hazard model adjusted for treatment group and baseline covariates. Objective response rate (ORR) was summarized by treatment group. PFS and ORR were defined according to RECIST criteria.

Results

NSCLC cells with KDR CNGs express elevated expression of VEGFR-2, p38, and mTOR pathway components

We compared protein levels of KDR/VEGFR-2 in NSCLC cell lines with or without KDR CNG by Western blot analysis. VEGFR-2 is expressed at relatively high levels in cell lines with KDR gene amplification compared with cell lines without KDR CNG (Fig. 1A). We next analyzed 49 NSCLC cell lines for expression of 188 proteins by RPPA to identify differential expression of key signal transduction molecules in cell lines with KDR CNG compared with those lacking amplification. The most differentially expressed protein was mTOR, a key regulatory protein in cell survival and metabolism (ref. 27; P = 0.008; Fig. 1B). In addition, KDR CNG was also associated with significantly increased expression of p38 (P = 0.025), a known activator of mTOR (28), and p70s6K (P = 0.048), a direct substrate of mTOR (Fig. 1B).

Figure 1.

VEGFR-2 signaling pathways in NSCLC cell lines with KDR CNGs. A, NSCLC cell lines with KDR CNG express elevated levels of VEGFR-2 compared with NSCLC cell lines without KDR CNG as determined by Western blot analysis. B, RPPA analysis shows increased expression of p38, mTOR, and p70s6K in NSCLC cell lines with KDR CNG compared to cell lines without KDR CNG. C, stimulation of H23 cells (KDR amplified) with VEGF (50 ng/mL) for 15 minutes results in activation of p38 and p70S6K. This effect is inhibited with VEGFR TKIs, sorafenib (SORA), vandetanib (VAN), or axitinib (AXIT). VEGF did not induce Akt activation in H23 cells. A431 cells treated with EGF serve as a positive control for activated Akt. D, treatment of A549 cells (KDR normal) with VEGF (50 ng/mL for 15 minutes) resulted in activation of Akt but not p38.

VEGFR TKIs inhibit VEGF-induced activation of p38 MAPK in NSCLC cells with KDR CNGs

Ligand-induced activation of KDR/VEGFR-2 on endothelial cells results in phosphorylation of p38 MAPK, a key signaling component in VEGF-mediated endothelial cell migration (29, 30). We stimulated H23 NSCLC cells (KDR CNG positive) with 50 ng/mL VEGF for 15 minutes alone or in the presence of increasing concentrations of the VEGFR TKIs, sorafenib, vandetanib, or axitinib; and protein lysates were analyzed by Western blot analysis. VEGF stimulation resulted in activation of p38 MAPK and p70s6K, and levels of phospho-p38 were diminished with the addition of VEGFR TKIs in a dose-dependent manner (Fig. 1C). Similarly, VEGF induced phosphorylation of p38 in Calu-1 cells (KDR CNG positive) and this effect was inhibited with the addition of VEGFR TKIs (Supplementary Fig. S1A). VEGF treatment did not induce activation of Akt in H23 cells (Fig. 1C). A431 cells treated with EGF served as a positive control for p-Akt. In contrast, in A549 NSCLC cells (KDR normal), VEGF treatment did not result in p38 phosphorylation, but did induce elevation of p-AKT (Fig. 1D). Because VEGF can bind and activate both VEGFR-2 and VEGFR-1, we evaluated VEGFR-1 levels by ELISA and found that VEGFR-1 expression was absent or minimal in all three cell lines (Supplementary Fig. S1B). SK-N-AS neuroblastoma cells served as a positive control. These data suggest that signaling pathways are differentially activated by VEGF in NSCLC cell lines with or without KDR CNGs.

Effect of VEGFR TKIs on proliferation and migration of NSCLC cells with KDR CNG

To investigate the effect of VEGFR inhibitors on the proliferation of NSCLC cell lines with or without KDR CNG, we performed high-throughput cell proliferation screening to evaluate the antitumor cell activity of the VEGFR TKIs, sorafenib and cediranib, on 49 NSCLC cell lines with or without KDR CNG. There was no significant difference in IC₅₀ values between cell lines with KDR CNGs and KDR normal cell lines (Fig. 2A).

Figure 2.

Direct effects of VEGFR TKIs on NSCLC cells with or without KDR CNG. A, the effect of VEGFR TKIs (sorafenib and cediranib) on viability of NSCLC cell lines with KDR CNG (copy number ≥4) or without KDR CNG (copy number = 2) was evaluated. There was no significant difference in IC₅₀ values between cell lines with or without KDR CNG. B, VEGFR TKIs inhibit the migration of NSCLC cells with KDR CNGs. H23 (KDR amplified), Calu-1 (KDR amplified), and A549 (KDR normal) NSCLC cells were plated in Boyden chambers in media containing 1 µmol/L cediranib (a pan-VEGFR inhibitor), sunitinib (TKI targeting VEGFR, PDGFR, and KIT), or imatinib (TKI targeting KIT, PDGFR, but not VEGFR) for 24 hours. Number of migrating cells was counted under 100× magnification. Data are graphed as mean ± SE; *, P < 0.05. C, the VEGFR TKI nintedanib inhibits the migration of H23 (KDR amplified) but not A549 (KDR normal) cells.

We previously demonstrated that in NSCLC cell lines with KDR CNG, siRNA targeting of KDR resulted in diminished tumor cell motility. Therefore, we investigated the effects of VEGFR TKIs on tumor cell migration in cell lines with or without KDR CNG using a Boyden chamber assay. In the KDR-amplified cell line, H23, treatment with cediranib (TKI targeting VEGFR-1, 2, 3) or sunitinib (TKI targeting VEGFR-1, -2, PDGFR, and c-Kit) significantly decreased the number of migrating cells (P < 0.05; Fig. 2B). In contrast, imatinib (TKI targeting PDGFR and c-Kit, but not VEGFR) had no effect on H23 tumor cell migration. VEGF stimulation induced a modest but significant (P < 0.05) increase in migration, likely due to the fact that H23 cells secrete VEGF in an autocrine manner as determined by ELISA (data not shown). In A549 cells, which are KDR normal, neither of the VEGFR TKIs affected the migratory capacity of the cells (Fig. 2B). Similar findings were observed using additional KDR-amplified cell lines, Calu-1 (Fig. 2B) and H1993 (Supplementary Fig. S2A). Likewise, inhibition of VEGFR signaling with axitinib (TKI targeting VEGFR-1, -2, -3, PDGFR, and KIT), sorafenib (TKI targeting VEGFR-2, -3, KIT, and PDGFR; Supplementary Fig. S2B and S2C), or nintedanib (TKI targeting VEGFR, FGFR, and PDGFR) decreased the migration of H23 (KDR CNG) cells but not A549 (KDR normal) cells (Fig. 2C)

VEGFR TKIs modulate HIF1α levels in NSCLC cells with KDR CNG

Activation of tyrosine kinases including VEGFR is known to induce HIF1α expression in normoxic conditions (31, 32). We previously demonstrated that NSCLC clinical samples with KDR CNG are more highly vascularized and express higher levels of HIF1α than tumors without KDR CNG (26). In normoxic conditions, HIF1α expression is elevated in NSCLC cell lines with KDR CNG compared with KDR normal cell lines, and in vitro targeting of KDR/VEGFR2 using siRNA diminishes HIF1α expression in cells with KDR CNG (26). Therefore, we sought to determine whether VEGFR TKIs modulate HIF1α levels in cells with KDR CNG. H23 (KDR amplified) cells were treated with the VEGFR inhibitor sunitinib (1 µmol/L) for 24 hours. Protein lysates were collected, and HIF1α levels were evaluated by ELISA. In KDR-amplified H23 cells, sunitinib (TKI targeting VEGFR-1, -2, PDGFR, and c-Kit) significantly decreased HIF1α expression, whereas imatinib (inhibits PDGFR and c-Kit, but not VEGFR-2) did not (P < 0.05; Fig. 3A). Similar findings were observed with an additional KDR-amplified cell line, H2085 (data not shown). In contrast, sunitinib treatment had no effect on HIFα levels in A549 cells (KDR normal; Fig. 3A).

Figure 3.

VEGFR TKIs modulate HIF1α and p-Met levels in NSCLC cells with KDR CNGs. A, H23 (KDR amplified) and A549 (KDR normal) NSCLC cells were treated with 1 µmol/L sunitinib (TKI targeting VEGFR, PDGFR, and KIT) or imatinib (TKI targeting KIT, PDGFR, but not VEGFR) for 24 hours. HIF1α levels were evaluated by ELISA. Sunitinib significantly reduced HIF1α levels, but imatinib did not. Sunitinib did not affect HIF1α levels in A549 cells. Data are graphed as mean ± SE; *, P < 0.05. B, H23 and Calu-1 (both KDR amplified) cells were treated with complete media containing bevacizumab (BEV), cediranib (CED), or vandetanib (VAN) for 24 hours. Western blot analysis shows p-Met levels were reduced following VEGFR pathway inhibition. C, Met is not constitutively activated in A549 (KDR normal) cells. D, Calu-1 cells were serum starved for 24 hours and then treated with VEGF or HGF (50 ng/mL) for 15 minutes to evaluate direct cross-talk between these receptors. HGF, but not VEGF, induced c-Met activation. E, H23 cells were treated with VEGF (50 ng/mL) alone or in combination with the p38 inhibitor SB203580 (1 µmol/L) or the c-Met inhibitor, EMD1214063 (1 µmol/L), and cell migration was evaluated by Boyden chamber assay. Inhibition of p38 or c-Med blocked VEGF-induced tumor cell migration.

VEGFR TKIs decrease Met signaling in NSCLC cells with KDR CNG

Because HIF1α modulates the expression of numerous proteins known to facilitate processes including angiogenesis and tumor cell invasiveness (33, 34), we sought to determine whether VEGFR TKIs may decrease the expression of proteins regulated by HIF1α. RPPA analysis indicated that VEGF pathway inhibition was associated with modulation of c-Met activity (data not shown). Therefore, we treated H23 and Calu1 cells (both with KDR CNG) with the VEGF monoclonal antibody, bevacizumab (4 or 40 µg/mL), or the VEGFR TKIs cediranib or vandetanib (1 or 10 µmol/L) for 24 hours. Protein lysates were collected and analyzed by Western blotting. VEGF pathway inhibition resulted in decreased phospho-c-Met in both Calu1 and H23 cell lines (KDR amplified; Fig. 3B). In A549 cells (KDR normal), phospho-Met was undetectable, and c-Met levels were unchanged with VEGF pathway inhibitors (Fig. 3C). To evaluate whether the effects of VEGF pathway inhibitors on phospho-Met levels in KDR-amplified cells was due to direct cross-talk between KDR/VEGFR-2 and c-Met, we treated Calu-1 cells with 50 ng/mL of VEGF or hepatocyte growth factor (HGF) and evaluated c-Met activation by Western blotting. HGF, but not VEGF, induced c-Met phosphorylation (Fig. 3D).

Met is known to be HIF1α regulated and is a key contributor to the invasive phenotype of NSCLC (35). Therefore, we treated H23 cells with VEGF alone or in combination with the c-Met inhibitor EMD1214063 and evaluated the effect on tumor cell migration. Pharmacologic inhibition of Met repressed VEGF-induced tumor cell migration (Fig. 3E). Moreover, given our observation that VEGF induced phosphorylation of p38 in NSCLC cells with KDR CNG and that p38 has been shown to be critical for the VEGF-induced migration of endothelial cells (36, 37), we treated H23 cells with VEGF alone or in combination with the p38 inhibitor, SB203580, and found that inhibition of p38 blocked VEGF-induced tumor cell migration (Fig. 3E).

KDR CNG does not predict clinical benefit from vandetanib in patients with NSCLC

To address whether KDR CNGs can identify patients likely to derive benefit from VEGFR TKIs, we screened archival lung tumor samples collected from consenting patients in the placebo-controlled phase III ZODIAC (NCT00312377) study of vandetanib (100 mg daily) plus docetaxel versus docetaxel alone in previously treated patients with NSCLC. In this study, the addition of vandetanib to docetaxel resulted in a statistically significant improvement in PFS but not OS (6). Overall there was adequate tissue for KDR CNG analysis from 288 patients (140 in the vandetanib arm and 148 in the placebo arm) out of 1,391 patients total. KDR CNG were detected in 44 of 288 patients (15%; Table 1). The 288 patients with adequate tissue included 162 patients with adenocarcinoma and 83 patients with squamous cell carcinoma; KDR CNG were detected in 24 of 162 (15%) and 14 of 83 (17%), respectively. We next assessed whether KDR CNG were predictive of clinical outcome. In patients harboring KDR CNG, the addition of vandetanib to docetaxel was associated with a slightly poorer PFS compared with placebo plus docetaxel (HR, 1.28; 95% CI, 0.64–2.55), and in the KDR-negative group, the HR for PFS was 0.99 (95% CI, 0.75–1.31), indicating that KDR was not predictive of PFS benefit for vandetanib (Table 2; Fig. 4A and B). For OS, among patients positive for KDR CNG, the addition of vandetanib to docetaxel was associated with a slightly poorer OS compared with placebo plus docetaxel (HR, 1.16; 95% CI, 0.49–2.73), and in the KDR-negative group, the HR for OS was 0.81 (95% CI, 0.57–1.16; Table 2; Fig. 4C and D). The ORR was also not improved with vandetanib plus docetaxel versus placebo plus docetaxel treatment in the KDR-positive group (4.3%). Although adequate tumor tissue was not available from this clinical dataset for investigating the relationship between KDR CNGs and cooccurring mutations in genes such as p53 that may influence sensitivity to hypoxic conditions and antiangiogenic therapy, using The Cancer Genome Atlas (TCGA) NSCLC dataset (38), we observed a significantly greater prevalence of p53 mutations in patients with KDR CNG compared with KDR CNG–negative patients (P = 0.001; Supplementary Table S1).

Table 1.

KDR CNG status in clinical specimens from the phase III ZODIAC trial

KDR CNG status	Vandetanib + docetaxel (N = 694)	Placebo + docetaxel (N = 697)
Positive	23	21
Negative	117	127
No sample	546	535
Assay failed	8	14

Table 2.

PFS in KDR CNG–positive and negative patients in the phase III ZODIAC trial

KDR CNG status	Treatment	N	Number of events	HR 95% CI
Positive	Vandetanib + docetaxel	23	21	1.28 (0.64–2.55)
	Placebo + docetaxel	21	19
Negative	Vandetanib + docetaxel	117	100
	Placebo + docetaxel	127	118	0.99 (0.75–1.31)

Figure 4.

Kaplan–Meier plots of PFS in the vandetanib plus docetaxel (A) or placebo plus docetaxel (B) treatment groups by KDR CNG status. Kaplan–Meier plots of OS in the vandetanib plus docetaxel (C) or placebo plus docetaxel (D) treatment groups by KDR CNG status.

Discussion

Inhibitors of the VEGF pathway, such as bevacizumab, and multitargeting VEGFR TKIs, such as vandetanib, have in some cases improved the clinical outcome in patients with advanced NSCLC. The benefits among all NSCLC patients have thus far been modest. Currently, there are no validated markers for VEGF pathway inhibitors in routine clinical use. Although VEGFR is typically expressed on the tumor-associated blood vessels, we and others have detected KDR CNGs in a fraction of lung cancer clinical specimens and cell lines (16, 17). Given the correlation between amplification and activating mutations in genes such as EGFR and FGFR and sensitivity to agents targeting these receptors, the identification of KDR CNGs in NSCLC tumor cells raised the possibility that this molecular aberration might serve as a biomarker of response to VEGFR TKIs. Several important findings emerged from this study. Our preclinical data indicate that in KDR-amplified cells, VEGF drives activation of p38, mTOR, and invasion pathways, but not survival pathways. Although VEGFR TKIs inhibit VEGF-mediated invasiveness and reduce HIF1α levels in these cells, they do not cause significant growth inhibition. Our analysis of KDR CNGs in clinical samples collected during a randomized phase III study shows that KDR amplification status is not a marker of clinical benefit from the VEGFR targeting agent, vandetanib. Although the power of the analysis is limited by the modest number of KDR CNG–positive patients in each arm and its retrospective nature, taken together, the results provide evidence that KDR CNG is not predictive of clinical benefit from vandetanib and that the overall PFS benefits observed for vandetanib-treated patients in this study is unlikely to be so solely due to benefit in the KDR CNG–positive population.

We show that NSCLC cell lines with KDR CNGs have increased expression of VEGFR2 and that in these cells VEGF stimulation triggers phosphorylation of p38 MAPK. Moreover, VEGF-driven tumor cell migration in KDR-amplified cells is p38 dependent. p38 MAPK has been shown to function as both an oncogenic protein and a tumor suppressor. As part of the stress-activated pathway, p38 positively regulates p53, antagonizes cell proliferation, and promotes apoptosis (39, 40). In contrast, elevated levels of phosphorylated p38 have been observed in various cancer types (41–44), and p38 activation is linked to cancer cell migration, production of inflammatory and angiogenic cytokines, and increased activity of NF-κB and HIF1α (45–48). Although the molecular mechanisms determining how p38 can have diametrically opposing roles is poorly understood, it is likely dependent on tumor cell type, tumor stage, and cross-talk with other signaling pathways.

We previously reported that in patients with KDR amplification, tumors were more highly vascularized and expressed elevated levels of HIF1α, a key activator of angiogenic programs (16). The role of various receptor tyrosine kinases in modulating HIF1α levels under normoxic conditions has been demonstrated. Specifically, EGFR has been shown to increase HIF1α levels in a hypoxia-independent, cell type–specific manner in prostate, lung, and breast cancer (34, 49–51), and the presence of a hypoxia-driven VEGF/VEGFR-1 autocrine loop modulating HIF1α has been described previously (31, 32). Here, we report that VEGFR TKIs decrease HIF1α levels in NSCLC cells with KDR CNGs but not in KDR normal tumor cells. This is consistent with our previous observation that sunitinib and sorafenib reduced HIF1α levels in normoxia or hypoxia in neuroblastoma cell lines (32). Given the role of HIF1α in chemoresistance, invasiveness, and metastasis, the effect of VEGFR TKIs on HIF1α levels may have additional effects on drug sensitivity and disease progression in patients with KDR amplification. In NSCLC tumors with KDR CNG, VEGFR TKIs may impair angiogenesis by both directly acting on endothelial cells and by decreasing HIF1α-regulated proangiogenic factors produced by the tumor cells.

c-Met is a receptor tyrosine kinase that promotes malignant transformation and tumorigenesis and is regulated by HIF1α (52–54). In NSCLC, high expression of the c-Met ligand, HGF, is associated with aggressive disease and a poor prognosis (54). We observed that blockade of KDR/VEGFR-2 signaling in KDR-amplified cells results in decreased c-Met activation. Hypoxia is known to increase HGF, presumably through HIF1α (53). VEGFR pathway inhibitors could potentially reduce HGF production though decreasing HIF1α.

Identification of predictive biomarkers is needed to select subsets of patients that may benefit from VEGFR TKIs. In a recent report by Zhang and colleagues, the authors analyzed the expression of VEGFA, FLT1/VEGFR1, and KDR in three independent colon cancer datasets (55) and found that high expression of all three factors was associated with a poor prognosis. A detailed analysis revealed that the VEGFA, FLT1, and KDR gene signature was only significant in patients with proficient mismatch repair status, wild-type but not mutant KRAS status, or mutant but not wild-type p53 status. Furthermore, this three gene signature predicted tumor response to the VEGF-targeting antibody, bevacizumab. Taken together, it may be useful to assess the impact of KDR CNG in NSCLC in the context of KRAS or p53 status; however, this analysis is not possible in the dataset reported here. Our analysis of p53 and KDR CNG status in patients with NSCLC from the TCGA dataset revealed increased prevalence of p53 mutations among patients with KDR amplification compared with KDR CNG–negative patients. Given the impact of p53 mutation status on tumor cell survival in low oxygen conditions, we cannot rule out that the lack of responsiveness to vandetanib reported here is a result of p53 mutations (56).

We recently showed that EGFR gene copy number or activating EGFR mutations may identify patients who receive greater benefit from vandetanib in combination with docetaxel in NSCLC (57). This is likely due, at least in part, to the activity of vandetanib against EGFR. In the current analysis of clinical specimens from the ZODIAC study, KDR amplification status did not predict clinical benefit from the VEGFR TKI, vandetanib. Here, our analysis was limited to patients with advanced stage, metastatic NSCLC. Because KDR CNGs may drive invasive but not survival pathways, this marker may be a better predictor of benefit in other disease settings, such as early-stage NSCLC. In addition, we and others have observed that TKIs may induce compensatory pathways that may abrogate the effects of target inhibition (58–60); therefore, other agents targeting this pathway, such as bevacizumab, aflibercept, or the VEGFR2-targeted monoclonal antibody ramicirumab, may yield different results. Although the focus of this study was KDR amplification, alternative tumor biomarkers might correlate with clinical response to VEGFR-targeting agents. For example, recent studies have suggested that bevacizumab may add greater benefit when combined with erlotinib in EGFR-mutant NSCLC as compared with EGFR wild-type NSCLC (61, 62). Further studies will be needed to identify predictive biomarkers for VEGFR TKIs and other agents targeting the VEGF in NSCLC.

Translational Relevance.

VEGFR inhibitors, including the multitargeted receptor kinase inhibitors vandetanib, pazopanib, sorafenib, and sunitinib, have been shown to prolong progression-free survival in patients with non–small cell lung cancer (NSCLC). These benefits have been modest, and thus far, biomarkers identifying patients likely to benefit have not yet been identified. Although expression of VEGFR2 (KDR) was initially thought to primarily occur on endothelium, we and others have recently reported that a subset of NSCLC tumors harbor DNA amplification of KDR/VEGFR2 and express VEGFR2. Here, we investigated whether tumor cell KDR amplification is a biomarker associated with benefit from VEGFR inhibitors. Our data suggest that VEGF activates distinct signaling pathways in cells with KDR amplification, and promotes an invasive phenotype. However, our analyses of specimens from a large, recently completed phase III study of docetaxel with or without vandetanib indicates that KDR amplification status is not a marker of clinical benefit from the VEGFR inhibitor vandetanib.