Purpose: To investigate the incidence of FGFR1 amplification in Chinese non–small cell lung cancer (NSCLC) and to preclinically test the hypothesis that the novel, potent, and selective fibroblast growth factor receptor (FGFR) small-molecule inhibitor AZD4547 will deliver potent antitumor activity in NSCLC FGFR1–amplified patient-derived tumor xenograft (PDTX) models.

Experimental Design: A range of assays was used to assess the translational relevance of FGFR1 amplification and AZD4547 treatment including in vitro lung cell line panel screening and pharmacodynamic (PD) analysis, FGFR1 FISH tissue microarray (TMA) analysis of Chinese NSCLC (n = 127), and, importantly, antitumor efficacy testing and PD analysis of lung PDTX models using AZD4547.

Results: The incidence of FGFR1 amplification within Chinese patient NSCLC tumors was 12.5% of squamous origin (6 of 48) and 7% of adenocarcinoma (5 of 76). AZD4547 displayed a highly selective profile across a lung cell line panel, potently inhibiting cell growth only in those lines harboring amplified FGFR1 (GI50 = 0.003–0.111 μmol/L). AZD4547 induced potent tumor stasis or regressive effects in four of five FGFR1-amplified squamous NSCLC PDTX models. Pharmacodynamic modulation was observed in vivo, and antitumor efficacy correlated well with FGFR1 FISH score and protein expression level.

Conclusions: This study provides novel epidemiologic data through identification of FGFR1 gene amplification in Chinese NSCLC specimens (particularly squamous) and, importantly, extends the clinical significance of this finding by using multiple FGFR1-amplified squamous lung cancer PDTX models to show tumor stasis or regression effects using a specific FGFR inhibitor (AZD4547). Thus, the translational science presented here provides a strong rationale for investigation of AZD4547 as a therapeutic option for patients with squamous NSCLC tumors harboring amplification of FGFR1. Clin Cancer Res; 18(24); 6658–67. ©2012 AACR.

This article is featured in Highlights of This Issue, p. 6569

Translational Relevance

Deregulated fibroblast growth factor receptor (FGFR) expression plays an important role in driving many cancers. In this report, we focus on squamous cell carcinoma, where there exists a desperate need for effective treatment therapies, and identify FGFR1 gene amplification in a cohort of Chinese patients with non–small cell lung cancer (NSCLC). Through the use of high-throughput cell line screening, we confirm the ability of the novel FGFR inhibitor, AZD4547, to modulate FGFR signaling and inhibit tumor cell proliferation only in lung tumor cell lines harboring amplified FGFR1. Importantly, we extend the clinical significance of these findings to show tumor stasis or regression effects in a panel of FGFR1-amplified squamous lung patient–derived tumor xenograft models (PDTX), but not in nonamplified models. Thus, our findings provide a strong rationale for the investigation of AZD4547 as a novel therapeutic option for patients with squamous NSCLCs harboring FGFR1 amplification.

Despite significant geographical variations, the overall global lung cancer incidence rate remains the highest amongst all cancer types (1). Disease mortality is of equal concern with average 5-year survival rates of about 15% in the United States (2). Non–small cell lung cancer (NSCLC) accounts for about 85% of lung cancer cases and includes the major histologic subtypes of adenocarcinoma, large cell carcinoma, and squamous cell carcinoma (SCC). Regardless of histology, treatment regimens have been dominated largely by platinum-based chemotherapeutics which typically prolong median progression-free survival for less than 1 year (3). However, recent advances in molecularly targeted treatment options, including EGFR inhibitors for EGFR mutation–positive tumors and anaplastic lymphoma kinase inhibitors for EML4-ALK fusion–positive tumors, have provided significant improvements in survival (4–6). Unfortunately, these genetic events are rare and limited almost exclusively to the adenocarcinomas of “never smoked” patients and thus, SCC, primarily a smokers' disease (7), still represents a disease of high unmet need both in terms of tractable genetic targets and more effective therapies.

Further molecular segmentation has been suggested recently through the discovery of genetic amplification of the fibroblast growth factor receptor 1 (FGFR1) gene in squamous cell lung clinical samples (8, 9). FGFR pathway signaling normally contributes to the physiologic processes of tissue repair, hematopoiesis, angiogenesis, and embryonic development; however, FGFs and FGFRs have emerged as driving oncogenes within a significant proportion of human tumors. Deregulation of FGFR signaling through multiple mechanisms (including gene amplification) has been documented within clinical samples of breast (10), multiple myeloma (11), noninvasive bladder (12), endometrial (13), gastric (14), and prostate cancers (15). Precedent for the success of targeting tumor-amplified tyrosine kinases as a therapeutic strategy has been shown clinically using agents such as trastuzumab [ERBB2 in breast cancer (ref. 16)] and cetuximab [EGFR in colorectal cancer (ref. 17)]. We have previously characterized the in vitro and in vivo activity of the novel, potent, and selective small-molecule FGFR inhibitor, AZD4547 (18).

Although lung cancer incidence and mortality rates continue to decline in the West, the same does not hold true for China and Japan, where the principle risk factors of smoking, air pollution, and an ageing population contribute to increasing incidence (19). Indeed, geographical differences also appear to exist in disease histology with a higher prevalence of adenocarcinoma in Japan than in most Western countries (20, 21). Tumor genetics also differ significantly with region, with a higher incidence of activating EGFR mutations in Chinese NSCLCs than in the West, and conversely a lower incidence of KRas mutation (20, 22).

Accordingly, in this study, we aimed to investigate the incidence of FGFR1 amplification in a Chinese NSCLC patient population. AZD4547 was initially used to screen a large panel of lung cancer cell lines and displayed potent antiproliferative activity in 2 cell lines containing amplified FGFR1. More importantly, we established translational significance by showing potent AZD4547 antitumor efficacy and pharmacodynamic activity in a panel of FGFR1-amplified squamous lung patient-derived tumor xenograft (PDTX) models.

AZD4547

Synthesis of N-[5-[2-(3,5-dimethoxyphenyl)ethyl]-2H-pyrazol-3-yl]-4-(3,5-diemthylpiperazin-1-yl)benzamide (AZD4547, AstraZeneca) has been described previously (18). The free base of AZD4547 (molecular weight = 463.6) was used in all preclinical studies. For in vitro studies, AZD4547 was prepared as a 10 mmol/L stock solution and diluted in the relevant assay media. For in vivo studies, AZD4547 was formulated in a 1% (v/v) solution of polyoxyethylenesorbitan monooleate (Tween 80) in deionized water. Animals were given AZD4547 or vehicle control once daily (qd) by oral gavage.

Cell culture

Cell lines were obtained from the American Type Culture Collection (ATCC), the German Resource Centre for Biological Material (DSMZ), or from internal collections as described previously (23). DMS114 and NCI-H1975 cell lines were routinely grown in RPMI-1640 supplemented with 10% (v/v) fetal calf serum (Biochrom AG) and 2 mmol/L l-glutamine (Invitrogen). Cell lines were genetically tested and authenticated using the StemElite ID System Kit (Promega) and were not cultured for more than 6 months before conducting the work described here.

In vitro antiproliferative cell panel screening

Cell line screening was conducted as described previously (23). Viability was determined after 96 hours by measuring cellular ATP content (CellTiter-Glo, Promega). Half-maximal inhibitory concentrations (GI50) were determined with the statistical data analysis software “R” with the package “IC50.”

FGFR1 FISH

The FGFR1 FISH probe was generated internally by directly labeling BAC (CTD-2288L6) DNA with Spectrum Red (Vysis, Cat # 30-803400). The CEP8 Spectrum Green probe (Vysis, Cat #32-132008) for the centromeric region of chromosome 8 was used as internal control. FISH assays were conducted on 4-μm dewaxed and dehydrated formaldehyde-fixed, paraffin-embedded (FFPE) sections. The SpotLight Tissue pretreatment Kit (Invitrogen, Cat #00-8401) was used for pretreatment (boiled in reagent 1 for ∼15 minutes then coated with reagent 2 for ∼10 minutes, minor time adjustments were made for individual samples). Sections and probes were co-denatured at 80°C for 5 minutes and then hybridized at 37°C for 48 hours. After a quick post-wash-off process (0.3%NP40/1xSSC at 75.5°C for 5 minutes, twice in 2×SSC at room temperature for 2 minutes), sections were finally mounted with 0.3 μg/mL 4′,6-diamidino-2-phenylindole (DAPI; Vector, Cat #H-1200), and stored at 4°C avoiding light for at least 30 minutes before scoring. FGFR1 gene and CEP8 signals were observed using a fluorescence microscope equipped with the appropriate filters allowing visualization of the intense red FGFR1 gene signals, the intense green chromosome 8 centromere signals, and the blue counterstained nuclei. Enumeration of the FGFR1 gene and chromosome 8 was conducted by microscopic examination of 50 tumor nuclei, which yielded a ratio of FGFR1 to CEP8. Tumors with FGFR1 to CEP8 ratio ≥2 or presence of ≥10% gene cluster were defined as amplified (AMP).

Analysis of FGFR1 mRNA expression

RNA samples from PDTX models were reverse transcribed to cDNA using High Capacity RNA-to-cDNA master mix (Applied Biosystems). FGFR1 mRNA expression was then determined by quantitative PCR (qPCR) assay using the ABI 7900HT platform. About 50 ng RNA input was used for each qPCR reaction, and the gene expression results were normalized using glyceraldehyde 3-phosphate dehydrogenase (GAPDH) levels.

Protein expression analysis

Cell lines were treated with AZD4547 or dimethyl sulfoxide (DMSO) control for 2 hours at 37°C. Frozen tumor fragments were lysed in 1× cell lysis buffer (Cell Signaling Technologies) containing phosphatase and protease inhibitors (Sigma) using a Fast Prep Homogenizer (MP Biomedicals). Immunoprecipitation studies were conducted by preclearing with protein G sepharose beads (Invitrogen) for 2 hours at 4°C. Cleared supernatants, containing beads and primary antibody, were then gently rocked overnight at 4°C, followed by centrifugation, washing in fresh lysis buffer and transfer directly to into SDS loading buffer. Western blotting was conducted using standard SDS-PAGE procedures and antibody incubation conducted overnight at 4°C. Antibodies were obtained from the following sources: FGFR1 (Epitomics), pFGFR (Y653/Y654) (CST), pFRS2 (Tyr436) (R&D Systems), pErk1/2 (Thr202/Tyr204), and GAPDH (Cell Signaling Technologies). Secondary antibodies were applied and immunoreactive proteins visualized using “SuperSignal West Dura” Chemiluminescence substrate according to the manufacturer's instructions (Pierce).

Immunohistochemistry

Antigen retrieval was conducted on FFPE tissues for 5 minutes in pH 6 retrieval buffer (S1699; Dako) followed by washing in running tap water for 5 minutes. Sections (3 μm) were rinsed in TBS with Tween (TBST) and incubated with endogenous peroxidase block on a LabVision autostainer for 10 minutes. Slides were washed twice in TBST and then incubated with primary antibody (FGFR1, Epitomics 2144-1, 1:50; pERK, Thr202/Tyr204, CST4376, 1:50; pS6 Ser240/244, CST2215, 1:100; CC3, CST9661, 1:100, respectively) for 60 minutes at room temperature and finally washed twice in TBST. The appropriate Envision or biotinylated goat anti-rabbit immunoglobulin secondary antibodies were used (DAKO E0432 or Vector Laboratories, Inc) and staining was detected using diaminobenzidine (K3468; Dako). For Ki67 immunohistochemical (IHC) analysis, ARK kit (DAKO K3954) was used to prevent cross-reaction with mouse tissue. Sections were incubated with biotinylated primary antibody (Ki67, DAKO M7240; 1: 100) for 15 minutes at room temperature and then washed twice in TBST. Following 15 minutes streptavidin–peroxidase treatment and washing in TBST, sections were developed and counterstained as described above.

For baseline expression or modulation detection, IHC scoring of FGFR1, phospho-S6, and phospho-Erk was calculated according to the following formula: scoring = 0 × [% cells with no staining (0)] + 1 × [% cells staining faint to barely visible (1+)] + 2 × [% cells staining weak to moderately (2+)] + 3 × [% cells staining strongly (3+)]. This method combines positive intensity and percentage tumor cell staining and was determined by 2 separate pathologists using microscopy. Quantification of Ki67- and CC3-positive signals was conducted using the Ariol system (Genetix).

In vivo efficacy studies using lung PDTX mouse models

A panel of PDTX mouse models was established as part of a previous study using patient NSCLC tissues acquired locally during resection (manuscript submitted to European Journal of Cardio-Thoracic Surgery). Prior written informed consent was obtained from all patients, and the study protocol was approved by the local hospital ethics committee. Eight- to 10-week-old female nude (nu/nu) mice (Vital River) were used for in vivo studies. All experiments using immunodeficient mice were carried out in accordance with the guidelines approved by Institutional Animal Care and Use Committees (IACUC). PDTX mouse models were established by directly implanting fresh surgical tumor tissue into immunodeficient mice. Briefly, PDTX tissue fragments (∼15 mm3) were implanted subcutaneously via Trocar needle into female nude mice. Tumor-bearing mice with a tumor size range of 100 to 200 mm3 were randomly divided into vehicle control or AZD4547 treatment groups (8 animals per group). Animals were treated orally by gavage needle. Subcutaneous tumors in nude mice and mice body weight were measured twice weekly. Tumor volumes were calculated by measuring 2 perpendicular diameters with calipers [formula: V = (length × width2)/2]. Percentage tumor growth inhibition [%TGI = 1 − [change of tumor volume in treatment group/change of tumor volume in control group) × 100] was used for the evaluation of antitumor efficacy. For tumor regression, in which the tumor volume after treatment was smaller than the tumor volume at the beginning, the following equation was used: T/T0 (%) = 100 × ΔT/T0, ΔT is change of tumor volume in treatment group. Statistical significance was evaluated using a one-tailed, two-sample t test. P < 0.05 was considered statistically significant. Tumor volume was calculated as described previously (24).

Potent in vitro AZD4547 antiproliferative activity correlates with FGFR1 gene amplification and pharmacodynamic activity in a lung cancer cell line panel screen

To characterize antiproliferative sensitivity to AZD4547, an in vitro screen was conducted using a comprehensive lung tumor cell line panel (23, 25). Cellular proliferation was assessed using a standard metabolism-based proliferation assay and GI50 values determined (Fig. 1A). Of 78 cell lines, only 2 (DMS114 and NCI-H1581) displayed a profound sensitivity to AZD4547, with GI50 values of 0.111 and 0.003 μmol/L, respectively. The DMS114 (small cell) and NCI-H1581 (large cell) lung cancer lines are well established and have been previously shown to harbor amplified FGFR1 (8). GI50 values for the remaining cell lines were all >4 μmol/L. An in vivo antitumor efficacy study of mice harboring H1581 xenografts treated daily with 12.5 mg/kg AZD4547 resulted in tumor regressions (Supplementary Fig. S1).

Figure 1.

Potent in vitro AZD4547 antiproliferative activity correlates with FGFR1 gene amplification and pharmacodynamic activity in a lung cancer cell line panel screen. A, GI50 values (y-axis) of AZD4547 across 78 lung cancer cell lines (x-axis). Only DMS114 and NCI-1581 lines are FGFR1-amplified (copy number ≥ 4). Cell line NCI-H1975 (used in B below) is marked with an asterisk. B, AZD4547 treatment inhibits FGFR signaling through p-FRS2 and p-Erk1/2 in DMS114, but not NCI-H1975. Cell lines were incubated for 2 hours with the stated concentrations of AZD4547 and then lysed and immunoblotted for the proteins indicated.

Figure 1.

Potent in vitro AZD4547 antiproliferative activity correlates with FGFR1 gene amplification and pharmacodynamic activity in a lung cancer cell line panel screen. A, GI50 values (y-axis) of AZD4547 across 78 lung cancer cell lines (x-axis). Only DMS114 and NCI-1581 lines are FGFR1-amplified (copy number ≥ 4). Cell line NCI-H1975 (used in B below) is marked with an asterisk. B, AZD4547 treatment inhibits FGFR signaling through p-FRS2 and p-Erk1/2 in DMS114, but not NCI-H1975. Cell lines were incubated for 2 hours with the stated concentrations of AZD4547 and then lysed and immunoblotted for the proteins indicated.

Close modal

To assess pharmacodynamic modulation of FGFR signaling using AZD4547 in vitro, one amplified (DMS114) and one nonamplified control cell line (NCI-H1975—EGFR L858R/T790M) were chosen for further study. FISH staining and further in vitro antiproliferative analysis confirmed FGFR1 amplification and AZD4547 sensitivity in the DMS114 line, but not in NCI-H1975 (Supplementary Fig. S2 and Supplementary Table S1). Pharmacodynamic modulation of FGFR signaling within these lines was assessed following a 2-hour incubation with AZD4547 and subsequent Western blotting with anti-sera raised against phospho-FRS2(Y436) and phospho-Erk1/2(T202/Y204), both well-established downstream markers of FGFR signaling. Clear, titratable inhibition of both p-FRS2 and p-ERK1/2 was observed in the DMS114 cell line, but not within NCI-H1975 (Fig. 1B). Thus, due to a proliferative dependence on FGFR1 gene amplification and signaling within the DMS114 cell line, strong pathway inhibition using AZD4547 led to potent antiproliferative activity which was not observed in the nonaddicted NCI-H1975 cell line.

FGFR1 gene amplification is a frequent occurrence in Chinese NSCLC patient samples

To establish the incidence of FGFR1 amplification in Chinese NSCLC patient samples, 127 tumor fragments were obtained locally at surgery and FFPE sections processed for FISH analysis using a specific FGFR1 gene probe. A total of 11 of 127 NSCLC samples (8.7%) were confirmed as FGFR1 gene–amplified (defined as an FGFR1/CEP8 gene probe ratio of ≥2 or cluster signals in ≥10% of tumor cells; Table 1). Of the 11 FGFR1-amplified tumor samples, 6 of 48 were of squamous origin (12.5%) and 5 of 76 were identified as adenocarcinoma (7%).

Table 1.

Incidence of FGFR1 gene amplification in Chinese NSCLC patient samples

NSCLC tumor histology
All (n = 127)SCC (n = 48)AC (n = 76)AC/SCC (n = 3)
FGFR1 amplification incidence [no. of samples (%)] 11 (8.7) 6 (12.5) 5 (7.0) 0 (0) 
NSCLC tumor histology
All (n = 127)SCC (n = 48)AC (n = 76)AC/SCC (n = 3)
FGFR1 amplification incidence [no. of samples (%)] 11 (8.7) 6 (12.5) 5 (7.0) 0 (0) 

NOTE: One hundred and twenty-seven patient NSCLC tumor fragments were obtained locally at surgery and FFPE sections processed for FISH analysis using a specific FGFR1 gene probe. FGFR1 FISH amplification incidence is shown here within SCC and adenocarcinoma (AC). Three samples were defined as mixed morphology (AC/SCC). FGFR1 amplification (FISH 6) was defined as an FGFR1/CEP8 gene probe ratio of ≥2 or cluster signals in ≥10% of tumor cells.

Statistical analysis of patient clinicopathologic parameters showed clear associations of FGFR1 amplification with male gender and “ever smoker” status (P = 0.0046 and 0.0022, respectively, χ2 likelihood ratio test; Table 2). To rule out any potential confounding effect of gender, the male-only samples were further analyzed for smoking status and FGFR1 amplification correlation. Despite not reaching statistical significance, the data did suggest a correlative trend between smoking status and FGFR1 amplification (P = 0.0573, χ2 likelihood ratio test). No correlations were found between FGFR1 amplification and patient age, tumor grade, or tumor stage. At the time of analysis, survival data was too immature to be able to determine the effect of FGFR1 amplification on patient survival.

Table 2.

Correlation analyses of FGFR1 amplification and clinicopathologic parameters

FGFR1 status
ParameterValueNonamplified n (%)Amplified n (%)Pa
Gender Male 81 (88) 11 (12) 0.0046 
 Female 38 (100) 0 (0)  
Stage 1,2 79 (92) 7 (2) 0.5650 
 3,4 68 (88) 7 (12)  
1,2 33 (92) 3 (8) 0.8624 
 3,4 68 (91) 7 (9)  
67 (94) 4 (6) 0.1061 
 1,2 34 (85) 6 (15)  
Histology SCC 42 (88) 6 (12) 0.2655 
 AC 71 (93) 5 (7)  
“Ever smoker” Yes 53 (84) 10 (16) 0.0022 
 No 63 (98) 1 (2)  
FGFR1 status
ParameterValueNonamplified n (%)Amplified n (%)Pa
Gender Male 81 (88) 11 (12) 0.0046 
 Female 38 (100) 0 (0)  
Stage 1,2 79 (92) 7 (2) 0.5650 
 3,4 68 (88) 7 (12)  
1,2 33 (92) 3 (8) 0.8624 
 3,4 68 (91) 7 (9)  
67 (94) 4 (6) 0.1061 
 1,2 34 (85) 6 (15)  
Histology SCC 42 (88) 6 (12) 0.2655 
 AC 71 (93) 5 (7)  
“Ever smoker” Yes 53 (84) 10 (16) 0.0022 
 No 63 (98) 1 (2)  

NOTE: Patient clinicopathologic information was obtained for 127 Chinese NSCLC tumor specimens and analyzed for correlations with FGFR1 amplification status.

aSignificance was determined using the χ2 likelihood ratio test (P ≤ 0.05).

AZD4547 treatment results in pharmacodynamic modulation of FGFR signaling and potent antitumor activity in FGFR1-amplified squamous PDTX models

To test preclinically the hypothesis that use of a selective FGFR pharmacologic inhibitor could offer therapeutic benefit to patients with NSCLCs harboring FGFR1-amplified tumors, we established a panel of primary SCC xenograft models derived directly from patient tumor material. Tissue sections from these models were characterized using hematoxylin and eosin (H&E) staining and FGFR1 FISH and IHC analysis. Of these, 5 models (L123, LC038, LC026, L121, and L133) were identified as FGFR1-amplified (FISH score 6) and (with the exception of model L133) high-level FGFR1 protein expression confirmed by IHC (denoted as “+++”; Fig. 2A). Model LC036 served as a negative control with normal FGFR1 gene copy number and undetectable protein expression. All models tested wild-type for EGFR, K-Ras and negative for the EML4-ALK gene fusion (data not shown).

Figure 2.

AZD4547 treatment results in pharmacodynamic modulation of FGFR signaling and tumor regression in an FGFR1-amplified patient-derived squamous cell lung cancer model. A, characterization of 5 FGFR1-amplified squamous cell lung cancer models (L123, LC038, LC026, L121, and L133) and one nonamplified squamous cell lung cancer model (LC036) using morphologic, IHC, and FISH staining techniques. FGFR1 IHC score and FGFR1 amplification status are embedded within the images (“AMP” denotes gene amplified). For FISH images, FGFR1 gene probe signals appear red, CEP8 signals are green, and DAPI-counterstained nuclei appear blue. Scale bars represent 50 μm for H&E/IHC images and 30 μm for FISH images. All images within each row are to the same scale. B, AZD4547 was administered by oral gavage once (qd) daily to nu/nu mice bearing established PDTX model (L121) xenograft fragments at the doses indicated. Tumor volumes are plotted against time. C, immunoprecipitation (IP)/Western blot (WB) analysis (IP: total FGFR1, WB: p-FGFR) showing modulation of p-FGFR1 within L121 tumor lysates following a single dose of 12.5 mg/kg AZD4547. Three animals were used for each timepoint and individual lanes represent tumor lysate from one animal. D, IHC analysis of L121 tumor sections showing modulation of FGFR-downstream signaling (p-Erk and p-S6) and proliferative (Ki67) and apoptotic markers (CC3). p-Erk and p-S6 analysis was conducted at 8 hours following a single dose of 12.5 mg/kg AZD4547, whereas CC3 and Ki67 data were collected at 24 and 48 hours post-dose, respectively. Quantified H scores were determined for each group (n = 3 animals/group) and statistical significance established for each marker using a one-tailed t test (P ≤ 0.05; data not shown). The scale bar represents 50 μm and all images are to the same scale.

Figure 2.

AZD4547 treatment results in pharmacodynamic modulation of FGFR signaling and tumor regression in an FGFR1-amplified patient-derived squamous cell lung cancer model. A, characterization of 5 FGFR1-amplified squamous cell lung cancer models (L123, LC038, LC026, L121, and L133) and one nonamplified squamous cell lung cancer model (LC036) using morphologic, IHC, and FISH staining techniques. FGFR1 IHC score and FGFR1 amplification status are embedded within the images (“AMP” denotes gene amplified). For FISH images, FGFR1 gene probe signals appear red, CEP8 signals are green, and DAPI-counterstained nuclei appear blue. Scale bars represent 50 μm for H&E/IHC images and 30 μm for FISH images. All images within each row are to the same scale. B, AZD4547 was administered by oral gavage once (qd) daily to nu/nu mice bearing established PDTX model (L121) xenograft fragments at the doses indicated. Tumor volumes are plotted against time. C, immunoprecipitation (IP)/Western blot (WB) analysis (IP: total FGFR1, WB: p-FGFR) showing modulation of p-FGFR1 within L121 tumor lysates following a single dose of 12.5 mg/kg AZD4547. Three animals were used for each timepoint and individual lanes represent tumor lysate from one animal. D, IHC analysis of L121 tumor sections showing modulation of FGFR-downstream signaling (p-Erk and p-S6) and proliferative (Ki67) and apoptotic markers (CC3). p-Erk and p-S6 analysis was conducted at 8 hours following a single dose of 12.5 mg/kg AZD4547, whereas CC3 and Ki67 data were collected at 24 and 48 hours post-dose, respectively. Quantified H scores were determined for each group (n = 3 animals/group) and statistical significance established for each marker using a one-tailed t test (P ≤ 0.05; data not shown). The scale bar represents 50 μm and all images are to the same scale.

Close modal

AZD4547 antitumor efficacy was tested first using the FGFR1-amplified model, L121. Tumor-bearing mice were randomly grouped and dosed orally, once daily with vehicle, 3.125, 6.25, or 12.5 mg/kg AZD4547 for a period of 15 days. Potent tumor regressions were observed in the 6.25 and 12.5 mg/kg treatment groups (TGI = 159% and 190%, respectively; P ≤ 0.0001), whereas tumor stasis followed by slow regrowth was observed in the 3.125 mg/kg treatment group (TGI = 84%, P = 0.0019; Fig. 2B). To confirm direct in vivo target engagement and modulation of tumor FGFR signaling by AZD4547, a separate study was used using L121 tumor–bearing mice, treated with a single dose of 12.5 mg/kg AZD4547. Tumors were excised at 0, 1, 2, 4, 8, 24, and 48 hours post-dose and processed for both IHC and Western blot analysis. Clear dynamic modulation of phospho-FGFR1 was detected by immunoprecipitation/Western blot analysis of L121 tumor lysates following AZD4547 dosing, with maximal inhibition occurring around 4 hours and signal recovery between 24 and 48 hours (Fig. 2C). Similarly, markers of downstream FGFR signaling, phospho-S6 and phospho-Erk, also showed significant modulation at 8 hours post-dose as determined by quantified IHC staining of L121 tumor sections (55% and 73% reductions in p-S6 and p-Erk staining, respectively, both P ≤ 0.05; Fig. 2D). Phenotypic markers were also assessed in this study and included Ki67 and cleaved caspase 3 (CC3)—proliferative and apoptotic markers, respectively. At 24 hours post-dose, significant induction of CC3 was observed (48% of cells IHC-positive, P ≤ 0.05), whereas at 48 hours, Ki67 staining was dramatically reduced in response to AZD4547 dosing (72% of cells IHC-negative, P ≤ 0.05). Dynamic modulation of these markers over multiple timepoints is shown in Supplementary Table S2.

Next, we assessed the antitumor efficacy of AZD4547 in 4 additional FGFR1-amplified PDTX models (L123, LC038, LC026, and L133) and used model LC036 to test the null hypothesis. Following implantation and establishment of tumor fragments in nude mice, randomized groups were dosed orally with AZD4547, once daily at either 12.5 or 25 mg/kg for 2 to 3 weeks. In 2 of the 4 FGFR1-amplified models, potent tumor regressions were observed (TGI = 199% and 134% in models L123 and LC026, respectively; both P ≤ 0.0001), whereas model LC038 displayed sustained tumor stasis (TGI = 94%; P ≤ 0.0001; Fig. 3). Interestingly, despite being characterized as FGFR1-amplified using FISH analysis, model L133 displayed relatively poor antitumor efficacy in response to AZD4547 treatment (TGI = 55% at 25 mg/kg AZD4547, P = 0.01), likely a consequence of very low FGFR1 protein expression. Importantly, minimal (nonsignificant) efficacy was observed in the negative control model, LC036 (TGI = 18%; P = 0.25) confirming the previously documented in vivo FGFR selectivity of AZD4547 (18). In all of the antitumor efficacy studies presented here, treatment with AZD4547 was well tolerated and did not result in significant body weight loss.

Figure 3.

AZD4547 displays potent antitumor efficacy in 3 of 4 additional FGFR1-amplified patient-derived squamous lung cancer models but is inactive in an FGFR1 nonamplified control model (LC036). AZD4547 was administered by oral gavage once (qd) daily to nu/nu mice bearing established patient-derived human lung tumor xenograft fragments at the doses indicated. Tumor volumes are plotted against time.

Figure 3.

AZD4547 displays potent antitumor efficacy in 3 of 4 additional FGFR1-amplified patient-derived squamous lung cancer models but is inactive in an FGFR1 nonamplified control model (LC036). AZD4547 was administered by oral gavage once (qd) daily to nu/nu mice bearing established patient-derived human lung tumor xenograft fragments at the doses indicated. Tumor volumes are plotted against time.

Close modal

AZD4547 antitumor efficacy correlates well with FGFR1 FISH score and protein expression level

To further inform our understanding of the relationship between FGFR1 gene and protein level with AZD4547 antitumor efficacy, Western blotting of PDTX model lysates was conducted using an FGFR1-specific antisera and the data combined with FISH, IHC, and efficacy data (Fig. 4). Two further FGFR1 nonamplified models (LC011 and LC022, FISH score 4 and FISH score 5, respectively) were included in this analysis. Notably, AZD4547 treatment at either 12.5 or 25 mg/kg gave potent tumor stasis or regression effects in 4 of 5 of the FGFR1-amplified models (94%–199% TGI). The 3 remaining nonamplified models (LC036, LC011, and LC022) displayed inferior antitumor efficacy ranging from 18% to 69% TGI using 25 mg/kg AZD4547. This trend in AZD4547 antitumor response based on gene amplification was also observed when using FGFR1 gene copy number (GCN) analysis. Potent stasis and regression effects were only observed in those models with FGFR1 GCN>6.

Figure 4.

AZD4547 antitumor efficacy correlates well with FGFR1 gene and protein expression levels. Model summary table displaying FGFR1 FISH score, gene copy number (GCN), IHC score, protein expression by Western blotting, and antitumor efficacy in response to 2 to 3 weeks, once daily oral AZD4547 treatment (25 or 12 mg/kg). Western blot data were obtained using FGFR1 and GAPDH antisera on fresh tumor fragment lysates. P values were calculated using a one-tailed t test. *, AZD4547 dosed at 12.5 mg/kg/qd.

Figure 4.

AZD4547 antitumor efficacy correlates well with FGFR1 gene and protein expression levels. Model summary table displaying FGFR1 FISH score, gene copy number (GCN), IHC score, protein expression by Western blotting, and antitumor efficacy in response to 2 to 3 weeks, once daily oral AZD4547 treatment (25 or 12 mg/kg). Western blot data were obtained using FGFR1 and GAPDH antisera on fresh tumor fragment lysates. P values were calculated using a one-tailed t test. *, AZD4547 dosed at 12.5 mg/kg/qd.

Close modal

Similarly, an analysis of FGFR1 protein expression using both IHC and Western blotting confirmed that the models in which AZD4547 induced tumor stasis or regression all showed strong FGFR1 protein expression (intense bands on Western blot or IHC score 3+). Model L133 is of interest as despite being FGFR1 gene amplified (FISH score 6, GCN = 6, IHC score 0–1), it displayed relatively low FGFR1 protein expression and consequently inferior sensitivity to AZD4547 compared with the other amplified models (55% TGI at 25 mg/kg, P = 0.01). Subsequent analysis using qPCR revealed that model L133 FGFR1 messenger RNA (mRNA) levels were barely detectable in comparison to high-level mRNA expression in the other FGFR1 gene–amplified models (Supplementary Fig. S3).

Thus, AZD4547 induced potent tumor stasis or regression in 4 of the 5 FGFR1 gene–amplified squamous lung PDTX models tested here, whereas in nonamplified models, the best response observed was partial tumor growth inhibition. AZD4547 antitumor efficacy correlates well with FGFR1 gene amplification (FISH score 6) and consequent high-level protein expression (IHC 3+).

This report describes the incidence of FGFR1 amplification in Chinese NSCLCs, and importantly, through our observations of potent antitumor efficacy in multiple patient-derived lung cancer xenograft models harboring FGFR1 gene amplification, provides strong translational support for the clinical use of AZD4547 in patients with tumors bearing amplification of FGFR1.

In light of previous reports highlighting geographical differences in tumor molecular genetics (20, 22), the observed enrichment of FGFR1 amplification incidence within our Chinese squamous NSCLC samples (12.5%) is notable from a molecular epidemiology perspective and is broadly consistent with previously published data on Western NSCLC cohorts (8, 9). Our Chinese data confirmed the well-known correlation of smoking status (“ever-smokers”) with SCCs (P = 0.0026, Fisher exact test; Supplementary Table S3). No correlations were found between FGFR1 amplification and any other clinicopathologic parameters. Interestingly, although lower than squamous, we also observed FGFR1 amplification within our Chinese adenocarcinoma samples (7%). Because of the novelty of this finding, we confirmed the classification of these samples through independent review. This has potential importance due to the higher incidence of lung adenocarcinoma in Asian populations and also the high smoking prevalence amongst Asian males. Accordingly, these data warrant further investigation and confirmation in a larger patient cohort.

We note that previous studies have identified FGFR1-containing amplicons of differing sizes (133–208 kbp; refs. 8, 26), thus raising the possibility that genes other than FGFR1 could be driving or contributing to tumorigenesis. Published data using short hairpin RNA (shRNA) in an 8p11-amplified lung cell line did not support tumorigenic involvement of the WHSC1L1 and FLJ43582 genes while confirming the role of FGFR1 in cell viability (8). While in some cases (e.g., such as case L133 reported here) genes other than FGFR1 (e.g., WHSC1L1; ref. 27) may be the functional target of the 8p amplicon, the profound antitumor efficacy observed only in the FGFR1-amplified primary models elicited by the selective FGFR inhibitor, AZD4547, supports our conclusion that proliferation within these models (and the patient tumors from which they were derived) is driven primarily by FGFR1 amplification and signaling.

From a molecular segmentation perspective, our data highlight the significance of FGFR1 amplification as a novel therapeutic target. Characterization of our Chinese NSCLC samples revealed a higher incidence of EGFR mutation (36.0%) and lower incidence of KRas mutation (4.4%) than Western NSCLCs (17% and 22%, respectively; ref. 20), but more importantly, our data suggest that these genetic aberrations are almost entirely mutually exclusive with FGFR1 amplification (Supplementary Table S4). This finding extended to our lung PDTX models (data not shown) and has clear implications for clinical trial design and patient stratification.

Cell line–derived xenografts have proven use as models for pharmacologic studies and as efficacy models in cases of oncogene addiction. However, the use of cell line xenografts is limited by the lack of molecular and cellular heterogeneity. PDTX models offer the promise of better disease models through increased diversity of molecular lesions and the preservation of 3-dimensional tumor stromal cell components and interactions (28). Importantly, amplification of the FGFR1 gene was maintained between the PDTX models used herein and the original patient tumor samples from which they were derived (Zhang JC, manuscript submitted). Furthermore, all studies were conducted using models which had undergone fewer than 8 serial in vivo fragment passages, thus maintaining distinct regions of tumor heterogeneity and tumor/stromal architecture. Accordingly, we believe that the AZD4547 antitumor efficacy data presented here have strong translational significance.

Within FGFR1-amplified PDTX models, once daily, oral dosing of AZD4547 resulted in either tumor stasis or rapid tumor regressions. Antitumor activity correlated with inhibition of tumor phospho-FGFR1 levels and downstream signaling through p-Erk and p-S6. Furthermore, antitumor efficacy correlated with inhibition of tumor Ki67 staining and induction of apoptosis. Although not specifically examined here, previous in vitro mechanistic studies using AZD4547 led us to speculate that the tumor cell apoptosis observed in model L121 is likely a consequence of AZD4547-induced BIM expression (18). With regard to inhibition of tumor angiogenesis, previous experience with AZD4547 leads us to conclude that such effects are unlikely to result from AZD4547 treatment. Published preclinical data using AZD4547 at efficacious dose levels have failed to show any effects on tumor angiogenesis as measured by CD31 IHC staining (18). Furthermore, we did not observe in vivo inhibition of kinase insert domain receptor (KDR) signaling, nor AZD4547 efficacy in models which are known to be sensitive to anti-KDR agents. Importantly, in 7 of 8 models, we were able to show a good correlation between FGFR1 gene level, protein expression, and efficacy in response to AZD4547. Specifically, only those models with FISH score 6 (GCN ≥ 10) and corresponding high-level protein expression (IHC 3+) showed tumor stasis or regression in response to AZD4547. In contrast, model L133 did not respond despite a FISH score of 6, but notably bore relatively low protein expression by IHC (score 0–1) and Western blotting. Further analysis using qPCR confirmed near undetectable levels of FGFR1 mRNA expression, relative to the other FGFR1-amplified models. Tumor heterogeneity is an unlikely explanation for this finding, as multiple FISH analyses of this model detected FGFR1 amplification throughout the entirety of each tumor section (data not shown). Hence, other genes (e.g., WHSC1L1) may be driving tumorigenesis in this model. Overall, this PDTX model dataset supports the potential for the use of FISH to select FGFR1 gene–amplified patients likely to benefit from treatment with the FGFR inhibitor AZD4547 and highlights that assessment of FGFR1 protein expression by IHC could provide an additional approach.

To conclude, our data highlight the occurrence of FGFR1 gene amplification in a cohort of Chinese NSCLC and enrichment within SCC. Moreover, our data suggest a correlative trend between smoking status and FGFR1 amplification. Furthermore, we show the ability of the novel and selective FGFR inhibitor, AZD4547, to drive tumor regressions in patient-derived models of FGFR1-amplified squamous cell lung disease where there is currently high unmet medical need. Taken together, these data support further investigation of AZD4547 as a targeted therapeutic option for patients with NSCLCs with tumors harboring genetic amplification of FGFR1. AZD4547 is currently being evaluated in phase I/II clinical trials.

P.D. Smith has ownership interest (including patents) in AstraZeneca. R.K. Thomas reports the following potential sources of conflict of interest: consulting and lecture fees (Sanofi-Aventis, Merck KGaA, Bayer, Lilly, Roche, Boehringer Ingelheim, Johnson & Johnson, AstraZeneca, Atlas-Biolabs, Daiichi-Sankyo, Blackfield); research support (AstraZeneca, Merck, EOS). He is also a founder and shareholder of Blackfield, a company involved in cancer genome services and cancer genomics–based drug discovery. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J. Zhang, K. Liu, Q. Ji, E. Kilgour, P.D. Smith, A.N. Brooks, R.K. Thomas, P.R. Gavine

Development of methodology: J. Zhang, L. Zhang, X. Su, M. Li, Y. Xu, K. Liu, G. Zhu, L. Tang, P.R. Gavine

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Zhang, L. Zhang, X. Su, M. Li, F. Malchers, Y. Xu, K. Liu, Z. Dong, L. Tang, R.K. Thomas

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Zhang, L. Zhang, X. Su, F. Malchers, X. Yin, Z. Dong, Z. Qian, L. Tang, P. Zhan, E. Kilgour, P.D. Smith, R.K. Thomas, P.R. Gavine

Writing, review, and/or revision of the manuscript: J. Zhang, X. Su, X. Yin, Z. Dong, G. Zhu, E. Kilgour, P.D. Smith, A.N. Brooks, R.K. Thomas, P.R. Gavine

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Zhang, X. Su, S. Fan, X. Yin

Study supervision: P.D. Smith, A.N. Brooks, P.R. Gavine

IHC studies: Y. Xu

The majority of this research was conducted and funded by AstraZeneca. R.K. Thomas is supported by the State of Nordrhein-Westfalen through the PerMed initiative (grant z1104me004a/005-1111-0025), by the German Ministry of Science and Education (BMBF) as part of the NGFNplus program (grant 01GS08100), by the Deutsche Forschungsgemeinschaft (DFG) through SFB832 (TP6) and TH1386/3-1, by the EU-Framework Programme CURELUNG (HEALTH-F2-2010-258677), by a Stand Up To Cancer Innovative Research Grant, a Program of the Entertainment Industry Foundation (SU2C-AACR-IR60109), by the Behrens-Weise Foundation, and by an anonymous foundation.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
GLOBOCAN 2008 (IARC)
. 
Section of Cancer Information
[cited 2012 Sep]. Available from
: http://www.iarc.fr/en/research-groups/sec1/index.php.
2.
Horner
MJ
,
Ries
LAG
,
Krapcho
M
,
Neyman
N
,
Aminou
R
,
Howlader
N
, et al
SEER cancer statistics review, 1975-2006
.
Bethesda, MD
:
National Cancer Institute
; 
2009
.
Available from
: http://seer.cancer.gov/csr/1975_2006/,
based on November 2008 SEER data submission, posted to the SEER website
.
3.
Janku
F
,
Stewart
DJ
,
Kurzrock
R
. 
Targeted therapy in non-small-cell lung cancer — is it becoming a reality?
Nat Rev Clin Oncol
2010
;
7
:
401
14
.
4.
Shaw
AT
,
Yeap
BY
,
Solomon
BJ
,
Riely
GJ
,
Gainor
J
,
Engelman
JA
, et al
Effect of crizotinib on overall survival in patients with advanced non-small-cell lung cancer harbouring ALK gene rearrangement: a retrospective analysis
.
Lancet Oncol
2011
;
12
:
1004
12
.
5.
Maemondo
M
,
Inoue
A
,
Kobayashi
K
,
Sugawara
S
,
Oizumi
S
,
Isobe
H
, et al
Gefitinib or chemotherapy for non-small-cell lung cancer with mutated EGFR
.
N Engl J Med
2010
;
362
:
2380
8
.
6.
Mitsudomi
T
,
Morita
S
,
Yatabe
Y
,
Negoro
S
,
Okamoto
I
,
Tsurutani
J
, et al
Gefitinib versus cisplatin plus docetaxel in patients with non-small-cell lung cancer harbouring mutations of the epidermal growth factor receptor (WJTOG3405): an open label, randomised phase 3 trial
.
Lancet Oncol
2010
;
11
:
121
8
.
7.
Khuder
SA
. 
Effect of cigarette smoking on major histological types of lung cancer: a meta-analysis
.
Lung Cancer
2001
;
31
:
139
48
.
8.
Weiss
J
,
Sos
ML
,
Seidel
D
,
Peifer
M
,
Zander
T
,
Heuckmann
JM
, et al
Frequent and focal FGFR1 amplification associates with therapeutically tractable FGFR1 dependency in squamous cell lung cancer
.
Sci Transl Med
2010
;
2
:
62ra93
.
9.
Dutt
A
,
Ramos
AH
,
Hammerman
PS
,
Mermel
C
,
Cho
J
,
Sharifnia
T
, et al
Inhibitor-sensitive FGFR1 amplification in human NSCLC
.
PLoS ONE
2011
;
6
:
e20351
.
10.
Ray
ME
,
Yang
ZQ
,
Albertson
D
,
Kleer
CG
,
Washburn
JG
,
Macoska
JA
, et al
Genomic and expression analysis of the 8p11-12 amplicon in human breast cancer cell lines
.
Cancer Res
2004
;
64
:
40
7
.
11.
Keats
JJ
,
Reiman
T
,
Belch
AR
,
Pilarski
LM
. 
Ten years and counting: so what do we know about t(4;14)(p16;q32) multiple myeloma
.
Leuk Lymphoma
2006
;
47
:
2289
300
.
12.
Billerey
C
,
Chopin
D
,
Ubriot-Lorton
MH
,
Ricol
D
,
Gil Diez de Medina
S
,
Van
RB
, et al
Frequent FGFR3 mutations in papillary non-invasive bladder (pTa) tumors
.
Am J Pathol
2001
;
158
:
1955
9
.
13.
Pollock
PM
,
Gartside
MG
,
Dejeza
LC
,
Powell
MA
,
Mallon
MA
,
Davies
H
, et al
Frequent activating FGFR2 mutations in endometrial carcinomas parallel germline mutations associated with craniosynostosis and skeletal dysplasia syndromes
.
Oncogene
2007
;
26
:
7158
62
.
14.
Jang
JH
,
Shin
KH
,
Park
JG
. 
Mutations in fibroblast growth factor receptor 2 and fibroblast growth factor receptor 3 genes associated with human gastric and colorectal cancers
.
Cancer Res
2001
;
61
:
3541
3
.
15.
Sahadevan
K
,
Darby
S
,
Leung
HY
,
Mathers
ME
,
Robson
CN
,
Gnanapragasam
VJ
. 
Selective over-expression of fibroblast growth factor receptors 1 and 4 in clinical prostate cancer
.
J Pathol
2007
;
213
:
82
90
.
16.
Piccart-Gebhart
MJ
,
Procter
M
,
Leyland-Jones
B
,
Goldhirsch
A
,
Untch
M
,
Smith
I
, et al
Trastuzumab after adjuvant chemotherapy in HER2-positive breast cancer
.
N Engl J Med
2005
;
353
:
1659
72
.
17.
Bardelli
A
,
Siena
S
. 
Molecular mechanisms of resistance to cetuximab and panitumumab in colorectal cancer
.
J Clin Oncol
2010
;
28
:
1254
61
.
18.
Gavine
PR
,
Mooney
L
,
Kilgour
E
,
Thomas
AP
,
Al-Kadhimi
K
,
Beck
S
, et al
AZD4547: An orally bioavailable, potent and selective inhibitor of the Fibroblast Growth Factor Receptor tyrosine kinase family
.
Cancer Res
2012
;
72
:
2045
56
.
19.
Du
Y
,
Cha
Q
,
Chen
X
,
Chen
Y
,
Huang
L
,
Feng
Z
, et al
An epidemiological study of risk factors for lung cancer in Guangzhou, China
.
Lung Cancer
1996
;
14
Suppl
:
S9
37
.
20.
Kris
MG
,
Johnson
BE
,
Kwiatkowski
DJ
,
Iafrate
AJ
,
Wistuba
II
,
Aronson
SL
, et al
Identification of driver mutations in tumor specimens from 1000 patients with lung adenocarcinoma: The NCI's lung cancer mutation consortium (LCMC)
.
J Clin Oncol
29, 2011
(
suppl; abstr CRA7506
).
21.
Subramanian
J
,
Govindan
R
. 
Lung cancer in never smokers: a review
.
J Clin Oncol
2007
;
25
:
561
70
.
22.
Wu
YL
. 
Molecular events in non-small cell lung cancer in Chinese
.
Ann Oncol
2011
;
22
(
Suppl 9
):
37
.
23.
Sos
ML
,
Michel
K
,
Zander
T
,
Weiss
J
,
Frommolt
P
,
Peifer
M
, et al
Predicting drug susceptibility of non-small cell lung cancers based on genetic lesions
.
J Clin Invest
2009
;
119
:
1727
40
.
24.
Wedge
SR
,
Kendrew
J
,
Hennequin
LF
,
Valentine
PJ
,
Barry
ST
,
Brave
SR
, et al
AZD2171: a highly potent, orally bioavailable, vascular endothelial growth factor receptor-2 tyrosine kinase inhibitor for the treatment of cancer
.
Cancer Res
2005
;
65
:
4389
400
.
25.
Sos
ML
,
Fischer
S
,
Ullrich
R
,
Peifer
M
,
Heuckmann
JM
,
Koker
M
, et al
Identifying genotype-dependent efficacy of single and combined PI3K- and MAPK-pathway inhibition in cancer
.
Proc Natl Acad Sci U S A
2009
;
106
:
18351
6
.
26.
Beroukhim
R
,
Mermel
CH
,
Porter
D
,
Wei
G
,
Raychaudhuri
S
,
Donovan
J
, et al
The landscape of somatic copy-number alteration across human cancers
.
Nature
2010
;
463
:
899
905
27.
Tonon
G
,
Wong
KK
,
Maulik
G
,
Brennan
C
,
Feng
B
,
Zhang
Y
, et al
High-resolution genomic profiles of human lung cancer
.
Proc Natl Acad Sci U S A
2005
;
102
:
9625
30
.
28.
Tentler
JJ
,
Tan
AC
,
Weekes
CD
,
Jimeno
A
,
Leong
S
,
Pitts
TM
, et al
Patient-derived tumour xenografts as models for oncology drug development
.
Nat Rev Clin Oncol
2012
;
9
:
338
50
.