Lung squamous cell carcinoma (SqCC) is a molecularly complex and genomically unstable disease. No targeted therapy is currently approved for lung SqCC, although potential oncogenic drivers of SqCC have been identified, including amplification of the fibroblast growth factor receptor 1 (FGFR1). Reports from a recently completed clinical trial indicate low response rates in patients treated with FGFR tyrosine kinase inhibitors, suggesting inadequacy of FGFR1 amplification as a biomarker of response, or the need for combination treatment. We aimed to develop accurate models of lung SqCC and determine improved targeted therapies for these tumors. We show that detection of FGFR1 mRNA by RNA in situ hybridization is a better predictor of response to FGFR inhibition than FGFR1 gene amplification using clinically relevant patient-derived xenograft (PDX) models of lung SqCC. FGFR1-overexpressing tumors were observed in all histologic subtypes of non–small cell lung cancers (NSCLC) as assessed on a tissue microarray, indicating a broader range of tumors that may respond to FGFR inhibitors. In FGFR1-overexpressing PDX tumors, we observed increased differentiation and reduced proliferation following FGFR inhibition. Combination therapy with cisplatin was able to increase tumor cell death, and dramatically prolonged animal survival compared to single-agent treatment. Our data suggest that FGFR tyrosine kinase inhibitors can benefit NSCLC patients with FGFR1-overexpressing tumors and provides a rationale for clinical trials combining cisplatin with FGFR inhibitors. Mol Cancer Ther; 16(8); 1610–22. ©2017 AACR.

Chemotherapy, radiotherapy, or surgery are the primary treatment strategies for non-small cell lung cancer (NSCLC), yet 5-year survival rates are still among the lowest when compared with other malignancies (1). Squamous cell carcinoma (SqCC), the second most common histopathologic subtype after adenocarcinoma (ADC), has a poor outcome with a 5-year survival rate of 14% (2). Driver mutations in lung ADC have been known for a number of years and have led to the successful development of targeted therapies for this subtype of cancer. Effective ADC treatment with EGFR inhibitors or anaplastic lymphoma kinase (ALK) inhibitors has relied on the specific stratification of patients based on the molecular characteristics of their tumors—mutations in EGFR predict patient response to EGFR inhibitors as do rearrangements in ALK for ALK inhibition (3, 4). In SqCC, although potential candidate driver mutations have been described, these have not yet translated to approved targeted therapies, likely due to an inability to accurately identify patients who may respond to these treatments.

Recent reports of somatic alterations in FGFRs in SqCC have generated great interest in the use of FGFR inhibitors in the clinic (5, 6). Although the FGFR1 8p12 locus is amplified in 20% of lung SqCC (7), FGFR1 carcinogenic point mutations and FGFR1 fusion genes are rare (8, 9). Aberrations in FGFR2–4 occur at low frequency, and only recently have driver mutations in FGFR3 (S249C) and FGFR3–TACC3 gene fusions been described in 7.4% of lung squamous cell carcinomas (10, 11). Results from phase I trials using the FGFR1-3 inhibitor BGJ398 showed a response rate of only 11% and a disease control rate of 50% in patients selected for FGFR1-amplified tumors (12). Mechanisms of primary resistance are still unclear and may be due to the inadequate selection of patients based on FGFR1 amplification, where levels of expression of FGFR1 and its ligands do not always correlate with gene amplification (6, 13, 14). FGFR1 may also not be the only driver of tumor growth in these carcinomas, requiring combination of FGFR inhibitors with other therapeutic approaches.

In this study, we aimed to identify new biomarkers that better predict response to FGFR-targeted therapy and determine which combination treatment may improve clinical outcomes for patients with FGFR1-altered SqCC. Use of preclinical models that reflect the complexity of patient tumors are necessary to evaluate novel therapies and to identify biomarkers that will predict drug efficacy. Cell lines are commonly used to test novel therapeutics, but do not often accurately model clinical samples, nor do they correlate with drug sensitivity in clinical trials (15, 16). Patient-derived xenografts (PDX) are generated from the direct engraftment of resected patient tumor samples into immunocompromised mice. These models have been successfully established for solid tumors, including breast, melanoma, prostate, and pancreatic cancers and have been shown to recapitulate the phenotype, molecular profile, and therapeutic response of the patient's tumors (17–19). PDXs therefore constitute a highly clinically relevant system to study human solid tumors, even more specifically in lung cancer where whole genome sequencing analyses have revealed the complex heterogeneity and high frequency of genetic alterations in lung SqCC (20–23).

We generated a bank of PDX models of lung cancer and showed that PDXs recapitulated the phenotype of the patient's tumor and conserved transcriptomic and genomic characteristics, including high mutagenic burdens and copy number alterations. SqCC PDXs were then used to evaluate response to BGJ398, an FGFR inhibitor (24), as single agent or in combination with targeted therapy and standard chemotherapy. Our results show that FGFR1 RNA expression is a better predictor of response to FGFR targeted therapy compared with FGFR1 gene amplification and that cisplatin potentiates the response to FGFR inhibitors in FGFR1-overexpressing tumors.

Human tumor collection and transplantation

Written informed consent was obtained from all patients by the Victorian Cancer BioBank prior to inclusion in the study, according to protocols approved by the Human Research Ethics Committee of the Walter and Eliza Hall Institute of Medical Research (WEHI; approval #10/04). Human lung tumors from surgical resections (ADC, SqCC, and other NSCLC) were either transplanted immediately or held intact for a maximum of 48 hours at 4°C in DMEM/F12 media (Gibco), supplemented with 1 mg/mL of penicillin and streptomycin (Invitrogen). Tumors were implanted subcutaneously into the flanks of NOD.SCIDprkdcIl2rγ−/− (NSG) mice. Human lung tumors from endobronchial ultrasound (EBUS) biopsy specimens (ADC, SqCC) were collected in PBS and spun at 335 RCF for 5 minutes at 4°C. Samples were resuspended in 50 to 150 μL of PBS, depending on the size of the biopsy specimen and an equal volume of Matrigel (BD Biosciences) was added. Biopsy specimens were then transplanted into NSG. Mice were put on antibiotics for 7 days postsurgery (Baytril, Bayer) before twice weekly monitoring for tumor development. NSG mice (8–12 weeks old males) were bred at the WEHI breeding facility. Mice were maintained in our animal facilities according to institutional guidelines. All animal experiments were approved by the WEHI Animal Ethics Committee (approval #2013.028). Mice were culled at ethical endpoint (tumor volume of 600 mm3 as calculated by width × length2/2) or if 1 year had passed with no tumor growth.

PDX tumor cell preparation

PDX tumor samples were minced then digested for 1 hour at 37 °C with 2 mg/mL collagenase (Worthington) in 0.2 % d-glucose (Sigma) in DPBS (Gibco). The cell suspension was strained through a 100-μm cell strainer and washed with 2% FCS-PBS, followed by red blood cell lysis to obtain a single cell suspension. Tumor cells were either processed immediately for in vitro assays or frozen in a 1:9 DMSO (Sigma Aldrich): FCS (Gibco) mix. Cells were stored in liquid nitrogen.

PDX tumor sphere in vitro assays

Freshly isolated single-cell tumor suspensions were resuspended in DMEM/F12 (Gibco) supplemented with 1 mg/mL penicillin/streptomycin (Invitrogen), B27 (Gibco), 4 μg/mL heparin (Sigma Aldrich), 100 ng/mL EGF (Sigma Aldrich), insulin–transferrin–selenium (Gibco), 0.4% BSA (Sigma Aldrich), and 10 ng/mL basic FGF2 (R&D Systems), hereafter referred to as tumor media. Serum-free conditions were used to promote tumor sphere formation. Cells were plated at a density of 20,000 cells per well in 96-well low attachment plates (Corning). Cells were left to recover for a minimum of 1 hour, and then drugs added at the following concentrations: BGJ398 1 μmol/L (Active Biochemicals) or cisplatin 5 μmol/L (Hospira). Cells were grown for 72 hours at 37°C in 5% CO2 and 5% O2 before cell viability was assessed by the MTS assay according to manufacturer's instructions (Promega).

PDX in vivo assays

PDXs at passage 4 were defrosted and washed with PBS before counting. 200,000 to 500,000 cells were then injected in a 50:50 PBS:matrigel mix into the flanks of NSG mice. Tumors were measured twice weekly and treatment started when tumors were between 70 and 120 mm3. Mice were assigned to treatment groups based on average tumor volume per group. Mice were culled either 48 hours after treatment start for short term signaling experiments or when tumors reached a volume of 600 mm3 for long-term survival studies. Tumors were collected and pieces were either fixed in 10% formalin for histologic analyses or snap frozen for protein analyses. Mice were treated with either 30 mg/kg BGJ398 (Active Biochemicals) by oral gavage 5 consecutive days a week for 5 weeks or vehicle (33% PEG300, 5% dextrose); alone or in combination with 4 mg/kg cisplatin by intraperitoneal injection once every 3 weeks (Hospira) or vehicle (PBS); or with 15 mg/kg PKI-587 (synthesized by WEHI chemists) intravenously once weekly for 5 weeks or vehicle (10% DMSO, 5% dextrose pH 3.5).

Immunostaining

Samples were fixed in formalin for 24 hours at room temperature then paraffin embedded. Slides were dewaxed using standard histology protocols. Antigen retrieval was performed using citrate buffer (10 mmol/L, pH 6) or high pH antigen retrieval solution (Vector). Sections were blocked in 10% goat serum, incubated with antibodies overnight at 4°C followed by HRP-conjugated secondary antibodies (Vector). Antibodies used were Ki67 (B56, BD Pharmingen), cleaved caspase-3 Asp175 (CC3, polyclonal, Cell Signaling Technology) TTF-1 (SPT24, Novocastra), or p63 (DAK-p63, Dako). For quantification of the Ki67 and CC3 immunostaining, treatment groups were first blinded. Ki67 positive or CC3 positive cells were counted in three different fields of view.

Fluorescence in situ hybridization

Formalin-fixed paraffin-embedded (FFPE) sections were baked at 60°C for 30 minutes then dewaxed. Slides were then placed in heat pretreatment solution (Invitrogen) for antigen retrieval (125°C, 17 psi for 2.5 minutes) and washed with water. Slides were digested with enzyme pretreatment reagent (Invitrogen) for 15 minutes at 3°C, washed and dehydrated with graded ethanol before air drying. FGFR1, CCND1, CEP11, and CEP8 probes (Abnova) were applied to the sections, denatured for 5 minutes at 80°C and allowed to hybridize overnight at 37°C under humid conditions in a Thermobrite Hybridiser (Vysis). Slides were rinsed in stringent buffer (Invitrogen), then placed in stringent buffer at 75°C for 5 minutes before washing in water. Slides were mounted in Vectashield with DAPI (Vector). Slides were imaged using an Olympus BX51 fluorescence microscope. Tumors were designated as FGFR1 or CCND1 amplified if ≥50% of cells had ≥5 FGFR1 or CCND1 signals or if the ratio of FGFR1 signal to CEP8 signal, or CCND1 signal to CEP11 signal was >2.

FGFR1 RNA in situ hybridization

RNA in situ hybridization (RNA ISH) was performed on FFPE sections using an RNAscope Assay with FGFR1-specific probes (Advanced Cell Diagnostics) according to the manufacturer's instructions. FGFR1 RNAhi samples were determined if the number of FGFR1 probe signals was greater than positive control probe signals.

Western blotting

Fresh tumors were snap-frozen and crushed in liquid nitrogen. Tissue was then lysed in KalbC buffer (1% Triton X-100, 50 nmol/L Tris, pH 7.5, 1 mmol/L EDTA, 150 mmol/L NaCl, 1 mmol/L NaV, 2 mmol/L NaF, Roche complete mini protease inhibitor cocktail, Roche PhosSTOP phosphatase inhibitor cocktail). Protein concentration was quantified using the Pierce BCA Protein Assay Kit (Thermo Scientific). Protein samples were then run on 10% Bis-Tris gels (Invitrogen) in 2-N-morpholino ethanesulfonic acid buffer (Invitrogen) and transferred onto PVDF membranes (Immobilon) in transfer buffer (20% methanol, 2.42 g/L Tris base, 11.26 g/L glycine). Membranes were blocked in 10% milk (Devondale) in 0.1% Tween20 PBS solution for 1 hour at room temperature. Primary antibodies were incubated overnight at 4°C in milk solution—phosphoAKT Ser473 (M89-61, BD Pharmingen), AKT (polyclonal, Cell Signaling Technology), β-actin (AC-15, Sigma), c-myc (D84C12, Cell Signaling Technology), phosphoERK Thr202/Tyr202 (polyclonal, Cell Signaling Technology), ERK (polyclonal, Cell Signaling Technology), or PTEN (138G6, Cell Signaling Technology). Secondary conjugated antibodies (Southern Biotech) were incubated for 1 hour at room temperature before developing (Amersham).

Molecular characterization of somatic alterations

Copy number profiling was performed using HumanOmniExpress-24 microarray (Illumina) and normalized using GenomeStudio at Australian Genome Research facility. LogR and B allele frequency values from GenomeStudio were used as input for OncoSNP (25) to perform copy number calling. RNA-seq libraries were prepared from total RNA using TruSeq v2 mRNA stranded kit (Illumina) and sequenced as 75 bp paired-end reads. Reads were aligned to human and mouse reference genomes using subjunc (26) and reads that had better mapping to the mouse genome were removed. Differential expression analysis was performed using limma-voom workflow (27) and gene ontology (GO) terms enrichment analysis was performed using goana function in BioConductor. Whole exome sequencing libraries were prepared using Nextera Rapid Capture Exome Kit (Illumina) with the standard capture region and sequenced as 80 bp paired-end reads. Reads were aligned to human and mouse reference genomes using BWA-mem (28) and mouse reads were removed using Xenomapper tool (29). Preliminary variant calling was performed using VarScan (30) with liberal settings. Variant filtering, analysis and visualizations were carried out using superFreq pipeline (https://github.com/ChristofferFlensburg/superFreq). RNA-seq data from the Cancer Genome Atlas (TCGA) samples were downloaded from (GEO, accession number GSE62944; ref. 31). For detailed description of protocols and analyses used, refer to supplementary methods. RNAseq, ExomeSeq and SNP array data have been deposited at the European Genome-phenome Archive (EGA, http://www.ebi.ac.uk/ega/), which is hosted at the EBI (European Bioinformatics Institute), under accession number EGA #EGAS00001002423. In the interest of reproducibility and to facilitate access to nonsensitive data for the rest of scientific community, we have deposited non-identifiable summary data, as well as analysis scripts, in the following repository: https://github.com/ophiothrix/PDX-genomics.

Statistical analysis

Data represent mean ± SEM. Student t tests were performed using GraphPad Prism software and applied to each experiment as described in the figure legends. A P value less than 0.05 was considered significant. To determine significance in Kaplan–Meier survival analyses, the Mantel–Cox test was used. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

NSCLC successfully engraft to generate PDX models

To develop a bank of clinically relevant models of NSCLC that reflect the complexity of patients' tumors, we generated PDX models. Tumors from surgically resected specimens of NSCLC (stage I and II; t0) were implanted subcutaneously in immunocompromised mice and the animals monitored for tumor growth. A total of 121 NSCLC were xenografted. Tumor take rate was highest for SqCC (42 samples, take rate 52.3%), followed by other NSCLC (including adenosquamous carcinoma and large cell carcinoma, 27 samples, 37%) then ADC (78 samples, take rate 15.4%) with a median latency for successful grafts to reach ethical endpoint of 155, 140, and 259 days, respectively (Supplementary Fig. S1A and Supplementary Table S1).

Tumor biopsies collected by EBUS-guided biopsies (ref. 32; stages II and IV) from metastatic lymph nodes displayed higher engraftment rates compared to surgery specimens in ADCs (15.4% vs. 44.4%; Supplementary Fig. S1B), whereas no significant differences were observed in engraftment rates between surgery and biopsy specimens in SqCC (52.3% vs. 33.3%; Supplementary Fig. S1C). PDXs from both surgery and biopsy specimens recapitulated the patient's tumor by histologic analysis (Supplementary Fig. S1D).

Given that FGFR1 amplification is mostly detected in SqCC, we focused our study on this subtype of lung cancer. Once the first grafted tumor (t1) reached ethical endpoint size, the graft was subsequently transplanted in mice to monitor the maintenance of the characteristics of the patient's tumor up to five passages (t5; Fig. 1A). Latency for SqCC tumor growth was reduced after the first passage from 112 to 71 days by the third passage, but was maintained thereafter over subsequent passages (Fig. 1B). Histologic analysis of the grafted tumors showed that the xenografts maintained the phenotype of the patient's tumor after at least four passages (Fig. 1C). Immunohistochemical analysis of clinical markers of lung cancer subtypes showed that SqCC xenografts expressed p63 but did not express the ADC marker TTF-1, and matched the corresponding patient's tumor (Fig. 1C). Overall, these data establish that our NSCLC PDXs encompass diverse histologic subtypes and disease stages and ultimately recapitulate the phenotypic characteristics of the patient's tumor. SqCC PDXs derived from surgical specimens engraft more readily than other subtypes of NSCLC and are amenable to expansion in mice.

Figure 1.

Human SqCC tumor samples can be passaged in mice and phenocopy the primary tumor. A, Schematic representing the passaging strategy and nomenclature for SqCC patient-derived xenografts. B, Kaplan–Meier analysis showing survival of mice transplanted with SqCC PDXs from successive passages. n = indicated number of patient samples, significance determined by Mantel–Cox test, **P < 0.01, ***P < 0.001. C, IHC staining for p63, a marker of SqCC, and TTF-1, a marker of ADC, for a representative SqCC patient tumor (926) and corresponding PDXs. Scale bar, 500 μm.

Figure 1.

Human SqCC tumor samples can be passaged in mice and phenocopy the primary tumor. A, Schematic representing the passaging strategy and nomenclature for SqCC patient-derived xenografts. B, Kaplan–Meier analysis showing survival of mice transplanted with SqCC PDXs from successive passages. n = indicated number of patient samples, significance determined by Mantel–Cox test, **P < 0.01, ***P < 0.001. C, IHC staining for p63, a marker of SqCC, and TTF-1, a marker of ADC, for a representative SqCC patient tumor (926) and corresponding PDXs. Scale bar, 500 μm.

Close modal

Lung SqCC PDXs recapitulate the molecular genotype of patient tumors

To further investigate whether the molecular characteristics of the patient's tumor were conserved in PDXs, we compared the transcriptome, gene copy number (GCN), and exome of three different patients and their corresponding PDXs (Supplementary Fig. S2A). To evaluate differences that could occur from xenotransplantation and to determine if the molecular characteristics of the patients' tumors were maintained in the samples used for drug studies, we focused our sequencing effort on comparing t0 (patient tumor), t1 (first passage), and t4 (passage used for drug studies). RNA sequencing showed that the three patient tumors, termed 926, 788, and 792, were molecularly distinct and that the PDXs clustered with their corresponding primary tumor (Fig. 2A and Supplementary Fig. S2B). In the top differentially expressed (DE) genes for each patient, upregulated and downregulated genes between t0 and t1 PDXs were conserved after multiple passages in mice, indicating a robust conservation of gene expression between patient tumors and all passages of PDXs (Fig. 2A). Although all PDXs from one patient clustered very tightly together, the patient's tumors were slightly removed from the PDXs for all three patients (Fig. 2A and Supplementary Fig. S2B). The consistency of this effect across patients suggests that common processes were affected after transplantation of the human tumor into mice, consistent with previous reports (33). Analysis of the DE genes between t1 and t0 across patients demonstrated a large overlap where for each patient, over 50% of DE genes were also detected in at least one other patient (Fig. 2B). GO terms enrichment analysis of the 132 overlapping DE genes between all patients showed a strong enrichment for GO terms related to immune response (Supplementary Fig. S2C; Supplementary Table S2). This is consistent with purification of the tumor cells during transplantation as human blood and stromal cells are no longer present in the PDXs. We detected only two DE genes between t4 and t1 tumors in PDX 926, whereas there were no DE genes in other PDXs, suggesting that the expression profile of PDXs remains stable after they have been established in the mouse. Comparison of the transcriptome of our PDXs with the transcriptome of 541 ADCs, 502 squamous cell carcinomas and 110 normal lung samples obtained from the TCGA consortium showed that the gene expression profiles of our SqCC primary tumors and PDXs were consistent with lung SqCC transcriptomes (Supplementary Fig. S2D).

Figure 2.

Genetic alterations in patient tumors are conserved in SqCC patient-derived xenografts. A, Unsupervised hierarchical clustering of the top 1000 most variable genes revealed high consistency of expression profiles between PDXs (passage t1–t6) and the corresponding primary tumor (t0, from patients 788, 792, or 926). B, Venn diagram depicting numbers of DE genes between patient tumors (t0) and corresponding PDXs (t1s) for patients 788, 792, and 926. C, Genome-wide map of copy-number alterations for three SqCC patients and corresponding PDXs. Red indicates amplification and blue indicates deletion according to scale of copy numbers. D, The proportion of the genome altered for primary tumors and corresponding PDXs (926, n = 6; 788, n = 9; 792, n = 8) compared with TCGA SqCC cohort (n = 502). E, Scatter plot of VAF for somatic (dots) and germline (crosses) variants between passage 4 PDX and primary tumor from patient 788. F, Scatter plot of VAF for somatic (dots) and germline (crosses) allele frequencies between two independent PDX passages 4 from patient 788 (left) or PDX passages 6 and 4 from patient 926 (right). Variants that are present at statistically different VAF between the samples are marked in red. Orange circles indicate protein altering variants and green circles indicate variants found in COSMIC database. The size of the dot or a cross is proportional to the sequencing coverage at the variant locus.

Figure 2.

Genetic alterations in patient tumors are conserved in SqCC patient-derived xenografts. A, Unsupervised hierarchical clustering of the top 1000 most variable genes revealed high consistency of expression profiles between PDXs (passage t1–t6) and the corresponding primary tumor (t0, from patients 788, 792, or 926). B, Venn diagram depicting numbers of DE genes between patient tumors (t0) and corresponding PDXs (t1s) for patients 788, 792, and 926. C, Genome-wide map of copy-number alterations for three SqCC patients and corresponding PDXs. Red indicates amplification and blue indicates deletion according to scale of copy numbers. D, The proportion of the genome altered for primary tumors and corresponding PDXs (926, n = 6; 788, n = 9; 792, n = 8) compared with TCGA SqCC cohort (n = 502). E, Scatter plot of VAF for somatic (dots) and germline (crosses) variants between passage 4 PDX and primary tumor from patient 788. F, Scatter plot of VAF for somatic (dots) and germline (crosses) allele frequencies between two independent PDX passages 4 from patient 788 (left) or PDX passages 6 and 4 from patient 926 (right). Variants that are present at statistically different VAF between the samples are marked in red. Orange circles indicate protein altering variants and green circles indicate variants found in COSMIC database. The size of the dot or a cross is proportional to the sequencing coverage at the variant locus.

Close modal

To evaluate copy number changes in our PDX cohort, we performed SNP array genotyping. All our samples displayed high levels of genomic instability with virtually no heterozygous diploid regions remaining (Fig. 2C). Notably, despite the high genomic instability of the primary tumors, the copy number profile remained largely conserved across passages where regions of focal gene amplification or loss were maintained in PDXs compared with the patient's tumor (Fig. 2C). Our PDXs had an average of 74.5% of their genome altered, which is consistent with high levels of genome alteration detected in the TCGA cohort (Fig. 2D). These data demonstrate our PDXs retain high levels of genetic instability, a key characteristic of lung SqCC.

Finally, exome sequencing was performed to evaluate somatic mutations in patient's tumors and PDXs. We observed a high consistency in the somatic variant allele frequency (VAF) between the primary tumor and the corresponding PDX after four passages (Fig. 2E) that remained highly conserved in independent t4 passages or after consecutive passages (Fig. 2F). For patient 788, for which we had matched normal DNA, we detected 474 somatic mutations in the primary tumor and ∼550 somatic mutations in each of two PDX at passage t4. Some de novo mutations were detected at relatively low VAF in PDX788 t4 (Fig. 2E, 16 and 25 de novo mutations for each of the two 788 PDXs). The limited number of de novo somatic mutations with VAF > 50% (2 for each PDX when compared with the primary tumor) suggests conservation of the mutational landscape between the primary tumor and corresponding PDXs. More specifically, point mutations present at high VAF in TP53 in the patient's tumors (VAF > 10%) were always preserved in the corresponding PDXs with no additional mutations present (Supplementary Table S3). An extra TP53 mutation was detected at low VAF (less than 10%) in p792 (T155P) and p926 (K24N) primary tumors. These were subsequently eliminated in PDXs (Supplementary Table S3). We also observed a high correlation of VAF for germline variants (Fig. 2E and F), further suggesting conservation of copy number profiles between the primary tumor and PDXs.

These results indicate that PDXs recapitulate the molecular characteristics of the patient's tumor and that limited genetic drift is observed after passaging of lung SqCC PDXs in mice.

FGFR1 RNA expression in tumor predicts response to FGFR inhibitors

Given that FGFR1 amplification has been described in 20% of lung SqCC (7), we performed FISH to determine FGFR1 amplification status in 36 SqCC PDXs. Results showed that 17% of our PDXs (6/36) have an FGFR1 amplification (Fig. 3A), consistent with data observed from a larger cohort (7). FISH and SNP array analysis further showed that FGFR1 amplification was maintained across passages in FGFR1-amplified tumors (Supplementary Fig. S3A and Supplementary Table S4). FGFR1 amplification is the molecular test currently utilized to recruit patients to clinical trials with FGFR inhibitors (5, 12). We therefore assessed response to FGFR inhibition therapy in two FGFR1amp (406 and 788) and two FGFR1WT PDXs (792 and 926; Fig. 3B). Treatment of tumor-bearing mice with the pan-FGFR inhibitor BGJ398 showed a reduction in tumor growth in PDX 788 (FGFR1amp), whereas FGFR1WT PDX 792 did not respond to therapy (Fig. 3B). Surprisingly, there was no response in PDX 406 (FGFR1amp), whereas robust inhibition of tumor growth occurred in PDX 926 (FGFR1WT; Fig. 3B). These data suggest that FGFR1 amplification is not an accurate predictor for a patient's response to FGFR inhibitor (FGFRi) therapy.

Figure 3.

FGFR1 RNA overexpression is a predictor of response to FGFR inhibition. A, Representative images of FGFR1 FISH in SqCC PDXs t1s (PDX 788, 406, and 926) and t0 (PDX 792). Red and green fluorescent signals indicate FGFR1 and control chromosome 8 (CEP8) probes, respectively. PDXs 792 and 926 show nonamplified signals, whereas 406 and 788 demonstrate gene amplification. B,In vivo response of SqCC PDXs to the FGFR inhibitor BGJ398 (30 mg/kg oral gavage every day for 20 days). Data represent mean ± SEM, n = 6 mice per treatment. Significance determined by Student t test; *, P < 0.05; **, P < 0.01. C, Western blot detecting expression levels of downstream effectors of FGFR signaling (AKT, ERK) and c-myc. Actin serves as the loading control. Untreated PDXs at the t3 to t5 passage number have been used. D, Analysis of FGFR1 RNA expression by RNA ISH, representative images from 4 SqCC PDXs t1s shown.

Figure 3.

FGFR1 RNA overexpression is a predictor of response to FGFR inhibition. A, Representative images of FGFR1 FISH in SqCC PDXs t1s (PDX 788, 406, and 926) and t0 (PDX 792). Red and green fluorescent signals indicate FGFR1 and control chromosome 8 (CEP8) probes, respectively. PDXs 792 and 926 show nonamplified signals, whereas 406 and 788 demonstrate gene amplification. B,In vivo response of SqCC PDXs to the FGFR inhibitor BGJ398 (30 mg/kg oral gavage every day for 20 days). Data represent mean ± SEM, n = 6 mice per treatment. Significance determined by Student t test; *, P < 0.05; **, P < 0.01. C, Western blot detecting expression levels of downstream effectors of FGFR signaling (AKT, ERK) and c-myc. Actin serves as the loading control. Untreated PDXs at the t3 to t5 passage number have been used. D, Analysis of FGFR1 RNA expression by RNA ISH, representative images from 4 SqCC PDXs t1s shown.

Close modal

MYC overexpression and amplification of the 11q13 locus that contains Cyclin D1 (CCND1) have been proposed as biomarkers predictive of response to FGFR inhibition in FGFR1amp tumors (14). In our FGFRamp tumors, FGFRi-sensitive (PDX 788) and FGFRi-resistant (PDX 406) PDXs expressed comparable levels of MYC, indicating that MYC is not a consistent marker to predict response to FGFR inhibition (Fig. 3C). FISH for CCND1 showed amplification of the gene in all PDXs 788, 926, and 406, regardless of their sensitivity to FGFRi (Supplementary Fig. S3B), indicating that CCND1 amplification also does not predict response to therapy. In addition, Western blot analysis of downstream effectors of FGFR signaling, phospho-ERK1/2 and phospho-AKT, did not show any correlation between level of expression of these proteins and response to therapy (Fig. 3C), indicating that activation of these pathways is not a predictor of response to BGJ398.

Given the paucity of antibodies to evaluate FGFR1 protein expression on clinical samples, we assessed whether levels of FGFR1 RNA expression, as detected by RNA-ISH, may be correlated with sensitivity to FGFR inhibitors. FGFRi-sensitive PDX 926, an FGFR1WT tumor, was found to express high levels of FGFR1 RNA by RNA-ISH (named FGFR1+++ from now), whereas FGFR1amp, FGFRi-resistant PDX 406 expressed low levels of FGFR1 RNA (Fig. 3D). RNA-ISH results were further confirmed by RT-qPCR in these four PDXs (Supplementary Fig. S3C). These results indicate that FGFR1 RNA expression is a better predictor of response to FGFR inhibition compared to FGFR1 amplification.

To evaluate the frequency of high FGFR1 RNA expression in NSCLC, we performed RNA-ISH on a NSCLC tissue microarray (TMA) containing 26 ADC, 32 SqCC, 4 adeno-squamous, and 4 large cell carcinomas according to the WHO classification of lung cancer (Supplementary Table S5). Of the 32 SqCCs tested, 8/24 FGFR1WT tumors (33%) were found to have high FGFR1 RNA expression, whereas 3/8 (38%) FGFR1amp tumors did not have FGFR1 RNA expression (Supplementary Table S6). Similar results were observed in our cohort of SqCC PDXs (Supplementary Table S7). Further examples of FGFR1WT tumors expressing high levels of FGFR1 RNA, or FGFR1amp tumors expressing low levels of FGFR1 RNA were identified in TCGA cohort of 503 SqCC tumors (Supplementary Fig. S3D). Interestingly, we found that other histologic subtypes of NSCLC also expressed high levels of FGFR1 RNA, although they did not demonstrate FGFR1 amplification (Supplementary Table S5). Analysis of patient survival data showed that there was no significant correlation between high FGFR1 RNA expression and overall survival or disease-free survival (Supplementary Fig. S4A and S4B).

These data suggest that selecting patients for FGFR inhibitor therapy based solely on FGFR1 amplification by FISH is not an accurate predictive marker for response to treatment. FGFR1 RNA expression appears to be a more valid approach to recruit patients that are likely to benefit from this therapeutic approach and may allow for patients with diverse histologic subtypes of lung cancer to receive effective FGFR-targeted therapy.

PI3K/FGFR combination treatment improves survival in some FGFR inhibitor-sensitive tumors

Our in vivo drug studies showed that BGJ398 treatment resulted in stabilization of the disease but did not reduce tumor burden (Fig. 3B). Histologic analysis of the tumor sections revealed an increase in tumor cell differentiation after BGJ398 treatment in inhibitor-sensitive PDXs 788 and 926, as evidenced by the increased presence of keratin pearls, and a significant reduction in tumor cell proliferation (Fig. 4A and B). A moderate increase in apoptosis was only observed in PDX 788 as detected by CC3 immunostaining (Fig. 4A and B). These results suggest that FGFR inhibition decreases cell proliferation and enhances differentiation, resulting in slower tumor growth, while only inducing a modest increase in cell death.

Figure 4.

FGFR inhibition in sensitive tumors decreases tumor cell proliferation and enhances tumor differentiation. A, Far left panel shows representative H&E for BGJ398-sensitive PDX 788 or 926 after treatment with BGJ398 for 14 days. Black arrowheads indicate keratin pearls. Middle and right panel shows representative immunohistochemistry for Ki67 (proliferation) and CC3 (apoptosis), respectively. Tumors collected 48 hours after first BGJ398 treatment of 30 mg/kg. B, Quantification of Ki67 and CC3 immunostaining for PDXs 788 (left) and 926 (right). Images were quantified by first blinding groups and counting Ki67 or CC3 positive cells in three different fields of view. n = 6 mice per group, data is mean ± SEM. Significance determined by Student t test; *, P < 0.05; **, P < 0.01; ***, P < 0.001. Scale bar, 500 μm. C, Western blot detecting expression levels of c-myc, p-AKT, and p-ERK in FGFR inhibitor-sensitive and inhibitor-resistant SqCC PDXs. Actin serves as the loading control. Tumors collected after 2 weeks of BGJ398 treatment (14 days at 30 mg/kg by oral gavage). Bottom panel shows quantification of the Western blots in top panel by densitometry. Data represent mean ± SEM, n = 6 mice per treatment. Significance determined by Student t test; *, P < 0.05.

Figure 4.

FGFR inhibition in sensitive tumors decreases tumor cell proliferation and enhances tumor differentiation. A, Far left panel shows representative H&E for BGJ398-sensitive PDX 788 or 926 after treatment with BGJ398 for 14 days. Black arrowheads indicate keratin pearls. Middle and right panel shows representative immunohistochemistry for Ki67 (proliferation) and CC3 (apoptosis), respectively. Tumors collected 48 hours after first BGJ398 treatment of 30 mg/kg. B, Quantification of Ki67 and CC3 immunostaining for PDXs 788 (left) and 926 (right). Images were quantified by first blinding groups and counting Ki67 or CC3 positive cells in three different fields of view. n = 6 mice per group, data is mean ± SEM. Significance determined by Student t test; *, P < 0.05; **, P < 0.01; ***, P < 0.001. Scale bar, 500 μm. C, Western blot detecting expression levels of c-myc, p-AKT, and p-ERK in FGFR inhibitor-sensitive and inhibitor-resistant SqCC PDXs. Actin serves as the loading control. Tumors collected after 2 weeks of BGJ398 treatment (14 days at 30 mg/kg by oral gavage). Bottom panel shows quantification of the Western blots in top panel by densitometry. Data represent mean ± SEM, n = 6 mice per treatment. Significance determined by Student t test; *, P < 0.05.

Close modal

Downstream signaling of FGFR occurs through both the MAPK/ERK and the PI3K–AKT–mTOR pathways. In all PDXs, protein expression levels of total AKT or ERK were not significantly altered by treatment with BGJ398 (Fig. 4C). We observed a reduction in the levels of c-myc and phosphorylated AKT (pAKT) in BGJ398-sensitive PDX 788 after treatment, whereas decreased expression of phosphorylated ERK (pERK) was detected in PDX 926 (Fig. 4C). There was no change in expression levels of c-myc, pAKT, or pERK in inhibitor resistant PDX 792. These data suggest that inhibition of either MAPK/ERK or PI3K/AKT is responsible for the delayed tumor growth mediated by BGJ398. To test whether blockade of PI3K/AKT/mTOR could potentiate the activity of FGFR inhibitors, we treated PDXs 788 and 926 (FGFRi-sensitive) with PKI587, a PI3K/mTOR dual inhibitor (34), alone or in combination with BGJ398. PKI587 had no effect as a single agent or in combination with BGJ398 in PDX 788. However, a survival advantage was observed with PKI587 alone, similar to BGJ398 single-agent therapy, in PDX 926 (Fig. 5A and B), a PDX in which FGFR inhibition did not alter the levels of p-AKT (Fig. 4C). Combination treatment with PKI587 and BGJ398 in PDX 926 increased mean survival by 19 days compared with single agent alone (mean survival of 26 days, 24 days, 45 days for PKI587 alone, BGJ398 alone or combination, respectively). Western blot analysis indicated a dual inhibition of pERK and pAKT with combination BGJ398/PKI587 treatment in PDX 926, but not PDX 788, most likely explaining this survival advantage (Fig. 5C). PTEN is a negative regulator of PI3K signaling and loss of PTEN has been described in 8% of lung SqCC (35). However, PTEN expression was detected in both PDXs (Supplementary Fig. S5A). Genomic analysis revealed PIK3CA amplification in PDX 788 but not in 926 (Supplementary Fig. S5B and S5C), consistent with previous studies showing that PIK3CA amplification did not predict response to PI3K inhibitors (36). PIK3CA mutations were not detected in either PDX (Supplementary Fig. S5C). We therefore concluded that BGJ398 treatment can block either ERK or AKT signaling, and that combination of BGJ398 with PI3K inhibitor is beneficial only in tumors where FGFR inhibition does not alter PI3K signaling.

Figure 5.

Combining PI3K inhibition with FGFR inhibition increased treatment efficacy in one PDX model. A and B,In vivo response of PDX 788 (A) and PDX 926 (B) to the FGFR inhibitor BGJ398 (30 mg/kg oral gavage 5 days a week for 5 weeks), PI3K inhibitor PKI587 (15 mg/kg i.v. once weekly for 5 weeks) or in combination. Left panel shows tumor volume after beginning of treatment, right panel shows Kaplan–Meier survival curve. Data represents mean ± SEM, n = 7–8 mice per treatment. Significance determined by Mantel–Cox test; *, P < 0.05; **, P < 0.01. C, Representative Western blot analysis showing the expression of p-AKT, AKT, p-ERK, ERK in tumors treated with BGJ398 and PKI587 as single agents or in combination. Tumors collected at ethical end point. Bottom panel shows quantification of the Western blots by densitometry. Data represent mean ± SEM, n = 8 mice per treatment. Significance determined by Student t test; *, P < 0.05; **, P < 0.01.

Figure 5.

Combining PI3K inhibition with FGFR inhibition increased treatment efficacy in one PDX model. A and B,In vivo response of PDX 788 (A) and PDX 926 (B) to the FGFR inhibitor BGJ398 (30 mg/kg oral gavage 5 days a week for 5 weeks), PI3K inhibitor PKI587 (15 mg/kg i.v. once weekly for 5 weeks) or in combination. Left panel shows tumor volume after beginning of treatment, right panel shows Kaplan–Meier survival curve. Data represents mean ± SEM, n = 7–8 mice per treatment. Significance determined by Mantel–Cox test; *, P < 0.05; **, P < 0.01. C, Representative Western blot analysis showing the expression of p-AKT, AKT, p-ERK, ERK in tumors treated with BGJ398 and PKI587 as single agents or in combination. Tumors collected at ethical end point. Bottom panel shows quantification of the Western blots by densitometry. Data represent mean ± SEM, n = 8 mice per treatment. Significance determined by Student t test; *, P < 0.05; **, P < 0.01.

Close modal

FGFR inhibition combined with cisplatin treatment improves overall survival in FGFR1-overexpressing tumors

To improve response to FGFR inhibition in further models, we evaluated the combination of BGJ398 with cisplatin, the first-line chemotherapy agent for lung SqCC. In vitro treatment of BGJ398-resistant and sensitive PDXs revealed all tumors were responsive to cisplatin treatment, although cell viability was only reduced by 38% to 18% (Fig. 6A and Supplementary Fig. S6A and S6B). However, only BGJ398-sensitive tumors cells displayed improved response to dual BGJ398 and cisplatin treatment in PDXs and H1581 cell line (Fig. 6A and Supplementary Fig. S6A and S6B). In this short-term culture in vitro, PDX cells proliferate minimally, explaining the modest effect of BGJ398, an inhibitor of cell proliferation. In vivo treatment further confirmed a marked improvement in overall survival in FGFR1-RNA-expressing, BGJ398-sensitive tumors with combination therapy (Fig. 6B), whereas no such survival benefit of combination therapy was observed in BGJ398-resistant tumors (Supplementary Fig. S6C). Immunostaining of PDX tumors with Ki67 revealed that cisplatin treatment reduced cell proliferation in PDX 926 and increased cell death in PDX 788 (Fig. 6C and D). However, only combination therapy resulted in robust increases in CC3 positive cells in both PDXs, indicating that cisplatin enhances the potency of BGJ398 by increasing tumor cell death. We suggest that blocking growth signaling through FGFR1 inhibition (BGJ398) and induction of DNA damage by an agent interfering with DNA replication (cisplatin) drives tumor cell apoptosis to result in improved overall survival.

Figure 6.

Combining cisplatin treatment with FGFR inhibition increases survival. A,In vitro cell viability of PDX cells cultured for 72 hours in the presence of cisplatin (5 μmol/L), BGJ398 (1 μmol/L), or the combination. Data represents mean ± SEM, n = 3 (PDX 788), and n = 6 (PDX 926) tumors per group. Significance determined by Student t test between DMSO control and single-agent treatment group; or between BGJ398 single-agent and combination treatment. B, Combination FGFR inhibitor (BGJ398, 30 mg/kg, 5 days a week for 5 weeks) and cisplatin therapy (4 mg/kg every 3 weeks) in PDX 788 and PDX 926, demonstrating a significant survival benefit of combination treatment compared to single-agent therapy in FGFRi-sensitive PDXs. Top panel shows tumor volume after beginning of treatment, bottom panel shows Kaplan–Meier survival curve. n = 8–9 mice per group, significance determined by Mantel–Cox test. **, P < 0.01; ***, P < 0.001. C, Immunostaining for Ki67 and CC3 in PDX 788 and PDX 926 treated with BGJ398 (30 mg/kg per day), cisplatin (4 mg/kg once), or the combination. Tumors collected 48 hours after first treatment. D, Quantification of the Ki67 and CC3 immunostaining for PDXs 788 (left) and 926 (right). Images were quantified by first blinding groups and counting Ki67 or CC3 positive cells in three different fields of view. n = 6 mice per group, data are mean ± SEM. Significance determined by Student t test. Scale bar, 500 μm.

Figure 6.

Combining cisplatin treatment with FGFR inhibition increases survival. A,In vitro cell viability of PDX cells cultured for 72 hours in the presence of cisplatin (5 μmol/L), BGJ398 (1 μmol/L), or the combination. Data represents mean ± SEM, n = 3 (PDX 788), and n = 6 (PDX 926) tumors per group. Significance determined by Student t test between DMSO control and single-agent treatment group; or between BGJ398 single-agent and combination treatment. B, Combination FGFR inhibitor (BGJ398, 30 mg/kg, 5 days a week for 5 weeks) and cisplatin therapy (4 mg/kg every 3 weeks) in PDX 788 and PDX 926, demonstrating a significant survival benefit of combination treatment compared to single-agent therapy in FGFRi-sensitive PDXs. Top panel shows tumor volume after beginning of treatment, bottom panel shows Kaplan–Meier survival curve. n = 8–9 mice per group, significance determined by Mantel–Cox test. **, P < 0.01; ***, P < 0.001. C, Immunostaining for Ki67 and CC3 in PDX 788 and PDX 926 treated with BGJ398 (30 mg/kg per day), cisplatin (4 mg/kg once), or the combination. Tumors collected 48 hours after first treatment. D, Quantification of the Ki67 and CC3 immunostaining for PDXs 788 (left) and 926 (right). Images were quantified by first blinding groups and counting Ki67 or CC3 positive cells in three different fields of view. n = 6 mice per group, data are mean ± SEM. Significance determined by Student t test. Scale bar, 500 μm.

Close modal

We established a bank of NSCLC PDXs from lung cancer surgery and trans-bronchial aspiration (TBNA) EBUS-guided biopsy specimens. In concordance with previous groups, we found SqCCs engraft more readily than ADCs and that more aggressive tumors generate PDXs with greater efficiency (37–40). We observed that samples obtained from EBUS biopsies of metastatic lymph nodes containing higher grade tumors grew more efficiently in mice than those established from early-stage carcinomas obtained by surgical resections. Use of these clinically relevant PDXs enabled us to identify FGFR1 RNA expression as a predictor of response to a FGFR tyrosine kinase inhibitor. In addition, we demonstrated that treatment with cisplatin improved response to FGFR inhibitors by increasing tumor cell death, as opposed to the tumor cell growth arrest observed in tumors treated with the FGFR inhibitor alone.

A possible concern when using PDXs is that substantial selection events could occur during establishment and maintenance of the PDX, which could dramatically alter the molecular composition of the PDX and render it nonrepresentative of the original tumor (41). Detailed molecular characterization of our cohort of SqCC PDXs demonstrated that lung SqCC PDXs preserve gene expression, copy number, and mutational landscapes of the original tumor across all four passages of tumors tested, consistent with previous studies (42, 43). PDXs displayed molecular complexity and genetic instability to the same extent as patient's SqCC tumors, further indicating that PDXs are clinically relevant models to evaluate novel therapeutic approaches. Patient tumors show human immune signature genes that were largely absent in PDX tumors, accounting for the most DE genes between patient and PDX tumors. PDXs also had enrichment of tumor specific gene mutations, possibly due to loss of human stromal elements in the resultant mouse tumors which tend to have a paucity of stroma (33, 44). PDXs therefore represented a renewable source of tumor material that recapitulate the complexity of the human tumors. Detailed molecular characterization of PDXs facilitates correlation between response to therapy and molecular features of the tumor.

FGFR1 amplification was observed in 6 of 36 SqCC PDXs, as detected by FISH and supported by SNP data, but did not always correlate with BGJ398 response, consistent with previous reports (13, 14) and results from phase I clinical trial studies (12). High levels of FGFR1 RNA expression, MYC overexpression, and 11q13 locus amplification that contains CCND1, FGF4, and FGF19 have been suggested as better biomarkers of response to FGFR inhibitors (6, 13, 14). However, we did not observe correlation between CCND1 amplification or MYC expression and sensitivity to FGFR inhibition in our PDX samples. In contrast, we observed that FGFR1 RNA expression predicted response to therapy as suggested by Wynes and colleagues, in a study using lung cancer cell lines (13). Our results confirm these data using clinically relevant PDX models and a more specific FGFR inhibitor. In our PDX cohort, an independent tissue microarray cohort and the TCGA SqCC cohort, we observed that FGFR1ampFGFR1 RNAlo patients who may not respond to therapy are common, as are FGFR1WTFGFR1 RNA+++ patients who may benefit from FGFR inhibition and who would otherwise not enter clinical trials. The identification of FGFR1 RNA+++ tumors in other subtypes of lung cancer is also in agreement with data from other groups (13, 19). It is unknown thus far what is driving aberrant FGFR1 transcription in the absence of a FGFR1 amplification, and further investigations into related microRNAs, transcription factors, or epigenetic regulators are necessary. Overall, our results demonstrate that stratification of NSCLC patients based on mRNA expression levels of FGFR1 would be a better approach to select patients for FGFR-targeted therapy.

Interestingly, high expression levels of FGFR1 did not predict patient outcome, similar to data derived from patients with FGFR1 amplifications (45, 46). This observation may suggest that FGFR1 is not the only driver of tumor growth in FGFR1-overexpressing tumors and that FGFR inhibitors should be combined with targeted therapy or chemotherapy to improve efficacy. Inhibitor-sensitive FGFR1+++ PDX 926 displayed reduced pERK protein expression upon BGJ398 treatment. In contrast, FGFR1amp PDX 788 had no reduction in pERK but decreased levels of pAKT and c-MYC after treatment with BGJ398. It is unknown if the distinct mechanisms of FGFR1 activation or if other aberrantly activated kinases are causing the differing downstream effects of FGFR inhibition in PDXs 788 and 926 and further experiments would be necessary to investigate such conclusions. Using two FGFRi-sensitive lung cancer cell lines, Singleton and colleagues suggested that combination therapy with mTOR inhibitor AZD2014 and FGFR inhibitor AZD4547 improved response compared to single agents (47). However, we observed response to FGFR inhibitor and PI3K/mTOR inhibitor combination only in PDX 926, a tumor with strong AKT activity that was not reduced by treatment with BGJ398 alone. These results indicate that this combination may not be relevant for all patients with FGFR1-overexpressing tumors and identification of the right biomarkers will be necessary to predict response to FGFR/PI3K/mTOR therapy. Amplification of PIK3CA was not a predictor of response to PI3K/mTOR inhibitor or combination therapy, consistent with previous studies (36). We also observed that FGFR inhibition resulted in reduced cell proliferation, enhanced cell differentiation, and caused a modest increase in cell apoptosis. This suggests FGFR1-addicted tumors may continue to grow on cessation of FGFR therapy, and combination therapies may be necessary to drive tumor cell death as opposed to tumor cell growth arrest. To increase cell death, we combined BGJ398 with cisplatin, the first-line chemotherapy agent for SqCC. Combination therapy increased cell death and prolonged survival, indicating that patients with FGFR1-overexpressing tumors may benefit from complementing first-line therapy with a FGFR tyrosine kinase inhibitor. The mechanism responsible for the increased cell death observed after FGFR and cisplatin combination treatment remains to be resolved. Association between genomic instability and response to DNA damaging agents such as cisplatin has been proposed in other tumor models such as pancreatic (48) and ovarian cancers (49). However, such a correlation could not be drawn from our PDX cohort due to all PDXs possessing high levels of genetic instability, and would require analysis of a larger number of samples to be conclusive.

In conclusion, we have established a bank of NSCLC PDXs that recapitulated the patient's tumors at the phenotypic and molecular level. We used these PDXs to identify high FGFR1 RNA expression as a biomarker of response to FGFR inhibition. Furthermore, FGFR1 RNA overexpression can easily be detected in small biopsy specimens by RNA ISH, indicating that this is a potentially clinically relevant biomarker. Finally we show that combination therapy with a FGFR tyrosine kinase inhibitor and cisplatin reduced tumor growth by decreasing cell proliferation and increasing cell death. Our study demonstrates the power of PDX models for dissecting complex molecular events behind drug sensitivity and the ability to use these models to identify appropriate biomarkers of response to targeted therapy.

B. Solomon is a consultant/advisory board member of AstraZeneca, Merck, Bristol Myers Squibb, Pfizer, Novartis, and Roche. No potential conflicts of interest were disclosed by the other authors.

Conception and design: C.E. Weeden, A.Z. Holik, B. Solomon, M.-L. Asselin-Labat

Development of methodology: C.E. Weeden, A.Z. Holik, R.J. Young, D.P. Steinfort, M.-L. Asselin-Labat

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.E. Weeden, A.Z. Holik, R.J. Young, S.B. Ma, S.B. Fox, P. Antippa, L.B. Irving, D.P. Steinfort, G.M. Wright, P.A. Russell, B. Solomon

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C.E. Weeden, A.Z. Holik, L.B. Irving, G.M. Wright, M.E. Ritchie, C.J. Burns, B. Solomon, M.-L. Asselin-Labat

Writing, review, and/or revision of the manuscript: C.E. Weeden, A.Z. Holik, R.J. Young, S.B. Ma, S.B. Fox, L.B. Irving, D.P. Steinfort, G.M. Wright, P.A. Russell, C.J. Burns, B. Solomon, M.-L. Asselin-Labat

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.E. Weeden, J.-M. Garnier, G.M. Wright, P.A. Russell, M.-L. Asselin-Labat

Study supervision: M.-L. Asselin-Labat

The authors thank Leanne Taylor from the Victorian Cancer Biobank for facilitating provision of human samples. We thank Leanne Scott and Melissa Hobbs for excellent animal care and Stephen Wilcox for technical support.

M.-L. Asselin-Labat is supported by a Viertel Foundation Senior Medical Researcher Fellowship. C.E. Weeden is supported by an Australian Post-Graduate Award and a Cancer Therapeutics CRC Top-Up Scholarship. This work was supported by grants from the Victorian Cancer Agency, the Cancer Therapeutics CRC, the Harry Secomb Foundation, the Ian Potter Foundation, the Victorian State Government Operational Infrastructure Support and Australian Government NHMRC IRIISS.

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.
Siegel
RL
,
Miller
KD
,
Jemal
A
. 
Cancer statistics, 2015
.
CA A Cancer J Clin
2015
;
65
:
5
29
.
2.
Australian Institute of Health and Welfare
.
Lung cancer in Australia
; 
2012
. Available from: http://www.aihw.gov.au/cancer/lung/.
3.
Chong
CR
,
Janne
PA
. 
The quest to overcome resistance to EGFR-targeted therapies in cancer
.
Nat Med
2013
;
19
:
1389
400
.
4.
Solomon
B
,
Varella-Garcia
M
,
Camidge
DR
. 
ALK gene rearrangements: a new therapeutic target in a molecularly defined subset of non-small cell lung cancer
.
J Thorac Oncol
2009
;
4
:
1450
4
.
5.
Tiseo
M
,
Gelsomino
F
,
Alfieri
R
,
Cavazzoni
A
,
Bozzetti
C
,
De Giorgi
AM
, et al
FGFR as potential target in the treatment of squamous non small cell lung cancer.
Cancer Treatment Rev
2015
;
41
:
527
39
.
6.
Weeden
CE
,
Solomon
B
,
Asselin-Labat
M-L
. 
FGFR1 inhibition in lung squamous cell carcinoma: questions and controversies
.
Cell Death and Discov
2015
:1;
15049
:
1
5
.
7.
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 Translat Med
2010
;
2
:
62ra93
3
.
8.
Helsten
T
,
Elkin
S
,
Arthur
E
,
Tomson
BN
,
Carter
J
,
Kurzrock
R
. 
The FGFR landscape in cancer: analysis of 4,853 tumors by next-generation sequencing
.
Clin Cancer Res
2016
;
22
:
259
67
.
9.
Wang
R
,
Wang
L
,
Li
Y
,
Hu
H
,
Shen
L
,
Shen
X
, et al
FGFR1/3 Tyrosine Kinase fusions define a unique molecular subtype of non-small cell lung cancer
.
Clin Cancer Res
2014
;
20
:
4107
14
.
10.
Majewski
IJ
,
Mittempergher
L
,
Davidson
NM
,
Bosma
A
,
Willems
SM
,
Horlings
HM
, et al
Identification of recurrent FGFR3 fusion genes in lung cancer through kinome-centred RNA sequencing
.
J Pathol
2013
;
230
:
270
6
.
11.
Liao
RG
,
Jung
J
,
Tchaicha
JH
,
Wilkerson
MD
,
Sivachenko
A
,
Beauchamp
EM
, et al
Inhibitor-sensitive FGFR2 and FGFR3 mutations in lung squamous cell carcinoma
.
Cancer Res
2013
;
73
:
5195
205
.
12.
Nogova
L
,
Sequist
LV
,
Perez Garcia
JM
,
Andre
F
,
Delord
J-P
,
Hidalgo
M
, et al
Evaluation of BGJ398, a fibroblast growth factor receptor 1-3 kinase inhibitor, in patients with advanced solid tumors harboring genetic alterations in fibroblast growth factor receptors: results of a global phase I, dose-escalation and dose-expansion study
.
J Clin Oncol
2017
;
35
:
157
65
.
13.
Wynes
MW
,
Hinz
TK
,
Gao
D
,
Martini
M
,
Marek
LA
,
Ware
KE
, et al
FGFR1 mRNA and protein expression, not gene copy number, Predict FGFR TKI sensitivity across all lung cancer histologies
.
Clin Cancer Res
2014
;
20
:
3299
309
.
14.
Malchers
F
,
Dietlein
F
,
Schottle
J
,
Lu
X
,
Nogova
L
,
Albus
K
, et al
Cell-autonomous and non-cell-autonomous mechanisms of transformation by amplified FGFR1 in lung cancer
.
Cancer Discov
2014
;
4
:
246
57
.
15.
Johnson
JI
,
Decker
S
,
Zaharevitz
D
,
Rubinstein
LV
,
Venditti
JM
,
Schepartz
S
, et al
Relationships between drug activity in NCI preclinical in vitro and in vivo models and early clinical trials
.
Br J Cancer
2001
;
84
:
1424
31
.
16.
Gillet
J-P
,
Calcagno
AM
,
Varma
S
,
Marino
M
,
Green
LJ
,
Vora
MI
, et al
Redefining the relevance of established cancer cell lines to the study of mechanisms of clinical anti-cancer drug resistance
.
Proc Nat Acad Sci
2011
;
108
:
18708
13
.
17.
Derose
YS
,
Wang
G
,
Lin
Y-C
,
Bernard
PS
,
Buys
SS
,
Ebbert
MTW
, et al
Tumor grafts derived from women with breast cancer authentically reflect tumor pathology, growth, metastasis and disease outcomes
.
Nat Med
2011
;
17
:
1514
20
.
18.
Pajic
M
,
Scarlett
CJ
,
Chang
DK
,
Sutherland
RL
,
Biankin
AV
. 
Preclinical strategies to define predictive biomarkers for therapeutically relevant cancer subtypes
.
Hum Genet
2011
;
130
:
93
101
.
19.
Zhang
J
,
Zhang
L
,
Su
X
,
Li
M
,
Xie
L
,
Malchers
F
, et al
Translating the therapeutic potential of AZD4547 in FGFR1-amplified non-small cell lung cancer through the use of patient-derived tumor xenograft models
.
Clin Cancer Res
2012
;
18
:
6658
67
.
20.
de Bruin
EC
,
McGranahan
N
,
Mitter
R
,
Salm
M
,
Wedge
DC
,
Yates
L
, et al
Spatial and temporal diversity in genomic instability processes defines lung cancer evolution
.
Science
2014
;
346
:
251
6
.
21.
Alexandrov
LB
,
Nik-Zainal
S
,
Wedge
DC
,
Aparicio
SAJR
,
Behjati
S
,
Biankin
AV
, et al
Signatures of mutational processes in human cancer
.
Nature
2013
;
500
:
415
21
.
22.
Campbell
JD
,
Alexandrov
A
,
Kim
J
,
Wala
J
,
Berger
AH
,
Pedamallu
CS
, et al
Distinct patterns of somatic genome alterations in lung adenocarcinomas and squamous cell carcinomas
.
Nat Genet
2016
;
48
:
607
16
.
23.
Weeden
CE
,
Chen
Y
,
Ma
SB
,
Hu
Y
,
Ramm
G
,
Sutherland
KD
, et al
Lung basal stem cells rapidly repair DNA damage using the error-prone nonhomologous end-joining pathway
.
PLoS Biol
2017
;
15
:
e2000731
.
24.
Guagnano
V
,
Furet
P
,
Spanka
C
,
Bordas
V
,
Le Douget
M
,
Stamm
C
, et al
Discovery of 3-(2,6-Dichloro-3,5-dimethoxy-phenyl)-1-{6-[4-(4-ethyl-piperazin-1-yl)-phenylamino]-pyrimidin-4-yl}-1-methyl-urea (NVP-BGJ398), a potent and selective inhibitor of the fibroblast growth factor receptor family of receptor tyrosine kinase
.
J Med Chem
2011
;
54
:
7066
83
.
25.
Yau
C
,
Mouradov
D
,
Jorissen
RN
,
Colella
S
,
Mirza
G
,
Steers
G
, et al
A statistical approach for detecting genomic aberrations in heterogeneous tumor samples from single nucleotide polymorphism genotyping data
.
Genome Biol
2010
;
11
:
R92
.
26.
Liao
Y
,
Smyth
GK
,
Shi
W
. 
The Subread Aligner: fast, accurate and scalable read mapping by seed-and-vote
.
Nucleic Acids Res
2013
;
41
:
e108
8
.
27.
Law
CW
,
Chen
Y
,
Shi
W
,
Smyth
GK
. 
voom: precision weights unlock linear model analysis tools for RNA-seq read counts
.
Genome Biol
2014
;
15
:
R29
.
28.
Li
H
. 
Aligning sequence reads, clone sequences and assembly contigs with BWA-MEM
.
ArXiv
. 
2013
;
1
:
1
3
.
29.
Wakefield
MJ
. 
Xenomapper: mapping reads in a mixed species context
.
JOSS
2016
;
1
.
30.
Koboldt
DC
,
Zhang
Q
,
Larson
DE
,
Shen
D
,
McLellan
MD
,
Lin
L
, et al
VarScan 2: Somatic mutation and copy number alteration discovery in cancer by exome sequencing
.
Genome Res
2012
;
22
:
568
76
.
31.
Rahman
M
,
Jackson
LK
,
Johnson
WE
,
Li
DY
,
Bild
AH
,
Piccolo
SR
. 
Alternative preprocessing of RNA-Sequencing data in the cancer genome atlas leads to improved analysis results
.
Bioinformatics
2015
;
31
:
3666
72
.
32.
Steinfort
DP
,
Hew
MJ
,
Irving
LB
. 
Bronchoscopic evaluation of the mediastinum using endobronchial ultrasound: a description of the first 216 cases carried out at an Australian tertiary hospital
.
Int Med J
2009
;
41
:
815
24
.
33.
Reyal
F
,
Guyader
C
,
Decraene
C
,
Lucchesi
C
,
Auger
N
,
Assayag
F
, et al
Molecular profiling of patient-derived breast cancer xenografts
.
Breast Cancer Res
2012
;
14
:
R11
.
34.
Mallon
R
,
Feldberg
LR
,
Lucas
J
,
Chaudhary
I
,
Dehnhardt
C
,
Santos
ED
, et al
Antitumor efficacy of PKI-587, a highly potent dual PI3K/mTOR kinase inhibitor
.
Clin Cancer Res
2011
;
17
:
3193
203
.
35.
Hammerman
PS
,
Lawrence
MS
,
Voet
D
,
Jing
R
,
Cibulskis
K
,
Sivachenko
A
, et al
Comprehensive genomic characterization of squamous cell lung cancers
.
Nature
2012
;
489
:
519
25
.
36.
Johan F Vansteenkiste MD
P
,
MD
J-LC
,
De Braud MD
F
,
MD
FG
,
De Pas MD
T
,
MD
JEG
, et al
Safety and efficacy of buparlisib (BKM120) in patients with PI3K pathway-activated non-small cell lung cancer
.
J Thorac Oncol
2015
;
10
:
1319
27
.
37.
Hao
C
,
Wang
L
,
Peng
S
,
Cao
M
,
Li
H
,
Hu
J
, et al
Gene mutations in primary tumors and corresponding patient- derived xenografts derived from non-small cell lung cancer
.
Cancer Lett
2014
;
357
:
179
185
38.
Ilie
M
,
Nunes
M
,
Blot
L
,
Hofman
V
,
Long-Mira
E
,
Butori
C
, et al
Setting up a wide panel of patient-derived tumor xenografts of non-small cell lung cancer by improving the preanalytical steps
.
Cancer Med
2014
;
4
:
201
11
.
39.
John
T
,
Kohler
D
,
Pintilie
M
,
Yanagawa
N
,
Pham
NA
,
Li
M
, et al
The ability to form primary tumor xenografts is predictive of increased risk of disease recurrence in early-stage non-small cell lung cancer
.
Clin Cancer Res
2011
;
17
:
134
41
.
40.
Leong
TL
,
Marini
KD
,
Rossello
FJ
,
Jayasekara
SN
,
Russell
PA
,
Prodanovic
Z
, et al
Genomic characterisation of small cell lung cancer patient-derived xenografts generated from endobronchial ultrasound-guided transbronchial needle aspiration specimens
.
PLoS ONE
2014
;
9
:
e106862
.
41.
Eirew
P
,
Steif
A
,
Khattra
J
,
Ha
G
,
Yap
D
,
Farahani
H
, et al
Dynamics of genomic clones in breast cancer patient xenografts at single-cell resolution
.
Nature
2015
;
518
:
422
6
.
42.
Wang
D
,
Pham
NA
,
Tong
J
,
Sakashita
S
,
Allo
G
,
Kim
L
, et al
Molecular heterogeneity of non-small cell lung carcinoma patient-derived xenografts closely reflect their primary tumors
.
Int J Cancer
2016
;
140
:
662
73
.
43.
Gao
H
,
Korn
JM
,
Ferretti
SEP
,
Monahan
JE
,
Wang
Y
,
Singh
M
, et al
High-throughput screening using patient-derived tumor xenografts to predict clinical trial drug response
.
Nat Med
2015
;
21
:
1318
25
.
44.
Hidalgo
M
,
Amant
F
,
Biankin
AV
,
Budinska
E
,
Byrne
AT
,
Caldas
C
, et al
Patient-derived xenograft models: an emerging platform for translational cancer research
.
Cancer Discov
2014
;
4
:
998
1013
.
45.
Russell
PA
,
Yu
Y
,
Young
RJ
,
Conron
M
,
Wainer
Z
,
Alam
N
, et al
Prevalence, morphology, and natural history of FGFR1-amplified lung cancer, including squamous cell carcinoma, detected by FISH and SISH
.
Mod Pathol
2014
;
27
:
1621
31
.
46.
Craddock
KJ
,
Ludkovski
O
,
Sykes
J
,
Shepherd
FA
,
Tsao
M-S
. 
Prognostic value of fibroblast growth factor receptor 1 gene locus amplification in resected lung squamous cell carcinoma
.
J Thorac Oncol
2013
;
8
:
1371
7
.
47.
Singleton
KR
,
Hinz
TK
,
Kleczko
EK
,
Marek
LA
,
Kwak
J
,
Harp
T
, et al
Kinome RNAi screens reveal synergistic targeting of MTOR and FGFR1 pathways for treatment of lung cancer and HNSCC
.
Cancer Res
2015
;
75
:
4398
406
.
48.
Waddell
N
,
Pajic
M
,
Patch
A-M
,
Chang
DK
,
Kassahn
KS
,
Bailey
P
, et al
Whole genomes redefine the mutational landscape of pancreatic cancer
.
Nature
2015
;
518
:
495
501
.
49.
Norquist
B
,
Wurz
KA
,
Pennil
CC
,
Garcia
R
,
Gross
J
,
Sakai
W
, et al
Secondary somatic mutations restoring BRCA1/2Predict chemotherapy resistance in hereditary ovarian carcinomas
.
J Clin Oncol
2011
;
29
:
3008
15
.

Supplementary data