Synthetic lethal screening reveals FGFR as one of the combinatorial targets to overcome resistance to Met-targeted therapy

Abstract

Met is a receptor tyrosine kinase that promotes cancer progression. In addition, Met has been implicated in resistance of tumors to various targeted therapies such as EGFR inhibitors in lung cancers, and has been prioritized as a key molecular target for cancer therapy. However, the underlying mechanism of resistance to Met targeting drugs is poorly understood. Here, we describe screening of 1310 genes to search for key regulators related to drug resistance to an anti-Met therapeutic antibody (SAIT301) by employing a siRNA-based synthetic lethal screening method. We found that knockdown of 69 genes in Met-amplified MKN45 cells sensitized the anti-tumor activity of SAIT301. Pathway analysis of these 69 genes implicated FGFR as a key regulator for anti-proliferative effects of Met targeting drugs. Inhibition of FGFR3 increased target cell apoptosis through the suppression of Bcl-xL expression, followed by reduced cancer cell growth in the presence of Met targeting drugs. Treatment of cells with the FGFR inhibitors substantially restored the efficacy of SAIT301 in SAIT301-resistant cells and enhanced the efficacy in SAIT301-sensitive cells. In addition to FGFR3, integrin β3 is another potential target for combination treatment with SAIT301. Suppression of integrin β3 decreased AKT phosphorylation in SAIT301-resistant cells and restores SAIT301 responsiveness in HCC1954 cells, which are resistant to SAIT301. Gene expression analysis using CCLE database shows cancer cells with high levels of FGFR and integrin β3 are resistant to crizotinib treatment, suggesting FGFR and integrin β3 could be used as predictive markers for Met targeted therapy and provide a potential therapeutic option to overcome acquired and innate resistance for the Met targeting drugs.

Introduction

Met, a typical receptor tyrosine kinase (RTK) present on cell surfaces, is overexpressed in various tumors and may contribute to the poor prognosis of several malignancies^1,2. Met functions to mediate a wide spectrum of signals driven by binding with its ligand hepatocyte growth factor/scatter factor (HGF/SF) and promotes cancer progression, metastasis, cancer cell migration and angiogenesis³. HGF/SF binding to Met induces Met dimerization followed by the activation of intracellular signal transduction such as MAPK and AKT phosphorylation^4,5. Among signal cascades induced by Met activation, AKT phosphorylation is related to cell survival⁶; treatment of cancer cells with an anti-Met antibody results in marked antiproliferative effect and a concomitant decrease of AKT phosphorylation and increase of apoptosis⁷.

Because of the relevance of Met to cancer biology, it has been a popular target for cancer drug development^8,9. Several Met-targeted drugs such as cabozantinib, onartuzumab and tivantinib are in phase 3 registration trials. Every Met targeted drug being tested in phase 3 clinical trials targets Met and other targets simultaneously or in combination with another drug¹⁰. Cabozantinib was approved by the United States FDA to treat medullary thyroid cancer; this drug targets Met and vascular endothelial growth factor receptor 2 (VEGFR2) simultaneously¹¹. Onartuzumab, an anti-Met monoclonal human antibody, and tivantinib, a small molecule Met inhibitor, are in phase 3 clinical trials in patients with non-small-cell lung cancer (NSCLC) in combination with erlotinib^12–14.

Met has been validated as an oncogenic kinase in preclinical models through the use of selective kinase inhibitors^15–17. Although these inhibitors may induce early responses, the emergence of drug resistance is common and limits their effectiveness¹⁸. The Met pathway was associated with acquired resistance to epidermal growth factor receptor (EGFR) inhibitors in EGFR mutant NSCLCs¹⁹. In turn, the activation of the HER family was shown to be responsible for the resistance of PHA665752, a Met specific inhibitor, in Met-addicted gastric cancer cells^20,21. It was also reported that resistance to Met targeting inhibitors can occur through MET point mutations, especially at Y1230²², MET gene amplification followed by KRAS over-expression in Met-addicted gastric and lung cancer cells²³, and over-expression of constitutively active SND1-BRAF fusion protein²⁴. In NSCLC, the mechanism of acquired resistance to EGFR/Met tyrosine kinase inhibitor was attributed to the activation of mammalian target of rapamycin (mTOR) and the Wnt signaling pathway²⁵.

However, the underlying mechanism of acquired or inherent resistance to Met targeted antibodies has not been fully elucidated^26–28. Although the relationship between Met and other RTKs in the survival of Met drug resistant cancer cells remains uncertain, it has been shown that Met inhibitor-driven resistance could be rescued by inactivation of fibroblast growth factor receptor (FGFR) by small molecules^29,30. Recently, many approaches have focused on discovering biomarkers for patient selection and exploring novel combination therapies³¹. To systematically identify targets whose inhibition would increase the response of cancer cells to Met inhibitors, we performed medium-throughput siRNA library synthetic lethal screening targeting genes associated with systems biology-derived EGFR and Met signaling pathways³². Here, we show that FGFR could have a role as an alternative driver kinase for Met because dependence on either FGFR or Met can be compensated by activation of the other kinase. Therefore, simultaneous inhibition of FGFR and Met or intervention at a common downstream effector such as AKT is required for effective Met targeted anti-cancer therapeutics.

Previous studies have shown that integrin β1 mediates EGFR drug resistance and its association with the Met signaling pathway in NSCLCs³³. Integrin β subunits are adhesion molecules involved in cell survival and cancer resistance to chemotherapy in breast cancers^34,35. Here, we identify significant crosstalk between integrin β3 and Met in HCC1954 breast cancer cells and investigate the mechanism of Met drug resistance related to integrin signaling. We also demonstrate that perturbation of integrin β3 and FGFR signaling significantly inhibits proliferation of SAIT301-resistant MKN45 cells. These data provide a strong rationale for the use of integrin β3 and FGFR inhibitors in Met-amplified tumors that have become resistant to selective Met inhibition, or to combined therapy to prevent these resistance mechanisms. Our findings demonstrate a specific crosstalk of integrin, FGFR and Met pathways and suggest the partial overlap of downstream signaling and common cellular effects of each pathway.

Results

Synthetic lethal screening to identify sensitizers of cellular response to a Met inhibitor

In order to identify molecular determinants that modulate cellular responses to Met-targeted therapies we developed a siRNA library and performed synthetic lethal screening using a Met-specific monoclonal antibody, SAIT301^7,36. Previously we reported that SAIT301 promotes Met degradation via a LRIG1-mediated pathway. SAIT301 treatment promoted the binding of Met with LRIG1, bypassing the Cbl-mediated Met degradation pathway which requires Met activation. This unique mechanism permits SAIT301 to induce Met degradation without triggering Met signaling activation, and consequently activate cellular apoptosis⁷. The siRNA library used in our studies comprised of siRNAs targeting 1310 genes. We used Met as a seed node to collect data from public archives reporting curated pathway information, protein-protein interactions (PPIs), association in protein complexes, and putative genes responsive to Met antibodies (Supplementary Figure S1). The data mining provided 828 genes in the Met-centered network. As there is good evidence of crosstalk amongst the Met and EGFR pathways, we included the 638 genes from an EGFR-centered network described by one of us (LMW)³². A total of 1310 genes comprised the final network, which included 156 genes shared by the two networks (Supplementary Figure S1).

The MKN45 gastric cancer cell line is dependent on Met signaling for proliferation and survival. This cell line was screened with the siRNA library targeting 1310 genes in combination with vehicle or SAIT301 at IC₂₀. Primary hits were identified as genes that, when knocked down, cause reduction of cell viability by >15% of normalized cell viability in the presence of SAIT301 compare with vehicle control (False Discovery Rate (FDR) < 20% and Sensitization Index (SI) < 0.85) in three independent experiments.

To further validate the positive hits, MKN45 cells were screened using the deconvoluted siRNAs that target candidate genes identified in the initial screening studies. Statistical analysis confirmed that 69 validated genes significantly increased the sensitivity of MKN45 cells to SAIT301 treatment (Table 1). Therefore, the 69 genes were considered to be potential mediators of resistance to SAIT301. Figure 1a demonstrates that the distribution of viability and SI of 69 hits was independent of the viability reduction induced by siRNA knockdown in the absence of antibody treatment. Most of the sensitizing hits were connected in a physically interacting network (Figure 1b), and the distribution of these hits was random in the context of the full library.

Table 1.

69 validated MET antibody-sensitizing genes. Hits are listed alphabetically by the official Entrez gene symbol

Gene Symbol	Entrez ID	Location	Type(s)	Drug(s)
AKT2	208	Cytoplasm	kinase	enzastaurin
AREG/AREGB	374	Extracellular Space	growth factor
ATP1A2	477	Plasma Membrane	transporter	digoxin, omeprazole, ethacrynic acid, perphenazine
BCL2L1	598	Cytoplasm	other
BCR	613	Cytoplasm	kinase	imatinib
BMPR1A	657	Plasma Membrane	kinase
CALR	811	Cytoplasm	transcription regulator
CASP1	834	Cytoplasm	peptidase
CASP2	835	Cytoplasm	peptidase
CCND2	894	Nucleus	other
CD151	977	Plasma Membrane	other
CD1D	912	Plasma Membrane	other
CD247	919	Plasma Membrane	transmembrane receptor	visilizumab, blinatumomab
CD3E	916	Plasma Membrane	transmembrane receptor	visilizumab, blinatumomab, muromonab-CD3
CDKN1B	1027	Nucleus	kinase
CDKN2C	1031	Nucleus	transcription regulator
CHRNA7	1139	Plasma Membrane	transmembrane receptor	varenicline, ABT-089, isoflurane,mecamyla mine,
CRK	1398	Cytoplasm	other
CTSD	1509	Cytoplasm	peptidase
CTTN	2017	Plasma Membrane	other
DIO1	1733	Cytoplasm	enzyme	propylthiouracil
DOK2	9046	Plasma Membrane	other
DUSP2	1844	Nucleus	phosphatase
E2F1	1869	Nucleus	transcription regulator
EGR1	1958	Nucleus	transcription regulator
EPB41L2	2037	Plasma Membrane	other
EPHB1	2047	Plasma Membrane	kinase
EPS15L1	58513	Plasma Membrane	other
FGFR3	2261	Plasma Membrane	kinase	pazopanib
FOS	2353	Nucleus	transcription regulator
GAB1	2549	Cytoplasm	other
GBP1	2633	Cytoplasm	enzyme
GRB7	2886	Plasma Membrane	other
HIC1	3090	Nucleus	transcription regulator
HOPX	84525	Nucleus	transcription regulator
HSF4	3299	Nucleus	transcription regulator
HSP90B1	7184	Cytoplasm	other	17-dimethylaminoethyla mino -17-demethoxygeldanamy cin, IPI-504, cisplatin
IL24	11009	Extracellular Space	cytokine
INSRR	3645	Plasma Membrane	kinase
ITGB3	3690	Plasma Membrane	transmembrane receptor	abciximab, TP 9201, EMD121974, tirofiban
KDM1A	23028	Nucleus	enzyme
KRAS	3845	Cytoplasm	enzyme
MAP3K11	4296	Cytoplasm	kinase
MCM2	4171	Nucleus	enzyme
MCM7	4176	Nucleus	enzyme
MYB	4602	Nucleus	transcription regulator
NFKB2	4791	Nucleus	transcription regulator
PARP1	142	Nucleus	enzyme	ABT-888, olaparib, INO-1001
PDX1	3651	Nucleus	transcription regulator
PLAU	5328	Extracellular Space	peptidase
PRKAB1	5564	Nucleus	kinase
PTPN11	5781	Cytoplasm	phosphatase
RAB5A	5868	Cytoplasm	enzyme
RAC1	5879	Plasma Membrane	enzyme
RAC2	5880	Cytoplasm	enzyme
RAF1	5894	Cytoplasm	kinase	CHIR-265, vemurafenib, regorafenib, sorafenib
RGS16	6004	Cytoplasm	other
RPS6KA1	6195	Cytoplasm	kinase
RPS6KA2	6196	Nucleus	kinase
SERPINA3	12	Extracellular Space	other
SGK1	6446	Cytoplasm	kinase
SIN3A	25942	Nucleus	transcription regulator
SOS1	6654	Cytoplasm	other
SPEN	23013	Nucleus	transcription regulator
SRF	6722	Nucleus	transcription regulator
STK3	6788	Cytoplasm	kinase
TNIP2	79155	Cytoplasm	other
TYR	7299	Cytoplasm	enzyme	hydroquinone, azelaic acid
WDR1	9948	Extracellular Space	other

Figure 1. Design and screening of a targeted library.

(a) Distribution of normalized viability and SI of 69 validated hits. The siRNAs for the 69 validated target genes are listed in the order of intrinsic impact on viability of MKN45 cells treated with medium (blue circles). Blue circles, normalized viability with the siRNA knockdown. Red circles, SI for validated hits with SAIT301. (b) Network of 69 validated hits (red squares) sensitizing to Met-targeting SAIT301 in the context of the full library (blue squares). Lines (edges) represent connections based on PPIs. Hits and genes from the starting set that were not connected in the network are shown below the network. (c) Heatmap of the Sensitization Index of the 69 validated genes in MKN45, H1993, A549, BxPC3, HCC827, NCI-N87, HCC1954 and RKO cells. SI on log scale is shown on the heatmap. Gene with SI <1 are depicted in green, and those with SI>1 are shown in red.

We next assessed the efficacy of these 69 putative sensitizers to SAIT301 treatment in seven other cell lines – NCI-N87 (gastric cancer), HCC827 (lung cancer), BxPC3 (pancreatic cancer), H1993 (lung cancer), A549 (lung cancer), RKO (colon cancer) as well as HCC1954 (breast cancer). As shown in Figure 1c, 24 genes sensitized at least three of these cell lines to the effects of a Met targeting antibody.

Pathway analysis of the targets from the screen implicates the FGFR and integrin pathways as key regulators of the efficacy of a Met targeting antibody

To delineate the functional associations and identify potential interacting partners of the 69 validated hits, we evaluated the hit network using Ingenuity Pathway Analysis (IPA) software (http://www.ingenuity.com/). Gene set enrichment analysis showed that seven canonical signaling pathways were utilized by the identified 69 hits (P<0.015). The genes involved in each individual pathway are summarized in Figure 2a. As these pathways were all statistically significantly enriched by the analysis, we examined the frequency of the genes shared by different pathways. Genes such as RAF1, AKT2, SOS1, KRAS (shaded in blue) are heavily shared by 6 out of 7 pathways, indicating that the RAS/RAF/MEK and PI3K/AKT signaling pathways are common downstream pathways. ITGB3, FGFR3, SGK1 (red, orange shades) are not shared by more than two signaling pathways, therefore representing the involvement of individual signaling pathways. Subsequent network analysis corroborated several interacting partners of Met signaling that potentially mediate resistance to Met inhibitors. Among them, highly ranked partners are cell surface receptors triggering intracellular signaling system, such as FGFR3, involved in the FGFR signaling pathway, and ITGB3, triggering the integrin signaling pathway. Depletion of these two genes significantly reduces cancer cell viability when employed in combination with SAIT301 treatment. IPA analysis of the distribution of connections indicates potential interactions amongst the FGFR, integrin and Met signaling pathways (Figure 2b), which may serve as compensatory pathways for escape from Met blockade.

Figure 2. Canonical pathway enrichment and network analysis of 69 validated hits reveal links between Met and FGFR, integrin signaling pathway.

(a) In silico analysis was performed using the Ingenuity Pathway Analysis (IPA) software and categorized the hits according to their proposed enrichment for canonical pathways. The top seven signaling pathways are plotted based on the number of genes identified within each pathway. (P<0.015). Hits were included in one signaling pathway (red shades); two signaling pathways (orange shades); three signaling pathways (yellow shades); four signaling pathways (green shades); and six signaling pathways (blue shades). (b) Interactions among hits and potential interaction partners were accessed using IPA. Nodes represent genes, with their shape representing the functional class of the gene product, and the edges indicate the biological relationship between the nodes. Hits are represented in pink outlined symbols, interaction partners not in the hit list in black.

Based on these findings, FGFR3 and ITGB3 were selected for further studies that explore the potential for SAIT301-based combination therapy.

FGFR3 suppression increases the anti-proliferative effects of SAIT301 by induction of apoptosis in MKN45 cells

FGFR signaling is known to be responsible for acquired resistance to RTK inhibitors, such as crizotinib in acute myeloid leukemia (AML)²⁹ and vemurafenib in melanoma³⁷. In the synthetic lethal screening studies described here, the growth inhibitory effect of SAIT301 was increased with concomitant knockdown of the FGFR3 gene in MKN45 cells compare with control siRNA transfected MKN45 cells (Figure 3a). Knockdown of FGFR3 gene also increased the growth inhibitory effect of other Met inhibitors such as PHA665752 and XL-184 (Figure 3b and 3c); PHA665752 is a small molecule inhibitor that specifically targets Met¹⁷, and XL-184 is a small molecule inhibitor that targets the Met, VEGFR2, and Ret kinases³⁸. We then investigated the mechanisms through which FGFR3 knockdown promotes the anti-tumor activity of SAIT301 in MKN45 cells. FGFR3 knockdown increased the apoptosis of MKN45 cells compare with treatment with SAIT301 alone (Figure 3d). Apoptosis induction by FGFR3 suppression was further confirmed by cleavage of PARP and caspase-3 (Supplementary Figure S2). It is reported that FGFR3 inhibition leads to apoptosis due to decreased expression of Bcl-2³⁹. We therefore examined the expression of Bcl-xL, which is transcribed by BCL2L1, one of the 69 sensitizers identified by our synthetic lethal screening. FGFR3 knockdown substantially enhanced reduction of Bcl-xL protein and SAIT301-induced PARP cleavage (Figure 3e). To determine the underlying signaling mechanism for the regulation of Bcl-xL expression, we tested the MAPK inhibitors (SB203580 and U0126) and an AKT inhibitor (MK-2206). Interestingly, a p38 inhibitor, SB203580, in the presence of a Met targeting antibody, suppressed Bcl-xL expression, followed by induction of apoptosis (Figure 3f). Consistent with this finding, BCL2L1 knockdown also increased the growth inhibitory effect of SAIT301 (Figure 3g). Additionally, FGFR signaling activation by bFGF treatment restored the growth inhibition induced by SAIT301 (Supplementary Figure S3), further confirming the role of FGFR activity in cellular sensitivity to a Met-targeting antibody. These data strongly support that FGFR signaling has an anti-apoptotic role in MKN45 cells by sustaining Bcl-xL expression, and FGFR suppression makes these cells susceptible to apoptosis induced by Met-targeting drugs.

Figure 3. FGFR3 suppression enhances the efficacy of an anti-Met antibody through decreased expression of Bcl-xL.

(a) The viability of MKN45 cells was measured by CTG assay after reverse-transfection of control or FGFR3 targeting siRNAs for 24 hours and treating with medium or 0.016 µg/ml of SAIT301 for 72 hours. The relative cell viability (%) represents percent growth compared to the control group (no antibody treatment). (b) The viability of MKN45 cells was measured by CTG assay after reverse-transfection of control or FGFR3 targeting siRNAs for 24 hours and treating with medium or 10 nM of PHA665752 for 72 hours. The relative cell viability (%) represents percent growth compared to the control group (no inhibitor treatment). (c) The viability of MKN45 cells was measured by CTG assay after reverse-transfection of control or FGFR3 targeting siRNAs for 24 hours and treating with medium or 62.5 nM of XL-184 for 72 hours. The relative cell viability (%) represents percent growth compared to the control group (no inhibitor treatment). (d) The apoptotic rate of MKN45 cells was determined by caspase activation using Caspase 3/7 Glo assay, after reverse-transfection of control or FGFR3 targeting siRNAs for 24 hours and treating with medium or 0.016 µg/ml of SAIT301 for 72 hours. Cell numbers were normalized in parallel wells using CTG assay. The relative caspase-3/7 activity (%) is represented as a percentage comparison to the control group (no antibody treatment). (e) Cleaved PARP and Bcl-xL were detected by immunoblotting assay in MKN45 cells. Twenty-four hours after reverse-transfection of control, FGFR3 or BCL2L1 targeting siRNAs, MKN45 cells were treated with 2 µg/ml of SAIT301 for 72 hours. GAPDH was a loading control. (f) Cleaved PARP and Bcl-xL were detected by immunoblotting assay with MKN45 cells treated with 2 µg/ml of SAIT301 alone or in combination with p38 inhibitor (SB203580, 10 µM) for 48 hours. (g) The viability of MKN45 cells was measured by CTG assay after reverse-transfection of control or BCL2L1 targeting siRNAs for 24 hours and treating with medium or 0.016 µg/ml of SAIT301 for 72 hours. The relative cell viability (%) represents percent growth compared to the control group (no antibody treatment).

Combinations of selective FGFR inhibitors and Met inhibitors synergistically inhibit cancer cell growth accompanied by induction of apoptosis

Because genetic knockdown of FGFR3 sensitizes MKN45 cells to SAIT301 treatment, we assessed whether pharmacological inhibitors of the FGFR family have similar effects. The small molecule tyrosine kinase inhibitor, PD173074, is a specific and potent inhibitor of the FGFR family⁴⁰. PD173074 potently inhibited tyrosine phosphorylation of FGFR in MKN45 cells and siRNAs directed against other FGFR family members also increased Met targeted antibody-mediated efficacy (Supplementary Figure S4). MKN45 cell viability was reduced by Met inhibition using SAIT301 alone, and this inhibition was significantly enhanced by the addition of PD173074 (combination index [CI] < 1; isobologram analysis; Figure 4a). PD173074 treatment also promoted the cell growth inhibitory effect of crizotinib, a small molecule targeting Met as well as ALK¹⁶, PHA665752, and XL-184 (Figure 4b, 4c and 4d). To further confirm this effect, we tested EBC1, a lung cancer cell line that is known to be addicted to Met signaling for its survival⁷. Cell growth inhibitory effect of SAIT301 was increased by the addition of PD173074 in EBC1 cells (combination index [CI] < 1; isobologram analysis; Figure 4e). Like MKN45 cells, EBC1 cells showed marked reduction of cell viability when treated with Met targeting inhibitors, PHA665752 and XL-184, in the presence of PD173074 compare with the treatment with either drug alone (Supplementary Figure S5a and S5b). We tested another selective FGFR inhibitor, NVP-BGJ398 (hereinafter referred to as BGJ398) that is currently in phase I clinical trials⁴¹. BGJ398 combined with SAIT301 also has a synergistic effect on growth inhibition of both EBC1 and MKN45 cells (combination index [CI] < 1; isobologram analysis; Supplementary Figure S6). We then determined if the effects of combining FGFR inhibitors with SAIT301 were due to induction of apoptosis. SAIT301-mediated cell apoptosis in EBC1 cells was dramatically enhanced with the addition of PD173074 or BGJ398 (Figure 4f). Furthermore, as measured by immunoblot analysis, PD173074 significantly induced PARP cleavage when combined with SAIT301 (Figure 4g). PD173074 alone had no effect on the protein levels of Bcl-xL, but exposure to the combination of PD173074 and SAIT301 induced a nearly complete ablation of Bcl-xL. As BCL2L1 knockdown improved SAIT301 efficacy in MKN45 cells, combined treatment with SAIT301 and ABT-263, a potent inhibitor of Bcl-xL⁴², synergistically inhibited cell growth of EBC1 cells (Figure 4h). Crizotinib also synergized with BGJ398 in EBC1 cells (Figure 4i and 4j). BGJ398 treatment increased the apoptotic rate induced by Met inhibitors, PHA665752 or XL-184 (Figure 4k and 4l). These data suggest that FGFR activation maintains anti-apoptotic regulation in Met-addicted cancer cells. We further confirmed the combination effect using a nude mouse heterotropic MKN45 xenograft model. SAIT301 demonstrated an inhibition of tumor growth, resulting in the reduction of tumor volume by 27.8%. Combination treatment of SAIT301 and BGJ398 resulted in the reduction of tumor volume by 41.2% (Supplementary Figure S7). These findings further support the concept that FGFR inhibitors and Met-targeted drugs cooperatively inhibit cancer cell growth.

Figure 4. FGFR inhibitor and anti-Met antibody synergize to activate apoptosis.

(a) The viability of MKN45 cells was measured by CTG assay after treating with SAIT301 (0.04 µg/ml) alone or in combination with the FGFR inhibitor (PD173074, 5 or 10 µM) for 72 hours. The relative cell viability (%) represents the percent growth as compared to the control group (no antibody and inhibitor treatment). The combination index (CI) < 1.0 indicates synergistic effects. (b) The viability of MKN45 cells was measured by CTG assay after treating with 16 nM of PHA665752 alone or in combination with the FGFR inhibitor (PD173074, 10 µM) for 72 hours. The relative cell viability (%) represents the percent growth as compared to the control group (no inhibitor treatment). (c) The viability of MKN45 cells was measured by CTG assay after treating with 4 nM of PHA665752 alone or in combination with the FGFR inhibitor (PD173074, 10 µM) for 72 hours. The relative cell viability (%) represents the percent growth as compared to the control group (no inhibitor treatment). (d) The viability of MKN45 cells was measured by CTG assay after treating with 4 nM of XL-184 alone or in combination with the FGFR inhibitor (PD173074, 10 µM) for 72 hours. The relative cell viability (%) represents the percent growth as compared to the control group (no inhibitor treatment). (e) The viability of EBC1 cells was measured by CTG assay after treating with 0.04 µg/ml of SAIT301 alone or in combination with the FGFR inhibitor (PD173074, 5 or 10 µM) for 72 hours. The relative cell viability (%) represents the percent growth as compared to the control group (no antibody and inhibitor treatment). (f) The apoptotic rate of EBC1 cells was determined by caspase activation using Caspase 3/7 Glo assay after treating with 0.08 µg/ml of SAIT301 alone or in combination with the FGFR inhibitors (PD173074 10 µM, BGJ398 5 µM) for 72 hours. Cell numbers were normalized in parallel wells using CTG assay. The relative caspase-3/7 activity (%) is represented as a percentage comparison to the control group (no antibody treatment). (g) Cleaved PARP and Bcl-xL were detected by immunoblotting assay with EBC1 cells treated with 0.4 µg/ml of SAIT301 alone or in combination with PD173074 (10 µM) for 72 hours. (h) The viability of EBC1 cells was measured by CTG assay after treating with 0.08 µg/ml of SAIT301 alone or in combination with the Bcl-xL inhibitor (ABT-263, 0.5 µM) for 72 hours. The relative cell viability (%) represents the percent growth as compared to the control group (no antibody and inhibitor treatment). (i) The viability of EBC1 cells was measured by CTG assay after treating with 10 nM of crizotinib alone or in combination with the FGFR inhibitor (BGJ398, 5 µM) for 72 hours. The relative cell viability (%) represents the percent growth as compared to the control group (no inhibitor treatment). (j) The apoptotic rate of EBC1 cells was determined by caspase activation using Caspase 3/7 Glo assay after treating with 20 nM of crizotinib alone or in combination with the FGFR inhibitor (BGJ398, 5 µM) for 72 hours. Cell numbers were normalized in parallel wells using CTG assay. The relative caspase-3/7 activity (%) is represented as a percentage comparison to the control group (no antibody treatment). (k) The apoptotic rate of EBC1 cells was determined by caspase activation using Caspase 3/7 Glo assay after treating with 20 nM of PHA665752 alone or in combination with the FGFR inhibitor (BGJ398, 5 µM) for 72 hours. Cell numbers were normalized in parallel wells using CTG assay. The relative caspase-3/7 activity (%) is represented as a percentage comparison to the control group (no antibody treatment). (l) The apoptotic rate of EBC1 cells was determined by caspase activation using Caspase 3/7 Glo assay after treating with 1 µM of XL-184 alone or in combination with the FGFR inhibitor (BGJ398, 5 µM) for 72 hours. Cell numbers were normalized in parallel wells using CTG assay. The relative caspase-3/7 activity (%) is represented as a percentage comparison to the control group (no antibody treatment).

MKN45-SAIT301 resistant cells display activation of FGFR3, which confers resistance to Met inhibition through reactivation of AKT signaling

Since acquired resistance is a general phenomenon in targeted therapy, identifying the relevant resistance mechanisms can improve the clinical benefit of targeted cancer drugs⁴³. We created SAIT301-resistant clones of MKN45 cells and EBC1 cells that originally were sensitive to this antibody⁷. The viability of SAIT301-resistant clones nos. 1 and 24 was not affected by SAIT301 treatment at up to 2 µg/ml concentrations that induced ~50% growth inhibition in parental MKN45 cells (Figure 5a). To characterize underlying molecular mechanisms of resistance in these SAIT301-resistant clones, we investigated the activation status of several signaling mediators. As shown in Figure 5b, the phosphorylation of FGFR3 and AKT is higher in SAIT301-resistant clones compare with the parental MKN45 cells. To test if insensitivity of resistant cells to SAIT301 is due to the activation of AKT, we performed cell viability assays with SAIT301 in the presence of an AKT inhibitor, MK-2206^44,45. AKT inhibition using MK-2206 re-sensitized MKN45-SAIT301 resistant cells to SAIT301 in a dose-dependent manner (Figure 5c and Supplementary Figure S8a).

Figure 5. FGFR activation is required for acquired resistance to anti-Met antibody.

(a) The viability of MKN45, MKN45-SAIT301 resistant clones #1 and #24 was measured by CTG assay after treating with various concentrations (0.016, 0.08, 0.4, and 2 µg/ml) of SAIT301 for 72 hours. The relative cell viability (%) represents the percent growth as compared to the control group (no antibody treatment). (b) Phosphorylation of FGFR3 and AKT was detected by immunoblot assay in MKN45, MKN45-SAIT301 resistant clones #1 and #24. GAPDH was a loading control. (c) The viability of MKN45-SAIT301 resistant clone #1 cells was measured by CTG assay after treating with various concentrations (0.016, 0.008, 0.4 and 2 µg/ml) of SAIT301 alone or in combination with the AKT inhibitor (MK-2206, 5 µM) for 72 hours. The relative cell viability (%) represents percent growth compared to the control group (no antibody and inhibitor treatment). (d) AKT phosphorylation was detected by immunoblot assay in MKN45-SAIT301 resistant clone #24 cells. Twenty-four hours after pre-treatment of FGFR inhibitor (PD173074, 10 µM), MKN45-resistant clone #24 cells were treated with medium or 2 µg/ml of SAIT301 for 1 hour. (e) The viability of MKN45-SAIT301 resistant clone #24 cells was measured by CTG assay after treating with various concentrations (0.016, 0.008, 0.4 and 2 µg/ml) of SAIT301 alone or in combination with the FGFR inhibitor (BGJ398, 5 µM). The relative cell viability (%) represents percent growth compared to the control group (no antibody and inhibitor treatment). (f) The viability of EBC1-SAIT301 resistant clone #20 cells was measured by CTG assay after treating with various concentrations (8, 12, 16 and 20 nM) of crizotinib alone or in combination with the FGFR inhibitor (BGJ398, 3 µM) for 72 hours. The relative cell viability (%) represents percent growth compared to the control group (no antibody and inhibitor treatment).

Because AKT is commonly involved in the signaling of several RTKs, FGFR could be an upstream RTK of activated AKT in SAIT301 resistant clones. As shown in Figure 5d, SAIT301-mediated inhibition of AKT phosphorylation was significantly enhanced by co-treatment with PD173074. To investigate the role of FGFR in acquired resistance for SAIT301, MKN45-SAIT301 resistant clone #24 cells were treated with SAIT301 and BGJ398. Concomitant inhibition of Met and FGFR signaling using SAIT301 and BGJ398 led to a dose-dependent inhibition of viability of MKN45-SAIT301 resistant cells (Figure 5e). Because parental MKN45 cells displayed decreased levels of Bcl-xL following FGFR inhibition (Figure 3e), we tested whether Met inhibition in combination with ABT-263 could inhibit the viability of MKN45-SAIT301 resistant cells. ABT-263 as well as BGJ398 significantly reduced the cell viability of MKN45-SAIT301 resistant #1 cells treated with SAIT301 (Supplementary Figure S8b). To confirm the role of FGFR function in acquired resistance to Met targeting therapy, we tested EBC1-SAIT301 resistant clones, and found that AKT phosphorylation was elevated in these cells (Supplementary Figure S9a). Combined inhibition of Met and FGFR led to a reduction of cell viability in a dose-dependent manner (Figure 5f and Supplementary Figure S9b). These data further support FGFR and its downstream pathway mediated signaling as a potentially important mechanism of acquired resistance to Met inhibition.

ITGB3 inhibition improves the anti-tumor efficacy of SAIT301 in cancer cells with acquired and innate resistance to Met inhibition

ITGB3 is another highly ranked hit from our synthetic lethal screening study, and ITGB3 is known to influence the development and progression of breast cancers^46,47. To investigate the relationship of ITGB3 expression to Met resistance, knockdown assays were conducted with MKN45-SAIT301 resistant cells. As shown in Figure 6a, ITGB3 suppression by siRNA enhanced the effects of SAIT301 on cell viability. A potent integrin β3 inhibitor, cilengitide^19,48, also enhanced the effects of SAIT301 in MKN45-SAIT301 resistant cells (Figure 6b). Combination treatment of cilengitide further enhanced SAIT301-mediated inhibition of AKT phosphorylation compared to treatment with SAIT301 alone (Figure 6c).

Figure 6. ITGB3 suppression increases anti-proliferative effects of anti-Met antibody.

(a) The viability of MKN45-SAIT301 resistant clone #1 cells was measured by CTG assay after reverse-transfection of control or ITGB3 targeting siRNAs for 24 hours and treating with various concentrations (0.016, 0.008, 0.4 and 2 µg/ml) of SAIT301 for 72 hours. (b) The viability of MKN45-SAIT301 resistant clone #1 cells was measured by CTG assay after treating with various concentrations (0.016, 0.008, 0.4 and 2 µg/ml) of SAIT301 alone or in combination with the integrin β3 inhibitor (cilengitide, 10 µM) for 72 hours. The relative cell viability (%) represents percent growth compared to the control group (no antibody and inhibitor treatment). (c) AKT phosphorylation was detected by immunoblot assay in MKN45-SAIT301 resistant clone #1 cells. Twenty-four hours after pre-treatment of the integrin β3 inhibitor (cilengitide, 10 µM), MKN45-resistant clone #1 cells were treated with medium or 2 µg/ml of SAIT301 for 1 hour. (d) The viability of HCC1954 cells was measured by CTG assay after reverse-transfection of control or ITGB3 targeting siRNAs for 24 hours and treating with various concentrations (0.016, 0.008, 0.4 and 2 µg/ml) of SAIT301 for 72 hours. (e) AKT phosphorylation was detected by immunoblot assay in HCC1954 cells. Forty-eight hours after reverse-transfection of control or ITGB3 targeting siRNAs, MKN45-resistant clone #1 cells were treated with medium or 2 µg/ml of SAIT301 for 30 minutes. (f) CCLE mRNA expression data with crizotinib treated 504 cell lines. Group A consists of cell lines with “low” expression of FGFR2, FGFR3 and ITGB3, and Group B has cell lines with “high” expression for at least one of the three genes.

HCC1954 breast cancer cells have high Met protein expression, and were employed to test the efficacy of Met targeting drugs⁴⁹. HCC1954 cells did not respond to SAIT301 treatment and were shown to have very low protein levels of FGFR3 but relatively high levels of integrin β3 (data not shown). It is noteworthy that a FGFR inhibitor, PD173074, yields no synergistic inhibition of proliferation with SAIT301 in these cells (data not shown). The reason might be low levels of FGFR3 phosphorylation compared to MKN45 cells in which FGFR inhibitors have synergistic effects with SAIT301 (Supplementary Figure S10). To investigate the role of ITGB3 in innate resistance to SAIT301, the effect of knockdown of ITGB3 was tested in HCC1954 cells. The combination of SAIT301 treatment and ITGB3 knockdown resulted in greater inhibition of cell growth (Figure 6d) and AKT phosphorylation (Figure 6e) as compared to SAIT301 treatment alone.

Additionally, to elucidate the potential implication of FGFR and ITGB3 in cancer cell responses to Met targeting therapy in general, we first analyzed IC₅₀ values for 504 cell lines following treatment with crizotinib (PF-2341066) from the CCLE database (http://www.broadinstitute.org/ccle/) and categorized each cell line into three tiers (low, medium and high) based on the expression levels of FGFR2, FGFR3, and ITGB3, with cutoffs at 33 and 67 percentile values. Cell lines that have “low” expression for all three genes were assigned to Group A, while Group B consisted of cell lines that have “high” expression for at least one of the three genes. The median IC₅₀ values for Group A and B are 6.81 and 8.00, respectively (p=0.045 by Mann-Whitney U test) (Figure 6f). These data provide molecular validation of ITGB3 or FGFR family as being synthetically lethal with a Met inhibitor under defined conditions.

DISCUSSION

As it is well known that repeated treatment with targeted cancer drugs eventually leads to drug resistance¹⁸, investigating the mechanism of acquired resistance is important in the development of cancer drugs. siRNA screening has been demonstrated as an effective method for identifying the potential mechanisms of acquired or inherent resistance to certain drugs^50,51. We developed a siRNA library centered to Met and EGFR signaling and performed synthetic lethal screening to search for key regulators related to drug resistance against an anti-Met therapeutic antibody. The siRNA library used in our studies was comprised of siRNAs targeting the 1310 Met- and EGFR-centered genes. By focusing on the statistically significant viability changes and by filtering off-target effects by requiring two or more siRNAs to replicate the effects, we were able to identify 69 genes connected to several signaling pathways that likely mediate SAIT301 resistance. By examining the frequency of the genes shared by these seven putative signaling pathways, we found many heavily shared hits (RAF1, AKT2, SOS1, KRAS) located at the mediator level of cross-talk involved in resistance, did not represent any specific pathway. In contrast, canonical pathway enrichment and network analysis of the 69 gene validated hit set revealed enrichment of the integrin and FGFR signaling pathways. ITGB3 and FGFR3 regulate the integrin and FGF signaling pathways, respectively, are well enriched surface molecules and are easily accessible to targeted therapies. Furthermore, potential interaction indicated compensatory pathways amongst the FGFR, integrin and Met signaling pathways.

Recent reports have referred to signaling crosstalk between FGFR and Met. In AML, FGFR1 is responsible for innate resistance to crizotinib by inducing HGF transcription²⁹. In turn, HGF secretion compensates for cancer cell growth inhibition by BGJ398, a selective FGFR inhibitor⁵². The results from the present study support previous findings, as we show that suppression of FGFR can significantly improve the anti-proliferative effects of Met targeting drugs in Met-addicted cell lines such as MKN45 (gastric cancer) and EBC1 (lung cancer). In MKN45, FGFR signaling has an anti-apoptotic function by retaining Bcl-xL expression through constitutively active p38, and suppresses the transcription of pro-apoptotic genes such as TRAIL and BIM (data not shown). Pharmacological inhibitors targeting FGFR synergize with Met targeting drugs in preventing cancer cell growth in vitro and in vivo. These data suggest that a Met targeting drug in combination with an FGFR inhibitor warrants further interrogation in combination therapy.

In the present study, we demonstrated that SAIT301-resistant clones obtained through prolonged treatment with SAIT301 exhibit enhanced FGFR3 activation and increased AKT phosphorylation. AKT activation appears to be a key mechanism for resistance to the Met-targeting antibody, and MK-2206, an AKT inhibitor, restores the sensitivity of resistant cells to SAIT301 (Figure 5c and Supplementary Figure S8a). FGFR signaling compensates for the inhibition of Met in 2 ways, by retaining anti-apoptotic molecules and by activating PI3K/AKT signaling. These data provide the rationale for further investigation combinations of Met and FGFR inhibitors to overcome acquired resistance.

We also examined integrin signaling, which emerged as another highly ranked resistance-promoting signaling pathway in the analysis of our screening studies. Met binds to integrin β3 and promotes the migration and invasion of mammary epithelial cells⁴⁷. Suppression of ITGB3 improved the efficacy of SAIT301 in vitro, through anti-proliferative effects but not by facilitating apoptosis in MKN45 cells (data not shown). Cilengitide, an integrin β3 inhibitor, also restored sensitivity to SAIT301 in SAIT301-resistant MKN45 cells. In addition, integrin signaling may be related to innate resistance to SAIT301. SAIT301 alone has no anti-proliferative effect on HCC1954, even though this cell line possesses a relatively high level of Met expression compare with other breast cancer cell lines. However, ITGB3 inhibition combined with SAIT301 had significant anti-tumor effects. Interestingly, transcript levels of ITGB3 are high in HCC1954 compare with other Met-addicted cell lines, such as MKN45 and EBC1 (data not shown). These findings imply that ITGB3 can be used as a putative predictive marker as well as a target for combination therapy employing SAIT301. Although future studies will be required to confirm the hypothesis that the expression of FGFR and ITGB3 can both predict the efficacy of SAIT301 and comprise a new combination treatment strategy, our synthetic lethal screening approach appears to be a useful way to identify potential candidates for combination with a tumor targeting monoclonal antibody that perturbs cancer cell signaling.

In summary, we have identified a number of key regulators of resistance to a Met targeting antibody by synthetic lethal screening. Among them, we have validated that FGFR and integrin are important components for resistance mechanism toward SAIT301, and possibly for Met inhibitors in general. Combining Met inhibitors with FGFR or integrin inhibitors may be useful to mitigate the development of clinical resistance. In addition, both FGFR and integrin may be useful as predictive markers for clinical response to Met inhibitors.

Materials and methods

siRNA library construction

We have previously reported on an EGFR network targeted library³² using EGFR as the seed protein. This network was involved in this study based on the evidence of crosstalk amongst the Met and EGFR pathways. A distinct Met-centered network containing 828 genes was developed using the open source software tool, Cytoscape with Met as the seed protein. Bioinformatic databases were mined for protein-protein interactions (PPIs), protein complexes, members of canonical pathways linked with Met and additional genes include genes encoding RTKs receptor tyrosine kinases and differentially expressed genes in SAIT301-treated NCI-H441 cells. These sources included BIND, BioGRID, DIP, HPRD, IntAct, MiMI, MINT, STRING, Biocarta, Linnea, Protein Lounge, STKE, and literature searches (Figure S1). This resulted in a siRNA library comprised siRNAs targeting 1310 genes.

Further testing of identified sensitizing genes in responsive and non-responsive cell lines

We further tested the 69 validated hits for sensitization to SAIT301 in three SAIT301 responsive cell lines (BxPC-3, HCC827, H1993) versus four SAIT301 non-responsive cell lines (A549, NCI-N87, RKO and HCC1954). Cells were transfected with siRNA pools comprised of two most effective siRNAs identified during the validation. SI and statistical significance were calculated as in Supplementary Information. All experiments were performed at least three times independently.

Network analysis with 69 validated hits

For canonical pathway enrichment, network analysis and gene interaction networks were generated using Ingenuity Pathway Analysis (IPA; Ingenuity systems, Redwood City, CA, USA; www.Ingenuity.com). 69 validated hits were uploaded into the application. MiMI plugin for Cytoscape was used to access the integrated molecular data and to retrieve interaction map that display protein interactions. All 69 genes were hierarchically clustered as distance measure with Cluster 3.0, and then visualized using Java TreeView (version1.1.6r2).

Generation of SAIT301-resistant MKN45 and EBC1 cells in vitro

To generate SAIT301 resistant clones, MKN45 and EBC1 cells were exposed to increasing concentrations (range 1–10 µg/ml) of SAIT301 over 3 months in vitro. To confirm that these clones are durably resistant to SAIT301, cell viability tests were performed after the resistant clones had been cultured in the presence or absence of SAIT301 for 6 weeks.

Cell viability and apoptosis assay

Cell viability test and apoptosis assay were performed as previously reported^7,36. Synergism for drug combination was quantified by the combination index (CI), where CI<1 and CI=0 indicate synergistic and additive effects, respectively. CI was calculated with CalcuSyn software (Biosoft, Cambridge, UK).