Rewiring ERBB3 and ERK signaling confers resistance to FGFR1 inhibition in gastrointestinal cancer harbored an ERBB3-E928G mutation

Recently, a large number of studies found that activation of ERBB3 (Erb-B2 receptor tyrosine kinase 3, also known as HER3) may be one of the major mechanisms underlying resistance to therapies that target the EGFR (epidermal growth factor receptor), HER2, and other receptor tyrosine kinases (RTKs) (Chen et al., 2011; Choi et al., 2012; Ross et al., 2018). Interestingly, mutations in ERBB3 are commonly reported in gastrointestinal (GI) cancer, with mutations identified in approximately 12% of stomach and colorectal cancer cases (Jaiswal et al., 2013). Here we also mined TCGA-generated data in the cBioPortal for Cancer Genomics regarding ERBB3 and its related RTKs (ERBB2, EGFR, VEGFR, IGF1R, MET and FGFR) family members) in GI cancer. Our analysis revealed that the ERBB3 gene is genetically altered in 9% (75 of total 797 queried samples) of GI samples analyzed, making ERBB3 the third most commonly altered gene after ERBB2 (15%) and EGFR (10%). Interestingly, unlike other RTKs, mutations in ERBB3 (the green color labeled) accounted for 73.3% of ERBB3 genetic alterations (55 mutations out of total 75 ERBB3 genetically altered samples) (Fig. S1), suggesting that mutations in ERBB3 might play a role in the progression of GI cancer. We therefore hypothesized that ERBB3 might play a central role in conferring resistance to commonly used targeted therapies in patients with GI cancer; moreover, we hypothesized that pharmacologically blocking ERBB3 might help reduce this resistance to tyrosine kinase inhibitors (TKIs), thereby delaying relapse and improving patient outcome. Here we demonstrate that ERBB3-E928G mutated cells are highly resistant to FGFR1 inhibition via an increased activation of ERBB3 and its downstream ERK signaling pathways, and combination of ERBB3 monoclonal antibody LJM716 and the specific FGFR1 inhibitor PD173074 synergistically suppressed the growth of ERBB3-mutated gastrointestinal cancer and potently overcome tumor refractory to FGFR1 inhibitors. We firstly investigated whether the presence of mutant ERBB3 proteins affects the response to clinically available TKIs in a subset of GI cancer cell lines. Specifically, we measured the effects of TKIs in four GI cancer cell lines carrying representative ERBB3 hotspot mutations, including CW-2 cells (ERBB3-E928G), KYSE150 cells (ERBB3D297Y), HCT116 cells (ERBB3-Q261*), HCT15 cells (ERBB3-N126K), and AGS control cell line with wildtype (WT) ERBB3 (Fig. 1A). Notably, CW-2 cells have significantly higher levels of phosphorylated ERBB3 (pERBB3) compared to the other four screened GI cancer cell lines (Fig. 1A). For each cell line, we measured cell viability in the presence of various concentrations of Gefitinib (an EGFR inhibitor), Lapatinib (a HER2 inhibitor), Apatinib (a VEGFR-2 inhibitor), Linsitinib (an IGF-1R inhibitor), Tivantinib (a c-Met inhibitor), and BGJ398 (an FGFR1-3 inhibitor). Results showed that only CW-2 cells, which harbor the ERBB3E928G mutation, were more resistant to the FGFR1-3 inhibitor BGJ398 compared to the other four cell lines (Fig. 1A). Results of the time course experiment validated that CW-2 cells were virtually unaffected by BGJ398 treatment compared with AGS control cells (Fig. 1B). Importantly, two FGFR4 inhibitors, BLU554 (Fig. 1C) and BLU9931 (Fig. S2), had a similar effect on CW-2 cells and AGS cells. Given that BGJ398 inhibits FGFR1, FGFR2 and FGFR3, we examined which FGFR underlies the resistance to BGJ398 in CW-2 cells. As shown in Fig. 1D, CW-2 cells are more resistant to the FGFR1-specific inhibitor PD173074 (Nguyen et al., 2013) compared to AGS cells. We also used RNAi to knock down FGFR1 expression in both CW-2 and AGS cells (Fig. S3). As shown in Fig. 1E, knocking down FGFR1 using two different shRNA constructs significantly reduced the viability of AGS cells compared to CW-2 cells. These results suggest that the ERBB3-E928G mutation underlies the cellular resistance to FGFR1 inhibitors. Thus, we hypothesized that the E928G mutation in ERBB3 kinase domain might coordinate with FGFR1 to maintain the growth and survival of CW-2 cells in the presence of FGFR1 inhibition.

mRNA was measured in CW-2 and AGS cells transfected with either a control shRNA or two shRNA constructs that target FGFR1 using real-time PCR analysis and was normalized to the control. * P < 0.05, ** P < 0.01, *** P < 0.001, and n.s., not significant. 5 Figure S4: ERBB3 knocking down efficiency in CW-2 cells. (A) ERBB3 mRNA was measured in CW-2 cells transfected with either a control shRNA or two shRNA constructs that target ERBB3. (B) pERBB3 and total ERBB3 in CW-2 cells were tested three days after transfection with the indicated shRNA constructs. * P < 0.05, ** P < 0.01, *** P < 0.001, and n.s., not significant.

Cell culture
Cells in this study including CW-2, KYSE150, HCT15, HCT116, and AGS were purchased from the Cell Bank at the Chinese Academy of Sciences (Shanghai, China).
Cells were cultured as instructed. All culture media contained 10% (v/v) fetal bovine serum (FBS), and all cells were cultured at 37°C in 5% CO2.

TKIs for use in in vitro and in vivo experiments
Except where indicated otherwise, all TKIs were obtained from Selleck Chemicals. in accordance with the manufacturer's instructions.

Foci formation assay
15 Cells (500 cells/well) were incubated in 6-well plates at 37°C for 24 hours, and then treated with drugs and DMSO as control; every 3 days, the culture medium was replaced with fresh medium containing the corresponding drugs. When the cells grew to the level of visible colonies (i.e., foci), the cells were washed once in phosphatebuffered saline (PBS) and fixed in 4% paraformaldehyde (Sigma-Aldrich) for 20 mins.
The cells were then stained with 0.5% crystal violet diluted in 20% methanol for another 15 mins at room temperature, and the number of colonies was counted using light microscopy.
The specific shRNA hairpin sequences used to knock down ERBB3 and FGFR1 are listed in Supplementary Table 1.

Quantitative real-time PCR
Total RNA was extracted with TRIzol (#3101-100, Pufei Biotech), and the concentration and purity were measured using a spectrophotometer. RNA was reversetranscribed using the PrimeScript RT Kit (#RR037A, Takara), and quantitative PCR was then performed using a CFX96 Real-Time System (Bio-Rad) with SYBR Green Supermix (#B21202, Biomake) in accordance with the manufacturer's instructions. The fold difference in gene expression was calculated using the 2-△△Ct method and is presented relative to GAPDH mRNA. All reactions were performed in triplicate, and specificity was monitored using melting curve analysis. The PCR primers used in this study are listed in Supplementary Table 2.

Western blot analyses
Total proteins were prepared by homogenizing the cells in RIPA buffer containing protease inhibitors. The homogenate was cleared by centrifugation at 4°C for 15 mins at 12,000 rpm, and the supernatant (containing the protein fraction) was collected.

Co-immunoprecipitation (Co-IP)
For PD173074 (20 mg/kg body weight). The health status of each animal was monitored daily, including any change in weight. After 3 weeks, the mice were euthanized and the xenograft tumors were harvested, photographed, and then fixed in 4% paraformaldehyde and embedded in paraffin for subsequent immunohistochemistry.

Immunohistochemistry
After removal as described above, the tumors were fixed overnight in 4% paraformaldehyde. The fixed tumors were then treated with gradient methanol to remove any air, embedded in paraffin, and sectioned at 4 μm using a rotary microtome.

Statistical analyses
Data were analyzed and plotted using SPSS 19.0 or GraphPad Prism v6.0 (La Jolla, CA), and all summary data are presented as the mean ± the standard deviation (SD).
Two groups were compared using the Student's t-test, and multiple groups were compared using a one-way ANOVA followed by Tukey's post hoc test. Differences with a P-value <0.05 were considered statistically significant.