Anti-tumor efficacy of theliatinib in esophageal cancer patient-derived xenografts models with epidermal growth factor receptor (EGFR) overexpression and gene amplification

Targeted therapy is not yet approved for esophageal cancer (EC). In this study, we first evaluated EGFR gene and protein expression in 70 Chinese EC patient tumor samples collected during surgery. We then established 23 patient-derived EC xenograft (PDECX) models and assessed the efficacy of theliatinib, a potent and highly selective EGFR inhibitor currently in Phase I clinical study, in 9 PDECX models exhibiting various EGFR expression levels. Immunohistochemical analysis showed that 50 patient tumor samples (71.4%) had high EGFR expression. Quantitative PCR showed that eight tumors (11.6%) had EGFR gene copy number gain, and fluorescence in situ hybridization (FISH) revealed that four tumors had EGFR gene amplification. These results suggest that EGFR protein may be overexpressed in many EC tumors without gene amplification. Also detected were rare hot-spot mutations in EGFR and PIK3CA, whereas no mutations were found in K-Ras or B-Raf. Theliatinib exhibited strong antitumor activity in PDECX models with high EGFR expression, including remarkable tumor regression in two PDECX models with both EGFR gene amplification and protein overexpression. However, the efficacy of theliatinib was diminished in models with PI3KCA mutations or FGFR1 overexpression in addition to high EGFR expression. This study demonstrates that theliatinib could potentially benefit EC patients with high EGFR protein expression without mutations or aberrant activities of associated factors, such as PI3KCA or FGFR1.


INTRODUCTION
Esophageal cancer (EC) is the eighth most common cancer worldwide and ranks in the top 5 in China in terms of patient cases and mortalities [1,2]. The 2012 Chinese cancer registry annual report shows that EC accounts for nearly 10% of all cancer deaths. Large numbers of EC patients were generally diagnosed in an advanced stage and had poor prognosis. Patients with unresectable or metastatic EC are typically treated with chemotherapy using a combination of 5-fluorouracil (5-FU) and cisplatin and have a median survival of less than one year [3]. There are limited salvage options for patients with refractory EC [4] and targeted therapies are not yet available for EC. Therefore, there is great scope and need for novel targeted therapeutic options for EC.
Overexpression of EGFR is commonly found in EC and is associated with poor prognosis [5]. Unlike colon or lung cancers, K-Ras mutations are less frequent in EC [6]. This suggests that inhibition of EGFR pathway maybe therapeutically beneficial to EC patients with EGFR overexpression. However, clinical trials with anti-EGFR antibodies or first generation tyrosine kinase inhibitors (TKIs) in advanced EC patients have been disappointing Research Paper www.impactjournals.com/oncotarget [7,8,9], partly due to lack of prospective patient selection. For example, in the phase III COG trial conducted in mainly esophageal adenocarcinoma patients, the overall survival rates for gefitinib and placebo controls were comparable (3.73 months for gefitinib and 3.67 months for placebo), although there were slight improvements in progression free survival, disease control rate and quality of life [10]. Subsequently, a TRANS-COG trial that was prospectively as part of the COG trial demonstrated that gefitinib significantly enhanced overall survival in patients carrying EGFR gene amplification (HR = 0.19, p = 0.007) and EGFR gene copy number gain (HR = 0.53, p = 0.042) [11]. A recent pre-clinical study showed that esophageal tumor cell lines with high polysomy of EGFR were sensitive to gefitinib [12]. These findings suggested that EGFR directed therapy could be beneficial to esophageal cancer patients with high EGFR protein expression or gene copy number. Although the TRANS-COG trial investigated the relationship between EGFR gene expression and anti-tumor response of gefitinib, EGFR protein expression was not reported. Therefore the correlation between EGFR protein expression and response to EGFR-TKI is not yet clear.
Theliatinib is a novel EGFR tyrosine kinase inhibitor that is currently being evaluated in phase I clinical trial in China (NCT02601248). We conducted a pre-clinical study to evaluate the anti-tumor activity of theliatinib in a panel of patient derived esophageal cancer xenograft (PDECX) models to determine the association between anti-tumor activity of theliatinib and different levels of EGFR expression in EC tumors.

Establishment of PDECX models
As of July 2014, 54 fresh EC samples were implanted subcutaneously into NOD-SCID mice, of which 23 (42.6%) grew for 3 consecutive passages (P3). Fourteen of the 23 well-growing PDECX models, including 11 with high EGFR expression (H score ≥ 200) were successfully profiled. Finally, top 9 of the 14 PDECX models (Table 2) with a broad range of EGFR expression levels (H score = 15~300) were selected to assess the correlation between EGFR expression levels and theliatinib sensitivity. Then, we compared the histological and molecular characteristics between PDECXs and the corresponding patient tumor samples and found no significant changes during the serial tumor tissue passages, indicating that the PDECX models closely represented the human tumors. Analyses of two representative models, PDECX1T0326 and PDECX1T0781 with their corresponding primary tumors are shown in Figure 2.

Theliatinib is a highly selective and potent ATP-competitive inhibitor of EGFR
The chemical structure of theliatinib is shown in Figure 3A [13]. Theliatinib is a highly potent EGFR inhibitor with 3~7 fold greater potency than erlotinib or gefitinib at both the enzyme and the cell level ( Figure 3B and Table 3). The enzyme kinetics studies demonstrated that theliatinib, gefitinib or erlotinib were all ATP-competitive inhibitors with Ki values 0.05, 0.35 and 0.38 nM, respectively against the wild type EGFR ( Figure 3C-3E

In vivo anti-tumor activity of theliatinib in multiple PDECX models
Nine PDECX models were selected to evaluate in vivo anti-tumor activity of theliatinib, including 7 with high EGFR expression (H score > 200), 2 (PDECX 1T0326 and PDECX 1T0950) with simultaneous EGFR gene amplification and high EGFR protein expression, 1 with medium EGFR expression (PDECX 1T0474, H score = 180) and 1 with low EGFR expression (PDECX 1T0773, H score = 15). We selected 15 mg/kg/day theliatinib and 20 mg/kg/day gefitinib doses for mice that were similar to those achieved in the clinics.
We observed that the two PDECX models with EGFR gene amplification (PDECX1T0326 and PDECX1T0950) were most sensitive to theliatinib treatment demonstrating tumor regression of 32% and 75%, respectively, at the end of study ( Figure 4 and Table 4). The PDECX1T0326 model demonstrated significant inhibition of phosphorylation of EGFR (p-EGFR) and its downstream targets, AKT and ERK (p-AKT and p-ERK) as shown in Figure 4B. In the PDECX1T0950 model, a dose response (2, 5 and 15 mg/kg) was observed ( Figure 4C), suggesting that anti-tumor activity was dependent on the level of inhibition of EGFR pathway activation. Gefitinib also displayed anti-tumor activity in these two models with EGFR gene amplification ( Figure 4A and Figure 4D). The 20 mg/kg/day dose of gefitinib administered to nude mice was similar to the 500 mg/day dose used in the phase III COG trial. In most PDECX models, the anti-tumor effect of 15 mg/kg theliatinib was significantly better than 20 mg/kg gefitinib ( Figure 4A, 4D and Figure 5).
Generally, the anti-tumor efficacy of theliatinib correlated with the level of EGFR expression in PDECX models ( Figure 5 and Table 4). Robust efficacy was observed in PDECX1T0781 and PDECX1T1315 that demonstrated high EGFR expression ( Figure 5A and Table 4). However, the efficacy of theliatinib was attenuated in PDECX1T0472 and PDECX1T0327 that also had high EGFR expression, probably due to other gene alterations, such as PIK3CA mutation in PDECX1T0472 and FGFR1 overexpression in PDECX1T0327 ( Figure 5A and Table 4). In fact, the PDECX1T0327 model demonstrated rapid tumor regression upon treatment with AZD4547 (12 mg/kg), a selective FGFR inhibitor, suggesting that FGFR may be a stronger driver in this tumor than EGFR. These results suggested that when multiple pathways are aberrantly activated in tumors, combination therapy would be necessary. In models   Figure 5B and Table 4).

DISCUSSION
In the present study, 93% (65/70) of Chinese esophageal tumor samples were squamous cell carcinoma (ESCC) in contrast to mostly adenocarcinoma in Caucasian population [14]. ESCC tumors are mostly situated in the upper two thirds of the esophagus and associated with smoking and alcohol, salty foods and nitrosamine in foods. In contrast, esophageal adenocarcinoma (EAC) tumors are located almost exclusively in the lower third of the esophagus and esophagogastric junction and are associated with Barrett's metaplasia and chronic gastroesophageal reflux [14,15]. Apart from the histological and epidemiological distinctions, ESCC and EAC differ in molecular features [16]. These differences suggest that the two cancer subtypes have distinct molecular pathogenesis and genetic alterations. However, the COG and TRANS-COG trials demonstrated that the tumor response to EGFR-TKI correlates with EGFR status rather than histological subtypes [10,11]. The preclinical results presented in our study further demonstrate that tumors with high EGFR expression with or without EGFR gene copy number gains result in sensitivity to theliatinib treatment. Therefore, our data demonstrates that theliatinib would be clinically beneficial to esophageal cancer patients with high EGFR expression.
Since the anti-tumor effect of theliatinib correlated with EGFR expression levels in the PDECX models, patient selection is important for beneficial outcomes. The two PDECX models with EGFR gene amplification were most sensitive to theliatinib treatment with rapid and robust tumor regression. The PDECX models without EGFR amplification but with high EGFR expression (IHC H score = 270~300) also responded strongly to theliatinib with tumor growth inhibition of 83~96%. In contrast, the PDECX models with low EGFR expression (IHC H score < 200) had low to moderate effect upon theliatinib treatment. Therefore, our study indicates that tumors with EGFR gene amplification and high EGFR expression may demonstrate improved object response rate (ORR), whereas tumors with high EGFR expression without EGFR gene amplification may demonstrate improved progression free survival (PFS) with or without ORR. Based on the preclinical findings of theliatinib, we propose that esophageal cancer patients with high EGFR levels (IHC H score ≥ 270) should potentially benefit from theliatinib treatment.
The presence of EGFR activating mutations is an established predictive biomarker in non-small cell lung cancer (NSCLC) with in-frame deletions of exon 19 (E746_A750) and L858R substitution in exon 21 The top 9 models were used to determine anti-tumor activity of theliatinib.
accounting for > 90% of all activating mutations [17]. However, these mutations are distinctly less common in esophageal cancer [18]. In the present study, exon 19 deletions and L858R mutation in exon 21 were not detected in any of the 70 patients. However, a novel EGFR H850L mutation in exon 21 was found in 1 patient and the biological significance of this mutation needs to be determined in future. Further, we found that in 27% (18/66) of cases, an EGFR SNP at codon 787 of exon 20 was present, similar to a previous finding in Japanese ESCC patients in which Q787Q SNP was reported to be associated with decreased overall survival in the patients that received chemoradiotherapy [19]. Interestingly, Q787Q SNP was also identified in head and neck cancer cell lines and patient samples and was associated with higher sensitivity to gefitinib [20]. In our anti-tumor studies, one PDECX model (1T0994, H score 230) that carried Q787Q SNP was insensitive to theliatinib.
Therefore, the molecular details between Q787Q SNP and response to EGFR-TKIs need to be investigated further.
Overall, EGFR hot-spot mutations were rare in our study. It is also well recognized that lung cancer cells carrying wild-type EGFR are less sensitive to EGFR TKIs than lung cancer cells with EGFR activating mutations [21,22]. This was responsible for failure of the first generation TKIs, gefitinib or erlotinib in lung cancer patients with wild type EGFR in earlier clinical trials. In comparison to erlotinib or gefitnib, theliatinib showed much stronger binding affinity to wild type EGFR and was more difficult to be replaced by ATP. This unique feature may result in better target engagement for theliatinib compared to erlotinib or gefitinib, leading to stronger antitumor activity in tumors with wild type EGFR activation due to gene amplification or protein overexpression.
Our study also demonstrates that aberrant function of other receptor tyrosine kinase receptors or somatic mutations in down-stream targets such as FGFR overexpression or PIK3CA mutation can influence drug sensitivity of EC to EGFR TKIs. The frequency of PIK3CA mutations in EC varies from 0% to 21% [23]. However, the predictive or prognostic role of the PIK3CA mutation in EC remains unclear. In our study, one EC sample with a PIK3CA E542K mutation on exon 9 and the corresponding PDECX1T0472 showed moderate sensitivity to theliatinib treatment with 84% TGI, which was lower than two other PDECX models with similar EGFR H scores, 1T0781 and 1T1315, without any PIK3CA mutation. This suggested that the PIK3CA hot-spot mutations may attenuate the sensitivity to EGFR inhibitors. Therefore, it would be interesting to investigate anti-tumor activity of theliatinib in combination with a PI3K inhibitor in the future. FGFR1 gene amplification was reported in ESCC with a frequency of 6% to 21% [24] and protein overexpression in 17% of ESCC [25]. However, the therapeutic potential of FGFR1 targeted therapy in ESCC has not been fully evaluated. The PDECX model 1T0327 with overexpression of both EGFR and FGFR1 protein showed only modest tumor growth inhibition to theliatinib treatment suggesting that FGFR1 overexpression could diminish theliatinib efficacy. Interestingly, treatment with a FGFR inhibitor AZD4547 induced rapid tumor regression, suggesting that FGFR1 overexpression was driving the tumor growth in the specific model. This is also the first report of antitumor activity of a FGFR TKI in a PDECX model with high FGFR1 expression. We propose to test the efficacy of FGFR TKI either alone or in combination with EGFR TKI to evaluate its therapeutic benefits in ESCC.
There are some limitations of our study. Firstly, the sample size was relatively small and may have inherent bias. Secondly, although Sanger sequencing is a costeffective method to detect gene mutations, it has several drawbacks such as low sensitivity that may result in not detecting novel mutations. Thirdly, in vivo theliatinib target inhibition study was carried out at 8h after treatment with 15 mg/kg theliatinib ( Figure 4B). We propose further investigating the target inhibition effect at 24 h to clearly establish the relationship between target inhibition and anti-tumor activity.    In summary, we observed high EGFR protein levels in majority of Chinese EC patient samples that we analyzed. Comparatively, EGFR gene amplification was less prevalent in these patients. Theliatinib, a novel EGFR TKI with strong affinity to wild type EGFR protein demonstrated dose-dependent anti-tumor activity in a panel of PDECX models with a generally good correlation between EGFR H score and tumor growth inhibition. Furthermore, aberrant activation or gene mutations of other targets such as PI3K and FGFR diminished the antitumor activity of the EGFR TKIs, especially, theliatinib. In conclusion, our data suggests that thelaitinib would be beneficial for patients that have high EGFR expression thereby proper patient selection strategy would result in enhanced drug efficacy. Therefore, in phase IB trials, it would be worthwhile to analyze the ORR, DCR and PFS parameters in EC patients with high EGFR protein expression (IHC H score ≥ 270, with or without EGFR gene amplification).

Chinese esophageal cancer patients in this study
Seventy esophageal tumor specimens from newly diagnosed patients were collected during surgical resection from the Shanghai Biobank Network of Common Human Tumor Tissue at the Changhai hospital, Shanghai, China. Two additional surgical EC samples used for PDECX models establishment were provided by Renji Hospital, Shanghai. Prior written informed consent was obtained from all patients.

PDECX models and anti-tumor efficacy studies
Seven-to nine week old NOD-SCID (NOD.CB17-Prkde<scid>/JSlac) immunodeficient or BALB/cASlacnu/nu male or female mice were obtained from Shanghai SLAC Laboratory Animal Co. Ltd or Shanghai LingChang BioTech Co. Ltd. Fresh tumor specimens from newly diagnosed patients were collected during surgery and separated into three parts for the following: (1) To prepare formalin fixed paraffin embedded (FFPE) sections; (2) For snap freezing in liquid nitrogen for DNA extraction and sequencing and (3) For subcutaneous implantation into NOD-SCID mice (P0) and subsequent passages in additional NOD-SCID or nude mice once the tumor size reached 800~1500 mm 3 . After several consecutive in vivo passages, the PDECX models (P3~P7) were used to evaluate the anti-tumor efficacy of theliatinib or gefitinib or AZD4547. All experiments in animals were performed in accordance with protocols approved by the Hutchison MediPharma Limited Animal Care and Use Committee (HMPLACUC).
When the average tumor volume reached the 250~500 mm 3 , mice were randomly divided into different experimental groups. Tumor-bearing mice were daily administered oral doses of test compounds (Theliatinib in 0.5% CMC-Na; Gefitinib in 0.5% Tween-80 and AZD4547 in 1% Tween-80 (tested only in the PDECX 1T0327 model)) or vehicle control. Body weight and tumor size of all mice was measured twice or thrice a week. Tumor volumes (TV) were calculated by measuring two perpendicular diameters with calipers (formula: TV = (length × width 2 )/2). Tumor growth inhibition (TGI) was calculated using the formula TGI= [1-(Vt -V0)drug treated / (Vt -V0)vehicle control]×100%. Statistical significance was determined by Student's t test and p < 0.05 was considered statistically significant.

Hematoxylin and Eosin (H&E) and EGFR-IHC staining and scoring
Tissues from EC patients were harvested and fixed in 10% buffered formalin for 24 h within 30 minutes after resection. Then, 4 μm tissue sections were cut and paraffinized before storage. H&E staining were performed following the routine procedures [26] and finally, diagnosed by pathologists at the Changhai hospital. IHC staining on patient tumor sections was performed with EGFR PharmDx (DAKO, K1494) on DAKO auto-stainer Link48 according to the manufacturer's instructions. Briefly, 4 μm tissue sections were prepared and dried for 1h at room temperature and then placed in a 56~60°C incubator for 1h. After deparaffinization and rehydration [26], sections were incubated with proteinase K solution for 5 minutes. Endogenous peroxidase activity was blocked with 3% hydrogen peroxide for 5 minutes and washed with the buffer followed by incubation with mouse monoclonal anti-human EGFR antibody (clone2-18C9c) for 30 minutes. Then, the sections were incubated with secondary goat anti-mouse antibody conjugated with horseradish peroxidase for 30 minutes followed by diaminobenzidine (DAB) substrate chromogen solution for 10 minutes. Monoclonal mouse IgG1 antibody was used as negative control. In each IHC staining run, a control slide provided by EGFR pharmDx ™ kit with IHC 0 and 2+ was also included. The tumor sections from PDECX samples were manually treated with EGFR antibody (Cell Signaling Technology, Cat#4267) followed by biotinylated secondary antibody and the DAB chromogen.

EGFR gene copy number (GCN) by qPCR
Genomic DNA was extracted from frozen tumor samples or FFPE tumor sections using QIAamp Mini kit (Qiagen, Valencia, CA) according to the manufacturer's instructions. qPCR was carried out in a 20 µL reaction mixture containing genomic DNA, primers, and SYBR Premix Ex Taq II (TaKaRa, Cat#RR820A) (Supplementary Table 3). The primers for EGFR were 5′-GAATTCGGATGCAGAGCTTC-3′ for forward and 5′-GACATGCTGCGGTGTTTTC-3′ for reverse. The primers for internal control MTHFR (Methylene Tetrahydrofolate Reductase) were 5′-CCATCTTCCTGCTGCTGTAACTG-3′ for forward and 5′-GCCTTCTCTGCCAACTGTCC-3′ for reverse. The EGFR gene copy number was normalized to NCI-H441 cells. Samples with EGFR GCN ≥ 2.0 were further validated by EGFR FISH.

EGFR gene amplification by FISH
The tumor specimens with EGFR GCN ≥ 2.0 identified by qPCR assay were further validated by FISH. In brief, 4 µm sections from FFPE samples were deparaffinized and rehydrated. Specimens were heated in the pre-treatment solution followed by proteolytic digestion using Pepsin (Dako FISH Accessary Kit, K5799). Tissue sections applied with EGFR/CEP7 FISH probe (Vysis, Order NO. 01N35-020) were denatured at 73°C for 5 minutes and hybridized at 37˚C overnight followed by a stringent wash with saline-sodium citrate containing Tween-20 (Dako FISH Accessory Kit, K5799). Finally, the specimens were mounted with fluorescence mounting medium (Invitrogen, P36935) containing 4′,6 diamidino-2-phenylindole (DAPI). The sections were visualized in a fluorescence microscope (Olympus BX53). The enumeration of EGFR gene and chromosome 7 was conducted in 50 tumor nuclei in each section by two independent observers. In case of discordance between the two observers, a third observer was involved. EGFR gene with ≥ 15 copies in ≥ 10% of cells or a gene/chromosome ratio per cell of ≥ 2.0 in homogenously stained regions was determined as EGFR gene amplification [28].

Hot spot mutation detection of EGFR, PIK3CA, K-Ras and B-Raf genes
Hotspots mutations in exon 19, 20 and 21 of EGFR gene, exon 9 and 20 of PIK3CA gene, exon 2 and 3 of K-Ras gene and exon 11 and 15 of B-Raf gene were detected in tumor samples by the Genewiz Inc., using ABI3730XL sequence analyzer. The primers for detecting hot spot mutations in EGFR, PIK3CA, K-Ras and B-Raf are listed in Supplementary Table 4.

EGFR kinase inhibition assay
EGFR kinase inhibition was determined using the Z´-LYTE ™ kinase assay kit-Tyr 4 peptide (Invitrogen, PV3193). Briefly, EGFR kinase (Invitrogen, PV3872) was dissolved in a reaction buffer composed of 50 mM HEPES pH 7.5, 0.01% BRIJ-35, 10 mM MgCl 2 and 1 mM EGTA. The final 10 µL of the kinase reaction mixture consisted of 5 ng kinase, 2 µM substrate peptide Try4, ATP and the test compounds theliatinib, gefitinib and erlotinib. ATP (10,50,200, 500, 800 and 1000 µM) was finally added to the reaction mixture to initiate the enzymatic reaction. The final concentration of DMSO in the assay was 2%. The reaction mixture was incubated at 25°C for 60 minutes in a 384 well plate. Then, 5 µL of Development Reagent B was added per well and incubated for further 60 minutes at 25°C. The fluorescent signal was read at emission wavelengths, 445 nm and 520 nm after excitation at 400 nm in a Victor3 multi label reader (PerkinElmer). The

EGFR phosphorylation inhibition in tumor cells by DELFIA assay
EGF stimulated EGFR phosphorylation was determined in A431 human epidermoid carcinoma cells using the modified DELFIA assay. Briefly, A431 cells (1.3 × 10 4 cells/well) were seeded overnight in a 96 well plate in 100 µL DMEM with 10% FBS. Then, the culture medium was removed and the cells were starved in FBS-free DMEM medium at 5% CO 2 and 37°C for 24 h. Further, the cells were treated with 10 µL of the test compounds (theliatinib, gefitinib and erlotinib) at different concentrations (300~0.137 nM, 3 fold gradient dilution) for 60 minutes at 37°C and 5% CO 2 . The control cells were treated with 10 µL of FBS-free DMEM medium containing 5% DMSO (final concentration of DMSO was 0.5%). Cells were stimulated by 20 ng/mL recombinant human EGF (BIOSOURCE, PHG0064) for 45 minutes. Then, the cell supernatant was discarded and the cells were lysed with 100 µL DELFIA lysis buffer. The plates with cell lysates were kept at -80°C overnight. Then, the lysates were thawed on ice with gently mixing, and 20 µL of lysates were added into the assay plates (PerkinElmer, AAAND-0001), pre-coated with monoclonal anti-EGFR capture antibody (R&D, AF231), followed by incubation for 1h at room temperature. The phosphorylated EGFR (p-EGFR) in the lysates was detected with the detection mixture containing DELFIA Eu-N1 labeled antiphospho-tyrosine antibody PT66 (PerkinElmer, Eu-PT66) and DELFIA enhancement solution for 1h at room temperature. Fluorescence signals were detected at 620 nm emission and 340 nm excitation by Victor3 multi label reader (PerkinElmer). All tested concentrations of theliatinib, gefitinib and erlotinib were repeated in duplicated wells.

Cell survival assay
A431 cells (1 × 10 4 cells/well) in exponential phase were seeded in duplicates in DMEM containing 10% FBS and incubated at 37°C and 5% CO 2 overnight. Then, 10 µL of test compounds (theliatinib, gefitinib and erlotinib) at tested concentrations (10~0.005 µM, 3 fold gradient dilution) were added into each well with the final concentration of DMSO at 0.5%. The cells were incubated for 48 h followed by further incubation for 1h after adding 10 µL/well CCK-8 solution (Dojindo, CK-04-13). Cell survival was determined by measuring the optical density at 450 nm using Labsystems Multiskan K3 (Thermo Fisher Scientific Inc.).