Discussion
Mouse xenografts derived from prolonged in vitro cultured cells have been the standard toolkit for cancer biologists for decades; however, the high failure rate of compounds entering clinical trials indicates a low clinical predictive power of traditional tumor model systems.
Compared with cell line derived xenografts, patient derived tumor xenografts models closely retain the histopathologic, genetic and phenotypic features of patients’ original tumors, and could better predict drug efficacy in humans [
8‐
10]. In the last few years, PDX models have been widely established in various tumor, including gastric cancer [
9,
11‐
14]. PDX models are now becoming a valuable tool for evaluating new anticancer drugs before the initiation of clinical trials. The main purpose of this study was to establish and characterize a panel of PDX GC models, mimicking the genetic alteration of gastric cancer patients and further use them as a platform for the evaluation of potential targeted medicine.
Recently, Choi et al. reported the genomic fidelities of the gastric PDX systems and identified factors correlated with engraftment success of PDX tumors. They found the tumor cell percentages in the implanted tissues were correlated with higher success rates [
28]. Zhang et al. established 32 PDGCX and the genetic characters of which were confirmed to be consistent with the original tumors [
29]. Here, a panel of 9 PDX models were established with a success rate of 28.1%, which is similar to previously reported data [
8]. When comparing the success rates between the first, second and third generations, we found the first implantation is a critical step in the process, while the success rate after the third generation nearly reached 100%.No significant correlation was observed between the engraftment achievement ratio and patients’ clinical feature.
However, a recent study revealed that prior chemotherapy may reduce the engraftment achievement ratio [
30]. Indeed, patients who showed complete response after pre-surgery chemotherapy were excluded from further implantation in our study. A vital question of the utility of PDX models is whether the passage of tumors in experimental mice alters their histopathological and molecular features. As shown above, only one model with cMet IHC 3+/FISH(+) changed to cMet IHC 0/FISH(-) (Additional file
2: Figure S1); the concordance rate of Her2,cMet and FGFR2 alteration between patients’ primary tumors and xenografts was 92.9% (Table
3). We speculate that tumor heterogeneity may be one of the possible cause of discordance between PDX models and donors. Previous studies have revealed that established PDX models are biologically stable when passaged in mice, in terms of global gene-expression patterns, mutational status, and drug responsiveness [
8,
10]. Therefore, PDX models, mimicking the histopathological and molecular features of patients, constitute a superior tool for the preclinical study of new emerging targeted therapies.
Approving trastuzumab for HER2-positive GC patients represents a milestone in gastric cancer targeted therapy [
15]. Recently, cMet and FGFR2, two other members of the RTK family, have been intensely investigated in gastric cancer. A high–resolution genomic analysis of a large cohort of gastric primary tumors revealed that approximately 4% and 9% gastric patients harbor cMet and FGFR2 amplification, respectively. The prevalence rates of cMet and FGFR2 alterations in Chinese gastric cancer patients have rarely been reported. Liu et al. reported that nearly 6.1% and 5.1% gastric cancer specimens harbor cMet or FGFR2 amplification in a cohort of Chinese gastric cancer, respectively. Here, we simultaneously studied the status of Her2, cMet and FGFR2 amplification in a cohort of Chinese GC patients and a panel of PDX models. Our data showed cMet overexpression and amplification in 19.6% and 4.3% of GC patients, respectively. Survival analysis revealed that patients with cMet protein IHC3+ or gene amplification have a poorer survival rate, in agreement with a recent study [
24]. In addition, FGFR2 amplification was harboured in 3.7% GC patients. FGFR2 amplification tended to indicate a lower survival rate, although a statistical significance was not obtained. A recent international multi-center study reported that FGFR2 amplification is associated with poor prognosis in gastric cancer. Her2 positive rate was 10.4%, which was consistent with previously reported data [
31].
Based on the above findings, we confirm that a significant proportion of gastric cancer patients harbour cMet and FGFR2 alterations, and may benefit from cMet and FGFR2 targeted therapies. A phase I study showed that patients with cMet-amplified gastroesophageal cancer treated with crizotinib experience significant tumor shrinkage [
32]. In addition, a preclinical study revealed that AZD4547 exerts marked antitumor effects on GC xenografts carrying FGFR2 gene amplification [
25]. Therefore, cMet or FGFR2 gene amplification seems to be a potential predictive biomarker for drug sensitivity; however, there is a possible correlation between gene amplification and protein expression as well as possible inconsistencies between IHC data and gene amplifications in terms of tumor take rates, growth kinetics and/or sensitivities to the respective treatment regimens. To further evaluate the translational significance of the above findings, the antitumor efficacy of crizotinib and AZD4547 was assessed on the panel of PDX models generated. As shown above, PDX models with cMet and FGFR2 amplification were highly sensitive to crizotinib, while others showed minimal or even no response. Interestingly, we found that combined treatment of crizotinib and AZD4547 enhances the antitumor effect of AZD4547 in G03 (FGFR2-amplifed and cMet non-amplified but IHC(2+)). The in vitro data showed obvious inhibition of Erk activation by crizotinib in the G03 derived cells, however, crizotinib monotherapy in vivo showed no significant tumor growth inhibition compared with the control group. There might be some mechanism causing this discrepancy which need further investigation. A recent study also demonstrated that activation of several receptor tyrosine kinases (RTKs), including EGFR, HER3 and MET, contributes to AZD4547 hyposensitivity in
FGFR2 amplified GC cells, and the rescue effect was abrogated by inhibiting these RTKs with their targeted tyrosine kinase inhibitors (TKIs) [
33]. Another study demonstrated that FGFR is one of the combinatorial targets to overcome resistance to cMet-targeted therapy in gastric cancer [
34]. The underlying mechanisms for the enhanced antitumor effect by combined treatment of crizotinib and AZD4547 in G03 is still unknown. By using the G03 xenograft derived cells, in vitro assay showed that a combination treatment of crizotinib and AZD4547 led to synergetic inhibition of MAPK/ERK pathway. Further biochemistry study on the GC cell lines with different status of cMet or FGFR2 amplification showed that the synergetic effect were obtained only in cMet or FGFR2 amplified cells, we speculated that co-targeting cMet and FGFR2 may exhibit a synergetic tumor inhibition through MAPK/ERK pathway. We observed the trans-phosphorylation of MET and FGFR2, however, the trans-phosphorylation were not consistent in the four cell lines(data not shown). The synergistic effect of the combo treatment of the crizotinib and FGFR2 inhibitor at the level of ERK phosphorylation is consistent in all the four different cell lines except the AGS cells which is negative for both receptor expression. We believe that the molecular mechanism underlying the synergistic effect of concomitant inhibition of the two parallel pathways, is more like to involve the downstream effectors of MET and FGFR2, but not the transphosphorylation of the two parallel receptors. Further studies are needed to explore the crosstalk between cMet and FGFR2 signaling pathway. Co-targeting cMet and FGFR2 may be a promising strategy for gastric cancer patients with amplification of cMet or FGFR2.