Development of preclinical models for a precision medicine pipeline
In order to implement a precision medicine strategy for the treatment of metastatic CRC, a CRC precision medicine pipeline was created. We first developed patient derived models of cancer including low passage cell lines and patient derived xenografts (PDX) for patients undergoing resection of their CRC liver metastasis or primary colon at Duke University under an Institutional Review Board (IRB)- and Institutional Animal Care and Use Committee (IACUC)-approved protocol. For each patient, matching cell lines and PDXs were developed as previously described [11, 12]. CRC057, CRC119, CRC240, CRC247 and 15–496 were derived from CRC liver metastasis, and 16–159 was derived from a primary colon cancer. Patient demographics are described in Fig. 1A, and Fig. 1B shows the histological features of PDXs (I-VI) and cell lines (VII-XII).
High-throughput drug screening in vitro and in vivo validation of cytotoxic chemotherapy agents
As the first step in identifying potential therapeutic targets, we performed a series of in vitro high-throughput drug screens using our patient-derived cell lines. The drug screen consisted of 119 FDA-approved small molecule inhibitors, and we first analyzed responses to commonly-used cytotoxic chemotherapeutic agents. In general, our CRC cell lines appear to be sensitive to anthracyclines (doxorubicin and epirubicin) and vinca alkaloids (vincristine and vinblastine) and resistant to platinum agents (cisplatin and carboplatin) and alkylating agents (ifosfamide and cyclophosphamide) (Fig. 2A). We next analyzed the response to standard-of-care cytotoxic agents used in the treatment of metastatic CRC, including oxaliplatin, irinotecan and 5-fluorouracil. CRC057 and 15–496 were found to be relatively sensitive to oxaliplatin, while CRC119, CRC240, CRC247 and 16–159 were found to be resistant (Fig. 2A). In contrast, CRC119 and 16–159 were relatively sensitive to irinotecan, while CRC057, CRC240, CRC247 and 15–496 were resistant (Fig. 2A).
To validate our in vitro screening results, we performed in vivo validation on matched PDX tumors. Mice at 10 weeks of age (~ 25 gram) were divided into 2 groups (control and treatment, consisting of 5 mice/group) and treated with oxaliplatin (10 mg/kg) and irinotecan (10 mg/kg) intraperitoneally five times a week. Consistent with our in vitro data, the CRC119 PDX tumor was sensitive to irinotecan (2-way ANOVA, p = 0.0002) and resistant to oxaliplatin treatment. Similarly, as predicted by our in vitro drug screen CRC240 PDX was resistant to both chemotherapeutic agents (Fig. 2B). No significant adverse events were seen. Finally, as previously described, 16–159 PDX was sensitive to irinotecan and resistant to oxaliplatin [14]. Together, these studies indicated that our screening and validation platform enabled rapid analysis of sensitivity and resistance to standard-of-care agents.
High-throughput drug screening identifies ponatinib as a novel therapeutic target
Next, to identify novel targeted agents to treat metastatic CRC, we mined our drug screen data for targeted therapies for which one or more cell lines were inhibited by > 50%. Interestingly, only a limited number of targeted therapeutic agents inhibited cell growth in vitro, including dabrafenib, trametinib, and ponatinib. Among these, ponatinib inhibited growth of 4/6 cell lines at > 50% (Fig. 3A). To further characterize the effect of ponatinib in CRC, drug sensitivity studies were performed on the cell lines to determine the IC50 of ponatinib. The estimated IC50 values were 0.7 µM for CRC057, 1.1 µM for CRC119 and 1.1 µM for CRC240 (Fig. 3B). To validate the efficacy of ponatinib inhibition in vivo, we used matched PDX models of the cell lines. CRC119, CRC240 and CRC057 were injected subcutaneously in the flanks of the mice (at 10 weeks of age and ~ 25 gram) as previously described [11, 12], were divided into 2 groups (control and treatment, consisting of 5 mice/group) and treated with 30 mg/kg oral ponatinib five times a week. Consistent with the in vitro results, CRC057, CRC119 and CRC240 were all found to be sensitive to ponatinib (2-way ANOVA, P < 0.0001) (Fig. 3C). No significant adverse events were seen. Together, these in vitro and in vivo studies indicate that using matched patient-derived cell lines and PDXs can provide a robust screening and in vivo validation platform to identify personalized therapies to treat CRC.
Targeting ponatib in CRC
Our personalized medicine pipeline identified ponatinib as a potentially effective agent to treat CRC. Ponatinib is a multi-kinase inhibitor that targets the fibroblast growth factor receptor (FGFR), platelet derived growth factor receptor (PDGFR), vascular endothelial growth factor receptor (VEGFR), SRC, and ABL [15]. As ponatinib is a multi-kinase inhibitor, in order to identify the main target of ponatinib in our cell lines, we re-analyzed our screen data to identify other agents that target similar pathways as ponatinib. We identified three other agents, including axitinib, a VEGFR and PDGFR inhibitor; sunitinib, a PDGFR inhibitor; and dasatinib, a SRC and ABL inhibitor [16–20]. Remarkably, CRC057, CRC119 and CRC240 were all resistant to axitinib, sunitinib, and dasatinib, suggesting that the mechanism of action of ponatinib in these three cell lines might be through inhibiting a common signaling pathway or pathways (Fig. 4A).
To identify the common pathway or pathways that drive CRC growth in these patient-derived cell lines, we next screened CRC057, CRC119 and CRC240 with specific inhibitors of ponatinib’s targets, including ABL, FGFR, PDGFR, SRC and VEGFR. However, all three cell lines were found to be resistant to the specific inhibitors (Fig. 4B), suggesting that ponatinib was showing its antitumor activity by targeting multiple signaling pathways.
Pre-treatment analysis of the cell lines with p-FGFR, p-VEGF, p-PDGFR, p-SRC and p-ABL antibodies showed that all three cell lines expressed a variety of intracellular tyrosine kinase receptors (Fig. 4C). The IC50 dose of ponatinib inhibited p-FGFR activity in CRC119 and CRC240; p-VEGFR activity in CRC057 and CRC119; p-PDGFR activity in CRC 119; and p-SRC activity in all three cell lines (Fig. 4C), suggesting that different signaling pathways are implicated in determining sensitivity to ponatinib. To further verify the cell growth inhibition by ponatinib, we next focused on downstream signaling pathways of FGFR, VEGFR, PDGFR and SRC including the PI3K/AKT/mTOR, RAS/RAF/MEK/ERK, and STAT pathways. Pre- and post- treatment western blot analysis showed that STAT pathways were consistently targeted in all three cell lines. In contrast, p-AKT increased in response to ponatinib treatment, suggesting that the PI3K/AKT/mTOR pathway was activated in CRC119. Similarly, p-ERK expression increased in CRC057 and CRC240, suggesting that the RAS/RAF/MEK/ERK pathways were activated in response to ponatinib treatment (Fig. 4D).
Determining the molecular predictor of sensitivity to ponatinib
To better understand the potential underlying genetic determinants of our patient-derived models of cancer to ponatinib, we performed RNA-Seq on the cell lines. Specifically, we found two mutations in the FGFR1 open reading frame, three mutations in the FGFR2 open reading frame and three mutations in the FGFR4 open reading frame in our six cell lines. We observed A254V and S429fs mutations in FGFR1 in CRC119 and 16–159, respectively. Similarly, we found P470L and W76R mutations in FGFR2 in CRC119 and CRC240, respectively, and M71T mutation in FGFR2 in 16–159. None of the cell lines were found to have mutation in FGFR3. In all six cell lines, we observed a P136L mutation in FGFR4. G388R mutation was also observed in FGFR4 in the CRC057, CRC240 and 16–159 cell lines. In addition to this, V10I mutation was found in FGFR4 in only the 16–159 cell line (Table 1). A complete list of all mutations found in FGFR is listed in Table S1. While all lines had synonymous, intronic, or upstream/downstream mutations in SRC, CRC119 also had mutations in the 5’ and 3’ untranslated regions, and both CRC119 and CRC247 had three mutations within the 3’ untranslated region as well as a 5’ splice site mutation within exon 2 of SRC. No mutations were found in VEGFR.