Prognostic and predictive role of EGFR pathway alterations in biliary cancer patients treated with chemotherapy and anti-EGFR

The association of anti-EGFR to gemcitabine and oxaliplatin (GEMOX) chemotherapy did not improve survival in biliary tract carcinoma (BTC) patients. Multiple mechanisms might be involved in the resistance to anti-EGFR. Here, we explored the mutation profile of EGFR extracellular domain (ECD), of tyrosine kinase domain (TKD), and its amplification status. EGFR mutational status of exons 12, 18–21 was analyzed in 57 tumors by Sanger sequencing. EGFR amplification was evaluated in 37 tumors by Fluorescent In Situ Hybridization (FISH). Kaplan-Meier curves were calculated using the log-rank test. Six patients had mutations in exon 12 of EGFR ECD and 7 in EGFR TKD. Neither EGFR ECD nor TKD mutations affected progression free survival (PFS) or overall survival (OS) in the entire population. In the panitumumab plus GEMOX (P-GEMOX) arm, ECD mutated patients had a worse OS, while EGFR TKD mutated patients had a trend towards shorter PFS and OS. Overall, the presence of mutations in EGFR or in its transducers did not affect PFS or OS, while the extrahepatic cholangiocarcinoma (ECC) mutated patients had a worse prognosis compared to WT. Nineteen out of 37 tumors were EGFR amplified, but the amplification did not correlate with survival. ECC EGFR amplified patients had improved OS, whereas the amplification significantly correlated with poor PFS (p = 0.03) in gallbladder carcinoma patients. The high molecular heterogeneity is a predominant feature of BTC: the alterations found in this work seem to have a prognostic impact rather than a predictive role towards anti-EGFR therapy.

Introduction Different strategies aimed at inhibiting EGFR with small molecules (erlotinib and gefitinib) or with monoclonal antibodies (cetuximab and panitumumab) have been developed over the years in many cancer types [1][2][3][4][5][6]. Panitumumab (Vectibix, Amgen), a fully human antibody directed against EGFR, was initially approved in KRAS wild type (WT) metastatic colorectal cancer (mCRC) patients refractory to previous chemotherapy [7,8]. In biliary tract carcinoma (BTC), preclinical evidence of antitumor activity [9] and the lack of compelling therapies suggested that the combination of standard chemotherapy and EGFR inhibitors could be an attractive option to improve patient outcome [10,11]. The randomized, open-label, phase II Vecti-BIL trial compared the efficacy of gemcitabine and oxaliplatin (GEMOX) chemotherapy with or without panitumumab (P) in KRAS WT advanced BTC (Clinical Gov Identifier NCT01389414). The study, which enrolled and stratified intrahepatic cholangiocarcinoma (ICC) and extrahepatic cholangiocarcinoma (ECC) including gallbladder carcinoma (GBC), revealed that the addition of panitumumab to the standard chemotherapy did not improve progression free survival (PFS), which was 5.3 months in experimental arm and 4.4 months in control arm. No differences were observed in overall survival (OS), being of 9.9 with GEMOX and 10.2 months with P-GEMOX [12]. Hence, we concluded that KRAS WT status was not sufficient to select patients who can achieve tumor response to anti-EGFR therapies. Over the years, the phenomenon of resistance to anti-EGFR therapies has been deeply studied, in particular in CRC. The panel of potential drivers of resistance was expanded and KRAS exons 3-4, in addition to exon 2, NRAS, PIK3CA, and BRAF analyses were introduced in the clinical practice [13][14][15]. Thus, we retrospectively analyzed the mutational status of these genes in patients enrolled in the Vecti-BIL study and we found that the presence of these mutations did not affect the response to treatments. Recently, new mechanisms of resistance to anti-EGFR antibodies have been recognized in mutations of EGFR exon 12 of the extracellular domain (ECD); in CRC it was demonstrated that they prevented the correct binding of anti-EGFR, reducing their activity [16]. Moreover, even if controversial, EGFR amplification seemed to be a predictive marker of prognosis and response to the anti-EGFR therapies in CRC [15,17]. EGFR amplification was also described in BTC [18,19], but its prognostic role is unknown. Overall, in both arms of the Vecti-BIL trial, there was a broad range of PFS and OS: in the experimental arm, PFS ranged from 1.1 to 21.3 months and OS from 2.7 to 34.9 months, while in the control arm PFS ranged between 1.1 to 15.4 months, and OS between 1.1 and 31.7 months.
Here, we extended the molecular analyses to the EGFR ECD and TKD mutation profiling, and to the EGFR amplification status to explain these differences, and to correlate them to the arm of treatment.

Patients
The Vecti-BIL trial (ClinicalTrials.gov Identifier: NCT01389414) enrolled 89 BTC patients selected for the absence of KRAS exon 2 mutations. Forty-five patients were assigned to receive GEMOX in association with panitumumab (ARM-A) and 44 patients GEMOX alone (ARM-B). All patients enrolled in the study have signed the Independent Ethical Committees (IEC) informed consent, which provided the authorization to perform molecular analyses of EGFR status and of its principle transducers on archival tumor tissues. The study was performed in accordance with the Declaration of Helsinki and was approved by the "Comitato Etico San Luigi Gonzaga di Orbassano". Of 89 patients enrolled in the study, 57 had formalin fixed paraffin embedded tissues (FFPE blocks or tissues slides) available for molecular analyses.

Mutational analysis
For all the specimens, the tumor areas, identified by a pathologist, were scraped from two tissue slides (5 micron thick). DNA was then extracted by using Qiamp DNA FFPE mini kit (Qiagen s. r.l. Milano, Italy) following manufacturers' instruction. Briefly, For DNA extraction, tissues were deparaffinized by xylene, rehydrated by ethanol, lysed with appropriate buffer and proteinase K at 56˚C for one hour and subsequently for another hour at 90˚C. Samples were then transferred on columns, washed with different buffers and then eluted in nuclease free water. DNA quantity was evaluated by Nanodrop (ThermoScientific Italia, Monza, Italy). The integrity of DNA was tested by a specific PCR for the housekeeping gene GAPDH. The EGFR exon 12 was amplified using nested PCR [16]. Exons 18-21 of EGFR were amplified by nested PCR using primers and conditions already described [20]. PCR products were purified using The Wizard SV Gel and PCR Clean-Up System kit (Promega Italia, Milano Italy). Each exon was sequenced using the BigDye Terminator Cycle sequence following the PE Applied Biosystem strategy and Applied Biosystems ABI PRISM3100 DNA Sequencer (Applied Biosystem, Forster City, CA). Sense and antisense sequences were obtained by using forward and reverse internal primers, respectively. All mutations were confirmed by performing two independent PCR amplifications.

Fluorescent in situ hybridization (FISH) analysis
EGFR gene status evaluation was performed by FISH on 3-μm thick tissue sections. Dual-color FISH assay was performed using LSI EGFR/CEP7 probes (Vysis, Abbott Laboratories, USA) following the manufacturer's instructions already described [21]. Fluorescent in situ hybridization signals were evaluated with a Zeiss Axioscope (Carl Zeiss SPA, Italy). Tumors with !4 copies of EGFR or gene amplification in !40% of cells were classified as FISH+.

Statistical analysis
KaplanMeier survival curves were calculated using the Graph Pad Prism 6 software. Molecular findings (mutational status of exon 12, 18-21 of EGFR, its transducers KRAS, NRAS, BRAF, PIK3CA,and FISH) were correlated with PFS an OS in the entire study population and separately in the two arms of treatment. The Cox proportional hazards regression model was used to identify prognostic factors. PFS and OS were calculated using Kaplan-Meier estimation and examined by log-rank test. The association between the radiological response and molecular findings was analyzed using the two tailed Chi-Square test (C.I. 95%). A p-value less than 0.05 was considered as statistically significant.

Mutational status of EGFR extracellular domain (ECD) and tyrosine kinase domain (TKD)
Six out of 57 patients (10.5%) displayed mutations in EGFR ECD (exon 12). Four patients belonged to the ARM-A and two patients to the ARM-B.  Table summarizes the mutations found, some of which have already been described in literature. In our previous work we described 3 patients with BRAF V600E mutations, 2 with NRAS mutations (A146S and Q61R), and 2 with PIK3CA E545K mutations [12]. One patient, the 71772, which harbored a PIK3CA mutation, was also mutated in EGFR ECD. Overall, we found that 18 out of 57 patients (31.6%) had mutations in EGFR or in its transducers. The mutations were homogeneously distributed between the two arms of treatment (8 patients in P-GEMOX arm, and 10 patients in GEMOX arm). Nine out of 18 mutated tumors (50%) were ICC, while 3 were ECC (16.7%), and 6 were GBC (33.3%).

EGFR ECD mutations are associated with a worsening survival in subgroups
Analyzing all the 57 patients, independently from the treatment, we found that the presence of EGFR ECD mutations did not significantly affect PFS (p = 0.81) and OS (p = 0.44) (Fig 3A  and 3B). In P-GEMOX arm, patients with EGFR ECD mutations had worse OS, even if not significantly (p = 0.1), while the PFS was not affected (p = 0.6) (Fig 3C and 3D). In the ICC subgroup, EGFR ECD mutated patients had impaired OS compared to WT patients (p = 0.14), but similar PFS (p = 0.92) (Fig 3E and 3F). Due to the small number of mutated patients, the analysis was not conducted for the GEMOX arm and for the other site subgroups.

EGFR TKD mutations have a predictive negative role in P-GEMOX treated patients
The survival analysis of EGFR TKD mutated versus WT patients in the entire cohort of the study did not reveal any difference in terms of PFS (p = 0.75) and OS (p = 1.0) (Fig 4A and   4B). Even though, by analyzing P-GEMOX treated patients, we found that EGFR TKD mutations caused a trend towards decrease of PFS (p = 0.06) and OS (p = 0.06) (Fig 4C and 4D). On the contrary, EGFR TKD mutated patients treated with GEMOX had a better PFS (p = 0.2), while no differences in OS was observed (p = 0.9) (Fig 4E and 4F). For GBC patients, the presence of EGFR TKD mutations did not affect the PFS or OS (data not shown). Due to

Wild-type status of EGFR and its transducers is not associated with efficacy of panitumumab
We analyzed if the presence of any mutation affecting EGFR and its main transducers (KRAS, NRAS, BRAF, PIK3CA) could influence survival in BTC patients. As shown in Fig 5A and 5B, PFS and OS were not impaired by the presence of mutations in EGFR and/or its transducers in the entire cohort (p = 0.49 and p = 0.93, respectively) as well as considering the two arms of treatment (Fig 5C-5F (Fig 6A and 6B). Similar results were obtained analyzing the two arms of treatment separately. Fig 6C and 6D shows the PFS and OS survival curves of P-GEMOX treated patients (p = 0.97 and p = 0.45, respectively) and Fig 6E and 6F represents the PFS and OS survival curves of GEMOX-treated patients (p = 0.68 and p = 0.46, respectively). Further, we stratified patients according to the site subgroup; as shown in  EGFR FISH+ patients had a better PFS (p = 0.07) and OS (p = 0.08) compared to FISHpatients (Fig 7C and 7D), even if not significantly. On the contrary, in GBC patients, FISH+ was associated to a PFS significantly worse (p = 0.03) and an OS slightly reduced (p = 0.23, Fig 7E and 7F). Finally, in ICC, FISH+ did not affect PFS (p = 0.54) and OS (p = 0.33, Fig 7A  and 7B).

Discussion
The role of EGFR in cholangiocarcinogenesis is well documented. Its high expression as well as the low percentage of mutations in its downstream effectors suggested that it could be a suitable target for molecular therapy. Even though, these premises did not reflect the success obtained in mCRC and now the scientific community agreed about the marginal role of anti-EGFR therapies in BTC. The lack of efficacy obtained by the addition of panitumumab to standard chemotherapy demonstrated in the Vecti-BIL trial confirmed previous results available in literature [22][23][24]. The difference in median PFS and OS obtained in the two arms of treatment were not statistically different, but each arm displayed a broad range of PFS and OS, suggesting that both prognostic and predictive factors may have affected the results. The predefined collection of tumor tissues of patients enrolled in the prospective trial allowed a successive investigation of the role of EGFR mutational status, both in ECD and TKD, and the EGFR amplification. We demonstrated that 6 patients harbored mutations in EGFR ECD, none of them previously described. Four mutated patients were enrolled in the P-GEMOX arm; their PFS ranged from 2 to 9 months and the OS between 5.7 to 15.2 months. The survival curves showed that EGFR ECD mutations may have a predictive negative role in patients treated with panitumumab, but considering the worst survival of ICC patients having EGFR ECD mutations as compared to WT, the impact of EGFR ECD mutations may be rather prognostic, at least in this site subgroup.
Seven out of 57 patients harbored mutations in the exons 18-21 of EGFR, coding for the TKD: of them, two were novel mutations, while the others were already described in literature in other cancer types [25][26][27][28]. In particular, the V786M is described as high sensitive to gefitinib treatment in bronchioloalveolar adenocarcinoma [29]. Interestingly, none of EGFR TKD mutated patients had a tumor response, even if the correlation is not statistically significant (p = 0.07, data not shown). Even if not affecting the survival in the entire cohort, EGFR TKD mutated patients treated with P-GEMOX had a worse prognosis compared to WT patients.
Although the biological impact of each of the molecular alterations investigated may be different, we have considered all the mutations affecting EGFR, NRAS, KRAS, BRAF, and PIK3CA as potentially involved in the resistance to anti-EGFR therapies in BTC. For this reason, we evaluated the effect of treatments in patients without any of the mutations, referred to as "quintuple negative" population. This population does not seem to have different prognosis, and received similar benefit with both GEMOX and P-GEMOX as compared with patients with any of the mutations. By subgroup analysis, ECC quintuple negative patients have a trend towards a better survival than mutated patients, whereas in other site subgroups, mutation analysis revealed no prognostic impact.
Thirty-seven patients had tumor sample available for EGFR amplification analysis by FISH. More than half of the tested samples resulted FISH+. EGFR was amplified in all site subgroups (ICC, ECC, and GBC). Overall, EGFR amplification did not correlate with PFS or OS. According to the site of the disease, we found that EGFR amplification significantly correlated with poor PFS in GBC patients, while in ECC there was a trend towards a better survival of FISH+. Of note, 18 out of 19 EGFR amplified patients had lymph-node infiltration, suggesting a more advanced disease. This finding is enforced by a work on gastroesophageal cancers, in which EGFR amplification seemed to correlate with worse prognosis and lymph node metastasis [30]. However, the role of EGFR amplification as prognostic biomarker remains controversial. In two studies in mCRC, authors demonstrated that EGFR amplification was associated with longer PFS and OS [31,32]; otherwise, in another work, EGFR amplification was not associated to prognosis [17].
The EGFR protein expression was not evaluated in our case series; previous studies demonstrated that EGFR is expressed in about 60% of BTC, ranging from 54% to 65%, and that the EGFR positivity does not affect the clinical outcome, independently from the treatment and site of origin [9,23,24,33]. In contrast, Yang and collaborators evaluated the EGFR expression in 175 BTC, demonstrating that EGFR expression is a negative prognostic factor in ICC, but not in ECC [19,34]. Therefore, the role of EGFR expression remains controversial, mainly due to the high heterogeneity of tumors analyzed.
The Vecti-BIL study included all KRAS exon 2 WT patients, but no other molecular stratification was performed at the time of enrollment, yielding to a highly heterogeneous population both in terms of site subgroups and of molecular alterations. Even if patients were selected for the absence of KRAS mutation, the antitumor effect of panitumumab was demonstrated neither in the overall population, nor in the molecular subgroups in which EGFR pathway was deemed suitable for effective inhibition. Most of the genetic alterations found seem to have a prognostic significance rather than a predictive role towards the anti-EGFR therapy. A high molecular heterogeneity seems to be a predominant feature of BTC, preventing the identification of well-defined subgroup of patients, which could benefit from targeted therapies. Due to the small cohort of patients analyzed, we can only speculate about prognostic or predictive role of EGFR status; nevertheless, the meta-analysis conducted by Chen and collaborators [35], which pooled four studies to examine the efficacy of anti-EGFR therapies combined to standard chemotherapy, revealed promising results of EGFR-targeted therapy in increasing the survival rate (PFS) of advanced BTC patients. These data keep the door open to the anti-EGFR therapies, provided that an accurate selection of patients is carried out. Moreover, it is highly recommended the identification of other biomarkers able to select a priori subgroups of patients who could benefit from these treatments. One of the suggested predictive biomarkers of response/resistance to anti-EGFR therapies is the expression of EGFR ligands; Luraghi and collaborators demonstrated that NRG1 is a marker of resistance to cetuximab, while EREG and AREG expression does not confer resistance to anti-EGFR therapies in mCRC xenospheres [36]. Further, another proposed mechanism is the presence of a specific oncogenic variant of EGFR, the EGFRvIII, which is constitutively active, without the requirement of ligand binding; in fact, it has been demonstrated that anti-EGFR antibodies are less effective in the presence of this variant in glioblastoma multiforme [37]. EGFR promoter methylation could be involved in the lack of response to anti-EGFR antibodies; in the work of Scartozzi, it has been demonstrated that mCRC patients treated with cetuximab displaying EGFR promoter methylation are less responder compared to unmethylated patients, with a significant worse prognosis [38]. This is only an overview of the main mechanisms involved in pathogenesis and drug resistance in BTC, evidencing the complexity of EGFR pathway and the major points of failure of anti-EGFR therapies. In fact, EGFR activation could be sustained also by a defective ubiquitination and subsequent internalization, which can cause a prolonged activation as well as a transactivation through COX-2/PgE2 signaling, as shown by Yoon and collaborators in BTC cell lines [39].
Since BTCs are rare disease, great efforts have been made in recent years in conducting prospective studies to investigate the efficacy of new therapies. Standard of care has been identified in advanced disease [40,41] and a variety of targeted therapies has been tested so far [11,12,[42][43][44]. The inclusion of patients with primary cancers arising in different part of the biliary tree is a compromise generally accepted in clinical trials. However, with the advent of new technologies as next-generation sequencing, it has become evident that mutational profiling of BTC varies within the tumor location. As a consequence, a neglected genetic heterogeneity can result in findings difficult to interpret. Thus, it is desirable that the genomic landscape of BTC will take the place of pathologic criteria based on the site of origin in future trials [45].

S1 Fig. Kaplan-Meier survival curves in EGFR or its transducers mutated vs WT patients.
A-B) association between the presence of any mutations and PFS and OS, respectively, in ICC patients. C-D) association between the presence of any mutation and PFS and OS, respectively, in ECC patients. E-F) association between the presence of any mutation and PFS and OS, respectively, in GBC patients. (TIF) S1