The landscape of targeted therapies for cholangiocarcinoma: current status and emerging targets

Cholangiocarcinoma (CCA) is a relatively rare malignancy that arises from the epithelial cells of the intrahepatic, perihilar and distal biliary tree. Intrahepatic CCA (ICC) represents the second most common primary liver cancer, after hepatocellular cancer. Two-thirds of the patients with ICC present with locally advanced or metastatic disease. Despite standard treatment with gemcitabine and cisplatin, prognosis remains dismal with a median survival of less than one year. Several biological plausibilities can account for its poor clinical outcomes. First, despite the advent of next generation and whole exome sequencing, no oncogenic addiction loops have been validated as clinically actionable targets. Second, the anatomical, pathological and molecular heterogeneity, and rarity of CCA confer an ongoing challenge of instituting adequately powered clinical trials. Last, most of the studies were not biomarker-driven, which may undermine the potential benefit of targeted therapy in distinct subpopulations carrying the unique molecular signature. Recent whole genome sequencing efforts have identified known mutations in genes such as epidermal growth factor receptor (EGFR), Kirsten rat sarcoma viral oncogene homolog (KRAS), v-raf murine sarcoma viral oncogene homolog (BRAF) and tumor protein p53 (TP53), novel mutations in isocitrate dehydrogenase (IDH), BRCA1-Associated Protein 1 (BAP1) and AT-rich interactive domain-containing protein 1A (ARID1A), and novel fusions such as fibroblast growth factor receptor 2 (FGFR2) and ROS proto-oncogene 1 (ROS1). In this review, we will discuss the evolving genetic landscape of CCA, with an in depth focus on novel fusions (e.g. FGFR2 and ROS1) and somatic mutations (e.g. IDH1/2), which are promising actionable molecular targets.


INTRODUCTION
Cholangiocarcinoma (CCA) comprises of malignancy arising from the intrahepatic, perihilar and distal biliary tree. Intrahepatic CCA (ICC) is the second most common primary hepatic malignancy, after hepatocellular carcinoma, and accounts for 10-20% of primary liver cancers [1,2]. The incidence and mortality rates of ICC have been rising worldwide in the past decade, whereas those of extrahepatic CCA (ECC) are either stable or decreasing [2]. In the Western countries, the annual incidence of ICC is 2.1 per 100,000 person years [3]. Chronic inflammation from liver fluke infestation, hepatitis B and C infections, primary sclerosing cholangitis and inflammatory bowel disease are the main risk factors of CCA [4]. Other less common etiologic factors include hepatolithiasis, cirrhosis, alcohol, smoking, fatty liver disease and cholelithiasis [1].
Only 10-15% of the patients with CCA are amenable to potentially curative surgery, as majority present at an advanced stage due to lack of effective screening strategies [5]. Despite resection, high recurrence rates of 50-60% persist, conferring a five-year overall survival (OS) of only 30% [5,6]. The high rate of relapse prompted a strong rationale for adjuvant therapies to improve survival. However, the available evidence remains conflicting as randomized adjuvant trials are still ongoing. A meta-analysis of 6,712 biliary tract cancer Review www.impactjournals.com/oncotarget (BTC) patients who received varying forms of adjuvant therapy (chemotherapy, radiotherapy, chemoradiotherapy) demonstrated no clear survival benefit with adjuvant treatment (Odds ratio (OR) 0.74, 95% Confidence interval (CI) 0.55-1.01; P = 0.06) [7]. Liver transplantation, though not considered as standard therapy for CCA, has also been explored in selected patients with early stage perihilar CCA, where complete resection is impossible due to vascular or biliary invasion. A meta-analysis of 605 CCA patients who underwent liver transplantation demonstrated a 5-year OS of 39%, with superior outcomes in those who underwent perioperative chemoradiotherapy (5-year OS 57%) [8].
Majority of the patients present at an advanced stage, with limited treatment options which include locoregional or systemic therapy. There has been a growing interest in various locoregional therapy modalities including transarterial chemoembolization, selective internal radiotherapy, external beam radiation or ablation in patients who present with liver-limited disease [9]. However, these therapies were evaluated in small retrospective series or single arm phase II trials, and thus limit generalizability. The current standard of care for first line treatment of unresectable CCA is the combination of gemcitabine and cisplatin, albeit with modest benefit [10]. The prognosis of patients with unresectable or metastatic CCA is universally poor, with a median OS of less than one year. The treatment complexity is further confounded by the presence of recurrent cholangitis or cholestasis, necessitating interventions for restoration of biliary drainage and long term antibiotics use, thus leading to delays in systemic treatment.
Notably, the conduct of phase III randomized controlled trials (RCTs) have been exceptionally challenging due to the rarity of CCA and its inherent anatomical, pathological and molecular heterogeneity. With the advent of whole genome sequencing, mutations in epidermal growth factor receptor (EGFR), Kirsten rat sarcoma viral oncogene homolog (KRAS), v-raf murine sarcoma viral oncogene homolog (BRAF) and tumor protein p53 (TP53) were unraveled. More recently, novel mutations in isocitrate dehydrogenase (IDH), BRCA1-Associated Protein 1 (BAP1) and AT-rich interactive domain-containing protein 1A (ARID1A), and novel fusions such as fibroblast growth factor receptor (FGFR2) and ROS proto-oncogene 1 (ROS1) were revealed. In this review, we will discuss the evolving genetic landscape and summarize the targeted therapies in CCA.

SYSTEMIC CHEMOTHERAPY
The standard of care for first line chemotherapy for advanced CCA is the combination of gemcitabine and cisplatin. The pivotal United Kingdom National Cancer Research Institute Advanced Biliary Cancer (ABC)-02 study reported superior survival with gemcitabine and cisplatin (GC), with a median OS of 11.7 months versus 8.1 months, and median progression free survival (PFS) of 8.0 months versus 5.0 months, when compared to gemcitabine alone [10]. Despite intensified evaluation of other chemotherapy combinations with fluorouracil, oxaliplatin or irinotecan, the improvement in survival has been marginal [11]. Currently, there is no standard secondline chemotherapy. In a systemic review of 761 patients, treatment with second-line chemotherapy attained a mean OS of 7.2 months (95% CI 6.2-8.2), PFS of 3.2 months (95% CI 2.7-3.7), response rate (RR) of 7.7% (95% CI 6.5-8.9) and disease control rate (DCR) of 49.5% (95% CI 41.4-57.7) [12]. However, these results need to be interpreted with caution. First, patients who receive second-line chemotherapy have better performance status, which may be associated with improved prognosis [13]. Second, only 15-25% of patients will be fit enough to receive second-line treatment [14]. Third, no RCTs have been included in this systemic review. Given the marginal advances with chemotherapy, emphasis has been shifted to molecularly targeted therapies, either as a single agent or in combination with chemotherapy.
Next generation sequencing (NGS) of 46 cancerrelated genes in 75 CC patients has highlighted anatomical variability in frequency of mutations [16]. Notably, it may be technically challenging to distinguish ICC and ECC based on pathology, and hence there may be inherent biases in these studies. The common genetic alterations in ICC include TP53 (30%), KRAS (24%), ARID1A (20%), IDH1 (18%) and MCL1 (16%), whereas for extrahepatic CCA, common aberrations include TP53 (45%), KRAS (40%), ERBB2 (20%), SMAD4 (25%), F-box/WD repeatcontaining protein 7 (FBXW7) (15%) and CDKN2A (15%). Furthermore, there were significant differences with regards to the prognostic significance of the above molecular markers, with TP53, KRAS and MTOR alterations predicting a worse prognosis in ICC, and BAP1, PBRM1 and chromatin modulating genes linked to a worse survival in ECC. A subsequent meta-analysis of 4,458 patients with the study of 102 individual markers revealed that genetic alterations of HER2 and TP53 were more common in ECC, and BCL-2, EGFR, SMAD4, p16 and VEGF-A were more frequent in ICC [33]. Table 1 summarizes the molecular aberrations in CCA. In the following section, we will highlight the known molecular aberrations in conjunction with their targeted therapies. Figure 1 depicts the key signaling pathways in the pathogenesis of CCA, and novel targeted therapies in development in CCA.  [15,32] Abbreviations: NR, not reported

EGFR/HER2
The EGFR family comprise of ERBB1-4, with ERBB1 (EGFR) and ERBB2 (HER2) being frequently implicated in the multi-step carcinogenesis of CCA [15]. Binding of EGF-ligands to the receptors induce homodimerization or heterodimerization, which in turn activates downstream signaling pathways (MAPK, PI3K/ AKT/MTOR and STAT) that regulates cell differentiation, migration, angiogenesis and survival. EGFR overexpression occurs in 11-27% of ICC and 5-19% of ECC, and has been associated with tumor recurrence and worsened survival [15,17,18]. Majority (77-79%) of EGFR overexpression in BTCs exhibit copy number gain, with activating mutations in EGFR being extremely rare [15]. Although no mutations have been reported in HER2, HER2 overexpression has been noted in 0-2% of ICC and 5-20% of ECC [18]. Preclinical studies have demonstrated that overexpression of HER2 in transgenic mouse models and orthotopic transplantation BDEneu models enhance the development of CCA, and provided consistent evidence of the oncogenic potential of EGFR [34].
Despite the strong rationale of targeting EGFR in BTCs and early interesting results with single arm phase II trials suggesting the benefits of EGFR inhibitors either as single agents or in combination with chemotherapy (Table 2), four completed randomized studies have failed to confirm the benefits of targeting EGFR in advanced BTCs. The only phase III trial of 133 patients with BTCs demonstrated that the addition of erlotinib to gemcitabineoxaliplatin (GEMOX) significantly improved RR, but did not demonstrate any benefit in survival, with a median OS of 9.5 months in both arms [35]. However, subgroup analyses showed that for patients with CCA, the addition of erlotinib to chemotherapy significantly prolonged median PFS by 2.9 months [5.9 months vs. 3.0 months (HR 0.73, 95% CI 0.53-1.00; P = 0.049)]. In a phase II study, the addition of cetuximab, a chimeric anti-EGFR monoclonal antibody to GEMOX did not confer a survival benefit in patients with advanced BTCs [36]. The median PFS was 6.1 months for the GEMOX and cetuximab arm, compared to 5.5 months in the GEMOX alone arm, and the median OS was 11.0 months and 12.4 months, respectively. In another study, patients who were stratified by KRAS status, received GEMOX with or without cetuximab [37]. The addition of Cetuximab to GEMOX was associated with a trend in improvement in PFS (6.7 months vs. 4.1 months; P = 0.05), but not OS (10.6 months vs. 9.8 months; P = 0.91). In addition, KRAS mutation did not predict for benefit in survival. The addition of another EGFR antibody, panitumumab to gemcitabine/cisplatin based chemotherapy did not improve survival in patients with advanced BTCs [38]. Additional biomarker-driven trials will provide further insight as most of the studies were conducted in patients who were unselected for KRAS mutation status or other signatures implicated in predicting response to EGFR therapy.

VEGF
The most potent angiogenic factor in perpetuating tumor growth and metastasis is the vascular endothelial growth factor. VEGF overexpression was observed in 54% of ICC and 59% of ECC, and has been shown to promote metastasis, tumor recurrence and confer a worse prognosis [15,18].
The efficacy of VEGF inhibitors has been investigated in several trials ( Table 2). Bevacizumab has been combined with GEMOX, erlotinib or gemcitabine and capecitabine, yielding a PFS of 4-8 months and OS of 10-13 months [39][40][41]. Five trials have investigated sorafenib, a multikinase inhibitor against VEGFR-2, VEGFR-3, RAF, platelet derived growth factor receptor (PDGFR) and stem cell factor (KIT), and did not report any significant benefit in survival [42][43][44][45][46]. Other VEGF inhibitors such as sunitinib [47] and vandetanib (ZD6474) [48] yielded disappointing results. Recently, Valle and colleagues reported the results of ABC-03 trial in which the addition of cediranib, a potent oral VEGFR 1-3 inhibitor, was evaluated in combination of gemcitabine/ cisplatin in advanced BTCs in a randomized phase II trial [49]. Of the 124 patients enrolled (62 in each arm), the addition of cediranib improved the response rate (44% in the cediranib arm and 19% in the placebo arm, P = 0.004) but did not improve the median PFS (8.0 months in cediranib arm and 7.4 months in placebo arm, HR 0.93, P = 0.72) or OS (14.1 months in cediranib arm and 11.9 months in placebo arm, HR 0.86, P = 0.44). Whether other antiangiogenic agents have any benefits in BTCs and whether any biomarkers have any predictive values in BTCs remain to be investigated.

MET
Binding of hepatocyte growth factor (HGF) to HGF receptor (c-MET) activates multiple key downstream signaling pathways such as the RAS/MAPK, PI3K/ AKT and JAK/STAT, which play critical roles in tumor proliferation and survival [50]. Activation of MET can arise via mutations or copy number amplification. Through gene expression profiling, increased c-MET expression was observed in 20-60% of ICC and 0-70% of ECC [20,50]. Accumulating evidence has established that MET overexpression is associated with a poor prognosis. There is emerging evidence that suggest MET aberration to be one of the mechanisms responsible for EGFR resistance [51]. This led to the evolution of MET inhibitors for CCA,

NOVEL ONCOGENIC DRIVERS
The advent of next generation sequencing techniques has further shaped the genomic landscape of CCA and enhanced our understanding of its pathogenesis. Recent discoveries include IDH1/2 mutations, FGFR2 and ROS1 fusions, and mutations in chromatin remodeling genes for example ARID1A and BAP1. We will further elaborate on these promising molecular targets.

IDH mutations
IDH1 and 2 alterations exist in several tumors including gliomas and more recently identified in BTCs through high throughput molecular profiling [15-17, 21, 25-27, 54]. IDH1 and IDH2 are metabolic enzymes that catalyze the oxidative decarboxylation of isocitrate to alpha-ketoglutarate [55]. IDH mutations enhance the conversion of alpha-ketoglutarate to 2-hydroxyglutarate (2-HG), an oncometabolite that inhibits α-ketoglutaratedependent enzymes responsible for DNA methylation, epigenetic regulation and call signaling. The accumulation of 2-HG in tumor tissue in turn promotes cell proliferation and survival.
The frequency of IDH mutations ranges from 16-36%, and is ubiquitously higher in ICC than ECC [15-17, 25-27, 54, 55]. IDH mutations were observed in 22-36% of ICC and only 0-7% of ECC, and may be associated with clear cell or poorly differentiated histology [26,55]. The prognostic significance of IDH mutations remains conflicting. In a cohort of 326 patients with resected ICC, IDH mutation was associated with longer time to recurrence and OS [27]. In addition, the authors observed enhanced p53 and DNA hypermethylation among patients with IDH mutations. In contrast, Jiao et al. demonstrated in a study of 32 patients with ICC that IDH mutations confer a worse prognosis when compared to those with IDH wild-type (3-year OS 33% vs. 81%; P = 0.003) [24]. However, this adverse finding may be due to the presence of a larger proportion of stage IV disease amongst the IDH mutants compared to IDH wild-type (50% vs. 15%).  [57]. Current IDH-inhibitor studies are in early clinical development (NCT02073994, NCT02381886 and NCT02273739). The preliminary results of a phase 1 trial of AG120 (IDH1 inhibitor) in 62 patients with IDH1 mutation positive solid tumors who had progressed on standard treatment was reported at the AACR-NCI-EORTC International Conference on Molecular Targets and Cancer Therapeutics 2015. There were no dose limiting toxicities, with anemia being the most frequent Grade 3 AE (5%). 1/20 (5%) CC patients attained PR and 11/20 (55%) attained SD. Reduction in circulating 2-HG level was observed ranging from 73% to 99%, and reduction in Ki67 staining was seen from 22% -96%. The expansion phase with 500 mg QD is underway (NCT02073994).

Fibroblast growth factor receptor (FGFR) 2 fusions
FGFR2, a member of the fibroblast growth factor family of receptors (FGFR 1-4), is located at chromosome 10q26 and mitigates cell differentiation, proliferation and apoptosis [58]. The oncogenic property of FGFR2 has been linked to loss of the carboxy terminus and ligand independent dimerization, leading to FGFR protein overexpression.
Whole exome sequencing and fluorescence in situ hybridization (FISH) have identified FGFR2 alterations primarily in 6%-50% of ICC and 0-5% of ECC [ in the FGFR pathway occurred in 13% of intrahepatic CCA and 5% of extrahepatic CCA , and that these alterations were associated with improved survival [16]. More recently, FGFR2 fusions have been detected in several studies (Table 3). These fusions are a product of the FGFR receptor (exons 1-19) and various partners (e.g. AHCYL1, BICC1, KCTD1 and TXLNA). The fusion protein is activated by the enforced dimerization of the respective partners with resultant intracellular domain tyrosine residue phosphorylation, and activation of downstream signaling pathways including MAPK, PIK3/ AKT/MTOR and JAK/STAT pathways [59]. There are marked variability in the frequency of FGFR2 fusions, ranging from 6-50% in ICC, and rarely in ECC. In a series of 102 patients with BTCs, Arai et al. observed FGFR2 fusions (FGFR2-AHCYL1 or FGFR2-BICC1) in 13.6% of ICC (9/66 ICC), and that inhibition of FGFR2 impeded activation of MAPK pathway, which is responsible for uncontrolled tumor growth [28]. Another study evaluated 152 CCA and 4 intraductal papillary biliary neoplasm of the bile duct with FISH, and reported FGFR2 translocation in 12/96 (13%) of ICC, with a female predominance [29]. Those who harbored FGFR2 translocations had improved cancer-specific survival (123 vs. 37 months) and superior DFS (125 months vs. 26 months). Furthermore, cholangiocarcinoma harboring FGFR2 translocation and concomitant KRAS mutation are only rarely reported [31]. Therefore, this association remains to be explored in larger cohorts to further assess if FGFR2 translocation work in synergy with KRAS mutation in promoting carcinogenesis in CCA. In a study comprising of 109 ICC, 40 ECC and 11 gallbladder cases, novel FGFR2 gene fusions (FGFR2-KCTD1 and FGFR2-TXLNA) and a new variation of FGFR2-BICC (Type 2) were reported [60]. Using NIH3T3 clones that express either wild-type or kinase-inactive mutant forms of FGFR2-KCTD1 or FGFR2-TXLNA, the Nakamura et al. showed that wild-type FGFR fusions, and not the mutant forms induce tumor growth in vivo via ligand-independent autophosphorylation and activation of the MAPK signaling pathway. In addition, there was marked inhibition of FGFR autophosphorylation and cell proliferation by the FGFR inhibitors (BGJ398 and In a genome-wide and whole transcriptome sequencing on 6 ICC samples with FGFR2 translocations in 3/6 (50%) patients, two out of three patients responded to FGFR2 inhibitors [30]. One patient with FGFR2-MGEA5 fusion was treated with ponatinib (a pan-FGFR inhibitor) and had a biochemical CA 19-9 response with shrinkage of tumor. Another patient with FGFR2-TACC3 fusion who previously achieved a partial response with pazopanib, and subsequently received ponatinib attained stable disease. These encouraging results suggest that FGFR2 has the potential to be an actionable molecular target, and that patients who harbor these alterations may benefit from tyrosine-kinase directed therapies. An ongoing phase 2 study of BGJ398 (a selective pan-FGFR inhibitor) in patients with advanced or metastatic CCA with FGFR genetic alterations reported promising efficacy (Javle MM et al, 2016 Gastrointestinal Cancer Symposium, J Clin Oncol 34, 2016 (suppl 4S; abstr 335)) The overall RR was 22% (8/36 evaluable patients) and DCR was 75% (27/26 patients). BGJ398 was generally well tolerated. The Grade 3/4 AEs include hyperphosphatemia (19%), hypophosphatemia (9%), hyponatremia (6%), and asymptomatic increased lipase (6%). This is a promising drug that warrants further investigation.

IMMUNE CHECKPOINT INHIBITORS
Immune checkpoints including cytotoxic T-lymphocyte-associated antigen (CTLA)-4, programmed cell death (PD)-1 receptor and its ligands (PD-L1, PD-L2) promotes T-cell anergy [61]. Increased levels of tumor-infiltrating CD8+ cytotoxic T cells and/or CD4+ T cells have been shown to be associated with improved prognosis in BTCs [62]. Given the success of ipilimumab (CTLA-4 monoclonal antibody), pembrolizumab and nivolumab (anti-PD-1 antibodies) in the treatment of metastatic melanoma [63,64], there has been growing interest of the benefit of immunomodulation in BTCs. In a preclinical study of intrahepatic CCA, Koido et al. showed that both gemcitabine and interferon -γ led to an upregulation of PD-L1, which suggest that treatment with PD-L1 blockade may be beneficial [65]. Studies have suggested that mismatch repair (MMR) deficient tumors are more responsive to PD-1 blockade than are MMR proficient tumors [66]. A phase II study demonstrated that pembrolizumab led to high RR in colorectal cancer patients with genetic defects in mismatch repair (MMR) [66]. The phase II study with pembrolizumab in MMR deficient non-colorectal gastrointestinal cancers (ampullary (n = 4), pancreas (n = 4), biliary (n = 3), small bowel (n = 3), and gastric (n = 3) cancers) is ongoing. An interim analysis reported an ORR of 50% and DCR of 70% in 10 evaluable patients. The OS was 21 months and PFS was not reached (Le DT et al, 2016 Gastrointestinal Symposium, J Clin Oncol 34, 2016 (suppl 4S;abstr 195)). There are currently no studies evaluating the efficacy of PD-1 inhibitors in CCA patients with microsatellite instability (MSI)-high versus MSI-stable tumors. The interim results of another phase 1b study of pembrolizumab (MK-3495) in patients with advanced BTC was presented at the European Cancer Congress 2015 (NCT02054806). Pembrolizumab was well tolerated with an ORR of 17.4% (95% CI, 5.0-38.8) in the 23 evaluable patients. 4/24 (16.7%) of the patients experienced a treatment-related grade 3 AE (anemia, autoimmune hemolytic anemia, colitis, and dermatitis). Currently, pembrolizumab is evaluated in combination with mFOLFOX6 in a phase 1/2 study at the University of Utah (NCT02268825).

LESS-ESTABLISHED MOLECULAR ABERRATIONS
There has been limited studies regarding the following molecular aberrations and additional studies are required to provide further insight.

ROS1
Elevated ROS expression has been observed in nonsmall cell lung cancer, glioblastoma and breast cancer [32]. ROS kinase fusions [between kinase domain of ROS and Fused in Glioblastoma (FIG) gene] has been described in 8.7% of patients with CCA [32]. These fusions further activate downstream effectors such as STAT3 and AKT. The FIG-ROS fusion driver gene has been shown to accelerate tumor growth in an orthotopic allograft mouse model, and that inactivation of the gene portends an antitumor effect [67].
Notably, TAE684 (an ALK inhibitor) has been shown to inhibit ROS kinase activity, with consequent cell inhibition and cell death in BaF3 cells expressing this fusion protein [32]. Given the success of crizotinib in attaining an impressive response rate of 48% in ROS1rearranged non-small cell lung cancer [68], similar studies in CCA are warranted to evaluate the potential benefit of targeted therapy in patients with ROS fusions. A phase II trial of crizotinib in patients with ALK, MET or ROS1 alterations is underway (NCT02034981).

PI3K/AKT/MTOR
Constitutive activation of the EGFR, HER2, MET and Insulin growth factor (IGF) receptor or disruption of the PTEN and SMAD4 triggers the downstream activation of PI3K/PTEN/AKT/mTOR signaling pathway [69,70]. Dysregulation of this pathway subsequently stimulates cell proliferation, angiogenesis and survival. Activation of this pathway in patients harboring EGFR, HER2 and MET overexpression has been reported in as high as 65% of tumors. The incidence of PIK3CA (a subunit of PI3K) hotspot mutations in CCA ranges from 5% to 34% [71]. Furthermore, increased expression of phosphor-AKT1 and phosphor-MTOR in intrahepatic CCA is positively correlated with prognosis and that this association was not modified by PTEN expression [72].
Dual inhibition of AKT and MTOR with MK-2206 and everolimus (RAD001) has been shown to enhance anti-proliferative effects in CCA [73]. More recently, increased efficacy was attained in-vitro by dual inhibition of the PI3K/AKT/MTOR and RAF/MEK/ERK pathway, which overcame resistance pathways [74]. A phase I trial of mFOLFOX6 and the oral PI3K inhibitor BKM120 in patients with advanced solid tumors (4/17 CCA) reported high toxicity rates, with 76 % of the patients experiencing a grade 3/4 AE [75]. The most common AEs were neutropenia, fatigue, leukopenia, hyperglycemia and thrombocytopenia. 1/4 of the CCA patients achieved SD. The combination of everolimus with gemcitabine and cisplatin was evaluated in 10 CCA and gallbladder cancers, of which 60% had SD [76]. Currently, MK2206 (AKT inhibitor) is being investigated in advanced refractory BTC (NCT01425879).

RAS/RAF/MEK/ERK
The RAS/RAF/MEK/ERK signal transduction pathway is frequently dysregulated in BTCs [77]. Activation of this pathway requires the binding of EGF, PDGF and cytokines to its receptors, with subsequent transactivation of downstream signaling cascade, leading to the end-phosphorylation of MEK1 and 2 and ERK-1 and ERK-2. MEK is an attractive target as ERK -1 and ERK-2 are the only known MEK substrates [70]. Gain of function mutations in KRAS constitutes one of the most frequent mutations in CCA, with the most frequent alteration in codon 12 [15]. The frequency of activating KRAS mutations ranges from 9%-40% [15][16][17]. KRAS has been associated with perineural invasion and poor prognosis [78]. In addition, there is marked anatomical variability in KRAS mutation, with KRAS mutations observed in 53.3% of perihilar-type, but only 16.7% of intrahepatic CCA. Notably, the incidence of KRAS mutations increases with disease stage. [79] Despite the recognized frequency of KRAS mutations, targeting this pathway remains challenging. Early evidence of efficacy of MEK inhibitor was reported in a single arm study of selumetinib in advanced BTCs [80]. Of the 28 patients enrolled, 3 patients had confirmed partial responses. In this study, no BRAF V600E mutations were found. Recently, the ABC-04 study of selumetinib in combination with gemcitabine and cisplatin in advanced or metastatic BTC (9/13 CCA) demonstrated a RR of 37.5%, a median PFS of 6.4 months and manageable toxicities [81].

BRAF
B-Raf is a proto-oncogene and is a key component of the RAS/RAF/MEK/ERK proliferation signaling pathway. The most common BRAF gene mutation found in human cancers is V600E, and exists in up to 22% of CCA in one report [82]. More importantly, BRAF and KRAS mutations are mutually exclusive. In a recent phase II "basket" study of vemurafenib in BRAF V600 mutated non-melanoma cancers, one patient with CCA achieved a durable PR of more than one year [83].

NFk-B
Several studies have suggested the NF-kB, a transcriptional nuclear factor, plays a critical role in tumor migration and treatment resistance in several tumors, although the evidence is not conclusive [84]. This stems from the observation that tumor proliferation can be kept in check via proteasome inhibition, which halts the clearance of pro-apoptotic factors. To date, the only proteasome inhibitor investigated was bortezomib and results were disappointing, with no objective response, median time to progression was 5.8 months and median OS was 9 months [85].

JAK/STAT cytokine pathway
Binding of pro-inflammatory cytokine, interleukin-6 (IL-6) to gp130 triggers the downstream activation of the JAK/STAT pathway, leading to the silencing of its inhibitor, suppressor of cytokine signaling-3 (SOCS3) [86]. This in turn accelerates inflammation, cell growth and tumor formation. This pathway has been noted in 70% of the inflammation subclass in ICC, characterized by activation of the STAT3 and cytokine pathways and improved prognosis. [87] Furthermore, the JAK2 inhibitor AZD1480 has been demonstrated to inhibit Stat3 signaling and exhibit anti-tumor efficacy in solid tumor cell lines [88].

Notch signaling pathways
The Notch signaling cascade is a highly conserved pathway, responsible for cell differentiation, apoptosis and cell survival. To date, there are four known Notch receptors and five ligands. Aberrant Notch signaling was first described in acute T-cell lymphoblastic leukemia, and subsequently in CCA [89,90]. Notch-mediated conversion of hepatocytes into biliary lineage has been shown to promote intrahepatic CCA formation and progression in a mouse model of ICC [91]. Notch 1 and 4 were noted to be more frequently expressed in tumor cells compared to normal tissue. The frequency of Notch expression in ICC for Notch 1, Notch 2, Notch 3 and Notch 4 were 82.9%, 56.1%, 39.0% and 34.1% respectively [92]. In addition, Notch 4 was found to be prognostic and Notch 1 overexpressed in large tumors. Furthermore, Notch overexpression has been demonstrated to predict sensitivity to 5-fluorouracil in vivo. The complex Notch signaling pathway warrants further understanding before the advent of novel Notch targeting agents.

Protein kinase A regulatory subunit 1 alpha (PRKAR1A) pathway
Protein kinase A is a cyclic AMP (cAMP)-dependent protein kinase and is part of the serine-threnonine protein kinase family. Activation of the PRKAR1A/PKAI pathway is found in various tumors, including CCA [93]. More recently, fusion genes comprising of cAMP-dependent protein kinase (PKA) and mitochondrial ATP synthase (ATP1B-PRKACA and ATP1B-PRKACB) were detected with resultant increased expression of PRKACA and PRKACB and activation of MAPK signaling [60]. The abrogation of PRKAR1A gene expression has been linked to significant cell inhibition and apoptosis of CC cells via suppression of the JAK/STAT, MAPK, PI3K/AKT and WNT/β-catenin pathway signaling. Drug evaluation with PKA inhibitor (isoquinoline H89) as well as sitespecific cAMP analogs (8-Cl cAMP and 8-Br cAMP) showed promising anti-proliferative effect in CCA cells, supporting the notion that PKA can potentially contribute as a drug target in CCA.

Wnt/β-catenin pathway
Aberrant genetic alterations of the Winglesstype MMTV integration site family (Wnt)/β-catenin signaling cascade has been implicated in tumorigenesis in several studies [94]. The Wnt signaling pathway is highly activated in CCA, and an inflammatory milieu comprising of inflammatory macrophages is required for its sustainability [95]. Furthermore, tumor regression in mouse and rat models were prompted with the introduction of Wnt inhibitors. The Wnt signaling pathway has also been postulated as one of the mechanisms responsible for chemoresistance in CCA [94]. GSK3β, a "destruction complex" phosphorylates and degrade β-catenin, leading to downregulation of the Wnt survival pathway. Recently, Huang et al. showed that β-escin, an active compound in horse chestnut (Aesculus hippocastanum) seed, could inhibit the GSK3β/β-catenin pathway and thus terminate cell growth [96]. Hence, the Wnt signaling pathway may represent another alternative target for ICC treatment.
Clinical studies with novel agents in early development are summarized in Table 4.

CONCLUSIONS
Advanced CCA portends a dismal prognosis despite standard treatment with gemcitabine and cisplatin. Given the modest benefits with chemotherapy alone and the anatomical, pathological and molecular heterogeneity, there is an unmet and imperative need for comprehensive genomic profiling to improve the understanding of the pathogenesis of CCA, with the aim of personalized treatment. To achieve this aim, we must overcome the mounting challenges, which include a lack of RCTs due to the rarity of CCA, and the inherent complexity due to interactions of the signaling pathways. Extensive collaborative efforts will be required to formulate adequately powered biomarker-driven trials to improve clinical outcomes. Results of the EGFR inhibitors have been disappointing. As the majority of the trials are performed in unselected population, it will be informative to conduct trials in patients enriched for the presence of molecular signatures implicated in predicting EGFR sensitivity to determine its efficacy. Given the promising early evidence of efficacy signal with IDH and FGFR2 inhibitors in early phase trials, additional studies should focus on novel strategies targeting IDH mutations and FGFR2 fusions. Furthermore, the identification of oncogenic addiction loops, or novel combination strategy that targets critical molecular pathways simultaneously will be paramount to improve the clinical outcome in CCA.

ACKNOWLEDGMENTS
Both the authors contributed equally to this manuscript.

CONFLICTS OF INTEREST
The authors declared that they do not have anything to disclose with respect to this manuscript.

FINANCIAL SUPPORT
Andrew X. Zhu received research funding support from the V Foundation for Cancer Research and Target Cancer Foundation.
Dawn Q. Chong received research funding support from the National Medical Research Council.