Molecular Pathogenesis and Targeted Therapies for Intrahepatic Cholangiocarcinoma

Abstract

Intrahepatic cholangiocarcinoma (iCCA) is a molecularly heterogeneous hepatobiliary neoplasm with poor prognosis and limited therapeutic options. The incidence of this neoplasm is growing globally. One third of iCCA tumors are amenable to surgical resection, but most cases are diagnosed at advanced stages with chemotherapy as the only established standard of practice. No molecular therapies are currently available for the treatment of this neoplasm. The poor understanding of the biology of iCCA and the lack of known oncogenic addiction loops has hindered the development of effective targeted therapies. Studies with sophisticated animal models defined IDH mutation as the first gatekeeper in the carcinogenic process and led to the discovery of striking alternative cellular origins. RNA- and exome-sequencing technologies revealed the presence of recurrent novel fusion events (FGFR2 and ROS1 fusions) and somatic mutations in metabolic (IDH1/2) and chromatin-remodeling genes (ARID1A, BAP1). These latest advancements along with known mutations in KRAS/BRAF/EGFR and 11q13 high-level amplification have contributed to a better understanding of the landscape of molecular alterations in iCCA. More than 100 clinical trials testing molecular therapies alone or in combination with chemotherapy including iCCA patients have not reported conclusive clinical benefits. Recent discoveries have shown that up to 70% of iCCA patients harbor potential actionable alterations that are amenable to therapeutic targeting in early clinical trials. Thus, the first biomarker-driven trials are currently underway.

Introduction

Intrahepatic cholangiocarcinoma (iCCA) is the second most common liver cancer following hepatocellular carcinoma (HCC), accounting for 5% to 10% of all primary liver malignancies with an annual incidence of 2 cases per 100,000 in Western countries (1, 2). At present, it is widely accepted that iCCA arises from the malignant transformation of the intrahepatic cholangiocytes and is anatomically distinguished from the extrahepatic biliary tract cancers (eCCA), which are known as perihilar (pCCA) and distal (dCCA), with the second-order bile ducts acting as the separation point (3).

During the past decade a growing interest has been expressed in iCCA due to a marked increase in both incidence and mortality rates (1, 4). Currently, surgical resection represents the sole curative treatment option in 30% to 40% of patients with 5-year survival of 20% to 40% (1, 5). The majority of iCCA patients have no underlying liver disease or known risk factors, which further hinders the development of screening strategies for early detection. In patients with advanced disease, the combination of gemcitabine and cisplatin has been shown to confer a survival advantage over gemcitabine alone and is currently proposed as the standard of practice (6). As opposed to HCC, to date there is no approved targeted molecular therapy for iCCA, and the identification of a first-line conclusive treatment remains an unmet need. Recently, the use of next-generation sequencing technologies has enabled the identification of recurrent actionable molecular alterations that hold the promise of improving the management of advanced iCCA patients. Herein, we provide an overview of the recent discoveries of new molecular targets that should ultimately lead to the development of more personalized therapeutic approaches.

Epidemiology and Risk Factors

iCCA is a devastating disease with poor prognosis. Several studies have reported global trends of increasing incidence and mortality for iCCA in contrast with decreasing rates for eCCA (7–10). iCCA presents more commonly at older age with a slight predominance in men (male to female ratio 1.2-1.5:1; ref. 1). There is a considerable geographic and demographic variation in the epidemiology of iCCA, which likely reflects distinct environmental and genetic predispositions. The incidence of iCCA is the highest in Southeast Asia and more specifically in Thailand (>80 cases per 100,000) and can be as low as 0.2 per 100,000 in some Western countries (1, 11). Even though the vast majority of iCCAs are sporadic, several risk factors have been identified. Historically, most of these risk factors have been established for CCA without distinguishing between iCCA and eCCA, despite the fact that increasing evidence supports the hypothesis that they represent distinct entities with marked differences in their genomic features and epidemiology (Table 1; refs. 3, 12–15). The most prevalent risk factors for HCC have also been significantly associated with iCCA but not with eCCA (Table 1), including cirrhosis and chronic hepatitis B and C infections (1, 11, 16–23). Other risk factors for iCCA include primary sclerosing cholangitis (PSC), biliary duct cysts, hepatolithiasis, and hepatobiliary flukes. Hepatolithiasis has been defined as a well-known risk factor for iCCA (up to 20%) in Asian countries but not in Western countries (11). Less-established risk factors with modest associations include inflammatory bowel disease, obesity, diabetes, and alcohol abuse (1, 11).

Table 1.

Main epidemiologic and molecular differences between iCCA and extrahepatic subtypes (pCCA-dCCA)

Gene or molecule	iCCA	pCCA-dCAA	References
Proportion of CCA cases	5%−20%	pCCA (50%−70%), dCCA (15%−20%)	(12–15)
Incidence rate	Increasing	Stable or slightly decreasing	(7–10)
Anatomic location	Intrahepatic biliary tract	Extrahepatic biliary tract pCCA (near origin of cystic duct) dCCA (lower half of large duct)	(3)
Differenctial risk factors (n = positive cases/total, % casesa)
Biliary lithiasisb	377/1,539 (24%)	289/549 (52%)	(17, 18, 20–23)
Cirrhosis	161/1,622 (10%)	23/712 (3%)	(17–23)
HCV	61/1,522 (4%)	11/712 (1.5%)	(17–21, 23)
HBV	129/1,411 (9%)	4/712 (0.6%)	(17–22)
Alcoholc	158/1,524 (10%)	37/712 (5%)	(17–22)
Molecular alterations (n = positive cases/total, % casesa)
Somatic mutations
TP53	99/606 (16%)	36/137 (26%)	(50–53, 56–62)
KRAS	165/885 (19%)	29/152 (19%)	(50–53, 56–62)
IDH1/2	143/951 (15%)	3/164 (2%)	(51–54, 56–62)
ARID1A	50/390 (13%)	20/137 (14%)	(51–54, 56–57, 59, 61–62)
BAP1	45/443 (11%)	3/164 (2%)	(51–54, 56–57, 59, 61–62)
BRAF	28/574 (5%)	0/137 (0%)	(50–51, 53–54, 55–59, 61)
EGFR	14/545 (3%)	3/151 (2%)	(50–51, 53–54, 55–59, 61)
Fusion proteins
FGFR2 fusions	71/307 (23%)	0/36 (0%)	(51, 56, 57, 72, 73, 75)
Chromosomal abberations (ampifications)d
17q11 (ERBB2)	0/170 (0%)	10/55 (18%)	(31, 66)
11q13 (FGF19, CCDN1, ORAOV1)	5/128 (4%)	NA	(31)

NOTE: Frequencies in iCCA have been calculated only in non–liver fluke cases.

Abbreviations: dCCA, distal cholangiocarcinoma; HBV, hepatitis B virus infection; HCV, hepatitis C virus infection; iCCA, intrahepatic cholangiocarcinoma; NA, not applicable; pCCA, perihilar cholangiocarcinoma.

^aThe percentage of cases has been calculated by considering the number of samples presenting the molecular alteration over the total number of samples analyzed in all cohorts (discovery and validation set of samples).

^bBiliary lithiasis includes patients with hepatolithiasis, cholelithiasis, and choledocholithiasis.

^cPatients with heavy alcohol consumption or alcoholic liver disease.

^dGenomic amplifications evaluated by FISH assay or copy number alteration by SNP array.

Cells of Origin

iCCA includes a group of histologically heterogeneous tumors with diverse cellular phenotypes and cell markers, which suggests the possible existence of multiple cells of origin (Fig. 1; ref. 24). In addition, the existence of mixed hepatocellular cholangiocarcinoma (HCC-iCCA) tumors (25), a subtype with predominance of stem cell features, points out the presence of a possible common cell of origin. Thus, iCCA is currently believed to derive from biliary epithelial cells (cholangiocytes) of the intrahepatic biliary tract, hepatic progenitor cells (HPC), or even mature hepatocytes.

Figure 1.

Schematic representation of multiple cells of origin in primary liver cancers. Hepatic progenitor cells (HPC) are located at canals of Hering (CoH) near the portal triads and are thought to have the potential to differentiate into hepatocytes and cholangiocytes. There is evidence that the differentiated hepatocytes can give rise to such cells. HCC and iCCA can develop from the neoplastic transformation of mature hepatocytes and cholangiocytes, respectively. In addition, HPC and its intermediate states are thought to be the common cell of origin for hepatocellular carcinoma (HCC), intrahepatic cholangiocarcinoma (iCCA), and mixed HCC-iCCA tumors [i.e., cholangiolocellular carcinoma (CLC)]. Furthermore, recent evidence supports the hypothesis that mature hepatocytes can transdifferentiate to cholangiocytes, leading to the development of iCCA.

All liver cells share a common embryonic origin, arising from bipotential progenitors known as hepatoblasts (26). However, in the adult liver, normal tissue turnover is mainly sustained by differentiated hepatocytes and cholangiocytes. Nevertheless, upon major injury, there is an expansion of cells in the region of the canals of Hering that have been proposed to be bipotent HPCs capable of differentiating into hepatocyte or cholangiocyte lineages (Fig. 1; ref. 27). Alternatively, hepatocytes can dedifferentiate into progenitor-like cells in response to acute injury (28, 29).

With this backdrop, the hypothesis that iCCA and HCC may share a common ancestor such as the HPCs has been an important subject of discussion during the past decade. Notably, emerging data point to an overlapping molecular profile between specific subclasses of iCCA and HCC tumors. Two independent studies (30, 31) have demonstrated that a subset of iCCA tumors are enriched with liver-specific stem cell gene signatures (30, 32, 33) and molecular subclasses of poor prognosis and aggressive phenotype of HCC (proliferation; ref. 34; and S2 subclass; ref. 35). Reciprocally, a subset of HCC samples expressing biliary cell markers (i.e., CK19 and CK7; ref. 36) or enriched by iCCA-like gene expression signatures (37) show overall survival rates similar to those for iCCA patients. In addition, cholangiolocellular carcinoma (CLC), a stem cell featured mixed HCC-iCCA tumor, shares similar histopathologic features with iCCA and CK19-positive HCC (12, 38). These data suggest HPC as a possible common ancestor for a subset of primary liver cancers. Alternatively, the mutations associated with these tumors may “reprogram” differentiated liver cells toward a progenitor-like state.

Recently, several studies using genetically engineered mouse models (GEMM) and primary progenitor cell models have shed light on the link between cell differentiation and iCCA pathogenesis. The expression of gain-of-function IDH mutations, commonly reported in iCCA, led to the inhibition of hepatocyte differentiation both in vitro and in vivo and caused the expansion of HPCs (39). In turn, combined IDH and KRAS mutations in GEMMs showed pronounced oncogenic cooperation, leading to the development of premalignant biliary lesions and subsequent progression to iCCA. These data implicate mutant IDH in the subversion of liver differentiation states and in the persistence of HPCs that are susceptible to the accumulation of additional oncogenic hits (Fig. 1). While these studies did not directly determine the origin of HPCs, they did point to expansion of progenitor-like cells as a key mechanism contributing to liver carcinogenesis. Similarly, mice with genetic alterations in Hippo pathway components in the liver (i.e., YAP, SAV1, MST1/2) show expansion of progenitor-like cells, followed by the development of both HCC and iCCA (40–42). In parallel, two independent studies demonstrated that differentiated hepatocytes have the potential to give rise to iCCA through the activation of NOTCH signaling (43, 44). Aberrant activation of NOTCH signaling has been described in both iCCA (60%) and HCC (30%) tumors (45, 46). Interestingly, in a GEMM with constitutive overexpression of NOTCH1, a subset of the HCC tumors presented progenitor-like cell features with a mixed biliary and hepatocytic phenotype (45). In contrast, a recent study revealed that iCCA originates from the transformation of biliary epithelial cells in the context of chronic injury and p53 inactivation (47). Collectively, it appears that iCCA can emerge from different liver cell types depending on the initial triggering mutation and/or environmental insult. Future studies are needed to fully define these routes to iCCA, and to understand their molecular underpinnings as well their relevance to different iCCA subtypes.

Molecular Pathogenesis

Over the past 15 years, major scientific breakthroughs that have significantly changed the management of human cancers have been driven by the discovery and successful therapeutic targeting of the so-called “oncogenic addiction loops.” The term “oncogene addiction” is used to define the dependency status of cancer cells on the activation or loss of specific genes. Several examples exist of the striking survival benefits obtained in BRAF-mutated melanomas treated with vemurafenib (48) or in lung cancer harboring ALK rearrangements and treated with crizotinib (49). Unfortunately, to date, no oncogene addiction loop has been reported in iCCA.

The molecular pathogenesis of iCCA is a complex process involving multiple genomic alterations and signaling pathway deregulations. Before the implementation of next-generation sequencing technologies, our knowledge of the role of mutations in iCCA was limited, encompassing recurrent activating mutations in KRAS (19%), low frequency mutations in BRAF (5%), and EGFR (3%), and widely varying reports of loss-of-function mutations in the tumor suppressor TP53 (16%, range 1%–38%; Tables 1 and 2; refs. 31, 50–64). While KRAS and TP53 mutations are relatively common in all CCA, mutations in IDH1/2 and BRAF are considerably more prevalent in iCCA (Table 1). Epigenetic alterations through promoter hypermethylation have also been described, and the most recurrent (>25%) affects p16INK4A/CDKN2, p14ARF, RASSF1A, APC, GSTP, and SOCS-3 (58). Inflammation-related signaling pathways, such as JAK–STAT3, and proliferation-related pathways, such as EGFR and HGF–MET signaling, show profound deregulation in iCCA (58). In addition, recent studies have proposed emerging roles for NOTCH and WNT signaling in iCCA pathogenesis. Furthermore, two independent whole-transcriptome analyses discerned the existence of two distinct molecular subclasses of iCCA (31, 50). Both studies identified a proliferation molecular subclass that defines tumors with activation of oncogenic signaling pathways, including RAS–MAPK, MET, and EGFR, and poor prognosis. In addition, approximately 40% of patients belong to the Inflammation subclass, characterized by enrichment of cytokine related pathways, constitutive activation of STAT3, and better prognosis (31).

Table 2.

Potential molecular alterations amenable for targeted therapies in iCCA

Gene or molecule	Type of alteration	No. of positive/total samples (frequency)a	References
Somatic mutations
Metabolic enzymes
IDH1/2	Activating mutations	143/951 (15%)	(51–54, 56–62)
Tyrosine kinase signaling
KRAS	Activating mutations	165/885 (19%)	(50–53, 56–62)
BRAF	Activating mutations	28/574 (5%)	(50–51, 53–54, 55–59, 61)
EGFR	Activating mutations	14/545 (3%)	(50–51, 53–54, 55–59, 61)
Chromatin-remodeling genes
ARID1A	Inactivating mutations	50/390 (13%)	(51–54, 56–57, 59, 61–62)
BAP1	Inactivating mutations	45/443 (11%)	(51–54, 56–57, 59, 61–62)
PBRM1	Inactivating mutations	34/443 (8%)	(51–54, 56–57, 59, 61)
Tyrosine kinase (TK) fusion proteins
FGFR2 fusions
FGFR2-BICC1	TK fusion protein	46/211 (22%)	(51, 56, 57, 72, 73, 75)
FGFR2-PPHLN1	TK fusion protein	17/153 (11%)	(51, 56, 57, 72, 73, 75)
FGFR2-AHCYL1	TK fusion protein	7/111 (6%)	(51, 56, 57, 72, 73, 75)
FGFR2-MGEA5	TK fusion protein	1/53 (2%)	(51, 56, 57, 72, 73, 75)
FGFR2-TACC3	TK fusion protein	2/53 (4%)	(51, 56, 57, 72, 73, 75)
FGFR2-KIAA1598	TK fusion protein	1/53 (2%)	(51, 56, 57, 72, 73, 75)
ROS fusions
ROS1 fusions	TK fusion protein	2/23 (9%)	(77)
Chromosomal aberrations
11q13 (FGF19, CCND1, ORAOV1)	High-level amplification	5/128 (4%)	(32)

^aThe frequency in iCCA has been calculated by considering the number of samples presenting the molecular alteration over the total number of samples for which the specific alteration has been evaluated (discovery and validation set of samples) in different studies. Frequencies in iCCA have been calculated only in non–liver fluke cases.

Emerging signaling pathways

NOTCH signaling.

The NOTCH signaling pathway is known to play an important role during embryonic development and is essential for a proper maturation of the liver architecture. Recently, NOTCH pathway deregulation has been implicated in induction of inflammation (65) and the development and progression of iCCA (66, 67). In human CCAs, upregulation of NOTCH1 and NOTCH4 has been reported in 82.9% and 56.1%, respectively, (46). In preclinical studies, liver-induced expression of NOTCH1 intracellular domain in mice resulted in the formation of iCCAs (67). Considering that a number of NOTCH inhibitors are currently under development, the NOTCH pathway may represent a novel amenable target in iCCA (Fig. 2). However, a recent study reported different effects of targeting NOTCH receptors in a mouse model of primary liver cancer driven by v-akt viral oncogene homolog (AKT) and neuroblastoma RAS viral oncogene homolog (NRAS; ref. 68). Interestingly, while the inhibition of NOTCH2 reduced tumor burden, NOTCH1 inhibition altered the relative proportion of tumor types, reducing HCC-like tumors but dramatically increasing CCA-like tumors (68). Thus, further studies are needed to understand the complex role of NOTCH signaling in primary liver cancer.

Figure 2.

Current and potential targeted therapies in intrahepatic cholangiocarcinoma. Tyrosine kinase receptor signaling: several growth factor signaling pathways (i.e., EGF/EGFR) have been reported to be aberrantly activated in iCCA. The specify binding of growth factors results in oligomerization and autophosphorylation of their receptors, followed by signaling through the RAS–MAPK and PI3K–AKT effector cascades. FGFR2 fusions: The presence of fusion partners in the cytoplasmic domain of FGFR2 results in constitutively active receptors that induce signaling through downstream signaling pathways. NOTCH signaling: Binding of ligands on the surface of neighboring cells to the extracellular domain of NOTCH receptors (NOTCH-R) induces proteolytic cleavage of the receptor, releasing its intracellular domain (NICD), which then translocates to the nucleus and regulates expression of target genes. WNT/β-catenin signaling: activation of frizzled (FZD) receptors by WNT ligands triggers the displacement of the regulatory APC/Axin/GSK3-complex, accumulation of β-catenin and induction of target genes. IDH signaling: Mutated IDH enzymes acquire the capacity to synthesize 2-hydroxygluterate (2-HG) from α-ketoglutarate (α-KG). 2-HG alters the activity of α-KG–dependent dioxygenase enzymes involved in multiple cellular processes, including cell differentiation, survival, and DNA methylation. Molecular targeted therapies have also been highlighted; drugs currently assessed in phase II clinical trials (red) and those evaluated in early clinical trials or preclinical studies (brown) are shown.

WNT signaling.

The WNT pathway is highly activated in the tumor epithelium of human CCAs and is often characterized by overexpression of the ligands WNT7B and WNT10A along with several target genes (69, 70). It has been demonstrated that inflammatory macrophages in the stroma surrounding the tumor are required for the maintenance of this highly activated WNT signaling status (69, 71). As recently demonstrated in two rodent models mimicking human iCCA, the WNT pathway was progressively activated during the course of iCCA development, and treatment in vitro and in vivo with WNT inhibitors (ICG001 and C59) successfully inhibited tumor growth (69). Considering the recent development of several pharmacologic WNT inhibitors and the absence of APC and CTNNB1 mutations in iCCA, the WNT pathway may represent another important clinical opportunity (Fig. 2).

Identification of Novel Drivers

Recent technological advancements have led to a better understanding of the genetic and molecular forces that drive human cancers. Significant progress has been made also in iCCA, where deep-sequencing studies have unveiled novel mutations (i.e., IDH1/2, ARID1A) and oncogenic fusion genes (ROS1 and FGFR2 fusions). In the following section, we highlight the most promising discoveries, with particular emphasis on those potentially amenable to targeted therapies (Table 2; Fig. 2).

Tyrosine kinase fusion genes

FGFR2 is a tyrosine kinase (TK) protein that acts as cell-surface receptor for fibroblast growth factors and plays an essential role in the regulation of cell proliferation, differentiation, migration, and apoptosis. Recently, several FGFR2 chromosomal fusions with multiple genomic partners have been identified in several cancers, including iCCA (Table 2; refs. 51, 56, 57, 72–75). All of these fusions contain the same portion of the FGFR2 receptor (exons 1–19) and are fused to different partners through genomic breakpoints within the same intronic region (e.g., BICC1, PPHLN1, CCDC6, MGEA5, TACC3). The oncogenic activation of these FGFR2 fusion proteins relies on the activation of the TK included in the rearrangement and involves enforced dimerization, subsequent transautophosphorylation, and activation of downstream signaling pathways (57, 72, 73). Transforming and oncogenic potential of FGFR2 fusions (FGFR2–BICC1, FGFR2–PPHLN1, FGFR2–AHCYL1, FGFR2–TACC3) has been proven in vitro (57, 72, 73, 76) and in vivo (72). Furthermore, the presence of FGFR2 fusions seems to predict higher sensitivity to selective FGFR2 inhibitors (57, 72, 73, 76). However, the relative oncogenic potential of the different FGFR2 fusions or their sensitivity to specific FGFR2 inhibitors remains unknown and should be extensively investigated in future studies. Screening of FGFR2 fusions in multiple studies by massive parallel sequencing technologies or FISH-based assay has revealed striking differences in the incidence of the FGFR2 fusion events with a range between 3% and 50% of iCCA patients (51, 56, 57, 72, 73, 75). FGFR2 fusions were found to be rare in mixed HCC-iCCA and mostly absent in HCC and eCCA (Table 1; refs. 57, 72). Thus, FGFR2 fusions are a novel hallmark of iCCA.

A significant association has been observed between the presence of FGFR2 fusions (FGFR2–PPHLN1, FGFR2–BICC1) and KRAS mutations and signaling pathway activation, suggesting a possible cooperative role in driving iCCA pathogenesis (57). Even though no clear association between presence of FGFR2 fusions and clinicopathologic parameters (e.g., gender, age, stage, and prognosis) has been identified across the multiple datasets, a large study conducted in Japan has suggested a significant association with viral hepatitis (72), and a female predominance was observed in a North American cohort (75). Larger epidemiologic studies need to be conducted to clarify such discrepancies.

Besides FGFR2 fusions, ROS1 kinase fusion proteins have been identified in 8.7% (2/23) of CCAs (77). Expression of FIG–ROS1 in NIH3T3 cells conferred transforming ability both in vitro and in vivo, which could be inhibited by specific targeting (77). Furthermore, the oncogenic potential of FIG–ROS has been recently validated in an orthotopic allograft mouse iCCA model harboring KRAS and TP53 mutations (78). FIG–ROS alone was also able to promote tumorigenesis, although with reduced penetrance and longer latency. Notably, preliminary data support the efficacy of therapeutic targeting of ROS1 kinase in vitro and in vivo with small ATP-competitive inhibitors (e.g., foretinib, crizotinib). Further investigation will be required to establish the frequency of ROS fusions across different iCCA patient populations and to evaluate the potential benefit of such therapies for patients with these translocated alleles.

New somatic alterations

The application of exome-sequencing technologies has led to the discovery of novel somatic mutations in the protein-coding region of several genes and has defined a mutational landscape of the disease. Interestingly, emerging data supports a different genetic profile between liver fluke–related and non–liver fluke related CCAs in terms of gene expression (79) and mutation profiles (80). Exome sequencing of 8 cases of liver fluke-related CCAs identified 10 novel mutated genes involved in histone modification, genomic instability, and G protein signaling (e.g., KMT2C, ROBO2, PEG3, and GNAS) and confirmed mutations in already known genes (TP53 and KRAS; ref. 80). A followup study was later conducted by the same group and profiled 209 CCAs collected from Asia and Europe, associated with Opisthorchis viverrini (n = 108) and non–O. viverrini–related etiologies (n = 101; ref. 52). In summary, these studies reveal that (i) TP53, SMAD4, KMT2C, and GNAS are more commonly mutated in O. viverrini–infected CCA cases; (ii) IDH1/IDH2 mutations are almost exclusive for non–O. viverrini–related iCCA; and (iii) fluke-related CCAs present a mean of 26 somatic mutations per tumor, compared with a mean of 16 mutations per tumor in CCA with other etiologies. In addition, whole-exome sequencing (WES) studies have led to the identification of somatic mutations in chromatin-remodeling genes, BAP1, ARID1A, and PBRM1—in iCCA (52, 54). Functional studies have revealed tumor-suppressive activity of BAP1 and ARID1A, further supporting the potential role of chromatin modulators in iCCA development (52). In particular, ARID1A encodes an accessory subunit of the SWI/SNF chromatin-remodeling complex and mutations in this gene have recently been identified in a wide variety of cancers. Silencing of ARID1A in CCA cell lines (including non–O. viverrini–associated and O. viverrini–associated) resulted in a significant increase of cell proliferation. Conversely, overexpression of wild-type ARID1A led to retarded cell proliferation confirming the tumor-suppressive role of this gene (52). The possibility that iCCA patients harboring mutations in these genes may benefit from treatment with histone deacetylase (HDAC) inhibitors, such as vorinostat or panobinostat, remains unclear and needs to be further explored.

IDH1 and IDH2 mutations have been reported in approximately 14% of iCCAs (Table 2). In a large cohort of iCCA cases (n = 326), IDH1/2 mutations were associated with better overall survival (60). In contrast, in a recent WES-based study, patients with IDH1 or IDH2 mutations had shorter survival compared with patients with wild-type IDH genes (3-year survival of 33% in IDH mutants vs. 81% in IDH wild-type; ref. 54). IDH1 and IDH2 mutations in iCCA and other cancer types cluster at the hotspots codons 132 and 172, respectively. IDH1 and IDH2 encode metabolic enzymes whose normal function is to interconvert the metabolic intermediate isocitrate to α-ketoglutarate (α-KG) in conjunction with the generation of NADPH. Mutations in IDH1 and IDH2 are always present in a heterozygous state with the wild-type allele and they result in the acquisition of an abnormal enzymatic activity, the reduction of α-KG to 2-hydroxyglutarate (2-HG). 2-HG has been designated as an “oncometabolite” that contributes to cancer formation by inhibiting multiple dioxygenase enzymes that require α-KG for their activity, resulting in altered cell differentiation, survival, and extracellular matrix maturation (Fig. 2). Abnormal DNA methylation and increased protein levels of TP53 are common features of tumors with IDH1 and IDH2 mutations (60). Furthermore, using in vitro stem cell systems and GEMMs, it has been demonstrated that mutant IDH mutations are able to promote iCCA formation by blocking hepatocyte differentiation and inducing proliferation of hepatic progenitors (39).

Management and Molecular Targeted Therapies

At present, the treatment of choice for iCCA when feasible is surgical resection (1), whereas liver transplantation remains controversial. Upon resection, the median overall survival is of around 3 years and recurrence occurs in up to 60% of patients, depending on several prognostic factors, among which tumor burden and lymphonodal status appear to be the most relevant (1, 16). The prognosis for patients diagnosed with unresectable disease is even more dismal, with a life expectancy around 1 year and actuarial probability of survival of 5% at 5 years (1, 58).

The lack of clinical trials conducted specifically in iCCA patients as opposed to all biliary tract cancers (BTC) and the limited number of patients studied are among the challenges that preclude clinical practice guidelines in establishing a standard of care for patients with advanced iCCA (1). Among the 112 clinical trials reported in advanced BTCs testing systemic therapies (81), the majority are single-arm phase II studies with low statistical power and unclear impact on overall survival. The current standard of practice for advanced-stage iCCA is represented by systemic chemotherapy with gemcitabine and cisplatin (6). Survival benefits favoring the combination arm as opposed to gemcitabine alone (11.7 vs. 8 months; ref. 6) were demonstrated in a subgroup analysis of patients with iCCA (n = 80) included in a large randomized phase III trial (n = 410, ABC-02) of patients with advanced and metastatic BTCs.

On the other hand, so far no molecular targeted therapy has been proven effective for iCCA or other biliary tract cancers. The results of few trials with targeted therapies as monotherapy (i.e., selumitinib) or in combination with chemotherapy (i.e., sorafenib plus gemcitabine, cetuximab plus gemcitabine–oxaliplatin) have been discouraging with limited effects on overall survival (1). In this sense, patient stratification based on molecular biomarkers (Table 2) may be essential for clinical success in treating iCCA patients. Toward this direction, the first clinical trials driven by biomarkers (e.g., FGFR2 aberrations and IDH1/2 mutations) in BTCs, including iCCA, are currently ongoing and their results are anxiously awaited (Fig. 2, Table 3). BGJ398, a selective FGFR inhibitor, has shown efficacy in vitro by blocking the neoplastic transformation and growth of cell lines expressing FGFR2 fusions (57). Clinical efficacy of BGJ398 is currently being investigated in a phase II multicenter single-arm study in adult patients with advanced or metastatic CCA harboring FGFR2 gene fusions or other FGFR genetic alterations who have failed chemotherapy (NCT02150967). Furthermore, promising preliminary data have been reported following treatment with ponatinib, a multikinase inhibitor, in 2 iCCA patients harboring FGFR2 fusions (FGFR2–TACC3, FGFR2–MGEA5), resulting in tumor size reduction (51). Currently, a pilot study with ponatinib is ongoing in BTC patients with FGFR2 fusions (NCT02265341). At the same time, based on demonstrated efficacy in preclinical studies, specific inhibitors for IDH1 (AG-120) and IDH2 (AG-221) are currently being investigated in phase I (NCT02073994) and phase I/II (NCT02273739) clinical trials, respectively (Table 3). In parallel, considering the emerging roles of NOTCH and WNT pathway activation in the pathogenesis of iCCA, the first clinical trials targeting these pathways using available specific inhibitors are expected to move forward (Fig. 2).

Table 3.

Ongoing clinical trials using targeted therapies^a

Treatment	Targets	Clinical trial phase	Number of trials
Biomarker driven
BGJ398	FGFR, ABL, FYN, KIT, LCK, LYN, YES	II	1
Ponatinib hydrochloride	BCR-ABL, VEGFR, PDGFR, FGFR, EPH, SRC, KIT, RET, TIE2, FLT3	II	1
AG-221	Mutated IDH2	I/II	1
AG-120	Mutated IDH1	I	1
Monotherapy
Cabozantinib (XL-184)	MET, VEGFR2, RET, c-KIT, FLT1/3/4, TIE2	II	1
Everolimus	mTOR	II	2
Sunitinib	VEGFR, PDGFR, KIT, FLT3, RET	II	1
Regorafenib	RET, RAF-1, VEGFR, KIT, BRAF (V600E), PDGFRB, FGFR1, TIE2	II	2
Celecoxib	COX	IV	1c
Trastuzumab	HER2-neu	II	1
LY2801653	c-MET, MST1R, FLT3, AXL, MERTK, TEK, ROS1, DDR1/2	I	1
BKM120	VPS34/mTOR/DNAPK/PI4Kβ	II	1
Lapatinib	ErbB2–4/EGFR/SRC	II	2
Selumetinib	MEK1/2	II	1
MK2206	AKT1–3	II	1
RAV12	RAAG12	I	1
PLX8394	BRAF	I/II	1
Combination
Selumetinib + MK-2206	MEK1 + AKT1–3	II	1
Bosutinib + capecitabine	ABL/SRC/c-KIT	I	1
AZD2171 + AZD0530	VEGFR/PDGFR/FGFR1/c-KIT + SRC/ABL/LCK/YES/EGFR/LYN	I	1
Pazopanib + GSK1120212	VEGFR/PDGFR/FGFR/KIT + MEK1/2	I	1
Cetuximab + erlotinib	EGFR	I/II	2c
Trastuzumab + tipifarnib	HER2-neu + FTI	I	1
Erlotinib + bevacizumab	EGFR + VEGFA	II	2
Combination with chemotherapy
Radiotherapy + bevacizumab	VEGFA	I	1
Chemotherapyb + veliparib	PARP1/2	I	1
Chemotherapy + bevacizumab	VEGFA	II	2c
Chemotherapy ± panitumumab	EGFR	II	5c
Chemotherapy ± vandetanib (ZD6474)	VEGFR, EGFR	I, II	2c
Chemotherapy + cediranib	VEGFR	II	1
Chemotherpy ± sorafenib	BRAF, VEGFR, PDGFR	I/II	2
Chemotherapyb ± cetuximab	EGFR	II	2c
Chemotherpyb + selumetinib	MEK1/2	I/II	1
Chemotherpy ± trametinib	MEK1/2	II	1
Chemotherapyb + sirolimus	mTOR	I	1
Chemotherapy + pazopanib	VEGFR/PDGFR/FGFR/KIT	II	1
Chemotherapy + AZD2171	VEGFR/PDGFR/FGFR1/c-KIT	II	1
Chemotherapyb ± CX-4945	CX2	I/II	1c
Chemotherapy + erlotinib	EGFR	I/II	3

Abbreviations: FGFR, fibroblast growth factor; KIT, c-kit proto-oncogene receptor tyrosine kinase; PDGFR, platelet-derived growth factor receptor.

^aInformation acquired from clinicaltrials.gov.

^bChemotherapy (standard of practice: gemcitabine and cisplatin).

^cRandomized controlled clinical trials.

Future Perspectives

The application of new technologies has led to a more accurate mapping of the genomic landscape of iCCA, a devastating disease with limited treatment options. Among the newly discovered molecular alterations, FGFR2 fusions and IDH1/2 mutations hold great promise for improving the future management and treatment of iCCA patients through the first biomarker-driven clinical studies currently ongoing. Whether FGFR2 aberrations may represent a novel oncogene addiction loop in iCCA still remains an unanswered question. Nevertheless, FGFR2 fusions have the potential to represent a new avenue of research for basic investigators and clinicians. Finally, the intriguing possibility of multiple cells of origin in iCCA deserves further investigation as a means to understand the mechanisms underlying the carcinogenesis process and to determine whether this can be of relevance in clinical application.