Cholangiocarcinoma — novel biological insights and therapeutic strategies

Abstract

In the past 5 years, important advances have been made in the scientific understanding and clinical management of cholangiocarcinoma (CCA). The cellular immune landscape of CCA has been characterized and tumour subsets with distinct immune microenvironments have been defined using molecular approaches. Among these subsets, the identification of ‘immune-desert’ tumours that are relatively devoid of immune cells emphasizes the need to consider the tumour immune microenvironment in the development of immunotherapy approaches. Progress has also made in identifying the complex heterogeneity and diverse functions of cancer-associated fibroblasts in this desmoplastic cancer. Assays measuring circulating cell-free DNA and cell-free tumour DNA are emerging as clinical tools for detection and monitoring of the disease. Molecularly targeted therapy for CCA has now become a reality, with three drugs targeting oncogenic fibroblast growth factor receptor 2 (FGFR2) fusions and one targeting neomorphic, gain-of-function variants of isocitrate dehydrogenase 1 (IDH1) obtaining regulatory approval. By contrast, immunotherapy using immune-checkpoint inhibitors has produced disappointing results in patients with CCA, underscoring the requirement for novel immune-based treatment strategies. Finally, liver transplantation for early stage intrahepatic CCA under research protocols is emerging as a viable therapeutic option in selected patients. This Review highlights and provides in-depth information on these advances.

Introduction

Cholangiocarcinoma (CCA) is an epithelial cell malignancy of the liver and biliary tract. Histopathologically, this cancer is characterized by features of cholangiocyte differentiation (cholangiocytes are the epithelial cells that line the intrahepatic and extrahepatic bile ducts). The incidence of CCA is increasing in several countries globally, and given its high lethality with 5-year overall survival (OS) ranging from 7% to 20%, this disease has attracted considerable scientific and clinical interest^1,2. Indeed, several outstanding reviews focusing on CCA have been published in the past few years^2,3. These reviews richly describe the three anatomical subsets of the disease — namely, perihilar cholangiocarcinoma (pCCA), intrahepatic cholangiocarcinoma (iCCA) and distal cholangiocarcinoma (dCCA) — and their epidemiology, histopathology, genetics and molecular drivers. These recent reviews also summarize the clinical diagnosis, staging and treatment of CCA. Yet given a virtual explosion of published manuscripts (>6,000 publications) since the last Review of CCA in this journal in 2018 (ref. 4), an update is warranted.

This update is focused on what we believe are the most pertinent scientific and clinical advances made in the past 5 years. CCA is a highly desmoplastic cancer characterized by a rich tumour immune microenvironment (TIME)². We highlight the deepening understanding of the TIME based on single-cell RNA sequencing (scRNA-seq) and immunophenotyping. The desmoplasia of CCA is the result of an abundance of cancer-associated fibroblasts (CAFs)⁵, which have not been fully described in prior reviews, and the nature and role of CAFs in CCA pathogenesis will be emphasized herein. Assessments of circulating tumour cells and cell-free DNA (cfDNA) are becoming important in the diagnosis and monitoring of this malignancy^6,7, and are therefore also covered in this Review. Potent oncogenic drivers of iCCA include fibroblast growth factor receptor 2 (FGFR2) gene fusions and neomorphic, gain-of-function variants of the genes encoding isocitrate dehydrogenase (IDH), specifically IDH1 (refs. 8,9). Importantly, both FGFR2 fusions and IDH1 mutations are targetable, and approved therapies are now available for the targeted treatment of CCAs harbouring these genetic aberrations^10–17. We discuss the clinical efficacy of these molecularly targeted therapies in detail, considering that information continues to evolve regarding their efficacy and sequential use within the treatment armamentarium for advanced-stage CCA. Finally, emerging information regarding the efficacy of liver transplantation for early stage iCCAs, especially those occurring in the context of chronic liver disease, are reviewed¹⁸. We hope that our focused update on these particularly germane topics will be of high value to scientists and clinicians studying CCA and caring for patients afflicted with this complex and nefarious cancer.

The TIME of CCA

Immunosuppressive barriers

CCAs are dense, desmoplastic tumours with a prominent stroma comprising CAFs and innate and adaptive immune cells, among other cell types, that can have tumour-promoting or tumour-suppressive roles depending upon the context¹⁹. The TIME of CCA is complex with intricate and extensive crosstalk between the various stromal and immune components¹⁹. Emerging literature suggests that the CCA microenvironment is ‘immune cold’, with a paucity of cytotoxic immune cells and an abundance of immunosuppressive elements, particularly myeloid cells^20–24. Indeed, CCAs have a decreased abundance of cytotoxic T lymphocytes (CTLs) and natural killer (NK) cells relative to adjacent tumour-free liver tissues, whereas the density of regulatory T (T_reg) cells is increased²⁴. Moreover, the expression of immune-checkpoint molecules such as PD-1 and its ligand PD-L1 as well as CTLA4 is increased on the tumour-infiltrating immune cells, indicating an immunosuppressive milieu^20,23–25. CCAs also have a high level of intratumoural transcriptomic heterogeneity, probably owing to the presence of multiple different cell types that promote immunosuppression²⁶. These immunosuppressive elements within the TIME (Fig. 1) facilitate tumour growth and progression, immune evasion and resistance to chemotherapy, and probably contribute to the limited efficacy of immune-checkpoint inhibitor (ICI) monotherapy in clinical trials^20,27,28.

Fig. 1 |. Therapeutically re-engineering the TIME of CCA to reduce tumour burden.

The cholangiocarcinoma (CCA) tumour immune microenvironment (TIME) is poorly immunogenic with an abundance of immunosuppressive cell types such as myeloid-derived suppressor cells (MDSCs), tumour-associated macrophages (TAMs) and regulatory T (T_reg) cells that foster tumour growth and progression. Immunotherapeutic strategies that have been investigated in preclinical studies using mouse models include neutralization of granulocyte–macrophage colony-stimulating factor (GMCSF), liver X-receptor (LXR) agonism to inhibit MDSCs combined with macrophage inhibition using anti-colony-stimulating factor 1 receptor (CSF1R) antibodies and immune-checkpoint inhibitors (ICIs), as well as the combination of ICIs with CD40 agonism or MEK inhibition with trametinib. These immunotherapeutic strategies have the potential to re-engineer the CCA TIME to an immunogenic one that is less favourable to the tumour. NK, natural killer.

Tumour-associated macrophages (TAMs) are the predominant immunosuppressive myeloid cell type in the TIME of CCA^20,21. Moreover, an abundance of TAMs is associated with poor patient outcomes²¹. A high ratio of monocytes to lymphocytes in the blood is an independent predictor of unfavourable outcomes in patients with CCA^21,29–31. Furthermore, patients with iCCAs harbouring a high percentage of macrophages expressing CD206, a scavenger receptor expressed on TAMs, and a low percentage expressing CD86, a co-stimulatory signal for T cells, have a higher risk of disease recurrence and worse OS following surgical resection than those with CD206-low, CD86-high tumours³².

TAMs mediate tumour initiation and progression via diverse pathways, and multiple mechanisms are involved in TAM accumulation within the TIME. Kupffer cells, which are resident macrophages in the liver, can facilitate cholangiocyte proliferation and oncogenic transformation by secreting TNF that triggers JNK signalling in the cholangiocellular compartment³³. Accordingly, JNK inhibition attenuates iCCA development in preclinical models³³. In addition, inflammatory macrophages in the TIME promote a WNT-high state that favours tumour growth and progression across several preclinical models of CCA³⁴. Consequently, either WNT inhibition or macrophage depletion markedly reduces the tumour burden in these models³⁴. TNF-like weak inducer of apoptosis (TWEAK) predominantly produced by immune cells, including macrophages, can induce secretion of monocyte chemoattractant protein 1 (MCP1, also known as CCL2) by CCA cells and thereby promotes macrophage recruitment to the CCA TIME and induces their polarization to a pro-tumour CD206⁺ TAM phenotype, in association with an increase in tumour growth³⁵. Indeed, although resident macrophages have an essential role in CCA development, the majority of TAMs in established murine tumours are recruited from peripheral monocytes²⁰. Expression of PD-L1 is an important mechanism of tumour immune evasion, and PD-L1 expression within the tumour front is associated with unfavourable OS following surgical resection of iCCA²³. Monocyte-derived TAMs in the CCA TIME abundantly express PD-L1 and are in fact the main source of PD-L1 in human CCAs, and these recruited PD-L1⁺ TAMs are essential for CCA progression in mouse models²⁰. However, genetic disruption of TAM recruitment or pharmacological inhibition of TAM function did not reduce CCA tumour burden owing to a compensatory emergence of another immunosuppressive myeloid cell population, myeloid-derived suppressor cells (MDSCs).

MDSCs are a heterogeneous population of immature myeloid cells categorized as either monocytic (M-MDSCs) or granulocytic (G-MDSCs), with the latter also referred to as polymorphonuclear MDSCs (PMN-MDSCs)³⁶. MDSCs foster tumour growth and have potent immunosuppressive properties that contribute to CTL and NK cell dysfunction in cancer^20,36. Notably, patients with CCA have higher circulating levels of CD14⁺CD11b⁺HLA-DR⁻ MDSCs than individuals without known cancer³⁷. Moreover, high levels of MDSCs in the peripheral blood are associated with resistance to ICIs in patients with melanoma³⁸. In preclinical models of colitis or primary sclerosing cholangitis, which is an important risk factor for CCA, intestinal barrier dysfunction and gut dysbiosis enable bacterial cells and endotoxins to accumulate in the liver²². Consequently, hepatocytes secrete the chemokine CXCL1, which recruits CXCR2⁺ PMN-MDSCs that accelerate the development of CCA²². This effect can be abrogated through antibiotic treatment or neutralization of CXCL1 (ref. 22). These findings highlight the interplay between the gut microbiota, intestinal barrier function and the CCA TIME.

Insights from multiomics analyses

Emerging single-cell and bulk multiomics studies have begun to illuminate the complex heterogeneity of the CCA TIME, and its relationship with distinct molecular subtypes of the disease and correlation with patient prognosis, as well as its modulation through immunotherapy. scRNA-seq of 144,878 cells from 14 iCCA specimens revealed two tumour subtypes based on differential expression of the markers S100P and SPP1 (also known as osteopontin)³⁹. Each subtype had a distinct TIME and correlated with a different patient prognosis. For example, S100P⁺SPP1⁻ iCCAs (termed perihilar large-duct-type tumours) had decreased infiltration of CD4⁺ T cells and CD56⁺ NK cells while having an increased abundance of CCL18⁺ macrophages and PD-1⁺CD8⁺ T cells relative to the other tumour subtype³⁹. By contrast, S100P⁻SPP1⁺ iCCAs (peripheral small-duct-type) had enhanced infiltration of SPP1⁺ macrophages³⁹. Moreover, S100P⁻SPP1⁺ iCCAs were associated with a less aggressive tumour phenotype (including lower levels of serum CA19–9 and CEA and tumour Ki67 expression, as well as a lower prevalence of lymph node metastasis and advanced-stage disease) and better OS (based on immunohistochemistry microarray data from 186 patients and RNA-seq data from 66 patients)³⁹. In another multiomics analysis of 151 iCCAs using proteomics analysis, whole-exome sequencing and scRNA-seq, the tumours were classified into three molecular subtypes: chromatin remodelling, metabolism and chronic inflammation⁴⁰. The chronic inflammation subtype was characterized by the presence of APOE⁺C1QB⁺ TAMs and associated with poor patient outcomes⁴⁰. By comparison, the metabolism subtype had frequent mutations in KMT2D (encoding histone-lysine N-methyltransferase 2D) and was associated with lower inflammatory activity⁴⁰.

Evolving data also suggest that the CCA TIME can influence response to immunotherapy. Multiplexed single-cell transcriptomic and epitope sequencing of 200,000 peripheral blood mononuclear cells from nine patients with CCA and eight matched donors without known cancer demonstrated high levels of a unique population of CD14⁺ monocytes, termed CD14_CTX cells, in patients with tumours resistant to anti-PD-1 antibodies⁴¹. The CD14_CTX cells had increased expression of immunosuppressive cytokines and chemotactic molecules, and induced expression of SOCS3 in CD4⁺ T cells, thereby inhibiting their functional activity. The presence of CD14_CTX cell is also of prognostic relevance, given that tumours expressing a CD14_CTX gene signature were associated with unfavourable OS in patients with CCA as well as in those with other ICI-refractory cancers⁴¹.

Overcoming the immunosuppressive TIME

Preclinical studies have provided insights into potential strategies to circumvent immunosuppressive barriers in the TIME of CCAs. TAMs and MDSCs seem to have an integral role in the CCA TIME; therefore, combined targeting of both of these cell populations is a rational therapeutic approach (Fig. 1). Activation of the transcription factor liver-X receptor (LXR) restricts the survival and abundance of MDSCs⁴². In a syngeneic orthotopic mouse model of CCA, pharmacological targeting of MDSCs using either an anti-Ly6G antibody or the LXR agonist GW3965 combined with targeting of TAMs using an anti-CSF1R antibody augmented the efficacy of anti-PD-1 antibodies, with consequent prolongation of survival²⁰. Neutralization of granulocyte–macrophage colony-stimulating factor (GMCSF) is another approach that has demonstrated promise in preclinical models of iCCA²¹. GMCSF has an integral role in myeloid cell programming, and higher expression of this cytokine is associated with unfavourable OS in patients with iCCA following surgical resection²¹. Accordingly, antibody-mediated GMCSF blockade curtailed tumour growth in an orthotopic model of CCA by decreasing TAM abundance and repolarizing immunosuppressive TAMs and MDSCs in a manner that facilitated the T cell response²¹.

CD40 is a receptor that is expressed on antigen-presenting cells, including dendritic cells (DCs) as well as macrophages, and promotes DC activation and macrophage re-education to an antitumour, inflammatory phenotype⁴³. In several different preclinical models of iCCA, combination of an agonistic anti-CD40 antibody with an anti-PD-1 antibody markedly reduced tumour burden, in association with increases in the abundance and activation of macrophages and DCs as well as cytotoxic cell types including CD8⁺ T cells, NK cells and CD4⁺ T cells; most of these cell types were functionally implicated in the antitumour response through depletion experiments⁴⁴. Furthermore, CD40 agonism combined with PD-1 antagonism enhanced the efficacy of gemcitabine–cisplatin chemotherapy, with consequent increased mouse survival compared to gemcitabine–cisplatin alone⁴⁴.

Other systemic therapies can also prime the TIME and potentiate the response to ICIs. For example, MEK inhibition with trametinib upregulates tumour cell expression of MHC class I (MHC I) molecules and PD-L1. Hence, the combination of trametinib and an anti-PD-1 antibody had enhanced antitumour efficacy compared with that of either agent alone in a mouse model of iCCA⁴⁵. In aggregate, the results of these preclinical studies support the concept that therapeutic strategies that modulate the CCA TIME by shifting the balance from immuno-suppressive factors in favour of cytotoxic elements can augment the antitumour response and the efficacy of ICIs (Fig. 1).

Desmoplasia and CAFs in CCA

One of the main features of CCA is its highly desmoplastic nature with an accumulation of CAFs and excessive deposition of extracellular matrix (ECM) in the TIME, which contributes to the poor prognosis of patients with this disease^27,46–49. CAFs intercommunicate with a wide variety of cell types including cancer cells, immune cells and endothelial cells, all interacting in a multidirectional network influencing cholangiocarcinogenesis^27,50. Indeed, CAFs are one of the most abundant cell types in the TIME of CCA and contribute to tumour neoangiogenesis, immunomodulation and stiffness by secreting growth factors, cytokines, exosomes and ECM proteins; therefore, CAFs have long been considered a putative target for therapy in CCA^5,48,49,51. However, therapeutic targeting of CAFs in other tumour types has thus far failed to improve patient outcomes, highlighting the need for a more meticulous characterization of these cells⁵². Indeed, despite their overall tumour-promoting role in CCA^5,51,53, specific CAF mediators have now been shown to exert tumour-promoting or tumour-restricting effects⁵, adding more complexity to an already intricate picture.

CAF heterogeneity

CAFs are heterogeneous in their origins, transcriptomes and functions (Fig. 2), as observed in several tumour types^46,48,54,55. Mainly originating from resident fibroblasts in the liver, CAFs in the CCA TIME demonstrate considerable plasticity; for example, becoming activated in response to the dynamic microenvironment during cholangiocarcinogenesis^5,51. Classically identified based on expression of α-smooth muscle actin (α-SMA), CAFs are also characterized by expression of collagen type I α1 chain (COL1A1), fibroblast activation protein, platelet-derived growth factor receptor-α (PDGFRα) and PDGFRβ, among other proteins^{5,26,52,54,56–59}. Heterogeneity in CAF biomarkers has been confirmed by data from both bulk RNA-seq and scRNA-seq transcriptomic studies, which also outline the coexistence of transcriptomically diverse CAF subtypes in the TIME of CCA^5,57,58. Despite the lack of a common nomenclature, the presence of specific CAF subtypes enriched in genes and pathways related to either a myofibroblastic or an inflammatory phenotype (myCAFs and iCAFs, respectively) has been confirmed in all such studies^5,57,58. Moreover, the identification of intermediate CAF populations arising from the same progenitor cell points towards the existence of transient rather than static CAF subtypes, thus linking CAF heterogeneity to CAF plasticity⁵. Despite the abundant evidence of the overall tumour-promoting role of CAFs in CCA^5,51,53, specific CAF mediators such as COL1A1 have been shown to restrain rather than support tumorigenesis^5,56,60, thus challenging the historical paradigm and highlighting the need to further investigate the functionality of specific CAF subtypes.

Fig. 2 |. Different shades of CAF heterogeneity in CCA.

Cholangiocarcinoma (CCA) is a highly desmoplastic tumour type characterized by a rich stroma with an abundance of fibroblasts and their molecular products. These cancer-associated fibroblasts (CAFs) are heterogeneous in several aspects including their cell of origin, mainly arising from hepatic stellate cells (HSCs) and portal fibroblasts (PFs). Moreover, transcriptomically diverse CAF subtypes have been described to coexist in the same tumour and include vascular CAFs (vCAF), inflammatory CAFs (iCAF), myofibroblastic CAFs (myCAF) and antigen-presenting CAFs (apCAF). CAF heterogeneity is also reflected in their functionality, with evidence for subtypes with tumour-promoting and/or tumour-restricting activities. Furthermore, diverse CAF subtypes are likely to interact differentially with tumour cells and other cells in the tumour microenvironment (TME) during tumorigenesis.

Therapeutic opportunities

New therapeutic strategies are urgently needed to improve outcomes in patients with CCA, and CAFs are considered an attractive target for therapy because they actively participate in cholangiocarcinogenesis^{3,27,46,49,59}. With the emerging evidence of CAF heterogeneity, however, targeting these cells becomes more challenging and makes strategies aimed at their global targeting infeasible⁵¹. As evidenced by data from studies reported in the past 2 years, interfering with the cancer cell–CAF crosstalk and the development of dual therapies aimed at targeting both CAFs and cancer cells seem to be the most promising therapeutic strategy^5,56,57,60. Mediators of this direct cross-talk that promote tumour growth include hyaluronan produced by myCAFs and hepatocyte growth factor (HGF) secreted by iCAFs, which can stimulate cancer cells via various hyaluronan receptors and MET, respectively^5,56. Nevertheless, myCAFs also secrete type I collagen that can counteract these tumour-promoting effects⁵, as shown through conditional deletion of Col1a1 in this cell type in mouse models of other desmoplastic cancers such as pancreatic ductal adenocarcinoma⁶⁰, probably by forming a physical barrier that mechanically restrains tumour invasion⁵⁶. These opposing effects highlight the complex role of CAFs in the TIME; a comprehensive understanding of these dual effects would guide the development of more effective therapeutic strategies. For example, therapeutic strategies that target tumour-promoting CAF mediators (such as HGF) while maintaining the effect of tumour-suppressive mediators (such as type I collagen) have the potential to transform CAFs from a pro-tumour to an antitumour cell population. Furthermore, stratification of CCAs based on patient-specific mutational profiles and TIME characteristics⁵⁷ might provide new insights to advance towards precision medicine. Therefore, a better understanding of CAF heterogeneity and plasticity is needed to develop new therapeutic approaches aimed at targeting CCA taking into consideration the complexity of its TIME.

Molecular diagnosis using blood and bile

Clinical gaps exists in the early diagnosis of CCA as well as the monitoring of disease progression and treatment responses. Utilization of blood-based liquid biopsy assays has shown promise in detecting multiple types of early stage cancers^61,62. Among various classes of liquid biopsy, circulating tumour DNA (ctDNA), a fraction of cfDNA originating from tumour cells, has been studied most extensively for early cancer detection. Indeed, ctDNA in the blood might serve as a diagnostic marker for CCA^6,63. A proof-of-concept retrospective study published in 2021 investigated a panel of ctDNA methylation markers for the detection of CCA specifically⁶. The methylated DNA markers were identified as differentially methylated regions in frozen CCA specimens versus matched adjacent benign biliary epithelia and liver parenchyma tissues, and a total of nine markers were selected from this tissue comparison⁶. Subsequently, this panel of methylated DNA markers was blindly tested in plasma cfDNA samples from 54 patients with pCCA and five with dCCA as well as from control groups comprising patients with primary sclerosing cholangitis or individuals without known liver disease (n = 22 and 95, respectively)⁶. The cfDNA methylation panel had a sensitivity of 76% for the detection of CCA at a specificity of 94%⁶. The panel correctly identified 64% of patients with CCAs amenable to transplantation or surgical resection and 63% of CCAs among patients with low serum CA19-9 levels (≤100 U/ml)⁶, highlighting the potential utility of ctDNA methylation assays for the detection of early stage CCA. In addition, methylation of OPCML and HOXD9 in cfDNA was assessed in a cohort of 40 patients and compared with a disease control group comprising patients with non-malignant biliary diseases such as cholelithiasis⁶³. The combination of OPCML and HOXD9 markers had a sensitivity and specificity of 62.5% and 100%, respectively, for the diagnosis of CCA⁶³. These findings require prospective validation in larger cohorts before methylated ctDNA markers can be used for routine clinical care.

Genetic and epigenetic alterations detected in cfDNA are strongly correlated with those present in tumour tissue^7,64; therefore, tumour-specific somatic mutations and epigenetic changes in plasma ctDNA could potentially be utilized for minimally invasive longitudinal monitoring of dynamic changes in tumour biology and surveillance for disease progression. In a multicentre retrospective analysis of 138 blood-derived cfDNA samples from 124 patients with CCA (70% of whom had iCCA) using a 73-gene panel, 105 samples (76%) had one or more detectable genetic alterations after excluding variants of unknown significance, and 76 samples (55%) had therapeutically relevant alterations, including BRAF mutations, ERBB2 (also known as HER2) amplifications, FGFR2 fusions or mutations and IDH1 mutations⁶⁵. Thus, ctDNA could be a valuable resource for tumour molecular profiling, especially when obtaining tumour tissue samples is technically challenging or unsafe, which is a major challenge in managing CCA. In addition, ctDNA can be used to monitor response to chemotherapy and targeted therapy. A proof-of-concept study of serial plasma cfDNA testing in eight patients with pancreaticobiliary cancers demonstrated that changes in cfDNA mutant allele fraction correlate well with changes in CA19–9 levels (Pearson’s r = 0.69 for interval slopes, r = 0.93 for interval differences), suggesting that ctDNA levels reflect changes in disease burden over time during treatment⁷. Furthermore, longitudinal profiling of ctDNA can be performed to track the emergence of resistance to chemotherapy and, therefore, potentially facilitate the timely use of second-line or third-line treatments for CCA^7,64. Thus, determination of cancer-derived cfDNAs present in blood is a promising and emerging technology for the management of patients with CCAs.

Confirmation of malignant biliary stricture is another major clinical challenge in patients with indeterminate biliary strictures detected using non-invasive imaging. The diagnostic work-up in these patients involves endoscopic retrograde cholangiopancreatography (ERCP), but the sensitivity of traditional methods such as ERCP-guided brush cytology or biopsy sampling for the detection of malignant biliary stricture is <40%^66,67. To address this diagnostic dilemma, molecular characterization of biliary brush specimens and/or bile has been evaluated as a tool for the diagnosis of malignancy. Fluorescence in situ hybridization (FISH) analysis of biliary brush specimens for chromosome 1q21, 7p12, 8q24 and 9p21 aberrations was found to improve the accuracy in diagnosing malignant biliary stricture, achieving 65% sensitivity and a specificity of 93%⁶⁷. Moreover, somatic KRAS mutations and/or loss-of-heterozygosity alterations across ten tumour-suppressor genes detected in DNA isolated from biliary brush specimens have been shown to be excellent diagnostic biomarkers for malignant biliary strictures⁶⁸. In this prospective study involving 100 patients with biliary strictures (41% with malignancy), routine FISH identified nine of 28 malignancies not detected on cytology analysis; mutation profiling of cell-free supernatant fluid of the brush specimen identified an additional eight malignancies not detected by FISH, increasing the overall sensitivity to 51% versus 32% with cytology alone, whilst maintaining 100% specificity⁶⁸. These data highlight the complementary role of these diagnostic assays. Considering the rapid evolution of next-generation sequencing (NGS) capacity, a single-centre prospective study in 252 patients with biliary stricture utilized BiliSeq, a targeted NGS panel encompassing 28 genes, to screen for mutations in biliary brush cytology samples with or without biliary biopsy specimens; the assay had good performance for the detection of CCA with 73% sensitivity and 100% specificity⁶⁹. Notably, BiliSeq had a sensitivity of 83% in patients with primary sclerosing cholangitis⁶⁹. For comparison, the sensitivity and specificity of elevated serum CA19–9 alone were 76% and 69%, respectively, and for pathological evaluation alone were 48% and 99%⁶⁹. In another prospective study, NGS-based mutational analysis of cfDNA from ERCP-derived bile samples using a 52-gene panel had a sensitivity of 96.4% and a specificity of 69.2% for the detection of CCA in 68 patients with suspicious biliary strictures⁷⁰.

Altered DNA methylation markers in biliary brush specimens and/or bile samples from patients with indeterminate biliary strictures might also improve the diagnosis of CCA. The first proof-of-concept case–control study in patients with CCA (n = 15 and 34 in the training and validation cohorts, respectively) and control patients with primary sclerosing cholangitis (n = 20 and 34) reported that a panel of four methylation markers (in CDO1, CNRIP1, SEPT9 and VIM) measured in DNA from biliary brush specimens had a sensitivity of 85% and a specificity of 98% for the detection of CCA (area under the receiver operating characteristic curve (AUC) 0.944)⁷¹. Several smaller case–control studies have shown the outstanding performance of DNA methylation biomarkers (for example, at loci associated with EMX1 or HOXD8) in biliary brush and/or bile samples, with AUC values of 0.98–1 for the detection of malignant biliary strictures^6,72. Similarly, profiling of microRNAs in bile samples has been investigated, with promising results for the detection of CCA (AUC 0.652, 0.730, 0.765, 0.975 and 0.975 for miR-92a-3p, miR-30d-5p, miR-944, miR-9 and miR-145*, respectively)^73,74.

The data from these initial biomarker studies have revealed the great promise of ctDNA as a novel biomarker for CCA detection. We anticipate that such liquid biopsy assays will further improve the diagnosis of CCA in routine clinical practice once larger-scale external validation studies are completed in the near future. Given that comprehensive molecular profiling is recommended in the routine clinical care of patients with advanced-stage CCA, ctDNA analysis might also serve as a minimally invasive test for the selection of molecularly targeted treatments. Plasma ctDNA will be particularly helpful in identifying patients with CCAs harbouring IDH1 mutations or FGFR2 fusions, given the availability of approved therapies predicated on these alterations⁶⁵. In an NGS-based analysis involving 1,671 patients with advanced-stage CCA, genetic alterations in cfDNA were detected in samples from 84%, with targetable alterations in cfDNA detected in samples from 44%⁷⁵. The concordance between cfDNA and tumour tissue analysis for the detection of IDH1 and BRAF^V600E mutations was 87% and 100%, respectively, but was only 18% for FGFR2 fusions⁷⁵. Although molecular analyses of blood, biliary brush and/or bile specimens can facilitate the implementation of precision medicine in patients with CCA, important challenges continue to limit widespread clinical use of such assays. First, numerous methods can be used to detect methylation markers or mutations in cfDNA, with substantial variation in sample type (for example, biliary brush cytology versus cell-free supernatant fluid versus bile), handling and processing, and analysis. Standardized ctDNA assay protocols that can capture the full spectrum of mutation and methylation with high sensitivity and specificity need to be developed and validated in order to ensure robust and generalizable results. Second, most current supporting evidence comes from smallcohort, single-centre, case–control studies. Given the low incidence of CCA, these issues necessitate large-scale, collaborative, multicentre prospective studies using standardized protocols.

Emerging molecular classifications of CCA

The prevalence of molecular profiling has enabled the definition of transcriptome-based classes of iCCA and pCCA^76–78. Importantly, the molecular subclasses might predict patient outcomes or response to certain therapeutic approaches. For example, a genomic and transcriptomic characterization of 104 surgically resected iCCA specimens revealed two distinct tumour classes associated with different clinical outcomes and with the absence or presence of mutations in KRAS or BRAF⁷⁶. Specifically, the 5-year OS in 51 patients with ‘cluster 1’ class tumours according to a transcriptomic classifier based on 238 significantly differentially expressed genes was 72% compared with 30% in 53 patients with cluster 2 tumours (P < 0.0007)⁷⁶. Patients with cluster 2 tumours also had a shorter median time to recurrence (13.7 ± 6.3 months versus 22.7 ± 18.1 months; P = 0.001)⁷⁶. Integration of KRAS/BRAF mutational status with the 238-gene classifier grouped all 18 evaluable patients with such mutations into the poor prognosis cluster 2 (ref. 76). Integrative molecular analyses have also been applied for the detection of biological classes of CCA with differential activation of certain signalling pathways. One such analysis of 149 iCCAs defined two principal tumour classes: an inflammation class (38%) and a proliferation class (62%)⁷⁸. Features of the inflammation class included activation of inflammatory signalling pathways, overexpression of cytokines and activation of STAT3, whereas the proliferation class was characterized by activation of oncogenic signalling pathways, KRAS/BRAF mutations, and gene expression signatures previously associated with poor outcomes in patients with hepatocellular carcinoma⁷⁸. Patients with proliferation class iCCAs had significantly worse OS than patients with inflammation class tumours (median 24.3 months versus 47.2 months; P = 0.048)⁷⁸. Similarly, an integrated genomic analysis of pCCA and dCCA tumour specimens (n = 189) enabled the definition of four distinct classes: metabolic, proliferation, mesenchymal and immune⁷⁷. Tumours of the metabolic class (comprising 19% of cases) had a hepatocyte-like phenotype with enrichment of genes related to bile acid metabolism⁷⁷. dCCAs were more likely to be categorized into the proliferation class (accounting for 23% of tumours overall), which was defined by mTOR signalling, HER2 alterations and enrichment for MYC targets⁷⁷. Patients classified in the mesenchymal class (47%) had unfavourable OS and enrichment of gene signatures associated with epithelial-to-mesenchymal transition⁷⁷. The immune class (11%) was characterized by high levels of lymphocyte infiltration and PD-1/PD-L1 expression, as well as other transcriptomic features associated with increased responsiveness to ICI in other tumour types⁷⁷.

The increasing emphasis in oncology on immunotherapeutic approaches has highlighted the integral role of the TIME and the potential clinical significance of immune-based classifications of CCA. A TIME-based classification developed using transcriptomics from a training set comprising 198 iCCA specimens identified four immune subclasses associated with distinct immune escape mechanisms and patient outcomes: immune-desert, immunogenomic, myeloid and mesenchymal⁷⁹. This classification was validated using an independent set of 368 iCCAs and through immunohistochemical analysis of a further 64 iCCA samples⁷⁹. In alignment with the concept that the CCA TIME is ‘immune cold’, the dominant subclass (present in approximately 45% of patients overall) was the immune-desert subset characterized by low expression of TIME gene signatures⁷⁹ (Table 1). By contrast, the immunogenomic subclass was enriched for infiltration of adaptive and innate immune cells and activation of inflammatory and immune-checkpoint pathways, and was associated with the most favourable patient outcomes⁷⁹. The myeloid subset featured enrichment of monocyte-derived and myeloid gene signatures, whereas the mesenchymal subclass was enrichment for fibroblast gene signatures⁷⁹. Similarly, a Stroma, Tumour and Immune Microenvironment (STIM) classification of iCCA, which as its name implies integrates stromal, tumour and immune microenvironmental elements, categorized the preponderance of patients into non-inflamed tumour classes (65%)⁵⁷. Of the five STIM classes of iCCA identified (Table 1), two inflamed classes were defined as ‘immune classical’ (accounting for 10% of patients) and ‘inflammatory stroma’ (~25%), with the latter characterized by a rich stroma (extensive desmoplasia), T cell exhaustion and KRAS mutations⁵⁷. The three non-inflamed classes are: (1) the ‘desert-like’ class (20%) with a paucity of immune cells but enrichment of T_reg cells; (2) the ‘hepatic stem-like’ class (35%) with abundant M2-like macrophages and enrichment for IDH and BAP1 mutations and FGFR2 fusions; and (3) the ‘tumour classical’ class (10%) characterized by cell cycle pathway activation and associated with unfavourable patient outcomes⁵⁷. Although these classifications assist in the broad categorization of CCA subtypes and support the notion that CCAs typically have an immunosuppressive TIME, their clinical utility is currently unclear. Prospective validation of these classifications is needed before they can be used in routine patient care. For example, clinical trials stratifying patients based on these classifications with correlation to particular treatments would help to clarify their potential utility in the management of CCA.

Table 1 |.

Immune classifications of iCCA based on RNA-sequencing data

Subclass	Proportion of iCCAs (%)	Key features	Prognostic association (median OS, months)
Classification based on immune gene expression signatures 79
Immune-desert	46–48	TME signatures with weak expression of immune and myofibroblast signatures	42
Immunogenomic	9–13	Signatures of recruited innate immune cells, adaptive immune cells and activated fibroblasts Activation of inflammatory pathways	73
Myeloid	13–19	Strong expression of monocyte-derived and/or other myeloid signatures Low expression of lymphoid signatures	25
Mesenchymal	22–28	Strong expression of fibroblast signatures	19
Stroma, Tumour and Immune Microenvironment (STIM) classification 57
Immune classical	~10	High immune infiltration (63%) Moderate stromal infiltration (27%)	–a
Inflammatory stroma	~25	Moderate immune infiltration (43%) High stromal infiltration (50%) Abundant desmoplastic reaction and ECM deposition High stiffness Activated inflammatory stroma T cell exhaustion	–
Hepatic stem-like	~35	Low immune infiltration (28%) Low stromal infiltration (17%) Abundant tumour-promoting macrophages	–
Tumour classical	~10	Low immune infiltration (22%) Low stromal infiltration (12%) Activation of cell cycle pathways	–
Desert-like	~20	Low immune infiltration (22%) Low stromal infiltration (11%) Regulatory T cell enrichment	–

ECM, extracellular matrix; iCCA, intrahepatic cholangiocarcinoma; OS, overall survival; TME, tumour microenvironment.

^aIn a multivariate analysis, the STIM classes were not independent predictors of outcome.

Advances in the treatment of CCA

Given that the objective of this Review is to provide an overview of pertinent clinical updates since our last Review in the journal⁴, here we focus on progress made over the past 5 years in systemic and non-systemic therapeutic strategies for CCA. Locoregional therapies including radiotherapy, transarterial chemoembolization and radioembolization were highlighted in our prior Review⁴. Since then, no major advances in locoregional treatments for CCA have been made, and thus such therapies are not included in this update.

Molecularly targeted therapies

Several high-throughput genomic sequencing analyses have elucidated the distinct mutational landscapes of iCCA, pCCA and dCCA, thus revealing differences in biology between these anatomical subtypes of the disease^80–84. Actionable alterations such as FGFR2 fusions and IDH1 mutations occur almost exclusively in iCCAs, whereas pCCAs and dCCAs more commonly harbour HER2 amplifications and KRAS mutations^80–84. Mutation frequencies among CCAs vary geographically and by aetiology, with Asian populations having lower frequencies of IDH1 mutations than non-Asian populations, and liver fluke-associated CCAs having a lower prevalence of IDH1 mutations and a higher prevalence of TP53 mutations than those with other aetiologies^85–87.

The identification of actionable genetic alterations in approximately 40–50% of iCCAs and 15–20% of pCCAs and dCCAs⁸⁸ has transformed the therapeutic landscape forpatients with advanced-stage CCA. Mutational profiling is usually performed on tumour tissue specimens, but given failure rates of up to 27% in CCA biopsies, liquid biopsy analysis of ctDNA might serve as an additional tool for identifying therapeutic targets, as described above^64,65,89. Genetic alterations that define the eligibility of patients with CCA for molecularly targeted therapies based on FDA approvals or recommendations in the National Comprehensive Cancer Network (NCCN) guidelines are discussed below⁹⁰.

FGFR2 fusions and other rearrangements.

FGFR signalling has key roles in various physiological processes and drives cell proliferation, growth and survival, thus making this pathway susceptible to hijack by cancer cells^91,92. FGFR2 fusions and other rearrangements, which were first reported in CCAs in 2013 and subsequently found to be present in 13–14% of iCCAs^93–95, constitute one of the most highly actionable alterations in this disease. The first two molecularly targeted therapies to gain regulatory approval for the treatment of CCA were the ATP-competitive, reversible FGFR inhibitors pemigatinib and infigratinib^13,14. These oral small-molecule inhibitors demonstrated overall response rates (ORRs) of 35.5% and 23.1%, respectively, in single-arm phase II trials involving patients with previously treated advanced-stage CCAs harbouring FGFR2 fusions or rearrangements^10,11 (Table 2), resulting in FDA Accelerated Approval for this indication in April 2020 and May 2021, respectively^13,14. However, the development of infigratinib has been halted according to a manufacturer’s press release⁹⁶. A third selective FGFR inhibitor, futibatinib, binds covalently and irreversibly to the P-loop cysteine residue in the FGFR kinase domain and achieved an ORR of 41.7%, with responses lasting a median of 9.7 months, in a similar phase II study involving patients with previously treated advanced-stage FGFR2-rearranged iCCA¹². These data led to FDA Accelerated Approval of futibatinib for this indication in September 2022 (ref. 15). The safety and efficacy of derazantinib, an FGFR1–3 inhibitor, has also been assessed in a phase II trial involving 103 patients with advanced-stage CCA harbouring an FGFR2 fusion who had received at least one line of prior treatment⁹⁷. In this study, derazantinib was associated with an ORR of 21.4%, a disease control rate of 74.8% and a median OS of 15.5 months⁹⁷.

Table 2 |.

Efficacy of molecularly targeted therapies in patients with CCA

Agent	Study phase	Number of patientsa	ORR (%)	DCR (%)	Median PFS (months)	Median OS (months)	Ref.
FGFR inhibitors
Pemigatinibb	II	107	35.5	82.2	6.9	NA	10
Infigratinibb	II	108	23.1	84.3	7.3	12.2	11
Futibatinibb	II	103	41.7	82.5	9.0	21.7	12
Derazantinib	I/II	103	21.4	74.8	8.0	15.5	97
IDH inhibitors
Ivosidenibb	III	124	2.4	53.2	2.7	10.8	16
BRAF inhibitors
Dabrafenib plus trametinib (MEK inhibitor)b,c	II	43	46.5	85.4	9.0	14.0	111
Vemurafenibc	II	9	33.3	NA	NA	NA	113
HER2-targeted agents
Trastuzumab plus pertuzumab (monoclonal antibodies)c,d	II	39	23.1	51.3	4.0	10.9	115
Trastuzumab deruxtecan (antibody–drug conjugate)c	II	24	36.4	81.8	4.4	7.1	116
Neratinib (pan-ErbB tyrosine kinase inhibitor)c	II	25 (11 with CCA)	16	28	2.8 (1.4 in CCA subgroup)	5.4 (5.4 in CCA subgroup)	118
Zanidatamab (bispecific antibody)c	I	17	47	65	NA	NA	119

CCA, cholangiocarcinoma; DCR, disease control rate; NA, not available; ORR, objective response rate; PFS, progression-free survival.

^aNumber of patients with target molecular aberrations: FGFR2 fusions and/or rearrangements for FGFR inhibitors, IDH1 R132 mutations for IDH inhibitors; BRAF^V600E mutations for BRAF inhibitors; and ERBB2 (also known as HER2) amplification and/or overexpression for HER2-targeted therapies.

^bFDA approved for use in patients with previously treated advanced-stage CCAs harbouring the target oncogenic alteration.

^cData for these agents come from patients with CCAs as well as other biliary tract cancers.

^dListed in National Comprehensive Cancer Network guidelines as subsequent-line therapy for patients with advanced-stage CCA following disease progression on recommended initial treatements⁹⁰.

Acquired resistance to FGFR inhibitors limits the effectiveness of these agents, which are consistently associated with median progression-free survival (PFS) durations of 7–9 months^10–12,97 (Table 2). Polyclonal secondary mutations in the FGFR2 kinase domain are a common resistance mechanism^98–101. Futibatinib has shown potent preclinical activity against many of these mutations and has shown clinical activity after progression of CCA on treatment with reversible FGFR inhibitors^101–103. Additional next-generation inhibitors, such as the FGFR2-selective inhibitor RLY-4008 (NCT04526106) and the multikinase inhibitor tinengotinib (also known as TT-00420; NCT04919642), are currently under investigation to assess their efficacy in overcoming resistance to prior FGFR inhibitors^104–106. RLY-4008 is a highly selective FGFR2 inhibitor that has activity against primary FGFR2 alterations and common resistance mutations¹⁰⁶. In a first-in-human study involving 35 patients with FGFR2-altered CCA, this agent had encouraging clinical activity, with >10% tumour shrinkage observed in nine (56%) of 16 patients with CCA previous treated with an FGFR inhibitor¹⁰⁶.

Other FGFR inhibitors under investigation in patients with CCA include the pan-FGFR inhibitors erdafitinib (NCT02699606) and KIN-3248 (which has activity against FGFR2 gatekeeper, molecular brake and activation loop mutations¹⁰⁵; NCT05242822), the multikinase inhibitor derazantinib(NCT03230318), the bivalent FGFR1–3 inhibitor tasurgratinib (E7090, which blocks FGF–FGFR interactions; NCT04238715), and the FGFR1–3 inhibitor HMPL-453 (NCT04353375). Moreover, phase III trials comparing the approved FGFR inhibitors against frontline gemcitabine and cisplatin chemotherapy are currently ongoing (pemigatinib (NCT03656536), infigratinib (NCT03773302) and futibatinib (NCT04093362)), as are early phase trials ofcombination therapies including FGFR inhibitors (gemcitabine and cisplatin plus pemigatinib or ivosidenib (NCT04088188) and derazantinib plus atezolizumab (NCT05174650)).

IDH1 mutations.

R132-mutant IDH1 acquires a neomorphic activity that converts α-ketoglutarate to 2-hydroxyglutarate, an oncometabolite that inhibits histone and DNA demethylases and results in widespread epigenetic alterations and oncogenesis¹⁰⁷. The frequency of IDH1 mutations in CCAs varies widely across studies and was found to be 13.1% in iCCAs and 0.8% in pCCAs and dCCAs in a systematic review of 45 studies⁸⁵. Ivosidenib, an orally administered small-molecule inhibitor of mutant IDH1, gained FDA approval for the treatment of chemorefractory IDH1-mutated CCA in August 2021 based on the results of a phase III trial showing an improvement in PFS over placebo (median 2.7 versus 1.4 months; HR 0.37, 95% CI 0.25–0.54; P < 0.0001)^16,17 (Table 2). OS was not significantly improved (HR 0.79, 95% CI 0.56–1.12; P = 0.093); however, the fact that 70% of patients in the placebo group crossed over to receive ivosidenib following disease progression is likely to have confounded this result¹⁶. Additional IDH1 inhibitors under investigation in patients with CCA include olutasidenib (FT-2102; NCT03684811), LY3410738 (NCT04521686) and HMPL-306 (NCT04762602). Given that IDH1 mutations can lead to deficiencies in the repair of double-stranded DNA damage^108,109, several ongoing trials of poly(ADP-ribose) polymerase (PARP) inhibitors are also including patients with IDH1-mutant CCA (olaparib (NCT03212274). olaparib plus the ATR inhibitor ceralasertib (NCT03878095) and olaparib plus durvalumab (NCT03991832)).

BRAF^V600E mutations.

BRAF^V600E mutations occur in up to 5% of biliary tract cancers (BTCs), primarily in iCCAs¹¹⁰. Treatment with the BRAF inhibitor dabrafenib in combination with the MEK inhibitor trametinib led to an ORR of 47% and a median PFS of 9.0 months in the BTC cohort of the phase II ROAR basket trial¹¹¹ (Table 2). In June 2022, this combination gained FDA approval for use in patients with previously treated non-colorectal solid tumours, thus including CCAs, harbouring a BRAF^V600E mutation and no satisfactory alternative treatment options, based on data from ROAR and several other trials¹¹². Data on BRAF inhibitors as monotherapy for BTCs are limited, although vemurafenib induced partial responses in three of nine patients with BRAF^V600E-mutant BTCs in another basket trial¹¹³.

HER2 alterations.

HER2 belongs to the ErbB family of receptor tyrosine kinases and is overexpressed more frequently in pCCAs and dCCAs (17.4%) and gallbladder cancers (19.1%) than in iCCAs (4.8%)¹¹⁴. The combination of two anti-HER2 antibodies, trastuzumab and pertuzumab, achieved an ORR of 23% and a median PFS of 4.0 months in 39 patients with HER2-amplified and/or overexpressing BTCs in the phase IIa MyPathway basket study¹¹⁵ (Table 2), gaining a listing in the NCCN guidelines as a recommended therapy for previously treated HER2-positive BTCs⁹⁰. The antibody–drug conjugate trastuzumab deruxtecan demonstrated an ORR of 36.4% (including two complete responses (CRs)), a median PFS of 4.4 months and a median OS of 7.1 months in a similar phase II trial population (n = 22)¹¹⁶, and a partial response was also seen in one of eight patients with HER2-low disease. Trastuzumab deruxtecan is being further explored in patients with BTCs in an ongoing phase II trial (NCT04482309). Another phase II trial evaluated the efficacy of combination folinic acid, 5-fluorouracil and oxaliplatin (FOLFOX) chemotherapy plus trastuzumab in the second-line setting in 34 patients with HER2-positive BTCs; the ORR was 29.4%, median PFS was 5.1 months and median OS was 10.7 months¹¹⁷.

Additionally, the oral, irreversible, pan-ErbB tyrosine kinase inhibitor (TKI) neratinib has been evaluated in patients with previously treated HER2-mutant CCA (as opposed to merely HER2-positive disease) as part of the SUMMIT basket trial¹¹⁸. Results from the phase II BTC cohort of this trial, which included 11 patients with CCAs, demonstrated an ORR of 16% in all 25 patients evaluated, with a median PFS of 2.8 months and median OS of 5.4 months (1.4 months and 5.4 months, respectively, in the CCA subgroup)¹¹⁸ (Table 2).

The novel bispecific anti-HER2 antibody zanidatamab, which targets the same epitopes as trastuzumab and pertuzumab, achieved an interim ORR of 47% and a median response duration of 6.6 months in 17 patients with advanced-stage HER2-positive BTCs¹¹⁹ (Table 2), and this agent is being further evaluated in studies including patients with CCAs and other BTCs (NCT04466891 and NCT03929666). Other therapeutic strategies currently being evaluated in trials involving patients with HER2-overexpressing CCAs include the combination of trastuzumab and the HER2 TKI tucatinib (NCT04579380) and novel antibody–drug conjugates (A166 (NCT03602079) and zanidatamab zovodotin (NCT03821233)).

NTRK fusions.

NTRK fusions occur in approximately 0.2% of CCAs and are highly actionable alterations in solid tumours¹²⁰. These NTRK fusions are therapeutically actionable using the TRK inhibitors larotrectinib and entrectinib, which received FDA approval agnostic of tumour histology based on efficacy across cancer types (summary ORR of 75% with larotrectinib and 57% with entrectinib); partial responses were observed in two of three patients with CCA included in the pivotal trials of these agents^121–124.

RET fusions.

RET fusions are also very rare in CCAs, with a prevalence of 0.15% in iCCA and 0.11% in pCCA/dCCA¹²⁵. The phase I/II ARROW trial evaluated the potent, selective of RET inhibitor pralsetinib in 29 patients with advanced-stage RET-altered solid tumours¹²⁶. Of the three patients with CCA included in this basket study cohort, two had a partial response and one had stable disease¹²⁶, gaining pralsetinib a listing on the NCCN guidelines⁹⁰. The safety and efficacy of selpercatinib, another highly selective RET inhibitor, are being evaluated in patients with RET fusion-positive cancers, including two patients with CCA, in the ongoing phase I/II LIBRETTO-001 basket trial¹²⁷.

Immunotherapies

The role of ICI-based immunotherapy for the treatment of CCA has changed drastically over the past few years¹²⁸. It is important to differentiate the role of immunotherapy for CCA according to two separate scenarios¹²⁹.

The first scenario is related to exploration of immunotherapies in a tumour-agnostic manner, whereby selection of patients relies on specific molecular alterations, regardless of the primary tumour site or histology. In this regard, the ICI pembrolizumab is indicated for use in patients harbouring CCAs with deficient mismatch repair (dMMR) and/or high microsatellite instability (MSI-H)^130–132. Pembrolizumab, an anti-PD-1 antibody, had an ORR of 53% (21% CR rate) in a cohort of 86 patients with advanced-stage dMMR solid tumours; four patients with CCA were included in this cohort and one had a CR, with the other three having stable disease¹³⁰. An updated analysis of the multicohort phase II KEYNOTE-158 study of pembrolizumab in patients with advanced-stage MSI-H/dMMR non-colorectal cancers included data from 22 patients with CCA¹³¹. In the CCA cohort, pembrolizumab was associated with an ORR of 40.8% (CR rate 13.5%), a median response duration of 30.6 months, a median PFS of 4.2 months and median OS of 19.4 months, with a 3-year OS of 30.3%¹³¹. In addition, ICIs have notable activity against cancers with a high tumour mutational burden (TMB). Indeed, pembrolizumab has also been granted histology-agnostic approval by the FDA for patients with a high TMB (≥10 mutations per megabase)¹³³, based on a reported ORR of 29% (CR rate 4%) in a cohort of 102 patients with previously treated advanced-stage solid tumours (not specified how many had CCA); responses were durable, lasting ≥12 months in 57% of patients and ≥24 months in 50%¹³⁴. Unfortunately, these tumour-agnostic approaches are applicable to only a very limited proportion of patients with CCA, given that <5% of this patient population have dMMR/MSI-H or TMB-high tumours¹²⁹. Therefore, developing and applying immunotherapies in a way that would enable a wider spectrum of patients to benefit is of substantial importance.

Indeed, the second scenario is the development of immunotherapy in unselected populations of patients with advanced-stage CCA. This approach is mainly based on translational research work suggesting a potential role for immune-directed therapies in the treatment of CCA and other BTCs, considering that markers of a pre-existing antitumour T cell response are generally associated with a favourable prognosis in patients with these cancers^135–138. Hence, first efforts in this regard were mainly focused on the use of monotherapy approaches with ICIs. Unfortunately, this strategy had limited efficacy in patients with BTCs, with an ORR of 5.8%, a median PFS of 2.0 months and median OS of 7.4 months with single-agent pembrolizumab in the phase II KEYNOTE-158 study²⁸. Moreover, the efficacy was not clearly better in patients with PD-L1-positive tumours (ORR 6.6% and 13.0% in KEYNOTE-158 and KEYNOTE-028, respectively)²⁸. Given these findings, alternative approaches have been pursued, mainly in the form of combining ICIs with cytotoxic chemotherapy.

Preliminary results from a retrospective study by Sun and colleagues¹³⁹ suggested better outcomes with the addition of anti-PD-1 antibodies to chemotherapy (38 patients) than with anti-PD-1 antibodies alone (20 patients) or chemotherapy alone (19 patients); the median PFS durations were 5.1 months, 2.2 months and 2.4 months, respectively, and the median OS was 14.9 months versus 4.1 months versus 6.0 months. Additional phase II studies in patients with previously untreated advanced-stage BTCs supported this rationale^140–142. In the largest of these phase II studies, Oh and colleagues¹⁴² recruited 124 patients to three different treatment arms: (1) chemotherapy (gemcitabine plus cisplatin) followed by concurrent chemotherapy and dual ICI therapy with the anti-PD-L1 antibody durvalumab and the anti-CTLA4 antibody tremelimumab from cycle 2 (30 patients); (2) concurrent chemotherapy plus durvalumab from cycle 1 (47 patients); or (3) concurrent chemotherapy plus durvalumab and tremelimumab from cycle 1 (47 patients). Whenever tremelimumab was administered, a maximum of four cycles were given. The primary end point of the study was ORR, which was highest in the arms in which ICIs were administered from cycle 1 onwards (ORRs of 50%, 72% and 70% in arms 1, 2 and 3, respectively); however, the median PFS was similar between arms (12.8 months, 11.8 months and 12.3 months, respectively)¹⁴². In view of a more favourable toxicity profile without compromised efficacy in the arm without tremelimumab, the combination of gemcitabine, cisplatin and durvalumab was selected for further assessment in the phase III TOPAZ-1 trial.

In TOPAZ-1, 685 patients with treatment-naive advanced-stage CCAs or gallbladder cancer were randomly assigned to receive gemcitabine and cisplatin plus either durvalumab or placebo¹⁴³. The primary end point was OS, which was significantly improved with durvalumab (median 12.8 months versus 11.5 months in the placebo group; 24-month OS 24.9% versus 10.4%; HR 0.80, 95% CI 0.66–0.97; P = 0.021)¹⁴³. The durvalumab group also had improvements in PFS (median 7.2 months versus 5.7 months; HR 0.75, 95% CI 0.63–0.89; P = 0.001) and ORR (26.7% versus 18.7%; OR 1.60, 95% CI 1.11–2.31)¹⁴³. This improved activity was reported without additional toxicity (for example, grade 3−4 adverse events in 75.7% versus 77.8% of patients)¹⁴³, and with a trend towards a longer time to deterioration in patient-reported global health status according to EORTC QLQ C30 scores favouring the durvalumab arm (HR 0.87, 95% CI 0.69–1.12; P = 0.279)¹⁴⁴. Interestingly, subgroup analyses did not reveal significant differences in outcomes depending on PD-L1 expression¹⁴³, primary tumour site¹⁴⁵, disease status (that is, initially unresectable versus recurrent)¹⁴⁶ or geographical area (Asian versus rest of the world)¹⁴⁷. On the basis of these results, the FDA approved the combination of gemcitabine, cisplatin and durvalumab in 2022 (ref. 148), and the latest NCCN guidelines for hepatobiliary cancer (version 1.2022)⁹⁰ recommend this regimen as the preferred first-line treatment option for advanced-stage CCA. Similarly, the KEYNOTE-966 trial, the largest phase III trial in patients with CCA to date, evaluated the combination of chemotherapy and ICI therapy in the first-line treatment of patients with unresectable, locally advanced or metastatic CCA or gallbladder cancer¹⁴⁹. In this randomized, double-blind trial, 533 patients were allocated to receive pembrolizumab plus gemcitabine and cisplatin and 536 patients to receive placebo plus gemcitabine and cisplatin¹⁴⁹. The primary end point was median OS, which was 12.7 months in the pembrolizumab plus gemcitabine and cisplatin group compared with 10.9 months in the placebo group (HR 0.83, 95% CI 0.72–0.95; one-sided P = 0.0034)¹⁴⁹. In contrast to the TOPAZ-1 trial, maintenance gemcitabine was allowed in both arms without a maximum duration while cisplatin was administered for a maximum of eight cycles¹⁴⁹. This may have accounted for the ORR being the same in both arms (29%)¹⁴⁹. On the basis of these two trials, the role of ICIs in the first-line treatment of advanced-stage CCA has been firmly established.

Several other clinical trials of ICIs are ongoing in patients with BTCs¹²⁸. Beyond their addition to chemotherapy, combinations of ICIs with TKIs have also been tested, albeit early results have been mediocre. For example, a small-cohort phase II study found an ORR of only 10% with the combination of lenvatinib plus pembrolizumab in patients with previously treated advanced-stage BTCs¹⁵⁰. Nevertheless, other ongoing clinical trials are exploring this potential alternative, chemotherapy-free ICI-based therapeutic strategy (NCT04211168).

Thus, the future of immunotherapy for CCA looks bright for the first time. Efforts to identify predictive biomarkers for patient selection and to understand the mechanisms of resistance will be paramount in order to realize the full potential of ICIs and other immunotherapies.

Liver transplantation for iCCA

Transplant oncology is a developing field that merges the disciplines of organ transplantation and oncology¹⁵¹. iCCA is one of the paradigms of transplant oncology, but remains a formal contraindication for liver transplantation outside clinical trials in most jurisdictions. Two distinct populations of patients with iCCA might benefit from liver transplantation: (1) patients with decompensated cirrhosis who have a liver mass with imaging characteristics that are not diagnostic of hepatocellular carcinoma and are found to have iCCA following biopsy sampling; and (2) those with multifocal and large iCCAs usually occurring in otherwise healthy livers.

In the past 10 years, enthusiasm to incorporate iCCA as an indication for liver transplantation has increased after data from several retrospective studies demonstrated excellent outcomes in patients with cirrhosis and with ‘very early’ and ‘early’ iCCAs (defined as single tumours ≤2 cm and ≤5 cm, respectively) who received a liver transplant (Table 3). Most of these studies involved patients either initially misdiagnosed with hepatocellular carcinoma or with an incidental iCCA mass found on explant pathology. A prospective phase II trial of liver transplantation in patients with cirrhosis and very early iCCA (NCT02878473) is now underway (Table 3); however, accruing patients has been challenging owing to the difficulty of listing such patients for transplantation and finding an allograft. Before liver transplantation can be recommended for such patients outside clinical trials, several questions remain to be answered. First, do patients with a known diagnosis of iCCA have similar outcomes to those in whom iCCA was diagnosed incidentally? Second, should these patients undergo neoadjuvant treatments? Third, should this transplant indication be limited to single iCCAs of ≤2 cm or could this be expanded to those with single larger tumours?

Table 3 |.

Summary of key contemporary studies of liver transplantation in patients with Icca

Study (year of publication)	Tumour size (n)	iCCAs discovered incidentally/initially misdiagnosed as HCC	Patients with cirrhosis (n)	OS	RFS	Recurrence rate	Chemotherapy treatment
Retrospective studies in patients with incidental or misdiagnosed small iCCAs, mainly associated with cirrhosis
Sapisochin et al. (2014)156	Single ≤2 cm (8) Multiple or single >2 cm (21)	23 (79%)/6 (21%)	29 (100%)	Overall: 79% at 1 year; 61% at 3 years; 45% at 5 years. ≤2 cm tumour group: 100% at 1 year; 73% at 3 years; 73% at 5 years. >2 cm tumour group: 71% at 1 year; 43% at 3 years; 34% at 5 years	NR	Overall: 11% at 1 year; 29% at 3 years; 29% at 5 years. ≤2 cm tumour group: 0% at 1 year; 0% at 3 years; 0% at 5 years. >2 cm tumour group: 26% at 1 year; 42% at 3 years; 42% at 5 years	None
Sapisochin et al. (2016)157	Single >2 cm or multiple tumours of any size (33)	8 (24%)/25 (76%)	33 (100%)	1 year: 79% 3 years: 50% 5 years: 45%	NR	1 year: 30% 3 years: 47% 5 years: 61%	None
	Single ≤2 cm (15)	7 (47%)/8 (53%)	15 (100%)	1 year: 93% 3 years: 84% 5 years: 65%	NR	1 year: 7% 3 years: 18% 5 years: 18%	None
Jung et al. (2017)158	Single ≤2 cm (4) Multiple ≤2 cm (2) Single >2 cm (10)	1 (6%)/14 (88%)	12 (13%)	1 year: 81% (63%a) 3 years: 52% (63%a) 5 years: 52%	NR	1 year:56% (50%a) 3 years:56% (50%a) 5 years:78%	None
De Martin et al. (2020)159	Largest nodule ≤5 cm (49; 24 patients with iCCA only combined with 25 patients with mixed HCC–iCCA for analyses)	5 (10%)/44 (90%)	49 (100%)	1 year: 90% (88%b, 92%c) 3 years: 76% (65%b, 87%c) 5 years: 67% (65%b, 69%c)	1 year: 87% (81%b, 78%c) 3 years: 79% (74%b, 69%c) 5 years: 75% (74%b, 55%c)	NR	None
Retrospective studies in patients with large multifocal iCCAs diagnosed prior to liver transplantation
Lunsford et al. (2018)160	Among 6 patients: median number of tumours 4 (IQR 3.0–5.8); median cumulative diameter 10.5 cm (IQR 7.0–13.5 cm)	NA	1 (17%)	1 year: 100% 3 years: 83.3% 5 years: 83.3%	1 year: 50% 3 years: 50% 5 years: 50%	1 year: 50% 3 years: 50% 5 years: 50%	Yes, neoadjuvant and adjuvant
Ito et al. (2022)161	22 of 31 patients (71%) had tumours larger ≥5 cm	NA	14 (45%)	1 year: 80% 3 years: 63% 5 years: 49% 5-year OS was 100% in patients who received both neoadjuvant chemotherapy and locoregional therapy, compared with 77%, 57% and 41% at 1, 3 and 5 years, respectively in those who did not or received neoadjuvant locoregional therapy alone	5 years 42%	NR	Yes, neoadjuvant, sometimes combined with locoregional therapy (10 patients received neoadjuvant locoregional therapy with or without neoadjuvant chemotherapy); 11 patients received adjuvant chemotherapy
Retrospective studies in patients with large multifocal iCCAs diagnosed prior to liver transplantation
McMillan et al. (2022)152	Among 18 patients: median number of tumours 2 (1–11); median tumour size 10.4 cm (2.5–19.9 cm)	NA	6 (33%)	1 year: 100% 3 years: 71% 5 years: 57%	3 years: 52%	7 (38.9%); median time to recurrence 11 months (range 4–42 months)	Yes, neoadjuvant and adjuvant
Ongoing prospective clinical trials
University Health Network Toronto phase II multicentre trial NCT02878473 (NA)	Single iCCA ≤2 cm with no extrahepatic disease or vascular invasion (target of 30 patients who are not candidates for tumour resection and have serum CA19.9 levels ≤ 100 ng/ml)	NA	100%	NR (5-year OS is primary end point)	NR (secondary end point)	NR (secondary end point)	NR
Oslo University phase II TESLA trial NCT04556214 (NA)	NA; histologically proven iCCA with no extrahepatic disease, vascular invasion or lymph node involvement (target of 15 patients who are not candidates for tumour resection)	NA	NR	NR (3-year OS is primary end point)	NR (secondary end point)	NR	Patients must have stable disease after ≥6 months of systemic chemotherapy or locoregional treatment
University Health Network Toronto phase II single-centre Trial NCT04195503 (NA)	NA; histologically proven iCCA with no extrahepatic disease, vascular invasion or lymph node involvement (target of 10 patients who are not candidates for tumour resection and have a potential live liver donor)	NA	NR	NR (5-year OS is primary end point; 1-year OS is secondary end point)	NR (5-year RFS is secondary end point)	NR	Patients must have stable disease or tumour regression lasting ≥6 months after systemic chemotherapy or locoregional treatment

HCC, hepatocellular carcinoma; iCCA, intrahepatic cholangiocarcinoma; IQR, interquartile range; NA, not applicable; NR, not reported; OS, overall survival; RFS, recurrence-free survival.

^aData for patients with iCCA alone (n = 8), as opposed to concurrent iCCA and HCC (n = 7) or mixed HCC–iCCA (n = 1).

^bData for patients with iCCA and mixed HCC–iCCA tumours >2 cm but ≤5 cm (n = 24).

^cData for patients with iCCA ≤2 cm and mixed HCC–iCCA tumours (n = 25).

With advances in the understanding of iCCA biology and improvements in patient management, the concept of considering liver transplantation has also been revisited for patients with unresectable iCCA, no extrahepatic disease and a sustained response to systemic and/or local therapies¹⁸. The largest prospective study to date involved a series of 65 patients referred for liver transplantation evaluation; 37 patients met the criteria and were listed, although five patients were subsequently excluded owing to conversion to resectable disease with the stipulated neoadjuvant chemotherapy, and 18 eventually received a liver transplant¹⁵². The results of that study demonstrated a 5-year OS of 57% and a 3-year disease-free survival of 52% after transplantation¹⁵². Several questions remain, and these results need to be externally validated, although they are encouraging and will serve as a baseline in future studies to identify the optimal candidates for liver transplantation¹⁵³. To our knowledge, several centres are now considering liver transplantation as an option for patients with unresectable locally advanced iCCA who have a sustained response to systemic and/or local therapies and no evidence of extrahepatic disease at any time point. With the development of molecularly targeted therapies for iCCA and a better understanding of mutations associated with a favourable prognosis, such as FGFR genetic aberrations¹⁵⁴, the proportion of patients who might be eligible for liver transplantation in the future might increase. When considering alternative treatments for these patients, the options are extremely limited owing to very poor outcomes with resection or hepatic artery infusion pump chemotherapy¹⁵⁵.

In the next few years, more data on liver transplantation for iCCA will be published and this treatment modality is expected to become available to patients at many centres. Until then, liver transplantation should be performed under strict research protocols with stringent inclusion criteria.

Conclusions

As is evident from this update, many crucial scientific and clinical questions remain despite the recent progress in CCA research and management. Although the TIME presents potential targets for the treatment of CCA, the limited antitumour activity of ICIs in this disease highlights the need for a more in-depth and fundamental understanding of the immunosuppressive cell populations. Further information will also be needed to stratify TIME-based subgroups for a precision therapy approach, and perhaps some subtypes such as the immune-desert tumours will prove difficult if not impossible to target effectively. Given the heterogeneity of CAFs in CCAs, targeting this cell population is also likely to prove more difficult than originally conceptualized. Better diagnostic modalities are necessary for pCCA and perhaps the emerging technology of measuring cfDNA and ctDNA will address this problem; the use of such assays would also greatly facilitate monitoring of disease activity following surgical resection or during systemic therapy. Indeed, we are optimistic that these liquid biopsy assays will achieve clinical utility in the management of CCA. Although an important advance, targeted therapies for patients with FGFR2 fusions or IDH1 mutations are often met with intrinsic resistance or a short durability of response owing to acquired resistance; combinatorial approaches will need to be examined in model systems and patients in order to improve outcomes. The use of neoadjuvant and adjuvant therapies in combination with surgery and/or locoregional therapies to improve outcomes is also an unmet need that warrants further studies. Finally, selection of patients who will benefit from liver transplantation will require national organ allocation systems to acknowledge patients with iCCA as candidates for this procedure. We look forward to future updates on CCA, which will hopefully report solutions to these unmet needs in patients with this disease.

Key points.

• Cholangiocarcinomas (CCAs) are highly desmoplastic cancers characterized by a tumour microenvironment that is poorly immunogenic, with an abundance of immunosuppressive cell types such as myeloid-derived suppressor cells and tumour-associated macrophages.

• The desmoplasia of CCAs is associated with an abundance of heterogeneous cancer-associated fibroblasts (CAFs), with several transcriptomically diverse CAF subtypes identified.

• Assessments of circulating tumour cells and cell-free tumour DNA are becoming important in the diagnosis and monitoring of CCA.

• Potent oncogenic drivers of intrahepatic CCAs (iCCAs) include fibroblast growth factor receptor 2 (FGFR2) gene fusions and neomorphic, gain-of-function variants of isocitrate dehydrogenase 1 (IDH1). FGFR2 fusions and IDH1 mutations are targetable, and approved molecularly targeted therapies are now available for the treatment of CCAs harbouring these genetic aberrations.

• The results of immune-checkpoint inhibitor monotherapy in patients with CCA have been disappointing, although the combination of gemcitabine, cisplatin and durvalumab has been approved in the first-line setting for the treatment of advanced-stage CCA.

• Liver transplantation is an emerging option for the subset of patients with early stage iCCA occurring in the setting of cirrhosis.