Cholangiocarcinoma: Pathologic and Molecular Classification in the Era of Precision Medicine

Multiple risk factors and predisposing conditions are associated with the development of CCA, some shared by all CCA subtypes, and others specific to 1 subtype or geographic region.^2,22  To date, these include bile duct cysts, primary sclerosing cholangitis (PSC), Caroli disease, cholelithiasis/choledocholithiasis, cirrhosis, viral hepatitis B and C, liver flukes (Clonorchis sinensis, Opisthorchis viverrini), and nonalcoholic fatty liver disease (NAFLD)/nonalcoholic steatohepatitis (NASH).² 

PSC is an autoimmune disease characterized by narrowing or obstruction of both intrahepatic and extrahepatic bile ducts.^2,24  Bile duct obstruction leads to intracellular accumulation of toxic bile acids that contribute to hepatocyte and cholangiocyte damage and inflammation.²⁵  Patients with PSC have an approximately 400-fold (standardized incidence rate, 398; 95% CI, 246–608)²⁶  to 1500-fold (relative risk, 1560; 95% CI, 780–2793; P < .001)²⁷  increased risk of developing CCA as compared with the general population. CCA may present within the first year of the diagnosis of PSC, with a reported range of presentation of 0.8 to 15.0 years.^27,28 

Worldwide, choledochal cysts, choledocholithiasis, and cirrhosis (regardless of cause) were identified as the strongest risk factors for CCA in a recent meta-analysis study.²⁹  Of these, choledochal cysts had the greatest contribution to the risk of both iCCA and eCCA, with pooled odds ratios (ORs) of 26.71 (95% CI, 15.80–45.16) and 34.94 (95% CI, 24.36–50.12), respectively.²⁹  Cirrhosis has a greater contribution to the risk of iCCA than eCCA.²⁹  Cholangitis and liver fluke infections are reported as the main risk factors for large-duct–type iCCA, and cirrhosis and viral hepatitis as major risk factors for small-duct–type iCCA.^1,30 

NAFLD/NASH also have been shown to be risk factors for CCA and may contribute to the increasing incidence of CCA worldwide.^31,32  A recent meta-analysis of 7 case-control studies showed a positive association between NAFLD and CCA (pooled OR, 1.95; 95% CI, 1.36–2.79).³²  When classified according to CCA subtypes, NAFLD was a stronger risk factor for iCCA (OR, 2.22; 95% CI, 1.52–3.24) than for eCCA (OR, 1.55; 95% CI, 1.03–2.33).³²  NASH was also reported as an independent risk factor for iCCA (OR, 3.36; 95% CI, 1.15–10.2).³³ 

Although the exact molecular mechanisms of tumor induction evoked in these disparate predisposing conditions remain to be resolved, it is speculated that tumorigenesis occurs in a proinflammatory background in the setting of biliary epithelial injury, chronic inflammation, and cytokine and growth factor upregulation, all of which contribute to DNA damage, with resulting mutations in cell cycle genes.^34–37  The current World Health Organization classification identifies 2 molecular subtypes of iCCA: the inflammation and proliferation subtypes.³⁸  The inflammation subtype shows signal transducer and activator of transcription 3 (STAT3) activation with overexpression of cytokines, and generally is associated with a better prognosis.³⁸  The proliferation subtype is more commonly associated with mutations in IDH, KRAS, B-Raf proto-oncogene serine/threonine-protein kinase (BRAF), and other oncogenic signaling pathways, as well as with chromosomal instability, and is associated with a worse prognosis.³⁸ 

The most common clinical presentation of eCCA is jaundice due to bile duct obstruction; however, jaundice is an initial symptom in only 10% to 15% of patients with iCCA.^3,39  Presenting symptoms for iCCA are often not site-specific and include abdominal pain, night sweats, asthenia, nausea, and weight loss.³⁹  Altered liver function test results are commonly found at the time of diagnosis in patients with CCA, but are not specific or indicative of malignancy.^11,39  Nevertheless, elevated total and direct bilirubin, alkaline phosphatase, gamma-glutamyl transferase, and 5′-nucleotidase suggest the presence of biliary obstruction.⁴⁰  Reduced albumin and elevated alkaline phosphatase, total bilirubin, direct bilirubin, and gamma-glutamyl transpeptidase have been correlated with shorter overall survival in patients with iCCA.⁴¹ 

Radiologically, it is challenging to distinguish iCCA from pCCA because of the propensity of pCCA tumors to involve both the large ducts at the liver hilum and the perihilar parenchyma.^3,42,43  Moreover, tumors within the biliary tract are usually difficult to access for sampling, leading to insufficient or inadequate tumor tissue for diagnosis and biomarker analysis.^3,44 

CCA arises from transformed epithelial cells including hepatic or biliary tree progenitor cells, mature cholangiocytes, and hepatocytes as potential cells of origin.⁴⁵  The pCCA variant of eCCA is localized to one or both hepatic ducts and/or the junction between the left and right ducts, whereas the dCCA variant is localized in the common bile duct (Figure 1, A).^1,4  Both eCCA subtypes generally exhibit a periductal growth pattern that presents as flat or poorly defined nodular sclerosing tumor with diffuse infiltration into adjacent structures, and less frequently show an intraductal growth pattern (intraductal papillary tumor).^4,46  iCCA may occur anywhere from the periphery to central and hilar segments of the liver, arising from virtually any ductal structure from large segmental bile ducts to smaller branches of the intrahepatic biliary tree (Figure 1, A).^1,4  Three gross iCCA tumor growth patterns are reported: mass forming, periductal infiltrating, and intraductal growing (Figure 1, B and C).^1,4  Mass forming is the most common growth pattern in iCCA, present in approximately 60% of patients.⁴⁷  Small-duct iCCA generally shows a mass-forming growth pattern, whereas large-duct–type iCCA usually shows periductal growth but may also occur as an intraductal tumor.⁴ 

The morphology of CCA includes the spectrum of extremely well-differentiated gland-forming tumors to poorly differentiated adenocarcinoma, all of which tend to demonstrate a desmoplastic stromal response and frequent perineural invasion.^3,46  Cuboidal or low columnar cells are typical in CCA; however, mucinous, clear cell, micropapillary, and other variants occur rarely.³  Differentiating iCCA from benign glandular proliferations and dCCA from pancreatic ductal adenocarcinoma may be challenging (see Diagnosis and Differential Diagnosis).^3,39,48 

The World Health Organization classification of digestive system tumors (5th edition) separates iCCA into 2 main subtypes: a small-duct type, which shares many etiologic, pathogenetic, and imaging characteristics with HCC, and a large-duct type resembling eCCA.⁴⁹ 

Small-duct iCCAs, originating from interlobular and septal bile ducts lined by cuboidal cholangiocytes,⁴  manifest histologically as adenocarcinomas composed of small tubular or acinar structures with nodular growth invading the liver parenchyma and with no or minimal mucin production (Figure 2).^1,45  Large-duct iCCA, originating from tall cylindrical cholangiocytes, often shows papillary architecture with mucin-producing columnar tumor cells.^1,4  Large-duct iCCA is commonly preceded by precancerous lesions such as biliary intraepithelial neoplasia or intraductal papillary neoplasia (Figure 2).⁴⁵ 

A novel “solid tubulocystic variant” of iCCA has been described recently (previously referred to as the cholangioblastic variant of iCCA).^50,51  This variant is an inhibin-positive carcinoma with most cases reported in younger women (17–44 years of age).⁵⁰  These rare tumors have distinctive solid, tubular, and cystic growth patterns, with tubules, cysts, and solid/nested pattern interrupted by lumina. Morphologically, these resemble thyroid follicular neoplasms and granulosa cell tumors.⁵⁰  The cytology is biphasic, with small basophilic cells and large, polygonal eosinophilic cells.⁵¹  The pattern of spread and morphologic features differs from iCCA; however, based on its primary hepatic location, strong albumin in situ hybridization staining, and glandular differentiation, it is designated as a variant of CCA.⁵⁰  Recently, a novel gene fusion, nipped-B–like protein-nucleus accumbens-associated protein 1 (NIPBL-NACC1), was reported in 4 patients with tumors that are morphologically similar to solid tubulocystic iCCA.⁵¹ 

The use of a consistent staging system is important for all tumors to predict prognosis, develop treatment plans, and evaluate therapeutics. The AJCC TNM classification defines cancer stages according to the extension of primary tumor (T), regional lymph node infiltration (N), and the presence of distant metastases (M).⁵²  The 8th edition of the AJCC staging manual separates staging criteria for CCA subgroups (iCCA, pCCA, and dCCA) and contains significant updates from prior editions.⁵²  In this review, updated criteria for T-stage are discussed below. As for all solid tumors using AJCC staging criteria, final stage designations in CCA encompass lymph node and distal metastasis status.⁴⁷  The reader is referred to the 8th edition AJCC Cancer Staging Manual for details.⁵³ 

According to the AJCC 8th edition iCCA criteria, the T1 category comprises solitary tumors without vascular invasion and is divided into 2 subcategories: 1a (≤5 cm) and 1b (>5 cm), to account for the prognostic impact of tumor size.^47,52  The T2 category encompasses solitary tumors with vascular invasion or multiple tumors with or without vascular invasion.^47,52  Tumors perforating the visceral peritoneum are T3 category and tumors directly invading local extrahepatic structure are T4.^47,52 

Considering pCCA T-stage criteria, high-grade biliary intraepithelial neoplasia is classified as carcinoma in situ (Tis). The T1 category applies to tumors confined to the bile duct, with extension up to the fibrous tissue or muscle layer.^52,54  T2 tumors invade beyond the wall of the bile duct to surrounding adipose tissue (T2a) or invade adjacent hepatic parenchyma (T2b). T3 tumors invade unilateral branches of the hepatic artery or portal vein.^52,54  Tumors meeting criteria for T4 invade the common hepatic artery, or the main portal vein or its branches bilaterally, or unilateral second-order biliary radicals with contralateral portal vein or hepatic artery involvement.^52,54 

T-stage criteria for dCCA distinguish T1, T2, and T3 category tumors by the depth of invasion into the bile duct wall (<5 mm, 5–12 mm, and >12 mm, respectively).^52,55  The T4 category applies to tumors involving the celiac axis, superior mesenteric artery, and/or common hepatic artery.^52,55 

The classic morphology and immunohistochemical staining profiles for CCA subtypes are shown in Figure 2. Cytokeratin (CK) stains of various molecular weights (eg, CK7, CK19, and CK20) and other markers usually expressed in adenocarcinomas of gastrointestinal and pancreatobiliary origin (eg, BerEP4, cytoplasmic epithelial membrane antigen, carcinoembryonic antigen, and CA19-9) are reliably positive in most CCAs, although entirely nonspecific.^56–58  Compared with small-duct–type iCCA, the large-duct–type iCCA is more likely to be positive for CA19-9 (57.1% versus 8.4%) and exhibit the CK7⁺, CK19⁺, CD56⁻ phenotype (71.4% versus 26.5%).⁵⁹  The immunohistochemistry profile for eCCA also includes keratins, mucins, and tumor-associated epithelial markers (Figure 2).⁶⁰  However, because of the possibility of false positives or false negatives associated with these markers, a diagnosis of CCA should not be made on the basis of the presence or absence of immunohistochemical markers alone.⁶¹ 

Distinguishing CCA from metastatic adenocarcinoma to the hepatobiliary tree or from benign biliary lesions, such as bile duct adenomas and reactive ductular proliferations, can be challenging, largely because of similarities in clinical presentation, overlapping histopathologic features, and the paucity of iCCA-specific tissue markers.^{4,14,43,58,61}  For these reasons, iCCA is frequently a diagnosis of exclusion.⁴³  Immunohistochemical and in situ hybridization markers commonly used to differentiate iCCA and metastatic adenocarcinoma are shown in Table 1.

Differentiating CCA from HCC is usually accomplished on hematoxylin-eosin sections. In histologically challenging samples, CCA may be differentiated from HCC by the absence of hepatocytic markers including hepatocyte paraffin 1 antigen and arginase 1, and by the absence of a canalicular staining pattern with CD10 or polyclonal carcinoembryonic antigen stains (Figure 2).^3,14,58,62  Hepatocyte nuclear factor-1β (HNF-1β), positive in virtually all iCCAs as well as other adenocarcinomas (pancreatic, gallbladder, and ovarian carcinomas), is positive in only 9.2% of HCC tumors.^63,64  Conversely, a lack of HNF-1β expression may be used to exclude iCCA in the setting of adenocarcinomas of unknown primary in the liver.⁶³  Albumin mRNA, detected by in situ hybridization, is a marker of hepatocellular differentiation and is positive in most iCCAs; therefore, it cannot be used to distinguish iCCA from HCC.^12,65  Characterization of albumin expression across multiple tumor types reveals positive albumin in situ hybridization in 100% of HCCs, 81% of iCCAs, 39% of gallbladder cancer, and 20% of lung cancer samples.⁶⁵  Expression of albumin mRNA is also reported in hepatoid pancreatic adenocarcinoma (33%) and acinar cell carcinoma (29%).⁶⁵  Of note, most eCCAs are negative for albumin.¹²  In situ hybridization for α-fetoprotein can be useful to distinguish between CCA (low sensitivity and specificity) and HCC (high specificity).^55,66  In addition, MOC31 (anti–epithelial cellular adhesion molecule antibody) is positive in most cases of CCA, while typically negative in HCC.¹²  However, MOC31 may be positive in 10% to 20% of cases of poorly differentiated HCC.¹² 

In the era of precision medicine, when confronted with a liver tumor biopsy, the key factor in the approach to diagnosis is to select the minimal number of stains necessary to render the diagnosis, working in a stepwise fashion that takes the clinical history and radiologic findings into account in what can be termed a personalized approach to diagnosis.⁶⁷ 

Serum biomarkers specific for CCA have not yet been conclusively identified.¹⁸  CA19-9, although not specific for CCA, has been used to support a CCA diagnosis in the proper clinical setting, with levels greater than 1000 U/mL associated with the presence of unresectable disease.^18,68  However, sensitivity for detecting disease is low, with CA19-9 elevations found in only 50% of patients with iCCA in some studies.^68,69  A recent study reported that doublecortin-like kinase 1 (DCLK1) was present in higher concentration in serum samples from patients with iCCA or pCCA than in serum samples from patients with HCC, PSC, or cirrhosis (P < .05), and was undetectable in samples from healthy individuals.⁷⁰  Although confirmatory research is needed, this suggests DCLK1 is a potential serum biomarker for the early CCA diagnosis.

Comprehensive gene profiling studies based on NGS methodologies are defining the genomic and transcriptomic landscape in CCA and may guide the selection and optimal use of targeted therapies.^15,71  Advancements in NGS technology have led to the development of numerous molecular diagnostic assays profiling large gene panels that further confirm histopathologic diagnosis within a short time frame (Table 2).^71,72  These assays detect a wide spectrum of genomic alterations including mutations, amplifications, substitutions, deletions, insertions, copy number alterations, fusions or other rearrangements, and other genomic signatures such as tumor mutational burden and microsatellite instability.⁷² 

Commonly altered genes in iCCA are IDH1, isocitrate dehydrogenase 2 (IDH2), FGFR2, BAP1, BRAF, AT-rich interactive domain-containing protein 1A (ARID1A), TP53, SMAD family member 4 (SMAD4), KRAS, neuroblastoma RAS viral oncogene homolog (NRAS), mouse double minute 2 homolog (E3 ubiquitin-protein ligase Mdm2; MDM2), phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha (PIK3CA), protein polybromo-1 (PBRM1), and erythroblastic oncogene B b2 receptor tyrosine kinase 2 (ERBB2), whereas the commonly altered genes in eCCA are KRAS, TP53, SMAD4, ARID1A, and ERBB2 (Figure 1, A).^{1,4,15,17–20,73–79}  Alterations in BAP1 and FGFR2 have been reported to be almost exclusively in iCCA.^15,17  Notably, large-duct–type iCCA demonstrates a high mutation frequency of KRAS, TP53, SMAD4, and MDM2, whereas IDH1/2 mutations, FGFR2 fusions, and BAP1 loss are more frequently observed in small-duct–type iCCA (Figure 1, A).^{1,4,20,74–78} IDH1 and IDH2 mutations are associated with poor prognosis, whereas FGFR2 fusions are associated with better prognosis in small-duct–type iCCA.³ 

The specific types of alterations that are targeted for FGFR2 include single-nucleotide variants, gene fusions or other rearrangements or translocations, and copy number amplifications.^80,81  These alterations can lead to dimerization, kinase activation, increased ligand-binding affinity, or altered cellular localization, and in turn, upregulated and/or aberrant FGFR signaling and subsequently increased tumor growth and survival.^80,81  Resistance to FGFR2 targeted therapy can be primary (lack of initial response to the treatment) or secondary. Secondary resistance can occur through occurrence of additional alterations at specific “gatekeeper” residues in the FGFR protein, hyperactivation of a non-FGFR receptor tyrosine kinase “bypass” signaling pathway, and gene fusion.^80,82  A study of 1863 CCA cases using NGS revealed a high prevalence of ERBB family genetic alterations in iCCA (7.6%) and eCCA (15.3%).⁸³  In patients with iCCA and eCCA, amplifications in ERBB2 (HER2) were observed in 2.8% and 4.0% of the cases, and ERBB3 (HER3) amplifications were observed in 0.5% and 0.8%, respectively.⁸³ PIK3CA mutations lead to dysregulation of the PI3K/AKT pathway and have been reported in 6% to 8% of CCA tumors.^84,85  Mutation frequencies of PIK3CA in pCCA were reported to be similar to those in iCCA, but notably higher than in dCCA.⁸⁶ 

Subclasses of CCAs have distinct multilevel molecular characteristics. The Cancer Genome Atlas consortium identified a distinct subtype of CCA (predominantly intrahepatic and fluke- and hepatitis-negative) enriched for IDH mutants with distinct molecular features including low chromatin modifier and high mitochondrial gene expression, increased mitochondrial DNA copy number, and potential epigenetic silencing of ARID1A.⁸⁷ 

Recently, multiplatform molecular characterization of eCCA in a cohort of 189 patients identified 4 novel transcriptome-based molecular classes and identified 25% of tumors with actionable genomic alterations having potential prognostic and therapeutic implications.⁴⁶  Among these molecular classes, the proliferation class is suggested as a strong candidate for ERBB2 inhibitor treatment. Programmed death receptor-1 (PD-1)/programmed death ligand-1 (PD-L1) blockade is recommended for the immune class.⁴⁶  According to the National Comprehensive Cancer Network guidelines on biliary tract cancers, molecular testing should be considered before initiation of primary therapy in patients with CCA with deficient mismatch repair/microsatellite instability-high tumors or a family history of breast cancer gene 1/2 (BRCA1/2) mutations, and may include tumor mutational burden and neurotrophic tyrosine receptor kinase (NTRK) fusion analysis.¹⁹ 

In addition to identifying potential therapeutic targets, genomic testing is essential to understand treatment responses to chemotherapy and immunotherapy.^16,88,89  A study using targeted sequencing and immunohistochemical staining to identify biomarkers associated with treatment response in biliary tract cancers found that responses to chemotherapy were more favorable in small-duct–type iCCA with mutations in BAP1 and IDH1 than in large-duct–type iCCA with KRAS and SMAD4 mutations.⁸⁹  Significantly longer progression-free survival (6.4 versus 3.2 months; hazard ratio, 3.70; 95% CI, 1.72–8.33; P < .001) and overall survival (21.9 versus 13.9 months; hazard ratio, 2.22; 95% CI, 1.03–4.76; P = .037) were reported in small-duct–type versus large-duct–type iCCA.⁸⁹  Tumors with alterations in DNA damage response pathway genes (BAP1, ATM serine/threonine kinase [ATM], IDH1/2) had a higher overall response rate than those with alterations in transforming growth factor-β pathway genes (SMAD4, transforming growth factor beta receptor 2 [TGFBR2], F-box and WD repeat domain-containing 7 [FBXW7], MYC proto-oncogene, BHLH transcription factor [MYC]; 50.0% versus 9.5%, P = .015).⁸⁹  These results suggest that the distinct molecular alterations in CCA subtypes may help predict response to chemotherapy.⁸⁹ 

Molecular profiling in patients treated with anti–PD-1/PD-L1 antibodies identified KRAS alteration and chromosomal instability in tumors as independent treatment-resistant factors.⁸⁹  Patients having at least 1 of these resistance factors had shorter median progression-free survival with rapid tumor growth and poor clinical benefit to immunotherapy than those with no resistance factors. A high density of tumor-infiltrating lymphocytes in the tumor center is associated with favorable response to immunotherapy in patients with CCA.⁸⁹  Of note, CCA with either KRAS alteration or chromosomal instability showed low density of tumor-infiltrating lymphocytes in the tumor center, indicating immune-suppressive tumor microenvironments in these tumors.⁸⁹ 

Targeted therapies that are approved by the US Food and Drug Administration or are in development for the treatment of advanced CCA include fibroblast growth factor receptor inhibitors (pemigatinib,⁹⁰  infigratinib⁹¹ [distribution of infigratinib will be discontinued March 2023],⁹²  and futibatinib⁹³), isocitrate dehydrogenase inhibitors (ivosidenib and dasatinib),⁹⁴  and NTRK inhibitors (entrectinib and larotrectinib)⁹⁵  (Table 3).

In a retrospective review of treatment outcomes in patients receiving matched therapy for actionable molecular alterations based on the European Society for Medical Oncology Scale for Clinical Actionability of Molecular Target, median overall survival was significantly improved in patients with CCA receiving matched therapy (22.6 months), compared with patients without actionable alterations (14.3 months).⁹⁶  Actionable alterations included mutations in IDH, FGFR2, and BRAF genes.⁹⁶ 

NGS-based assays can be performed by using tumor tissue samples or whole blood as sources for nucleic acids.⁷²  Liquid biopsy profiling (eg, circulating tumor cells, circulating tumor DNA, RNAs or exosomes from peripheral blood samples) has also been suggested as an alternative to traditional tissue biopsies, since blood sampling is a minimally invasive procedure that can be easily repeated to examine the change in tumor-specific diagnostic, prognostic, or predictive biomarkers over time.^97,98  In a recent study on circulating tumor DNA screening in patients with advanced gastric cancer, FGFR2 amplifications were more frequently detected by circulating tumor DNA sequencing (7.7%) than by tissue analysis alone (2.6%–4.4%).⁹⁹  Despite these advantages, a lack of standardized methods for collection and analysis of liquid biopsy samples has led to variability and lack of reproducibility across studies,⁹⁷  such that liquid biopsy is not currently used in standard clinical practice.⁹⁸ 

As genomic alterations differ between iCCA and eCCA, and between the histologic subtypes within iCCA,⁷⁹  genomic screening tests may complement and provide further confirmation of the histopathologic diagnosis. Molecular characterization of CCAs using NGS-based assays can help identify CCA subtypes and shed light on tumor prognosis and treatment response.^18,79  Targeted therapies have become an important part of the treatment of CCA based on genomic findings,^18,79  supporting the practice of performing genetic testing early in the care of the patient, to identify patients who may benefit from these therapies.^18,79  A proposed diagnostic algorithm from clinical presentation to therapeutic intervention in patients with CCA is provided in Figure 3.

CCA is a group of clinically, pathologically, and genetically heterogeneous tumors that are increasing in incidence and are associated with limited survival because of frequent detection in late disease stage. Correct diagnosis of CCA requires high clinical suspicion and the astute application of histopathology and immunohistochemical/in situ hybridization testing that allows the distinction of CCA from other tumor types. Advances in knowledge of the CCA mutational landscape now place this tumor squarely within the growing group of neoplasms for which patient outcome may be improved by routine molecular profiling. Since almost 50% of patients with CCA have potentially targetable genomic alterations,^15,16  tumor genomic analysis should be considered in patients with CCA when feasible. A collaborative approach between oncologists and pathologists will optimize best practices in the use of tumor samples for patient care.

Medical writing assistance was provided by Ciara Duffy, PhD, CMPP (Evidence Scientific Solutions, Sydney, Australia), and Madeeha Aqil, PhD, MWC (Envision Pharma Group, Philadelphia, Pennsylvania), funded by Incyte Corporation.