The most frequently mutated genes in the tumor tissue of BTCs are TP53, KRAS, CDKN2A/B, and SMAD4 (18–24). The same studies reported low prevalence (usually <5%) of targetable mutations such as IDH1/2, FGFR2, BRAF, PIK3CA, and NTRK. It is difficult to estimate the frequency of each genetic alteration for the entire group (EHC, IHC, and GBC) as the study populations were different in these studies. Among BTCs, there are some noticeable differences in the detected mutations, which are summarized in Table 1 (22, 25–34).

Targetable mutations with approved therapies (FGFR2 and IDH1/2) are more common in IHC, and IDH1/2 mutations rarely occur in EHC or GBCs (32, 33). Eight genes identified as potential drivers for IHC are TP53, KRAS, IDH1, PTEN, ARID1A, EPPK1, ECE2, and FYN (30). In multifocal metastatic IHC, the SNPs and CNVs in the primary are often concordant with the intrahepatic metastasis (same segment and <2 cm from the primary) and satellite lesions (different segment and >2 cm from the primary) (35). Clinically, this may indicate that multifocal IHCs originate from common progenitor cells and can be considered for surgical resection.

EHCs rarely have targetable mutations and are more likely than other BTCs to harbor a KRAS mutation. Precursor lesions, intraductal tubular papillary neoplasm (ITPN), predominantly originate in intrahepatic ducts and have very few mutations (36). Intraductal papillary neoplasm of the bile duct (IPNB) is usually localized in extrahepatic ducts and has mutational profiles similar to EHC. Overall, the precursor lesions and invasive CCA have overlapping mutations with few exceptions: ROBO2 mutations exist only in invasive CCA, and CTNNB1 are identified in ITPN and IPNB (36). TP53 is the most common mutation in GBC, while PIK3CA is the least prevalent (32, 33). HER2 alterations are common in GBC compared with CCAs. CNVs in CDKN2A, TP53, MDM2 proto-oncogene, and CCD1 genes and HER2 amplifications increased with the development GBC from its precursor lesions (gallstones, low-grade/high-grade dysplasia) (37). These distinct mutational profiles among BTCs can help localize the origin of the tumor and tailor therapy for individual patients.

Mutation detection in tumor tissue or blood helps identify the cell signaling pathways that play a pivotal role in carcinogenesis, drug resistance, and prognosis. Tools such as Ingenuity Pathway Analysis (IPA) were used in previous studies to correlate the genomic variations with specific signaling pathways (27). The major pathways and associated gene alterations in BTCs are as follows: i) FGF pathway with FGFR mutations; ii) mTOR with mutations such as FBXW7, PIK3CA, PTEN, NF1, NF2, PIK3R1, STK11, TSC1, and TSC2; iii) MAP/ERK pathway with KRAS, MYC, BRAF, EGFR, MAP2K1, MAP3K1, and NRAS; iv) DNA damage repair (DDR) pathway with MSH6, BRCA1, BRCA2, BAP1, ATM, MLH1, and MSH2; and v) chromatin remodeling (CR) modification pathway with BAP1, ARID1A, and PBRM1 (27).

About 19% of BTCs have DDR gene alteration mutations and usually co-exist with CR alterations such as ARID1A and PBRM1 mutations (38). These tumors tend to have a high tumor mutational burden (TMB) and a worse prognosis (38, 39). cAMP-dependent signaling activation is another pathway common for all three types of BTCs (25). Nepal et al. classified IHCs based on the three common mutations, IDH, KRAS, and TP53, or undetermined, and showed their potential predictive value in cell lines. IDH-mutated IHCs are rich in metabolic pathways such as glutathione metabolism and the citrate cycle and respond to metabolic modulators such as IDH1 inhibitors. KRAS-mutated IHCs are rich in immune-related pathways and actin cytoskeleton rearrangement and may benefit from microtubule modulators or immunotherapy. TP53-mutated tumors are rich in cell cycle dysregulation (MAPK, WNT, and p53 signaling) and may benefit from topoisomerase inhibitors. In IHC without IDH, TP53, and KRAS mutations, the mTOR pathway is predominant and may benefit from mTOR inhibitors. The relation between IDH-mutated tumors and metabolic enzymes was observed in other studies too, but rigorous preclinical and clinical studies are needed before this classification can be used in clinical practice (32).

Mutational profiling can also aid in identifying the etiology of BTCs. Fluke-positive (Opistharchis viverrine and Clonorchis sinensis related) tumors predominantly have KRAS, TP53, KMT2C (MLL3), ROBO2, RNF43, PEG3, GNAS, SMAD4, BRCA1/2, and HER2 compared with fluke-negative tumors (40–42). BAP1, IDH1/2, and FGFR mutations are frequent in the latter group (41, 42). In IHCs, HBsAg-seropositive patients most likely have TP53 mutations, and KRAS mutations are common in HBsAg-seronegative patients (30). IDH alteration-positive IHCs usually do not have any underlying risk factors such as infections, bile duct cysts, alcohol/tobacco, or Thorotrast exposure (33).

The somatic mutations with potential prognostic value are summarized in Table 2 (20, 22, 26–28, 32, 42–45). Detection of FGFR and/or IDH1 in IHC and PBRM1 and/or BAP1 in EHC is associated with a good outcome. Alternatively, PIK3CA in EHC and specific mutations in IHC, such as TP53, KRAS, CDK2NA/B, EGFR, and PBRM1, are bad prognostic markers. Mutations with prognostic value are not well established for GBCs. CCAs with mismatch repair (MMR) deficiency and low mesothelin levels have poorer outcomes compared with those with MMR proficient and high mesothelin expression (median OS: 14.5 vs. 30.0 months, p = 0.05) (46). In the same study, CCA with MMR deficiency alone had a trend toward the worst prognosis (median overall survival or OS: 19.2 vs. 28.1 months, p = 0.07). In IHC, Zhu et al. have demonstrated an interesting relation between mutations and pathological features: a) IDH1 alone or IDH/IDH2 is associated with bilobar invasion of the tumor; b) KRAS alone with positive margins (R1) and direct invasion of surrounding organs and KRAS, NRAS, or BRAF with R1; and c) NRAS with intrahepatic metastasis (47).

Studies have shown that patients with targetable mutations have better response and survival advantages if treated with the appropriate targeted therapy (20, 21). Therefore, it is crucial to attempt to identify targetable mutations when possible. For patients with no targetable mutations, there are no established mutations with predictive value. EHCs with overexpression of programmed cell death protein 1 (PD-1)/programmed death-ligand 1 (PD-L1) and higher lymphocyte infiltration respond better to ICIs (48). There are no mutations that can predict response in patients receiving chemotherapy either. With limited treatment options and poor outcomes in BTCs, this area needs further study. One area where this would be used is in the neoadjuvant setting where the lack of treatment response to chemotherapy could be identified upfront, and alternate therapy such as targeted agents or immunotherapy could be used instead (49).

Epigenetic changes are modifications in the genome not involving the nucleotide sequence (50). They can be a) DNA hypo- or hypermethylation; b) histone modification including methylation, phosphorylation, acetylation, and SUMOylation; c) CR; and d) RNA-associated silencing. In this section, the focus will be on DNA-methylation markers. Comparing the methylation profiles of malignant vs. normal or precancerous tissues can provide insight into pathways of malignant transformation and cells of origin in BTCs (36). Integrating methylation profiles with clinical and/or mutational profiles has prognostic and predictive value (28, 29, 42, 51).

In CCAs, frequently methylated genes are APC, DAPK, E-cadherin/CDH1, GSTP, RASSF1A, hMHL1, MGMT, p15^INK4b , p16^INK4a , p^14ARF , p73, 14-3-3 sigma, SOCS-3, EGFR, and RAR-β (52, 53). RASSF1A, HOXA1, HPP1, CDH1, and NEUROG1 are predominantly methylated genes in EHC, while CHFR, GSTP1, IGF2, MGMT, MINT31, p14, and RBP1 are predominant in IHC (52, 54). MGMT promoter methylation level is high, and protein expression is low in IHC compared with normal tissues (55). MGMT inhibition promotes cell proliferation via p21, P27, and Cyclin E. Low-expression MGMT in tumor tissues is associated with worse clinicopathological features and outcomes. Similarly, promoter methylation of APC, p16, and TIMP3 have a good prognostic value in IHC (56). EHCs with lymph node metastasis have higher CpG island loci and hypermethylation of T1G1 gene (54). A data mining study in an extensive database identified genes associated with methylation pathways in GBC (FGA, F2, HAO1, CFH, PIPOX, ITIH4, GNMT, MAT1A, MTHFD1, HPX, CTH, EPHX2, HSD17B6, AKR1C4, CFHR3, ENNP1, and NAT2) (57). Among them, the methylated genes were FGA, CFH, F2, HPX, and PIPOX. When validated in the clinical samples, FGA, CFH, F2, HPX, and PIPOX were high compared with controlled tissues but not significantly different. Multiple studies identified numerous genes methylated in GBC, and a set of genes consistently represented in these studies is as follows: APC, SHP1, 3-OST-2, FHIT, p16, SEMA3B, and CDH13 (58).

Integration of the genomic (somatic mutations, CNVs, and gene expression) and epigenomic data (methylation) has been attempted to better understand the disease processes. In such studies, the tumors are classified into groups, or clusters, to identify high-risk populations. In one such study, IHCs were divided into four clusters using the iClusterPlus platform (42). Cluster 1 mainly had fluke-positive IHCs rich in TP53, ARID1A, BRCA1/2, and H3K27me3 promoter mutations; had high expression of HER2, AKT1, and EZH2; and low expression of TET2 and CpG island hypermethylation. Clusters 2 and 3 had a mix of fluke-positive and fluke-negative tumors. Cluster 2 tumors were rich in TP53 mutations and had high expression of CTNNB1, WNT5B, and AKT1. Cluster 3 was rich in immune-related pathways and had the highest CNVs. Cluster 4 had tumors rich in BAP1, IDH1/2, and FGFR mutations; high FGFR1–4 gene expression; and CpG shore island hypermethylation. Cluster 4 had a better prognosis over clusters 1–3. Qui et al. divided BTCs into six clusters based on the degree of methylation (51). The cluster with a high methylation rate had the highest CNV and a worse prognosis. Lower methylation rate tumors had higher BCR/TCR diversity, immune cell infiltration, and PD-L1 and cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) mRNA expression and, hence, potentially respond to ICI better. Interestingly, hypermethylated genes participate in DNA-binding transcription activity, and hypomethylated genes are involved in transmembrane receptor and ion binding.

In another study, the authors integrated gene expression data, signaling pathways, chromosomal abnormalities, mutations, and poor prognosis signatures derived from previous studies in HCC and tyrosine kinase inhibitors in CCA (28, 59–61). They divided the IHCs into two classes: “proliferative” and “inflammatory”. The proliferative class accounted for 62% of the study population. These tumors had KRAS and BRAF mutations; more oncogenic pathways such as RAS/MAPK and MET; poor prognostic signatures; specific CNVs, amplifications at 11q13.2, deletions at 14q22.1; and moderate to poor differentiation with intraneural invasion on histology and poor survival. In the inflammatory class, tumors had activated inflammatory signaling pathways, overexpression of cytokines (interleukin or IL-3, IL-4, IL-6, IL-10, IL-17A, and CCL19), and STAT 3 activation; no poor prognostic signatures, KRAS or BRAF mutations; and well-differentiated tumors with favorable features and good survival.

IDH-mutant CCAs have a distinct molecular profile, according to a study that integrated genomic (somatic mutations and CNV) and epigenomic data (DNA-methylation) with mRNA expression (29). They have higher mitochondrial DNA copy numbers, low expression of chromatin modifiers, and elevated expression of mitochondrial genes compared with IDH-WT tumors. Comparing 103 IHCs with matched controls, Zhou et al. identified three pathways (transforming growth factor-b/Smad signaling pathway along with known Ras/phosphatidylinositol-4,5-bisphosphate 3-kinase signaling and p53/cell cycle signaling) along with genes involved in epigenetic regulation and oxidative phosphorylation more frequent in IHCs (30). Using samples from patients in Chile (which has the highest incidence of GBC), Brägelmann et al. were able to identify methylation changes during the progression from gallstone disease to dysplasia and then to GBC (37). They identified stages of progression through a sequence of early (gallstone disease and low-grade dysplasia), intermediate (high-grade dysplasia), and late (GBC) stages. In particular, methylation of genes involved in WNT signaling, Hedgehog signaling, and tumor suppression increased with tumor grade. CNVs also increased along with tumor grade (as mentioned above).

Circulating cell-free DNA (cfDNA) usually refers to nucleic acids (fragments of DNA) detected in the peripheral blood released secondary to apoptosis or necrosis (62, 63). In cancer patients, a significant proportion of cfDNA comes from normal cells of the body, a small part related to tumors, coming from primary tumors, metastatic sites, or circulating tumor cells (CTC), and is called circulating tumor-DNA (ctDNA) (64). In the last 5 years, liquid biopsies (detecting mutations in cfDNA) have gained popularity in oncology practice as they may be the only source for tumor DNA in certain cancers such as BTCs, where getting adequate tissue for sequencing is difficult. Among BTCs, acquiring tissue for EHCs is more difficult than for IHC and GBCs. In BTC, limited studies in this area indicate a high mutation detection rate (74%–85%) regardless of whether patients are treatment naive or on treatment at the time of sample collection and have reliable concordance with mutations in the tumor tissue (50%–74%) (21, 65–67). Concordance in the matched samples (blood and tissue from the same patient) depends on the source of the tissue (higher with metastatic vs. primary site) and also the tumor type [higher in IHC (92%) compared with EHC or GBC] (67).

The cfDNA has a dependable sensitivity and specificity in diagnosing BTC (68, 69). The frequency of the detected mutations in cfDNA in BTC patients is the same as those detected in the tumor tissue (21, 67). Due to limited studies, it is hard to conclusively identify cfDNA mutational profile differences between IHC, EHC, and GBCs. In one study with 69% IHCs, 18% EHC, and 13% GBC, FGFR and ARID1A were the most prevalent alteration in IHC and EHC, respectively. CDK6, APC, and SMAD4 alterations were common in GBCs. TP53 and KRAS were the most prevalent in all three groups (70). BTCs in younger patients (<50 years) have more FGFR2, PIK3CA, MET, and BRAF mutations compared with older patients (≥50 years) (70). TP53 (67% vs. 35%, p = 0.6) was predominant in the older population.

Detection of clinically relevant and targetable mutations for approved therapies is better in ctDNA than tissue, and it ranges between 30% and 40% in advanced CCAs (65, 71). In a study with 71 BTCs with detectable ctDNA, 75% of the patients had targetable mutations (both on- and off-label for BTC), indicating that this will be a valuable tool in the future (21). Variant allelic frequency (VAF) represents the percentage of mutant reads divided by the total number of reads coverage at a specific genomic position (72). VAF in ctDNA at the baseline correlates well with the tumor burden in CCA but does not have any predictive or prognostic value (65, 67). cfDNA integrity (ratio of ALU247 and ALU115) had a reliable diagnostic and prognostic value (unfavorable clinicopathological features) in GBC (68). Short segments of DNA originally characterized by Arthrobacter luteus (Alu) restriction endonucleases are ALU units (69).

CTC detection is low in CCA, especially in early-stage (stages I and II) tumors, and detection rates vary with the thresholds for positive tests (CTC ≥ 2 vs. CTC ≥ 5). In a study reported in 2016, detection rates for the threshold of ≥2 and ≥5 were 17% and 9%, respectively, and most of them were advanced-stage cancers (73). Both the thresholds had significant prognostic value in the entire group. In subgroup analyses, both maintained prognostic value in metastatic CCA and EHC. A threshold of ≥5, in addition, had prognostic value in the IHC group and trended toward significance in non-metastatic CCAs. Li et al. proposed the concept of ctDNA fingerprinting in eight tumor types, including CCA (74). The authors began with whole exosome sequencing (WES) in patients before the treatment (surgery or systemic therapy or locoregional therapy) appropriate to the disease. Then, a patient-specific panel with high-frequency clonal population clusters was designed and followed in the patients after treatment. Two entities followed in subsequent tests were ctDNA content fraction (CCF) and the fold change in CCF. The CCF is higher, and the fold change increased in patients who progress compared with patients with stable disease or response. In patients receiving selective internal radiation, a reduction in the CNVs was observed in patients with IHC (75). Platinum-based therapy showed a potential benefit in BTCs with DDR gene mutations in a small study (65). The benefit was seen in partial response rate, radiological response, and increased overall survival.

In one of the earlier studies that proved the feasibility of detecting DNA-methylation markers in CCA, a four-gene panel with HOXA1, PKRCB, CYP26C1, and PTGDR was proposed to differentiate CCA from healthy controls with sensitivity/specificity of 83%/93% (76). A couple of years later, higher cfDNA methylation levels of OPCML and HOXA9 genes, individually or together, also proved to have a good diagnostic value in distinguishing CCA from benign biliary diseases (77). The sensitivity/specificity of a combined marker is 62%/100%. Non-specific methylation levels in cfDNA do not have good diagnostic utility in GBC, with sensitivity/specificity around 55%/50% (68). We need robust studies to take the methylation markers in cfDNA to clinical practice.

Examining cytology specimens derived from endoscopic retrograde cholangiopancreatography (ERCP) to identify malignant cells is the standard test to diagnose malignancy. The addition of fluorescence in situ hybridization (FISH) and mutational profile testing to cytology helps to diagnose BTCs in patients who present with biliary strictures (78–81). Mutational profiling with KRAS and other prominent tumor-suppressor genes alone can increase the sensitivity by 56% and diagnostic yield to 100% when done in conjunction with cytology (82).

DNA-methylation markers in the biliary brushings can also be a tool for diagnosing CCA, and many studies reported using four to six gene panels for this purpose. Andersen et al. suggested that a four-gene panel, namely, CDO1, CNRIP1, SEPT9, and VIM, detected in 45%–77% of the samples had a sensitivity and specificity of 85% and 98%, respectively (83). Prachayakul et al. reported using the methylation index (MI) of two genes, HOXA1 and NEUROG1 (84). Each has better sensitivity and accuracy than cytology and CA 19-9 in differentiating CCA from benign diseases. The methylation markers were positive even in patients with negative cytology and normal CA 19-9 levels. The combined sensitivity/accuracy of MI, cytology, and CA 19-9 was 97%/91%. Yang et al. reported good sensitivity and specificity of methylated EMX1 for differentiating EHC from primary sclerosing cholangitis (PSC) and non-PSC controls (85). In the same study, HOXA1 had a sensitivity of 100% differentiating CCA from non-PSC controls, but not in PSC controls. Parsi et al. showed that 80% (N = 10) of NTCs have positive methylation for one of three genes, CCND2, NPTX2, and TFPI2 (86).

The ctDNA isolated from the bile using the techniques employed for plasma cfDNA (NGS) has shown some encouraging results (87, 88). In bile, mutations can be detected in 60%–75% of BTC patients, with specificity close to 100%. Concordance rates between bile ctDNA and tissue DNA are higher than plasma cfDNA (88%). p^16INK4a is a tumor-suppressor gene frequently associated with many cancers, including CCA (89). In bile samples, promoter methylation of p^16INK4a is positive in over 50% of CCA compared with just 6% in benign diseases (cholelithiasis) and normal bile ducts (90). Concordance with the tissue samples for p^16INK4a promoter methylation was as high as 86%. Just as with mutation profiling of biliary brushings, few panels for methylation markers in bile have been reported in the literature. In EHC, methylation of the five-gene panel with CCND2, CDH13, GRIN2B, RUNX3, and TWIST1 has a better sensitivity than cytology (83% with methylation markers vs. 46% in cytology) (91). Similarly, Zhang et al. proposed a six-gene panel (DKK3, p16, SFRP2, DKK2, NPTX2, and ppENK) to diagnose pancreatobilary cancers with sensitivity/specificity/accuracy of 77.27%/77.78%/77.50% (92). In summary, expanding precision medicine to bile and cytology brushings can improve diagnostic testing and help identify better prognostic and predictive biomarkers.

Non-DNA molecular markers include messenger RNA (mRNA), non-coding RNA (ncRNA), proteins, and metabolites. mRNA and ncRNA are within the scope of precision medicine. The ncRNA refers to RNA that is not translated into protein and can be detected in the blood and is being extensively studied in tumors for diagnostic and prognostic purposes (93, 94). They can be classified bases on their size into a) long non-coding RNAs or lcRNAs that are more than 200 bases; and b) small non-coding RNAs or sncRNAs that have up to 200 bases which are further classified into small nucleolar RNAs (snoRNAs), small nuclear RNAs (snRNAs), Piwi-RNAs (piRNAs), and microRNA or miR. Extracellular vesicles are lipid membrane-bound spheres released from cells into body fluids (blood, saliva, bile, and urine) comprising materials shed from the cells, including proteins, nucleic acids, and metabolites (95, 96). They are a good source for molecular markers in the blood. They are of three kinds: microvesicles, exosomes, and apoptotic bodies (97, 98).

Among the RNA molecular markers, miR is the preferred biomarker among all the ncRNAs in oncology as it has a proven role in pathways implicated in malignant transformation and can be detected/measured easily in blood, bile, and other body fluids; it is a more stable nucleic acid making it more reliable (94, 99–101). Even though it is still in the experimental stage, the current landscape of prominent non-DNA markers in blood and urine is summarized in Table 3 (44, 102–128).

In summary, NGS of the tumor tissue (when available) is widely used to identify the targetable mutations in the current clinical practice for managing BTCs. There is a strong clinical need to develop novel biomarkers. On the clinical side, we may see the expansion of precision medicine in three different directions soon: firstly, identifying more reliable DNA-molecular markers such as DNA-methylation markers and non-DNA markers such as mRNA, miRNA, and lncRNA; secondly, making detection of the molecular markers in non-invasive sources of genetic material such as blood (ctDNA and EV), bile, and urine feasible with acceptable reliability and accessible in day-to-day clinical practice; and finally, integrating different molecular makers from different sources and stratifying in high prognostic and predictive value. Non-invasive sources will also help in monitoring the treatment response. Novel techniques to detect ctDNA such as surface-enhanced Raman scattering (SERS) and biosensors are being studied, which may compete and replace traditional techniques (polymerase chain reaction and NGS) (129–133).

Gemcitabine/cisplatin (GC) combination is the systemic therapy of choice in advanced BTC, and capecitabine is the recommended adjuvant treatment (in case of negative margins) (134, 135). The NCCN guidelines recommend FOLFOX based on the ABC-06 trial, and the difference in OS was minimal even though statistically significant (6.2 months in the FOLFOX group vs. 5.3 months in the supportive therapy group) (136). GC and the nab-paclitaxel combination seem promising, especially in the first line (49, 137). In a recently presented trial, nanoliposomal irinotecan (nal-IRI) plus 5-FU/LV was compared with 5-FU/LV in a phase IIb randomized trial (138). The PFS (7.1 months for the combination vs. 1.4 months, p = 0.019), OS (8.6 vs. 5.5 months, p > 0.001), and ORR (14% vs. 5.8%) were better for the combination therapy which is encouraging. The serious adverse events (≥ grade 3) were very high in the combination group (77% vs. 31%).

In FGFR2 fusions or rearrangements, pemigatinib and infigratinib are recommended (139). IDH1 or IDH1 mutant BTCs can be treated with ivosidenib (14). Neurotrophic tyrosine receptor kinase (NTRK) gene rearrangement-positive tumors respond to NTRK inhibitors such as larotrectinib and entrectinib (10, 11). Dabrafenib and trametinib combination was recently approved for BRAF V600E mutation (140, 141). The list of targeted therapies is summarized in Table 4 . The published trials are cited, and the identifier is added to unpublished trials.

Treating disease by modulating (suppressing or activating) the immune system refers to immunotherapy (153). ICI is the most commonly used for immunotherapy in oncology and is used in almost all solid and hematological malignancies. In clinical practice, after progression in GC, targeted therapy or immunotherapy is considered if feasible. In patients with microsatellite instability (MSI-H), pembrolizumab is recommended based on the KEYNOTE-158 study with 22 advanced BTCs. The objective response rate was observed in 41% of BTCs with two patients with complete response (154). Correlation between TMB and response to ICI is difficult to assess as none of the BTCs enrolled in KEYNOTE-158 had higher TMB (≥10 mutations/Mb). PDL-1 overexpression does not seem to have any effect on responses either (155, 156). Nivolumab and ipilimumab combination did not have encouraging results either, with just 23% ORR (157). The current NCCN recommendation is to use nivolumab for refractory advance BTCs that are not MSI-H irrespective of TMB, and PDL-1 expression is weak (156).

Currently, many trials are combining ICI with either chemotherapy or targeted therapy or locoregional therapies such as transarterial chemoembolization (TACE), cryotherapy, radiofrequency ablation (RFA), and radiotherapy (as reported in Table 5 ). Traditional ICIs target either PD-1 or PD-L1 or CTLA-4. Bintrafusp alfa, a bifunctional antibody targeting TGF-β (transforming growth factor-β) and PD-L1, had success in the early phase trials (162). Vaccines with individual peptides such as MUC1, Wilms tumor 1 (WT1), or multiple in treating advanced BTC peptides had limited success previously but not enough to pursue it forward (170–175). There is some evidence that chimeric antigen receptor-modified T cells (CART) against epidermal growth factor receptor (EGFR) and CD133 and tumor-infiltrating lymphocytes are effective in managing refractory CCA (176–178).