Genomic profiling, prognosis, and potential interventional targets in young and old patients with cholangiocarcinoma

ABSTRACT Molecular mechanisms behind potentially inferior prognosis of old cholangiocarcinoma (CCA) patients are unclear. Prevalence of interventional targets and the difference between young and old CCA patients are valuable for promising precision medicine. A total of 188 CCA patients with baseline tumor tissue samples were subgrouped into the young (≤45 years) and old (>45 years) sub-cohorts. Somatic and germline mutation profiles, differentially enriched genetic alterations, and actionable genetic alterations were compared. An external dataset was used for the validation of molecular features and the comparison of overall survival (OS). Compared to young patients, KRAS alterations were more common in old patients (P = .04), while FGFR2 fusions were less frequent (P = .05). TERT promoter mutations were exclusively detected in old patients. The external dataset (N = 392) revealed no significant difference in OS between young and old patients; however, old patient-enriched KRAS (hazard ratio [HR]: 1.96, 95% confidence interval [CI]: 1.37–2.80) and TERT alterations (HR: 2.03, 95% CI: 1.22–3.38) were associated with inferior OS. Approximately 38.3% of patients were identified of actionable oncogenic mutations indicative of a potential response to targeted therapy or immunotherapy. Actionable FGFR2 fusions (P = .01) and BRAFV600E (P = .04) mutations were more frequent in young females than old patients. The enrichment of KRAS/TERT alterations in CCA patients over 45 years resulted in inferior OS. Approximately one-third of CCA patients were eligible for targeted therapy or immunotherapy given the actionable mutations carried, especially young females.


Introduction
Cholangiocarcinoma (CCA), with three subtypes including intrahepatic cholangiocarcinoma (iCCA), perihilar cholangiocarcinoma (pCCA) and distal cholangiocarcinoma (dCCA), 1 is the second most common primary hepatic malignancy, accounting for approximately 15% primary liver tumors 2 . pCCA and dCCA are also collectively referred to as extrahepatic CCA (eCCA) 3 . Several well-known risk factors for CCA have been identified, such as alcohol consumption, smoking, hepatitis B/C virus infection, genetic alterations related to DNA repair, and the potential influence of obesity as well as drugs. 4,5 CCA patients are usually asymptomatic in early stage, 6,7 resulting disease being diagnosed in advanced stage and a poor prognosis. For CCA patients eligible for resection treatment followed by adjuvant chemotherapy, their median overall survival (mOS) and relapse-free survival (mRFS) are 51.1 and 24.4 months, respectively, 8 with a high relapse rate. 9,10 On the other hand, the expected mOS and median progression-free survival (mPFS) are 11.7 and 8.0 months, respectively, for CCA patients receiving palliative systemic chemotherapy due to unresectable disease. 8 The influence of patient age on CCA prognosis is inconsistent across various studies, with unclear potential molecular mechanisms behind. Several previous studies have reported the negative association between patient age and clinical outcomes. In the United States, old iCCA patients ≥45 years had worse 5-year survival rates when compared to young patients under 45. 11 One study focused on iCCA demonstrated that old patients had significantly inferior OS than young patients, and old age was identified as an independent prognostic factor in multivariate analyses. 12 On the other hand, in a Canadian study including 200 iCCA and 191 pCCA patients, it has been revealed that similar survival benefits from surgery and palliative chemotherapy were observed irrespective of patient age. 13 Similar results were reported in the sub-analysis of 13 prospective trials on advanced biliary cancer receiving palliative chemotherapy. 14 A number of previous sequencing studies have been completed, discovering the hotspot IDH mutation in iCCA, identifying the actionable FGFR2 fusion mutation, as well as emphasizing the genomic complexity of CCA across subtypes. [15][16][17] Nevertheless, few research works have comprehensively investigated genetic alterations in old and young CCA patients separately.
Molecular profiling of tumor tissue samples is able to guide the development of treatment option for CCA patients with advanced diseases, even though chemotherapy is currently recommended as the standard of care. The ABC−02 trial reported that patients with locally advanced or metastatic CCA, gallbladder cancer, or ampullary cancer receiving the combination of cisplatin and gemcitabine as the first-line palliative chemotherapy achieved significant survival advantage, compared to those treated with gemcitabine alone. 18 After disease progression on first-line chemotherapy, the addition of FOLFOX (folinic acid, 5-FU and oxaliplatin) to active symptom control is recommended as the standard of care for second-line treatment, owing to superior OS as well as increased 6-month and 12-month OS rates. 19 In addition to chemotherapy, targeted therapies could be options for CCA patients with locally advanced or metastatic disease, such as pemigatinib and infigratinib for patients with FGFR2 fusion or other rearrangement, 20,21 ivosidenib for previously treated patients carrying IDH−1 mutations, 22 etc. Immunotherapy might be another effective treatment option, while a subsequent trial was unable to confirm the efficacy of pembrolizumab in biliary tract carcinoma (BTC), with a responding rate as low as 6%. 23,24 Therefore, it is meaningful to well understand the proportion of CCA patients harboring each potential intervention target and the differences between age groups.
Herein, this research aimed to comprehensively studied somatic and germline mutation profiles, as well as differentially expressed genes between young and old CCA patients. CCA prognosis data from one external data set were then explored. Prevalence of actionable mutations in young and old CCA patients was also investigated separately.

Patients
Participants were retrospectively included from the database of Nanjing Geneseeq Technology Inc., between August 2016 and December 2020. Main inclusion criteria were: 1) adults ≥18 years old; 2) with pathologically confirmed CCA; 3) with baseline tumor tissue samples within 90 days after initial diagnosis and prior to systemic treatment. TNM stages in CCA were determined according to the 8 th edition of the American Joint Committee on Cancer classification. Demographics and clinical characteristics of participants, including age, gender, treatment history, the location of CCA, and family history of BTC and/or hepatocellular carcinoma (HCC), were obtained from the database of Nanjing Geneseeq Technology Inc. According to a previous study focused on a Chinese cohort in which a cutoff value of 45 years was used to identified young CCA patients, 25 our patients whose ages at initial diagnosis ≤45 years were grouped into the young subgroup, and patients over 45 years at initial diagnosis were grouped into the old subgroup. The procedures of this study were approved by the Medical Ethics Committee of Nanjing Geneseeq Medical Laboratory (NSJB-MEC−2022-02), and each patient provided written informed consent.

DNA extraction, library preparation, and NGS data processing
Genomic profiling of baseline tumor tissue samples was performed using targeted next-generation sequencing (NGS) covering 425 cancer-related genes at a centralized Clinical Laboratory Improvement Amendments-certified, College of American Pathologists-accredited clinical laboratory (Nanjing Geneseeq Technology Inc). Genomic DNA from baseline formalin-fixed paraffin-embedded (FFPE) samples was extracted with QIAamp DNA FFPE Tissue Kit (QIAGEN); genomic DNA from leukocyte controls was extracted using DNeasy Blood and Tissue Kit (QIAGEN), from peripheral blood centrifuged at 1,800× g for 10 min at room temperature within 2 h after collection. Sequencing libraries were prepared using KAPA Hyper Prep Kit (KAPA Biosystems), and targeted enrichment was performed with customized xGen lockdown probes panel (Integrated DNA Technologies), Human cot−1 DNA (Life Technologies) and xGen Universal blocking oligos (Integrated DNA Technologies). Libraries were sequenced on Illumina Hiseq NGS platforms (Illumina). All procedures were conducted following the manufacturers' instructions.
FASTQ file quality was controlled using Trimmomatic, removing leading/trailing low-quality (reading <15) or N bases. 26 Sequencing data were then aligned to the reference human genome (build hg19) and processed using the Picard suite and the Genome Analysis Toolkit (GATK). 27,28 A somatic mutation, filtered for common single nucleotide polymorphisms and germline mutations, was retained when it had at least 0.5% mutant allele frequency, at least three unique reads on different strands with good quality scores, and not present in public databases (Exome Variant Server, 1000 Genomes Project, and Exome Aggregation Consortium) at a population frequency > 1%. Gene fusions and copy number variations (CNV) were analyzed using FACTERA and ADTEx, respectively. 29,30 and manually reviewed in Integrative Genomics Viewer Software (IGV, Broad Institute). The cutoffs of retaining CNV were 1.6 for CNV gain and 0.6 for CNV loss.

Statistical analysis
Fisher's exact test and two-sample t-test were performed to compare the frequencies and means of variables between young and old CCA patients, respectively. For survival data, Kaplan-Meier curves for OS were generated, and log-rank tests were used to compare differences. Cox proportional hazards models were fitted to estimate hazard ratios (HR) and 95% confidence intervals (CI), and the proportionality of hazards was assessed using log(−log) survival plots. Data were analyzed using R software (version 4.0.3), and the survival package.
Differential enrichment analysis for signaling pathways in which altered genes were located was conducted using KEGG pathway enrichment analysis. All quoted P-values were twotailed, with values less than 0.05 considered to be statistically significant.
From one bi-institutional study, 31  external data set were downloaded from cBioPortal for Cancer Genomics (https://www.cbioportal.org/study/summary?id= ihch_msk_2021). Intriguingly, in this external cohort, the proportion of patients with TMB level ≥ 6 muts/Mb was only 11%, and no significant association between patient age and TMB level was observed (Table S1). Although no KRAS alteration was detected in the young subgroup, the prevalence of KRAS alterations in old patients was only 11%, being much lower than in our study cohort (34%) (Figure 2c). TERT promoter mutations were exclusively detected in old patients, which was consistent with the finding in our study cohort. Interestingly, none of PTEN, NFE2L2, and NOTCH1 alterations were detected in the young subgroup of the external cohort, while the proportion of old patients carrying altered PTEN (3%), NFE2L2 (2%), and NOTCH1 (1%) were similar to our study cohort. The association between patient age and FGFR2 fusions detection remained significant; however, the prevalence of FGFR2 fusion in this external cohort was higher than that in our study cohort (13% vs. 3%, P < .001).
Germline mutations were identified in three (12%) young patients and eight (5%) old patients in our study cohort. Except RECQL4 and SDHA, other mutated genes were all located in the DNA damage response pathway (Figure 2d). KEGG analysis demonstrated that alcoholism and alcoholic liver relevant genes were frequently altered among young patients; however, genetic alterations related to nonalcoholic fatty liver and gene repair were enriched in the old subgroup (Figure 2e). The result of KEGG analysis could potentially reveal distinctive risk factors for CCA in young and old people.

Inferior overall survival due to genetic alterations rather than patient age
The mOS of the external cohort was 32.0 (95% CI: 26.8-36.6) months. Compared to patients with wide-type KRAS gene, patients harboring altered KRAS genes had significantly inferior OS (mOS: 17.8 vs. 33.6 months, HR: 1.96, 95% CI: 1.37-2.80, Figure 3a). Similar results were obtained in patients with mutated TERT genes (mOS: 16.9 vs. 32.7 months, HR: 2.03, 95% CI: 1.22-3.38, Figure 3b). We were not able to observe significant differences in OS across patients with different FGFR2 mutation status. The mOS of patients with FGFR2 fusions, with other types of FGFR2 alterations, and without FGFR2 alteration were 48.2, 23.8, and 30.7 months respectively. The hazard of death was 14% lower a b c d  in CCA patients harboring FGFR2 fusion when compared to those without FGFR2 alteration, whereas the 95% CI (0.59-1.27) covered 1 ( Figure S1a). Similarly, there was no significantly different survival outcomes between patients with IDH1 mutations and without (mOS: 35.6 vs. 30.8 months, HR: 0.81, 95% CI: 0.61-1.09, Figure S1b). TP53-mutated CCA patients displayed significantly poorer OS than those with wild-type TP53 genes (mOS: 13.7 vs. 36.6 months, HR: 2.23, 95% CI: 1.67-2.99, Figure S1c). Furthermore, a trend was observed that higher TMB level might be associated with poorer OS (mOS: 26.7 vs. 34.1 months, HR: 1.25, 95% CI: 0.98-1.60, Figure S1d). We then tried confirming the previously reported association between patient age and OS. The mOS of 359 old patients and 33 young patients were 30.8 (95% CI: 26.5-36.6) months and 36.1 (95% CI: 26.7-67.8) months, respectively, without significant difference in OS between two subgroups (HR: 1.11, 95% CI: 0.71-1.73, Figure 3c). KRAS and TERT alterations, both enriched in old patients and associated with inferior OS, were potential confounders or effect modifiers of the relationship between patient age and OS. Thus, patients were further grouped according to their ages and KRAS/TERT gene alterations, and stratified analyses were conducted. For patients without altered KRAS/TERT genes, old patients (Group 2) display similar OS to young patients (Group 1) (mOS: 33.9 vs. 36.1 months, HR: 0.99, 95% CI: 0.63-1.56, Figure 3d). The association of KRAS/TERT alterations with OS remained significant when old patient with altered KRAS/TERT genes (Group 3) were compared to those without (Group 2) (mOS: 17.8 vs. 33
The proportions of patients with the potential of receiving targeted therapy were similar between young and old patients (36% vs. 27%, P = .35, Figure S2a). Females were more likely to be eligible for targeted therapy irrespective of their ages (young subgroup: 60% vs. 0%, P = .002; old subgroup: 37% vs. 20%, P = .02). Treatments targeting FGFR2 fusions (20% vs. 2%, P = .01) and BRAF V600E (13% vs. 1%, P = .04) mutations might benefit higher percentages of young females than old patients ( Figure S2b). Of note, no young male CCA patients were identified with FGFR2 fusions or BRAF V600E mutations, and ERBB2 and BRCA actionable mutations were exclusively observed in the old subgroup. Although KRAS mutations were more frequently detected in old patients than in young patients, the proportions of KRAS G12C inhibitors eligible patients were close between two subgroups (2% vs. 4%, P = .51).

Discussion
We addressed a comprehensive comparison of molecular features, prognosis data, and actionable mutations for target treatment between young and old CCA patients. Compared to 25 young patients whose CCA development mechanism was probably related to alcoholism consumption, CCA might attribute to DNA repair issues or nonalcoholic fatty liver in 163 old patients. Although no significant difference in OS was observed between young and old CCA patients, KRAS and TERT promoter mutations, which displayed higher prevalence in old patients, were associated with inferior OS. Nevertheless, compared to old patients, the proportion of patients who were potentially eligible for target treatment was likely to be larger in young patients, especially young females, due to more common FGFR2 fusions and BRAF mutations. Previous studies have not drawn consistent conclusions on whether old CCA patients have poorer prognosis than young patients. Based on 11,127 iCCA patients enrolled between 1995 and 2004, higher 1-year (HR: 1.77, 95% CI: 1.56-2.02) and 5-year all-cause mortalities (HR: 1.65, 95% CI: 1.49-1.84) were observed in patients at least 45 years old, compared to patients below 45. 11 Age was also identified as a separate factor predicting prognosis among gallbladder cancer and iCCA patients proceed with gemcitabine and S−1 combination chemotherapy as first-line palliative treatment. 32 Similarly, patient age could serve as a prognostic factor for hilar CCA patients treated with surgical resection. 33 In present study, we did not detect a significantly inferior OS among old patients, which was consistent with a couple of previous studies. For instance, in one study including 136 CCA patients, OS was not strongly associated with patient age (HR: 1.01, 95% CI: 0.99-1.03). 34 Another study focused on CCA patients treated with resection neither detected an obvious relationship between prognosis and age. 35 Of note, in a part of previous studies, patient age was treated as a numeric variable and included in regression models directly; however, in other studies where patient age was used as a categorical variable, different thresholds were applied over patient age to identify young and old patients. Moreover, the CCA subtype compositions differed across studies, and several studies focused on BTC also included patients with gallbladder cancers. Therefore, the heterogeneity of age group threshold and subtype of CCA cases might result in inconsistent conclusions on the association of patient age with prognosis.
In our study, KRAS and TERT alterations were enriched in old patients over 45 and inferior OS was associated with altered KRAS or TERT gene, which could partially explain the worse OS in old patients observed in some previous studies. One study where 81% of 195 total patients were diagnosed with iCCA also demonstrated that altered KRAS gene, detected in 13% patients, was negatively related to OS (P = .026). That study also showed KRAS alterations occurred with greater frequency in eCCA; however, the enrichment of KRAS alterations in old patients was unable to be confirmed due to lacking relevant data. 16 In another study including 85 (85/123, 69%) iCCA patients and investigating ctDNA from their blood samples, a trend of higher prevalence of altered KRAS genes among patients ≥50 years was observed, even though the difference between early-onset (age <50 years) and old patients was not statistically significant (approximately 33% vs. 21%). 36 Conversely, TERT promoter mutations appeared to be more commonly detected in early-onset patients than old patients (approximately 10% vs. 2%). In comparison with early-onset patient, TP53 mutation prevalence was relatively high in old patients (67% vs. 35%). Although we did not observe such an enrichment of mutated TP53 genes in the old subgroup of our study cohort, a strong association between TP53 mutations and poorer OS was observed. Similar negative association between TP53 mutation and clinical outcomes was also observed in BTC patients undergoing surgery treatment 37 We demonstrate, for the first time to our knowledge, the potentially distinctive prognosis between young and old CCA patients might be rationalized by different prevalence of genetic alterations, whereas larger study cohorts with diverse ethnic groups are required to further confirm our findings.
Development of drugs targeting genetic alterations in CCA has got great achievements since the genomic profiling studies being launched. Ivosidenib is one potential option for iCCA patients, of whom IDH1/2 mutations are frequently identified, after chemotherapy-refractory. In one phase III randomized controlled trial including 185 post-chemotherapy iCCA patients, significantly improved PFS was observed in patients receiving ivosidenib, in comparison with placebo (HR: 0.37, 95% CI: 0.25-0.54). 22 For patients harboring FGFR1/2/3 fusion, detected in approximately 5% patients in our study cohort, several FGFR inhibitors have been developed. 20,38 Also, the combination of BRAF and MEK inhibitors, dabrafenib plus trametinib, showed promising treatment effects among CCA patients carrying BRAF V600E mutations. 39 However, owning to the FGFR2 fusion less commonly identified among fluke-related CCA patients. 40 as well as the low prevalence of BRAF V600E mutations, targeted therapy plans should be developed depending on personal genomic profiles. Moreover, over 10% patients in our study cohort harboring BRCA1/2 mutations, ERBB2 or MET CNV gain, for which potential targeted therapy might be available. It was worth noting that only approximately 10% altered KRAS genes were KRAS G12C and most KRAS G12X mutations were other than KRAS G12C , suggesting the studies in which the efficacy of KRAS G12C inhibitors could be investigated in extended cohorts. Old patients might get great benefit from these studies, as they were observed with higher KRAS mutation prevalence than young patients. For immunotherapy, a group of studies have indicated the modest efficacy in CCA, and a phase III clinical trial of pembrolizumab combined with gemcitabine and cisplatin is ongoing. 41 Although chemotherapy is currently serving as the standard of care for CCA patients not eligible for surgical resection, personalized targeted therapy and immunotherapy guided by genomic profiles exhibited improved clinical outcomes when compared to routine chemotherapy (objective response rate: 87.5% vs. 25%, P < .001; mOS: not reached vs. 6.5 months, HR: 0.10, 95% CI: 0.02-0.48). 42 Our study does have limitations, mainly including unknown CCA subtypes, and considerable missing data of treatment history. As a result, we were neither able to conduct stratified analyses for each CCA subtype separately nor to provide a landscape view of CCA management in real world. Another limitation was the lack of clinical stages at initial diagnosis, leading to the difficulty of controlling for the potentially confounding effect of clinical stages on OS. We had no prognosis OS data in our own study cohort, and main findings in prognosis was based on the external cohort mainly consisting of CCA cases in western countries. However, genomic landscape and etiology of CCA might vary in different countries, such as the higher proportion of microsatellite instability-high CCA patients in Asian countries, 43 resulting in potentially limited generalizability of the conclusion in prognosis. Additionally, due to the lack of clinical data, we were neither able to show Carbohydrate Antigen 19-9 levels of our study cohort nor to compare the difference between young and old patients. Finally, we could not assess hepatitis B/C virus infection history because of the retrospective structure of our study, and the sample size of young patients below 45 years was relatively small. Further studies with more young patients and complete hepatitis data were needed.

Conclusions
Previously reported inferior prognosis in old CCA patients might be rationalized by more frequent KRAS/TERT alterations related to poor OS. Over 35% patients were suitable for potential personalized targeted therapy or immunotherapy, and young females were more likely to be eligible for treatment targeting FGFR fusions or BRAF V600E mutations.

Authors' contributions
YC, and FX conceived and design the study. JW, YS, JC, JL, XZ, JP, XS, YT, and QO analyzed data and interpreted results. All authors wrote, revised and reviewed the manuscript. YC, and FX supervised the whole project. All authors read and approved the final manuscript.