Purpose:

Biliary tract cancers, which are rare and aggressive malignancies, are rich in clinically actionable molecular alterations. A major challenge in the field is the paucity of clinically relevant biliary tract cancer models that recapitulate the diverse molecular profiles of these tumors. The purpose of this study was to curate a collection of patient-derived xenograft (PDX) models that reflect the spectrum of genomic alterations present in biliary tract cancers to create a resource for modeling precision oncology.

Experimental Design:

PDXs were derived from biliary tract cancer samples collected from surgical resections or metastatic biopsies. Alterations present in the PDXs were identified by whole-exome sequencing and RNA sequencing. PDXs were treated with approved and investigational agents. Efficacy was assessed by change in tumor volume from baseline. Event-free survival was defined as the time to tumor doubling from baseline. Responses were categorized at day 21: >30% decrease in tumor volume = partial response, >20% increase in tumor volume = progressive disease, and any non-partial response/progressive disease was considered stable disease.

Results:

Genomic sequencing demonstrated key actionable alterations across this cohort, including alterations in FGFR2, isocitrate dehydrogenase I, ERRB2, PIK3CA, PTEN, and KRAS. RNA sequencing demonstrated fusions and expression of antibody–drug conjugate targets, including TROP2, HER2, and Nectin4. Therapeutic matching revealed objective responses to approved and investigational agents that have been shown to have antitumor activity clinically.

Conclusions:

In this study, we developed a catalog of biliary tract cancer PDXs that underwent comprehensive molecular profiling and therapeutic modeling. To date, this is one of the largest collections of biliary tract cancer PDX models and will facilitate the development of personalized treatments for patients with these aggressive malignancies.

Translational Relevance

There has historically been a limited number of relevant models for developing drugs targeting biliary tract cancers, which are aggressive, hard-to-treat malignancies. We have developed 31 biliary tract cancer patient–derived xenograft (PDX) models that have genomic and expression alterations in clinically actionable genes. As proof of principle, we demonstrate that clinically relevant drugs can cause objective responses in these models. This work demonstrates the potential of PDXs for testing biomarker-matched therapies. This PDX collection can serve as a resource to guide the development of personalized treatments for patients with these aggressive malignancies.

Biliary tract cancers, including intrahepatic cholangiocarcinoma (ICCA), extrahepatic CCA (ECCA), and gallbladder cancer, are aggressive malignancies that are within close proximity anatomically but are molecularly distinct (14). Only 20% of patients with biliary tract cancers present with surgically resectable disease, and their post-resection median survival is estimated to be 18 to 30 months (5, 6). With the growing utilization of next-generation sequencing in clinical practice, it has become apparent that biliary tract cancers are rich in actionable alterations which may be targeted either directly or indirectly with investigational or approved agents (79). Genomic profiling efforts have revealed mutations, insertion/deletions (indels), and/or copy-number variations (CNV) in several key actionable genes, including FGFR2, isocitrate dehydrogenase I (IDH1), ERBB2, KRAS, and BRAF (2, 10). This realization has transformed modern cancer care for patients with biliary tract cancers in the advanced/metastatic setting and has led to the discovery of new agents, some of which have demonstrated enhanced clinical activity in selected patients.

The doublet of gemcitabine and cisplatin in combination with immunotherapy has become the standard-of-care for patients with unresectable disease, though clinical outcomes remain poor, emphasizing the need for novel therapeutic strategies (11). Over the past several years, the treatment landscape has evolved for patients who have progressed on standard-of-care agents. The FDA has approved two FGFR inhibitors (pemigatinib and futibatinib) for patients with FGFR2 fusions/rearrangements; a third FGFR2 inhibitor (infigratinib) was approved, but the approval was withdrawn. In addition, the IDH1 inhibitor ivosidenib has been approved for patients with susceptible IDH1 mutations (1214). The tumor-agnostic approvals of larotrectinib for patients with TRK fusions, pembrolizumab for high tumor mutational burden, and dabrafenib and trametinib for BRAF V600E–mutant tumors provide further genomically matched therapeutic options (1517). Several other agents at various stages of clinical development are also being studied, including the anti–HER2-bispecific antibody zanidatamab, anti-HER2 antibody combination trastuzumab + pertuzumab, and HER2-directed antibody–drug conjugate (ADC) trastuzumab deruxtecan (T-DXd), among others (1823). Despite these advances, many patient tumors do not have alterations linked to currently approved therapies or have tumors which have several genomic drivers, highlighting the need for further preclinical study.

A driving force for precision oncology is preclinical modeling with patient-derived xenograft (PDX) models, as they recapitulate the molecular and 3D architectural characteristics of human tumors (24). As biliary tract cancers are rare cancers, there is a paucity of tumor samples and clinically relevant models of these tumors, which represents a major challenge in the study of targeted therapies for patients with advanced disease (25). To gain insight into these rare cancers and develop a resource of PDX models that could be shared with the scientific community, we developed a collection of biliary tract cancer xenografts from patients with ICCA, ECCA, and gallbladder cancer. Furthermore, each PDX tumor underwent comprehensive molecular profiling to identify functional alterations in actionable genes for therapeutic matching. We hypothesized that our collection of biliary tract cancer PDX models would enable the modeling of precision oncology with the goal of developing biomarker-selected treatment strategies to facilitate clinical translation.

Patients and consent process

Patients with biliary tract cancers provided written informed consent for tissue acquisition for PDX development by image-guided biopsy or surgery. This study was approved by the Institutional Review Board at the University of Texas MD Anderson Cancer Center (MDACC; PA14-0353) and conducted in accordance with the Declaration of Helsinki and the U.S. Common Rule. Demographic, clinical, and pathologic characteristics were retrospectively abstracted from electronic medical records and a prospectively maintained clinical genomic database.

Clinical genomic sequencing data

The clinical sequencing data on patients, when available, were obtained by retrospective review of clinical records as well as a prospectively maintained genomic testing database.

Genomic sequencing results were obtained from two publicly available data sets downloaded from the cBioPortal for Cancer Genomics (https://www.cbioportal.org/). Data from the International Cancer Genome Consortium cohort (26) included patients with CCA and ECCA, and data from the MSK-IMPACT (Memorial Sloan Kettering–Integrated Mutation Profiling of Actionable Cancer Targets) cohort (27) included biliary tract cancer data (ICCA, ECCA, and gallbladder cancer). Genes with the highest frequency of alterations (mutations, indels, and CNVs) were reported in descending order.

Whole-exome sequencing

Fragments of flash-frozen PDX tissues were lysed in buffer containing protease K, and homogenized. DNA was then extracted using Qiagen DNA Mini Kit as per the manufacturer’s protocol. Normal DNA was also purified from whole blood samples using Qiagen Blood Mini Kit as per the manufacturer’s protocol. Whole-exome sequencing (WES) of samples was performed using the Institute of Personalized Cancer Therapy Cancer Genomics Laboratory, NIH/NCI, or Translational Research to AdvanCe Therapeutics and Innovation in ONcology platform at the MDACC. The FASTQ files of the WES samples are aligned to the reference genome (human Hg19) using Burrows-Wheeler Aligner (BWA; RRID: SCR_010910; ref. 28). For sequences shorter than 100 bp, we use BWA with three mismatches, including two in the first 40 seed regions. For sequences longer than 100 bp, we use BWA-MEM with a 31 bp seed length. The aligned BAM files are then processed for mark duplication, realignment, and recalibration using Picard (RRID: SCR_006525) and Genome Analysis Toolkit (29) before any downstream analyses. Germline mutation calls are made using Platypus (30), somatic mutation calls using MuTect (31), and indel calls using Pindel (RRID: SCR_000560; ref. 32). Copy-number analysis is conducted using ExomeLyzer (33), followed by CBS segmentation (34).

RNA sequencing

Frozen tumor fragments from 17 early-passage PDX models were placed in lysis buffer and homogenized. Total RNA was isolated from the tumor lysate using Norgen BIOTEK Total RNA Purification Plus Kit. PicoGreen (Invitrogen) was used to quantify genomic RNA which was then quality assessed using the 2200 TapeStation (Agilent). Reverse transcription was performed to convert RNA from each sample into double-stranded cDNA using Ovation RNA-Seq System V2 Kit (NuGEN). Libraries were constructed using KAPA kits, and NimbleGen whole-exome V3 probes were used to capture gene expression profiles. The FASTQ files of RNA samples are processed using both STAR (35) with a two-step alignment procedure and TopHat (36) in conjunction with Cufflinks (37). Additional RNA quantification is performed using HTSeq (v0.11.0; ref. 38) with htseq-count (v2.0.2). The gene counts are then normalized and used to fit the model with the R package DESeq2 (39).

To compare RNA sequencing (RNA-seq) of PDXs to The Cancer Genome Atlas (TCGA) data, we randomly selected 30 samples from each of 17 cancer types in the TCGA database (https://gdac.broadinstitute.org/) and assembled their RNA-seq profiles. This included bladder urothelial carcinoma (BLCA), breast invasive carcinoma (BRCA), cholangiocarcinoma (CHOL), colon adenocarcinoma (COAD), rectum adenocarcinoma (READ), glioblastoma multiforme (GBM), squamous cell carcinoma of the head and neck (HNSC), kidney renal clear cell carcinoma (KIRC), kidney renal papillary cell (KIRP), low-grade glioma (LGG), liver hepatocellular carcinoma (LIHC), lung adenocarcinoma (LUAD), lung squamous cell carcinoma (LUSC), ovarian serous cystadenocarcinoma (OV), prostate adenocarcinoma (PRAD), skin cutaneous melanoma (SKCM), stomach adenocarcinoma (STAD), thyroid carcinoma (THCA), and uterine corpus endometrial carcinoma (UCEC). A panel of 857 genes was selected as having the highest pan-cancer variability. Our PDX RNA-seq data were then batch-corrected into the TCGA data using the combat function in the R package sva. Multiple Experiment Viewer (https://webmev.tm4.org/about) was then used to conduct hierarchical clustering on the log2 transformed, median-centered data.

Gene fusions

Structural variations in cancer-related genes were assessed by four fusion callers: MapSplice, FusionMap, Defuse, and TopHat-Fusion (Supplementary Table S1). Fusions which were confirmed in two or more fusion callers were reported. Functional annotations of structural variants were performed. Select fusions were validated by PCR. Based on the information of genomic fusion caller analysis derived from WES results, PCR primers to detect the fusion genes were designed (Supplementary Table S2). RNA samples were extracted from PDX tissues and then subjected to reverse transcription, followed by PCR using different annealing temperatures. PCR products were observed by DNA gels.

Biomarker actionability assessment

Actionability assessments were made by the MD Anderson Precision Oncology Decision Support (PODS) team, as previously described (40, 41). A gene is considered therapeutically actionable if there are (i) convincing data at least at the preclinical level that alteration(s) of the gene predict sensitivity or resistance to a therapy that is accessible by FDA approval or clinical trial, or (ii) alteration(s) of the gene are being selected for or are required for enrollment in an open clinical trial (42). Within the gene-level, actionability is further hierarchically specified at the alteration type and functional levels to customize actionability between activating fusions or inactivating mutations. Within alteration types, specific alterations are researched by trained PODS scientists who examine databases, literature records, and other resources to provide a narrative interpretation of the variant. Furthermore, PODS scientists provide structured data calls on the alteration’s functional effect on the protein and overall actionability based on functional and therapeutic implications pertaining to a specific drug, drug class, or clinical trial eligibility (yes, potentially, unknown, or no).

IHC

For assessment of HER2 and PTEN in PDX models, tumors from early-passage PDXs were collected, then rapidly fixed in 10% neutral buffered formalin for 24 hours, and then washed with 70% ethanol. Tissues were formalin-fixed and paraffin-embedded by the MDACC Research Histology Core Facility. A tissue microarray of biliary tract cancer PDXs was made for the purpose of biomarker assessment. HER2 and PTEN IHC assessments were performed on tissue microarray sections (three 1.0 mm representative cores per tumor) or individual tumor sections using by the MDACC Clinical Laboratory Improvement Amendments–certified laboratory and reviewed by a clinically trained pathologist. Her2 expression was scored using standardized guidelines for HER2 analysis adjusted from the 2013 American Society of Clinical Oncology/College of American Pathologists guidelines for the HercepTest scoring system for gastric adenocarcinoma (43).

PDX establishment

Animal experiments were approved by the Institutional Animal Care and Use Committee at the University of Texas MDACC. Tumor fragments from image-guided or surgical biopsies were implanted into the flanks of highly immunodeficient NOD/SCID gamma mice, followed by passaging into athymic nu/nu mice for PDX development. Confirmation of tumor origin was performed using short tandem repeat DNA fingerprinting, as described previously. Once early-passage PDX tumors were of adequate size, they were passaged into athymic nu/nu mice for experimental testing.

In vivo testing

Treatments were initiated when most tumors were between 200 and 400 mm3 in size. Groups were formed with equal average starting size within each experiment. Pemigatinib was dosed at 1 mg/kg by oral gavage once daily. Futibatinib was dosed at 5, 7.5, or 15 mg/kg orally once daily. Neratinib was dosed at 10 mg/kg orally once daily. Trastuzumab was dosed at 30 mg/kg (loading dose) followed by 15 mg/kg (maintenance dose) by i.p. injection once per week. Pertuzumab was dosed at 30 mg/kg (loading dose) followed by 15 mg/kg (maintenance dose) by i.p. injection once per week. Zanidatamab was dosed at 4, 8, or 16 mg/kg by i.v. tail vein injection twice per week. Ivosidenib was dosed at 150 mg/kg orally twice per day. BAY 1895344 was dosed at 20 mg/kg orally twice per day 3 days on/4 days off. Everolimus was dosed at 1 mg/kg orally once daily. Alpelisib was dosed 50 mg/kg orally once daily. Copanlisib was dosed at 10 mg/kg by i.v. injection once daily (2 days on/5 days off schedule). Palbociclib was dosed at 50 mg/kg orally once daily. Abemaciclib was dosed at 50 mg/kg orally once daily. Enzalutamide was dosed at 50 mg/kg orally once daily (5 days on/2 days off schedule). Bicalutamide was dosed at 20 mg/kg orally once daily. Gemcitabine was dosed at 100 mg/kg by i.v. injection twice per week. Cisplatin was dosed at 5 mg/kg by i.p. injection twice per week. Capecitabine was dosed at 360 mg/kg orally once daily. Binimetinib was dosed at 6 mg/kg orally once daily. T-DXd was dosed at 5.4 mg/kg by i.v. injection every 3 weeks, sacituzumab govitecan was dosed at 20 mg/kg by i.v. injection twice a week, and enfortumab vedotin was dosed at 3 mg/kg by i.v. injection weekly.

Materials and reagents

Zanidatamab was provided in frozen vials as a gift from Zymeworks Inc. Futibatinib was provided as a gift from Taiho Oncology. BAY 1895344 and copanlisib were provided as a gift from Bayer. Neratinib was provided as a gift from Puma Biotechnology. Pemigatinib, ivosidenib, everolimus, alpelisib, palbociclib, abemaciclib, bicalutamide, and enzalutamide were obtained from the NCI Developmental Therapeutics Program. Trastuzumab, pertuzumab, gemcitabine, cisplatin, capecitabine, T-DXd, sacituzumab govitecan, and enfortumab vedotin were obtained from the MD Anderson Pharmacy. The following antibodies were purchased for IHC staining of PDX tissues: HER2 (Ventana, #790-2991 RRID: AB_2335975), TROP2 (Abcam, #ab214488, RRID: AB_2811182), NECTIN4 (Abcam, ab192033), and PTEN (Agilent, #M3627, RRID AB_2174185). Probes for CDKN2A/B FISH were purchased from Abbott Molecular/Vysis Products.

Statistical analysis

Statistical comparisons and figures were performed using Prism version 8 (GraphPad, RRID: SCR_002798) and statistical package for the social sciences Statistics Version 24 (international business machines corporation, RRID: SCR_002865). Tumor volume (TV) was calculated as TV (mm3) = [(width)2 × length]/2, and percent change in TV from baseline was calculated as (TV, day 0 − TV, day X)/TV, day 0 × 100%. The relative treatment-to-control ratio was calculated as (TV, day 21/TV, day 0)/(Vc, day 21/Vc, day 0), in which t = treatment and c = control. Event-free survival (EFS-2) was defined as the day on which TV doubled in size from baseline. Log-rank test (GraphPad) was used to compared EFS-2 curves.

Data availability

Sequencing data are available at www.ncbi.nlm.nih.gov/bioproject/PRJNA1166965. The data generated in this study are available upon request from the corresponding author.

Characteristics of established biliary tract cancer xenograft models

PDX models were developed from patients with biliary tract cancer (Fig. 1A). Of the 97 tumor samples implanted, a total of 31 unique PDXs were successfully established from 28 patients. Three patients had multiple xenografts developed from their tumors at different time points during their clinical care. When stratified by primary tumor location, there were 21 (67.7%) ICCA, 5 (16.1%) ECCA, and 5 (16.1%) gallbladder cancer xenografts established (Fig. 1B). A major characteristic of this biliary tract cancer PDX collection is that most models were derived from biopsy samples (30 of 31) from metastatic sites (27 of 31), and the majority of patients had received at least one line of chemotherapy (28 of 31). PDXs were developed from patients who had received several different treatment modalities: 28 (90.3%) PDXs were developed from patients who had received chemotherapy, 16 (51.6%) from patients who had received targeted therapy, 9 (29.0%) from patients who had received radiotherapy, 5 (16.1%) from patients who had received immunotherapy, and 3 (9.7%) from patients who had received no therapy (Fig. 1C). The characteristics, treatment histories, and clinical biomarker results for patient with matched PDXs are shown in Table 1 and Supplementary Table S3. In Supplementary Fig. S1, we show hematoxylin and eosin images of five patient–PDX pairs in which a patient sample for a simultaneous biopsy to the PDX was available.

Figure 1.

Characteristics of established biliary tract cancer xenograft models. A, Tumor fragments were obtained by image-guided biopsy or surgical resection from patients with ICCA, ECCA, and gallbladder cancer and implanted into NSG mice for PDX generation. DNA and RNA were sequenced, and alterations in cancer-related genes were assessed. Functional annotations of somatic mutations and therapeutic matching were conducted by the MD Anderson PODS team. PDX testing was carried out in athymic nude mice. B, Distribution of PDXs generated by primary tumor (ICCA, ECCA, and gallbladder cancer). C, Types of therapies administered prior to PDX development. (A, Created with BioRender.com.) NSG, NOD/SCID gamma.

Figure 1.

Characteristics of established biliary tract cancer xenograft models. A, Tumor fragments were obtained by image-guided biopsy or surgical resection from patients with ICCA, ECCA, and gallbladder cancer and implanted into NSG mice for PDX generation. DNA and RNA were sequenced, and alterations in cancer-related genes were assessed. Functional annotations of somatic mutations and therapeutic matching were conducted by the MD Anderson PODS team. PDX testing was carried out in athymic nude mice. B, Distribution of PDXs generated by primary tumor (ICCA, ECCA, and gallbladder cancer). C, Types of therapies administered prior to PDX development. (A, Created with BioRender.com.) NSG, NOD/SCID gamma.

Close modal
Table 1.

Characteristics of patients and corresponding patient-derived biliary tract cancer xenografts.

CharacteristicICCAECCAGallbladder cancerTotal
N%N%N%N%
Age (years)  
 Median, range 58 (30–77) 71 (50–72) 70 (46–71) 60 (30–77) 
Sex  
 Female 10 47.6% 80.0% 80.0% 18 58.1% 
 Male 11 52.4% 20.0% 20.0% 13 41.9% 
Ethnicity  
 Asian 4.8% 0.0% 0.0% 3.2% 
 Hispanic or Latino 4.8% 20.0% 20.0% 9.7% 
 White 19 90.5% 80.0% 80.0% 27 87.1% 
Sampling method  
 Biopsy 20 95.2% 100.0% 100.0% 30 96.8% 
 Surgery 4.8% 0.0% 0.0% 3.2% 
Primary vs. metastasis  
 Metastasis 18 85.7% 80.0% 100.0% 27 87.1% 
 Primary 14.3% 20.0% 0.0% 12.9% 
Tissue site  
 Abdominal wall 9.5% 0.0% 40.0% 12.9% 
 Axilla 0.0% 0.0% 20.0% 3.2% 
 Liver 16 76.2% 40.0% 40.0% 20 64.5% 
 Lung 4.8% 20.0% 0.0% 6.5% 
 Neck 4.8% 0.0% 0.0% 3.2% 
 Pelvis 0.0% 20.0% 0.0% 3.2% 
 Peritoneum 0.0% 20.0% 0.0% 3.2% 
 Retroperitoneum 4.8% 0.0% 0.0% 3.2% 
Prior chemotherapy         
 No 9.5% 20.0% 0.0% 9.7% 
 Yes 19 90.5% 80.0% 100.0% 28 90.3% 
Prior targeted therapy         
 No 38.1% 80.0% 60.0% 15 48.4% 
 Yes 13 61.9% 20.0% 40.0% 16 51.6% 
Prior immunotherapy         
 No 18 85.7% 60.0% 100.0% 26 83.9% 
 Yes 14.3% 40.0% 0.0% 16.1% 
Prior radiotherapy         
 No 14 66.7% 80.0% 80.0% 22 71.0% 
 Yes 33.3% 20.0% 20.0% 29.0% 
CharacteristicICCAECCAGallbladder cancerTotal
N%N%N%N%
Age (years)  
 Median, range 58 (30–77) 71 (50–72) 70 (46–71) 60 (30–77) 
Sex  
 Female 10 47.6% 80.0% 80.0% 18 58.1% 
 Male 11 52.4% 20.0% 20.0% 13 41.9% 
Ethnicity  
 Asian 4.8% 0.0% 0.0% 3.2% 
 Hispanic or Latino 4.8% 20.0% 20.0% 9.7% 
 White 19 90.5% 80.0% 80.0% 27 87.1% 
Sampling method  
 Biopsy 20 95.2% 100.0% 100.0% 30 96.8% 
 Surgery 4.8% 0.0% 0.0% 3.2% 
Primary vs. metastasis  
 Metastasis 18 85.7% 80.0% 100.0% 27 87.1% 
 Primary 14.3% 20.0% 0.0% 12.9% 
Tissue site  
 Abdominal wall 9.5% 0.0% 40.0% 12.9% 
 Axilla 0.0% 0.0% 20.0% 3.2% 
 Liver 16 76.2% 40.0% 40.0% 20 64.5% 
 Lung 4.8% 20.0% 0.0% 6.5% 
 Neck 4.8% 0.0% 0.0% 3.2% 
 Pelvis 0.0% 20.0% 0.0% 3.2% 
 Peritoneum 0.0% 20.0% 0.0% 3.2% 
 Retroperitoneum 4.8% 0.0% 0.0% 3.2% 
Prior chemotherapy         
 No 9.5% 20.0% 0.0% 9.7% 
 Yes 19 90.5% 80.0% 100.0% 28 90.3% 
Prior targeted therapy         
 No 38.1% 80.0% 60.0% 15 48.4% 
 Yes 13 61.9% 20.0% 40.0% 16 51.6% 
Prior immunotherapy         
 No 18 85.7% 60.0% 100.0% 26 83.9% 
 Yes 14.3% 40.0% 0.0% 16.1% 
Prior radiotherapy         
 No 14 66.7% 80.0% 80.0% 22 71.0% 
 Yes 33.3% 20.0% 20.0% 29.0% 

Landscape of genomic alterations in biliary tract cancer xenograft models

To investigate the landscape of genomic alterations in our collection of biliary tract cancer PDX models, we performed WES on early-passage PDXs. We filtered somatic alterations (mutations, CNVs, indels, and fusions) across the PDXs using a list of cancer-related genes that are considered “actionable,” which indicated that alterations in such genes may be targeted either directly or indirectly with approved or investigational agents. For the overall cohort, the most prevalent somatic alterations were found in TP53, FAT1, BCL2, ROS1, JAK2, MTOR, FGFR2, and NRG1. When stratified by primary tumor location, the most frequent alterations in ICCA were TP53, FGFR2, BRCA2, FAT1, ROS1, NOTCH3, and MAP3K4; in ECCA were TP53, ROS1, NRG1, KRAS, AR, and MAP2K2; and in gallbladder cancer were ERBB2, TP53, JAK2, CDK12, and RARA (Fig. 2). We detected several alterations reported in previously published genomic sequencing studies (10), as well as those found in data from the International Cancer Genome Consortium and MSK-IMPACT data sets accessed via the cBioPortal for cancer genomics (https://www.cbioportal.org; Supplementary Fig. S2A and S2B; refs. 26, 27), However, our PDX series was enriched for actionable alterations such as HER2 (ERBB2) amplifications and FGFR2 alterations, as the PDXs were mainly developed from patients referred for treatment in an investigational cancer therapeutics program.

Figure 2.

Landscape of genomic alterations identified in biliary tract xenograft models. WES was performed on early-passage PDXs to identify alterations in cancer-related genes. Mutations, indels, CNVs, and FGFR2 structural variations/fusions were reported. PDXs were stratified by primary tumor (left to right): ICCA, ECCA, and gallbladder cancer.

Figure 2.

Landscape of genomic alterations identified in biliary tract xenograft models. WES was performed on early-passage PDXs to identify alterations in cancer-related genes. Mutations, indels, CNVs, and FGFR2 structural variations/fusions were reported. PDXs were stratified by primary tumor (left to right): ICCA, ECCA, and gallbladder cancer.

Close modal

Specific alterations in the actionable genes were then characterized as “actionable,” “potentially actionable,” “not actionable,” or “unknown,” as previously described (40), for the purpose of identifying candidate models for in vivo therapeutic matching. We found that 29 of 31 (93.5%) biliary tract cancer PDXs had one or more alterations in actionable genes in a broad range of therapeutic targets, including those that have FDA-approved indications (FGFR2 and IDH1) and others under study (HER2, CDKN2A/B, ATM, KRAS, CCND1, PIK3CA, and PTEN). Supplementary Table S4 details alterations in cancer-related genes identified on PDX WES, and Supplementary Table S5 shows alterations (including those which are not actionable or unknown) in cancer-related genes (44).

A total of 76 patients who had PDX implantations had clinical genomic testing; of these, 25 patients had PDX development. Supplementary Figure S3 demonstrates the genomics of patients with and without PDX growth. Supplementary Table S6 shows comparison of frequency of genomic alterations in patients with and without PDX growth. Of 25 patients with PDX growth, 8 had HER2-altered tumors compared with 5 of 46 who did not (P = 0.016; Χ2 analysis)

Twenty-five patients who had PDX models generated had genomic testing using clinical sequencing platforms. Supplementary Figure S4 demonstrates the comparison of alterations seen in PDXs and patients. These data included three patients who had PDXs generated in more than one time point. When patient and PDX genomics were compared, many of the key actionable alterations were preserved. Although there were some discordances as well, it is harder to interpret whether these demonstrate genomic evolution with PDX implantation or intervening therapy as the clinical genomic testing was not necessarily conducted on a synchronously obtained sample.

Efficacy of targeting actionable alterations in biliary tract xenografts

Based on the genomic profiling results and actionability of alterations, we selected PDX models to assess the monotherapy efficacy of molecularly matched therapies. We assessed agents (either investigational or approved) targeting a broad range of genomic drivers, including agents with and without FDA-approved indications in biliary tract cancer.

FGFR2 inhibitors pemigatinib and futibatinib both have been FDA-approved for CCA with FGFR2 fusions/rearrangements. We selected several ICCA models, two of which were developed from patients who previously received FGFR inhibitor therapy, with FGFR2 fusions, and assessed the efficacy of FGFR inhibitors as monotherapy (Fig. 3A). PDX.003.048, which was developed after FGFR inhibitor (pemigatinib) treatment, displayed sensitivity to both pemigatinib (EFS-2: 11 days control vs. not estimable days pemigatinib; P = 0.003) and futibatinib (EFS-2: 11 days control vs. not estimable days futibatinib; P = 0.003). PDX.003.304, which was developed from an FGFR inhibitor–naïve patient, also showed sensitivity to pemigatinib (EFS-2: 9 days control vs. 41 days pemigatinib; P = 0.001) and futibatinib (EFS-2: 9 days control vs. not estimable days futibatinib; P ≤ 0.001). Additionally, PDX.003.071, which was developed from a patient treated previously with FGFR inhibitors, most recently futibatinib, regressed when treated with futibatinib (EFS-2: 24.5 days control vs. not estimable days futibatinib; P ≤ 0.001).

Figure 3.

Efficacy of targeting genomic drivers with monotherapy in biliary tract xenografts. Curves represent percent change in TV from baseline on the y-axis and experimental day on the x-axis. A, FGFR2 fusion models tested with futibatinib (15 mg/kg once daily) and/or pemigatinib (1 mg/kg, once daily). B, Model with ERBB2 mutations and/or CNVs tested with trastuzumab [30 mg/kg (loading dose) followed by 15 mg/kg (maintenance dose) once per week], pertuzumab [30 mg/kg (loading dose) followed by 15 mg/kg (maintenance dose) once per week], and the combination. The right was treated with increasing doses of zanidatamab [this experiment was previously published and was included with permission (46)]. C,IDH1-mutant model tested with ivosidenib (150 mg/kg twice per day). D,ATM-mutant model tested with the ATR inhibitor BAY 1895344 (20 mg/kg twice per day, 3 days on/4 days off). E, Models with PI3K–AKT–mTOR pathway alterations tested with everolimus (1 mg/kg once daily), alpelisib (50 mg/kg once daily), or copanlisib (10 mg/kg, 2 days on/5 days off). F, Models with cell-cycle aberrations tested with the CDK4/6 inhibitors abemaciclib (50 mg/kg once daily) or palbociclib (50 mg/kg once daily). G, Model with an activating AR mutation tested with the AR inhibitors enzalutamide (50 mg/kg, 5 days on/2 days off) and bicalutamide (20 mg/kg once daily). Amp, amplification; Del, deletion.

Figure 3.

Efficacy of targeting genomic drivers with monotherapy in biliary tract xenografts. Curves represent percent change in TV from baseline on the y-axis and experimental day on the x-axis. A, FGFR2 fusion models tested with futibatinib (15 mg/kg once daily) and/or pemigatinib (1 mg/kg, once daily). B, Model with ERBB2 mutations and/or CNVs tested with trastuzumab [30 mg/kg (loading dose) followed by 15 mg/kg (maintenance dose) once per week], pertuzumab [30 mg/kg (loading dose) followed by 15 mg/kg (maintenance dose) once per week], and the combination. The right was treated with increasing doses of zanidatamab [this experiment was previously published and was included with permission (46)]. C,IDH1-mutant model tested with ivosidenib (150 mg/kg twice per day). D,ATM-mutant model tested with the ATR inhibitor BAY 1895344 (20 mg/kg twice per day, 3 days on/4 days off). E, Models with PI3K–AKT–mTOR pathway alterations tested with everolimus (1 mg/kg once daily), alpelisib (50 mg/kg once daily), or copanlisib (10 mg/kg, 2 days on/5 days off). F, Models with cell-cycle aberrations tested with the CDK4/6 inhibitors abemaciclib (50 mg/kg once daily) or palbociclib (50 mg/kg once daily). G, Model with an activating AR mutation tested with the AR inhibitors enzalutamide (50 mg/kg, 5 days on/2 days off) and bicalutamide (20 mg/kg once daily). Amp, amplification; Del, deletion.

Close modal

Recently, there has been great interest in targeting HER2 in HER2-amplified or HER2-expressing biliary tract cancer. The combination of two HER2-targeted mAbs, pertuzumab and trastuzumab, has recently been shown to have antitumor efficacy in HER2+ biliary tract cancer (19) and was incorporated into the National Comprehensive Cancer Network guidelines. In the ECCA model PDX.003.085, which has both amplification by next-generation sequencing and overexpression of HER2 (3+) as well as an activating ERBB2 mutation (ERBB2 S310Y), trastuzumab prolonged EFS compared with untreated control, whereas pertuzumab had no substantial antitumor activity (EFS-2: 5 days control vs. 28 days trastuzumab, P = 0.035; vs. 9 days pertuzumab, P = 0.220; Fig. 3B). Although neither single-agent treatment was able to achieve stable disease, the combination of trastuzumab + pertuzumab induced tumor regression in this model. Notably, we have recently shown that PDX.003.085 is highly sensitive to zanidatamab, a bispecific anti-HER2 antibody that has clinical efficacy in HER2+ biliary tract cancer in phase I and phase II trials (18, 45, 46). In PDX.003.085, all three dose levels of zanidatamab caused statistically significant prolongation of EFS (EFS-2: 16.5 days control vs. 28.5 days zanidatamab 4 mg/kg, P = 0.034; vs. not estimable days 8 mg/kg, P = 0.001; vs. not estimable days 16 mg/kg, P = 0.001; Fig. 3B). Importantly, this model regressed when treated with zanidatamab at both 8 and 16 mg/kg.

Two ICCA models were found to have IDH1 R132C mutations, PDX.022 and PDX.013). We treated these models with the IDH1 inhibitor ivosidenib (Fig. 3C). Ivosidenib had limited effect on either model and did not stabilize tumor growth for a significant time. PDX.022 showed a modest change in the time to tumor doubling with ivosidenib (EFS-2: 4 days control vs. 8 days ivosidenib; P = 0.044), and PDX.013I showed no difference in growth between study arms (EFS-2: 18 days control vs. 18 days ivosidenib; P = 0.446).

We also assessed the activity of agents targeting genomic drivers that have been less studied clinically in patients with biliary tract cancer, including ATM alterations, PI3K–AKT–mTOR pathway alterations, and cell-cycle aberrations. Because ATM loss has been proposed as a potential marker of ATR inhibitor response, we treated PDX.003.277, an ECCA model with ATM mutations and ATM loss, by IHC with the ATR inhibitor BAY 1895344 (Fig. 3D). BAY 1895344 treatment resulted in some growth inhibition and borderline significant prolonged time to tumor doubling (EFS-2: 7 days control vs. 18 days BAY 1895344; P = 0.027).

Next, we tested the effect of PI3K and mTOR inhibitors on two models with PI3K–AKT–mTOR pathway alterations (Fig. 3E). PDX.002 has an inactivating PTEN frameshift mutation; and in this model, mTOR inhibitor everolimus caused a doubling of EFS-2 (EFS-2: 18.5 days control vs. 36 days everolimus; P = 0.007). In PDX.003.277, which has a PIK3CA activating mutation (E545K), there was statistically significant but modest growth inhibition with PI3K inhibitors copanlisib (EFS-2: 7 days control vs. 13 days copanlisib; P = 0.031) and the α-specific PI3K alpelisib (EFS-2: 7 days control vs. 23 days alpelisib; P = 0.013). In these experiments, targeting PI3K/mTOR signaling was not able to achieve stable disease or regression.

We next tested the effects of CDK4/6 inhibitors abemaciclib and palbociclib against models with cell-cycle pathway aberrations C (Fig. 3F). PDX.022, an ICCA model with amplification of cyclin D, was not sensitive to abemaciclib (EFS-2: 5 days control vs. 9 days abemaciclib; P = 0.754). PDX.003.048, an ICCA model with CDKN2A/B deletion, was growth inhibited by palbociclib (EFS-2: 11 days control vs. 31 days palbociclib; P = 0.003), but stable disease or tumor regression was not achieved.

PDX.003.100, which has an androgen receptor mutation (R630Q), was treated with two androgen receptor inhibitors (Fig. 3H). Neither enzalutamide (EFS-2: 10.5 days control vs. 14.5 days enzalutamide; P = 0.283) nor bicalutamide (EFS-2: 10.5 days control vs. 12.5 days bicalutamide; P = 0.550) had a significant effect on growth of PDX.003.100.

In summary, PDX modeling demonstrated what we have already observed in clinical medicine. Not all “actionable” alterations are created equal, and whereas targeting FGFR2 fusions and HER2 amplification leads to clear tumor responses, targeting many other alterations either had no or limited efficacy without tumor regression. Also of note, IDH inhibition was not associated with objective responses, a finding also observed in clinical trials (12). A summary of treatment-to-control ratios and median EFS-2 data for in vivo experiments is shown in Supplementary Fig. S5 and Supplementary Table S7.

Assessing combinations of targeted therapies with chemotherapy in biliary tract xenograft models

Based on our monotherapy testing experiments, we sought to assess combinations of genomically matched therapies with chemotherapeutic agents. The combination of neratinib and capecitabine has been FDA-approved for metastatic HER2+ breast cancer. We thus tested PDX.003.085, the model with an activating ERBB2 mutation, amplification, and overexpression with a combination of neratinib and capecitabine (Fig. 4A). Both neratinib and capecitabine led to statistically significant prolongation of EFS (Supplementary Table S7). The combination extended EFS-2 further compared with the single agents, leading to tumor regression.

Figure 4.

Assessing combinations of targeted therapies with chemotherapy in biliary tract xenograft models. Curves represent percent change in TV from baseline on the y-axis and experimental day on the x-axis (left), waterfall plot with % change in TV from baseline on a select experimental day (middle), and EFS curves with an event determined as the day of tumor doubling from baseline (right). A, Combination of neratinib (10 mg/kg once daily) and capecitabine (360 mg/kg once daily) in a model with ERBB2 amplification, HER2 overexpression, and activating ERBB2 S310Y mutation. B, Combinations of pemigatinib (1 mg/kg once daily) with gemcitabine (100 mg/kg twice per week) and cisplatin (5 mg/kg twice per week) in a model with FGFR2-BICC1 fusion developed from an FGFR inhibitor–naïve patient. C, Combinations of ivosidenib (IDH1 inhibitor) with 100 mg/kg twice per week) and cisplatin (5 mg/kg twice per week) in a model with IDH1 R132C mutation developed from an IDH1 inhibitor–naïve patient. Amp, amplification.

Figure 4.

Assessing combinations of targeted therapies with chemotherapy in biliary tract xenograft models. Curves represent percent change in TV from baseline on the y-axis and experimental day on the x-axis (left), waterfall plot with % change in TV from baseline on a select experimental day (middle), and EFS curves with an event determined as the day of tumor doubling from baseline (right). A, Combination of neratinib (10 mg/kg once daily) and capecitabine (360 mg/kg once daily) in a model with ERBB2 amplification, HER2 overexpression, and activating ERBB2 S310Y mutation. B, Combinations of pemigatinib (1 mg/kg once daily) with gemcitabine (100 mg/kg twice per week) and cisplatin (5 mg/kg twice per week) in a model with FGFR2-BICC1 fusion developed from an FGFR inhibitor–naïve patient. C, Combinations of ivosidenib (IDH1 inhibitor) with 100 mg/kg twice per week) and cisplatin (5 mg/kg twice per week) in a model with IDH1 R132C mutation developed from an IDH1 inhibitor–naïve patient. Amp, amplification.

Close modal

Because the combination of gemcitabine + cisplatin remains the standard-of-care in advanced/metastatic biliary tract cancer, there is considerable clinical interest in combining chemotherapy with targeted agents. We tested the combination of pemigatinib with gemcitabine + cisplatin in PDX.003.304, the ICCA model with a FGFR2-BICC1 fusion that had not been previously exposed to FGFR inhibitor therapy in the patient (Fig. 4B). PDX.003.304 was sensitive to pemigatinib, gemcitabine, and gemcitabine + cisplatin (EFS-2: 13 days control vs. not estimable pemigatinib, P = 0.007; vs. 33 days vs. gemcitabine, P = 0.007; vs. not estimable days gemcitabine + cisplatin, P = 0.007). Both gemcitabine + cisplatin and gemcitabine + cisplatin + pemigatinib induced tumor regression, with no implants doubling in either treatment arm, but with this experimental design, the addition of pemigatinib did not confer benefit over gemcitabine + cisplatin (P ≥ 0.99).

We also tested the combination of ivosidenib with gemcitabine + cisplatin in PDX.022, a model with an IDH1 mutation (Fig. 4C). PDX.022 demonstrated limited sensitivity to ivosidenib, gemcitabine, cisplatin, or gemcitabine + cisplatin, and the combination of ivosidenib with gemcitabine + cisplatin did not result in significantly enhanced antitumor activity (P = 0.071), with continued tumor progression.

Targeting genomic drivers and co-alterations with combinatorial strategies in biliary tract xenograft models

Rational combination strategies include targeting both genomic drivers and co-alterations or other oncogenic parallel/bypass pathways. In a previous work, we have shown synergy between HER2 inhibition and CDK4/6 inhibitors (47). We used PDX.003.085, which has both ERBB2 alterations and CDKN2A/B loss (ERBB2 amplification, IHC 3+, S310Y; CDKN2A/B deletion), to assess the combination of the HER2-targeted zanidatamab plus palbociclib therapy (Fig. 5A). The combination of palbociclib + zanidatamab did not further enhance antitumor efficacy over low-dose zanidatamab alone (EFS-2: 9 days zanidatamab vs. 6 days zanidatamab + palbociclib; P = 0.996).

Figure 5.

Targeting genomic drivers and co-alterations with combinatorial strategies in biliary tract xenograft models. Curves represent percent change in TV from baseline on the y-axis and experimental day on the x-axis (left), waterfall plot with % change in TV from baseline on a select experimental day (middle), and EFS curves with an event defined as the day of tumor doubling from baseline (right). A, Combinations of zanidatamab (4 mg/kg twice per week) and palbociclib (50 mg/kg once daily) in a model with ERBB2 amplification/mutation, HER2 overexpression, and CDKN2A/B deletion. B, Combinations of futibatinib (7.5 mg/kg once daily) and palbociclib (50 mg/kg once daily) in a model with FGFR2-BICC1 fusion and CDKN2A/B loss. C, Combinations of binimetinib (6 mg/kg once daily) and palbociclib (50 mg/kg once daily) in a KRAS G12D–mutant model. Amp, amplification; Del, deletion.

Figure 5.

Targeting genomic drivers and co-alterations with combinatorial strategies in biliary tract xenograft models. Curves represent percent change in TV from baseline on the y-axis and experimental day on the x-axis (left), waterfall plot with % change in TV from baseline on a select experimental day (middle), and EFS curves with an event defined as the day of tumor doubling from baseline (right). A, Combinations of zanidatamab (4 mg/kg twice per week) and palbociclib (50 mg/kg once daily) in a model with ERBB2 amplification/mutation, HER2 overexpression, and CDKN2A/B deletion. B, Combinations of futibatinib (7.5 mg/kg once daily) and palbociclib (50 mg/kg once daily) in a model with FGFR2-BICC1 fusion and CDKN2A/B loss. C, Combinations of binimetinib (6 mg/kg once daily) and palbociclib (50 mg/kg once daily) in a KRAS G12D–mutant model. Amp, amplification; Del, deletion.

Close modal

We also tested the combination of FGFR inhibition and CDK4/6 inhibition in PDX.003.048 (FGFR2-BICC1, CDKN2A/B deletion; Fig. 5B). Both futibatinib and futibatinib + palbociclib led to prolonged tumor regression, but the combination was not an improvement over single-agent futibatinib as both regimens achieved durable disease control (EFS-2: not estimable days futibatinib vs. not estimable days futibatinib + palbociclib; P > 0.999).

The combination of MEK and CDK4/6 inhibitors for KRAS-mutant tumors (48) is being explored, including in the NCI ComboMATCH trial, so we tested this combination in PDX.003.277, a model with an activating KRAS mutation (G12D). We found that neither binimetinib nor palbociclib had single-agent activity (EFS-2: 11 days control vs. 19 days binimetinib, P = 0.580; vs. 17 days palbociclib, P = 0.406), but binimetinib + palbociclib led to stable disease and almost a tripling of EFS compared with untreated control, although this EFS difference was not statistically significant versus single agents (Fig. 5C).

RNA-seq analysis for detection of fusions, the actionable transcriptome, and expression of ADC targets

We also performed RNA-seq on the PDX models. We used four different algorithms to screen for potential gene fusion. Interestingly, there was limited overlap in RNA fusion calls by different algorithms (Fig. 6A). A total of 109 fusions were called by more than one algorithm. We validated a set of these fusions with RT-PCR, including fusions detected in the patients’ clinical testing (e.g., FGFR1-BICC in PDX.003.071) as well as several fusions not reported in the clinical patient reports (FGFR2-BICC in PDX.003.153 and HER2-CDK12 in PDX.003.285; Supplementary Fig. S6).

Figure 6.

The actionable transcriptome and expression of ADC targets. A, Venn diagram showing the overlap of fusions called between for different algorithms. B, Expression of actionable alterations in biliary tract cancer PDXs. C, Expression of selected ADC targets in biliary tract cancer PDXs. (D) Expression of TROP2, NECTIN4, and HER2 by IHC in PDX.003.184. E, Antitumor activity of TROP2 ADC sacituzumab govitecan, enfortumab vedotin, and T-DXd in PDX.003.184. Curves represent percent change in TV from baseline on the y-axis and experimental day on the x-axis (left), waterfall plot with % change in TV from baseline on a select experimental day (middle), and EFS curves with an event determined as the day of tumor doubling from baseline (right).

Figure 6.

The actionable transcriptome and expression of ADC targets. A, Venn diagram showing the overlap of fusions called between for different algorithms. B, Expression of actionable alterations in biliary tract cancer PDXs. C, Expression of selected ADC targets in biliary tract cancer PDXs. (D) Expression of TROP2, NECTIN4, and HER2 by IHC in PDX.003.184. E, Antitumor activity of TROP2 ADC sacituzumab govitecan, enfortumab vedotin, and T-DXd in PDX.003.184. Curves represent percent change in TV from baseline on the y-axis and experimental day on the x-axis (left), waterfall plot with % change in TV from baseline on a select experimental day (middle), and EFS curves with an event determined as the day of tumor doubling from baseline (right).

Close modal

We also analyzed expression of actionable transcriptomic alterations (Fig. 6B) and demonstrated that several actionable alterations were differentially expressed between PDXs from different anatomic sites, including HER2 mRNA that was overexpressed in gallbladder cancer and FGFR2 mRNA that was overexpressed in ICCA.

Five of the patients also had RNA-seq performed on a tumor sample. We selected the genes under a FDR of 0.01 and fold change ≥2 and determined that 468 genes were differentially expressed (Supplementary Table S8). Under a FDR of 0.1, and fold change ≥2, 18 genes were in potentially actionable genes (Supplementary Fig. S7). To confirm that our PDX models are similar to human CCA, we also compared our PDX gene expression data to a TCGA panel consisting of 510 samples representing 17 cancer types. Hierarchical clustering showed that our PDX models clustered solely with CHOL samples and were far from LIHC samples, validating their origin (Supplementary Fig. S8).

Although several ADCs are FDA-approved clinically, few clinical trials have focused on biliary tract cancer. We therefore assessed the expression of ADC targets. We found that several models had upregulation of ADC target genes which may be associated with sensitivity to ADC therapy, including HER2, TACSTD2 (TROP2), and NECTIN4, among others (Fig. 6C). We selected one PDX model, PDX.003.184, with RNA expression of all three targets for testing. IHC demonstrated protein expression of all three targets (Fig. 6D). We then tested this model with three ADCs targeting different targets: HER2 (T-DXd), NECTIN4 (enfortumab vedotin), and TROP2 (sacituzumab govitecan). All three ADCs had significant antitumor activity [(EFS-2: 13 days control vs. not estimable days T-DXd, enfortumab vedotin, and TROP2 ADC sacituzumab deruxtecan (P = 0.000; Fig. 6E)].

Clinical genomic profiling efforts in biliary tract cancer have identified several key alterations in compelling targets for cancer therapy, though the lack of clinically relevant biliary tract cancer models is a barrier to moving therapies forward into clinical development (25). In the present study, we developed a unique collection of biliary tract cancer PDX models to serve as a platform for strategic development of novel therapeutic strategies and clinical translation. This robust collection includes ICCA, ECCA, and gallbladder cancer models and underwent comprehensive clinical and molecular characterization. Furthermore, in vivo, we utilized therapeutic matching to assess the efficacy of several targeted agents (both approved and investigational) in monotherapy and in combination with other targeted agents/chemotherapy to determine if antitumor efficacy could be enhanced. Additionally, we performed transcriptomic analysis to identify expression of genomic drivers/tumor suppressors and targets for ADC therapies.

Our study represents one of the largest molecularly and phenotypically characterized collections of biliary tract cancer models available and will serve as an invaluable resource for the development of much needed novel therapeutic strategies for patients with these aggressive cancers. Recently, there have been prior publications that have described biliary tract cancer PDX collections. Leiting and colleagues reported the generation of a biliary tract cancer PDX collection from Mayo Clinic. They reported that unlike our series, most of their PDXs were derived from surgical samples, and only 3 of 47 models were derived from biopsies. Furthermore, they reported a lower engraftment rate for biopsies versus surgical samples (23% vs. 59%; P = 0.03). The Mayo experience was updated by Lynch and colleagues, who reported that 61 of 161 biliary tumors successfully implanted for an engraftment rate of 37.9% (49, 50). More recently, Serra-Camprubí and colleagues (51) reported a collection of 19 PDXs, mostly derived from biopsies on patients with unresectable metastatic CCA, with a take rate of 38.8%. They were able to use these models to determine genotype–phenotype correlation in the context of PARP inhibitor sensitivity. They demonstrated that whereas a model with bilallelic BRCA2 was sensitive to PARP inhibition, models bearing several other putative markers of sensitivity, such as those with IDH1, BAP1, and ARID1A/B, were not, demonstrating the potential utility of PDXs for biomarker testing and validation (51).

Our experiments demonstrated disease stabilization or durable regressions when drug classes approved or shown to be active in biliary tract cancer, including FGFR inhibitors pemigatinib and futibatinib in FGFR fusions and HER2 inhibitors zanidatamab and the combination of pertuzumab and trastuzumab in HER2-positive biliary tract cancer (13, 1821). In contrast, many other approaches had limited antitumor activity; similarly, in the clinic, several other precision oncology approaches have shown limited single-agent activity in monotherapy to date. Notably, our study was not designed as a co-clinical trial, and thus we were unable to perform a correlation analysis between antitumor activity seen in patients and PDXs derived from their tumors. However, we have recently reported that in a series of PDXs established from pretreatment biopsies on patients undergoing treatment with zanidatamab, associations were observed between antitumor activity in PDXs and matched patients in seven of eight co-clinical models tested. Thus, taken together, our work suggests that testing of antitumor activity in PDXs is indeed able to demonstrate signal of activity with novel agents, and agents that are associated with clinical responses in the clinic often have durable disease control or regression in PDXs.

Aberrations in FGF/FGFR signaling have become compelling targets for therapy as 3% to 16% of patients with ICCA have FGFR2 fusions/rearrangements, which has been marked by the recent FDA approvals of the FGFR1, FGFR2, and FGFR3 inhibitors pemigatinib, infigratinib, and futibatinib (13, 14, 20, 52). In our study, we found that several ICCA models with FGFR2 fusions demonstrated robust sensitivity to both futibatinib and pemigatinib in vivo.

Although FGFR inhibitors have demonstrated robust activity in FGFR2 fusion CCA, a clinical challenge is the presence of intrinsic or development of acquired resistance (53). De novo acquired resistance mutations in FGFR2 have been well characterized (54, 55). Serra-Camprubi and colleagues (51) reported a pair of PDXs, developed from a patient with CCA bearing a FGFR2 fusion, developed before FGFR inhibitor treatment and after progression. The tumoroids developed from the pretreatment PDX was more sensitive to pemigatinib compared with tumoroids from the postprogression PDX, and the postprogression PDX demonstrated a FGFR2 V464L mutation, which has been associated with acquired resistance (51, 54). In contrast, we observed response to FGFR inhibitors both in FGFR inhibitor–naïve PDXs and in our PDXs generated from patients after FGFR inhibitor exposure. Both models of FGFRi-exposed PDXs were generated from image-guided biopsies targeting tumors that were progressing on treatment. However, on WES, none of these PDX models had known genomic mechanisms of resistance to FGFR inhibitors, suggesting that in their absence, there may be some plasticity of FGFR resistance and cells with time during PDX development and passage may resensitize.

A study by Silverman and colleagues (56) identified that patients FGFR2 fusions and co-alterations in TP53, CDKN2A/B, or PBRM1 had worse progression-free survival durations when treated with pemigatinib. We thus hypothesized that the combination of a CDK4/6 inhibitor could enhance the efficacy of FGFR inhibitors in patients with FGFR fusions with concomitant inactivating CDKN2A/B alterations We tested this hypothesis by treating a PDX model with a FGFR2-BICC1 fusion and CDKN2A/B deletion with futibatinib alone, the CDK4/6 inhibitor palbociclib alone, or their combination. In this experiment, both futibatinib monotherapy and in combination with palbociclib led to durable tumor regression. Therefore, our experiment could not appropriately address our question. Further experiments are needed to see whether the combination could enhance antitumor activity, potentially by testing lower doses of futibatinib and by designing longer-term experiments to see whether the combination overcomes the likelihood of or time to the relapse. Further work is also needed to see if the combination would have antitumor activity after progression on FGFR inhibitors.

Mutant IDH1 is another important target for biliary tract cancer identified in approximately 10% to 20% of patients with ICCA (5759). The results from the randomized phase III ClarIDHy trial demonstrated modest clinical benefit with the mutant IDH1 inhibitor ivosidenib in patients with CCA with IDH1 mutations, leading to its FDA approval (12, 60). We tested two models (PDX.022 and PDX.013) with inactivating IDH1 R132C mutations with ivosidenib and found limited antitumor activity in monotherapy. We then tested ivosidenib combinations with gemcitabine + cisplatin and found that whereas the greatest tumor growth inhibition was found in gemcitabine + cisplatin + ivosidenib, regression was not observed. A phase 3 trial is ongoing to study this combination (NCT04088188). There is growing evidence to suggest that mutant IDH1 plays a critical role in the immune microenvironment, which may explain why minimal efficacy was observed using an immunodeficient xenograft model (57).

HER2 expression/amplification is found in ∼10% to 20% of patients with ECCA and gallbladder cancer (10, 61, 62). The phase 2 MyPathway basket trial assessed the combination of trastuzumab + pertuzumab in patients with HER2-positive metastatic biliary tract cancer and demonstrated an objective response rate (ORR) of 23% (19). The MyPathway trial was a single-arm study, and therefore the contribution of components could not be demonstrated. In our study, we demonstrated enhanced activity with the combination of trastuzumab + pertuzumab over either single agent. Several novel anti-HER2 agents are also in clinical development for patients with HER2-expressing biliary tract cancers, including the HER2-bispecific antibody zanidatamab which demonstrated an ORR of 38% in patients with HER2-positive biliary tract cancer in the phase I study and an ORR of 41.3% in patients with HER2 2+ and 3+ biliary tract cancers (18, 45). In our study, we also demonstrated the antitumor activity of zanidatamab. In the PUMA phase II clinical trial, the activity of neratinib in biliary cancer has been more modest. However, as when we tested the combination of neratinib + capecitabine, a combination that is FDA-approved for HER2-positive breast cancer, this combination was also found to be active in a biliary tract cancer model, suggesting that capecitabine combinations may be worth exploring in the setting of HER2-positive or HER2-mutant biliary tract cancer (47, 63).

There is great interest in combining targeted therapies to enhance standard options. In our study, we tested this concept with only a few experiments. FGFR fusion–bearing PDX.003.304 was already quite sensitive to gemcitabine/cisplatin, and inhibiting FGFR inhibition did not statistically enhance EFS. IDH1-mutant PDX022 was not sensitive to gemcitabine-cisplatin, and adding IDH inhibition did not significantly enhance antitumor activity. However, the enhanced antitumor activity seen with neratinib/capecitabine in HER2-amplified and -mutant PDX.003.085 highlights that at least in some contexts, targeted therapy can enhance activity of chemotherapy. Thus, further preclinical modeling is needed to determine whether targeting oncogenic drivers can enhance activity of different chemotherapeutics used in the care of patients with biliary tract cancer.

Our transcriptomic analysis demonstrated both potentially actionable fusions and variability on expression of potential therapeutic targets. The variability in detected fusions with different informatic algorithms highlights a challenge in using RNA-seq for discovery of fusions. However, this approach identified not only known FGFR2 fusions but also one that had not been detected with targeted DNA-based testing in the Clinical Laboratory Improvement Amendments environment in the patient, as well as other potentially actionable fusions such as fusions in HER2.

Our transcriptomic data also nominated several ADCs that showed efficacy in vivo in our biliary tract cancer models. Our in vivo data suggest that ADCs with both antitubulin payloads (such as vedotin, payload of enfortumab vedotin) and topoisomerase inhibitors (such as deruxtecan and SN38, payloads of T-DXd and sacituzumab govitecan, respectively) may have antitumor activity in biliary tract cancer. Notably the expression of some ADC targets seems to differ based on tumor types; for example, HER2 was expressed more in gallbladder models, CLDN18 in extrahepatic models, and CDH6 and VTCN1 (B7H4) in ICCA models (Fig. 6). Furthermore, patients differed in their ADC target expression pattern, suggesting that multiplex profiling may be able to help personalize ADC choice.

There are several limitations to this study. First, these PDXs were generated from a population of patients in a single academic cancer center in the United States, many of which had previously been evaluated for phase I trials at our institution, thus likely representing a cohort enriched for actionable genomic alterations. Second, this cohort of PDXs was mostly developed from patients with locally advanced/metastatic rather than localized early-stage disease. Third, many patients had previously been treated with chemotherapy and/or targeted therapies, which may result in selective pressures which alter the genomic profiles of these tumors. Fourth, we performed a therapeutic testing of a broad panel of genomically matched therapies. We demonstrated genotype–phenotype correlations, and our results suggested novel therapeutic strategies. However, we did not perform a deep analysis of mechanisms of action or determine mechanisms of sensitivity or resistance. Fifth, we have also been able to generate PDX-derived spheroids and cell lines from some models but have not yet determined the correlation between in vitro models and in vivo antitumor activity in PDXs. Lastly, these xenograft experiments were conducted in immunodeficient mice, which limits the ability to study the effects of immune modulation on in vivo responses.

In total, we describe the generation, molecular characterization, and enrollment in rational preclinical trials of 31 novel biliary tract cancer xenograft models that capture a diverse set of actionable genomic alterations. We envisage a research community-based effort to identify and test novel therapeutic combinations, leveraging the genomic and transcriptomic data of our cohort. In particular, we note that our models that lack currently actionable mutations provide an opportunity to generate and mine additional nongenomic data to nominate novel rational therapeutic interventions. Indeed, our observed preclinical activity of pathway-targeted agents indicates that these PDX models can be utilized for the modeling of precision oncology and may facilitate clinical translation.

T.P. DiPeri reports grants from the NIH/NCI during the conduct of the study. K.W. Evans reports grants from the NCI during the conduct of the study. H. Wang reports grants from the NIH during the conduct of the study. T.A. Yap reports other support from the University of Texas at MD Anderson Cancer Center and Seagen; personal fees from AbbVie, Acrivon, Adagene, Almac, Aduro, Amgen, Amphista, Astex, Athena, Atrin, Avenzo, Avoro, Axiom, Baptist Health Systems, BioCity Pharma, Bloom Burton, Boxer, BridGene Biosciences, Bristol Myers Squibb, C4 Therapeutics, Calithera, Cancer Research UK, Carrick Therapeutics, Circle Pharma, Cybrexa, Daiichi Sankyo, Dark Blue Therapeutics, Debiopharm, Diffusion, Duike Stree Bio, 858 Therapeutics, EcoR1 Capital, Ellipses Pharma, Entos, FoRx Therapeutics AG, Genesis Therapeutics, Genmab, Glenmark, GLG, Globe Life Sciences, Grey Wolf Therapeutics, GSK, GUidepoint, Ideaya Biosciences, Idience, Ignyta, I-Mab, Ipact Therapeutics, Istitut Gustave Roussy, Intellisphere, Janssen, Joint Scientific Committee for Phase I Trials in Hong Kong, Kyn, Kyowa Krin, Lumanity, MEI Pharma, Mereo, Merit, Monte Rosa Therapeutics, Natera, Nested Therapeutics, Nexys, Nimbus, Novocure, Odyssey Therapeutics, OHSU, OncoSec, Ono Pharma, Onxeo, PanAngium Therapeutics, Pegascy, PER, Piper-Sandler, Prolynx, Protai Bio, Radiopharma Theranostics, reTORbio, Ryvu Therapeutics, SAKK, Schrodinger, Servier, Synnovation, Synthis Therapeutics, TCG Crossover, TD2, Terremoto Biosciences, Tessellate Bio, Theragnostics, Terns Pharmaceuticals, Thryv Therapeutics, Toremo, Tome, Trevarx Biomedical, Varian, Veeva, Versant, Vibilome, Voronoi Inc, Xinthera, Zai Labs, and ZielBio; grants and personal fees from Artios, AstraZeneca, Bayer, Beigene, Blueprint, Clovis, EMD Serono, F-Star, Immunesensor, Merck, Pfizer, Pliant Therapeutics, Prelude Therapeutics, Repare, Roche, Sanofi, and Tango; and grants from BioNTech, Bristol Myers Squibb, Boundless Bio, Constellation, CPRIT, Cyteir, Department of Dfense, Eli Lilly, Exelixis, Forbius, GlaxoSmithKline, Genentech, Gilead, Golfers Against Cancer, Haihe, Ideaya, Insilico Medicine, Ionis, Ipsen, Jounce, Karyopharm, KSQ, Kyowa, Mirati, Novartis, NIH/NCI, Ribon Therapeutics, Regeneron, Rubius, Scholar Rock, Seattle Genetics, Synnovation, Tesaro, V Foundation, Vivace, Zenith, and Zentalis during the conduct of the study. J. Rodon reports personal fees from Cancer Core Europe, Hummingbird Yingli, Merus, AadiBioscience, ForeBio, Amgen, MonteRosa, Debio, Incyte, Bridgebio Pharma, and Vall d’Hebron Institute of Oncology/Cancer Core Europe; other support from Loxo Oncology, Ellipses Pharma, Molecular Partners, IONCTURA, Sardona, Mekanistic, Amgen, Merus, MonteRosa, Aadi, and Bridgebio; and non-financial support from Cancer Core Europe, Symphogen, BioAlta, Pfizer, Kelun-Biotech, GlaxoSmithKline, Taiho, Roche Pharmaceuticals, Hummingbird Yingli, Bicycle Therapeutics, Merus, AadiBioscience, ForeBio, Loxo Oncology, Hutchinson MediPharma, Ideaya, Amgen, Tango Therapeutics, Mirati, Linnaeus Therapeutics, MonteRosa, Kinnate, Debio, BioTheryX Loxo Oncology, Storm Therapeutic, Beigene, MapKure, Relay, Novartis, FusionPharma, C4 Therapeutics, Scorpion Therapeutics, Incyte, Fog Pharmaceuticals, Tyra, Nuvectis Pharma, Bridgebio Pharma, 3H Pharmaceuticals, Hummingbird, AstraZeneca, Yingli, and Vall d’Hebron Institute of Oncology/Cancer Core Europe outside the submitted work. F. Meric-Bernstam reports personal fees from AstraZeneca Pharmaceuticals, Becton Dickinson, Calibr (a division of Scripps Research), Daiichi Sankyo, Dava Oncology, Debiopharm, EcoR1 Capital, eFFECTOR Therapeutics, Elevation Oncology, Exelixis, GT Aperion, Incyte, Jazz Pharmaceuticals, LegoChem Biosciences, Lengo Therapeutics, Menarini Group, Molecular Templates, Protai Bio, Ribometrix, Roche, Tallac Therapeutics, Tempus, Zymeworks, Biovica, Cybrexa, FogPharma, Guardant Health, Harbinger Health, Karyopharm Therapeutics, LOXO Oncology, Mersana Therapeutics, OnCusp Therapeutics, Sanofi Pharmaceuticals, Seagen, Theratechnologies, Zentalis Pharmaceuticals, and Dava Oncology; grants from Jazz Pharmaceuticals, Zymeworks, Aileron Therapeutics, Inc., AstraZeneca, Bayer Healthcare Pharmaceutical, Calithera Biosciences Inc., Curis Inc., CytomX Therapeutics Inc., Daiichi Sankyo Co. Ltd., Debiopharm International, eFFECTOR Therapeutics, Genentech Inc., Guardant Health Inc., Klus Pharma, Takeda Pharmaceutical, Novartis, Puma Biotechnology Inc., and Taiho Pharmaceutical Co.; and other support from European Organization for Research and Treatment of Cancer (EORTC), European Society for Medical Oncology (ESMO), Cholangiocarcinoma Foundation, and Dava Oncology outside the submitted work. No disclosures were reported by the other authors.

T.P. DiPeri: Conceptualization, data curation, formal analysis, validation, investigation, visualization, writing–original draft, writing–review and editing. K.W. Evans: Data curation, formal analysis, validation, investigation, visualization, methodology, writing–review and editing. S. Scott: Data curation, investigation. X. Zheng: Data curation, software, formal analysis, investigation, visualization, writing–review and editing. K. Varadarajan: Data curation, formal analysis, investigation, writing–review and editing. L.N. Kwong: Data curation, formal analysis, investigation, visualization, writing–review and editing. M. Kahle: Data curation, formal analysis, investigation, writing–review and editing. H.S. Tran Cao: Resources, data curation, writing–review and editing. C.-W. Tzeng: Resources, supervision, writing–review and editing. T. Vu: Resources, data curation, formal analysis, writing–review and editing. S. Kim: Resources, data curation, formal analysis, writing–review and editing. F. Su: Resources, data curation, formal analysis, writing–review and editing. M.G. Raso: Data curation, formal analysis, supervision, investigation, visualization, writing–review and editing. Y. Rizvi: Data curation, formal analysis, investigation, visualization, writing–review and editing. M. Zhao: Data curation, formal analysis, investigation, visualization, methodology, writing–review and editing. H. Wang: Data curation, formal analysis, writing–review and editing. S.S. Lee: Investigation, writing–review and editing. T.A. Yap: Investigation, writing–review and editing. J. Rodon: Investigation, writing–review and editing. M. Javle: Resources, supervision, funding acquisition, project administration, writing–review and editing. F. Meric-Bernstam: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

We would like to sincerely thank the following personnel for their contributions to this manuscript: Susanna E. Brisendine for manuscript formatting and preparation; The University of Texas MDACC’s Department of Veterinary Medicine and Surgery for excellent care of animals through the duration of these studies; The University of Texas MDACC Research Histology Core Facility for tissue processing and embedding; and The Mayo Cytogenetics Core, including Dr. Patricia T. Greipp, D.O., and Ryan A. Knudson, who performed CDKN2A/B FISH analysis for this study and is supported, in part, by the Mayo Clinic Comprehensive Cancer Center Grant, funded by the NCI (P30CA15083). This work was supported by the following: NIH Training of Academic Surgical Oncologists (Award #: 5T32CA009599-32), the University of Texas PDX Development and Trial Center (Award #: 5U54CA224065-04), the Stewart Mather Fund, Center for Clinical and Translational Science (Awards #: 5UL1TR003167 and 1UM1TR004906), the Experimental Therapeutics Clinical Trials Network (4UM1CA86688), and the MDACC support grant (Award #: P30 CA016672).

Note: Supplementary data for this article are available at Clinical Cancer Research Online (http://clincancerres.aacrjournals.org/).

1.
Nakamura
H
,
Arai
Y
,
Totoki
Y
,
Shirota
T
,
Elzawahry
A
,
Kato
M
, et al
.
Genomic spectra of biliary tract cancer
.
Nat Genet
2015
;
47
:
1003
10
.
2.
Javle
M
,
Bekaii-Saab
T
,
Jain
A
,
Wang
Y
,
Kelley
RK
,
Wang
K
, et al
.
Biliary cancer: utility of next-generation sequencing for clinical management
.
Cancer
2016
;
122
:
3838
47
.
3.
Lowery
MA
,
Ptashkin
R
,
Jordan
E
,
Berger
MF
,
Zehir
A
,
Capanu
M
, et al
.
Comprehensive molecular profiling of intrahepatic and extrahepatic cholangiocarcinomas: potential targets for intervention
.
Clin Cancer Res
2018
;
24
:
4154
61
.
4.
Khan
SA
,
Thomas
HC
,
Davidson
BR
,
Taylor-Robinson
SD
.
Cholangiocarcinoma
.
Lancet
2005
;
366
:
1303
14
.
5.
Bridgewater
J
,
Galle
PR
,
Khan
SA
,
Llovet
JM
,
Park
JW
,
Patel
T
, et al
.
Guidelines for the diagnosis and management of intrahepatic cholangiocarcinoma
.
J Hepatol
2014
;
60
:
1268
89
.
6.
DeOliveira
ML
,
Cunningham
SC
,
Cameron
JL
,
Kamangar
F
,
Winter
JM
,
Lillemoe
KD
, et al
.
Cholangiocarcinoma: thirty-one-year experience with 564 patients at a single institution
.
Ann Surg
2007
;
245
:
755
62
.
7.
Bogenberger
JM
,
DeLeon
TT
,
Arora
M
,
Ahn
DH
,
Borad
MJ
.
Emerging role of precision medicine in biliary tract cancers
.
NPJ Precis Oncol
2018
;
2
:
21
.
8.
Churi
CR
,
Shroff
R
,
Wang
Y
,
Rashid
A
,
Kang
HC
,
Weatherly
J
, et al
.
Mutation profiling in cholangiocarcinoma: prognostic and therapeutic implications
.
PLoS One
2014
;
9
:
e115383
.
9.
Chakravarty
D
,
Johnson
A
,
Sklar
J
,
Lindeman
NI
,
Moore
K
,
Ganesan
S
, et al
.
Somatic genomic testing in patients with metastatic or advanced cancer: ASCO provisional clinical opinion
.
J Clin Oncol
2022
;
40
:
1231
58
.
10.
DiPeri
TP
,
Javle
MM
,
Meric-Bernstam
F
.
Next generation sequencing for biliary tract cancers
.
Expert Rev Gastroenterol Hepatol
2021
;
15
:
471
4
.
11.
Valle
J
,
Wasan
H
,
Palmer
DH
,
Cunningham
D
,
Anthoney
A
,
Maraveyas
A
, et al
.
Cisplatin plus gemcitabine versus gemcitabine for biliary tract cancer
.
N Engl J Med
2010
;
362
:
1273
81
.
12.
Abou-Alfa
GK
,
Macarulla
T
,
Javle
MM
,
Kelley
RK
,
Lubner
SJ
,
Adeva
J
, et al
.
Ivosidenib in IDH1-mutant, chemotherapy-refractory cholangiocarcinoma (ClarIDHy): a multicentre, randomised, double-blind, placebo-controlled, phase 3 study
.
Lancet Oncol
2020
;
21
:
796
807
.
13.
Abou-Alfa
GK
,
Sahai
V
,
Hollebecque
A
,
Vaccaro
G
,
Melisi
D
,
Al-Rajabi
R
, et al
.
Pemigatinib for previously treated, locally advanced or metastatic cholangiocarcinoma: a multicentre, open-label, phase 2 study
.
Lancet Oncol
2020
;
21
:
671
84
.
14.
Javle
M
,
Roychowdhury
S
,
Kelley
RK
,
Sadeghi
S
,
Macarulla
T
,
Weiss
KH
, et al
.
Infigratinib (BGJ398) in previously treated patients with advanced or metastatic cholangiocarcinoma with FGFR2 fusions or rearrangements: mature results from a multicentre, open-label, single-arm, phase 2 study
.
Lancet Gastroenterol Hepatol
2021
;
6
:
803
15
.
15.
Drilon
A
,
Laetsch
TW
,
Kummar
S
,
DuBois
SG
,
Lassen
UN
,
Demetri
GD
, et al
.
Efficacy of larotrectinib in TRK fusion-positive cancers in adults and children
.
N Engl J Med
2018
;
378
:
731
9
.
16.
Marabelle
A
,
Fakih
M
,
Lopez
J
,
Shah
M
,
Shapira-Frommer
R
,
Nakagawa
K
, et al
.
Association of tumour mutational burden with outcomes in patients with advanced solid tumours treated with pembrolizumab: prospective biomarker analysis of the multicohort, open-label, phase 2 KEYNOTE-158 study
.
Lancet Oncol
2020
;
21
:
1353
65
.
17.
Subbiah
V
,
Lassen
U
,
Élez
E
,
Italiano
A
,
Curigliano
G
,
Javle
M
, et al
.
Dabrafenib plus trametinib in patients with BRAF(V600E)-mutated biliary tract cancer (ROAR): a phase 2, open-label, single-arm, multicentre basket trial
.
Lancet Oncol
2020
;
21
:
1234
43
.
18.
Meric-Bernstam
F
,
Beeram
M
,
Hamilton
E
,
Oh
DY
,
Hanna
DL
,
Kang
YK
, et al
.
Zanidatamab, a novel bispecific antibody, for the treatment of locally advanced or metastatic HER2-expressing or HER2-amplified cancers: a phase 1, dose-escalation and expansion study
.
Lancet Oncol
2022
;
23
:
1558
70
.
19.
Javle
M
,
Borad
MJ
,
Azad
NS
,
Kurzrock
R
,
Abou-Alfa
GK
,
George
B
, et al
.
Pertuzumab and trastuzumab for HER2-positive, metastatic biliary tract cancer (MyPathway): a multicentre, open-label, phase 2a, multiple basket study
.
Lancet Oncol
2021
;
22
:
1290
300
.
20.
Meric-Bernstam
F
,
Bahleda
R
,
Hierro
C
,
Sanson
M
,
Bridgewater
J
,
Arkenau
HT
, et al
.
Futibatinib, an irreversible FGFR1-4 inhibitor, in patients with advanced solid tumors harboring FGF/FGFR aberrations: a phase I dose-expansion study
.
Cancer Discov
2022
;
12
:
402
15
.
21.
Goyal
L
,
Meric-Bernstam
F
,
Hollebecque
A
,
Valle
JW
,
Morizane
C
,
Karasic
TB
, et al
.
Futibatinib for FGFR2-rearranged intrahepatic cholangiocarcinoma
.
N Engl J Med
2023
;
388
:
228
39
.
22.
Meric-Bernstam
F
,
Makker
V
,
Oaknin
A
,
Oh
DY
,
Banerjee
S
,
González-Martín
A
, et al
.
Efficacy and safety of trastuzumab deruxtecan in patients with HER2-expressing solid tumors: primary results from the DESTINY-PanTumor02 phase II trial
.
J Clin Oncol
2024
;
42
:
47
58
.
23.
Sweeney
CJ
,
Hainsworth
JD
,
Bose
R
,
Burris
HA
,
Kurzrock
R
,
Swanton
C
, et al
.
MyPathway human epidermal growth factor receptor 2 basket study: pertuzumab + trastuzumab treatment of a tissue-agnostic cohort of patients with human epidermal growth factor receptor 2-altered advanced solid tumors
.
J Clin Oncol
2024
;
42
:
258
65
.
24.
Dobrolecki
LE
,
Airhart
SD
,
Alferez
DG
,
Aparicio
S
,
Behbod
F
,
Bentires-Alj
M
, et al
.
Patient-derived xenograft (PDX) models in basic and translational breast cancer research
.
Cancer Metastasis Rev
2016
;
35
:
547
73
.
25.
Loeuillard
E
,
Fischbach
SR
,
Gores
GJ
,
Ilyas
SI
.
Animal models of cholangiocarcinoma
.
Biochim Biophys Acta Mol Basis Dis
2019
;
1865
:
982
92
.
26.
Jusakul
A
,
Cutcutache
I
,
Yong
CH
,
Lim
JQ
,
Huang
MN
,
Padmanabhan
N
, et al
.
Whole-genome and epigenomic landscapes of etiologically distinct subtypes of cholangiocarcinoma
.
Cancer Discov
2017
;
7
:
1116
35
.
27.
Zehir
A
,
Benayed
R
,
Shah
RH
,
Syed
A
,
Middha
S
,
Kim
HR
, et al
.
Mutational landscape of metastatic cancer revealed from prospective clinical sequencing of 10,000 patients
.
Nat Med
2017
;
23
:
703
13
.
28.
Li
H
,
Durbin
R
.
Fast and accurate short read alignment with Burrows-Wheeler transform
.
Bioinformatics
2009
;
25
:
1754
60
.
29.
DePristo
MA
,
Banks
E
,
Poplin
R
,
Garimella
KV
,
Maguire
JR
,
Hartl
C
, et al
.
A framework for variation discovery and genotyping using next-generation DNA sequencing data
.
Nat Genet
2011
;
43
:
491
8
.
30.
Rimmer
A
,
Phan
H
,
Mathieson
I
,
Iqbal
Z
,
Twigg
SRF
, et al;
WGS500 Consortium
.
Integrating mapping-assembly- and haplotype-based approaches for calling variants in clinical sequencing applications
.
Nat Genet
2014
;
46
:
912
18
.
31.
Cibulskis
K
,
Lawrence
MS
,
Carter
SL
,
Sivachenko
A
,
Jaffe
D
,
Sougnez
C
, et al
.
Sensitive detection of somatic point mutations in impure and heterogeneous cancer samples
.
Nat Biotechnol
2013
;
31
:
213
19
.
32.
Ye
K
,
Schulz
MH
,
Long
Q
,
Apweiler
R
,
Ning
Z
.
Pindel: a pattern growth approach to detect break points of large deletions and medium sized insertions from paired-end short reads
.
Bioinformatics
2009
;
25
:
2865
71
.
33.
Zhang
J
,
Fujimoto
J
,
Zhang
J
,
Wedge
DC
,
Song
X
,
Zhang
J
, et al
.
Intratumor heterogeneity in localized lung adenocarcinomas delineated by multiregion sequencing
.
Science
2014
;
346
:
256
9
.
34.
Olshen
AB
,
Venkatraman
ES
,
Lucito
R
,
Wigler
M
.
Circular binary segmentation for the analysis of array-based DNA copy number data
.
Biostatistics
2004
;
5
:
557
72
.
35.
Dobin
A
,
Davis
CA
,
Schlesinger
F
,
Drenkow
J
,
Zaleski
C
,
Jha
S
, et al
.
STAR: ultrafast universal RNA-seq aligner
.
Bioinformatics
2013
;
29
:
15
21
.
36.
Kim
D
,
Salzberg
SL
.
TopHat-fusion: an algorithm for discovery of novel fusion transcripts
.
Genome Biol
2011
;
12
:
R72
.
37.
Trapnell
C
,
Williams
BA
,
Pertea
G
,
Mortazavi
A
,
Kwan
G
,
van Baren
MJ
, et al
.
Transcript assembly and quantification by RNA-seq reveals unannotated transcripts and isoform switching during cell differentiation
.
Nat Biotechnol
2010
;
28
:
511
5
.
38.
Anders
S
,
Pyl
PT
,
Huber
W
.
HTSeq-a Python framework to work with high-throughput sequencing data
.
Bioinformatics
2015
;
31
:
166
9
.
39.
Love
MI
,
Huber
W
,
Anders
S
.
Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2
.
Genome Biol
2014
;
15
:
550
.
40.
Kurnit
KC
,
Dumbrava
EEI
,
Litzenburger
B
,
Khotskaya
YB
,
Johnson
AM
,
Yap
TA
, et al
.
Precision oncology decision support: current approaches and strategies for the future
.
Clin Cancer Res
2018
;
24
:
2719
31
.
41.
Johnson
A
,
Khotskaya
YB
,
Brusco
L
,
Zeng
J
,
Holla
V
,
Bailey
AM
, et al
.
Clinical use of precision oncology decision support
.
JCO Precis Oncol
2017
;
2017
:
PO.17.00036
.
42.
Johnson
A
,
Zeng
J
,
Bailey
AM
,
Holla
V
,
Litzenburger
B
,
Lara-Guerra
H
, et al
.
The right drugs at the right time for the right patient: the MD Anderson precision oncology decision support platform
.
Drug Discov Today
2015
;
20
:
1433
8
.
43.
Hofmann
M
,
Stoss
O
,
Shi
D
,
Büttner
R
,
van de Vijver
M
,
Kim
W
, et al
.
Assessment of a HER2 scoring system for gastric cancer: results from a validation study
.
Histopathology
2008
;
52
:
797
805
.
44.
Mosele
F
,
Remon
J
,
Mateo
J
,
Westphalen
CB
,
Barlesi
F
,
Lolkema
MP
, et al
.
Recommendations for the use of next-generation sequencing (NGS) for patients with metastatic cancers: a report from the ESMO Precision Medicine Working Group
.
Ann Oncol
2020
;
31
:
1491
505
.
45.
Harding
JJ
,
Fan
J
,
Oh
DY
,
Choi
HJ
,
Kim
JW
,
Chang
HM
, et al
.
Zanidatamab for HER2-amplified, unresectable, locally advanced or metastatic biliary tract cancer (HERIZON-BTC-01): a multicentre, single-arm, phase 2b study
.
Lancet Oncol
2023
;
24
:
772
82
.
46.
DiPeri
TP
,
Evans
KW
,
Wang
B
,
Zhao
M
,
Akcakanat
A
,
Raso
MG
, et al
.
Co-clinical trial of novel bispecific anti-HER2 antibody zanidatamab in patient-derived xenografts
.
Cancer Discov
2024
;
14
:
828
45
.
47.
Zhao
M
,
Scott
S
,
Evans
KW
,
Yuca
E
,
Saridogan
T
,
Zheng
X
, et al
.
Combining neratinib with CDK4/6, mTOR, and MEK inhibitors in models of HER2-positive cancer
.
Clin Cancer Res
2021
;
27
:
1681
94
.
48.
Meric-Bernstam
F
,
Ford
JM
,
O’Dwyer
PJ
,
Shapiro
GI
,
McShane
LM
,
Freidlin
B
, et al
.
National cancer Institute combination therapy platform trial with molecular analysis for therapy choice (ComboMATCH)
.
Clin Cancer Res
2023
;
29
:
1412
22
.
49.
Leiting
JL
,
Murphy
SJ
,
Bergquist
JR
,
Hernandez
MC
,
Ivanics
T
,
Abdelrahman
AM
, et al
.
Biliary tract cancer patient-derived xenografts: surgeon impact on individualized medicine
.
JHEP Rep
2020
;
2
:
100068
.
50.
Lynch
IT
,
Abdelrahman
AM
,
Alva-Ruiz
R
,
Fogliati
A
,
Graham
RP
,
Smoot
R
, et al
.
Cancer “avatars”: patient-derived xenograft growth correlation with postoperative recurrence and survival in pancreaticobiliary cancer
.
J Am Coll Surg
2023
;
237
:
483
500
.
51.
Serra-Camprubí
Q
,
Verdaguer
H
,
Oliveros
W
,
Lupión-Garcia
N
,
Llop-Guevara
A
,
Molina
C
, et al
.
Human metastatic cholangiocarcinoma patient-derived xenografts and tumoroids for preclinical drug evaluation
.
Clin Cancer Res
2023
;
29
:
432
45
.
52.
Li
J
,
Lu
H
,
Ng
PK
,
Pantazi
A
,
Ip
CKM
,
Jeong
KJ
, et al
.
A functional genomic approach to actionable gene fusions for precision oncology
.
Sci Adv
2022
;
8
:
eabm2382
.
53.
Krook
MA
,
Lenyo
A
,
Wilberding
M
,
Barker
H
,
Dantuono
M
,
Bailey
KM
, et al
.
Efficacy of FGFR inhibitors and combination therapies for acquired resistance in FGFR2-fusion cholangiocarcinoma
.
Mol Cancer Ther
2020
;
19
:
847
57
.
54.
Goyal
L
,
Saha
SK
,
Liu
LY
,
Siravegna
G
,
Leshchiner
I
,
Ahronian
LG
, et al
.
Polyclonal secondary FGFR2 mutations drive acquired resistance to FGFR inhibition in patients with FGFR2 fusion-positive cholangiocarcinoma
.
Cancer Discov
2017
;
7
:
252
63
.
55.
Goyal
L
,
Shi
L
,
Liu
LY
,
Fece de la Cruz
F
,
Lennerz
JK
,
Raghavan
S
, et al
.
TAS-120 overcomes resistance to ATP-competitive FGFR inhibitors in patients with FGFR2 fusion-positive intrahepatic cholangiocarcinoma
.
Cancer Discov
2019
;
9
:
1064
79
.
56.
Silverman
IM
,
Hollebecque
A
,
Friboulet
L
,
Owens
S
,
Newton
RC
,
Zhen
H
, et al
.
Clinicogenomic analysis of FGFR2-rearranged cholangiocarcinoma identifies correlates of response and mechanisms of resistance to pemigatinib
.
Cancer Discov
2021
;
11
:
326
39
.
57.
Zhu
Y
,
Kwong
LN
.
IDH1 inhibition reawakens the immune response against cholangiocarcinoma
.
Cancer Discov
2022
;
12
:
604
5
.
58.
Lapin
M
,
Huang
HJ
,
Chagani
S
,
Javle
M
,
Shroff
RT
,
Pant
S
, et al
.
Monitoring of dynamic changes and clonal evolution in circulating tumor DNA from patients with IDH-mutated cholangiocarcinoma treated with isocitrate dehydrogenase inhibitors
.
JCO Precis Oncol
2022
;
6
:
e2100197
.
59.
Farshidfar
F
,
Zheng
S
,
Gingras
MC
,
Newton
Y
,
Shih
J
,
Robertson
AG
, et al
.
Integrative genomic analysis of cholangiocarcinoma identifies distinct IDH-mutant molecular profiles
.
Cell Rep
2017
;
18
:
2780
94
.
60.
Zhu
AX
,
Macarulla
T
,
Javle
MM
,
Kelley
RK
,
Lubner
SJ
,
Adeva
J
, et al
.
Final overall survival efficacy results of ivosidenib for patients with advanced cholangiocarcinoma with IDH1 mutation: the phase 3 randomized clinical ClarIDHy trial
.
JAMA Oncol
2021
;
7
:
1669
77
.
61.
Javle
M
,
Churi
C
,
Kang
HC
,
Shroff
R
,
Janku
F
,
Surapaneni
R
, et al
.
HER2/neu-directed therapy for biliary tract cancer
.
J Hematol Oncol
2015
;
8
:
58
.
62.
Uzunparmak
B
,
Haymaker
C
,
Raso
G
,
Masciari
S
,
Wang
L
,
Lin
H
, et al
.
HER2-low expression in patients with advanced or metastatic solid tumors
.
Ann Oncol
2023
;
34
:
1035
46
.
63.
Saura
C
,
Oliveira
M
,
Feng
YH
,
Dai
MS
,
Chen
SW
,
Hurvitz
SA
, et al
.
Neratinib plus capecitabine versus lapatinib plus capecitabine in HER2-positive metastatic breast cancer previously treated with ≥ 2 HER2-directed regimens: phase III NALA trial
.
J Clin Oncol
2020
;
38
:
3138
49
.
This open access article is distributed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0) license.