Egan, Jan B.

e16185 Background: Comprehensive sequencing efforts have identified fibroblast growth factor receptor (FGFR) translocations in numerous cancers. FGFR2 fusions were recently identified as an actionable therapeutic target in ̃17% of cholangiocarcinoma patients, predominantly found in patients with intrahepatic cholangiocarcinoma. Genomic partners for FGFR fusions are variable. In addition, obtaining high yield tissue biopsies for CCA remains challenging and tumor tissue available for molecular analyses is scarce. When combined, these two factors make detection of FGFR fusions in CCA challenging. An accurate molecular assay to detect FGFR fusions in CCA could improve utilization of FGFR-targeted therapies. Methods: The objective of our study was to develop an approach for targeted RNA analysis to improve sensitivity for FGFR fusion detection from limited tumor material. We adapted the sensitive targeted digital sequencing (TARDIS) approach recently developed in our lab for analysis of RNA. This approach utilizes a combination of open-ended targeted amplification followed by ligation, enabling detection of tyrosine kinase fusions without prior knowledge of the precise sequence of the fusion breakpoint or identity of the fusion partner. For evaluation of analytical performance, we analyzed RNA from a CCA organoid model with a known FGFR2 fusion, and Seraseq Fusion RNA Mix v3 Reference Material with two known FGFR3 fusions. Results: Without prior knowledge the partner or coordinates of the fusion breakpoint, we detected an FGFR2-KIF5C fusion in RNA from a CCA organoid model. The fusion breakpoint coordinates predicted by TARDIS-RNA were validated by comparison with RNA-Seq. To assess sensitivity and quantitative performance of the assay, we analyzed a serial dilution of the reference RNA sample from Seraseq with concentration of RNA molecules representing candidate fusions verified by digital PCR. Using TARDIS-RNA to analyze replicates with 5 ng RNA input, we were able to detect candidate fusions represented by as few as 2.7 fusion molecules on average. Concentrations of fusion molecules measured using the two methods were highly correlated (R = 0.96). Conclusions: These results demonstrate sensitivity and quantitative performance of a targeted RNA sequencing assay to detect and quantify FGFR fusions in tumor specimen from intrahepatic cholangiocarcinoma. On-going work is focused on further evaluating assay performance and characterizing FGFR fusions using additional tissue specimen.

AbstractBackgroundThe goal of this study is to evaluate germline genetic variants in African American men with metastatic prostate cancer as compared to those in Caucasian men with metastatic prostate cancer in an effort to understand the role of genetic factors in these populations.MethodsAfrican American and Caucasian men with metastatic prostate cancer who had germline testing using multigene panels were used to generate comparisons. Germline genetic results, clinical parameters, and family histories between the two populations were analyzed.ResultsA total of 867 patients were included in this retrospective study, including 188 African American and 669 Caucasian patients. There was no significant difference in the likelihood of a pathogenic or likely‐pathogenic variants (PV/LPVs) between African American and Caucasian patients (p = .09). African American patients were more likely to have a variant of unknown significance than Caucasians (odds ratio [OR] = 1.95; p < .0001). BRCA1 PV/LPVs were higher in African Americans (OR = 4.86; p = .04). African American patients were less likely to have a PV/LPV in non‐BRCA DNA repair genes (OR = 0.30; p = .008). Family history of breast (OR = 2.09; p = .002) or ovarian cancer (OR = 2.33; p = .04) predicted PV/LPVs in Caucasians but not African‐Americans. This underscores the limitations of family history in AA men and the importance of personal history to guide germline testing in AA men.ConclusionsIn metastatic prostate cancer patients, PV/LPVs of tested genes did not vary by race, BRCA1 PV/LPVs were more common in the African American subset. However, PV/LPVs in non‐BRCA DNA repair genes were less likely to be encountered in African Americans. Family history associated with genetic testing results in Caucasians only.

Abstract Motivation Functions of cancer driver genes vary substantially across tissues and organs. Distinguishing passenger genes, oncogenes (OGs) and tumor-suppressor genes (TSGs) for each cancer type is critical for understanding tumor biology and identifying clinically actionable targets. Although many computational tools are available to predict putative cancer driver genes, resources for context-aware classifications of OGs and TSGs are limited. Results We show that the direction and magnitude of somatic selection of protein-coding mutations are significantly different for passenger genes, OGs and TSGs. Based on these patterns, we develop a new method (genes under selection in tumors) to discover OGs and TSGs in a cancer-type specific manner. Genes under selection in tumors shows a high accuracy (92%) when evaluated via strict cross-validations. Its application to 10 172 tumor exomes found known and novel cancer drivers with high tissue-specificities. In 11 out of 13 OGs shared among multiple cancer types, we found functional domains selectively engaged in different cancers, suggesting differences in disease mechanisms. Availability and implementation An R implementation of the GUST algorithm is available at https://github.com/liliulab/gust. A database with pre-computed results is available at https://liliulab.shinyapps.io/gust. Supplementary information Supplementary data are available at Bioinformatics online.

Abstract Primary central nervous system lymphoma (PCNSL) is an aggressive and incurable variant of non-Hodgkin lymphoma (NHL) that is confined to the central nervous system. Most PCNSL (90%) are part of the immune-privileged site-associated diffuse large B-cell lymphomas (DLBCL). Unlike nodal DLBCL, only a limited number of genetic studies have been performed in PCNSL, partly due to lack of available tissue specimens. Because of the fragmented knowledge of the genomic basis, it is still a matter of debate whether they differ from systemic DLBCL with respect to their molecular features and pathogenesis and also if there is a CNS specific signature. We performed a comprehensive genomic study in a cohort of 22 immunocompetent (HIV- and EBV-) PCNSL. DNA was extracted from FFPE tissues (N=15) and frozen tissue (N=7). Samples were analyzed using a combination of aCGH (N=18), exome sequencing (N=10), mate-pair genome sequencing (N=2) and targeted sequencing (N=8). We found a complex karyotype with a median of 21 copy-number abnormalities (range 10–47), 6 structural abnormalities, 6 frameshift indels and 67 nonsynonymous mutations. By integrating mutation and copy number data we found a group of genes recurrently mutated and/or deleted in PNCSL. Remarkable findings were the high prevalence of MYD88 activating mutations (L265P/M232T/V217F), found in 69% of cases and the biallelic loss of CDKN2A (60%). A subset of recurrent abnormalities was exclusively found in PCNSL, and not being previously identified in systemic DLBCL. Thus, 11% of PCNSL have biallelic inactivation of TOX (a regulator of T-cell development) and PRKCD (protein kinase C delta), while another 17% of cases show focal monoallelic deletions/mutations in these genes. Finally, recurrent mutations have been identified in ATM, which have not been found in nodal DLBCL. Several other genes affected have been previously identified in nodal DLBCL, such as biallelic loss of TNFAIP3 (16%), PRDM1 (16%), GNA13, TMEM30A, B2M and CD58 (11% each), activating mutations of CD79B (28%) and CARD11 (19%) and translocations of BCL6 (22%). Components of the NF-kB pathways were altered in >90% of PNCSL. Pathway analysis also showed an enrichment of networks associated with immune response, proliferation, regulation of apoptosis and lymphocyte differentiation and activation. Finally, we searched for associations between genetic alterations and clinical outcome. We showed that deletions of 6q21 (PRDM1) and 6q23 (TNFAIP3) were both associated with shorter overall survival (p=0.007 and p=0.03, respectively). In summary, we report a genomic background in PCNSL similar to post-GC DLBCL but reinforcing the existence of a subset of abnormalities specific to PCNSL, suggesting their potential relevance in the disease pathogenesis. Additionally, the results obtained from FFPE samples are encouraging and larger archival tissue collections can now be analyzed in order to complement the still fragmented knowledge we have of the genetic basis of the disease. Disclosures: Stewart: Onyx: Consultancy, Research Funding; Millenium: Honoraria, Research Funding; Celgene: Honoraria; BMS: Honoraria. Fonseca:Medtronic: Consultancy; Otsuka: Consultancy; Celgene: Consultancy; Genzyme: Consultancy; BMS: Consultancy; Lilly: Consultancy; Onyx: Consultancy, Research Funding; Binding Site: Consultancy; Millennium: Consultancy; AMGEN: Consultancy; Cylene: Research Funding; Prognostication of MM based on genetic categorization of the disease: Patents & Royalties.

Abstract Abstract 2810 Poster Board II-786 We wished to explore the genetic events associated with disease progression and development of drug resistance in multiple myeloma (MM). To do so 11 patients were studied in whom at least two (range 2-3) temporally distinct samples of tumor DNA and RNA were available. The baseline genetic initiating event was defined for all patients (3 were genetic high risk; one with t(14;16) two with t(4;14)) as well as the gene expression profile (GEP) defined risk score using the Little Rock 70 or 17 gene panel (only one, the t(14;16) was GEP defined high risk). High resolution array CGH and gene expression were then performed on each sample. Of the 8 patients with a “low risk” tumor initiating event and low risk GEP score, 6 patients had no, or only one, copy number abnormality (CNA) change between the two temporally distinct MM samples. In stark contrast the 3 genetic high risk at baseline had between 17 and 40 distinct CNA changes at the time of progression. For all 11 patients 89 CNA were acquired with progression whereas 19 previously abnormal regions disappeared suggesting clones with these abnormalities were extinguished by the therapy received. In total we detected 0-40 CNA changes between the various timepoints, median 1, mean 10.7. The acquisition of new CNA was much more common than the loss of CNA. We then focused more specifically on the t(4;14) patient with the highest number of CNA changes. This patient has a well documented clinical course of having a sustained two year VGPR to Len/dex and then progressing while still taking Len/dex. Comparison of the pre and post-Len/dex samples identified 40 CNA changes(the most of any pair studied to date). Only six CNA were shared between the two samples, which included deletions of chr4, 9, 12, 13, and X plus a t(4;14) translocation. These likely represent the initiating “driver” tumor events. The new CNA we identified originated from both remodeled genomic changes and the emergence of unique changes, indicating a new tumor clone had emerged while the previously dominant clone had regressed (e.g. a deletion of a large segment of chromosome 8 at diagnosis was no longer observed in the relapse sample). The newly acquired CNA encompassed 3968 genes (13.7% of the genes in the genome), however, only 1235 of these genes (4%) were expressed in this patient at diagnosis (1188 in the typical myeloma patient). Since 1235 genes is still a large number we hypothesized that whole genome sequencing (WGS) would help elucidate the mechanism of lenalidomide resistance. We isolated DNA from germline tissue and CD138 purified tumor cells including: diagnostic, first relapse and second relapse samples. Utilizing SOLiD (Applied Biosystems, Foster City, CA) sequencing technology, we have completed fragment library WGS on both the germline and the final tumor samples. Quality control measures report the average number of sequence reads per start point to be less than 1.2, indicating the library is primarily composed of unique molecules. In addition, approximately 40% of the sequence reads map uniquely to the genome. Together these quality measures indicate our sample libraries are complex and provide good representation of the genome. Data on the whole genome sequence of myeloma at diagnosis and at the time of progression will be presented. Disclosures: Bergsagel: Celgene: Consultancy.

Abstract Abstract 420 Primary central nervous system lymphoma (PCNSL) is a rare and aggressive non-Hodgkin lymphoma that is confined to the CNS because of a poorly understood neurotropism. Most of PCNSL (90%) are part of the immune-privileged site-associated DLBCL (IPDLBCL). IPDLBCL consist in late–germinal center or post–germinal center lymphoid cells but that show very distinct characteristics that separate them from systemic DLBCL. It is still a matter of debate whether the PCNSL differ from nodal DLBCL with respect to molecular features and pathogenesis and also if there is a genomic signature specific of PCNSL. Only few genetic studies have been performed in PCNSL, partly due to the rarity of the tumors and the limited amount of available tissue. To gain insight into the genomic basis of PCNSL, we performed an integrated, high-throughput, genomic analysis in 17 immunocompetent, EBV- and HIV- cases. B-cell differentiation status was characterized by immunostains for CD10, MUM-1, and BCL-6. Either frozen samples or formalin fixed embedded paraffin sections from 17 PCNSL were studied by array-based comparative genomic hybridization (aCGH) using Sureprint G3 (1 million probes) array. Massively parallel whole-exome sequencing was performed in 4 of these cases. Additionally, 2 cases were analyzed by mate-pair whole genome sequencing searching for chromosomal breakpoints. Sanger DNA sequencing was used for validation. All cases were characterized by complex genomic aberrations with a median of 21 copy-number abnormalities (CNA, range 10–49), 4 structural abnormalities, 6 frameshift indels and 99 nonsynonymous exonic mutations. Focal deletion affecting CDKN2A (9p21) was the most common CNA, found in 14 of 17 cases (82%); with 6 of these cases (35%) having homozygous deletion. The second most frequent CNA involved the HLA genes (6p21), found in 11 of 17 (65%) cases; 4 of them (23%) with homozygous deletions. We identified recurrent CNA and mutations in several genes previously found in systemic DLBCL. Thus, PRDM1 (BLIMP1) was deleted or mutated in 47% of cases and the translocation IgH-BCL6 was found in 30% of PCNSL. Furthermore, recurrent mutations were found in NF-kB genes CD79B (75%, 3 of 4 cases analyzed), MYD88 (70%, 7 of 10), TNFAIP3 (50%, 2 of 4) and CARD11 (50%, 2 of 4). Additionally, recurrent abnormalities were found in B2M, BCL7A, CD58, CIITA, ETV6, GNA13, PAX5, TMEM30A and TP53. Nevertheless, we identified several recurrent genetic alterations not described in systemic DLBCL. TOX (a regulator of T-cell development) and TBL1XR1 (a negative regulator of the NF-kB and Wnt pathways) were either deleted or mutated in 30% of PNCSL, something not previously described in systemic DLBCL. Additionally, chromosomal breakpoints either in DLGAP1 or DLGAP2 (play a role in neuronal cell signaling) were found in 18% of PCNSL but not in systemic DLBCL. Moreover, mutations in ATM (master controller of cell cycle checkpoint), BAI3 (inhibitor of brain angiogenesis), BTG2 (cell cycle arrest), KDM6B (histone demethylase), PKC family members PRKCD/PRKCDE, genes from the histone cluster, the protocadherin family and the WD repeat domain were found in 10% to 50% of PCNSL. Pathway analysis including the most commonly affected genes in PCNSL showed an enrichment of networks associated with immune response, NF-kB pathway, proliferation, regulation of the apoptosis and lymphocyte differentiation and activation. In summary, we show evidence of a highly complex genome and identified a subset of genes with potential relevance in PCNSL pathogenesis. The genomic profile described here reinforces the existence of a specific molecular signature in PCNSL, thus helping to genetically differentiate this entity from the nodal DLBCL and related lymphomas. Disclosures: Stewart: Millenium: Consultancy, Honoraria, Research Funding; Onyx: Consultancy; Celgene: Consultancy. Fonseca:Medtronic: Consultancy; Otsuka: Consultancy; Celgene: Consultancy; Genzyme: Consultancy; BMS: Consultancy; Lilly: Consultancy; Onyx: Consultancy; Binding site: Consultancy; Millenium: Consultancy; AMGEN: Consultancy; Celgene : Research Funding; Onyx: Research Funding; prognostication of MM based on genetic categorization of the disease: Prognostication of MM based on genetic categorization of the disease Patents & Royalties.

Abstract Abstract 573 Using high throughput RNA interference screening on 6,722 druggable genes we previously identified XPO1/CRM1 as one of the 50 most vulnerable targets in Multiple Myeloma (MM)1. XPO1 knockdown proved lethal in MM cell lines, but had no effect on human embryonic kidney (293) cells or lung cancer (A549) cells, showing that XPO1 is a specific myeloma vulnerability, and that myeloma cell survival is dependent upon XPO1 expression. XPO1 encodes the protein exportin 1, a nuclear transport protein that exports tumor suppressor proteins from the nucleus, where they are active, to the cytoplasm, where they become inactive. We next analyzed XPO1 in MM via gene expression profiling (GEP). XPO1 expression is up-regulated as the disease progresses: patients with active MM have a higher level of XPO1 compared to normal plasma cells (p<0.04) and to patients with monoclonal gammopathy of undetermined significance or smoldering MM (p<0.0001). The highest levels were in human MM cell lines. TC classification revealed highest levels in t(11;14) and lowest levels in t(4;14) disease. Selective inhibitors of nuclear export (SINE) compounds have recently been developed that irreversibly inhibit XPO1/CRM1 and its nuclear export function. One such inhibitor, KPT-276, decreased the viability of all 12 MM cell lines tested in vitro, as shown by MTT assay. After 72 hours of drug treatment, a median IC50 value of approximately 175 nM (range 30–1000 nM) was observed. No synergy with other commonly used anti-MM therapeutics was observed in vitro. In contrast, the drug had little effect in 8 solid tumor cell lines with the exception of the B cell lymphoma line Ramos. KPT-276 was also consistently active in inducing apoptosis against MM primary patient samples. Using an IC80 dose of KPT-276, drug-treated samples had a reduced population of cells in S phase (8%) compared to cells treated with DMSO (21%). Using the vkappa*myc transgenic MM model, KPT-276 reduced monoclonal spikes (by a mean of 56%) in all mice treated orally with 150 mg/kg dose three times per week for 4 weeks. Furthermore, KPT-276 significantly reduced tumor growth in a xenograft MM1.S mouse model. GEP was performed in the presence or absence of drug in two different MM cell lines. Two genes of probable relevance, cell division cycle 25 homolog A (CDC25A) and Bromodomain-containing protein 4 (BRD4), were dysregulated by SINE treatment. Both are involved in cell cycle control and have been linked to MYC. RT-PCR and western blotting confirm that MYC, CDC25A and BRD4 are down-regulated, as soon as six hours, after treatment with KPT-276. KPT-276 has shown marked anticancer activities against B cell malignancies in vitro and is active and tolerated in Phase I canine studies. KPT-330, a close analog of KPT-276, is currently in Phase 1 studies in human with advanced hematological and solid tumors. Disclosures: Schmidt: Karyopharm: Research Funding. McCauley:Karyopharm Therapeutics Inc: Employment. Shacham:Karyopharm Therapeutics: Employment. Kauffman:Karyopharm Therapeutics Inc: Employment. Stewart:Millenium: Consultancy, Honoraria, Research Funding; Onyx: Consultancy; Celgene: Consultancy.

Abstract Abstract 3968 To better understand the genetic events associated with extra-medullary multiple myeloma (MM) we completed whole exome, long-insert whole genome and RNA sequencing on a patient with t(14;16) and neck adenopathy, as well as whole genome sequencing on two additional patients with respectively: t(4;14) and plasma cell leukemia, and del 17 with chest wall mass. All patients had multi-drug resistant disease and del 13. The frequency of somatic non-synonymous single nucleotide variants (SNV) ranged from 119 to 498, including an average of 15 truncating SNV, as well an average of 75 structural variants (range 2–147). Importantly, in one patient a truncating mutation, Q99*, on one allele and minor clone point mutation, R283K, on the second allele was observed in CRBN, a gene whose expression is required for effective lenalidomide and pomalidomide anti-myeloma activity. In the same patient a non-synonymous mutation, E171K, was also observed in PSMG2, a proteasome assembly protein, and a non-synonymous mutation, G369A, in NR3C1, a glucocorticoid receptor. Subsequently we have sequenced CRBN in 27 patients and 7 cell lines and have found no further examples of mutation in CRBN but have previously reported a bi-alleleic focal deletion in the MM.1Sres cell line. Mutations in NR3C1 have not previously been described in other MM genomes. Other, potentially clinically relevant mutations in the three patients were observed in FGFR3, PRDM1, STAT5B, NRAS, KRAS, PIK3CA, NF1, ATM, NFKB2, CKS1B and ERBB4. Although the three patients did not share a unifying event, a number of mutated genes observed were shared with other published myeloma genomes including RB1, TP53, MLL3, JAG2, VAV3 and SETD2. We also looked at pathway activation across the three genomes with the most statistically mutated pathways including VEGF signaling and activation, cell adhesion and chemokines, cytoskeleton remodeling, integrins in carcinoma progression, TP53 signaling and NOTCH signaling. In conclusion, we have identified, for the first time, the presence of a mutation in CRBN in a myeloma patient as well as mutations in PSMG2 and NR3C1, likely associated with drug resistance. Disclosures: Schmidt: Karyopharm: Research Funding. Carpten:Life Technologies: Research Funding. Stewart:Millenium: Consultancy, Honoraria, Research Funding; Onyx: Consultancy; Celgene: Consultancy.

Abstract Purpose: Cholangiocarcinoma (CCA) is a highly lethal malignancy that often presents at advanced stages. Gemcitabine/cisplatin (GP) represents the standard-of-care but provides only a modest response rate and survival benefit. Loss of the CDK4/6 inhibitor CDKN2A is a prevalent genetic event in CCA, thus CDK4/6 inhibitors, FDA-approved in some breast cancer regimens, are hypothesized to have a role in treating CCA. Clinical challenges to target cell cycle dysregulation and exploit cytotoxic chemotherapy include predictive biomarkers for selecting patients most likely to respond. We hypothesized autophagy is involved in therapeutic resistance to CDK4/6 inhibitors, and plays a role in the activity of GP in CCA. Thus, we investigated CDK4/6 inhibitors as monotherapy, and combined with GP in CCA, measuring multiple readouts pertaining to drug synergy, putative pathway biomarkers and autophagy. Methods: The activity of 3 FDA approved CDK4/6 inhibitors, palbociclib, ribociclib and abemaciclib, as monotherapy (N=24), and combined with GP (N=10), was assessed in CCA cell lines by measuring relative cell number with Cell TiterGlo, and apoptosis with Annexin/PI and cleaved caspase3 quantification via flow cytometry. To characterize RB status for biomarker potential, protein levels of CCND1, CDK4, CDK6, RB, and CDKN2A were measured by western blot. Autophagy was measured using an autophagy flux assay, LysoTracker confocal staining, and by quantifying protein levels of LC3B, p62 and Beclin1A. Results: 22/24 (92%) of CCA cell lines were found to have at least one defect likely relevant to CDK4/6 inhibition: CDKN2A (92%), Rb1 (63%), CDK4 (23%) and cyclin D1 (32%). Protein levels of CDKN2A and RB1 showed moderate correlation with CDK4/6 inhibitor activity. Abemaciclib was the most potent CDK inhibitor in vitro (IC50≤0.5uM; palbociclib IC50&gt;1uM; ribociclib IC50&gt;5uM). Combined abemaciclib and GP showed higher cytotoxicity and synergy at lower doses compared to chemotherapy alone or single-agent abemaciclib. Compared to CDK4/6 inhibitors or chemotherapy alone, combination therapy enhanced growth inhibition, cell cycle arrest and caspase-dependent apoptotic cell death. Abemaciclib triggered autophagy mediated by significant increases in reactive oxygen species (ROS). Combination therapy reduced ROS production and thus autophagy, as seen by reduced autophagic flux, reduced LysoTracker mean fluorescence intensity, a decrease in the expression of LC3B-II, Beclin1A and an increase in p62. Conclusion: These results suggest CDK4/6 inhibitors interact synergistically with GP in CCA and mechanistically disrupt autophagy. This suggests potential to further explore CDKN2A and RB as biomarkers for preclinical sensitivity in CCA. Pre-clinical data supports ongoing clinical studies of this concept, as well as planned clinical studies of GP and alternative CDK4/6 inhibitors. Citation Format: Mansi Arora, James Bogenberger, Nagalo Bolni, Yumei Zhou, Latha Kannan, Jan Egan, Raquel Yokoda, Daniel H. Ahn, Mitesh J. Borad. Synergy of cyclin dependentkinase 4/6 inhibitors with cytotoxic chemotherapy in cholangiocarcinoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 4417.

293 Background: Cholangiocarcinoma (CCA) is a deadly cancer with ~20% five year survival. Obtaining tissue to confirm diagnosis of CCA is often challenging. This also limits the use of tumor genotyping that could guide potential molecularly targeted treatment (such as FGFR inhibitors for a subset of cases). We investigated whether circulating tumor DNA (ctDNA) analysis in plasma is feasible for noninvasive tumor genotyping in patients with advanced CCA. Methods: We collected and processed plasma samples for ctDNA analysis from 17 patients with CCA at presentation and from 11 healthy volunteers. Following cell-free DNA extraction and quality assessment using digital PCR, we performed low-pass, shallow whole genome sequencing (sWGS). We analyzed sequencing data to identify copy number aberrations and quantify tumor fraction in plasma using recently published computational approaches. Results: Median total cfDNA concentration in patients with CCA was 7,003 genome equivalents (GEs)/mL plasma (range: 1,512- 107,707 GEs/mL) and significantly higher than healthy volunteers (two tailed t-test p < 0.04, median: 1,658 GEs/mL, range: 661-4,938 GEs/mL). We successfully sequenced all tumor libraries, generating a median of 173 million paired-end reads. ctDNA was detectable in 10/17 samples using sWGS. In these 10 samples, median tumor fraction in plasma DNA was 7.1% (range: 3.9%-44.6%). We found recurrent copy number aberrations expected in CCA including 1q gain in 80%, 7p gain in 50%, 1p loss in 70% and 3p loss in 50% of cases. Conclusions: ctDNA analysis for noninvasive tumor genotyping is feasible in patients with advanced cholangiocarcinoma. On-going work is focused on expanding these results using targeted and whole-exome sequencing.