Patients were identified over a four-year period starting in April 2015 and were deemed eligible for the study if they had a confirmed histological diagnosis of ICC. Written informed consent for tumor profiling was obtained from each patient upon their first admission to Fudan University Shanghai Cancer Center (FUSCC). The study protocol was approved by the FUSCC ethics committee (No. 218-1611 and No. 050432-4-1911D).
The clinical data and NGS results for 122 patients with ICC were available at the time of analysis. Overall survival (OS) and progression-free survival (PFS) rates were collected.
Survival data of 139 patients accepted curative surgery for ICC in the same center was used.
Sample collection and preparation:
Previously collected fresh tissue and blood samples were used in this study. The tissues were obtained through laparoscopic surgery or core needle biopsy. The fresh tissue was soaked in 5 times the volume of 4% formaldehyde solution within 30 min. A wax block was made within 24 h after soaking the tissue, and it was sent to the pathologist for diagnosis and review. The specimens were sent to the laboratory for NGS detection within 48 h at 4–8 °C. Twenty milliliters of peripheral blood was drawn and sent to the laboratory within 48 h at 15–35 °C.
Tissue samples with an estimated tumor purity <10% based on histopathological assessment were deemed insufficient for sequencing. The standard amount of DNA input was 250 ng, and the minimum input was 50 ng in cases for which the DNA quality was limited. Matched germline DNA from prospectively collected blood samples was analyzed for all patients.
Tissue and plasma DNA isolation and purification:
Genomic DNA (gDNA) was extracted from formalin-fixed, paraffin-embedded (FFPE) samples using the GeneRead DNA FFPE Kit (Qiagen, USA), and gDNA was extracted from the white blood cell samples using the DNA Blood Midi/Mini kit (Qiagen, USA). The quality of purified DNA was assayed by gel electrophoresis and quantified by the Qubit® 4.0 fluorometer (Life Technologies, USA).
Library construction and bioinformatics analysis:
Purified gDNA was first fragmented into DNA pieces approximately 200-300 bp in size using an enzymatic method (5X WGS Fragmentation Mix, Qiagen, USA). After end repair, tailing and T-adaptor ligation by polymerase chain reaction (PCR) was used to generate a prelibrary, and the products were then subjected to exon capture. Captured fragments were subsequently purified and hybridized by a 417-gene panel (supplementary table 1). FASTP7 was used to trim adapters and remove low-quality sequences to obtain clean reads, which were aligned to the Ensemble GRCh37/hg19 reference genome by BWA8. PCR duplicates were processed by GenCore9, and consensus reads were generated. SAMtools10 was utilized for the detection of single-nucleotide variations (SNVs), insertions and deletions, and Human Genome Variation Society (HGVS) variant descriptions were annotated by ANNOVAR11 software. After annotation, SNVs with a PopFreqMax > 0.05 were excluded, and nonsynonymous SNVs with a variant allele frequency (VAF) > 0.5% or a VAF > 0.1% in cancer hotspots collected from the patient database were retained for further analysis.
The microsatellite instability (MSI) statuses of all tissue samples were determined, and this score was used to classify the samples into three groups, MSI-high, ≥2 unstable microsatellite loci; MSI-low, only 1 instable locus; and microsatellite stable (MSS), no locus instability. The MSI-high results were further confirmed by PCR validation.
The tumor mutational burden (TMB) was estimated by somatic nonsynonymous mutations per megabase of the panel sequences examined.
Pathway enrichment was conducted in KEGG website (www.kegg.jp).
Immunohistochemical analysis of PD-L1:
The tissue was placed in a 4% paraformaldehyde solution for 12 h. Then, the tissues were dehydrated with 75% and 95% absolute ethanol each for 1.5 h at 60–70 °C. Next, the tissues were soaked in 40 °C dichloromethane for 4 h and then rehydrated with absolute ethanol, 95% ethanol, 75% ethanol and distilled water for 1 h, 0.5 h, 1 h and 1 h, respectively, at 60–70 °C. The samples were then embedded in paraffin wax, and ultrathin sections (5 µm) were cut using an ultramicrotome (Lecia RM2126RT, Germany), mounted on glass slides, and stained with hematoxylin and eosin for analyzing tissue structures using an upright fluorescence microscope (Eclipse TE2000-S, Nikon, Japan).
We performed immunohistochemical studies to evaluate programmed death ligand 1 (PD-L1) expression on tumor cells (TCs) and immune cells (ICs) using the Ventana SP263 assay with the Ventana BenchMark GX system (Roche/Ventana Medical Systems, Tucson, USA) according to the recommended protocol. A rabbit monoclonal anti-human PD-L1 antibody (clone SP263, Roche/Ventana) was used. The cellular staining pattern for the VENTANA PD-L1 (SP263) antibody is membranous and/or cytoplasmic staining of tumor cells. Immune cells demonstrate linear membranous, diffuse cytoplasmic, and/or punctate staining. The slides were immersed in acetone (3 min) and xylene (10 min) to remove the coverslip; the sections were then rehydrated with alcohol in decreasing concentrations and immersed in distilled water. Antigen retrieval was performed with Cell Conditioner 1 for 64 min against SP263. The sections were then incubated with the specific primary antibody for 16 min against SP263. Subsequently, the sections were treated with the OptiView HQ Linker for 8 min and the OptiView HRP Multimer for 8 min. Finally, counterstaining was performed with Mayer's hematoxylin and Scott's tap water bluing reagent. The evaluation of the stained tissue sections was performed by two investigators who had no knowledge of the patients’ clinical status. Cases with discrepancies were jointly re-evaluated until a consensus was reached. PD-L1 expression was calculated as the percentage of membrane staining on TCs or ICs in the overall area of the tumor, regardless of intensity.
Gemcitabine-based chemotherapy was used as the first-line chemotherapeutic treatment in this study, and combinatory strategies included cisplatin, oxaliplatin and capecitabine. Some patients accepted additional target/immune therapies (detailed in supplementary table 2). Patients refused the suggested target/immune therapies would accept chemotherapy alone. The Common Terminology Criteria for Adverse Events (CTCAE) 4.0 criteria were used to evaluate adverse events.
Follow-up was conducted every 8 weeks at the lowest frequency. Enhanced abdominal CT/MR scans and serum CA19-9 levels were examined, and Response Evaluation Criteria in Solid Tumors (RECIST) 1.1 criteria were used to evaluate therapeutic efficacy.
To evaluate the association of clinical characteristics or genes, Fisher’s exact test was performed. Odds ratios and false discovery rate (FDR)-corrected P values were also calculated. PFS was calculated using the Kaplan-Meier method, and the Chi-square test was used to compare therapeutic efficiencies between patients treated with different strategies and between those with different genetic alterations. The PFS rate was calculated as the time from the treatment start date to the date of progression or death. For patients who underwent surgical procedures, recurrence after curative resection or progression after palliative surgery was considered progression. Patients alive and without progression were censored to the date of the last follow-up.