DAGLβ is the principal synthesizing enzyme of 2-AG and promotes aggressive phenotype of intrahepatic cholangiocarcinoma via AP-1/DAGLβ/miR4516 feedforward circuitry

Abstract

The endocannabinoid system (ECS) is dysregulated in various liver diseases. Previously, we had shown that the major endocannabinoid 2-arachidonoyl glycerol (2-AG) promoted tumorigenesis of intrahepatic cholangiocarcinoma (ICC). However, biosynthesis regulation and clinical significance of 2-AG remain elusive. In the present study, we quantified 2-AG by gas chromatography/mass spectrometry (GC/MS) and showed that 2-AG was enriched in patients with ICC samples as well as in thioacetamide-induced orthotopic rat ICC model. Moreover, we found that diacylglycerol lipase β (DAGLβ) was the principal synthesizing enzyme of 2-AG that significantly upregulated in ICC. DAGLβ promoted tumorigenesis and metastasis of ICC in vitro and in vivo and positively correlated with clinical stage and poor survival in patients with ICC. Functional studies showed that activator protein-1 (AP-1; heterodimers of c-Jun and FRA1) directly bound to the promoter and regulated transcription of DAGLβ, which can be enhanced by lipopolysaccharide (LPS). miR-4516 was identified as the tumor-suppressing miRNA of ICC that can be significantly suppressed by LPS, 2-AG, or ectopic DAGLβ overexpression. FRA1 and STAT3 were targets of miR-4516 and overexpression of miRNA-4516 significantly suppressed expression of FRA1, SATA3, and DAGLβ. Expression of miRNA-4516 was negatively correlated with FRA1, SATA3, and DAGLβ in patients with ICC samples. Our findings identify DAGLβ as the principal synthesizing enzyme of 2-AG in ICC. DAGLβ promotes oncogenesis and metastasis of ICC and is transcriptionally regulated by a novel AP-1/DAGLβ/miR4516 feedforward circuitry.

NEW & NOTEWORTHY Dysregulated endocannabinoid system (ECS) had been confirmed in various liver diseases. However, regulation and function of 2-arachidonoyl glycerol (2-AG) and diacylglycerol lipase β (DAGLβ) in intrahepatic cholangiocarcinoma (ICC) remain to be elucidated. Here, we demonstrated that 2-AG was enriched in ICC, and DAGLβ was the principal synthesizing enzyme of 2-AG in ICC. DAGLβ promotes tumorigenesis and metastasis in ICC via a novel activator protein-1 (AP-1)/DAGLβ/miR4516 feedforward circuitry.

INTRODUCTION

Cholangiocarcinoma originates from the neoplastic transformation of cholangiocyte that line the intra- and extrahepatic bile ducts. The incidence of intrahepatic cholangiocarcinoma (ICC) is increasing worldwide as revealed by The American Cancer Society and The National Central Cancer Registry of China (NCCR; 1–4). Characteristics of well-recognized risk factors of ICC are chronic inflammation caused by biliary infections or cholestasis. Therapeutic challenges posed by ICC are daunting as onset of ICC is usually insidious and only less than 30% of patients with ICC are eligible candidates for curative resection. Moreover, ICC is refractory to conventional chemotherapy and radiotherapy (5–7). Therefore, development and identification of novel therapeutic targets of ICC remain scientific rationale in clinical practice.

Endocannabinoids (ECs) including arachidonoyl ethanolamine (AEA) and 2-arachidonoyl glycerol (2-AG) are endogenous signaling lipids as mediated by the bioactive effects of Cannabis sativa (Δ⁹-tetrahydrocannabinol; THC). However, these molecules are not just significant factors in the context of cannabis but are a feature of other biological phenomena. AEA and 2-AG are generated on demand in response to elevated intracellular calcium or metabotropic receptor activation. The endocannabinoid system (ECS) consists of cannabinoid receptors, their endogenous ligands (endocannabinoids), and enzymes involved in EC biosynthesis and degradation (8). Dysregulated ECS had been demonstrated in various liver diseases including nonalcoholic fatty liver disease (NAFLD), liver fibrosis, and hepatocellular carcinoma (9). Previously, we had revealed that AEA and 2-AG exerted opposing effects on ICC proliferation; 2-AG promoted ICC tumorigenesis via cannabinoid (CB) receptors-independent, lipid raft-dependent activation of the Notch 2/Presenilin-2 pathway (8, 10). However, biosynthesis regulation and clinical significance of 2-AG in ICC remain elusive (10, 11). Biosynthesis of 2-AG involves the hydrolysis of arachidonic acid-esterified diacylglycerols (DAGs) at the sn-1 position mainly by two evolutionary homologous diacylglycerol lipases, diacylglycerol lipase α (DAGLα) and diacylglycerol lipase β (DAGLβ). Though both DAGLs synthesize 2-AG, production of 2-AG by DAGLs exhibits cell type and tissue specificity. DAGLα enriches in synapse-rich regions of central nervous systems where it involves in regulating synaptic activity and neuroinflammation, whereas DAGLβ mainly enriches in cells of mononuclear phagocytic system and peripheral tissues where it involves in regulating inflammatory responses (12). Genetic knockout of DAGLβ in mice leads to moderate reduction of 2-AG in brain, whereas induces ∼90% reductions of 2-AG and 70% reductions of arachidonic acid in the liver (13, 14). Until now, regulation and function of DAGLβ in ICC have not yet been elucidated.

In the present study, we have shown that DAGLβ was the principal synthesizing enzyme of 2-AG in ICC. DAGLβ can promote tumorigenesis and metastasis of ICC in vitro and in vivo. The expression of DAGLβ was significantly upregulated and positively correlated with tumor aggressiveness and poor prognosis in patients with ICC. Transcription of DAGLβ is regulated by an activator protein-1 (AP-1)/DAGLβ/miR4516 feedforward circuitry, which can be augmented by lipopolysaccharide (LPS).

MATERIALS AND METHODS

Patients and Clinical Samples

Patients with pathologically confirmed intrahepatic cholangiocarcinoma (n = 87) from the First Affiliated Hospital of Sun Yat-sen University and the Sun Yat-sen University Cancer Center from May 2008 to October 2020 were enrolled in clinicopathological correlation analysis. Three different tissue cores (1.5 mm in diameter) reflecting the typical region of ICC were sampled from each patient’s paraffin blocks. These paraffin tissue cores of ICC were assembled into tissue microarrays by the MTA1-manual tissue arrayer (Beecher Instruments, Inc.; 15). Patients’ follow-up was carried out by two independent physicians unaware of the study. The overall survival (OS) was defined as the time interval from surgery to any cause of death. The relapse-free survival (RFS) was defined as the time interval from surgery to clinical evidence of tumor recurrence. Paired tumor and adjacent normal intrahepatic bile duct (>3 cm distant from the tumor) from 27 patients with ICC from August 2019 to April 2020 were obtained by microdissection before RNA or protein extraction. This study was approved by the Ethical Committee of the First Affiliated Hospital of Sun Yat-sen University along with the Sun Yat-sen University Cancer Center (Code: 2021[170] and RDDB2020000851). Informed consent was obtained from all patients.

Integrated Bioinformatics Analysis of Cholangiocarcinoma

Microarray data sets (GSE32225/45001/26566/34166/89749/132305) and RNA-sequencing data set (GSE107943) of the Gene Expression Omnibus (GEO) database, ICC proteome data set (OEP001105) of the NODE database, and the cholangiocarcinoma data set (CHOL) of The Cancer Genome Atlas (TCGA) database were used for external validation analysis. The probe expression matrixes were annotated and normalized as gene expression intensities in the log-transformed format of raw value for microarray or FPKM/TPM for RNA-seq before the cross-platform integration. Expression data were integrated by Perl script and batch-corrected by the Rank-In online platform (http://www.badd-cao.net/rank-in/; 16). Then T test was used to compare difference of DAGLβ expression between tumor and adjacent normal tissue.

Cell Lines and Materials

Human ICC cell lines (CCLP-1 and HuCCT1), gall bladder cancer cell line (Mz-ChA-1), and immortal benign cholangiocyte cell line (H69) were gifts from Dr. G. Fitz (University of Texas Southwestern Medical Center, Dallas, TX), Dr. A. J. Demetris (University of Pittsburg, PA), and Dr. G. J. Gores (Mayo Clinic, Rochester, MN), respectively. Human ICC cell lines (RBE and QBC939), human myeloid leukemia cell line THP-1, and human normal immortal hepatocyte cell line LO2 were purchased from Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences (Shanghai, PR China). The primary human intrahepatic cholangiocytes HIBEC was purchased from Sciencell (Carlsbad, CA). Culture media was complemented with 10% fetal bovine serum and 1% penicillin-streptomycin (Gibco) in a 5% CO₂ humidified incubator (Thermo Fisher Scientific) at 37°C.

SP600125, pharmacological inhibitors of JNK (HY-12041, MedChemExpress), KT109, inhibitors of DAGLβ (SML1364, Sigma-Aldrich), lipopolysaccharide (L6529, Sigma-Aldrich), and Matrigel Matrix (356234 and 354248, Corning) were applied in the functional experiments. Lentiviral open reading frame (ORF) expression clones for DAGLβ (EX-T3603-Lv207) and empty control vector (EX-NEG-Lv207), shRNA lentiviral particles targeting DAGLα (LPP-HSH111196-LVRH1MP), DAGLβ (LPP-HSH095957-LVRH1MP), and scrambled control (LP516-025) were designed by Genecopoeia. ORF expression plasmid for c-Jun (SC118762) was purchased from Origene, and ORF expression plasmid for FRA1 (HG12926-G) was purchased from Sino Biological (Beijing, PR China). The siRNA of c-Jun and FRA1 (No. SIGS0003638-1 and No. SIGS0008462-1), the miR-4516 mimic, and miR-4516 inhibitor (miR10019053, miR214423153601) were designed by Ribobio (Guangzhou, PR China). Antibodies for DAGLβ (D4P7C, Cat. No. 12574), c-Jun (60A8, Cat. No. 9165), FRA1 (D80B4, Cat. No. 5281), Stat3 (D3Z2G, Cat. No. 12640), PCNA (D3H8P, Cat. No. 13110), Ki67 (8D5, Cat. No. 9449), and β-tubulin (Cat. No. 2146) were purchased from Cell Signaling Technology. Antibodies for DAGLβ (ab214233), DAGLα (ab81984), monoglyceride lipase (MAGL) (ab119777), CK19 (ab52635), matrix metalloproteinase (MMP)-9 (ab76003), MMP-2 (ab86007), and GAPDH (ab8245) were purchased from Abcam. Bouin’s fixative solution (PH0976) was purchased from Phygene (Fujian, PR China). d-Luciferin (C3654) was purchased from APExBIO (Houston, TX).

RNA Extraction and Quantitative Real-Time PCR

Extraction of total RNA from ICC samples or cell lines was performed with TRIzol reagent (Invitrogen). After determination of RNA concentration, cDNA synthesis was performed by SureScript First-Strand cDNA Synthesis Kit, whereas miRNA was synthesized by All-in-One miRNA qRT-PCR Detection Kit 2.0 (GeneCopoeia). qRT-PCR was performed with BlazeTaq One-Step SYBR Green RT-qPCR Kit or All-in-One miRNA qRT-PCR Detection Kit 2.0 (GeneCopoeia) and analyzed by the ViiA 7 Real-Time PCR System (Applied Biosystems). Primers for ACTB (Cat. No. HQP108762), DAGLβ (Cat. No. HQP005337), DAGLα (Cat. No. HQP018537), MAGL (Cat. No. HQP001606), YY1 (Cat. No. HQP018570), RBPJ (Cat. No. HQP009574), NFIC (Cat. No. HQP067069), FRA1 (Cat. No. HQP019708), c-Jun (Cat. No. HQP009853), RNU6-2 (Cat. No. HmiRQP9001), and hsa-miR-4516 (Cat. No. HmiRQP2159) were purchased from GeneCopoeia.

Western Blot Analysis

Total protein from tissue samples or cell lines were extracted by RIPA buffer (Beyotime) supplemented with Protease Inhibitor Cocktail (CoWin Biosciences). Quantification of total protein was performed with bicinchoninic acid (BCA) protein assay kit (Thermo Scientific). For each sample, 10 μg of protein suspended in RIPA buffer was loaded onto a 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). After electrophoresis, total proteins were transferred to a 0.45-μm polyvinylidene difluoride membrane. Membranes were further blocked in 1× protein free rapid blocking buffer (Epizyme Biotech) for 10–15 min. After sequential incubation with primary antibodies (overnight at 4°C) and corresponding secondary antibodies (1 h at room temperature), chemiluminescent horseradish peroxidase substrate assay (Millipore) was used for detection of antibodies binding. GAPDH and β-tubulin were used as internal controls. Quantification of protein expression was performed with Image J software.

Quantification of 2-AG by Gas Chromatography/Mass Spectrometry

1-Stearoyl-2-arachidonoyl-sn-glycerol (SAG), precursor of 2-arachidonoyl glycerol (2-AG), was applied to evaluate the biosynthesis of 2-AG by ICC cell. Standards of 2-AG (Cat. No. 62160) and SAG (Cat. No. 10008650) were purchased from Cayman. The parameter of the gas chromatography/mass spectrometry was calibrated and the standard curves of each standards were prepared.

For sample preparation, 5 mg of ICC tumor tissues or adjacent normal intrahepatic bile duct were ground in liquid nitrogen and resuspended with 1 mL of sterile double-distilled water (for other type of sample: 1 mL of bile or 1 mL of serum with EDTA were acquired from patients with ICC; 5 × 10⁶ culture cells were resuspended with 1 mL of ddH₂O or 1 mL of culture supernatant in T25 flask). The mix was transferred to a 15-mL centrifuge tube, and 6 mL of liquid mixture of trichloromethane/methanol (4:2 vol/vol) was added, vortexed on ice, and centrifuged at 800 g for 15 min at 4°C. Transferred the lower liquid phase into a fresh 15-mL centrifuge tube. Dried the sample under nitrogen and resolved the sample with 1 mL of precold ethanol and then added 9 mL of precold sterile distilled water, adjusted the pH to 3.0. Total lipids were extracted by solid-phase extraction with a C-18 Sep-Pak cartridge (Cat. No. 12113001, Agilent). Briefly, samples were pipetted into the activated C-18 Sep-Pak cartridge and vacuum aspirated. After extraction, total lipids were dissolved by 3 mL of ethyl acetate/acetonitrile (1:1 vol/vol) and dried under nitrogen. A 1-μL internal reference standard of 2-arachidonoyl glycerol-d5 (2-AG-d5, 5 μg/mL) was added, and 40 μL of silylating agent-dimethylisopropylsilyl imidazole (Cat. No. D1596, TCI, Tokyo, Japan) were added to the aforementioned lipid sample. Silanization was performed in dark for ∼1 h at room temperature. Quantification of 2-AG was performed by gas chromatography/mass spectrometry (GC/MS; QP2010 Ultra High-End GC-MS, Shimadzu, Japan) with parameters listed in Table 1 (17). The characteristic ion peaks of 2-AG and the corresponding deuterated internal reference standard (2-AG-d5) are shown in Supplemental Fig. S1. 2-Arachidonoyl glycerol-d5 (Cat. No. 362162) and 1-stearoyl-2-arachidonoyl-d8-sn-glycerol (Cat. No. 10009872) were purchased from Cayman (Ann Arbor, MI).

Table 1.

Parameters of characteristic ion peaks of each reference standard in gas chromatography/mass spectrometry

Reference Standard	m/z Quantifier Ion	m/z Qualifier Ions	Retention Time, min
2-AG	535	479	18.35
2-AG-d5	540
1-AG	535	479	18.70

2-AG, 2-arachidonoyl glycerol; 2-AG-d5, 2-arachidonoyl glycerol-d5.

Cell Transfection

For transient transfection of siRNA, plasmid DNA, miRNA mimic/inhibitor, lipofectamine 3000 (Invitrogen, Carlsbad, CA) were used. Briefly, when cells grew on T25 flask to 60% confluency, transfection complex was prepared and incubated for 15 min and then added to the cells. Forty-eight hours after transfection, cells were used for further analysis. For stable lentiviral transduction of vectors expressing DAGLβ or mock control, shRNA constructs targeting DAGLβ or scramble control, lentivirus packaging was used for infection following the methodology described previously (15). 293T cells were used for lentivirus packaging, and the supernatant was collected after 48 h to transfect HuCCT1 cell or RBE cell. Twenty-four hours after transfection, puromycin was used to screen resistant clones (exposure to puromycin for 3–5 days during which selection occurred).

Cell Counting Kit-8 Assay

CCK-8 assay was used to evaluate cellular proliferation. Briefly, cells were seeded into 96-well plates at a density of 2,000 cells/well in sextuplicate. Cell viability was measured for 4 days using the CCK-8 system (Dojindo, Japan). After incubation with CCK-8 solution at 37°C for 2 h, the optical density (OD) values at 450 nm were determined by microplate reader.

Soft Agar Colony Formation Assay

Cells were trypsinized and resuspended to prepare a single-cell suspension for the experiment. First, a bottom layer of agar with complete media was poured and solidified (1.4% agar). Then an upper layer with a specified number of cells was suspended in a medium-agar mixture (0.7% agar). When the agar became solid, 1 mL of complete medium was added to the cell layer. The size and number of colonies were measured and counted after 2 wk of incubation (18).

Wound Healing Assay

Cells were cultured in six-well plates at 37°C. When the cells grew at 90% confluence, an artificial straight scratch wound was drawn at the middle bottom of each well by using a sterile 1,000-μL pipet tip. Images were captured every other day and Image J software was used for analysis.

Transwell Cell Migration and Invasion Assay

Both of the cell migration and invasion were performed by 24-well transwells (8-μm pore size; Falcon). Cells (5 × 10⁴) were seeded into the upper chamber with or without Matrigel mixture for invasion or migration assay, respectively. For invasion assay, the chambers were incubated with Matrigel overnight at 37°C. Complete culture medium of 750 μL was added to the lower chambers, whereas 200 μL of serum-free medium containing cells was added to the upper chambers. After incubation for 24 h, medium in the upper chambers was removed and rinsed with PBS. The chambers were then incubated with 4% paraformaldehyde and stained with 0.1% crystal violet. Cells retained in the upper layer of the chambers were removed by cotton swabs. Cells were counted in three random microscopic fields at high magnification. All experiments were repeated three times to reach statistical difference.

Immunohistochemistry

Immunoreactivity of DAGLβ or other proteins were detected in the aforementioned tissue microarrays as well as paraffin-embedded tissue sections using corresponding primary antibodies, then assessed quantitatively by the mean intensity of the area of interest under the representative field by Image Pro-Plus 7.0 software (Media Cybernetics, Rockville, MD). Expression of DAGLβ in each sample was evaluated by mean intensity of the three cores or tissue sections and grouped into score 0 (<25% of mean intensity of control group), score 1 (25%–50% of mean intensity of control group), score 2 (50%–75% of mean intensity of control group), and score 3 (75%–100% of mean intensity of control group). The high and low DAGLβ expression groups were then divided according to the median score. Immunohistochemical analysis was conducted by three independent pathologists who were blinded to the related clinical data.

Construction of Full-Length and Sequentially Truncated Promoter and Site-Mutated Promoter of DAGLβ

The full-length and sequentially truncated promoters of DAGLβ were generated by PCR amplification (primers were listed in Table 2). Promoter activities were determined by a transient dual-luciferase reporter assay involving the full-length and sequentially truncated promoters constructed into the dual luciferase reporter plasmid. Transcription factor binding site was determined using the construct with site-mutated cis-elements of the full-length promoter of DAGLβ. Site-mutated construct of DAGLβ promoter was generated by Quick-Change Site-Directed Mutagenesis Kit (Stratagene No. 2001518, La Jolla, CA) as instructed.

Table 2.

PCR primers for cloning the full-length and sequential truncated of DAGLβ promoter

	Primers	Product Size, bp	Mapping Site from TSS, bp
Full length	F: 5′- CAGGGTGCAGGCTTTATCTG-3′	1,676	−1,422 to +254
	R: 5′- AAGCTTCACGACCAGCTCGAAGAACC-3′
Truncate a	F: 5′- CTCGAGAGAGCCATACGCTGCCTTAA-3′	785	−531 to +254
	R: 5′- AAGCTTCACGACCAGCTCGAAGAACC-3′
Truncate b	F: 5′- CTCGAGCCTGGGAACAGAGCGAGACT-3′	498	−244 to +254
	R: 5′- AAGCTTCACGACCAGCTCGAAGAACC-3′
Truncate c	F: 5′- CTCGAGGATTTCACCGACCGGCAGAC-3′	359	−105 to +254
	R: 5′- AAGCTTCACGACCAGCTCGAAGAACC-3′

DAGLβ, diacylglycerol lipase β; TSS, transcription start site.

Bioinformatic Annotation for DAGLβ Promoter

The human DAGLβ promoter region was identified by UCSC Genome Browser (http://genome.ucsc.edu/). To predict transcription factors binding to the DAGLβ promoter, bioinformatics retrieval was performed in ENCODE database and PROMO database (factors predicted within a dissimilarity margin less than or equal to 15%). Prediction of transcription factors binding sites of the DAGLβ promoter was performed in the JASPAR 2022 database (https://jaspar.genereg.net/).

DNA Pull-Down Assay

The DNA pull-down kit was purchased from BersinBio (Guangzhou, PR China). Briefly, after preparation of the desulfurized biotin-labeled DNA probes mapping to the DAGLβ promoter region (−1,422 bp to +254 bp), we conducted the DNA pull-down assay with aforementioned DNA probes and streptavidin magnetic beads. Proteins pulled down by DNA probes were further dissociated by elution buffer and separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis. The special bands corresponding to the silver staining were excised from the gel and analyzed using liquid chromatography-mass spectrometry. The amino acid sequences of peptides were analyzed on the MASCOT search engine (http://www.matrixscience.com/) in the SwissProt database.

Chromatin Immunoprecipitation Assay

The DNA sequence bound to c-Jun and FRA1 protein was performed with chromatin immunoprecipitation using CUT&RUN Assay Kit (CST No. 86652, Danvers, MA). To put it simply, protein-DNA complexes were cross-linked, immunoprecipitated, and purified, and then the DAGLβ promoter sequence was amplified by qPCR. The primers are shown in Table 3.

Table 3.

PCR primers for ChIP-qPCR and CUT&RUN pull-down assay

Gene	Primer Sequence	Mapping Site from TSS, bp	Product Size, bp
DAGLβ	F: 5′- TAAATTAGCCGGGAGTGTGG-3′	−990	174
	R: 5′- TTAAGGCAGCGTATGGCTCT-3′
DAGLβ	F: 5′- AGGGTGCAGGCTTTATCTGA-3′	−288	154
	R: 5′- CTGAGCGCTTTTCAGTACCC-3′

TSS, transcription start site.

Dual-Luciferase Reporter Assay

The full-length or series of truncated promoter fragments and mutated promoter sequence of DAGLβ gene were cloned into pGL3-BASIC plasmid. To explore the core region of DAGLβ promoter and examine whether c-Jun binds to DAGLβ promoter, a Dual-Luciferase Reporter Kit (Promega No. E1910, Madison, WI) was performed as instructed (19).

TAA-Induced Orthotopic Model of ICC in Rat

Adult (8-wk old) male Sprague-Dawley (SD) rats (330–370 g, n = 15) were used in the present study. Thioacetamide (TAA; 500 mg/L; 0.05%) was orally fed with sweetened drinking water every day (20). Food and water were available ad libitum. During induction of ICC development, the rats were weighed weekly. After 28 wk, the rats were euthanized and the liver and serum samples were harvested.

Subcutaneous Xenograft Model of ICC in Mice

All animal experiments were performed with protocols reviewed and approved by the Insights and Considerations for Ethics (ICE) for Clinical Research and Animal Trails of the First Affiliated Hospital, Sun Yat-sen University, Guangzhou, China (Approval Code: [2021]170). Adult (6- to 8-wk old) male NOD/SCID/gamma-null (NSG) mice were obtained from the Laboratory Animal Center of First Affiliated Hospital, Sun Yat-sen University, Guangzhou, China. Firefly luciferase was stably transfected in RBE or HuCCT1 tumor cells, which were stably expressing DAGLβ or mock control, as well as shDAGLβ or shScramble. Viable tumor cells were collected and resuspended in 150 μL of PBS/mouse with 1:1 (vol/vol) Matrigel mix. Tumor cells (150 μL) (∼8 × 10⁶ to 1 × 10⁷ cells) were subcutaneously inoculated in the left axilla of the mice. One-hundred-fifty microliters with the same amount of tumor cells expressing mock control or shScramble were subcutaneously inoculated in the right axilla. Noninvasive bioluminescence imaging (BLI) was performed with the multimodal imaging system for living animals (AniView100, Guangzhou, PR China). Briefly, mice were injected intraperitoneally with d-luciferin (150 mg/kg) and were imaged 10 min after injection. Photon intensities were quantified to monitor the development of xenografted tumor (21). Tumors were measured with a caliper every 3 days, and tumor volume was calculated according to the formula: L × W²/2, where L and W are the tumor’s length and width, respectively.

In Vivo Multiorgan Metastasis Model of ICC in Mice

Intrahepatic metastasis model of ICC.

NSG mice (4- to 6-wk old) were anesthetized with pentobarbital (50 mg/kg ip). The surgical site was sterilized and a 1-cm incision was made to access the abdominal cavity. The spleen was exposed by a slight pressure. Firefly luciferase were stably transfected in HuCCT1 tumor cells, which were stably expressing DAGLβ or mock control, as well as shDAGLβ or shScramble. Viable tumor cells were collected and resuspended in PBS. Tumor cells (1 × 10⁶) in 40 μL of PBS of each study group were injected into the lower pole of the spleen. The abdominal wall was sutured closed by 4-0 coated Vicryl plus Antibacterial/polyglactin 910. Heating pad was prepared to maintain body temperature until fully recovered. All mice were weighed once every 3 days, and intrahepatic metastatic lesions of ICC were measured with multimodal imaging system for living animals (AniView100, Guangzhou, PR China) as aforementioned (21).

Bilateral lung metastasis model of ICC.

Tumor cells (1 × 10⁶) resuspended in 100 μL of PBS of aforementioned study groups were injected through tail veins of NSG mice (4 to 6 wk old). All mice were weighed once every 3 days, and lung metastatic lesions of ICC were measured with multimodal imaging system for living animals (AniView100, Guangzhou, PR China).

Peritoneal metastasis model of ICC.

Tumor cells (1 × 10⁶) resuspended in 100 μL of PBS of aforementioned study groups were injected into the peritoneal cavity of NSG mice (4- to 6-wk old). All mice were weighed once every 3 days, and lung metastatic lesions of ICC were measured with multimodal imaging system for living animals (AniView100, Guangzhou, PR China).

MicroRNA Array Analysis

Total RNAs were extracted from RBE cells treated with 2-AG (10 μM for 2 wk, n = 3) or control (10 μM DMSO for 2 wk, n = 3) using TRIzol reagent (Invitrogen). Total microRNAs were further isolated with RNeasy mini kit (QIAGEN) following manufacturer’s instructions. Microarray array analysis was performed according to Exiqon’s manual. The threshold of differently expressed microRNAs was |log2 (fold change)| > 1 and P < 0.05.

Statistical Analysis

All experiments were repeated at least three times unless otherwise noted. Statistical analyses were performed with SPSS 24.0 and GraphPad Prism 7.0 (La Jolla, CA). All data are presented as the means ± SD. Comparisons were performed using Student’s unpaired t test, Mann–Whitney’s U test, or one-way ANOVA. Survival curves were estimated using Kaplan–Meier survival analysis, subsequently evaluated by Log Rank test. A two-tailed P < 0.05 was considered statistically significant.

RESULTS

DAGLβ is the Main Synthetase of 2-AG in ICC and is Closely Related with Adverse Prognosis

Previously, we had demonstrated that 2-AG, the principal endocannabinoid, promoted tumorigenesis of cholangiocarcinoma via CB receptors-independent, lipid raft-dependent activation of the Notch 2/Presenilin-2 pathway (8). However, biosynthesis, expressional regulation, and clinical significance of 2-AG in ICC remain elusive. In the present study, we demonstrated that 2-AG was enriched in ICC cell lines and tumor parenchyma (TAA-induced orthotopic ICC model of rat and clinical samples; Fig. 1, A and B, and Supplemental Fig. S1), which was quantified by gas chromatography/mass spectrometry (GC/MS). To explore the role of DAGLα and DAGLβ in 2-AG biosynthesis in ICC, we knocked down the expressions of DAGLα and DAGLβ by shRNA in RBE cells, respectively. We showed that DAGLβ was the principal synthetase of 2-AG and that 2-AG synthesis was reduced by ∼75% in RBE cells with DAGLβ knockdown, after incubation with 1-stearoyl-2-arachidonoyl-sn-glycerol (SAG), the precursor of 2-AG (Fig. 1C).

Figure 1.

Evidence for increased 2-arachidonoyl glycerol (2-AG) synthesis and elevated diacylglycerol lipase β (DAGLβ) expression in intrahepatic cholangiocarcinoma (ICC). A: total lipids of clinical patients with ICC samples (n = 8) and of thioacetamide (TAA)-induced orthotopic model of ICC in rat tissues (n = 15) were extracted by solid-phase extraction, and 2-AG was quantified by gas chromatography/mass spectrometry (GC/MS). 2-AG was significantly enriched in tumor parenchyma. B: total lipids of the cell lysate of human ICC and gall bladder carcinoma cell lines (Mz-CHA-1, CCLP-1, RBE, and QBC939) and of human normal cholangiocytes (H69 and HIBEC) were extracted by solid-phase extraction, and 2-AG was quantified by GC/MS (n = 3). 2-AG was significantly elevated in ICC cell lines. C: expressions of diacylglycerol lipase alpha (DAGLα) and DAGLβ were knocked down by shRNA in RBE cells, respectively. Protein expressions of DAGLα and DAGLβ were determined by Western blot (n = 3). 2-AG synthesis in RBE cells was induced by incubation with 1-stearoyl-2-arachidonoyl-sn-glycerol (SAG), the precursor of 2-AG. Total lipids of the cell lysate and supernatant were extracted by solid-phase extraction, respectively, and 2-AG was quantified by GC/MS. 2-AG synthesis was significantly reduced by about 75% in the cell lysate and reduced by about 65% in the supernatant of cells with DAGLβ knockdown, whereas 2-AG synthesis was reduced by about 50% both in the cell lysate and in the supernatant of cells with DAGLα knockdown (n = 3). D: total RNAs and proteins were extracted from frozen clinical ICC samples. mRNA expression and protein expression of DAGLβ were determined by qRT-PCR (n = 17) and Western blot (n = 10), respectively. Expression of DAGLβ was significantly increased in clinical ICC patients’ samples. E: total RNAs and proteins were extracted from human ICC and GBC cell lines (Mz-CHA-1, CCLP-1, RBE, QBC939, HuCCT1) and from normal human cholangiocytes (H69 and HIBEC). mRNA expression and protein expression of DAGLα, DAGLβ, and monoglyceride lipase (MAGL) were determined by qRT-PCR and Western blot, respectively. Expression of DAGLα, DAGLβ, and MAGL were elevated in ICC. F: representative images of TAA-induced orthotopic model of ICC in rat and the normal control are shown. There were diffuse tumor nodules in the liver of the TAA treatment group. Hematoxylin-eosin (H&E) staining and immunohistochemistry staining analysis were performed to determine DAGLβ expression in formalin-fixed paraffin-embedded tissue sections of orthotopic ICC model in rat. Protein expression of DAGLβ was significantly increased in TAA-induced orthotopic model of ICC in rat. Scale bars = 50 μm, ×200; Scale bars = 25 μm, ×400. CK19, cytokeratin 19. Error bars represent means ± SD; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, according to Student’s t test.

Expressions of DAGLβ were increased in biliary tract cancer (BTC) cell lines and clinical ICC samples when comparing their nonpathological counterparts (human immortalized cholangiocyte cell line H69 and human intrahepatic biliary epithelial cells HIBEC; clinical adjacent normal tissue or biliary epithelium; Fig. 1, D and E). Elevated DAGLβ expression was also demonstrated in tumor parenchyma of TAA-induced orthotopic ICC model of rat (Fig. 1F; 20). Furthermore, we performed correlation analyses of DAGLβ expression with clinical pathophysiological parameters in a clinical tissue microarray of 87 patients with ICC (Table 4; Fig. 2A). We demonstrated that DAGLβ immunoreactivity was elevated in ICC accompanied with hepatolithiasis (P = 0.03) and positively correlated with carcinoembryonic antigen (CEA) (P = 0.00), tumor size (P = 0.03), lymph node metastasis (P = 0.05), and advanced tumor, lymph node and metastasis status (TNM) stage (III-IV; P = 0.01; Table 5). Survival analysis also revealed that DAGLβ expression predicted shorter recurrence-free survival and overall survival (Fig. 2B). External validation of DAGLβ elevation in ICC was performed by integrative analysis of the TCGA and GEO databases, as revealed in the cholangiocarcinoma data set (CHOL) of TCGA database and in GSE32225/107943/45001 data sets of GEO database after batch normalization. Heterogeneity of DAGLβ expression in cholangiocarcinoma (GSE26566/34166/89749) or in extrahepatic cholangiocarcinoma (GSE132305) was not significant (Fig. 2C). Moreover, survival analysis of the CHOL data set of TCGA (n = 30), ICC data set (GSE89749) of GEO database (n = 103), and ICC proteome data set (OEP001105) of the NODE database (n = 165) revealed that elevated DAGLβ predicted trend of adverse prognosis [CHOL data set of TCGA, P = 0.773, maybe due to small sample size and, therefore, this particular correlation should be viewed with caution; ICC data set (GSE89749), P = 0.052; ICC proteome data set (OEP001105), P = 0.047] (Fig. 2D). Overall, these data suggested that DAGLβ was the principal synthesizing enzyme of 2-AG and DAGLβ might exert an oncogenic role in ICC.

Table 4.

Demographics and clinical characteristics of patients with ICC in present study

Characteristics	Values
Age, year (means ± SD)	56.98 ± 12.34
Gender (male/female)	50/37
Hepatolithiasis (−/+)	58/29
Primary sclerosing cholangitis (−/+)	85/2
HBsAg (−/+)	58/29
HBcAb (−/+)	53/34
CA19-9 (≤35 U/L/>35 U/L)	34/53
CEA (≤5 μg/L/>5 μg/L)	50/37
Tumor size (≤5 cm/>5 cm)	37/50
Tumor differentiation (<moderate/≥moderate)	42/45
Number of tumor (single/multiple)	62/25
Lymph node metastasis (−/+)	59/28
AJCC 8th TNM stage (I, II/III, IV)	48/39

ICC, intrahepatic cholangiocarcinoma.

Figure 2.

Diacylglycerol lipase β (DAGLβ) is closely related with adverse prognosis in patients with intrahepatic cholangiocarcinoma (ICC). A: IHC analysis of DAGLβ expression was performed in formalin-fixed paraffin-embedded tissue sections of clinical ICC samples (n = 87). IHC staining intensity of DAGLβ expression was divided into four grades as scoring (0 to 3; 0 = low, 3 = high). Scale bars = 250 μm, ×40; scale bars = 100 μm, ×100; scale bars = 25 μm, ×400. B: patients with ICC of present cohort were divided into two groups according to mean IHC staining intensity of DAGLβ expression (staining score 0 and 1: low expression group, n = 36; staining score 2 and 3: high expression group, n = 37). The correlations between DAGLβ expression and patients with ICC prognosis [relapse-free survival (RFS) and overall survival (OS)] were performed by Kaplan–Meier analysis and log-rank tests. C: bioinformatic analysis of DAGLβ mRNA expression was performed in the cholangiocarcinoma data set (CHOL) of The Cancer Genome Atlas (TCGA) database (cholangiocarcinoma, n = 33, normal bile duct, n = 8); ICC data sets (GSE32225/107943/45001, ICC, n = 189; normal bile duct, n = 6; surrounding normal liver tissues, n = 37) cholangiocarcinoma (CCA) data sets (GSE26566/34166/89749, CCA, n = 228; normal bile duct, n = 12; surrounding normal liver tissues, n = 59) and extrahepatic cholangiocarcinoma (ECC) data set (GSE132305, ECC, n = 182; normal bile duct, n = 38) of the Gene Expression Omnibus (GEO) database were also applied. DAGLβ mRNA expression data of ICC data sets (GSE32225/107943/45001) and CCA data sets (GSE26566/34166/89749) of the GEO database were integrated by Perl script and batch-corrected by the Rank-In online platform. D: external validation of correlation between DAGLβ expression and patients’ prognosis was performed in the CHOL data set of TCGA (n = 30), ICC data set (GSE89749) of GEO database (n = 103), and ICC proteome data set (OEP001105) of the NODE database (n = 165). Error bars represent means ± SD; *P < 0.05, ***P < 0.001, ****P < 0.0001, according to Student’s t test.

Table 5.

Correlation analyses of DAGLβ expression with clinical pathophysiological parameters in patients with ICC

Characteristics	Number of Patients		P Value
	Low DAGLβ expression	High DAGLβ expression
Gender			0.45
Male	27	23
Female	17	20
Age			0.91
≤57	22	21
>57	22	22
Hepatolithiasis			0.03
Negative	34	24
Positive	10	19
Primary sclerosing cholangitis			>0.99
Negative	43	42
Positive	1	1
HBsAg			0.22
Negative	32	26
Positive	12	17
HBcAb			0.06
Negative	31	22
Positive	13	21
CA19-9			0.09
≤35 U/L	21	13
>35 U/L	23	30
CEA			0.00
≤5 μg/L	32	18
>5 μg/L	12	25
Tumor size			0.00
≤5 cm	25	12
>5 cm	19	31
Tumor differentiation			0.59
<Moderate	20	22
≥Moderate	24	21
Number of tumor			0.86
Single	31	31
Multiple	13	12
Lymph node metastasis			0.05
Negative	34	25
Positive	10	18
AJCC 8th TNM stage			0.01
I/II	30	18
III/IV	14	25

DAGLβ, diacylglycerol lipase β; ICC, intrahepatic cholangiocarcinoma. Bold letters represented P values < 0.05.

AP-1 Binds to the DAGLβ Promoter and Promotes Its Transcription

To investigate the transcriptional regulation of DAGLβ, the core promoter region of DAGLβ was determined by dual luciferase reporter assay (Promega). First, we cloned the full-length (−1,422 bp to ±254 bp) and sequential truncated promoter of DAGLβ by PCR in RBE cells; clone products were constructed into luciferase reporter plasmids, respectively. Results revealed that the core promoter region of DAGLβ was located in −244 bp to −105 bp from the transcription start site (TSS; chr7: 6,448,256–6,448,117; Fig. 3A). To identify transcription factors binding to the DAGLβ promoter, we performed desulfurized biotin-labeled DNA pull-down assay in RBE cells with DNA probe mapping −1,422 bp to +254 bp of DAGLβ promoter. Protein pulled down by DNA probe were further identified by liquid chromatography tandem mass spectrometry (LC/MS). There were 734 proteins identified by LC/MS. By integration analysis of present DNA pull-down assay and bioinformatics analysis of DAGLβ promoter in ENCODE database and PROMO database, four candidate transcription factors were identified: c-Jun, NFIC, RBPJ, and YY1 (Fig. 3B). Each of these four candidate transcription factors was knocked down by si-RNA in RBE cells and results indicated that c-Jun regulated DAGLβ transcription (Fig. 3C and Supplemental Fig. S2A). Moreover, bioinformatics analysis was performed in JASPAR database and results showed that there were two conservative binding motifs of c-Jun located around −288 bp and −990 bp of DAGLβ promoter. Dual luciferase reporter assay validated that c-Jun promote DAGLβ transcription in RBE cells (Fig. 3D). We also performed CUT&RUN pull-down assay and result showed that c-Jun directly bound to DAGLβ promoter (Fig. 3E). Activator protein-1 (AP-1), a heterodimeric transcription factor, consists of various pair of members from the JUN family and FOS family (22). Among members of the FOS family, FRA1 (also named as FOSL1) recently had been demonstrated to promote tumorigenesis in KRAS-mutant ICC (23). Remarkably, we knocked down FRA1 expression by si-RNA significantly suppressed DAGLβ transcription (Supplemental Fig. S2B); CUT&RUN pull-down assay further validated that FRA1 directly bound to DAGLβ promoter, at the same time predicted binding sites of c-Jun (Supplemental Fig. S2C). Overexpression of c-Jun and FRA1 also significantly promoted DAGLβ expression (Fig. 3F and Supplemental Fig. S2D). Taken together, we demonstrated that AP-1 (c-Jun/FRA1) was bona fide transcription factor of DAGLβ.

Figure 3.

Activator protein-1 (AP-1) enhances diacylglycerol lipase β (DAGLβ) transcription, and lipopolysaccharide (LPS) promotes DAGLβ expression in intrahepatic cholangiocarcinoma (ICC). A: full-length (−1,422 bp to +254 bp) and sequential truncated promoter of DAGLβ were cloned by PCR in RBE cells; clone products were constructed into pGL3-BASIC plasmid luciferase reporter plasmids, respectively. Results of dual luciferase reporter assay revealed that the core promoter region of DAGLβ located in −244 bp to −105 bp from the transcription start site (TSS). B: to identify transcription factors binding to the DAGLβ promoter, protein pull-down assay with desulfurized biotin-labeled DNA probe mapping −1,422 bp to +254 bp of DAGLβ promoter was performed in RBE cells. Proteins pulled down by DNA probe were shown by sliver staining and were further identified by liquid chromatography tandem mass spectrometry (LC/MS). By integration analysis of present DNA pull-down assay and bioinformatics analysis of DAGLβ promoter in ENCODE database and PROMO database, four candidate transcription factors were identified: c-Jun, NFIC, RBPJ, and YY1. C: c-Jun expression was knocked down by si-RNA in RBE cells. qRT-PCR analysis of DAGLβ mRNA expression revealed that knockdown of c-Jun significantly suppressed DAGLβ transcription (n = 5). The amino acid sequences of peptides identified by liquid chromatography tandem mass spectrometry (LC/MS) in B were analyzed on MASCOT search engine in the SwissProt database. Results confirmed that c-Jun was pulled down by DNA probes of DAGLβ promoter. Bioinformatics analysis of JASPAR database showed that there were two binding motifs of c-Jun located at −288 bp and −990 bp of DAGLβ promoter. D: the full-length and mutated promoter sequence of DAGLβ were cloned into pGL3-BASIC plasmid. RBE cells were transient transfected with above plasmid and c-Jun plasmid. Results of dual luciferase reporter assay showed that c-Jun significantly promoted DAGLβ transcription. E: DNA agarose electrophoresis images of CUT&RUN pull-down assay in the present study. Results showed that c-Jun directly bound to two sites (S-288 and S-990) of the DAGLβ promoter. Quantification of c-Jun binding to the DAGLβ promoter was performed by real-time PCR (n = 3). F: mRNA expression and protein expression of DAGLβ were analyzed by qRT-PCR (n = 5) and Western blot (n = 3) in RBE cells cotransfected with c-Jun and FRA1 plasmids. Results showed that AP-1 (c-Jun and FRA1) significantly promoted DAGLβ transcription. G: effect of LPS treatment (for 48 h) on mRNA expression and protein expression of diacylglycerol lipase alpha (DAGLα), DAGLβ, c-Jun, and FRA1 were determined by qRT-PCR (n = 5) and Western blot (n = 3). Results showed that LPS significantly promoted transcription of DAGLβ, c-Jun, and FRA1, whereas there was no significant promoting effect of LPS on DAGLα expression. H: JNK inhibitor SP600125 (10 μM) abrogate promoting effect of LPS on DAGLβ expression. mRNA expression and protein expression of DAGLβ, c-Jun, and FRA1 were determined by qRT-PCR (n = 5) and Western blot (n = 3). I: expression of Notch 2 and Presenilin-2 were significantly upregulated by overexpression of DAGLβ in ICC cell lines. Knockdown of DAGLβ suppressed expression of Notch 2 and Presenilin-2. Protein expression of Notch 2 and Presenilin-2 were determined by Western blot (n = 3). J: proliferation of ICC cell lines were stimulated by LPS treatment (for 48 h), which can be suppressed either by JNK inhibitor SP600125 (10 μM) or DAGLβ specific inhibitor KT109 (10 μM). Cell proliferation was assessed by CCK8 assay (n = 6). Error bars represent means ± SD; *P < 0.05, ***P < 0.001, ****P < 0.0001, according to Student’s t test.

LPS Promotes DAGLβ Expression in ICC Via Activation of JNK Signaling Pathway

Several studies had demonstrated that serum level of 2-AG was elevated in patients suffering from septic shock while the underlying mechanism remained unknown (24–26). Increasing evidence suggested that chronic inflammation increased the risk of ICC (27–30). In the present study, we also demonstrated that DAGLβ immunoreactivity was elevated in patients with ICC accompanied with hepatolithiasis. Therefore, we would like to investigate whether inflammation caused by bacterial infection would lead to elevation of DAGLβ expression. Lipopolysaccharide (LPS) is the major virulence factor of gram-negative bacteria and triggers inflammation mainly via TLR4/MyD88 pathway and downstream JNK/c-Jun pathway (31). We demonstrated that the expression of DAGLβ, c-Jun, and FRA1 was promoted by LPS dose dependently in RBE and HuCCT1 cells (Fig. 3G and Supplemental Fig. S2E). This promoting effect of LPS on DAGLβ expression was also noticed in murine macrophage in the BioGPS database (14) and in human macrophages [peripheral blood mononuclear cell (PBMC)-derived macrophage and THP-1 cell-derived macrophage], and human immortalized normal hepatocyte LO2 cell in the present study (Supplemental Fig. S2G). Expression of Notch 2 and Presenilin-2 were also markedly upregulated by overexpression of DAGLβ in ICC cell lines, as previously we had demonstrated that 2-AG promoted activation of the Notch 2/Presenilin-2 pathway (8). Knockdown of DAGLβ suppressed the expression of Notch 2 and Presenilin-2 (Fig. 3I). Furthermore, proliferation of RBE and HuCCT1 cells was stimulated by LPS treatment (for 48 h), which can be suppressed either by pretreatment of JNK inhibitor SP600125 (10 μM) or DAGLβ specific inhibitor KT109 (10 μM; Fig. 3J; 14). Taken together, these results showed that LPS promoted the expression of DAGLβ through JNK signaling pathway.

DAGLβ Promotes Tumorigenesis and Metastasis of ICC In Vitro and In Vivo

To further determine the pathophysiological role of DAGLβ in ICC development, stable DAGLβ-overexpressed or DAGLβ-knocked down cell lines were established by lentiviral transfection with expression vector or shRNA vector of DAGLβ in RBE and HuCCT1 cells, respectively. Functional studies were performed in vitro and in vivo. Results showed that DAGLβ significantly promoted ICC proliferation either in normal culture condition or in nutrition-deprived condition (glucose-deprived or glutamine-deprived, data not shown) as evaluated by CCK8 assay and soft agar colony formation assay. Either shRNA knockdown or specific inhibitor of DAGLβ significantly suppressed ICC proliferation (Fig. 4, A–C and Supplemental Fig. S3, A–C and F–G). Moreover, DAGLβ promoted migration and invasion of ICC as revealed by wound healing assay and transwell assay; either knockdown or inhibition of DAGLβ significantly abrogate these promoting effects (Fig. 4, D and E and Supplemental Fig. S3, D, E, H, and I).

Figure 4.

Diacylglycerol lipase β (DAGLβ) promotes tumorigenesis and metastasis of intrahepatic cholangiocarcinoma (ICC) in vitro and in vivo. A: stable DAGLβ-overexpressed or DAGLβ-knocked down cell lines were established by lentiviral transfection with expression vector or shRNA vector of DAGLβ in HuCCT1 cells. mRNA expression and protein expression of DAGLβ were determined by qRT-PCR (n = 5) and Western blot (n = 3). B: DAGLβ overexpression significantly promoted HuCCT1 cells proliferation, whereas knockdown of DAGLβ suppressed ICC proliferation in normal culture condition. Cell proliferation was assessed by CCK8 assay (n = 6). C: DAGLβ overexpression significantly enhanced whereas knockdown of DAGLβ suppressed HuCCT1 cells anchorage-independent proliferation, as assessed by soft agar colony formation assay (n = 3). Both colony size and colony number of each group were measured. Representative images of colonies are shown. Scale bars = 100 μm, ×100; scale bars = 50 μm, ×200. D: DAGLβ overexpression significantly promoted migration and invasion of HuCCT1 cells, whereas knockdown of DAGLβ significantly abrogate these promoting effects. Cell migration and invasion were assessed by transwell assay without or with Matrigel matrix (n = 3). Representative images of transwell assay were shown. Scale bars = 100 μm, ×100. E: DAGLβ overexpression significantly promoted migration of HuCCT1 cells, whereas knockdown of DAGLβ significantly suppressed this promoting effect as revealed by wound healing assay (n = 3). Representative images of wound healing assay are shown. Scale bars = 250 μm, ×40. F: firefly luciferase was stably transfected in HuCCT1 cell lines with DAGLβ overexpression or mock lentiviral vector as in A. Subcutaneous xenograft model of ICC was established in NOD/SCID/gamma-null (NSG) mice (n = 5). Bioluminescence imaging was applied to evaluate xenografted tumor development. Photon flux from these mice were evaluated that was injected intraperitoneally with d-luciferin (150 mg/kg), and photon intensities were quantified as units of photons/s/cm²/steradian (n = 5). Total weight and volume of xenografted tumor were also evaluated. Viable xenografted tumors were confirmed by fluorescence microscopy. Protein expression of DAGLβ in xenografted tumors was analyzed by Western blot (n = 3). Representative images of IHC staining of DAGLβ, CK19, Ki67, and PCNA in xenografted tumors are shown. Scale bars = 250 μm, ×40. G: firefly luciferase was stably transfected in HuCCT1 cell lines as in A. Bilateral lung metastasis model of ICC was established in NSG mice via tail veins injection with aforementioned cell lines expressing luciferase (n = 3). Bioluminescence imaging was applied as in F to evaluate lung metastasis. Representative images of gross specimen of lungs fixed with Bouin’s solution and hematoxylin-eosin (H&E) staining of tissue sections are shown. Red arrow indicated micro-metastatic lesions and red dotted line indicated macro-metastatic lesions in lung. Scale bars = 250 μm, ×40; scale bars = 50 μm, ×200. H: firefly luciferase was stably transfected in HuCCT1 cell lines as in A. Intrahepatic metastasis model of ICC was established in NSG mice via spleen injection with aforementioned cell lines expressing luciferase (n = 3). Bioluminescence imaging was applied as in F to evaluate liver metastasis. Representative images of gross specimen of liver and spleen and H&E staining of tissue sections were shown. Red arrow indicated micro-metastatic lesions and red dotted line indicated macro-metastatic lesions in liver. Scale bars = 100 μm, ×100; scale bars = 250 μm, ×40; scale bars = 50 μm, ×200. I: firefly luciferase was stably transfected in HuCCT1 cell lines as in A. Peritoneal metastasis model of ICC was established in NSG mice via peritoneal cavity injection with aforementioned cell lines expressing luciferase (n = 3). Bioluminescence imaging was applied as in F to evaluate peritoneal metastasis. Representative images of gross specimen of peritoneum and abdomen, and HE staining of tissue sections are shown. Red arrow indicated micro-metastatic lesions and dotted line indicated macro-metastatic lesions in peritoneum and abdominal smooth muscle. Scale bars = 50 μm, ×200; scale bars = 25 μm, ×400. Error bars represent means ± SD; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, according to Student’s t test.

Firefly luciferase was stably transfected in aforementioned ICC cell lines in which DAGLβ expression was genetically modulated. We applied in vivo bioluminescence imaging to evaluate the effect of DAGLβ on ICC development. Subcutaneous xenograft model of ICC and in vivo multiorgan metastasis models of ICC (intrahepatic metastasis, lung metastasis, and peritoneal metastasis) were established in NOD/SCID/gamma-null (NSG) mice. Results showed that DAGLβ significantly promoted ICC proliferation and multiorgan metastasis in vivo; whereas knockdown of DAGLβ effectively suppressed ICC proliferation and metastasis (Fig. 4, F–I and Supplemental Fig. S4, A–E). Moreover, the expression of proliferation markers (K_i-67 and PCNA) and prometastatic factors (MMP2 and MMP9; 32) were significantly upregulated in DAGLβ-overexpressed ICC xenografts and ICC cell lines; Knockdown of DAGLβ suppressed the expression of aforementioned markers (Fig. 4F and Supplemental Figs. S4A and S5). Taken together, DAGLβ significantly promotes tumorigenesis and metastasis of ICC in vitro and in vivo.

AP-1/DAGLβ/miRNA-4516 Foster a Feedforward Circuitry to Promote ICC Development

microRNAs (abbreviated as miRNAs) were endogenous molecules like endocannabinoids, had been demonstrated as important regulators and biomarkers of various cancers (33, 34). Therefore, we would like to explore miRNAs and the downstream signaling pathway that can be regulated by 2-AG and DAGLβ. RBE cells were incubated with 2-AG (10 μM, refreshed every 48 h) for 2 wk, mimicking the chronic inflammatory microenvironment caused by a bacterial infection (24–26). Differentially expressed miRNAs were quantified by microarray (Exiqon miRCURY LNA miRNA array) and hierarchical clustering of miRNAs was performed (access to GSE188863). Our results showed that 48 miRNAs were significantly differentially expressed after 2-AG incubation (44 miRNAs upregulated and 4 miRNAs downregulated; Fig. 5A). Among four miRNAs downregulated by 2-AG, expression of miR-4516 was the lowest and reduced by ∼84%. Consistently, miR-4516 expression can be suppressed by DAGLβ overexpression or LPS incubation in RBE or HuCCT1 cells; whereas miR-4516 expression can be enhanced by knockdown of DAGLβ (Fig. 5, B and C and Supplemental Fig. S6, A and B).

Figure 5.

FRA1/diacylglycerol lipase β (DAGLβ)/miRNA-4516 forms a positive feedforward circuitry to promote intrahepatic cholangiocarcinoma (ICC) development. A: total RNAs were extracted from RBE cells treated with 2-arachidonoyl glycerol (2-AG; 10 μM for 2 wk, n = 3) or control (10 μM DMSO for 2 wk, n = 3) and total microRNAs were isolated for microarray analysis. The threshold of differently expressed microRNAs was |log2 (fold change)| > 1 and P < 0.05. Quantification and graphical visualization of differentially expressed miRNAs were performed by heat map and volcano plot analysis. Results showed that among four miRNAs downregulated by 2-AG, expression of miR-4516 was the lowest and reduced by about 84%. B: stable DAGLβ-overexpressed or DAGLβ-knocked down cell lines were established as in Fig. 4A. miR-4516 expression in each group was determined by qRT-PCR (n = 5). C: HuCCT1 cells were treated with the indicated doses of lipopolysaccharide (LPS; for 24 h). miR-4516 expression in each group was determined by qRT-PCR (n = 5). D: HuCCT1 cells were transfected with miR-4516 mimic or inhibitor. mRNA expression and protein expression of DAGLβ, Stat3, and FRA1 of each group were determined by qRT-PCR (n = 5) and Western blot (n = 3). E: HuCCT1 cells were transfected with miR-4516 mimic or control. Cell proliferation of each group was assessed by CCK8 assay (n = 6). F: HuCCT1 cells were transfected with miR-4516 mimic or control. Cell migration of each group was assessed by wound healing assay (n = 3). Scale bars = 250 μm, ×40. G: HuCCT1 cells were transfected with miR-4516 mimic or control. Cell migration and invasion of each group were assessed by transwell assay without or with Matrigel matrix (n = 3). Scale bars = 100 μm, ×100. H: mRNA expression and protein expression of STAT3, c-Jun, FRA1, MMP9, MMP2, and DAGLβ in clinical ICC samples (n = 6) were determined by qRT-PCR and Western blot. miR-4516 expression in each sample was determined by qRT-PCR. Correlation analysis between DAGLβ and each of aforementioned genes and miR-4516 was performed. I: scheme showing the proposed mechanism of transcriptional regulation of DAGLβ in ICC. Activator protein-1 (AP-1; heterodimers of c-Jun and FRA1) directly bound to DAGLβ promoter and promoted transcription of DAGLβ, which could be enhanced by lipopolysaccharide (LPS) via JNK signaling pathway. FRA1 and STAT3 were targets of miRNA-4516 that could be suppressed by DAGLβ. Suppression of miRNA-4516 by DAGLβ would derepress FRA1 expression, which subsequently constituted an AP-1/DAGLβ/miR4516 feedforward circuitry in ICC (Created with BioRender.com). Error bars represent means ± SD; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, according to Student’s t test.

miRNAs had been demonstrated to function in RNA silencing and posttranscriptional regulation of gene expression (33, 34). We further determined the pathophysiological effect of miR-4516 and its downstream target in ICC, by literature review and in silico miRNA target databases analysis (TargetScan, http://www.targetscan.org; miRDB, http://www.mirdb.org). FRA1 and STAT3 were identified as putative targets of miR-4516 as 3′ untranslated region (UTR) of FRA1 mRNA and 3′UTR of STAT3 mRNA contained miR-4516 recognition sequence (35, 36). We pharmacologically modulated miR-4516 expression in RBE and HuCCT1 cells by transfection of miR-4516 mimics or inhibitor. Results showed that miR-4516 mimics suppressed mRNA expression of FRA1 (0.24 ± 0.03, P < 0.01), STAT3 (0.17 ± 0.05, P = 0.03), and DAGLβ (0.20 ± 0.05, P = 0.01), whereas miR-4516 inhibitor enhanced the expression of aforementioned targets (Fig. 5D and Supplemental Fig. S6C); miR-4516 mimics significantly suppressed ICC cells proliferation, migration, and invasion in vitro (Fig. 5, E–G, and Supplemental Fig. S6, D–F). Consistent with previous results showing that miR-4516 expression can be suppressed by DAGLβ overexpression or LPS incubation in ICC cells, we further demonstrated that expressions of FRA1 and STAT3, as well as c-Jun were significantly elevated by DAGLβ overexpression; whereas expressions of aforementioned genes can be suppressed by DAGLβ knockdown (Supplemental Fig. S6G). Moreover, in clinical ICC samples, we also demonstrated that there were positive correlations between expressions of DAGLβ and aforementioned targets of miR-4516 (FRA1 and STAT3), as well as c-Jun, MMP2, and MMP9; whereas there was a negative correlation between expressions of DAGLβ and miR-4516 (Fig. 5H). Taken together, we showed that AP-1/DAGLβ/miR4516 forms a feedforward circuitry to promote ICC development.

DISCUSSION

The present study demonstrated that 2-AG was enriched in ICC, and DAGLβ was the principal synthesizing enzyme of 2-AG in ICC. DAGLβ was significantly upregulated in ICC and was closely related with advanced stage and poor prognosis of ICC. Moreover, DAGLβ promoted ICC proliferation, migration, and invasion in vitro and in vivo. Mechanically, we showed that AP-1 (c-Jun/FRA1) regulated DAGLβ transcription, which could be significantly augmented by LPS; FRA1 and STAT3 were targets of miRNA-4516, which could be suppressed by 2-AG incubation or DAGLβ overexpression, indicating there was an AP-1/DAGLβ/miR-4516 positive feedforward circuitry in ICC.

Dysregulated ECS had been demonstrated in various liver diseases including NAFLD, liver fibrosis, and hepatocellular carcinoma (9). Previously, we had showed that 2-AG exerted protumorigenic effect via CB receptors-independent, lipid raft-dependent activation of the Notch 2/Presenilin-2 signaling pathway (8, 10). Furthermore, in the present study, we showed that the expression level of 2-AG was elevated in ICC. As endocannabinoids (AEA and 2-AG) are short-lived endogenous substance (less than 15 min) that are generated on demand by elevated intracellular calcium or metabotropic receptor activation, our results indicate that there are persistent stimulators in the desmoplastic and hypovascularized tumor microenvironment of ICC. Chronic inflammation accompanied with risk factors of ICC (hepatolithiasis, Caroli disease, liver fluke, etc.) may exert stimulating effects on 2-AG biosynthesis. In accordance with previous clinical studies that serum level of 2-AG was elevated in patients with septic shock (24–26), our results further demonstrated that LPS, the major virulence factor of sepsis, significantly promoted the expression of DAGLβ via JNK signaling pathway in a time- and dose-dependent manner in ICC. Either inhibition of DAGLβ or JNK significantly suppressed ICC tumorigenesis and invasiveness. Collectively, these results underlined a potential therapeutic vulnerability to inhibition of the DAGLβ/2-AG pathway in ICC.

2-AG is the major EC in the body and is the hydrolytic production of arachidonic acid-esterified DAGs at the sn-1 position by DAGLα or DAGLβ. Production of 2-AG by DAGLs exhibits cell type and tissue specificity that DAGLα enriches in synapse-rich regions of central nervous systems, whereas DAGLβ enriches in mononuclear phagocytic system and peripheral tissues where it involves in regulating inflammatory responses (12). After synthesis, 2-AG can be further degraded by monoglyceride lipase (MAGL), as well as cyclooxygenase-2 (COX-2) and lipoxygenase (LOX) isozymes to generate free fatty acid, arachidonic acid, eicosanoids, and prostanoids (14, 37). MAGL, COX-2, and the oxidative products of 2-AG (arachidonic acid and prostaglandin E2) have been demonstrated as tumorigenic stimulators (8). In DAGLβ genetic knockout mice model, there was dramatic reductions of 2-AG (90%) and arachidonic acid (70%) in the liver (13, 14). Consistent with result of animal model, we demonstrated that knockdown of DAGLβ significantly reduced 2-AG expression by ∼75% in ICC cells incubated with 2-AG precursor SAG, indicating DAGLβ was the principal synthesizing enzyme of 2-AG in ICC.

DAGLβ had been shown to play an important role in promoting inflammation and regulating endogenous lipid network both in central nervous system and peripheral tissues by biosynthesis of inflammatory mediators (14, 37–41). However, regulation and function of DAGLβ in cancer development remained elusive. Recently, a study had reported that estrogen-related receptor γ (ERRγ) regulated DAGLα, DAGLβ, and transferrin receptor 2 (TFR2) transcription in hepatocyte and served as a therapeutic target of alcoholic liver disease in mice (42, 43). Moreover, in the present study, we provided molecular mechanisms of DAGLβ transcription in ICC. We demonstrated that there were two c-Jun binding motifs in DAGLβ promoter including one core region (chr7:6,448,256–6,448,117) closed to the TSS. We also demonstrated that heterodimeric transcription factor AP-1 (c-Jun/FRA1) directly bound to these sites and regulated DAGLβ transcription in response to LPS. These AP-1 binding sites contained dense H3K27Ac mark (often found near enhancer or highly accessible chromatin regions) that was recorded in ENCODE database (not shown in the present study) reflecting active transcriptional activities in these regions. AP-1 is a collective term referring to heterodimeric transcription factors consisting of Jun, Fos, or ATF (activating transcription factor) subunits. AP-1 has been demonstrated to be involved in cellular proliferation, transformation, and death (22). In recent comprehensive studies of ICC, both c-Jun and FRA1 had been shown to be crucial oncogenic factors in KRAS-mutant ICC (23, 44). Consistently, in the present study, we also demonstrated that AP-1 (c-Jun/FRA1) upregulated DAGLβ transcription and promoted tumorigenesis and metastasis of KRAS-mutant ICC cell lines (RBE and HuCCT1) in vitro and in vivo.miRNAs are a type of endogenous noncoding RNAs (21–24 nucleotides in length) that function in RNA silencing or posttranscriptional regulation of gene expression.

miRNAs and their regulatory circuitries are found to be conservative throughout phylogeny (45, 46). Moreover, we and others have found that miRNAs function as important regulators and biomarkers of various human liver diseases including ICC (19, 33, 34, 47–49). In the present study, we demonstrated that 2-AG incubation or DAGLβ suppressed miR-4516 expression in ICC cells; a negative correlation between expressions of DAGLβ and miR-4516 also had been shown in clinical samples. Downregulation of miR-4516 decompressed the expression of its targets FRA1 and STAT3, which had been demonstrated as oncogenes in ICC (23, 35, 36, 50, 51). It should be noted that miR-4516 could exert different effect on cancer development. For instance, miR-4516 suppressed pancreatic cancer development via downregulating orthodenticle homeobox 1 (52); miR-4516 decreases proliferation and invasion of breast cancer by antagonizing oncogenic LncRNA PART1 (53); moreover, miR-4516 could suppress hepatocellular carcinoma by antagonizing oncogenic LncRNA LSINCT5 (54). Conversely, miR-4516 significantly promoted glioblastoma (GBM) by directly targeting protein tyrosine phosphatase, nonreceptor type 14 (PTPN14) to activate the Hippo signaling pathway (55). Nonetheless, in the present study, we demonstrated that miR-4516 exerted tumor suppressive effect on ICC. We showed that miR-4516 mimics suppressed the expression of FRA1, STAT3, and DAGLβ in ICC cells; moreover, miR-4516 mimics significantly suppressed ICC proliferation, migration, and invasion in vitro. Taken together, these results suggested the promising therapeutic potential of miR-4516 or targeting the AP-1/DAGLβ/miR-4516 positive feedforward circuitry in the treatment of ICC. The underlining regulatory mechanism of miR-4516 in ICC, including competing endogenous RNA (ceRNA) network, is being investigated by our group.

In summary, we demonstrate DAGLβ as the principal synthesizing enzyme of 2-AG in ICC and promotes tumorigenesis and metastasis via a novel AP-1/DAGLβ/miR-4516 feedforward circuitry. These results support DAGLβ as a reasonable therapeutic target and prognostic biomarker for ICC.

DATA AVAILABILITY

Data will be made available upon reasonable request.

SUPPLEMENTAL DATA

Supplemental Figs. S1–S6: https://doi.org/10.6084/m9.figshare.22852946.

GRANTS

This work was supported by the National Natural Science Foundation of China under grants 81201919 and U1813204 (to L. Huang), and the Natural Science Foundation of Guangdong Province Grant 2017A030313495 (to L. Huang).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

M.M., J.L., and L.H. conceived and designed research; M.M., G.Z., B.T., G.Z., W.Z., Y.S., J.L., B.X., Z.W., J.C., M.H., Y.H., Y.L., D.C., and L.H. performed experiments; M.M., G.Z., B.T., G.Z., Q.S., W.Z., Y.S., J.L., B.X., Z.W., J.C., M.H., C.Y., J.Y., Y.H., Y.L., D.C., Y.H., S.D., L.L., J.L., and L.H. analyzed data; M.M., G.Z., B.T., G.Z., Q.S., W.Z., Y.S., J.L., B.X., Z.W., J.C., M.H., C.Y., J.Y., Y.H., Y.L., D.C., Y.H., S.D., L.L., J.L., and L.H. interpreted results of experiments; M.M., G.Z., B.T., G.Z., Y.S., J.Y., Y.H., Y.L., D.C., J.L., and L.H. prepared figures; M.M., G.Z., B.T., G.Z., Q.S., C.Y., J.Y., Y.H., D.C., Y.H., S.D., L.L., J.L., and L.H. drafted manuscript; M.M., M.H., C.Y., J.Y., D.C., Y.H., S.D., L.L., J.L., and L.H. edited and revised manuscript; M.M. and L.H. approved final version of manuscript.