Deubiquitination of DNM1L by USP3 triggers the development and metastasis of gallbladder carcinoma

Abstract

Background

Patients diagnosed with gallbladder carcinoma (GBC) accompanied by hepatic metastasis exhibit unfavorable prognoses generally. Mitochondrial dysfunction promotes cellular transformation and cancer cell survival implicating its importance in cancer development. Previous studies have indicated that dynamin 1 like (DNM1L) is a key mediator of mitochondrial fission. However, whether DNM1L regulates mitochondrial homeostasis in GBC remains unknown.

Methods

The morphological changes of mitochondria were investigated by transmission electron microscopy and mitoTracker red staining. Co-immunoprecipitation assay was performed to detect the interaction of ubiquitin-specific protease-3 (USP3) and DNM1L. The cell-derived xenograft and liver metastasis tumor models were established to validate the function of DNM1L in vivo. The metabolomics data from transcriptomics/metabolomics were analyzed to identify the differentially expressed genes/metabolites of DNM1L in GBC.

Results

DNM1L exhibited a marked upregulation in clinical GBC tissues compared to the adjacent tissues, and it promoted proliferation, invasiveness, and migration capability of GBC cells by inducing mitochondrial dysfunction. Mice subcutaneously injected with DNM1L overexpression cells exhibited elevated intrahepatic metastatic nodules within their livers. USP3, a deubiquitinating enzyme, was demonstrated to directly interact with DNM1L and it specifically cleaved the K48-linked polyubiquitin chains to deubiquitinate and stabilize DNM1L. By integrating two omics, we found several altered pathways and speculated that DNM1L disturbed DNA synthesis and glycine, serine, threonine, and pyrimidine metabolism pathways.

Conclusion

Our findings suggest that DNM1L is a promising clinical target for GBC treatment and that focusing on DNM1L may provide new insights into GBC strategy.

Introduction

Gallbladder carcinoma (GBC) is the most common and aggressive tumor of the biliary system, and chronic infection of the gallbladder and/or environmental exposure to specific chemicals are common risk factors [1]. GBC is characterized by lack of symptoms at the initial stage leading to late diagnosis, accompanied by the invasion to the liver because of the close anatomical location [2]. Broadly, the poor prognosis of overall 5-year survival is 50% for stage I cancers and 3% for stage IV cancers [3, 4]. Elucidating potential therapeutic targets can provide implications for therapeutic strategies to inhibit rapid proliferation and hepatic metastasis behaviors.

Dynamin 1 like (DNM1L), also known as dynamin-related protein 1 (DRP1) or dynamin-like protein 1 (DLP1), primarily plays a crucial role in the regulation of fission processes in both mitochondria and peroxisomes [5]. It belongs to the dynamin family of GTP-binding proteins, which contains four domains: an N-terminal GTPase domain, a middle domain, a variable domain, and a C-terminal GTPase effector domain [6] Prior studies indicate that the dysregulated expression and gene variation of DNM1L cause Alzheimer's disease, neonatal-onset encephalopathy, and rheumatoid arthritis [7,8,9]. The carcinogenic properties of DNM1L have also been reported in maintaining cancer cell stemness or chemotherapy resistance [10, 11]. Srinivasan et al. suggested that dysfunctional mitochondria induced retrograde signaling from mitochondria to the nucleus, bringing global changes in nuclear gene expression and phenotypic changes in cells [12]. Mitochondrial retrograde signaling has received considerable attention in recent years, and whether it is also involved in DNM1L-induced phenotypic changes of GBC cells is unknown. We hypothesized that we may find some evidence from a joint analysis of the transcriptome and metabolome.

As a dynamin-related cytosolic protein, the stability of the DNM1L protein is directly related to its function [13, 14]. Our previous research has indicated that deubiquitinase ubiquitin-specific protease-3 (USP3) was highly expressed in GBC tissues and it played a pivotal role in the progression and metabolism of GBC [15]. USP3 is a key deubiquitylation enzyme implicated in various human diseases, therefore, the intricate and context-dependent regulatory networks associated with USP3 still need further investigation [16]. Existing studies have shown that USP3 promotes cancer progression and metastasis through its deubiquitination function [17]. Ubibrowser, a computational predictive system, was used to investigate the human E3 ubiquitin-protein ligase-substrate interaction network. It predicted that USP3 possibly deubiquitinated DNM1L through the ubiquitylation sites for substrate protein. However, to date, it is unclear whether USP3 affects the protein stability of DNM1L.

In this study, we showed that upregulated DNM1L induced mitochondrial dysfunction, mitochondrial over-fission, and mtDNA stress of GBC cells, and the manipulation of DNM1L significantly promoted the growth and liver metastasis of GBC. Mechanistic exploration revealed that DNM1L expression is upregulated through USP3-mediated deubiquitylation. Omics analysis shows that DNM1L may disturb DNA synthesis and amino acid metabolism. Therefore, the regulation and biological function of DNM1L make it a candidate molecular target for treating GBC.

Materials and methods

GBC patients

GBC tissues (n = 8) and para-carcinoma tissues were collected from GBC patients during surgery, and the patients consented to participate in this study and provided informed consent. The samples were collected from The First Affiliated Hospital of Zhengzhou University during July 2019 to the end of December 2021. This study was approved by the Ethics Committee of The First Affiliated Hospital of Zhengzhou University, and all procedures were performed following the Declaration of Helsinki.

Cell culture, transfection, and treatment

The human GBC cell lines, including GBC-SD and NOZ cells, were purchased from the iCell Bioscience (Shanghai, China). The HEK293T cells were bought from Zhong Qiao Xin Zhou Biotechnology (Shanghai, China). GBC-SD cells were cultivated in Roswell Park Memorial Institute (RPMI) 1640 medium (Servicebio, Beijing, China) supplemented with 10% fetal bovine serum (FBS, Tianhang biotechnology, Huzhou, China). NOZ and HEK293T cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) (Servicebio, Beijing, China) supplemented with 10% FBS. All cells were maintained in a humidified incubator under a 5% CO₂ atmosphere at 37 °C.

In the current study, a doxycycline (Dox)-inducible lentiviral vector system was employed to induce the expression of DNM1L messenger RNA (mRNA) or short hairpin RNA (shRNA). The full-length coding sequences (CDS) of DNM1L were inserted into the pLVX-TetOne-Puro vector, and the DNM1L-specific shRNA (shDNM1L) or a non-silencing shRNA (shNC) were constructed into the Tet-pLKO-puro vector. GBC-SD and NOZ cells were infected with the packaged virus, after which the cells were treated with 200 μg/mL puromycin (Thermo Fisher Scientific, Massachusetts, USA) to construct stable cell lines. After that, GBC-SD and NOZ cells were treated for 48 h with 1 µg/mL Dox (Macklin, Shanghai, China) to induce the overexpression or knockdown of DNM1L. Cells without Dox treatment were used as negative controls.

The CDS of USP3 was inserted into the pCDNA3.1 vector to overexpress USP3 (USP3^OE), and its empty vector was used as a control. NOZ cells were then transfected with the recombinant constructs via Lipofectamine 3000 (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s instructions. For knockdown experiments, NOZ cells were transfected with USP3 siRNA (USP3^KD) or negative control siRNA (siNC). The siRNA transfections were performed by using Lipofectamine 3000. Forty-eight hours after the transfection, NOZ cells were treated with 20 μg/mL protein synthesis inhibitor cycloheximide (CHX, Aladdin, Shanghai, China) for 0, 3, 6, and 9 h, respectively, and the cell lysates were then subjected to western blot analysis. The information about shRNA/siRNA was shown in Table S1.

Cell growth, cycle, and apoptosis assays

CCK-8 assay was performed 48 h after Dox treatment to assess the cell viability by the CCK-8 cell counting kit (KeyGEN Biotech, Nanjing, China) according to the manufacturer’s instructions. Absorbance at a wavelength of 450 nm was detected with a microplate reader (BioTek Instruments, Winooski, VT, USA). Each assay was conducted in triplicate, with a total of three repetitions for each.

For cell cycle analysis, GBC-SD and NOZ cells were collected and fixed with 70% ethanol, followed by stained with propidium iodide (PI) at 37 °C for 30 min in the dark by Cell Cycle Analysis Kit (KeyGEN Biotech) according to the manufacturer's protocols. After incubation, the samples were analyzed by a flow cytometer (Agilent, Santa Clara, CA, USA).

An apoptosis assay kit (KeyGEN Biotech) was used to detect cell apoptosis in GBC-SD and NOZ cells. In brief, the cells were collected by centrifugation at 150 g for 5 min. After aspirating the supernatant, they were washed twice with phosphate-buffered solution (PBS)and resuspended by the addition of 500 μL of Binding Buffer. Annexin V-FITC (5 µL) and PI staining (5 µL) were added to the cells for 15 min in the dark. The mixed samples were detected by using a flow cytometer (Agilent).

EDU staining

EDU assay was conducted to visualize the proliferating cells. In brief, GBC cells were subjected to treatment with EDU at a concentration of 10 μM for 2 h prior to fixation. They were then permeabilized with 0.5% TritonX-100 for 20 min at room temperature. Subsequently, the cells were stained using the Click-iT Plus EDU Imaging Kit (KeyGEN Biotech) in accordance with the manufacturer's instructions. Cells with fluorescence were observed under a microscope (Olympus, Tokyo, Japan) at 400 × magnification.

Migration and invasion assays

Cell migration and invasion assays were performed using 8-μm transwell filters in 24-well plates (LABSELECT, Anhui, China). GBC-SD and NOZ cells (5 × 10³ cells for migration assays; 2 × 10⁴ cells for invasion assays) were plated in the upper chamber, whose membrane was coated with (invasion assays) or without (migration assays) Matrigel (Corning, New York, USA). After 24 h-incubation, cells on the membrane of the chambers were fixed with 4% polyformaldehyde (Aladdin) at room temperature for 20 min and stained with crystal violet (Amresco, Solon, OH, USA) for 5 min. The migration and invasion of cells were photographed under a microscope (Olympus) at 200 × magnification, after which the cells were counted in five randomly selected fields in each well.

Detection of MMP2 and ATP levels

GBC-SD and NOZ cells (1 × 10⁶ cells) were centrifuged at 300 g for 10 min with subsequent removal of pellets. The total MMP2 concentration in the supernatant was determined using the Human MMP-2 ELISA Kits (LIANKE Biotech, Hangzhou, China) according to the manufacturer’s protocols. Absorbance measurements were conducted at a wavelength of 450 nm, subtracting the values measured at 570 nm, utilizing a microplate reader (BioTek Instruments).

ATP level was determined using the ATP Assay Kits (Beyotime Biotechnology, Shanghai, China). Briefly, 1 × 10⁶ cells were lysed in 200 μL of lysis buffer and centrifuged at 12,000 g for 5 min to collect the supernatant. The prepared ATP assay solution according to the manufacturer’s instructions was added to each sample. Finally, the ATP concentration was calculated based on the generated ATP standard curve, and the ATP level was expressed as (nmol/mg prot).

Detection of reactive oxygen species (ROS)

Cellular ROS levels were detected by the fluorescent probe DCFH-DA in Reactive Oxygen Species Assay Kits (Biosharp, Hefei, China). In brief, cell suspension with the concentration of 2 × 10⁴ cells was seeded in six-well plates. The cells were collected after the centrifugation at 150 g for 5 min and the cells were then incubated with 1 mL DCFH-DA for 30 min at 37℃. After incubation, the cells were washed with PBS and then resuspended in 500 µL PBS. The detection of ROS was performed using a flow cytometer (Agilent).

Mitochondrial observation

Transmission electron microscopy and Mitotracker red staining were performed to observe the morphological changes of mitochondria, respectively. The cell sections were stained with uranyl acetate and lead citrate and analyzed with an H-7650 electron microscope (Hitachi, Tokyo, Japan).

Mitotracker Red is affected by the inner mitochondrial membrane proton gradient when it enters the mitochondria. In this study, GBC cells were stained with MitoTracker Red (Myaokang Biotechnology, Shanghai, China) according to the manufacturer’s instructions to explore the distribution and activity of mitochondria. The state and activity of mitochondria were visualized using a microscope (Olympus) at 400 × magnification.

Determination of mitochondrial DNA (mtDNA) copies

Cytosol fractions of cells were isolated using the cell mitochondria isolation kit (Abcam, Cambridge, UK) following the manufacturer’s instructions. The mtDNA copy number was measured by determining the ratio of mitochondrial encoded NADH dehydrogenase subunit 1 (MT-ND1) to the nuclear control gene beta-actin (ACTB) using a quantitative real-time polymerase chain reaction (qRT-PCR) assay.

Nude mouse xenograft experiments

The animal experiments were approved by the Ethics Committee of The First Affiliated Hospital of Zhengzhou University. The 4-week-old male BALB/c nude mice were obtained from Huachuang Xinnuo Pharmaceutical Technology (Taizhou, China) and housed in standard conditions. Twenty-eight nude mice were randomly divided into two groups (pLVX-TetOne-DNM1L and Tet-shDNM1L). To construct a mouse xenograft model, mice were subcutaneously injected into 1 × 10⁶ NOZ cells transfected with pLVX-TetOne-DNM1L and Tet-shDNM1L, respectively, at the right axillary region. The subcutaneous xenograft tumors were visualized one week after injection. Each group was subsequently subdivided into two subgroups with five mice per subgroup. The mice in experimental condition were administered drinking water containing 5% sucrose and 2 mg/mL of Doxorubicin (Dox), while the mice receiving sucrose drinking water without Dox served as the control group. During the observation period, the length and width of the tumor were measured and recorded every week for 28 days. Tumor volumes were calculated according to the formula: (length × width²)/2. TUNEL assay was performed with xenograft tumor tissue for detection of cell apoptosis using the In Situ Cell Death Detection Kits (Roche, Basel, Switzerland).

Immunohistochemical (IHC) staining

IHC staining was conducted on tumor tissue sections collected from nude mice to investigate the histological characteristics. The tumor sections underwent deparaffinization using xylene, followed by rehydration in a series of graded ethanol solutions. To inhibit endogenous peroxidase activity, the sections were incubated with 3% hydrogen peroxide (Sinopharm, Shanghai, China) for 15 min at room temperature. Subsequently, the sections were incubated with Ki67 antibody (1:50 dilution) (Abcam) and HRP Goat Anti-Rabbit IgG (1:100 dilution, Solarbio, Beijing, China) overnight at 4 °C. After counterstaining with hematoxylin (Solarbio), the sections were dehydrated and mounted. Images were captured at a magnification of 400 × .

In vivo metastasis assays

We also constructed an intrasplenic injection model for in vivo metastasis assays. NOZ cells transfected with pHIV-Luc-ZsGreen vectors were injected into the spleens of mice. The spleens were oppressed for 2 min following splenectomy. The tumor metastasis was visualized at 28 days post-injection by in vivo bioluminescence imaging using IVIS Spectrum Imaging System (Perkin Elmer, Massachusetts, USA) after intraperitoneally injecting 4.0 mg D-Luciferin for 10 min. The liver tissues of nude mice were collected, and the number of tumor nodules was counted. Haematoxylin and eosin staining (HE) was performed in paraffin-embedded liver tissue sections. Briefly, the sections were stained with hematoxylin (Solarbio) for 5 min and eosin (Sangon, China) for 3 min. Sections were observed with a microscopy (Olympus) at a magnification of 200 × .

RNA isolation and qRT-PCR

Total RNA was extracted from clinical samples and GBC cells using the TRIpure extraction kits (BioTeke, Beijing, China), and the quality of extracted RNA was measured using NanoDrop 2000 (Thermo Fisher Scientific, Massachusetts, USA). Following reverse transcription utilizing BeyoRT II M-MLV reverse transcriptase (Beyotime Biotechnology), qRT-PCR analysis was conducted using a real-time detection system employing SYBR Green for detection. The relative gene expression levels were normalized to β-actin and analyzed quantitatively using the 2^−ΔΔCt method. The primers used were listed as below: DNM1L-forward: 5'-GGAAATAATAAGGGAGTAAG-3'; DNM1L-reverse: 5'-TGATGTTGCCATATCTGTA-3'; USP3-forward: 5'-TTAGCCCAGAGTCCTTATT-3'; USP3-reverse: 5'-AGCCGTGACAACAGTAGAT-3'.

Western blot

Total protein was extracted using lysis buffer containing protease and phosphatase inhibitors (Beyotime Biotechnology), respectively. Protein samples were separated on SDS-PAGE gel with different concentrations and transferred to polyvinylidene fluoride membranes (Abcam). After blocking with 5% nonfat milk, those membranes were incubated with β-actin (1:1000 dilution; Santa Cruz, California, USA), DNM1L (1:400 dilution; Santa Cruz), USP3 (1:1000 dilution; Proteintech, Wuhan, China), Cyclin E (1:2000 dilution; Proteintech), p21 (1:3000 dilution; Proteintech), cleaved PARP antibody (1:1000 dilution; Cell signaling technology, Boston, USA), and cleaved caspase 3 antibody (1:1000 dilution; Cell signaling technology, Boston, USA). The membranes were incubated with the second antibody (1:5000 dilution; Beyotime Biotechnology) at room temperature for 1 h. Protein bands were visualized using ECL chemiluminescence kits (Beyotime Biotechnology). Western blots with independent biological replicates were repeated at least three times.

Co-immunoprecipitation (Co-IP) and ubiquitination assay

Co-IP assay was performed to detect the combination of endogenous USP3 and DNM1L via using the Co-IP kit (Pierce Biotechnology, Rockford, USA) according to the manufacturer’s instructions. In brief, DNM1L antibody (Santa Cruz) was immobilized on AminoLink plus coupling resin to generate the antibody-crosslinked resin complexes. IgG was used as negative control. Cell lysates were incubated with the resin in which the antibody had been immobilized for 2 h at the room temperature. Cells lysed with the lysis buffer were served as positive control. The immunoprecipitated samples was eluted using an elution buffer and subsequently analyzed via western blot assay.

To verify the domains of DNM1L binds to USP3, various HA-tagged DNM1L deletion mutants (including HA-1–302 aa, HA-303–643 aa, and HA-644–736 aa) or HA-full constructs (HA-1–736 aa) and Flag-USP3 constructs were generated in the pcDNA3.1 vector. HEK293T cells were transiently co-transfected with HA-DNM1L plasmid and Flag-USP3 construct for 48 h. Co-IP assay was performed in transfected HEK293T cells to determine the binding of USP3 and DNM1L.

For the ubiquitination assays, the transfected NOZ cells were treated with the proteasomal inhibitor MG132 (MedChemExpress, Monmouth Junction, NJ, USA) for 6 h. After that, Co-IP assay was performed to detect the ubiquitinated level of DNM1L, and the immunoprecipitated proteins were then subjected to western blot using DNM1L antibody (Santa Cruz), anti-ubiquitin antibody (Abcam), and anti-ubiquitin K48 antibody (Abcam).

Immunofluorescence staining

GBC-SD and NOZ cells were fixed and then permeabilized with Triton X-100 (Beyotime Biotechnology). Following blocking, the cells were incubated with primary antibodies against DNM1L (1:50 dilution; Proteintech) and USP3 (1:50 dilution; Proteintech) at 4 °C overnight. The proteins were visualized by incubation with secondary antibody (Cell Signaling Technology) for 1 h at room temperature, and the nuclei were stained with DAPI (Aladdin). Imaging was performed using confocal microscopy (Nikon, Tokyo, Japan).

Transcriptomics and metabolomics detection

Transcriptomics and metabolomics analysis were performed by Suzhou Panomike Biomedical Technology (Suzhou, China). For global transcriptome analysis, RNA was extracted using standard methods and cDNA libraries were subjected to high-throughput sequencing using the Illumina Sequencing platform. The libraries were then aligned to the GRCh38 reference genome (Ensembl database) using the HISAT2 (v2.2.1). HTSeq software (v2.0.2) was used to calculate read counts per transcript. DEseq2 software (v1.28.0) was performed to identify differently expressed genes (DEGs), and p value < 0.05 and |log2 (foldchange)|> 1 was set as the threshold for significantly differential expression. For metabolome analysis, metabolites were extracted by addition of 1 mL of acetonitrile/methanol/H₂O (2:2:1 v/v/v). Next, the metabolites were analyzed by using a Thermo Q Exactive mass spectrometer coupled to a Thermo Vanquish ultra-high performance liquid chromatography system. The Proteowizard4 software (v3.0.8789) was used to convert RAW files to.mzXML files, and XCMS software (v3.12.0) was used for peak alignment, peak filtering, and peak filling. Metabolites were then identified by using public databases, including human metabolome database (HMDB), MassBank, LipidMaps, mzcloud, and KEGG. Orthogonal partial least squares-discriminant analysis (OPLS-DA) was constructed using R “ropls” package (v1.30.0), and differential expressed metabolites (DEMs) were identified using a significance level of p < 0.05 and the variable importance in the projection (VIP) > 1. The principal component analysis (PCA) and orthogonal partial least squares-discriminant analysis (OPLS-DA) were applied to reflect the inter- and intra-group variations.

Personalized transcriptomic analyses

R package plug-in clusterProfiler was used to implement Gene Ontology (GO) annotations and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways analysis for DEGs. Protein–protein interaction (PPI) network analysis for DEGs was performed using the STRING database. Cytoscape (version 3.9.1) was employed to construct a correlation network. Gene set enrichment analysis (GSEA) was conducted by the MAVTgsa package of R [18].

Personalized metabolomics analysis

The heatmap showed DEMs expression in each sample and their belonging to 15 classes as indicated. Functional pathways in which DEMs participate were identified and annotated on KEGG and Reactome databases. Furthermore, pathways with no less than 2 detected metabolites were used for differential abundance (DA) score calculation. The DA scores of each pathway were calculated based on the formula: DA score = (Metabolites Increased-Metabolites Decreased)/Measured Metabolites in Pathway, which referred to Hakimi’s research [19]. The DA score varies from − 1 to 1. A score of 1 indicates that all metabolites in a pathway increased in abundance, while a score of − 1 indicates that all metabolites decreased.

Comprehensive analysis of transcriptomics and metabolomics

To collect more biological information, we conducted a combined transcriptome and metabolome analysis from four different perspectives. (1) Pearson correlation analysis was used to analyze the correlations between DEGs and DEMs, in which the results that met correlation coefficients > 0.5 and p < 0.05 were entered into consideration. (2) By conducting KEGG enrichment analysis on both metabolome and transcriptome, we identified common pathways for subsequent investigation. The DEGs and DEMs with correlations in each pathway were used to construct the co-expression network, a process executed using Cytoscape software. (3) We performed a 2-way Orthogonal Partial Least Square with Discriminant (O2PLS-DA) analysis to integrate two omics data using the OmicsPLS package of R. This method decomposes the variation present in the two data matrices into three parts, the joint variation between the two datasets, the orthogonal variation that is unique to each dataset and noise. (4) Based on the enriched KEGG terms and the molecular interaction/reaction relationship in the KEGG pathway map, we drew a custom KEGG path diagram to visually show some of the major changes in the study. Finally, the results of these four parts validated and supported each other and contributed to a more comprehensive understanding of the effect of DNM1L on the progression of GBC.

Measurement of glutamic acid

The level of glutamic acid was assayed using the Glutamic Acid Colorimetric Assay Kit (Elabscience, Wuhan, China) Briefly, the cell pellet was resuspended with PBS and subjected to the sonication procedure. The sonicated samples were centrifuged at 3,100 g for 10 min, and 0.5 mL supernatant was taken for the following detection according to manufacturers’ protocol. Optical density (OD) measurements of the culture medium at a wavelength of 570 nm were conducted utilizing a spectrophotometer (Yoke Instrument, Shanghai, China).

Statistical analysis

Statistical analysis was performed using GraphPad Prism software (La Jolla, CA, USA). All values are expressed as mean ± standard deviation (SD). Differences were assessed using a two-tailed unpaired test or ANOVA for comparison of two or multiple groups, respectively. p < 0.05 was considered significant.

Results

High expression of DNM1L promoted GBC cell proliferation and inhibited cell apoptosis

Our analysis revealed an elevated level of DNM1L in clinical GBC samples compared to corresponding para-carcinoma tissues. The finding was corroborated by the publicly available dataset GSE76633, which showed a significant upregulation of DNM1L, with a Log₂FC of 1.4 (Fig. S1a-c). Next, we designed the gain- or loss-function assays in GBC-SD and NOZ cells without (− Dox) or with (+ Dox) Dox induction (Fig. 1a). As shown in Fig. S2a-c, the pLVX-TetOne-DNM1L, and Tet-DNM1L-shDNM1L system successfully induced DNM1L overexpression or silencing in cells after treating with Dox for 48 h.

Fig. 1

DNM1L promoted GBC cell proliferation and inhibited cell apoptosis. a The gain- and loss-function assay were designed using the Dox-inducible system in GBC-SD and NOZ cells. b After treating with or without Dox for 48 h, the cell viability of GBC-SD and NOZ cells was detected by CCK-8 assay and represented as OD values at 450 nm. c EDU staining was used to detect cell proliferation, and d the EDU-positive cells (red fluorescence) were quantified. e The distribution of cells across the various phases of the cell cycle was assessed using flow cytometry. f Western blot analysis for cell cycle markers, including Cyclin E and p21. g Quantitative analysis of the cell cycle markers. h Statistical data showed the cell percentage of early apoptosis (AnnexinV-FITC⁺PI⁻ cells) and late apoptosis (AnnexinV-FITC⁺PI⁺ cells). i Western blot analysis for the cleaved PARP and caspase 3. n = 3 per group. *p < 0.05, **p < 0.01, ***p < 0.001

We found that DNM1L overexpression positively regulated cell viability, no matter whether GBC-SD or NOZ cells (Fig. 1b). The results of EDU staining revealed that DNM1L overexpression notably increased the number of EDU-positive cells, suggesting the pro-proliferative role of DNM1L in GBC cells (Fig. 1c-d). In addition, DNM1L upregulation decreased the cell percentage of the G1 phase and increased the cell percentage of the S phase, whereas DNM1L knockdown had the opposite effect (Fig. 1e). The level of cell cycle marker Cyclin E was positively regulated by DNM1L. DNM1L also inhibited the expression of cell cycle marker p21 (Fig. 1f-g). Results from flow cytometry showed that DNM1L knockdown significantly increased cell percentage of early apoptosis and late apoptosis (Fig. 1h). Moreover, DNM1L knockdown upregulated the expression of cleaved PARP and cleaved caspase 3, suggesting the activity of the apoptosis pathway (Fig. 1i). These results revealed the promoting role of DNM1L in the GBC progression and development.

DNM1L enhanced the migration and invasion potential of GBC cells

To further assess the biological function of DNM1L in GBC, we carried out transwell assays to evaluate cell mobility. As shown in Fig. 2a, DNM1L overexpression increased the cells in the lower chamber of the culture dish, and DNM1L knockdown decreased the cells in the lower chamber of the culture dish. Thus, the upregulation of DNM1L confers enhanced migratory and invasive abilities to GBC-SD and NOZ cells. However, DNM1L silencing revealed a decrease in the mobility of these cells (Fig. 2b-d). As shown in Fig. 2e, DNM1L overexpression (Dox +) resulted in a significant increase in the enzymatic activity of MMP2 compared to cells without DNM1L overexpression (Dox-). Moreover, the loss of DNM1L induced by Dox decreased the enzymatic activity of MMP2. The above results illustrated that DNM1L accelerates GBC cell motility.

Fig. 2

DNM1L enhanced the invasiveness and migration capability of GBC cells. a-b The transwell assay without matrigel treatment detected the number of migrated cells. c-d The lower chamber of the transwell invasion model was collected and cultured with crystal violet dye. The number of invaded cells were counted. e Enzyme-linked immunosorbent assay (ELISA) analysis detected the concentration of MMP2 in the supernatant of DNM1L-mediated cells. n = 3 per group. *p < 0.05, **p < 0.01

DNM1L induced the mitochondrial dysfunction of GBC cells

Mitochondrial homeostasis was related to cancer development, including initiation, growth, survival, and metastasis stages (Fig. 3a). In this study, DNM1L overexpression significantly inhibited the production of ATP and increased cellular ROS in GBC-SD and NOZ cells (Fig. 3b-c). Following transmission electron microscope analyses showed that normal mitochondria in the cells without Dox treatment were short rod-like and densely packed in the periphery of the nuclear envelope and Dox-induced DNM1L overexpression altered mitochondrial morphology. Specifically, the overexpression of DNM1L caused mitochondrial fragmentation and it made mitochondria appear small and round under the Dox treatment (Fig. 3d). To further investigate the morphology and structural characteristics of the mitochondrial network, Mitotracker Red staining assay was performed to label mitochondria. The images and quantitative data showed that DNM1L overexpression induced mitochondrial fragmentation (Fig. 3e-f). Previous study suggests that a reduction in mtDNA copy number is typically correlated with an elevated risk of various cancers, and an increase in mtDNA level may be associated with a diminished risk of certain types of cancer [20]. As shown in Fig. 3g, the results suggested that DNM1L overexpression markedly decreased mtDNA content compared to control cells. These data revealed that DNM1L mediated mitochondrial dysfunction in GBC cells.

Fig. 3

DNM1L disrupted mitochondrial homeostasis and induced mitochondrial dysfunction in GBC. a Four important indicators of mitochondrial homeostasis were associated with the cancer progression. b The content of ATP in cells was detected and corrected according to the protein concentration. c Flow cytometry was performed to detect the ROS level of GBC cells. d The morphology of mitochondria was observed by transmission electron microscopy. e The mitochondrial network stained with Mitotracker red to identify mitochondrial fragmentation was visualized under microscopy, and f the fluorescent intensity was shown. g Cytosolic mtDNA copy number was quantitated via qRT-PCR assay in GBC cells. n = 3 per group. *p < 0.05, **p < 0.01, ***p < 0.001

DNM1L drove GBC tumorigenesis and cancer metastasis

Given that DNM1L contributed to the growth and malignant behaviors by in vitro assay, we further explored whether DNM1L augmented GBC cell-derived xenograft growth and metastasis. Next, the cell-derived xenograft models and liver metastasis tumor models in nude mice were established (Fig. 4a). DNM1L expression in the tissues of xenograft tumors was verified by western blot assay (Fig. S2d). As shown in Fig. 4b, DNM1L overexpression enhanced the tumor volume by around 50% of the control group without Dox treatment. On the contrary, DNM1L knockdown decreased the tumor volume by around 33% of the control group without Dox treatment. IHC staining showed that the tumor tissues from mice with DNM1L overexpression had higher expression of Ki67, suggesting that DNM1L increased the malignant degree of GBC tumor (Fig. 4c). The tumor sections with DNM1L silencing had more TUNEL-positive cells, suggesting that DNM1L inhibited GBC cell apoptosis (Fig. 4d).

Fig. 4

DNM1L promoted carcinogenesis of GBC cells and liver metastasis in mice. a The flow diagram showed the main operation and time for the cell-derived xenograft model and splenic injection liver metastasis experiment. b The xenografts from the mice injected with NOZ cells were collected and their volumes were calculated. c IHC staining showed the expression of Ki67 in different tumor sections. d TUNEL staining of tumor xenograft sections. e Representative images of luciferase signals in mice at 28 days after intrasplenic injection with NOZ cells. f Representative livers were shown and black arrows indicated the metastasis nodules. g Metastatic nodules were counted in the intrasplenic injection model. (h) HE staining detected the tumor cells in the live sections. n = 5 per group.**p < 0.01, ***p < 0.001

The mice with DNM1L overexpression had increased luciferase signals and intrahepatic metastatic nodules in their livers, whereas DNM1L knockdown prevented cancer cell metastasis and growth (Fig. 4e). The more abundant and enlarged liver metastatic foci in response to DNM1L overexpression. In contrast, the mice with DNM1L knockdown reduced metastatic foci by at least half (Fig. 4f). Taken together, these data strongly suggested that DNM1L promoted tumorigenesis and cancer metastasis.

USP3 physically interacted with and stabilized DNM1L

Next, we successfully transfected plasmids carrying the CDS region or siRNA of USP3 into GBC-SD and NOZ cells (Fig. S2e-f). USP3 overexpression or knockdown had no significant effects on the mRNA expression of DNM1L (Fig. 5a). Inversely, USP3 positively regulated the DNM1L protein level (Fig. 5b). Co-IP assay showed the combination of USP3 and DNM1L in GBC-SD and NOZ cells (Fig. 5c). In addition, the yellow fluorescence observed in the merged images indicated a distinct co-localization of USP3 and DNM1L within the cytoplasm, while no co-localization was detected in the nucleus (Fig. S3). Upon the treatment of CHX, the protein stability of DNM1L in USP3-mediated NOZ cells was detected. We discovered that USP3 overexpression was associated with a significantly extended DNM1L half-life after treatment with CHX, and USP3 downregulation accelerated the protein degradation of DNM1L (Fig. 5d). As shown in Fig. 5e, ectopic USP3 expression decreased ubiquitin and K48-linked ubiquitin that combined with DNM1L protein by the proteasome inhibitor MG132, whereas USP3 knockdown impaired DNM1L deubiquitylation. To identify the domain of DNM1L responsible for interaction with USP3, various DNM1L truncation mutants were generated in HEK293T cells to test their ability to interact with USP3. We found that USP3 had strong interaction with full-length DNM1L and the dynamin-type G domain of DNM1L (1–302 aa), but no interaction with the domain containing 303–643 aa and 644–736 aa of DNM1L (Fig. 5f). These findings suggested that the USP3-DNM1L interaction mediated the function of GBC cells.

Fig. 5

A USP3 interacted with DNM1L and enhanced its stability. a qRT-PCR and b western blot analysis for the DNM1L expression after USP3 overexpression or knockdown. c Co-IP assay detected the direct interaction between USP3 and DNM1L in GBC cells. d NOZ cells transfected with the indicated plasmids were treated with cycloheximide (CHX) for 0, 1, 3, 6, and 9 h, and they were collected for western blot analysis. e USP3 mediated deubiquitination and K48-linked deubiquitination of DNM1L were detected. Co-IP assays were performed with the indicated antibodies. (f) To identify the region of each protein responsible for the interaction, HA-tagged DNM1L WT or truncation mutants were co-expressed with Flag-USP3 in HEK293 cells. n = 3 per group. **p < 0.01, ***p < 0.001

The effect of USP3 in promoting the malignant activities of GBC cells partly was achieved through its regulation of DNM1L

To fully explore the interplay between USP3 and DNM1L, we examined the role of USP3 in inducing malignant activities through DNM1L. As shown in Fig. 6a-c, USP3 overexpression significantly promoted cell migration and invasion compared to empty vector-transfected cells, whereas Dox-induced DNM1L silencing suppressed the capabilities. In addition, USP3 had a promoting role in mitochondrial fission, which was abolished by the presence of DNM1L silencing (Fig. 6d). We also found that the cells knocking down USP3 presented the inhibitory effects on cell migration and invasion, while DNM1L overexpression restored the trends (Fig. 6e-f). These results underscore the key role of DNM1L in a USP3-induced enhancement of proliferation, migration, and invasion capacities of GBC cells.

Fig. 6

USP3 affected malignant behaviors by regulating DNM1L. a-c Migratory and invasive capability and cell viability of NOZ cells were assessed in different groups. d The Mitotracker red staining showed that USP3 regulated the mitochondrial homeostasis of NOZ cells by increasing the DNM1L level. e–f Migratory and invasive capability of NOZ cells were assessed in different groups. n = 3 per group. **p < 0.01, ***p < 0.001

DNM1L overexpression induced the disturbance of the signaling pathway of GBC cells based on the transcriptomic analysis

Transcriptomic analysis was performed in NOZ cells overexpressing DNM1L and its control cells to screen DEGs targeting DNM1L (Fig. 7a). A total of 1253 upregulated genes and 1521 downregulated genes between DNM1L-overexpressing cells and control cells at the criteria of |Log₂FC|> 1 and p < 0.05 (Fig. 7b) The results from PCA plot indicated a notable distinction among the various groups (Fig. 7c). A heatmap showed the top 20 upregulated and downregulated genes, among which AAK1, ATF3, and PCNA were typical cancer markers (Fig. 7d). The most pathways from GSEA belonged to main categories, such as human diseases, organismal systems, and metabolism. Importantly, DNM1L overexpression influenced the cellular processes, including cell growth and death, transport and catabolism, cellular community, and cell motility (Fig. 7e). Furthermore, the PPI network showed that the DNA replication and p53 signaling had more interaction with other pathways, as evidenced by the number of enriched genes (Fig. 7f).

Fig. 7

Characteristics of RNA profiling in NOZ cells with DNM1L overexpression. a NOZ cell samples with DNM1L overexpression or control cells were analyzed using RNA sequencing. b Pie charts showed the percentage of genes for each trait. c PCA was used to pre-process the raw data and describe sample characteristics. d Heatmap showed the expression of the top 20 upregulated and downregulated genes. e GSEA results based on the KEGG database were displayed in the pie charts. f The PPI network of key pathways was shown

DNM1L induced metabolic reprogramming by controlling the glutamic acid level

As the understanding of the complexity of tumor biology increases, metabolic reprogramming has been seen as a hallmark of malignancy and may serve as a target for therapeutic interventions [21]. However, it remains unclear whether the DNM1L-promoted progression of GBC was associated with metabolic reprogramming. Thus, we conducted metabolome analysis using the same samples as those employed in the transcriptome analysis (Fig. 8a). The plots of PCA and OPLS-DA supervised pattern recognition method showed a high degree of QC polymerization (Fig. 8b-c). In total, 31 DEMs were identified with p < 0.05 and VIP > 1, and they were enriched in the amino acid, fatty acid, and lipid metabolic pathways that were the characteristic changes in metabolism [22,23,24] (Fig. S4a-b). These findings indicated that the overexpression of DNM1L indeed led to alterations in the cell-intrinsic metabolism of GBC cells. Moreover, the potential marker metabolites might offer novel perspectives and new insights into the underlying mechanisms. Based on the above, the expression and class information of 31 DEMs were shown in the sample clustering heatmap (Fig. 8d). We discovered that most DEMs belonged to the class of carboxylic acids and derivatives, and those parameters were shown in a bubble diagram (right of Fig. 8d). Among that, only L-glutamic acid and succinic acid with p values < 0.001, whereas L-glutamic acid had a higher VIP value than succinic acid. Thus, we speculated that L-glutamic acid may be the key metabolite during the DNM1L-promoted malignant phenotype, which was verified by Glutamic Acid Assay Kit. As presented in Fig. 8e, the reduced glutamic acid level was detected in cells overexpressing DNM1L, which was consistent with the metabolome data.

Fig. 8

Characteristics of metabolism profiling in NOZ cells with DNM1L overexpression. a NOZ cell samples with DNM1L overexpression or not were analyzed using non-targeted metabolomics. b PCA and c OPLS-DA methods were used to analyze the features of samples in positive and negative ion models. d Heatmap showed the expression and class of 31 DEMs (p < 0.05 and VIP > 1) that responded to DNM1L overexpression. The bubble diagram showed the p value and VIP value that determined the importance of each metabolite under this class. e The content of glutamic acid in NOZ cells was detected. n = 3 per group. **p < 0.01

Discussion

USP3 is involved in multiple cancer progression and metastasis through its deubiquitinating enzyme activity [17, 25]. Based on its zinc-finger Ub-binding domain (ZnF-UBP) and a catalytic domain of the Ub-specific protease (USP) class, the ZnF-UBP domain of USP3 mediates monoubiquitinated H2A (uH2A)-USP3 interaction, which affected DNA damage response [26]. In this study, we found that USP3 was bound with the dynamin-type guanine nucleotide-binding domain (G_DYNAMIN_2) of DNM1L. The deubiquitinase USP3 had similar result in hepatocellular carcinoma [27]. Therefore, we speculated that G_DYNAMIN_2 might be an important domain for DNM1L to bind to USP3. To detail, overexpression of USP3 increased DNM1L stability by inhibiting the K48-like ubiquitination of DNM1L (Fig. 5d-e). Consistently, USP3 removed the K48-linked ubiquitination on ASC and strengthened its stability by blocking proteasomal degradation [28]. We believe that additional deubiquitinases may also possess the capability to deubiquitinate DNML1. We intend to further investigate this possibility and will disseminate our findings in subsequent research endeavors.

DNM1L functions as an oncogene in GBC: the results point to the function of DNM1L as a promoter role of the proliferation, invasion, and metastasis of GBC via widely regulated transcriptome and metabolome. The function of DNM1L as a critical GTPase for mitochondrial fission was nonnegligible. Mitochondrial DNA (mtDNA) copy number is a surrogate index of mitochondrial activity, which reflects the biogenesis and function of mitochondria. Previous study suggests that a reduction in mtDNA copy number is usually associated with an increased risk of the cancers [20, 29]. Consistent with our research, DNM1L was demonstrated to disturb mitochondrial function and mediated mitochondrial fission of GBC cells (Fig. 3). Furthermore, we observed significant apoptosis events after DNM1L silencing, including the enriched expression of cleaved caspase3 and PARP, which is due to that DNM1L-dependent mitochondrial fission dictated apoptotic susceptibility [30, 31]. In vivo assays verified that DNM1L overexpression not only promoted the xenograft formation but enhanced the hepatic metastases of GBC (Fig. 4). Some studies indicated that GBC patients complicated with hepatic invasion had poorer prognoses than those without invasion in long-term follow-ups [32]. In a similar vein, Zhao et al. demonstrated that DNM1L enhanced metastatic potential in human-invasive breast carcinoma [33]. These findings proved that DNM1L might serve as a new prognostic factor influencing the long-term survival of patients with advanced GBC.

In Fig. 8d, the clustering heatmap not only showed the concentration of each DEM but also displayed the class information that was identified based on the HMDB. The carboxylic acids and derivatives class include more cancer-related DEMs, such as succinic acid-involved succinylation was the metabolic bridge between cancer and immunity [34], targeting valine catabolism to inhibit metabolic reprogramming in prostate cancer [35], and ornithine as a possible marker of cancer [36]. Thus, L-Glutamic acid may be the key metabolite during DNM1L-promoted gallbladder cancer malignant phenotype, The role of glutamic acid in GBC is unclear, but glutamic acid has been shown to play an important role as an anticancer agent in a variety of cancers [37]. To ensure comprehensibility, we have only presented the fundamental information. In fact, we are investigating the function of these DEGs and DEMs in GBC development by joint analysis for the transcriptome and metabolome, and we will share these findings in the future.

Romani et al. reported that DNM1L-induced abnormal mitochondrial dynamics resulted in activation of the oxidation-antioxidation system, including increased cystine uptake and glutathione metabolism [11]. Moreover, a previous study demonstrated that glutamine metabolism was related to the ferroptosis of GBC cells [38]. In addition, DNM1L has been shown to facilitate the reprogramming of glucose metabolism, thereby markedly augmenting the Warburg effect in ovarian cancer cells [39]. These findings indicated that DNM1L played a tumor-promoting role in GBC by regulating several metabolism pathways, which in turn promoted the malignant behaviors of GBC cells. This study used transcriptomics data paired with metabolomics data to provide insight into the changes induced by DNM1L on GBCs.

Several major conclusions can be drawn from the findings presented in this study (Fig. 9). First, DNM1L emerged as a significant contributor to GBC development, manifesting as promoting tumorigenesis, invasiveness, and migration capability by modulating mitochondrial dynamics. Second, we confirmed that USP3 specifically targeted the K48-linked ubiquitination of DNM1L, thereby increasing the stability of the DNM1L protein. Furthermore, a comprehensive analysis integrating two omics approaches revealed major metabolic pathways and potential biomarkers related to DNM1L overexpression in GBC. These findings partly explain the rapid growth and poor prognosis of GBC.

Fig. 9

Schematic diagram of the function and mechanism of DNM1L in the malignant progression of GBC