USP15 is aberrantly highly expressed in GC tissues and indicative of malignant progression

To evaluate whether USP15 is meaningful in cancer, we first looked into public databases. The Cancer Genome Atlas (TCGA) database showed that USP15 was amplified in many types of tumors, including GC (Fig. S1A). The mRNA level of USP15 in most GC cell lines was upregulated based on the Cancer Cell Line Encyclopedia (CCLE) database (Fig. S1B). In both the TCGA-STAD and PUCH-STAD cohorts, USP15 expression levels were higher in gastric cancer (GC) tissues compared to normal tissues (Fig. 1A). In addition, patients with a high level of USP15 had poorer overall survival than those in the low expression group (Fig. 1B). As USP15 was found upregulated in GC, we confirmed its clinical significance via using immunohistochemistry. The staining of USP15 protein ranged from low to high and located in the cytoplasm (Fig. 1C), which indicated that USP15 expression was markedly increased in GC tissue sections, whereas USP15 staining was low or negative in adjacent normal tissue sections. Consistent with previous studies, the staining score of USP15 was significantly associated with tumor-node-metastasis (TNM) stage. Interestingly, the expression level of USP15 was elevated in the group of patients who were nonresponse to neoadjuvant chemotherapy (NCT) (Fig. 1D). Furthermore, both CCLE database and Genomics of Drug Sensitivity in Cancer (GDSC) database showed that USP15 expression was negative correlated with the efficacy of multiple chemotherapeutic drugs (Fig. 1E and Fig. S1C). These results suggested that abnormal expression of USP15 might be closely related to the progression of GC.

Fig. 1

High expression of USP15 is a marker for poor prognosis and chemotherapy resistance during GC progression. A The expression of USP15 was analyzed in GC tissues and adjacent normal tissues from The Cancer Genome Atlas (TCGA) cohort and Peking University Cancer Hospital (PUCH) cohort. ***P < 0.001 calculated by unpaired Student’s t-test. B Kaplan–Meier plots for the overall survival rate of patients with GC in group of USP15 high or low expression level in the TCGA cohort and KUGH cohort (GEO access number: GSE26253). The P-values as indicated are calculated by the log-rank test. C Representative images of immunohistochemical (IHC) staining of USP15 in GC tissues and adjacent normal tissues provided by Biobank of Peking University Cancer Hospital. Scale bar as indicated. USP15 staining scores were further evaluated by pathologists and correlated with clinicopathological characteristics. D Analysis of USP15 expression in endoscopic samples with the efficacy differences in patients receiving neoadjuvant therapy (NAT). The raw data derived from NAT cohort of PUCH (https://doi.org/10.1126/sciadv.aay4211, NGDC access number: HRA000067). E Correlation analysis of the sensitivity of chemotherapeutic drugs and the expression level of USP15 in GC cells provided by Cancer Cell Line Encyclopedia (CCLE) database and Genomics of Drug Sensitivity in Cancer (GDSC) database. The expression levels of USP15 were normalized by the z-score method to better demonstrate. -Log10AUC represents the sensitivity of drug, and lower values indicate that the cells are less sensitive to the drug. The P-values as indicated are calculated by Pearson correlation analysis

Establishment of stable GC cell lines and evaluating the functions of USP15 in vitro

To determine the cellular functions of USP15, the expression level of USP15 was detected among several cell lines. The results showed that both mRNA and protein levels of USP15 were significantly higher in SGC7901 and BGC823 cells than others (Fig. 2A and B). We synthesized two shRNA sequences to knockdown USP15 in GC cells and constructed an ectopic expression plasmid of USP15, and further established stable cell lines. The efficiency of either knockdown or overexpression of USP15 was detected by western blot (Fig. 2C). IncuCyte system for live-cell imaging and analysis was used to determine tumor cells proliferation activity, and knockdown of USP15 was consistently inhibited the cell proliferation in both SGC7901 and BGC823 cells (Fig. 2D). In addition, the colony formation ability was notably suppressed in cells with lower expression (Fig. 2E). In line with these results, the activity of DNA replication in USP15 knockdown cells was markedly decreased (Fig. 2F), which further confirmed that USP15 was involved in the regulation of aberrant proliferation signaling in GC cells. Interestingly, USP15 knockdown resulted in enhancement of DNA breakage in GC cells, as validated by the TUNEL assay (Fig. 2G). The activation of apoptotic process was confirmed via Annexin-V/PI staining and detection of death-associated proteins abundance (Fig. S2A and B).

Fig. 2

Establishing cell models with knockdown or overexpression of USP15 is performed to determine the effect of USP15 on the malignant phenotype. Relative USP15 expression levels in cell lines derived from normal gastric mucosa and primary GC by A qRT-PCR analysis and (B) western blot analysis. *P < 0.05 vs. GES-1. C Upper panel: SGC7901 and BGC823 cells were transfected with USP15 shRNAs to establish the stable knockdown cell lines. Lower panel: NCI-N87 cells were transfected with Flag-tagged USP15 ectopic expression plasmid to establish the stable overexpression cell line. The efficiencies were further confirmed by western blot analysis. β-Tubulin was used as an internal control. D Cell proliferation was measured by IncuCyte live cell. E Colony formation assay. Number of colonies was further counted. F The activity of DNA replication was detected by EdU staining. Scale bar = 500 μm. *P < 0.05 vs. sh-NC. G DNA strand breakage was determined by TUNEL staining. Scale bar = 100 μm. The quantitative measurements of the percentage of either EdU-stained positive cells (red) or TUNEL-stained positive cells (green) were presented in column shown in the right panel. Data are expressed as mean ± SD. *P < 0.05 vs. sh-NC, n = 12 independent experiments

As high expression level of USP15 was associated with metastasis in patients with primary GC, we assumed that USP15 might endow GC cells with invasive behavior. Both wound healing and invasion assays confirmed that knockdown of USP15 significantly inhibited migration and invasion (Fig. S2C and D). Besides, USP15 knockdown partially reversed the remodeling of the epithelial-mesenchymal transition (EMT), as promoting the expression of E-cadherin and reducing the expression of N-cadherin and Vimentin (Fig. S2E). Collectively, knockdown of USP15 in gastric cancer cells could lead to lower level of tumor progression.

Knockdown of USP15 enhances the antitumor effect of chemotherapeutic drugs and distal colonization in vivo

Our previous results found that knockdown of USP15 could improve the sensitivity of chemotherapeutic drugs against GC cells, especially platinum drugs (Fig. S3A). To establish the functional importance of USP15 on the progression of GC, nude mice were subcutaneously injected with SGC7901 cells (sh-NC, sh-1#), respectively. After 1 week, xenograft tumor models could be observed growing by eyes nearly at the same time. We further divided the nude mice into four groups and treated with Oxaliplatin (L-OHP, 5 mg/Kg, i.p.) and saline respectively, twice per week, lasting for 4 weeks. Although inhibition of USP15 expression did not effectively prevent tumor growth, USP15 knockdown indeed enhanced the antitumor effect of L-OHP, which was similar with the results of in vitro (Fig. 3A-C). Moreover, the expression level of USP15 remained low in xenograft tumor tissues of the knockdown group and was consistent with the trend of proliferation markers expression (Fig. 3D and Fig. S3C). Next, we evaluated the effect of USP15 knockdown on tumor metastatic colonization in nude mice. As usual, SGC7901-luc cells stably transfected with either sh-NC or sh-1# were injected into nude mice via the tail vein. Bioluminescent imaging was performed at day 14, 21, 28, 35 and only slight distant metastases were found in mice of the USP15 knockdown group (Fig. 3E and F). These data indicated that USP15 aggressively promoted malignant progression and is expected to be a therapeutic target for GC.

Fig. 3

Knockdown of USP15 enhances the antitumor effect of oxaliplatin and inhibits distal colonization in xenograft models. A Xenograft models were established as previously described. When tumors were large enough (approximately 100 mm3), both sh-NC group and sh-1# group were further randomized to treatment arms. Xenograft tumor volumes measured every seven days for five weeks after treated with either oxaliplatin (L-OHP) or saline. At the end of experiment, the mice were sacrificed, and tumors were collected. B Representative images of xenograft tumors. C Tumor weight was measured. Data are expressed as mean ± SD. *P < 0.05 vs. sh-NC, n = 5 independent experiments. D IHC staining for Ki67, USP15, HKDC1 expression, and H&E staining of xenograft tumors. E The effects of USP15 knockdown on metastatic colonization through blood circulation. Either SGC7901-luc sh-NC cells or sh-1# cells were injected intravenously into mice. At each indicated time, mice were injected with D-luciferin and bioluminescence imaging was performed. Lumina intensity indicates the ability of tumor cell metastasis. F Representative images of Bouin-fixed lung specimens of distal colonization models and numbers of metastatic nodules in lung per mouse for each group were counted. *P < 0.05 vs. sh-NC, n = 6 independent experiments

USP15 plays an important role in glycolytic remodeling in GC progression

To predict the potential downstream regulatory mechanisms of USP15 signaling axis, we performed a combined multi-omics analysis through transcriptome profiling, targeted metabolome profiling, and interacting protein identification profiling. The transcriptome sequencing identified a set of differentially expressed genes after USP15 knockdown and presented these genes via volcano plot and heatmap (Fig. 4A and Fig. S4A) The result of functional enrichment analysis showed high confidence of genes enriched in the function of ubiquitination, degradation of the extracellular matrix, pyruvate metabolism, and glycolysis (Fig. 4B and S4B). Notably, knockdown of USP15 significantly reduced the production of glycolytic process in GC cells, including lactic acid, fructose-6-phosphate (F-6-P), and glucose-6-phosphate (G-6-P), which was consistent with the canonical molecules of these pathways significantly modulated (Fig. 4C and D and Fig. S4C). Besides, low expression of USP15 has a lower metabolism signature score in TCGA-STAD cohort, especially for glycolysis and OXPHOS activity (Fig. S4D). Identification of USP15-interacting proteins was performed using CoIP-MS based proteomics, and silver staining was conducted to distinguish the differential protein bands in the resultant immunoprecipitants (Fig. 4E). These immunoprecipitants was next analyzed by MS, and the unique proteins captured by USP15 antibody were compared to previous studies that reported USP15 interaction (Fig. 4F and Table S4). In line with transcriptomics and metabolomics results, CoIP-MS analysis showed that some core regulators in the glycolytic pathway bind to USP15, including HKDC1 and IGF2BP3 (Fig. 4G and Fig. S4E), suggesting that USP15 might be directly involved in the ubiquitination modification of these proteins.

Fig. 4

Prediction of potential biological functions and targets of USP15. A RNA sequencing analysis was performed to detect differentially expressed genes in GC cells following USP15 knockdown, and volcano plot was displayed with tumor driver genes (e.g., EGF, CD274) and metabolic key enzymes (e.g., ALDOA, LDHA) as indicated. B Functional enrichment analysis based on GSEA was performed. Knockdown of USP15 significantly affect the regulation of multiple glucose metabolic pathways, including glycolysis, oxidative phosphorylation (OXPHOS). C Heatmap of the average order of magnitude of central carbon metabolites. D Simplified schematic overview of central carbon metabolism in Homo sapiens, with heatmap of the log2 fold change of average metabolite and mRNA levels in sh-1# cells versus sh-NC cells. Two-tailed T-test, adjusted P-value (False Discovery Rate (FDR)) = 0.05. Metabolites and related genes colored light grey were not detected or unchanged. Co-enzymes and substrates are not included. E Co-immunoprecipitation experiments were performed using SGC7901 cell lysates with anti-USP15 antibody, or with IgG as negative control. The proteins were resolved on SDS-PAGE and stained with silver staining. F The specific peptides in the USP15 complex were resolved using MS, and the bands of potential targets supported by MS assay were indicated by arrows shown in the (E). G Analysis of protein–protein interaction network of USP15 based on CoIP-MS

Knockdown of USP15 suppresses glycolytic activity and impairs mitochondrial functions

The above indicated that USP15 may be an important component for the regulation of glycolytic activity in GC cells, thereby we measured the effects of inhibition of USP15 expression on glycolysis and mitochondrial OXPHOS processes. USP15 knockdown not only resulted in blocking of glycolysis and glycolytic capacity, but also decreased the levels of both basal O2 consumption rate (OCR) and max OCR (Fig. 5A), indicating compromised aerobic glycolysis and aerobic oxidation. As the integrity of mitochondria is important for maintaining the OXPHOS process as well as tumor cell viability, we analyzed the functional and structural integrity of mitochondria in GC cells after USP15 knockdown. The JC-1 probe was used to detect the mitochondrial membrane potential to indicate the functionality of the mitochondria. The results showed that USP15 knockdown induced an abnormal reduction in the mitochondrial membrane potential, as the fluorescent pattern changed from a punctate red fluorescence to a diffuse green fluorescence (Fig. 5B and Fig. S5A). Besides, the number of cells with abnormal mitochondria increased significantly after knockdown of USP15 (Fig. 5C and Fig. S5B). Importantly, transmission electron microscope imaging clearly observed the abnormalities of mitochondrial structure in GC cells caused by USP15 knockdown (Fig. 5D and Fig. S5C). Elevated intracellular ROS levels and aberrant expression of mitochondria-related proteins further indicated the mitochondrial damage in GC cells following USP15 knockdown (Fig. 5E and Fig. S6).

Fig. 5

Knockdown of USP15 inhibits glycolytic activity and causes both functional and structural damage to mitochondria. A Both ECAR and OCR of stable USP15-knockdown or negative control SGC7901 cells were measured. For ECAR, the Seahorse automatically filled each well with 10 mmol/L glucose, 1 µmol/L oligomycin, and 50 mmol/L 2-DG successively. Basal glycolysis was determined after addition of glucose, and glycolytic capacity was calculated after addition of oligomycin. For OCR, 1 µmol/L oligomycin, 1 µmol/L FCCP, and 0.5 µmol/L rotenone were automatically injected successively. Basal OCR was determined prior to addition of oligomycin, and maximal respiratory capacity was determined by subtracting nonmitochondrial OCR, calculated after rotenone injection to maximal OCR upon FCCP uncoupling and maximal electron transport. Data are expressed as mean ± SD. *P < 0.05 vs. sh-NC, n = 12 independent experiments. B Mitochondrial membrane potential of stable USP15-knockdown or negative control SGC7901 cells was detected by JC-1 probe. The red fluorescence represents the mitochondrial aggregate JC-1, and the green fluorescence indicates the monomeric JC-1. Scale bar = 200 μm. C Changes of mitochondrial membrane potential were further measured quantitatively by FACS analysis. Erastin was used to trigger the mitochondrial stress in GC cells. D Representative transmission electron microscope (TEM) images of the morphological and subcellular structure of different groups. Red arrows represent mitochondria with abnormal structure, and black arrows represent mitophagy. Scale bar = 1 μm. E Intracellular ROS level was assayed by FACS analysis using DCFH-DA fluorescent probe. Data are expressed as mean ± SD. ***P < 0.001 vs. sh-NC, n = 10 independent experiments

The glycolytic regulators HKDC1 and IGF2BP3 are the core interacting targets of USP15

Combined with the multi-omics results, we hypothesized that hexokinase HKDC1 and IGF-binding protein IGF2BP3 are the key targets of USP15 downstream. The CoIP experiments were performed to verify the interaction of USP15 with HKDC1 or IGF2BP3. The results confirmed that both HKDC1 and IGF2BP3 interacted with USP15 directly (Fig. 6A and Fig. 6B), and those proteins colocalize at the subcellular level (Fig. 6C and Fig. S7). Given that USP15 is a regulator on the glycolysis, we wonder whether USP15 would be a deubiquitinase for both HKDC1 and IGF2BP3. We further determined the role of USP15 in the expression of HKDC1 and IGF2BP3. Interestingly, knockdown of USP15 reduced the protein level of HKDC1 without affecting IGF2BP3 (Fig. 6D). After blocking the proteasome activity by treating cells with MG132, the decreased protein level of HKDC1 induced by USP15 knockdown was rescued, indicating that USP15 is responsible for the stability of HKDC1 protein. Importantly, USP15 knockdown notably increased the accumulation of ubiquitinated HKDC1 (Fig. 6E). Considering the synthesis pathway, cycloheximide (CHX), a protein synthesis inhibitor, was used to detect the half-life of protein. Western blot analysis showed that USP15 knockdown decreased the half-life of HKDC1 rather than IGF2BP3 (Fig. 6F).

Fig. 6

USP15 interacts with both HKDC1 and IGF2BP3. A Left panel: Cell lysates were extracted from SGC7901 cells. Immunoprecipitated with anti-USP15 antibody, or with IgG as negative control. Right panel: Cell lysates were extracted from N87 Flag-USP15 cells or Flag tagged empty vector (Flag-EV) transfected cells. Immunoprecipitated with anti-Flag. HKDC1, IGF2PB3, and USP15 were further detected by western blot analysis. B Immunoprecipitation experiments were performed using SGC7901 cell lysates with anti-HKDC1 antibody (left panel) or anti-IGF2BP3 antibody (right panel), IgG was used as negative control. HKDC1, IGF2PB3, and USP15 were further detected by western blot analysis. C Colocalization analysis by dual-color confocal imaging. Scale bar = 50 μm. D Western blot analysis was performed to determine the expression levels of HKDC1 and IGF2BP3 following USP15 knockdown. Data are expressed as mean ± SD. *P < 0.05 vs. sh-NC, n = 6 independent experiments. E GC cells were co-transfected with HA-Ub and USP15 shRNAs. MG132, a proteasome inhibitor, was used to treated with among USP15 sh-1#, sh-2#, and sh-NC cells. After 4 h treated with 5 μM MG132, immunoprecipitation experiments were performed using these cell lysates with anti-HKDC1 antibody. The ubiquitination level of HKDC1 was detected by western blot using anti-HA antibody. F Cycloheximide (CHX, 50 μM), a protein synthesis inhibitor, was used to treated with USP15 sh-1# and sh-NC cells. The accumulation levels of both HKDC1 and IGF2BP3 were evaluated by western blot analysis. β-Tubulin was used as an internal control. Data are expressed as mean ± SD. *P < 0.05 vs. sh-NC, n = 6 independent experiments

To further evaluate whether the enzymatic activity of USP15 is essential for the deubiquitination of HKDC1, we transfected USP15 knockdown cell lines with expression vectors for USP15 variants, USP15-298A (loss of enzyme activity) and USP15-812A (loss of activity towards polyubiquitin). Ectopic expression of USP15-298A mutant resulted in decreased protein level of HKDC1 (Fig. S8A). The up-regulated level of ubiquitination of HKDC1 caused by repression of USP15 was reversed by USP15 wild-type, but not by the 298A mutant, whereas USP-812A mutant showed differently (Fig. S8B). As expected, the attenuation of glycolytic activity in GC cells was rescued by overexpression of USP15 wild type (USP15-WT), rather than 298A mutant (Fig. S9A). Importantly, overexpression of USP15 mutants deficient in enzymatic activity in USP15 knockdown cells failed to enhance the activities of both proliferation and DNA replication in GC cells (Fig. S9B and C). In addition, only USP15-WT was able to inhibit apoptosis and maintain mitochondrial stability in GC cells (Fig. S10A and B). All these results suggest that deubiquitination of HKDC1 requires the enzymatically active USP15, which regulates glycolytic activity.

HKDC1 is a key target of USP15 to regulate glycolytic activity

To further determined whether HKDC1 is essential for the regulation of glycolytic activity by USP15 in GC, we used HKDC1 blocker or siRNA to interfere with HKDC1 function or expression. Overexpression of HKDC1 enhanced glycolytic activity and can be reversed by HKDC1 blocker in GC cells (Fig. 7A and B and Fig. S11). Moreover, HKDC1 blocker decreased the activates of both proliferation and DNA replication (Fig. 7C and D), which were enhanced by overexpression of HKDC1. Similarly, HKDC1 silencing in NCI-N87 cells reversed the enhancement of glycolytic activity induced by USP15 up-regulation (Fig. S13A and B). Besides, in line with the results of HKDC1 blocker, the cell viability was also suppressed by HKDC1 silencing (Fig. S13C and D). These data confirmed that HKDC1 is required for USP15 regulated glycolytic activity in GC.

Fig. 7

USP15 regulated glycolysis through HKDC1. A Western blot analysis was performed to determine the expression levels of USP15, HKDC1 and His-tag after treating with HKDC1 blocker. β-Tubulin was used as an internal control. B ECAR of USP15-sh-1#, USP15-sh-His-HKDC1, USP15-sh-His-HKDC1 + blocker and empty vector cells were measured. Data are expressed as mean ± SD. *P < 0.05 vs. empty vector, #P < 0.05 vs. USP15-sh-1#, $P < 0.05 vs. USP15-sh + His-HKDC1, n = 8 independent experiments. C Cell proliferation was measured by IncuCyte system. D The activity of DNA replication was detected by EdU staining. Scale bar = 500 μm. Data are expressed as mean ± SD. *P < 0.05 vs. empty vector, #P < 0.05 vs. USP15-sh-1#, $P < 0.05 vs. USP15-sh-His-HKDC1, n = 12 independent experiments