SPTBN1 abrogates renal clear cell carcinoma progression via glycolysis reprogramming in a GPT2-dependent manner

ccRCC and adjacent normal renal samples were obtained by radical nephrectomy from the First Affiliated Hospital of Nanjing Medical University (Jiangsu Province Hospital) between 2005 and 2018. Detailed clinicopathological information of involved patients was listed in Table 3. All diagnoses were confirmed by senior pathologists independently. Informed consent was provided by all patients. The study design and protocol were approved by the ethics committee of the First Affiliated Hospital of Nanjing Medical University (Jiangsu Province Hospital).

IHC was performed as previously described [40, 41]. Briefly, the primary antibodies were diluted as follows: anti-SPTBN1 (1:150, Abcam, USA) and anti-GPT2 (1:200, Proteintech, China). TMA was constructed to analyze a total of 180 ccRCC samples. To assess the degree of protein staining expression level, TMA staining signals were quantified using H-score system [42] (H-Score = ∑(I × Pi), I = intensity of staining, and Pi = percentage of stained tumor cells, ranging from 0 to 300). The expression level of protein was separately and independently assessed by two experienced pathologists blinded to the clinical outcomes. According to the average expression level of IHC, the SPTBN1 in ccRCC patients was considered as low expression (H-score < 70) and high expression (H-score ≥ 70). The GPT2 was considered as low expression (H-score < 60) and high expression (H-score ≥ 60).

RCC cell lines (786-O, 769-P, ACHN, Caki-1) and human renal tubular epithelial cell line (HK-2) were purchased from the Type Culture Collection of the Chinese Academy of Sciences (Shanghai, China) and cultured in RPMI 1640 (786-O, 769-P), McCoy's 5A (Caki-1), DMEM (ACHN) and DMEM/F12 (HK-2) (Gibco, Thermo Fisher Scientific, USA) containing 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin (Gibco, Thermo Fisher Scientific, USA). The lentiviral vectors containing short hairpin RNAs targeting SPTBN1 (shSPTBN1) and negative control (shNC) were constructed and transfected.2-Deoxy-D-glucose (2-DG; Selleck, China), a glycolysis inhibitor, was applied to inhibit aerobic glycolysis. Cells were transfected with small interfering RNA (siRNA) and overexpression SPTBN1 plasmid (oeSPTBN1) using Lipofectamine 3000 (Invitrogen, Thermo Fisher Scientific, USA).

Total RNA was extracted using TRIzol Reagent (Sigma-Aldrich, Merck, Germany). HiScript III All-in-one RT SuperMix (Vazyme, China) was used for cDNA synthesis according to the manufacturer’s instructions. QRT-PCR was performed with SYBR qPCR Master Mix (Vazyme, China) using StepOne Plus (Applied Biosystems, USA) and LightCycler 480 PCR instrument (Roche Diagnostics, Switzerland). The primers and siRNA Oligo used were listed in Additional file 1: Table S1.

Total proteins were extracted using RIPA buffer supplemented with Halt protease and phosphatase inhibitor cocktail (Thermo Fisher Scientific, USA). Then, total 20 μg of protein was resolved using SDS-PAGE gels (4%-12%) and transferred onto polyvinylidene fluoride (PVDF) membranes (Millipore, USA). The membranes were blocked in TBST containing 5% skim milk powder for 2 h at room temperature. The membranes were incubated with the primary antibodies overnight at 4 °C, then with the secondary antibodies conjugated to horseradish peroxidase (HRP) for 1.5 h at room temperature, respectively. Band signals were detected by chemiluminescence system (Bio-Rad, USA).

Pretreated cells were counted and seeded into 96-well plates. Cell proliferation was measured using the CCK-8 Cell Counting Kit (Vazyme, China). The absorbance was measured at 450 nm with a microplate reader, following a 2-h incubation at 37 °C.

For the colony formation assay, pretreated cells were seeded into 6-well plates and incubated for 10–15 days. Colonies were fixed in 4% paraformaldehyde for 20 min and stained with 0.1% crystal violet for further analysis.

Cells were labeled using PI/RNase Staining Buffer (BD Biosciences, USA) according to the manufacturer’s protocol and followed by flow cytometry (BD Biosciences, USA). The results were further analyzed using Modfit software.

In the migration assay, pretreated cells were seeded into the upper 24-well Transwell chambers with serum-free medium. Matrigel medium (Corning, USA) was applied for invasion assay. The medium containing 20% FBS was added to the bottom chamber. After incubation at 37 °C for 24 or 36 h, the cells were fixed in 4% paraformaldehyde, stained with 0.1% crystal violet, and captured in five randomly selected fields by a microscope. All the experiments were repeated for at least three times.

All the mouse experiments were approved by the Institutional Animal Care and Use Committee (IACUC) of Nanjing Medical University (No. IACUC-1908003). Briefly, total 2 × 10⁷ 786-O cells with stably knockdown-SPTBN1 (shSPTBN1) and negative control cells were collected and suspended with PBS and Matrigel (1:1, Corning, USA), then subcutaneously injected into 4-week-old female BALB/c nude mice. The tumor formation and volume were measured every other day. The formula of tumor volume was calculated as follows: Tumor volume = length * width² / 2. Xenograft tumor samples were fixed in 4% paraffin for further research.

786-O and Caki-1 cells were treated with 5 μg/mL Actinomycin D (Selleck, China) for 0, 2, 4, 6, 8, or 10 h. Total RNA was harvested for qRT-PCR experiments. The level of GPT2 transcript was normalized to that of β-actin control, and the relative half-life of GPT2 was calculated using GraphPad software.

RNA immunoprecipitation (RIP) assay was performed with Magna RIP RNA-Binding Protein Immunoprecipitation Kit (Millipore, USA) according to manufacturer’s protocol. Briefly, a total of 2*10⁷ 786-O and Caki-1 cells were lysed in RIP lysis buffer, then incubated with magnetic beads conjugated with anti-SPTBN1 or anti-Rabbit IgG antibodies for 6 h at 4 °C temperature. After centrifugation and washing, the supernatant was treated with Proteinase K. Finally, the immunoprecipitated RNA was purified and extracted using TRIzol Reagent for further qRT-PCR analysis.

786-O shNC and 786-O shSPTBN1 cells were applied for RNA-sequencing in biological duplicates. For each sample, 1.5 μg of RNA was generated for library preparation using NEBNext Ultra RNA Library Prep Kit for Illumina (NEB, USA). The library preparation was accomplished on an Illumina Hiseq-4000 platform (Allwegene, Beijing, China). Raw data in fastq format were processed for quality control and further analyzed using “DESeq” R package [43, 44] (Benjamini and Hochberg’s algorithm). Genes with adjusted P-value (q-value) < 0.05 were considered as differentially expressed.

786-O and Caki-1 ccRCC cells were transfected with pEZX-FR02, a luciferase reporter containing wild-type or mutated GPT2-3’UTR region. The ratios of firefly and Renilla luciferase activities were calculated after 48 h of transfection using Dual-Luciferase Assay System (Promega, USA).

Assay kits (Nanjing Jiancheng Bioengineering, China) were used to measure the glucose concentrations, lactic acid production, and ATP level in 786-O and Caki-1 ccRCC cell lines. Concentration was calculated and normalized to cell number.

TCGA-KIRC cohort, GSE6344, GSE40435, GSE46699, GSE53757, GSE66270, GSE105261, together with single-cell RNA-seq GSE152938 and GSE156632 datasets were employed in this study [45‐54]. Cancer Cell Line Encyclopedia (CCLE), CPTAC and UALCAN databases were also applied [55‐57]. The proteogenomic expression data in Chinese ccRCC patients were downloaded from the supplementary materials of Qu [58]. For single-cell RNA-seq analysis, R package “Seurat” was used to process data [54, 59‐61]. The cells expressing > 20% mitochondria-related genes, > 5% of hemoglobin reads, less than 500 genes, or greater than 10,000 genes were filtered out, then we normalized and rescaled the expression matrix. Next, cell clusters were obtained through the UMAP method. Finally, we used the “SingleR” package to annotate the cell clusters [62].

Statistical analysis was run on SPSS and R software. Data were expressed with the means ± SD. The independent t-test was used to compare continuous variables that exhibited normal distributions. The Wilcoxon test was used to compare the continuous variables that were not normally distributed. Kaplan–Meier survival analysis and Cox hazard regression model were employed to sift out factors predicting the overall survival, disease-specific survival, disease-free interval, and progression-free interval prognostic factors. Correlations analysis was performed by Pearson and Spearman methods. All experiments were repeated at least three times. All statistical tests were two-sided, and P < 0.05 was considered statistically significant. ns: No significant; *:P < 0.05; **:P < 0.01; ***:P < 0.001; ****:P < 0.0001.

Seven Spectrin-family genes with significantly abnormal expression profiles in ccRCC tissues were screened out from the TCGA-KIRC, along with the clinicopathological parameters of related patients [15, 16, 28, 29]. The flowchart of screening process was illustrated in Fig. 1A. Our results confirmed that SPTA1 and SPTBN5 were up-regulated in ccRCC tissues, while SPTAN1, SPTB, SPTBN1, SPTBN2, and SPTBN4 were down-regulated in ccRCC tissues (Fig. 1B, C). Meanwhile, we analyzed the distribution of Spectrin-family genes using datasets of single-cell RNA-sequencing (Additional file 1: Figure S1A-E). SPTBN1 was mainly enriched in endothelial cells and tissue stem cells, and SPTBN2 in endothelial cells and natural killer cells (Fig. 1D, E). Then, we conducted univariate and multivariate cox regression analyses to assess the prognostic value of Spectrin-family genes. Among these candidate genes, only SPTBN1 and SPTBN2 were significantly associated with overall survival (OS), disease-specific survival (DSS), and progression survival interval (PFI) (Additional file 1: Figure S2A–C). In addition, the correlations between Spectrin-family genes were analyzed (Additional file 1: Figure S3A). Functional enrichment analysis demonstrated that Spectrin-family genes were mainly involved in the interaction between L1 and Ankyrins (Additional file 1: Figure S3B). After targeting small interfering RNA to SPTBN1 and SPTBN2, we identified SPTBN1 as a candidate target in ccRCC among Spectrin-family genes (Fig. 1F). Finally, our pan-cancer analysis discovered that SPTBN1 was significantly down-regulated in the tissue samples of most cancer subtypes, compared to that in adjacent normal samples (Additional file 1: Figure S4A). Based on the Cancer Cell Line Encyclopedia (CCLE) dataset, we observed that SPTBN1 was highly expressed in kidney cancer, compared to that in most of other solid tumors [63] (Additional file 1: Figure S4B). These findings indicated that SPTBN1 might act as a tumor suppressor role in ccRCC.

To further investigate the tumor-suppressing role of SPTBN1 in ccRCC, firstly we detected the expression of SPTBN1 in multiple databases. The results showed that SPTBN1 was significantly decreased in both paired and unpaired samples from the TCGA-KIRC cohort (P < 0.001; Fig. 1G). This expression patterns was further validated in external GEO databases, including GSE66270, GSE105261, GSE6344, GSE40435, GSE46699 and GSE53757. (Additional file 1: Figure S5A–F). CPTAC and FUSCC proteome databases also demonstrated that the protein level of SPTBN1 was downregulated in ccRCC tissues (P < 0.001; Fig. 1H).

We further explored the expression of SPTBN1 in clinical ccRCC samples by IHC, qRT-PCR and WB. IHC staining results confirmed that the expression of SPTBN1 was remarkably decreased in tumor tissues (Additional file 1: Figure S5G). Besides, the receiver operating characteristic (ROC) curve analysis showed a high diagnostic value of SPTBN1 for ccRCC (AUC:0.692; 95% CI = 0.644–0.741; Fig. 1I). Moreover, qRT-PCR and WB also demonstrated the same results (NJMU ccRCC cohort 1: N = 53; Fig. 1J, K). In accordance with the findings in TCGA and GEO datasets, SPTBN1 expression was significantly down-regulated in ccRCC cell lines (786-O, 769-P, ACHN, and Caki-1) at both mRNA and protein levels, compared with that in renal tubular normal epithelial cell line HK-2 (P < 0.05; Fig. 1L).

We observed that low expression of SPTBN1 was positively correlated with high age, TNM stage, and tumor grade (P < 0.05, Fig. 3A; Table 1). Meanwhile, the analysis based on CPTAC and FUSCC proteome databases revealed low SPTBN1 expression in high-grade ccRCC patients (P < 0.05; Additional file 1: Figure S6A, B). In univariate and multivariate cox regression analyses, SPTBN1 served as an independent predictive factor for the survival of ccRCC patients (Univariate: HR = 0.532, 95% CI = 0.444 − 0.637, P < 0.001; Multivariate: HR = 0.647, 95% CI = 0.489–0.854, P = 0.002; Fig. 2B). We also examined the correlation between SPTBN1 and survival in subgroups of different clinicopathological characteristics, discovering that a low expression of SPTBN1 might predict a poor prognosis (Additional file 1: Figure S7A–D). The logistic regression analysis was adopted to detail the correlativity between SPTBN1 expression and clinicopathological characteristics (Table 2). Moreover, a nomogram based on TNM stage and SPTBN1 was developed to predict the 1-, 3-, and 5-year OS of each ccRCC patient (Fig. 2C). Kaplan–Meier analysis further indicated that a low SPTBN1 expression was correlated with worse overall survival, disease-specific survival, and progression-free survival in the TCGA-KIRC cohort (Fig. 2D). According to SPTBN1 expression in our tissue microarray from NJMU TMA cohort (N = 180), we divided the patients into high-SPTBN1 and low-SPTBN1 expression subgroups (Fig. 3E; Table 3). The results showed that a low SPTBN1 expression was associated with poor overall survival and relapse-free survival (P < 0.01; Fig. 3F). Simultaneously, we observed that SPTBN1 was predominantly down-regulated in ccRCC patients with advanced clinicopathological characteristics, which was consistent with the results of TCGA-based analysis (Table 3). Taken together, SPTBN1 possessed a high prognostic value for ccRCC patients.

To further investigate the biological function of SPTBN1 in ccRCC, we successfully established SPTBN1-knockdown and SPTBN1-overexpression cell models (786-O and Caki-1) and validated by qRT-PCR and WB (Additional file 1: Figure S8A, B, P < 0.05). Cell counting kit-8 (CCK-8) assay indicated that SPTBN1 knockdown significantly increased cell proliferation ability (Fig. 3A). Colony formation assay was also employed to determine the long-term impact of SPTBN1 on cells proliferation. Colony formation was more evident in SPTBN1 knockdown group than in control group, while SPTBN1 overexpression reversed this change (Fig. 3B). Furthermore, flow cytometry showed that SPTBN1 knockdown increased the proportion of 786-O and Caki-1 cells in the S phase and decreased that in the G0-G1 phase, while SPTBN1 overexpression resulted in opposite effect (Fig. 3C). Additionally, Transwell migration and invasion assay demonstrated that SPTBN1 knockdown turned CCRCC cells more aggressive and metastatic (Additional file 1: Figure S9A).

In addition, to determine the influence of SPTBN1 in vivo, we established subcutaneously xenograft tumor models using nude mice (Fig. 3D). Knocking down SPTBN1 dramatically accelerated tumor growth, as quantified by tumor size and tumor volume (Fig. 3D). Accordingly, IHC experiments further revealed that Ki-67 and PCNA, indicators of proliferation, were notably up-regulated in tumors derived from shSPTBN1 786-O cells, indicating more rapid tumor growth (Fig. 3E). Collectively, SPTBN1 could retard the G1/S progression and suppress the proliferation of ccRCC cells both in vitro and in vivo.

In order to clarify the underlying mechanism of SPTBN1 involved in RCC cell proliferation, we performed RNA-sequencing on shNC and shSPTBN1 786-O cells. A total of 4491 up-regulated genes and 4179 down-regulated genes were identified (Fig. 4A). Gene ontology (GO) functional enrichment analysis revealed their enrichment in metabolic process-related pathways (Fig. 4B). Subsequently, we performed gene set enrichment analysis (GSEA) using both TCGA-KIRC and our RNA-seq dataset. Interestingly, Reactome Glucose Metabolism and Reactome Glycolysis pathways were significantly associated with the biological function of SPTBN1 (Fig. 4C). To verify this hypothesis, we detected the expression level of rate-limiting enzymes involved in glycolytic process. Unexpectedly, glycolysis-associated genes (PKM, HKII, PFKFB3, ENO2, SLC2A1, PFKP, and LDHA) were dramatically up-regulated after SPTBN1 knockdown, whereas SPTBN1 overexpression reversed this influence (Fig. 4D–G). Consequently, we observed that knockdown of SPTBN1 significantly enhanced lactate concentrations (Fig. 4H), glucose consumption rates (Fig. 4I) and ATP levels (Fig. 4J) in ccRCC cell lines, while SPTBN1 overexpression substantially reversed this increase (Fig. 4H–J).

Ultimately, supplementation with 2 mM 2-deoxy-D-glucose (2-DG), a specific inhibitor of glycolysis, could partially block glycolysis-induced proliferation of ccRCC cells. SPTBN1 knockdown dramatically prevented 2-DG from suppressing 786-O and Caki-1 cells. The half-maximal inhibitory concentration (IC50) data revealed that SPTBN1 knockdown increased the IC50 value and weakened the sensitivity to 2-DG, but SPTBN1 overexpression displayed contradictory effects (Fig. 4K). Taken together, that 2-DG could partially repress the malignant proliferation of SPTBN1-knockdown ccRCC cells (Fig. 4L).

We next explored the downstream genes of SPTBN1 in the glycolytic pathway. We screened out the GPT2 gene based on log₂FC (Fold Change) and P-value (Fig. 5A). qRT-PCR experiments further confirmed that GPT2 expression was significantly elevated in SPTBN1-knockdown 786-O and Caki-1 cells, while GPT2 mRNA levels fell significantly after SPTBN1 overexpression (Fig. 5B, C). The correlation analysis in the TCGA-KIRC database indicated that the expression levels of SPTBN1 and GPT2 were negatively correlated (Pearson: R = − 0.176, P < 0.001; Spearman: R = − 0.203, P < 0.001; Fig. 5D). In addition, the same results were observed in NJMU ccRCC cohort (N = 53; Pearson: R = − 0.703, P < 0.001; Spearman: R = − 0.668, P < 0.001; Fig. 5E). TMA immunohistochemical data also revealed a negative correlation between SPTBN1 and GPT2 (NJMU ccRCC TMA cohort2: N = 180; Pearson: R = − 0.274, P < 0.001; Spearman: R = − 0.292, P < 0.001; Fig. 5F, G).

Subsequently, we evaluated the expression and prognostic value of GPT2 in ccRCC patients. qRT-PCR confirmed that GPT2 mRNA was obviously up-regulated in NJMU ccRCC cohort (N = 53, P < 0.001; Fig. 6A). Likewise, GPT2 up-regulation was observed in IHC (Additional file 1: Figure S10A). Meanwhile, qRT-PCR and WB showed that GPT2 expression increased significantly in various types of CCRCC cells, compared with that in HK2 cells (Fig. 6B). The ROC curve analysis showed an AUC of 0.816 (95%CI = 0.770–0.861; Fig. 6C). Kaplan–Meier survival analysis indicated that the prognosis (OS, DSS, and PFI) of patients with high GPT2 expression was significantly worse (OS: P = 0.014; DSS: P = 0.022; PFI: P = 0.003; Fig. 6D), demonstrating the excellent prognostic value of GPT2. According to our tissue microarray from NJMU TMA cohort (N = 180), we divided patients into high- and low-GPT2 expression subgroups. The results showed that high GPT2 expression predicted longer overall survival and relapse-free survival (P < 0.01; Fig. 6E). CCK-8 assay demonstrated that GPT2 knockdown significantly decreased cell proliferation (Fig. 6F). Colony-formation efficiency was reduced after GPT2 depletion (Fig. 6G). GPT2 knockdown significantly suppressed lactate concentration, glucose consumption rate and ATP level in ccRCC cell lines (Fig. 6H, I). In conclusion, GPT2, as a downstream gene of SPTBN1, promoted ccRCC progression via glycolysis.

Considering the interaction between SPTBN1 and GPT2 we ascertained above, we suspected whether depletion of SPTBN1 could alter GPT2 expression. The mRNA expression level of SPTBN1 was negatively correlated with that of GPT2, indicating SPTBN1 regulated GPT2 at the transcriptional level. Given that SPTBN1 is known as an RNA-binding protein, we hypothesized that SPTBN1 functions as an RBP binding to the 3’UTR region of GPT2 [64, 65]. Control and SPTBN1-deficient 786-O and Caki-1 cells were stimulated with actinomycin D for 0, 2, 4, 6, 8, and 10 h. Then, mRNA stability was monitored using qRT-PCR. This analysis revealed that SPTBN1-depleted cells predominantly contained more stable GPT2 mRNAs than control cells, whereas SPTBN1 overexpression suppressed GPT2 mRNA level in 786-O and Caki-1 cells (Fig. 7A). Congruently, RIP experiments in 786-O and Caki-1 (with SPTBN1 overexpression) cells further demonstrated that SPTBN1 could bind to GPT2 (Fig. 7B). In the dual-luciferase reporter assay, 786-O and Caki-1 CCRCC cells were transfected with pEZX-FR02 plasmid which contained mutated GPT2 3’-UTR region (GPT2-MUT) and wild type of GPT2 3’-UTR region (GPT2-WT). Compared with mutated GPT2 3’-UTR reporter plasmids, SPTBN1 could significantly reduce the luciferase activity of the wild-type GPT2 3’-UTR, which was in concordance with our hypothesis (Fig. 7C). These findings confirmed that SPTBN1 could function as an RNA-binding protein to regulate the stability of GPT2 mRNA by binding to its 3’-UTR regions.

We investigated whether silencing SPTBN1 can reverse the malignant proliferation of ccRCC cells through directly modulating GPT2. CCK-8 assay revealed that GPT2 knockdown attenuated the oncogenic effect of SPTBN1 knockdown (Fig. 7D). Likewise, GPT2 knockdown abolished the ability of SPTBN1 in promoting colony formation (Fig. 7E). It was further validated that GPT2 knockdown could partly reduce ATP concentration, lactate production, and glucose uptake, implying that the glycolysis caused by SPTBN1 depletion in ccRCC cells (Fig. 7F, G). Collectively, SPTBN1 suppressed ccRCC development through degrading GPT2 to reprogram glyolysis (Fig. 8).