Background
Renal cell carcinoma (RCC) is diagnosed in up to 300 000 people worldwide each year, with clear cell RCC (ccRCC) as the predominant histological subtype (∼80% of all RCCs) [
1‐
3]. ccRCC originates from the epithelial cells of the proximal convoluted tubule in the nephron and the inactivation of the
von Hippel Lindau (VHL) gene is the early event to drive the disease pathogenesis [
4,
5]. VHL serves in the E3 ubiquitin ligase complex, which mediates α subunits of hypoxia-inducible factors (HIF1α and 2α) for proteasomal degradation, while the VHL inactivation leads to the stabilization of HIFαs even under normoxia, thereby resulting in pseudohypoxic phenotype and metabolic reprogramming, including aberrant glycolysis, nucleotide and lipid biosynthesis [
5,
6]. In addition, alterations also occur frequently in other genes encoding metabolic enzymes [
6]. All these aberrations together give rise to the abnormal increase of certain metabolites with oncogenic function (so-called oncometabolites). For instance, the loss of the
L-2-hydroxyglutarate dehydrogenase (L2HGDH) gene, and gain of the
malate dehydrogenases (MDHs) or
lactate dehydrogenases (LDHs) genes result in the accumulation of L-2-hydroxyglurate (L-2-HG), a
bona fide oncometabolite that inhibits a group of enzymes of α-ketoglutarate-dependent dioxygenases including the ten-eleven translocation (TET) family of 5-methylcytosine (5mC) hydroxylases, and histone lysine or RNA demethylases [
6‐
10]. The L-2-HG-mediated inhibition of TET activity triggers widespread DNA cytosine hypermethylation in ccRCC, thereby promoting aggressive disease [
6,
7,
11,
12]. However, it remains poorly defined how such aberrant DNA hypermethylation contributes to ccRCC progression and which genes are targeted by L-2-HG.
GA-binding protein A (GABPA) is the ETS family transcription factor, and it forms a complex with its partner either GABPB1 or GABPB2 to regulate target gene transcription [
13,
14]. GABPA has long been shown to play oncogenic roles in the pathogenesis of leukemia, prostate cancer, glioblastoma and other malignancies [
13,
15‐
17]. More recently, GABPA was further identified as a key transcription factor to activate the mutated telomerase reverse transcriptase (TERT) promoter [
18]. TERT promoter mutations occur in various types of cancer and create de novo ETS binding motifs recognized by the GABPA-containing complex [
18]. The interplay between the mutated TERT promoter and GABPA induces the expression of TERT, a rate-limiting component for telomerase activity required for immortalization and malignant transformation [
14,
18‐
20]. In glioblastoma cells harboring the TERT promoter mutation, GABPB1 depletion led to reduced TERT expression coupled with diminished telomerase activity followed by progressive telomere attrition and eventual loss of oncogenic potential [
17]. Given all the above findings, GABPA or its partner GABPB1 has been suggested as a therapeutic target for tumors bearing TERT promoter mutations [
17].
TERT promoter mutations occur in up to 20% of RCCs [
4,
21], and it is currently unclear which effects GABPA exerts on ccRCC. The present study is designed to address this issue by combining the ccRCC-specific metabolic reprogramming. By doing so, we observe that GABPA acts as a tumor suppressor by stimulating TGFBR2 transcription and TGFβ signaling, while the oncometabolite L-2-HG epigenetically inhibits GABPA expression, disrupting the GABPA-TGFβ loop to drive ccRCC aggressiveness. Clinically, higher GABPA is significantly associated with longer survival of ccRCC patients. These findings may have important implications in understanding ccRCC pathogenesis and precision oncology.
Methods
Patients and specimens
The present study included 31 patients with ccRCC whose tumors and noncancerous adjacent renal tissues (NTs) were available. NT samples were examined to exclude cancer cell contamination or local metastasis. Specimens were freshly frozen at -80 °C until use. Clinical information for this cohort of patients was summarized in Table S
1. In addition, the tissue microarray (TMA), containing 90 ccRCC tissues together with patient clinical characteristics and follow-up data on OS (endpoints: dead or alive) (Table S
2), was obtained from Shanghai Outdo Biotech (Shanghai, China). The study was approved by Shandong University Qilu Hospital Ethics Committee.
The Cancer Genome Atlas (TCGA), GEO and EMBL-EBI cohorts of ccRCC and GEO ChIP-sequencing
The data for the TCGA cohort of ccRCC were downloaded via
https://portal.gdc.cancer.gov in Jan. 2021 [
22]. Patient clinical characteristics were presented in Table S
3. mRNA abundances were expressed as RSEM (RNA-Seq by Expectation Maximization). DNA methylation was expressed as β values (the ratio of signal intensity between methylated and unmethylated CpGs). GSEA and KEGG analyses were performed using the ‘clusterProfiler’ R package.
GSE96015 and GSE105431 data were downloaded from GEO datasets, which contain the GABPA ChIP-seq results obtained from HepG2 cells (GSM2527661 and GSM2527662) and K562 cells (GSM2825959 and GSM2825960), respectively [
25].
Cell lines and cell culture
ccRCC-derived cell lines A498 (purchased from American Type Culture Collection, Manassas, VA) and 786-O cells (Shanghai Institute of Biochemistry and Cell Biology Cell Bank/Stem Cell Bank, Shanghai, China) were used in the present study. Cells were cultured in RPMI-1640 medium (Thermo Fisher Scientific, Waltham, MA) supplemented with 10% FBS (Thermo Fisher Scientific), 100 U/ml penicillin, 100 μg/ml streptomycin and 4 mM L-glutamine. Cells were analyzed for mycoplasma infection every six months.
Pyrosequencing for DNA methylation analyses
Genomic DNA was extracted using QIAamp DNA blood Mini Kit (Qiagen, Hilden, Germany) and then converted by Sodium Bisulfite using EpiTect Bisulfite Kits (Qiagen, Hilden, Germany). PCR amplification was performed with GABPA methylation region-specific primers. The PCR product was purified by binding to streptavidin-coated Sepharose beads (GE Healthcare, UK), denaturation and washing. Then the sequencing primer was annealed to the purified PCR fragment followed by pyrosequencing in a PyroMark Q96 (Qiagen). The primer sequences are listed in Table S
4.
5-hmC assays
Quest 5-hmC DNA ELISA kits (Zymo Research) were used to assess 5-hmc levels according to the manufacturer’s protocol. Anti–5-hydroxymethylcytosine polyclonal antibody was pre-coated on the bottom of wells, and 100 ng genomic DNA was denatured and then added. Anti-DNA HRP antibody and HRP developer were employed to detect DNA bound to the anti–5-hmC Ab. Greenish-blue color was analyzed in the wells by a plate reader at 405- to 450-nm detection. The levels of 5-hmC were expressed as absorbance.
Cell proliferation analyses
Proliferation of ccRCC-derived cells was monitored and analyzed every 8 h for a total of 72 h using IncuCyte S3 Live-Cell Analysis System (Essen Bioscience, Ann Arbor, MI, USA). The changes of phase area confluence represent the cell proliferation.
Flow cytometry
For cell cycle analyses, cells were fixed with 70% ethanol at + 4 °C overnight and stained with RNAse A (0.5 μg)-containing Propidium Iodide (50 μg/ml). Cell cycle distribution was determined using flow cytometry with ModFit (BD Biosciences, Franklin Lakes, NJ). The assessment of ccRCC stem cell markers CD44 and CD105 was performed by staining cells with FITC-conjugated anti-CD44 (BioLegend) and anti-CD105 (BD Biosciences) antibodies, and fluorescence signals were measured as the expression level of CD44 and CD105, respectively. CD90 expressed was assessed as above using CD99 antibody (BioLegend).
Cells (3000/well) were cultured in ultra-low-attachment 96-well plates (Corning Life Sciences) with 100 μl RPMI-1640/10 mM HEPES serum-free medium supplemented with cocktails of following growth factors: 10 ng/mL bFGF (PeproTech Nordic, Stockholm, Sweden) and 20 ng/mL EGF (PeproTech Nordic). Fresh medium was supplemented every 3 days. Fifteen days later, the spheroid colonies were examined under light microscopy, counted and photographed. Single spheroid of ccRCC-derived cells was monitored and analyzed by using IncuCyte 3D Single Spheroid Assays (Essen Bioscience, Ann Arbor, MI, USA). The single spheroid was examined under the IncuCyte, measured and photographed.
Wound healing assays
Migration of ccRCC-derived cells was monitored and analyzed every 8 h for a total of 40 h using Incucyte Cell Migration Scratch Wound Analysis (Essen Bioscience, Ann Arbor, MI, USA). The changes of the wound region represent the cell migration.
Matrigel-coated invasion assays
Fifty μl of matrigel (Corning Life Sciences, Flintshire, UK) was first loaded to the bottom of the upper chamber. The cell suspension containing 5.0 × 104 cells/ml was prepared in the serum-free medium and then seeded into the upper chamber. The low chamber contained RPMI-1640 medium with 20% FBS. Cells were incubated at 37 °C for 48 h. Cells invading through the matrigel were stained with crystal violet, counted under the microscope and photographed.
Total RNA was extracted with Trizol-Reagent (Thermo Fisher Scientific), and reversely transcribed using High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific). qPCR was performed in QuantStudio 7 Flex Real-Time PCR System using SYBR Green (Thermo Fisher Scientific). Levels of target mRNA were calculated based on the ΔCT values and normalized to human β2-M expression. Primers used in this study are documented in Table S
4.
RNA-sequencing
Total RNA was extracted from the GABPA-knockdown and control 786-O cells, and a cDNA library was prepared according to the standard Illumina RNA-seq protocol. The raw data was available from GSE165728. FeatureCounts v1.5.0-p3 was used to count the read numbers mapped to each gene. A fold change > 1.5 and FDR < 0.05 were set as the thresholds for identifying DEGs.
Western blot analysis
Proteins were extracted using Pierce RIPA Buffer (Thermo Scientific) with 1% Phenylmethanesulfonyl fluoride (Sigma-Aldrich) and quantified with DC Protein Assay (Bio-Rad). Thirty µg of proteins were separated in Mini-PROTEAN TGX Gels (Bio-Rad) and transferred to PVDF membranes using Trans-Blot Turbo Transfer Pack (Bio-Rad). Membranes were blocked with 5% non-fat milk diluted in TBST, and then incubated with primary antibodies and secondary antibodies before imaged with Clarity Max Western ECL Substrate (Bio-Rad, 1,705,062) and ChemiDoc MP Imaging System (Bio-Rad). The following primary antibodies were used: GABPA (Santa Cruz), L2HGDH (Novus), LDHB (Santa Cruz), MDH2 (Santa Cruz), cMYC (Santa Cruz), TGFBR2 (Abcam), SMAD2/3 and p-SMAD2/3 (Sigma-Aldrich), CDKN1A (Cell Signaling Technologies), CDH1 (Cell Signaling Technologies), vimentin (Cell Signaling Technologies) and Actin (Santa Cruz). Secondary antibodies include Goat Anti-Mouse IgG (H + L)-HRP Conjugate and Goat Anti-Rabbit IgG (H + L)-HRP Conjugate (Bio-Rad).
Immunofluorescence
Control and GABPA-depleted A498 and 786-O cells were treated with 4% paraformaldehyde followed by 5% BSA. The cells were then incubated with FITC-conjugated GABPA and Alexa Fluor® 594-conjugated TGFBR2 antibodies (Santa Cruz) at room temperature for 2 h, respectively. Cells were finally counterstained with Hoechst and imaged under a fluorescence microscopy.
TGFBR2 promoter constructs were made by Shanghai Integrated Biotech Solutions Co. Ltd. Putative GABPA binding sites on the TGFBR2 promoter regions were identified using the Consite software and its mutant variant made using a Site-Directed Mutagenesis Kits (Thermo Fisher Scientific). The above promoter reporters were transfected into A498 and 786-O cells, or co-transfected with GABPA expression vectors. Cells were then harvested, and luciferase activity was determined using a dual luciferase reporter assay system (Promega, Madison, USA). The target promoter-driven firefly luciferase activity was normalized to the renilla activity included in the kit.
Chromatin immunoprecipitation (ChIP)
SimpleChIP® Enzymatic Chromatin IP Kit (Cell Signaling Technologies) was used according to the protocol provided. In brief, cells were crosslinked with formaldehyde. Chromatin digestion was performed with micrococcal nuclease and analyzed by agarose gel. For GABPA binding to the TGFBR2 promoter, antibodies against GABPA (Sigma-Aldrich) were added into the digested samples and incubated overnight at 4 °C with rotation. For 5hmC at the GABPA methylation region spanning cg08521263, antibodies against 5hmC (Abcam) were added. IgG antibodies were included as negative controls in ChIP assays. Protein G magnetic beads were used to precipitate the DNA-antigen–antibody complex followed by the elution of chromatin from Antibody/Protein G magnetic beads and reversal of cross-links. Spin columns were used to purify DNA and the collected DNA was amplified using PCR with specific primers (Table S
4).
Immunohistochemistry (IHC)
Paraffin embedded slides were deparaffinized and rehydrated followed by antigen-retrieval using citric acid buffer. Endogenous peroxidase was deactivated by H2O2. Slides were blocked using 10% goat serum and incubated with the corresponding primary antibodies overnight at 4 °C. After incubation with secondary antibodies for 45 min at room temperature, DAB staining (Thermo Fisher Scientific) was used to detect the antigen–antibody binding. The primary antibodies used were: GABPA (ProteinTech) and TGFBR2 (Abcam). The slides were examined by two of the co-authors (ZF and NZ) and mean values of GABPA and TGFBR2 positive cells were presented based on the results from two observers. For each slide, a total of 200 cells in two fields were analyzed.
Nude mice and tumor cell injection
Six-week-old male athymic BALB/c nude mice were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd./Charles River Laboratories, Beijing, China, and used to evaluate the in vivo effect of GABPA on metastasis and growth. Two millions of 786-O cells expressing ectopic GABPA (786-O/GABPA) and control cells with empty vectors (786-O/Control) were injected into nude mice via the tail vein (4/group) and subcutaneously (10/group), respectively. Mice via vein injection were killed after 8 weeks and lungs were collected for evaluation of tumor seeding or metastasis. Mice with subcutaneous injection were monitored for tumor growth, and when tumors were palpable, their sizes were measured every 4 days for 28 days. Mice were then killed, and tumors were isolated for growth and IHC analyses. The study was approved by Shandong University Qilu Hospital Ethics Committee.
Statistical analyses
All statistical analyses were performed using IBM SPSS Statistics version 24 (IBM, Armonk, NY). Based on the distribution of data, Student’s t-test, Mann–Whitney U-test, Kruskal Wallis test and Chi2-test or Fisher’s exact test were used for analysis. Spearman’s Rank-Order Correlation coefficient was applied to determine correlation coefficients R. Survival analyses were performed with log-rank test. Overall survival (OS) disease-free survival (DFS) was visualized with Kaplan–Meier plots. Univariate and multivariate Cox regression analyses were used to determine the effect of various predictor variables on OS and PFS. p-values < 0.05 were considered as statistically significant.
Discussion
ccRCC as a metabolic disease exhibits extensive metabolic reprogramming including downregulation of the tricarboxylic acid (TCA) cycle and upregulation of fatty acid synthesis, the pentose phosphate pathway and glutamine transporters, thereby giving rise to aggressive disease and poor prognosis [
6,
28]. The aberrant accumulation of L-2-HG, a
bona fide oncometabolite, has been shown as an important driver for ccRCC progression [
6]. Indeed, approximately 1/3 of ccRCC patients already have distant dissemination at presentation, which is in general resistant to conventional chemotherapy and radiotherapy [
29]. Although numbers of drugs targeting cellular/molecular pathways have been applied for ccRCC, but patient response to these treatments is limited. Thus, the delineation of ccRCC pathogenesis and identification of new therapeutic targets is demanding tasks. Here we observed a widespread downregulation of GABPA expression in ccRCC and identified TGFBR2 as a direct target of GABPA through which the TGFβ signal was activated to inhibit ccRCC progression. Moreover, GABPA is the downstream effector epigenetically silent by L-2-HG, which disrupted the GABPA-TGFβ loop. Restoring GABPA expression strongly inhibits in vivo metastatic and carcinogenic abilities of ccRCC cells. These results exemplify how oncometabolites rewires a GABPA-mediated signaling and erases its tumor suppressive function for cancer development/progression (Fig.
7G).
TGFBR2 is an established tumor suppressor in several human malignancies [
26,
27], however, its role in ccRCC is unclear, although its downregulation has long been observed in ccRCC [
30,
31]. The
TGFBR2 gene is localized at chromosome 3p24.1 where LOH is widespread in ccRCC [
22]. Intriguingly, the
TGFBR2 LOH does not affect its expression (Fig. S
8), indicating that other regulatory mechanisms are more important. In the present study, we provide strong evidence demonstrating TGFBR2 as the direct target of GABPA: (i) GABPA knockdown and over-expression down- and up-regulates TGFBR2 expression, respectively; (ii) GABPA knockdown and over-expression inhibits and stimulates TGFBR2 promoter activity, respectively; moreover, the GABPA binding motif mutation leads to a dramatic reduction in the basic TFGBR2 promoter activity, indicating a key effect of GABPA on TGFBR2 transcription; (iii) The ChIP assay shows the occupancy of GABPA on the TGFBR2 promoter. Similar results were also obtained from the analysis of ChIP-seq databases. TGFBR2 exhibits a tumor suppressive function in ccRCC and attenuates TGFβ response. Consistently, GABPA depletion significantly inhibits the response of ccRCC cells to TGFβ treatment, which includes diminished SMAD2/3 phosphorylation and increased MYC expression, while reduced CDKN1A expression. These TGFβ downstream effectors consequently result in altered proliferation and invasion of ccRCC cells.
2-HG exists as two isoforms including L-2-HG and D-2-HG, and both and α-KG can be converted with each other by specific enzymes [
6]. Isocitrate dehydrogenases (IDH1/2) and MDH1/2 or LDHB catalyze α-KG into D-2-HG and L-2-HG, respectively, whereas D2HGDH and L2HGDH oxidize D-2-HG and L-2-HG to α-KG [
6]. Gain-of-function of IDH mutations occurs most frequently in low-grade glioma, cartilaginous tumors, intrahepatic cholangiocarcinoma, and certain hematological malignancies; and thus, the accumulation of D-2-HG occurs in these tumors [
6]. In contrast, IDH mutations are rare in ccRCC [
22]. Instead, loss of L2GHDH or gain of MDHs and LDHB is widespread, which leads to the generation of excess L-2-HG [
10,
12,
22]. In addition, pseudohypoxic phenotype-mediated metabolic reprogramming may also contribute to increased L-2-HG [
32,
33]. α-KG is a co-substrate for α-KG/Fe(II)-dependent dioxygenases, while L-2-HG acts as an α-KG antagonist to competitively inhibit α-KG/Fe(II)-dependent dioxygenases among which are the TET family of 5mC hydroxylases [
6]. The diminished TET activity subsequently results in aberrant DNA methylation in ccRCC [
12]. Our findings demonstrate a causal relationship between GABPA hypermethylation/gene silencing and L-2-HG accumulation; the GABPA hypermethylation is coupled with reduced 5hmC, demonstrating impaired TET activity by L-2-HG in ccRCC cells. Of note, ectopic L2HGDH expression or MDH2 and LDHB depletion robustly increased 5hmC, but only inhibited the methylation of cg08521263 moderately. Likely, the demethylation of other CpGs at the
GABPA loci synergizes with the demethylated cg08521263 to promotes GABPA expression. Indeed, there exist other 14 CpGs at the
GABPA loci, and 12 of them are largely unmethylated, while the rest two CpGs (cg00106744 and cg21890848) are methylated at low
/intermediate levels in tumors. Further studies are required to thoroughly elucidate the role of L-2-HG-mediated methylation in inhibiting GABPA expression. In addition, we recognize the global effect of both 5-AZA and L-2-HG on DNA methylation and 5hmC levels, and specific approaches such as dCas9-TET combined with GABPA promoter targeting have been planned to address these issues in our future studies.
Although we identified that the GABPA-TGFBR2 nexus plays a key suppressive role in ccRCC aggressiveness, it remains possible that GABPA may restrain aggressive ccRCC by regulating other targets. For instance, DICER1, a ribonuclease functioning in the microRNA (miRNA) processing machinery, is transcriptionally activated by GABPA through which the invasive phenotype and metastasis of thyroid cancer are inhibited [
34]. In bladder cancer, GABPA activates the
Fox1A and
GATA3 genes, thereby promoting cancer cell differentiation [
19]. Thus, further studies are required to thoroughly delineate various effects of GABPA on ccRCC progression. On the other hand, the strong association between GABPA and TGFBR2 is observed in many types of cancer based on the TCGA data analyses, which indicates a broad implication of the present findings in oncogenesis.
Paradoxically, GABPA has long been documented to act as an oncogenic factor, especially for its critical role in activating the mutated TERT promoter and inducing TERT expression [
13,
14,
18]. TERT, as the catalytic component of telomerase, is not only required for infinite proliferation of cancer cells by stabilizing telomere length, but also operative in promoting invasion, metastasis and other cancer hallmarks via a telomere lengthening-independent function [
19,
20]. Indeed, glioblastoma cells knocked-out of GABPB1, the GABPA partner required for activation of the mutated TERT promoter, was shown to undergo telomere shortening, senescence or apoptosis and eventual loss of tumorigenesis [
17]. Therefore, targeting GABPB1 or GABPA for telomerase-based therapy has been suggested as a novel anti-cancer strategy [
17,
35]. However, evidence has recently accumulated that GABPA function may be context-dependent, and it acts as a tumor suppressor in several types of cancer, despite its stimulatory effect on TERT transcription [
19,
34,
36‐
39]. Similarly, GABPB1 downregulation due to the gene hypermethylation is observed in thyroid carcinoma and its inhibition promotes invasion of thyroid cancer cells, too [
40]. These unprecedented findings suggest that it should be cautious in targeting GABPA or GABPB1 for cancer intervention.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.