Integrated genomic and proteomic analyses identify PYGL as a novel experimental therapeutic target for clear cell renal cell carcinoma

Abstract

Sunitinib, the first-line targeted therapy for metastatic clear cell renal cell carcinoma (ccRCC), faces a significant challenge as most patients develop acquired resistance. Integrated genomic and proteomic analyses identified PYGL as a novel therapeutic target for ccRCC. PYGL knockdown inhibited cell proliferation, cloning capacity, migration, invasion, and tumorigenesis in ccRCC cell lines. PYGL expression was increased in sunitinib-resistant ccRCC cell lines, and CP-91149 targeting the PYGL could restore drug sensitivity in these cell lines. Moreover, chromatin immune-precipitation assays revealed that PYGL upregulation is induced by the transcription factor, hypoxia-inducible factor 1α. Overall, PYGL was identified as a novel diagnostic biomarker by combining genomic and proteomic approaches in ccRCC, and sunitinib resistance to ccRCC may be overcome by targeting PYGL.

1. Introduction

Age-standardized incidence rate for kidney cancers was 49.6 per 100,000 men and women, and 4.05 million disability-adjusted life years associated with kidney cancer was recorded worldwide in 2019 [1]. Renal cell carcinoma (RCC) is mainly divided into five major tissue subtypes of clear cell, papillary, chromophobe, collecting duct, and unclassified RCCs [2,3]. Clear cell renal cell carcinoma (ccRCC) stands out as the most prevalent subtype, constituting approximately 75% of cases [4]. Localized RCC could be treated by surgery, such as partial or radical nephrectomy. However, about 30% of patients with localized ccRCC will eventually develop metastatic disease after nephrectomy [5]. In addition to surgery, there are also cytokine therapy, immunotherapy, and targeted drug therapy. Sunitinib, an inhibitor targeting VEGFR2, PDGFR-β, c-KIT, and FLT3, is the first-line targeted drug for the treatment of metastatic renal cell carcinoma (mRCC) [6]. However, 30% of patients with mRCC are intrinsically resistant to sunitinib, and the remaining 70% who respond to sunitinib initially will eventually become resistant to sunitinib and experience disease progression in 6–15 months [7].

The field of high-throughput screening measuring gene expression profiling, such as mRNA expression microarrays, RNA-sequencing (RNA-seq), and proteomic mass spectroscopy, is expanding rapidly. These advances have enhanced the understanding of cancer development and progression. Unlike the study of a single gene, protein, reaction, or metabolite, current technologies, such as genomics, proteomics, metabolomics, and bioinformatics, enable systematic reviews and meta-analyses that could improve our understanding of cancers. However, RNA-seq and microarrays demonstrate relatively low concordance in genome-wide gene expression profiling, particularly in low-abundance transcripts. Additionally, the correlation between gene expression at the mRNA level and its corresponding protein level is often inadequate in human diseases [8,9]. Based on the above factors, several recent studies have highlighted the important roles of integrated genomic and proteomic analyses of gene expression in cancer research [[10], [11], [12]]. Although some previous studies have used a single high-throughput gene expression analysis technique to study ccRCC [[13], [14], [15], [16]], combining genomic and proteomic analyses for the discovery of ccRCC biomarkers is still poorly studied.

In this study, the Gene Expression Omnibus (GEO) and The Cancer Genome Atlas (TCGA) databases were utilized to detect differentially expressed genes (DEGs) between ccRCC and normal kidney samples. Subsequently, the Human Protein Atlas (HPA), a comprehensive database leveraging antibody proteomics, was employed to identify potential ccRCC markers by examining protein expression profiles of numerous human proteins. Differentially expressed proteins detected by mass spectrometric technique were also analyzed from previous research [14]. We performed integrated genomic and proteomic analyses to identify novel tumor markers in ccRCC.

PYGL, a gene responsible for catalyzing the cleavage of alpha-1,4-glucosidic bonds, facilitates the release of glucose-1-phosphate from liver glycogen reserves [17]. We performed an integrated analysis of publicly available datasets and identified PYGL as a novel ccRCC-associated gene. PYGL expression was validated in this study, and its correlation with clinical features was examined. Furthermore, we studied the oncogenic functions of PYGL and found a strong positive correlation between high PYGL expression and sunitinib resistance. Mechanistically, PYGL may in part be regulated by hypoxia-inducible factor 1-alpha (HIF-1α) to exert its effects by promoting cancer cell survival and growth. Together, our findings revealed that PYGL could be a novel therapeutic target, and PYGL inhibition might reverse acquired resistance to sunitinib in ccRCC cells.

2. Material and methods

2.1. Gene expression profiles of ccRCC in GEO

Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih.gov/geo) is a public functional genomics data repository, which provides free distribution and shares access to comprehensive datasets [18]. We downloaded four microarray-based gene expression databases from GEO datasets to conduct an analysis of gene expression profiles in both ccRCC and normal kidney tissues.

2.2. DEGs of ccRCC in TCGA cohort

The Cancer Genome Atlas (TCGA; https://tcga-data.nci.nih.gov/tcga/) is an archive of tumor samples with detailed clinical information. The RNA-Seq data was obtained at TCGA for the kidney renal clear cell carcinoma (KIRC) projects and the DEGs were analyzed from those data.

2.3. RCC marker identified from the Human Protein Atlas (HPA)

The Human Protein Atlas (HPA) is a publicly available database (www.proteinatlas.org), which utilizes antibodies for immunofluorescence staining on cell lines (detecting the spatial distribution of genes at a subcellular level) and immunohistochemistry on tissue microarrays (exploring the distribution of the protein expression in normal and cancer tissues). There are 12 RCC samples determined in the HPA database and the immunohistochemistry staining of each antibody are classified as High, Medium, Low and Not detected. These proteins with high staining in ≥1/3 samples or medium staining in ≥2/3 samples were considered as highly expressed proteins in RCC. Furthermore, these proteins with no staining in ≥2/3 samples or low staining in ≥3/4 samples were defined as lowly expressed proteins in RCC.

2.4. Patient specimens

Human ccRCC and matching nontumor tissues were collected from patients treated with nephrectomy in Renji Hospital, shanghai with written informed consent. These samples used for research were approved by the ethics review committees of the institutional review boards of the Renji Hospital, Shanghai Jiao Tong University School of Medicine.

2.5. Quantitative real-time PCR and western blot analysis

Real-time PCR and Western blot assays were performed as described previously [19]. Details of primer sequences of primers were shown in Supplementary Table S1. β-actin and GAPDH were used as the loading control. The antibodies used for western blotting are listed in Supplementary Table S2.

2.6. Immunofluorescence staining

The following commercial primary antibodies were used for immunofluorescence: TMEM45A (1:100, Abcam, ab166899), and PYGL (1:200, Sigma, HPA000962). The detailed protocol was described previously [19].

2.7. Immunohistochemistry (IHC) staining of tissue microarray (TMA)

Tissue microarray (TMA) were prepared by Shanghai Outdo Biotech Co.,Ltd. PYGL antibody used for IHC staining was the same as immunofluorescence. Staining scores were based on the staining intensity and the percentage of cells that stained positively. The staining intensities were classified as 0 (total absence of staining), 1+ (weak staining), 2+ (moderate staining), and 3+ (strong staining). The percentage of positively stained cells were divided into four categories: 0 = 0%, 1 < 25%, 2 = 26–50%, and 3 > 50%. The final staining scores were calculated by multiplying the percentage of positive cells by the intensity and the scores ranged from 0 to 9.

2.8. Cell culture

All cell lines were purchased from ATCC and supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin at 37° in humidified 5% CO₂ atmosphere. 293T was cultured in DMEM medium, while other cell lines were cultured in 1640 medium. The hypoxic conditions are achieved by adding different concentrations of cobalt chloride (CoCl₂ · 6H₂O, Sigma Aldrich 769495-100G) or by flushing the hypoxic chamber (Billups Rothenberg) with a gas mixture of 94% N₂, 5% CO₂ and 1% O₂.

2.9. Plasmids, transfection, and lentivirus

PYGL shRNA was cloned into Tet-regulated lentiviral expression vector LT3REVIR [20]. For virus packaging, the packaging plasmids psPAX2 and pMD2.G were co transfected into 293T cells with either the experimental group plasmid LT3GEPIR-shRNA-PYGL (Supplementary Table 3) or the control group plasmid LT3GEPIR. According to the previous description, lentivirus production was achieved through a method based on polyethyleneimine [21]. 16–18 h after transfection, replace fresh DMEM medium with 10% FBS. After 48 h of transfection, the lentivirus can be obtained by 0.45-μm membrane (Merck Millipore) filtration of the culture medium. Infect ccRCC cells with collected lentivirus and then pass through 4 μg/ml puromycin for 7 days until stable transfected cells are produced.

2.10. MTT Cell proliferation assays

Cells were seeded in 96-well plates (2 × 10³ cells/well) and incubated for 24–120 h. After the incubation period, 10 μl of the MTT labeling reagent (A600799-0001, Sangon Biotech, Shanghai) at a final concentration of 0.5 mg/mL was added to each well. Incubate the 96-well plates for 4 h and add 150 μL DMSO into each well. The absorbance at 490 nm was measured by a microplate reader (ThermoFisher, Waltham, MA, USA).

2.11. Cell cycle analysis

Cell cycle analysis was performed according to the manufacturer's instructions (Propidium Iodide (PI) RNase staining buffer, BD PharMingen; #550825). Fluorescent intensities were acquired on a FACSCalibur (BD Biosciences) and the percentage of population in each phase of the cell cycle was calculated by ModFit LT 5.0 cell cycle analysis software.

2.12. Transwell migration and invasion assay

Cell migration and invasion assays were measured using transwell chambers with membrane pore size of 8.0 μm. Add 200 ng/mL matrix gel (BD Biosciences, Franklin Lakes, NJ, USA) to the chamber for measuring invasion. Inoculate cells into the upper chamber (3422, Corning; 1x105 cells/well). Add RPMI-1640 medium containing 10% FBS to the bottom chamber. After 48 h of cultivation, the invading cells adhered to the surface below the membrane were fixed with 4% paraformaldehyde and stained with 0.1% crystal violet. Observe and take photos of invading cells through a microscope.

2.13. Chromatin immunoprecipitation assay (ChIP)

ChIP assay was performed using a SimpleChIP Enzymatic Chromatin IP Kit (Cell Signaling) according to the manufacturer's instructions. The PCR primers are indicated in Supplementary Table S1.

2.14. In vivo xenograft assay

Mix ccRCC cancer cell suspension (2 × 10⁶ cells) with Matrix in a 1:1 ratio and inject it subcutaneously on the right side of 6-week-old male BALB/C nude mice (Nanjing Cavans Biotechnology Co., Ltd., China). Starting from 14 days after vaccination, measure and record the volume of tumors every 3 days. Tumor volume = (tumor length x tumor width 2)/2. Collect tumors and take photos 54 days after vaccination. All mice were raised under specific pathogen free conditions. All animal experiments have been approved by the Animal Experiment Ethics Committee of South China University.

2.15. Drug treatment

Sunitinib and PYGL inhibitor CP-91149 were obtained from MCE (Neodesha, KS, USA). ccRCC cells were seeded into 96-well plates. 12 h later, the cells were treated with different concentrations of sunitinib, CP91149, and sunitinib in combination with CP91149 or an equal volume of DMSO. Three days after this treatment, the optical density (OD) value of each well was determined using a spectrophotometer.

2.16. Pathway analysis

Obtain pathway analysis between protein-protein interaction (PPIs) networks and DEGs in the STRING database.

2.17. Drawing Venn diagrams

Venn diagrams were drawn using the Web-based utility Venn Diagram Generator (http://bioinformatics.psb.ugent.be/webtools/Venn/).

2.18. Statistical analyses

Perform statistical analysis using SPSS version 22.0. Whether the difference between the two groups is statistically significant is tested by Student's t-test. Survival analysis was carried out using the Kaplan-Meter method and compared by the log-rank test. Analyze the differences between multiple xenograft groups using one-way ANOVA. The value of P < 0.05 was considered statistically significant. ns = not significant (p > 0.05); *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001; ****p ≤ 0.0001.

3. Results

3.1. Identification of DEGs in GEO and TCGA datasets

A workflow of the bioinformatics analysis is shown in Fig. 1A to illustrate the basic ideas of this study. The schematic flow chart displays that multiple microarray datasets containing four GEO databases, a TCGA cohort, the HPA database, and Atrih's study data were comprehensively analyzed to identify potential ccRCC biomarkers. More information about the databases included in our study is presented in Supplementary Table S4.

Fig. 1.

Integrates multi-omics datasets used to identify novel markers of ccRCC. A, Schematic flow chart demonstrating databases and techniques used to identify ccRCC biomarkers. B, The number of samples was included among four GEO databases and TCGA cohort. C and D, The total number of genes and DEGs detected in four GEO databases and TCGA cohort, respectively. “FC” is an abbreviation of fold change. E and F, Venn diagrams showing the overlap of high-1000 and low-1000 ranked genes among five datasets. G, Heatmap were used to show 144 highly expressed genes from TCGA cohort, which were overlapped by five datasets.

GEO datasets and a TCGA cohort were analyzed with bioinformatics tools to identify candidate biomarkers in ccRCC. Fig. 1B and C shows the number of samples and genes detected among the multiple microarray datasets and TCGA cohort, respectively. The number of DEGs is displayed in Fig. 1D. The figure reveals that the datasets and cohort were different from each other in terms of mining DEGs in ccRCC. The top 1000 upregulated and 1000 downregulated genes at each GEO database and TCGA cohort were selected for further analysis (Supplementary Table S5). Venn diagrams show that 144 highly expressed genes and 124 lowly expressed genes in ccRCC were shared by the datasets (Fig. 1E and F). Based on the transcriptomic profiling of TCGA, the DEGs shared by the GEO datasets and TCGA cohort were used to generate a heatmap of expression profiles between ccRCC and normal adjacent tissues (Fig. 1G). The results indicated that the overlapping genes were more likely to be seen as potential ccRCC biomarkers.

Given that 144 highly expressed genes were overlapped between the microarray datasets and TCGA cohort in the Venn diagram, these genes were used to explore the potential molecular mechanisms of ccRCC carcinogenesis. Protein–protein interaction (PPI) networks were constructed and pathway analysis was performed to reveal the molecular mechanism underlying ccRCC development. The results showed the top 20 pathways associated with ccRCC development, including phagosomes, cell adhesion molecules, chemokine signaling pathway, and HIF-1 signaling pathway (Supplementary Fig. S1).

3.2. Identifying biomarkers for ccRCC from the HPA and Atrih data

The HPA database was utilized to analyze the differentially expressed proteins distinguishing ccRCC from normal kidney samples to investigate the biomarkers identified by proteomics methods. RCC is a kidney cancer that originates from proximal tubular cells [22]. A hematoxylin–eosin-stained section of the renal specimen showed the basic structure and organization of the kidney (Fig. 2A). The number of proteins detected in renal glomeruli and tubules with different expression levels from those in the HPA database is shown in Fig. 2B and C, respectively. Fig. 2D displays the number of proteins with high or low expression in RCC (the definition of high or low expression in RCC is presented in the Methods section). RCC is derived from the proximal renal tubular epithelium [23,24]; hence, differentially expressed proteins were analyzed between kidney tubule and RCC tissues. Subsequently, Venn diagrams were used to identify the differentially expressed proteins of RCC in the HPA database. A total of 204 proteins with high expression and 3108 proteins with low expression in RCC were identified through comparisons with normal kidney tubules as shown in Fig. 2E and F, respectively. The protein levels of VIM and FABP7 in ccRCC cells were higher than those in normal kidney cells (Fig. 2G). This finding was consistent with previous reports [[25], [26], [27]]. ATP6V1B1 (ATPase H + Transporting V1 Subunit B1) was highly expressed in renal tubular epithelium while its expression was lost in ccRCC (Fig. 2G, right). What's more, Atrih data that used mass-spectrometric technique to identify potential diagnostic markers of RCC were presented, and it demonstrated that FABP7, VIM and PYGL were the potential biomarkers of ccRCC (Fig. 2H). Together, our results indicated that HPA database, as well as mass spectrometric method were effective methods for identifying ccRCC biomarkers.

Fig. 2.

Proteomics used to identify biomarkers for RCC. A, Basic diagram of normal kidney, which includes proximal tubule, distal tubule, and glomerulus. B and C, The Human Protein Atlas (HPA) includes the number of proteins with different expressions in kidney glomeruli and tubules, respectively. The level of protein expression/antibody staining is divided into four categories: High, Medium, Low, and Not detected. D, The number of proteins selected in RCC for biomarker screening. The proteins with high expression in ≥1/3 RCC samples or medium expression in ≥2/3 RCC samples were considered as those with high expression in RCC and included in the following differential analysis. Similarly, the proteins with no expression in ≥2/3 RCC samples or low expression in ≥3/4 RCC samples were considered as those with low expression in RCC. E and F, Venn diagrams show the number of highly expressed proteins and lowly expressed proteins in RCC, respectively. G, Potential biomarkers of RCC discovered by the HPA database. H, Quantitative proteomic revealing 409 differentially expressed proteins in ccRCC compared to normal kidney tissue samples. This data was derived from Atrih data.

3.3. Comparison of the differences between TCGA cohort and the HPA database in identifying ccRCC biomarkers

The DEGs of ccRCC determined in the TCGA cohort and the differentially expressed proteins detected in the HPA database were compared against each other to explore the correlation of the biomarkers identified by genomic and proteomic analyses. Among the 191 proteins with high expression in RCC from the HPA database, the levels of 74 protein expression were consistent with the corresponding mRNA levels detected in the TCGA cohort (Supplementary Table S6). Nevertheless, 84 proteins between the TCGA cohort and HPA database showed no remarkable correlations. Surprisingly, inverse correlations were found between mRNA and protein levels among 33 proteins. The results suggest that not more than 40% of proteins with high expression in ccRCC are consistent with the respective mRNA expression levels.

3.4. Diagnostic and prognostic value of TMEM45A and PYGL for ccRCC

Multiple datasets were used to identify ccRCC biomarkers; thus, we wanted to find the markers shared by the different datasets. Fig. 3A shows the highly expressed genes overlapping among the microarray datasets, TCGA cohort, Atrih data, and the HPA database. The figure demonstrated that the highly expressed genes shared by all datasets in ccRCC were FABP7, VIM, and PYGL. TMEM45A and PYGL were identified as novel ccRCC biomarkers according to the findings in Fig. 3A. Then, we explored the diagnostic and prognostic values of the two genes in ccRCC. TMEM45A mRNA expression had strong positive correlations with ccRCC development (Supplementary Fig. S2A–E), tumor size (Supplementary Fig. S2F), lymph node metastasis (Supplementary Fig. S2G), distant metastasis (Supplementary Fig. S2H), advanced tumor stage (Supplementary Fig. S2I), high-grade tumors (Supplementary Fig. S2J), residual tumor (Supplementary Fig. S2K), and shorter overall survival rates (Supplementary Fig. S2L). However, the absence of TMEM45A protein expression was observed in ccRCC from the HPA database (Supplementary Fig. S2M). The mRNA levels of TMEM45A in ccRCC and matched normal kidney tissues were determined by real-time PCR to further confirm the initial findings of the inverse correlation in TMEM45A mRNA and protein expression (Supplementary Fig. S2N). Immunofluorescence staining and Western blot analysis were also performed to investigate the levels of TMEM45A protein in ccRCC and matched normal kidney tissues (Supplementary Fig. S2O and P). The results further confirmed the presence of inverse correlations between TMEM45A mRNA and protein expression in ccRCC samples.

Fig. 3.

PYGL is a novel marker of ccRCC identified by multiple datasets. A, Highly expressed genes shared by Microarray Datasets, TCGA cohort, Atrih Data, and the HPA database, respectively. B–F, PYGL mRNA expression among four GEO databases and TCGA cohort in ccRCC compared with NAT (normal adjacent tissues). G, Protein levels of PYGL were compared between kidney and kidney cancer derived from the HPA database. H–O, The relationships between PYGL expression and clinical characteristics including tumor size, lymph nodes metastasis, distant metastasis, tumor stage, tumor residual, tumor grade, disease-free survival (DFS), and overall survival rates (OS), respectively.

Next, we investigated whether PYGL could be a novel therapeutic target for ccRCC. The mRNA and protein levels of PYGL considerably increased in human ccRCC samples compared with those in adjacent normal renal tissues (Fig. 3B–G). Unfortunately, PYGL mRNA levels had no correlations with tumor size (Fig. 3H), lymph nodes metastasis (Fig. 3I), distant metastasis (Fig. 3J), advanced tumor stage (Fig. 3K), residual tumor (Fig. 3L), high-grade tumors (Fig. 3M), shorter disease-free survival rates (Fig. 3N), and shorter overall survival rates (Fig. 3O).

3.5. Validation of PYGL expression and prognostic value in patients with ccRCC

The results of bioinformatics analysis indicated that PYGL is a good diagnostic marker but not a prognostic marker in ccRCC (Fig. 3B–O). We performed real-time PCR experiments with 14 pairs of freshly collected human ccRCC specimens and their matched normal tissues to further confirm these finding. The results demonstrated that PYGL was distinctively highly expressed in ccRCC (Fig. 4A). Similar results were obtained in the immunofluorescence staining assays and immunohistochemical analysis of tissue microarrays (Fig. 4B–E). Consistent with the bioinformatics analysis results (Fig. 3H–O), we found that the scoring of PYGL immunohistochemistry had no correlations with tumor size, lymph node status, overall TNM stage, histologic malignancy grading, and the poorer overall survival of patients with ccRCC (Fig. 4F–J). These results suggest that although increased PYGL expression may be correlated with ccRCC development, this gene does not directly promote ccRCC progression in our findings.

Fig. 4.

Association between PYGL expression, clinical characteristics, and prognosis in patients with ccRCC were validated in our laboratory. A-C, The level of expression of PYGL was measured by real-time PCR, immunofluorescence, and western blot analysis between ccRCC and NAT. D, Classification of samples according to the staining score of PYGL on tissue microarray of ccRCC. E, PYGL proteins levels compared between ccRCC and NAT. F-J, The correlation between PYGL expression and clinical features and prognosis in patients with ccRCC.

3.6. PYGL promotes ccRCC cells growth in vitro and in vivo

We detected the gene expression patterns of PYGL in several cell lines to investigate the biological function of PYGL in ccRCC (Fig. 5A and B). We then employed small hairpin RNA (shRNA) expression vector systems against PYGL in Caki-1 and ACHN cells. Real-time PCR and Western blot assay showed that shRNA targeting PYGL rearkably decreased the mRNA and protein expression levels of PYGL, respectively (Fig. 5C–E). Data showed that proliferation and clone formation ability were remarkably inhibited in Caki-1 and ACHN cells infected with PYGL-shRNA compared with the control (Fig. 5F–I). Next, the migration and invasion abilities of Caki-1 and ACHN cells were compared between PYGL-shRNA and the control. The results showed that migration and invasion were considerably reduced in the PYGL-shRNA group compared with the control group (Fig. 5J–M).

Fig. 5.

Knockdown of PYGL by shRNA inhibits the growth of ccRCC in vitro and in vivo. A-B, PYGL mRNA and protein expression in the five ccRCC cell lines. C-E, Quantitation shRNA-induced Knockdown by real-time PCR and western blot. Four different shRNA vectors directed against PYGL and one scrambled control were designed. These vectors were transfected into both Caki-1 cells and ACHN cells. F-G, Cell viability in the Control and PYGL-shRNA groups, as evaluated by the methyl thiazole tetrazolium (MTT) assay. H–I, Clonogenic assay of Caki-1 cells and ACHN cells transfected with control shRNA or PYGL-specific shRNAs (shPYGL-3, shPYGL-3). J-K, Transwell matrigel migration assays of Caki-1 cells and ACHN cells after transfected with negative control and PYGL-shRNAs. L-M, Transwell matrigel invasion assays of Caki-1 cells and ACHN cells after transfected with negative control and PYGL-shRNAs. N–O, Cell cycle analysis distribution between control shRNA and PYGL-shRNA groups in Caki-1 cells. P-Q, Growth of murine tumor xenografts with PYGL-specific shRNAs Caki-1 cells versus a scrambled control. Two-sided t-test was used to the last time point (p < 0.001). R, The photograph of xenograft tumors in control and PYGL-shRNAs group.

Cell cycle status was determined using flow cytometry. The results showed that shRNA-mediated PYGL downregulation increased the percentages of Caki-1 and ACHN cells in the G0/G1 phase (Fig. 5N–O and Supplementary Fig. S3). We performed xenograft tumor assays using Caki-1 cells stably transfected by PYGL-shRNA lentiviruses to verify the role of PYGL in ccRCC development in vivo. The data displayed that the PYGL-shRNA group had considerably lower tumor volume compared with the control group (Fig. 5P–Q). In addition, Western blot results confirmed that the protein expression of PYGL was decreased among the PYGL-knockdown xenograft tumors (Fig. 5R).

3.7. PYGL-mediated induction of the acquired resistance to sunitinib in ccRCC

Sunitinib, a multitargeted tyrosine kinase inhibitor, is the first-line therapy for metastatic ccRCC. Unfortunately, acquired resistance against sunitinib can develop in nearly all responsive patients and represents a major cause of treatment failure. Next, we evaluated the association between PYGL expression and the acquisition of a phenotype resistant to sunitinib in ccRCC. Two human renal cancer cell lines, namely, Caki-1 and ACHN, were treated with increasing doses of sunitinib for 2 years to develop sunitinib-conditioned cell lines, which were named Caki-1-R and ACHN-R, respectively (Fig. 6A and B). Interestingly, the data showed a positive correlation between acquired resistance to sunitinib and increased PYGL expression (Fig. 6C). Expectedly, short-term adjuvant sunitinib treatment could also promote PYGL expression PYGL in Caki-1, ACHN, and A498 cell lines (Fig. 6D–G and Supplementary Fig. S4A–C). The IC50 (the concentration of drug causing half inhibition of cell growth) of sunitinib was compared in the control shRNA and PYGL-specific shRNA-transduced cells to determine whether PYGL mediated the transition from sunitinib-sensitive to sunitinib-resistant phenotypes. Our results showed that the cells with PYGL downregulation demonstrated a remarkable decrease in IC50 values (Fig. 6H and I).

Fig. 6.

Targeting PYGL overcomes the sunitinib resistance ccRCC cell lines. A-B, Establishment of two sunitinib-resistant cell lines (Caki-1-R and ACHN-R) from human renal cell lines (Caki-1 and ACHN). C, The mRNA levels for PYGL were detected in sunitinib -resistant cell line, when compared to its sunitinib -sensitive counterpart. D-E, Caki-1 and ACHN cells were treated with different doses of sunitinib (0.25, 1, 4, and 8 μM) for 0.5, 2, and 4 h, and then the levels of PYGL mRNA were measured by real-time PCR. F, PYGL protein levels were examined in Caki-1 cells and ACHN cells after treatment with different doses of sunitinib for 12 h. G, PYGL protein levels were detected in Caki-1 cells and ACHN cells after treatment with different doses of sunitinib (3 μM)) at five different time points. H–I, shRNA-mediated knockdown of PYGL enhanced sensitivity to sunitinib in Caki-1 cells and ACHN cells. J-K, Drug synergy matrix of sunitinib and CP-91149 in Caki-1-R and ACHN-R cells. L-M, Dose-response curves were compared between sunitinib and the combination of sunitinib with CP-91149 in Caki-1-R and ACHN-R cells. N–O, The combination index (CI) of sunitinib and CP-91149 was calculated for CP-91149 in Caki-1-R and ACHN-R cells. Values of combination index between 0.7 and 1.0 represent slight synergism and values less than 0.3 represent strong synergism, while combination index values between 0.3 and 0.7 indicate synergism.

Four cell lines, namely, Caki-1-R, ACHN-R, Caki-1, and ACHN, were treated with sunitinib combined with CP-91149 (PYGL inhibitor) to explore the combinational effect of sunitinib and CP-91149 (Fig. 6J and K and Supplementary Fig. S4D–E). Quantal dose–response curves displayed that the combination treatment was more effective in killing cells than single-drug treatment (Fig. 6L–M and Supplementary Fig. S4F–G). Based on the median effect equation [28], the combination index value was calculated in ccRCC cell lines exposed to different drug combination schemes. One interesting observation is that the combined treatment had greater synergistic effects on the sunitinib-resistant cell lines (Caki-1-R and ACHN-R) than on the parent cell lines (Caki-1 and ACHN, Fig. 6N–O and Supplementary Fig. S4H–I). Furthermore, the combination of sunitinib and CP-91149 remarkably increased DNA damage and apoptosis in ccRCC cells compared with any single study (Supplementary Fig. S4J). These data suggest that high PYGL expression in ccRCC cells confers acquired resistance to sunitinib, and CP-91149 targeting PYGL could restore drug sensitivity in sunitinib-resistant ccRCC cell lines.

3.8. HIF-1α up-regulates the expression of PYGL through binding to PYGL promoter

GSEA was performed based on PYGL mRNA expression in the TCGA ccRCC cohort to further investigate the molecular mechanism underlying PYGL's potential involvement in ccRCC development. The GSEA results revealed that hypoxia and glycolysis signaling pathways were enriched in the higher PYGL expression group compared with the lower PYGL expression group (Fig. 7A and Supplementary Fig. S5A). Bioinformatics analysis of GEO datasets also identified that PYGL was associated with the hypoxia signaling pathway in renal epithelial cells and other cancer cell lines (Fig. 7B and C and Supplementary Fig. S5B). The schematic diagram in Fig. 7D shows the canonical HIF-1α-binding motif, and the sequence from the human PYGL promoter region exhibits potential HIF-1α binding sites (Fig. 7E and Supplementary Fig. S5C). 293T and Caki-1 cell lines were used to compare HIF-1α and PYGL expression under normoxia and hypoxia to further confirm that PYGL is regulated by HIF-1α in ccRCC. The results demonstrated that HIF-1α, glycolysis-related genes (GLUT1 and LDHA), and PYGL were dramatically upregulated in the cell lines when the cells were exposed to intermittent hypoxia (Fig. 7F–I and Supplementary Fig. S5D–G). Western blot analysis of the cells treated with hypoxia also supported the hypothesis that PYGL activity was regulated by HIF-1α (Fig. 7J).

Fig. 7.

HIF-1α up-regulates the expression of PYGL by directly binding to the PYGL promoter A, GESA revealed that hypoxia signaling pathway was enriched in the PYGL-higher expression group compared to the PYGL-lower expression group from the TCGA KIRC dataset. B–C, GEO datasets showed that the expression of PYGL was increased in response to hypoxia in renal epithelial and HeLa cells, respectively. D, HIF-1α binding motif provided by the JASPAR database. E, A simplified schematic demonstrating the potential hypoxia response elements (HRE1 and HRE2) and truncated sites in the human PYGL promoter region. F-J, 293-T and Caki-1 cell lines were cultured in 1% oxygen for 0, 6, 12, 24, and 48 h and then the HIF-1α and PYGL mRNA and protein expression were respectively detected by Real-time PCR and Western blotting. K, The expression of HIF-1α and PYGL protein significantly increased in a dose-dependent fashion by CoCl2. L, The expression of HIF-1α and PYGL protein significantly increased in a time-dependent manner by CoCl2 (100 μM). M-N, Routine PCR amplification of DNA fragments (HRE1 and HER2) immuno-precipitated by anti-HIF1α. Caki-1 cells were co-cultured with CoCl2 (200 μM) for 24 h and subjected to ChIP assay. O, Bar graph represents the fold enrichment of HIF1α ChIP DNA (HRE1 and HER2) compared with IgG control by real-time PCR.

CoCl2 was used to mimic the hypoxia model in ccRCC cells. Western blot showed a remarkable increase in HIF-1α and PYGL expression under CoCl₂ treatment in Caki-1 and ACHN cell lines in a dose-dependent and time-dependent manner (Fig. 7K and L). Furthermore, real-time PCR demonstrated higher levels of PYGL, GLUT1, and LDHA expression after CoCl₂ treatment, which appeared to be increased by hypoxia compared with normoxia (Supplementary Fig. S5H–M). Our results showed that in the TCGA ccRCC (KIRC) cohort, PYGL expression was positively correlated with gene expression including GLUT1 and LDHA, as expected (Supplementary Figure S5N-4O). Chromatin immunoprecipitation was performed with Caki-1 extract and 1 μg of normal rabbit IgG or 5 μg anti-HIF1 alpha antibody to further assess whether HIF-1α directly interacted with the PYGL promoter region to upregulate PYGL expression in the ccRCC cell lines. The results confirmed that HIF-1α could bind to the PYGL promoter to activate PYGL expression (Fig. 7M − O and Supplementary Fig. S6A–D). Together, the results suggest that PYGL expression is partly regulated by HIF-1α under hypoxia.

3.9. PYGL is a mediator of epithelial-mesenchymal transition (EMT)

EMT is a biological process in which epithelial cells are transformed into mesenchymal cells, and it plays an important role in carcinogenesis, metastasis, and therapy resistance. TCGA data resource revealed that the EMT pathway was the one of the 10 most remarkably altered pathways associated with high PYGL expression (Fig. 8A). Interestingly, morphological examination showed that PYGL knockdown alleviated EMT morphological changes (Fig. 8B). Importantly, we found that cell morphology changes were accompanied by EMT marker changes, in which the expression of N-cadherin, Snail, Twist, and Vimentin was decreased and the expression of E-cadherin was increased (Fig. 8C–E). Together, our findings revealed that PYGL is a novel candidate oncogene in ccRCC and is involved in an EMT-like process, which may confer resistance to sunitinib. However, further research is warranted to support this hypothesis.

Fig. 8.

PYGL is involved in the regulation of epithelial-mesenchymal transition (EMT) in ccRCC A, TCGA-KIRC database was employed and GSEA was performed, and it showed EMT molecules were highly enriched in the PYGL-higher expression group. B, Mesenchymal-epithelial transition (MET)-related morphological alterations detected in PYGL knockdown Caki-1 cells and ACHN cells. Cell morphological changes were shown in the phase contrast image. C-D, The mRNA expression levels of EMT-associated genes were detected by Real-time PCR between doxycycline-inducible control or PYGL-targeted shRNAs in Caki-1 cells and ACHN cells, respectively. E, Western blot analyses showing the protein level of EMT-related genes of control or PYGL-shRNAs Caki-1 cells and ACHN cells. F, A schematic model to summarize the findings of this study.

4. Discussion

The application of bioinformatics in tumor research is becoming increasingly widespread, but the variability issue between different database sets is also becoming more prominent. Therefore, the joint analysis between multiple omics used in this study can more efficiently screen out potential target genes.

The HPA contains protein expression data from 17 different forms of human cancer and normal human tissues. Therefore, this database was used for screening RCC biomarkers. The results demonstrated that genes, such as VIM, FABP7, and ATP6V1B1, were of high value as diagnostic biomarkers in RCC. VIM was suggested is associated with metastasis in RCC [29]. According to reports, FABP7 affects RCC cell proliferation depending on cell phenotype [30,31]. At present, the role of ATP6V1B1 in renal cell carcinoma has not been reported, which can be further identified in the future.

Moreover, the DEGs identified by GEO and TCGA datasets were compared with the differentially expressed proteins mined from the HPA database. The results indicated that more than 60% of proteins with high expression in ccRCC were discordant between the mRNAs and proteins (as described in Supplementary Table S6) because of post-translational modifications, such as phosphorylation and ubiquitination, influencing mRNA–protein correlation.

In this study, TMEM45A and PYGL were identified as novel biomarkers in ccRCC by integrated genomic and proteomic analyses. TMEM45A has been reported to be highly expressed in various cancers [[32], [33], [34]]. We discovered a negative correlation between the expression of TMEM45A mRNA and protein. However, the mRNA and protein levels of PYGL were remarkably upregulated in human ccRCC samples compared with their adjacent normal renal tissues. So we focused on the question of whether PYGL is a candidate diagnostic and prognostic biomarker for ccRCC.

No considerable differences were found between clinicopathologic characteristics and PYGL expression in ccRCC. These data strongly support the hypothesis that PYGL promotes ccRCC development but is not linked to ccRCC progression. As expected of an oncogene, PYGL knockdown (shPYGL) resulted in the inhibition of cell proliferation, cloning capacity, migration, invasion, and cell cycle arrest.

Sunitinib, a multitargeted tyrosine kinase inhibitor, has been established as the first-line treatment for advanced RCC. However, sunitinib resistance is a major challenge, and mechanistic insight into the possible underlying mechanisms of resistance is limited. Here, our results demonstrated that PYGL expression was remarkably increased in the sunitinib-resistant ccRCC cell lines compared with the parent cell lines. Importantly, PYGL inhibition could restore the sunitinib sensitivity of ccRCC cell lines.

In order to further explore the mechanism of action of PYGL in renal cell carcinoma, We have demonstrated through bioinformatics and molecular biology experiments that HIF-1α is a transcription factor of PYGL. HIF-1α plays an important role in ccRCC development [35,36]. Targeting the HIF-1α-PYGL signaling pathway may provide new ideas for the treatment of ccRCC.

Given that EMT is associated with cancer stem cells and acquired resistance to sunitinib [37,38], so we further confirmed through experiments that PYGL is a mediator of EMT in ccRCC cell lines. We speculate that PYGL may be involved in the potential mechanism of sunitinib resistance induced by EMT. However, more validation procedures are needed to support this hypothesis.

In summary, we used integrated genomic and proteomic analyses and identified PYGL as a specific tumor marker may have important value in the diagnosis and targeted therapy of ccRCC. Importantly, the work presented here suggests that targeted PYGL inhibitors could overcome resistance to sunitinib, we will focus on investigating the specific molecular mechanisms by which PYGL promotes sunitinib resistance in renal cell carcinoma, providing new therapeutic targets for sunitinib resistant patients.

Ethics approval and consent to participate

Human ccRCC and matching nontumor tissues used for research were approved by the ethics review committees of the institutional review boards of the Renji Hospital, Shanghai Jiao Tong University School of Medicine. And all animal experiments were approved by the Committee on the Ethics of Animal Experiments of University of South China (No: USCKF201711K13).

Patient consent for publication

Human ccRCC and matching nontumor tissues were collected from patients treated with nephrectomy in Renji Hospital, shanghai with written informed consent.

CRediT authorship contribution statement

Mingyong Li: Writing – review & editing, Visualization, Funding acquisition, Conceptualization. Guoqiang Zhu: Validation, Investigation. Yiqi Liu: Writing – review & editing, Investigation. Xuefeng Li: Writing – review & editing, Investigation, Funding acquisition. Yuxia Zhou: Validation, Investigation. Cheng Li: Validation, Investigation. Minglei Wang: Writing – review & editing, Resources, Formal analysis. Jin Zhang: Resources. Zhenping Wang: Validation, Investigation. Shuangfeng Tan: Validation, Investigation. Wenqi Chen: Validation, Supervision, Investigation. Hu Zhang: Writing – original draft, Validation, Methodology, Investigation, Data curation.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article:

figs1Pathway analysis was performed among 144 highly expressed genes to explore the underlying molecular mechanisms in the development of ccRCC.

figs2The clinical application value of TMEM45A for ccRCC. A-E, TMEM45 mRNA expression among four GEO databases and TCGA cohort in ccRCC compared with NAT. F-L, The relationships between TMEM45A expression and clinical characteristics including tumor size, lymph nodes metastasis, distant metastasis, tumor stage, tumor grade, tumor residual, and overall survival rates, respectively. M, Protein levels of TMEM45 were compared between kidney and kidney cancer derived from the HPA database. N-P, The level of expression of TMEM45A was measured by real-time PCR, immunofluorescence, and western blot analysis between ccRCC and matching nontumor tissues.

figs3Cell cycle analysis distribution between control shRNA and PYGL-shRNA groups in ACHN cells. A, The effect of knockdown of PYGL expression on the cell cycle distribution. B, Percentages of cells in the different phases are shown in the bar graph.

figs4PYGL inhibition abrogates sunitinib resistance in ccRCC cell lines. A, A498 cells were treated with different doses of sunitinib (0.25, 1, 4, and 8μM) for 0.5, 2, and 4 h, and then the levels of PYGL mRNA were measured by Real-time PCR. B, PYGL protein levels were examined in A498 cells after treatment with different doses of sunitinib for 12 h. C, PYGL protein levels were detected in A498 cells after treatment with different doses of sunitinib (3 μM)) at five different time points. D-E, Drug synergy matrix of sunitinib and CP-91149 in Caki-1 and ACHN cells. F-G, Dose-response curves were compared between sunitinib and the combination of sunitinib with CP-91149 in Caki-1 and ACHN cells. H-I, The combination index (CI) of sunitinib and CP-91149 was calculated for CP-91149 in Caki-1 and ACHN cells. J, Combination of sunitinib and CP-91149 enhances DNA damage and apoptosis in ccRCC cells. PARP, PCNA, H2A·X, and Cleaved Caspase-3 were detected after treatment of sunitinib, CP-91149, and the combination of sunitinib and CP-91149 in Caki-1 and ACHN cells.

figs5Glycolysis-associated genes including PYGL, GLUT1, and LDHA are up-regulated in ccRCC cells under hypoxia. A, GSEA identified glycolysis as a very important pathway to the PYGL-higher expression group. B, It showed that the knockdown of HIF-1α and HIF-2α decreased the expression of PYGL from GSE3188 data sets. C, A simplified schematic demonstrating the potential hypoxia response elements (HRE3 and HRE4) and truncated sites in the human PYGL promoter region. D-G, 293-T and Caki-1 cell lines were cultured in 1% oxygen for 0, 6, 12, 24, and 48 h, and then the mRNA of GLUT1, and LDHA were respectively detected by Real-time PCR. H-M, A dose- and time-dependent activation of PYGL, GLUT1 and LDHA mRNAs by CoCl2 were observed in Caki-1 cells and ACHN cells. N, Correlation between the PYGL mRNA levels and GLUT1 mRNA levels among TCGA-KIRC tumors. O, Correlation between the PYGL mRNA levels and LDHA mRNA levels among TCGA-KIRC tumors.

figs6Routine PCR and real-time PCR amplification of DNA fragments (HRE3 and HER4) immuno-precipitated by anti-HIF1α. Caki-1 cells were co-cultured with CoCl2 (200 μM) for 24 h and subjected to ChIP assay.