Elevated transcriptional levels of aldolase A (ALDOA) associates with cell cycle-related genes in patients with NSCLC and several solid tumors
© The Author(s). 2017
Received: 24 February 2016
Accepted: 27 December 2016
Published: 7 February 2017
Aldolase A (ALDOA) is one of the glycolytic enzymes primarily found in the developing embryo and adult muscle. Recently, a new role of ALDOA in several cancers has been proposed. However, the underlying mechanism remains obscure and inconsistent. In this study, we tried to investigate ALDOA-associated (AA) genes using available microarray datasets to help elucidating the role of ALDOA in cancer.
In the dataset of patients with non-small-cell lung cancer (NSCLC, E-GEOD-19188), 3448 differentially expressed genes (DEGs) including ALDOA were identified, in which 710 AA genes were found to be positively associated with ALDOA. Then according to correlation coefficients between each pair of AA genes, ALDOA-associated gene co-expression network (GCN) was constructed including 182 nodes and 1619 edges. 11 clusters out of GCN were detected by ClusterOne plugin in Cytoscape, and only 3 of them have more than three nodes. These three clusters were functionally enriched. A great number of genes (43/79, 54.4%) in the biggest cluster (Cluster 1) primarily involved in biological process like cell cycle process (P a = 6.76E-26), mitotic cell cycle (P a = 4.09E-19), DNA repair (P a = 1.13E-04), M phase of meiotic cell cycle (P a = 0.006), positive regulation of ubiquitin-protein ligase activity during mitotic cell cycle (P a = 0.014). AA genes with highest degree and betweenness were considered as hub genes of GCN, namely CDC20, MELK, PTTG1, CCNB2, CDC45, CCNB1, TK1 and PSMB2, which could distinguish cancer from normal controls with ALDOA. Their positive association with ALDOA remained after removing the effect of HK2 and PKM, the two rate limiting enzymes in glycolysis. Further, knocking down ALDOA blocked breast cancer cells in the G0/G1 phase under minimized glycolysis. All suggested that ALDOA might affect cell cycle progression independent of glycolysis. RT-qPCR detection confirmed the relationship of ALDOA with CDC45 and CCNB2 in breast tumors. High expression of the hub genes indicated poor outcome in NSCLC. ALDOA could improve their predictive power.
ALDOA could contribute to the progress of cancer, at least partially through its association with genes relevant to cell cycle independent of glycolysis. AA genes plus ALDOA represent a potential new signature for development and prognosis in several cancers.
KeywordsALDOA Transcripts Cell cycle Cancer
In mammalian tissues, three aldolase isozymes (A, B and C), encoded by three different genes, are differentially expressed during development. Aldolase A (ALDOA), also known as fructose-bisphosphate aldolase A, is one of the glycolytic enzymes that catalyze the reversible conversion of fructose-1, 6-bisphosphate to glyceraldehyde-3-phosphate and dihydroxyacetone phosphate. ALDOA is primarily found in the developing embryo and adult muscle, and contributes to various cellular functions and biological process related to muscle maintenance, regulation of cell shape and mobility, striated muscle contraction, actin filament organization and ATP biosynthetic process. ALDOA deficiency probably results in myopathy and hemolytic anemia [1–3].
Recently, a new role of ALDOA has been proposed, given that ALDOA is highly expressed in a variety of malignant cancers, including human lung cancer , osteosarcoma , colorectal cancer , oral squamous cell carcinomas  and hepatocellular carcinomas . It could serve as a diagnostic and prognostic marker. Although elevated ALDOA level has been observed in these tumors, the underlying mechanism remains obscure and inconsistent. Some assumed that since glycolysis in rapidly growing tumor cells was up to 200 times faster than those of their normal tissues (Warburg effect), ALDOA expression would be also increased as an enzyme of this process . However, others demonstrated that ALDOA probably played a non-metabolic role to facilitate cell proliferation . Although RNA interference of ALDOA has been shown to inhibit cell proliferation in Ras-transformed NIH-3 T3 cells, there was no report on the potential mechanism or associated genes relevant to the role of ALDOA played in cell proliferation .
In the glycolysis pathway, ALDOA functions at the fourth step, followed by glyceraldehyde-3-phosphate dehydrogenase (GAPDH), another glycolytic enzyme at the sixth step. GAPDH is regarded as a housekeeping gene, but its expression is not always constant, especially in cancer, and has been supposed to correlate with cell cycle-dependent genes . However, little was known about the associated genes of ALDOA and its mechanism other than glycolysis, albeit its potential role in tumors. In this study, we tried to investigate ALDOA-associated genes (AA genes) with the utilization of publicly available microarray datasets, which would help elucidating the role for ALDOA played in cancer. Since the effect of ALDOA on lung cancer was reported in the previous studies [4, 12, 13], microarray datasets focused on lung cancer was used to detect AA genes, which were verified in other tumors.
Gene expression microarray datasets
Six independent datasets from ArrayExpress and GDS website. Except for the RNAseq data (normalized RSEM expression), other gene expression microarray datasets were normalized using RMA with R-package “affy”
Non-small-cell lung cancer
Identification of differentially expressed genes
Raw data retrieved from ArrayExpress were then normalized using Robust Multi-array Analysis (RMA) with “affy” R-Package (version 3.2.2), and the normalized expression values represented the probe set intensity on a log-2 scale. The expression levels of more than two probes standing for the same gene were averaged. Moderated t-statistic was carried out with “limma” R-Package to indentify differentially expressed genes (DEGs) between different disease statuses in each dataset. Both adjusted P value (Benjamini & Hochberg, P a) and Fold Change (FC) were obtained and only the genes with P a value < 0.05 and |FC| > 1.5 were selected as DEGs.
Construction of gene co-expression network with AA genes
Pearson’s correlation coefficient (r) was calculated. If P values of correlation coefficients between ALDOA and other DEGs in the cancer group were larger than 0.05, these DEGs were regarded as AA genes. Then, if r value between each two AA genes was larger than 0.7 in the cancer group, these two genes were connected and considered for ALDOA-associated gene co-expression network (GCN) construction. GCN is undirected with each node corresponding to one gene. Given a significant co-expression relationship exists between a pair of nodes, they are connected with an edge. GCN is of biological interest since co-expressed genes are likely controlled by the same transcriptional regulatory program, functionally related, or members of the same pathway . Besides, network analysis was performed on GCN in Cytoscape (version 3.2.1). The genes with highest degree and betweenness were considered as hub genes. Degree is defined as the number of links incident upon a node. Betweenness, an indicator of a node's centrality in a network, is equal to the number of shortest paths from all vertices to all others that pass through that node. A node with high betweenness has a large influence on the transfer of items through the network .
Partial correlation analyses between ALDOA and other genes
Given the possibility that the concurrence of up-regulated ALDOA and AA genes could be a mutual effect of accelerated glycolysis in cancer, we revaluated their relationship by partial correlation analysis, which measures the degree of association between two random variables in statistics, removing the possible influences of glycolysis on the co-existence of ALDOA and AA genes.
As reported, there are three important rate limiting enzymes, namely hexokinase, phosphofructokinase and pyruvate kinase, controlling the flux of glycolysis. Hexokinase (HK) is an uninversally expressed enzyme catalyzing the conversion of glucose into glucose-6-phosphate. Phosphofructokinase (PFK) is responsible for the phosphorylation of fructose-6-phosphate yielding fructose-1,6-bisphosphate. Pyruvate kinase (PK) catalyzes the last step of glycolysis transferring a phosphate group from phosphoenolpyruvate to ADP, producing one molecule of pyruvate and one molecule of ATP. HK2, PFKM and PKM encode the muscle-type isozymes respectively of these three enzymes. Here, the partial correlation coefficients (r p) between ALDOA and other genes at the transcriptional level were calculated using R-package ‘corpcor’ when separately removing the effect of these genes.
Network clustering and identification of cluster function
ClusterOne plugin in Cytoscape is designed to discover densely connected and possibly overlapping regions within the Cytoscape network (eg., Protein-Protein interaction network) by a greedy procedure adding or removing vertices to find groups with high cohesiveness . In this study, we utilized ClusterOne to detect highly interconnected region (Cluster) of the ALDOA-associated GCN. The functional enrichment analysis was performed using DEVID bioinformatics resource (version 6.7), to explore the well known database: Gene Ontology (GO) database. We especially annotated the clusters with GO BP (Biological Process), MF (Molecular Function), and CC (Cellular Component) terms. For multiple hypothesis tests, P a was obtained with bonferroni method.
Patients and tissue homogenate preparation
Frozen breast tumors stored in RNAlater® Solution (P/N: AM7021, Ambion, USA) were obtained from 16 patients diagnosed and operated in the Cancer Hospital of Shantou University Medical College in 2015. All patients did not receive radiotherapy or chemotherapy before surgery resection. Informed consent for the use of their samples was obtained from all the patients. This study was approved by the medical ethics committee of the Cancer Hospital of Shantou University Medical College. Tissue samples about 500 mg were homogenized in 1 ml TRIzol (Cat No. 15596–026, Invitrogen, USA) using a tissue homogenizer, and then supernatant was extracted for mRNA quantification after centrifugation.
Quantitative real-time polymerase chain reaction (qRT-PCR)
Primers of ALDOA, CDC20, CDC45 and CCNB2 for qRT-PCR. Primer Premier 5.0 was used to design the primers, and Primer-BLAST tool of NCBI was used to ensure the accuracy and specificity of these primers
Relationship of ALDOA with several hub genes
Spearman’s rank correlation coefficients (r s) were calculated between ALDOA and each of hub genes at the transcriptional level obtained from qRT-PCR mentioned above. P < 0.05 was considered as statistical significance.
Cell lines and culture conditions
One human breast cancer cell line SKBR3 were purchased from the Culture Collection of the Chinese Academy of Sciences, Shanghai, and maintained in DMEM (high glucose) (Gibco, Thermo Fisher Scientific Inc., California, USA) supplemented with 10% fetal bovine serum (FBS, Biological Industry, Kibbutz Beit Haemek, Israel) at a 37 °C, 5% CO2 incubator.
Western blotting (WB)
Cells were lysed with a lysis buffer and PMSF (Beyotime, Shanghai, China) on ice for 30 min and centrifuged at 12000 rpm for 15 min at 4 °C. Cell lysates (50 μg) were electrophoresed on 12% SDS polyacrylamide gel and transferred onto a PVDF membrane. After blocking with Tris buffered saline containing 0.05% Tween 20 (TBST) and 5% non-fat milk for 1 h at room temperature, the filters were washed 3 times × 5 min with TBST and then incubated with either rabbit anti-ALDOA polyclonal antibody (1:2000, Code No. ab71433, abcam, Cambridge, UK), or mouse anti β-Tubulin monoclonal antibody (1:1000, Code No. HC101-01, transgene, Illkirch Graffenstaden Cedex, France) diluted in blocking buffer at 4 °C overnight. After 3 times× 5 min wash with TBST, the membranes were incubated with horseradish peroxidase-labelled antirabbit (1:5000, Novus Biologicals, Littleton, USA) or antimouse (1:5000, Santa Cruz Biotechnology, Santa Cruz, USA) IgG at room temperature for 2 h, and washed with TBST. The blots were visualized with chemiluminescence.
Knockdown of ALDOA expression by small interfering RNA (siRNA)
SKBR3 with abundant ALDOA expression, was used to test the effect of ALDOA on cell cycle progression. We tried four different pairs of siRNAs designed targeting ALDOA specifically and purchased from GenePharmagps (Shanghai, China), only one of them substantially reduced the expression level of ALDOA. Therefore we used this siRNA oligonucleotide targeting the coding sequence of human ALDOA mRNA (siALDOA) and one scramble control siRNA (siNC) was served as control. The sense and antisense strand sequences of siALDOA are 5′-GCCUUGCCUGUCAAGGAAATT−3′ and 5′-UUUCCUUGACAGGCAAGGCTT−3′, respectively, the sense and antisense strand sequences of siNC are 5′-UUCUCCGAACGUGUCACGUTT-3′ and 5′- ACGUGACACGUUCGGAGAATT−3′, respectively. When cells growing in DMEM (high glucose) supplemented with 10% FBS were at a 40–50% confluence, siALDOA or siNC was added for transfection at a ratio of 75pmol siRNA: 7.5uL Lipofectamine 3000TM (Invitrogen, Thermo Fisher Scientific Inc., California, USA), 48 h after transfection, the transfectants were either lysed to check the efficiency of knockdown by WB, or switched to glucose-free DMEM for additional 8 h culture.
Flow cytometry for cell cycle analysis
Considered that the role of ALDOA in cancer could be an epiphenomenon of glycolysis (Warburg effect), we performed flow cytometry assays to examine the effect of ALDOA on cell cycle progression in the absence of glucose. Briefly, 48 h after transfection with either siALDOA or siNC, SKBR3 cells were then cultured in glucose-free DMEM (Gibco, Thermo Fisher Scientific Inc., California, USA) for additional 8 h to block the initiation of glycolysis, and cells were then collected and fixed with 70% ethanol at 4 °C overnight, then washed twice with ice-cold PBS. After that, 1 mg/ml RNaseA (Sigma-Aldrich Co., St Louis, MO, USA) was added at 37 °C, followed by propidium iodide (PI) staining for 30 min in the dark. BD AccuriTM C6 flow cytometer was used to measure the DNA contents. Each experiment was repeated at least three times.
Hierarchical clustering and survival analysis
Hierarchical clustering was performed to cluster samples with hub genes plus ALDOA, in order to determine whether these genes could also distinguish tumors from controls for other cancer types.
DEGs were identified between NSCLC and controls
The dataset E-GEOD-19188 based on the HG-U133_Plus_2 array, available from ArrayExpress database, contained 91 NSCLC tumors and 65 normal lung tissues (Table 1). Linear models identified 3448 genes differently expressed between NSCLC tumors and normal lung tissues (|FC| > 1.5 and P a < 0.05), which were regarded as DEGs. Of these, 1479 genes including ALDOA were up-regulated in tumors and the remaining 1969 genes were down-regulated.
AA genes were recognized in NSCLC
To recognize DEGs whose expression in the tumors were associated with ALDOA expression (AA genes), Pearson’s correlation coefficient was conducted within the NSCLC cancer cohort (91 arrays). Consequently, 1200 DEGs with P < 0.05 were identified as AA genes, of which the expression of 710 genes was positively correlated with ALDOA expression (the range of r values: 0.21 to 0.61), while 490 genes were negatively correlated with ALDOA (the range of r values:−0.54 to−0.21). Previous studies suggested genes with positively correlated expression profiles were much more likely to share similar functional annotations than genes with negatively correlated expression profiles . Therefore, only these 710 AA genes were used for ALDOA-associated GCN construction.
ALDOA-associated GCN was constructed in NSCLC
In order to construct ALDOA-associated GCN, for each pair of AA genes, we calculated Pearson’s correlation coefficients over again between their mRNA expression profiles. Totally, 1619 gene pairs with r ≥ 0.7 turned out to be interrelated and connected in the network. Thus the ALDOA-associated GCN finally included 182 nodes and 1619 edges.
ALDOA-associated GCN was clustered and annotated
Functional enrichment of AA genes in clusters. Functional enrichment analysis was performed for cluster annotation using DEVID bioinformatics resource, exploring the well known database: Gene Ontology (GO) database. We especially annotated the clusters with GO BP (Biological Process), MF (Molecular Function) and CC (Cellular Component) terms
P a values*
cell cycle process
mitotic cell cycle
M phase of meiotic cell cycle
positive regulation of
ubiquitin-protein ligase activity during mitotic cell cycle
structural constituent of cytoskeleton
structural molecule activity
calcium ion binding
Top ranked 10 genes with highest values of either degree or betweenness. Network analysis in Cytoscape was carried out on GCN. Moderated t-test was performed for examining differential expression of these genes between NSCLC cases and controls (E-GEOD-19188). Pearson’s correlation coefficients were calculated between ALDOA and these genes
P a valueb
P a valued
AA genes involved in cell cycle could be novel gene signatures for lung cancer and other solid tumors
The cohort from TCGA-LUAD, which included 30 LUAD patients and 30 normal controls randomly sampled from the whole dataset were clustered according to expression of hub genes and ALDOA. Class 1 was associated with controls (30/35, 85.7%), while Class 2 was preferentially enriched for LUAD (25/25, 100%, Fig. 3a). Only 5 LUAD cases (5/60, 8.3%) were misclassified into Class 1.
The cohort from E-GEOD-9750 containing 33 cervical cancer patients and 24 normal controls (normal cervix epithelium), were clustered. Class 1 was mainly enriched for controls, as it has 23/29 normal controls (79.3%) and 6 cervical cancer patients. On the other hand, Class 2 centralized 27 patients with cervical cancers and 1 normal controls (Fig. 3b). Totally, 7 samples (7/57, 12.3%) were wrongly classed into the opposite.
The cohort from E-GEOD-21422, which includes 14 breast cancer patients (5 with invasive ductal breast cancer (IDC) and 9 with ductal carcinoma in situ (DCIS)) and 5 normal controls, were clustered. Class 1 had 5 normal controls (71.4%) and 2 DCIS, while Class 2 had 7 DCIS and 5 IDC (100%, Fig. 3c).
The cohort from E-GEOD-14520 containing 30 hepatocellular carcinoma patients and 30 normal controls (liver non-tumor tissue) randomly selected from the original dataset were clustered. Class 1 was preferentially associated with normal controls (29/32, 90.6%). On the other hand, Class 2 was enriched for hepatocellular carcinoma patients, as it was composed of 27 cases (27/28, 96.4%) and 1 normal controls (Fig. 3d).
Positive relationship of ALDOA with hub genes is not regulated by glycolysis
Partial correlation analyses for reassessing the relationship of ALDOA and hub genes. Partial correlation analyses were performed in R-package ‘corpcor’ to reassess the relationship between ALDOA and hub genes (the top ranked 10 genes with highest values of either degree or betweenness)
r p a
r p c
Relationship of ALDOA with several hub genes were verified in breast tumors
ALDOA might affect the cell cycle progression independent of glycolysis
ALDOA might improve the predictive power of hub genes for NSCLC
3-year and 5-year OS rates of patients with different expression level of AA genes with or without ALDOA expression. Z-test was used to compare OS rates between different groups
CDC20 H ALDOA H
MELK H ALDOA H
PTTG1 H ALDOA H
CDC45 H ALDOA H
CCNB1 H ALDOA H
TK1 H ALDOA H
In this study, we utilized the available microarray datasets from the public database, ArrayExpress and GDC, to evaluate transcriptional levels of ALDOA and AA genes in solid tumors. The dataset used for detecting the DEGs and AA genes relied on an independent NSCLC cohort (E-GEOD-19188), but not aggregated data for avoiding false positive and/or negative correlation resulted from combining different datasets from various sources. Our identification of statistically significant changes in ALDOA expression in NSCLC enabled the evaluation of the gene expression profile that associated with ALDOA. In this study, AA genes in tumors were identified through ALDOA-associated GCN construction and clustering, to help explaining the role of ALDOA in tumors.
ALDOA has been known as the sole aldolase isozyme in red blood cells and skeletal muscle and is necessary for the production of adenosine triphosphate (ATP) in erythrocytes and muscle fibers . Recently, increasing evidences have shown that ALDOA could express in cancer cells, and its contribution to carcinogenesis in some tumors has been proposed both in vivo and in vitro. The role of ALDOA in lung cancer has been widely studied. Rho et al.  has pointed that ALDOA protein was up-regulated in human lung adenocarcinomas compared to normal pulmonary tissue, which was consistent with the result of Lin’s paper . Du et al.  further showed that ALDOA protein may induce epithelial-mesenchymal transition and promote cell migration in lung squamous cell carcinoma. Additionally, in previous studies, positive effect of ALDOA on initialization and progression of other cancers, such as colorectal cancer, oral tumor, osteosarcoma and hepatocellular carcinoma, had also been demonstrated [5–8]. Moreover, in vitro, ALDOA mRNA levels were down-regulated after glioma cell line SHC-44 cells treated with all-trans retinoic acid . Compared with melanocytes, the mRNAs of ALDOA were highly expressed in human melanoma cell lines G361 . In our paper, we also confirmed that ALDOA might contribute to tumorgenesis with aberrant mRNA levels in NSCLC. All these above have demonstrated that elevated ALDOA expression might be a potential biomarker for cancer diagnosis.
However, the mechanism of ALDOA in cancer remains unknown. Although some papers suggested that this was correlated to glycolysis, Ritterson and Tolan  have shown that silencing ALDOA drastically decreased the rate of cancer cell proliferation, and this did not greatly interfere with cellular energy metabolism. Although ALDOA is localized primarily in the cytoplasm, nuclear localization of ALDOA might be a common feature of proliferating cells, including cancer cells. All these indicated another potential mechanism other than glycolysis that ALDOA might involved in for carcinogensis. As a supplement to these reports, our analysis further showed that a majority of AA genes (over 50%) in the biggest cluster (Cluster 1) out of ALDOA-associated GCN were enriched for biological process relevant to cell cycle control, which demonstrated that ALDOA mRNA expression in NSCLC probably involved in cell proliferation. This result turned out to be consistent with the finding of Mamczur’s paper , which pointed to ALDOA as a factor involved in the regulation of cells proliferation in lung cancer cell. Mamczur and his colleagues further showed that ALDOA tended to co-existed with the expression of MKI67, a marker of proliferation. Interestingly, our paper also displayed a positive relationship between ALDOA and MKI67 (r = 0.275, P a = 0.046), and MKI67 was included in Cluster 1 (Fig. 2).
Moreover, we performed network analysis of GCN, and in view of the degree and betweenness of nodes, CDC20, MELK, PTTG1, CCNB2, CDC45, CCNB1, TK1 and PSMB2 were identified as the hub genes (Table 4), all of which directly or indirectly are involved in cell cycle control. Except for PSMB2, the relationship between these hub genes and cancer has been reported by previous studies. TK1, the most ALDOA relevant gene in GCN (r = 0.503), is a key kinase in the one-step salvage pathway, participates in DNA synthesis and is therefore closely related to the S-phase of the cell cycle. CDC45 (r = 0.41) is crucial for the initiation as well as the elongation process of eukaryotic DNA replication. Both of them have been found to be upregulated in several tumors and associated with proliferating cell populations [27–29]. The APC/C’s main function is to trigger the transition from metaphase to anaphase by tagging specific proteins for degradation. CDC20 (r = 0.40) is a regulatory protein that activates the APC. PTTG1 (r = 0.36) is an APC substrate that associates with a separin until activation of the APC. Upregulation of these two genes was associated with aggressive progression and poor prognosis in several tumors [30–32]. CCNB1 [33, 34] and CCNB2 [35, 36] might contribute to G2/M transition, and function as an oncogene and serve as a potential therapeutic target. MELK is a cell cycle-dependent protein kinase that belongs to the KIN1/PAR-1/MARK family. MELK overexpression has been detected in various human tumors . Given that a significant correlation between these genes and ALDOA was observed in our paper, it was supposed that ALDOA probably served as a key role in cell cycle regulation.
Our analysis also demonstrated up-regulated transcripts of ALDOA and hub genes could mostly distinguish tumors from controls not only in NSCLC, but also in other tumors, namely cervical cancer, breast cancer and hepatocellular carcinoma, suggesting that transcription of ALDOA might contribute to increased cell cycle-related cell proliferation, and be an important and probably universal step in carcinogenesis. We further detect transcripts of ALDOA and several hub genes in breast tumors by RT-qPCR, and Pearson’s correlation analysis demonstrated a positive relationship of ALDOA with CCNB2 and CDC45, but not CDC20 and TK1, even a trend observed between them. Although intimate correlation of ALDOA with several genes relevant to carbohydrate metabolism, such as LDHA (r = 0.605, P a < 0.001), PFKP (r = 0.598, P a < 0.001), GPI (r = 0.595, P a < 0.001), TPI1 (r = 0.548, P a < 0.001), GAPDH (r = 0.517, P a < 0.001), PGK1 (r = 0.447, P a = 0.001), PGAM1 (r = 0.411, P a = 0.002), ENO2 (r = 0.394, P a = 0.004), PKM (r = 0.363, P a = 0.007) and ENO1 (r = 0.361, P a = 0.008) was found for NSCLC, no cluster enriched into GO terms relevant to carbohydrate metabolism, indicating that ALDOA participated in carcinogenesis more likely through cell cycle control other than glycolysis. However, it might be partially due to limited number of metabolic genes found to be correlated with ALDOA, and it is unsuitable for ClusterOne used in our paper since this program tended to detect clusters with a large size of nodes.
Given that the positive relationship between ALDOA and AA genes indicated by Pearson’s correlation analysis could not exclude the possibility that it might be an entirely mutual consequence of the high energy demands required for rapid growth (Warburg effect). Thus, we reassessed their association by partial correlation analyses using GSE19188 and TCGA-LUAD by removing the potential influences of the two ALDOA-associated rate limiting enzymes of glycolysis (HK2 and PKM) [38–40]. We found that the significant association between ALDOA and most of hub genes remained, especially that between ALDOA and TK1 was seen both in these two datasets (GSE19188: r p = 0.408/0.424; TCGA-LUAD: r p = 0.442/0.296). Although RT-qPCR mentioned above failed to indicate a positive relationship between them (probably due to the limited sample sizes), we still supposed TK1 might be a promising target when studying the mechanism of ALDOA in cancer in future. Additionally, knocking down ALDOA in SKBR3 cells blocked cells at G0/G1 under minimized glycolytic condition, suggesting that ALDOA could contribute to the progress of cancer, at least partially through its association with genes relevant to cell cycle independent of glycolysis.
Several hub genes found in our paper have been proposed to predict poor prognosis of NSCLC in previous studies, such as CDC20 , MELK , PTTG1 , CCNB1  and TK1 . However, our study have indicated that these genes in combination with ALDOA could dramatically improve the predictive power for NSCLC prognosis. As shown in Figure 6 and Table 6, patients simultaneously with high expression of ALDOA and AA genes had a significantly lower survival rates than patients only with high expression of AA gene, especially for 5-years OS rates.
Our study has displayed ALDOA mRNA upregulation in cancers, confirmed its seemingly universal effect on carcinogenesis. Positive association between ALDOA and hub genes relevant to cell cycle remained even after minimizing the effect of glycolysis, indicating that ALDOA might contribute to cell proliferation of cancer, at least partially independent of glycolysis. AA genes, especially the hub genes would help to elucidate the non-glycolytic related functions of ALDOA in cancer. ALDOA might be a potential diagnostic and prognostic factor for cancer since ALDOA could distinguish tumors from controls and dramatically improve the predictive power of AA genes for poor survival.
- AA genes:
Ductal carcinoma in situ
Differentially expressed genes
Gene co-expression network
Invasive ductal breast cancer
Non-small-cell lung cancer
- P a :
Adjusted P values
This research was supported by the grants of (1) the National Natural Science Foundation of China (No. 81272931, 81572588); (2) Guangdong Provincial Sic-Tech Program (No. 2010B031600133, No. 2011B031800323); (3) Shantou University Medical College Clinical Trial Uplift Program (No. 201423).
The data analyzed during the current study are available in two public domain repositories. Datasets with identifier(s) ‘E-GEOD-19188’ (ref. 14), ‘E-GEOD-37745’ (ref. 15), ‘E-GEOD-9750’ (ref. 16), ‘E-GEOD-21422’ (ref. 17), ‘E-GEOD-14520’ (ref. 18) were derived from the public resource ArrayExpress (http://www.ebi.ac.uk/arrayexpress/). Another dataset ‘TCGA-LUAD’ was available at GDC Data Portal (https://gdc-portal.nci.nih.gov/projects/TCGA-LUAD).
FZ performed all the statistical analyses and drafted the manuscript. JDL designed and performed the experiments. XYZ helped to draft and revise the manuscript. YXZ and CQH collected the samples. GJZ and XJC helped to revise the manuscript. YKC conceived of the study, participated in revise the manuscript and supervised the work. All authors read and approved the final manuscript.
The authors declare that they have no competing interests.
Consent for publication
Ethics approval and consent to participate
Informed consent for the use of frozen breast tumors was obtained from all the patients. This study was approved by the medical ethics committee of the Cancer Hospital of Shantou University Medical College.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
- Knull HR, Walsh JL. Association of glycolytic enzymes with the cytoskeleton. Curr Top Cell Regul. 1992;33:15–30.View ArticlePubMedGoogle Scholar
- Tochio T, Tanaka H, Nakata S, Hosoya H. Fructose-1,6-bisphosphate aldolase A is involved in HaCaT cell migration by inducing lamellipodia formation. J Dermatol Sci. 2010;58:123–9.View ArticlePubMedGoogle Scholar
- Yao DC, Tolan DR, Murray MF, Harris DJ, Darras BT, Geva A, et al. Hemolytic anemia and severe rhabdomyolysis caused by compound heterozygous mutations of the gene for erythrocyte/muscle isozyme of aldolase, ALDOA(Arg303X/Cys338Tyr). Blood. 2004;103:2401–3.View ArticlePubMedGoogle Scholar
- Du S, Guan Z, Hao L, Song Y, Wang L, Gong L, et al. Fructose-bisphosphate aldolase a is a potential metastasis-associated marker of lung squamous cell carcinoma and promotes lung cell tumorigenesis and migration. PLoS One. 2014;9:e85804.View ArticlePubMedPubMed CentralGoogle Scholar
- Chen X, Yang TT, Zhou Y, Wang W, Qiu XC, Gao J, et al. Proteomic profiling of osteosarcoma cells identifies ALDOA and SULT1A3 as negative survival markers of human osteosarcoma. Mol Carcinog. 2014;53:138–44.View ArticlePubMedGoogle Scholar
- Peng Y, Li X, Wu M, Yang J, Liu M, Zhang W, et al. New prognosis biomarkers identified by dynamic proteomic analysis of colorectal cancer. Mol Biosyst. 2012;8:3077–88.View ArticlePubMedGoogle Scholar
- Lessa RC, Campos AH, Freitas CE, Silva FR, Kowalski LP, Carvalho AL, et al. Identification of upregulated genes in oral squamous cell carcinomas. Head Neck. 2013;35:1475–81.PubMedGoogle Scholar
- Hamaguchi T, Iizuka N, Tsunedomi R, Hamamoto Y, Miyamoto T, Iida M, et al. Glycolysis module activated by hypoxia-inducible factor 1alpha is related to the aggressive phenotype of hepatocellular carcinoma. Int J Oncol. 2008;33:725–31.PubMedGoogle Scholar
- Mamczur P, Gamian A, Kolodziej J, Dziegiel P, Rakus D. Nuclear localization of aldolase A correlates with cell proliferation. Biochim Biophys Acta. 1833;2013:2812–22.Google Scholar
- Ritterson Lew C, Tolan DR. Targeting of several glycolytic enzymes using RNA interference reveals aldolase affects cancer cell proliferation through a non-glycolytic mechanism. J Biol Chem. 2012;287:42554–63.View ArticlePubMedGoogle Scholar
- Wang D, Moothart DR, Lowy DR, Qian X. The expression of glyceraldehyde-3-phosphate dehydrogenase associated cell cycle (GACC) genes correlates with cancer stage and poor survival in patients with solid tumors. PLoS One. 2013;8:e61262.View ArticlePubMedPubMed CentralGoogle Scholar
- Lin CC, Chen LC, Tseng VS, Yan JJ, Lai WW, Su WP, et al. Malignant pleural effusion cells show aberrant glucose metabolism gene expression. Eur Respir J. 2011;37:1453–65.View ArticlePubMedGoogle Scholar
- Rho JH, Roehrl MH, Wang JY. Glycoproteomic analysis of human lung adenocarcinomas using glycoarrays and tandem mass spectrometry: differential expression and glycosylation patterns of vimentin and fetuin A isoforms. Protein J. 2009;28:148–60.View ArticlePubMedGoogle Scholar
- Hou J, Aerts J, den Hamer B, van Ijcken W, den Bakker M, Riegman P, et al. Gene expression-based classification of non-small cell lung carcinomas and survival prediction. PLoS One. 2010;5:e10312.View ArticlePubMedPubMed CentralGoogle Scholar
- Botling J, Edlund K, Lohr M, Hellwig B, Holmberg L, Lambe M, et al. Biomarker discovery in non-small cell lung cancer: integrating gene expression profiling, meta-analysis, and tissue microarray validation. Clin Cancer Res. 2013;19:194–204.View ArticlePubMedGoogle Scholar
- Scotto L, Narayan G, Nandula SV, Arias-Pulido H, Subramaniyam S, Schneider A, et al. Identification of copy number gain and overexpressed genes on chromosome arm 20q by an integrative genomic approach in cervical cancer: potential role in progression. Genes Chromosomes Cancer. 2008;47:755–65.View ArticlePubMedGoogle Scholar
- Kretschmer C, Sterner-Kock A, Siedentopf F, Schoenegg W, Schlag PM, Kemmner W. Identification of early molecular markers for breast cancer. Mol Cancer. 2011;10:15.View ArticlePubMedPubMed CentralGoogle Scholar
- Roessler S, Jia HL, Budhu A, Forgues M, Ye QH, Lee JS, et al. A unique metastasis gene signature enables prediction of tumor relapse in early-stage hepatocellular carcinoma patients. Cancer Res. 2010;70:10202–12.View ArticlePubMedPubMed CentralGoogle Scholar
- Stuart JM, Segal E, Koller D, Kim SK. A gene-coexpression network for global discovery of conserved genetic modules. Science. 2003;302:249–55.View ArticlePubMedGoogle Scholar
- Melak T, Gakkhar S. Comparative Genome and Network Centrality Analysis to Identify Drug Targets of Mycobacterium tuberculosis H37Rv. Biomed Res Int. 2015;2015:212061.View ArticlePubMedPubMed CentralGoogle Scholar
- Nepusz T, Yu H, Paccanaro A. Detecting overlapping protein complexes in protein-protein interaction networks. Nat Methods. 2012;9:471–2.View ArticlePubMedPubMed CentralGoogle Scholar
- Wang S, Xie Y, Yang X, Wang X, Yan K, Zhong Z, et al. Therapeutic potential of recombinant cystatin from Schistosoma japonicum in TNBS-induced experimental colitis of mice. Parasit Vectors. 2016;9:6.View ArticlePubMedPubMed CentralGoogle Scholar
- Allocco DJ, Kohane IS, Butte AJ. Quantifying the relationship between co-expression, co-regulation and gene function. BMC Bioinformatics. 2004;5:18.View ArticlePubMedPubMed CentralGoogle Scholar
- Lee YJ, Galoforo SS, Berns CM, Tong WP, Kim HR, Corry PM. Glucose deprivation-induced cytotoxicity in drug resistant human breast carcinoma MCF-7/ADR cells: role of c-myc and bcl-2 in apoptotic cell death. J Cell Sci. 1997;110(Pt 5):681–6.PubMedGoogle Scholar
- Zeng Y, Yang Z, Han YY, You C. Impact of all-trans retinoic acid on gene expression profile of glioblastoma cell line SHG-44. Ai Zheng. 2008;27:482–90.PubMedGoogle Scholar
- Suzuki A, Iizuka A, Komiyama M, Takikawa M, Kume A, Tai S, et al. Identification of melanoma antigens using a Serological Proteome Approach (SERPA). Cancer Genomics Proteomics. 2010;7:17–23.PubMedGoogle Scholar
- Chen G, He C, Li L, Lin A, Zheng X, He E, et al. Nuclear TK1 expression is an independent prognostic factor for survival in pre-malignant and malignant lesions of the cervix. BMC Cancer. 2013;13:249.View ArticlePubMedPubMed CentralGoogle Scholar
- Korkmaz T, Seber S, Okutur K, Basaran G, Yumuk F, Dane F, et al. Serum thymidine kinase 1 levels correlates with FDG uptake and prognosis in patients with non small cell lung cancer. Biomarkers. 2013;18:88–94.View ArticlePubMedGoogle Scholar
- Pollok S, Bauerschmidt C, Sanger J, Nasheuer HP, Grosse F. Human Cdc45 is a proliferation-associated antigen. FEBS J. 2007;274:3669–84.View ArticlePubMedGoogle Scholar
- Ding ZY, Wu HR, Zhang JM, Huang GR, Ji DD. Expression characteristics of CDC20 in gastric cancer and its correlation with poor prognosis. Int J Clin Exp Pathol. 2014;7:722–7.PubMedPubMed CentralGoogle Scholar
- Wu WJ, Hu KS, Wang DS, Zeng ZL, Zhang DS, Chen DL, et al. CDC20 overexpression predicts a poor prognosis for patients with colorectal cancer. J Transl Med. 2013;11:142.View ArticlePubMedPubMed CentralGoogle Scholar
- Demeure MJ, Coan KE, Grant CS, Komorowski RA, Stephan E, Sinari S, et al. PTTG1 overexpression in adrenocortical cancer is associated with poor survival and represents a potential therapeutic target. Surgery. 2013;154:1405–16. discussion 16.View ArticlePubMedPubMed CentralGoogle Scholar
- Begnami MD, Fregnani JH, Nonogaki S, Soares FA. Evaluation of cell cycle protein expression in gastric cancer: cyclin B1 expression and its prognostic implication. Hum Pathol. 2010;41:1120–7.View ArticlePubMedGoogle Scholar
- Aaltonen K, Amini RM, Heikkila P, Aittomaki K, Tamminen A, Nevanlinna H, et al. High cyclin B1 expression is associated with poor survival in breast cancer. Br J Cancer. 2009;100:1055–60.View ArticlePubMedPubMed CentralGoogle Scholar
- Takashima S, Saito H, Takahashi N, Imai K, Kudo S, Atari M, et al. Strong expression of cyclin B2 mRNA correlates with a poor prognosis in patients with non-small cell lung cancer. Tumour Biol. 2014;35:4257–65.View ArticlePubMedGoogle Scholar
- Albulescu R. Elevated cyclin B2 expression in invasive breast carcinoma is associated with unfavorable clinical outcome. Biomark Med. 2013;7:203.PubMedGoogle Scholar
- Jiang P, Zhang D. Maternal embryonic leucine zipper kinase (MELK): a novel regulator in cell cycle control, embryonic development, and cancer. Int J Mol Sci. 2013;14:21551–60.View ArticlePubMedPubMed CentralGoogle Scholar
- Song GQ, Zhao Y. Kisspeptin 10 inhibits the Warburg effect in breast cancer through the Smad signaling pathway: both in vitro and in vivo. Am J Transl Res. 2016;8:188–95.PubMedPubMed CentralGoogle Scholar
- Sun H, Zhu A, Zhang L, Zhang J, Zhong Z, Wang F. Knockdown of PKM2 Suppresses Tumor Growth and Invasion in Lung Adenocarcinoma. Int J Mol Sci. 2015;16:24574–87.View ArticlePubMedPubMed CentralGoogle Scholar
- Luo W, Semenza GL. Emerging roles of PKM2 in cell metabolism and cancer progression. Trends Endocrinol Metab. 2012;23:560–6.View ArticlePubMedPubMed CentralGoogle Scholar
- Kato T, Daigo Y, Aragaki M, Ishikawa K, Sato M, Kaji M. Overexpression of CDC20 predicts poor prognosis in primary non-small cell lung cancer patients. J Surg Oncol. 2012;106:423–30.View ArticlePubMedGoogle Scholar
- Li Y, Tang H, Sun Z, Bungum AO, Edell ES, Lingle WL, et al. Network-based approach identified cell cycle genes as predictor of overall survival in lung adenocarcinoma patients. Lung Cancer. 2013;80:91–8.View ArticlePubMedPubMed CentralGoogle Scholar
- Li H, Yin C, Zhang B, Sun Y, Shi L, Liu N, et al. PTTG1 promotes migration and invasion of human non-small cell lung cancer cells and is modulated by miR-186. Carcinogenesis. 2013;34:2145–55.View ArticlePubMedGoogle Scholar
- Cooper WA, Kohonen-Corish MR, McCaughan B, Kennedy C, Sutherland RL, Lee CS. Expression and prognostic significance of cyclin B1 and cyclin A in non-small cell lung cancer. Histopathology. 2009;55:28–36.View ArticlePubMedGoogle Scholar
- Xu Y, Liu B, Shi QL, Huang PL, Zhou XJ, Ma HH, et al. Thymidine kinase 1 is a better prognostic marker than Ki-67 for pT1 adenocarcinoma of the lung. Int J Clin Exp Med. 2014;7:2120–8.PubMedPubMed CentralGoogle Scholar