• microarray;
  • microRNA;
  • tumor;
  • meta-analysis;
  • miR-154


  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

The expression profiles of microRNAs (miRNAs) are associated with the initiation and progression of human tumors. DNA microarrays are widely used to explore the expression patterns of miRNAs. Because of the limited sample size and experimental expense, the statistical power of individual research projects is not sufficient to yield a robust conclusion. However, collected microarray datasets of expression profiles provide opportunities to compile the information of individual studies. Our study carried out a comprehensive meta-analysis of miRNA expression microarray datasets from 28 published tumor studies; it comprises 33 comparisons and nearly 4,000 tumor and corresponding nontumorous samples. This work reports 52 miRNAs as common signatures that are dysregulated in tumors. In addition to the commonly altered miRNAs, five solid cancers displayed specific tissue patterns of altered miRNAs as well. The meta-analysis also revealed some novel tumor-related miRNAs such as hsa-miR-144, hsa-miR-130b, hsa-miR-132, hsa-miR-154, hsa-miR-192 and hsa-miR-345. We further validated the expression pattern of hsa-miR-154 in human hepatocellular carcinoma by RT-PCR. Restoration of intracellular miR-154 inhibited tumor cell malignance and the G1/S transition in cancer cells. Both bioinformatic prediction and western blotting demonstrated that miR-154 could target CCND2. In addition, expression patterns of miR-154 were inversely correlated with those of CCND2 in hepatocellular carcinoma. Overall, this study used a large-scale data analysis to identify a qualified list of miRNAs that are consistently changed in tumors, which could lead to a better understanding of human tumor etiology.

Changes in gene expression occur in many biological processes such as development, aging and cancer. The RNA microarray is a high-throughput tool that can be used to survey the expressing pattern of thousands of genes. During the past decade, studies have increasingly focused on the changes in gene expression patterns of human cancers with microarray techniques. These studies have resulted in a flood of data that has been deposited in the Gene Expression Omnibus (GEO) and Array Express public databases.1, 2 Although reanalysis of the expression profile data as a whole remains a challenge, meta-analysis of multiple studies is a reasonable way to determine the genetic markers of tumors.3

MicroRNAs (miRNAs) are a class of small noncoding, single-strand RNAs of ∼22 nucleotides that are conserved in many species. In animals, miRNAs repress multiple downstream genes by binding to their imperfect complementary sites on the prime untranslated regions at their 3′ ends (3′-untranslated region; 3′-UTR) or through miRNA-directed mRNA cleavage.4 Moreover, recent work has suggested that miRNAs regulate targets while oscillating between repression and activation during the cell cycle.5 MiRNA profiles of cancer versus normal tissues have been investigated by microarray or real-time PCR methods. It is worth noting that each tumor has shown significantly different miRNA profiles compared to normal cells from the same tissue.6 The expression patterns of hundreds of tumors and corresponding normal controls have produced a specific view of dysregulated miRNAs in human tumors. In a separate study, it was suggested that only changes with 10-fold or greater differences in miRNA expression levels were significant.7 However, much smaller changes may have great effects because single miRNAs can impact multiple targets.8 Therefore, it is urgent to reanalyze the data using meta-analysis techniques with increased statistical power, which should yield more valid and more informative results than analysis of a single experiment.

Therefore, the aim of our article is to identify miRNA changes in tumors by pooling the available miRNA expression datasets. Our study used 28 public studies from GEO or Array Express to analyze tumor versus corresponding nontumorous tissues. We performed a comprehensive study to characterize the profile of dysregulated miRNAs in tumors. Although most of these miRNAs are well-known for their role in cancer, our systematic analysis also provides new clues about cancer-related miRNAs. To confirm the meta-results of miRNA expressing patterns, we focused on a novel cancer-related miRNA: miR-154. Previous studies have suggested that the genomic locus of miR-154 is often lost in cancers.9–11 We observed the downregulation of miR-154 in hepatocellular carcinoma (HCC) tissues. We also characterized the antimalignance role of miR-154 in a cell model of HCC. We further identified that miR-154 directly targets cyclin D2 (CCND2), which is essential for the control of cell cycle progression. These results confirmed the universal and specific changes of miRNAs in carcinogenesis, which should provide a direction for the future study of miRNA profiles in human cancers.

Material and methods

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Data collection

We searched the Pubmed, GEO and EBI Array Express databases for articles and datasets published before January, 2010 with the following key words: cancer, miRNA or microRNA and microarray. We retrieved studies that compared cancer to corresponding normal tissues to detect miRNA expression differences. The microarray data were downloaded from the public database or obtained from the authors directly. A study with the quantitative method of real-time PCR was also included.

Data processing

Except for the study with normalized data deposited in the public database, the background signal was subtracted from the raw data of each study.12 The mean variance among arrays was normalized by the VSN method.13 The data quality was checked based on the scheme curve of the signal values of each array. Arrays with extreme high or low signal values were discarded. Before combining the expression data of each study, the probes designated as blanks or controls were removed. Nonhuman miRNA probes were also removed, and the ID of each probe was converted according to miRBASE at the Sanger Center.14 An average expression value from the array repeat spots in the microarray experiments was estimated based on the correlation between the spots.15 We used the inverse-variance model to combine the effect sizes of each miRNA compared between cancer and normal tissues across studies.16 These procedures directly extracted the expression information from the raw data for each probe. The between-study heterogeneity for each miRNA was evaluated by the random effect model. Effect size analysis for each miRNA was also stratified by cancer type if the cancer study numbers reached three or more. The nonbiological variations among studies were adjusted using empirical Bayes methods. The adjusted expression values were used for hierarchical clustering based on a centered Pearson and average linkage measures in R.17 The p values were adjusted by false discovery rate (FDR) for each miRNA. p < 0.05 was considered significant. All analyses were carried out in R version 2.10.1 and Bioconductor version 2.6.18

Cell lines, tissue specimens and RNA oligoribonucleotides

Two hepatoma cell lines (HepG2 and Hep3B) were used in our study. HepG2 was cultured in DMEM supplemented with 10% fetal bovine serum (FBS), whereas Hep3B was maintained in 1640 with 10% FBS. Fresh surgical specimens of HCC, including tumor tissues and the neighboring nontumor liver tissues, were obtained from HCC patients at Zhongshan Hospital, Shanghai, China. All of the samples were immediately frozen in liquid nitrogen after surgery and then later stored at −80°C before further analysis. The miR-154 mimic (UAGGUUAUCCGUGUUGCCUUCG) and negative control (NC, UUCUCCGAACGUGUCACGU) RNA were purchased from Genepharma. The NC RNA mimic was not homologous to any human genomic sequence.

Colony formation assay and cell proliferation assay

Cells at 30–50% confluences were separately transfected with miR-154 mimic and NC RNA using Lipofectamine 2000 (Invitrogen). Twenty-four hours later, the cells were seeded onto a six-well cell plate at a density of 1,500 cells in each well in duplicate. Colonies were identified by crystal violet staining following 10–14 days of culture. The results are representative of three independent experiments. For the cell proliferation assay, cells were seeded onto a 96-well cell plate at a density of 800 cells each well. During a 6-day culture period, cells were subjected to the CCK-8 assay (Dojindo) every day. The cell growth curve was expressed as the absorbance at 490 nm (n = 5), which was read by a microtiter reader (Bio-Rad).

Cell cycle analysis

Cells were harvested 36 hr after transfection, washed twice with PBS and treated with PI staining solution (0.03% Triton X-100, 5 mg/ml RNase A and 10 μg/ml propidium iodide) for 30 min protected from light. The treated cells were analyzed on a flow cytometer to determine DNA synthesis and cell cycle status (FACSCalibur, BD Biosciences). At least 10,000 cells were acquired for each sample. The results are representative of three independent experiments with triplicate samples for each condition.

Quantitative real-time PCR

Paired HCC and bordering nontumor tissues were obtained from patients undergoing resection of HCC. Total RNA was extracted from tissues using Trizol reagent (Invitrogen) according the manufacturer's protocol. One or two micrograms of RNA was applied for reverse transcription using specific RT primers for miR-154 and U6. Appropriately diluted cDNA was used in a 20-μL real-time PCR reaction in triplicate for each gene. The expression of mature miR-154 was detected by hairpin-primer-based mature miRNA assays (Genepharma). The amount of miR-154 relative to that of tRNA was determined by the function of 2−ΔCT, where ΔCT = (CTmiR-154 − CTU6). To detect the expression of CCND2 in HCC tissues, real-time PCR analysis was carried out using SYBR Green Supermix kit (Takara) with the iCycler detection system (Bio-Rad). Primers for CCND2 were forward, 5′-CCGACAACTCCATCAAGCCTCAG-3′; and reverse, 5′-TGCCAGGTTCCACTTCAACTTCC-3′. Primers for GAPDH were forward, 5′-GAAGGTGAAGGTCGGAGTC-3′; and reverse, 5′-GAAGATGGTGATGGGATTTC-3′. The relative amount of CCND2 to GADPH was determined by the function of 2−ΔCT, where ΔCT = (CTCCND2 – CTGADPH).

Western blot

Protein samples were separated by SDS-PAGE and then transferred onto nitrocellulose membranes. After blocking with 5% milk in TBST (20 mM Tris-HCl, pH 7.6, 137 mM NaCl, and 0.01% Tween-20) for 1 h at room temperature, the membranes were incubated with specific antibodies against different proteins at 4°C overnight, followed by incubation with a horseradish peroxidase-conjugated secondary antibody. Immunoreactivity was visualized by chemiluminescence (Santa Cruz). The related antibodies we used included anti–β-actin (Sigma) and rabbit anti-CCND2 (ProteinTech).


  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Study characteristics

Twenty-eight studies7, 19–38 including 33 comparisons between tumor and nontumor tissue were identified. Supporting Information Table 1 shows the details of these datasets. Our study combined the miRNA expression data from 1,843 tumor tissue samples and 1,097 nontumor tissue samples, including 16 cancer types and corresponding nontumor controls. In total, 1,150 probes were enrolled for the effect size analysis, and 879 miRNAs were investigated in more than three studies. Nearly 50% of these probes were designed according to the passenger strand sequences of the miRNA genes.

Common miRNA signatures in human tumors

As shown in Figure 1a, the profiles of a group of miRNAs were altered in cancer tissues from the datasets comprising 33 comparisons and nearly 4,000 tumor and nontumor samples. Fifty-two miRNA genes were differentially expressed between various tumors and corresponding normal tissues (Table 1). The listed miRNAs were present in more than 10 studies, and the FDR-adjusted p-values were below the threshold level of 0.05. These dysregulated miRNAs consisted of 29 downregulated miRNAs and 23 upregulated miRNAs in cancers. Figure 1b shows a heat map that depicts the different expression profiles of the tumor and nontumor tissues. The heat map was determined using unsupervised clustering of 40 top-altered miRNAs of 1,037 samples, which were the subset of the total datasets obtained after qualified normalization for between-studies variation. Most miRNAs were thought to be involved in carcinogenesis in the previous studies. We also discovered novel miRNAs that are dysregulated in tumors. For example, hsa-miR-144, hsa-miR-130b, hsa-miR-132, hsa-miR-154, hsa-miR-192 and hsa-miR-345 had significantly dysregulated in tumors. Our data also suggest that members of some conserved miRNA families were upregulated or downregulated simultaneously between cancer and noncancer tissues. For example, hsa-miR-125a/125b were normally downregulated in cancers, whereas hsa-miR-18a/b, hsa-miR-25/92a, hsa-miR-103/107 and has-miR-17-5p/20/93/106 were upregulated. It has been reported that some miRNAs are contained in a single long transcript as a polycistronic gene. Such clustered miRNAs showed a similar expression pattern across the genome in tumors (Fig. 1c). The miR-143/145 cluster in chromosome 5 and the miR-99a/let-7c cluster in chromosome 21 were downregulated in tumors, whereas the miR-182/183 cluster in chromosome 7 was upregulated. The other three clusters were evolutionarily conserved as the consequence of ancient gene duplication.

thumbnail image

Figure 1. The microRNA expression profile altered in tumors. (a) A summary plot of the pooled effect sizes of miRNAs and 95% confidence interval sorted by P-FDR; the IDs of the top 5 overexpressed and top 5 under-expressed miRNAs are labeled. (b) Unsupervised clustering of samples based on miRNA expression for 612 patients and 425 normal controls from 12 datasets. The interstudy expression variations were adjusted using empirical Bayes methods. The enrolled samples were generally classified into two groups: tumor (red bar) versus normal (blue bar) with distinctive expression patterns. (c) The locus of the altered miRNAs in the genome. The “X” symbols show the chromosome positions of the top-ranked miRNAs altered in tumors. If more than two miRNAs were located within a 10-kb window in the genome, these miRNAs were considered to make up a miRNA cluster. The arrowhead matches indicates the upregulated or downregulated status for each cluster. [Color figure can be viewed in the online issue, which is available at]

Download figure to PowerPoint

Table 1. MiRNAs upregulated- or down-regulated in human tumors
inline image

Correlation analysis of the guide and passenger strands of dysregulated miRNAs

It is well-known that the passenger strands are degraded at the end of miRNA mature process, and our data also showed that the mean expression level of the passenger strand was significantly lower than that of the guide strand for each miRNA (t-test, p < 0.001). It is unknown, however, whether the miRNA expression patterns for both guide strands and passenger strands are altered in tumors. The correlation between the fold changes of the guide strands and those of the passenger strands of miRNAs was determined from the Pearson coefficient (Fig. 2). For 52 dysregulated miRNAs, the Pearson's correlation was 0.61 (p < 0.01), demonstrating a good correlation between the dysregulated expression profiles of the guide and passenger strands of the miRNAs. For example, both the guide and passenger strand of has-miR-1, hsa-miR-126 and hsa-mir-183 were significantly dysregulated in multiple tumors. For other miRNAs with significantly dysregulated guide strands, we observed a similar expression trend for the passenger strands, although the statistical significance did not reach the threshold level because these strands have been rarely investigated (data not shown). However, some miRNA passenger strands (e.g., hsa-miR-93, has-miR-25 and has-miR-191) did not show a dysregulated expression pattern, despite the significant dysregulation of the guide strand. For hsa-miR-30e and hsa-miR-132, we only observed that their passenger strands were significantly dysregulated in tumors. These data suggest that the strands of some precursor miRNAs (guide strand or passenger strand) are independently regulated in tumors, which adds more complexity to the activity of miRNAs.

thumbnail image

Figure 2. Scatter plot of fold changes of guide strand and passenger strand for dysregulated miRNA in tumors. The x-axis and y-axis indicate the fold changes of passenger strands and guide strands, respectively, of altered miRNA in tumors. The correlation between the fold changes of the guide strand and passenger strand of each miRNA was determined from the Pearson coefficient. The dotted line shows the linear fit predicting fold changes of guide strands from passenger strands. Although the plot shows good correlation of the dysregulation pattern of the two strands (p < 0.01), some miRNAs only display a dysregulated guide strand (up ellipse) or a dysregulated passenger strand (low ellipse).

Download figure to PowerPoint

Tissue patterns of miRNAs altered in five solid tumors

Considering the tissue-specific expression patterns of genes, we further reanalyzed the miRNA expression pattern for cancers investigated by more than three studies. These cancer types included tissue origins of breast, colon, liver, lung and prostate. Besides the commonly dysregulated miRNAs, the results showed that many specific miRNAs were changed in each cancer (Table 2). For breast cancer, colon cancer and HCC, more miRNAs were upregulated than downregulated. On the other hand, lung cancer and prostate cancer displayed much more downregulation of miRNAs. For prostate cancer, the number of downregulated miRNAs was nearly 5-fold greater than that of upregulated miRNAs. Interestingly, more than 60% of the changed miRNAs in HCC were not shared by the other four cancers listed in the Table 2. For the other cancers, 25% of the altered miRNAs were not shared. Thus, in addition to the miRNAs that were commonly dysregulated in cancers, there were some specific miRNAs that play a role only in certain cancers.

Table 2. The expression patterns of miRNAs altered in five solid tumors
inline image

MiR-154 was downregulated in cancer

In our study, some novel miRNAs were suggested to associate with cancer through the data-mining of published RNA-microarray data. MiR-154, one of the miRNAs altered in tumors, attracted our attention. The genomic locus of miR-154 (14q32) was frequently lost in several cancers.9–11 This chromosomal abnormality would cause the aberrant expression of miRNAs in this region, including miR-154.39, 40 According to the meta-results from 30 comparisons, miR-154 was significantly downregulated in tumors (p < 0.01, Fig. 3a). Stratified analysis according to cancer types showed that miR-154 was significantly downregulated in prostate cancer (five studies, p < 0.01, tau2 = 0.24) and in HCC (four studies, p < 0.05, tau2 = 0.02). Although the statistical power of this meta-analysis was sufficient to detect the changed miRNA between tumor and nontumor tissues, we further validated the altered expression of miR-154 in cancer using real-time PCR. Thus, we detected the expression of the mature miR-154 in HCC tissue and adjacent nontumor tissues. In the 25 paired tissues evaluated, miR-154 was downregulated in 21 paired samples (84%, Fig. 3b). In addition, nearly 48% (12/25) of HCC tissues showed a >50% decrease in miR-154 levels when compared to the adjacent nontumor tissues. These results suggest that the loss of miR-154 could play an active role in the carcinogenesis of HCC or other cancers.

thumbnail image

Figure 3. MiR-154 is downregulated in tumors and inhibits tumor cell malignancy. (a) Forest plots of miR-154 expression fold changes (log2 scale) and corresponding 95% confidence intervals (CIs). MiR-154 was significantly downregulated in the meta-results (p = 0.001) of over 30 comparisons between tumors and corresponding controls. The left column shows the cancer type and GEO or Array Express database accession ID for each dataset. (b) The reduced expression levels of mature miR-154 were validated in 25 paired HCC tissues and adjacent nontumor tissues by real-time PCR. The expression levels of miR-154 were normalized against those of U6 RNA. (c, d) Overexpression of miR-154 in cancer cells (HepG2 and Hep3B) inhibits the colony formation and cell growth rate. (e) Hep3B cells transfected with miR-154 show G1 phase arrest compared to controls. (* p < 0.05, ** p < 0.01).

Download figure to PowerPoint

MiR-154 inhibited tumor cell malignancy and the G1/S transition in vitro

As miR-154 expression was frequently reduced in tumors, we further examined the possible role of miR-154 overexpression in cancer cell malignancy. First, we carried out the colony formation assay on two liver cancer cell lines (HepG2 and Hep3B) by transfecting them with a control RNA duplex (NC) or miR-154 duplex. MiR-154 duplex-transfected cells displayed a reduced number of colonies compared to NC transfectants in two cancer cell lines (Fig. 3c). We also determined that overexpression of miR-154 can inhibit the colony formation of a cervical cancer cell line (HeLa; data not shown). Additionally, the growth rate of miR-154-transfected cells was significantly reduced compared to that of NC cells (Fig. 3d). Considering the antimalignancy role of miR-154, we further examined whether miR-154 affects the cell cycle of cancer cells. For HepG2 cells transfected with miR-154, restriction of G1/S progress was observed in comparison to controls (Fig. 3e, p < 0.05). Thus, these data suggest that miR-154 could inhibit tumor cell malignance by controlling the cell cycle.

MiR-154 directly targeted CCND2 in cancer cells

Because we had shown that miR-154 reduced tumorigenicity and inhibited the G1/S transition in cancer cells, we predicted its downstream target using TargetScan and miRanda algorithms to account for the underlying mechanism. Bioinformatics analysis revealed that CCND2 was a putative target for miR-154 (Fig. 4a). The 3′-UTR of CCND2 mRNA contained a complementary site for the seed region of miR-154, and this region has been evolutionarily conserved in animals (Fig. 4b). To determine whether miR-154 could regulate CCND2 at the protein level, HepG2 cells were transfected with miR-154 mimics and analyzed by western blot. The miR-154 mimics could reduce CCND2 protein expression relative to the control in HegG2 cells (Fig. 4c). As miR-154 was downregulated in HCC tissues, we further examined the expression pattern of CCND2 in HCC. For the majority (73%) of the 22 paired HCC tissues and adjacent nontumor tissues, CCND2 was significantly overexpressed in the cancer tissues (Fig. 4d, p < 0.05).

thumbnail image

Figure 4. CCND2 is directly targeted by miR-154. (a) Putative binding site of miR-154 in the cyclin D2 3′-UTR region was predicted by TargetScan and miRanda. (b) The binding site of miR-154 in cyclin D2 3′-UTR is evolutionary conserved across several species. Alignment of multiple DNA sequences was built using MULTIZ.46 Alignment shading was processed by TEXshade software.47 (c) The CCND2 protein levels were reduced by transfection of miR-154 in HepG2 cells, as determined by western blot. Protein levels were normalized against those of β-actin and presented relative to the levels obtained for the NC control. (d) CCND2 mRNA was upregulated in HCC tissues compared to adjacent tissues. The expression levels were normalized against those of GADPH.

Download figure to PowerPoint


  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

The etiology of cancer remains to be elucidated despite many efforts to understand it. Accumulating evidence has suggested that miRNAs play an active role in controlling development, differentiation and carcinogenesis. The efficacy of the microarray technique in miRNA research has produced large expression datasets of miRNA changes in tumors. Although Bargaje et al. conducted a comparison of miRNA expression profiles across mammalian tissues,41 there has been little work on the interplatform meta-analysis of miRNA associated with human tumors. Therefore, we combined the relevant datasets and processed them using meta-analysis techniques. In our study, we collected 33 datasets of miRNA chips from 28 studies that included 1,843 cancer samples and 1,097 noncancer samples. The statistical power of our integrated analysis should overcome the problems in an individual study. Our results showed that many miRNA genes were differentially expressed in malignant tissues compared to normal ones. Furthermore, the expression patterns of some miRNAs were consistently altered in more than one cancer. Thus, these miRNAs can be considered as common signatures for tumor status.

Meta-analysis uses a set of classic statistical methods to combine the results of several independent but related studies. Key issues in the meta-analysis of microchip data consist of analyzing the datasets in an integrated way, eliminating the variations among samples and processing the data using powerful statistic techniques.3 We applied the following methods to make this study feasible. First, we renamed the probe IDs of each dataset according to the nomenclature of miRNAs at the Sanger Center, which assisted the interplatform combination. Second, although the designs and platforms of different studies may cause nonbiological variations across studies, the energizing statistical methods have been applied to effectively correct the interstudy variation.17 Finally, several statistical techniques can be used to conduct meta-analyses based on direct published results or indirect raw data. We chose the inverse-variance model to combine the effect sizes of each study. As a comprehensive method to integrate the raw information of all available genes from the oligonucleotide chips, this approach considers the impact of each study and results in biologically interpretable measures.42 Nevertheless, we acknowledge that we failed to rule out underlying internal batch effects for each dataset. Batch effects are usually caused by various factors such as different experimental conduction times, different batches of chemical reagents and the different sources of biological samples. Thus, normalization of raw data is inadequate for some datasets because of unavailable batch information.

Previous studies noted the altered expression pattern of miRNAs in different cancers. Hierarchical clustering of the samples using miRNA profiles paralleled the developmental origins of the tissues.43 In addition to the miRNAs commonly altered in tumors, we also explored the tissue expression pattern of other miRNAs in five common solid tumors. Our results suggest that each tumor showed specific expression patterns besides the commonly changed miRNAs. Nearly 60% of miRNAs are altered in HCC, whereas fewer miRNAs are changed in other tumors. Future work on these specific miRNAs should explore their role in HCC, which is known for its poor prognosis.

The advantage of meta-analyzing the expression patterns from multiple datasets is that the increase in statistical power revealed some novel cancer-related miRNAs. In this study, we confirmed the meta-results of miRNA expression patterns for a novel cancer-related miRNA. This study found that miR-154 was significantly downexpressed in tumors, and we further confirmed this altered pattern in HCC specimens. As miR-154 is located in a genomic region that is often deleted in tumors, the reduced expression of miR-154 may promote the progress of cancer. Although a previous study suggested that miR-154 could associate with cancers such as squamous cell carcinoma of the tongue, the detailed accounting for the association remains unclear.44 We characterized the antimalignancy role of miR-154 in an HCC model, and our data revealed that miR-154 directly targets CCND2 and causes G1 arrest in cancer cells. Interestingly, in contrast to the reduced expression of miR-154 in HCC, tumor tissues also showed an increased expression of CCND2. The main role of cyclin D2 protein is to promote cell cycle progression from the G1 to S phase. Previous studies suggested that increased transcription of CCND2 is an early event in neoplastic transformation of the colon.45 Our data provide a new hint that loss control of miR-154 over CCND2 may cause runaway proliferation of cancer cells. Thus, these experimental analyses suggest that loss of miR-154 in tumors was partially responsible for the malignancy of cancer cells in an HCC model.

In conclusion, for the first time, we have performed a comprehensive meta-analysis of miRNA genes in a systemic way to identify the common miRNA signatures of human cancers. Our results provide a high-quality list of miRNAs that are changed in tumors, which warrants future studies to further explain the role of miRNAs in tumorigenesis.


  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

We would like to thank all authors of the studies that provided the raw datasets for this research. Thanks to Prof. Xianmei Yang and Dr. Shuyu Zhang for valuable comments on the manuscript.


  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information
  • 1
    Brazma A, Hingamp P, Quackenbush J, Sherlock G, Spellman P, Stoeckert C, Aach J, Ansorge W, Ball CA, Causton HC, Gaasterland T, Glenisson P, et al. Minimum information about a microarray experiment (MIAME)-toward standards for microarray data. Nat Genet 2001; 29: 36571.
  • 2
    Barrett T, Troup DB, Wilhite SE, Ledoux P, Rudnev D, Evangelista C, Kim IF, Soboleva A, Tomashevsky M, Edgar R. NCBI GEO: mining tens of millions of expression profiles—database and tools update. Nucleic Acids Res 2007; 35: D7605.
  • 3
    Moreau Y, Aerts S, De Moor B, De Strooper B, Dabrowski M. Comparison and meta-analysis of microarray data: from the bench to the computer desk. Trends Genet 2003; 19: 5707.
  • 4
    Bartel DP. MicroRNAs: genomics, biogenesis, mechanism, and function. Cell 2007; 131: 1129.
  • 5
    Vasudevan S, Tong Y, Steitz JA. Switching from repression to activation: microRNAs can up-regulate translation. Science 2007; 318: 19314.
  • 6
    Calin GA, Croce CM. MicroRNA signatures in human cancers. Nat Rev: Cancer 2006; 6: 85766.
  • 7
    He H, Jazdzewski K, Li W, Liyanarachchi S, Nagy R, Volinia S, Calin GA, Liu CG, Franssila K, Suster S, Kloos RT, Croce CM, et al. The role of microRNA genes in papillary thyroid carcinoma. Proc Natl Acad Sci USA 2005; 102: 1907580.
  • 8
    Calin GA, Croce CM. MicroRNA-cancer connection: the beginning of a new tale. Cancer Res 2006; 66: 73904.
  • 9
    Bando T, Kato Y, Ihara Y, Yamagishi F, Tsukada K, Isobe M. Loss of heterozygosity of 14q32 in colorectal carcinoma. Cancer Genet Cytogenet 1999; 111: 1615.
  • 10
    Cazals-Hatem D, Rebouissou S, Bioulac-Sage P, Bluteau O, Blanche H, Franco D, Monges G, Belghiti J, Sa Cunha A, Laurent-Puig P, Degott C, Zucman-Rossi J. Clinical and molecular analysis of combined hepatocellular-cholangiocarcinomas. J Hepatol 2004; 41: 2928.
  • 11
    Yoshimoto T, Matsuura K, Karnan S, Tagawa H, Nakada C, Tanigawa M, Tsukamoto Y, Uchida T, Kashima K, Akizuki S, Takeuchi I, Sato F, et al. High-resolution analysis of DNA copy number alterations and gene expression in renal clear cell carcinoma. J Pathol 2007; 213: 392401.
  • 12
    Smyth GK. Linear models and empirical bayes methods for assessing differential expression in microarray experiments. Stat Appl Genet Mol Biol 2004; 3: Article 3.
  • 13
    Huber W, von Heydebreck A, Sultmann H, Poustka A, Vingron M. Variance stabilization applied to microarray data calibration and to the quantification of differential expression. Bioinformatics 2002; 18 ( Suppl 1): S96104.
  • 14
    Griffiths-Jones S, Grocock RJ, van Dongen S, Bateman A, Enright AJ. miRBase: microRNA sequences, targets and gene nomenclature. Nucleic Acids Res 2006; 34: D1404.
  • 15
    Smyth GK, Michaud J, Scott HS. Use of within-array replicate spots for assessing differential expression in microarray experiments. Bioinformatics 2005; 21: 206775.
  • 16
    Ramasamy A, Mondry A, Holmes CC, Altman DG. Key issues in conducting a meta-analysis of gene expression microarray datasets. PLoS Med 2008; 5: e184.
  • 17
    Johnson WE, Li C, Rabinovic A. Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics 2007; 8: 11827.
  • 18
    Gentleman RC, Carey VJ, Bates DM, Bolstad B, Dettling M, Dudoit S, Ellis B, Gautier L, Ge Y, Gentry J, Hornik K, Hothorn T, et al. Bioconductor: open software development for computational biology and bioinformatics. Genome Biol 2004; 5: R80.
  • 19
    Ambs S, Prueitt RL, Yi M, Hudson RS, Howe TM, Petrocca F, Wallace TA, Liu CG, Volinia S, Calin GA, Yfantis HG, Stephens RM, et al. Genomic profiling of microRNA and messenger RNA reveals deregulated microRNA expression in prostate cancer. Cancer Res 2008; 68: 616270.
  • 20
    Arndt GM, Dossey L, Cullen LM, Lai A, Druker R, Eisbacher M, Zhang C, Tran N, Fan H, Retzlaff K, Bittner A, Raponi M. Characterization of global microRNA expression reveals oncogenic potential of miR-145 in metastatic colorectal cancer. BMC Cancer 2009; 9: 374.
  • 21
    Avissar M, Christensen BC, Kelsey KT, Marsit CJ. MicroRNA expression ratio is predictive of head and neck squamous cell carcinoma. Clin Cancer Res 2009; 15: 28505.
  • 22
    Blenkiron C, Goldstein LD, Thorne NP, Spiteri I, Chin SF, Dunning MJ, Barbosa-Morais NL, Teschendorff AE, Green AR, Ellis IO, Tavare S, Caldas C, et al. MicroRNA expression profiling of human breast cancer identifies new markers of tumor subtype. Genome Biol 2007; 8: R214.
  • 23
    Ferretti E, De Smaele E, Po A, Di Marcotullio L, Tosi E, Espinola MS, Di Rocco C, Riccardi R, Giangaspero F, Farcomeni A, Nofroni I, Laneve P, et al. MicroRNA profiling in human medulloblastoma. Int J Cancer 2009; 124: 56877.
  • 24
    Godlewski J, Nowicki MO, Bronisz A, Williams S, Otsuki A, Nuovo G, Raychaudhury A, Newton HB, Chiocca EA, Lawler S. Targeting of the Bmi-1 oncogene/stem cell renewal factor by microRNA-128 inhibits glioma proliferation and self-renewal. Cancer Res 2008; 68: 912530.
  • 25
    Gramantieri L, Ferracin M, Fornari F, Veronese A, Sabbioni S, Liu CG, Calin GA, Giovannini C, Ferrazzi E, Grazi GL, Croce CM, Bolondi L, et al. Cyclin G1 is a target of miR-122a, a microRNA frequently down-regulated in human hepatocellular carcinoma. Cancer Res 2007; 67: 60929.
  • 26
    Guo Y, Chen Z, Zhang L, Zhou F, Shi S, Feng X, Li B, Meng X, Ma X, Luo M, Shao K, Li N, et al. Distinctive microRNA profiles relating to patient survival in esophageal squamous cell carcinoma. Cancer Res 2008; 68: 2633.
  • 27
    Iorio MV, Visone R, Di Leva G, Donati V, Petrocca F, Casalini P, Taccioli C, Volinia S, Liu CG, Alder H, Calin GA, Menard S, et al. MicroRNA signatures in human ovarian cancer. Cancer Res 2007; 67: 8699707.
  • 28
    Mattie MD, Benz CC, Bowers J, Sensinger K, Wong L, Scott GK, Fedele V, Ginzinger D, Getts R, Haqq C. Optimized high-throughput microRNA expression profiling provides novel biomarker assessment of clinical prostate and breast cancer biopsies. Mol Cancer 2006; 5: 24.
  • 29
    Navon R, Wang H, Steinfeld I, Tsalenko A, Ben-Dor A, Yakhini Z. Novel rank-based statistical methods reveal microRNAs with differential expression in multiple cancer types. PLoS One 2009; 4: e8003.
  • 30
    Pineau P, Volinia S, McJunkin K, Marchio A, Battiston C, Terris B, Mazzaferro V, Lowe SW, Croce CM, Dejean A. miR-221 overexpression contributes to liver tumorigenesis. Proc Natl Acad Sci USA 2010; 107: 2649.
  • 31
    Raponi M, Dossey L, Jatkoe T, Wu X, Chen G, Fan H, Beer DG. MicroRNA classifiers for predicting prognosis of squamous cell lung cancer. Cancer Res 2009; 69: 577683.
  • 32
    Schaefer A, Jung M, Mollenkopf HJ, Wagner I, Stephan C, Jentzmik F, Miller K, Lein M, Kristiansen G, Jung K. Diagnostic and prognostic implications of microRNA profiling in prostate carcinoma. Int J Cancer 2010; 126: 116676.
  • 33
    Schetter AJ, Leung SY, Sohn JJ, Zanetti KA, Bowman ED, Yanaihara N, Yuen ST, Chan TL, Kwong DL, Au GK, Liu CG, Calin GA, et al. MicroRNA expression profiles associated with prognosis and therapeutic outcome in colon adenocarcinoma. JAMA 2008; 299: 42536.
  • 34
    Su H, Yang JR, Xu T, Huang J, Xu L, Yuan Y, Zhuang SM. MicroRNA-101, down-regulated in hepatocellular carcinoma, promotes apoptosis and suppresses tumorigenicity. Cancer Res 2009; 69: 113542.
  • 35
    Tanaka M, Oikawa K, Takanashi M, Kudo M, Ohyashiki J, Ohyashiki K, Kuroda M. Down-regulation of miR-92 in human plasma is a novel marker for acute leukemia patients. PLoS One 2009; 4: e5532.
  • 36
    Ueda T, Volinia S, Okumura H, Shimizu M, Taccioli C, Rossi S, Alder H, Liu CG, Oue N, Yasui W, Yoshida K, Sasaki H, et al. Relation between microRNA expression and progression and prognosis of gastric cancer: a microRNA expression analysis. Lancet Oncol 2010; 11: 13646.
  • 37
    Volinia S, Calin GA, Liu CG, Ambs S, Cimmino A, Petrocca F, Visone R, Iorio M, Roldo C, Ferracin M, Prueitt RL, Yanaihara N, et al. A microRNA expression signature of human solid tumors defines cancer gene targets. Proc Natl Acad Sci USA 2006; 103: 225761.
  • 38
    Yanaihara N, Caplen N, Bowman E, Seike M, Kumamoto K, Yi M, Stephens RM, Okamoto A, Yokota J, Tanaka T, Calin GA, Liu CG, et al. Unique microRNA molecular profiles in lung cancer diagnosis and prognosis. Cancer Cell 2006; 9: 18998.
  • 39
    Agueli C, Cammarata G, Salemi D, Dagnino L, Nicoletti R, La Rosa M, Messana F, Marfia A, Bica MG, Coniglio ML, Pagano M, Fabbiano F, et al. 14q32/miRNA clusters loss of heterozygosity in acute lymphoblastic leukemia is associated with up-regulation of BCL11a. Am J Hematol 2010; 85: 5758.
  • 40
    Choi HJ, Lee H, Kim H, Kwon JE, Kang HJ, You KT, Rhee H, Noh SH, Paik YK, Hyung WJ, Kim H. MicroRNA expression profile of gastrointestinal stromal tumors is distinguished by 14q loss and anatomic site. Int J Cancer 126: 164050.
  • 41
    Bargaje R, Hariharan M, Scaria V, Pillai B. Consensus miRNA expression profiles derived from interplatform normalization of microarray data. RNA 2009; 16: 1625.
  • 42
    Elo LL, Lahti L, Skottman H, Kylaniemi M, Lahesmaa R, Aittokallio T. Integrating probe-level expression changes across generations of Affymetrix arrays. Nucleic Acids Res 2005; 33: e193.
  • 43
    Lu J, Getz G, Miska EA, Alvarez-Saavedra E, Lamb J, Peck D, Sweet-Cordero A, Ebert BL, Mak RH, Ferrando AA, Downing JR, Jacks T, et al. MicroRNA expression profiles classify human cancers. Nature 2005; 435: 8348.
  • 44
    Wong TS, Liu XB, Wong BY, Ng RW, Yuen AP, Wei WI. Mature miR-184 as potential oncogenic microRNA of squamous cell carcinoma of tongue. Clin Cancer Res 2008; 14: 258892.
  • 45
    Bartkova J, Thullberg M, Slezak P, Jaramillo E, Rubio C, Thomassen LH, Bartek J. Aberrant expression of G1-phase cell cycle regulators in flat and exophytic adenomas of the human colon. Gastroenterology 2001; 120: 16808.
  • 46
    Blanchette M, Kent WJ, Riemer C, Elnitski L, Smit AF, Roskin KM, Baertsch R, Rosenbloom K, Clawson H, Green ED, Haussler D, Miller W. Aligning multiple genomic sequences with the threaded blockset aligner. Genome Res 2004; 14: 70815.
  • 47
    Beitz E. TEXshade: shading and labeling of multiple sequence alignments using LATEX2 epsilon. Bioinformatics 2000; 16: 1359.

Supporting Information

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Additional Supporting Information may be found in the online version of this article.

IJC_25818_sm_suppinfo.pdf98KSupporting Information.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.