Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Meta-analysis of adrenocortical tumour genomics data: novel pathogenic pathways revealed

Abstract

Sporadic adrenocortical tumours are common, but their pathogenesis is poorly elucidated. In this study, we present a meta-analysis and review of gene expression microarray and comparative genome hybridization (CGH) studies performed to date on these tumours, including our own data. Data of whole genome microarray studies from altogether 164 tumours (97 benign, 67 malignant) and 18 normal tissues were reclassified and reanalysed. Significant gene sets and cytogenetic changes from publications without available genomic data were also examined including 269 benign, 215 malignant tumour and 30 normal tissues. In our experimental study, 11 tumour and four normal samples were analysed by parallel mRNA and CGH profiling. Data were examined by an integrative bioinformatics approach (GeneSpring, Gene Set Enrichment Analysis and Ingenuity Pathway Analysis softwares) searching for common gene expression changes and paralleling chromosome aberrations. Both meta-analysis of available mRNA and CGH profiling data and our experimental study revealed three major pathogenetic pathways: (1) cell cycle, (2) retinoic acid signalling (including lipopolysaccharide/Toll like receptor 4 pathway), (3) complement system and antigen presentation. These pathways include novel, previously undescribed pathomechanisms of adrenocortical tumours, and associated gene products may serve as diagnostic markers of malignancy and therapeutic targets.

Introduction

Adrenocortical tumours (ACT) are common; their prevalence may reach up to 5–7% in pathological series. Most of them are benign (adrenocortical adenoma, ACA), and the majority of these are hormonally inactive (IA); however, rare aldosterone- or cortisol-producing (APA or CPA) tumours are associated with significant morbidity and mortality. The prognosis of the rare adrenocortical cancer (ACC) is poor with an overall 5-year survival below 30% (Libè et al., 2007).

Several molecular pathways involved in adrenocortical malignancy have been revealed including overexpression of insulin-like growth factor 2 (IGF-2), mutations of tumour protein p53 (TP53) and alterations of Wnt/β-catenin and cAMP-mediated signalling (Soon et al., 2008). Despite these findings and the ongoing intensive investigation of adrenocortical tumorigenesis, we are far from an integrative view.

Studies on gene expression (mRNA profiling) and chromosome aberrations (comparative genome hybridization, CGH) by high-throughput bioinformatics-based approaches have revealed numerous pathomechanisms and molecular characteristics in several tumours. However, the overall interpretation of these findings and the selection of biologically relevant pathomechanisms is difficult (Hong and Breitling, 2008).

A number of mRNA and CGH (both classic and microarray-based) profiling studies have been performed on adrenocortical tumours to date that showed various alterations in gene expression and chromosome aberration patterns. Gene expression alterations may be used for the differentiation of benign and malignant tumours (de Fraipont et al., 2005; Soon et al., 2009) and for determining ACC prognosis (de Reyniès et al., 2009). Beside steady alterations observed in all settings (for example, overexpression of IGF-2 mRNA in ACC), considerable differences can be found in different studies.

We hypothesized that by reclassifying and reanalyzing publicly available gene expression and CGH data then subjecting these to pathway analysis, biologically relevant and previously unknown pathogenic pathways may be identified. Common significant gene expression changes between different microarray studies and corresponding mRNA expression and chromosomal alterations were searched for, that is, chromosome losses and gains paralleling under- or overexpression of genes by comparing Gene Set Enrichment Analysis (GSEA) results and data on cytogenetic changes (CGH and FISH (fluorescent in situ hybridization)).

To confirm the feasibility of this in-silico meta-analysis of mRNA and CGH profiling results, we performed an own pilot study including parallel mRNA and CGH profiling on 11 tumour samples and four normal adrenocortical tissues (NA) by microarray-based approaches. By applying the recently published methods for the comparison of simultaneous cDNA microarray and CGH data (Skawran et al., 2008), the implication of chromosomal aberrations in adrenocortical tumour pathogenesis could be revealed.

Results

Meta-analysis of gene expression data

The ACA–ACC comparison could be performed with the largest data sets, as malignancy was determined in all studies analysed. By statistical analysis of microarray studies performed by Giordano et al. (2009) and Tömböl et al. (2009), we have identified 554 and 450 significant gene expression changes between NAs and ACAs, respectively: 46 of these were common. In the same studies, 2509 and 2564 significant gene expression changes between NAs and ACCs were found and 1017 of these were common. Comparison of ACA and ACC groups was possible in three studies: 1929, 1251 and 1904 significant gene expression changes were found in the studies of de Reyniès et al. (2009), Giordano et al. (2009) and Tömböl et al. (2009), respectively: 402 expression changes were common in all (Supplementary Figure 1, Supplementary Table 1). As these microarray studies were performed on different microarray platforms using independent tumour sets, these common genes may be regarded as cross-validated.

Since the clinical profiles of samples were unavailable in one major study, the comparison of hormonal activity-related gene expression profiles between different studies was not feasible in all cases. Common significant gene expression changes in the studies of Giordano et al., (2009) and Tömböl et al., (2009) were identified: 15 between NAs and IAs, 17 between NAs and CPAs, and only 6 between IAs and CPAs. 305 significant gene expression changes between IAs and cortisol-secreting ACCs and 253 gene expression changes between CPA and cortisol-secreting ACC samples were found (Supplementary Table 2).

Correlation of GSEA results retrieved from publicly available microarray and cytogenetic data

GSEA was performed on NA–ACC and ACA–ACC comparisons. As potential adrenomedullary contamination may interfere with the analysis of normal adrenocortical samples and chromosomal aberrations are more infrequent in ACAs than in ACCs (Kjellman et al., 1996) (Supplementary Table 3), we focused on the ACA–ACC comparison. All chromosomal regions found to be differentially enriched by GSEA in the publicly available ACC sample sets have already been published previously (Table 1). All significantly enriched chromosomal regions in these studies were upregulated in ACCs suggesting the frequent gain of these chromosome regions. Chromosome region losses were often observed in ACCs by cytogenetic studies (Supplementary Table 3), however, our GSEA approach failed to reveal downregulated chromosomal regions. This observation highlights the limitations of indirect gene expression and cytogenetic data comparisons, and supports the necessity of simultaneous mRNA and CGH profiling approaches.

Table 1 Comparison of the common GSEA results of the publicly available whole genome mRNA expression data sets and the literature findings on cytogenetic changes in ACCs

Analysis of parallel mRNA and CGH profiling in our own experimental study

With the direct comparison of GSEA and CGH results performed on the same samples, overlaps between chromosome aberrations revealed by CGH and significantly enriched chromosome regions identified by GSEA were established. CGH data are presented in Supplementary Table 4 and accessible at GEO under accession number: GSE17396. Among the 101 aberrant chromosomal regions detected in at least 75% of the investigated ACC samples by CGH, 46 regions were found to be significantly enriched in the same samples by GSEA (Figure 1a and Supplementary Table 5). Despite the low sensitivity (45.54%), the specificity of GSEA to identify aberrant chromosome regions was 84.86%. Among the 3815 genes harboured on these 46 aberrant chromosome regions, we found 1736 genes by applying Leading Edge Analysis and then loaded these genes onto IPA. About 15% of significant gene expression changes could be related to chromosomal rearrangements.

Figure 1
figure1

(a) Correlation of gene expression changes revealed by GSEA with CGH data in our experimental study. (b) Pie diagram representing the distribution of gene expression changes revealed by pathway analysis of publicly available microarray data sets.

Pathway analysis

Fifteen canonical pathways have been found to be significantly altered in at least two different microarray studies (Supplementary Table 6). Significant gene sets in NA–ACC and ACA–ACC comparisons overlapped with retinoic acid signalling, cell cycle, complement system and several metabolomic processes (Figure 1b).

IPA analysis of overlapping CGH and GSEA data sets in our experimental study using in silico identified leading edge subsets mostly revealed gene expression alterations involved in the regulation of cell cycle G2/M and G1/S checkpoints that were highly concordant with the results of microarray meta-analysis (including cyclins E1 and E2 (CCNE1 and CCNE2), G2/mitotic-specific cyclin-B1 (CCNB1), cell division cycle 25 homologue C (CDC25C), myc proto-oncogene protein gene (C-MYC)).

Discussion

Pathway meta-analysis of available transcriptomics and CGH data as well as pathway analysis performed in our pilot study involving parallel CGH and transcriptomics profiling on the same samples revealed three major pathogenetic pathways that is, (1) damage of cell cycle, (2) retinoic acid signalling (3) antigen presentation and complement system.

Gene Ontology analysis has already been performed in a few microarray studies (Fernandez-Ranvier et al., 2008a, 2008b), which identified 25 genes out of 314 at chromosome 11q13 that are mainly involved in the regulation of cellular growth and apoptosis. Giordano et al. (2009) associated the list of differentially expressed genes with chromosomal instability and ‘functional aneuploidy’ that underlines the relevance of parallel gene expression and CGH profiling. The gene expression profiles established by the large-scale study by de Reyniès et al. (2009) could be associated with pathways of DNA replication, mitotic cell cycle and immune response.

Whereas several findings support the relevance of cell cycle damage in adrenocortical tumorigenesis, no data have been published on the involvement of retinoic acid signalling to date. Here, we discuss the findings established by our meta-analysis and experimental CGH-transcriptomics study and present a review of these and previous observations.

Cell cycle

Damage of cell cycle has been described in the pathogenesis of several tumours, and there are some reports describing expression changes of genes involved in G1/S and G2/M transition in ACCs. Overexpression G1 cyclins (cyclin E), and G1 cell division protein kinases (CDK2, CDK4) was reported in sporadic ACCs (Bourcigaux et al., 2000). Overexpression of CCNE1 and CCNE2 was found in four of the investigated microarray studies (Giordano et al., 2003, 2009; Tömböl et al., 2009; de Reyniès et al., 2009). GSEA of microarray studies revealed significant enrichment of overexpressed genes on chr19q12, where CCNE1 is mapped (Giordano et al., 2009; Tömböl et al., 2009; de Reyniès et al., 2009). Gain of chr19q12 has already been described (Supplementary Table 3). GSEA of microarray studies by de Reyniès et al. (2009), Giordano et al. (2009) and Tömböl et al. (2009) revealed significant enrichment of overexpressed genes on chr12q13, where CDK2 is harboured. Gain of chr12q13 is often found in ACCs (Supplementary Table 3). Stephan et al. (2008) reported a 391 kb-long segment on chr12q13.2 that correlated with poor survival, CDK2 is one of the 25 genes mapped to this chromosome segment.

It is of interest to note that C-MYC overexpressed in several other tumours was found to be underexpressed in three microarray data sets analysed (Giordano et al., 2009; Tömböl et al., 2009; de Reyniès et al., 2009). In our direct comparison of CGH and GSEA results, its underexpression was correlated with loss of chr8q24, where C-MYC is harboured.

Overexpression of several genes involved in the regulation of G2/M phase transition was noted in ACCs and correlated with chromosome gains: CCNB1 and cell division protein kinase 7 (CDK7) on chr5q13, ubiquitin C (UBC) on chr12q24, and Mdm2 p53-binding protein homologue (MDM2) on chr12q15 (Giordano et al., 2003, 2009; Tömböl et al., 2009; de Reyniès et al., 2009). Overexpression of CCNB1 in ACCs was validated in the recent study by Soon et al., (2009) as well.

Several other members of G2/M transition were significantly overexpressed in ACCs without corresponding chromosomal alterations in at least three microarray studies, including CCNB2, CDC2, CDC25B and topoisomerase II alpha (TOP2A) (Giordano et al., 2003, 2009; Tömböl et al., 2009; Soon et al., 2009; de Reyniès et al., 2009). Overexpression of TOP2A has already been reported and considered as a diagnostic marker for ACCs (Giordano et al., 2003). TOP2A is a major target for a variety of antineoplastic agents, including etoposide and doxorubicin used in ACC chemotherapy (Patalano et al., 2009). Resistance to TOP2A targeting drugs in cell lines has been associated with reduced expression of the protein, mutations of binding sites and aberrant localization in the cytoplasm. Exportin (XPO1) mediates nuclear export of several proteins including TOP2A. In ovarian cancer, high cytoplasmic immunostaining of TOP2A was linked to XPO1 overexpression and associated with borderline significance for decreased overall survival (Faggad et al., 2009). We have observed significant overexpression of XPO1 in ACCs in two studies (Giordano et al., 2009; de Reyniès et al., 2009), however, aberrant cytoplasmic expression of TOP2A has not been reported yet in ACCs (Figure 2).

Figure 2
figure2

Pathways with altered gene expression patterns in the regulation G1/S and G2/M checkpoint in ACCs.

Retinoic acid signalling

Retinoids are specific ligands of retinoid acid receptors (RARs) and retinoid X receptors (RXR). RARs bind both all-trans retinoic acid and 9-cis retinoic acid while RXRs bind only 9-cis retinoic acid (Shimizu et al., 2009). During embryonic development, the adrenal gland may be the most important source of serum retinoic acid before liver maturation (Haselbeck et al., 1997). Retinoic acid concentration in the adult adrenal cortex is about half of that found in the liver, that may be by far sufficient for intraadrenal RAR and RXR activation (Haselbeck et al., 1997). Retinoids are involved in the pathogenesis of several tumours, and are used for cancer therapy and prevention. 9-cis retinoic acid treatment inhibited the growth of the human adrenocortical cancer cell line NCI-H295 (Ferruzzi et al., 2005).

The second oxidation step of retinoic acid synthesis catalysed by aldehyde dehydrogenase 1A enzyme family members (ALDH1A1-3) is rate-limiting (el Akawi and Napoli, 1994). We found decreased ALDH1A1 and ALDH1A3 mRNA expression in ACC relative to ACA in several microarray studies (Giordano et al., 2003, 2009; Velázquez-Fernández et al., 2005; Laurell et al., 2009; Soon et al., 2009; de Reyniès et al., 2009). Decreased ALDH1A1 has been validated by QRT-PCR in two studies (Fernandez-Ranvier et al., 2008b; Laurell et al., 2009). Significant underexpression of RXR alpha (RXRA) was observed in ACCs in one study (Tömböl et al., 2009). By the direct comparison of CGH and GSEA results in our experimental study, decreased expression of RXR beta (RXRB) could be correlated with the loss of chr6p21 where RXRB is harboured. Based on these findings, retinoic acid production and action might be reduced in ACC, but this hypothesis awaits experimental confirmation.

Cholesterol and lipid metabolism: RXR/LXR and RXR/PPARG signalling

Retinoic acid bound RXRs may heterodimerize with other nuclear receptors including liver x receptors (LXR) and peroxisome proliferator-activated receptors (PPAR) (Shimizu et al., 2009). Since these receptors are involved in the regulation of drug metabolism and transport, changes in their concentrations may affect the effective drug dose during tumour therapy or may even be involved in adrenocortical carcinogenesis (Scripture et al., 2005). The adrenal gland expresses LXRA and LXR beta (LXRB, NR1H2). LXRs have pivotal roles in adrenal cholesterol metabolism, and influence steroid hormone production (Cummins et al., 2006). LXR alpha (LXRA, NR1H3) was significantly underexpressed in ACCs in three studies (Giordano et al., 2009; Tömböl et al., 2009; de Reyniès et al., 2009) that may be related to the already described loss of chr11p11 (Figure 3; Supplementary Table 3).

Figure 3
figure3

Pathways with altered gene expression patterns involved in LXR/RXR signalling in ACCs.

PPAR-γ (PPARG) regulates expression of genes involved in glucose and lipid homeostasis and is implicated in the pathogenesis of several tumours (Ondrey, 2009). PPARG is expressed in normal adrenals, benign and malignant ACTs (Betz et al., 2005), and in the NCI-H295 cell line (Ferruzzi et al., 2005). Treatment of NCI-H295 cells with PPARG activator thiazolidinediones decreased growth, invasiveness and induced their differentiation. Co-treatment with 9-cis retinoic acid and thiazolidinedione had additive effects in vitro (Betz et al., 2005; Ferruzzi et al., 2005). Expression of several genes targeted by LXRA and PPARG has been confirmed in the adrenal cortex, including ATP-binding cassette A1 and G1 (ABCA1, ABCG1), apolipoprotein E and C1 (APOE, APOC1), sterol regulatory element-binding transcription factor 1 (SREBF1), scavenger receptor-B1 (SCARB1) (Cummins et al., 2006).

We have observed the underexpression of all these target genes in ACC. Overexpression of the LXRA target CD36 (thrombospondin receptor) gene was revealed in two studies (Tömböl et al., 2009; de Reyniès et al., 2009). Significant enrichment of overexpressed genes was found on chr7q21 by GSEA, where the CD36 gene is located. The gain of chr7q21 has been described previously (Supplementary Table 3).

Overexpression of CD36 may result in enhanced cholesterol influx from high and low density lipoprotein molecules (HDL and LDL) to adrenocortical cells, however, underexpression of SCARB1 that was observed by Slater et al., (2006) may interfere with cholesterol influx from HDL. Other genes involved in cholesterol efflux including ABCA1 and ABCG1 showed significant underexpression in several studies (Velázquez-Fernández et al., 2005; Giordano et al., 2009; Tömböl et al., 2009; Soon et al., 2009; de Reyniès et al., 2009). SREBF1 (a key regulator of genes involved in fatty acid synthesis and production of ester substrates) expression was found to be significantly lower in ACCs in four studies (Giordano et al., 2003, 2009; Tömböl et al., 2009; de Reyniès et al., 2009). APOE enhances cholesterol ester accumulation in adrenocortical cells (Thorngate et al., 2002). APOE was significantly underexpressed in ACCs in four adult microarray studies (Giordano et al., 2009; Tömböl et al., 2009; Soon et al., 2009; de Reyniès et al., 2009) and in childhood ACCs, as well (West et al., 2007). These gene expression alterations affecting proteins are involved in the regulation of adrenocortical cholesterol and lipid metabolism may be related to the observation that ACCs are relatively lipid-poor in comparison with their benign counterparts (Figure 3).

Bacterial lipopolysaccharide (LPS) recognition via Toll-like receptor 4 (TLR4)

The TLR receptor family is related to interleukin 1 receptor (IL-1R) family. LPS-mediated TLR4 activation affects retinoic acid signalling, as it suppresses RXRA-mediated gene expression in the liver (Ghose et al., 2004). Several other molecules are required for the recognition of LPS by TLR4 including CD14 and lymphocyte antigen 96 (LY96) (Pålsson-McDermott and O'Neill 2004). TLR4 is expressed on human adrenocortical cells, and it is involved in immune-neuroendocrine crosstalk (Zacharowski et al., 2006). LPS directly stimulates cortisol secretion by NCI-H295 cells (Vakharia and Hinson, 2005), and IL-6 and IL-8 expression by human adrenocortical cells (Kanczkowski et al., 2009). We found significant underexpression of TLR4 in ACCs in two studies (Giordano et al., 2009; Tömböl et al., 2009) and confirmed TLR4 expression changes by QRT–PCR (data not shown). CD14 and LY96 expression was decreased in several studies.

The proinflammatory cytokine IL-1 may induce the activation of hypothalamic–pituitary–adrenal axis (Turnbull and Rivier, 1999). IL-1-mediated signalling is triggered by the binding of IL-1 to interleukin-1 receptor type 1 (IL1R1), whereas interleukin-1 receptor type 2 (IL1R2) is regarded as a decoy receptor (Subramaniam et al., 2004). We observed significant underexpression of IL1R1 in ACCs in four studies (Giordano et al., 2003, 2009; Slater et al., 2006; de Reyniès et al., 2009), and the overexpression of IL1R2 in one (Tömböl et al., 2009).

Complement system and antigen presentation

Alterations of complement system member gene expression were often observed in the investigated microarray studies. Significant underexpression of complement components involved in both classical and alternate routes of activation, for example, complement 1 q subcomponent, A and B chains (C1QA, C1QB), complement factors D and H (CFD, CFH) etc. was observed in three studies (Giordano et al., 2009; Tömböl et al., 2009; de Reyniès et al., 2009). SERPING1 (serpin peptidase inhibitor, clade G (C1 inhibitor), member (1) underexpression was validated by QRT–PCR and it is considered as a potential prognostic marker of malignancy (Fernandez-Ranvier et al., 2008a, 2008b) (Supplementary Figure 2).

Major histocompatibility complex (MHC) class II molecules are involved in the presentation exogenous antigens in the immune system, and are expressed by the zona reticularis of the human adrenal cortex (Wolkersdörfer et al., 2005). ACCs lack MHCII proteins and MHCII expression is regarded as a marker for benign lesions (Marx et al., 1996). We have observed underexpression of several MHCII antigens in ACCs (Supplementary Table 1) (Giordano et al., 2009; Tömböl et al., 2009; de Reyniès et al., 2009). By the direct comparison of GSEA and CGH results in our experimental study, we have found that MHCII underexpression correlated with the loss of chr6p21 where the cluster of these genes is harboured.

Significant underexpression of MHCII genes was observed in cortisol-producing ACTs compared with hormonally inactive ACTs. Beside the downregulation of MHCII antigens in CPA samples, overexpression of MHCI class I B (HLA-B) was observed in the study by Giordano et al. (2009). These expression changes might be related to a shift from MHCII to MHCI-mediated antigen presentation. The pathophysiological relevance of these findings is unclear.

It is hard to interpret these findings in the lack of data on the physiological or pathological relevance of these genes in adrenocortical functioning or tumours. Altogether, alterations of cytokine signalling, complement system and antigen presentation pathways suggest the involvement of immune mechanisms in ACT pathogenesis.

In conclusion, three major pathogenetic pathways were revealed both by the meta-analysis of available mRNA profiling and cytogenetic data and by our pilot study involving mRNA and CGH profiling on the same samples. The observation that the same pathways were found by both approaches supports the feasibility of in silico meta-analysis. As pathway analysis represents a more comprehensive form of analysis than the molecular studies of individual gene expression and cytogenetic alterations, this approach may reveal novel, previously unknown pathogenic pathways. As the therapeutic repertoire for the treatment of adrenocortical cancer is very limited, our results may pave the way for further in vitro studies aimed at the development of novel agents affecting these pathways.

Materials and methods

Meta-analysis of available gene expression data sets

Data sets

Whole genome raw mRNA expression data of 164 tumour (97 ACA and 67 ACC) and 18 NA samples from Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo) and ArrayExpress (http://www.ebi.ac.uk) were collected (Ye et al., 2007; de Reyniès et al., 2009, Giordano et al., 2009, Tömböl et al., 2009), reclassified and analysed (Table 2). Data on hormonal activity were unavailable in the study by de Reyniès et al., (2009).

Table 2 Characteristics of the analyzed microarray data sets

Further significant gene sets from studies where raw gene expression data were unavailable have been obtained from Oncomine Research (http://www.oncomine.org) (Giordano et al., 2003) and seven papers found by literature search (PubMed, http://www.ncbi.nlm.nih.gov/pubmed) (Table 3). These included 30 NA, 182 ACA and 91 ACC samples. Characteristics of these gene sets are presented in Table 3. Cytogenetic (CGH and FISH) data on adrenocortical tumours were acquired from 10 publications found by literature search (Kjellman et al., 1996; Figueiredo et al., 1999, 2005; Russell et al., 1999; Zhao et al., 1999, 2002; Dohna et al., 2000; Sidhu et al., 2002; Bertherat et al., 2003; Stephan et al., 2008). These included 87 ACA and 124 ACC samples (Supplementary Table 3). The CGH data set presented by Stephan et al. (2008) was accessible at GEO, and these samples were reclassified and reanalysed by applying DNA Analytics software (version 4.0.85, Agilent Technologies, Santa Clara, CA, USA).Altogether, gene expression and cytogenetic data from 42 NAs, 350 ACA and 270 ACC samples were available.

Table 3 Characteristics of significant gene sets in publications where raw whole genome microarray data were unavailable

Statistical analysis of mRNA profiling studies

Identification of gene sets differentially expressed among NA, ACA and ACC groups and among NA, IA, CPA, APA, hormone-secreting and hormonally inactive ACC groups was carried out by one-way ANOVA. Comparison of hormone-secreting and hormonally inactive benign and malignant ACTs was performed by a two-way ANOVA. Each ANOVA was followed by Tukey's post hoc test for all pairwise multiple comparisons. An unpaired t-test was used for the NA–APA and ACA–ACC comparisons for two studies where only these two groups were examined (Ye et al., 2007; de Reyniès et al., 2009). Fold change filter was set to 2-fold. Furthermore, Benjamini–Hochberg multiple testing correction was performed in each analysis to minimize false-positive cases. All statistical analyses were performed by Genespring GX 10.0 Software (Agilent Tech Inc., Santa Clara, CA, USA) Supplementary Table 7 summarises the statistical tests applied.Common significant expression changes between different studies were collected by an own programme written in C++ programme language by applying Code::Blocks software (Code::Blocks, Release 8.02, www.codeblock.org).

Gene set enrichment analysis (GSEA)

GSEA is a novel computational approach developed for the analysis of fine gene expression alterations: it analyzes gene expression data by rank statistics, and determines whether a particular set of genes is over- or underrepresented in the samples compared (Subramanian et al., 2005). GSEA was performed by GSEA Software v2.0 (www.broad.mit.edu) in pairwise comparisons of NA vs ACA, NA vs ACC and ACA vs ACC. Gene expression results derived from microarray experiments were correlated with chromosome gene sets. GSEA was performed using gene set permutation type as default and the number of permutation was set to 1000. Statistical significance levels were defined as nominal P-value <0.05 and false discovery rate (FDR) <0.25.

Correlation of GSEA results retrieved from publicly available microarray and cytogenetic data

To identify genes with the most relevant expression changes on an aberrant chromosome region, Leading Edge Analysis using GSEA Software v2.0 (www.broad.mit.edu) was performed. The leading edge subset of genes can be interpreted as the core that accounts for the significant chromosomal gene set's enrichment signal. By this approach, genes harboured on aberrant chromosome regions showing the most prominent expression changes could be identified. These gene sets were loaded onto Ingenuity Pathway Analysis (IPA) to identify pathways affected by chromosomal aberrations in ACTs.

Pilot study on ACTs involving parallel mRNA and CGH profiling

CGH was performed on 4 ACC and 7 ACA (4 IA and 3 CPA) samples that have already been studied by mRNA profiling by Tömböl et al. (2009) (GEO accession number: GSE14922). The study was approved by the Ethical Committee of the Hungarian Health Council, and informed consent was obtained from all patients involved. Tissue digestion, labelling, hybridization and data analysis of genomic DNA were performed according to the Agilent Technologies (Santa Clara, CA, USA) protocol version 2.0 for 105K arrays. The sample and reference (G1521 female DNA, Promega, Madison, WI, USA) CGH105A oligo microarray slides were scanned with Agilent Microarray scanner and data were extracted with Feature Extraction software version 9.5.1.1 including dye normalization (linear lowess). DNA Analytics software was used (version 4.0.85, Agilent Technologies, Santa Clara, CA, USA) for data analysis. The starting and ending points of the aberrations were confirmed by the ADM-2 algorithm with 6.0 threshold.

Gene expression alterations paralleling chromosome aberrations were searched for as previously described in the meta-analysis section. By applying Leading Edge Analysis, genes showing the highest gene expression alterations on aberrant chromosome regions could be selected and these were loaded onto IPA.

Pathway analysis

Ingenuity Pathways Analysis Knowledge Base is updated with current results published in journal articles and other data sources. By applying the Ingenuity Pathway Analysis (IPA) software (Ingenuity Systems, www.ingenuity.com; Redwood City, CA, USA), gene sets can be functionally annotated and biologically relevant pathways may be identified.

Gene sets established both by meta-analysis of mRNA, GSEA and CGH data sets and by our pilot experimental study were subjected to IPA and results were compared with each other.

Accession codes

Accessions

GenBank/EMBL/DDBJ

References

  1. Bertherat J, Groussin L, Sandrini F, Matyakhina L, Bei T, Stergiopoulos S et al. (2003). Molecular and functional analysis of PRKAR1A and its locus (17q22–24) in sporadic adrenocortical tumors: 17q losses, somatic mutations, and protein kinase A expression and activity. Cancer Res 63: 5308–5319.

    CAS  PubMed  Google Scholar 

  2. Betz MJ, Shapiro I, Fassnacht M, Hahner S, Reincke M, Beuschlein F . (2005). Peroxisome proliferator-activated receptor-gamma agonists suppress adrenocortical tumor cell proliferation and induce differentiation. J Clin Endocrinol Metab 90: 3886–3896.

    CAS  Article  Google Scholar 

  3. Bourcigaux N, Gaston V, Logié A, Bertagna X, Le Bouc Y, Gicquel C . (2000). High expression of cyclin E and G1 CDK and loss of function of p57KIP2 are involved in proliferation of malignant sporadic adrenocortical tumors. J Clin Endocrinol Metab 85: 322–330.

    CAS  PubMed  Google Scholar 

  4. Cummins CL, Volle DH, Zhang Y, McDonald JG, Sion B, Lefrançois-Martinez AM et al. (2006). Liver X receptors regulate adrenal cholesterol balance. J Clin Invest 116: 1902–1912.

    CAS  Article  Google Scholar 

  5. de Fraipont F, El Atifi M, Cherradi N, Le Moigne G, Defaye G, Houlgatte R et al. (2005). Gene expression profiling of human adrenocortical tumors using complementary deoxyribonucleic acid microarrays identifies several candidate genes as markers of malignancy. J Clin Endocrinol Metab 90: 1819–1829.

    CAS  Article  Google Scholar 

  6. de Reyniès A, Assié G, Rickman DS, Tissier F, Groussin L, René-Corail F et al. (2009). Gene expression profiling reveals a new classification of adrenocortical tumors and identifies molecular predictors of malignancy and survival. J Clin Oncol 27: 1108–1115.

    Article  Google Scholar 

  7. Dohna M, Reincke M, Mincheva A, Allolio B, Solinas-Toldo S, Lichter P . (2000). Adrenocortical carcinoma is characterized by a high frequency of chromosomal gains and high-level amplifications. Genes Chromosomes Cancer 28: 145–152.

    CAS  Article  Google Scholar 

  8. el Akawi Z, Napoli JL . (1994). Rat liver cytosolic retinal dehydrogenase: comparison of 13-cis-, 9-cis-, and all-trans-retinal as substrates and effects of cellular retinoid-binding proteins and retinoic acid on activity. Biochemistry 33: 1938–1943.

    CAS  Article  Google Scholar 

  9. Faggad A, Darb-Esfahani S, Wirtz R, Sinn B, Sehouli J, Könsgen D et al. (2009). Topoisomerase IIalpha mRNA and protein expression in ovarian carcinoma: correlation with clinicopathological factors and prognosis. Mod Pathol 22: 579–588.

    CAS  Article  Google Scholar 

  10. Fernandez-Ranvier GG, Weng J, Yeh RF, Shibru D, Khafnashar E, Chung KW et al. (2008a). Candidate diagnostic markers and tumor suppressor genes for adrenocortical carcinoma by expression profile of genes on chromosome 11q13. World J Surg 32: 873–881.

    Article  Google Scholar 

  11. Fernandez-Ranvier GG, Weng J, Yeh RF, Khanafshar E, Suh I, Barker C et al. (2008b). Identification of biomarkers of adrenocortical carcinoma using genomewide gene expression profiling. Arch Surg 143: 841–846.

    Article  Google Scholar 

  12. Ferruzzi P, Ceni E, Tarocchi M, Grappone C, Milani S, Galli A et al. (2005). Thiazolidinediones inhibit growth and invasiveness of the human adrenocortical cancer cell line H295R. J Clin Endocrinol Metab 90: 1332–1339.

    CAS  Article  Google Scholar 

  13. Figueiredo BC, Cavalli LR, Pianovski MA, Lalli E, Sandrini R, Ribeiro RC et al. (2005). Amplification of the steroidogenic factor 1 gene in childhood adrenocortical tumors. J Clin Endocrinol Metab 90: 615–619.

    CAS  Article  Google Scholar 

  14. Figueiredo BC, Stratakis CA, Sandrini R, DeLacerda L, Pianovsky MA, Giatzakis C et al. (1999). Comparative genomic hybridization analysis of adrenocortical tumors of childhood. J Clin Endocrinol Metab 84: 1116–1121.

    CAS  PubMed  Google Scholar 

  15. Ghose R, Zimmerman TL, Thevananther S, Karpen SJ . (2004). Endotoxin leads to rapid subcellular re-localization of hepatic RXRalpha: A novel mechanism for reduced hepatic gene expression in inflammation. Nucl Recept 2: 4.

    Article  Google Scholar 

  16. Giordano TJ, Kuick R, Else T, Gauger PG, Vinco M, Bauersfeld J et al. (2009). Molecular classification and prognostication of adrenocortical tumors by transcriptome profiling. Clin Cancer Res 15: 668–676.

    CAS  Article  Google Scholar 

  17. Giordano TJ, Thomas DG, Kuick R, Lizyness M, Misek DE, Smith AL et al. (2003). Distinct transcriptional profiles of adrenocortical tumors uncovered by DNA microarray analysis. Am J Pathol 162: 521–531.

    CAS  Article  Google Scholar 

  18. Haselbeck RJ, Ang HL, Deltour L, Duester G . (1997). Retinoic acid and alcohol/retinol dehydrogenase in the mouse adrenal gland: a potential endocrine source of retinoic acid during development. Endocrinology 138: 3035–3041.

    CAS  Article  Google Scholar 

  19. Hong F, Breitling R . (2008). A comparison of meta-analysis methods for detecting differentially expressed genes in microarray experiments. Bioinformatics 24: 374–382.

    CAS  Article  Google Scholar 

  20. Kanczkowski W, Zacharowski K, Wirth MP, Ehrhart-Bornstein M, Bornstein SR . (2009). Differential expression and action of Toll-like receptors in human adrenocortical cells. Mol Cell Endocrinol 300: 57–65.

    CAS  Article  Google Scholar 

  21. Kjellman M, Kallioniemi OP, Karhu R, Höög A, Farnebo LO, Auer G et al. (1996). Genetic aberrations in adrenocortical tumors detected using comparative genomic hybridization correlate with tumor size and malignancy. Cancer Res 56: 4219–4223.

    CAS  PubMed  Google Scholar 

  22. Laurell C, Velázquez-Fernández D, Lindsten K, Juhlin C, Enberg U, Geli J et al. (2009). Transcriptional profiling enables molecular classification of adrenocortical tumours. Eur J Endocrinol 161: 141–152.

    CAS  Article  Google Scholar 

  23. Libè R, Fratticci A, Bertherat J . (2007). Adrenocortical cancer: pathophysiology and clinical management. Endocr Relat Cancer 14: 13–28.

    Article  Google Scholar 

  24. Marx C, Wolkersdörfer GW, Brown JW, Scherbaum WA, Bornstein SR . (1996). MHC class II expression—a new tool to assess dignity in adrenocortical tumours. J Clin Endocrinol Metab 81: 4488–4491.

    CAS  PubMed  Google Scholar 

  25. Ondrey F . (2009). Peroxisome proliferator-activated receptor gamma pathway targeting in carcinogenesis: implications for chemoprevention. Clin Cancer Res 15: 2–8.

    CAS  Article  Google Scholar 

  26. Pålsson-McDermott EM, O′Neill LA . (2004). Signal transduction by the lipopolysaccharide receptor, Toll-like receptor-4. Immunology 113: 153–162.

    Article  Google Scholar 

  27. Patalano A, Brancato V, Mantero F . (2009). Adrenocortical cancer treatment. Horm Res 71: 99–104.

    CAS  PubMed  Google Scholar 

  28. Russell AJ, Sibbald J, Haak H, Keith WN, McNicol AM . (1999). Increasing genome instability in adrenocortical carcinoma progression with involvement of chromosomes 3, 9 and X at the adenoma stage. Br J Cancer 81: 684–689.

    CAS  Article  Google Scholar 

  29. Scripture CD, Sparreboom A, Figg WD . (2005). Modulation of cytochrome P450 activity: implications for cancer therapy. Lancet Oncol 6: 780–789.

    CAS  Article  Google Scholar 

  30. Shimizu M, Takai K, Moriwaki H . 2009. Strategy and mechanism for the prevention of hepatocellular carcinoma: phosphorylated retinoid X receptor alpha is a critical target for hepatocellular carcinoma chemoprevention. Cancer Sci 100: 369–374.

    CAS  Article  Google Scholar 

  31. Sidhu S, Marsh DJ, Theodosopoulos G, Philips J, Bambach CP, Campbell P et al. (2002). Comparative genomic hybridization analysis of adrenocortical tumors. J Clin Endocrinol Metab 87: 3467–3474.

    CAS  Article  Google Scholar 

  32. Skawran B, Steinemann D, Weigmann A, Flemming P, Becker T, Flik J et al. (2008). Gene expression profiling in hepatocellular carcinoma: upregulation of genes in amplified chromosome regions. Mod Pathol 21: 505–516.

    CAS  Article  Google Scholar 

  33. Slater EP, Diehl SM, Langer P, Samans B, Ramaswamy A, Zielke A et al. (2006). Analysis by cDNA microarrays of gene expression patterns of human adrenocortical tumors. Eur J Endocrinol 154: 587–598.

    CAS  Article  Google Scholar 

  34. Soon PS, McDonald KL, Robinson BG, Sidhu SB . (2008). Molecular markers and the pathogenesis of adrenocortical cancer. Oncologist 13: 548–561.

    CAS  Article  Google Scholar 

  35. Soon PS, Gill AJ, Benn DE, Clarkson A, Robinson BG, McDonald KL et al. (2009). Microarray gene expression and immunohistochemistry analyses of adrenocortical tumors identify IGF-2 and Ki-67 as useful in differentiating carcinomas from adenomas. Endocr Relat Cancer 16: 573–583.

    CAS  Article  Google Scholar 

  36. Stephan EA, Chung TH, Grant CS, Kim S, Von Hoff DD, Trent JM et al. (2008). Adrenocortical carcinoma survival rates correlated to genomic copy number variants. Mol Cancer Ther 7: 425–431.

    CAS  Article  Google Scholar 

  37. Subramaniam S, Stansberg C, Cunningham C . (2004). The interleukin 1 receptor family. Dev Comp Immunol 28: 415–428.

    CAS  Article  Google Scholar 

  38. Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA et al. (2005). Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci USA 102: 15545–15550.

    CAS  Article  Google Scholar 

  39. Thorngate FE, Strockbine PA, Erickson SK, Williams DL . (2002). Altered adrenal gland cholesterol metabolism in the apoE-deficient mouse. J Lipid Res 43: 1920–1926.

    CAS  Article  Google Scholar 

  40. Tömböl Z, Szabó P, Molnár V, Wiener Z, Tölgyesi G, Horànyi J et al. (2009). Integrative molecular-bioinformatics study of human adrenocortical tumors: microRNA, tissue specific target prediction and pathway analysis. Endocr Relat Cancer 16: 895–906.

    Article  Google Scholar 

  41. Turnbull AV, Rivier CL . (1999). Regulation of the hypothalamic-pituitary-adrenal axis by cytokines: actions and mechanisms of action. Physiol Rev 79: 1–71.

    CAS  Article  Google Scholar 

  42. Vakharia K, Hinson JP . (2005). Lipopolysaccharide directly stimulates cortisol secretion by human adrenal cells by a cyclooxygenase-dependent mechanism. Endocrinology 146: 1398–1402.

    CAS  Article  Google Scholar 

  43. Velázquez-Fernández D, Laurell C, Geli J, Höög A, Odeberg J, Kjellman M et al. (2005). Expression profiling of adrenocortical neoplasms suggests a molecular signature of malignancy. Surgery 138: 1087–1094.

    Article  Google Scholar 

  44. West AN, Neale GA, Pounds S, Figueredo BC, Rodriguez Galindo C, Pianovski MA et al. (2007). Gene expression profiling of childhood adrenocortical tumors. Cancer Res 67: 600–608.

    CAS  Article  Google Scholar 

  45. Wolkersdörfer GW, Marx C, Brown J, Schröder S, Füssel M, Rieber EP et al. (2005). Prevalence of HLA-DRB1 genotype and altered Fas/Fas ligand expression in adrenocortical carcinoma. J Clin Endocrinol Metab 90: 1768–1774.

    Article  Google Scholar 

  46. Ye P, Mariniello B, Mantero F, Shibata H, Rainey WE . (2007). G-protein-coupled receptors in aldosterone-producing adenomas: a potential cause of hyperaldosteronism. J Endocrinol 195: 39–48.

    CAS  Article  Google Scholar 

  47. Zacharowski K, Zacharowski PA, Koch A, Baban A, Tran N, Berkels R et al. (2006). Toll-like receptor 4 plays a crucial role in the immune-adrenal response to systemic inflammatory response syndrome. Proc Natl Acad Sci USA 103: 6392–6397.

    CAS  Article  Google Scholar 

  48. Zhao J, Roth J, Bode-Lesniewska B, Pfaltz M, Heitz PU, Komminoth P . (2002). Combined comparative genomic hybridization and genomic microarray for detection of gene amplifications in pulmonary artery intimal sarcomas and adrenocortical tumors. Genes Chromosomes Cancer 34: 48–57.

    CAS  Article  Google Scholar 

  49. Zhao J, Speel EJ, Muletta-Feurer S, Rütimann K, Saremaslani P, Roth J et al. (1999). Analysis of genomic alterations in sporadic adrenocortical lesions. Gain of chromosome 17 is an early event in adrenocortical tumorigenesis. Am J Pathol 155: 1039–1045.

    CAS  Article  Google Scholar 

Download references

Acknowledgements

This work was supported by the Hungarian Scientific Research Fund (OTKA, PD72306) and the Hungarian Ministry of Health (ETT 040/09)

Author information

Affiliations

Authors

Corresponding author

Correspondence to P Igaz.

Ethics declarations

Competing interests

The authors declare no conflict of interest.

Additional information

Supplementary Information accompanies the paper on the Oncogene website

Supplementary information

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Szabó, P., Tamási, V., Molnár, V. et al. Meta-analysis of adrenocortical tumour genomics data: novel pathogenic pathways revealed. Oncogene 29, 3163–3172 (2010). https://doi.org/10.1038/onc.2010.80

Download citation

Keywords

  • adrenocortical tumour
  • meta-analysis
  • mRNA profiling
  • comparative genome hybridization
  • pathway analysis

Further reading

Search

Quick links