Physiological and toxicological transcriptome changes in HepG2 cells exposed to copper

Min Ok Song, Jianying Li, Jonathan H. Freedman

Abstract

Copper is an essential trace element; however, at supraphysiological levels, it can be extremely toxic. Microarray data from HepG2 cells exposed to 100, 200, 400, and 600 μM copper for 4, 8, 12 and 24 h were generated and analyzed. Principal components, K-means, and hierarchical clustering, interactome, and pathway mapping analyses indicated that these exposure conditions induce physiological and toxicological changes in the HepG2 transcriptome. As a general trend, when the level of toxicity increases, the number and diversity of affected genes, Gene Ontology categories, regulatory pathways, and complexity of interactomes increase. Physiological responses to copper include transition metal ion binding and responses to stress/stimulus, whereas toxicological responses include apoptosis, morphogenesis, and negative regulation of biomolecule metabolism. The global gene expression profile was overlaid onto biomolecular interaction networks and signal transduction cascades using pathway mapping and interactome identification. This analysis indicated that copper modulates signal transduction pathways associated with MAPK, NF-κB, death receptor, IGF-I, hypoxia, IL-10, IL-2, IL-6, EGF, Toll-like receptor, protein ubiquitination, xenobiotic metabolism, leukocyte extravasation, complement and coagulation, and sonic hedgehog signaling. These results provide insights into the global and molecular mechanisms regulating the physiological and toxicological responses to metal exposure.

  • signal transduction pathways
  • interactome
  • Gene Ontology
  • transition metal
  • Cytoscape

the transition metal copper plays important physiological roles, serving as a cofactor in enzymes that modify neuropeptides, generate cellular energy, detoxify oxygen-derived radicals, mobilize iron, coagulate blood, and cross link connective tissue (44, 54). At higher than physiological concentrations, however, copper has a destructive potential toward cellular macromolecules. Copper participates in redox reactions that can generate reactive oxygen species (ROS), which damage lipids, proteins, and DNA. Copper can also directly bind to protein sulfhydryl and amino groups, leading to structural and functional modifications (13, 30, 38, 41, 68). Finally, copper can bind to DNA to form adducts and is involved in chromatin condensation (11, 58). Thus, it is critical for organisms to maintain homeostatic concentrations of copper, because abnormally high or low levels can lead to pathological conditions (8).

Cells defend against copper-induced toxicity by activating the transcription of stress-responsive genes, which encode proteins that repair intracellular damage or remove the metal (23, 51, 64, 67, 68). The mechanism by which copper modulates the expression of many of these genes is not yet clear. Copper can activate transcription through metal- and oxidative stress-responsive signal transduction pathways involving PKC and MAPKs (48). The activation of copper-responsive gene transcription may also be mediated by NF-κB signaling (64).

Global gene expression profiling, through the use of DNA microarrays, allows the monitoring of changes in the expression of thousands of genes and subsequently identifies novel regulatory pathways. The majority of the studies examining the genomic response to copper exposure have focused on the toxicological response (50, 65). There is a paucity of data, however, on the genomic response to physiological levels of copper.

We propose that copper modulates the activity of multiple intracellular signal transduction pathways to affect transcription. Furthermore, the pathways affected by toxic concentrations of copper may be unique from those affected by physiological levels. In the present study, transcriptomes were generated using HepG2 cells exposed to four concentrations of copper (100, 200, 400, and 600 μM) for four time periods (4, 8, 12, and 24 h). These conditions were selected based on our previous cytotoxicity results (64) and copper levels reported from human case studies. In normal human serum, copper levels range from 18.1 to 31.5 μM (39, 49, 53); however, they can be elevated under pathophysiological conditions. Elevated serum copper levels of 1,300 μg/100 ml (205 μM) were reported in a patient with hypercupremia associated with multiple myeloma (42). Hepatic copper concentrations as high as 1,142 μg/g dry tissue in a Wilson's disease patient and 4,788 μg/g dry tissue from an individual in the terminal stages of Indian childhood cirrhosis have been observed (27).

In humans, acute copper toxicity is rare; however, elevated and toxic levels of copper can be encountered as a result of environmental exposure, genetic defects, and certain neoplastic diseases (8, 26). There are several genetic diseases of copper metabolism that are characterized by elevated levels of intracellular hepatic copper: Wilson's disease, Indian childhood cirrhosis, and idiopathic copper toxicosis. Patients with these diseases present hepatic copper levels at milligram/gram concentrations (61). In addition to genetic disorders of copper metabolism, other pathological conditions, including hepatic necrosis, cholestatic cirrhosis, bile duct proliferation, hepatitis, and hepatocellular carcinoma, are associated with elevated copper levels (19, 20, 25). Environmental exposures to elevated copper levels have been reported to be as high as 160 μM in drinking water and up to 90 mM in rivers (1). In the present study, analysis of the copper transcriptomes from HepG2 cells revealed that at lower concentrations (100 and 200 μM) copper modulated the expression of genes associated with physiological adaptive responses and at higher concentrations (400 and 600 μM) copper induced toxicological responses. The present study provides insights into global, molecular mechanisms associated with copper intoxication as well as mechanisms by which cells maintain normal physiological levels of this essential metal.

MATERIALS AND METHODS

Cell culture, RNA isolation, and microarray hybridization

HepG2 cells (human hepatoma cell line, no. HB-8065, American Type Culture Collection) were grown in MEM supplemented with 10% FBS, 100 μM nonessential amino acids, 1 mM sodium pyruvate, 100 U/ml penicillin, and 100 μg/ml streptomycin (Life Technologies, Gaithersburg, MD). Cells were maintained in a humidified incubator at 37°C under 5% CO2. To prepare RNA for microarray experiments, HepG2 cells were grown until they were 50% confluent. Cells were then treated with 100, 200, 400, or 600 μM copper sulfate (Sigma-Aldrich Chemical, St. Louis, MO) for 4, 8, 12, or 24 h. These times and concentrations corresponded to exposures between a 5% lethal dose and the 50% lethal dose (64). Three independent total RNA samples were isolated from untreated and treated cells using RNeasy Mini kits following the manufacturer's instructions (Qiagen, Valencia, CA). The quality of the purified RNA was determined using a BioAnalyzer (Agilent Technologies, Palo Alto, CA), and samples were then stored at −80°C until use.

For microarray hybridizations, 100 ng of total RNA from copper-treated cells were amplified and labeled with Cy3 fluorescent dye, and a common reference pool (nontreated cells) was amplified and labeled with Cy5 using Agilent Technologies Low RNA Input Linear Amplification Labeling kit following the manufacturer's protocol. The quantity and purity of the fluorescently labeled cRNAs were evaluated using a Nanodrop ND-100 spectrophotometer (Nanodrop Technologies, Wilmington, DE), and the size distribution was evaluated using an Agilent Bioanalyzer. Equal amounts of Cy3- and Cy5-labeled cRNAs were then hybridized to Agilent's Human Microarray (∼22,000-k features) for 17 h at 65°C. The hybridized microarrays were then washed and scanned using an Agilent G2565BA scanner. Data were then extracted using Agilent Feature Extraction software. A total of 96 microarrays were analyzed in this study: four copper concentrations × four time points × three biological replicates × two dye-swap replicates. This resulted in six microarrays for each treatment condition.

Transcriptome Data Analysis

GeneSpring (version 7, Agilent Technologies) was used to identify genes that showed significant changes in gene expression with any treatment. Before the statistical modeling, a full-scale data quality assessment was applied to ensure a satisfied quality level. For the global normalization of raw microarray data, per spot- and per chip intensity-dependent (Lowess) normalization were applied (70). A two-sample t-test between treated and control samples was applied on the normalized dataset; the false discovery rate (FDR) was applied as a P value correction to handle a possible multiple testing issue. The statistical significance threshold was set at FDR P values of ≤0.05. To ensure the biological significance and comparability across platforms, a 1.5- or 2-fold change in the level of expression was also used as another cutoff (63). Differentially expressed genes were further defined to satisfy both criteria. Microarray data presented in this publication have been deposited in the National Center for Biotechnology Information's Gene Expression Omnibus (GEO) (21) and are accessible through GEO Series Accession No. GSE9539.

Principal component analysis.

Principal component analysis (PCA) was performed using the Partek Genomics Suite (Partek, St. Louis, MO) (36). Gene expression data were preprocessed and normalized as described above, and short-wide format data matrixes were constructed either with a complete microarray data set (all-gene list) or a differentially regulated genes dataset (≥2-fold, FDR P ≤ 0.05). PCA was performed based on the correlation matrix, which was standardized to a mean of 0 and a SD of 1. The number of principal components was determined through a Scree plot to ensure that sufficient variability had been captured (∼80% or above for differentially expressed genes). The three components with the largest eigenvalues were plotted.

Cluster analysis.

Hierarchical (average linkage) and K-means clustering of 2,132 differentially expressed genes (>1.5-fold, P ≤ 0.001, in at least 4 of 16 conditions) was performed using Cluster 3.0 (18). As similarity measures, correlation (uncentered) was used for hierarchical clustering, and Euclidean distance was used for K-means clustering. For K-means clustering, the number of clusters (K) was 13 for genes and 2 for experimental conditions. We combined clusters 12 and 13 together because they had similar expression profiles and enriched Gene Ontology (GO) categories. Clustering results were visualized with Java TreeView 1.0.7 (59). GO analysis of the genes in the various clusters was performed using the Gene Ontology Tree Machine (University of Tennessee and Oak Ridge National Laboratory) (72).

Ingenuity pathway analysis.

The Ingenuity Pathway Analysis (IPA) platform was used to identify significant canonical or functional pathways from the IPA library of pathways (http://www.ingenuity.com). A Fisher's exact test was used to determine the probability that the association between the copper-responsive genes and the canonical or functional pathway occured by chance alone. The initial expression data matrix consisted of 12,266 genes, which corresponds to genes with P ≤ 0.05 in at least 1 of 16 conditions (see Supplemental Material, Additional Data File 1).1 The Benjamini and Hochberg FDR was applied as a multiple testing correction (10). This list was uploaded to Ingenuity, and a cutoff of >1.5-fold change was then applied. IPA identified significantly differentially regulated genes (focus genes) and then overlaid them onto the Ingenuity Pathway Knowledge Base. Canonical pathways and functional networks associated with these genes were generated based on their connectivity.

Interactome analysis.

Cytoscape with the jActiveModule plug-in was used to identify neighborhoods in the networks associated with differentially expressed genes (34, 62). The merged human interactome developed by Garrow (http://www.cytoscape.org/cgi-bin/moin.cgi/Data_Sets), which contains 10,344 nodes and 53,526 interactions, was used in this analysis. To identify interactomes, a list of 12,266 genes that had P ≤ 0.05 in at least 1 of 16 conditions and had expression data for all 16 conditions was used (see Supplemental Material, Additional Data File 1). Two expression data matrixes were generated: 1) a concentration-based matrix, in which the genes identified under each of the four exposure times at a single concentration were combined, and 2) a time-based matrix, in which the genes identified under each of the four concentrations at a single time were combined.

RESULTS

Gene Expression Profile

The analysis of transcriptomes of HepG2 cells exposed to 100, 200, 400, or 600 μM copper for 4, 8, 12, or 24 h identified significantly differentially expressed genes. A total of 2,257 unique genes were differentially expressed by >2-fold (FDR P ≤ 0.05) among the 16 exposure conditions: 1,088 upregulated genes and 1,169 downregulated genes. Exposure to 600 μM copper for 12 h increased the expression of the largest number of genes (791 genes), whereas exposure to 100 μM copper for 12 h affected the fewest (6 genes; Table 1 and Supplemental Material, Additional Data File 2).

View this table:
Table 1.

Summary of differentially expressed genes

Most of the genes upregulated by lower copper concentrations (100 and 200 μM) at all exposure times were metallothionein (MT) isoforms, which may be a physiological response to maintain copper homeostasis (50). Heat shock proteins (HSPs) were also upregulated at 200 μM copper. In addition, HSPs showed the highest fold change after exposure to 400 and 600 μM copper for 4 and 8 h. Upregulated genes at the higher copper concentrations with the largest fold changes in expression (within the top 35) included Bcl-2-associated athanogene 3 (BAG3), suppressor of cytokine signaling 3 (SOCS3), IL-8, and growth arrest and DNA damage-inducible-γ (GADD45G) (apoptosis); the glutamate-cysteine ligase modifier subunit (GCLM) (cysteine metabolism); very-low-density LDL receptor (VLDLR) (lipoprotein binding); IL-8 and cysteine-rich angiogenic inducer 61 (CYR61) (morphogenesis); dual-specificity phosphatase (DUSP)1, DUSP5, and DUSP13 (MAPK signaling); Fos (FOS), EGF receptor 1 (EGR1), v-maf musculoaponeurotic fibrosarcoma oncogene homolog B (MAFB), nuclear receptor 4A1 (NR4A1), paired like homeodomain factor 1 (PROP1), transforming growth factor-β1 (TGFB1), basic leucine zipper transcription factor, ATF-like (BATF), and CaM-binding transcription activator 2 (CAMTA2) (regulation of transcription); and chemokine (C-X-C motif) ligand 2 (CXCL2), DnaJ (Hsp40) homolog B1 (DNAJB1), HSPA1A, HSPA1L, HSPA6, HSPH1, neutrophil cytosolic factor 2 (NCF2), and adrenomedullin (ADM) (response to stress) (Supplemental Material, Additional Data File 2). Most of these biological processes or molecular functions are associated with toxicological responses.

Significantly downregulated genes were identified only at the higher copper concentrations (400 and 600 μM) except for one gene, THC1991570 [G protein-coupled receptor 110 (GPR110)], which was identified at 200 μM (24 h), suggesting that the suppression of transcription is a toxicological response. The genes that showed the largest fold decrease in expression included apolipoprotein A-V (APOA5), apolipoprotein C-III (APOC3), lipase A (LIPA), solute carrier family 27A2 (SLC27A2), and phospholipase A2, group XIIB (PLA2G12B) (lipid metabolism and transport); arginase 1 (ARG1) and phenylalanine hydroxylase (PAH) (organic acid metabolism); fibrinogen α-chain (FGA), fibrinogen β-chain (FGB), fibrinogen γ-chain (FGG), erythropoietin (EPO), and angiopoietin-like 1 (ANGPTL1) (circulation, coagulation, and wound healing); haptoglobin (HP) and haptoglobin-related protein (HPR) (hemoglobin binding); ankyrin repeat domain 15 (ANKRD15), BRCA1-associated RING 1 (BARD1), and centromere protein F (CENPF) (regulation of the cell cycle); and serpin peptidase inhibitor A4 (SERPINA4), serpin peptidase inhibitor I1 (SERPINI1), and reversion-inducing cysteine-rich protein with kazal motifs (RECK) (serine -type endopeptidase inhibitor activity) (Supplemental Material, Additional Data File 2).

PCA

To examine the relationship among the 16 treatment conditions, PCA was performed using both whole microarray and differentially expressed gene datasets. The first three principal components were visualized, and the genes that significantly contributed to these principal components were identified. For the whole microarray dataset, the first three principal components explained only 58.9% of the variability in the data; however, a separation between the lower (100 and 200 μM) and higher (400 and 600 μM) copper concentrations along the first principal component was observed (Fig. 1A). This trend was more clearly demonstrated in the analysis of the differentially expressed gene dataset, where the first three principal components represented 88.7% of the variability (Fig. 1B). The lower copper concentrations were tightly grouped in both analyses. These results suggested that there are similar transcriptional profiles after exposure to the low copper concentrations. In contrast, the PCA results suggested that the expression profiles for higher copper concentrations were different from those for lower copper concentrations. Furthermore, the expression profile for 400 μM was dissimilar to that of the 600 μM profile.

Fig. 1.

Principal component analysis of copper-responsive genes. A and B: three-dimensional representations of the first three principal components for all genes (A) and two views of those that have a ≥2-fold (false discovery rate P ≤ 0.05) change in expression (B). Copper concentrations are represented by red symbols (100 μM), blue symbols (200 μM), green symbols (400 μM), and purple symbols (600 μM). Cells were exposed to the four concentrations of copper for 4 h (triangles), 8 h (squares), 12 h (diamonds), and 24 h (hexagons).

Sixty genes that had the largest contribution to each principal component (both positive and negative) using the differentially expressed gene dataset were identified (Supplemental Material, Additional Data File 3). GO analysis showed that the genes contributing to the principal components were mostly associated with cellular lipid metabolism, cholesterol biosynthesis, complement activation, fatty acid binding, lipid transport, mitosis, positive regulation of transcription, protein folding, small GTPase-mediated signal transduction, and xenobiotic metabolism. These molecular functions and biological processes were related with a variety of physiological and toxicological responses and were consistent with those of differentially expressed genes (Tables 2 and 3).

View this table:
Table 2.

Significantly enriched GO categories for upregulated genes

View this table:
Table 3.

Significantly enriched GO categories for downregulated genes

GO Analysis of Differentially Expressed Genes

GO analysis was used to place the gene expression data into a biological and functional context. Summaries of the enriched GO categories (P ≤ 0.005) for up- and downregulated genes are shown in Tables 2 and 3, respectively. The genes that are contained within the GO categories can be found in the Supplemental Material (Additional Data File 4). The number and diversity of enriched GO categories increased in a dose- and time-dependent manner for both up- and downregulated genes. That is, the number of differentially expressed genes increased and the cognate number of enriched GO categories increased as copper exposures went from physiological to toxicological.

At lower copper concentrations (100 and 200 μM), transition metal ion binding was the only GO category identified for upregulated genes at all exposure times. For genes upregulated by 400 and 600 μM copper at all exposure times, enriched GO categories also included transition metal ion binding and responses to stress/stimulus. In addition, several GO categories were significantly enriched at both high concentrations, including cysteine metabolism, kinase inhibitor activity, transcription corepressor activity, and cell cycle arrest at 4 h of exposure; cysteine metabolism, cell fate determination, MAPK phosphatase activity, protein folding, and protein dimerization activity at 8 h of exposure; MAPK phosphatase activity, cysteine metabolism, glutathione biosynthesis, muscle cell differentiation, and regulation of apoptosis at 12 h of exposure; and cell fate determination, hemopoiesis, positive regulation of cell proliferation, and protein dimerization activity at 24 h of exposure (Table 2). There were also enriched GO categories that mapped to specific treatment conditions: oxygen and ROS metabolism and IGF binding (400 μM, 8 h); ubiquitin-protein ligase activity (600 μM, 8 h); endosome organization and biogenesis (600 μM, 12 h); and angiogenesis and glycerol kinase activity (600 μM, 24 h). LDL receptor activity, ribosome assembly, and small GTPase-mediated signal transduction were the enriched GO categories that mapped at two or three conditions of higher copper concentrations (Table 2).

For genes downregulated by 400 and 600 μM copper, most of the identified GO categories were related to biomolecule metabolism including binding and transport, polysaccharide metabolism, alcohol metabolism, lipid transport and metabolism, amino acid derivative metabolism, and hormone metabolism (Table 3). A larger number of enriched GO categories identified for the 400 μM copper treatment also mapped at 600 μM copper. In addition, as the exposure time increased, there was a concomitant increase in the diversity of GO categories and the number of genes contained in each category. Some of the enriched GO categories were mapped at specific conditions: negative regulation of the Wnt receptor signaling pathway (400 μM, 8 h); platelet activation (400 μM, 12 h); phosphoinositide-mediated signaling (600 μM, 8 h); inositol 1,4,5-trisphosphate receptor activity (600 μM, 12 h); and B cell-mediated immunity, complement activation, and xenobiotic metabolism (600 μM, 24 h). Hemoglobin binding and DNA repair were mapped at two conditions of higher copper concentrations (Table 3).

These results clearly show that copper toxicity results in the disruption of biomolecule metabolism, regulation of the cell cycle and transcription, and affects the expression of genes associated with multiple intracellular signal transduction pathways.

Cluster Analysis

K-means clustering for 16 experimental conditions and 2,312 differentially expressed genes resulted in 12 clusters for genes and 2 clusters for experimental conditions (Fig. 2). The experimental conditions of 400 and 600 μM copper (at all exposure times except 4 h) and 100 and 200 μM copper formed separate clusters. These clusters corresponded to physiological and toxicological responses to copper exposure.

Fig. 2.

K-means clustering of copper-responsive genes. K-means clustering was performed with 2,132 genes, which were differentially regulated by >1.5-fold in at least 4 of 16 conditions and have >80% expression data in 16 conditions, using Cluster 3.0. Clustering results were visualized with Java Treeview 1.0.7. The experimental cluster is shown on the horizontal axis and each gene cluster is marked by a number (clusters 1–12). We performed Gene Ontology analysis with gene lists from each cluster through the Gene Ontology Tree Machine. Gene ontologies and associated genes can be found in Table 4 and the Supplemental Material (Additional Data File 5), respectively.

Genes that were downregulated by 400 and 600 μM copper at 12 and 24 h of exposure grouped in clusters 1 and 8. Cluster 1 includes GO categories of alcohol dehydrogenase activity, steroid biosynthesis, cell cycle arrest, and estrogen and xenobiotic metabolism (Table 4; the genes associated with each cluster are listed in Supplemental Material, Additional Data File 5). Genes in cluster 8 are related to lipoprotein metabolism, glutathione transferase activity, and blood pressure regulation (Table 4). Another group of genes that were downregulated by 400 and 600 μM copper at 4, 8, and 12 h exposures were in cluster 2. Their biological and molecular functions included phosphoinositide-mediated signaling, cell cycle, and nucleotide biosynthesis (Table 4).

View this table:
Table 4.

Enriched GO categories for K-means clusters

Genes that were upregulated by 600 μM copper at 4, 8, and 12 h exposure grouped into cluster 3 and had biological functions: ubiquitin cycle, transcriptional repressor activity, chemokine activity, and response to stimulus (Table 4). MT genes grouped into cluster 5 showed the highest sensitivity to copper exposure and were upregulated under all treatment conditions. Cluster 7 was composed of genes upregulated predominately by 600 μM copper for 12 and 24 h. It was enriched in GO categories of IκB kinase/NF-κB cascade, PKC inhibitor activity, regulation of cell growth, cytoskeletal protein binding, and nucleotide biosynthesis (Table 4). Genes that were upregulated by 400 and 600 μM of copper at most of the exposure times grouped into clusters 9 and 11. Genes in cluster 9 included muscle cell differentiation, regulation of cell growth, IGF binding, regulation of transcription, and MAPK phosphatase activity. Biological and molecular functions of the genes in cluster 11 included cell fate determination, notch binding, activation of MAPKKK activity, cell cycle arrest, cysteine metabolism, protein kinase inhibitor activity, and LDL receptor activity.

Many of the downregulated genes grouped together in clusters 6 and 12. Each of these clusters had GO categories associated with immune function, such as histamine receptor activity in cluster 6 and the immune response in cluster 12.

Average linkage hierarchical clustering of the dataset used in the K-means analysis was also performed (Fig. 3). As shown in the experimental conditions dendrogram, the higher (400 and 600 μM) and lower (100 and 200 μM) copper concentrations were clearly separated, which is consistent with PCA and K-means clustering. At higher copper concentrations, upregulated genes formed seven clusters and down-regulated genes formed three clusters (Fig. 3). All of the clusters, except clusters 4 and 5, have correlations of >0.94 on the gene array dendrogram. Correlations for clusters 4 and 5 were 0.84 and 0.90, respectively.

Fig. 3.

Hierarchical clustering of copper-responsive genes. Average linkage hierarchical clustering was performed with the dataset used in the K-means analysis (2,132 genes, >1.5-fold change in expression in at least 4 of 16 conditions and have >80% expression data in 16 conditions). Clustering results were visualized with Java Treeview 1.0.7. The gene cluster dendrogram is shown on the vertical axis, and the experimental cluster dendrogram is on the horizontal axis. Gene Ontology analysis was performed with gene lists from each cluster through the Gene Ontology Tree Machine. The results of this analysis are shown in Table 5 and the Supplemental Material (Additional Data File 5).

Enriched GO categories for the hierarchical clusters are shown in Table 5 (genes associated with each cluster are listed in Supplemental Material, Additional Data File 5). While most of the GO categories identified in the hierarchical clusters overlap with those from K-means clustering, there were several that appeared only in the hierarchical clusters. Clusters of downregulated genes had GO categories of electron transport and EGF receptor activity. GO categories of upregulated genes included calcium-dependent protein binding, cellular morphogenesis, chromosome organization and biogenesis, glutathione disulfide oxidoreductase activity, G protein-coupled receptor binding, histone deacetylase activity, protein modification, and regulation of proteolysis.

View this table:
Table 5.

Enriched GO categories for hierarchical clusters

Integration of Microarray Data Into Protein Interaction Networks

Interactomes were identified based on copper concentration or exposure time using Cytoscape and the jActiveModule (34). The top 15 interactomes for each copper concentration or each exposure time are shown in the Supplemental Material (Supplemental Tables 1 and 2, respectively). GO analysis of the core genes in the copper concentration interactomes showed that the genes were associated with cell ion homeostasis and protein biosynthesis at lower copper concentrations (100 and 200 μM) and responses such as sulfur compound biosynthesis, MAPK signaling, and transcriptional repressor activity at higher concentrations (400 and 600 μM) (Supplemental Material, Supplemental Table 1). Genes of interactomes at higher copper concentrations included IL-8, heme oxygenase (decycling) 1 (HMOX1), BRCA1, and ubiquitin-conjugating enzyme E2C (UBE2C), which were involved in xenobiotic metabolism, protein ubiquitination, leukocyte extravasation, and hypoxia signaling. Similar to what was observed above, as the concentration of copper increased, there was an increase in the number of genes linked to each of the core genes. GO analysis of interactomes by exposure time showed some exposure time-specific GO categories, including positive regulations of the IκB kinase/NF-κB cascade at 4 and 8 h and structural constituent of the ribosome at 8 h (Supplemental Material, Supplemental Table 2).

The first neighbors of each module (≥1.5-fold up- or downregulated) were also identified, and GO analysis of the neighbors was performed (Supplemental Material, Supplemental Tables 1 and 2 and Additional Data File 6). The number of differentially expressed first neighbors per each active module increased in a dose- and exposure time-dependent manner, similar to what was observed above. The enriched GO categories for each copper concentration or exposure time from the interactome analysis were consistent with those obtained using the differentially regulated genes (Tables 2 and 3).

Pathway Mapping

IPA revealed that copper significantly (P ≤ 0.05) affected the expression of genes participating in various canonical signaling pathways (Table 6). IL-10 signaling (anti-inflammatory action) was the most prevalent and was identified in 9 of 16 conditions. EGF, neurotrophin/tyrosin receptor kinase (TRK), and IL-2 pathways were identified only at lower copper concentrations (100 and 200 μM). Toll-like receptor, IL-10, IL-6, hypoxia, and IGF-I signaling mapped to both low and high copper concentrations. The primary focus genes modulating these signaling pathways at lower copper concentrations were JUN and FOS (Fig. 4; Supplemental Material, Additional Data File 7). Death receptor, xenobiotic metabolism, and protein ubiquitination signaling were identified only at higher copper concentrations. This further supports the hypothesis that copper modulates the expression of genes associated with toxic responses at higher concentrations. Leukocyte extravasation and sonic hedgehog signaling also mapped to higher copper concentrations. This is the first observation that copper exposure may modulate leukocyte extravasation and sonic hedgehog signaling pathways.

Fig. 4.

Representative Ingenuity Pathway Analysis (IPA) networks. Jun and Fos centered networks identified by IPA showing the interaction between the significantly regulated genes at 400 μM copper after 4 h of exposure. Associated genes can be found in the Supplemental Material (Additional Data File 7). See text for gene descriptions.

View this table:
Table 6.

Significant canonical pathways

Functional networks with a significance of P ≤ 0.05 were also identified. These networks were associated with cellular development, cellular growth and proliferation, gene expression, cell death, cell cycle, cell signaling, lipid metabolism, amino acid metabolism, carbohydrate metabolism, cell-to-cell signaling and interactions, DNA replication, recombination and repair, posttranslational modification, and free radical scavenging (Table 7; Supplemental Material, Additional Data File 7).

View this table:
Table 7.

Significant functional networks

DISCUSSION

Copper is an essential metal; however, excessive levels can lead to intracellular toxicity and pathologies. Transcriptomes were generated from HepG2 cells exposed to four concentrations of copper for four time periods. Expression profiling showed that the number of up- and downregulated genes increased as concentration and/or exposure time increased (Table 1). There was also a concomitant increase in the number and diversity of GO categories and interacting partners (Tables 2 and 3 and Supplemental Material, Supplemental Tables 1 and 2). Principal components, K-means and hierarchical clustering, Cytoscape, and Ingenuity pathway analyses indicated that the exposure conditions induce physiological responses at low (100 and 200 μM) copper concentrations and toxicological responses at high (400 and 600 μM) copper concentrations.

Physiological Responses to Copper Exposure

HepG2 cells were exposed to levels of copper that occur in the environment and that humans may encounter (1). Approximately 10 and 20 genes were upregulated at 100 and 200 μM copper, respectively. At these concentrations, the number of differentially expressed genes and the diversity of GO categories were unaffected by exposure time (Tables 13). This suggests that cells can adequately accommodate these concentrations of copper to maintain the metal at homeostatic, nontoxic levels. In addition, the lack of time dependence in the response observed at 100 and 200 μM copper suggests that longer exposures would not alter the expression pattern. That is, metal chelation by MTs and the activity of copper transporters are sufficient to protect cells and cells will not be overwhelmed at these concentrations.

The dominant physiological response to copper exposure was an increase in the copper binding capacity. Genes encoding the metal-binding protein MT were the most responsive/sensitive to copper exposure. Changes in multiple MT genes (MT1A, MT1B, MT1E, MT1F, MT1G, MT1J, MT1K, MT1X, and MT2A) were observed at low copper concentrations at all exposure times. To maintain homeostatic levels of copper, cells use a combination of metal-regulated import, export, and sequestration mechanisms (15). In most organisms, MTs play central roles in the homeostasis of essential metals such as zinc and copper (17, 24, 66). Additionally, pathway mapping showed that neurotrophin/TRK and EGF signaling were affected at lower copper concentrations. Genes that are associated with these pathways include FOS, JUN, and phosphoinositide 3-kinase regulatory subunit 3 (PIK3R3) (Supplemental Material, Additional Data File 7), factors that have been implicated in the regulation of MT transcription (3, 16).

Toxicological Responses to Copper

HepG2 cells were exposed to supraphysiological levels of copper, which may occur in the environment and in cases of human genetic disease. At high copper concentrations, time significantly affected the numbers of differentially expressed genes (Table 1), suggesting that the toxicogenomic response to copper is a product of both concentration and exposure time. In addition, the number of downregulated genes increased only at toxic levels, suggesting that the suppression of gene expression by copper may be a toxicological response.

The cellular and molecular mechanisms underlying copper-regulated gene expression and toxicity have been investigated in yeast, mouse fibroblasts, and rodent strains with mutations in Atp7b (5, 28, 33, 69). Atp7b−/− mice demonstrate intracellular copper accumulation, low serum oxidase activity, and increased copper excretion in the urine and liver pathology, similar to Wilson's disease patients (12, 32). Transcriptome analysis of the Atp7b−/− mouse liver revealed copper-induced alterations of lipid metabolism and cholesterol homeostasis, which are also observed in Wilson's disease patients (33). In addition, MT genes and genes associated with the cell cycle and chromosome structure were upregulated. Genes encoding proteins involved in cholesterol metabolism were significantly downregulated in the Atp7b−/− liver (32, 33). Similar changes were observed in the levels of expression for these genes in copper-exposed HepG2 cells. MT genes were significantly upregulated at all copper concentrations. Pathways associated with lipid metabolism, cholesterol synthesis, and the cell cycle were significantly mapped at only higher copper concentrations (400 and 600 μM).

Transcriptome changes in HepG2 cells after exposure to 100 μM copper for 0–72 h have also been studied (50). This study (50) revealed that after 24 h of exposure, copper significantly upregulated genes involved in heavy metal detoxification (MTs), oxidative stress, protein modification/renaturation/ubiquitination, electron transport, signaling, and glutathione biosynthesis (>3-fold, P < 0.05). The expression data presented in this report agreed with this study in terms of heavy metal detoxification, where MTs were significantly upregulated at all 16 conditions. However, most of the other significant upregulated genes in the other study (50) were only affected after treatments with higher copper concentrations (200, 400, and 600 μM). HSP genes (protein modification) and HMOX1 (oxidative stress) were upregulated by 200 μM or higher copper concentrations. Furthermore, most of the genes involved in glutathione biosynthesis, the ubiquitin pathway, and signaling in the previous study (50) were upregulated only by high copper concentrations (400 and 600 μM) in the present study. The difference in response may be attributed to differences in cell culturing. Muller et al. (50) used serum-free medium (MEM supplemented with l-glutamine) for incubation with copper, whereas we used MEM with 10% heat-inactivated FBS. Cells treated with metal in serum-containing medium may be exposed to a lower concentration of metal than those in serum-free medium. There are components in serum, including α-fetoprotein and albumin, that bind copper with high affinities (4, 55). This effectively reduces the amount of copper that is available for cellular uptake.

Toxic concentrations of copper upregulated genes associated with transcription regulation, apoptosis, the MAPK cascade, and morphogenesis and downregulated genes involved in the regulation of DNA replication, DNA damage response/signal transduction, and biomolecule metabolism (Tables 2 and 3). A continued examination of the relation between copper exposure and these processes will provide insights into the mechanisms of copper hepatotoxicity.

Molecular Mechanisms of Copper Toxicity

When HepG2 cells are exposed to toxic concentrations of copper, there is a delay in cell cycle progression and an increase in cell death (7). In trout hepatocytes, copper exposure leads to cell death through ROS formation (46). In the present study, genes implicated in caspase activity, cell cycle arrest, and cyclin-dependent protein kinase inhibition were significantly upregulated. Likewise, genes associated with caspase inhibitor activity and negative regulation of caspase activity were significantly downregulated (Tables 4 and 5 and Supplemental Material, Additional Data File 5). Interactome, clustering, IPA, and GO analyses indicated that copper modulates death receptor and TGF-β1 signaling pathways. TGF-β signaling cooperates with the death receptor apoptotic pathway (Fas and TNF) and intracellular modulators of apoptosis (p38 and NF-κB) (22, 60). Thus, our results suggest that copper-induced apoptosis may be caused by modulation of the death receptor cascade and TGF-β1 signaling.

There is a paucity of data describing the adverse effects of copper exposure on mammalian development, and the molecular mechanisms have not been elucidated. In pregnant rats, copper exposure caused retardation of embryonic growth and differentiation, particularly affecting the neural tube (29). Copper is also a potent teratogen for amphibians (31, 45). There are several reports describing the effects of copper on invertebrate development. Copper induces developmental abnormalities or arrest in the sea urchin, oyster, crab, sea squirt, and in insects (2, 9, 40, 56, 57, 71). Clustering and GO analyses showed that at high concentrations, copper upregulated genes associated with cell fate determination and differentiation, embryonic development, and cell polarity, including IFN-related developmental regulator 1 (IFRD1), myeloid cell leukemia sequence 1 (MCL1), jagged 1 (JAG1), delta-like 1 (DLL1), tissue inhibitor of metalloproteinase 1 (TIMP1), MAFB, adenylate cyclase-associated protein 1 (CAP1), FYVE, RhoGEF, PH domain-containing 6 (FGD6), and IL-11 (Tables 4 and 5 and Supplemental Material, Additional Data File 5). JAG1 and DLL1 participate in the notch signaling pathway, which affects the implementation of differentiation, proliferation, and apoptotic programs, providing a general developmental tool to influence organ formation and morphogenesis (6, 52). In addition, IPA revealed that copper could significantly modulate sonic hedgehog signaling, which is critical in vertebrate development (14). These results suggest that the modulation of notch and sonic hedgehog signaling may be components of molecular mechanisms underlying copper-induced developmental abnormalities.

Effect of Copper on Signal Transduction Pathways and Transcription

Previous studies have confirmed the effects of metals on the MAPK and NF-κB pathways (for reviews, see Refs. 48 and 68). In addition to these pathways, clustering analyses combined with GO and IPA indicated that copper modulates hypoxia, Toll-like receptor, IGF-I, EGF, death receptor, TGF-β, notch, and sonic hedgehog signaling pathways (Tables 6 and 7). Copper significantly upregulated the expression of transcription factors including FOS, FOSB, FOSL1, JUN, JUNB, MAFB, MAFK, MAFG, and ATF3, which regulate cell development and differentiation. These results suggest that copper toxicity may be a consequence of its ability to disrupt the normal activity of multiple intracellular signal transduction pathways and transcription factors.

Toxic concentrations of copper also caused a suppression of gene expression. Clustering and GO analysis suggested that the mechanistic cause of this suppression includes the upregulation of genes associated with histone deacetylase activity (HDAC4 and HDAC10). The expression of jumonji domain-containing 2A (JMJD2A), a trimethylation-specific demethylase for histone and a transcriptional repressor, also increased. Copper also caused increased expression of Bcl-2-associated transcription factor 1 (BCLAF1), DNA damage-inducible transcript 3 (DDIT3), E2F transcription factor 6 (E2F6), and ring finger protein 12 (RNF12), which are transcription repressors and corepressors. Copper can induce histone hypoacetylation by directly inhibiting histone acetyltransferase activity or via oxidative stress (37, 43). Thus, copper-induced transcriptional repression may be caused by changes in the chromatin structure and alterations in the transcriptional machinery. Additional analyses of target genes and pathways related with the transcriptional repression will further explain the effect of copper on transcription.

Conclusions

In this report, the genomic responses to physiological and toxicological levels of copper were defined. At physiological levels, the primary response is an increase in the cell's capacity to bind/sequester copper. At toxic levels, there is a disruption in cell signaling, suppression of transcription, and increased cell death. It is possible that similar transcriptional responses would be observed in cells exposed to low or physiological concentrations of zinc. Similar to copper, intracellular homeostatic levels of zinc are maintained through the action of MT and zinc transporters (47). Thus, increases in MT mRNA levels would be expected in cells exposed to zinc. Changes in the levels of expression for genes involved in transport, ion homeostasis, metabolism, and responses to oxidative stress were observed in studies comparing transcriptome profiles of yeast exposed to copper and zinc. However, significantly different responses in genes associated with ribosome biogenesis and carbohydrate and glucose metabolism were observed. In addition, deletome analysis indicates that genes involved in vacuole organization are essential for survival in the presence of zinc, while copper-binding transcription factors are required to protect cells from copper toxicity (35). These results suggest that copper and zinc will have overlapping but distinct transcriptional profiles.

GRANTS

This work was supported (in part) by National Institute of Environmental Health Sciences Grants U19-ES-011375, P42-ES-010356, and Z01-ES-102045 and by the Intramural Research Program of the National Institutes of Health. RNA labeling, microarray hybridization, and data extraction were performed by Cogenics (Morrisville, NC).

Footnotes

  • 1 Supplemental Material for this article is available online at the Physiological Genomics website.

  • Address for reprint requests and other correspondence: J. H. Freedman, Laboratory of Molecular Toxicology, National Institute of Environmental Health Sciences, Mail Drop E1-05, PO Box 12233, 111 T.W. Alexander Dr., Research Triangle Park, NC 27709 (e-mail: freedma1{at}niehs.nih.gov).

REFERENCES

View Abstract