Global Gene Expression Analysis Reveals Differences in Cellular Responses to Hydroxyl- and Superoxide Anion Radical–Induced Oxidative Stress in Caco-2 Cells

Abstract

Reactive oxygen species–induced oxidative stress in the colon is involved in inflammatory bowel diseases and suggested to be associated with colorectal cancer risk. However, our insight in molecular responses to different oxygen radicals is still fragmentary. Therefore, we studied global gene expression by an extensive time series (0.08, 0.25, 0.5, 1, 2, 4, 8, 16, or 24 h) analyses in human colon cancer (caco-2) cells after exposure to H₂O₂ or the superoxide anion donor menadione. Differences in gene expression were investigated by hybridization on two-color microarrays against nonexposed time-matched control cells. Next to gene expression, correlations with related phenotypic markers (8-oxodG levels and cell cycle arrest) were investigated. Gene expression analysis resulted in 1404 differentially expressed genes upon H₂O₂ challenge and 979 genes after menadione treatment. Further analysis of gene expression data revealed how these oxidant responses can be discriminated. Time-dependent coregulated genes immediately showed a pulse-like response to H₂O₂, while the menadione-induced expression is not restored over 24 h. Pathway analyses demonstrated that H₂O₂ immediately influences pathways involved in the immune function, while menadione constantly regulated cell cycle–related pathways Altogether, this study offers a novel and detailed insight in the similarities and differences of the time-dependent oxidative stress responses induced by the oxidants H₂O₂ and menadione and show that these can be discriminated regarding their modulation of particular colon carcinogenesis–related mechanisms.

hydroxyl radicals, superoxide anion radicals, oxidative stress, caco-2 cells, global gene expression

The balance between oxidant and antioxidant activities determines the extent of oxidative damage induced by reactive oxygen species (ROS) in cells or tissues. Oxidative stress may occur in almost any tissue, but particularly, the imbalance between high oxidant exposure and antioxidant capacity in the colon has been linked to increased cancer risk (Hussain et al., 2003; Sanders et al., 2004). Especially, patients with inflammatory bowel diseases, accompanied by oxidative stress (Pavlick et al., 2002), are at increased risk for developing colorectal cancer (Itzkowitz and Yio, 2004). The most relevant forms of ROS, namely the hydroxyl (HO^•) and superoxide anion ${(\text{O}}_2^{·-})$ radicals, have preferred cellular targets and react with different kinetics with antioxidants enzymes (Winterbourn and Metodiewa, 1999). Low levels of oxidative stress activate the transcription of genes encoding proteins that participate in the defense against oxidative injuries, oxidative damage repair mechanisms, and apoptosis (D’Autreaux and Toledano, 2007). In the early response to increased oxidative stress in mammalian cells, the nuclear factor (erythroid-derived 2)-like 2 (NFE2L2 or NRF2) represents a crucial sensor (D’Autreaux and Toledano, 2007). After nuclear accumulation of NRF2 (Dinkova-Kostova et al., 2005; Kobayashi and Yamamoto, 2006), transcriptional activation via the antioxidant-responsive element results in the induction of antioxidants enzymes, including the superoxide dismutase (SOD1, 2, and 3), catalase (CAT), and glutathione peroxidase. Also, the small G-protein Ras is an upstream signaling element during oxidative stress. It activates activator protein 1 and nuclear factor-κB (NF-κB), which in turn are involved in the activation of other antioxidant genes, including thioredoxin (Morel and Barouki, 1999). In case the first line of defense against oxidative damage is insufficient to prevent the induction of DNA damage, several other (signaling) processes are initiated. First, the cells can repair the oxidant-induced DNA damage by base excision repair (Slupphaug et al., 2003). Furthermore, in order to allow more time for repair of damages, cell proliferation can be blocked at several phases of the cell cycle, such as G₁–S transition, S phase, and G₂–M transition (Clopton and Saltman, 1995).

Since our insight in the molecular pathways involved in the biological response to different oxygen radicals is still fragmentary and we do not understand the cascades of biological effects induced by oxidative stress, we are unable to assess differential health risks of increased cellular levels of HO^• and ${\text{O}}_2^{·-}$⁠. Therefore, we compared global gene expression changes in human colon carcinoma (caco-2) cells, as an in vitro model for the target organ in vivo, by extensive time series analyses using DNA microarrays following exposure to H₂O₂ or menadione, resulting in respectively the formation of HO^• (Wilms et al., 2008) or ${\text{O}}_2^{·-}$ radicals (Giardina and Inan, 1998; Zhao et al., 2005). In order to ascertain intracellular radical formation, electron spin resonance (ESR) analysis was performed. To associate gene expression modification with phenotypic markers of oxidative stress, whole-genome gene expressions were related to oxidant-induced DNA damage (8-oxodG levels) as well as apoptosis and cell cycle progression.

MATERIALS AND METHODS

Cell Culture and Treatment

The caco-2 human colorectal adenocarcinoma cell line (American Type Culture Collection, Manassas, VA) was grown in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum, 1% nonessential amino acids, 1% sodium pyruvate, and 1% penicillin/streptomycin at 37°C and 5% CO₂. Cells were grown until 80% confluency before medium was replaced by DMEM without supplements containing 100μM menadione or 20μM H₂O₂. For microarray analyses, time-matched control cells were treated in an identical manner without addition of the oxidants. Samples were taken after 0.08, 0.25, 0.5, 1, 2, 4, 8, 16, or 24 h. Because ESR measurements showed radical scavenging activity by the supplements, these were supplied 45 min after oxidant addition. At the indicated time points, cells were fixed in 2 ml ice-cold methanol for flow cytometry or medium was replaced by 1 ml Trizol for RNA/DNA isolation.

ESR Spectroscopy

Caco-2 cells were washed with PBS, preincubated for 30 min with 50mM 5,5-dimethyl-1-pyrolline N-oxide (DMPO), and washed with PBS again. After exposure to H₂O₂ or menadione as described above, cells were scraped. ESR spectra from cells were recorded on a Bruker EMX 1273 spectrometer with instrumental conditions and quantification of DMPO^•-OH peak signals as described before (Briede et al., 2005).

Flow Cytometric Analyses

Analyses of cell cycle profiles and apoptosis were performed as previously described (Staal et al., 2007). Cells were stained with propidium iodide; apoptotic cells were visualized by means of the primary antibody M30 CytoDeath (Roche, Penzberg, Germany) and fluorescein isothiocyanate-conjugated anti-mouse Ig as secondary antibody (DakoCytomation, Glostrup, Denmark). Data analysis was done using CellQuest software (version 3.1; Becton Dickinson, San Jose). M30 CytoDeath positive (apoptotic) and negative (nonapoptotic) signals were displayed as percentage of total cells with WinMDI 2.8 (http://facs.scripps.edu/software.html). Cell cycle profiles were analyzed using ModFit LT for Mac (version 2.0).

RNA and DNA Isolation

RNA was isolated from the Trizol solutions according to the producer’s manual and purified with the RNeasy mini kit (Qiagen Westburg bv., Leusden, The Netherlands). After removal of the aqueous phase during RNA isolation using Trizol, remaining phases were used for DNA isolation according to manufacturer’s protocol. RNA and DNA quantity were measured on a spectrophotometer, and RNA quality was determined on a BioAnalyzer (Agilent Technologies, Breda, The Netherlands). Only RNA samples that showed clear 18S and 28S peaks and with a RNA integrity number > 8 were used.

8-oxodG Analyses

Analysis of 8-oxodG was performed as described earlier (Briede et al., 2004). DNA extraction and workup were performed in the presence of a combination of EDTA and 8-hydroxyquinoline in order to prevent artificial 8-oxodG formation. After extraction, DNA was digested into deoxyribonucleosides by treatment with nuclease P1 (0.02 U/μl) and alkaline phosphatase (0.014 U/μl). The digest was then injected into a high performance liquid chromatography with electrochemical detection system. The mobile phase consisted of 10% aqueous methanol containing 94mM KH₂PO₄, 13mM K₂HPO₄, 26mM NaCl, and 0.5mM EDTA. The detection limit was 1.5 residues/10⁶ 2′-deoxyguanosine (dG). Ultraviolet absorption of dG was simultaneously monitored at 260 nm.

Real-Time PCR

Quantitative real-time (RT) PCR was performed as reported (Staal et al., 2007). All PCRs were performed in duplicate. β-Actin messenger RNA was used as reference. The RT-PCR was run on the MyiQ Single-Color Real-Time PCR Detection System (Bio-Rad Laboratories). The following forward and reverse primers were used (operon, 5′–3′ sequences): SOD1 GTGGTCCATGAAAAAGCAGATGA (forward) and CACAAGCCAAACGACTTCCA (reverse), SOD2 ATCAGGATCCACTGCAAGAA (forward) and CGTGCTCCCACACATCAATC (reverse), catalase GACTGACCAGGGCATCAAAAA (forward) and CGGATGCCATAGTCAGGATCTT (reverse), and β-actin CCTGGCACCCAGCACAAT (forward) and GCCGATCCACACGGAGTACT (reverse). Gene expression is calculated using the ΔΔCt log method and log base 2 transformed (Livak and Schmittgen, 2001) for correlation analysis with gene expression as measured by microarrays.

Microarrays

Labeling and hybridization.

Labeling and hybridization of RNA samples were done according to Agilent’s manual for microarrays with minor modifications (Agilent Technologies). Each treated and time-matched control RNA sample (0.5 μg) was transcribed into complementary DNA and labeled with Cyanine 3 (Cy3) or Cyanine 5 (Cy5), respectively. When dye incorporation was above 7 pmol/μg RNA, 2 μg complementary RNA of the treated and time-matched control samples was applied on G4110B 22K and G4112F 4 × 44K Agilent Human Oligo Microarray. Slides were scanned on a GenePix 4000B (Molecular Devices, Sunnyvale) with fixed laser power (100%) and photomultiplier tube gain. For each exposure, two biological repeats per time point were performed. Also, swapped Cy3 and Cy5 dye labeling was done. In total, this setup of 2 (oxidant exposure) × 2 (biological replicates) × 2 (dye swap) × 9 (time points) resulted in 72 hybridizations.

Image analysis and processing.

Images were processed with ImaGene 6.0 software (BioDiscovery Inc., Los Angeles) to quantify spot signals. Irregular spots were flagged. Data from ImaGene were transported into GeneSight software version 4.1.6 (BioDiscovery Inc.). For each spot, background was subtracted and flagged spots as well as spots with a net expression level below 20 in both channels were omitted. Data were log base 2 transformed, and locally weighted scatterplot smoothing normalization was applied. Expression difference with the time-matched control was calculated, and if more than one probe of a gene was present on the array, the replicates were combined while omitting outliers (> 2 SD). Only the probe sets present at both types of array platforms were selected.

Gene Expression Data Analyses

Differentially expressed genes.

Genes were selected for which all four replicate hybridizations (two biological experiments with two hybridizations) resulted in a log base 2 expression difference of more than 0.5 or all four less than −0.5 (all replicates in the same direction) at at least one time point. For calculation of the gene expression into a fold change per time point, the values of the four hybridizations were averaged. Using selection criteria based on fold change showed to be more reproducible if compared to criteria based on statistical parameters (Guo et al., 2006). The web tool GenePattern (Reich et al., 2006) (http://genepattern.broad.mit.edu/gp/pages/index.jsf) was used for clustering and visualization of heat maps and cluster patterns.

Time series analyses.

For identification of genes coregulated time dependently and clustering with the phenotypic parameters, the software tool “Short Time-series Expression Miner” (STEM, version 1.1.2b; http://www.cs.cmu.edu/∼jernst/stem/) was used (Ernst and Bar-Joseph, 2006). The clustering algorithm assigns each gene passing the filtering criteria (maximally four missing values, minimum correlation between experiments 1 and 2 of 0.3) to these profiles based on correlation coefficients and permutation analyses (n = 50). Profiles are clustered by correlation analyses. For clustering, 8-oxodG and cell cycle distribution levels were transformed into log 2 base ratios.

Functional interpretation of significantly differentially expressed gene sets.

MetaCore (GeneGo, San Diego, CA) was used to identify and visualize the involvement of the differentially expressed genes in the biological processes that may be affected at the level of pathways, by selecting significant pathways with a p value < 0.01.

Statistical Analyses of RT-PCR, 8-oxodG Levels, and Cell Cycle Effects

Data are presented as means ± SD. Statistical analyses of changes in apoptosis, 8-oxodG levels, or cell cycle phases and correlations between RT-PCR and microarray data were performed using SPSS for Windows 14.0 software (SPSS Inc., Chicago, IL) using a Student’s t-test or Spearman correlation with statistical significance set at p < 0.05.

RESULTS

Effects on Cellular ROS, 8-oxodG Levels, and Changes in Cell Cycle

Caco-2 cells pretreated with the spin trap DMPO were exposed to H₂O₂ or menadione and analyzed with ESR spectroscopy for cellular radical formation. Increasing concentrations of H₂O₂ resulted in dose-dependent formation of DMPO^•-OH (Fig. 1A). Based on the detected increase in cellular HO^• formation in combination with the absence of the induction of apoptosis, a noncytotoxic concentration of 20μM H₂O₂ was chosen. The absence of apoptosis is a crucial factor in our experimental setup because this would subsequently result in loss of cells from the investigated cell populations. Next, it was detected by ESR that increasing concentrations of menadione showed a ${\text{O}}_2^{·-}$-derived DMPO^•-OH signal at 100μM and 45-min incubation time (Fig. 1B), in combination with the absence of apoptosis, so this concentration was selected.

FIG. 1.

Levels of cellular hydroxyl radical formation after exposure to different concentrations of H₂O₂ (A) and time-dependent oxygen radical formation after exposure to 100μM menadione (B) as measured by ESR in combination with spin trapping by DMPO.

Compared to the constant basal level (20 × 10⁻⁶ 8-oxodG/dG) in control cells, H₂O₂ resulted in an immediately significant increased peak in 8-oxodG at 15 min that rapidly decreased to baseline level followed by a smaller but again significantly elevated level at 16 h after exposure (Fig. 2A). In contrast, the highest significantly increased level of 8-oxodG after menadione exposure was found at 2 h, while the significant increased level detected at 15 min was much lower as compared to H₂O₂ exposure at 15 min.

FIG. 2.

Levels of ratio of 8-oxodG/dG (A) and cell cycle distribution of caco-2 cells after exposure to 20μM H₂O₂ (B) or 100μM menadione (C). Venn diagram (D) showing overlapping and unique genes in the differential expressed gene data sets after the exposure to 20μM H₂O₂ or 100μM menadione. The genes present in different clusters are described in the Supplementary Data. Dendogram (E) of the result of the hierarchal clustering of the identical genes in the data set after exposure to 20μM H₂O₂ (H₂O₂) or 100μM menadione (Men) over all time points.

Analyses of cell cycle distributions showed that 16 h after H₂O₂ exposure, the number of cells in G₁ phase significantly decreased as compared to control cells (Fig. 2B). At 24 h, the percentage of cells in the S phase significantly decreased and in the G₁/M phase significantly increased, reflecting progression of cells previously in cell cycle arrest. After menadione exposure (Fig. 2C) at 24 h, a significant increase in the number of cells in the S phase, concomitantly with a significant decrease of cells in the G₁ phase, was detected. This suggests an arrest of cells in S phase and/or a stimulation of G₁ cells to go into S phase at 24 h. Altogether, this indicates that cell cycle arrest took place much later in menadione-exposed caco-2 cells than in H₂O₂-exposed cells.

Differentially Expressed Genes

Gene expression analysis resulted in 1404 differentially expressed genes upon H₂O₂ challenge and 979 genes after menadione treatment (see Supplementary Data for their time-dependent expression levels). Visualization of the relative expression ratios and all arrays in a heat map shows a high reproducibility in the expression profiles between the various replicate experiments and arrays (Supplementary Data). In total, 297 genes appeared commonly modified after both oxidant exposures (Fig. 2D), including the antioxidant enzyme CAT. Pathway analysis (Table 1) indicated that the significant regulated pathways are vitamin B7 and peroxisome proliferator-activated receptor (PPAR) regulation of metabolism, induction of transcription via hypoxia inducible factor 1 (HIF1) activation and other apoptosis pathways, and the Wnt signaling pathway. The significant pathways exclusively regulated in the gene set induced by H₂O₂ were amino acid metabolism and the immune response pathway oncostatin M signaling via mitogen-activated protein kinase (MAPK). The number of pathways significantly induced exclusively by menadione was twice as large, 24, and included responses to extracellular responses transcription and cell cycle processes. Hierarchal clustering of the expression of these 297 genes over all time points (Fig. 2E) showed that H₂O₂- and menadione-induced gene expression modification clustered completely apart.

TABLE 1

Significantly Regulated Pathways Induced by the Common 297 Genes. Using the Criteria Described in the “Materials and Methods” section, 297 Overlapping Genes Were Found for Both H₂O₂ and Menadione Exposure. A Summary of Significantly (p < 0.01) Regulated Pathways and Related Cellular Processes as Indicated by MetaCore Is Shown

Pathway	Cellular process
Vitamin B7 (biotin) metabolism
PPAR regulation of lipid metabolism	Regulation of lipid metabolism
Role of Akt in hypoxia-induced HIF1 activation	Transcription
Anti-apoptotic TNFs/NF-κB/IAP pathway	Apoptosis
WNT signaling pathway. Part 2	Response to extracellular stimulus
Regulation of lipid metabolism via LXR, NF-Y, and SREBP	Transcription
Nitrogen metabolism	Amino acid metabolism
Role of IAP proteins in apoptosis	Transcription
Anti-apoptotic TNFs/NF-κB/Bcl-2 pathway	Apoptosis
Activin A signaling regulation	Response to extracellular stimulus
RXR-dependent regulation of lipid metabolism via PPAR, RAR, and VDR	Transcription
Transcription regulation of amino acid metabolism	Transcription
CDC42 in cellular processes	Small GTPase-mediated signal transduction

Correctness of expression patterns for the antioxidant enzymes SOD1, SOD2, CAT, and cell cycle enzyme CDKN1A were confirmed by RT-PCR, resulting in a significant correlation (Spearman, R > 0.67, p < 0.05) between RT-PCR and microarray detected gene expression.

Pathways Analysis per Time Point

The number of significantly (p < 0.01) regulated pathways after H₂O₂ exposure (Table 2A) is highest at the earlier time points, up to 1 h. This is in accordance with the pattern of time-dependent formation of the 8-oxodG formation. Immune response was one of the most abundant regulated pathways over all time points, next to amino acid metabolism, cell cycle, apoptosis, and cell adhesion pathways. Particularly, interesting is the early presence of the WNT pathway at 0 and 4 h, taken into account that this is important in colon cancer development. Compared to H₂O₂, menadione resulted in more pathways that were more constantly enhanced over all time points (Table 2B). Pathways that regulate the cell cycle were most abundantly changed at all time points. Both the Activin A signaling regulation pathway and the vitamin B7 (biotin) metabolism were highly influenced at almost all investigated time points.

TABLE 2

General and Specific Pathways per Time Point. Summary of the Significant Pathways (p < 0.01) per Time Point in the Differentially Expressed Genes After H₂O₂ (A) or Menadione (B) Exposure as Indicated by MetaCore, Including the Exact Number of Pathways

Time (h)	Number of pathways	Summary of pathways
A
0.08	8	(Regulation of) amino acid metabolism
		Immune response: MIF-JAB1 and IL3 signaling
		Development: Notch signaling and WNT signaling pathway
0.25	15	Cell adhesion: ECM remodeling
		Immune response: MIF-mediated glucocorticoid regulation
		Development: glucocorticoid receptor and Notch signaling
		Apoptosis and survival: lymphotoxin-beta receptor signaling
		Transcription regulation of amino acid metabolism
0.5	17	Transcription: role of Akt in hypoxia-induced HIF1 activation and amino acid metabolism
		Development: growth hormone and gluocorticoid receptor signaling
		Immune response: oncostatin M, IL2, and IL3 activation and signaling pathway
		Oxidative stress: role of ASK1 under oxidative stress
		Apoptosis and survival: role of IAP proteins in apoptosis
		Cell cycle: nucleocytoplasmic transport of CDK/cyclins
1	6	Transcription: amino acid metabolism
		Immune response: oncostatin M, IL2, and IL3 signaling
		Development: leptin and EGF signaling
2	4	Transcription: amino acid metabolism
		Immune response: oncostatin M signaling via MAPK in human cells
		Cholesterol biosynthesis
		Development: EGF signaling pathway
4	5	Transcription: amino acid metabolism
		Cell cycle: chromosome condensation in prometaphase
		Immune response: oncostatin M signaling via MAPK in human cells
		Development: WNT signaling pathway
8	8	Polyamine, arginine, and biotin metabolism
		Development: glucocorticoid receptor signaling
		Transcription: role of Akt in hypoxia-induced HIF1 activation
		Apoptosis and survival: role of IAP proteins in apoptosis
		Cell cycle: initiation of mitosis
		Cell adhesion: ECM remodeling
16	5	Arginine metabolism
		Development: EPO-induced Jak-STAT pathway
		Urea cycle
		Immune response: antigen presentation by MHC class I
		Cell adhesion: ECM remodeling
24	6	Arginine, polyamine, CTP/UTP, and ATP/ITP metabolism
		Urea cycle
		Cell cycle: initiation of mitosis
B
All		Cell cycle: various pathways
All except 2 h		Vitamin B7 (biotin) metabolism
0.08	18	Signal transduction: Activin A and AKT signaling regulation
		Glutathione metabolism
		Cytoskeleton remodeling: CDC42 in cellular processes
		Transcription: formation of Sin3A and NuRD and amino acid metabolism
0.25	16	Signal transduction: Activin A signaling regulation
		Cytoskeleton remodeling: CDC42 in cellular processes
		Transcription: Sin3A and NuRD complexes, amino acid metabolism, and role of HP1
		Development: TGF-beta receptor signaling and TPO in cell process
0.5	14	Transcription: formation of Sin3A and NuRD complexes, amino acid metabolism, androgen receptor nuclear signaling, and role of HP1
		Signal transduction: Activin A signaling regulation
1	15	Signal transduction: Activin A and Akt signaling
		Glutathione and nitrogen metabolism
		Cytoskeleton remodeling: CDC42 in cellular processes
		Transcription: role of Akt in hypoxia-induced HIF1 activation
		Development: WNT signaling pathway
2	8	Spindle assembly and chromosome separation
		Transport: RAN regulation pathway
		Transcription: role of HP1 family in transcriptional silencing
4	10	Signal transduction: Activin A signaling regulation
		Cytoskeleton remodeling: CDC42 in cellular processes
		Transcription: formation of Sin3A and NuRD complexes
		Apoptosis and survival: role of IAP proteins in apoptosis
8	14	Signal transduction: AKT and Activin A signaling regulation
		Cytoskeleton remodeling: CDC42 in cellular processes
		Development: Notch signaling pathway and NF-κB activation
		Bile acid biosynthesis
		IMP and ATP/ITP metabolism
16	19	Signal transduction: Akt 1 and Activin A signaling regulation
		Cytoskeleton remodeling: CDC42 in cellular processes
		Transcription: formation of Sin3A and NuRD complexes and their role in transcription regulation
		Development: Notch, VEGF, IGF, and angiopoietin signaling
		Transcription: P53 signaling pathway
		Transport: Rab-9 regulation pathway
24	18	Amino acids, nitrogen, and triacylglycerol metabolism
		Signal transduction: Activin A signaling regulation
		Cytoskeleton remodeling: CDC42 in cellular processes
		Transcription_Formation of Sin3A and NuRD complexes and p53 Development: VEGF, Notch signaling pathway, and NF-κB activation
		Bile acid biosynthesis

Time-Dependent Coregulated Genes

Temporal differences in gene expression were further investigated with STEM (Ernst and Bar-Joseph, 2006). H₂O₂ exposure resulted in 998 genes that were significantly assigned to six different clusters of gene expression curves (Fig. 3A). Clusters of genes were identified that were upregulated at early and/or late time points (clusters 1, 2, and 5) or downregulated (clusters 0, 3, and 4) at all time points. All clusters showed profiles characterized by an immediately upregulation (Fig. 3A, cluster 5) or downregulation of genes at the early time points (Fig. 3A, clusters 0, 1, 2, and 4). Menadione exposure resulted in four different clusters containing 680 different genes (Fig. 3B). The clusters of genes could be divided completely either upregulated or downregulated during almost all time points (clusters 0 and 1) or downregulated at early (cluster 3) or late time points (cluster 2). Interestingly, while all clusters of coregulated genes for H₂O₂ exposure mainly showed extensive changes in gene expression immediately after exposure, the clusters identified after menadione exposure showed that expression was more induced not earlier than hours after exposure.

FIG. 3.

Average expression curves for the clusters of genes generated by STEM from caco-2 cells after various exposure times to 20μM H₂O₂ (A) or 100μM menadione (B).

Specific pathways regulated by the gene clusters were identified (Table 3). For both oxidant exposures, regulation of genes that are involved in the process of cell cycle regulation were found, especially in clusters 0, 2, 3, 4, and 5 for H₂O₂ and cluster 1 for menadione exposure. Other clusters shared in common pathways involved in nucleotide metabolism, transcription, and translation, confirming the fact that gene expression related processes are initiated. Again, the Activin A signaling regulation pathway and vitamin B7 (biotin) metabolism are found in clusters found for both oxidants. Interestingly, in cluster 0, after H₂O₂ exposure, genes involved in DNA double-strand break repair were found, a process that is known to be initiated by H₂O₂. Also, uniquely for H₂O₂ exposure, one cluster contained genes that are involved in the CXCR4 signaling pathway (cluster 1) and epidermal growth factor (EGF) signaling (cluster 3) pathway. Unique pathways for menadione exposure included glutathione and vitamin E metabolism (cluster 0).

TABLE 3

Summary of the Pathways of the Gene Expression Cluster Profiles. Significantly (p < 0.01) Regulated Pathways as Indicated by MetaCore of the Gene Expression Profiles Found by STEM Shown in Figure 2. Tables Show a Summary of the Pathways and Cellular Processes After H₂O₂ (A) or menadione (B)

Cluster	Number	Pathways	Processes
A
0	4	Chromosome condensation in prometaphase and metaphase checkpoint	Cell cycle
		Mechanism of DSB repair	DNA damage
		Clathrin-coated vesicle cycle	Transport
1	1	CXCR4 signaling	G-coupled protein signaling/immune response
2	9	ATP, CTP/UTP. GTP metabolism	Nucleotide metabolism
		Role of APC in cell cycle regulation	Cell cycle
		Aspartate, aspargine, and polyamine metabolism	Metabolism
		Role of Akt in hypoxia-induced HIF1 activation	Transcription, hypoxia
3	1	EGF signaling pathway	Growth factor, development
4	3	Regulation of translation initiation	Translation
		Transition and termination of DNA replication	Cell cycle
		Nucleocytoplasmic transport of CDK/cyclins	Cell cycle
5	4	Vitamin B7 metabolism	Metabolism
		Activin A signal regulation	Cell cycle
		CDC42 in cellular processes	Cytoskeleton remodeling
		TPO signaling via JAK-STAT pathway	Development
B
0	12	Iso(leucine), valine, trypthophane, and proline metabolism	Amino acid metabolism
		Glutathione and vitamin E metabolism	Vitamin metabolism
		Propionate and triglycerol metabolism, mitochondrial long chain and peroxisomal branched chain fatty acid oxidation	Lipid metabolism
		Notch signaling and activating pathway	Development
1	18	Vitamin B7 metabolism	Vitamin metabolism
		Activin A signal regulation	Cell cycle
		CDC42 and ATM./ATR regulation of G1/S checkpoint	Cell cycle
		Parkin in and ubiquitin proteasomal pathway	Proteolyse
		P53, NF-κB, and NGF activation of NF-κB, VEGF signaling	Transcription
2	4	Cholesterol, cortisone, bile acid, androstenedione, and testosterone metabolism	Metabolism
3	8	WNT signaling	Development
		Activin A in cell differentiation	Development
		GTP, CTP/UTP, and ATP metabolism	Nucleotide metabolism
		Endothelin-1/EDNRA signaling	Development
		Androgen receptor nuclear signaling	Transcription

STEM also offers the ability to evaluate similarities and correlations in time-dependent gene expression between different experiments. No correlating clusters were found, but a significant correlation between time-dependent expressions of 30 genes was found. Pathway analysis resulted in only one affected pathway, being chromosome condensation in the prometaphase as part of the cell cycle processes. This again showed that the similarities between time-dependent gene expression profiles are limited but both oxidants share in common the ability to influence cell cycle–related processes. In addition, we also used STEM to analyze the time-dependent coregulated genes in the 297 commonly modified data set induced by both oxidant exposures (Fig. 2D). For H₂O₂ exposure, this resulted in three clusters of which two clusters (clusters 1 and 2; Supplementary Data) again showed profiles characterized by an immediately upregulation or downregulation of genes at the early time points. No cluster of time-dependent coregulated genes was obtained for menadione exposure, but this again confirmed the absence of similarities between time-dependent gene expression profiles for both oxidant exposures.

Correlations with 8-oxodG Levels and Changes in Cell Cycle

STEM was also used to identify the correlation between time-dependent gene expression patterns and changes in 8-oxodG formation and in cell cycle phases. For log ratio transformed 8-oxodG levels (Table 4A and B), this analysis resulted in eight genes for H₂O₂ and three genes for menadione exposure, with no common genes, again demonstrating that the transcriptomic responses to H₂O₂ and menadione over 24 h are clearly different. A number of these genes differentially expressed after H₂O₂ treatment is reported to be under control, or involved, in cellular oxidative stress processes. Cytochrome b5 belongs to a family of electron transport haemoproteins and is involved in the lipid peroxidation cycles induced by oxygen radicals. It is shown before that an inhibition of steroid 5-alpha reductase by finasteride results in enhancement of cellular H₂O₂ production (Cayatte et al., 2006).

TABLE 4

Genes that Clustered in the Same Gene Expression Cluster Profiles with 8-oxodG Levels or Cell Cycle Phases. Correlating Genes Clustered by STEM with the 8-oxodG Levels Induced by Exposure to 20μM H₂O₂ (A) or 100μM menadione (B). Genes of Which the Time-Dependent Expression Is Clustered by STEM with Cell Cycle by Exposure to 20μM H₂O₂ (C) or 100μM menadione (D) Including Induction of Significant (p < 0.01) Pathways as Indicated by MetaCore

Gene symbol	Description
A
CYB5A	Cytochrome b5 type A
SRD5A2L	Steroid 5 alpha-reductase 2-like
KLHL13	kelch-like 13
APH1B	Anterior pharynx defective 1 homolog B
IFIT1	Interferon-induced protein with tetratricopeptide repeats 1
SDSL	Serine dehydratase-like
PEPD	Peptidase D
OAT	Ornithine aminotransferase, nuclear gene encoding mitochondrial protein
B
SH3BGRL3	SH3 domain–binding glutamic acid-rich protein-like 3
FRMD4A	FERM domain containing 4A
DKFZP434H132	DKFZP434H132 protein
Cell cycle phase	Genes	Pathways	Summary of pathways
C
S	1	0	—
G2/M	45	1	Arginine metabolism
G1	29	0	—
D
S	74	4	P53 signaling, androstenedione, testosterone and bile acid synthesis and metabolism, triglycerol metabolism
G2/M	1	0	—
G1	311	15	Amino acid metabolism, Notch signaling (for NF-κB), PPAR regulation of lipid metabolism, cholesterol and vitamin E metabolism, fatty acid oxidation, bile acid biosynthesis, cadherin-mediated cell adhesion

As presented in Table 4C and D, time-dependent expression of many genes correlated with changes in the different phases of the cell cycle (S, G₂/M, and G₁). For H₂O₂ exposure, the gene related to S phase, ADAMTS19, could not be related directly to cell cycle processes. Also, arginine metabolism was the only pathway related to the G₂/M phase, and no pathways compromised the genes related to the G₁ phase. In contrast, genes as well their induced pathways correlating to changes in cell cycle phases after menadione exposure were very relevant for the oxidant exposure and included p53 signaling, Notch signaling for NF-κB, and vitamin E metabolism. The gene for which the expression pattern correlated to the G₂/M phase is MCM5, and this minichromosome maintenance component 5 is also known as cell division cycle 46, which actively participates in the process of cell cycle. Comparison of the gene sets correlating with cell cycling for both exposures showed that these shared two genes in common (CDKN1C, a cyclin-dependent kinase inhibitor and SLC2A4RG, a transcription factor), which both correlated to changes in the G₁ phase for H₂O₂ and S phase for menadione exposure.

DISCUSSION

ESR spectrometry confirmed that H₂O₂ and menadione induced cellular oxygen free radical formation. Both types of oxidants are used to make the comparison between the induction of different kind of cellular ROS (Dunning et al., 2009; Pocsi et al., 2005). Menadione induces the formation of cellular ${\text{O}}_2^{·-}$ (Zhao et al., 2005), including in colon cancer cells (Giardina and Inan, 1998). We here measured that the kinetics of 8-oxodG formation induced by both oxidants in caco-2 cells were different. It is possible that time-dependent cellular conversion of ${\text{O}}_2^{·-}$ into H₂O₂ by SOD, followed by the formation of much more effective DNA-oxidizing HO^• radicals, resulted in this highest level of 8-oxodG 2 h after menadione exposure. But extensive studies to the time-dependent formation of 8-oxodG by H₂O₂ or menadione in caco-2 cells were not published before. The baseline levels between 20 and 30 per 10E6 dG in caco-2 cells are higher than what is normally found in other cells. Although many studies on oxidative damage have been performed in caco-2 cells, almost all of them have been using the comet assay and report damage levels in arbitrary units, which does not allow comparison of absolute 8-oxodG levels. The procedures in order to reduce potential DNA oxidation during extraction were carefully validated, and baseline levels of 8-oxodG were measured in previous studies using the same methodologies. Only one study with menadione in rat hepatocytes reported the absence of 8-oxodG formation in rat hepatocytes up to 2 h after incubation (Fischer-Nielsen et al., 1995). Analysis of cell cycle effects indicated a similar temporal difference between H₂O₂ and menadione. Altogether, this shows that the formation of 8-oxodG by both oxidants is followed by cell cycle arrest, but both effects are induced at a slower rate after the increase of intracellular ${\text{O}}_2^{·-}$⁠.

This difference in temporal response to both oxidants is reflected at the gene expression level: Analysis of the time-dependent coregulated genes revealed that the cells responded immediately on HO^• by an early upregulation or downregulation of genes at early time points. Genes in these clusters are specifically involved in transcription, cell cycle, cell differentiation, and metabolic processes. But uniquely for H₂O₂ exposure, genes involved in DNA double-strand break repair, CXCR4 signaling pathway, and EGF signaling pathway were found. The chemokine receptor CXCR4 is constitutively expressed in caco-2 cells (Dwinell et al., 2004), involved in colon carcinogenesis, and proposed as a prognostic factor for gastrointestinal tumor formation (Schimanski et al., 2008). The EGF receptor is tyrosine kinase receptor involved in growth, proliferation, and apoptosis signaling of cells, and activation is observed in gastrointestinal carcinogenesis.

In contrast, an increase in intracellular ${\text{O}}_2^{·-}$ caused gradual increase in upregulated or downregulated genes up to 24 h, while again, cell cycle and metabolic processes are biological functions in these clusters. Unique pathways for menadione exposure included glutathione and vitamin E metabolism, pathways involved in the antioxidant response toward oxidative stress. Altogether, the difference in gene expression patterns for both oxidants explains the separation of the exposures in the hierarchal clustering of the 297 commonly regulated genes. The pulse-like response toward H₂O₂, including recovery within 1 h, while no recovery was observed with menadione, is completely comparable to the gene expression effects previously described for the fungus Aspergillus nidulans (Pocsi et al., 2005). Comparison of pathways showed that (secondary) metabolism also was found to be induced by both oxidants in this study. A previously performed comparison between different prokaryotic and eukaryotic species indicated heat-shock proteins as the only evolutionary conserved H₂O₂-induced proteins (Vandenbroucke et al., 2008); however, only one time point was investigated.

The sequential effects caused by the oxidants are revealed in detail by pathway analysis of the gene list per time point. Within 5 min after H₂O₂ exposure, the cells responded to oxidative stress by showing an immune-like response, via macrophage migration inhibitory factor (MIF)-Jun activating binding protein (JAB1) and interleukin (IL)3. The cytokine MIF interacting with JAB1 antagonizes cell cycle regulation (Kleemann et al., 2000), while IL3 modulates colon cancer cell growth in vitro (Block et al., 1992). The cell cycle pathway indeed appears hereafter. MIF regulation is also seen 15 min after H₂O₂ exposure. Also, Notch signaling is immediately induced, and this signaling network is involved in colon carcinogenesis (Katoh and Katoh, 2007) and aging (Carlson et al., 2008). After 30 min, apoptosis signal–regulating kinase 1 (ASK1) is affected, being a MAP kinase kinase kinasepreferentially activated in response to oxidative stress and plays pivotal roles in a wide variety of cellular responses (Matsuzawa and Ichijo, 2008). After 30 min, gene and pathway expression responses dampen, although cell cycle pathways and immune response, especially via oncostatin M signaling, carries on. Oncostatin M is related to IL31 induction and seen in inflammatory bowel diseases (Dambacher et al., 2007). In contrast to the pulse-like modulation of gene and pathway expression profiles, amino acid metabolism is a pathway that is continuously affected over 24 h.

Menadione exposure continuously modulates the cell cycle pathways and vitamin B7 (biotin) metabolism. This vitamin is a prosthetic group for carboxylases, plays a role in lipid metabolism, and biotin is via biotinylation a function as a keeper of gene expression. Deficiency of biotin results in increased mitochondrial ROS formation via an impaired respiration (Depeint et al., 2006). Furthermore, Activin A is almost continuously present in the signal transduction pathway. Activin is a proinflammatory mediator and overexpressed in gastrointestinal inflammation (Werner and Alzheimer, 2006). In contrast to H₂O₂ exposure, however, no (immediate) responses of immune pathways or pathways specifically assigned to oxidative stress were revealed. Interestingly, H₂O₂ immediately and menadione after 1 h induced the WNT signaling pathway, which is a key player in colon carcinogenesis (Gmeiner et al., 2008).

Microarray analysis of the effect of 75μM H₂O₂ on the expression of 12,931 genes in caco-2 cells at 30-min incubation time was done before (Herring et al., 2007), and this resulted in the selection of 87 affected genes. Comparison of the genes showed 29 genes (33%) are similar if leaving out specific subtypes of proteins. At the functional level identical pathways, i.e., cell development and cell cycle were identified if compared with our results at T = 0.5 h. Comparison with the time dependently induced probe set found after exposing human A549 lung cells to the H₂O₂-generating system (Dandrea et al., 2004) showed that 12 of 51 genes (24%) were similar of which most genes are involved in cell cycle processes.

Genes as well their induced pathways correlating to changes in 8-oxodG and cell cycle phases were relevant for these type phenotypical parameters. Comparison of both gene sets for the cell cycle revealed the presence of a cyclin-dependent kinase inhibitor (CDKN1C) that is clearly involved in the regulation of cell cycle processes.

Altogether, we now report as a novel finding that although a large number of identical genes were affected by both oxidants, no similarity in the temporal expression of these genes was found. Comparison of the profiles of coregulated genes showed that gene expression in caco-2 cells immediately reacted by a pulse-like response to the HO^• formation, while formation of ${\text{O}}_2^{·-}$ results into gene expression that is not restored in 24 h. Pathway analyses identified the sequential involvement of gene expression in various cell responses and demonstrated that the difference in the modulation of gene expression is also reflected in regulation of pathways by HO^• and ${\text{O}}_2^{·-}$⁠: H₂O₂ immediately influences pathways involved in the immune function, while menadione constantly regulated cell cycle–related pathways. Altogether, the temporal response of caco-2 cells toward two oxidants is different at all levels of gene expression: temporal patterns, influenced pathways, and phenotypical anchoring, clearly demonstrating that HO^• and ${\text{O}}_2^{·-}$ need to be differentially associated with oxidative stress–related colon diseases, including cancer.

This strongly emphasizes that for the biological interpretation of genome wide gene expression, studying time-dependent effects is inevitable. Taken into account that the cellular defense capacity against oxidative stress in the human body is strongly influenced by the uptake of dietary antioxidants in the colon, it would be interesting to translate the oxidants-induced temporal gene expression profiles into the beneficial health effect of antioxidants in food on colon cells. For example, a comparison showed that the genes metallothionein 1G, cyclin G1, activating transcription factor 3, and ornithine decarboxylase 1 are in both differentially expressed by oxidative stress in caco-2 cells and in colonic tissue of colorectal patients after intake of a high vegetable-containing diet (van Breda et al., 2004)

FUNDING

Netherlands Toxicogenomics Centre (NTC), subsidized by the Netherlands Genomics Initiative (NGI)/Netherlands Organization for Scientific Research (NWO) (Grant nr 050-060-510).

We thank E. A. M. Bolderman and T. H. Yuen for their contribution and support with the pilot experiments that were performed for the setup of this study. Raw data of the microarray experiments are available for reviewers in GEO via the internet link: http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE15327.