Aristolochic acid (AA) is the causative agent of urothelial tumors associated with AA nephropathy and is also implicated in the development of Balkan endemic nephropathy-associated urothelial tumors. These tumors contain AA-characteristic TP53 mutations. We examined gene expression changes in Hupki (human TP53 knock-in) mice after treatment with aristolochic acid I (AAI) by gavage (5 mg/kg body weight). After 3, 12 and 21 days of treatment gene expression profiles were investigated using Agilent Whole Mouse 44K Genome Oligo Array. Expression profiles were significantly altered by AAI treatment in both target (kidney) and nontarget (liver) tissue. Renal pathology and DNA adduct analysis confirmed kidney as the target tissue of AAI-induced toxicity. Gene ontology for functional analysis revealed that processes related to apoptosis, cell cycle, stress response, immune system, inflammatory response and kidney development were altered in kidney. Canonical pathway analysis indicated Nfκb, aryl hydrocarbon receptor, Tp53 and cell cycle signaling as the most important pathways modulated in kidney. Expression of Nfκb1 and other Nfκb-target genes was confirmed by quantitative real-time PCR (qRT-PCR) and was consistent with the induction of Nfκb1 protein. Myc oncogene, frequently overexpressed in urothelial tumors, was upregulated by AAI on the microarrays and confirmed by qRT-PCR and protein induction. Collectively we found that microarray gene expression analysis is a useful tool to define tissue-specific responses in AAI-induced toxicity. Several genes identified such as TP53, Rb1, Mdm2, Cdkn2a and Myc are frequently affected in human urothelial cancer, and may be valuable prognostic markers in future clinical studies.
The herbal drug aristolochic acid (AA), which is derived from Aristolochia species, has been associated with the development of a novel human nephropathy, designated as aristolochic acid nephropathy (AAN), and its associated urothelial cancer.1, 2 AA is a mixture of structurally related nitro-phenanthrene carboxylic acids, the major components being aristolochic acid I (AAI) and aristolochic acid II (AAII). Herbal drugs derived from Aristolochia spp. have been known since antiquity and remain in use today, particularly in Asia but also in Central America. AAN was first reported in Belgian women who had consumed Chinese herbs as part of a weight-loss regimen in 1991 and was traced to the ingestion of A. fangchi inadvertently included in the slimming pills.2 Within a few years of taking the pills, Belgian patients with AAN had developed a high risk of upper tract urothelial carcinoma (about 50%) and, subsequently, bladder urothelial carcinoma.3, 4 Using the highly sensitive 32P-postlabeling method, exposure to AA was demonstrated by the identification of specific AA-DNA adducts in urothelial tissue of the Belgian patients.3, 5, 6 Subsequently, AA-DNA adducts have been shown to be suitable biomarkers of exposure to AA among AAN cases elsewhere in Europe and in Asia.7–9 As a consequence, the sale of products containing AA has been banned in many countries. However, a number of herbal products containing Aristolochia species continue to be advertised for sale on the internet.10
In Europe, exposure to A. clematitis has been linked to Balkan endemic nephropathy (BEN) and its associated urothelial cancer.11, 12 This nephropathy is endemic in certain rural areas of Serbia, Bosnia, Croatia, Bulgaria and Romania. BEN is clinically and morphologically very similar to AAN.13 Indeed, the detection of AA-specific DNA adducts was reported in patients with BEN and in individuals with end-stage renal disease living in areas endemic for BEN,11, 14 demonstrating that dietary exposure to AA is a significant risk factor for the development of the disease.
The most abundant DNA adduct detected in patients with AAN/BEN is 7-(deoxyadenosin-N6-aristolactam I (dA-AAI) and AA-initiated carcinogenesis in humans is associated with characteristic AT → TA transversion mutations in the TP53 tumor suppressor gene.11, 15, 16 Furthermore, AT → TA transversions were also induced in human TP53 in vitro using the HUF (Hupki murine fibroblast) assay, in which mouse embryonic fibroblasts from Hupki (human TP53 knock-in) mice were treated with AAI.17, 18 Since this type of mutation in TP53 is relatively rare in human cancer, but has been observed in urothelial tumor samples of patients with BEN, these human and experimental data strongly implicate the involvement of AA in the etiology of BEN-associated cancer.11, 18
Toxicogenomics can provide the means to define relationships between toxicological end-points and gene expression patterns, predict toxic responses and identify mechanisms of toxicity of environmental agents such as AA.19–21 In this study, we treated Hupki mice with AAI at a dose that has previously been shown to be carcinogenic in NMRI mice after repeated oral administration.22 Gene expression in the kidney (target organ) and liver (nontarget organ) was analyzed by microarray technology to identify genes altered by AA exposure and to explore potential candidate genes that are involved in AA-induced carcinogenesis.
Natural plant extract AA (Roth, Karlsruhe, Germany) was subjected to 3 sequential recrystallizations from boiling dimethylformamide-water. The orange crystals obtained were dried under vacuum yielding around 98% AAI as determined by reverse-phase HPLC. Analysis by high-field 1H-NMR and mass spectrometry gave results fully consistent with the structure of AAI. AAI was converted to its sodium salt by neutralization with aqueous sodium hydroxide. The solution was thoroughly mixed and taken to dryness under reduced pressure, yielding the sodium salt AAI quantitatively as a water-soluble deep-red amorphous powder.
Hupki mice (Trp53tm1/Holl, homozygous for the knock-in TP53 allele harboring human TP53 sequences in the 129/Sv background) were purchased from The Jackson Laboratory (Bar Harbor, ME) and bred in-house. Animals were kept under standard conditions with food and water ad libitum.
Animal treatment and sample collection
Female Hupki mice (2−4 months old) were randomly assigned to treatment groups (4/group) and treated daily with 5 mg/kg body weight (bw) AAI by gavage according to a protocol published previously22 for 3, 12 or 21 days. Control animals were treated with vehicle (water) only. Mice were killed 24 hr after the final dose. All animal experiments were carried out under license in accordance with the law, and with local ethical approval.
Organs (forestomach, glandular stomach, kidney, liver, lung, colon, spleen and pancreas) were collected, portions snap-frozen in liquid nitrogen and stored at −80°C until further analysis. For histopathology (21-day group only) organ sections (liver, kidney and bladder) were fixed in PBS containing 4% paraformaldehyde for subsequent paraffin embedding and sectioning.
DNA isolation and DNA adduct analysis by 32P-postlabeling
Genomic DNA was isolated by a standard phenol/chloroform extraction method. DNA adducts were determined using the nuclease P1 enrichment version of the 32P-postlabeling method as described previously.21, 23 In brief, DNA samples (4 μg) were digested with micrococcal nuclease (240 mU) and calf spleen phosphodiesterase (60 mU) in digestion buffer containing 20 mM sodium succinate and 10 mM calcium chloride (pH 6) for 3 hr at 37°C in a total volume of 10 μl. For nuclease P1 enrichment, the digests were incubated with 4 μg nuclease P1 in 3 μl of a buffer containing 0.8 M sodium acetate (pH 5) and 2 mM zinc chloride for 30 min at 37°C. The reaction was terminated by the addition of 3 μl Tris base (427 mM). DNA digests were then 32P-labeled by adding 4 μl of a mixture consisting of 400 mM bicine (pH 9.5), 200 mM MgCl2, 300 mM dithiothreitol, 10 mM spermidine, 50 μCi [γ-32P]ATP (∼7000 Ci/mmol; MP Biomedicals, Cambridge, UK) and 6 U T4 polynucleotide kinase (NEB, Hitchin, Herts, UK), and incubated for 30 min at 37°C. Chromatographic conditions for thin-layer chromatography on polyethyleneimine-cellulose plates (10 cm × 20 cm; Macherey-Nagel, Düren, Germany) were21: D1, 1.0 M sodium phosphate, pH 6; D3, 3.5 M lithium-formate, 8.5 M urea, pH 4; D4, 0.8 M LiCl2, 0.5 M Tris-HCl, 8.5 M urea, pH 9. After chromatography, thin-layer sheets were scanned using a Packard Instant Imager (Dowers Grove, IL) and DNA adduct levels (RAL, relative adduct labeling) were calculated from adduct cpm, the specific activity of [γ-32P]ATP and the amount of DNA (pmol of DNA-P) used. Results were expressed as DNA adducts/108 nucleotides. Urothelial DNA samples from patients with AAN were included in the analysis for comparison.3
For histopathological examination, 7-micron sections of liver, kidney and bladder were cut and stained with hematoxylin-eosin (H&E) and periodic-acid Schiff (PAS). Slides were randomized and analyzed blind by light microscopy. Sections were examined at low (100−200×) and high (400×) magnification for the presence of cellular injury, dysplasia/neoplasia, inflammation and fibrosis.
RNA extraction and gene expression profiling
Tissue (∼30 mg) was resuspended in 300 μl Trizol reagent (Invitrogen, Paisley, UK) in a 2-ml Eppendorf tube and homogenized with an Ultra-turrax T25 S7 homogenizer (Janke & Kunkel, Staufen, Germany). After 5 min at room temperature 180 μl chloroform was added, the tube was shaken for 30 sec and kept at room temperature for another 2 min. The tube was centrifuged at >13,000g for 20 min at 4°C and the supernatant was transferred to a Qiagen RNeasy Mini Kit column (Qiagen, Crawley, UK). Total RNA was extracted according to the manufacturer's instructions. RNA was quantified using GeneQuant pro UV/VIS Spectrophotometer (Biochrom, Cambridge, UK), and its integrity was determined using a 2100 BioAnalyzer (Agilent Technologies, South Queensferry, UK). RNA that had a RNA Integrity Number (RIN) >6.5 was used for microarray and quantitative real-time RT-PCR (qRT-PCR) analysis. RNA samples were stored at −80°C until use.
Gene expression analysis was carried out using Agilent Whole Mouse 44K Genome Oligo Array (Agilent Technologies). This specific array represents 41,000+ mouse genes and transcripts, all with public domain annotations. Total RNA (5 μg) was reverse transcribed into cDNA by incorporating a T7 oligo-dT promoter primer prior to the generation of fluorescent cRNA using an Agilent Quick Amp Labeling Kit (Agilent Technologies). The labeled cRNA was purified using a Qiagen RNeasy Mini Kit (Qiagen) and quantified using a NanoDrop ND-1000 instrument (Thermo Fisher Scientific, Waltham, MA). In these experiments, the reference design was used where Universal Mouse Reference RNA (UMRR; Stratagene, La Jolla, CA) was hybridized to every sample. Cy3- (reference) and Cy5-labeled (sample) cRNAs were hybridized to the array using a Gene Expression Hybridization Kit (Agilent Technologies). The hybridization was incubated in Agilent SureHyb chambers for 17 hr in a Rotisserie Hyb Oven set to 65°C and rotating at 10 rpm. The microarray slides were washed according to the manufacturer's instructions and then scanned on an Axon B400 Scanner (Axon Instruments, USA).
Microarray data analysis
For initial analysis the multi-image TIFF files generated by the scanner were loaded into BlueFuse software (BlueGnome Limited, Cambridge, UK) to adjust the initial grid positions and to optimize the spot finding in the image. Raw data generated from BlueFuse software were imported into GeneSpring v-7.2 software (Agilent Technologies) for modulation analysis without background subtraction. Within GeneSpring, the data were subjected to LOWESS normalization. Spots that were flagged in BlueFuse software with low confidence were removed from further analysis. Natural transformed averaged Cy5/Cy3 ratios were used to identify modulated genes that had a Welch's t-test p-value of <0.05 assigned. Natural transformed data were used for any statistical algorithms performed within GeneSpring such as hierarchical cluster analysis (HCA) and principal component analysis (PCA). Average values of Cy5/Cy3 ratios were used to identify modulated genes by fold-change analysis with a cutoff of 1.5.
The web-accessible program DAVID (Database for Annotation, Visualization and Integrated Discovery) was used for functional annotation of modulated genes. Signaling pathway analysis was carried out with the online software Pathway Miner. Gene association network analysis was performed with Ingenuity Pathway Analysis software (Ingenuity Systems, Redwood City, CA).
The gene expression data discussed in this publication have been deposited in the National Center for Biotechnology (NCBI) Gene Expression Omnibus database (http://www.ncbi.nlm.nih.gov/geo/) and are accessible through GEO Series accession number GSE18246.
Two-step reverse transcription-PCR was used to generate cDNA for relative analysis using real-time fluorescent PCR on an ABI PRISM 7900HT Sequence Detection System (Applied Biosystems, Foster City, CA) performed as previously described.24, 25 To detect the expression of selected target genes 20× Assays-On-Demand™ gene expression primers and probes (Applied Biosystems) were used. Endogenous Gapdh was used with the following probes: Nfκb1-Mm00476361_m1, Ccng1-Mm00438084_m1, Fos-Mm00487425_m1, Myc-Mm00487803_m1, Mdm2-Mm00 487656_m1, Stat3-Mm00456961_m1, Bcl3-Mm00504311_m1, Bcl6-Mm00477633_m1, Nqo1-Mm00500821_m1, Bbc3-Mm00519268, Cdkn1a-Mm0103209 and TP53-Hs01034 253_m1. Relative gene expression was calculated using the comparative threshold cycle (CT) method.
Western blot analysis
Tissue (∼30 mg) was homogenized in 300 μl T-Per buffer (Pierce, Rockford, IL) supplemented with protease (Complete tablets; Roche Diagnostics, Burgess Hill, UK) and phosphatase (BD Biosciences, Oxford, UK) inhibitors. The protein solution was centrifuged for 20 min at >13,000g (4°C) and the supernatant was saved. The protein concentration was determined with a BCA Kit from Pierce and protein extracts were diluted to a final concentration of 1 mg/ml.
Samples containing 10 μg of total proteins were separated by sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) on a 4–12% Bis-Tris Criterion XT Precast gel (Bio-Rad, Hercules, CA) and then transferred to a nitrocellulose membrane (Sigma, St. Louis, MO). For Western blotting, the following primary antibodies were used: NQO1 antibody at 1:4.000 (Sigma, N5288), anti-human p53 antiserum at 1:1000 (Novocastra, Newcastle upon Tyne, UK; CM-5), Nfκb p65 antibody at 1:1000 (Santa Cruz Biotechnology, Santa Cruz, CA; C-20), Cdkn1a (p21) antibody 1:1000 (BD, 556431), GAPDH antibody at 1:40.000 (Chemicon, MAB374) and c-Myc antibody at 1:500 (Abcam, Cambridge, UK; 11917). Anti-Cyp1a1 antisera used at 1:100026 was a generous gift from Colin J. Henderson (Biomedical Research Centre, Dundee, UK). Monoclonal antibody to detect GAPDH was used as a loading control. The immunoreactive proteins were detected by secondary anti-rabbit or anti-mouse antibodies (1:5000) coupled with horseradish peroxidase (Cell Signaling Technology, Danvers, MA). Bands were visualized using an enhanced chemoluminescene (ECL) SuperSignal West Pico detection kit according to the manufacturer's instructions (Pierce) and quantified with a Molecular Imager® Chemi Doc™ XRS imaging system (Bio-Rad).
DNA adduct analysis
DNA adduct formation was evaluated by 32P-postlabeling. Adduct levels in all organs were greater at 12 days than at 3 days, but no further increase was apparent at 21 days (Fig. 1). After 21 days of treatment highest DNA binding was observed in kidney (319.9 ± 82.7 adducts per 108 nucleotides), followed by forestomach, but the level of binding was 7-fold lower (45.4 ± 16.0 adducts per 108 nucleotides). DNA adduct levels in other organs ranged between 20 and 40 adducts per 108 nucleotides. No DNA adduct formation was detected in tissues from control animals treated with vehicle (water) only (data not shown). The adduct patterns induced by AAI in various organs were qualitatively similar to that found in patients with AAN, consisting of 2 major adduct spots (spots 1 and 2) and 1 minor adduct spot (spot 3).3, 5, 6 These adducts were identified previously5, 6 as 7-(deoxyadenosin-N6-yl)aristolactam I (spot 1; dA-AAI), 7-(deoxyguanosin-N2-yl)aristolactam I (spot 2; dG-AAI) and 7-(deoxyadenosin-N6-yl)aristolactam II (spot 3; dA-AAII).
After 21 days livers from both AAI-treated and control mice had normal histology (Figs. 2a and 2b). Although control kidneys appeared normal (Fig. 2c), all AAI-treated kidneys showed tubular injury and mild tubulointerstitial fibrosis (Fig. 2d) and 1 had occasional enlarged epithelial or endothelial nuclei (at least twice the size of normal nuclei). Tubular injury was characterized by focal loss of epithelial brush border on PAS stain, with epithelial flattening and occasional intraluminal casts. Tubulointerstitial fibrosis was characterized by widening of the interstitium between tubules, with accumulation of edema and interstitial matrix. The scattered enlarged nuclei were not associated with either structural changes or increased proliferation, and therefore did not constitute dysplasia. The bladders of AAI-treated mice showed subepithelial lymphoid aggregates, but no dysplasia (data not shown).
Gene expression analysis.
Microarray analysis identified a total of 701 genes in target organ kidney (213 after 3 days; 155 after 12 days and 354 after 21 days) and 650 genes in nontarget organ liver (204 after 3 days; 207 after 12 days and 266 after 21 days) with significantly (t-test; p < 0.05) altered expression after AAI treatment at 1 or more time points. After applying a fold-change cutoff of 1.5, 409 genes were modulated in kidney (Fig. 3a,a′), with 172 (106 ↑, 66 ↓), 86 (66 ↑, 20 ↓) and 167 (104 ↑, 63 ↓) genes altered after 3, 12 and 21 days, respectively (Fig. 3a,a″). Among those genes, 16 were modulated at 2 time points. In liver, 237 genes were modulated after applying a fold-change cutoff of 1.5 (Fig. 3a,a′), with 37 (19 ↑, 18 ↓), 81 (50 ↑, 31 ↓) and 127 (91 ↑, 36 ↓) genes altered after 3, 12 and 21 days, respectively (Fig. 3a,a″′). Among those genes, 8 were modulated at 2 time points. No gene was modulated at all 3 time points either in kidney or in liver. When the kidney gene list was compared to the liver gene list, only 13 genes were found to be commonly altered (Fig. 3a,a′), indicating that the gene expression profiles strongly depend on the tissue origin.
HCA and PCA.
HCA is used to classify different data groups into subsets (clusters) according to distance measurement. We used HCA to visualize clusters of samples corresponding to tissue origin and treatment period (Fig. 3b). The HCA is based on the expression profiles of 409 genes significantly modulated by at least 1.5-fold in kidney and 237 genes in liver. Furthermore, we also used PCA to cluster the samples (Fig. 3c). PCA uses the principal sources of variance in data sets to display it graphically here in plotting the overall data 2-dimensionally. HCA demonstrated that samples were grouped together according to tissue origin and AAI treatment period, suggesting a large difference between the expression profiles of the 2 tissues. Using PCA, we found that in kidney only a marginal variance difference was observed between AAI-treated groups and control (80.3% variance for component 1 and 7.5% variance for component 2) (Fig. 3c,c′), whereas in the liver no variance difference was found for principal component 1 (91.0 %) but a weak difference for principal component 2 (3.4%) (Fig. 3c,c″), suggesting only minor differences in the expression patterns in each of the tissues.
Functional annotation of AAI-modulated genes.
We used several bioinformatics approaches to analyze functional involvement of AAI-modulated genes. First, gene ontology (GO) analysis revealed that apoptosis (GO:0006915; 29 genes in kidney and 16 genes in liver) and cell differentiation (GO:0030154; 56 genes in kidney and 30 genes in liver) were among the biological processes altered in both organs (Supporting Information Table S1). In liver, only lipid metabolic processes (GO:0006629) involved the largest number of modulated genes (16 genes). In kidney, several processes related to the immune system (GO:0002376; 32 genes), cell cycle (GO:0007049; 28 genes), stress response (GO:0006950; 27 genes), response to wounding (GO:0009611; 15 genes), inflammatory response (GO:006954; 12 genes) and kidney development (GO:0001822; 5 genes) were identified. In these processes genes such as Bcl6, Bbc3 and Cycs were modulated in both organs whereas others such as Cdkn1a, Stat3, Ccng1, Fos, Nqo1 and Rb1 were altered only in kidney.
We also carried out canonical pathway analysis and most pathways indicated were altered in both organs including apoptosis and Nfκb signaling. Several interesting pathways may be activated differently in kidney including aryl hydrocarbon receptor signaling (Nqo1, Aldh1a1, Mgst2, Ccnd2, Cdkn1a, Myc, Nfia, Rb1 and Tfdp1), Tp53 signaling (Bbc3, Ccnd2, Ccng1, Cdkn1a, Cdkn2a, Mdm2, Rb1 and Wt1) and cell cycle (Ccnd2, Cdkn1a, Cdkn2a, Myc, Rb1 and Tfdp1) and in liver including oxidative phosphorylation (Atp5h, Atp5l, Cox17, Cox7b, Ihpk2, Ndufb3, Ndufb6 and Sdhb).
The interactions of AAI-modulated genes were investigated using gene association network (GAN) analysis (Supporting Information Table S2). Two of these networks generated by 409 genes in kidney are shown in Figure 4. As shown in Figure 4a, GAN analysis identified 26 genes that can regulate Nfκb activity or can be regulated by Nfκb (in red or green) indicating that the main function of this network is inflammatory response and that Nfκb is considered to be the network core molecule. Similarly, other core genes including Myc (see Fig. 4b), Rb1 (see Fig. 4b), Fos, Cdkn1a, Hnf4a, Stat3 and Tp53 were indicated in kidney, some of which (Hnf4a, Rb1, Tp53) are also suggested in liver.
qRT-PCR is a more sensitive and specific measure of gene expression and was used for kidney to validate a number of interesting expression changes supported by the microarray data. Twelve genes (Nfκb1, Ccng1, Fos, Myc, Nqo1, Mdm2, Stat3, TP53, Bcl3, Bcl6, Bbc3 and Cdkn1a) were selected to be measured by qRT-PCR. Generally, the qRT-PCR changes confirmed the alterations identified by microarray (Fig. 5), but whereas the microarray analysis indicated gene expression changes for many of those genes only at 1 particular time point, qRT-PCR often showed the level of expression changes to increase over time. Although it is well known that qRT-PCR is a more accurate measure than microarray to study gene expression changes, it is noteworthy that some of the changes observed by qRT-PCR were only marginally below the cutoff on the microarray. Our results are in line with other studies where the gene expression levels can be quite varied at different time-points by microarray analysis.19, 27
Western blot analysis
GAN analysis revealed that Nfκb and Myc are potential biomarkers of AA exposure in kidney. Using Western blot analysis, we confirmed the induction of both Nfκb1 and c-Myc at the protein level in a time-dependent manner (Fig. 6). As AAN urothelial atypia have been associated with the overexpression of p53,29 we assessed its accumulation in kidney after AAI treatment. However, p53 protein was not detectable at any of the time points investigated (Fig. 6). In contrast, clear, although weak, Cdkn1a induction was observed in kidney at later time points (12 and 21 days; Fig. 6). Since AAI is known to be metabolically activated by microsomal cytochrome P450 (CYP) 1A1 and cytosolic NAD(P):quinone oxidoreductase (NQO1),30 we investigated Cyp1a1 and Nqo1 expression in renal tissue. Whereas no expression of Cyp1a1 was found in kidney, Nqo1 expression was strongly induced after AAI treatment at all time points (Fig. 6).
Humanized mouse models may be useful in toxicological research in addressing the problem of species differences in tumor development. The Hupki mouse model offers the opportunity to study the formation of mutations in human TP53 sequences in tumors of experimental animals after carcinogen treatment.31 These mice have wild-type character of the Hupki allele, which harbors WT human TP53 sequences from intron 3–9 in place of the corresponding murine sequences of the endogenous mouse Tp53 gene. Hupki mice homozygous for the knock-in allele do not differ in tumor response from their counterparts with murine Tp53.32 Upon irradiation gene expression profiles of homozygous Hupki mice were in concordance with profiles from WT mice31; expression patterns were highly similar in the 2 genotypes, both from untreated and from irradiated mice. The HUF assay has already been shown to be a powerful tool to induce and select mutations in human TP53 sequences of mammalian cells exposed to AAI in vitro.17, 18, 33 In this assay, AAI induced predominantly AT → TA transversion mutations, which are typically observed in TP53 in urothelial tumors from patients with BEN and AAN exposed to AA.11, 15 Similarly, these mutations have been observed in H-ras in tumors of rodents treated with AA.34, 35 Thus, Hupki mice may also be a promising model to examine AA-induced mutations in human TP53 in experimental animals in vivo. In this study, we have assessed DNA damage and gene expression changes induced by short-term AAI treatment in Hupki mice. In a separate, ongoing study Hupki mice were treated with AAI using the same experimental protocol to monitor tumor development and to screen AAI-induced tumors for TP53 mutation.
We based our treatment protocol on a previous carcinogenicity study in which female NMRI mice were treated with 5 mg/kg bw AA (80% AAI, 20% AAII) over a period of 3 weeks.22 In this study, tumors were observed in multiple organs including forestomach, kidney and lung within 56 weeks after the start of treatment. For treatment, we used purified AAI only. As expected for the short-term duration of the treatment (up to 3 weeks), no neoplastic lesions were observed by histopathology in this study. Although the hepatic pathology was normal, the renal pathology revealed tubular injury and tubulointerstitial fibrosis after AAI treatment. Further, we found that DNA binding in kidney was around 10-fold higher than in liver. These results are consistent with kidney being a target tissue and liver being a nontarget tissue of AAI-induced toxicity.
Chemical carcinogenesis is a multistep process and the formation of DNA adducts as promutagenic lesions is considered necessary but not sufficient for tumor development. Previously, we showed that specific DNA damage due to AA in urothelial cells and cell-specific alterations of protein levels might impair physiological processes (e.g., receptor-mediated endocytosis of low-molecular-weight proteins).36, 37 More recently, we found that organic anion transporters (OATs) mediate uptake of AA into proximal tubule cells and thereby participate in renal cell damage.38 Altered gene expression is another mechanism that may explain the specificity of AA-induced oncogenesis in view of the wider tissue distribution of AA-DNA adducts. Although we observed AAI-induced gene expression profiles in both the target organ kidney and in the nontarget organ liver, the profiles were very different for the 2 tissues. Similar findings have been seen in Big Blue rats treated with 10 mg/kg bw AA for 3 months,20 indicating that these changes appear to be related to tissue-specific effects of AA. The number of genes whose expression was significantly altered after treatment was much smaller in Hupki mice (this study) than in the rat study.20 However, in the Big Blue rat study none of the expression changes were investigated at the protein level.
Apoptosis was among the biological processes altered in both organs and this likely reflects DNA damage induced by AAI in both organs. One gene that was altered (upregulated) in both organs was Cycs, the gene encoding cytochrome c. The release of cytochrome c from mitochondria is essential for apoptosis occurring in response to cellular stress and damage.39 Tubular atrophy is a pathological feature in AAN and AA-induced apoptosis in tubular cells may be responsible for it.40 Besides Cycs, genes involved in the apoptotic process in kidney included Bbc3, Bcl3, Myc and Cdkn1a and their strong upregulation was confirmed by qRT-PCR. It is noteworthy that for the selected genes the qRT-PCR data confirmed their alteration identified by microarray.
Early growth response 1 (EGR1) has important tumor suppressor properties, e.g., via the direct induction of the epithelial cell suppressor TGFβ1 and the direct induction of p53 to promote apoptosis.41 We previously found that EGR1 was one of the most highly upregulated genes in human colorectal HCT116 cells after exposure to AAI in vitro21 and upregulation of Egr1 was now also observed in kidney after AAI treatment in vivo. In the study with HCT116 cells, a small set of genes was identified after AAI exposure that discriminated TP53 status in 2 isogenic HCT116 cell lines, 1 expressing wild-type TP53 (HCT116 TP53-WT) and the other with this gene knocked-out (HCT116 TP53-null).21 Among the TP53-regulated genes were CDKN1A and CCNG1. In this study, Cdkn1a and Ccng1 were strongly upregulated after AAI treatment, which is consistent with results in Eker and corresponding wild-type rats treated with 10 mg/kg bw AA (41% AAI, 56% AAII) for up to 14 days, where Cdkn1a and Ccng1 were upregulated in a time-dependent manner.19 However, at the protein level only weak induction of Cdkn1a was observed in this study. CDKN1A encodes a potent cell cycle inhibitor, regulating cell cycle progressing at the G1 and G2 check-points in response to a variety of stress stimuli.42CCNG1 encodes cyclin G1, which has growth inhibitory activity.43 In AAN urothelial atypia have been associated with the overexpression of p53,29 indicating its cellular accumulation in response to AA-induced DNA damage. In HCT116 TP53-WT cells, we previously found p53 accumulation after AAI exposure.21 Accumulation of p53 protein has been observed following DNA damage in Hupki mouse skin after UV exposure,44 but in this study, we did not find accumulation of the protein in kidney after AAI treatment.
Most of the biological processes that were altered in kidney only were related to the immune system, inflammatory response, stress response or cell cycle. Since AA is a potent nephrotoxin, it is expected that AAI treatment triggers a strong immunological and inflammatory response in the kidney.45 Genes involved in those responses included Bcl3, Bcl6, Smad1 and Stat3 and their upregulation was confirmed by qRT-PCR. Using GAN analysis we identified Nfκb as a network core molecule, although its gene expression was not changed on the microarray. However, using qRT-PCR, we found a clear overexpression of Nfκb1, and induction of Nfκb1 protein was confirmed by Western blotting. The NFκB signaling pathway plays a critical role in regulating innate and adaptive immune responses.46 In the immune system, tight regulation of NFκB signaling is crucial for maintaining the normal function of immune cells and avoidance of tumorigenesis.46 Interestingly, Bcl3 is an IκB family protein that is involved in transcriptional regulation of a number of NfκB-target genes. In addition RelB, which belongs to the Nfκb family, was also upregulated on the microarray in renal samples. Furthermore, it is noteworthy that abnormally high expression of both BCL3 and BCL6 can lead to the development of various B-cell leukemia.46 Recently, NFκB2 has been shown to be a transcriptional regulator of TP53, and consequently regulates TP53-target genes such as CDKN1A and BBC3 (PUMA) via a DNA-binding independent mechanism.47Bbc3 and Cdkn1a were both altered in this study and therefore Nfκb2 may be another candidate gene for further investigations. Finally, NFκB can function as a tumor promoter in inflammation-associated cancer48 and thus may be important in the development of AA-associated urothelial cancer.
Using GAN analysis, we identified Myc, Rb1, Mdm2, Cdkn2a and Tfdp1 as core genes in a functional network related to cancer and cell cycle. The retinoblastoma tumor suppressor gene (Rb1) gene was the most upregulated gene identified by microarray analysis. Deregulation of pathways in which RB1 functions is common in most types of human cancers.49 The Rb1-encoded protein (pRb) is well known as a general cell-cycle regulator, and this activity is critical for pRB-mediated tumor suppression. A recent study in Wistar rats treated with 10 mg/kg bw AA (41% AAI, 56% AAII) for 3 months showed that AA treatment increased the expression of cyclins and cyclin-dependent kinases (cdks), and the association of cyclins/cdks which promoted the activation of Rb, resulting in a decrease of the Rb/E2f complex.50 These results suggest that AA-induced urothelial proliferation of urinary bladder in rats is through cell-cycle progression. Furthermore, the cell-cycle progression was mediated via the induction of cyclin D1/cdk4 and/or cyclinE/cdk2 activity as well as the increasing phosphorylation of Rb, which may act as a possible neoplastic stimulus.50 In this context, it is noteworthy that the transcription factor Dp1 (encoded by Tfdp1) is critical for the activity of E2f; E2f-regulated genes are crucial for the progression of cells from G1 to S phase of the cell cycle.51 Moreover, Dp1 is a direct target of Arf, a tumor suppressor that is mutated or inactivated in many human tumors and linked to the p53 tumor suppressor pathway. Arf is a physiological regulator of Mdm2,52 and Mdm2 was strongly upregulated in kidney after AAI treatment. Although mutations in TP53 are found in many highly malignant human urothelial tumors,53 a subset of advanced urothelial cancers without TP53 mutations harbor MDM2 amplifications.54 Thus, the normal function of p53 may be abrogated by MDM2 gene amplification in urothelial cancers. Overexpression and amplification of the c-MYC gene have also been frequently observed in advanced human urothelial carcinoma.53, 55, 56 In this study, c-Myc was clearly overexpressed in the kidney after AAI treatment, and Myc induction was also clearly seen at the protein level. This is consistent with recent results that show that MYC was overexpressed in HCT116 TP53-WT cells after exposure to AAI. Besides many other cellular functions, MYC is required in normal cells for cell cycle competence, whereas in tumor cells it is overexpressed and functions as the angiogenic switch.57 Finally, another oncogene with upregulated expression in AAI-treated kidney samples was Fos, and FOS has been linked to progression in human cancers.58 In contrast, we found previously that FOS was downregulated in HCT116 TP53-WT cells after exposure to AAI.
In summary, the tumor suppressor genes TP53, RB1 and CDKN2A as well as the oncogene MYC have been shown to be frequently affected in urothelial cancer in humans.59 In this study, expression of all those genes were found to be modulated or targets in the kidney of Hupki mice after AAI treatment. Apart from TP53, these genes have not yet been investigated in clinical AAN samples, but our findings suggest that they may have some relevance for AAN-associated urothelial carcinomas. It is possible that these genes might be involved in the development and progression of AA-induced urothelial tumors, and may be valuable prognostic markers in future clinical studies. The potential for the markers (e.g., MYC) identified in mice to be of value in human clinical studies is currently being tested in urothelial tissue of patients with AAN.
We thank the Cancer Research UK DNA Microarray Facility and Drs. Ian Giddings and Daniel Brewer for advice and help with the microarray data depository. We also thank Ms. Maria L. Simões (Institute of Cancer Research) for excellent technical assistance. J.Z. received a PhD studentship from the Institute of Cancer Research. V.M.A., H.H.S. and D.H.P. are members of ECNIS (Environmental Cancer Risk, Nutrition and Individual Susceptibility), a network of Excellence operating with the European Union 6th Framework Program, Priority 5; “Food Quality and Safety” (Contract No. 513943).