1186
Chem. Res. Toxicol. 2008, 21, 1186–1196
Chemical Profile Transcriptomic Analysis of Nephrotoxicity Induced by Cephaloridine, a Representative Cephalosporin Antibiotic Masatomo Rokushima,*,† Kae Fujisawa,‡ Naoko Furukawa,‡ Fumio Itoh,‡ Toru Yanagimoto,† Ryou Fukushima,‡ Akiko Araki,† Manabu Okada,‡ Mikinori Torii,‡ Ikuo Kato,‡ Jun Ishizaki,‡ and Kazuo Omi† DiscoVery Research Laboratories, Shionogi & Co., Ltd., 12-4, Sagisu 5-chome, Fukushima-ku, Osaka 553-0002, Japan, and DeVelopmental Research Laboratories, Shionogi & Co., Ltd., 1-1, Futaba-cho 3-chome, Toyonaka, Osaka 561-0825, Japan ReceiVed December 31, 2007
Cephaloridine (CER) is a classical β-lactam antibiotic that has long served as a model drug for the study of cephalosporin antibiotic-induced acute tubular necrosis. In the present study, we analyzed gene expression profiles in the kidney of rats given subtoxic and toxic doses of CER to identify gene expression alterations closely associated with CER-induced nephrotoxicity. Male Fischer 344 rats were intravenously injected with CER at three different dose levels (150, 300, and 600 mg/kg) and sacrificed after 24 h. Only the high dose (600 mg/kg) caused mild proximal tubular necrosis and slight renal dysfunction. Microarray analysis identified hundreds of genes differentially expressed in the renal cortex following CER exposure, which could be classified into two main groups that were deregulated in dose-dependent and high dose-specific manners. The genes upregulated dose dependently mainly included those involved in detoxification and antioxidant defense, which was considered to be associated with CER-induced oxidative stress. In contrast, the genes showing high dose-specific (lesion-specific) induction included a number of genes related to cell proliferation, which appeared to reflect a compensatory response to CER injury. Of the genes modulated in both manners, we found many genes reported to be associated with renal toxicity by other nephrotoxicants. We could also predict potential transcription regulators responsible for the observed gene expression alterations, such as Nrf2 and the E2F family. Among the candidate gene biomarkers, kidney injury molecule 1 was markedly upregulated at the mildly toxic dose, suggesting that this gene can be used as an early and sensitive indicator for cephalosporin nephrotoxicity. In conclusion, our transcriptomic data revealed several characteristic expression patterns of genes associated with specific cellular processes, including oxidative stress response and proliferative response, upon exposure to CER, which may enhance our understanding of the molecular mechanisms behind cephalosporin antibioticinduced nephrotoxicity. Introduction Cephaloridine (CER)1 is a first-generation cephalosporin antibiotic that has a broad spectrum of activity against Gramnegative and Gram-positive bacteria (1). This antibiotic is wellknown to cause renal damage in humans and experimental animals, which is characterized by acute necrosis of proximal tubular epithelial cells (S2 segment target) in the renal cortex (2). Although CER was withdrawn from the market due to nephrotoxicity, it has long served as a model drug for the study of cephalosporin antibiotic-induced renal tubule damage (3). * To whom correspondence should be addressed. Tel: (81)6-64585861 ext. 5576. Fax: (81)6-6455-2099. E-mail: masatomo.rokushima@ shionogi.co.jp. † Discovery Research Laboratories. ‡ Developmental Research Laboratories. 1 Abbreviations: CER, cephaloridine; ROS, reactive oxygen species; PKCδ, protein kinase C δ; MEK, mitogen and extracellular signal-regulated kinase kinase; ERK, extracellular signal-regulated protein kinase; FDVE, the degradation product of the anesthetic sevoflurane; BUN, blood urea nitrogen; FDR, false discovery rate; GO, gene ontology; RT-PCR, reverse transcription polymerase chain reaction.
CER is actively transported from the blood into the tubular cells by organic anion transporters in the basolateral membrane but does not readily move to the tubular lumen through the brush border membrane, resulting in considerable intracellular accumulation of the drug (4). This condition leads to the generation of free radicals such as reactive oxygen species (ROS), which provoke depletion of reduced glutathione and increased lipid peroxidation; the latter most likely causes the observed histopathological lesions (5, 6). The generation of free radicals is reported to occur in mitochondria, which is preceded by activation of protein kinase C δ (PKCδ) and its translocation into mitochondria (7, 8). It has also been reported that the mitogen and extracellular signal-regulated kinase kinase (MEK)/ extracellular signal-regulated protein kinase (ERK) pathway and the cAMP pathway are involved in the free radical-mediated nephrotoxicity of CER as well (9, 10). Thus, the cellular and molecular mechanisms by which this antibiotic causes injury to the proximal tubular cells have been partly elucidated; however, little is known about the gene expression alterations underlying CER-induced in vivo nephrotoxicity.
10.1021/tx800008e CCC: $40.75 2008 American Chemical Society Published on Web 05/23/2008
Transcriptomics of Cephaloridine Nephrotoxicity
Toxicogenomics is an interdisciplinary field that combines transcriptomics using microarray technology with conventional toxicology (11, 12). This approach holds promise for elucidating molecular mechanisms underlying compound-induced toxicity and identifying gene expression signatures that allow more effective safety assessment of chemicals or new drug candidates (13–16). Nephrotoxicity is one of the types of toxicity for which toxicogenomic approaches have been most successfully applied. This is because several studies have shown that altered gene expression could be used to discern regional specific damage of nephrotoxicity and to detect subtle kidney injury without overt phenotypes defined by traditional toxicological assays, such as clinical chemistry. For example, Amin and co-workers analyzed microarray data of the kidney of rats treated with cisplatin, gentamicin, and puromycin and predicted the mild proximal tubular damage caused by puromycin, which is thought to specifically injure the glomerular podocytes (14). In addition, Thukral et al. examined gene expression profiles derived from the kidney of rats exposed to six model nephrotoxicants and identified a subset of genes that enabled the prediction of pathological outcome (17). Kharasch et al. showed that gene expression alteration was a much more sensitive indicator of the nephrotoxicity induced by the degradation product of the anesthetic sevoflurane (FDVE) as compared to classical toxicological parameters (18). The candidate gene markers reported to be associated with renal toxicity in rodents include kidney injury molecule 1, clusterin, and osteopontin (19). Some of such potential biomarkers have also been shown to be inducible in nonhuman primate models of nephrotoxicity, suggesting that these candidates are applicable across multiple species (19). Furthermore, excreted proteins of some putative gene markers are detectable in the urine and hence are being validated as diagnostic indicators for kidney injury in preclinical and clinical studies (20). In this study, we investigated gene expression alterations in the kidney cortex of rats treated with mildly toxic and subtoxic doses of CER for 24 h. A toxic dose is defined as a dose that causes histopathologically apparent tubular necrosis and/or changes in the two clinical chemistry parameters of renal function: blood urea nitrogen (BUN) and creatinine. Both toxic and subtoxic dose levels were employed to distinguish between genes deregulated specifically at the toxic dose and those affected in a dose-dependent manner. Alternatively, short-term exposure was selected to examine the early response of altered gene expression, which could allow us to evaluate the utility of candidate biomarkers as early indicators of cephalosporin nephrotoxicity. The aims of the present study were (i) to identify altered gene expression closely associated with CER-induced acute tubular necrosis, (ii) to explore gene biomarkers that could be applicable to the detection of acute CER nephrotoxicity, and (iii) to gain more insight into the molecular mechanism(s) of CER toxicity.
Materials and Methods Chemicals and Materials. CER of clinical grade was synthesized and prepared inhouse. Saline was from Otsuka Pharmaceutical Factory (Tokushima, Japan). Pentobarbital sodium and the bloodcollecting tube containing heparin sodium were from Terumo (Tokyo, Japan). RNAlater was from Ambion (Austin, TX). QIAzol Lysis Reagent, MagAttract RNA Cell Mini M48 Kit, and RNeasy Mini Kit were from Qiagen (Valencia, CA). Cyanine 3-CTP was from PerkinElmer (Wellesley, MA). Low RNA Input Linear Amplification Kit and Whole Rat Genome Oligo Microarray were purchased from Agilent Technologies (Santa Clara, CA).
Chem. Res. Toxicol., Vol. 21, No. 6, 2008 1187 Animals and Animal Care. Male Fischer 344 rats (7 weeks old) were purchased from Charles River Laboratories Japan (Kanagawa, Japan) and housed in plastic cages in an environmentally controlled room (12 h light-dark cycle, 23 ( 3 °C, and 50 ( 20% relative humidity). The animals had ad libitum access to certified rodent chow (Oriental Yeast, Tokyo, Japan) and water and were acclimatized under these conditions for a week before experimentation. Animal maintenance and treatment were performed according to the principles outlined in the Guide for the Care and Use of Laboratory Animals prepared by the Japanese Association for Laboratory Animal Science and our institution. Study Design and Administration of Chemicals. Rats were randomly assigned to dose groups according to body weight (three rats/group). CER was dissolved in saline and intravenously administered as a single dose of 0 (vehicle only), 150, 300, or 600 mg/kg. The high dose (600 mg/kg) was selected based on the literature where a similar dose level caused proximal tubular necrosis in male Fischer 344 rats (21). At 24 h after the administration, the animals were sacrificed under anesthesia with pentobarbital sodium. At necropsy, a blood sample was drawn from the posterior vena cava and put into a vacuum blood-collecting tube containing heparin sodium for clinical chemistry examination. The rats were then euthanized by exsanguination, and both kidneys were collected and weighed. Transverse sections of the kidney cortex were fixed in 10% neutral buffered formalin for histopathological evaluation. Other portions of the cortex were immediately soaked in RNAlater and stored at -80 °C until the RNA was isolated. Histopathology and Clinical Chemistry Analysis. The formalin-fixed kidney sections were trimmed and embedded in paraffin. Sections of approximately 3 µm in thickness were taken and stained with hematoxylin and eosin (H&E). Histopathological examinations of the tissue sections were conducted under a light microscope. Plasma was obtained from the blood sample collected with heparin sodium. BUN and creatinine levels were measured for the plasma using an automatic analyzer 7170 (Hitachi, Tokyo, Japan). RNA Isolation and Labeled cRNA Preparation. The kidney cortex was disrupted and homogenized in QIAzol Lysis Reagent with TissueLyser (Qiagen). Total RNA was isolated from the homogenates using a BioRobot M48 Workstation (Qiagen) in combination with MagAttract RNA Cell Mini M48 Kit following the manufacturer’s instructions, including a DNase digestion step. RNA concentration was determined by absorbance at 260 nm, and the integrity of the RNA was checked using an Agilent 2100 Bioanalyzer (Agilent Technologies). Labeled cRNA targets were prepared using Low RNA Input Linear Amplification Kit according to the “One-Color MicroarrayBased Gene Expression Analysis” protocol provided by the manufacturer (Agilent Technologies). Briefly, 500 ng of total RNA was reverse-transcribed using a primer containing oligo-dT and a T7 promoter. Using the resultant cDNAs as templates, cRNA targets were synthesized by in vitro transcription in the presence of cyanine 3-CTP. The resulting labeled cRNAs were purified using RNeasy Mini Kit and quantified by absorbance at 260 nm. Microarray Analysis. Cyanine 3-labeled cRNA targets were hybridized to Whole Rat Genome Oligo Microarray (Agilent Technologies, G4131F, 41,012 probes) at 65 °C for 17 h following the procedures recommended by the manufacturer. After hybridization, the microarray slide was washed, dried, and scanned on an Agilent DNA microarray scanner (Agilent Technologies). Hybridization images were quantified using Feature Extraction Software 9.1 (Agilent Technologies), and background-subtracted signal values were utilized for the following analyses. The entire microarray data are publicly available through the GEO database (http://www. ncbi.nlm.nih.gov/geo/) with accession number GSE10034. Raw microarray data were processed and analyzed mainly with GeneSpring GX 7.3.1 (Agilent Technologies). For per array normalization, the 50th percentile of the signal values taken from each microarray was used as the normalizing reference. Per gene normalization was based on calculating fold expression levels relative to the median of the control animals (n ) 3). We chose
1188
Chem. Res. Toxicol., Vol. 21, No. 6, 2008
Rokushima et al.
Table 1. Conventional Toxicological Parametersa plasma CER dose (mg/kg)
tubular necrosis (score)
BUN (mg/dL)
creatinine (mg/dL)
kidney weight (g/100 g BW)
control 150 300 600
0 (0-0) 0 (0-0) 0 (0-0) 2 (2-2)
23 ( 2 25 ( 1 22 ( 2 31 ( 0b
0.25 ( 0.01 0.27 ( 0.01 0.24 ( 0.02 0.47 ( 0.04b
0.75 ( 0.02 0.78 ( 0.04 0.80 ( 0.02 0.87 ( 0.03b
a Tubular necrosis is expressed as a median (range) of severity score, which consists of a range from 0 to 4 (normal, minimal, mild, moderate, and marked, respectively). BUN and creatinine levels and kidney weight are presented as means ( SD (n ) 3). Kidney weight is shown as relative weight (sum of weights of the left and right kidneys relative to 100 g of body weight). b Significantly different from the control group (p < 0.01).
the probes with significant signal above background (termed as “gIsWellAboveBG” in Feature Extraction Software) in all of the 12 samples to eliminate data with low reliability. The probes that passed this filter (23154 probes) were then analyzed for differential expression. To determine cutoff value of fold change for differential expression, we performed power prediction analysis (22). As a result, we concluded that filtering gene expression ratios using 1.5fold change would identify significant genes in our experimental condition (Supplementary Method). Then, we selected probes with expression levels of 1.5-fold change or more relative to the control in all of three rats within individual dose groups. The resulting probe lists were further evaluated for statistical significance using an unpaired two-sample t test (Welch’s t test) with the false discovery rate (FDR) multiple testing correction (FDR < 0.05) (23). The genes for which the annotation was unclear were excluded from Tables 2 and 3, where representative deregulated genes are presented. K-means clustering of deregulated probes was performed with 1000 iterations using standard correlation as a similarity measure. Gene ontology (GO) analysis was conducted using the “GO Ontology Browser” function implemented in the software. Nonredundant gene sets for lists of probes belonging to individual clusters in the K-means clustering were generated and used for GO analysis because these probe lists included multiple probes for a single gene, which would lead to misidentification of overrepresented GO categories. The GO categories containing over 2000 probes on the microarray or having no unique deregulated genes were removed from the resulting lists to exclude ambiguous or highly similar categories, respectively. The expression data of probes with significant signal above background in all samples (23154 probes) were also subjected to network analysis using PathwayExpert software and the ResNet database (Ariadne Genomics, Rockville, MD). This allows construction of molecular networks based on a comprehensive database of interactions reported in literature (24). Briefly, the averaged log2 ratios (treated/control) of the probes were loaded into PathwayExpert and mapped to their corresponding gene objects by using EntrezGene ID as an identifier. The transcription regulators having significantly modulated gene targets were searched for using the “Find Significant Regulators” algorithm in the “Analyze Experiment” function, with “PromoterBinding” employed as the network type. The resulting putative transcription factors with a p < 0.01 were considered. Statistics. Data for BUN and creatinine levels and relative kidney weight are expressed as means ( SD (n ) 3), and Dunnett’s multiple comparison test was used to evaluate the differences between control and treated groups. Statistical analyses except for those of microarray data were performed using SAS software.
Results (Histo)pathology and Clinical Chemistry Analyses. Male Fischer 344 rats were intravenously injected with CER at three dose levels (150, 300, and 600 mg/kg) and sacrificed after 24 h. To confirm the incidence of acute kidney injury, we conducted
conventional toxicological assays. In histopathological examination, no lesions were seen in the kidney of rats treated with the low and middle doses (150 and 300 mg/kg) of CER as well as vehicle (saline) alone. The high dose (600 mg/kg) of CER induced mild necrosis of proximal tubular epithelium in the cortex in all of three rats with uniform severity (Figure 1 and Table 1). Clinical chemistry analysis revealed slight but significant increases in BUN and creatinine levels only at the high dose of CER, which is consistent with the histopathological results (Table 1). A significant increase in relative kidney weight was also observed at the high dose. Thus, on the basis of these traditional toxicological assays, we concluded that only the high dose of CER had provoked mild but significant renal failure phenotypes. The low and middle doses of CER caused no changes associated with nephrotoxicity and, therefore, were considered to be subtoxic doses. Microarray Analysis. Global gene expression profiles in the kidney cortex of rats given a single dose of CER were obtained using whole-genome oligonucleotide microarrays. We first extracted differentially expressed probes (genes) at individual dose levels based on the criterion of a 1.5-fold change relative to the control and FDR
