ToxSci Advance Access originally published online on December 14, 2006
Toxicological Sciences 2007 96(1):101-114; doi:10.1093/toxsci/kfl184
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Determination of Phospholipidosis Potential Based on Gene Expression Analysis in HepG2 Cells
UCB Pharma SA, Non-Clinical Development, Chemin du Foriest, 1420 Braine-l'Alleud, Belgium
1 To whom correspondence should be addressed. Fax: +32 (0) 2-3862798. E-mail: franck.atienzar{at}ucb-group.com.
Received September 11, 2006; accepted November 27, 2006
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Phospholipidosis (PLD) is characterized by an intracellular accumulation of phospholipids in lysosomes and the concurrent development of concentric lamellar bodies. Recently, H. Sawada et al. (2005, Toxicol. Sci. 83, 282–292) identified 17 genes as potential biomarkers of PLD in HepG2 cells. The present study was undertaken to determine if this set of genes measured by quantitative PCR could be validated in the same cell line. The objective was also to investigate the dose-response relationship to further validate the assay and to select the concentrations to use for screening activities. In a first experiment (one concentration tested), out of the 17 genes, the best gene biomarkers of PLD (i.e., 11 genes) were selected for practical screening reasons. Based on these genes, 91.6% (i.e., 11 of 12) of the compounds known to induce PLD were identified as positive and all the negative compounds (i.e., five of five) were also confirmed. When the data obtained in the first experiment were compared to the data by Sawada et al., (2005) the coefficient of correlation calculated was slightly higher than 75%. In the second experiment (26 compounds [all 17 compounds from the first experiment plus 9 other compounds] tested at a minimum of three concentrations), 93.3% (14/15) of the compounds known to induce PLD were identified as such and all the negative controls (six compounds) were also confirmed. Three compounds likely to induce PLD were identified as positive in our assay. Finally, two compounds for which no data are available were also tested. When both experiments 1 and 2 were compared, the coefficient of correlation for 16 compounds tested at the same concentrations reached 87.7%. In conclusion, the present study further confirms the utility of gene expression in HepG2 cells to identify a potential to induce PLD. Finally, based on the data presented, researchers are encouraged to use a range of minimum three concentrations (e.g., 12.5, 25, and 50µM) to screen for PLD in the human HepG2 cell line.
Key Words: phospholipidosis; HepG2 cells; gene expression; biomarkers; screening activities.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Phospholipidosis (PLD) is characterized by an intracellular accumulation of phospholipids in lysosomes and the concurrent development of concentric lamellar bodies. It is often observed following treatment with cationic amphiphilic drugs (CADs). These drugs, which are characterized by a hydrophobic ring structure and a hydrophilic side chain with a charged cationic amine group, can interact with lipid cell components and cause lipid storage disorders. CADs that penetrate into the lysosomes will become protonized and will then be trapped in the lysosomal acidic medium (Lüllmann et al., 1978
). The phospholipids in those complexes can be protected against enzymatic attack from the phospholipases which require negatively charged substrate to operate (Joshi et al., 1989). Other theories to explain PLD imply the inhibition of phospholipase activity, increased synthesis of phospholipids, and inactivation of lysosomal enzymes due to an increase in lysosomal pH resulting from the accumulation of the CAD within the organelle or a combination of these actions (Kodavanti and Mehendale, 1990; Lüllmann et al., 1978). CADs are used in many therapeutic indications such as local anesthetics, antimalarial agents, antibiotics, antiarrhythmics, antidepressants, and others (Kodavanti and Mehendale, 1990; Schneider et al., 1997). As early as 1979, Lüllmann-Rauch (1979) reported that 30 drugs were known to induce PLD in humans, animals, or cell cultures. In addition, genetic disorders (e.g., Niemann-Pick and Tay-Sachs diseases) as well as non-CADs (e.g., xenobiotics, hormones, cofactors, and other agents) may alter the metabolism of the cells and indirectly lead to PLD (Kodavanti and Mehendale, 1990; Schneider et al., 1997).
Up to now, there is no definitive evidence that the presence of CAD-induced PLD per se is detrimental to animals or humans (Reasor et al., 2006). Camus and Mehendale (1986) reported that despite massive induction of pulmonary PLD, there were only minor effects on lung function. The administration of amiodarone induced increased hepatic density and fibrosis in addition to PLD in different animal models and human (Cantor et al., 1987). In most cases, the toxicological implications of PLD are unknown (Reasor, 1989). The only circumstance where a causal relationship between the presence of PLD and tissue dysfunction has been suggested is with gentamycin (a well-studied aminoglycoside) and nephrotoxicity. Indeed, gentamycin-induced PLD has been observed in conjunction with renal toxicity (Laurent et al., 1990) and the inhibition of the development of PLD inhibits nephrotoxicity (Samadian et al., 1993). The possible toxicological consequences of PLD are an area of great interest both to the Food and Drug Administration (FDA) and the pharmaceutical industry. In 2004, the FDA established a PLD working group whose objective is to determine whether compounds-inducing PLD pose a potential clinical risk. Consequently, a PLD database is being populated with clinical and preclinical data on compounds from in-house FDA reviews. The results will hopefully enable to come to a decision resulting in a PLD guidance that will be useful to the pharmaceutical industry. Nevertheless, very recently, Reasor et al. (2006) indicated that from a regulatory perspective and consistent with the task of determining drug safety, PLD has been considered as an adverse finding, whether justified or not.
The presence of foamy cells as detected by histopathological analyses only suggests that PLD may have occurred. The standard confirmatory method for the detection of PLD on ex vivo samples is electron microscopy (Drenckhahn et al., 1976). However, the method is expensive, time consuming, and not dedicated to medium and high-throughput screenings. In addition, drug-induced PLD detected by electron microscopy could be discovered only after subchronic to chronic intake of compounds. In silico approaches have been successfully used to predict drug-induced PLD. For instance, the data presented by Ploemen et al. (2004) support the use of simple physicochemical calculations of ClogP and pKa to discriminate rapidly between compounds suspected of being PLD inducers. Nevertheless, there is a clear need to develop in vitro methodologies to enable rapid assessment during the drug development and help in the selection of the best candidates. PLD can be rapidly assessed by measuring the binding of dyes to the phospholipids. For instance, different cellular models (e.g., primary hepatocytes, HepG2, U-937, CHO-K1, and CHL/IU cell lines, spleen macrophages), dyes (e.g., nile red, NBD-PC), and methodologies (e.g., flow cytometry, fluorescence microscopy, gene expression) have been successfully used to detect PLD under in vitro conditions (Casartelli et al., 2003; Gum et al., 2001; Kasahara et al., 2006; Morelli et al., 2006; Sawada et al., 2005; Ulrich et al., 1991). Recently, Sawada et al. (2005), who developed a rapid and sensitive in vitro screening test for drug-induced PLD based on gene expression analysis in HepG2 cells, identified 17 genes as potential markers of PLD. To increase the throughput of the assay, the PCR-based assay was transferred into a 96-well microplate-based multiple mRNAs measuring assay (Sawada et al., 2006).
In a previous study, we showed that gene expression analysis generated with an array containing a restricted set of genes chosen as relevant markers of toxicity and metabolism allowed us to cluster 11 different hepatotoxicants (de Longueville et al., 2003). In the same spirit, the present study was undertaken to determine if we could (1) confirm the results published by Sawada et al. (2005) using the same human cell line, (2) assess the relevance of the model using various concentrations of compounds known to induce PLD or not, and (3) determine the reliability and reproducibility of the assay.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Chemicals
The following drugs (acetaminophen, amiodarone, amitriptyline,
-naphthyl-isothiocyanate [ANIT], AY9944, ß-naphtoflavone [BNF], chlorpromazine, clomipramine, clozapine, disopyramide, erythromycin, flecainide, fluoxetine, haloperidol, isoniazid, ketoconazole, ofloxacin, perhexiline, procainamide, sotalol, tamoxiphen, tetracycline, thioridazine, valproic acid, and zimelidine) were purchased from Sigma (St Louis, MO). Phenobarbital was purchased from CERTA (Braine-l'Alleud, Belgium).
Cell Culturing
The human hepatocellular carcinoma cell line (HepG2) was purchased from the European Collection of Cell Cultures (Salisbury, UK). Cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 50 U/ml penicillin, 50 µg/ml streptomycin, 2 mmol/l L-glutamine, and nonessential amino acid solution at 37°C in a humidified 5% CO2/95% air atmosphere. Medium was changed twice a week. Cells were passaged as needed using 0.5% trypsin-EDTA. All the cell culture solutions, mentioned in this section, were purchased from BioWhittaker Inc. (Walkersville, MD). The experiments were performed with HepG2 cells passaged between 10 and 35 times. Due to our screening activities, the responses of the HepG2 cells to reference compounds are regularly checked in the context of cytotoxicity and PLD experiments.
Concentrations Tested and Drug Treatment
HepG2 cells were plated and cultured overnight at a concentration of 106 cells (2 ml) per well (six-well format). The final concentration of dimethylsulfoxide (vehicle; Sigma) was 0.5%. Cells were incubated in a 37°C incubator in an atmosphere of 5% CO2/95% air and were exposed for 24 h to the compounds. The cells were then rinsed twice with few milliliters of phosphate-buffered saline at 37°C and placed at – 80oC prior to total RNA extraction.
Experiment 1.
The cells were exposed to a single concentration (X = 8.3 or 25µM) of positive and negative compounds according to the concentrations used by Sawada et al. (2005). The compounds tested were acetaminophen, amiodarone, amitriptyline, AY9944, chlorpromazine, clomipramine, clozapine, erythromycin, flecainide, fluoxetine, ketoconazole, ofloxacin, perhexiline, sotalol, tamoxiphen, thioridazine, and zimelidine. For more information on the compounds and doses tested refer to Tables 1 and 2, respectively.
|
|
Experiment 2.
The HepG2 cells were exposed to the same compounds as in experiment 1 but at minimum three different concentrations (X/2, X, and 2X) as well as to additional compounds (ANIT, BNF, disopyramide, haloperidol, isoniazid, phenobarbital, procainamide, tetracycline, and valproic acid). For more information on the compounds tested refer to Table 1. The concentrations of the negative drugs (compounds not tested in experiment 1) were 800, 400, and 200µM unless cytotoxicity occurred at these doses. In this case, lower concentrations were tested. For the positive drugs and compounds likely to induce PLD, concentrations in the range 2–30µM were tested except for tetracycline (6.25–800µM). For more information about the tested doses refer to Table 3.
|
Cytotoxicity Determination
No cytotoxicity assays were performed for the selection of the doses as there was sufficient available information on the reference compounds in the literature. In the context of screening, it is advisable to measure cytotoxic endpoints with proprietary compounds before evaluating PLD potential. Nevertheless, the amount of total RNA extracted, which correlates directly with the number of viable cells, gives some clear indication about cytotoxicity effects.
Total RNA Extraction and Reverse Transcription
RNA extraction was performed with the RNeasy Qiagen kit (Qiagen, Hilden, Germany). A DNAse step was included to reduce genomic DNA contamination. The integrity of the total RNA was estimated with the 2100 Bioanalyzer (Agilent, Palo Alto, CA). Quantification of the total RNA was performed with the Nanodrop (Nanodrop Technologies, Wilmington, DE). RNA samples were stored at – 80°C until assayed. Concentration of RNA samples was adjusted to 250 ng/µl, and reverse transcription (RT) was performed in triplicate with 2 µg RNA and oligo-dT oligonucleotide primer in a final volume of 20 µl. Pooled samples were then quality checked with the 2100 Bioanalyzer.
Quantitative PCR (SYBR green)
Technical validation.
Quantitative PCR (QPCR) experiments were performed in triplicate with different amounts of template (RT samples diluted 10, 40, 160, 640, 2560, and 10240 times) (data not shown). Negative, genomic, and mRNA controls were also included (in duplicate). Quantitative real-time PCR was performed using 4 µl of the RT dilutions, 12 µl of 2x master mix solutions (Stratagene, La Jolla, CA), 4 µl of primers (300µM for Sigma Proligo primers but unknown for Qiagen primers), and 0.1 µl of ROX (diluted 125 times) (Stratagene). All the primers were ordered from Qiagen (Hilden, Germany) except AP1S1, TAGLN, and FLJ10055 which were ordered from Sigma Proligo (Paris, France). For AP1S1, TAGLN, and FLJ10055, the technical validation criteria (see below) were not met with the Qiagen primers and consequently other primers had to be tested. Technical information (e.g., sequences, concentrations) of the Qiagen primers is unknown; the sequences of the proligo primers were those used by Sawada et al. (2005). QPCR measurements were performed with the MX3000p PCR machine (Stratagene) with the following conditions: 15 min denaturation at 95°C followed by 40 cycles of denaturation at 94°C for 20 s, annealing temperature of 55°C for 30 s, and extension at 72°C for 30 s. Denaturation curves were produced after the 40 cycles to check the specificity of the reactions. A pair of primers was considered validated when the efficacy of amplification was comprised between 90–110% and with a minimum r2 of 0.98. Further experiments were only conducted with validated primers.
Experiments 1 and 2.
In the first experiment, 17 genes identified as potential biomarkers of PLD by Sawada et al. (2005) (plus one housekeeping gene) were quantified by QPCR (SYBR green). In the second experiment, the genes validated during the first experiment (11 gene biomarkers of PLD plus 1 housekeeping gene) were measured. RT samples were diluted 20 times. Quantitative real-time PCR was performed with the conditions described in the technical validation section. Denaturation curves were produced after the 40 cycles to check the specificity of the reactions. Fluorescence emission was detected for each PCR cycle and the threshold cycle (CT) values were determined. The CT value was defined as the actual PCR cycle when the fluorescence signal increased above the background threshold. Average CT values from triplicate PCR reactions were normalized to average CT values for housekeeping gene (transferrin receptor gene [TFR]) from the same cDNA preparations. The calculations were done as follows: ratio = 2–
CT and CT = (CT gene 1 treated – CT TFR gene 1 treated) – (CT gene 1 control – CT TFR gene 1 control).
Rules for the classification of the gene as potential biomarkers of PLD.
The 17 genes measured during the first experiment were classified according to two parameters (i.e., scores 1 and 2). A cutoff value of 1.5 was chosen arbitrarily for the calculation of both scores which are presented in Table 2.
Score 1 is based on the gene expression level of the compounds that induced PLD in the paper by Sawada et al. (2005) (i.e., PLD pathology scores [PPSs] = +, ++, or +++) with the following rules:
For the first 14 genes (generally upregulated by compounds inducing PLD), if ratio:
- > 1.5
score = 1
- < 1.5
score = 0.
- < 1.5
For the last three genes (generally downregulated by compounds inducing PLD), if ratio:
- < – 1.5 score = 1
- > – 1.5
score = 0.
- > – 1.5
Score 2 is based on the gene expression level of the compounds that did not induce PLD in the paper by Sawada et al. (2005) (i.e., PPS = –) with the following rules:
For the first 14 genes (generally upregulated by compounds inducing PLD), if ratio:
- < 1.5 score = 1
- > 1.5
score = 0.
- > 1.5
For the last three genes (generally downregulated by compounds inducing PLD), if ratio:
- < – 1.5 score = 0
- > – 1.5
score = 1.
- > – 1.5
The overall score = score 1 x score 2. The highest score = 12 (maximum score 1) x 5 (maximum score 2) = 60. A classification is also given in Table 2 according to the overall score. The highest and lowest overall scores correspond to the best and worst gene biomarkers of PLD. An example of the calculation of scores 1 and 2 is given for MGC4171 and AP1S1 in Table 2.
Calculation of the PLD Index
The use of the drugs known to induce PLD in the present study revealed that the nine following genes (C10orf10, p8, ASNS, FRCP1, SERPINA3, INHBE, NRB02, MGC4171, and FLJ10055) were upregulated, whereas the two following genes (AP1S1 and TAGLN) were downregulated. In the present paper, the gene expression figures represent the ratio "mRNA level in treated/mRNA level in control" (normalized with the TFR housekeeping gene). The PLD index (PI) can be defined as follows:
PI = average of the level of expression of the 11 genes biomarkers of PLD with the following rules:
Upregulated genes by the drugs known to induce PLD, ratio:
- > 1 score = level of expression
- Negative value
score = 0.
- Negative value
Downregulated genes by the drugs known to induce PLD, ratio:
- < – 1 score = – (level of expression) (i.e., a positive value)
- Positive value
score = 0.
- Positive value
Brief Gene Description of the 11 Genes Biomarkers of PLD
C10orf10 (chromosome 10 ORF 10): gene expressed in adipose tissue and induced by fasting. FRCP1/FNDC4/FLJ22362 (fibronectin type III domain): no information available. SERPINA3 (serine or cysteine proteinase inhibitor): a plasma protease inhibitor. NRB02 (nuclear receptor subfamily 0, group B): orphan nuclear receptor, regulation of cholesterol metabolism. MGC4171/GDPD3/MGC4171/FLJ22603 (glycerophosphodiester phosphodiesterase): protein implicated in glycerol metabolism. FLJ10055/WIPI1 (WD40 repeat protein interacting with phosphoinositides of 49 kDa): key components of many essential biological functions. The protein contains a conserved motif for interactions with PLDs. TAGLN (transgelin): an actin cross-linking protein. p8 (p8 protein): protein implicated in metastasis 1, induction of apoptosis and cell growth. ASNS (asparagine synthetase 440): protein involved in the synthesis of asparagine. INHBE (inhibin, beta E): protein implicated in cell growth and/or maintenance. AP1S1 (adaptor-related protein complex 1, sigma 1 subunit): links clathrin to receptors in coated vesicles.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Table 1 gives some information on the potential of the drugs used in this study to induce PLD or not according to literature data. It also presents some of the results obtained in experiments 1 and 2
Rules for the Classification of the Compounds
Based on our data (first and second experiments) and on the paper by Sawada et al. (2005) and see also the "Discussion" section, we consider that a compound can potentially induce PLD when (1) PI is higher than 1.25 at a maximum concentration of 50µM and when (2) a dose-response relationship is obtained. More explanations are provided at the end of the results section.
Experiment 1: HepG2 Cells Exposed to a Single Concentration of Compounds
Selection of the gene biomarkers of PLD.
The expression of 17 genes (candidate markers of PLD) identified by Sawada et al. (2005) was measured in HepG2 cells exposed to 17 compounds known to induce PLD or not (Table 2). The best gene biomarker of PLD was classified according to two parameters (i.e., scores 1 and 2 as described in "Materials and Methods" section). According to the overall score (i.e., score 1 x score 2), a classification is also reported. "1" and "17" correspond to the best and worst gene biomarkers of PLD (Table 2). Most of the genes responded well to the compounds known to induce PLD. In particular, the following genes (INHBE, P8, C10orf10, and TAGLN) were induced or repressed (minimum 1.5 times) by at least 11 of the 12 positive compounds (Table 2). In addition, the expression of these genes either did not vary much or did not follow the trend induced by the positive compounds when the HepG2 cells were exposed to drugs not inducing PLD (Table 2). The data show that the genes classified from 1 to 14 (Table 2) could have been initially selected as potential gene markers of PLD. However, for practical reasons, the best 11 genes were selected according to the overall score presented in Table 2 plus 1 housekeeping gene (i.e., the TFR gene) because these 12 genes could be measured in two 96-well plate formats. The selected genes were MGC4171, NR0B2, INHBE, P8, SERPINA3, ASNS, C10orf10, FLJ10055 and FRCP1 (genes generally upregulated by compounds known to induce PLD), and AP1S1 and TAGLN (genes generally downregulated by compounds known to induce PLD) (Table 2). Other genes (classified at position 15, 16, and 17) do not appear to be good markers of PLD because they responded only to a limited number of compounds (e.g., SLC2A3 and PHYH) or due to reproducibility problems (i.e., ASAH). Table 2 gives also an overview of the genes selected by Sawada et al. (2005). To help the interpretation of the data, the PI was calculated based on the selected 11 genes (see "Materials and Methods" section).
Compound classification.
Following the rules mentioned above for the compound classification, 91.6% of the compounds known to induce PLD (i.e., 11 of 12) were identified as positive in agreement with the PPSs reported in Sawada et al. (2005) (Table 2). Positive PPS reflects the formation of myelin-like bodies in lysosomes identified by electron microscopy. In our experiment, tamoxiphen (a drug known to induce PLD) which had a PI of 1.13 (at 8.3µM) is classified as negative (PI < 1.25). All the negative drugs (i.e., five of five) were also identified as negative at 25µM (Table 2). Table 2 shows also a comparison of the PI calculated with our data (based on our set of 11 genes) and the data of Sawada et al. (2005). The coefficient of correlation calculated was slightly higher than 75%. Overall, this shows that similar results were obtained in both experiments (experiment 1 and data by Sawada et al. [2005]) although the same set of genes was not measured.
Experiment 2: HepG2 cells Exposed to a Minimum of Three Concentrations of Compounds
The objective was to test the effect of a range of concentrations but also to further validate the assay by using more compounds (i.e., 26 chemicals in total vs. 17 in the first experiment). Out of the 15 compounds known to induce PLD (Table 3), all of them showed potential to induce PLD at low concentrations in the range 4–16µM except erythromycin which was positive only at 100µM. Thus, for the positive controls, we reproduced with our system the data for 93.3% of the compounds (14/15). Chemicals such as ANIT, sotalol, and tetracycline are likely to induce PLD (although it is not mentioned in the literature that such compounds do so). Our data clearly indicate that these three compounds (ANIT, tetracycline, and sotalol) have the potential to induce PLD at 50µM (Table 3). All negative controls (six compounds) were also confirmed in the range 50–800µM (Table 3). Finally, there were two remaining compounds (BNF and isoniazid) for which no PLD data or other relevant information was available due to lack of information in the literature (Table 1). Our results clearly show that BNF has the potential to induce PLD at 12.5 and 25µM and that isoniazid was identified as negative up to 800µM (Table 3). The most potent compounds identified with our assay are amitriptyline, chlorpromazine, clomipramine, clozapine, fluoxetine, ketaconazole, perhexiline, thioridazine, and tamoxiphen, as PI higher than three were obtained at 16 or 25µM (except for thioridazine; PI = 2.58 but at 8µM) (Table 3). The most potent drug was perhexiline with a PI of 11.21 at 16µM (Table 3).
Correlation between experiments 1 and 2 and reproducibility of data.
Based on the PI values, the coefficient of correlation for the 17 compounds tested in experiments 1 and 2 at the same concentration was of 70.2 %. Nevertheless, the coefficient reached 87.7% when the clozapine data was discarded (correlation for 16 compounds) (Fig. 1). Indeed, for unknown reasons the PI induced by clozapine in the second experiment was much higher than in the first experiment (i.e., 5.13 vs. 2.15 at 25µM). Overall, this shows that similar results were obtained in both experiments.
|
" corresponds to clozapine which is identified as an outlier. The symbol "•" corresponds to the other 16 compounds. The equation of the line is: y = 0.87x + 0.33 with a coefficient of correlation of 87.7% (r2 = 0.877) when clozapine is excluded from the analysis.To further evaluate the reproducibility of the assay, HepG2 cells were exposed to 12.5, 25, and 50µM amitriptyline in six different experiments (Table 4). Based on the PI values, the coefficient of correlation among the experiments was of 98.5 ± 1.6 % (average ± SD) which indicates again that comparable results were obtained when experiments are performed at different times. Nevertheless, Table 4 also revealed that the level of response of certain genes can be quite different. For instance, the P8 gene was induced approximately 35, 13, 16, 71, 16, and 19 times in experiments A, B, C, D, E, and F, respectively (Table 4).
|
Justification of the rules for the classification of compounds.
Figure 2 displays the PI values obtained in experiment 2 with the positive and negative drugs. These compounds have been shown to induce the development of concentric lamellar bodies or not in HepG2 cells after the 3-day treatment (Sawada et al. 2005
). The compounds which are likely to induce PLD or those for which no information is available have not been included in Figure 2. All the negative compounds tested had a PI below than 1.25; the highest PI value was for valproic acid at 800µM (PI = 1.18). On the other hand, all the positive compounds had PI values higher than 1.25 at least at one concentration and at three successive concentrations for the majority of the compounds. Figure 2 also shows that clear dose-response relationships were observed for the positive compounds.
|
Based on our data (first and second experiments) and on the data by Sawada et al. (2005
; see also the "Discussion" section), we consider that a compound can potentially induce PLD when (1) PI is higher than 1.25 at a maximum concentration of 50µM and when (2) a dose-response relationship is obtained. |
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
The present study was undertaken to determine if a set of 17 genes, identified as biomarkers of PLD by Sawada et al. (2005)
, could be used to detect a potential for drug-induced PLD in HepG2 cells. In addition, the objective was to test the effect of a range of concentrations for the validation of the assay and to select the most suitable concentrations to use for screening activities. Out of the 17 genes, 12 genes (11 biomarkers of PLD plus 1 housekeeping gene) were selected for practical screening reasons as these 12 genes can be measured in two 96-well plate formats. These genes were selected because they responded well to the compounds known to induce PLD. In addition, the expression of the selected genes either did not vary much or did not follow the trend induced by the positive compounds when the HepG2 cells were exposed to compounds not inducing PLD. Some of the genes were discarded because they responded well to only a limited number of compounds or due to reproducibility problems. Nevertheless, some of the genes (such as LSS and FABP1 ranked on position 12 and 13, respectively) that we discarded after experiment 1 could potentially be good biomarkers of PLD. For instance, LSS with a score of 24 seems to be as good as SERPINA3 (position 11 with a score of 25) according to experiment 1 [Table 2]). Out of the 17 candidate genes, Sawada et al. (2005) selected 12 genes which showed significant concordance with lamellar myelin-like body formation. The first experiment allowed us to select the best 11 marker genes of PLD (out of the 17 genes) and seven of these markers (i.e., C10orf10, FRCP1, SERPINA3, NRB02, MGC4171, TAGLN, and FLJ10055) were also selected by Sawada et al. (2005). To help the interpretation of the data, we calculated the PI which measures a potential to induce PLD and takes into account the 11 selected genes at a global level. Although the selection of genes made by Sawada et al. (2005) and us were not exactly the same, we found a coefficient of correlation of 75% when PI obtained in our first study was compared to the PI calculated with the data of Sawada et al. (2005). Thus, this shows that similar results were obtained in both experiments.
Out of the 17 positive compounds (which induced the formation of lamellar myelin-like bodies in HepG2 cells after 72 h treatment) screened by Sawada et al. (2005), all had a PI value (called PLD mRNA score in the paper by Sawada et al.) higher than 1.5. Out of the 13 negative compounds, 10 had a PI value lower than 1.25; the three other PI values were close to 1.5 (Sawada et al., 2005). It is noteworthy that there was a significant correlation between the PI values and the pathological scores obtained from electron microscopic analysis of the HepG2 cells (Sawada et al. 2005). Thus, the higher the PI value the more likely that the formation of lamellar myelin-like bodies in HepG2 cells is severe. In our study, the six negative compounds tested had a PI below 1.25. On the other hand, all the positive compounds had PI values higher than 1.25 at three successive concentrations for the majority of the compounds, and dose-response relationships were observed. Based on our data (first and second experiments) and on Sawada et al. (2005) data, we consider that a compound can potentially induce PLD when (1) PI is higher than 1.25 at a maximum concentration of 50µM and when (2) a dose-response relationship is obtained (see also below for more explanations). Following this rule, in the first experiment, 91.6% of the compounds known to induce PLD (i.e., 11 of 12) were identified as such. Tamoxiphen, a compound known to induce PLD, was classified as negative (in experiment 1 as PI was lower than 1.25 at 8.3µM. However, in the second experiment, tamoxiphen obtained a PI of 3.37 at 16µM which clearly indicates that this compound is positive in our assay. Such case clearly points out the necessity to test a range of concentrations when the potential to induce PLD is evaluated. In the first experiment, all the negative drugs (i.e., five of five) were also identified as such at 25µM. Nevertheless, it is noteworthy that compounds such as flecainide and erythromycin are not true negative compounds as they clearly induce PLD but at higher concentrations (Table 1). They were considered as negative because, at the concentrations tested, flecainide and erythromycin did not induce the formation of lamellar myelin-like bodies in lysosomes scored by electron microscopy after 72 h of treatment in HepG2 cells (Sawada et al. 2005).
In the second experiment, the objective was to test the effect of a range of concentrations but also to further validate the assay by using more compounds (26 compounds [all 17 compounds from the first experiment plus 9 new drugs]). 93.3% of the compounds known to induce PLD (i.e., 14 of 15) were identified as such. A dose-response relationship was also observed in most cases. Erythromycin, a compound known to induce PLD, showed a potential to induce PLD at high concentration (100µM). Thus, erythromycin was classified as a negative drug (according to our rules; see above) may be because the compound was badly absorbed as it is quite hydrophilic and/or because the HepG2 cell has low metabolic capacities (considering that the metabolites could induce PLD and not the parent compound). Chemicals such as ANIT, sotalol, and tetracycline are likely to induce PLD even though no evidence of this has been published. ANIT induced vacuole formation in mice liver, and high levels of phospholipids were measured in the blood (Chisholm et al., 1999); tetracycline can bind to phospholipids (Argast and Beck, 1984) and gene expression data from sotalol-exposed HepG2 cells suggest that this drug has the potential to induce PLD (Sawada et al. 2005). Our data clearly indicate that these three compounds have the potential to induce PLD at 25 or 50µM (Table 3). Our results also show that BNF has the potential to induce PLD from 12.5µM and that isoniazid was identified as negative up to 800µM. Nevertheless, the potential of BNF and isoniazid to induce PLD is unknown according to the literature data. Finally, all negative controls (six compounds), were also confirmed in the range 50–800µM. When both experiments 1 and 2 were compared, the coefficient of correlation for 16 compounds tested at the same concentrations reached 87.7% (based on PI values). In addition, an excellent correlation (98.5 ± 1.6%; average ± SD) was obtained when the HepG2 cells were exposed to 12.5, 25, and 50µM of amitriptyline in six different experiments (based on PI values). This shows that similar results are obtained when the same experiment is performed and thus that the endpoints measured are reliable. Thus, overall the present study further validates the use of gene expression in HepG2 cells to detect potential to induce PLD as initially demonstrated by Sawada et al. (2005).
Recently, Kasahara et al. (2006) examined the suitability of various cell types (ARLJ301, HepG2, Human and rat primary hepatocytes, CHO-K1, CHL/IU, and J744A) for detecting PLD based on the binding of a dye to phospholipids. CHO-K1 was considered to be the better cell line to detect PLD with this approach. The HepG2 cell line was reported to be unsuitable to detect PLD because it showed little response to a set of five CADs (Kasahara et al. 2006). This is not in agreement with the present study or the study of Sawada et al. (2005) as the HepG2 cell line was successfully used to detect a potential to induce PLD by a wide range of CADs. A first explanation could be that the methodology applied by Kasahara is not suitable on HepG2 cells, whereas gene expression is. Another explanation could be related to the fact that the HepG2 cell line may be not homogenous among laboratories. Indeed, a study showed that some HepG2 cell lines (used as such) were not always derived from the original HepG2 cells possibly due to cross-contamination during cloning of cell lines (Van Pelt et al. 2003). In addition, the study of six sublines of HepG2 cells (based on growth rates, morphology, plasma protein synthesis, etc) revealed that one subline represented a variant of HepG2 cells with an altered phenotype (Iwasa et al. 1990). There is considerable evidence to indicate that the pharmacokinetics and metabolism of drugs play an important role in the etiology of drug-induced PLD (Kodavanti and Mehendale, 1990). For instance, drugs that are rapidly metabolized fail to induce PLD (Joshi et al. 1989). Nevertheless, some CADs such as amiodarone produce lipophilic metabolites that have the same affinity for the tissue as the parent drugs and also produce PLD (Young and Mehendale, 1987). Thus, the fact that the HepG2 cell line has low metabolic capacities is an advantage or a drawback depending on the fact that either the parent compounds or the metabolites induce PLD.
Based on the data presented, researchers are encouraged to use a range of minimum three concentrations to screen for PLD for three main reasons. First, this allows studying dose-response relationships. Second, compounds can be easily ranked for their PLD potential based on a range of concentrations. Third, if only a single concentration is used one may draw wrong conclusions. For instance, in our assay, at 8µM, compounds known to induce PLD such as amiodarone and tamoxiphen have a PI of 0.90 and 1.23, respectively (negative assay, Table 3). But at 16µM, the results clearly indicate that these compounds have the potential to induce PLD (PI of 1.93 and 3.37 for amiodarone and tamoxiphen, respectively) (Table 3). For the screening of compounds, we currently use the following concentrations: 12.5, 25, and 50µM. Fifty µM was chosen as the maximum concentration because at 25µM some compounds known to induce PLD such as flecainide and zimelidine were negative or had a PI just higher than 1.25 (Table 3). In addition, compounds likely to induce PLD such as sotalol and tetracycline showed, according to our data, a potential to induce PLD from 50µM. Finally, we do not recommend using much higher concentrations because cytotoxic effects could occur and some negative drugs could induce nonspecific PLD at higher concentrations (although this was never observed with some of the negative compounds up to 800µM).
From a mechanistic point of view, the inhibition of lysosomal enzyme transport and phospholipase activity together with enhanced PLD synthesis could trigger PLD. Increased cholesterol biosynthesis is considered to be an indirect trigger and other transporter genes as well as genes that control the cell cycle may also be indirectly involved (Sawada et al. 2005). Some of the biomarker genes that were validated in this study clearly belong to the categories previously described. For instance, the hypothetical protein MGC4171 is implicated in the inhibition of lysosomal phospholipase activity (Sawada et al. 2005). AP1S1 is involved in the transport of newly synthesized lysosomal enzymes between the trans-golgi network and lysosomes (Zhu et al. 1999). The up- and downregulation of MGC4171 and AP1S1, respectively, is likely to lead to an accumulation of phospholipids in the lysosomes and consequently to PLD. Other genes are also interesting candidates as they may explain directly or indirectly the accumulation of PLDs in the lysosomes. For instance, FLJ10055 is known to interact with phospholipids, SERPINA3 is a plasma protease inhibitor, and NRB02 is implicated in the regulation of cholesterol metabolism.
As electron microscopic detection of PLD is not adapted for high/medium-throughput screening systems in drug discovery, it is essential to develop in vitro methodologies to help in the selection of the best drug candidates. Two general in vitro approaches have been developed so far. A first one consists in measuring the binding of dyes to the phospholipids by flow cytometry or fluorescence microscopy (Casartelli et al. 2003; Gum et al. 2001; Kasahara et al. 2006; Morelli et al. 2006; Ulrich et al. 1991). A second approach is based on the measurement of gene biomarkers of PLD (Sawada et al. 2005). The choice of the methodology for screening activities should be based on sensitivity, rapidity, and easiness to perform the assays. Reasor et al. (2006) were not really in favor of the toxicogenomic approach due to the high cost, bioinformatic challenge, unclear advantages over existing fluorescent methodologies, and lack of a uniform response among target genes. We have chosen to investigate the toxicogenomic approach due to the sensitivity of the methodology and our expertise in the field. The gene expression data clearly suggest that some drugs (e.g., haloperidol and quinidine) have the potential to induce PLD (and this is the case) although electron microscopy evaluation was negative (Sawada et al. 2005). This is because a certain amount of time is required before the manifestation of PLD at the lysosomal level can be detected by electron microscopy. Indeed, changes in gene expression are the first events that will occur in the cells in response to compounds. For instance, Sawada et al. (2005) revealed that gene expression changes occurred already in cells exposed to chemicals after 6 h of treatment. Changes could be even detected at an early stage and it is uncertain whether the current fluorescent methodologies could do so. Nevertheless, based on our data and those published by Kasahara et al. (2006), after sufficient incubation time (e.g., 24 h) it appears that the fluorescent methodologies and toxicogenomic approach are of similar sensitivity. Since a PI is simply calculated based on a set of 11 genes without any further investigations, we do not consider that there is a bioinformatic challenge. This means that the overall response of the gene set is taken into account and, consequently, there is no need to find a single biomarker gene which would respond to all compounds known to induce PLD. Finally, whatever the methodologies chosen, it is clear that the objective should be to rank compounds and that an absolute predictivity is not achievable.
In conclusion, the present study further confirmed the utility of gene expression in HepG2 cells to identify potential to induce PLD. Indeed, the majority of the compounds tested (20 [15 positive and 6 negative compounds] of 21) were correctly classified. The study also showed that reproducible results (correlation of 87.7%) were generated when a set of 16 compounds was tested in two independent experiments. Finally, the present study clearly indicates as well that it is recommended to use a range of concentrations (e.g., 12.5, 25, and 50µM) to screen for the PLD potential of unknown compounds.
|
ACKNOWLEDGMENTS |
|---|
We would like to thank Etienne Hanon (UCB Pharma SA) for the elaboration of an Excel macro for the PI calculations.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Argast M and Beck CF. (1984) Tetracycline diffusion through phospholipid bilayers and binding to phospholipids. Antimicrob. Agents Chemother. 26:263–265.
Bockhardt H and Lullmann-Rauch R. (1980) Zimelidine-induced lipidosis in rats. Acta Pharmacol. Toxicol. (Copenh) 47:45–48.[Medline]
Camus P and Mehendale HM. (1986) Pulmonary sequestration of amiodarone and desethylamiodarone. J. Pharmacol. Exp. Ther. 237:867–873.
Cantor JO, Keller S, Mandl I, Turino GM. (1987) Increased synthesis of elastin in amiodarone-induced pulmonary fibrosis. J. Lab. Clin. Med. 109:480–485.[Web of Science][Medline]
Casartelli A, Bonato M, Cristofori P, Crivellente F, Dal Negro G, Masotto I, Mutinelli C, Valko K, Bonfante V. (2003) A cell-based approach for the early assessment of the phospholipidogenic potential in pharmaceutical research and drug development. Cell Biol. Toxicol. 19:161–176.[CrossRef][Web of Science][Medline]
Chisholm JW, Nation P, Dolphin PJ, Agellon LB. (1999) High plasma cholesterol in drug-induced cholestasis is associated with enhanced hepatic cholesterol synthesis. Am. J. Physiol. 276:1165–1173.
de Longueville F, Atienzar FA, Marcq L, Dufrane S, Evrard S, Wouters L, Leroux F, Bertholet V, Gerin B, Whomsley R, et al. (2003) Use of a low-density microarray for studying gene expression patterns induced by hepatotoxicants on primary cultures of rat hepatocytes. Toxicol. Sci. 75:378–392.
Drenckhahn D, Kleine L, Lullmann-Rauch R. (1976) Lysosomal alterations in cultured macrophages exposed to anorexigenic and psychotropic drugs. Lab. Invest. 35:116–123.[Web of Science][Medline]
Gonzalez-Rothi RJ, Zander DS, Ros PR. (1995) Fluoxetine hydrochloride (Prozac)-induced pulmonary disease. Chest 107:1763–1765.
Gum RJ, Hickman D, Fagerland JA, Heindel MA, Gagne GD, Schmidt JM, Michaelides MR, Davidsen SK, Ulrich RG. (2001) Analysis of two matrix metalloproteinase inhibitors and their metabolites for induction of phospholipidosis in rat and human hepatocytes. Biochem. Pharmacol. 62:1661–1673.[CrossRef][Web of Science][Medline]
Iwasa F, Galbraith RA, Sassa S. (1990) Phenotypic variation in human HepG2 hepatoma cells: alterations in cell growth, plasma protein synthesis and heme pathway enzymes. Int. J. Biochem. 22:303–310.[CrossRef][Web of Science][Medline]
Joshi UM, Rao P, Kodavanti S, Lockard VG, Mehendale HM. (1989) Fluorescence studies on binding of amphiphilic drugs to isolated lamellar bodies: relevance to phospholipidosis. Biochim. Biophys. Acta 1004:309–320.[Medline]
Kasahara T, Tomita K, Murano H, Harada T, Tsubakimoto K, Ogihara T, Ohnishi S, Kakinuma C. (2006) Establishment of an in vitro high-throughput screening assay for detecting phospholipidosis-inducing potential. Toxicol. Sci. 90:133–141.
Kodavanti UP and Mehendale HM. (1990) Cationic amphiphilic drugs and phospholipid storage disorder. Pharmacol. Rev. 42:327–354.[Web of Science][Medline]
Laurent G, Kishore BK, Tulkens PM. (1990) Aminoglycoside-induced renal phospholipidosis and nephrotoxicity. Biochem. Pharmacol. 40:2383–2392.[CrossRef][Web of Science][Medline]
Lullmann-Rauch R. (1974) Lipidosis-like ultrastructural alterations in rat lymph nodes after treatment with tricyclic antidepressants or neuroleptics. Naunyn. Schmiedebergs Arch. Pharmacol. 286:165–179.[CrossRef][Web of Science][Medline]
Lullmann-Rauch R. (1979) Drug-induced lysosomal storage disorders. Font. Biol. 48:49–130.
Lullmann H, Lullmann-Rauch R, Wassermann O. (1978) Lipidosis induced by amphiphilic cationic drugs. Biochem. Pharmacol. 27:1103–1108.[CrossRef][Web of Science][Medline]
Montenez JP, Van Bambeke F, Piret J, Brasseur R, Tulkens PM, Mingeot-Leclercq MP. (1999) Interactions of macrolide antibiotics (Erythromycin A, roxithromycin, erythromycylamine [Dirithromycin], and azithromycin) with phospholipids: Computer-aided conformational analysis and studies on acellular and cell culture models. Toxicol. Appl. Pharmacol. 156:129–140.[CrossRef][Web of Science][Medline]
Morelli JK, Buehrle M, Pognan F, Barone LR, Fieles W, Ciaccio PJ. (2006) Validation of an in vitro screen for phospholipidosis using a high-content biology platform. Cell Biol. Toxicol. 22:15–27.[CrossRef][Web of Science][Medline]
Pakuts AP, Parks RJ, Paul CJ, Bujaki SJ, Mueller RW. (1990) Ketoconazole-induced hepatic lysosomal phospholipidosis: The effect of concurrent barbiturate treatment. Res. Commun. Chem. Pathol. Pharmacol. 67:55–62.[Web of Science][Medline]
Pessayre D, Bichara M, Degott C, Potet F, Benhamou JP, Feldmann G. (1979) Perhexiline maleate-induced cirrhosis. Gastroenterology 76:170–177.[Web of Science][Medline]
Ploemen JP, Kelder J, Hafmans T, van de Sandt H, van Burgsteden JA, Saleminki PJ, van Esch E. (2004) Use of physicochemical calculation of pKa and CLogP to predict phospholipidosis-inducing potential: A case study with structurally related piperazines. Exp. Toxicol. Pathol. 55:347–355.[Web of Science][Medline]
Reasor MJ. (1989) A review of the biology and toxicologic implications of the induction of lysosomal lamellar bodies by drugs. Toxicol. Appl. Pharmacol. 97:47–56.[CrossRef][Web of Science][Medline]
Reasor MJ and Davis ME. (1985) Prevention of chlorphentermine-induced pulmonary phospholipidosis in rats by phenobarbital. Drug Metab. Dispos. 13:192–196.[Abstract]
Reasor MJ, Hastings KL, Ulrich RG. (2006) Drug-induced phospholipidosis: Issues and future directions. Expert Opin. Drug Saf. 5:567–583.[CrossRef][Web of Science][Medline]
Reasor MJ and Kacew S. (2001) Drug-induced phospholipidosis: Are there functional consequences? Exp. Biol. Med. 226:825–830.
Samadian T, Dehpour AR, Amini S, Nouhnejad P. (1993) Inhibition of gentamicin-induced nephrotoxicity by lithium in rat. Histol. Histopathol. 8:39–47.
Sawada H, Takami K, Asahi S. (2005) A toxicogenomic approach to drug-induced phospholipidosis: Analysis of its induction mechanism and establishment of a novel in vitro screening system. Toxicol. Sci. 83:282–292.
Sawada H, Taniguchi K, Takami K. (2006) Improved toxicogenomic screening for drug-induced phospholipidosis using a multiplexed quantitative gene expression Array Plate assay. Toxicol. In Vitro 20:1506–1513.[Web of Science][Medline]
Schneider P, Korolenko TA, Busch U. (1997) A review of drug-induced lysosomal disorders of the liver in man and laboratory animals. Microsc. Res. Tech. 36:253–275.[CrossRef][Web of Science][Medline]
Tang W, Borel AG, Fujimiya T, Abbott FS. (1995) Fluorinated analogues as mechanistic probes in valproic acid hepatotoxicity: Hepatic microvesicular steatosis and glutathione status. Chem. Res. Toxicol. 8:671–682.[CrossRef][Web of Science][Medline]
Ulrich RG, Kilgore KS, Sun EL, Cramer CT, Ginsburg LC. (1991) An in vitro fluorescent assay for the detection of drug-induced cytoplasmic lamellar bodies. Toxicol. Methods 1:89–105.
Van Pelt JF, Decorte R, Yap PS, Fevery J. (2003) Identification of HepG2 variant cell lines by short tandem repeat (STR) analysis. Mol. Cell. Biochem. 243:49–54.[CrossRef][Web of Science][Medline]
Waters E, Wang JH, Redmond HP, Wu QD, Kay E, Bouchier-Hayes D. (2001) Role of taurine in preventing acetaminophen-induced hepatic injury in the rat. Am. J. Physiol. Gastrointest. Liver Physiol. 280:1274–1279.
Whitehouse LW, Menzies A, Mueller R, Pontefract R. (1994) Ketoconazole-induced hepatic phospholipidosis in the mouse and its association with de-N-acetyl ketoconazole. Toxicology 94:81–95.[CrossRef][Web of Science][Medline]
Xia Z, Ying G, Hansson AL, Karlsson H, Xie Y, Bergstrand A, DePierre JW, Nassberger L. (2000) Antidepressant-induced lipidosis with special reference to tricyclic compounds. Prog. Neurobiol. 60:501–512.[CrossRef][Web of Science][Medline]
Yoshida Y, Arimoto K, Sato M, Sakuragawa N, Arima M, Satoyoshi E. (1985) Reduction of acid sphingomyelinase activity in human fibroblasts induced by AY-9944 and other cationic amphiphilic drugs. J. Biochem. (Tokyo) 98:1669–1679.
Young RA and Mehendale HM. (1987) Effect of cytochrome P-450 and flavin-containing monooxygenase modifying factors on the in vitro metabolism of amiodarone by rat and rabbit. Drug Metab. Dispos. 15:511–517.[Abstract]
Zhu Y, Traub LM, Kornfeld S. (1999) High-affinity binding of the AP-1 adaptor complex to trans-golgi network membranes devoid of mannose 6-phosphate receptors. Mol. Biol. Cell 10:537–549.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||