Dissecting drug-induced in vitro cytotoxicity and metabolic dysfunction in conditionally immortalized human proximal tubule cells

Acute kidney injury accounts for 20% of all hospitalized adults, and 14 to 26% is drug-induced, emphasizing the importance of proper nephrotoxicity assessment. The ‘gold standard’ MTT assay is widely used to measure cell viability, but depends on cellular metabolic activity. Consequently, MTT may not be most optimal to assess cytotoxicity, as nephrotoxicity often involves mitochondrial dysfunction. We compared MTT with a direct cell death assay based on a compromised plasma membrane permeability. Mature conditionally immortalized proximal tubule epithelial cells were dose- (0.1-1,000 µM) and time- (0.5, 1, 2, 4, 8 and 24 hours) dependently exposed to a selection of prototypic nephrotoxicants. Dose-dependent reductions in cellular metabolic activity were stronger compared to declines in uorescence-based cell death, most prominently for cisplatin (1.6 ± 2.0% and 68 ± 4% (mean ± SEM), respectively) and chloroacetaldehyde (2.13 ± 0.05% and 61.0 ± 0.8%). Similar, but more pronounced time-dependent effects were observed, particularly for sanguinarine. We show that assessing cellular metabolic activity by MTT provides a composite readout of cellular metabolic activity and cell death. A nuclear staining approach is preferable when assessing nephrotoxicity of metabolically active compounds. We recommend both assays during drug development to discriminate between metabolically active versus non-active compounds.


Introduction
A steadily increasing number of commonly used drugs as well as various experimental compounds have been associated with acute kidney injury (AKI). AKI is a major disease burden, affecting 20 percent of hospitalized adults worldwide 1,2 . Mortality rates are over 50 percent in severe cases and less severe manifestations are associated with chronic kidney disease 3 . Various prospective cohort studies indicated that drug-induced mechanisms explain 14 to 26 percent of AKI cases, emphasizing the importance of renal toxicity of drugs [4][5][6] . Early detection of drug-induced toxicity, and particularly renal toxicity, is also an essential part of lead-optimization in drug development, as it in uences the success of a candidate compound to proceed in the developmental process. However, lowering the high drug attrition rates in clinical drug development phases remains a key challenge for pharmaceutical companies.
Despite the reduction in drug candidate failures due to poor pharmacokinetic pro les, issues regarding e cacy and safety still lead to high attrition rates [7][8][9][10] . This has been con rmed by an inventory of the main reasons for drug attrition from four leading pharmaceutical companies, showing a substantial contribution of safety-related (i.e., non-clinical toxicology) causes 10 . Especially nephrotoxicity is often observed in clinical studies that was not detected earlier during non-clinical development 11 . Hence, improved non-clinical testing of therapeutic candidates for their nephrotoxic potential is warranted to reduce attrition rates in clinical development phases 12 .
Obtaining accurate and reliable results from in vitro cytotoxicity assays is therefore of vital importance. In the 1980s, Mosmann developed a cell viability and proliferation assay, measuring the reductive activity as enzymatic conversion of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-2H-tetrazolium bromide (MTT) into dimethyl sulfoxide (DMSO)-soluble formazan crystals by dehydrogenases, which is nowadays often recognized as the 'golden standard' for cytotoxicity assessment 12,13 . However, the MTT assay determines cellular metabolic activity, as it mainly relies on the activity of mitochondrial dehydrogenases, in particular succinate dehydrogenase 14 . Consequently, this method may not be suitable to distinguish drug-induced metabolic suppression from reduced cell viability. This limitation has been described for various known inhibitors of cellular metabolism in three human breast carcinoma cell lines 12 . Similarly, radiation-induced cellular metabolic hyperactivation biases the ability of the MTT assay to assess cell viability 14 . Moreover, MTT reduction can be in uenced by other enzymes linked with cellular redox status, like glutathione S-transferases, further warranting cautious interpretation of results obtained with MTT using metabolically active compounds 15 . Finally, chemical reduction of MTT has been found in other organelles such as endoplasmic reticulum and may also explain MTT reduction by avonoids and polyphenols 12,16−19 .
Similar limitations of the MTT assay may also be relevant for the assessment of renal toxicity, as mitochondrial dysfunction is described for a vast number of nephrotoxic drugs 20 . Proximal tubule cells, which are most often affected in renal toxicity, are metabolically very active and have a high mitochondrial content 21 . So far, it has not been investigated whether MTT is the most suitable assay to assess the nephrotoxic potential of drugs. Therefore, we compared the MTT assay with the determination of cell viability by a method independent of cell metabolism, based on cell membrane permeability 22 .
We compared both methods in a renal proximal tubule cell model, ciPTEC-OAT1, exposed to a selection of prototypic nephrotoxic drugs in a dose-and time-dependent manner. We demonstrate that, compared to the uorescence-based approach, assessment of cellular metabolic activity by means of MTT does not exclusively assess cellular toxicity, but rather provides a composite readout of cell death and decreased metabolic rate.

Results
MTT assay shows a stronger dose-dependent decrease upon exposure to nephrotoxic compounds compared to uorescence-based cell death assessment Comparison of nephrotoxicity detected by MTT and our uorescence microscopy method was performed using four compounds, which previously have shown to induce nephrotoxic effects associated with mitochondrial dysfunction. These compounds cover a range of different drug classes and include the antineoplastic drug cisplatin, the antiretroviral drug tenofovir, the Na + /K + -ATPase inhibitor sanguinarine, and chloroacetaldehyde, a cytotoxic metabolite of the antineoplastic drug ifosfamide. ciPTEC expressing the organic anion transporter 1 (OAT1) cells were exposed for 24 hours to a dose-range of all four compounds. Cisplatin reduced cellular metabolic activity to 1.6 ± 2.0% whereas uorescence-based cell death only lowered viable cells to 68 ± 4% (100 µM; Fig. 1A), suggesting that it substantially impedes cellular metabolic activity in addition to cytotoxicity (p assay < 0.0001). For tenofovir a reduction to 32 ± 6% and 43 ± 9% in viable cells was observed using MTT and uorescence microscopy, respectively (1,000 µM), which did not signi cantly differ between both assays (p assay = 0.56; Fig. 1B). Sanguinarine reduced cell viability to 1.4 ± 0.1% and 20 ± 10% (10 µM; Fig. 1C) using the MTT or uorescence-based cell death analysis, respectively, with a statistically signi cant difference between both assays (p assay < 0.0001). At increasing chloroacetaldehyde concentrations cell viability signi cantly differed when measured by MTT compared to our uorescence-based cell death assay (p assay < 0.0001; 2.13 ± 0.05% and 61.0 ± 0.8%, respectively; 1,000 µM; Fig. 1D). These observations imply that previously described toxic effects determined using the MTT assay of at least cisplatin, sanguinarine and chloroacetaldehyde, could have been in uenced by compound interference with the metabolizing capacity of the cell. Consequently, it emphasizes the essential difference in readout between both assays described and warrants careful interpretation of results.
Nephrotoxic compounds time-dependently interfere with cellular metabolic activity and uorescence-based cell death Remarkably, no differences between both assays could be observed for the prototypic nephrotoxic drug tenofovir. It has, however, been described that drugs can decrease cellular metabolic activity before cellular viability is compromised 27 . Consequently, we investigated whether at earlier timepoints differences between both methods could show different results for tenofovir and all other compounds. These time-dependent effects were measured after 0.5, 1, 2, 4, 8 and 24 hours. To this end, ciPTEC-OAT1 cells were exposed to IC 70 and IC 90 drug concentrations (calculated from dose-dependent effects on cellular metabolic capacity, see Table 1), as the commonly used IC 50 concentration is likely subjected to variation due to its position in the steepest part of the curve. All prototypic nephrotoxicants showed a signi cant time-dependent reduction in cell viability as measure of cellular metabolic activity and uorescence-based cell death (Fig. 2). For cisplatin and tenofovir ( Fig. 2A and 2B, respectively), this effect was only observed 24 hours after incubation. Also, both assays signi cantly differed 24 hours after incubation for cisplatin (p < 0.01; Fig. 2A) and 8 hours after incubation for tenofovir (p < 0.05; Fig. 2B). The most prominent difference was noted for sanguinarine (Fig. 2C). A signi cant reduction in cell viability was observed 2 hours after incubation (p < 0.0001), whereas a signi cantly lower cell viability as determined by uorescence microscopy occurred only 4 hours after incubation (p < 0.05). Overall, cellular metabolic activity showed a stronger reduction compared to the uorescence-based approach (p assay < 0.0001), indicating that sanguinarine primarily interferes with metabolic capacity, while cell death originates secondary to a reduced metabolic activity. Remarkably, 8 hours after compound exposure, cellular metabolic activity was very low (1.7 ± 0.2%), whereas the nuclear staining approach showed a residual viability of 72 ± 3%, which could suggest activation of other compensatory cellular metabolic pathways that maintain cells viable. Chloroacetaldehyde showed a reduction in cell viability as measure of both cellular metabolic activity and uorescence-based cell death, starting 8 hours after compound incubation (Fig. 2D). After 24 hours of incubation, cellular metabolic activity was reduced to a viability of 29 ± 5% (p < 0.0001), whereas according to the uorescence-based assay, cell death decreased to only 72 ± 9% of control (p < 0.01). Similar, but less prominent effects of compound exposure on cellular metabolic activity and uorescence-based cell death could be observed for exposure to IC 70 -concentrations (Fig. 3), at which all four compounds signi cantly decreased cellular viability. Table 1 Overview of prototypic nephrotoxicants with calculated IC 70 -and IC 90 -concentrations. IC 70 -and IC 90 -concentrations are de ned as inhibitory concentrations that result in 70% and 90% of the maximal inhibitory effect on cellular metabolic activity, respectively. IC 70 and IC 90 concentrations were based on and calculated using IC 50 -values of mean cellular metabolic activity dose-response curve ttings (Fig. 1). 95% con dence bands were plotted in curve tting graphs and used to calculate con dence intervals for IC 70

Discussion
Drug-induced kidney injury is one of the most frequently observed adverse effects and often leads to high attrition rates during clinical drug development stages 1,11 . This warrants the use of proper in vitro strategies to assess potential toxic drug effects early in the development of new drugs. Importantly, many drugs are known to disturb mitochondrial function 28 , which may bias the interpretation of results obtained by commonly used tetrazolium salt-based methods that rely on enzymatic activity, such as the MTT assay. Here, we compared the toxicity of four nephrotoxic metabolically active drugs by the MTT assay and a uorescence-based method, which is not dependent on cellular metabolism. Differences in dose-and time-dependent toxicity patterns were observed between both assays, in which the selected drugs showed a stronger and earlier interference with cellular metabolic capacity.
The observed strong inhibitory effect of cisplatin on cellular metabolic activity has previously been related to apoptosis through mitochondrial pathways, as generation of (mitochondrial) oxidative stress appeared critical in its mechanism of cytotoxicity [29][30][31][32][33] . A similar strong reduction in the MTT assay has been reported before 34 , and a weaker inhibitory effect using our uorescence-based cell death assay is in line with the observed cisplatin-induced mitochondrial dysfunction. For tenofovir, a comparable timedependent effect on cellular metabolic capacity and cytotoxicity was detected. The slightly stronger metabolic response compared to uorescence-based cell death con rms previous observations that tenofovir primarily targets mitochondria (e.g., ultrastructural defects and mtDNA depletion) of the renal proximal tubule [35][36][37][38] . The consequent disruption of mitochondrial biogenesis would explain reduced abundance of mitochondrial enzymes, including dehydrogenases. Sanguinarine also demonstrated a strong reduction in cellular metabolic activity, whereas uorescence-based cell death was less severely affected. Serious toxicity of sanguinarine has been recognized before 39 and time-dependent differences between both assays align with previously observed mitochondrial effects, including depolarization and inhibition of respiration 40,41 . Finally, the strong metabolic effects of the ifosfamide metabolite chloroacetaldehyde is in agreement with previously observed toxicity in rabbit renal proximal tubule cells, in which ATP levels were dose-dependently reduced 42 . Moreover, chloroacetaldehyde seems to inhibit hexokinase 43 , which further explains the strong cellular metabolic effects observed in our study. Overall, our results are in accordance with previously reported cytotoxicity of the drugs studied, and differences between both assays can be explained by effects of the various nephrotoxicants on cellular metabolism.
The observed strong dose-and time-dependent patterns in our study are, however, expected not to be speci c to the MTT assay. Similar tetrazolium-based assays, including XTT, WST-1/8 and CCK-8, all rely on enzymatic conversion of a substrate by metabolic capacity of the cell 13,44 , these are therefore expected to show similar patterns after exposure to metabolically active drugs. Consequently, these enzymatic assays are not preferred in assessing cellular effects of metabolically active compounds, as they provide a composite readout consisting of cytotoxic and metabolic effects. Besides our uorescence-based method, other non-metabolism-dependent approaches, including Annexin-V and TUNEL assays have extensively been applied to examine cell death, also in high-throughput settings 45,46 .
Apoptosis detection by both methods is, however, limited as compared to Yo-Pro-1, which makes the latter a more sensitive approach to identify cytotoxicity in drug development 47,48 .
Our observation that cellular metabolic capacity is time-dependently affected highlights the importance of differentiating between metabolically-dependent and -independent cytotoxicity assays. These observations are especially important for drugs with a metabolic target, as such primary mitochondrial toxicants are expected to decrease mitochondrial dysfunction before cell viability is affected, as we observed for sanguinarine. Interestingly, time-dependent comparison of (nephrotoxic) compounds on cellular toxicity, could provide an ideal experimental setup to differentiate between primary metabolic active drugs and drugs that inhibit cellular metabolism via indirect mechanisms (i.e., secondary mitochondrial toxicants), including inhibitors of other metabolic pathways. Consequently, such a combination of assays may be instrumental to gain insight into the mechanisms underlying renal adverse drug events.
To conclude, methods measuring cellular metabolic activity, like the MTT assay, do not exclusively assess cellular toxicity. They rather provide a composite readout of cell death and decreased metabolic rate, as opposed to cytotoxicity assessed using assays that do not rely on cellular metabolism.
Consequently, the latter cell death assay is suitable for assessing toxicity of metabolically active compounds. A combination of assays would be preferable for initial toxicity screening in drug development to discriminate between metabolically active and non-active drugs. Selection of the proper cell viability assessment methods in early drug development stages or when assessing toxicity of existing drugs is vital, and will contribute to the development of drugs with low nephrotoxic potential and improved safety pro le.

Compounds
Cisplatin and chloroacetaldehyde solution were purchased from Sigma-Aldrich (Zwijndrecht, The Netherlands). Sanguinarine (chloride) was from Cayman Chemical (Ann Arbor, MI, USA) and tenofovir was obtained from Santa Cruz Biotechnologies, Inc. (Santa Cruz, CA, USA).

Statistical analysis
Curve-tting and statistical data analysis were performed using GraphPad prism v5.03 (GraphPad Software Inc., San Diego, CA). Data were normalized to untreated (vehicle) control ciPTEC-OAT1. Results were plotted after background subtraction using nonlinear regression with four parameters and variable slope, constraining the bottom to greater than 0.0 on log-transformed x-values. Statistical signi cance of dose-and time-dependent data was determined by one-way ANOVA, followed by Dunnett's post hoc analysis to correct for multiple comparison, and considered being signi cant when p < 0.05. Differences between assay types (p assay ) were statistically analyzed using two-way ANOVA and corrected for multiple comparison by Bonferroni's post hoc analysis. All data is presented as mean ± SEM of at least three independent experiments (n = 3), performed with ve or six experimental replicates. Figure 1 Comparison of MTT and uorescence-based cell death methods to assess dose-dependent reduction of ciPTEC cellular viability. Cellular metabolic activity (□) and uorescence-based cell death (•) were determined in ciPTEC-OAT1 after 24 hours incubation with a serial √10-dilution of nephrotoxic compounds (A) cisplatin, (B) tenofovir, (C) sanguinarine chloride and (D) chloroacetaldehyde. All results were normalized to unexposed controls. Statistical analyses: one-way ANOVA corrected for multiple comparison by Dunnett's post hoc analysis to compare differences between vehicle control and exposed conditions * p < 0.05, ** p < 0.01, *** p < 0.001; and two-way ANOVA corrected for multiple comparison by Bonferroni's post hoc analysis to compare differences between assays # p < 0.05, ### p < 0.001. P-value corresponding to overall signi cance between assays is indicated for each compound. Mean ± SEM; n=3 independent experiments. Time-dependent effects upon exposure of ciPTEC-OAT1 to IC90-concentrations of nephrotoxic drugs on cellular metabolic activity and uorescence-based cell death. Mature ciPTEC-OAT1 were exposed to IC90concentrations of nephrotoxic drugs for 0.5, 1, 2, 4, 8 or 24 hours, after which uorescence-based cell death (■) or cellular metabolic activity (■) assays were performed. (A) cisplatin, (B) tenofovir, (C) sanguinarine, (D) chloroacetaldehyde. For IC90-values see table 1. All results were normalized to unexposed controls. Statistical analyses: one-way ANOVA corrected for multiple comparison by Dunnett's post hoc analysis to compare differences between vehicle control and exposed conditions * p < 0.05, ** p < 0.01, *** p < 0.001; and two-way ANOVA corrected for multiple comparison by Bonferroni's post hoc analysis to compare differences between assays # p < 0.05, ## p < 0.01, ### p < 0.001. P-value corresponding to overall signi cance between assays is indicated for each compound. Mean ± SEM; n=3 independent experiments.