An in vitro method for the prediction of renal proximal tubular toxicity in humans

Abstract

The kidney is one of the major target organs for drug-induced toxicity. The renal proximal tubule is frequently affected due to its roles in drug transport and in concentrating the glomerular filtrate. Drug-induced kidney injury is associated with increased morbidity and mortality of patients. During drug development, nephrotoxicity is typically detected only late, which leads to high costs for the pharmaceutical industry. A central problem is the lack of pre-clinical models with high predictability. Regulatory accepted or validated in vitro models for the prediction of nephrotoxicity are not available. We developed a novel in vitro model for the prediction of renal proximal tubular toxicity in humans. It employs human primary renal proximal tubular cells and the expression levels of interleukin (IL)-6 and IL-8 were used as the endpoint. The model was evaluated with 41 well-characterized drugs and chemicals. The median values of the major performance metrics (balanced accuracy, sensitivity, specificity, positive predictive value, negative predictive value and area under the curve of the receiver operating characteristic curve) ranged between 0.76 and 0.85. This revealed that the predictability of the model is high and it would be expected that in ∼76%–85% of the cases where compounds were predicted as positives or negatives the predictions would be correct. Altogether, the data suggest that the model would allow the prediction of drug-induced proximal tubular toxicity at early pre-clinical stages during drug development. Future work will aim at further validating this model and adapting it to recently developed renal proximal tubular-like cells derived from human pluripotent stem cells.

---

Introduction

The kidney is one of the major target organs for drug-induced toxicity. Nephrotoxic drugs and chemicals can induce acute kidney injury (AKI), or chronic kidney disease and subsequently end stage renal disease (ESRD).^1–3 AKI and ESRD patients have increased morbidity and mortality and depend on dialysis.^1,4,5 About 5% of all hospitalized patients and ∼20%–30% of ICU patients develop AKI, and ∼20%–25% of these cases are due to nephrotoxic drugs.^2–4 Whenever alternative and new drugs become available, their nephrotoxic potential is often underestimated,⁶ which leads again to clinical complications, as in the case of COX2 inhibitors.⁷

Typically, nephrotoxicity is detected only late during drug development and accounts for 2% of drug attrition during pre-clinical studies and 19% in phase 3.⁸ Also, due to the large functional reserve of the kidney, nephrotoxic effects often become obvious only after regulatory approval. A recent example is tenofovir, which injures the renal proximal tubules.^9,10 Altogether, the problems outlined above are associated with increased risks for patients and subjects enrolled in clinical trials as well as substantial costs for the health care system and the pharmaceutical industry.

One major problem is the lack of pre-clinical models with high predictability. The predictability of animal models is compromised by interspecies variability, and there are other problems such as high costs and low throughput. Further, legislation changes in the EU (REACH and the 7th Amendment of the Cosmetics Directive) and new initiatives in the USA (ToxCast and Tox21) have increased the interest in in vitro models. Regulatory accepted or validated in vitro models for the prediction of nephrotoxicity in humans are currently not available. Major difficulties are related to the identification of appropriate cell types and endpoints.^11–13

In the kidney the cells of the renal proximal tubule (PT) are a major target for drug-induced toxicity due to their roles in glomerular filtrate concentration and the transport of drugs and organic compounds.^2,3 PT-derived cell lines, such as the human and porcine cell lines HK-2 (human kidney-2) and LLC-PK1 (Lewis lung cancer-porcine kidney 1), have been frequently applied in in vitro nephrotoxicology. However, immortalized cells are less sensitive than human primary renal proximal tubular cells (HPTC)¹⁴ and insensitive to well-known nephrotoxicants,¹³ which is due to functional changes and changes in drug transporter expression associated with immortalization.^15–17 Further, endpoints that are associated with general cytotoxicity, such as cell death, metabolic activity or ATP depletion, are not useful in addressing organ-specific toxicity. A recent study measuring ATP-depletion in liver-, kidney PT- and heart-derived cell lines treated with hepatotoxic, nephrotoxic and cardiotoxic compounds found that the majority of compounds had similar effects in all three cell lines.¹⁸

The European and US regulatory agencies in charge of the validation and acceptance of alternative methods (European Centre for the Validation of Alternative Methods (ECVAM) and the National Toxicology Program Interagency Center for the Evaluation of Alternative Toxicological Methods/Interagency Coordinating Committee on the Validation of Alternative Methods (NICEATM/ICCVAM)) are currently not involved in any activities on the validation of methods for in vitro nephrotoxicology. The ECVAM has funded one pre-validation study¹⁹ which used 15 drugs. Other models for in vitro nephrotoxicology that have been developed since then during the last 10 years^20–24 have been tested with limited numbers of drugs and are of unclear predictability. A recently developed high-throughput mitochondrial nephrotoxicant assay is based on rabbit cells,²⁵ which raises issues concerning interspecies variability. This applies also to a model employing PT freshly isolated from murine kidneys.^23,24 Both models would still require the use of animals.

Here, we developed a novel in vitro model for the prediction of human renal PT toxicity. The model employed HPTC and the expression levels of interleukin (IL)-6 and IL-8 were used as the endpoint. The model was evaluated with 41 well-characterized drugs and chemicals. The results revealed that the predictability of this model is high and is in the range of about 76%–85%.

Materials and methods

Test compounds

41 compounds were tested. The nature of these compounds, as well as their classification into different groups, is shown in Table 1. Compounds 3–5, 8, 14, 18–20, 23, 30, 34 and 37 were obtained from Merck (Darmstadt, Germany). Compound 1 was purchased from PAA Laboratories GmbH (Pasching, Austria). Compound 10 was obtained from ChemService (West Chester, PA, USA) and compound 22 was purchased from Tocris Bioscience (Bristol, UK). All other test compounds were purchased from Sigma-Aldrich (St. Louis, MO, USA). Where possible, stock solutions (10 mg ml⁻¹) of the test compounds were prepared with biotechnology grade water (1st Base, Singapore), which applied to the following compounds: 1, 2, 4–6, 9–18, 23, 25, 28, 30–36, and 40. Otherwise, stock solutions (6.8 mg ml⁻¹–100 mg ml⁻¹ depending on the solubility of the individual compound) were prepared with dimethyl sulfoxide (DMSO; Sigma-Aldrich; compounds 3, 7, 8, 19, 22, 27 and 41) or ethanol (compounds 20, 21, 24, 26, 29, 37 and 39). Vehicle controls were performed with the respective solvents. All stock solutions were stored in the dark at 4 °C. Stock solutions of metal oxides and inorganic salts (compounds 11–16 and 18) were stored for up to 6 months. No stock solution of an organic compound was stored for longer than 3 months and most stock solutions were consumed much faster during the comprehensive test series.

Table 1 Test compounds and highest expression levels of IL-6 and IL-8 in HK-2 and LLC-PK1 cells. The table lists the 41 test compounds, which were divided into three groups. Group 1 (compounds 1–22) represents nephrotoxins that directly damage PT. Group 2 (compounds 23–33) comprises nephrotoxins that do not directly damage PT and injure the kidney by different mechanisms. Group 3 (compounds 34–41) represents non-nephrotoxic compounds. HK-2 and LLC-PK1 cells were exposed to these compounds at concentrations ranging from 1 μg ml⁻¹ to 1000 μg ml⁻¹. The table lists the highest expression levels of IL-6 and IL-8 that were observed at any given concentration of a drug within this range. The numbers show the mean fold expression ±s.d. (n = 3) relative to the vehicle control. The highest expression levels shown here in Table 1 are highlighted in the ESI Tables S7–S10, which display in detail the expression levels obtained at all the drug concentrations tested

No.	Compound	HK-2		LLC-PK1
		IL-6	IL-8	IL-6	IL-8
1	Gentamicin	1.2 ± 0.1	1.3 ± 0.5	5.2 ± 1.0	2.7 ± 0.1
2	Tobramycin	1.4 ± 0.2	1.1 ± 0.1	1.2 ± 0.0	2.3 ± 0.2
3	Rifampicin	12.8 ± 1.7	5.1 ± 0.2	1.4 ± 0.2	8.4 ± 0.1
4	Tetracycline	8.6 ± 1.3	18.8 ± 5.3	4.8 ± 0.3	8.9 ± 0.3
5	Puromycin	120.5 ± 26.1	30.5 ± 2.7	313.7 ± 31.4	839.4 ± 305.9
6	Cephalosporin C	2.6 ± 0.3	1.9 ± 0.3	1.1 ± 0.0	1.6 ± 0.2
7	5-Fluorouracil	3.6 ± 0.8	14.3 ± 1.2	1.8 ± 0.1	10.1 ± 0.7
8	Cisplatin	1.7 ± 0.1	2.5 ± 0.1	1.7 ± 0.1	1.2 ± 0.4
9	Ifosfamide	1.2 ± 0.1	1.7 ± 0.1	1.0 ± 0.1	1.7 ± 0.0
10	Paraquat	1.2 ± 0.1	1.4 ± 0.5	6.3 ± 0.3	20.3 ± 2.0
11	Arsenic(III) oxide	5.9 ± 0.5	2.0 ± 1.0	0.9 ± 0.1	6.9 ± 0.3
12	Bismuth(III) oxide	1.3 ± 0.0	3.0 ± 0.0	1.8 ± 0.1	4.3 ± 0.9
13	Cadmium(II) chloride	2.8 ± 0.8	3.7 ± 0.7	12.9 ± 7.2	12.2 ± 4.3
14	Copper(II) chloride	16.1 ± 1.1	1.2 ± 0.1	3.7 ± 0.2	37.0 ± 4.4
15	Germanium(IV) oxide	0.7 ± 0.1	1.5 ± 0.2	2.4 ± 0.0	2.5 ± 0.1
16	Gold(I) chloride	3.5 ± 0.5	1.9 ± 0.4	2.8 ± 0.0	1.6 ± 0.3
17	Lead acetate	2.0 ± 0.4	2.1 ± 0.1	2.5 ± 0.2	16.3 ± 4.3
18	Potassium dichromate	0.6 ± 0.0	0.6 ± 0.1	0.9 ± 0.0	1.4 ± 0.2
19	Tacrolimus	7.8 ± 0.4	2.8 ± 0.1	7.4 ± 1.0	301.3 ± 27.5
20	Cyclosporin A	1.2 ± 0.1	1.6 ± 0.3	2.4 ± 0.6	4.9 ± 1.2
21	Citrinin	14.7 ± 1.5	1.7 ± 0.0	1.7 ± 0.5	2.6 ± 0.3
22	Tenofovir	2.0 ± 0.3	3.9 ± 0.5	1.5 ± 0.1	4.0 ± 0.3
23	Vancomycin	0.8 ± 0.0	1.2 ± 0.2	0.9 ± 0.1	1.2 ± 0.1
24	Phenacetin	0.7 ± 0.1	4.7 ± 0.6	1.3 ± 0.1	16.1 ± 0.4
25	Acetaminophen	1.2 ± 0.2	1.2 ± 0.1	2.3 ± 0.4	1.8 ± 0.3
26	Ibuprofen	0.9 ± 0.5	0.5 ± 0.2	2.1 ± 0.5	2.9 ± 1.0
27	Furosemide	1.0 ± 0.1	8.3 ± 0.7	2.6 ± 0.1	13.7 ± 1.2
28	Lithium chloride	1.6 ± 0.0	1.4 ± 0.1	1.0 ± 0.0	1.0 ± 0.1
29	Lindane	1.5 ± 0.6	0.5 ± 0.0	0.9 ± 0.2	1.2 ± 0.1
30	Ethylene glycol	1.3 ± 0.1	1.6 ± 0.1	1.1 ± 0.1	2.6 ± 0.1
31	Valacyclovir	1.4 ± 0.0	1.7 ± 0.3	1.6 ± 0.3	1.3 ± 0.0
32	Lincomycin	0.9 ± 0.2	1.1 ± 0.2	1.6 ± 0.1	2.1 ± 0.1
33	Ciprofloxacin	1.1 ± 0.1	1.2 ± 0.1	8.2 ± 2.6	9.1 ± 0.6
34	Ribavirin	1.3 ± 0.1	1.3 ± 0.1	1.8 ± 0.0	0.8 ± 0.0
35	Glycine	1.3 ± 0.1	1.5 ± 0.0	0.9 ± 0.1	1.0 ± 0.0
36	Dexamethasone	1.0 ± 0.1	0.8 ± 0.1	0.8 ± 0.0	0.7 ± 0.1
37	Melatonin	0.8 ± 0.0	0.3 ± 0.0	2.4 ± 0.3	2.6 ± 0.2
38	Levodopa (DOPA)	1.5 ± 0.3	1.2 ± 0.1	0.8 ± 0.0	1.1 ± 0.0
39	Triiodothyronine	2.4 ± 0.6	10.2 ± 3.5	7.7 ± 0.6	164.4 ± 4.0
40	Acarbose	1.2 ± 0.1	1.0 ± 0.0	1.1 ± 0.0	1.1 ± 0.0
41	Atorvastatin	35.0 ± 4.3	82.4 ± 8.2	0.9 ± 0.0	11.0 ± 1.0

---

Cell culture

HPTC were purchased from the American Type Culture Collection (ATCC, Manassas, VA, USA; HPTC 1) or were isolated from nephrectomy samples (HPTC 2–4) as described.²⁶ Commercial HPTC (HPTC 1) were used at passage (P) 4 and P 5, and HPTC 2–4 were used at P 3 and P 4. Nephrectomy samples were derived from tumor patients and areas with normal tissue were selected for HPTC isolation after examination by a pathologist. Respective anonymized normal tissue samples were obtained from the Tissue Repository of the National University Health System (NUHS, Singapore). HK-2 and LLC-PK1 cells were purchased from ATCC. The different cell types were cultivated as described in the culture media recommended by the vendors.¹⁴ The culture medium used for HPTC contained 0.5% fetal bovine serum. The Institutional Review Board's approval for the work with human kidney samples (DSRB-E/11/143) and the cell types (NUS-IRB Ref. Code: 09-148E) used has been obtained. All cells had been cryopreserved before use.

To ensure proper cell quality and marker expression patterns, all batches of HPTC were assessed by using quantitative real-time polymerase chain reaction (qPCR) using 31 marker genes (Fig. S1†). The expression of some markers was confirmed at the protein level by immunostaining and immunoblotting (Fig. S2†). For procedures, antibodies and primers (see ref. 27 and 28 and Table S11†).

qPCR

Cells were seeded into 24-well microplates (Nalgene Nunc, Penfield, NY, USA) at a density of 50 000 cells per cm². Cells were cultivated for 72 h and were then treated for 16 h with the test compounds. Total RNA was isolated using NucleoSpin® RNA II (Macherey-Nagel, Düren, Germany) or an RNeasy® Mini kit (Qiagen, Hilden, Germany). cDNA synthesis was performed using the SuperScript® III First Strand Synthesis System (Invitrogen, Carlsbad, CA, USA) and a MyCycler® thermal cycler (Bio-Rad, Hercules, CA, USA). qPCR (up to 40 cycles) was then performed with the 7500 Fast Real-Time PCR System (Applied Biosystems, Carlsbad, CA, USA). Procedures were carried out according to the manufacturers’ instructions with the software included in the device. Sequence Detection Software 7500 Fast version 2.0.5 was used for data analysis. Gene expression levels were determined with the 2 − ΔΔCT method. Primers were designed with Primer Express Software version 3.0. Details of the primers used (purchased from Sigma-Aldrich) are provided in ref. 27 and 28 and ESI Table S11.†

High content screening (HCS)

HPTC and HK-2 cells were seeded into 96-well microplates (Becton Dickinson, Franklin Lakes, NJ, USA) at a density of 50 000 cells per cm² and LLC-PK1 cells were seeded at 16 000 cells per cm². Cells were cultivated for 72 h and were then treated for 16 h with the test compounds. After fixation for 10 min with 3.7% formaldehyde in phosphate-buffered saline, cell nuclei were stained with 4′,6-diamidino-2-phenylindole (Merck) and imaged with the ImageXpress Micro High Content Screening System (Molecular Devices, Sunnyvale, CA, USA). 3 replicates were tested per cell type, drug and concentration. From each of the 3 wells, 9 images were acquired. Cell nuclei were counted on each individual image and from these data the average cell numbers per well were derived. Data acquisition and analysis was performed by MetaXpress 2.0 (Molecular Devices, Sunnyvale, CA, USA).

Data analysis

Microsoft Office Excel 2003 was used for all calculations. Compounds were defined as positive in the in vitro model and predicted as PT-specific nephrotoxins if the increase of expression of at least one of the marker genes (IL-6 or IL-8) was equal to or higher than the threshold value at any of the compound concentrations tested. Threshold values between 0.3 and 4.0 were examined. Standard definitions are illustrated and provided in the ESI Fig. S3.† True positives (TP) were defined as PT-specific nephrotoxins in humans (Table 1, compounds 1–22, group 1) that gave positive results in the in vitro model. True negatives (TN) were defined as non-nephrotoxic compounds (Table 1, group 3, compounds 34–41) or nephrotoxic compounds that do not damage the PT in humans (Table 1, group 2, compounds 23–33) that gave negative results in the in vitro model. The sensitivity was calculated by dividing the number of TP by the total number of PT-specific nephrotoxins (group 1, compounds 1–22). The specificity was calculated by dividing the number of TN by the total number of non-PT-damaging compounds (groups 2 and 3, compounds 23–41). Balanced accuracy was defined as the mean of sensitivity and specificity. The positive predictive value (PPV) was calculated by dividing the number of TP by the total number of positives identified by the in vitro model. The negative predictive value (NPV) was calculated by dividing the number of TN by the total number of negatives identified by the in vitro model. The concordance with clinical data was calculated in the following way: TP + TN/total number of 41 drugs. When percentages were provided the numbers were multiplied by 100%. The receiver operating characteristic (ROC) curves were generated by plotting sensitivity against (1-specificity) at all threshold values ranging from 0.3 to 4.0 (the same threshold values as in Fig. 3).

The unpaired t-test (Microsoft Office Excel 2010) was used for statistics. The normal distribution of the data was confirmed using SigmaStat (3.5) (Systat Software Inc., Chicago, IL, USA).

Results and discussion

Model design and endpoints

We selected HPTC to avoid issues related to animal cells and cell lines. All batches of HPTC were routinely characterized by microscopic examination and by qPCR, which was used to determine the expression levels of 31 different marker genes (see Fig. S1† and the Materials and methods section). For some markers, proper expression at the protein level was confirmed by immunostaining and immunoblotting (Fig. S2†). These analyses ensured a proper and comparable cell phenotype and quality. Cells were cultivated in normal uncoated multi-well plates. Uncoated tissue culture polystyrene sustains HPTC performance better than other materials with or without extracellular matrix coating.^29,30 Cells were seeded at a high density and cultivated for three days before drug treatment to allow the formation of a differentiated epithelium, which was confirmed in control experiments (data not shown) as described before.¹⁴ The state of cell differentiation is of central importance for obtaining cell type-specific responses.

First, we determined suitable endpoints by assessing the expressional behavior of different marker genes related to PT injury. These included the mesenchymal marker vimentin (VIM). Kidney injury molecule-1 (KIM-1) and neutrophil gelatinase-associated lipocalin (NGAL) are up-regulated in the tubular epithelium after injury and are potential novel biomarkers for the early detection of AKI.^31–35 Interleukin (IL)-18, which is up-regulated in the PT epithelium in diseased and injured kidneys, might be a useful biomarker for detecting kidney toxicity.^31,36,37

IL-6 and IL-8 are expressed in PT and PT-derived cells in vivo and in vitro^14,38–41 and play a central role in pro-inflammatory processes, which occur after injury. Different studies demonstrated up-regulation of IL-6 and IL-8 in injured and diseased kidneys.^42–44 It is thought that pro-inflammatory cytokines play a central role in the pathophysiology of AKI, including nephrotoxin-induced AKI.⁴⁵ Further, significant up-regulation of IL-6 after exposure to nephrotoxins has been demonstrated in a kidney culture model employing purified PTs.²⁴

To determine the effects of nephrotoxins on these six marker genes in vitro two different batches of HPTC were treated with high doses of gentamicin and CdCl₂ and expression levels were analyzed by qPCR. All individual results were normalized to the expression levels of glyceraldehyde 3-phosphate dehydrogenase (GAPDH), which were consistent with cell numbers (ESI Fig. S4†). If the expression of a specific marker gene would be suitable as an endpoint the gene should display relatively low expression levels in untreated cells and high levels of induction in response to nephrotoxins. In addition, the gene should be consistently up-regulated in different batches of HPTC and in response to different nephrotoxins. Fig. 1 shows that these criteria were best fulfilled by IL-6 and IL-8. From the marker genes tested IL-8 displayed the highest levels of up-regulation after treatment with nephrotoxins.

Fig. 1 Marker gene expression in response to nephrotoxins. Two different batches of HPTC (1 and 4) derived from different donors were treated with 2.5 mg ml⁻¹ gentamicin (light grey bars) and 10 μg ml⁻¹ CdCl₂ (dark grey bars; vehicle control: white bars). High doses of these nephrotoxins were applied in order to exclude that a lack of response was due to a dosing problem. The relative expression levels of the marker genes indicated on the x-axis were determined by qPCR. The bars show the mean fold expression compared to the vehicle control ±standard deviation (s.d.; n = 3). The means of the vehicle controls from each experiment were set to 1. Significant differences (P < 0.05) in comparison to the vehicle control are indicated by asterisks.

Also NGAL showed consistent up-regulation, but the levels of up-regulation ranged only between 1.8-fold and 3.5-fold and were lower than the levels of up-regulation of IL-6 and IL-8. VIM was up-regulated in only one cell batch. KIM-1 and IL-18 were up-regulated in response to only one compound in one cell batch. The observed low level or lack of up-regulation of NGAL (Lipocalin-2) and KIM-1, respectively, was consistent with the results of other in vitro models employing human renal proximal tubular cells and a PT-derived cell line (PREDICT-IV, 3rd and 4th Project Periodic Reports, June 30, 2012; http://www.predict-iv.toxi.uni-wuerzburg.de/periodic_reports/4th_annual_report/). It should be noted that primary cells, which are obtained by disruption of the organ, always display some degree of injury response. We observed that VIM, NGAL and KIM-1 were already expressed at relatively high levels in untreated control cells (Fig. S1†), which is consistent with previous results in the case of VIM.⁴⁶ This might explain the lack of or only moderate up-regulation after treatment with nephrotoxins. From all marker genes tested, the expression levels of IL-6 and IL-8 were the lowest in untreated HPTC and were consistently <0.1% of GAPDH expression.

Predictive performance analyzed with 41 compounds

Next, we determined the response to 41 well-characterized drugs and chemicals. Most of the 41 compounds used here were drugs that are routinely and widely applied in clinical practice. Some compounds, like CdCl₂ or lindane, are well-characterized environmental toxins and for all of the compounds, a wealth of human and animal in vivo and in vitro data is available. As a starting point we selected compounds from published lists^2,3,18,19 that classify compounds with regard to their nephrotoxicity in humans and their effects on various parts of the kidney and nephron. We then made an extensive literature search (PubMed) and also used Google and the ChemIDplus Advanced database (focusing on human clinical data; http://chem.sis.nlm.nih.gov/chemidplus/ProxyServlet?objectHandle=DBMaint&actionHandle=default&nextPage=jsp/chemidheavy/ResultScreen.jsp&ROW_NUM=0&TXTSUPERLISTID=0015663271) to get further information on each selected compound and to confirm its classification.

22 compounds were classified as nephrotoxins that are known to directly damage the PT (group 1, Table 1, compounds 1–22; some of these drugs have also different negative effects on the kidney in addition to PT-specific injury). Further, the 41 compounds included 11 nephrotoxins that do not directly damage the PT and have other effects on the kidney (group 2, Table 1, compounds 23–33). In addition, 8 non-nephrotoxic drugs were included (group 3, Table 1, compounds 34–41). The assay was performed with three batches of HPTC derived from different donors and endpoints were the expressions of IL-6 and IL-8 determined by qPCR. For comparison, all experiments were also performed with HK-2 and LLC-PK1 cells (Table 1).

Drug exposure was performed for 16 hours after cultivating the cells for 3 days at confluent density. Initially, we tested a wide range of concentrations covering 5 orders of magnitude and ranging from 0.01 μg ml⁻¹ to 1000 μg ml⁻¹. As usually no drug-induced changes (in comparison to the controls) were observed at the 2 lowest concentrations, we narrowed the range down and concentrations of 1 μg ml⁻¹, 10 μg ml⁻¹, 100 μg ml⁻¹ and 1000 μg ml⁻¹ were tested in all cases. Thus, the widest useful range of concentrations was applied in all experiments, with a lack of drug-induced changes at concentrations below the lower limit and compromised solubility of many compounds at concentrations exceeding the upper limit. All results were normalized to the vehicle control and expressed as fold change of IL-6 and IL-8 expression.

Dose–response curves obtained with HPTC and three drugs selected from each group are shown in Fig. 2. Detailed results on IL-6 and IL-8 expression levels for each cell batch/type and each drug at every concentration tested are listed in the ESI Tables S1–S10. The highest levels of IL-6 and IL-8 expression determined for each drug and cell batch/type at any given concentration of a drug within the range tested are highlighted in the ESI Tables S1–S10. These highest expression levels are summarized in Tables 1 and 2. The results showed that different drugs had different effects on the expression of IL-6 and IL-8, and some drugs induced both marker genes, while other drugs induced only one or none of the marker genes (Fig. 2). Expression of at least one marker gene was often substantially increased after exposure to PT-specific nephrotoxins (group 1, Tables 1 and 2, Fig. 2), whereas gene expression levels typically remained low at all concentrations after exposure to drugs from groups 2 and 3 (Tables 1 and 2, Fig. 2). Overall comparable results were obtained when secretion of IL-6 and IL-8 was assessed by using ELISA (data not shown). However, many of the drugs used inhibited protein synthesis and therefore qPCR data were more reliable.

Fig. 2 Dose–response curves. HPTC 1 were exposed to PT-specific nephrotoxins (group 1, left-hand panels), nephrotoxins that are not toxic for the proximal tubule (group 2, middle) or non-nephrotoxic compounds (group 3, right-hand panels) at the concentrations indicated on the x-axis (note the logarithmic scale). The figure shows the expression levels of IL-6 (grey graphs) and IL-8 (black graphs) relative to the expression levels of the vehicle control (mean ± s.d.). In the case of cisplatin, massive cell death occurred at the highest concentration tested.

Table 2 Highest expression levels of IL-6 and IL-8 in HPTC. Three different batches of HPTC derived from different donors were exposed to the 41 test compounds at concentrations ranging from 1 μg ml⁻¹ to 1000 μg ml⁻¹. The table lists the highest expression levels of IL-6 and IL-8 that were observed at any given concentration of a drug within this range. The numbers show the mean fold expression ±s.d. (n = 3) relative to the vehicle control. The highest expression levels shown here in Table 2 are highlighted in the ESI Tables S1–S6, which display in detail the expression levels obtained at all the drug concentrations tested

No.	HPTC 1		HPTC 2		HPTC 3
	IL-6	IL-8	IL-6	IL-8	IL-6	IL-8
1	16.9 ± 0.2	8.6 ± 1.3	1.7 ± 0.1	2.0 ± 0.1	1.6 ± 0.1	1.9 ± 0.3
2	8.0 ± 0.9	8.9 ± 1.2	1.4 ± 0.0	1.5 ± 0.1	1.4 ± 0.1	2.0 ± 0.1
3	38.9 ± 3.0	110.8 ± 39.0	3.6 ± 0.7	13.1 ± 3.0	5.3 ± 0.4	7.3 ± 2.7
4	6.3 ± 0.7	1.5 ± 0.1	2.3 ± 0.3	3.6 ± 0.8	23.6 ± 2.8	35.5 ± 4.4
5	8.5 ± 1.0	9.5 ± 0.6	79.5 ± 1.7	146.1 ± 3.1	80.4 ± 2.7	46.4 ± 1.2
6	3.9 ± 0.1	2.3 ± 0.2	3.6 ± 0.3	9.2 ± 0.2	1.4 ± 0.1	2.4 ± 0.6
7	3.6 ± 0.8	2.4 ± 0.2	4.8 ± 0.6	3.8 ± 0.9	12.0 ± 1.8	5.4 ± 0.4
8	3.8 ± 0.6	20.1 ± 2.2	27.9 ± 0.8	32.7 ± 0.9	12.1 ± 0.4	21.9 ± 0.8
9	1.7 ± 0.2	1.5 ± 0.2	1.3 ± 0.1	1.8 ± 0.4	2.9 ± 0.3	2.6 ± 0.2
10	3.1 ± 0.7	16.3 ± 4.6	7.5 ± 0.0	13.2 ± 0.3	11.8 ± 0.5	9.7 ± 0.2
11	3.0 ± 0.1	6.0 ± 0.1	1.6 ± 0.0	25.7 ± 3.8	31.4 ± 12.8	22.8 ± 4.0
12	1.1 ± 0.3	11.9 ± 6.2	1.9 ± 0.3	22.5 ± 2.0	4.0 ± 0.1	9.6 ± 1.8
13	6.6 ± 0.5	9.7 ± 2.5	12.9 ± 0.6	165.2 ± 14.7	10.8 ± 1.2	13.3 ± 1.4
14	10.4 ± 2.9	7.5 ± 4.4	12.4 ± 1.4	119.0 ± 5.4	0.9 ± 0.2	9.9 ± 1.1
15	1.7 ± 0.5	2.9 ± 1.7	3.6 ± 0.2	5.7 ± 0.2	2.2 ± 0.2	3.4 ± 0.5
16	10.0 ± 1.3	2.2 ± 0.7	13.6 ± 1.0	8.3 ± 0.5	5.8 ± 2.5	3.5 ± 1.0
17	3.5 ± 0.4	3.3 ± 0.1	8.6 ± 0.4	23.8 ± 2.9	2.3 ± 0.2	2.8 ± 0.4
18	1.7 ± 0.2	3.8 ± 1.0	0.8 ± 0.0	2.7 ± 0.3	0.6 ± 0.2	10.1 ± 0.9
19	24.6 ± 5.7	3.9 ± 1.3	37.6 ± 0.7	29.3 ± 4.3	38.1 ± 1.2	25.0 ± 0.8
20	1.1 ± 0.2	5.6 ± 1.3	3.4 ± 0.7	67.4 ± 3.1	1.3 ± 0.1	17.8 ± 1.6
21	3.9 ± 0.7	2.3 ± 0.7	1.3 ± 0.1	4.9 ± 0.1	2.3 ± 0.3	1.5 ± 0.1
22	1.6 ± 0.1	4.3 ± 0.4	0.7 ± 0.0	1.0 ± 0.0	1.1 ± 0.1	2.0 ± 0.2
23	1.4 ± 0.2	1.6 ± 0.3	1.4 ± 0.2	1.8 ± 0.4	1.0 ± 0.4	0.8 ± 0.2
24	1.2 ± 0.1	1.3 ± 0.1	0.8 ± 0.1	0.9 ± 0.2	6.2 ± 1.0	3.6 ± 0.3
25	1.3 ± 0.0	1.5 ± 0.1	2.9 ± 0.2	2.6 ± 0.1	0.8 ± 0.1	0.9 ± 0.0
26	2.6 ± 0.8	3.1 ± 1.4	16.1 ± 1.2	4.3 ± 0.3	2.9 ± 0.1	1.5 ± 0.3
27	1.7 ± 0.4	1.9 ± 0.5	2.6 ± 0.2	2.6 ± 0.2	9.0 ± 0.6	8.2 ± 0.7
28	0.7 ± 0.1	0.5 ± 0.1	17.4 ± 0.8	20.5 ± 0.8	1.1 ± 0.2	1.3 ± 0.3
29	1.0 ± 0.3	1.1 ± 0.4	0.8 ± 0.0	1.5 ± 0.1	0.9 ± 0.2	1.2 ± 0.0
30	1.0 ± 0.1	1.1 ± 0.1	1.8 ± 0.1	2.0 ± 0.1	1.0 ± 0.2	1.2 ± 0.3
31	1.7 ± 0.4	1.3 ± 0.1	1.6 ± 0.3	1.2 ± 0.2	1.4 ± 0.2	1.3 ± 0.2
32	1.3 ± 0.1	1.2 ± 0.0	1.3 ± 0.1	1.4 ± 0.1	1.0 ± 0.2	1.1 ± 0.2
33	4.7 ± 0.6	5.3 ± 1.1	1.3 ± 0.1	1.6 ± 0.2	2.9 ± 0.1	3.2 ± 0.3
34	1.5 ± 0.1	1.4 ± 0.1	1.2 ± 0.1	1.4 ± 0.1	1.2 ± 0.3	1.7 ± 0.4
35	1.7 ± 0.2	2.0 ± 0.1	2.7 ± 0.8	1.3 ± 0.0	1.0 ± 0.1	1.2 ± 0.1
36	0.9 ± 0.1	1.1 ± 0.1	2.7 ± 0.1	2.9 ± 0.0	1.4 ± 0.2	1.7 ± 0.3
37	1.2 ± 0.1	1.5 ± 0.3	1.1 ± 0.3	1.4 ± 0.2	0.9 ± 0.1	1.3 ± 0.1
38	1.8 ± 0.0	1.3 ± 0.1	1.2 ± 0.2	3.1 ± 0.2	0.6 ± 0.1	1.3 ± 0.2
39	3.4 ± 1.0	1.8 ± 0.1	3.1 ± 0.3	24.0 ± 3.4	3.4 ± 0.3	5.1 ± 0.3
40	1.6 ± 0.2	1.5 ± 0.1	1.2 ± 0.0	1.3 ± 0.0	1.3 ± 0.1	1.3 ± 0.1
41	1.0 ± 0.3	3.6 ± 1.2	1.0 ± 0.0	1.6 ± 0.1	128.5 ± 21.1	38.2 ± 7.1

---

In order to classify a result obtained with a specific drug and cell type/batch as positive or negative it was determined whether the highest expression level (mean; Tables 1 and 2) was equal to or higher than a threshold value. A drug was classified as positive and predicted as PT-specific nephrotoxin if the highest increase in gene expression (Tables 1 and 2) of at least one of the markers (IL-6 and IL-8) was equal to or higher than the threshold value. As it was unclear which threshold value might be most appropriate, this analysis was performed for a range of threshold levels from 0.3 to 4.0.

Examples that illustrate the processing of the data are shown in Tables 3 and 4. Both tables display the same data set obtained with HPTC 1. This data set is identical with the HPTC 1 data set in Table 2 and shows the highest expression levels of IL-6 and IL-8. A threshold of 2.0 (Table 3) or 3.5 (Table 4), respectively, was applied to this data set, and all values that were equal to or higher than the threshold were highlighted. If the expression level of at least one of the marker genes was equal to or higher than the threshold the test result was classified as positive, and the classification is indicated in Tables 3 and 4. Based on these results, the sensitivity (the number of positive test results from group 1 compounds (TP)/total number of 22 group 1 compounds) and the specificity (number of negative test results from group 2 and 3 compounds (TN)/total number of 19 group 2 and 3 compounds) were calculated.

Table 3 Example for the thresholding procedure, determination of positive and negative results and calculation of sensitivity and specificity. The HPTC 1 data set shown here is identical with the respective data set in Table 2. A threshold of 2.0 was applied. All IL-6 and IL-8 expression values that were equal to or above this threshold level are highlighted (bold). If for a given drug the expression value of at least one marker gene was equal to or above the threshold (bold) a result was classified as positive (+). If for a given drug the expression levels of both marker genes were below the threshold the result was classified as negative (−). Based on these positive and negative results the sensitivity and specificity were calculated. Sensitivity was defined as the number of positive group 1 drugs (TP) divided by the total number of 22 group 1 drugs. Specificity was defined as the number of negative group 2 and 3 drugs (TN) divided by the total number of 19 group 2 and 3 drugs

	HPTC 1		Threshold = 2.0
	IL-6	IL-8	+/− Test	Sensitivity/Specificity
1	16.9 ± 0.2	8.6 ± 1.3	+	Sensitivity = 21/22 = 96% positive
2	8.0 ± 0.9	8.9 ± 1.2	+
3	38.9 ± 3.0	110.8 ± 39.0	+
4	6.3 ± 0.7	1.5 ± 0.1	+
5	8.5 ± 1.0	9.5 ± 0.6	+
6	3.9 ± 0.1	2.3 ± 0.2	+
7	3.6 ± 0.8	2.4 ± 0.2	+
8	3.8 ± 0.6	20.1 ± 2.2	+
9	1.7 ± 0.2	1.5 ± 0.2	−
10	3.1 ± 0.7	16.3 ± 4.6	+
11	3.0 ± 0.1	6.0 ± 0.1	+
12	1.1 ± 0.3	11.9 ± 6.2	+
13	6.6 ± 0.5	9.7 ± 2.5	+
14	10.4 ± 2.9	7.5 ± 4.4	+
15	1.7 ± 0.5	2.9 ± 1.7	+
16	10.0 ± 1.3	2.2 ± 0.7	+
17	3.5 ± 0.4	3.3 ± 0.1	+
18	1.7 ± 0.2	3.8 ± 1.0	+
19	24.6 ± 5.7	3.9 ± 1.3	+
20	1.1 ± 0.2	5.6 ± 1.3	+
21	3.9 ± 0.7	2.3 ± 0.7	+
22	1.6 ± 0.1	4.3 ± 0.4	+
23	1.4 ± 0.2	1.6 ± 0.3	−	Specificity = 14/19 = 74% negative
24	1.2 ± 0.1	1.3 ± 0.1	−
25	1.3 ± 0.0	1.5 ± 0.1	−
26	2.6 ± 0.8	3.1 ± 1.4	+
27	1.7 ± 0.4	1.9 ± 0.5	−
28	0.7 ± 0.1	0.5 ± 0.1	−
29	1.0 ± 0.3	1.1 ± 0.4	−
30	1.0 ± 0.1	1.1 ± 0.1	−
31	1.7 ± 0.4	1.3 ± 0.1	−
32	1.3 ± 0.1	1.2 ± 0.0	−
33	4.7 ± 0.6	5.3 ± 1.1	+
34	1.5 ± 0.1	1.4 ± 0.1	−
35	1.7 ± 0.2	2.0 ± 0.1	+
36	0.9 ± 0.1	1.1 ± 0.1	−
37	1.2 ± 0.1	1.5 ± 0.3	−
38	1.8 ± 0.0	1.3 ± 0.1	−
39	3.4 ± 1.0	1.8 ± 0.1	+
40	1.6 ± 0.2	1.5 ± 0.1	−
41	1.0 ± 0.3	3.6 ± 1.2	+

---

Table 4 Example for the thresholding procedure, determination of positive and negative results and calculation of sensitivity and specificity. The HPTC 1 data set is identical with the data set shown in Table 3. Here in Table 4 a different threshold level of 3.5 was applied to this data set. For detailed explanations see the legend of Table 3

	HPTC 1		Threshold = 3.5
	IL-6	IL-8	+/− Test	Sensitivity/Specificity
1	16.9 ± 0.2	8.6 ± 1.3	+	Sensitivity = 20/22 = 91% positive
2	8.0 ± 0.9	8.9 ± 1.2	+
3	38.9 ± 3.0	110.8 ± 39.0	+
4	6.3 ± 0.7	1.5 ± 0.1	+
5	8.5 ± 1.0	9.5 ± 0.6	+
6	3.9 ± 0.1	2.3 ± 0.2	+
7	3.6 ± 0.8	2.4 ± 0.2	+
8	3.8 ± 0.6	20.1 ± 2.2	+
9	1.7 ± 0.2	1.5 ± 0.2	−
10	3.1 ± 0.7	16.3 ± 4.6	+
11	3.0 ± 0.1	6.0 ± 0.1	+
12	1.1 ± 0.3	11.9 ± 6.2	+
13	6.6 ± 0.5	9.7 ± 2.5	+
14	10.4 ± 2.9	7.5 ± 4.4	+
15	1.7 ± 0.5	2.9 ± 1.7	−
16	10.0 ± 1.3	2.2 ± 0.7	+
17	3.5 ± 0.4	3.3 ± 0.1	+
18	1.7 ± 0.2	3.8 ± 1.0	+
19	24.6 ± 5.7	3.9 ± 1.3	+
20	1.1 ± 0.2	5.6 ± 1.3	+
21	3.9 ± 0.7	2.3 ± 0.7	+
22	1.6 ± 0.1	4.3 ± 0.4	+
23	1.4 ± 0.2	1.6 ± 0.3	−	Specificity = 17 /19 = 90% negative
24	1.2 ± 0.1	1.3 ± 0.1	−
25	1.3 ± 0.0	1.5 ± 0.1	−
26	2.6 ± 0.8	3.1 ± 1.4	−
27	1.7 ± 0.4	1.9 ± 0.5	−
28	0.7 ± 0.1	0.5 ± 0.1	−
29	1.0 ± 0.3	1.1 ± 0.4	−
30	1.0 ± 0.1	1.1 ± 0.1	−
31	1.7 ± 0.4	1.3 ± 0.1	−
32	1.3 ± 0.1	1.2 ± 0.0	−
33	4.7 ± 0.6	5.3 ± 1.1	+
34	1.5 ± 0.1	1.4 ± 0.1	−
35	1.7 ± 0.2	2.0 ± 0.1	−
36	0.9 ± 0.1	1.1 ± 0.1	−
37	1.2 ± 0.1	1.5 ± 0.3	−
38	1.8 ± 0.0	1.3 ± 0.1	−
39	3.4 ± 1.0	1.8 ± 0.1	−
40	1.6 ± 0.2	1.5 ± 0.1	−
41	1.0 ± 0.3	3.6 ± 1.2	+

---

In this way, the sensitivity and specificity were calculated for every cell type and batch at 7 different threshold levels. The results are summarized in Table 5 and are graphically displayed in Fig. 3. In addition, Fig. 3 shows the overall concordance with clinical data.

Fig. 3 Sensitivity, specificity and overall concordance with clinical data. The figure displays graphically the percentages for sensitivity and specificity shown in Table 5. In addition, the figure shows the overall concordance with clinical data. Calculations were performed separately for the three different batches of HPTC as well as for HK-2 and LLC-PK1 cells. Thresholds (x-axis) ranged from 0.3 to 4.0. 80% are indicated by a dotted line to facilitate comparisons.

Table 5 Determination of true positives (TP), true negatives (TN), sensitivity and specificity. TP were defined as true PT-specific nephrotoxins (22 drugs, group 1) that were correctly detected as positives by our assay. TN were defined as non-PT-specific nephrotoxins and non-nephrotoxic drugs (19 drugs; groups 2 and 3) that remained negative in our assay. How positive and negative assay results were obtained at different threshold levels is shown in Tables 3 and 4. TP and TN were determined at the indicated threshold levels ranging from 0.3 to 4.0 and the numbers are displayed. From these numbers, the percentages of sensitivity (TP/total number of group 1 drugs × 100%) and specificity (TN/total number of group 2 + 3 drugs × 100%) were calculated. The percentages of sensitivity and specificity are displayed in brackets together with the respective numbers of TP (sensitivity) and TN (specificity). The numbers of TP and TN and respective percentages of sensitivity and specificity were determined for all cell types and batches based on the results from all 41 drugs. The percentages of sensitivity and specificity shown here are graphically displayed in Fig. 3

Thresholds		HK-2	LLC-PK1	HPTC 1	HPTC 2	HPTC 3
0.3	TP	22 (100%)	22 (100%)	22 (100%)	22 (100%)	22 (100%)
	TN	0 (0%)	0 (0%)	0 (0%)	0 (0%)	0 (0%)
1.5	TP	18 (82%)	21 (96%)	22 (100%)	21 (96%)	22 (100%)
	TN	9 (47%)	7 (37%)	6 (32%)	5 (26%)	10 (53%)
2.0	TP	15 (68%)	18 (82%)	21 (96%)	19 (86%)	21 (96%)
	TN	15 (79%)	9 (47%)	14 (74%)	10 (53%)	13 (68%)
2.5	TP	14 (64%)	17 (77%)	21 (96%)	18 (82%)	17 (77%)
	TN	15 (79%)	11 (58%)	16 (84%)	11 (58%)	13 (68%)
3.0	TP	12 (55%)	14 (64%)	20 (91%)	17 (77%)	15 (68%)
	TN	15 (79%)	14 (74%)	16 (84%)	15 (79%)	14 (74%)
3.5	TP	11 (50%)	14 (64%)	20 (91%)	17 (77%)	14 (64%)
	TN	15 (79%)	14 (74%)	17 (90%)	16 (84%)	15 (79%)
4.0	TP	8 (36%)	14 (64%)	15 (68%)	16 (73%)	14 (64%)
	TN	15 (79%)	14 (74%)	18 (95%)	16 (84%)	15 (79%)

---

The results revealed that a threshold value of 3.5 was most suitable for two of the HPTC batches (Fig. 3; HPTC 1 and 2). At this threshold level, sensitivity, specificity and overall concordance with clinical data were ∼90% (HPTC 1) and ∼80% (HPTC 2). If the same threshold value (3.5) was applied to the third batch (HPTC 3) the specificity was still ∼80%, whereas sensitivity and overall concordance with clinical data were ∼64% and ∼71%, respectively. As expected, these data revealed some inter-donor variability between the different batches of primary cells.

Compared to HPTC the optimal threshold levels for HK-2 and LLC-PK1 cells were different and overall the data revealed that the predictability was lower when these cell lines were used instead of HPTC (Fig. 3). For instance, at threshold values of 3 and above, which were most suitable for LLC-PK1 cells, all of the values remained below 80% and ranged between ∼64% and ∼74%.

To further analyze the predictability we calculated the ROC curves and determined the area under the curve (AUC) values. Fig. 4 shows the ROC curves obtained with each cell batch/line and the results for either IL-6 or IL-8 or the combination of markers are displayed. The respective AUC values are shown in Table 6. The results confirm that the predictability was higher when HPTC were used (compared to HK-2 and LLC-PK1 cells), and this applied to the mean and median values as well as to each single batch of HPTC. The use of a combination of both markers only slightly improved the results in comparison to the use of IL-8 alone. The AUC values obtained with the marker combination ranged from 0.71 (HK-2) to 0.94 (HPTC 1).

Fig. 4 ROC curves. For each cell batch/type the ROC curves were calculated for each single marker or the combination of both markers. The respective AUC values are summarized in Table 6. For comparison panel F displays simultaneously the ROC curves (a combination of both markers) obtained for all the different cell batches/types tested.

Table 6 AUC values. The table provides the AUC values of the ROC curves (Fig. 4) for every cell batch/type tested. For HPTC also the mean and median values are shown. AUC values were determined separately for either IL-6 or IL-8 or the combination of these two markers. AUC values >0.5 represent a predictive model that is better than chance

Cell type	AUC
	IL-6/IL-8	IL-6	IL-8
HPTC 1	0.94	0.85	0.90
HPTC 2	0.81	0.72	0.80
HPTC 3	0.82	0.71	0.84
HPTC mean	0.85	0.76	0.85
HPTC median	0.82	0.72	0.84
HK-2	0.71	0.74	0.68
LLC-PK1	0.73	0.65	0.72

---

The most important performance metrics (balanced accuracy, sensitivity, specificity, PPV, NPV and AUC values) are summarized in Table 7. For Table 7 and further analyses we used the combined expression data of IL-6 and IL-8 as endpoints. The mean PPV for HPTC was 0.85, which means that 85% of the time a compound was called out as a PT-specific toxin correctly. The respective values for HK-2 and LLC-PK1 cells were 0.73 and 0.74, respectively. The mean NPV of HPTC was 0.79. Here, the values for the cell lines were 0.6 (HK-2) and 0.67 (LLC-PK1).

Table 7 Performance metrics. The table summarizes the values for the following performance metrics: balanced accuracy (defined as the average between sensitivity and specificity), sensitivity, specificity, PPV, NPV and AUC. In the case of sensitivity, specificity, PPV and NPV the values obtained at a threshold value of 3.5 (see Fig. 3) are displayed. With respect to the AUC values the results obtained with a combination of both markers are provided. These values are identical with those in Table 6 and are displayed here again for completeness

Cell type	Balanced accuracy	Sensitivity	Specificity	PPV	NPV	AUC
HPTC 1	0.90	0.91	0.90	0.91	0.94	0.94
HPTC 2	0.81	0.77	0.84	0.85	0.76	0.81
HPTC 3	0.71	0.64	0.79	0.78	0.68	0.82
HPTC mean	0.81	0.77	0.84	0.85	0.79	0.85
HPTC median	0.81	0.77	0.84	0.85	0.76	0.82
HK-2	0.65	0.50	0.79	0.73	0.60	0.71
LLC-PK1	0.69	0.64	0.74	0.74	0.67	0.73

---

Impact of endpoints

Next, we compared assay performance when either IL-6/IL-8 expression or cell death was used as the endpoint. Cell numbers were determined by high content screening (HCS) in order to quantify cell death. Table 8 summarizes the results. Using IL-6/IL-8 expression as the endpoint, 91% (20/22) of the PT-specific nephrotoxins gave a positive result and were correctly predicted when HPTC 1 were used. In contrast, substantial cell death, which allowed us to calculate IC₅₀ values, was observed in only 42% (8/19) of the cases tested. Even when all of these cases, where >50% cell death occurred, would be classified as positives, the sensitivity would not be better than chance. The same applied to the results obtained with HK-2 and LLC-PK1 cells, where with respect to group 1 compounds substantial cell death was observed in 43% (6/14) and 53% (8/15) of the cases tested.

Table 8 Comparison of drug effects on IL-6/IL-8 expression and cell numbers. HPTC 1, HK-2 and LLC-PK1 cells were exposed to the 41 test compounds. Data on IL-6/IL-8 expression were based on previous results (Tables 1 and 2). A result was defined as positive (+) when expression of at least one marker showed at any concentration an increase of 3.5-fold or above. If marker expression values remained below 3.5-fold the result was classified as negative (−). IC₅₀ values were calculated based on cell numbers determined by HCS. A value of >1000 μg ml⁻¹ was assigned if cell viability was >50% at the highest concentration of a compound (1000 μg ml⁻¹). Cell numbers were not determined (ND) in some cases

No.	HPTC 1		HK-2		LLC-PK1
	IL-6/IL-8	IC50 (μg ml^−1)	IL-6/IL-8	IC50 (μg ml^−1)	IL-6/IL-8	IC50 (μg ml^−1)
1	+	>1000	−	>1000	+	>1000
2	+	>1000	−	>1000	−	>1000
3	+	707	+	632	+	795
4	+	>1000	+	ND	+	>1000
5	+	>1000	+	ND	+	ND
6	+	47	−	94	−	38
7	+	ND	+	>1000	+	469
8	+	ND	−	ND	−	ND
9	−	>1000	−	>1000	−	>1000
10	+	>1000	−	>1000	+	69
11	+	21	+	7	+	19
12	+	>1000	−	>1000	+	>1000
13	+	4	+	ND	+	9
14	+	147.0	+	116	+	79
15	−	>1000	−	ND	−	>1000
16	+	96	+	ND	−	ND
17	+	>1000	−	ND	+	>1000
18	+	23	−	14	−	ND
19	+	45	+	44	+	45
20	+	>1000	−	>1000	+	ND
21	+	ND	+	ND	−	ND
22	+	>1000	+	>1000	+	ND
23	−	>1000	−	>1000	−	>1000
24	−	>1000	+	>1000	+	>1000
25	−	>1000	−	>1000	−	>1000
26	−	742	−	678	−	945
27	−	>1000	+	>1000	+	>1000
28	−	>1000	−	>1000	−	>1000
29	−	>1000	−	>1000	−	>1000
30	−	>1000	−	>1000	−	>1000
31	−	>1000	−	>1000	−	>1000
32	−	>1000	−	ND	−	>1000
33	+	>1000	−	>1000	+	>1000
34	−	>1000	−	>1000	−	>1000
35	−	>1000	−	>1000	−	>1000
36	−	>1000	−	>1000	−	>1000
37	−	>1000	−	>1000	−	>1000
38	−	>1000	−	>1000	−	71
39	−	>1000	+	>1000	+	>1000
40	−	>1000	−	>1000	−	>1000
41	+	>1000	+	>1000	+	>1000

---

These results support the idea that endpoints measuring general cytotoxicity might not be useful for organ-specific assays. It should be noted that cell death, as measured here by determining the cell numbers, is also measured by other widely used assays such as the neutral red uptake assay or the MTT assay (the MTT assay measures metabolic activity, which is often used as an indirect indicator of cell numbers). In the light of these results we refrained from addressing other endpoints measuring general cytotoxicity.

Here, we developed an in vitro model for the prediction of PT-specific toxicity in humans. The model was based on HPTC and tested with 41 compounds. For comparison, HK-2 and LLC-PK1 cells were also assessed. When three batches of HPTC were used the mean and median values for the major performance metrics (balanced accuracy, sensitivity, specificity, PPV, NPV and AUC values) ranged between 0.76 and 0.85. These results show that the predictability of the model is high and it would be expected that in ∼76%–85% of the cases where compounds were predicted as positives or negatives the predictions would be correct. Future work on the validation of the model will involve larger numbers of compounds and prospective studies.

Two major features distinguish our model from other models for the prediction of PT-specific toxicity: (i) the use of HPTC and (ii) the use of IL-6/IL-8 expression as endpoints. Another difference is the application of confluent epithelia, but the impact of this parameter is currently unclear. Our results show that both, the use of HPTC and of the selected endpoint, were important. The key performance metrics only ranged between 0.50 and 0.79 when HK-2 or LLC-PK1 cells were applied instead of HPTC. Also, the sensitivity was severely compromised when cell death was used as the endpoint instead of IL-6/IL-8 expression.

IL-6 and IL-8 are expressed by a large variety of cell types in response to a broad variety of stimuli and injury mechanisms (see, for instance ref. 47–50 and citations therein) and thus the specificity of our model might be surprising. However, potential contributions from other cell types do not play a role in an in vitro model based on one cell type. As long as no other parameters than HPTC injury increase marker gene expression in the in vitro model the response would be expected to be specific. It has been demonstrated before that IL-6 and IL-8 are increased in injured and diseased kidneys^42–44 and increasing evidence suggests that pro-inflammatory cytokines play a central role in the pathophysiology of AKI, including nephrotoxin-induced AKI.⁴⁵ Further, it is known that PT and PT-derived cells express IL-6 and IL-8 in vivo and in vitro^14,38–41 and significant up-regulation of IL-6 after exposure to nephrotoxins has been demonstrated in a kidney culture model employing purified PTs.²⁴ These findings are in agreement with the observation that IL-6 and IL-8 are specifically up-regulated in the in vitro model described here when HPTC were injured by compounds damaging this cell type.

Our results showed that HPTC were more suitable than widely used cell lines, but HPTC are affected by inter-donor variability and the cell source is limited. Further, it is important to note that HPTC in vitro always display a certain degree of de-differentiation, which possibly reflects an injury-like state after cell isolation. This might explain the lack of substantial up-regulation of novel AKI biomarkers after exposure to PT-specific nephrotoxins, as at least some of these biomarkers appear to be already expressed at high levels in untreated HPTC in vitro. The different state of the cells in vitro and in vivo might also be one of the reasons for false-positive and false-negative results. Currently no method has been described that could be used to maintain HPTC in vitro in a fully differentiated in vivo-like state. Established cell lines are obviously not a viable alternative to primary cells, and a lack of up-regulation of novel AKI biomarkers after nephrotoxin exposure was also observed when a more recently developed cell line was used (PREDICT-IV, 4th Project Periodic Report, June 30, 2012). Stem cell-derived cells are currently the most promising alternative to primary cells. We recently developed the first method for the differentiation of human embryonic stem cells (hESC) into HPTC-like cells.²⁷ Currently it is under investigation whether hESC-derived HPTC-like cells could be applied instead of HPTC in the in vitro model described here.

At the current state our model provides binary information and does not reveal dose-related information. Clinical data are available for all of the compounds tested. Currently the questions whether and how dose-related information could be obtained and how correlation analyses of in vitro and clinical data could be performed in the most useful ways are under further investigation. This involves a careful consideration and comparison of the clinical endpoints that could be useful for correlation analyses. Additional work on these aspects goes beyond the limits of a single study and the results of correlation analyses will be provided when this work has been completed in appropriate ways. It is also important to note that drug-induced nephrotoxicity is often idiosyncratic and not dose-dependent.

Our current results suggest that our model would allow prediction of PT-specific toxicity in humans with high accuracy at an early pre-clinical stage, which is currently not possible. This would provide additional valuable information during hit-to-lead discovery and lead optimization and would allow one to investigate underlying mechanisms of PT-specific toxicity at an early stage. Pre-clinical results reliably predicting PT-specific toxicity would also help to design clinical studies and to decide whether a more extensive and more frequent clinical safety assessment would be required or patients with an increased risk of nephrotoxicity (e.g. advanced age, diabetes) should be excluded from Phase II studies until more information has been obtained. In this regard it is interesting to note that tetracycline consistently gave positive results in our model. This drug has usually no obvious nephrotoxic effects, but can induce AKI and ESRD in patients with pre-existing kidney disease.^51,52 Novel biomarkers for detecting kidney toxicity^31–35 would be expected to be valuable in clinical studies where careful monitoring of nephrotoxic effects, as predicted by pre-clinical models, should be performed.

Conclusions

We have developed an in vitro model for the prediction of renal proximal tubular toxicity in humans. The model is based on HPTC and the endpoint is increased expression of IL-6 and/or IL-8. The predictability of the model is high and in the range of ∼76%–85%. The results suggest that this model would allow prediction of renal proximal tubular toxicity at an early pre-clinical stage during drug development. This would be important for developing safer drugs and early detection of nephrotoxicity would also help to save substantial costs during drug development. Future work will aim at further validating the model in prospective studies and adapting it to human stem cell-derived renal cells.

Acknowledgements

We are grateful to Prof. Dr Anantharaman Vathsala (NUHS) and Dr Hoo Yee Tiong (NUHS) for continuous support and valuable advice. Anonymized human kidney tissue samples were provided by the National University Health System (NUHS) Tissue Repository, Singapore. We thank Dr Thomas Paulraj Thamboo (NUHS) for pathological examination of the tissue samples and Prof. Carol Pollock (The University of Sydney, Australia) for help with the HPTC isolation procedure. We thank Dr Angeline Goh (NUHS) for advice on data interpretation. This work was supported by a grant from the Joint Council Office (Agency for Science, Technology and Research, Singapore) Development Program and the Institute of Bioengineering and Nanotechnology (Biomedical Research Council, Agency for Science, Technology and Research).

References

1. E. M. Levy, C. M. Viscoli and R. I. Horwitz, JAMA, 1996, 275, 1489–1494.
2. D. Choudhury and Z. Ahmed, Nat. Clin. Pract. Nephrol., 2006, 2, 80–91.
3. X. Guo and C. Nzerue, Cleve Clin. J. Med., 2002, 69, 289–290 , 293–4, 296–7 passim.
4. K. Nash, A. Hafeez and S. Hou, Am. J. Kidney Dis., 2002, 39, 930–936.
5. B. S. Moffett and S. L. Goldstein, Clin. J. Am. Soc. Nephrol., 2011, 6, 856–863.
6. C. C. Szeto and K. M. Chow, Ren Fail., 2005, 27, 329–333.
7. M. A. Perazella, Hosp. Pract. (Minneap), 2001, 36, 43–46 , 55–46.
8. W. S. Redfern, L. Ewart, T. G. Hammond, R. Bialecki, L. Kinter, S. Lindgren, C. E. Pollard, R. Roberts, M. G. Rolf and J. P. Valentin, Toxicologist, 2010, 114, 231.
9. H. Izzedine, M. Harris and M. A. Perazella, Nat. Rev. Nephrol., 2009, 5, 563–573.
10. P. P. Kapitsinou and N. Ansari, J. Med. Case Rep., 2008, 2, 94.
11. W. Pfaller and G. Gstraunthaler, Environ. Health Perspect., 1998, 106(Suppl 2), 559–569.
12. P. Prieto, Altern. Lab. Anim., 2002, 30(Suppl 2), 101–106.
13. Y. Wu, D. Connors, L. Barber, S. Jayachandra, U. M. Hanumegowda and S. P. Adams, Toxicol. In Vitro, 2009, 23, 1170–1178.
14. Y. Li, Y. Zheng, K. Zhang, J. Y. Ying and D. Zink, Nanotoxicology, 2012, 6, 121–133.
15. P. H. Bach, D. K. Obatomi and S. Brant, In vitro methods for nephrotoxicity screening and risk assessment, in In vitro methods in pharmaceutical research, ed. J. V. Castell and M. J. Gomez-Lechon, Academic Press Ltd, San Diego, 1997.
16. M. Bens and A. Vandewalle, Pfluegers Arch., 2008, 457, 1–15.
17. S. E. Jenkinson, G. W. Chung, E. van Loon, N. S. Bakar, A. M. Dalzell and C. D. Brown, Pfluegers Arch., 2012, 464, 601–611.
18. Z. Lin and Y. Will, Toxicol. Sci., 2012, 126, 114–127.
19. T. Duff, S. Carter, G. Feldman, G. McEwan, W. Pfaller, P. Rhodes, M. Ryan and G. Hawksworth, Altern. Lab. Anim., 2002, 30(Suppl 2), 53–59.
20. W. Li, D. F. Choy, M. S. Lam, T. Morgan, M. E. Sullivan and J. M. Post, Toxicol. in Vitro, 2003, 17, 107–113.
21. W. Li, M. Lam, D. Choy, A. Birkeland, M. E. Sullivan and J. M. Post, Toxicol. in Vitro, 2006, 20, 669–676.
22. A. Limonciel, L. Aschauer, A. Wilmes, S. Prajczer, M. O. Leonard, W. Pfaller and P. Jennings, Toxicol. in Vitro, 2011, 25, 1855–1862.
23. A. I. Astashkina, B. K. Mann, G. D. Prestwich and D. W. Grainger, Biomaterials, 2012, 33, 4712–4721.
24. A. I. Astashkina, B. K. Mann, G. D. Prestwich and D. W. Grainger, Biomaterials, 2012, 33, 4700–4711.
25. C. C. Beeson, G. C. Beeson and R. G. Schnellmann, Anal. Biochem., 2010, 404, 75–81.
26. D. A. Vesey, W. Qi, X. Chen, C. A. Pollock and D. W. Johnson, Methods Mol. Biol., 2009, 466, 19–24.
27. K. Narayanan, K. M. Schumacher, F. Tasnim, K. Kandasamy, A. Schumacher, M. Ni, S. Gao, B. Gopalan, D. Zink and J. Y. Ying, Kidney Int., 2013, 83, 593–603.
28. F. Tasnim, K. Kandasamy, J. S. Muck, M. S. Bin Ibrahim, J. Y. Ying and D. Zink, Tissue Eng. Part A, 2012, 18, 262–276.
29. M. Ni, J. C. Teo, M. S. Ibrahim, K. Zhang, F. Tasnim, P. Y. Chow, D. Zink and J. Y. Ying, Biomaterials, 2011, 32, 1465–1476.
30. M. Ni, P. K. Zimmermann, K. Kandasamy, W. Lai, Y. Li, M. F. Leong, A. C. Wan and D. Zink, Biomaterials, 2012, 33, 353–364.
31. J. V. Bonventre, V. S. Vaidya, R. Schmouder, P. Feig and F. Dieterle, Nat. Biotechnol., 2010, 28, 436–440.
32. F. Dieterle, F. Sistare, F. Goodsaid, M. Papaluca, J. S. Ozer, C. P. Webb, W. Baer, A. Senagore, M. J. Schipper, J. Vonderscher, S. Sultana, D. L. Gerhold, J. A. Phillips, G. Maurer, K. Carl, D. Laurie, E. Harpur, M. Sonee, D. Ennulat, D. Holder, D. Andrews-Cleavenger, Y. Z. Gu, K. L. Thompson, P. L. Goering, J. M. Vidal, E. Abadie, R. Maciulaitis, D. Jacobson-Kram, A. F. Defelice, E. A. Hausner, M. Blank, A. Thompson, P. Harlow, D. Throckmorton, S. Xiao, N. Xu, W. Taylor, S. Vamvakas, B. Flamion, B. S. Lima, P. Kasper, M. Pasanen, K. Prasad, S. Troth, D. Bounous, D. Robinson-Gravatt, G. Betton, M. A. Davis, J. Akunda, J. E. McDuffie, L. Suter, L. Obert, M. Guffroy, M. Pinches, S. Jayadev, E. A. Blomme, S. A. Beushausen, V. G. Barlow, N. Collins, J. Waring, D. Honor, S. Snook, J. Lee, P. Rossi, E. Walker and W. Mattes, Nat. Biotechnol., 2010, 28, 455–462.
33. T. Ichimura, C. C. Hung, S. A. Yang, J. L. Stevens and J. V. Bonventre, Am. J. Physiol. Renal. Physiol., 2004, 286, F552–F563.
34. W. S. Waring and A. Moonie, Clin. Toxicol., 2011, 49, 720–728.
35. J. Mishra, Q. Ma, A. Prada, M. Mitsnefes, K. Zahedi, J. Yang, J. Barasch and P. Devarajan, J. Am. Soc. Nephrol., 2003, 14, 2534–2543.
36. D. Liang, H. F. Liu, C. W. Yao, H. Y. Liu, C. M. Huang-Fu, X. W. Chen and S. H. Du, Nephrology (Carlton), 2007, 12, 53–61.
37. K. Miyauchi, Y. Takiyama, J. Honjyo, M. Tateno and M. Haneda, Diabetes Res. Clin. Pract., 2009, 83, 190–199.
38. Z. I. Niemir, H. Stein, A. Ciechanowicz, P. Olejniczak, G. Dworacki, E. Ritz, R. Waldherr and S. Czekalski, Am. J. Kidney Dis., 2004, 43, 983–998.
39. J. S. Gerritsma, P. S. Hiemstra, A. F. Gerritsen, W. Prodjosudjadi, C. L. Verweij, L. A. Van Es and M. R. Daha, Clin. Exp. Immunol., 1996, 103, 289–294.
40. J. S. Gerritsma, C. van Kooten, A. F. Gerritsen, A. M. Mommaas, L. A. van Es and M. R. Daha, Exp. Nephrol., 1998, 6, 208–216.
41. S. C. Tang, J. C. Leung and K. N. Lai, Contrib. Nephrol., 2011, 170, 124–134.
42. M. Araki, N. Fahmy, L. Zhou, H. Kumon, V. Krishnamurthi, D. Goldfarb, C. Modlin, S. Flechner, A. C. Novick and R. L. Fairchild, Transplantation, 2006, 81, 783–788.
43. D. N. Grigoryev, M. Liu, H. T. Hassoun, C. Cheadle, K. C. Barnes and H. Rabb, J. Am. Soc. Nephrol., 2008, 19, 547–558.
44. D. Tramma, M. Hatzistylianou, G. Gerasimou and V. Lafazanis, Pediatr. Nephrol., 2012, 27, 1525–1530.
45. A. Akcay, Q. Nguyen and C. L. Edelstein, Mediators Inflammation, 2009, 2009, 137072.
46. G. Elberg, S. Guruswamy, C. J. Logan, L. Chen and M. A. Turman, Cell Tissue Res., 2008, 331, 495–508.
47. T. A. Luger and T. Schwarz, J. Invest. Dermatol., 1990, 95, 100S–104S.
48. P. Trayhurn, C. A. Drevon and J. Eckel, Arch. Physiol. Biochem., 2011, 117, 47–56.
49. M. Rossol, H. Heine, U. Meusch, D. Quandt, C. Klein, M. J. Sweet and S. Hauschildt, Crit. Rev. Immunol., 2011, 31, 379–446.
50. R. A. Swerlick and T. J. Lawley, J. Invest. Dermatol., 1993, 100, 111S–115S.
51. C. S. Miller and G. J. McGarity, J. Am. Dent. Assoc., 2009, 140, 56–60.
52. M. E. Phillips, J. B. Eastwood, J. R. Curtis, P. C. Gower and H. E. De Wardener, Br. Med. J., 1974, 2, 149–151.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c3tx50042j

---