Metabolic impact of genetic and chemical ADP/ATP carrier inhibition in renal proximal tubule epithelial cells

Mitochondrial dysfunction is pivotal in drug-induced acute kidney injury (AKI), but the underlying mechanisms remain largely unknown. Transport proteins embedded in the mitochondrial inner membrane form a significant class of potential drug off-targets. So far, most transporter-drug interactions have been reported for the mitochondrial ADP/ATP carrier (AAC). Since it remains unknown to what extent AAC contributes to drug-induced mitochondrial dysfunction in AKI, we here aimed to better understand the functional role of AAC in the energy metabolism of human renal proximal tubular cells. To this end, CRISPR/Cas9 technology was applied to generate AAC3−/− human conditionally immortalized renal proximal tubule epithelial cells. This AAC3−/− cell model was characterized with respect to mitochondrial function and morphology. To explore whether this model could provide first insights into (mitochondrial) adverse drug effects with suspicion towards AAC-mediated mechanisms, wild-type and knockout cells were exposed to established AAC inhibitors, after which cellular metabolic activity and mitochondrial respiratory capacity were measured. Two AAC3−/− clones showed a significant reduction in ADP import and ATP export rates and mitochondrial mass, without influencing overall morphology. AAC3−/− clones exhibited reduced ATP production, oxygen consumption rates and metabolic spare capacity was particularly affected, mainly in conditions with galactose as carbon source. Chemical AAC inhibition was stronger compared to genetic inhibition in AAC3−/−, suggesting functional compensation by remaining AAC isoforms in our knockout model. In conclusion, our results indicate that ciPTEC-OAT1 cells have a predominantly oxidative phenotype that was not additionally activated by switching energy source. Genetic inhibition of AAC3 particularly impacted mitochondrial spare capacity, without affecting mitochondrial morphology, suggesting an important role for AAC in maintaining the metabolic spare respiration. Supplementary Information The online version contains supplementary material available at 10.1007/s00204-023-03510-7.


Introduction
Acute kidney injury (AKI) is a known major disease burden, affecting 20% of hospitalized adults worldwide (Ishimoto and Inagi 2016; Siew and Davenport 2015). Mortality rates are over 50 percent in severe cases and less severe manifestations are associated with chronic kidney disease (Hoste et al. 2015). A steadily increasing number of commonly used drugs have been associated with AKI (Palevsky et al. 2013) and various prospective cohort studies indicated that drug-induced mechanisms explain 14-26% of all AKI cases, emphasizing the importance of renal toxicity assessment (Mehta et al. 2004; Moffett and Goldstein 2011;Perazella 2018). Besides a major burden for the healthcare system, drug-induced AKI also has implications for the Frans G. M. Russel and Tom J. J. Schirris have contributed equally to this work. pharmaceutical industry as drug attrition in development, clinical and post-marketing phases is often due to renal toxicity (Gai et al. 2020;Hoste et al. 2015;Mehta et al. 2004;Uchino et al. 2005;van der Meer et al. 2010).
Mitochondrial dysfunction has shown to play a pivotal role in AKI (Bhatia et al. 2020). This is compatible with the kidneys being among the most energy-demanding organs in the human body, consuming approximately 10 percent of the total oxygen supply. Accordingly, tubular cells are mitochondria rich, to meet ATP demands (e.g. to sustain activity of ATP-consuming pumps and transporters) required for physiological function. In particular, the proximal tubule contains a large amount of mitochondria and is characterized by the relatively high protein expression levels of solute transporters (Gai et al. 2020;Huls et al. 2008;Koepsell 2013;Launay-Vacher et al. 2006;Motohashi and Inui 2013;Prasad et al. 2016;Smeets et al. 2004;Soltoff 1986). The latter facilitate high transport rates of solutes, drugs and their (toxic) metabolites, contributing to high intracellular drug exposure. This could augment the probability of druginduced mitochondrial dysfunction, eventually resulting in renal insufficiency caused by proximal tubular cell death through ATP depletion and increased levels of oxidative stress (Gai et al. 2020;Heidari 2019).
The inner mitochondrial membrane (IMM) is virtually impermeable to most molecules and ions, which is essential for maintaining the proton gradient and the ensuing inwarddirected proton-motive force to drive ATP generation by the oxidative phosphorylation system (OXPHOS). However, to link cytosolic and mitochondrial metabolic pathways, molecules, inorganic ions and cofactors need to be exchanged between both compartments. This is facilitated by a plethora of transport proteins embedded in the IMM (Jaiquel Baron et al. 2021;Palmieri et al. 2020;Ruprecht and Kunji 2020), which account for more than five percent of the mitochondrial proteome. In principle, these transporters represent a significant class of potential drug off targets, which have largely been unexplored. Most of these transporters belong to the mitochondrial carrier family (SLC25) and enable import or efflux of solutes important for many cellular processes (Kunji 2011;Palmieri and Monne 2016). In addition, other transporters are located on the IMM include the pyruvate carrier (Herzig et al. 2012;Kunji 2011), five ATPbinding cassette (ABC) transporters (Lill and Kispal 2001), a Mg 2+ -transporter, a Na + /Ca 2+ -exchanger, a Ca 2+ -uniporter, an aquaporin and five sideroflexins (Kunji 2011;Palmieri and Pierri 2010). The SLC25 family consists of a vast number of isoforms which transport the same substrates but vary in expression from ubiquitous to organ-or disease-specific patterns (Gutierrez-Aguilar and Baines 2013; Haitina et al. 2006;Palmieri 2004). A canonical member of the SLC25 family is the mitochondrial ADP/ATP carrier (AAC) that mediates electrogenic exchange of ADP 3− from the cytosol into the mitochondrial matrix against ATP 4− to supply energy to the cytosol (Klingenberg 2008;Kunji et al. 2016a;Ruprecht and Kunji 2019). In humans four AAC isoforms are identified, which are encoded by the SLC25A4, SLC25A5, SLC25A6 and SLC25A31 genes in a tissue-specific manner. They cycle between a cytoplasmic-and matrix-open state, to enable bidirectional access to the substrate binding site. Due to their central role in cellular energy metabolism and more specifically in ADP/ATP exchange, they form a potential off-target in drug-induced mitochondrial dysfunction underlying proximal tubular toxicity and AKI (Jaiquel Baron et al. 2021).
Here, we aimed to better understand the function of AAC in cellular energy metabolism in human conditionally immortalized proximal tubule epithelial cells (ciPTEC). To investigate this, we applied CRISPR/Cas9 to generate AAC3 −/− cells, the most abundant AAC isoform expressed in ciPTEC (Palmieri 2004;Stepien et al. 1992). This AAC3 −/− cell model was subsequently characterized based on function and morphology. Previously, overexpression models in Lactococcus lactis, Saccharomyces cerevisiae or proteoliposomes have been developed to study drug-induced AAC inhibition (Jaiquel Baron et al. 2021;Todisco et al. 2016;Zhang et al. 2016). However, these models overexpress the protein of interest without considering the complete and complex human cellular physiology, and the potential interactions between AAC and other mitochondrial or cellular pathways. To explore whether our cell model could be useful in providing insights into (mitochondrial) adverse effects of drugs with a suspicion towards AAC-mediated mechanisms, we exposed wild-type and AAC3 −/− cells to established AAC inhibitors and measured cellular metabolic activity and respiratory capacity.

Single guide RNA design
Single guide RNAs (sgRNAs) targeting exon 1 and 2 of the Homo sapiens gene AAC3 (SLC25A6) were designed for clustered regularly interspaced short palindromic repeat (CRISPR/Cas9) using CHOPCHOP and crispor.tefor. net (hg38/GRCh38 was used as reference genome). sgR-NAs were selected based on predicted efficiency score and absence of off-targets with > 3 mismatches ( Table 1). The oligonucleotide pair for each sgRNA was phosphorylated and annealed using T4 Polynucleotide Kinase (EK0031, Thermofisher Scientific) in a thermocycler at 37 °C for 30 min, followed by 5 min at 95 °C and 5 °C/min decrease to 25 °C, according to the manufacturer's instructions. Plasmid pX333-GFP (kindly provided by Eric Verschuren, Dept. of Physiology, Radboudumc, Addgene, #64,073) was linearized at 37 °C for 1 h using BbsI (R0539, New England Biolabs, Ipswich, MA, USA), or BsaI (R0535, New England Biolabs) restriction enzymes. Ligation of annealed oligonucleotides into the linearized pX333-GFP plasmid was performed using T4 ligase (New England Biolabs) overnight at 16 °C, following instructions by the manufacturer. Heat shock at 42 °C transformed the ligation mix into TOP10F competent cells. The next day, colonies with pX333 containing sgRNAs were inoculated and plasmids were isolated using GenEluteTM Plasmid Miniprep Kit (Sigma-Aldrich) according to manufacturer's instructions. As the plasmid enables dual expression of sgRNAs, sequencing resulted in selection of plasmids containing both sgRNAs (forward primer exon 1/2 sgRNA, see Table 1). Insertion of T2A-eGFP after NLS-Cas9-NLS, allowed GFP-positive FACS sorting after transfection. Prior to transfection ciPTEC-OAT1 exons 1-4 were sequenced to confirm wild-type sequences (Table 1).

Functional ADP and ATP transport assessment
Cells were cultured and matured as described above. Mature cells were pelleted and resuspended in ice cold 10 mM Tris-HCl, pH 7.6, for isolation of mitochondrial enriched fractions. Cell suspension was mechanically homogenized (eight times) using a Potter-Elvehjem homogenizer on ice.
To increase protein yield, a maximum of 3 million cells 1 3 was homogenized at once. The homogenate was transferred to ice cold 1.5 M sucrose solution, mixed and centrifuged for 10 min at 2 °C and 600g. To obtain mitochondrialenriched protein fractions, supernatant was spun down for 10 min at 2 °C and 14,000g, followed by resuspension of pellet in 10 mM Tris-HCl, pH 7.6 complemented with cOmplete™ Mini EDTA-free protease inhibitor, PhosS-TOP™ and 0.1 mg/mL RNase-free DNase I (Roche, Basel, Switzerland). Mitochondrial protein in the isolated fraction was determined after three freeze/thaw cycles in liquid nitrogen, to break mitochondrial membranes, using Pierce™ BCA protein assay kit (Thermofisher Scientific) according to the manufacturer's instructions.

Cellular respiration and real-time ATP production rate measurements
To investigate cellular metabolism in both genetically and chemically inhibited cells, cellular oxygen consumption rates were measured using extracellular flux analyses (Seahorse XF96 Agilent, Santa Clara, CA, USA). 10,000 cells/well were seeded in 96-wells Seahorse XF96 cell culture microplates and matured as described above. Routinely culturing cells in high glucose conditions can inhibit mitochondrial function, known as the Crabtree effect, previously described for rapidly proliferating cancer cells (e.g., HepG2), thymocytes and human primary muscle cells (Aguer et al. 2011;Marroquin et al. 2007;Rossignol et al. 2004). Although a predominant oxidative phenotype for ciPTEC was previously suspected (Vriend et al. 2019), we attempted to stimulate mitochondrial respiration by replacing the glucose in the medium with galactose. While the glycolytic metabolism of glucose yields two net ATP molecules, pyruvate production via metabolism of galactose yields no ATP, and consequently increases reliance on oxidative phosphorylation for energy production (Marroquin et al. 2007). The effects of AAC interactors on mitochondrial respiration were evaluated in glucose-free DMEM (Sigma, D5030) complemented with 10 mM galactose, 0.64 mM MgCl 2 , 14.3 mM NaHCO 3 , 15 mM HEPES, 0.5 mM sodium pyruvate, and 2.5 mM glutamine, supplemented with ciPTEC growth factors mentioned above, pH 7.4 and 100 µM AAC inhibitors BKA, CATR, CD437, suramin or 0.1% DMSO for 12 h at 37 °C and 5% (v/v) CO 2 . Galactose-rich (10 mM) medium containing AAC inhibitors was replaced one hour before initiation of the experiment by NaHCO 3 -free DMEM containing 10 mM galactose, 0.5 mM sodium pyruvate, 31.7 mM NaCl,

Citrate synthase activity
Mitochondrial mass was determined by citrate synthase activity in matured ciPTEC-OAT1 wild type and AAC3 −/− , as previously described (Srere 1969). In short, cells were seeded and matured in transparent/clear flat bottom 96-wells plates as described above. Cells were lysed by 0.33% (v/v) Triton X-100 in 10 mM Tris-HCl, pH 7.6, followed by three freeze-thaw cycles in liquid nitrogen. Absorption was measured spectrophotometrically at 412 nm after addition of five volumes of 1 mM 5,5′-Dithiobis(2-nitrobenzoic acid) (DTNB, D8130, Sigma) with 10% (v/v) Triton X-100 in MilliQ, and 300 µM acetyl coenzyme A (A2181, Sigma), in a 25 °C prewarmed microplate reader (BioRad) using a kinetic protocol of medium mixing for 10 s with 20 s interval, followed by twenty cycles of 15 s interval without mixing. Total activity was measured after addition of 10 µL 10 mM oxaloacetic acid (A4126, Sigma) using the same kinetic protocol. Data analysis was performed in GraphPad Prism 9.0.0. For each measurement, slopes of at least five time points were defined, subtracted, and plotted.

Mitochondrial morphology and function by TMRM
Cells were seeded and matured as described above in black/ clear bottom flat 96-wells plates, after which mitochondrial morphology and membrane potential were assessed microscopically. Cells were washed once in HBSS, subsequently fluorescently stained with 25 nM tetramethylrhodamine methyl ester (TMRM) and co-stained with 0.75 µg/ mL Hoechst 33342 in HBSS for 30 min at 37 °C, followed by image acquisition (40 × objective) and 12 × 12 montage using Becton Dickinson (BD) Pathway 855 microscope (BD Bioscience, Breda, The Netherlands), as described before (Iannetti et al. 2016). Acquired images were processed and quantified using Image-Pro Plus 6.0 software (Media Cybernetics, Rockville, MD, USA), as described previously (Iannetti et al. 2016). Average data was obtained in N = 3 experiments, consisting of 20 replicates each, and 8 cells/ well were analyzed.

Statistical analysis
Statistical data analysis was performed using GraphPad Prism 9.0.0 (GraphPad Software Inc., San Diego, CA). All data were normalized to wild type ciPTEC-OAT1 and presented as mean ± SEM of at least three independent experiments (N = 3), performed with five or six experimental replicates as indicated in the figure legends, unless stated otherwise. One-way ANOVA with Dunnett's post hoc test or two-way ANOVA with Bonferroni post hoc test to correct for multiple comparison were used to evaluate differences between cell lines or responses, as indicated. Concentrationdependent effect of AAC inhibitors on cellular metabolic activity were fitted using nonlinear regression analysis.

AAC3 −/− knockout cells show a reduced ADP/ ATP transporter activity and cellular oxidative metabolism
Human AAC has four different isoforms, for which expression patterns vary in a tissue-and disease-specific manner Fig. 1 Development and validation of CRISPR/Cas9-mediated knockout of AAC3 in ciPTEC cells. Quantitative PCR was applied to assess the mRNA expression levels of the four human isoforms of AAC in conditionally immortalized proximal tubule epithelial cells (ciPTEC), before (A) and after (B) application of CRISPR/Cas9, for wild-type ciPTEC-OAT1 (black), AAC3 −/− -1 (dark grey) and AAC3 −/− -2 (light grey), relative to GAPDH expression. Two independently generated AAC3 knockout cell lines were included in the experiments. In addition, protein levels of AAC1, AAC2 and AAC3 were determined by Western Blot (C) in which COX5A was used as loading control. AAC4 was excluded from analysis as qPCR revealed no mRNA expression, also in line with expression patterns described before (Dolce et al. 2005;Jang et al. 2008;Kunji et al. 2016b;Stepien et al. 1992). Functional AAC-mediated transport of ADP (D) and ATP (E) was investigated, using radioactivity and bioluminescence. Mitochondrial respiration was assessed by the mito stress test (Agilent), as described by the manufacturer, performed in culture medium containing 10 mM glucose, using the Seahorse XF Analyzer to evaluate oxygen consumption rates (OCR) and extracellular acidification rate (ECAR) of the different mitochondrial complexes (F and G), upon following injection of complex specific inhibitors oligomycin A (2.5 µM, complex V), carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP, 1 µM) and rotenone/ antimycin A (1 µM/2.5 µM). Basal respiration (orange) consists of respiratory capacities under resting conditions, maximal respiration (green) is the maximal respiratory capacity of a cell. Spare capacity (red) was determined by subtracting basal respiration from maximal respiration (F and H). Cellular ATP production rate was assessed by the ATP rate assay, using the Seahorse XF Analyzer (I). Data was corrected for cell count, assessed by fluorescence microscopy after Hoechst staining. Significance was determined by one-way ANOVA with Dunnett's post hoc analysis, *p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001, mean ± SEM, N = 3 or 4 independent experiments ◂ (Chevrollier et al. 2011;Chevrollier et al. 2005;Dolce et al. 2005;Jang et al. 2008;Kunji et al. 2016a;Stepien et al. 1992). Previously, the Human Protein Atlas Project identified AAC3 as most abundantly expressed isoform in renal cells. Similarly, we found AAC3 to be the predominant isoform in our renal ciPTEC cell model (Fig. 1A). Next, to better understand the function of AAC3 in cellular energy metabolism, the CRISPR/Cas9 approach was used to independently generate two AAC3 −/− proximal tubule epithelial cell lines, followed by genetic characterization of the knockout cell lines ( Fig. 1 and Supplemental Fig. 1). First, quantitative PCR showed a significant reduction of AAC3-specific mRNA levels in both AAC3 −/− cell lines (Fig. 1B), whereas AAC1 and AAC2 were not affected. Sanger sequencing confirmed that targeting exon 1 resulted in two cell lines (AAC3 −/− -1 and AAC3 −/− -2, Supplemental Fig. 1) with severely shortened proteins of 61 and 60 amino acids, respectively, compared to 298 amino acids in wild type. Consequently, protein function was predicted to be disrupted. Protein expression was absent for AAC3 (~ 33 kDa), and unaffected for AAC1 (Fig. 1C), which suggests that no functional protein was properly folded. Remarkably, AAC2 could not be detected at ~ 33 kDa (Fig. 1C). However, a dimeric band (~ 66 kDa) was observed (Supplemental Fig. 2), which could be due to the detergent and its effect on the oligomeric state of membrane proteins in sample preparations, as discussed previously . Next, we functionally characterized the AAC3 −/− and wild-type cell lines, to determine the impact of the truncated proteins. With respect to transporter activity, AAC-dependent ADP and ATP transport were assessed by measurement of direct import of ADP, using detection of radiolabeled ADP and an indirect method of ATP export using a bioluminescence assay, respectively. Both approaches demonstrated a reduced AAC-dependent ADP import and ATP export transporter activity, indicating that also on a functional level, gene editing did affect protein function ( Fig. 1D and E). However, a substantial residual AAC activity was observed in AAC3 −/− . To exclude that this residual transport was not due to the presence of fractionated mitochondria as a result of the preparation (isolation) procedure, this was evaluated by high-resolution respirometry using the Oxygraph-2 k, as previously published (Lanza and Nair 2009). The integrity of the outer mitochondrial membrane was not affected (Supplemental Fig. 4). Although the OCR remained high upon inhibition of ATP synthase (State 4 respiration), which represents the respiration required to maintain the mitochondrial membrane potential during uncoupling, an equivalent reduction in respiration rate from state 3 to state 4 was observed, at least for wild type and AAC3 −/− cells. This suggests that the lower ADP and ATP transport activity in AAC3 −/− (Fig. 1D and E) does not result from an increase in proton leakage in knockout cells compared to wild type. Alternatively, this observation could be explained by functional compensation through AAC1 and AAC2 (i.e. a higher transport rate), and not upregulation or increased protein expression, as mRNA and protein levels were not elevated for the different AAC isoforms (Fig. 1B/C and Supplemental Fig. 2). Unfortunately, isoform-specific alterations in AAC activity are difficult to assess, because of their highly similar structure, function and substrate selectivity (Kunji 2012;Kunji and Robinson 2006). To further evaluate the effects of AAC3 −/− on mitochondrial and cellular metabolism, we investigated oxygen consumption and ATP production rates in wild-type and AAC3 −/− cells. Compared to wild type, both AAC3 −/− cell lines showed a significant decrease in basal and maximal respiration (OCR), while low basal ECAR levels did not differ between cell lines (Fig. 1F, G and Supplemental Fig. 3A), suggesting that cells mainly use OXPHOS-mediated ATP generation and do not compensate for the absence of AAC3 by increased glycolytic activity. Spare respiratory capacity was reduced in the AAC3 −/− cells (Fig. 1H). Moreover, ATP production rates were significantly lower in AAC3 −/− cells compared to wild type, suggesting mitochondrial dysfunction (Fig. 1I). To further probe the effects on oxidative metabolism, we used galactose as an energy source instead of glucose in an attempt to stimulate mitochondrial metabolism (Fig. 2) (Aguer et al. 2011). Under these conditions, maximal OCR was reduced compared to basal levels, indicating that under non-stressed conditions (i.e., basal respiration) cells used their complete oxidative respiratory capacity and were not able to adapt under stressed (i.e., maximal respiration) conditions ( Fig. 2A and B). Remarkably, the basal OCR of these cells in galactose was similar compared to glucose conditions, suggestive of an overall oxidative phenotype of these cells independent of the energy source. It was indeed previously shown that ciPTEC cells overexpressing the organic anion transporter 1 (ciPTEC-OAT1) are characterized by reduced extracellular lactate levels, while intracellular ATP levels were retained, which was attributed to the shift towards an oxidative phenotype in these cells (Vriend et al. 2019). This is in accordance with our observation that basal ECAR was low in both glucose-and galactose-conditioned media ( Fig. 2B and Supplemental Fig. 3). The low glycolytic capacity may be explained by the metabolic nature of renal proximal tubular cells, which has been described to mainly depend on fatty acid β-oxidation (Simon and Hertig 2015). However, this is only indirect evidence, as a correct interpretation of ECAR is challenging due to confounding interference by other sources of extracellular acidification, such as the reduction of exogenous pyruvate to lactate (Schmidt et al. 2021). Further research is required to understand why respiration rates remained unchanged in these cells under galactose conditions.
The importance of the spare capacity used under these conditions is shown by the fact that under uncoupled conditions (i.e., maximal respiration) both AAC3 −/− cell lines fail to thrive because respiration collapses even below basal rates ( Fig. 2A-C). In line with this finding, ATP production rate was significantly reduced in AAC3 −/− -1 (Fig. 2D).

AAC3 −/− affects mitochondrial mass but not their morphology
AAC deficiency has previously been associated with an increased mitochondrial mass (i.e., citrate synthase activity) ) and, therefore, we investigated mitochondrial content in our AAC3 −/− cell lines. Remarkably, we found that citrate synthase (CS) activity per cell was significantly reduced in both AAC3 −/− , while the total number of cells was unaffected ( Fig. 3A and B). To estimate whether mitochondrial mass was indeed decreased in the AAC3 knockout cell lines, we measured the number of mitochondrial objects per cell using TMRM staining and fluorescence microscopy. The number of mitochondria per cell was significantly reduced in AAC3 −/− -1 and appeared lower in AAC3 −/− -2 (Fig. 3C). Quantification of mitochondrial size (Fig. 3D) allowed calculation of the total mitochondrial area per cell, which was not significantly reduced in the AAC knockout cell lines (Fig. 3E). This may suggest that the observed decrease in mitochondrial mass measured by CS activity was relatively small. It also supports the idea that the observed residual ADP/ATP exchange activity is indeed explained by functional compensation of the other AAC isoforms, described above. Analysis of mitochondrial TMRM staining intensity revealed no differences between cell lines (Fig. 3F), whereas mitochondrial aspect ratio was slightly reduced in AAC3 −/− -2 cells (Fig. 3G). No difference in mitochondrial formfactor was observed (Fig. 3H). This suggests that the mitochondrial membrane potential is maintained by a mechanism that is independent of AAC3 or OCR, which requires further investigation.

Oxidative conditions increase susceptibility to AAC inhibition
As described above, the AAC3 knockout cell lines have residual ADP/ATP exchange and respiratory capacity, which may point towards functional compensation by other AAC isoforms. Consequently, we aimed to investigate the effects of AAC inhibitors on cellular metabolism in wild type and AAC3 −/− . Under standard glucose conditions, cell-permeable AAC inhibitors BKA and CD437 reduced the metabolic activity measured by MTT ( Fig. 4A and B) (Jaiquel Baron et al. 2021). Interestingly, switching the carbon source to galactose, making these cells solely reliant on mitochondrial oxidative capacity, enhanced metabolic inhibition. CATR, the most potent AAC inhibitor (Vignais et al. 1966), also  A and B). Spare capac-ity was determined by subtracting basal from maximal respiration rates (C). Cellular ATP production was assessed by measuring the ATP rate using the Seahorse XF Analyzer (D). Data was corrected for cell count, assessed by fluorescence microscopy after Hoechst staining. Significance was determined by one-way ANOVA with Dunnett's post hoc analysis, *p < 0.05, **p < 0.01 and ***p < 0.001, mean ± SEM, N = 3 independent experiments 1 3 showed a significant concentration-dependent reduction in cellular metabolic activity in wild-type cells (Fig. 4C). The previously reported AAC inhibitor suramin (Jaiquel Baron et al. 2021), did not cause a significant reduction in metabolic capacity (Fig. 4D), but this may have been due to its poor cell membrane permeability caused by its high molecular weight (1.297 Da) and large negatively charged surface areas (Herzig et al. 2012). Cells cultured in medium containing galactose showed a reduced metabolic capacity compared to 'standard' culture medium containing glucose ( Fig. 4A-D, p-values between assay are indicated). The observed significant reduction in cellular metabolic activity at 10 μM CATR may relate to the strong inhibitory capacity of CATR on AAC activity, reducing ADP import from the cytosol into the mitochondrial matrix. As a result, OXPHOS complex V is not active and the ATP remaining in Fig. 3 Quantitative assessment of mitochondrial morphology and function in ciPTEC-OAT1 and AAC3 −/− . ciPTEC-OAT1 (black), AAC3 −/− -1 (dark grey) and AAC3 −/− -2 (light grey) were matured and evaluated on a functional and morphological level (A). Citrate synthase (CS) activity was determined (B). In addition, cells were evaluated by fluorescence microscopic assessment using 25 nM tetramethylrhodamine methyl ester (TMRM), as described before (Iannetti et al. 2016). Co-staining with Hoechst was performed to correct for cell count. The number of mitochondria per cell (C), mitochondrial size (D), total mitochondrial area per cell (E), mitochondrial membrane potential (F), aspect ratio (a measure of mitochondrial length: ratio between major and minor axis) (G) and mitochondrial roundness (measure of mitochondrial length and degree of branching) (H) were evaluated. Significance was determined by one-way ANOVA with Dunnett's post hoc analysis, *p < 0.05 and **p < 0.01, mean ± SEM, N = 3 independent experiments the mitochondrial matrix may initiate reverse electron transfer (RET). Consequently, succinate dehydrogenase (complex II) could operate in the reverse mode. As MTT is a substrate of (mitochondrial) dehydrogenase activity, it may explain its reduced conversion. However, additional data supporting this hypothesis is lacking.
Since the effects of AAC inhibition were most pronounced in galactose medium, we further investigated the dose-dependent effects of all four AAC interactors in wild type and both AAC3 −/− ciPTEC clones (Fig. 5A-D). Our hypothesis was that AAC3 knockout cells would be less sensitive compared to wild type as they lack the pharmacological target (i.e., AAC). Comparison of wild-type cells with AAC3 −/− only revealed a statistical difference for CATR in AAC3 −/− -1. The lack of differences between knockouts and wild type suggests again that the other AAC isoforms could compensate for a possible decreased sensitivity.

Genetic and chemical AAC inhibition reduce mitochondrial respiration
To characterize the AAC3 −/− cells in more detail, we examined whether they recapitulated chemical inhibition of AAC, in particular the effects on mitochondrial respiration, as observed in Fig. 1F-H and Fig. 2A-C. Wild-type and AAC3 −/− ciPTEC were exposed to AAC inhibitors in 10 mM galactose medium to boost oxidative metabolism, followed by evaluation of mitochondrial respiratory capacity using the Seahorse XF Analyzer. All compounds, except for suramin, significantly reduced basal and maximal OCR in wild-type cells, compared to vehicle controls ( Fig. 6A and B). Interestingly, chemical AAC inhibition showed a significantly stronger reduction in OCR compared to the two vehicleexposed AAC3 −/− cell lines (p < 0.05), supporting the idea that the residual mitochondrial respiration observed in our genetically removed AAC3 model may be explained by functional compensation of the remaining AAC isoforms, as described above. This is further supported by the fact that mitochondrial transport carriers, including AAC, are known for their highly conserved residues in the substrate binding site and the large sequence homology (Bhatia et al. 2020;Kunji and Robinson 2006;Pebay-Peyroula et al. 2003;Robinson et al. 2008;Ruprecht et al. 2014. In addition, comparison of the isoforms demonstrated only minor differences in IC50-values for CD437 (Zhang et al. 2016). Therefore, chemical inhibition is not expected to Fig. 4 Concentration-dependent effects of AAC inhibitors on cellular metabolic activity in ciPTEC wild type using the glucose / galactose assay. Wild-type ciPTEC-OAT1 cells were matured and exposed to increasing concentrations of AAC inhibitors bongkrekic acid (BKA, A), CD437 (B), carboxyatractyloside (CATR, C), suramin (D) or 0.1% DMSO for 24 h in medium containing 10 mM glucose (light blue) or galactose (dark blue) at 37 °C and 5% (v/v) CO 2 , followed by measurement of cellular metabolic activity by means of MTT. All results were normalized to DMSO vehicle controls. Statistical analyses: two-way ANOVA corrected for multiple comparison by Bonferroni's post hoc analysis to compare differences in metabolic activity between medium composition conditions. p value corresponding to overall significance between media composition is indicated for each compound (p Glu/Gal ), *p < 0.05 and ****p < 0.0001. Mean ± SEM; N = 3 independent experiments 1 3 be selective for one isoform, resulting in equal reduction of AAC1, 2, 3, and 4 functions. Our genetic knockout model was designed to specifically target the most abundant isoform AAC3, which could result in functional compensation by AAC1, 2 and 4, as observed in our study. Next, we calculated the resulting respiratory spare capacity, for which a subtle, but stronger effect was observed in AAC3 −/− compared to chemically-inhibited wild-type cells (Fig. 6C). However, under all conditions there was no significant spare capacity when ETC was maximally stimulated compared to basal respiration levels, suggesting that metabolic compensation to restore cell function is impaired under energy-demanding conditions (viz. galactose and not glucose as carbon source). Moreover, specific deletion of AAC3 does not increase the sensitivity of cells for AAC inhibitors (Supplemental Fig. 5). Recapitulating the exact similar effect of chemical AAC inhibition in a genetic model might be challenging. Although in yeast a quadruple knockout has resulted in a viable cell (Hatanaka et al. 1999), it is expected to be technically difficult to generate a full knockout of all human isoforms.

Conclusions
In this study, we show that ciPTEC-OAT1 cells predominantly depend on oxidative metabolism for cellular function and that replacement of the energy source (viz. galactose instead of glucose) does not stimulate mitochondrial respiration. In addition, AAC is involved in maintaining mitochondrial OCR, as both basal and maximal OCR decrease upon genetic and chemical inhibition of AAC3. Particularly under energy demanding conditions, this may have consequences for cellular function. Chemical inhibition of AAC by established inhibitors reduces basal and maximal respiration even stronger compared to genetically inhibited AAC3, suggesting a role for functional AAC compensation in our AAC3 −/− model.
Acknowledgements The authors thank Dr. Eric Verschuren for sharing knowledge with respect to CRISPR/Cas9-mediated knockout generation, Dr. Rob Woestenenk for assistance with FACS cell sorting of GFP-positive cells and the microscopic imaging center and sequencing facility at Radboud University Medical Center for providing imaging and sequencing services.

Fig. 5
Concentration-dependent effects of AAC inhibitors on cellular metabolic activity in ciPTEC wild type and AAC3 −/− . Mature wild-type ciPTEC-OAT1 cells (black), AAC3 −/− -1 (dark grey) and AAC3 −/− -2 (light grey) were cultured, matured and exposed to increasing concentrations of AAC inhibitors bongkrekic acid (BKA, A), CD437 (B), carboxyatractyloside (CATR, C), suramin (D) or 0.1% DMSO for 24 h in medium containing 10 mM galactose, followed by measurement of cellular metabolic activity by means of MTT. All results were normalized to DMSO vehicle controls. Sta-tistical analyses: one-way ANOVA corrected for multiple comparison by Dunnett's post hoc analysis to compare differences between vehicle control and exposed conditions within a cell line: *p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001, indicated in the upper table of each graph) and two-way ANOVA corrected for multiple comparison by Bonferroni's post hoc analysis to compare overall significant differences between knockout cell line and wild type concentration-dependent responses, as indicated (p AAC3 Genetic and chemical AAC3 inhibition reduce mitochondrial respiration. Mature wild-type ciPTEC-OAT1 cells (black and orange/ red) were exposed to 0.1% DMSO (black) or 100 µM AAC inhibitors BKA (light orange), CD437 (middle orange), CATR (dark orange) and suramin (red) medium containing 10 mM galactose at 37 °C and 5% (v/v) CO 2 . Generated AAC3 knockouts (AAC3 −/− -1, dark grey and AAC3 −/− -2, light grey) were incubated with 0.1% DMSO in medium containing 10 mM galactose at 37 °C and 5% (v/v) CO 2 . After 12 h, mitochondrial respiration was evaluated using the Seahorse XF Analyzer. Oxygen consumption rates for all mitochondrial complexes were investigated. Data shown represents basal (A) and maximal respiration (B) and spare capacity (C). Data was corrected for cell count, assessed by fluorescence microscopy after Hoechst staining. Significance was determined by one-way ANOVA with Dunnett's post hoc analysis, *p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001, mean ± SEM, N = 3 independent experiments, comparing all conditions to wild type (black). Results of basal and maximal respiration and spare capacity shown in panels A-C for wildtype (black) and AAC3 −/− (dark and light grey) are also presented in Fig. 2