Mapping Adverse Outcome Pathways for Kidney Injury as a Basis for the Development of Mechanism-Based Animal-Sparing Approaches to Assessment of Nephrotoxicity

Mally, Angela; Jarzina, Sebastian

doi:10.3389/ftox.2022.863643

ORIGINAL RESEARCH article

Front. Toxicol., 15 June 2022

Sec. In Vitro Toxicology

Volume 4 - 2022 | https://doi.org/10.3389/ftox.2022.863643

This article is part of the Research Topic Adverse Outcome Pathways and New Approach Methodologies: Evolution, Opportunities, and Challenges View all 9 articles

Mapping Adverse Outcome Pathways for Kidney Injury as a Basis for the Development of Mechanism-Based Animal-Sparing Approaches to Assessment of Nephrotoxicity

Angela Mally*

Sebastian Jarzina

Department of Toxicology, University of Würzburg, Würzburg, Germany

In line with recent OECD activities on the use of AOPs in developing Integrated Approaches to Testing and Assessment (IATAs), it is expected that systematic mapping of AOPs leading to systemic toxicity may provide a mechanistic framework for the development and implementation of mechanism-based in vitro endpoints. These may form part of an integrated testing strategy to reduce the need for repeated dose toxicity studies. Focusing on kidney and in particular the proximal tubule epithelium as a key target site of chemical-induced injury, the overall aim of this work is to contribute to building a network of AOPs leading to nephrotoxicity. Current mechanistic understanding of kidney injury initiated by 1) inhibition of mitochondrial DNA polymerase γ (mtDNA Polγ), 2) receptor mediated endocytosis and lysosomal overload, and 3) covalent protein binding, which all present fairly well established, common mechanisms by which certain chemicals or drugs may cause nephrotoxicity, is presented and systematically captured in a formal description of AOPs in line with the OECD AOP development programme and in accordance with the harmonized terminology provided by the Collaborative Adverse Outcome Pathway Wiki. The relative level of confidence in the established AOPs is assessed based on evolved Bradford-Hill weight of evidence considerations of biological plausibility, essentiality and empirical support (temporal and dose-response concordance).

1 Introduction

There is general consensus that comprehensive understanding of how chemicals and drugs cause adverse effects is key to the development of human relevant, animal-sparing solutions for safety testing. Adverse Outcome Pathways (AOPs) represent a formal description of the mechanistic linkage between a molecular initiating event, a series of intermediate steps and key events at different levels of biological organization, and an adverse outcome. While the concept of mode-of-action (or AOPs) has been a key aspect of human cancer risk assessment for some time, it was recently adopted by the Organisation for Economic Co-operation and Development (OECD) as a pragmatic tool which may facilitate transition of chemical safety assessment from measurement of apical endpoints in animals to toxicity prediction based on mechanistic information (Vinken, 2013). Identification of key events and systematic mapping of AOPs for a given hazard endpoint can form the basis for the development of alternative tests (in vitro, lower organisms, refined in vivo) as part of a science-based integrated testing strategy to eventually replace conventional guideline studies (Tollefsen et al., 2014; Sakuratani et al., 2018). To promote implementation of this concept into chemical safety assessment, the OECD has published guidance documents for the development, assessment and reporting of AOPs as well as use of AOPs to support Integrated Approaches to Testing and Assessment (IATAs) (OECD, 2017b; OECD, 2017a).

The kidney and in particular the proximal tubule epithelium presents one of the key target sites of chemical-induced injury. A wide-range of structurally diverse chemicals, including heavy metals, fungal toxins and drugs are known to cause kidney injury. The particular susceptibility to toxic insult is due to a number of physiological factors that jointly lead to high renal exposure to xenobiotics present in the systemic circulation. The kidneys receive roughly 25% of the cardiac output and hence the rate of delivery of toxicants to the kidney is high. The kidneys’ ability to concentrate solutes further increases exposure of renal cells to xenobiotics (Pfaller and Gstraunthaler, 1998; Khan and Alden, 2002). Uptake of xenobiotics into kidney tubule cells and intrarenal bioactivation to toxic metabolites is facilitated by active transporters and drug metabolizing enzymes that are abundantly expressed particularly throughout the proximal tubule, which renders this segment of the nephron especially susceptible to toxicity (Pfaller and Gstraunthaler, 1998; Khan and Alden, 2002). Acute or chronic damage to proximal tubule cells can lead to kidney dysfunction and ultimately acute or chronic kidney failure. Moreover, chronic cytotoxicity and compensatory regenerative hyperplasia is a well-established mode-of-action by which some chemicals cause kidney tumor formation (Lock and Hard, 2004). Considering the varied chemical nature of nephrotoxic compounds (Pfaller and Gstraunthaler, 1998), it is perhaps not surprising that multiple mechanisms can lead to proximal tubule damage and loss of kidney function. Structurally diverse chemicals may interact with an equally diverse number of molecular targets, and these molecular initiating events (MIE) may each trigger a cascade of molecular and cellular events (Key events, KE) that ultimately result in kidney injury.

In line with recent OECD activities on the use of AOPs in developing Integrated Approaches to Testing and Assessment (IATAs), the overall aim of the present work was to contribute to building a network of AOPs leading to kidney injury through development and critical evaluation of AOPs. Here, we focused on three distinct mechanisms by which certain chemicals or drugs may cause nephrotoxicity and systematically captured the current mechanistic understanding in a formal description of AOPs in accordance with the harmonized terminology provided by the Collaborative Adverse Outcome Pathway Wiki (AOP Wiki; https://aopwiki.org/), a central repository for all AOPs developed as part of the OECD AOP Development Effort. The AOPs considered here are initiated by 1) inhibition of mitochondrial DNA polymerase γ (mtDNA Polγ), 2) receptor mediated endocytosis and lysosomal overload, and 3) covalent protein binding. Human and experimental data on selected chemical stressors for each AOP were identified via Pubmed literatur search and assembled to support the sequence of events leading to kidney injury. The relative level of confidence in the established AOPs was assessed based on evolved Bradford-Hill weight of evidence considerations of biological plausibility, essentiality and empirical support (temporal and dose-response concordance) provided by Becker et al. (2015) and OECD guidance documents for developing and assessing AOPs (OECD, 2017b; 2018) (Box 1).

BOX 1 |

Considerations of biological plausibility, essentiality, empirical support, and quantitative and temporal understanding of KERs, including criteria to define the weight of evidence (WoE) (Becker et al., 2015; OECD, 2017b; 2018) [not available in Crossref]

Considering that implementation of the AOP conceptual framework for translation of mechanistic data into regulatory decisions requires quantitative understanding of the relationships between key events within an AOP (Box 1), information on quantitative relationship between two pairs of KEs - as far as available–is assembled and data gaps that need to be filled in order to move from qualitative descriptions of AOPs to quantitative AOPs are highlighted.

2 Inhibition of Mitochondrial Deoxyribonucleic Acid Polymerase γ Leading to Kidney Toxicity (AOP-256)

This Adverse Outcome Pathway describes the sequential key events that link inhibition of mitochondrial DNA polymerase γ (mtDNA Polγ) to kidney toxicity. Nucleoside and nucleotide (nucleos(t)ide) analogs, which are widely used as antiviral drugs for the effective treatment of viral infections, including human immunodeficiency virus (HIV) and chronic hepatitis B virus infections, may act as chemical stressors for this pathway. As structural analogs of substrate nucleotides, these drugs act as chain terminators of viral DNA synthesis via competitive inhibition of reverse transcriptase or viral DNA polymerases, thereby blocking virus replication. Besides targeting viral enzymes, nucleos(t)ide antiviral agents may also interact with human DNA polymerases, which may lead to moderate to life-threatening adverse drug reactions, including peripheral neuropathy, myopathy, lactic acidosis, and acute and chronic kidney injury (Lewis and Dalakas, 1995; Johnson et al., 2001; Fontana, 2009; Fung et al., 2014).

2.1 Nephrotoxicity Associated With Long-Term Intake of Acyclic Nucleoside Phosphonates

The acyclic nucleoside phosphonates (ANPs) adefovir, tenofovir and cidofovir (Figure 1) were introduced into drug therapy of viral infections 15–20 years ago. Compared to existing antiviral drugs, this new class of antivirals offered a broad-spectrum activity against DNA viruses and retroviruses and lower risk of resistance development. However, long-term therapy with ANPs was subsequently found to cause renal proximal tubulopathy and even acute kidney injury. Based on its in vivo antiretroviral potency (Balzarini et al., 1989), adefovir [9-(2-phosphonylmethoxyethyl)adenine; PMEA] and its prodrug adefovir dipivoxil were originally developed for the treatment of HIV infections and cytomegaly virus (CMV) disease (James, 1997). While initial clinical studies reported effective antiretroviral activity and safety of adefovir dipivoxil (125 mg/d) in patients with advanced HIV infections (Deeks et al., 1997), a subsequent multi-center, randomized, double-blind and placebo-controlled trial in adult patients infected with HIV revealed an increased incidence of nephrotoxic effects in patients receiving adefovir (120 mg/d), characterized primarily by elevations in serum creatinine or hypophosphatemia (Kahn et al., 1999). Considering the risk for serious kidney toxicity in long-term use, the U.S. Food and Drug Administration (FDA) denied approval of adefovir for the treatment of HIV (Highleyman, 1999). However, at lower doses, adefovir dipivoxil was subsequently approved by the FDA for the treatment of hepatitis B, although safety concerns due to nephrotoxicity remain. Tenofovir [9-(2-Phosphonyl-methoxypropyly)adenine; PMPA], respectively its prodrug tenofovir disoproxil fumarate (TDF), obtained FDA approval for the treatment of HIV-1 infections in combination with other antiretroviral medicines in 2002, and subsequently for the treatment of chronic hepatitis B in adults in 2008. While tenofovir is now widely used as a first-line therapy against HIV and hepatitis B virus (HBV) infections, long-term treatment with tenofovir is associated with renal toxicity manifested as proximal tubule dysfunction, renal Fanconi syndrome or even acute kidney injury (Woodward et al., 2009; Agarwala et al., 2010; Herlitz et al., 2010; Gara et al., 2012; Hall, 2013; Sise et al., 2015). Cidofovir [(S)-l-(3-hydroxy-2-phosphonylmethoxypropyl)cytosine, HPMPC] is an acyclic nucleotide analog of deoxycytidine that was approved in 1996 for i.v. treatment of acquired immunodeficiency syndrome (AIDS) associated CMVretinitis in adults. Dose-dependent nephrotoxicity was found to be the major dose-limiting toxicity related to cidofovir treatment, and cases of acute renal failure were reported after treatment with as few as one or two doses (Lalezari et al., 1995; Group, 1997; The Studies of Ocular Complications of AIDS Research Group in collaboration with the AIDS Clinical Trials Group, 2000).

FIGURE 1

FIGURE 1. Chemical structure of the acyclic nucleoside phosphonates (ANPs) adefovir, tenofovir and cidofovir.

2.2 Mechanism of Acyclic Nucleoside Phosphonate Induced Nephrotoxicity

While the sequence of molecular events leading to ANP associated nephrotoxicity appears to present a universal mechanism by which nucleos(t)ide analogs may cause toxicity in a wide range of organs and tissues, including liver, heart, muscle and the nervous system (Lewis and Dalakas, 1995; Lewis et al., 2003; Fontana, 2009; Fung et al., 2014), the particular susceptibility of the kidney, respectively the proximal tubule, to ANP toxicity is a result of transporter-mediated uptake into proximal tubule cells, leading to intracellular accumulation of ANPs. Organic anion transporter 1 (OAT1) and to a lesser extent organic anion transporter 3 (OAT3) located at the basolateral membrane of proximal tubule cells are recognized as the major membrane carriers for uptake of ANPs (Figure 2) (Uwai et al., 2007; Zhang et al., 2015; Lash et al., 2018). In support of this, inhibition of basolateral membrane transporters has been shown to reduce ANP nephrotoxicity (Lalezari et al., 1995). In addition, recent data demonstrate ANP uptake into primary human kidney cells via apical carriers, potentially OAT4 or organic anion-transporting polypeptides (OATPs) (Lash et al., 2018). Thus, the tissue-specificity of ANP toxicity appears to be determined predominantly by toxicokinetics and renal handling of these drugs. Once taken up into kidney cells, the phosphonate analogs are transported across the mitochondrial membrane prior or subsequent to metabolic conversion into the active triphosphate form via nucleotide kinases present in mitochondria and the cytosol (Robbins et al., 1995; Lewis et al., 2003; Izzedine et al., 2005; Uwai et al., 2007; Kohler et al., 2011). While designed to inhibit viral reverse transcriptase and DNA polymerases with high efficiency, ANPs may also interact with human DNA polymerases, including mitochondrial DNA Pol γ, which is essential for mitochondrial DNA replication (Figure 2). Phosphorylated ANPs compete with endogenous deoxyribonucleotides for incorporation into DNA, thereby inhibiting mitochondrial DNA Pol γ and consequently mtDNA replication. As a result, mtDNA, which encodes 13 components of the electron transport chain essential to oxidative phosphorylation, is depleted. This leads to impaired mitochondrial function, i.e. reduced respiration, electron leakage and energy decline, and ultimately cell death (Figure 2) (Perazella, 2010; Fernandez-Fernandez et al., 2011). Although additional pathways are discussed as potential contributors to mitochondrial toxicity of nucleos(t)ide analogs (Apostolova et al., 2011), there is sufficient evidence from in vitro and in vivo studies in humans and rodents to support mitochondrial dysfunction as a consequence of inhibition of mt Pol γ dependent mtDNA replication as the primary mechanism of ANP induced proximal tubule injury.

FIGURE 2

FIGURE 2. Active transport of acyclic nucleoside phosphonates into proximal tubule cells predominantly via basolateral influx carriers (OATs and OAT3) and subsequent disruption of proximal tubular mitochondrial function due to inhibition of mitochondrial DNA Pol γ (modified from (Fernandez-Fernandez et al., 2011)). Adverse outcome pathway of inhibition of mitochondrial DNA Pol γ leading to kidney toxicity.

2.3 The Adverse Outcome Pathway of Inhibition of mtDNA Polymerase γ Leading to Kidney Toxicity

Based on these mechanistic considerations, the sequence of key events (KE) leading to kidney injury as an adverse outcome can be described as inhibition of mt Pol γ as the molecular initiating event (MIE), leading to mitochondrial DNA (mtDNA) depletion (KE1), mitochondrial dysfunction (KE2) and proximal tubule cell toxicity (KE3) (Figure 2). Evidence for inhibition of mitochondrial DNA polymerase γ leading to kidney toxicity as an adverse outcome primarily comes from experimental in vitro and in vivo studies on tenofovir, adefovir and cidofovir that serve as chemical stressors for this pathway, as well as from clinical trials and reports of patients treated with ANPs (Tables 1A–C). Collectively, these studies show a strong association between mitochondrial toxicity and ANP induced nephrotoxicity (Tanji et al., 2001; Cote et al., 2006; Kohler et al., 2009; Lebrecht et al., 2009; Herlitz et al., 2010; Ramamoorthy et al., 2014), with some studies also demonstrating concomitant mtDNA depletion (Tanji et al., 2001; Kohler et al., 2009; Lebrecht et al., 2009; Kohler and Hosseini, 2011). Additional support for mtDNA depletion and mitochondrial dysfunction as down-stream events of mt Pol γ inhibition is derived from studies on nucleos(t)ide analogs that induce mitochondrial toxicity in other target organs via the same principle mechanism. Moreover, there is a wealth of data that link point mutations in the gene encoding for the catalytic subunit of Pol γ with a wide range of human mitochondrial disorders that typically affect tissues with high energy requirement with varying symptoms and severity (Nurminen et al., 2017). In the following sections, evidence supporting the KEs and KE relationships (KERs) in this AOP will be presented, followed by a critical assessment of the AOP in terms of temporal and dose-response concordance, essentiality of key events, biological plausibility, coherence, and consistency of the experimental evidence.

TABLE 1A

TABLE 1A. Evidence from human, animal and in vitro studies on tenofovir or its prodrug tenofovir disoproxil fumarate (TDF) supporting the key events and qualitative concordance of KEs within this AOP (n/a = no data available).

TABLE 1B

TABLE 1B. Evidence from human, animal and in vitro studies on adefovir or its prodrug adefovir dipivoxil (ADV) supporting the key events and qualitative concordance of KEs within this AOP (n/a = no data available).

TABLE 1C

TABLE 1C. Evidence from human, animal and in vitro studies on cidofovir supporting the key events and qualitative concordance of KEs within this AOP (n/a = no data available).

Molecular Initiating Event: Inhibition of mtDNA Polymerase γ

As structural analogs of normal nucleotides that lack the 3′-OH group of the deoxyribose moiety, antiviral nucleos(t)ides were designed as alternative substrates for viral DNA polymerases that block virus replication by preventing chain-elongation. As an undesirable extension of their pharmacological action, antiviral nucleos(t)ides also interact with host DNA polymerases. Among the cellular replicative DNA polymerases, mitochondrial DNA Pol γ, which is responsible for maintenance of mtDNA, has been shown to be most sensitive to the inhibitory effects of these drugs, although nuclear DNA polymerases such as DNA polymerase α and β may also be affected. Numerous in vitro and in vivo studies document inhibitory effects of a wide range of nucleoside and nucleotide reverse transcriptase inhibitors on mtDNA Pol γ at concentrations achieved in vivo (reviewed in (Kakuda, 2000)). While there are significant differences in the ability of individual antiviral nucleos(t)ides to become incorporated into DNA by Pol γ, quantitative prediction of the overall inhibitory effect on mtDNA replication and subsequent mitochondrial toxicity is complicated by the 3′-5′exonuclease activity of Pol γ, which catalyses removal of incorporated nucleotides. This is exemplified by the case of zidovudine (3′-azido-3′-deoxythymidine), a drug that is a comparatively poor substrate for incorporation into mtDNA by Pol γ, which may still effectively block mtDNA replication due to inefficient excision of dideoxynucleotides and hence persistence in mtDNA (Lim and Copeland, 2001; Lim et al., 2003). In addition to the intrinsic 3′-5′exonuclease activity of Pol γ, a recent study also identified Pol β, previously thought to be exclusively located in the nucleus, as a major mtDNA repair enzyme (Prasad et al., 2017). Thus, the ability of nucleos(t)ide analogs to inhibit Pol β presents a further modifying events in this pathway, with progression to the next key event, i.e. mtDNA depletion, depending on the rate of nucleotide incorporation vs. the rate of removal by Pol γ (and presumably also by Pol β) relative to the time required to replicate mtDNA.

Using purified mtDNA Pol γ and activated calf thymus DNA as a primer template, adefovir, tenofovir and cidofovir were all shown to inhibit mammalian DNA polymerases α, β, and γ (Tables 1A–C). The kinetic inhibition constants (K_i) values of the diphosphates of the three nucleoside phosphonates against Pol γ were 0.97, 59.5 and 299 µM for adefovir, tenofovir and cidofovir, respectively, and with the exception of adefovir - significantly higher than K_i values of some of the other nucleos(t)ide analogs such as 2′,3′-dideoxycytidine (0.034 µM) or zidovudine (18.3 µM) (Cherrington et al., 1994; Cherrington et al., 1995). The kinetic inhibition constants against mammalian DNA polymerases β were 70.4, 81.7 and 520 µM for adefovir, tenofovir and cidofovir, respectively (Cherrington et al., 1994; Cherrington et al., 1995). The lower inhibitory activity of tenofovir and cidofovir against human DNA polymerases compared to adefovir and some of the other antiviral nucleos(t)ide analogs were considered to be in line with the relatively lower toxicity of tenofovir and cidofovir. Similarly, a toxicity index calculated based on single turnover kinetic studies using reconstituted human Pol γ holoenzyme to measure the rates of incorporation and exonuclease removal also suggested relatively low mitochondrial toxicity of tenofovir as compared to some other drugs, e.g., 2′,3′-dideoxycytidine (Johnson et al., 2001; Lee et al., 2003).

The inhibitory effects of antiviral drugs on Pol γ dependent mtDNA replication resemble mitochondrial genetic diseases associated with inactivating mutations in the gene encoding Pol γ. Pathogenic mutations in the catalytic subunit of Pol γ cluster into five distinct regions involving the active site, residues of the upstream DNA binding channel, and regions responsible for regulating polymerase vs. exonuclease activity and enzyme processivity (Nurminen et al., 2017). The clinical manifestations of Pol γ syndromes comprise a continuum of phenotypic abnormalities with varying degree of severity, age of onset and tissues affected. There is a close relationship between the age of onset and the severity of the symptoms, i.e. the earlier the onset, the more severe the condition. These range from prenatally-fatal to severe early childhood multi-system disorders such as Alpers-Huttenlocher syndrome (AHS), a progressive neurodegenerative disorder accompanied by disturbed hepatocellular function and tissue-specific DNA depletion (liver > skeletal muscle, heart) that progressively leads to psychomotor regression, epilepsy and liver failure, to adult-onset milder diseases such as progressive external ophthalmoplegia (Cohen et al., 1993; Nurminen et al., 2017). The latter initially presents with weakness of the eye muscles but may also involve other multisystemic features including generalized mitochondrial myopathy with ragged-red fibers, ataxia, axonal sensory-motor polyneuropathy, sensorineural hearing loss, depression, and lactic acidosis (Cohen et al., 1993). Childhood myocerebrohepatopathy spectrum (MCHS) is another Pol γ-related disorder that presents in the first few months of life with developmental delay, lactic acidosis, myopathy and further symptoms such as frequent vomiting, hearing loss, liver failure, pancreatitis and renal tubular acidosis (Cohen et al., 1993).

Key Event 1: mtDNA Depletion

As Pol γ is essential for mtDNA replication, a gradual decrease in mtDNA is an obvious and biologically plausible consequence of sustained inhibition of Pol γ. While there are no reports on the effect of cidofovir on mtDNA content (Table 1C), experimental in vitro and in vivo studies demonstrate reduced mtDNA copy numbers associated with decreased expression of proteins involved in mtDNA replication in kidney tubule cells in response to adefovir and tenofovir (Tables 1A, B). A reduction in the ratio of mitochondrial to nuclear DNA was also reported in proximal tubule cells of an HIV-infected individual maintained on highly active antiretroviral therapy that included adefovir dipivoxil for 7 months (Tanji et al., 2001). In contrast to these studies, Birkus et al. found no effect of tenofovir on mtDNA content in HepG2 cells, skeletal muscle cells and human renal proximal tubule epithelial cells (Birkus et al., 2002). It needs to be emphasized, however, that uptake of tenofovir into cells was not verified in this study, either by directly measuring intracellular levels of tenofovir or by characterization of cells with regard to expression of relevant drug transporters. Considering that tenofovir toxicity depends on transporters that mediate cellular uptake (Uwai et al., 2007; Zhang et al., 2015; Lash et al., 2018), it is questionable if sufficiently high intracellular concentrations to inhibit Pol γ and block mtDNA replication were achieved in this model.

The causal relationship between inhibition of Pol γ and loss of mtDNA is further supported by studies investigating the mechanism of toxicity of nucleos(t)ide analogs in other cells and tissues. For instance, a significant reduction in mtDNA was observed in muscle biopsies of zidovudine-treated HIV positive patients with myopathy as compared non-HIV-patient controls (Arnaudo et al., 1991). Inhibition of mtDNA synthesis and loss of cell number was also observed in a T-lymphoid leukemic cell line (Molt-4) treated with several anti-HIV and anti-HBV nucleoside analogs (d4T, 3′-deoxy-2′,3′-didehydrothymidine; FLT, 3′-fluoro-3′-deoxythynidine; ddC, 2′,3′-dideoxycytidine), which were also identified as potent inhibitors of Pol γ. Similar to the study on tenofovir by Birkus et al. (Birkus et al., 2002), a number of potent Pol γ inhibitors did not cause significant effects on mtDNA synthesis and cell viability (Martin et al., 1994). Based on these findings, the authors concluded that there was no clear quantitative or qualitative correlation between the inhibition of isolated Pol γ and inhibition of mitochondrial DNA synthesis in vitro, and moreover that these data are not predictive of in vivo toxicity (Martin et al., 1994). It is however important to stress that toxicokinetics, most notably cellular uptake of the tested antivirals, were not considered in this assessment. Thus, it is possible that some of the most potent inhibitors of Pol γ identified in a cell-free assay failed to induce mtDNA depletion and cytotoxicity in this cell model simply because of insufficient cellular uptake (Martin et al., 1994).

Experimental evidence for functional inhibition of mitochondrial Pol γ as the underlying cause of mtDNA depletion and associated mitochondriopathies also comes from animal studies. Functional knockout of Pol γ in mice leads to complete loss of mtDNA and embryonic lethality in mice (Hance et al., 2005). Similarly, Pol γ function was demonstrated to be essential for maintenance of mtDNA and development in Drosophila melanogaster (Iyengar et al., 2002). In contrast, polg−/− mutant zebrafish carrying mutations within the polymerase domain survived up to 4 weeks post-fertilization, but showed delayed growth and regenerative defects accompanied by a gradual decrease in mtDNA that correlated with impaired basal and maximal FCCP (carbonyl cyanide p-trifluoro-methoxyphenyl hydrazone)-uncoupled respiration (Rahn et al., 2015). This study also revealed tissue specific differences in the basal levels of mtDNA copy numbers per cell in wildtype animals, with the tail region of zebrafish containing higher levels of mtDNA compared to the region containing the gills, heart and internal organs, and yet lower levels in the central nervous system (CNS) region containing eyes and the brain. Moreover, the degree of mtDNA depletion upon Pol γ knockout was shown to differ between tissues, with the tissue most severely depleted being the organ fraction (mtDNA content in polg−/− 14% of wildtype), followed by CNS (38% of wildtype) and finally tail as the least affected region (52% of wildtype) (Rahn et al., 2015). Considering the finding that the organ fraction, which contains e.g. the liver, along with CNS were most affected by Pol γ knockout, it was suggested that the polg−/−zebrafish model closely resembles human Pol γ-associated mitochondrial diseases that typically present with first symptoms in organs with high energy demand, i.e., CNS and the liver.

Differences in the rate of mtDNA synthesis between tissues are also likely to be an important determinant of tissue-specific responses to Pol γ inhibition. In vivo mtDNA labeling with BrdU in adult wild-type mice showed that BrdU was more rapidly incorporated into mtDNA in the brain as compared to the liver (Fuke et al., 2014), suggesting more rapid mitochondrial biogenesis in the brain.

Key Event 2: Mitochondrial Dysfunction

Although the vast majority of proteins localized in mitochondria is encoded by nuclear DNA, the mitochondrial genome is essential for oxidative energy metabolism as all 13 polypeptides coded for by mitochondrial genes are subunits of complexes of the respiratory chain/oxidative phosphorylation system. It is therefore inevitable that depletion of mtDNA leads to mitochondrial dysfunction. Evidence for mitochondrial dysfunction as a key event in this AOP comes from human, animal and in vitro studies treated with ANPs as chemical stressors for this pathway (Tables 1A–C). Mitochondrial changes indicative of impaired mitochondrial function are typically described as mitochondrial enlargement with disrupted crystal architecture and reduced activity of mtDNA-encoded respiratory chain subunits. These alterations were frequently reported to occur concomitant with mtDNA depletion and cytotoxicity, further supporting a direct link between upstream and downstream key events (Kohler et al., 2009; Lebrecht et al., 2009; Kohler and Hosseini, 2011; Zhao et al., 2017). Likewise, studies in Pol γ deficient animal models demonstrate a close correlation between loss of mtDNA induced by Pol γ inactivation and altered mitochondrial function (e.g. impaired basal and maximal FCCP-uncoupled respiration (Rahn et al., 2015)). Although genetic knockdown of Pol γ typically affects different tissues than ANPs which specifically target the kidney in a transporter-dependent manner, these studies provide substantial evidence for a causal relationship between mtDNA depletion and mitochondrial dysfunction.

Key Event 3: Cytotoxicity

Mitochondria are not only critical for cellular metabolism and energy production that are fundamental to cell viability, they also act as signalling organelles and as such play a key role in cellular life-and-death decisions. Mitochondria participate in both the extrinsic and intrinsic pathway of apoptosis, the latter of which involves opening of the mitochondrial outer membrane and subsequent release of pro-apoptotic factors such as cytochrome C from mitochondria. Interference with the energy-producing function of mitochondria, e.g. through impairment of oxidative phosphorylation as a result of decreased mtDNA content, leads to adenosine triphosphate (ATP) depletion and consequently disturbed cellular function that culminates in necrosis.

While there are multiple mechanisms by which drugs and chemicals can target mitochondria and impair mitochondrial ATP synthesis (e.g. uncoupling of the mitochondrial respiratory chain, inhibition of ATP synthesis, damage to mtDNA, interference with mtDNA replication), mitochondrial toxicity is well established as a key cause of toxicity of a wide range of drugs and chemicals that affect different target organs, including the liver, heart, skeletal muscle, central nervous system, and the kidney. Similarly, mitochondrial dysfunction caused by inherited or sporadic mutations in mtDNA is considered to play a critical role in the pathogenesis of a range of diseases, including acute and chronic kidney injury that involve damage to the proximal tubule (Martin-Hernandez et al., 2005; Emma et al., 2012; Che et al., 2014; Hall and Schuh, 2016). Within the kidney, tubule cells and particularly those of the proximal tubules, are particularly vulnerable to mitochondrial dysfunction. Active transport of solutes in the proximal tubule requires large amounts of ATP generated predominantly via mitochondrial oxidative phosphorylation. To meet the high energy demand, proximal tubule cells contain numerous large mitochondria. While it is generally acknowledged that mitochondrial dysfunction may lead to activation of cell-death pathways, evidence for a mechanistic link between mitochondrial dysfunction caused by inhibition of mtDNA Pol γ and proximal tubule toxicity comes from in vitro and in vivo studies that demonstrate loss of cell viability, dilated proximal tubules and degenerative changes affecting proximal tubules in experimental animals and humans treated with ANPs (Tables 1A–C).

Adverse Outcome: Kidney Toxicity

Through excretion of metabolic wastes and regulation of acid-base balance, electrolyte concentrations and extracellular fluid volume, the kidney plays a key role in maintaining whole-body homeostasis. Functional integrity of the proximal tubule, which contributes to fluid, electrolyte, and nutrient homeostasis by reabsorbing approximately 60–80% of filtered solute and water as well as virtually all of the filtered nutrients (e.g., glucose and amino acids) and low-molecular-weight proteins, is critical for whole-kidney function. Consequently, injury to the proximal tubule will lead to a decline in kidney function, although minor proximal tubule changes may not cause significant effects on renal function due to the kidney’s functional reserve and capacity to regenerate. Numerous drugs and chemicals are known to cause nephrotoxicity primarily by killing proximal tubule cells. Depending on the nature and severity of the insult, altered tubule or whole kidney function may be evidenced by altered renal handling of electrolytes (e.g., sodium, phosphate, calcium, bicarbonate), an increase in urinary glucose, amino acids and low-molecular-weight proteins indicative of impaired tubular reabsorption, and a rise in blood urea nitrogen (BUN) and serum creatinine (sCr). Such changes are evident in experimental animals and patients treated with ANPs that act as chemical stressors for this AOP (Tables 1A–C). For instance, tubular proteinuria and increased urinary phosphate, potassium, and bicarbonate excretion accompanied by reduced serum phosphate and potassium were observed in rats treated with TDF at 600 mg/kg bw for 5 weeks (Ramamoorthy et al., 2014). In another study in rats, a dose-dependent decrease in creatinine clearance was observed (Liborio et al., 2008). Reduced creatinine clearance accompanied by increased sCr and BUN were also reported in mice treated with TDF and adefovir dipivoxil (Zhao et al., 2017). In humans, kidney toxicity associated with intake of ANPs is predominantly characterized by glucosuria and proteinuria, hypophosphatemia, and increased sCr (Kahn et al., 1999; Izzedine et al., 2004; Woodward et al., 2009; Agarwala et al., 2010; Herlitz et al., 2010; Gara et al., 2012; Lin et al., 2017; Vora et al., 2017).

2.4 Assessment of the Adverse Outcome Pathway of Inhibition of mtDNA Polymerase γ Leading to Kidney Toxicity

The relative level of confidence in the overall AOP was assessed based on evolved Bradford-Hill weight of evidence considerations provided by Becker et al. (2015) and OECD guidance documents for developing and assessing AOPs (OECD, 2017b; OECD, 2018) (Box 1).

Biological Plausibility

Mitochondrial DNA replication relies on Pol γ activity. As detailed in Section 2.3, sustained inhibition of Pol γ inevitably leads to reduced mtDNA synthesis and in consequence to a gradual decrease in mtDNA. Loss of mtDNA is thus an obvious and biologically plausible consequence of inhibition of Pol γ (Table 2). All 13 polypeptides encoded by mtDNA are subunits of complexes of the respiratory chain/oxidative phosphorylation system that are required for maintaining mitochondrial function. Biological plausibility for the KER between depletion of mtDNA and mitochondrial dysfunction is therefore considered high (Table 2). It is also well established that mitochondrial function is vital for cell survival, particularly in cells with a high energy demand such as proximal tubule cells. Finally, it is well established and supported by an extensive body of evidence that proximal tubule cell injury impairs kidney function. Thus, the level of confidence in the biological plausibility of all key event relationships (KERs) within the proposed AOP can be considered as high (Table 3).

TABLE 2

TABLE 2. Dose-Time Concordance of KEs based on rodent studies with tenofovir disoproxil fumarate (TDF) as a specific stressor for the adverse outcome pathway of inhibition of mitochondrial DNA polymerase γ leading to kidney toxicity (n.d. = not determined; n/a = no data available).

TABLE 3

TABLE 3. Weight-of-evidence analysis of KERs in the adverse outcome pathway of inhibition of mitochondrial DNA polymerase γ leading to kidney toxicity.

Essentiality of Key Events

While there are numerous studies to demonstrate that blocking Pol γ function via pharmacological inhibition, genetic knock-out or mutational inactivation is detrimental to cells as it leads to mtDNA depletion and reduced mitochondrial function, there do not appear to be any experimental studies to investigate if restoration of Pol γ function (e.g., through overexpression of Pol γ) maintains mtDNA copy numbers and mitochondrial function. However, studies on arterial aging in mice show that restoring mtDNA copy numbers through overexpression of the mitochondrial helicase Twinkle (Tw⁺) preserves arterial mitochondrial respiration in aging mice (Foote et al., 2018). Similarly, in a mouse model of volume overload-induced heart failure, increased mtDNA copy numbers in transgenic mice overexpressing human transcription factor A of mitochondria (TFAM) or Twinkle helicase afforded cardioprotection through maintaining mitochondrial enzymatic activities (Ikeda et al., 2015). These data provide experimental support for the essentiality of mtDNA copy number for mitochondrial respiration. Maintaining mitochondrial function has been recognized as a promising therapeutic target for the treatment of acute kidney injury (Hall and Schuh, 2016). Strategies to increase mitochondrial biogenesis, e.g., through activation or overexpression of peroxisome proliferator-activated receptor-γ coactivator-1α (PGC-1α) that acts as a master regulator of mitochondrial biogenesis, have been demonstrated to restore mitochondrial activity and/or kidney function in ischemia-reperfusion induced renal injury or drug-induced acute kidney injury (Sivarajah et al., 2003; Rasbach and Schnellmann, 2007; Liborio et al., 2008; Funk and Schnellmann, 2013; Jesinkey et al., 2014; Jesse et al., 2014). Collectively, these studies provide good evidence that mitochondrial dysfunction and kidney toxicity as the adverse outcome can be prevented by maintaining mtDNA levels, which requires mitochondrial biogenesis and hence Pol γ mediated mtDNA replication. Moreover, recovery of kidney function was reported in patients after discontinuation of ANP treatment (Kahn et al., 1999; Woodward et al., 2009; Herlitz et al., 2010; Gara et al., 2012). Based on direct evidence illustrating essentiality for at least one of the important KEs, the level of confidence for essentiality of KEs in this AOP can thus be considered as high (Table 2).

Empirical Evidence: Dose-Response and Temporal Concordance

Overall, there is only very limited data on dose-related effects of ANPs in vitro and in vivo. In HK-2 cells and primary mouse renal tubule cells treated with TDF and ADV at 300 and 1,000 µM for 5 days, irregularly shaped mitochondria accompanied by reduced mitochondrial respiration and ATP production were observed at both concentrations (Zhao et al., 2017). While treatment with 1,000 µM TDF and ADV resulted significant inhibition of cell growth, decrease in cell viability and induction of apoptosis, treatment with 300 µM of each ANP did not significantly affect cell viability (Zhao et al., 2017). These data provide evidence that mitochondrial dysfunction (KE2) occurs at equal and lower concentrations of adefovir and tenofovir than cytotoxicity (KE3). Unfortunately, mtDNA copy number and expression of proteins involved in mtDNA replication were only investigated at 1,000 µM of TDF and ADV, at which they were significantly altered. Although these data show that all KE were impacted at the same concentration, it is not possible to conclude that KE1 (mtDNA depletion) occurs at lower concentration than the downstream KEs. There are no in vivo studies investigating dose-related effects of ANPs on the proposed KEs, i.e., all studies conducted in experimental animals so far are limited to a single dose group per study (Table 3). Cross-study comparison to establish dose-response concordance is not possible due to variations in experimental design, including species, strain, dose and treatment duration. In studies in mice given TDF at a dose of 5 mg/kg bw for 5 weeks, loss of mtDNA and mitochondrial dysfunction were observed in the absence of proximal tubule injury and impaired kidney function (Kohler et al., 2009; Kohler and Hosseini, 2011), suggesting that either higher doses or prolonged treatment may be required to trigger the final KE and the AO in this AOP (Table 3). In another study, effects on mtDNA replication machinery, proximal tubule injury and kidney toxicity were evident in mice given TDF at 10 mg/kg bw for 8 weeks, while mitochondrial function was not assessed (Table 3) (Zhao et al., 2017). In rats, all 3 KE and kidney toxicity as the adverse outcome were observed after treatment with TDF (100 mg/kg bw) for 8 weeks (Table 3) (Lebrecht et al., 2009). Mitochondrial dysfunction, proximal tubule injury and kidney toxicity were also evident in rats given TDF at 600 mg/kg bw for 5 weeks, but mtDNA copy number was not assessed in this study (Table 3) (Ramamoorthy et al., 2014). Collectively, the available in vitro and in vivo studies conducted using chemical stressors for this AOP generally demonstrate effects on KEs across the entire AOP at equal doses/concentrations of the stressor, with some evidence for upstream events occurring at lower concentrations than downstream KEs. There are no data that would disagree with the assumption of dose-response concordance.

As can be seen from Table 3, there are no in vivo time-course studies on tenofovir or other specific stressors for this AOP, and hence temporal concordance for the entire sequence of events cannot be demonstrated. However, there is a large body of evidence to demonstrate that changes in mitochondrial bioenergetics and dynamics precede proximal tubule damage in kidney injury induced by nephrotoxic drugs and chemicals (Lock et al., 1993) as well as in diabetic kidney disease (Coughlan et al., 2016).

Based on the criteria for assessing AOP (Box 1), evidence for dose-response concordance with relevant stressors but lack of time-course studies to demonstrate temporal concordance, the level of confidence for empirical support for the KERs in this AOP is considered moderate (Table 2).

Weight-Of-Evidence Analysis

Based on the high level of confidence in the biological plausibility of KERs, strong support for essentality of the KEs provided i.e. by experimental studies demonstrating mtDNA depletion and reduced mitochondrial function in response to genetic knock-out or mutational inactivation of Pol γ, and moderate empirical support for the KER in this AOP, the overall weight-of-evidence of this AOP can be considered as high (Table 2).

2.5 Quantitative and Temporal Understanding of Key Event Relationships

Based on the available literature, there is at present little or no quantitative information on the response-response relationship between two pairs of KEs in the AOP of inhibition of Pol γ leading to depletion of mtDNA, but experiments are underway within the Risk-IT project to define these.

KER1: Inhibition of Pol γ Leading to Depletion of mtDNA

As outlined above, the quantitative relationship between inhibition of Pol γ and mtDNA depletion is still poorly defined. Efforts to predict a compounds inhibitory effect on mtDNA replication based on its inhibitory activity against Pol γ did not yield satisfactory results. In establishing response-response relationships, several aspects that determine the KER need to be considered. Firstly, inhibitory effects on Pol γ are typically assessed in cell-free systems, whereas studying effects on mtDNA replication require intact cells. Nominal concentrations added to a cell culture system may not adequately reflect concentrations at the molecular target, e.g., due to active transport, drug metabolism, or binding to plastic. Thus, adequate understanding of the in vitro toxicokinetics of the chemical stressors is needed to extrapolate from a cell-free to a cell-based assay. Second, the overall effect of a chemical stressor on Pol γ mediated mtDNA replication as the downstream KEs depends on its persistence in mtDNA, which is a function not only of the stressors ability to bind to Pol γ and become incorporated into mtDNA (feed-forward loops), but also on the rate of excision of nucleotides by the intrinsic proofreading 3′-5′exonuclease activity of Pol γ (Johnson et al., 2001; Lim and Copeland, 2001; Lim et al., 2003), which presents a feed-back loop (Figure 3). Moreover, recent evidence suggest that Pol β also contributes to base excision repair in mammalian mitochondria (Prasad et al., 2017). Thus, the ability of a chemical stressor to inhibit Pol β is also likely to influence its persistence in mtDNA and thus its overall effect on mtDNA replication. Finally, the rate of mitochondrial biogenesis is a critical determinant. If the stressor is removed more quickly from mtDNA than is required for mtDNA to replicate, mtDNA copy number may not be affected. Similarly, exposure to a stressor for only a short period of time may not be sufficient to trigger mtDNA depletion and subsequent mitochondrial toxicity in this AOP. The maximum lifetime of mitochondria in the kidney cortex has been estimated to be 15 days (Pfeifer and Scheller, 1975). This may explain why effects on KEs downstream of Pol γ inhibition in kidneys of patients and experimental animals generally occur only after continuous exposure for several weeks. This temporal delay between the MIE and the first KE in this AOP is also important to consider when developing in vitro test related to KEs in this AOP.

FIGURE 3

FIGURE 3. Schematic presentation of a universal AOP of Pol γ inhibition leading to adverse outcomes in kidney and extrarenal tissues, with predicted exposure at target site, feed-forward and feedback mechanisms, and potential modulating factors as determinants of the tissue-specific biological response to a chemical stressor.

While it is known that mtDNA content and rates of mitochondrial biogenesis differ between tissues, presumably due to tissue-specific energy demands, recent evidence suggests that there are also tissue-specific differences in the mode of mtDNA replication (Herbers et al., 2018). Since the sequence of KEs in this AOP is also relevant to nucleos(t)ide analog toxicity in extrarenal tissues including liver, heart, muscle and the nervous system, tissue-specific rates and modes of mitochondrial biogenesis may be important determinants of the tissue-specific downstream response to a particular nucleos(t)ide analog in addition to toxicokinetic factors as previously discussed (Figure 3).

Moreover, replication of mtDNA and mitochondrial biogenesis are complex processes that are regulated by a range of factors, including nitric oxide (NO), sirtuins, mitogen-activated protein kinase (p38 MAPK), AMP-activated protein kinase (AMPK) and calcium/calmodulin-dependent protein kinase IV (CaMKIV) (Jornayvaz and Shulman, 2010). This suggest that temporal or inter-individual variation in the activity of these pathways may act as modulating factors of the relationship between MIE and KE1. Estrogens are also known to be involved in the control of mitochondrial biogenesis, and thus sex-differences in the KER may exist (Klinge, 2017; Ventura-Clapier et al., 2017).

KER2: Depletion of mtDNA Leading to Mitochondrial Dysfunction

It is plausible to assume that the level of depletion of mtDNA required to cause mitochondrial toxicity may be cell- and tissue-specific, with metabolically active cells such as kidney tubule cells being most susceptible. There is a lack of quantitative information on the extent of mtDNA depletion required to induce a significant change in mitochondrial function.

While energy decline is thought to be an immediate consequence of mtDNA depletion, it may further increase mitochondrial stress through generation of oxidative stress that may cause mutations in the mitochondrial genome. It is also possible that an increased mtDNA mutation load due to impaired proofreading activity of Pol γ may contribute to mitochondrial dysfunction. Moreover, nucleos(t)ide analogs often also interact with Pol β, which plays a key role in mtDNA repair and maintenance of mitochondrial genome stability (Prasad et al., 2017). Thus, it needs to be considered that a further pathway initiated by inhibition of DNA Pol β by the very same chemical stressor, leading to increased mtDNA mutations and subsequently altered mitochondrial function, may combine with Pol γ inhibition to cause mitochondrial disturbance (Figure 3).

There are also numerous factors independent of chemical stressors of this AOP (e.g., age, genetic make-up, disease, drug-induced mitochondriopathies, co-exposure) that may affect mitochondrial function and increase the susceptibility of mitochondria to mitotoxicity induced by mtDNA depletion. These modulating factors are depicted in Figure 3.

KER3: Mitochondrial Dysfunction Leading to Cytotoxicity

Mitochondrial dysfunction is characterized by a reduced efficiency of oxidative phosphorylation and reduced synthesis of high-energy molecules, such as adenosine-5′-triphosphate (ATP). Expression of toxicity in response to a decline in mitochondrial function may be influenced by the cellular dependence on mitochondrial function, which is known to vary between tissues. Clearly, proximal tubule cells depend on cellular respiration and mitochondrial ATP production to fuel active transport of solutes. However, there is no systematic assessment as to how much decline in mitochondrial function or ATP depletion may be tolerated by a proximal tubule cell before it commits to apoptosis or necrosis. Rather, assays determining mitochondrial activity such as the [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide] (MTT) assay (succinate dehydrogenase activity) or ATP content (e.g., Cell Titer-Glo^®) are widely used as cytotoxicity assays based on the assumption that mitochondrial activity is related to the number of viable cells. On the other hand, there are numerous studies employing more than one cytotoxicity assay that show poor correlation between cell viability assays measuring mitochondrial activity vs. other end-points such as compromised plasma membrane integrity (e.g., lactate dehydrogenase (LDH) release), with mitochondrial activity assays generally reported to be more sensitive indicators of cytotoxicity than other endpoints (Fotakis and Timbrell, 2006; Xu et al., 2008). This is consistent with mitochondrial dysfunction being a distinct KE that generally precedes cell demise (Vinken and Blaauboer, 2017). It is important to point out that antivirals as chemical stressors for this AOP may have more than one target and thus several independent pathways may contribute to the overall outcome. Integration into nuclear DNA, telomere shortening, nuclear DNA hypermethylation, interference with ATP synthesis and transport of nucleotides into cellular compartments have all been suggested as possible mechanisms unrelated to mitochondrial dysfunction due to mtDNA Pol γ inhibition (Ho et al., 2000; Moyle, 2000). Biotransformation of cidofovir has been shown to give rise to cidofovir-phosphocholine, which may interfere with synthesis or degradation of membrane phospholipids based on its structural similarity with arabinofuranosyl-cytosine 5′-diphosphocholine (Eisenberg et al., 1998; Ho et al., 2000).

KER4: Cytotoxicity (Proximal Tubule Cell) Leading to Kidney Toxicity (Impaired Kidney Function).

Due to the functional reserve of the kidney, homeostasis may be maintained even in the presence of severe kidney damage. It is generally accepted that 70–80% of the renal epithelial mass must be lost before significant changes in serum creatinine (sCr) and blood urea nitrogen (BUN) occur (Pfaller and Gstraunthaler, 1998; Amin et al., 2004). Moreover, even though sCr and BUN are widely used as indicators of renal function in the clinic and in preclinical safety assessment, they are recognized as relatively insensitive markers that only start to rise when renal function is significantly impaired (approximately 50%) (Pfaller and Gstraunthaler, 1998; Amin et al., 2004). Markers related to renal handling of electrolytes, glucose and proteins, including urinary low-molecular weight proteins such as β₂-microglobulin, cystatin C and neutrophil gelatinase-associated lipocalin (NGAL) may be more sensitive indicators of proximal tubule function, but despite numerous in vivo studies investigating nephrotoxic effects of drugs and chemicals, there does not appear to be a systematic quantitative assessment as to the extent of proximal tubule injury required to cause significant changes in these markers. However, a multiscale mathematical model was recently developed and applied to prediction of gentamicin-induced kidney injury based on urinary excretion of kidney injury molecule-1 (Kim-1) (Gebremichael et al., 2018) (see Section 3.5). The authors suggest that the developed model should be generalizable to proximal tubule injury induced by nephrotoxins irrespective of their primary mechanism.

3 Receptor-Mediated Endocytosis and Lysosomal Overload Leading to Kidney Toxicity (AOP-257)

This Adverse Outcome Pathway describes the sequence key events that link receptor mediated endocytosis and lysosomal overload to kidney toxicity. Polybasic drugs and compounds with peptidic structure (e.g., aminoglycosides, glycopeptides, polymyxins) (Figure 4), as well as urinary proteins that act as ligands for multiligand, endocytic receptors (megalin, cubilin) expressed at the brush-boarder of renal tubule cells are efficiently taken up into proximal tubule cell via receptor-mediated endocytosis (Moestrup et al., 1995; Khan and Alden, 2002; Verroust et al., 2002; Thevenod, 2003; Schnellmann, 2013; Liu W. J. et al., 2015). Due to low lysosomal pH, endocytosed compounds may be trapped within lysosomes and accumulate in this organelle, leading to disruption of lysosomal function and eventually permeabilization of lysosomal membranes with release of reactive oxygen species and cytotoxic lysosomal enzymes (Figure 5) (Khan and Alden, 2002; Schnellmann, 2013; Liu D. et al., 2015).

FIGURE 4

FIGURE 4. Chemical structures of tobramycin as an example of an aminoglycoside, vancomycin as a glycopeptide, and polymyxin B and its less toxic analogue polymyxin B nonapeptide.

FIGURE 5

FIGURE 5. Uptake of polybasic drugs and compounds with peptidic structure via multiligand, endocytic receptors (megalin, cubilin) expressed at the brush-boarder of renal tubule cells. Endocytosed compounds may accumulate within lysosomes, leading to lysosomal swelling and disruption of lysosomes. Release of reactive oxygen species, cytotoxic lysosomal enzymes and possibly also endocytosed compounds may trigger cell death via necrosis or apoptosis. This may occur as a direct consequence (e.g., via massive proteolysis) or secondary to mitochondrial dysfunction. Adverse outcome pathway of receptor mediated endocytosis and lysosomal overload leading to kidney toxicity.

3.1 Nephrotoxicity Associated With Lysosomal Accumulation of Ligands of Endocytic Receptors Expressed at the Brush-Boarder of Proximal Tubule Cells

A pivotal function of the renal tubule epithelium is to reabsorb and recycle filtered proteins, carrier-bound vitamins and trace elements from the glomerular ultrafiltrate in order to retrieve nutrients (Eshbach and Weisz, 2017). Cells lining the proximal convoluted tubule are highly specialized for apical endocytosis due to abundant expression of the multiligand endocytic receptors megalin and cubilin at their brush-boarder. Megalin and cubilin ligands include vitamin carrier proteins (e.g., Vitamin D-binding protein, retinol binding protein, transcobalamin), other carrier proteins (e.g., albumin, hemoglobin, liver-type fatty acid-binding protein, metallothionein, transferrin), lipoproteins, enzymes and enzyme inhibitors, immunoglobulin light chains, as well as a number of drugs, including aminoglycosides and polymyxins (Eshbach and Weisz, 2017). Under physiological conditions, endocytosed physiological substrates are efficiently degraded by lysosomal proteases. However, lysosomal swelling and rupture leading to damage of proximal tubule cells may occur if lysosomal degradation is blocked or overwhelmed by substrate overload (Figure 5).

Overproduction of proteins or glomerular injury resulting in increased glomerular permeability and leakage of plasma proteins into urine leads to excessive protein reabsorption and overload of proximal tubule cells. Sustained proteinuria is thus recognized not only as a marker of renal dysfunction but also as a key mediator of tubular injury that contributes to progression of a range of human kidney diseases (Zoja et al., 2004; Erkan et al., 2005). Direct evidence for tubulotoxic effects of excess urinary protein comes from experimental models of albumin or light chain (LC) protein overload nephropathy (Eddy, 1989; Liu et al., 2008; Fang et al., 2018) as well as from in vitro studies with isolated kidney tubule cells (Erkan et al., 2001; Li et al., 2008).

Glycopeptide antibiotics, aminoglycosides and polymyxins (Figure 4) are drugs that exhibit bactericidal activity against Gram-positive and/or Gram-negative bacteria. Due to their polybasic structure, these drugs mimic endogenous ligands of endocytic receptors and highjack the endocytic system to enter proximal tubule cells (Avedissian et al., 2019). Polymyxins were first discovered in 1947 as antimicrobial agents with high activity against Gram-negative bacteria, including Pseudomonas aeruginosa, Klebsiella pneumoniae and Haemophilus influenzae (Stansly et al., 1947) based on their ability to disrupt the outer and inner membranes of Gram-negative bacteria after binding to lipopolysaccharide (LPS). Clinical signs of nephrotoxicity such as albuminuria and increased blood non-protein nitrogen were soon after reported in patients administered polymyxin B (Stansly, 1949), limiting their clinical use by the early 1970s. Despite the high risk of nephrotoxicity with conventional doses, polymyxins play an increasingly important role today as salvage therapy of life-threatening multidrug-resistant bacterial infections (Wertheim et al., 2013). The incidence of acute kidney injury following systemic polymyxin therapy with polymyxin B or colistin is reported to range beween 18 and 61% (Justo and Bosso, 2015).

In contrast to polymyxins, the glycopeptide vancomycin and aminoglycosides are mainstream therapy of serious bacterial infections. However, drug-induced kidney injury is a common and dose-limiting adverse effect of vancomycin and aminoglycoside antibiotic treatment. The incidence of acute kidney injury in patients treated with aminoglycoside antibiotics is reported to range between 10–33%, with the wide variation reflecting the characteristics of the population (Lopez-Novoa et al., 2011; McWilliam et al., 2017). The incidence of vancomycin-associated nephrotoxicity is reported to range up to 43%, depending on the target population (Jeffres, 2017).

3.2 Mechanism of Proximal Tubule Injury Induced by Aminoglycosides, Polymyxins, Vancomycin and Protein Overload

The key mechanism of kidney injury induced by aminoglycosides is fairly well established (Tulkens, 1989; Lopez-Novoa et al., 2011). Aminoglycosides are rapidly eliminated via urinary excretion (80% within 24 h). Following glomerular filtration, however, 5–10% of the dose are reabsorbed and accumulate within the renal cortex (Antoine et al., 2010). The proximal tubule, and specifically the proximal convoluted tubule, presents the critical target of aminoglycoside nephrotoxicity. The site-specificity of renal toxicity is consistent with the abundant expression of megalin and cubilin, which facilitate efficient uptake into epithelial cells of this nephron segment. Following receptor-mediated endocytosis, aminoglycosides accumulate within the endosomal compartment, particularly within lysosomes (Silverblatt and Kuehn, 1979). Due to their cationic structure, aminoglycosides bind to membrane phospholipids, e.g. within lysosomes, and alter their function (e.g., inhibition of A1, A2, C1 phospholipases). Accumulation of aminoglycosides within lysosomes eventually leads to permeabilization of lysosomal membranes and release of lysosomal content and free aminoglycosides into the cytosol. While cytosolic aminoglycosides have been suggested to directly interfere with the mitochondrial electron transport chain and mitochondrial energy production, and to activate the intrinsic apoptotic pathway, release of cathepsin proteases and reactive oxygen species from lysosomes is on its own detrimental to cells (Lopez-Novoa et al., 2011). Depending on their concentration, cathepsins can induce cell death by either apoptosis through cleavage of caspases and activation of Bid or necrosis through massive proteolysis. Besides lysosomal overload as the major pathway involved in the mechanism aminoglycoside toxicity, it has also been suggested that accumulation of aminoglycosides within the endoplasmic reticulum (ER) may induce ER stress by interfering with protein synthesis and protein folding (Lopez-Novoa et al., 2011).

Following glomerular filtration of polymyxins, renal tubule cells reabsorb 90% of the drug and thus polymyxins accumulate substantially within proximal tubule cells located within the renal cortex (Azad et al., 2019). Although there is some evidence to suggest that polypeptide transporters (PEPT1 and PEPT2) contribute to cellular uptake of polymyxins, megalin-mediated reabsorption is considered to play a key role in accumulation of polymyxins within proximal tubule cells (Yun et al., 2015; Azad et al., 2019). While there is limited data on intracellular distribution of polymyxins in kidney cells (Azad et al., 2019), recent studies in human alveolar epithelial cells demonstrate co-localization of polymyxin B with early endosomes, lysosomes, and mitochondria (Ahmed et al., 2019). Similar to aminoglycosides, polymyxin B was shown to affect release of the hydrolytic lysosomal enzyme N-acetyl-β-glucosaminidase from lysosomes in vitro (Powell and Reidenberg, 1983). Although there are as yet no mechanistic studies investigating the causal relationship between polymyxin mediated disruption of lysosomal function and death, the evident similarities in renal handling and lysosomal localization suggests that polymyxins act at least in part via the same mechanism as aminoglycosides. Besides lysosomal toxicity, polymyxins have been suggested to cause oxidative stress and apoptosis via mitochondrial, death receptor, and endoplasmic reticulum pathways (Dai et al., 2014; Azad et al., 2019), yet the interlinkage between these effects remains unclear.

Compared to aminoglycosides and polymyxins, the mechanism of vancomycin induced nephrotoxicity is less well studied. However, based on vancomycin being a ligand for megalin and lyososomal accumulation of vancomycin within the S1 and S2 segment of the proximal tubule (Beauchamp et al., 1992; Fujiwara et al., 2012), a similar mechanism can be assumed.

The mechanism of proximal tubule injury induced by protein overload initiated by receptor-mediated endocytosis of urinary proteins induced is also linked to lysosomal dysfunction and membrane permeabilization. Protein overload leads to increased lysosomal number and volume, impaired lysosome-mediated proteolytic degradation as a result of defective lysosomal acidification, and finally lysosomal membrane permeabilization (Liu W. J. et al., 2015). Activation of NF-κB in tubular epithelial cells is thought to play an important role in the progression of tubulointerstitial injury by promoting initerstitial infiltration of mononuclear cells, interstitial edema, and fibrosis (Zoja et al., 2004).

3.3 The Adverse Outcome Pathway of Receptor Mediated Endocytosis and Lysosomal Overload Leading to Kidney Toxicity

Binding to multiligand, endocytic receptors expressed at the brush-boarder of renal tubule cells, resulting in proximal tubule cell uptake via receptor-mediated endocytosis can be defined as the molecular initiating event (MIE) in this AOP. Although toxicokinetics are typically not considered as part of an AOP, the molecular interaction between ligand and receptor appears to be essential for the lysosomal accumulation of chemical stressors and subsequent disturbance of lysosomal function (KE1), disruption of lysosomes (KE2) and proximal tubule cell toxicity (KE3) (Figure 5).

Evidence for receptor-mediated endocytosis and lysosomal overload leading to kidney toxicity as an adverse outcome primarily comes from experimental in vitro and in vivo studies on aminoglycosides, polymyxins, and low molecular weight urinary proteins that serve as chemical stressors for this pathway (Tables 4A–C). In the following sections, evidence supporting the key events and key event relationships in this AOP is presented, followed by a critical assessment of the AOP in terms of temporal and dose-response concordance, essentiality of key events, biological plausibility, coherence, and consistency of the experimental evidence.

TABLE 4A

TABLE 4A. Evidence from human, animal and in vitro studies on aminoglycosides supporting the key events and qualitative concordance of KEs within this AOP (n/a = not data available).

TABLE 4B

TABLE 4B. Evidence from human, animal and in vitro studies on polymyxins supporting the key events and qualitative concordance of KEs within this AOP (n/a = not data available).

TABLE 4C

TABLE 4C. Evidence from human, animal and in vitro studies on vancomycin supporting the key events and qualitative concordance of KEs within this AOP (n/a = not data available).

Molecular Initiating Event: Receptor-Mediated Endocytosis

Receptor-mediated endocytosis via binding to the multi-ligand receptor system megalin/cubilin constitutes the principle pathway of cellular uptake of polybasic drugs and low molecular weight proteins (LMWP) from the glomerular filtrate. The interaction of polybasic drugs and LMWPs with the receptor located at the brush border is facilitated by the negative charge of acidic membrane phospholipids and interaction of basic residues of the ligand with negatively charged receptor domains (Moestrup et al., 1995). The site-specific toxicity of polybasic drugs and LMWPs to the proximal tubule, the proximal convoluted tubule, corresponds with the abundant expression of megalin and high endocytic capacity within this nephron segment (Schuh et al., 2018). Receptor-mediated endocytosis is primarily responsible for delivery of polybasic drugs and LMWPs to lysosomes and subsequent disturbance of lysosomal function that ultimately leads to kidney toxicity. There are several lines of in vitro and in vivo evidence that support receptor-mediated endocytosis as the key initiating event: 1) stressors for this AOP are ligands of megalin, 2) uptake and subsequent toxicity of stressors for this AOP can be prevented by competitive inhibitors and indirectly by statins, which block post-translational prenylation of guanosine-5′-triphosphate (GTP)-binding proteins are required for megalin-mediated endocytosis (Antoine et al., 2010; McWilliam et al., 2018). 3) Loss of megalin, e.g., through megalin knockout, protects from accumulation and toxicity of stressors. Experimental evidence from in vitro and in vivo studies on aminoglycosides, polymyxins and vancomycin as chemical stressors of this AOP are summarized in Tables 4A–C. In addition to these chemical stressors, silencing of megalin and cubulin has been shown to inhibit myeloma light chain endocytosis and reduce the toxicity of myeloma light chains (Li et al., 2008), which are exessively produced in multiple myeloma and may cause proximal tubule alterations through overload of the endocytic process (Batuman, 2007). Similarly, siRNA mediated knockdown of megalin and cubilin was shown to block albumin-induced tubular injury (Liu D. et al., 2015). Although there are as yet no data related to the MIE in humans, it is noted that a phase IIa randomized controlled clinical trial investigating prevention of aminoglycoside-induced kidney injury by rosuvastatin in children with cystic fibrosis has been initiated (McWilliam et al., 2018).

Key Event 1: Disturbance, Lysosomal Function

There is substantial evidence from in vitro and in vivo studies that accumulation of endocytosed drugs and LMWPs within lysosomes leads to ultrastructural changes, including increased number and size of lysosomes in renal proximal convoluted cells (Houghton et al., 1978b; Kaloyanides and Pastoriza-Munoz, 1980; Feldman et al., 1982; Mandal and Bennett, 1988; Toyoguchi et al., 1997). These ultastructural changes are considered to occur as a consequence of substrate overload or reduced lysosomal proteolytic capacity due to compound binding to lysosomal phospholipid membrane and inhibition of phospholipases. Interference with phospholipid metabolism results in accumulation of phospholipids with formation of lysosomal myeloid bodies, i.e. concentric multilaminar phospholipid membrane whorls. Experimental in vitro and in vivo studies demonstrating inhibition of phospholipases, phospholipidosis and myeloid body formation by chemical stressors for this AOP as well as detection of myeloid bodies in renal biopsies and urinary sediment of patients treated with aminoglycosides support disturbed lysosomal function as a key event in this AOP (Tables 4A–C). In addition to interference with lysosomal phospholipid metabolism, lysosomal protein overload through excessive exposure of proximal tubule cells to ligands of the endocytic receptor may also lead to altered lysosomal function that expresses itself in hyaline droplet formation. Recent data suggest that changes in the tertiary structure of albumin may interfere with lysosomal proteolysis (Medina-Navarro et al., 2019). This is reminiscent of stabilization of α2u-globulin through binding of chemicals to α2u-globulin, resulting in resistance of the α2u-globulin–chemical complexes to lysosomal degradation and subsequent protein droplet formation in α2u-nephropathy (Lehman-McKeeman et al., 1990).

Key Event 2: Disruption, Lysosome

While reduced release of N-acetyl-β-d-glucosaminidase suggestive of lysosomal membrane stabilization may be an initial response to chemical stressors of this AOP (Powell and Reidenberg, 1982; Powell and Reidenberg, 1983), swelling of lysosomes due to intralysosomal accumulation of chemical stressors and macromolecules (protein, phospholipids) ultimately leads to lysosomal membrane permeabilization or lysosome rupture. It has been suggested that impaired phospholipid metabolism may increase the hydrophobicity of the lysosomal membrane, thereby interfering with transport of water-soluble products across the lysosomal membrane and subsequent osmotic disruption of lysosomes (Powell and Reidenberg, 1982; Powell and Reidenberg, 1983). In addition to the evidence for chemical stressors presented in Tables 4A–C, proximal tubule toxicity induced by albumin and urinary proteins has been shown to involve lysosomal membrane permeabilization and lysosome rupture (Liu D. et al., 2015; Liu W. J. et al., 2015). As a result of lysosomal disruption, lysosomal enzymes such as cathepsins are released into the cytosol (Lopez-Novoa et al., 2011), which may be evident by reduced lysosomal cathepsin activity or immunoreactivity (Li et al., 2000).

Key Event 3: Increase, Cytotoxicity (Renal Tubular Cell)

The link between disruption of lysosomes and cell death is well established (Bursch, 2001; Turk et al., 2002; Guicciardi et al., 2004). Leakage of lysosomal proteases such as cathepsins may trigger apoptosis directly through activation of pro-caspases or indirectly via promoting release of cytochrome C from mitochondria, whereas extensive lysosomal rupture results in necrosis. Evidence for this comes from studies demonstrating that controlled lysosomal rupture induced by a lysosomoropic detergent causes cathepsin release prior to apoptosis (Li et al., 2000). This study also shows that changes in mitochondrial membrane potential occur seondary to lysomal rupture (Li et al., 2000). There is ample evidence from in vitro experiments and studies in animals and humans that demonstrate proximal tubule toxicity of stressors of this AOP (Tables 4A–C). On a cautionary note, establishment of stable cell lines often involves use of aminoglycosides as selection antibiotics and thus renal cell lines generated via this protocol may be resistant to aminoglycoside toxicity.