Investigating Cell Type Specific Mechanisms Contributing to Acute Oral Toxicity

The replacement of animals in acute systemic toxicity testing remains a considerable challenge. Only animal data are currently accepted by regulators, including data generated by reduction and refinement methods. The development of integrated approaches to testing and assessment (IATA) is hampered by an insufficient understanding of the numerous toxicity pathways that lead to acute systemic toxicity. Therefore, central to our work has been the collection and evaluation of the mechanistic information on eight organs identified as relevant for acute systemic toxicity (nervous system, cardiovascular system, liver, kidney, lung, blood, gastrointestinal system, and immune system). While the nervous and cardiovascular systems are the most frequent targets, no clear relationship emerged between specific mechanisms of target organ toxicity and the level (category) of toxicity. From a list of 114 chemicals with acute oral in vivo and in vitro data, 97 were identified with target organ specific effects, of which 94% (91/97) were predicted as acutely toxic by the 3T3 neutral red uptake cytotoxicity assay and 6% (6/97) as non-toxic. Although specific target organ mechanisms of toxicity could in some cases explain the false negative prediction obtained with the cytotoxicity assay, in general it is difficult to explain in vitro misclassifications only on the basis of mechanistic information. This analysis will help to prioritize the development of adverse outcome pathways for acute oral toxicity, which will support the assessment of chemicals using mechanistically informed IATA. This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 International license (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution and reproduction in any medium, provided the original work is appropriately cited. 1 https://www.piscltd.org.uk/acute-systemic-toxicity/ (accessed 28.06.2018).


Introduction
Acute systemic toxicity after oral, dermal or inhalation exposure requires that the substance becomes bioavailable at the target site and induces lethality through general toxicity or specific mechanism.This means that kinetic factors, and mainly absorption, are important determinants of toxicity (EURL ECVAM, 2015).In addition, if the damage involves interference with homeostatic mechanisms at the organ system level, non-exposed tissues and vital organs can also be affected (Gennari et al., 2004;Andrew, 2013).
The assessment of acute systemic toxicity is a core component of the safety assessment of substances in the context of EU and international legislations (Hamm et al., 2017).Information requirements vary depending on the type of substance subject to regulation and the region (EU, 2006(EU, , 2008(EU, , 2009a(EU, ,b, 2012)).The regulatory landscape in the USA was reviewed during two workshops cosponsored by The NTP Interagency Center for the Evaluation of Alternative Toxicological Methods (NICEATM), the People for the Ethical Treatment of Animals (PETA) International Science Consortium Ltd., and the Physicians Committee for Responsible Medicine (PCRM) (Hamm et al., 2017, Clippinger et al., 2018a;Strickland et al., 2018).The relevant information is available at the PETA website1 .In preclinical drug development, however, these studies are no longer required by default to support first clinical trials in humans (Robinson et al., 2008;ICH, 2009;Chapman et al., 2010).
One of the main uses of acute systemic toxicity data, is classification and labelling (Seidle et al., 2010;Graepel et al., 2016;Buesen et al., 2016;Strickland et al., 2018).Within the EU, the CLP (Classification, Labelling and Packaging) Regulation (EU, 2008) is used to classify chemicals on the basis of acute oral toxicity into four toxicity categories [categories 1 to 4 of the United Nations Globally Harmonised System of Classification and Labelling (UN GHS)].While CLP does not require animal testing, the classification criteria are based on data derived from animal tests (conducted for example under other pieces of legislation), including reduction and refinement methods for the oral, dermal, and inhalation routes (OECD TGs 402,403,420,423,425,433,436).Most of the standard in vivo tests use lethality as the endpoint, even though this has been widely criticised both on animal welfare and scientific grounds (Zbinden and Flury-Roversi, 1981;Hoffmann et al., 2010;Prieto et al., 2013a).
Basal cytotoxicity is certainly a key factor in many prevalent toxicological modes-of-action associated with acute health effects.It covers many general mechanisms of toxicity common to most cell types that can lead to organ failure, including for example, disruption of cell membrane structure or function, inhibition of mitochondrial function, disturbance of protein turnover, and disruption of metabolism and energy production (Gennari et al, 2004;NIH, 2009, Andrew, 2013).This is the reason why the utility of in vitro cytotoxicity assays to predict acute oral toxicity has been extensively investigated (Ekwall, 1999;Halle, 2003;NIH, 2006;Prieto et al., 2013a,b).Recently, Vinken and Blaauboer (2017) have proposed application of an adverse outcome pathway (AOP) framework for basal cytotoxicity consisting of three consecutive steps, including initial cell injury, mitochondrial dysfunction and cell death.The outcome of the basal cytotoxicity was then suggested by the authors as the first step of a tiered strategy aimed to evaluate the toxicity of new chemical entities.Further, in a second step, more specific types of toxicity could be evaluated.
One of the better known and standardised in vitro methods for basal cytotoxicity is the 3T3 Neutral Red Uptake (NRU) assay (DB-ALM protocol 1392 ; Stokes et al., 2008).The use of data from the NRU cytotoxicity assay within a Weight-of-Evidence (WoE) assessment is one of the choices for adapting the standard information requirements for acute oral toxicity, as described in the last update of the ECHA's guidance on Information Requirements and Chemical Safety Assessment.This WoE adaptation proposed by ECHA applies primarily to low toxicity substances (i.e., those which are not to be classified for acute toxicity) and it is based on an in depth analysis of the REACH database (Gissi et al., 2017, ECHA, 2017).Nevertheless, the limitations of the in vitro cytotoxicity assay, such as the lack of metabolic competence of the 3T3 cells and difficulty to capture specific mechanisms of action relating to interaction with specific molecular targets in certain tissues, need to be considered when building a WoE case for the purposes of REACH (Buesen et al., 2018, Gissi et al., 2018).
In addition to the assessment of basal cytotoxicity, it is also important to identify cell types and in vitro endpoints that are indicative of cell-type specific toxicities, with a view to incorporating such endpoints into Integrated Approaches to Testing and Assessment (IATA), as proposed in the EURL ECVAM strategy to replace, reduce and refine the use of animals in the assessment of acute mammalian systemic toxicity (EURL ECVAM, 2014).As defined by the OECD, IATA are pragmatic, science-based approaches for chemical hazard or risk characterization that rely on an integrated analysis of existing information in a WoE assessment coupled with the generation of new information, if required (OECD, 2016a).An iterative approach that preferably relies on mechanistic information or available AOPs is followed to answer a defined question in a specific regulatory context, taking into account the acceptable level of uncertainty associated with the decision making (OECD, 2016a;Sachana and Leinala, 2017).The importance of understanding the mechanisms of acute toxicity was further recognised during an international workshop in which a group of experts discussed alternatives approaches for identifying acute systemic toxicity (Hamm et al., 2017;Clippinger et al., 2018a).A better theoretical and mechanistic understanding of acute systemic toxicity would be useful to developers of test methods and other predictive tools as well as to validation and regulatory bodies.
Mechanisms involved in cellular failure and susceptible functions compromised in organ failure were discussed at an ECVAM workshop on strategies to replace in vivo acute systemic toxicity testing (Gennari et al, 2004).Several fundamental cellular processes common to many organ systems were identified, including energy production and metabolism (mitochondrial function and glycolysis), transportation of molecules, membrane integrity and secretion of molecules (enzymes, proteins, hormones, neurotransmitters).A number of key events associated with acute human poisoning were further identified in an ICCVAM/ECVAM/JaCVAM workshop on acute chemical safety testing (NIH, 2009) and it was agreed that mechanistic information could be used to develop more predictive in vitro test methods.A report, commissioned by the US Department of Defense, lists several of the cellular targets or molecular targets that are often associated with the acute lethal or debilitating effects of chemicals.This includes changes in neurotransmission function, altered ion flow, increased permeability of cellular membranes, altered bioenergetics, altered oxygen transport, oxidative stress and reactive oxygen species (ROS) formation, damage to DNA and subcellular systems, and immune-mediated effects (NRC, 2015).Hamm et al. (2017) and Clippinger et al. (2018b) have also reported some of the known mechanisms involved in acute systemic toxicity as part of ongoing activities in the US.
However, despite all the efforts made over the past 20 years in the area of acute systemic toxicity, relevant AOPs, mechanistically informed alternative methods, and IATA for acute systemic toxicity have not been adequately developed.This is partially due to the lack of a complete mechanistic understanding of the key acute toxicity pathways in humans, specific for different cell types (e.g.neuronal, cardiac, liver or kidney).
This study describes the analysis of mechanistic information collected on eight potential organs (i.e., nervous system, cardiovascular system, liver, kidney, lung, blood, gastrointestinal system (GI) and immune system) identified as relevant for acute systemic toxicity and using a set of chemicals inducing acute toxicity after oral exposure.This work will support the development of AOPs and IATA in the area of acute systemic toxicity, and will inform the development and application of mechanistically relevant new approach methodologies.

2
Materials and methods

Collection of mechanistic information
Information was collected on the eight potential target organs identified as relevant for acute systemic toxicity during the ECVAM workshop on acute systemic toxicity (Gennari et al., 2004): liver, blood, kidney, cardiovascular system, central and peripheral nervous system (CNS/PNS), lung, immune system, and GI.In safety pharmacology studies, the cardiovascular, respiratory and central nervous systems are assessed in a core battery since they are considered vital organs or systems, the functions of which are acutely critical for life (ICH, 2000).
In order to approach the ambitious task of mapping mechanisms specific for these potential target organs, a three-step approach was taken to identify the potential pathways of target organ toxicity.1.Based on a literature review using *target organ* and *acute toxicity* and *mechanism* as key words, commonly recognised pathways of toxicity were identified for each target organ/system.Information was derived from published literature, toxicology handbooks, short descriptions of reference compounds used in the EU FP6 project ACuteTox3 and internet databases [HSDB4 , INCHEM5 , pubchem6 , PubMed7 , Scopus8 , Google Scholar9 ].The pathways were then organised and visualised according to the target organ/system, the cell type, the effect and the mechanism.In this context, the effect refers to any adverse reaction that could be observed or measured (in vivo) and the mechanism refers to the molecular or cellular process that is interrupted by chemical stressors and leads to the observed adverse effect.2. In the second phase the "completeness" of the theoretical pathways of toxicity that were developed in phase 1 was probed.In order to do so, we consulted the in-house database and selected chemicals that were shown to be acutely toxic.For these chemicals, a thorough literature search was conducted to identify the target organ and mechanism of toxicity, searching first for *chemical* AND *acute toxicity* and *mechanism*, followed by *chemical* AND *Target organ*.These mechanisms were then added into the generated "maps" if they were not already present.Chemicals for which both in vivo acute oral toxicity data and in vitro cytotoxicity data were available were selected.On the basis of reference in vivo oral LD50 data and of the 2000 mg/kg body weight threshold introduced by the CLP Regulation, all compounds with an acute LD50 mean value below or equal to 2000 mg/kg were identified as acutely toxic, whereas those with an acute oral LD50 mean value above 2000 mg/kg were identified as non-acutely toxic.Only the chemicals that fell into our group of toxic chemicals were considered in this second phase.3.In a third phase, chemicals from the group of non-toxic chemicals that, nevertheless, had been assigned a harmonised classification (Annex VI of EU CLP Regulation) were identified and selected.The mechanisms collected and shown in this report are not intended to be exhaustive.

Selection of chemicals with in vivo LD50 values and in vitro cytotoxicity data
In the in-house database 178 test chemicals had oral LD50 values (see supplementary file10 ) that were collected from publicly available databases (e.g.ChemIDplus, IUCLID, RTECS, and HSDB), Merck index, EU Risk Assessment Reports, Sax's Dangerous Properties of Industrial Materials, and the published literature.According to the calculated mean LD50 values, 112 test chemicals were assigned to an EU CLP acute oral toxicity category and 66 remain as non-toxic (i.e., no category assigned because the LD50 was higher than 2000 mg/kg).Eleven out of the 66 non-classified chemicals had an official acute oral classification.
In vitro cytotoxicity data were available for 177 test chemicals that had been screened in the following international projects: NICEATM/ECVAM validation study (NIH, 2006), the EU FP6 project ACuteTox (Prieto et al., 2013a) and the ECVAM validation study (Prieto et al., 2013b).The list of chemicals used in each study is available through the JRC Chemical Lists of Information System (CheList11 ).
When the two sets were compared, in vitro cytotoxicity data were not available for two compounds, formaldehyde and carbon tetrachloride, and in vivo oral LD50 data were not found for benz(a)anthracene.Therefore, the final common set contained 176 chemicals.

Overall analysis of mechanistic maps
The mechanistic information collected following the three-step strategy was visualised in maps according to the eight organs/systems.The layout and structure across organs and systems was harmonised and analysed as shown below.Three maps were created per target organ/system.A first map illustrated the mechanisms found based on information collected from literature (step 1 under methods).The second map was an updated version based on the information collected from the in-house list of selected compounds (steps 2 and 3 under methods).The final harmonised version of each organ/system was shown by the third map (Fig. 1-8).
Information on mechanisms of toxicity was collected for 114 out of the 123 oral acutely toxic chemicals (see section 2.2).In terms of target organ/systems, the overall analysis summarised in Figure 9 shows that, according to the information found, the nervous and cardiovascular systems are the most frequent targets (67 and 39 chemicals, respectively) followed by liver, kidney, lung, gastrointestinal system, blood and immune system (31, 30, 24, 18, 11, and 3 chemicals, respectively).Twenty six chemicals appear to target single organs, in particular the nervous systems (12 chemicals).Seventy five chemicals affect more than one organ/system and thirteen chemicals affect all organs (non-specific target organ effects) (Fig. 10).Indirect effects were reported for 9 chemicals with multi-organ/system effects: 6 on the lung, 1 on the kidney and 4 on the cardiovascular system (Fig. 9).General cytotoxicity mechanisms were cited for 72 chemicals and target organ/system specific effects for 40 chemicals (11 chemicals acting on single organ/system and 29 on multiple targets) (see Tab. 1).For pentachlorobenzene and tetramethylthiuram monosulphide, the specific mechanism of acute toxicity was not found.
Tables 2-9 provide an overview of the specific target organ/system mechanisms leading to acute toxicity according to the information collected from the literature.The list of mechanisms shown is neither exhaustive nor definitive.
Referring to organ specific mechanisms of toxicity, interference with neurotransmitters and/or neurotransmission and impairment of propagation of electrical activity are among the main reported mechanisms for chemicals that target the nervous system.In particular, many chemicals interfere at the level of receptors and ion channel function.
Chemicals that target the cardiovascular system often interfere with ion balance/signalling/membrane potential of the cell and with intracellular signalling mechanisms.
For many of the chemicals that damage the liver after an acute insult, mechanisms such as depletion of free radical scavengers, ROS production, lipid peroxidation (grouped under oxidative stress induced inflammation) and necrosis, were reported.
Alterations in kidney tubule cell structure (accumulation in proximal tubular cells, loss of tubular epithelial barrier and/or tight junctions), alterations in tubule cell metabolism (interference with ion balance), tubular obstruction (impaired Na + and water reabsorption, distal cast formation, crystal deposition) and alterations in cell viability (necrosis) are among the most reported mechanisms leading to acute renal failure.
The in vivo classification for acute oral toxicity (i.e. the assigned CLP acute oral toxicity categories based on the collected mean oral LD50 values) of the chemicals acting via specific mechanisms of toxicity at organ/system level was evaluated in view of the information found in the literature for each chemical.Table 10 summarises the outcome of this analysis confirming that the nervous and cardiovascular systems are the most frequent targets for chemicals inducing acute oral toxicity.

Analysis of mechanistic information and in vitro 3T3 NRU cytotoxicity results
A total of 97 chemicals, for which in vivo and in vitro cytotoxicity data were available, have been identified with target organ/system specific effects, of which 91 were predicted as acutely toxic by the 3T3 NRU cytotoxicity assay (LD50 ≤ 2000 mg/kg).Table 11 summarises the prediction of the acute oral toxicity (EU CLP toxicity categories) by the in vitro cytotoxicity assay for these chemicals.
Figure 11 complements Table 11 by adding the collected mechanistic information.It also shows the percentage of chemicals identified in vitro as acutely toxic acting through cell type specific mechanisms of toxicity and via general cytotoxic mechanisms when the acute oral toxicity category was correctly predicted (50 chemicals), under-predicted (32 chemicals), and over-predicted (9 chemicals) by the in vitro cytotoxicity assay.
An overview of the information collected with regard to specific target organ/system and general cytotoxicity for the chemicals that are correctly assigned to the CLP acute oral toxicity category, under-predicted and over-predicted by the in vitro cytotoxicity assay, respectively, is provided in Tab.S1, S2, and S312 .
Among the 50 correctly predicted chemicals, 29 act through some general mechanisms of cytotoxicity (58%) and 21 only via cell type specific mechanisms of toxicity (42%).
Among the 32 under-predicted chemicals, 20 act through a general mechanism of cytotoxicity (63%) and only 12 via cell type specific mechanisms of toxicity (38%).
Among the 9 chemicals with toxicity category over-predicted by the cytotoxicity assay, 6 act through general mechanism of cytotoxicity (67%) and 3 only via specific mechanisms of toxicity (33%).
Among the 6 chemicals falsely predicted as non-classified (Fig. 12), two act through some mechanism of general cytotoxicity (33%) and four act only via cell type specific mechanisms of toxicity (67%), as shown in Tab.S4 12 .Fig. 11: Specific target organ/system toxicity and general cytotoxicity reported for the 91 chemicals predicted by the 3T3 NRU cytotoxicity assay as positive chemicals (i.e.LD50 ≤ 2000 mg/kg) Bars in each group represent from left to right: nervous system, cardiovascular system, liver, kidney, lung, gastrointestinal system, blood, immune system, general cytotoxicity, only specific target organ/systems.Cat.: acute oral toxicity category.

Acute oral toxicity category 1
From the compounds assigned to the acute oral toxicity category 1 in vivo (fatal if swallowed), three (i.e., brucine, disulfoton, and physostigmine) target the nervous system and act via specific mechanisms (e.g., inhibition of cholinesterase, antagonism of glycine receptor).Among the remaining compounds, 1-phenyl-2-thiourea has the lung as target organ and needs bioactivation, and triethylenemelamine affects the immune system.General mechanisms of cytotoxicity have also been reported for these two compounds.For brucine, necrosis was as the mechanism responsible for kidney tubular cell damage.When in vivo and in vitro mean values are compared, all compounds are misclassified by the in vitro cytotoxicity assay.
Triethylenemelamine is under-classified by one toxicity category and the other four compounds by 3 toxicity categories.

Acute oral toxicity category 2
Among the compounds assigned in vivo to the acute oral toxicity category 2 (fatal if swallowed), only colchicine is correctly predicted by the cytotoxicity assay.Colchicine inhibits microtubule formation and, thus, effectively inhibits mitosis, which is a general mechanism of toxicity.This mechanism of toxicity is also reported as the one responsible for the toxicity at the level of the nervous system and the liver.Digoxin and aconitine, which were predicted as false negatives by the cytotoxicity assay, act via specific mechanisms of toxicity such as interference with transporter enzymes (e.g.Na+-K+-ATPase) and calcium channels.Digoxin targets the cardiovascular system while aconitine acts on both the nervous and the cardiovascular system (Qiu et al., 2008;Chan, 2009;Prassas et al., 2011;Sun et al., 2014).
The compounds assigned to the toxicity category 2 in vivo but under predicted in vitro act mainly via specific target organ toxicity mechanisms.For D-amphetamine, parathion and strychnine, the central nervous system is the main target and specific mechanisms of toxicity were identified (i.e., inhibition of acetylcholinesterase, antagonism of glycine receptor, stimulation of the release of norepinephrine and dopamine).The kidney is the target for ochratoxin A, a well-known nephrotoxic agent acting on tubular cells (Ozbek, 2012).General cytotoxicity is also reported for ochratoxin A, and necrosis has been found as the mechanism underlying strychnine effects on kidney tubular cell damage.Epinephrine hydrogen tartrate targets the cardiovascular system, while warfarin targets both the cardiovascular system and blood cells (Hanley, 2004;Klaassen, 2001).

Acute oral toxicity category 3
General mechanisms of toxicity were also described for the chemicals correctly assigned in vitro to the acute oral toxicity category 3 (toxic if swallowed).Only for ethyl chloroacetate no general mechanisms of toxicity were reported, although it is a moderate irritant via the oral route.For the compounds under predicted in vitro by only one toxicity category, mechanisms of general cytotoxicity were reported.For sodium salt of chloroacetic acid and pentachlorophenol also the mechanisms identified at target organ level were general mechanisms of toxicity such as interference with mitochondrial function (CNS), membrane disruption and/or interference with macromolecules, depletion of free radical scavenger (such as glutathione, catalase) content in liver tissue, and compromised mitochondrial respiration of tubular cells (kidney).Dichlorvos and theophylline, act through general mechanisms of cytotoxicity at CNS level, and also via specific mechanisms of toxicity.Seventy three per cent of these under predicted compounds are linked to the CNS as the target system (i.e., GABAA receptor agonist, blockade of the GABAAreceptor coupled sodium channel, interference with the normal flux of Na+ and K+ ions across the axon membrane during neuronal signalling, antagonism of N-methyl-D-aspartate (NMDA) receptors and inhibition of serotonin/norepinephrine reuptake, agonist of nicotinic cholinergic receptors, neurotransmitter clearance from synaptic cleft).Verapamil, barium chloride, and theophylline act at the level of the heart by different mechanisms (blocking the calcium channel and binding to the cytosolic surface of the channel; interference with potassium channels, interference with intracellular signalling mechanisms, such as enzymatic activity e.g.phosphodiesterases and protein kinases).Fat accumulation in hepatocytes was also described as liver specific mechanism of toxicity of barium chloride (Ananda et al., 2013).

Acute oral toxicity category 4
General mechanisms of toxicity are reported for 49% of the compounds correctly assigned to the acute oral toxicity category 4 (harmful if swallowed).For several of these compounds, general cytotoxicity was identified at the target organ/level alone or in addition to other specific mechanisms of toxicity [e.g.acetylsalicylic acid uncouples mitochondrial oxidative phosphorylation and also inhibits Kreb's cycle dehydrogenases at CNS level (Ekwall et al., 1998); valproic acid alters the activity of the GABA neurotransmitter by increasing the inhibitory activity of GABA through inhibition of GABA degradation, inhibition of GABA transaminobutyrate, increased GABA synthesis, decreased turnover and inhibition of the GABA reuptake by the glia and synaptic mechanisms.It also interferes with cellular metabolic processes, interacts with membrane ion channels (Sztajnkrycer, 2002;Chateauvieux et al., 2010) and induces oxidative stress by compromising the antioxidant status of the neuronal tissue (Chaudhary and Parvez, 2012); caffeine in the CNS competitively antagonises adenosine receptors, inhibits phosphodiesterase, stimulates catecholamine release, and increases free calcium and intracellular cAMP (Fredholm et al., 1999); orphenadrine chloride competitively antagonise acetylcholine binding at the neuroreceptor sites and induces necrosis in liver (Sangster et al., 1978;Ekwall, et al., 1998)].The nervous and the cardiovascular system appeared as targets for 70% and 54% of the harmful compounds, respectively.Among the compounds acting via specific mechanisms of toxicity, 65% (13/20) targeted both the nervous and the cardiovascular system.
Four harmful compounds were falsely predicted in vitro as non-acutely toxic.Of those, isoniazid and paraldehyde act via specific mechanisms such as interference at the level of CNS receptors and ion channels function, depletion of GABA (isoniazid), impairment of the propagation of electrical activity in the CNS (paraldehyde), and inflammation of the GI mucosa (paraldehyde) (Gilman, 1985;Carpentier et al., 1992).The harmful effects of ethylene glycol mainly result from the accumulation of its more toxic metabolites (Hess et al., 2004).Diethylene glycol is metabolised to 2hydroxyethoxyacetaldehyde by alcohol dehydrogenase oxidation, then to 2-hydroxyacetic acid (HEAA) by aldehyde dehydrogenase.HEAA causes acidosis, renal failure, and neurologic dysfunction.It is thought that the parent compound is toxic as well (Schep et al., 2009).The formation of toxic metabolites will be missed in the in vitro cell system due to the lack of metabolic competence of the 3T3 cells.This could explain, at least in part, the misclassification by the in vitro approach.

Discussion
The exercise reported here served several purposes.First of all, the mechanistic information collected was visually summarised, allowing the direct comparison across target organs/systems.Secondly, by organising acute toxic effects by their mechanism and cell type(s) we could start to associate known acutely toxic compounds with the different mechanisms.Finally, organising information in this manner should facilitate the development of AOPs and IATA, for example by identifying properties that are requisites of in vitro testing systems for specific target organ toxicity testing.
This work also aimed to identify specific (complementary) mechanisms of acute toxicity that are perhaps not covered by the validated 3T3 NRU cytotoxicity assay.Therefore, in our analysis we tried to address: (i) whether chemicals can be identified and classified based on either positive or negative specific effects on target organ(s) (according to the 2000 mg/kg threshold) using the 3T3 NRU assay; (ii) which organs are the most frequent targets; and (iii) whether the triggered pathways of toxicity are conserved across organs.
In the overall analysis, 97 chemicals were identified with target organ specific effects, 94% (91/97) of them were predicted as acutely toxic by the in vitro cytotoxicity assay and 6% (6/97) as non-toxic.When comparing the positive (i.e.acutely toxic) and negative (i.e.non-acutely toxic) in vitro predictions with those of the in vivo study, it turned out that all six negatives were false predictions (false negatives), while 55% of the positive predictions were correctly predicted, according to a CLP acute toxicity category, 35% under predicted and 10% over predicted.When evaluating the performance of any alternative approach for the purposes of regulatory classification, it is also necessary to consider the uncertainty associated with both the in vivo and the in vitro data.Actually, the analysis of consistency in classification published by Hoffmann et al (2010) showed that conventional in vivo acute oral toxicity tests are intrinsically imprecise themselves and, about 44% of the substances would ambiguously occur within the limits of two adjacent classification categories (with at least 90% probability).A discussion of in vivo and in vitro data variability is outside the scope of this paper.Therefore, for the purpose of the mechanistic analysis shown and discussed here, the assignment of the compounds to the CLP acute oral toxicity categories was made based on the collected mean values (in vivo and in vitro).Another major source of uncertainty, not analysed here due to lack of information, is the role of ADME (absorption, distribution, metabolism, and excretion) in determining the acute toxicity category.ADME has also been identified as a source of uncertainty in many OECD IATA case studies that are based on new approach methodologies.Several regulatory bodies have published guidance on the identification, characterisation and reporting of uncertainty (SCHEER, 2018).
A closer look at the chemicals acting through the specific target organs has not revealed a clear pattern with regard to which specific mechanisms of target organ toxicity are representative of compounds in the different CLP acute oral toxicity categories.For instance, approximately the same percentage of compounds acting through mechanisms of neurotoxicity was found in each acute oral toxicity category (i.e.55% of the highly toxic chemicals allocated to toxicity categories 1 and 2, 50% of the toxic chemicals in category 3, and 56% of the harmful chemicals in category 4).A similar situation holds true for chemicals that act through mechanisms of cardiovascular toxicity, which were allocated to toxicity categories 2, 3 and 4 (33%, 19%, and 40%, respectively).Mechanisms of nephrotoxicity were also found for chemicals in all toxicity categories.Mechanisms of liver, lung and blood toxicity were described for a small percentage of the highly toxic compounds (8% -21%).Based on these results, it can be concluded that mechanisms of toxicity specific for each organ can be triggered by compounds that belong to the different CLP acute oral toxicity categories.This is not surprising since the CLP categorisation is based on potency, which can result from both toxicokinetic and toxicodynamic factors.
From the information collected and the analysis presented it can be concluded that general cytotoxicity is an important determinant of acute systemic toxicity.Overall, the majority of the analysed chemicals (63 %) causing acute lethal toxicity act via some general (rather than organ specific) mechanisms of toxicity.The nervous and the cardiovascular systems are the most frequent targets, with changes in neurotransmission and altered ion flow being important mechanisms often associated with acute neurotoxicity and cardiotoxicity, respectively.
It is well recognised that the use of basal cytotoxicity alone to determine the acute toxicity of a chemical may not always be enough and, furthermore, it depends on the chemical's kinetic behaviour and/or its specific mechanisms of toxicity.These features may need to be considered in order to correctly estimate the in vitro concentration causing toxicity that could be compared with the concentration that the target cells in vivo would be exposed to (EU FP6 project ACuteTox).In silico tools such as the Virtual Cell Based Assay (VCBA) can be used to simulate the distribution of the chemicals in the in vitro system (Zaldivar Comenges et al., 2017).By comparing the simulated and the nominal IC50 concentrations of the dissolved chemical, the influence of the in vitro kinetics on the cytotoxicity result may be anticipated.In addtion, in vivo kinetics is an important determinant of acute systemic toxicity that requires further investigation (Graepel et al., 2017;Duarte Lopes Mascarenhas Proença et al, 2017).In silico tools such as Physiologically Based Kinetic (PBK) models (Paini et al., 2017) can be used to simulate the kinetics and distribution of chemicals in vivo.
Although specific target organ mechanisms of toxicity could in some cases explain the false negative prediction obtained with the cytotoxicity assay, in general it is difficult to explain in vitro misclassifications only on the basis of mechanistic information.Therefore, in addition to kinetic considerations, in vitro misclassifications could be also linked to the number of acute oral toxicity categories under CLP and the associated LD50 ranges, which are not based on a particular mechanistic rationale.Indeed, the outcome of the classification analysis carried out in the context of the EU FP6 project ACuteTox, indicated that it is challenging to make a clear distinction between acute oral toxicity categories 1, 2, and 3 based on in vitro concentration-response data (Kinsner-Ovaskainen et al., 2013) and, therefore, three levels of toxicity [i.e. level 1: combination of categories 1 to 3, level 2: category 4, and level 3 (non-classified)] were proposed (Prieto et al., 2013a).A similar grouping was also considered by Norlén et al. (2012) in an investigation of the predictive performances of five alternative approaches for the assessment of acute oral toxicity.Overall, the value of the CLP classification into four acute oral toxicity categories could be challenged.
Building on all the collected/generated information it would be worth trying to develop an alternative way of classifying chemicals for acute oral toxicity based mainly on cytotoxicity and kinetic information, and complemented, if needed, with relevant organ specific mechanisms of toxicity.Based on the mechanistic knowledge discussed in this paper we propose to integrate in vitro assays anchored to the most frequent mechanisms of acute toxicity specific for each organ (CNS, heart, liver and kidney) into an IATA.An IATA can include defined approaches, i.e. formalised decision-making approaches that apply fixed data interpretation procedures to data generated with a defined set of information sources (OECD, 2016b).In this regard, an in vitro cytotoxicity assay would be used together with specific target tissue toxicity mechanisms tested by assays permitting evaluation of neurotoxicity (as the most sensitive), followed by cardiotoxicity, hepatotoxicity and kidney toxicity.Such a battery of tests should be designed to allow assessment of the compounds based on their cytotoxicity (e.g., based on the 3T3 NRU assay) and organ specific mechanisms.In the broader context of IATA, these in vitro mechanistic data should be integrated with additional sources of information (QSAR, read-across, in chemico, human data, in vivo data etc.) including, where appropriate, exposure and ADME information.
The development of AOPs relevant to acute neurotoxicity, cardiotoxicity, hepatotoxicity etc. is already ongoing.However, as indicated, they are at different stages of development (Tab.S5 12 ).It is worth noting that some of the relevant AOPs are not specific to acute toxicity but nevertheless include key events that are relevant to acute exposure effects.Interestingly, some of the chemicals identified in these AOPs as triggers of molecular initiating events overlap with the chemicals reviewed in this report.The mechanistic information provided in this paper should inform the development of AOPs relevant to acute systemic toxicity, as well as AOP-informed IATA.The further development of AOPs and IATA should focus on the major target organs identified, i.e. the CNS, heart, liver and kidney.

Fig. 1 :
Fig. 1: Target organ blood -visualization of mechanisms leading to acute systemic toxicity CO: carbon monoxide

Fig. 6 :
Fig. 6: Immune system -visualization of mechanisms leading to acute systemic toxicity NFAT: nuclear factor of activated T-cells; NF-kB: nuclear factor kappa-light-chain-enhancer of activated B cells

Tab. 11 :
Fig. 12: Specific target organ/system toxicity and general cytotoxicity reported for the 6 chemicals falsely predicted as negatives by the 3T3 NRU cytotoxicity assay GI: gastrointestinal system

Tab. 1: Chemicals acting through general cytotoxic mechanisms and/or specific mechanisms of toxicity General cytotoxic mechanisms Specific mechanism of toxicity
a https://www.micromedexsolutions.com/home/dispatch accessed on 09/07/2018.Login required.Tab.

10: Contribution of specific target organ/system mechanisms of toxicity to the in vivo acute oral toxic category of chemicals
a CLP