Description of the matrix
The collection of observations has generated a heterogeneous dataset (Online Resource 1) that considers all three categories described above, and where scientific knowledge is mapped across the various toxicity endpoints and the ten KCs of carcinogens, regardless of the level of standardisation of test methods and systems used or, biological level of organisation (i.e., molecular, cell, tissue or organism).
As such, the existing knowledge is organised, irrespective of specific indications on the regulatory use of each test method, and provides an indication of the distribution of available information. This allows to establish the differential contribution to the properties of carcinogens of the different toxicity endpoints, mainly in terms of mechanistic toxicity information (Fig. 2). The observed effects are parameters (biomarkers, enzyme activities, final test results) that define the KCs of carcinogens and describe the different toxicity effects.
As shown in Fig. 2, the endpoints toxicokinetics, skin sensitisation, genotoxicity, acute systemic toxicity, phototoxicity, immunotoxicity, toxicity to reproduction, repeated dose toxicity and carcinogenicity, are the major components in any toxicological regulatory dossier (e.g., ICH S4 2000; ICH S8 2006; ICH S1 2012; ICH S2(R1) 2012; ECHA R7a 2017), and contribute differently to the KCs of carcinogens. Notably, endocrine disrupting properties, despite not being a toxicity endpoint per se but a specific mode of action leading to various toxicity endpoints, have also been considered since, based on the criteria outlined in Commission Regulations (Regulation (EU) 2017/2100; Regulation (EU) 2018/605), specific tests have been proposed recently for the identification of endocrine disruptors in the context of EU Regulations for plant protection and biocidal products (Regulation (EC) 1107/2009; Regulation (EU) 528/2012; ECHA EFSA Guidance 2018).
Figure 2 shows that when considering the sources of information available to identify each toxicity endpoint, there is the possibility to cover more than one KC. For example, examining available studies that describe the toxicokinetics (TK) of a chemical, we identified a number of protocols, as described in OECD TG 417 (OECD 417 2010) and more recent ones, including various in vitro test systems (> 10 protocols). These studies are not only able to measure absorption, distribution, metabolism and excretion (ADME) properties but also parameters that can be related to the KCs of carcinogens, such as: act as an electrophile either directly or after metabolic activation; modulate receptor-mediated effects; induce oxidative stress; induce chronic inflammation. In addition, TK information applicable to these KCs, can be also predicted by means of physiologically-based kinetic (PBK) models. An in-house review of PBK models developed in the past 10 years (2009–2019) ((Lu et al. 2016); the WUR University (NL) collection; PubMed), has indeed provided a number of models designed to describe drug-drug interaction (DDI) or drug- or chemical-response analysis, distribution in target tissues or chemical carcinogens exposure analysis whose predictions are applicable to different KCs (Fig. 2 and Online Resource 1).
The introduction of PBK modelling in the matrix can help to predict systemic exposure from external exposures but also to integrate the information across various test methods along the KCs. Moxon and colleagues have recently described the application of PBK modelling to the NGRA based exclusively on NAMs for dermally applied consumer products and were able to provide conservative estimate of the maximal blood concentration (Cmax) for three case studies (Moxon et al. 2020).
Information sources available for skin sensitisation (7 in vitro studies, 4 different in vivo studies, QSAR models and available AOPs) may contribute to the description of many KCs ( i.e., act as an electrophile either directly or after metabolic activation, induce chronic inflammation, cell proliferation and cell death) except, induce epigenetic alterations and immortalisation.
Likewise, in the case of genotoxicity, the information derived from in vitro studies (n = 11), in vivo studies (n = 9), scrutinised so far, together with available QSAR models and AOPs may cover almost all the KCs of carcinogens, except immortalisation as reported in Online Resource 1. Certainly, all the assays aimed at the identification of the genotoxicity endpoint contribute mainly to the characteristics of: being genotoxic and alter DNA repair or cause genomic instability.
The group of available test methods (approximately, 10 protocols) in use for testing immunotoxicity contributes instead only to some of the KCs, being mainly specific to: induce chronic inflammation, be immunosuppressive and cell proliferation and cell death. It is worth noting that the majority of immunotoxicity studies are mainly recommended in safety guidelines for pharmaceuticals (FDA 2006; ICH S8 2006). This does not exclude that parameters related to the immune system are evaluated through several toxicity studies across different toxicity endpoints, as detailed in Online Resource 1.
As previously reported, the mechanistic knowledge derived by literature search and collected for acute systemic toxicity is a valuable starting-point to inform other adverse outcomes (Madia et al. 2020; Prieto and Graepel 2018). In the context of this exercise, it is possible to evaluate the extent to which such mechanisms could play a role after repeated dose exposure scenarios and eventually inform the KCs of carcinogens. Thus, the information derived from in vitro (n = 4) and in vivo studies (n = 7), QSAR models and available assays from the ToxCast program (EPA) may contribute to many of the KCs of carcinogens (Fig. 2). While the major contribution is for induce chronic inflammation and cell proliferation, cell death and nutrient supply KCs, the group of acute systemic toxicity tests can also inform: act as an electrophile either directly or after metabolic activation, induce oxidative stress, induce chronic inflammation, be immunosuppressive and, modulate receptor-mediated effects.
Similarly, the five protocols for the identification of phototoxicity hazard contribute mainly to induce chronic inflammation and cell proliferation, cell death and nutrient supply KCs. However, by exploiting their potential, some of these protocols can also inform: act as an electrophile either directly or after metabolic activation, to be genotoxic and induce oxidative stress.
Notably, by deconstructing available test methods and approaches for toxicity to reproduction and target organ toxicity after repeated exposures (mainly in vivo studies), we were able to identify a number of observed effects (in vivo observations). Studies available for both categories of toxicity are highly informative for the KCs of carcinogens induce chronic inflammation, be immunosuppressive, and modulate receptor-mediated effects and cell proliferation, cell death and nutrient supply. The overall pattern of information contributing to these KCs is substantially similar between the two toxicity endpoints.
The grouping of test methods scrutinised so far for the above categories of toxicity has enabled to describe in a qualitative dimension the distribution of information (Fig. 2). It also enabled to identify major contributors to the knowledge of the carcinogenic potential of substances but also areas of consistent lack of knowledge in terms of observed effects and hence available assays.
For example, KCs describing major mechanisms involved in toxicity outcomes such as oxidative stress, chronic inflammation, and alterations in cell growth can be detected by means of different test methods and test systems. They are routinely evaluated in a number of studies from in silico to in vitro and to in vivo. Instead, alter DNA repair or cause genomic instability, induce epigenetic alterations and immortalisation are still not fully incorporated within available regulatory toxicity studies and rarely investigated, despite their key role in carcinogenesis. In agreement with our observations, Krewski and colleagues (Krewski et al. 2020), analysing the KCs associated with 86 Group 1 human carcinogens reviewed by IARC, reported that information on epigenetic alterations derives mainly from human studies, both in vitro and in vivo, mostly epidemiological investigations. For alter DNA repair or cause genomic instability, epigenetic alterations and immortalisation investigations, a conspicuous number of assays and methodologies are available and in use routinely in the research field. However, as for other applications, i.e. new methodologies and “omics” techniques which are currently shaping cancer biology research (Nature various 2020), are not applied yet on a routine base in the regulatory context.
Mechanistic information provided by in vivo studies and new approach methodologies
As summarised in Fig. 2, the matrix built over the collection of observed effects allows the alignment of toxicity information in terms of mechanistic knowledge provided by each single study. This helps to visualise where relevant information is stored and how it can be shared across different toxicity endpoints and more importantly, whether it can be used to inform one toxicity endpoint from another (Madia et al. 2020).
A number of parameters can be observed and are included per single in vivo study: general clinical observations, food consumption, toxicokinetic data, clinical biochemistry parameters, histopathology, ideally performed on every single organ, urinalysis, and/or other specific parameters as macroscopic developmental and reproductive effects, depending on the study endpoint and relative study design (Online Resource 1). The majority of these observations, even if not mechanistic per se, can be used to derive mechanistic information based on evidence and to define the KCs of carcinogens. However, most of the toxicity information provided by in vivo studies across different endpoints is highly redundant. The in vivo studies reported in the analysis (Online Resource 1) repeatedly inform some of the KCs of carcinogens, in particular inflammation, immune-response, receptor-mediated effects, cell proliferation, cell death and nutrient supply (Fig. 3a). However, these are mainly defined by the following observations: acute/subacute inflammatory infiltrate (acinar and/or interstitial) from histopathology; blood/serum clinical biochemistry data, including total and absolute differential leukocyte counts in serum, urinalysis, body and organ weight, clinical signs and food consumption, tissue/cell proliferation, hyperplasia, hypertrophy, cytotoxicity, necrosis from histopathology (Online Resource 1).
Mechanistic studies, mainly in vitro, include fewer observations per single study but they can inform multiple KCs, thus reducing redundancy of information (Fig. 3b). In this case, mechanisms and specific key events at the molecular level can be investigated to provide a detailed understanding of the toxicological mode of action (Malarkey and Hoenerhoff 2013) that conventional in vivo studies may not provide. Derived information is more heterogeneous than that derived from the in vivo counterpart. The observed effects reviewed so far and included in Table 1 are highly redundant and NAMs capable to identify them are not all in place within the regulatory context. Nevertheless, NAMs included in this first exercise show the opportunity to enrich mechanistic information across multiple KCs and multiple toxicity endpoints. This is for example the case of specific chemical properties or known key molecular players, i.e., transcription factors, regulators, mediators of effects whose function relates to different KCs and inform different endpoints (the example of the Nrf2-Keap1-ARE signalling pathway is detailed in Box 1). Furthermore, they may be equally described by means of in vitro, in silico, or more recently developed ‘omics’ approaches applied to different endpoints (Online Resource 1). Along these lines, Baltazar and colleagues recently illustrated the application of NGRA for the safety assessment of systemic toxicity of cosmetic products to a case study (coumarin) that included the use of integrated information across different toxicity endpoints and by means of various methodologies: information predicted from PBK models and in silico alerts, data from genotoxicity studies such as the Toxtracker (Hendriks et al. 2016) test method, cell stress panel, and high-throughput transcriptomics (HTTr) (Baltazar et al. 2020).
The matrix built over the observed effects can serve as a repository and a guide to identify information relevant to the properties of carcinogens. This gives the opportunity, on the basis of a mechanistic read across, to select available ad hoc test methods that can be used to avoid redundancy of testing but also to identify where relevant information is missing. This is shown in the matrix, as mentioned above, for the KCs of induce epigenetic alterations and immortalisation. Also, in the case of the KC to be immunosuppressive, the number of available test methods is limited in their use and application and are not yet sufficient to cover specific mechanisms of immunosuppression (Online Resource 1 and Fig. 2 and Fig. 3).
Box 1. Nrf2-Keap1-ARE signalling pathway role
across toxicity endpoints
Skin sensitisers, particularly cysteine-reactive skin sensitisers, have been shown to induce protective genes regulated by Nrf2-Keap1-ARE regulatory pathway (Kleinstreuer et al. 2018). The Keratinosens test method (OECD 442D 2018) is based on this principle. Similarly, the Sens-is test method is proposed to monitor the expression of a panel of 65 genes, including NRF2 in Reconstructed human Epidermis (RhE) for irritancy and sensitisation (Cottrez et al. 2015).
The Nrf2 signalling pathway represents one of the main cell defence mechanisms (Leinonen et al. 2014; Basak et al. 2017) and is considered a master regulator of redox homeostasis. It has been shown to play a role in different neurodegenerative diseases, aging, diabetes, photo-oxidative stress, cardiovascular disease, inflammation, pulmonary fibrosis, acute pulmonary injury, and also cancer (Kansanen et al. 2013; Jaramillo and Zhang 2013).
The NrF2-Keap1 transcriptional activation has been reported to be elicited in response to liver and kidney toxicants (Limonciel et al. 2018; Xu et al. 2019). As such, the activation of the Nrf2 response is relevant to skin sensitisation as well as other toxicity endpoints as genotoxicity, acute toxicity and/or repeated dose toxicity, and carcinogenicity.
In vitro genotoxicity tests include as well directly or indirectly the analysis of Nrf2-Keap1-ARE regulatory pathway (e.g. GreenScreen, Toxtracker, DNA multiflow). In the Toxtracker test method, for example, the Nrf2 signalling activation is determined to investigate whether oxidative stress may contribute to the genotoxic and cytotoxicity profile of a compound (Hendriks et al. 2016).
The Nfr2 transcription factor is one of the 36 biomarkers included in the cellular stress panel proposed as part of the next generation risk assessment (NGRA) approach for systemic toxicity testing designed for cosmetic ingredients by Hatherell and colleagues (Hatherell et al. 2020).
Organising the toxicity information for three substances into the matrix
To evaluate whether it is possible to organise toxicity information as shown in Fig. 2 and Fig. 3 in a real scenario, we populated the matrix with the information provided in publicly available toxicological dossiers for three rich-data substances. We chose the plant protection product Linuron and two industrial chemicals, Hydroquinone and 1,2-dichloroethane. These substances have also been used to elaborate on the concept of cross endpoint evaluation (Madia et al 2020).
Linuron (CAS no.: 330-55-2) is a herbicide, with harmonised classification as possibly carcinogenic (cat. 2) and toxic for reproduction (cat. 1B) in accordance with CLP Regulation (Regulation (EC) 1272/2008). According to the Final Renewal Report Commission Staff Working Document, Linuron is considered to have endocrine disrupting properties in accordance with Annex II to Plant Protection Products Regulation (Regulation (EC) 1107/2009). Information available for Linuron was extracted from EFSA Draft Assessment Report (DAR) and Renewal Assessment Report (RAR) (EFSA pesticides Dossiers 2020).
As reported in the DAR, a number of studies considered for the final evaluation were quite old and several results were conclusive but not sufficient for classification. Some of the most recent studies instead included in the RAR and reporting toxicology and metabolism data, were proprietary information and, as such not disclosed. Approximately more than 30 toxicity studies, regarded as valid on the base of data and experimental design quality, were summarised in the report. These were reported in the matrix and the information provided from each single study was aligned to the KCs of carcinogens (Online Resource 2).
Induce chronic inflammation, alter cell proliferation and alter nutrient supply and cell death, were confirmed to be the most investigated KCs across various toxicity studies.
The repeated dose toxicity studies (five in total) were very informative, they were performed under GLP guidelines and with a good data reporting. However, the 72.5% of the "type information" provided, was the same across the different repeated dose toxicity studies and was related to similar observed effects. Neither ED properties nor immunotoxicity conventional studies were performed.
It is worth noting that a large part of standard TGs were not filled within the table, since this information was not detailed in the DAR.
As from DAR summary evaluation: Linuron undergoes metabolic activation, is not genotoxic, induces oxidative stress and protein reaction (methaemoglobin), inflammatory response, cell proliferation, cell death (cytotoxicity) and toxicity effects to reproduction, derived by two studies (OECD 416 2001; OECD 414 2018). However, detailed mechanistic information able to describe potential receptor-mediated effects, specific to androgenic effects, and details on the carcinogenic potential was provided mainly by several additional supplementary studies both in vitro and in vivo. Interestingly, such additional studies despite not being standard studies (i.e., no TGs available) and not being performed under GLP, provided a more diverse and less redundant pattern of information (in terms of different observed effects) as compared to repeated dose toxicity studies for two specific KCs such as receptor-mediated effects and alter cell proliferation, cell death and nutrient supply, thus enriching toxicity information from 1 to 2 fold (Online Resource 2).
1,2-dichloroethane (CAS no.: 107-06-2) is an industrial chemical. According to CLP Regulation (Regulation (EC) 1272/2008), this substance may cause cancer (cat. 1B), is harmful if swallowed, causes serious eye irritation, causes skin irritation and may cause respiratory irritation.
The information for 1,2 dichloroethane derived from ECHA registration dossier (C&L Inventory 2020) (last modified 16th April 2019).
In the ECHA database, about 65 study reports for 1,2-dichloroethane were included. Several studies did not fulfil completely the requirements from internationally accepted guidelines or were not fully reported. For this reason, only key studies with a score of reliability of 1 or 2Footnote 1 (ECHA R4 2011) were considered (a total of 13 studies) (Online Resource 3).
For this chemical, genotoxicity was the most informative and studied toxicity endpoint. Moreover, the information reported aligned with several observed effects describing the 10 KCs of carcinogens.
Particularly, the study reported as DNA damage [Comet], performed under GLP, was fully detailed and information-rich. The study included also a number of observations not strictly related to the standard OECD TG. Interestingly, the comet assay was performed also on mammary gland tissue. Very little information instead, was reported for the repeated dose toxicity study, despite the complexity of the study protocol.
The majority of mechanistic and informative data derived from toxicity studies performed in compliance with GLP procedures but not following any official test guideline (OECD TG). This was also the case for a specific cancer study (key study 3) investigating in detail 1,2-dichloroethane carcinogenic effects on mammary gland tissue after inhalation exposure. This resulted as the most informative study among those reported in the dossier and specifically more informative that the two conventional cancer studies also reported. Observed effects provided by the study covered almost all the KCs of carcinogens and included information on exposure markers linked to DNA damage. Among others, study parameters measured included cage side and clinical observations, feed consumption, body weights/body weight gains, oestrous evaluations, serum prolactin levels, measurement of reduced (GSH) and oxidised (GSSG) glutathione, DCE-glutathione conjugates S-(2-Hydroxyethyl)glutathione hydrochloride (HESG) and S,S’-Ethylene-bis glutathione (EBG), DNA adducts, 8-Hydroxy-2′-deoxyguanosine (8-OH dG) and S-(2-guanylethyl) glutathione (GEG) in mammary and liver tissue, Comet assay (mammary tissue), morphometric evaluation of mammary gland structure, cell proliferation (Ki-67), and histopathology (mammary tissue).
Hydroquinone (CAS no.: 123-31-9) is an industrial chemical. According to CLP Regulation (Regulation (EC) 1272/2008), this substance is harmful if swallowed (Acute Tox cat. 4), is suspected to cause cancer (cat. 2) and to be a mutagen (cat. 2), is a skin sensitiser (cat. 1), causes serious eye damage (cat. 1) and is very toxic to aquatic life (cat. 1).
The information available in the ECHA dossier includes more than 50 studies across different toxicity endpoints. Key studies and supporting evidence with a score of reliability of 1 or 2 were considered (a total of 40 studies, plus several in vivo toxicokinetic studies) (Online Resource 4).
Electrophilicity property partly explains the strong skin sensitising and mutagenic effect of hydroquinone (Madia et al. 2020). A number of studies, mainly new in vitro methodologies, available in the ECHA dossier for skin sensitisation and genotoxicity endpoints provide a substantial portion of the substance mechanistic information that align to almost all the KCs of carcinogens and, as such, inform other toxicity endpoints. We also reported toxicity information from the more recent Toxtracker in vitro genotoxicity assay that included a number of non-genotoxic endpoints (i.e., oxidative stress, protein damage, cellular stress/ER stress pathway) associated with increased cancer hazard thus, covering multiple KCs of carcinogens. Toxicity information provided by several in vivo studies especially for acute and repeated dose toxicity was not detailed. Neither ED properties nor immunotoxicity conventional studies were performed. The 2-year cancer study (OECD 453 2018) included in the dossier was instead, informative, providing also data on relevant biomarkers of exposure effect (DNA adducts, 8-OHdG bio-product), cell proliferation and morphology, apoptosis and other observations linked to the KCs. As for the two chemicals reported above, mechanistic information on tumour promotion, cell proliferation, DNA synthesis and lipid peroxidation specific to various target organs (e.g., urinary bladder, kidney, liver) were only provided by several additional non-standard studies included in the dossier. Those included also human studies, in vitro and epidemiological investigations (Online Resource 4).
Even if details of each single study were not available in the registration dossiers or in the assessment reports, it is evident that for a number of studies, mainly in vivo, there is a redundancy of similar observations and consistent lack of information in terms of observations for various endpoints. More importantly, the majority of mechanistic information is provided by additional non-standard studies not performed in compliance with GLP. Nevertheless, the information collected for the above substances showed that the observed effects provided by various studies across different toxicity endpoints can be indeed organised and integrated in a structured way on the basis of specific toxicity properties such as the ten KCs of carcinogens.