A case example of a radiation-relevant adverse outcome pathway to lung cancer

Abstract

Background

Adverse outcome pathways (AOPs) describe how a measurable sequence of key events, beginning from a molecular initiator, can lead to an adverse outcome of relevance to risk assessment. An AOP is modular by design, comprised of four main components: (1) a Molecular Initiating Event (MIE), (2) Key Events (KEs), (3) Key Event Relationships (KERs) and (4) an Adverse Outcome (AO).

Purpose

Here, we illustrate the utility of the AOP concept through a case example in the field of ionizing radiation, using the Organisation for Economic Cooperation and Development (OECD) Users' Handbook. This AOP defines a classic targeted response to a radiation insult with an AO of lung cancer that is relevant to radon gas exposure.

Materials and methods

To build this AOP, over 500 papers were reviewed and categorized based on the modified Bradford-Hill Criteria. Data-rich key events from the MIE, to several measurable KEs and KERs related to DNA damage response/repair were identified.

Results

The components for this AOP begin with direct deposition of energy as the MIE. Energy deposited into a cell can lead to multiple ionization events to targets such as DNA. This energy can damage DNA leading to double-strand breaks (DSBs) (KE1), this will initiate repair activation (KE2) in higher eukaryotes through mechanisms that are quick and efficient, but error-prone. If DSBs occur in regions of the DNA transcribing critical genes, then mutations (KE3) generated through faulty repair may alter the function of these genes or may cause chromosomal aberrations (KE4). This can impact cellular pathways such as cell growth, cell cycling and then cellular proliferation (KE5). This will form hyperplasia in lung cells, leading eventually to lung cancer (AO) induction and metastasis. The weight of evidence for the KERs was built using biological plausibility, incidence concordance, dose-response, time-response and essentiality studies. The uncertainties and inconsistencies surrounding this AOP are centered on dose-response relationships associated with dose, dose-rates and radiation quality.

Conclusion

Overall, the AOP framework provided an effective means to organize the scientific knowledge surrounding the KERs and identify those with strong dose-response relationships and those with inconsistencies. This case study is an example of how the AOP methodology can be applied to sources of radiation to help support areas of risk assessment.

Keywords:

• Adverse outcome pathway
• key events
• molecular initiating event
• key event relationships
• lung cancer

Introduction

Decades of research have been conducted to understand the impact of environmental exposures on human health, producing vast amounts of data across many disciplines. Organizing these data in a meaningful way to support risk assessment and decision-making has been a significant challenge. To address this, the OECD formalized the adverse outcome pathway (AOP) framework to consolidate knowledge into measurable key events (KEs) which can be causally linked to disease (OECD 2018a). AOPs are hypothetical, modular representations of series of relationships linking chemical interactions with molecules in cells to phenotypic effects (Ankley et al. 2010). Thus, AOPs provide a structured approach to utilizing data from various test methods in order to inform how chemicals cause toxic effects. This is of benefit to regulatory decision-making and facilitates the identification of knowledge gaps for further focused research (Ankley et al. 2010).

Construction of an AOP involves identifying a molecular initiating event (MIE) and mapping out KEs that lead to the adverse outcome (AO) of interest to risk assessment. The KEs are sequential, measurable and linked through key event relationships (KERs), which define the structural and functional relationships (OECD 2018b). Each KE is necessary but not individually sufficient to explain an outcome. KERs are a critical element of the AOP. They are derived from compiling mechanistic understanding (biological knowledge) and empirical evidence that support the notion that the upstream KE is related through quantitative, qualitative or semi-quantitative evidence to the downstream KE. The quality of an AOP is highly dependent on the strength of underlying evidence that supports its KERs. AOPs are not stressor specific because this would otherwise hinder the power the pathway has in risk assessment (OECD 2018b). Therefore, the evidence to build an AOP can come from any type of stressor and is not limited to one stressor. The KEs and KERs of an AOP are build in a modular fashion; i.e. they are stand-alone units. This allows simple updating of independent modules as new data/tests are produced, and reuse of modules in other AOPs. A network of related AOPs can emerge that share modules, creating a powerful tool for prediction of AOs that has many potential applications in risk assessment.

The AOP concept as represented by the OECD framework is relatively new for the radiation community. Globally, more efforts are being put forth to bring awareness to the OECD AOP program. During a recent workshop held in Canada, the role of the AOP concept to support the needs of the radiation community was discussed (Chauhan et al. 2019). The workshop highlighted how AOPs can link biological endpoints to AOs to support risk-based decision making. This framework can also help identify knowledge gaps where evidence is lacking. By identifying knowledge gaps, topics of future research can be prioritized and designed to a higher degree of specificity. Due to the AOPs being posted online, they serve as a living document, evolving as knowledge becomes available and readily informing the design of future studies.

To consolidate knowledge surrounding the role of radiation stressors in adverse outcomes, we have developed an AOP using a case example of lung cancer which is relevant to radon gas exposure. Exposure to radon gas has been shown to be associated to lung cancer (Samet and Eradze 2000; Samet 2006). Following the energy deposition from radiation such as radon, we demonstrate how a sequential chain of key events related to the subsequent DNA damage and (mis)repair can directly lead to lung cancer. There are extensive data to show that the deposited energy from stressors such as radon gas can induce deleterious lesions to the DNA structure by delivering localized energy. This cytogenetic damage is an important biological effector to the downstream development of lung cancer. Using the guidance outlined in the OECD Users' Handbook, we aimed to develop an AOP centered on this widely understood KE, of DNA damage. Here we provide detailed information on adjacent KERs and associated empirical evidence in the form of the biological plausibility, quantitative and essentiality studies that formed the weight of evidence to support the components of this AOP.

Methods

OECD guidance

The AOP was developed using the criteria outlined in the OECD Users' Handbook Supplement and revised guidelines (OECD 2018a, 2018b), The complete AOP is detailed in the AOP-wiki (www.aopwiki.ca, AOP-272, 2019). This report specifically provides information on the empirical evidence that was used to support the KERs. Any additional information on the KEs and non-adjacent KERs can be found in the AOP-Wiki (www.aopwiki.ca, AOP-272, 2019).

Criteria for building AOP

A literature scan was conducted using various search engines (Google Scholar, PubMed, MEDLINE, and JSTOR) to identify studies investigating the relationship between radiation exposures and their association with the AO (lung cancer) from 1950–2018. The endpoints measured in these studies were used to inform the development of a simplified pathway from the MIE to the AO. The recommended practices featured in the AOP Users Handbook were applied in building this AOP as follows:

1. The AOP contained only one MIE (starting point) and AO (end-point).

2. The AOP was mapped in the most simplified and directed means to achieve the AO without the inclusion of biological noise (e.g. detailed metabolic pathway processes).

3. The evidence used to support the AOP was specific to the MIE and not to the stressor (i.e. data from stressors other than radiation could be used to provide evidence to support the KERs).

4. All of the proposed KEs were essential (but not necessarily sufficient on their own) to achieve the AO.

5. Empirical evidence used to support each of the KERs was in the form of dose-responses, incidence-response and temporal-response data.

6. Empirical evidence was also in the form of semi-quantitative data and qualitative data.

7. Inconsistencies and uncertainties in the KERs were described.

8. The domain of applicability for each of the KERs were outlined.

Criteria for identifying studies to support key event relationships

The information and data used to support the KERs were derived from comprehensive and authoritative reviews written by experts in the field and original, high-quality research articles on cell-lines, animals and humans. A summary of the terms and phrases used to conduct the literature search for the KEs (including the MIE and the AO) is shown in Table 1. The inclusion of key words such as biological mechanism, dose-response, temporal response, timeframe, knock-in, knock-out, feedback loop, and modulating factors were added to literature searches pertaining to the KERs.

Table 1. Summary of key words used in the literature search for each of the KEs (including the MIE and the AO).

AOP Key Event	Key words used for literature search
Direct Deposition of Energy (DDoE)	ionizing radiation, alpha particles, beta particles, gamma ray, X-ray, ions, electrons, radon, energy deposition, high LET, low LET, relative biological effectiveness, Monte-Carlo simulation, fluorescent nuclear track detectors.
Double Strand Break (DSB)	DNA double strand break, DNA damage, detect double strand breaks, measure double strand breaks, histone H2A variant (H2AX), comet assay, pulsed field gel electrophoresis, DNA cleavage assays.
Inadequate repair	non-homologous end-joining (NHEJ), homologous recombination (HR), alternative end joining (alt-EJ), mechanism, inaccurate, imprecise, incorrect, fidelity, erroneous, detect inadequate DNA repair, measure inadequate DNA repair, alkyl adduct retention, DSB repair reporters, comet assay, pulsed field gel electrophoresis, primary rat hepatocyte DNA repair assay, flow cytometry.
Mutations	point mutation, missense, nonsense, frameshift, deletion, insertion, duplication, detect mutations, measure mutations, TK mutation assay, HPRT mutation assay, PCR, transgenic rodent mutation assay, genetic sequencing.
Chromosomal aberrations (CA)	ring chromosomes, dicentric chromosomes, acentric fragments, sister chromatid exchanges, translocations, micronuclei, nucleoplasmic bridges, copy number variants (CNV), detect chromosomal aberrations, measure chromosomal aberrations, fluorescent in situ hybridization, cytokinesis block micronucleus assay, genetic sequencing.
Cell proliferation	unrestricted proliferation, tumorigenesis, neoplasm, hyperplasia, abnormal proliferation, growth, detect cell proliferation, measure cell proliferation, nucleoside analog incorporation assays (e.g. BrdU, EdU), dye assays, CyQuant cell proliferation assay.
Lung cancer	lung carcinoma, lung adenocarcinoma, malignant, tumor, cellular transformation, radon-induced lung cancer, tobacco, asbestos, detect lung cancer, measure lung cancer, magnetic resonance imaging, computed tomography, sputum analysis, xenograft assay.

Review articles describing decades of historical research served as a source of biological plausibility, lending confidence to the AOP. For original research, only studies where sufficient information relating to the radiation type used, dose, tissue, time-point, animal model, experimental procedures, and experimental results were considered to assess empirical data for the various components of the AOP. Each study was evaluated to ensure it met the quality control questions related to details on experimental protocols, exposure conditions, sufficient biological replicates, robust statistical approach and conclusions clearly presented. The types of studies that were used included knock-out, knock-in, inhibitor, dose-response, temporal-response and epidemiological. Data to support the KERs were predominantly from stressors that were ionizing radiation-specific; however, in certain cases chemical stressors were also used to support KERs, when evidence from a radiation insult was lacking. The AOP-Wiki was reviewed to identify existing KEs and KERs that could be leveraged to help support the development of this AOP. An AOP on alkylation of DNA leading to heritable mutations (AOP-15, 2019) was identified as containing relevant modules that could be expanded upon for our purposes.

Criteria to determine overall weight of evidence

The modified Bradford Hill criteria (Fedak et al. 2015) were used to assess the overall weight of evidence for this AOP. Briefly, the Bradford Hill criteria are based on nine criterion that includes: (1) strength of association, (2) consistency of response within populations, (3) specificity, (4) temporality (exposure must precede outcome), (5) evidence of dose-response relationship, (6) biological plausibility (epidemiology of an outcome must be related to biology), (7) coherence (biological changes at a molecular level can be observed at higher organizational level), (8) experiment (removal of the stressor declined the outcome) and lastly, analogy (if a strong association is made from one exposure type to a disease, the evidence required for other similar exposure types can be lowered). Among the nine criteria, biological plausibility is weighted the strongest, the remaining criteria can support the relationships and provide a means to identify knowledge gaps. By using these criteria each KER was given an overall assessment on the weight of evidence from strong, moderate to weak.

Results

Summary of the proposed AOP

This AOP focuses on a classic targeted response of energy directly deposited onto the DNA molecule. Figure 1 illustrates the adjacent and non-adjacent KEs and KERs that comprise the AOP. We note that the existing AOP on alkylation of DNA leading to heritable mutations (AOP-15, 2019) had three KEs that could be expanded upon for the purpose of supporting this AOP: DNA damage, inadequate repair, and increases in mutations.

Figure 1. The AOP of direct deposition of energy leading to DNA damage, culminating in the adverse outcome, lung cancer. Labeled in each box are key events (KEs), joined and labeled (in red) by key event relationships (KERs). Nonadjacent relationships are denoted by the dashed lines (not discussed here).

The MIE for this AOP was chosen to be the direct deposition of energy which is relevant to all radiation insults and distinguishable from chemical stressors. Direct energy deposited onto a cell can lead to multiple ionization events to targets such as DNA. The data convincingly show that DNA is a critical target. Radiation-induced energy deposition events interact with DNA strands with the possibility of inducing DNA damage in the form of strand breaks (KE1). There is an extensive array of lesions that can be induced including single strand breaks (SSBs), double strand breaks (DSBs), base damage, mismatches, DNA-DNA cross links and DNA-protein cross-links. DSBs are the most unstable and detrimental to the genome and will be the focus of this AOP. A cell will move to repair the DSBs through the activation of DNA DSB repair machinery. In higher eukaryotes, this can occur through many mechanisms but non-homologous end joining (NHEJ) is the preferred route for DSBs, particularly for noncycling cells in G1 phase of the cell cycle where no homologous chromosomes exist. The process is quick and efficient, but error prone (KE2). If DSBs occur in regions of the DNA transcribing critical genes, then mutations generated through faulty repair may alter the function of these genes (KE3) or may even cause chromosomal aberrations (KE4), resulting in genomic instability and a path toward cancer. This genomic instability will alter many gene products involved in several different cellular pathways such as cell growth, cell cycling, and apoptosis. With these alterations, cell proliferation will prosper (KE5) and form hyperplasia in lung epithelial cells, eventually, toward a path of lung cancer induction and the metastasis of the cancer (AO). Within our AOP-Wiki entry, full descriptions of how each of these KEs work and the methods used to measure them are described, and we direct the reader there for this information (AOP-272, 2019). In the following discussion, we describe the domain of applicability for the AOP and provide a detailed description of the evidence and knowledge supporting each of the KERs. Note that we present information relating to adjacent KERs; non-adjacent KERs are used in the full, aforementioned AOP-Wiki entry as they provide additional support for the overall AOP.

Domain of applicability

This AOP is relevant to all mammals and is independent of sex (Eymin and Gazzeri 2010; Barron et al. 2014; Kurgan et al. 2017). The pathway leading to the development of lung cancer often occurs during adulthood but may be applicable at earlier life stages (Liu et al. 2015) and is independent of sex. In humans, however, genetic abnormalities/mutations suggestive of lung cancer risk seem to be influenced by confounders such as smoking history, age, sex, ethnicity and genotype (Lim et al. 2009; Sanders and Albitar 2010; Kim et al. 2012; Paik et al. 2012; Lloyd et al. 2013; Cortot et al. 2014; Minina et al. 2017).

Molecular initiating event

The MIE represents the first critical physical interaction within the cell through the direct deposition of energy onto the DNA molecule. The word ‘direct’ is used to distinguish the MIE from instances of ‘indirect’ damage to the DNA from reactive species generation through early dispersion of radiolytic products. It was hypothesized that the distributions and/or type of DSBs may be different from a chemical insult and it was therefore important to differentiate this effect by including a MIE that was specific to radiation. As research surrounding the link between the deposition of energy and the induction of DSBs advances, more critical KEs between the MIE and DSBs could emerge, and it would become important to capture these in the AOP. Direct deposition of energy refers to events where subatomic particles or electromagnetic waves of sufficient energy cause direct ionization to genetic material through which they transverse (Kaczmarska et al. 2017). The energy of these subatomic particles or electromagnetic waves is dependent on the source and type of radiation. When this energy is deposited, it may cause the ejection of electrons from atoms and molecules, thereby breaking different types of chemical bonds and ionizing the atoms and molecules. Not all electromagnetic radiation is ionizing; the incident radiation must have sufficient energy to free electrons from the atom or the molecule’s electron orbitals. Many of the downstream biochemical effects of energy deposition are due to the initial geometry of the physical energy deposition event, known as the track structure (Desouky et al. 2015). Thus, the ionization produced is usually localized on a track. Energy deposition can affect cells at the molecular, tissue and organ level.

KER1: Direct deposition of energy (Mie) leading to DNA double strand breaks (KE1)

It is well established that ionizing radiation can cause damage to the DNA in the form of strand breaks through direct (focus of this AOP) and indirect mechanisms (reviewed in Lomax et al. 2013). A significant fraction of strand breaks can also be generated by the splitting of water molecules near DNA (indirect). This can result in a significant quantity of reactive oxygen species in the form of hydroxyl free radicals. These short-lived but highly reactive molecules can attack nearby DNA, producing single-strand DNA breaks; if several single-strand breaks are in close proximity to each other, they may spontaneously convert into a DSB (Ward 1988; Yamaguchi et al. 2005). DSBs are considered the most lethal and deleterious type of DNA damage. If DSBs are left unrepaired, they can drive the cell toward apoptosis or senescence. If they are misrepaired, the DSBs can lead to clastogenic lesions, resulting in genomic instability and potentially tumorigenesis through the activation of oncogenes or inactivation of tumor-suppressor genes (Raynard et al. 2017).

The most direct path to generate DSBs entails a collision between a high-energy particle or photon (e.g. X-rays and gamma-rays) and a strand of DNA. The high-energy subatomic particles interact with the orbital electrons of the DNA; this results in ionization and excitation (Joiner and Van Der Kogel 2009) ultimately breaking the phosphodiester backbone. Additionally, excitation of secondary electrons in the DNA allows for a cascade of ionization events to occur, which can lead to the formation of multiple damage sites (Joiner and Van Der Kogel 2009). High linear energy transfer (LET) radiation, such as alpha particle radiation, can deposit larger quantities of energy within a single track than low LET radiation, such as gamma-ray radiation. High LET radiation has a propensity to produce clustered damage, meaning that as the radiation transverses the media, it induces complex DNA damage comprised of closely spaced DNA lesions (Hada and Georgakilas 2008). The complexity and yield of clustered DNA damage increases with ionizing density (Ward 1988; Goodhead 2006).

The mechanisms of this KER (biological plausibility) are widely understood and there are many reviews available on this topic. Results from in vitro (Sutherland et al. 2000; Rogakou et al. 1999; de Lara et al. 2001; Rothkamm and Lobrich 2003; Kuhne et al. 2005; Sudprasert et al. 2006; Beels et al. 2009; Grudzenski et al. 2010; Antonelli et al. 2015; Shelke and Das 2015; Fernandez-Antoran et al. 2019), in vivo (Sutherland et al. 2000; Rube et al. 2008; Beels et al. 2009; Grudzenski et al. 2010), ex vivo (Rube et al. 2008; Flegal et al. 2015) and simulation studies that used biological experimental data (Charlton et al. 1989) demonstrate that there is a dose-dependent increase in DSBs with increasing deposition of energy across a wide range of ionizing radiation (iron ions, X-rays, ultrasoft X-rays, gamma-rays, photons, and alpha particles) and doses (1 mGy–100 Gy). There is also considerable evidence suggesting a time concordance. A number of different models have documented DSBs at 10–30 minutes after exposure to ionizing radiation (Rogakou et al. 1999; Rube et al. 2008; Kuefner et al. 2009; Beels et al. 2009, Grudzenski et al. 2010; Antonelli et al. 2015), with evidence of DSBs occurring as early as 3–5 minutes postirradiation (Rogakou et al. 1999; Rothkamm and Lobrich 2003; Rube et al. 2008; Grudzenski et al. 2010). In the majority of cases, the number of DSBs decline to near baseline levels by 24 hours after irradiation (Rothkamm and Lobrich 2003; Rube et al. 2008; Grudzenski et al. 2010; Antonelli et al. 2015), with the sharpest decrease documented within the first 5 hours (Rogakou et al. 1999; Rube et al. 2008; Kuefner et al. 2009; Grudzenski et al. 2010; Shelke and Das 2015). DSBs were found to be more persistent when they were induced by higher LET radiation (Antonelli et al. 2015). Table 2 summarizes the quantitative evidence of the aforementioned studies for this KER. Reviews do not appear in the table.

Table 2. Summary of articles forming the empirical evidence (EE) for KER1 which is derived using dose concordance (EE:DC), time scale (EE:TS) and essentiality studies (Ess).

Reference	Relationship Summary
Charlton et al. (1989)	(EE:DC): Simulated model showing increased DSBs/54 nucleotide pairs as DDoE increases in the range 75–400 eV.
Sutherland et al. (2000)	(EE:DC): Observed dose-concordance in DSBs in human cells irradiated with iron ions in the range 0–0.70 Gy. Yield of 2.4 DSBs / per Mbp per Gy.
Rogakou et al. (1999)	(EE:DC): Radiation doses of 0.6 Gy and 2 Gy to human cells resulted in 10.1 and 12.2 DSBs per nucleus on average (0.6 Gy), respectively; increasing to 24 and 27.1 DSBs per nucleus (2 Gy). (EE:TS): DSBs present at 3 min, persisted from 15–60 min, and then were decreased to almost baseline levels by 180 min.
Rothkamm and Lobrich (2003)	(EE:DC): X-ray irradiation of human cells at 1 mGy–100 Gy. Dose-dependent linear increase in DSBs at 35 DSBs per cell per Gy. (EE:TS): DSBs observed within 3 min, decreased over 24 h.
Beels et al. (2009)	(EE:DC): Quadratic fit relationship between X-rays and DSBs at 0–6 mGy and a linear fit at 6–500 mGy from human cell data. (EE:TS) DSBs present 15 minutes after exposure to X-rays.
Kuefner et al. (2009)	(EE:DC): Single radiation dose to human cells resulted in a peak in DSBs 15-min post-irradiation. Fractionated dosing resulted in a peak after completion of the dosing. (EE:TS): 2/3 of DSBs decreased by 2.5 h.
Flegal et al. (2015)	(EE:DC): Exposure of human cells to gamma-radiation at 0.1 and 2 Gy resulted in an increase in the number of DSBs, with a much higher increase with 2 Gy compared to 0.1 Gy.
Shelke and Das (2015)	(EE:DC): Linear, positive relationship between radiation dose and DSBs in human cells exposed to X-rays. (EE:TS): DSBs decreased fastest within the first hour following irradiation with a rate of 0.6% repair per min with a slower rate of repair at subsequent time periods.
Sudprasert et al. (2006)	(EE:DC): Positive relationship between the radiation dose and the quantity of DSBs in human cells exposed to gamma-rays.
Kuhne et al. (2005)	(EE:DC): Linear increase in DSBs in human cells with dose from 0–70 Gy, Gamma-rays: 6.1 ± 0.2 x 10^−9 DSBs per base pair per Gy, X-rays: 7 ± 0.2 x 10^−9 DSBs per base pair per Gy, and Carbon-K-shell X-rays: 12.1 ± 1.9 x 10^−9 DSBs per base pair per Gy.
Grudzenski et al. (2010)	(EE:DC): Linear increase in DSBs according to dose from 0–200 mGy, with ∼21 foci per cell. In rodent cells, a linear increase in DSBs after 10 mGy, 100 mGy and 1 Gy doses. (EE:TS): In rodent cells, DSBs present 5 min-post irradiation, decreased over time. In mice organs, DSBs were present at 10 min, sharply decreased by 5 h, disappeared by 24 h for 100 mGy and 1 Gy doses, persisted at a 10 mGy dose.
De Lara et al. (2001)	(EE:DC): Linear increase in DSBs in rodent cells after irradiation with ultrasoft titanium and carbon K-shell X-rays and gamma rays.
Rube et al. (2008)	(EE:DC): Linear dose-dependent increase in DSBs in the brain, small intestine, lung and heart of mice after whole-body irradiation with 0.1–1.0 Gy of radiation.
Antonelli et al. (2015)	(EE:DC): Linear dose-dependent increase in the number of DSBs in human cells from 0–1 Gy for gamma-rays and alpha particles 30-min post-irradiation. (EE:TS): Maximum number of DSBs observed at 30 min for all radiation types, followed by a time-dependent decrease in the number of DSBs over 24 h.
Fernandez-Antoran et al. (2019)	(EE:DC): Direct response relationship between foci / cell and radiation doses between 16 mGy–1 Gy using a cesium gamma irradiator and animal model.
Mosconi et al. (2011)	(EE:TS): DSBs induced in human cells at 25 sec after interaction of an alpha particle and 110 sec after interaction with protons with the cell. Foci numbers plateau by 200 sec post-alpha particle irradiation. and ∼800 sec after proton radiation.

Full details of study can be found in AOP wiki. The quantified values are extrapolated directly from the indicated references.

KER2: double-strand breaks (KE1) leading to inadequate repair (KE2)

The maintenance of DNA integrity is essential for genomic stability. As such, when DNA DSBs occur, the cell reacts by initiating repair machinery. To fix this potentially lethal damage, eukaryotic cells have evolved a variety of repair pathways that rely on homologous and illegitimate recombination, also called nonhomologous DNA end joining (NHEJ), which may induce small scale mutations and chromosomal aberrations (Pfeiffer et al. 2000). Homologous recombination (HR) is used by cells when there is a homologous piece of DNA. In mitotic cells NHEJ occurs throughout all phases of the cell cycle, whereas HR is largely restricted to the S and G2 phases when the sister chromatid is available to mediate the repair process (Raynard et al. 2017). The type of DSB repair pathway that activates can vary widely among species, between different cell types of a single species, and during different cell cycle phases (Shrivastav et al. 2008). NHEJ is initiated in G1 and early S phases of the cell cycle (Lieber et al. 2003) and is preferentially used to repair DSB damage (Godwin et al. 1994), as it is more rapid than HR (Iliakis 1991; Jeggo 1998; Mao et al. 2008). In higher-order eukaryotes such as humans, NHEJ is relied upon when homology of DNA is scarce or unavailable, as NHEJ is highly flexible and can deploy multiple enzymes to process the various types of DNA ends (Pannunzio et al. 2018). NHEJ can occur through one of two subtypes: canonical NHEJ (C-NHEJ) or alternative non-homologous end joining (alt-NHEJ). C-NHEJ, as the name suggests, simply ligates the broken ends back together. In contrast, alt-NHEJ occurs when one strand of the DNA on either side of the break is resected to repair the lesion (Betermier et al. 2014). Both DSB repair mechanisms are error-prone, meaning insertions and deletions are sometimes formed due to the DSBs being repaired imperfectly (Thurtle-Schmidt and Lo 2018). However, alt-NHEJ is considered more error-prone than C-NHEJ, as studies have shown that it more often leads to chromosomal aberrations (Zhu et al. 2002; Guirouilh-Barbat et al. 2007; Simsek and Jasin 2010). The roles of various DNA repair proteins in the context of DSBs are highlighted in reviews by Chang et al. (2017) and van Gent et al. (2001), with discussions focusing on the various effects of losing some of these proteins in cells, mice and humans (van Gent et al. 2001).

There is evidence in the literature of dose/incidence concordance between the occurrence of DSBs and the incidence of inadequate DNA repair upon exposure to radiation. DNA repair is tightly coupled to DSBs. Visually, immunofluorescence has demonstrated a co-localization of DNA repair proteins with DSB foci in response to a radiation stressor (Paull et al. 2000; Asaithamby and Chen 2009; Dong et al. 2017). In studies examining cellular responses to increasing doses of radiation, which is known to evoke DNA DSBs (Dikomey and Brammer 2000; Lobrich et al. 2000; Rothkamm and Lobrich 2003; Kuhne et al. 2005; Asaithamby and Chen 2009; Bracalente et al. 2013), there were resulting dose-dependent increases in DSB repair foci (Asaithamby and Chen 2009), non-repaired DSBs (Dikomey and Brammer 2000), DSB misrepair rates (McMahon et al. 2016), and misrejoined DSBs (Kuhne et al. 2005; Rydberg et al. 2005), as well as a dose-dependent decrease in the total DSB rejoining (Lobrich et al. 2000). Furthermore, only 50% of the rejoined DSBs were found to be correctly repaired after high doses (Lobrich et al. 2000). When an inhibitor of a DNA repair protein was added to cells prior to exposure to a radiation stressor, DNA repair foci were not formed post-irradiation (Paull et al. 2000), and there were significant increases in DSBs at 6 hours and 12 hours after the radiation treatment (Dong et al. 2017). Similarly, there have been several knock-out/knock-down studies in which cells lacking a DNA repair protein have been exposed to a radiation stressor. As a result, DSBs were found to persist in these cells longer than in the wild-type cells (Rothkamm and Lobrich 2003; Bracalente et al. 2013; McMahon et al. 2016; Dong et al. 2017), and there was an increase in incorrectly rejoined DSBs (Lobrich et al. 2000). In one striking example, a human cell line lacking DNA ligase IV had DSBs that were still present approximately 240–340 hours post-irradiation (McMahon et al. 2016).

Complete recovery or repair may require days or weeks and data suggests that, even at moderately high doses, if the dose rate is low enough or the time interval between doses is adequate, the DNA will be fully repaired. In general, it seems that most DSB repair occurs within the first 3–6 hours after irradiation, based on expression of DNA repair proteins (Asaithamby and Chen 2009; Dong et al. 2017) and the observed fraction of unrepaired DSBs (Rothkamm and Lobrich 2003; Pinto and Prise 2005). A summary of the studies contributing to this KER is shown in Table 3.

Table 3. Summary of articles forming the empirical evidence for KER2 which is derived using dose concordance (EE:DC), time scale (EE:TS) and essentiality studies (Ess).

Reference	Relationship Summary
Dikomey and Brammer (2000)	(EE:DC): Dose-dependent increase in the number of DSBs in human cells, and those left unrepaired from 0–80 Gy and 0–180 Gy, respectively. (EE:TS): Time-dependent decrease in the number of DSBs until 12 h and then a plateau.
Dong et al. (2017)	(EE:DC): 5 Gy of X-ray irradiated mouse cells results in increased DSBs and activation of DSB repair proteins. (EE:TS): DSBs appear within 0.5 h and are resolved by 12 h. DSB repair proteins activated within 0.25 h. (Ess): Knock-out studies demonstrate that Ku 70 and DNA PKcs play role in decreasing the number of radiation-induced DSBs in a timely fashion. Inhibition study shows DNA-PK is necessary to decrease radiation-induced DSBs.
Rothkamm and Lobrich (2003)	(EE:DC): Irradiation of human cells with 1 mGy to 100 Gy X-rays reveal dose-dependent linear increase of DSBs per cell (35 DSBs per cell per Gy). (EE:TS): DSB induction within 3 min, declined over 24 h or until ∼0.1 foci per cell. (Ess): In DNA ligase IV-deficient cells, radiation induced the same number of DSBs as per WT, but repair kinetics slower such that more DSBs remained after 24 h.
Pinto and Prise (2005)	(EE:TS): Irradiation of human cells with 115 Gy X-rays or 100 Gy alpha-particles produce an increase in DSBs at 10 min post-irradiation. Repair of the DSBs progress over 48 h.
McMahon et al. (2016)	(EE:DC): Increase in DSB misrepair rate with increasing dose from 0–80 Gy in mouse and human cells. (EE:TS): Time-dependent decline in the DSB fraction over 20 h in normal human and mouse cells. (Ess): Mouse and human cell lines with defective components of NHEJ had slower DSB repair kinetics than wilde type cells.
Kuhne et al. (2005)	(EE:DC): Linear increase in DSBs and a linear-quadratic increase in misrejoined DSBs, suggesting that there are more DSBs and less correct DSB repair with increasing doses of radiation in human cells exposed to gamma rays.
Lobrich et al. (2000)	(EE:DC): Linear dose-dependent increase in the number of DSBs with increasing X-radiation dose. (EE:TS): Decrease in DSBs over time, with approximately 50% of DSBs being correctly repaired. (Ess): Ataxia telangiectasia patient cells more incorrectly repaired DSBs than wild type or heterozygous cells at smaller radiation fractions (5 and 10 Gy, given 16 and 8 times).
Rydberg et al. (2005)	(EE:DC): Curvilinear increase in the number of mis-rejoined DSBs in human cells exposed to increasing dose of X-rays, more linear response for high LET particles (which had higher numbers of mis-repaired DSBs relative to X-rays).
Bracalente et al. (2013)	(EE:DC): Significant linear dose-response 30 min post-irradiation with increasing number of DSBs from 0–2 Gy.in rodent cells exposed to gamma radiation (EE:TS): Significant increase in DSBs at 30 min relative to controls, and a decline in the number of DSBs from 30 min to 6 h due to DSB repair. (Ess): Knocking out DNA-PKcs, a DNA repair protein, resulted in persisting DSBs and therefore decreased DNA repair relative to wild type cells.
Asaithamby and Chen (2009)	(EE:DC): Linear increase in both DSBs and DNA repair in human cells exposed with increasing gamma radiation dose. (EE:TS): DNA repair foci, which were demonstrated to be reflective of DSB foci, appeared at 3–5 min post-irradiation, increased until 30 min, declined to 45% of original amount by 2 h, and were mostly resolved by 8 h.
Paull et al. (2000)	(EE:TS): DSBs that colocalized with repair proteins present within 30 min post-irradiation, though different proteins colocalized at different times in human cells exposed to gamma rays. (Ess): Inhibition of DNA-PK prior to irradiation prevented formation of DNA repair foci. In cells deficient in DNA repair, there were increased DSBs present before irradiation (gamma rays).
Rube et al. (2008)	(Ess): DSBs more persistent in multiple tissue types after irradiation when DNA repair was inadequate.
Kuhne et al. (2005)	(EE:TS): At 80 Gy and higher, the percentage of DSB misrejoining remains at approximately 50%. At 24 h post-irradiation, the frequency of misrejoined DSBs was at 50% for X-ray and alpha particle irradiation; For 80 Gy of alpha particle irradiation, the % of DSB misrejoining after 24 h remained constant around 80% for 10 days after an 80 Gy dose of alpha particles.

Full details of study can be found in AOP wiki. The quantified values are extrapolated directly from the indicated references.

KER3: inadequate repair (KE2) to mutations (KE3)

The preferred method for DSB repair is NHEJ in human somatic cells (Petrini et al. 1997; Mao et al. 2008). However, this mechanism is error-prone and has the potential to create mutations from DNA lesions (Little 2000). NHEJ is considered error-prone because it does not use a homologous template to repair the DSB. The NHEJ mechanism involves many proteins that work together to bridge the DSB gap by overlapping single-strand termini that are usually less than 10 nucleotides long (Anderson 1993; Getts and Stamato 1994; Rathmell and Chu 1994). Inherent in this process is the introduction of errors. These errors can be retained resulting in damaged DNA that is then used as a template during DNA replication. There is evidence in the literature suggesting a dose/incidence concordance between inadequate DNA repair and increases in mutation frequencies. In response to increasing doses from a radiation stressor, dose-dependent increases in both measures of inadequate DNA repair and mutation frequency have been found. In an analysis that combined results from several different studies conducted using in vitro cell-lines, the rate of DSB misrepair was revealed to increase in a dose-dependent fashion from 0 to 80 Gy, with the mutation rate also similarly increasing from 0–6 Gy (McMahon et al. 2016). Additionally, using a plant model, it was shown that increasing radiation dose from 0–10 Gy resulted in increased DNA damage as a consequence of inadequate repair. Mutations were observed 2–3 weeks post-irradiation (Ptacek et al. 2001). Moreover, increases in mutation densities were found in specific genomic regions of human cancer samples (namely promoter DNAse I-hypersensitive sites (DHS) and 100 bp upstream of transcription start sites (TSS)) that were also found to have decreased DNA repair rates attributable to inadequate nucleotide excision repair (NER) (Perera et al. 2016).

Interestingly, mutation rates have been shown to increase as the required DNA repair becomes more complex. Upon completion of DSB repair in response to radiation and treatment with restriction enzymes, more mutations were found in cases where the ends were non-complementary and thus required more complex DNA repair (1–4% error-free) relative to cases where ends were complementary (34–38% error-free) (Smith et al. 2001). Knock-out/knock-down studies suggest that there is a strong relationship between the adequacy of DNA repair and mutation frequency. In all examined cases, deficiencies in proteins involved in DNA repair resulted in altered mutation frequencies relative to wild-type cases. There were significant decreases in the frequency and accuracy of DNA repair in cell lines deficient in key proteins involved in DNA repair, including LIG4 (Smith et al. 2003) and Ku80 (Feldmann et al. 2000); rescue experiments performed with these two cell lines further confirmed that inadequate DNA repair proteins was the cause of the observed decreases in repair frequency and accuracy (Feldmann et al. 2000; Smith et al. 2003). In primary Nibrin-deficient mouse fibroblasts, there was increased spontaneous DNA damage relative to wild-type controls, suggestive of inadequate DNA repair. Using the corresponding Nibrin-deficient and wild-type mice, in vivo mutation frequencies were also found to be elevated in the Nibrin-deficient animals (Wessendorf et al. 2014). Furthermore, mutation densities were differentially affected in specific genomic regions in cancer patients depending on their XPC status. Specifically, mutation frequencies were increased in XPC-wild-type patients at DNase I-hypersensitive sites and 100 bp upstream of transcriptional start site relative to cancer patients lacking functional XPC (Perera et al. 2016). There is evidence in the literature suggesting a time concordance between the initiation of DNA repair and the occurrence of mutations. For simple ligation events, mutations were not evident until 12–24 hours, whereas DSB repair was evident at 6–12 hours. For complex ligation events, however, mutations and DSB repair were both evident at 12–24 hours. As the relative percent of DNA repair increased over time, the corresponding percent of error-free rejoining decreased over time in both ligation cases, suggesting that overall DNA repair fidelity decreases with time (Smith et al. 2001). A summary of the studies discussed for KER3 is provided in Table 4.

Table 4. Summary of articles forming the empirical evidence for KER3 which is derived using dose concordance (EE:DC), time scale (EE:TS) and essentiality studies (Ess).

Reference	Relationship Summary
Amundson and Chen (1996)	(Ess): Decreased DNA repair was observed in a human cell line with decreased DNA repair capabilities resulting in at least a 4-fold increase in the frequency of mutations relative to normal repair capacity cells.
McMahon et al. (2016)	(EE:DC): From 0–80 Gy of X-rays, there is a positive relationship between mis-repair and radiation dose; From 0–6 Gy, there is a positive relationship between mutation rate and radiation dose.
Smith et al. (2001)	(EE:TS): Repaired plasmids were first detected between 6–12 h for simple ligation events and between 12–24 h for complex ligation events; this first period was when the most error-free rejoining occurred. After this initial period of repair until its completion at 48 h, repair became increasingly more erroneous such that mutations were found in more than half of the repaired plasmids at 48 h.
Smith et al. (2003)	(Ess): Without reliable DSB repair proteins (hypomorphic mutations or knock-out of LIG4), repair did not occur as efficiently nor as accurately as it would if the wild-type protein was present.
Feldmann et al. (2000)	(Ess): Without reliable DSB repair proteins (Ku −/−), repair does not occur as efficiently nor as accurately as it would if the wild type protein was present.
Ptacek et al. (2001)	(EE:TS): DNA damage was evident immediately following gamma radiation with 30 Gy (within 30 sec of irradiation), with 51.7 min required to repair 50% of damage, 4 h required to repair 80% of damage, and damage completed repaired by 24 h; mutations were present 2–3 wk post-irradiation.
Wessendorf et al. (2014)	(Ess): Knock-out of reliable DSB repair proteins , repair does not occur as efficiently as it would if the wild type protein was present.
Perera et al. (2016)	(EE:DC): Inverse correlation between nucleotide excision repair (NER) and mutation density at transcription start site; Also significant negative correlation between NER and mutation density within DNase I-hypersensitive sites and flanking region; (Ess): When NER is globally inhibited in the genome (in XPC−/− patients), parts of the genome that are normally heavily mutated in cancer are not affected; But when NER is functioning properly (in XPC wild type patients), these same regions are highly mutated due to regional depletion of NER; Knocking -out a component of the repair pathway changes the mutation densities in specific genomic locations in cancer patients.

Full details of study can be found in AOP wiki. The quantified values are extrapolated directly from the indicated references.

KER4: inadequate repair (KE2) to chromosomal aberrations (KE4)

Chromosomal aberrations are physical changes in the structure of a chromosome that may or may not lead to downstream effects (reviewed in Nikjoo et al. 1999). They can be the result of terminal and interstitial deletions and inversions or rearrangements between two or more chromosomes. It is generally accepted that mechanisms involved in the repair of DSBs are highly correlated with the formation of chromosomal aberrations. Through NHEJ, the ends of chromosomes can be ligated together; this ligation may cause translocations (van Gent et al. 2001). If the ends of two different chromosomes are ligated, this has the potential to fuse genes, including genes that are critical to normal tissue function or oncogenes. It has also been observed that incorrect joining of the exchanged ends of two or more DSBs leads to a loss of the DNA sequences of varied size at the breakpoints (Povirk 2006). Characteristic hallmarks of NHEJ include the high frequency of small terminal deletions, the apparent splicing of DNA ends at micro-homologies, and gap-filling on aligned DSB ends (Povirk 2006). The most convincing support for this is derived from the analysis of chromosome instability syndromes, rare human autosomally inherited disorders which are associated with dramatically increased frequency of chromosomal aberrations. In these syndromes, components of NHEJ repair protein complexes have been shown to be mutated (Pfeiffer et al. 2000). Other phenomena, such as chromothripsis, also highlight the role of NHEJ in chromosomes being joined randomly through incorrect repair of DNA DSBs on a single chromosome (Leibowitz et al. 2015; Rode et al. 2016). There is also the possibility that some oncogenes can fuse due to incorrect repair of DSBs, which amplifies chromosome instability (Seki et al. 2015). Furthermore, this seemingly desperate ligation of DSB ends tends to lead to the pairing of two incompatible ends from different chromosomes, causing genomic rearrangements known as translocations (Wang-Michelitsch and Michelitsch 2007). In human cells, NHEJ has been shown to be directly responsible for the generation of translocations (Rothkamm et al. 2001; Weinstock et al. 2006; Lieber et al. 2010; Ghezraoui et al. 2014). Interestingly, in mice classical non-homologous end joining (C-NHEJ) has been shown to suppress translocations, suggesting that errors arise from alternative non-homologous end joining (alt-NHEJ) (Simsek and Jasin 2010). Despite these negative consequences, it has been suggested that from a biological perspective, it is more advantageous for the cell to rejoin DSBs and risk chromosomal aberrations rather than not ligating the DSBs at all; non-ligated DSBs are more toxic and could easily lead to cell death via apoptosis (Hoeijmakers 2001).

Studies have also shown that the loss of core NHEJ-associated proteins leads to an increased occurrence of chromosomal aberrations and translocations in human and rodent cells (Karanjawala et al. 1999; Rothkamm et al. 2001; Jeggo and Lobrich 2015). Such inhibitions or defects in DSB repair can increase the sensitivity of the cells to stressors such as ionizing radiation and chemical genotoxicants, as reviewed by Rassool and Tomkinson (2010). Rejoining of DSBs in cells deficient in key components of the NHEJ complex (e.g. ligase 4 or X-ray cross-complementing protein 4) has shown an increase in the number of translocations (Ghezraoui et al. 2014). Evidence has also emerged demonstrating that deregulated NHEJ may promote carcinogenesis by increasing genomic instability due to inappropriate repair (reviewed in Lin et al. 2014 and Sishc and Davis 2017). Micronuclei and translocation formation have also been shown to increase when DSB repair is inhibited (Chernikova et al. 1999; McMahon et al. 2016). Treatment of human cells with methylating agents also has been shown to trigger a significant increase in chromosomal abnormalities due to an increased frequency of incorrect DSB repair (Galloway 1995). This trend was also observed with other DSB-inducing agents when specific cell cycle checkpoints were inhibited (Janssen et al. 2011). Active overexpression of core NHEJ factors may also lead to chromosomal instability (Sishc and Davis 2017). This can arise if DSBs occurring on genes that produce core NHEJ factors are improperly repaired, leading to mutations which can either inactivate or overexpress these genes. In many kinds of myeloid leukemias, large deletions were found to be associated with an increased activity of NHEJ proteins (Gaymes et al. 2002). It is thought that Ku, a key protein of DNA repair, drives the overactive expression of the NHEJ pathway. Thus aberrant Ku activity is a candidate mechanism for inducing chromosomal instability (Gaymes et al. 2002). A full summary of studies is provided in Table 5.

Table 5. Summary of articles forming the empirical evidence for KER4 which is derived using dose concordance (EE:DC), time scale (EE:TS) and essentiality studies (Ess).

Reference	Relationship Summary
Cornforth and Bedford (1985)	(EE:TS): Repair-deficient human cells displayed prematurely condensed chromosome fragments within the first 10–15 hours, with no further changes within15–20 hours. (Ess): Loss of a protein involved in DNA repair resulted in less repair of DNA breaks and more chromosome fragments.
Wilhelm et al. (2014)	(Ess): Knock-out of either RAD51 or BRCA2 results in homologous recombination-deficient cells with increased levels of CAs (in the form of anaphase chromatin bridges) and aberrant mitosis.
Simsek and Jasin (2010)	(EE:TS): Translocations evident within 72 h of endonuclease treatment, inducing DSBs in wild type and knock-out cells. (Ess): Inhibition of NHEJ through knocking-out either Ku70 or Xrcc4 resulted in higher CA frequencies.
Patel et al. (1998)	(Ess): Knock-out of BRCA2 significantly affects DSB repair such that CAs were increased.
Chernikova et al. (1999)	(EE:DC): Negative, linear relationship between DNA repair and Wortmannin dose. There was a positive, approximately linear relationship between Wortmannin dose and micronuclei (MN )frequency. (EE:TS): DNA repair was evident by 3 h and MN by 24 h post-irradiation and post-Wortmannin treatments. (Ess): Inhibition of DNA-PK (a member of NHEJ) resulted in dose-dependent decreases in repair and increases in MN.
Lin et al. (2014)	(EE:DC): Positive relationship between radiation dose and CAs or sister chromatid exchanges (SCEs) for both gamma and alpha radiation in wild type cells and most NHEJ mutant cell lines. (Ess): Deficiency in NHEJ repair generally resulted in increased CAs and SCEs in response to radiation, but not for HR-deficient cells.
McMahon et al. (2016)	(EE:DC): Positive relationship between radiation dose and DSB misrepair rates, radiation dose and total CAs, dicentrics and large deletions. (EE:TS): CAs were evident within 10 min of radiation, and increased from 0–20 min. CAs levels plateau between 20–40 min and remain constant until 100 min. (Ess): Sharply increased levels of radiation-dose dependent CAs in NHEJ defective cells vs normal.
Karanjawala et al. (1999)	(Ess): NHEJ has an important role in preventing CAs, as cells heterozygous for Ku86 or DNA ligase IV had increased levels of CAs vs wild-type cells, while Ku86 −/− cells had even higher CA levels.
Gaymes et al. (2002)	(Ess): Reducing the availability of various NHEJ proteins decreases the ligation efficiency in myeloid leukemia cells and healthy controls; Misrepair frequency was higher in myeloid leukemia cells with known CAs relative to healthy, normal cells.
Thomas et al. (2003)	(EE:DC): Positive relationship between the radiation dose and the frequency of CAs.
Balajee et al. (2014)	(EE:DC):Positive relationship between the radiation dose and the frequency of MN.
Arlt et al. (2013)	(EE:DC): Positive relationship between the radiation dose and the frequency of de novo copy number variants per cell when cells were replated immediately following irradiation.
George et al. (2009)	(EE:DC): For complex aberrations in HF19, AT and NBS cells, the best fit was a quadratic relationship for both gamma-rays and iron ions; For simple aberrations induced by iron ions, there was a linear relationship; In M059K and M059J cells, a linear response was found for simple aberrations for both types of radiation, while the models were poorly fit overall due to downward curvature at the higher doses for complex aberrations.
Suto et al. (2015)	(EE:DC): Positive, linear quadratic relationship between the radiation dose and the aberration yield. Dose-response curve able to be used to predict doses from same donor and a different donor.
Tucker et al. (2005)	(EE:DC): Positive relationship between radiation dose and number of CAs. (EE:TS): CAs were present by 2 day post-irradiation, and declined by 7 days in all cases.

Full details of study can be found in AOP wiki. The quantified values are extrapolated directly from the indicated references.

KER5 and KER6: mutations (KE3)/chromosomal aberration (KE4) leading to cell proliferation (KE5)

Clastogenic lesions such as mutations and chromosome aberrations can occur in genes regulating cell cycling processes. If this occurs, it can lead to unnatural and potentially detrimental levels of increased cell proliferation. Persistent clastogenic lesions can induce cell proliferation through three different mechanisms (Vogelstein and Kinzler 2004). The first is through activation of proto-oncogenes via gain of function (GOF) mutations. This activation can occur through gene amplification or subtle intragenic mutations that alter regulatory sites within the gene, thereby placing these genes under the regulation of constitutively activated genes, such as MYC (Vogelstein and Kinzler 2004; Varella-Garcia 2009) that regulate cellular proliferation. The second mechanism involves the inactivation of tumor suppressor genes (TSGs) through a loss of function (LOF) mutation. These TSG inactivations usually occur due to epigenetic events such as methylation of promoter regions, missense mutations at residues essential for protein function, or point mutations resulting in the formation of truncated proteins (Vogelstein and Kinzler 2004). Unlike proto-oncogenes, which only require mutations in one copy of the gene to activate the oncogene, TSGs require mutations in both copies of the gene for inactivation (Vogelstein and Kinzler 2004). The final mechanism for induction of cell proliferation is the alteration of a stability gene, which directly or indirectly affects the activation status of proto-oncogenes or TSGs. When these stability genes, also known as caretakers, are mutated, they promote tumorigenesis in a manner that differs from that of oncogenes and TSGs. Genes involved in mismatch repair, nucleotide-excision repair, and base-excision repair, are prominent examples of stability genes. These stability genes are responsible for repairing subtle mistakes in critical genes made during normal DNA replication and through exposure to mutagens such as ionizing radiation or chemical agents (Vogelstein and Kinzler 2004). A few of the most important critical genes in terms of controlling cell proliferation are KRAS, EGFR, and TP53. Empirical data obtained from a number of direct sources strongly support the idea that an increase in mutations or chromosomal aberrations in these specific genes increases cell proliferation. The majority of the evidence has been derived from studies that have examined lung tissue from patients with lung cancer. The studies either used genotyping or knock-in and knock-out constructs that manipulate TP53, p53 proteins, or other genes and proteins that have an influence on p53 pathways. Thus, the evidence primarily takes the form of essentiality studies rather than dose, incidence and temporal concordance. For both KER5 and KER6, a summary of the studies and evidence to support mutations and chromosomal aberrations as indicators later cell proliferation is provided in Tables 6 and 7 respectively.

Table 6. Summary of articles forming the empirical evidence for KER5 which is derived using dose concordance (EE:DC), time scale (EE:TS) and essentiality studies (Ess).

Reference	Relationship Summary
Schabath et al. (2016)	(Ess): Mutations in TP53 are associated with enhanced cellular proliferation.
Duan et al. (2008)	(Ess): Mutations that completely abolish p53 function result in increased proliferation rates both before and after irradiation-induced damage. Mutant p53 that is still functional (specifically human p53(273H)) may still retain some wild-type function.
Lang et al. (2004)	(Ess): Mouse embryonic fibroblasts (MEF) with mutant p53 (specifically p53(515A)) have higher rates of cellular proliferation than cells completely lacking p53.
Ventura et al. (2007)	(Ess): Mutations in p53 that disrupt p53 functions allow tumor growth to proceed unhindered.
Li and Xiong (2017)	(Ess): Mutations in Cul9 or complete absence of p53, Arf, Cul9 or both Cul9 and p53 resulted in increased proliferation rates relative to wild-type cells.
Hundley et al. (1997)	(Ess): Lack of p53 in the transgenic ras mice resulted in increased cell proliferation, increased S and G2M cells, and decreased G1 cells in in vivo tumors.
Sato et al. (2017)	(Ess): A combination of oncogene activation (Kras or EGFR) and TSG deactivation (p53 and pRB) are essential to induce high rates of cell proliferation, which can be inhibited by MEK inhibitors (MEK has role in cell signaling downstream of Kras). EGFR inhibitors affected LE cells with activated EGFR, but not those with activated Kras (EGFR is upstream of Kras).
Jarvis et al. (1998)	(Ess): Loss of nuclear BRCA1 expression resulted in an increased rate of cellular proliferation.
Geng et al. (2017)	(Ess): Lack of SPOP in the prostate significantly increased cellular proliferation, as seen in mice and humans.
Fernandez-Antoran et al. (2019)	(Ess): p53 mutant progenitors outcompete their wild-type neighbors after low-dose ionizing radiation.

Full details of study can be found in AOP wiki.

Table 7. Summary of articles forming the empirical evidence for KER6 which is derived using dose concordance (EE:DC), time scale (EE:TS) and essentiality studies (Ess).

Reference	Relationship Summary
Li et al. (2007)	(Ess): Loss of JJAZ or addition of the JAZF1-JJAZ1 fusion (in presence of normal JJAZ1) results in decreased proliferation, while addition of the JAZF1-JJAZ1 fusion in the absence of normal JJAZ1 results in increased cellular proliferation. (EE:TS): In all cases where there is a difference in cell proliferation rates between genotypes, differences were evident by 3 days in culture.
Irwin et al. (2013)	(Ess): Cell proliferation linked to the Philadelphia (Ph) chromosome may be inhibited by a pan-ErbB family kinase inhibitor and a more specific ErbB1/ErbB2 kinase inhibitor; pErbB2 expression may also be significantly decreased by a pan-ErbB kinase inhibitor. (EE:TS): By 96 hours, cell proliferation rates in Ph + ALL cells exposed to ErbB kinase inhibitors were decreased relative to untreated Ph + ALL cells.
Guamerio et al. (2016)	(Ess): f-circRNAs, which results from chromosomal translocations that create fusion genes, are associated with significant increases in cellular proliferation rates. Inhibition of these f-CircRNAs significantly decreases the cellular proliferation rates. (EE:TS): Differences in cellular proliferation rates were evident by 2 days in culture in all cases.
Stopper et al. (2003)	(Ess): Addition of an estrogen receptor ligand/agonist (estradiol) resulted in increased frequencies of MN and increased cellular proliferation, while addition of an estrogen receptor antagonist did not cause a change in MN frequency or cell proliferation relative to controls. Accordingly, there are more estradiol-treated cells in the S-phase of the cell cycle and fewer in the G2/M phase relative to vehicle-treated controls. (EE:TS): Addition of estradiol increases the rate of cell proliferation such that differences are obvious by 100 hr and almost 4-fold higher at 216 h, and there are more MN present at the same cell number and at the same time point relative to untreated controls.
Soda et al. (2007)	(Ess): Gene fusions (caused by an inversion) within chromosome 2p were shown to drive cell proliferation, a positive linear growth is maintained in human cells carrying the gene fusion even in the absence of IL-3 which is required for growth, and in inhibition experiments, cell proliferation greatly declines with increasing inhibitor concentration. (EE:TS): Linear growth observed for all transfected cells in the presence of IL-3 over 7 days, while death in the absence of IL-3 occurs within 3 days. Presence of gene fusions maintained linear growth in the presence and absence of IL-3. Addition of an inhibitor to cells carrying gene fusion in the absence of IL-3 resulted in cell death by 5 days at the highest concentration.

Full details of study can be found in AOP wiki. The quantified values are extrapolated directly from the indicated references.

Knock-in and knock-out chromosomal translocation studies that manipulate cells to become genetically unstable and oncogenic have shown unrestricted proliferation. The two most extensively studied translocations are (1) the juxtaposition of the MYC gene to sequences from one of three immunoglobulin loci in Burkitt’s lymphoma, mouse plasmacytoma, and rat immunocytoma (reviewed in both Cory 1986 and Mes-Masson and Witte 1987), and (2) the fusion of the c-ABL oncogene with the BCR sequence in Philadelphia chromosome-positive chronic myelogenous leukemia (reviewed in Mes-Masson and Witte 1987). In general, the MYC family of cellular oncogenes encodes transcription regulators that are associated with cell proliferation and death regulation (Gostissa et al. 2009). Translocations involving MYC genes thus ultimately lead to uncontrolled proliferation. The association between increased CAs and unrestrained cell growth is a common theme among a variety of tumorgenic pathways involving the liver, breasts, and lungs, as evidenced by empirical data from transgenic mouse studies (reviewed in both Fowlis and Balmain 1993 and Cory 1986).

KER7: cell proliferation (KE5) leading to lung cancer (AO)

Uncontrollable growth and division of cells can lead to hyperplasia and the formation of neoplasms, which will continue to grow and ultimately form tumors. As thoroughly discussed in a review by Eymin and Gazzeri (2010), there are many different factors and signals that must be considered when examining tumorigenesis. A full description of this KER is provided in the AOPwiki, including components that occur in parallel to cell proliferation that are considered hallmarks of cancer (Boss et al. 2014). For example, balanced structural changes in genetic material (when the genetic material is exchanged evenly) are thought to initiate pathogenetic changes in cancer, while unbalanced structural changes (when genetic material is not exchanged evenly) may be causing secondary changes in cancer progression. Both of these processes involve uncontrollable cell proliferation (Mitelman 2005), which is the cornerstone of tumor development. Of note, is that many stimuli that are hyperplastic and toxic are not carcinogenic, (Hoel et al. 1988). Once cell proliferation accelerates and the abnormal growth pattern of cells continues, the formation of neoplasms ensues. As tumors develop and mature, they have the potential to become malignant and propagate throughout the body in a process known as metastasis. This relationship is still being studied. This uncontrollable and ultimately pathogenic cell proliferation may occur in any cells of the body; when these processes occur in cells specific to the lung, the result is lung cancer. There are multiple types of lung cancers that can be classified by histology. The two overarching types, non-small cell lung cancer (NSCLC) and small cell lung cancer (SCLC), are very invasive diseases with high metastatic potential (Hirsch et al. 1982; Kelly and Bunn 1998). The progression of the premalignant lesions to these different lung cancer types differs slightly, but the general overall process is similar (Wistuba et al. 1999; Gradowski et al. 2007). A summary of the studies supporting a link between cell proliferation and lung cancer is provided in Table 8.

Table 8. Summary of articles forming the empirical evidence for KER7 which is derived using dose concordance (EE:DC), time scale (EE:TS) and essentiality studies (Ess).

Reference	Relationship Summary
Sun et al. (2016)	(Ess): ZIC5 results in significant decreases in cell proliferation rates, decreased colony formation, and slower tumor growth. (EE:TS): In NSCLC cells with knock-down of ZIC5, the cell cycle was affected by 48 hours post-transfection, cell proliferation was significantly decreased by 96 hours post-transfection, and colony formation was significantly decreased by 2 weeks postinjection. In nude mice, differences in tumor growth rates were evident by 3 weeks postinjection, and tumor sizes and proliferation rates were significantly different at time of sacrifice (6 weeks postinjection).
Pal et al. (2013)	(Ess): Delphinidin treatment inhibits proliferation in two different NSCLC cell lines, as evidenced by declines in levels of pro-proliferative signaling molecules and declines in cell viability. Xenograft tumors also show dose-dependent decline in tumor volume and reduced markers of cell proliferation. (EE:TS): Differences in EGFR, pEGFR, VEGFR2 and pVEGFR2 levels were evident at 3 hours post-inhibitor treatment, while dose-dependent differences in PI3K/p110, PI3K/p85, pAKT, pERK1/2, pJNK1/2, pp38, PCNA and cyclin D1 after 48 hours. Differences in tumor volumes were evident from 13–16 days post-injection.
Tu et al. (2018)	KER5 - (Ess): MEFs with mutant p53 (specifically p53(515A)) have higher rates of cellular proliferation than cells completely lacking p53, this is known to induce tumors in animal models.
Lv et al. (2012)	KER5 - (Ess): Mutations in p53 that disrupt p53 functions inducing cell proliferation and tumor growth to proceed unhindered.
Kassie et al. (2008)	KER5 - (Ess): Mutations in Cul9 or complete absence of p53, Arf, Cul9 or both Cul9 and p53 result in increased proliferation rates relative to wild-type cells, this is known to induce tumors in animal models.
Zhong et al. (2005)	KER5 - (Ess): Lack of p53 in transgenic ras mice results in increased cell proliferation, increased S and G2M cells, and decreased G1 cells in in vivo tumors.
Warin et al. (2014)	(Ess): Decreased tumor volumes and weights in addition to increases in apoptosis and decreased cell proliferation are observed after treatment with suspected tumor inhibitors 6S and M2. (EE:TS): Tumors evident within 1 week of NSCLC injection, and continued to grow over 7 weeks.

Full details of study can be found in AOP wiki. The quantified values are extrapolated directly from the indicated references.

Overall assessment of the AOP

A full detailed summary of the overall assessment of the AOP, including inconsistencies is provided in the AOP-wiki (https://aopwiki.org/aops/272), and the reader is directed there for further details. Overall, decades of scientific radio-biological research provide data to support a strong biological plausibility for the KERs outlined in this AOP. Research has shown that radiation induced carcinogenesis can evolve from clastogenic lesions in the form of chromosomal aberrations and mutations leading to cell differentiation/proliferation and eventually neoplastic transformation. Through construction of this AOP, KERs that are less supported than others were identified (Figure 2). A general trend was identified: there was more support of upstream KERs (e.g. KER1, KER2 and KER3) over downstream KERs (e.g. KER5, KER6 and KER7). These downstream KERs are, by their nature, more difficult to assess given the need for well-designed long-term studies to characterize them, limiting the amount of available data. Indeed, in the case of KER7 (increased cell proliferation leading to lung cancer) this time period can extend beyond several decades. The distribution of support between biological plausibility and empirical evidence for each of the KERs is also highlighted and shows that in the case of KER3 and KER6 there is an apparent deficiency of empirical evidence necessary to complement the respective level of biological plausibility. As such, the AOP framework informs us that this particular AOP would benefit from further study of inadequate repair leading to mutations (KER3) and the impact of varying levels of chromosomal aberrations on cell proliferation (KER6).

Figure 2. Bar chart illustrating the number of articles contributing to the support metrics: biological plausibility and empirical evidence for each adjacent KER in the AOP presented here.

There was no single study that provided evidence to support all the KERs. For several of the KERs, there were limited data to directly demonstrate the impact of KEupstream to KEdownstream. This was particularly evident for inadequate repair leading to the generation of chromosomal aberrations and mutations leading to cell proliferation. Essentiality studies, using knock out/in and inhibitors were generally moderate/weak in level across the KERs. The majority of the upstream KERs had sufficient data to show dose-response concordance, however studies to strongly support time- and incidence-concordance were generally lacking and provided inconsistent results. In addition, throughout the KERs, discrepancies in data to support the KERs were evident, particularly for studies using low doses. There was also a lack of consistent studies that allowed for a quantitative understanding of this AOP, highlighting a clear knowledge gap worth exploring through directed research initiatives.

Discussion

Considerable mechanistic dose–response data have been generated in the radiation field, particularly in the area of clastogenic lesions. These data have been compiled and captured in an AOP to the most simplified, modular path to lung cancer from a molecular initiating event of deposited energy on DNA. This AOP is supported through KERs for which there is biological plausibility and available empirical evidence. Although it is clear that our proposed AOP is not the only route to the AO, it does represent a classic targeted response of radiation insult on a cell. The empirical evidence to support this pathway is strong and probabilistic. As per AOP conventions, the pathway does not describe every mechanism and alteration that is ultimately involved in radiation-associated carcinogenesis. Instead, KEs that are routinely measured using modern and conventional assays are described. For this reason, not all of the KEs that are hallmarks of cancer, that is, evasion, angiogenesis etc. are mapped out, but as they are critical events they can be developed separately. This AOP will be the first to use a MIE that is radiation-specific and therefore can act as a foundational AOP to build networks of radiation-specific responses. Networks can evolve to multiple AOs with additional KEs that incorporate nontargeted effects, immune and adaptive responses, in parallel.

While this AOP is applicable to other types of radiation-induced cancers, lung cancer was selected as the AO due to its relevance to radon risk assessment and its broader applicability to the chemical field. Lung cancer is a major public health problem world-wide, killing an estimated 1.5 million people annually (https://www.wcrf.org/dietandcancer/cancer-trends/worldwide-cancer-data). Although smoking is the leading cause of lung cancer, numerous environmental sources are also important contributors including radon, asbestos, air pollution and arsenic (Hubaux et al. 2012). Some of these stressors can act synergistically to increase risk, particularly among smokers. It has been shown that the histological lung profile of individuals that are smokers is quite different from nonsmokers exposed to high radon levels (Egawa et al. 2012). This is in part due to the complexity of each stressor, in terms of its interaction with cells at the molecular level. The AOP concept may help better understand the mode of action of different stressors at the cellular level that have a common adverse outcome.

Despite the decades of research in the area of radiation and DNA damage, a major challenge in developing this AOP was in finding the required components (i.e. essentiality, temporal, incidence and dose concordance) to provide strong empirical evidence to help support the KERs. Across all KERs, studies were lacking that used a broad dose-range. Most studies conducted analysis at one time-point and there were limited studies that supported the essentiality criteria. This was particularly evident for the KERs of inadequate repair to mutations/CA and mutations/CA to cellular proliferation. The non-adjacent KERs (i.e. DDoE to CA or DDoE to mutations), were generally more well supported. Furthermore, no single study encompassed all the KERs proposed in this AOP. In addition, there were considerable discordant results across KEs simply due to the MIE as its outcome is dependent on factors such as cell type, dose, dose-rate, and radiation quality. These factors can influence the amount and type of damage, which in turn can affect the probability to drive a path forward to cancer. The principle knowledge gap arose from the lack of data in the form of essentiality studies, using inhibitors and knock-in/out genes. As well for a number of KERs, there were minimal dose-response and temporal response data in well-conducted animal studies. There is also a range of uncertainty on how confounders such as lifestyle, health status, and radiosensitivities affect an individual’s path to an AO. Additional KEs may need to be added in parallel as our knowledge in these areas improves. These challenges can drive research priorities in the future.

An overall assessment of this AOP shows that there is strong biological plausibility and moderate empirical evidence to suggest a qualitative link between deposition of energy on DNA to the final AO of lung cancer. This evidence has been derived predominately from decades of research using laboratory studies and through mathematical simulations of cell-based models. These studies have shown both dose- and temporal-response relationships for select KEs. The quantitative thresholds to initiate each of the KEs have been shown to vary with factors such as the cell type, dose rate of exposure and radiation quality. Thus, an absolute amount of deposited energy (MIE) to drive a KE forward to a path of cancer is not yet definable. This is particularly relevant to low doses and low dose-rates of radiation exposure where the biology is interplayed with conflicting concepts of hormesis, hypersensitivity and the linear no threshold theory. Furthermore, due to the stochastic nature of the MIE, it remains difficult to identify specific threshold values of DSBs needed to overwhelm the DNA repair machinery to cause ‘inadequate’ DNA repair leading to downstream genetic abnormalities and eventually cancer. With a radiation stressor, a single hit to the DNA molecule could drive a path forward to lung cancer; however, this is with low probability. Conversely, at much higher doses, a cell will induce apoptosis and may not be driven to cancer induction. Although empirical modeling of cancer probability vs. mean radiation dose and time to lethality, does provide a good visualization of the effective thresholds (Raabe 2011), practically, there is still considerable uncertainty surrounding the connection of biologically contingent observations and stochastic energy deposition. Future work may focus on developing more precise quantitative and predictive models to help address these types of uncertainties.

This foundational AOP will initiate the building of networks and feedback loops that will provide further evidence linking the essential events to lung cancer, including genome alterations, oxidative stress, and metabolomics effectors. This will require efforts from the larger radiation community. As the empirical evidence to support these areas becomes stronger, a better representation of events to lung cancer will emerge. By identifying uncertainties and inconsistencies in the literature, research can be directed to address knowledge gaps, which can later help refine the pathway. It is our goal, with this AOP to motivate radiation researchers to use this framework for bringing together research data, exchanging knowledge and identifying research priority areas in the low-dose ionizing radiation field. Long-term, this AOP alongside others in the radiation field will help to identify key events common to chemical stressors and multiple adverse outcomes, which will be important to help refine risk assessment. In all, by building more radiation-relevant AOPs, the AOP framework will have a bigger role in supporting the practice of radiation protection.

Abbreviations
AOPs	=	Adverse outcome pathways
MIE	=	Molecular Initiating Event
KEs	=	Key Events
KERs	=	Key Event Relationships
AO	=	Adverse Outcome
OECD	=	Organisation for Economic Cooperation and Development
NHEJ	=	non-homologous end-joining
HR	=	homologous recombination
alt-EJ	=	alternative end joining
DSBs	=	double strand breaks
CNV	=	copy number variants
LET	=	linear energy transfer
TSG	=	tumor suppressor genes
LOF	=	loss of function

Acknowledgments

The authors would like to acknowledge Ruth Wilkins, Sami Qutob and Daniel Beaton for critical review and Baki Sadi, Fatemeh Nabavi, Sandra Noble, Nadine Adam, and Deepti Bijlani for reference checking and final editing of the AOP.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Notes on contributors

Vinita Chauhan

Vinita Chauhan is a research scientist at Health Canada.

Samantha Sherman

Samantha Sherman was a research assistant from the University of Ottawa.

Zakaria Said

Zakaria Said was a fourth year Honor student from the University of Ottawa.

Carole L. Yauk

Carole L. Yauk is a research scientist at Health Canada.

Robert Stainforth

Robert Stainforth was a post-doctoral fellow at Health Canada.