Accepted Manuscript

Fingerprinting the reactive toxicity pathways of 50 drinking water disinfection by-products

Daniel Stalter, Elissa O’Malley, Urs von Gunten, Beate I. Escher

PII: S0043-1354(15)30445-0

DOI: 10.1016/j.watres.2015.12.047

Reference: WR 11745

To appear in: Water Research

Received Date: 13 May 2015

Revised Date: 2 December 2015

Accepted Date: 28 December 2015

Please cite this article as: Stalter, D., O’Malley, E., von Gunten, U., Escher, B.I., Fingerprinting thereactive toxicity pathways of 50 drinking water disinfection by-products, Water Research (2016), doi:10.1016/j.watres.2015.12.047.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service toour customers we are providing this early version of the manuscript. The manuscript will undergocopyediting, typesetting, and review of the resulting proof before it is published in its final form. Pleasenote that during the production process errors may be discovered which could affect the content, and alllegal disclaimers that apply to the journal pertain.

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

1

Fingerprinting the reactive toxicity pathways of 50 drinking water 1

disinfection by-products 2

Daniel Stalter1,2*, Elissa O’Malley1, Urs von Gunten2,3, Beate I. Escher1,4,5 3

1National Research Centre for Environmental Toxicology, Entox, The University of 4

Queensland, Brisbane, Australia 5

2Eawag, Swiss Federal Institute of Aquatic Science and Technology, Duebendorf, Switzerland 6

3School of Architecture, Civil and Environmental Engineering (ENAC),Ecole Polytechnique 7

Fédérale de Lausanne (EPFL), CH-1015 Lausanne, Switzerland 8

4Department of Cell Toxicology, UFZ – Helmholtz Centre for Environmental Research, 9

Leipzig, Germany 10

5Department of Environmental Toxicology, Center for Applied Geosciences, Eberhard Karls 11

University, Tübingen, Germany 12

*corresponding author: daniel.stalter@eawag.ch; Überlandstrasse 133, 8600 Dübendorf, 13

Switzerland; tel. +41 58 765 6828, fax +41 58 765 50 28. 14

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

2

Abstract 15

A set of nine in vitro cellular bioassays indicative of different stages of the cellular toxicity 16

pathway was applied to 50 disinfection by-products (DBPs) to obtain a better understanding 17

of the commonalities and differences in the molecular mechanisms of reactive toxicity of 18

DBPs. An Eschericia coli test battery revealed reactivity towards proteins/peptides for 64% 19

of the compounds. 98% activated the NRf2-mediated oxidative stress response and 68% 20

induced an adaptive stress response to genotoxic effects as indicated by the activation of 21

the tumor suppressor protein p53. All DBPs reactive towards DNA in the E. coli assay and 22

activating p53 also induced oxidative stress, confirming earlier studies that the latter could 23

trigger DBP’s carcinogenicity. The energy of the lowest unoccupied molecular orbital ELUMO 24

as reactivity descriptor was linearly correlated with oxidative stress induction for 25

trihalomethanes (r2 = 0.98) and haloacetamides (r2 = 0.58), indicating that potency of these 26

DBPs is connected to electrophilicity. However, the descriptive power was poor for 27

haloacetic acids (HAAs) and haloacetonitriles (r2 < 0.06). For HAAs, we additionally 28

accounted for speciation by including the acidity constant with ELUMO in a two-parameter 29

multiple linear regression model. This increased r2 to > 0.80, indicating that HAAs’ potency is 30

connected to both, electrophilicity and speciation. Based on the activation of oxidative 31

stress response and the soft electrophilic character of most tested DBPs we hypothesize 32

that indirect genotoxicity—e.g., through oxidative stress induction and/or enzyme 33

inhibition—is more plausible than direct DNA damage for most investigated DBPs. 34

The results provide not only a mechanistic understanding of the cellular effects of DBPs but 35

the effect concentrations may also serve to evaluate mixture effects of DBPs in water 36

samples. 37

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

3

Keywords 38

Disinfection byproduct, oxidative stress, p53 activation, QSAR, electrophilicity, mutagenicity 39

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

4

1. Introduction 40

Drinking water disinfection effectively reduces waterborne diseases (Akin et al. 1982) but 41

the formation of disinfection by-products (DBPs) has raised concerns due to their potential 42

adverse health effects (Cantor et al. 2010, Komaki et al. 2014, Pals et al. 2013, Plewa et al. 43

2010, Plewa et al. 2008b). Some of the 600 to 700 known DBPs have been toxicologically 44

characterized with in vitro bioassays and a few with chronic in vivo studies demonstrating 45

that some DBPs are potent cytotoxicants, genotoxicants, and carcinogens (Plewa et al. 2010, 46

Plewa et al. 2008b, Richardson et al. 2007). However, the mechanisms through which DBPs 47

cause adverse health effects have not yet been entirely elucidated (Nieuwenhuijsen et al. 48

2009). 49

Among the 85 DBPs reviewed by Richardson et al. (2007), 68 were genotoxic 50

(Nieuwenhuijsen et al. 2009) indicating DNA damage as contributing factor for adverse 51

health effects. In addition, the toxicity pathway of some DBPs has been shown to involve 52

oxidative stress induction (Neale et al. 2012, Nieuwenhuijsen et al. 2009, Pals et al. 2013, 53

Plewa et al. 2010), which could contribute to adverse health effects since oxidative stress is 54

involved in an increasing list of major human diseases such as cancer, cardiovascular 55

diseases, or neurodegenerative diseases (Caro and Cederbaum 2004). 56

The various possible cellular toxicity pathways of reactive chemicals can be 57

rationalized within the framework of adverse outcome pathways (Fig. 1; Ankley et al. 2010). 58

In the cell, reactive chemicals can interact with DNA, proteins/peptides, and membrane 59

lipids via covalent binding (Fig. 1). Such molecular initiating events (MIEs) lead to a diverse 60

set of intermediate effects, for instance, gene damage, disruption of the redox balance, or 61

inhibition of enzymes. These intermediate effects are funneled into the activation of a small 62

number of crucial adaptive stress responses (Simmons et al. 2009), of which oxidative stress 63

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

5

response, p53-mediated response, and inflammation are the most relevant related to 64

toxicity caused by reactive (electrophilic) chemicals (Fig. 1). If the stress responses cannot 65

compensate intermediate effects, the cell goes into p53-mediated programmed cell death 66

(apoptosis) or, alternatively, succumbs to the lesions. The toxicity pathway of reactive 67

chemicals involves a network of processes (Fig. 1). Attack of proteins by soft electrophilic 68

DBPs, for example, can lead to enzyme inhibition and depletion of glutathione, which causes 69

a multitude of subsequent effects, such as generation of reactive oxygen species (ROS), 70

impaired DNA repair, or disturbed DNA replication. This can indirectly result in genotoxic 71

effects (Komaki et al. 2014, Komaki et al. 2009, Pals et al. 2011). 72

<<Figure 1>> 73

To obtain an effect fingerprint of reactive DBPs we applied a set of in vitro bioassays 74

(Fig. 1) with the major emphases on the molecular initiating event of covalent interaction 75

with proteins/peptides versus DNA, direct and indirect genotoxicity assays as well as 76

reporter gene assays indicative of adaptive stress responses to genotoxicity (activation of 77

p53), oxidative stress (activation of NRf2), and inflammation (activation of NF-κB). We used 78

chemical descriptors for reactivity to rationalize the observed effects. 79

Objectives of this study were (i) to fingerprint the adverse effects of a set of 47 80

commonly found DBPs and three drinking water contaminants (Table S1 of the SI: 81

halomethanes HMs; halonitromethanes HNMs, haloacetonitriles HANs, haloketones HKs, 82

haloacetic acids HAAs, chloral hydrate CH, haloacetamides HAcAms, nitrosamines, and the 83

halofuranone 3-chloro-4-(dichloromethyl)-5-hydroxy-5H-furan-2-one or MX) by use of cell-84

based bioassays, (ii) to relate the effect concentrations (ECs) to reactive properties of DBPs 85

in order to better understand dominant MIEs, and (iii) to provide a unified data base of 86

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

6

effect potencies for future research, e.g., for studies on toxicity of DBPs in mixtures (Escher 87

et al. 2013, Yeh et al. 2014). 88

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

7

2. Materials and Methods 89

2.1. Chemicals 90

We selected 47 commonly detected drinking water DBPs as well as 1,1-dichloroethene (1,1-91

dCE), dichloromethane (dCM), and bromochloromethane (BCM) as frequently detected 92

halogenated drinking water contaminants, to cover a wide range of different chemical 93

groups. Details about the tested chemicals, including abbreviations, supplier, and purity are 94

collated in Table S1 of the supporting information (SI). Methanolic stock solutions of all 95

DBPs were stored at −80°C. We used methanol as solvent because it showed the highest 96

effect concentration (i.e., lowest effect) in the AREc32 assay compared to DMSO, ethanol, 97

and MTBE (Escher et al. 2012). In addition, DMSO degrades readily and the degradation 98

products are reactive and can react with the dissolved organics (Balakin et al. 2006). This is 99

especially problematic for DBPs which are themselves reactive chemicals and may react with 100

DMSO. More information about the selected chemicals is stated in Section S1 of the SI. 101

2.2. Bioassays 102

Each of the eight bioassays applied in this study (Table 1) is described in more detail in the SI 103

(Section S2). The bacterial cytotoxicity assay using Aliivibrio fischeri (formerly termed Vibrio 104

fischeri) bioluminescence inhibition was selected because of its sensitivity to DBPs (Farré et 105

al. 2013, Neale et al. 2012, Yeh et al. 2014) and because a Quantitative Structure-Activity 106

Relationship (QSAR) was available to predict the ECs of those chemicals that act as baseline 107

toxicants in this assay (Tang et al. 2013). We used this assay to quantify the toxic ratio (TR) 108

of QSAR-predicted baseline EC to experimental EC. When the experimental EC falls close to 109

the toxicity predicted with the baseline toxicity QSAR model then it can be concluded that 110

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

8

chemicals act via non-polar or polar narcosis. When the toxicity is considerably higher (i.e., 111

TR>10), it can be concluded that effects must be exerted via reactive or specific mechanisms 112

of action (Tang et al. 2013, Verhaar et al. 1992). 113

A set of E. coli strains was applied to classify the compounds according to their reactivity 114

with proteins/peptides, DNA, or as unspecific reactive, i.e., reactive with both, proteins and 115

DNA (Harder et al. 2003a, Tang et al. 2012). Glutathione (GSH) is an antioxidant tri-peptide, 116

which acts as a reducing agent to inactivate oxidative compounds such as reactive oxygen 117

species or its thiol-group on the cysteine reacts with soft electrophiles via nucleophilic 118

substitution reactions. Accordingly, the GSH-deficient E. coli strain MJF335 (GSH−) is more 119

susceptible toward soft electrophilic attacks and reactive oxygen species than its parent 120

strain (MJF276, called here GSH+) and hence the latter shows greater ECs when the primary 121

MIE occurs via protein attack (Harder et al. 2003a, Harder et al. 2003b). Therefore, the ratio 122

of cytotoxicity EC50 of the GSH+ and GSH− strains (toxic raXo TRGSH) has been used to assess 123

the tendency of toxicants to react with proteins (Harder et al. 2003a). This assay was 124

performed according to Tang et al. (2012) in 96-well plates. The growth inhibition was 125

calculated by dividing the optical density (OD) at 600 nm in the sample by the OD600 in the 126

solvent control. 127

Complementarily, we assessed the cytotoxicity of a DNA repair-deficient strain (MV4108 or 128

DNA−) versus the parent strain (MV1161 or DNA+). Genotoxic compounds result in higher 129

cytotoxicity in the DNA− strain compared to DNA+. The resulting toxic ratio (TRDNA) has been 130

used together with TRGSH to assess if a toxicant reacts specifically with DNA, proteins or non-131

specifically with both (Harder et al. 2003a, Harder et al. 2003b). A TRDNA > 1.2 indicates a 132

preferred reactivity toward hard biological nucleophiles like DNA while a TR > 1.2 for both 133

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

9

assays (GSH± and DNA±) indicates an unspecific reactivity (Harder et al. 2003a). The 134

DNA+/DNA− assay (Harder et al. 2003a) was adapted to a microplate version based on the 135

procedure for the GSH+/GSH− assay (Tang et al. 2012). Nitrosamines were excluded from 136

these assays because they require metabolic activation to induce effects. 137

The bacterial umuC assay was selected as an initial screening tool for genotoxicity and to 138

assess effects of metabolic activation by use of mammalian liver enzymes (Reifferscheid et 139

al. 1991). This assay detects the activation of the cellular SOS-response, which is a global 140

response to DNA damage to induce DNA repair mechanisms, and hence indirectly detects 141

genotoxicity. Additionally, Reifferscheid and Heil (1996) demonstrated that the umuC assay 142

showed a high concordance with the Ames assay of about 90% when five Ames tester 143

strains were applied. 144

The Ames fluctuation assay with the Salmonella strain YG7108 was selected to assess the 145

mutagenic effects of nitrosamines (Magdeburg et al. 2014) because these compounds 146

appeared to be the least potent DBPs in the other assays and the strain YG7108 reacts 147

particularly sensitive to nitrosamines (Yamada et al. 1997). 148

The human MCF7 cell-based AREc32 assay targets the activation of the oxidative stress 149

response pathway NRf2-ARE (Escher et al. 2013). The Nrf2-ARE stress response pathway has 150

been demonstrated in previous studies to be an important toxicity pathway of mono-HAAs 151

(Attene-Ramos et al. 2010, Pals et al. 2013) and appears to play a central role for the toxicity 152

of many more DBPs (Neale et al. 2012, Nieuwenhuijsen et al. 2009). 153

The ARE-bla assay on oxidative stress response (Life-Technologies 2006) was employed in 154

addition to the AREc32 assay because it is based on a different cell line (HepG2 liver cells) 155

which might reveal cell specific differences in the response. Additionally, different reporter 156

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

10

gene constructs may lead to a different responsiveness of assays depending on 157

promoter/enhancer construction, ARE orientation, and other factors (Shukla et al. 2012). 158

The p53-bla assay (Yeh et al. 2014), derived from HCT-116 human colon carcinoma cells, was 159

applied because activation of p53 has been discussed as marker for genotoxic properties of 160

chemicals (Duerksen-Hughes et al. 1999). P53 can be stabilized and thus activated by a 161

variety of stress events, such as genotoxic chemicals, ionizing radiation, hypoxia, mitotic 162

spindle damage, hyperproliferation, or ribonucleotide depletion (Lavin and Gueven 2006, 163

Simmons et al. 2009). Most of these stress events are related to potential DNA damage or 164

impairment of DNA replication and p53 regulates the expression of a wide range of genes 165

involved in DNA repair, cell growth arrest, or apoptosis. Additionally, loss of p53 function 166

occurs during the development of most tumor types (Toledo and Wahl 2006). As a 167

consequence, p53 activation can be regarded as an indicator for genotoxic properties of 168

chemicals with a high degree of sensitivity and specificity (Duerksen-Hughes et al. 1999, van 169

der Linden et al. 2014). Furthermore, Attene-Ramos et al. (2010) found an altered 170

expression of genes of the p53 pathway through mono-HAAs, and hence p53 induction can 171

be regarded as a promising tool to characterize our set of DBPs. 172

The NF-κΒ-bla assay (Life-Technologies 2009) for inflammatory responses was employed 173

because of the cross-regulation of NF-κB and p53 as well as its role in the genotoxicity 174

response (Wu and Miyamoto 2007). Furthermore, a toxicogenomic analysis with human 175

cells by Pals et al. (2013) suggested a possible inflammatory response as a result of mono-176

HAA exposure. 177

For all volatile DBPs (HMs, HNMs, HANs, HKs, nitrosamines) the bioassays were conducted 178

free of headspace (Stalter et al. 2013) to reduce the loss of analyte during exposure. The 179

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

11

non-volatile or semi-volatile DBPs (HAAs, HAcAms, MX, CH) were assessed by use of the 180

conventional setup (Escher et al. 2013, Tang et al. 2013). Both, the conventional and 181

headspace-free setups produced comparable results for non-volatile DBPs (Stalter et al. 182

2013). For the GeneBLAzer cell lines (ARE-bla, p53-bla, NF-κB-bla) the headspace-free setup 183

could not be established because initial trials without headspace led to a very large 184

variability when using 384-well microplates with 120 μL total volume. Instead we used for all 185

DBPs low-volume 384-well plates (50 μL per well, Corning, USA) to reduce the headspace 186

and sealed the plates with aluminum microplate sealing tape. The loss of the volatile DBPs 187

was predicted with a mass balance-partitioning model (SI, Section S3; Liu et al. 2013). 188

According to these model predictions, the reduction of the headspace from 67% of the total 189

well volume to 20% reduced the partitioning to the headspace by 66–87%. After the 190

modification, only the most volatile compound (1,1-dCE) exceeded a mole fraction of 5% in 191

the headspace (Table S2). All compounds were dissolved in methanol before dosing. The 192

final concentration of methanol was 1% across all treatments. Effects of the test compounds 193

were normalized to medium containing 1% methanol, which did not exhibit effects different 194

from the medium control. Each compound was analyzed in two to five independent 195

experiments with two to four replicates each. The analytes were prepared in 8-point 2-fold 196

dilutions. Only for the AREc32 assay we applied a 16-point 1.4-fold dilution. The tested 197

concentration ranges are summarized in Table S3. 198

For the Microtox and the E. coli assays the assessment endpoint was the effect 199

concentration (EC), which induced 50% inhibition of bioluminescence (Microtox) or cell 200

density (E. coli assays). The EC50 was derived from a log-logistic concentration-effect curve 201

as described in Escher et al. (2008). For oxidative stress response, p53 pathway activation, 202

NF-κB signaling activation, and genotoxicity we used the induction ratio (IR). IR is defined as 203

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

12

the ratio of effect of the sample divided by the average effect observed in the controls. The 204

revertant ratio RR was analogously used for mutagenicity assessed with the Ames assay 205

(Escher et al. 2014). The effect concentration that elicits an IR or RR of 1.5 (ECIR1.5 or ECRR1.5; 206

i.e., 1.5-fold or 50% effect increase compared to the negative control), is the assessment 207

endpoint for these assays, calculated by use of linear regression (Escher et al. 2014). We 208

selected this endpoint because it has been demonstrated previously that the ECIR1.5 is a 209

statistically sound concept and significantly differs from the negative control (Escher et al. 210

2014). In case the cytotoxicity interfered with the induction before the IR 1.5 threshold was 211

reached, an extrapolated value is given but explicitly marked as “extrapolated” in Fig. 2 and 212

Table S4. Extrapolated values could be important for follow-up studies as even DBPs, which 213

do not exceed the effect threshold, could contribute to mixture effects (Escher et al. 2013). 214

For all GeneBLAzer assays (ARE-bla, p53-bla, and NF-κB-bla) the cytotoxicity was assessed 215

simultaneously by use of resazurin with a final concentration of 20 μM (Yeh et al. 2014). 216

Resazurin is a sensitive oxidation-reduction indicator and is widely used to assess cell 217

viability (O'Brien et al. 2000). We observed no interferences between resazurin/resorufin 218

fluorescence measurements and fluorescence measurements of the LiveBLAzer™ substrate 219

CCF4 (including its product) used for the GeneBLAzer assays. Concentrations above the EC10 220

for cytotoxicity were not included in the evaluation of the activation of the target endpoint. 221

2.3. Chemical descriptors 222

We used the software Spartan ’14 (V1.1.4; Wavefunction, Inc., Irvine, California, USA) to 223

calculate the energy of the lowest unoccupied molecular orbital ELUMO by use of Hartree-224

Fock 3-21G and MMFF geometry, the US-EPA online tool SPARC to calculate pKa values, EPI 225

suite for log Kow, and further literature data to extract chemical descriptors to support the 226

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

13

quantitative structure-activity relationship (QSAR) analyses in the discussion (details on 227

descriptors are compiled in Table S6). 228

2.4. Statistical analysis 229

We used Prism (Version 6.04; GraphPad Software, Inc., La Jolla, CA, USA) for statistical 230

analyses to calculate ECs as well as for correlation, linear regression, and two-parameter 231

multiple linear regression analyses. Data are stated as arithmetic mean ± standard deviation 232

(SD) in all plots and as mean and coefficient of variation (CV = SD/mean) in all Tables. 233

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

14

3. Results & Discussion 234

3.1. Ranking of effect potencies 235

The ECs of all DBPs in selected bioassays are summarized in Fig. 2, ranked within each 236

chemical group according to their potency for oxidative stress induction in the AREc32 assay 237

as most responsive test for most of the compounds. Detailed results of all bioassays are 238

listed in Tables S4, S5, and Fig. S1–S3. The ECs were comparable to previous studies within 239

one order of magnitude (Stalter et al. 2013, Yeh et al. 2014), which confirmed the 240

reproducibility of the assays. 241

The effect pattern followed the trends observed by Plewa et al. (2002, 2008a, 2010, 242

Plewa and Wagner 2009, 2004a, 2004b) with a general toxicity ranking of iodinated 243

DBPs > brominated DBPs > chlorinated DBPs. Also a tendency toward lesser toxicity with 244

increasing number of halogens per molecule (Plewa et al. 2002) could be confirmed in most 245

cases. 246

In the sections below, the ECs are compared within major chemical groups and 247

between the assays in relation to the cellular toxicity pathway. For this it must be kept in 248

mind that a comparison of assays is qualitative because the absolute sensitivity is also an 249

intrinsic property of the cell lines. It is influenced by cell-specific toxicokinetics, in particular 250

metabolic activity, as well as the make-up of the respective bio-molecular reporter tools 251

(Shukla et al. 2012). Thus, responsiveness of a bioassay is not necessarily to be equated with 252

sensitivity of the associated biological endpoint. Any correlation between the EC values of 253

two bioassays (Figures 3 and S6 to S8) will have a slope and an intercept and r2 for the 254

linearity of the regression. The intercept can be regarded as a measure of intrinsic 255

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

15

difference of the bioassays, while the slope is rather a measure of the relative sensitivity. 256

Important is to analyze if there is a linear correlation (i.e., if the r2 is high) or not. 257

<<Figure 2>> 258

Substantial data variability was noted in particular for volatile DBPs presumably due 259

to partial volatilization of analytes during dosing and before the plate sealing (Table 2, S4, 260

S5). The variability in the p53-bla assay was generally high, most likely as a result of the 261

small volumes of the wells in the 384-well plates, which requires pipetting of lower volumes 262

with potentially greater pipetting errors in case manual pipetting is performed. Also, 263

activation of p53 often occurs close to cytotoxic concentrations, which could have 264

compromised the induction data. 265

3.2. General nature of DBP-related molecular initiating events 266

3.2.1. Predominant reactive toxicity of DBPs 267

If the EC50 in the Microtox assay is more than 10-times greater than the QSAR-predicted EC50 268

(toxic ratio TR > 10), it can be concluded that the compound’s toxicity is exerted via other 269

mechanisms not taken into account by the baseline toxicity model, such as reactive or 270

specific mechanisms (Tang et al. 2013, Verhaar et al. 1992). 60% of the 50 compounds were 271

> 10 times more toxic than the baseline toxicity, which constitutes the minimum toxicity any 272

compound exhibits (Fig. S7). Together with the electrophilic nature of DBPs the high TR 273

values point to reactive modes of action rather than specific receptor-mediated effects. The 274

TR in the Microtox assay is only a screening indicator because the effect measured is the 275

bioluminescence inhibition after 30 min of incubation, which is related to energy depletion 276

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

16

and not to cell death. The test is not long enough for any defense mechanisms to be 277

triggered. 278

TR was greatest for MX, HNMs, dibromoacetonitrile (dBAN), bromochloroacetonitrile 279

(BCAN), iodoacetic acid (IAA), BAA (bromoacetic acid), and dibromoacetamide (dBAcAm) 280

with TR > 1000, which indicated a very high reactive toxicity. Reactive toxicity is caused by 281

reactive chemicals via unselective reaction with certain chemical structures in biomolecules 282

(Verhaar et al. 1992) and includes reactions with proteins, membrane lipids, and DNA 283

leading to a wide spectrum of reactive toxic mechanisms. 284

3.2.2. Reactivity with proteins/peptides versus DNA 285

The E. coli assays allowed an initial classification if electrophiles react preferentially with 286

proteins/peptides, DNA, or non-specifically with both (Harder et al. 2003a). Harder et al. 287

(2003b) also demonstrated that chemicals with a TRGSH >1.2 had high reactivity constants in 288

nucleophilic substitutions by the model biological soft nucleophile glutathione and 289

chemicals with a TRDNA >1.2 showed correlations between the effect concentrations and the 290

reaction rate constants with the model nucleotide 2′-deoxyguanosine. Therefore, activity in 291

the E. coli assay is a direct indicator whether the MIE is via protein or DNA interaction even 292

if not all of the reactivity may be translated into an adverse outcome in eukaryotic cells due 293

to defense mechanisms and adaptive stress responses. 294

For the GSH+/− strains a TRGSH > 1.2 for 71% of all tested compounds (all iodinated 295

trihalomethanes (THMs), HNMs, HKs, eight out of twelve HAAs, nine out of 10 HAcAms, and 296

MX; Table S5; Fig. S4, S5) indicated that cytotoxic effects were caused via protein 297

interaction, which suggests a soft electrophilic character of most DBPs. Chloroacetic acid 298

(CAA), trichloroacetonitrile (tCAN), BCAN, dBAN, and CH had a TRGSH < 1.2 despite their 299

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

17

potential to deplete GSH in mammalian cells and the cell-protective role of GSH when 300

mammalian cells are exposed to these DBPs (Chan et al. 2006, Chen et al. 2013, Lipscomb et 301

al. 2009). This indicated that more compounds interact with GSH and other soft 302

nucleophiles (proteins/peptides) than identified with this assay. Despite the potential for 303

false-negative results, this assay allows a fast and simple first-tier screening test for GSH 304

reactive compounds. 305

The TRDNA > 1.2 in the E. coli DNA +/− strains pointed to reactive toxicity toward DNA 306

for 31% of the 45 tested DBPs (Table S5; Fig. S4, S5). 12 out of the 14 DBPs with TRDNA > 1.2 307

also resulted in TRGSH > 1.2 indicating an unspecific reactivity toward DNA and proteins 308

(Harder et al. 2003a). 309

GSH conjugation can also increase toxicity of DBPs as shown previously for 310

dihalomethanes (Thier et al. 1993). This would result in both, GSH depletion in the GSH+ 311

strain as well as decreased toxicity in the GSH− strain. This should result in a TRGSH < 1 as the 312

GSH+ strain would be more sensitive to such compounds. This is the case for the two 313

dihalomethanes in our study, dCM and BCM. 314

These results are consistent with the predominantly soft-electrophilic nature of 315

DBPs, which implies electrophilic attacks of proteins/peptides as most likely MIE and a 316

potential to subsequently damage/inhibit proteins/enzymes. A soft electrophilic mechanism 317

of action has been shown for some DBPs in previous studies. HANs, for instance, affect the 318

glutathione (GSH) defense mechanism by depleting GSH through direct reaction and by 319

inhibiting glutathione-S-transferase (Lipscomb et al. 2009). GSH depletion has also been 320

previously observed for HAcAms (Chen and Stevens 1991), HKs (Merrick et al. 1987), and MX 321

(Yuan et al. 2006). 322

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

18

3.3. Bacterial genotoxicity - the effect of toxicokinetics/metabolism on DBP toxicity 323

3.3.1. Activation of SOS response in the umuC assay by DBPs 324

In the umuC assay 42 of the 50 compounds activated the bacterial SOS response (Fig. S8, 325

Table 2, Table S4). MX was the most potent compound in the umuC assay with an ECIR1.5 of 326

10−8 M. The second potent compound was tribromonitromethane (tBNM) with an ECIR1.5 of 327

10−6 M. MX and HNMs were more potent in the bacterial assay umuC compared to the 328

human cell-based p53 assay (Fig. 2, umuC vs. p53-bla). MX has been shown to be more 329

potent in bacterial assays than in mammalian cells (Plewa et al. 2002) but HNMs were more 330

potent in Chinese hamster ovary cells than in the Ames test (Kundu et al. 2004, Plewa et al. 331

2004a). 332

The SOS response was not induced by IAA, BAA, and CAA, despite their induction of 333

mutagenic effects in the Ames assay (Kargalioglu et al. 2002) and mammalian-cell based test 334

systems (Pals et al. 2011, Plewa et al. 2004b). This difference is most likely a result of 335

cytotoxic effects masking the reporter gene induction in the umuC assay since induction 336

ratios were excluded for data evaluation when cytotoxicity exceeded 50% (Reifferscheid et 337

al. 1991). Because of such limitations of the umuC assay and because mammalian cell lines 338

are generally more suitable to assess potential human health effects, we discuss 339

genotoxicity in more detail in the section about activation of p53. Here we focus on the 340

effect of metabolic enzymes on the genotoxicity of DBPs by use of bacterial test systems 341

(umuC and Ames) as established models with the rat liver extract S9. 342

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

19

3.3.2. Metabolism as confounding factor for DBP toxicity 343

In vitro assays are excellent diagnostic tools to study the cellular toxicity pathways of 344

chemicals but the measured effect concentrations cannot be directly translated to effects in 345

vivo. In particular toxicokinetics (i.e., uptake, biotransformation, distribution, and 346

elimination of a toxicant) are different in complex organisms compared to isolated cell lines 347

as well as among different cell-types and can thus be an important confounding factor for 348

the interpretation of results. The effect of metabolic enzymes is important to assess as 349

biotransformation can increase (activate) or decrease the toxicity of compounds. Therefore, 350

mammalian cell-based test systems could lead to false negative results due to 351

biotransformation processes. S9 contains mammal-specific cytochrome P450 isoforms as 352

well as glutathione S-transferases for drug metabolism. In the umuC assay, S9 addition did 353

not increase the potency (Table 2, S4 and Fig. S8). In contrast, the potency decreased for 354

some DBPs after addition of S9. This effect was especially pronounced for HAAs and also for 355

some HMs indicating biotransformation through metabolic enzymes. Thus, the umuC results 356

with and without S9 addition confirmed that no metabolic activation is required for most of 357

the analyzed DBPs and metabolizing enzymes partly reduced toxicity through 358

biotransformation. 359

3.3.3. Potency of N-Nitrosamines 360

N-Nitrosamines had low activity in all assays with N-nitrosodimethylamine (NDMA) as one of 361

the least potent compounds (Fig. 2, Table S4). Bacterial genotoxicity assessed with umuC did 362

not increase with S9, except for N-nitrosodiethylamine (NDEA). Nitrosamines are known to 363

be potent carcinogens with NDMA as its most potent representative but metabolic 364

activation is required for these compounds to exert genotoxic effects. The negative results 365

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

20

in the umuC assay were most likely a result of effective mechanisms to repair DNA lesions 366

caused by alkylating agents (Yamada et al. 1997). In the Ames assay, only tester strains 367

deficient of methyltransferases which repair such lesions, are highly responsive to N-368

nitrosamines (Yamada et al. 1997). Therefore, we included the Ames test with the strain 369

YG7108 in our study, which reacts particularly sensitive to N-nitrosamines after metabolic 370

activation with S9 (Wagner et al. 2012). The resulting rank order of potency was NDMA > 371

NDEA > N-nitrosodi-n-butylamine (N-But) > N-nitrosopiperidine (N-Pip) > N-372

nitrosomorpholine (N-Morph), which is consistent with a previous study (Wagner et al. 373

2012). 374

The test on activation of p53 as mammalian genotoxicity marker showed no effect or 375

very low potencies with extrapolated ECs for NDMA, NDEA, and N-But (Fig. 2). The HCT-116 376

cell line used for the p53-bla assay is supposed to be metabolically active (Hulikova et al. 377

2013) but it might lack sufficient phase 1 monooxygenase activity to bioactivate N-378

nitrosamines. However, another explanation could be the expression of DNA 379

methyltransferases, which is an effective cellular defense mechanism against alkylating 380

agents and its expression varies greatly depending on the cell type (Christmann et al. 2011). 381

These results emphasize limitations of the test systems used in this study for compounds 382

that induce effects only after metabolic activation because the required conditions might 383

not be appropriately simulated. For such compounds an underestimation of their potency in 384

vivo has to be taken into account. 385

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

21

3.4. Cellular stress responses on intermediate effects 386

3.4.1. NF-κB as inflammatory stress response marker 387

None of the DBPs in our study activated the NF-κB signaling pathway. In a previous study, 388

NF-κB induction after t-butylhydroquinone (tBHQ) exposure of HepG2 cells was observed by 389

use of an NF-κB binding site oligonucleotide probe (Pinkus et al. 1996). This is remarkable as 390

the NF-κB induction was attributed to the tBHQ-mediated formation of ROS such as �OH 391

radicals (Pinkus et al. 1996). As mono-HAAs lead potentially to ROS generation as well and 392

subsequently to activation of proteins characteristic for an inflammatory response (Pals et 393

al. 2013) we expected the activation of NF-κB as key transcriptional driver of inflammatory 394

responses (Simmons et al. 2009). The absence of effect in the NF-κB-bla assay for all of the 395

tested compounds indicated that these compounds do not trigger NF-κB/inflammation 396

under the given test conditions. 397

3.4.2. Activation of the oxidative stress response pathway 398

The human cell test on oxidative stress induction AREc32 revealed that 49 out of the 399

selected 50 compounds (including four extrapolated values) activated the oxidative stress 400

response (Fig. 2). The HAN dBAN was the most potent DBP for oxidative stress induction in 401

our study followed by diiodoacetamide (dIAcAm) and dibromochloroacetamide (dBCAcAm; 402

Fig. 2). HANs were generally among the most potent DBPs in this assay (Fig. 2) with ECIR1.5 403

from 10−7 M to 10−5 M. HAcAms were another potent group of DBPs for oxidative stress 404

induction (AREc32) with ECIR1.5 down to 10−7 M (dIAcAm). Also HKs, iodinated and 405

brominated HAAs (IAA, BAA, chloroiodoacetic acid (CIAA), bromoiodoacetic acid (BIAA)), 406

HNMs, and iodinated THMs reached ECs in the 10−6 to 10−5 M range. The ECIR1.5 in the 407

AREc32 and the EC50 in the Microtox assay revealed a similar pattern (Fig. 3A), while only 408

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

22

results for the group of HMs were well correlated (Spearman’s ρ = 0.93), indicating a 409

relationship between oxidative stress induction and cytotoxicity. Such a relationship is 410

plausible because greater reactivity of compounds leads to an increased potential for 411

oxidative stress (e.g., via GSH depletion), which eventually leads to cell death. However, 412

other DBP groups showed a high variability without a clear correlation. 413

<<Figure 3>> 414

Some DBPs (HAAs and some HAcAms) were equally potent in the AREc32 and ARE-415

bla assay (Fig. 3B). However, while only trichloroacetic acid (tCAA) did not induce oxidative 416

stress in both assays, 13 DBPs (five HMs, HNMs, HKs, and four nitrosamines) were active in 417

AREc32 but not active in the ARE-bla assay (Table 2, S4). Additionally, for most other DBPs 418

the ECIR1.5 in the AREc32 assay was up to one order of magnitude lower compared to the 419

ARE-bla assay indicating a lesser responsiveness of the latter. The difference was especially 420

pronounced for HMs and HANs, which were consistently > 10 times less potent in the ARE-421

bla. These differences might be explained by the metabolic capabilities of the HepG2 cells, 422

which are the basis of ARE-bla, seemingly indicating a complete or partial metabolic 423

inactivation of the respective DBPs. However, it has been demonstrated previously that 424

activity and expression of phase 1 metabolic enzymes is very low in HepG2 cells (Wilkening 425

et al. 2003). Thus, metabolic transformation of DBPs to less toxic compounds is very 426

unlikely. According to Shukla et al. (2012), differences in the reporter gene construct 427

between ARE-bla and another oxidative stress assay (ARE-luc) may explain discrepancies in 428

the results and predispose the ARE-bla to false positives and ARE-luc to false negatives. 429

According to this hypothesis, the AREc32 assay might be more prone to false positive results 430

than the ARE-bla assay. 431

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

23

For both assays we used ARE-driven transcription of a reporter gene as tool for 432

measuring induction of oxidative stress. It should be noted that the Nrf2-ARE signaling 433

pathway responds to both electrophilic attack as well as ROS-induced oxidative stress 434

(Kensler et al. 2007). Thus, we cannot directly differentiate between these two mechanisms. 435

However, also initial depletion of GSH through the soft-electrophilic attack of DBPs is likely 436

to increase ROS concentrations because GSH acts as important ROS-scavenger and ROS are 437

continuously formed as by-products of the normal cell-metabolism. 438

DBP-induced ROS induction or oxidative damages have been reported in several 439

studies. HAN exposure increased ROS and caused lipid peroxidation (Lipscomb et al. 2009). 440

Exposure of cells to mono-HAAs led to inhibition of the glycolysis enzyme glyceraldehyde 3-441

phosphate dehydrogenase GAPDH (Pals et al. 2011). GAPDH inhibition resulted in reduced 442

ATP and pyruvate concentrations, which in turn led to ROS formation, oxidative stress, and 443

subsequent DNA damage (Dad et al. 2013). Exposure to iodoacetamide (IAcAm) led to lipid 444

peroxidation and addition of antioxidants prevented lipid peroxidation as well as IAcAm-445

induced cell death (Chen and Stevens 1991). These examples demonstrate the potential of 446

DBPs to directly and indirectly cause oxidative stress. In this study we additionally 447

demonstrated the oxidative stress inducing potency of many more HAcAms, in particular di-448

HAcAms, as well as di-HAAs, iodinated THMs, HNMs, and HKs (Table 2, S4; Fig. S1, S2). Also 449

other DBP classes, such as halobenzoquinones, have been shown to induce oxidative stress 450

(Li et al. 2015). 451

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

24

3.4.3. Activation of the tumor suppressor protein p53 as mammalian genotoxicity 452

marker 453

The tumor suppressor protein p53 is activated in healthy cells as an adaptive stress 454

response to DNA lesions and can arrest the cell cycle to prevent the replication of cells with 455

high levels of genomic damage which otherwise can lead to neoplasms. Therefore, 456

activation of p53 can be used to predict genotoxic properties of chemicals (Duerksen-457

Hughes et al. 1999). 458

34 of the 50 selected compounds (including six extrapolated values) induced p53 (Fig. 2). 459

Activation of p53 is used as marker for mammalian genotoxicity. Some of the p53 inducing 460

DBPs have previously been identified as carcinogens. HANs, for instance, induced tumors 461

after topical application with the tumor-initiating activity ranking of dBAN > BCAN > tCAN = 462

dichloroacetonitrile (dCAN) (Bull et al. 1985). More recently, dBAN was shown to be 463

carcinogenic in mice and rats (NTP 2010). Bromochloroacetamide (BCAcAm) and 464

dichloracetamide (dCAcAm) are predicted mutagens and carcinogens (Bull et al. 2011). 465

Studies on IAcAm found a co-carcinogenic activity in a mouse skin assay (Gwynn and 466

Salaman 1953) and tumor promoting behavior together with nitrosamides (Fukushima et al. 467

1977, Takahashi et al. 1976). MX is one of the most potent carcinogenic DBPs in rats and CH 468

is carcinogenic in rodents (Richardson et al. 2007). In vivo studies also documented 469

carcinogenic effects for dibromoacetic acid (dBAA), dichloroacetic acid (dCAA), and tCAA; 470

however, not for CAA (Richardson et al. 2007). This reveals an inconsistency between in 471

vitro and in vivo studies since CAA is an in vitro genotoxicant but does not induce cancer 472

whereas the weakest in vitro genotoxicants (or non-genotoxicants) dCAA and tCAA are 473

carcinogenic in vivo. This indicates that non-genotoxic carcinogenic mechanisms could be 474

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

25

the primary mode of action for the carcinogenicity of these two HAAs (Richardson et al. 475

2007). However, because activation of p53 is regarded as a strong indicator for tumor 476

inducing properties (Duerksen-Hughes et al. 1999), in particular the carcinogenicity of 477

HAcAms, HKs, di-HAAs, HANs, and ITHMs should be further investigated. 478

All compounds that activated p53 also induced oxidative stress, which occurred at 479

lower concentrations in most cases (Fig. 3C). Some HAcAms, HNMs, and CH were > 10 times 480

more potent in AREc32. A similar trend between p53-bla and oxidative stress was observed 481

for the ARE-bla assay with only MX, tCAN, and triiodomethane (tIM) being > 10 times more 482

responsive in the p53-bla assay and dBCAcAm, dIAcAm, and CH being more potent in the 483

ARE-bla assay (Fig. S9). The comparison between activation of p53 and induction of 484

oxidative stress (Fig. 3C, S9) showed a high variability without a clear correlation. However, 485

a similar trend indicates a relationship between oxidative stress and p53 induction for some 486

of the DBPs. A potential causality is plausible as ROS generation can lead to increased DNA 487

damage and increasing cancer risk (Caro and Cederbaum 2004). A causal relationship 488

between oxidative stress induction (via GAPDH inhibition) and genotoxicity was 489

demonstrated previously for mono-haloacetic acids (Dad et al. 2013, Pals et al. 2013) while 490

p53 induction was observed for BAA and IAA (Attene-Ramos et al. 2010). Other mechanisms 491

for indirect genotoxic effects involve the cell cycle alteration induced by HANs (Komaki et al. 492

2014). Generally, the soft-electrophilic character of most DBPs makes such indirect 493

genotoxic effects more plausible than direct DNA damage, which requires hard-electrophilic 494

attacks. 495

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

26

3.5. Reactivity descriptors versus effects to assess MIEs 496

3.5.1. Relationship between reactivity descriptors and effects 497

The effect fingerprinting showed that reactive toxicity is the main toxicity pathway for DBPs. 498

DNA-reactive and protein/peptide reactive chemicals can be further characterized by their 499

electrophilic properties with ELUMO as a measure of electrophilicity (Schultz et al. 2006). ELumo 500

is a common electrophilicity descriptor frequently used for QSAR analyses of reactive 501

compounds (Burrow and Modelli 2013, Geiss and Frazier 2001, Huang et al. 2007, Schultz et 502

al. 2006). Calculated ELUMO values differ significantly depending on the applied model while 503

the relative ELUMO energies within a series of congeners are commonly quite comparable 504

(Burrow and Modelli 2013). Accordingly, only ELUMO values from the same chemical group 505

should be considered in QSAR studies (Burrow and Modelli 2013). It was not the goal of this 506

study to develop predictive QSAR models, but to apply QSARs as diagnostic tools to assess if 507

the MIE of DBPs involves more likely a direct covalent interaction with DNA or proteins. 508

Linear regression analyses of DBP reactivity versus ECs were used to evaluate the 509

relationship between reactivity and observed effects. A significant correlation between 510

reactivity and ECs for genotoxicity endpoints suggests a direct interaction between DBP and 511

DNA. In contrast, a significant correlation between reactivity and ECs for oxidative stress 512

induction without a correlation with genotoxicity indicates a direct protein interaction as a 513

more likely MIE. 514

Most DBPs analyzed in this study have soft-electrophilic characteristics such as 515

relatively low ELUMO values (Table S6) and good leaving groups (e.g., halogens) (Schultz et al. 516

2006). Such compounds have a high affinity to electrons during primary reactions, 517

preferably with soft nucleophiles (proteins/peptides) (Schultz et al. 2006). The main 518

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

27

exception is the group of nitrosamines, which lacks a good leaving group and hence is not 519

further discussed in this section. 520

3.5.2. Relationship between trihalomethanes’ toxicity and electrophilicity 521

Linear regressions of reactivity and ECIR1.5 for oxidative stress induction demonstrated that 522

ELUMO is a suitable descriptor for THMs’ biological activity with r2 = 0.98 (AREc32, Table 3, 523

Fig. 4A, green circles) and r2 = 0.93 (ARE-bla, Fig. 4B, green circles). 524

<<Figure 4>> 525

Oxidative stress induction by THMs has been demonstrated previously by use of an 526

assay with rat hepatocytes together with a good correlation between ELUMO and oxidative 527

stress induction with comparable r2-values (Geiss and Frazier 2001). ECs from the 528

genotoxicity assays also correlated significantly with ELUMO (r2 = 0.96 for p53-bla, Fig. 4C, 529

green circles, and 0.85 for umuC (data not shown)) indicating initial interactions with both 530

DNA and proteins as possible MIE. These results demonstrated that the biological effects of 531

THMs in our bioassays are directly related to the compounds’ electrophilicity. 532

3.5.3. Relationship between haloacetic acids’ potency and electrophilicity as well as 533

acidity 534

For HAAs, ELUMO did not correlate with the bioassay results (Table 3, Fig. 4; r2 ≤ 0.06). The 535

correlation in the single linear regression analysis did not improve through use of ELUMO 536

calculated for the deprotonated form of the HAA molecules, which dominate at 537

physiological pH (r2 ≤ 0.04; Table S6). However, after inclusion of pKa values together with 538

ELUMO as independent variables and our effect data (log 1/ECIR1.5) in a two-parameter 539

multiple linear regression model, the r2 reached 0.66 for AREc32, 0.61 for ARE-bla, and 0.67 540

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

28

for p53-bla (Table S7). The use of ELUMO for the deprotonated form increased the r2 in the 541

multiple linear regression model to ≥ 0.80 for these three bioassays (Table S7). This confirms 542

results of a previous study using mouse embryo effect concentrations in a two-parameter 543

multiple linear regression for a group of 11 HAAs (r2 = 0.75; Richard and Hunter 1996). Such 544

a good agreement with the previous study can be expected because of good correlations 545

between the embryo toxicity data and the effect concentrations of AREc32, ARE-bla, and 546

p53-bla with r2-values from 0.89 to 0.97 (Fig. S11). This emphasizes the potential relevance 547

of the applied mammalian cell-based in vitro bioassays to in vivo systems. These analyses 548

confirmed that the potency of HAAs is governed by at least two factors (Richard and Hunter 549

1996). The first factor, the pKa, largely defines the pH dependent speciation and 550

consequently influences bioavailability to the cells. The second factor, electrophilic 551

properties, likely defines the intrinsic toxicity of HAAs through covalent interactions. The 552

ability of the multiple linear regression to describe relative potency variations of 11 HAAs 553

also suggests that this group shares a similar MIE (Richard and Hunter 1996). This indicates 554

that not only mono-HAAs but also di- and tri-substituted HAAs induce genotoxic effects 555

indirectly via soft electrophilic attack of proteins. While a proposed MIE of mono-HAAs 556

exists (Dad et al. 2013, Dad et al. 2011, Pals et al. 2013, Pals et al. 2011), more research is 557

required to identify the exact MIEs, of di-, and tri-HAAs. 558

3.5.4. Relationship between haloacetamides’ toxicity and electrophilicity 559

The only other group of DBPs in our study that resulted in significant correlations of ECs for 560

oxidative stress induction with ELUMO were HAcAms. The maximum r2 found for the 561

regression of HAcAms’ ECs with ELUMO was 0.58 for AREc32 and 0.47 for ARE-bla (p < 0.05; 562

Table 3; Fig. 4, purple squares). The inclusion of log Kow in a two-parameter multiple linear 563

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

29

regression did not improve the r2-values (r2 = 0.43). The p53 induction did not correlate with 564

the reactivity descriptors (r2 = 0.03) indicating that direct DNA interaction is an unlikely MIE. 565

This is consistent with the classification of HAcAms with the E. coli assays as GSH reactive. 566

The weaker correlation of reactivity descriptors and oxidative stress compared to THMs 567

pointed to an indirect mode of action, e.g., via the formation of more biologically active 568

biotransformation products. 569

3.5.5. Relationship between haloacetonitriles’ toxicity and electrophilicity 570

For HANs, ELUMO was a poor descriptor for oxidative stress induction resulting in non-571

significant linear regressions (Table 2; Fig. 4, green hexagons). The significant linear 572

regression with ECs for activation of p53 (r2 = 0.96; Table 3) was driven by the EC for dCAN 573

only and was therefore regarded as irrelevant. The mechanism of action of HANs involves 574

the biotransformation to highly reactive intermediates (Lipscomb et al. 2009). Therefore, 575

chemical descriptors of the parent compounds are probably rather limited to predict 576

biological effects. However, these results are based on four DBPs only and the inclusion of 577

more compounds is desirable to increase the robustness of potential descriptors for 578

biological effects. 579

4. Conclusion 580

Our results suggest that most of the analyzed DBPs cause genotoxic effects indirectly. The 581

soft-electrophilic nature of DBPs favors soft-electrophilic attacks of proteins and/or peptides 582

over hard-electrophilic attacks of DNA. Resulting ROS generation, ROS accumulation 583

through preceding GSH depletion or enzyme inhibition can lead indirectly to genotoxic 584

effects. This is supported by (i) the large percentage of analyzed DBPs which activated the 585

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

30

NRf2-mediated oxidative stress response (98%) and (ii) the significant correlation between 586

reactivity descriptor (ELUMO) and oxidative stress for THMs and HAcAms together with lower 587

or insignificant r2-values for linear regressions with genotoxicity endpoints (activation of 588

p53). 589

Most analyzed DBPs (68%) activated p53, an indicator for tumor inducing properties, 590

demonstrating that in particular the carcinogenicity of HAcAms, HKs, di-HAAs, HANs, and 591

ITHMs should be further investigated. 592

593

Acknowledgements 594

Janet Tang and Matti Lang are acknowledged for helpful suggestions. Eva Glenn, Ruby Yeh, 595

and Jeffrey Molendijk (Entox) are thankfully acknowledged for experimental assistance. We 596

also thank Jennifer Bräunig (Entox) and Erik Prochazka (Smartwater) for reviewing the 597

manuscript. This research was supported by a Marie Curie International Outgoing 598

Fellowship within the 7th European Community Framework Program (PIOF-GA-2012-329169) 599

and by the Australian Research Council (FT100100694 and DP140102672). 600

601

Appendix A. Supporting Information 602

Additional information on the tested DBPs, the applied bioassays, and the bioassay results is 603

available via http://www.journals.elsevier.com/water-research/. 604

605

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

31

References 606

Akin, E.W., Hoff, J.C. and Lippy, E.C. (1982) Waterborne outbreak control - which disinfectant. 607 Environmental Health Perspectives 46(DEC), 7-12. 608

609 Ankley, G.T., Bennett, R.S., Erickson, R.J., Hoff, D.J., Hornung, M.W., Johnson, R.D., Mount, D.R., 610 Nichols, J.W., Russom, C.L., Schmieder, P.K., Serrrano, J.A., Tietge, J.E. and Villeneuve, D.L. (2010) 611 Adverse outcome pathways: a conceptual framework to support ecotoxicology research and risk 612 assessment. Environmental Toxicology and Chemistry 29(3), 730-741. 613

614 Attene-Ramos, M.S., Wagner, E.D. and Plewa, M.J. (2010) Comparative human cell toxicogenomic 615 analysis of monohaloacetic acid drinking water disinfection byproducts. Environmental Science & 616 Technology 44(19), 7206-7212. 617

618 Balakin, K.V., Savchuk, N.P. and Tetko, I.V. (2006) In silico approaches to prediction of aqueous and 619 DMSO solubility of drug-like compounds: Trends, problems and solutions. Current Medicinal 620 Chemistry 13(2), 223-241. 621

622 Bull, R.J., Meier, J.R., Robinson, M., Ringhand, H.P., Laurie, R.D. and Stober, J.A. (1985) Evaluation of 623 mutagenic and carcinogenic properties of brominated anf chlorinated acetonitriles - by-products of 624 chlorination. Fundamental and Applied Toxicology 5(6), 1065-1074. 625

626 Bull, R.J., Reckhow, D.A., Li, X.F., Humpage, A.R., Joll, C. and Hrudey, S.E. (2011) Potential 627 carcinogenic hazards of non-regulated disinfection by-products: Haloquinones, halo-cyclopentene 628 and cyclohexene derivatives, N-halamines, halonitriles, and heterocyclic amines. Toxicology 286(1-629 3), 1-19. 630

631 Burrow, P.D. and Modelli, A. (2013) On the treatment of LUMO energies for their use as descriptors. 632 SAR and QSAR in Environmental Research 24(8), 647-659. 633

634 Cantor, K.P., Villanueva, C.M., Silverman, D.T., Figueroa, J.D., Real, F.X., Garcia-Closas, M., Malats, N., 635 Chanock, S., Yeager, M., Tardon, A., Garcia-Closas, R., Serra, C., Carrato, A., Castano-Vinyals, G., 636 Samanic, C., Rothman, N. and Kogevinas, M. (2010) Polymorphisms in GSTT1, GSTZ1, and CYP2E1, 637 disinfection by-products, and risk of bladder cancer in spain. Environmental Health Perspectives 638 118(11), 1545-1550. 639

640 Caro, A.A. and Cederbaum, A.I. (2004) Oxidative stress, toxicology, and pharmacology of CYP2E1. 641 Annual Review of Pharmacology and Toxicology 44, 27-42. 642

643 Chan, K., Leung, C., Poon, R. and O'Brien, P.J. (2006) Role of carbonyl metabolizing enzymes in the 644 hepatocyte cytotoxicity of chloral hydrate and tribromoethanol anesthetics. Book series: 645 Enzymology and Molecular Biology of Carbonyl Metabolism; 12th Meeting on Enzymology and 646 Molecular Biology of Carbonyl Metabolism 12, 171-180. 647

648

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

32

Chen, C.H., Chen, S.J., Su, C.C., Yen, C.C., Tseng, T.J., Jinn, T.R., Tang, F.C., Chen, K.L., Su, Y.C., Lee, 649 K.I., Hung, D.Z. and Huang, C.F. (2013) Chloroacetic acid induced neuronal cells death through 650 oxidative stress-mediated p38-MAPK activation pathway regulated mitochondria-dependent 651 apoptotic signals. Toxicology 303(1), 72-82. 652

653 Chen, Q. and Stevens, J.L. (1991) Inhibition of iodoacetamide and tert-butylhydroperoxide toxicity in 654 LLC-PK1 cells by antioxidants - a role of lipid peroxidation in alkylation induced cytotoxicity. Archives 655 of Biochemistry and Biophysics 284(2), 422-430. 656

657 Christmann, M., Verbeek, B., Roos, W.P. and Kaina, B. (2011) O-6-Methylguanine-DNA 658 methyltransferase (MGMT) in normal tissues and tumors: Enzyme activity, promoter methylation 659 and immunohistochemistry. Biochimica Et Biophysica Acta-Reviews on Cancer 1816(2), 179-190. 660

661 Dad, A., Jeong, C.H., Pals, J.A., Wagner, E.D. and Plewa, M.J. (2013) Pyruvate remediation of cell 662 stress and genotoxicity induced by haloacetic acid drinking water disinfection by-products. 663 Environmental and Molecular Mutagenesis 54(8), 629-637. 664

665 Dad, A., Pals, J., Wagner, E. and Plewa, M.J. (2011) Genotoxic mechanism of haloacetic acid drinking 666 water disinfection byproducts. Environmental and Molecular Mutagenesis 52, S69-S69. 667

668 Duerksen-Hughes, P.J., Yang, J. and Ozcan, O. (1999) p53 induction as a genotoxic test for twenty-669 five chemicals undergoing in vivo carcinogenicity testing. Environmental Health Perspectives 670 107(10), 805-812. 671

672 Escher, B.I., Allinson, M., Altenburger, R., Bain, P.A., Balaguer, P., Busch, W., Crago, J., Denslow, N.D., 673 Dopp, E., Hilscherova, K., Humpage, A.R., Kumar, A., Grimaldi, M., Jayasinghe, B.S., Jarosova, B., Jia, 674 A., Makarov, S., Maruya, K.A., Medvedev, A., Mehinto, A.C., Mendez, J.E., Poulsen, A., Prochazka, E., 675 Richard, J., Schifferli, A., Schlenk, D., Scholz, S., Shiraish, F., Snyder, S., Su, G.Y., Tang, J.Y.M., van der 676 Burg, B., van der Linden, S.C., Werner, I., Westerheide, S.D., Wong, C.K.C., Yang, M., Yeung, B.H.Y., 677 Zhang, X.W. and Leusch, F.D.L. (2014) Benchmarking organic micropollutants in wastewater, recycled 678 water and drinking water with in vitro bioassays. Environmental Science & Technology 48(3), 1940-679 1956. 680

681 Escher, B.I., Bramaz, N., Mueller, J.F., Quayle, P., Rutishauser, S. and Vermeirssen, E.L.M. (2008) 682 Toxic equivalent concentrations (TEQs) for baseline toxicity and specific modes of action as a tool to 683 improve interpretation of ecotoxicity testing of environmental samples. Journal of Environmental 684 Monitoring 10(5), 612-621. 685

686 Escher, B.I., Dutt, M., Maylin, E., Tang, J.Y.M., Toze, S., Wolf, C.R. and Lang, M. (2012) Water quality 687 assessment using the AREc32 reporter gene assay indicative of the oxidative stress response 688 pathway. Journal of Environmental Monitoring 14(11), 2877-2885. 689

690 Escher, B.I., van Daele, C., Dutt, M., Tang, J.Y.M. and Altenburger, R. (2013) Most oxidative stress 691 response in water samples comes from unknown chemicals: the need for effect-based water quality 692 trigger values. Environmental Science & Technology 47(13), 7002-7011. 693

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

33

694 Farré, M.J., Day, S., Neale, P.A., Stalter, D., Tang, J.Y.M. and Escher, B.I. (2013) Bioanalytical and 695 chemical assessment of the disinfection by-product formation potential: Role of organic matter. 696 Water Research 47(14), 5409-5421. 697

698 Fukushima, S., Hibino, T., Shibata, M., Tsuda, H., Shirai, T., Takahashi, M. and Ito, N. (1977) Effect of 699 iodoacetamide or TWEEN-60 on methylnitrosocyanamide carcinogenesis in rat glandular and 700 forestomach. Gann 68(6), 813-818. 701

702 Geiss, K.T. and Frazier, J.M. (2001) QSAR modeling of oxidative stress in vitro following hepatocyte 703 exposures to halogenated methanes. Toxicology In Vitro 15(4-5), 557-563. 704

705 Gwynn, R.H. and Salaman, M.H. (1953) Studies on co-carcinogenesis - SH rectors and other 706 substances tested for co-carcinogenic action in mouse skin. British Journal of Cancer 7(4), 482-489. 707

708 Harder, A., Escher, B.I., Landini, P., Tobler, N.B. and Schwarzenbach, R.P. (2003a) Evaluation of 709 bioanalytical assays for toxicity assessment and mode of toxic action classification of reactive 710 chemicals. Environmental Science & Technology 37(21), 4962-4970. 711

712 Harder, A., Escher, B.I. and Schwarzenbach, R.P. (2003b) Applicability and limitation of QSARs for the 713 toxicity of electrophilic chemicals. Environmental Science & Technology 37(21), 4955-4961. 714

715 Huang, C.P., Wang, Y.J. and Chen, C.Y. (2007) Toxicity and quantitative structure-activity 716 relationships of nitriles based on Pseudokirchneriella subcapitata. Ecotoxicology and Environmental 717 Safety 67(3), 439-446. 718

719 Hulikova, A., Harris, A.L., Vaughan-Jones, R.D. and Swietach, P. (2013) Regulation of intracellular pH 720 in cancer cell lines under normoxia and hypoxia. Journal of Cellular Physiology 228(4), 743-752. 721

722 Kargalioglu, Y., McMillan, B.J., Minear, R.A. and Plewa, M.J. (2002) Analysis of the cytotoxicity and 723 mutagenicity of drinking water disinfection by-products in Salmonella typhimurium. Teratogenesis 724 Carcinogenesis and Mutagenesis 22(2), 113-128. 725

726 Kensler, T.W., Wakabayash, N. and Biswal, S. (2007) Cell survival responses to environmental 727 stresses via the Keap1-Nrf2-ARE pathway. Annual Review of Pharmacology and Toxicology 47, 89-728 116. 729

730 Komaki, Y., Marinas, B.J. and Plewa, M.J. (2014) Toxicity of drinking water disinfection by-products: 731 cell cycle alterations induced by the monohaloacetonitriles. Environmental Science & Technology 732 48(19), 11662-11669. 733

734

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

34

Komaki, Y., Pals, J., Wagner, E.D., Marinas, B.J. and Plewa, M.J. (2009) Mammalian Cell DNA Damage 735 and Repair Kinetics of Monohaloacetic Acid Drinking Water Disinfection By-Products. Environmental 736 Science & Technology 43(21), 8437-8442. 737

738 Kundu, B., Richardson, S.D., Swartz, P.D., Matthews, P.P., Richard, A.M. and DeMarini, D.M. (2004) 739 Mutagenicity in Salmonella of halonitromethanes: a recently recognized class of disinfection by-740 products in drinking water. Mutation Research-Genetic Toxicology and Environmental Mutagenesis 741 562(1-2), 39-65. 742

743 Lavin, M.F. and Gueven, N. (2006) The complexity of p53 stabilization and activation. Cell Death and 744 Differentiation 13(6), 941-950. 745

746 Li, J.H., Wang, W., Moe, B., Wang, H.L. and Li, X.F. (2015) Chemical and Toxicological 747 Characterization of Halobenzoquinones, an Emerging Class of Disinfection Byproducts. Chemical 748 Research in Toxicology 28(3), 306-318. 749

750 Life-Technologies (2006) CellSensor™ ARE-bla Hep G2 cell-based assay protocol. 751 Invitrogen/Lifetechnologies assay protocol K1208.pps. 752 http://tools.lifetechnologies.com/content/sfs/manuals/cellsensor_AREblaHepG2_man.pdf. 753

754 Life-Technologies (2009) CellSensor® NFκB-bla THP-1 cell-based assay protocol. 755 Invitrogen/Lifetechnologies assay protocol K1176.pps. 756 http://tools.lifetechnologies.com/content/sfs/manuals/CellSensor_NFkBbla_THP1_man.pdf. 757

758 Lipscomb, J.C., El-Demerdash, E. and Ahmed, A.E. (2009) Haloacetonitriles: metabolism and toxicity. 759 Reviews of Environmental Contamination and Toxicology 198, 169-200. 760

761 Liu, F.F., Peng, C., Escher, B.I., Fantino, E., Giles, C., Were, S., Duffy, L. and Ng, J.C. (2013) Hanging 762 drop: an in vitro air toxic exposure model using human lung cells in 2D and 3D structures. Journal of 763 Hazardous Materials 261, 701-710. 764

765 Magdeburg, A., Stalter, D., Schlüsener, M., Ternes, T. and Oehlmann, J. (2014) Evaluating the 766 efficiency of advanced wastewater treatment: target analysis of organic contaminants and (geno-767 )toxicity assessment tell a different story. Water Research 50(0), 35-47. 768

769 Merrick, B.A., Smallwood, C.L., Meier, J.R., McKean, D.L., Kaylor, W.H. and Condie, L.W. (1987) 770 Chemical reactivity, cytotoxicity, and mutagenicity of chloropropanones. Toxicology and Applied 771 Pharmacology 91(1), 46-54. 772

773 Neale, P.A., Antony, A., Bartkow, M.E., Farre, M.J., Heitz, A., Kristiana, I., Tang, J.Y.M. and Escher, B.I. 774 (2012) Bioanalytical assessment of the formation of disinfection byproducts in a drinking water 775 treatment plant. Environmental Science & Technology 46(18), 10317-10325. 776

777

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

35

Nieuwenhuijsen, M.J., Grellier, J., Smith, R., Iszatt, N., Bennett, J., Best, N. and Toledano, M. (2009) 778 The epidemiology and possible mechanisms of disinfection by-products in drinking water. 779 Philosophical Transactions of the Royal Society a-Mathematical Physical and Engineering Sciences 780 367(1904), 4043-4076. 781

782 NTP (2010) Toxicology and Carcinogenesis Studies of Dibromoacetonitrile (CAS No. 3252-43-5) in 783 F344/N Rats and B6C3F1 Mice (Drinking Water Studies), National Toxicology Program Technical 784 Report TR-544. NIEHS; National Toxicology Program, Research Triangle Park, N.C. 785

786 O'Brien, J., Wilson, I., Orton, T. and Pognan, F. (2000) Investigation of the Alamar Blue (resazurin) 787 fluorescent dye for the assessment of mammalian cell cytotoxicity. European Journal of Biochemistry 788 267(17), 5421-5426. 789

790 Pals, J., Attene-Ramos, M.S., Xia, M.H., Wagner, E.D. and Plewa, M.J. (2013) Human cell 791 toxicogenomic analysis linking reactive oxygen species to the toxicity of monohaloacetic acid 792 drinking water disinfection byproducts. Environmental Science & Technology 47(21), 12514-12523. 793

794 Pals, J.A., Ang, J.K., Wagner, E.D. and Plewa, M.J. (2011) Biological mechanism for the toxicity of 795 haloacetic acid drinking water disinfection byproducts. Environmental Science & Technology 45(13), 796 5791-5797. 797

798 Pinkus, R., Weiner, L.M. and Daniel, V. (1996) Role of oxidants and antioxidants in the induction of 799 AP-1, NF-kappa B, and glutathione S-transferase gene expression. Journal of Biological Chemistry 800 271(23), 13422-13429. 801

802 Plewa, M.J., Kargalioglu, Y., Vankerk, D., Minear, R.A. and Wagner, E.D. (2002) Mammalian cell 803 cytotoxicity and genotoxicity analysis of drinking water disinfection by-products. Environmental and 804 Molecular Mutagenesis 40(2), 134-142. 805

806 Plewa, M.J., Muellner, M.G., Richardson, S.D., Fasanot, F., Buettner, K.M., Woo, Y.T., McKague, A.B. 807 and Wagner, E.D. (2008a) Occurrence, synthesis, and mammalian cell cytotoxicity and genotoxicity 808 of haloacetamides: an emerging class of nitrogenous drinking water disinfection byproducts. 809 Environmental Science & Technology 42(3), 955-961. 810

811 Plewa, M.J., Simmons, J.E., Richardson, S.D. and Wagner, E.D. (2010) Mammalian cell cytotoxicity 812 and genotoxicity of the haloacetic acids, a major class of drinking water disinfection by-products. 813 Environmental and Molecular Mutagenesis 51(8-9), 871-878. 814

815 Plewa, M.J. and Wagner, E.D. (2009) Mammalian cell cytotoxicity and genotoxicity of disinfection by-816 products. 134. 817

818 Plewa, M.J., Wagner, E.D., Jazwierska, P., Richardson, S.D., Chen, P.H. and McKague, A.B. (2004a) 819 Halonitromethane drinking water disinfection byproducts: chemical characterization and 820 mammalian cell cytotoxicity and genotoxicity. Environmental Science & Technology 38(1), 62-68. 821

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

36

822 Plewa, M.J., Wagner, E.D., Muellner, M.G., Hsu, K.M. and Richardson, S.D. (2008b) Disinfection by-823 products in drinking water: occurrence, formation, health effects, and control. Karanfil, T., Krasner, 824 S.W. and Xie, Y. (eds), pp. 36-50, American Chemical Society, Washington. 825

826 Plewa, M.J., Wagner, E.D., Richardson, S.D., Thruston, A.D., Woo, Y.T. and McKague, A.B. (2004b) 827 Chemical and biological characterization of newly discovered lodoacid drinking water disinfection 828 byproducts. Environmental Science & Technology 38(18), 4713-4722. 829

830 Reifferscheid, G. and Heil, J. (1996) Validation of the SOS/umu test using test results of 486 831 chemicals and comparison with the Ames test and carcinogenicity data. Mutation Research-Genetic 832 Toxicology 369(3-4), 129-145. 833

834 Reifferscheid, G., Heil, J., Oda, Y. and Zahn, R.K. (1991) A microplate version of the Sos/Umu-test for 835 rapid detection of genotoxins and genotoxic potentials of environmental-samples. Mutation 836 Research 253(3), 215-222. 837

838 Richard, A.M. and Hunter, E.S. (1996) Quantitative structure-activity relationships for the 839 developmental toxicity of haloacetic acids in mammalian whole embryo culture. Teratology 53(6), 840 352-360. 841

842 Richardson, S.D., Plewa, M.J., Wagner, E.D., Schoeny, R. and DeMarini, D.M. (2007) Occurrence, 843 genotoxicity, and carcinogenicity of regulated and emerging disinfection by-products in drinking 844 water: a review and roadmap for research. Mutation Research-Reviews in Mutation Research 636(1-845 3), 178-242. 846

847 Schultz, T.W., Carlson, R.E., Cronin, M.T.D., Hermens, J.L.M., Johnson, R., O'Brien, P.J., Roberts, D.W., 848 Siraki, A., Wallace, K.B. and Veith, G.D. (2006) A conceptual framework for predicting the toxicity of 849 reactive chemicals: modeling soft electrophilicity. SAR and QSAR in Environmental Research 17(4), 850 413-428. 851

852 Shukla, S.J., Huang, R.L., Simmons, S.O., Tice, R.R., Witt, K.L., VanLeer, D., Ramabhadran, R., Austin, 853 C.P. and Xia, M.H. (2012) Profiling environmental chemicals for activity in the antioxidant response 854 element signaling pathway using a high throughput screening approach. Environmental Health 855 Perspectives 120(8), 1150-1156. 856

857 Simmons, S.O., Fan, C.Y. and Ramabhadran, R. (2009) Cellular stress response pathway system as a 858 sentinel ensemble in toxicological screening. Toxicological Sciences 111(2), 202-225. 859

860 Stalter, D., Dutt, M. and Escher, B. (2013) Headspace-free setup of in vitro bioassays for the 861 evaluation of volatile disinfection by-products. Chemical Research in Toxicology 26, 1605-1614. 862

863

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

37

Takahashi, M., Shirai, T., Fukushima, S., Hahanouchi, M., Hirose, M. and Ito, N. (1976) Effect of fundic 864 ulcers induced by iodoacetamide on development of gastric tumors in rats treated with n-methyl-n'-865 nitro-n-nitrosoguanidine. Gann 67(1), 47-54. 866

867 Tang, J.Y.M., Glenn, E., Thoen, H. and Escher, B.I. (2012) In vitro bioassay for reactive toxicity 868 towards proteins implemented for water quality monitoring. Journal of Environmental Monitoring 869 14(3), 1073-1081. 870

871 Tang, J.Y.M., McCarty, S., Glenn, E., Neale, P.A., Warne, M.S.J. and Escher, B.I. (2013) Mixture effects 872 of organic micropollutants present in water: Towards the development of effect-based water quality 873 trigger values for baseline toxicity. Water Research 47(10), 3300-3314. 874

875 Thier, R., Taylor, J.B., Pemble, S.E., Humphreys, W.G., Persmark, M., Ketterer, B. and Guengerich, F.P. 876 (1993) Expression of mammalian glutathione-S-transferase 5-5 in Salmonella typhimurium TA1535 877 leads to base-pair mutations upon exposure to dihalomethanes. Proceedings of the National 878 Academy of Sciences of the United States of America 90(18), 8576-8580. 879

880 Toledo, F. and Wahl, G.M. (2006) Regulating the p53 pathway: in vitro hypotheses, in vivo veritas. 881 Nature Reviews Cancer 6(12), 909-923. 882

883 van der Linden, S.C., von Bergh, A.R.M., van Vught-Lussenburg, B.M.A., Jonker, L.R.A., Teunis, M., 884 Krul, C.A.M. and van der Burg, B. (2014) Development of a panel of high-throughput reporter-gene 885 assays to detect genotoxicity and oxidative stress. Mutation Research-Genetic Toxicology and 886 Environmental Mutagenesis 760, 23-32. 887

888 Verhaar, H.J.M., Vanleeuwen, C.J. and Hermens, J.L.M. (1992) Classifying environmental polutants. 1. 889 Structure-activity-relationships for prediction of aquatic toxicity. Chemosphere 25(4), 471-491. 890

891 Wagner, E.D., Hsu, K.-M., Lagunas, A., Mitch, W.A. and Plewa, M.J. (2012) Comparative genotoxicity 892 of nitrosamine drinking water disinfection byproducts in Salmonella and mammalian cells. Mutation 893 Research 741(1-2), 109-115. 894

895 Wilkening, S., Stahl, F. and Bader, A. (2003) Comparison of primary human hepatocytes and 896 hepatoma cell line HEPG2 with regard to their biotransformation properties. Drug Metabolism and 897 Disposition 31(8), 1035-1042. 898

899 Wu, Z.H. and Miyamoto, S. (2007) Many faces of NF-KB signaling induced by genotoxic stress. Journal 900 of Molecular Medicine 85(11), 1187-1202. 901

902 Yamada, M., Matsui, K., Sofuni, T. and Nohmi, T. (1997) New tester strains of Salmonella 903 typhimurium lacking O-6-methylguanine DNA methyltransferases and highly sensitive to mutagenic 904 alkylating agents. Mutation Research-Fundamental and Molecular Mechanisms of Mutagenesis 905 381(1), 15-24. 906

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

38

907 Yeh, R.Y.L., Farré, M.J., Stalter, D., Tang, J.Y.M., Molendijk, J. and Escher, B.I. (2014) Bioanalytical and 908 chemical evaluation of disinfection by-products in swimming pool water. Water Research 59(0), 172-909 184. 910

911 Yuan, J., Liu, H., Zhou, L.H., Zou, Y.L. and Lu, W.Q. (2006) Oxidative stress and DNA damage induced 912 by a drinking-water chlorination disinfection byproduct 3-chloro-4-(dichloromethyl)-5-hydroxy-913 2(5H)-furanone (MX) in mice. Mutation Research-Genetic Toxicology and Environmental 914 Mutagenesis 609(2), 129-136. 915

916

Tables 917

Table 1: Bioassays used for the effect fingerprinting of 50 DBPs; OD: optical density.

bioassay test species (strain/cell line) endpoint detected signal

Microtox Aliivibrio fischeri cytotoxicity bioluminescence as indicator for

cell viability

E. coli ±GSH Escherichia coli MJF335 (GSH−)

and MJF276 (GSH+)

interaction with

proteins/peptides

OD at 600 nm as indicator for cell

density and descriptor of cell

growth

E. coli ±DNA Escherichia coli MV4108 (DNA−)

and MV1161 (DNA+)

interaction with DNA OD at 600 nm as indicator for cell

density and descriptor of cell

growth umuC assay Salmonella typhimurium

(TA1535/pSK1002)

activation of SOS-

response, ±S9 for

metabolic activation

OD at 420 nm as marker for

conversion of substrate by the

reporter enzyme

Ames assay Salmonella typhimurium

(YG7108)

mutagenicity, with S9

(only for nitrosamines)

OD at 415 nm to detect the color

change of the pH indicator which

indicates growth of revertants

AREc32 AREc32 cell line, based on

human breast cancer cell line

MCF7

activation of the oxidative

stress response pathway

NRf2-ARE

bioluminescence as indicator of

the reporter enzyme luciferase

ARE-bla ARE-bla HepG2 cell line, based on hepatocellular carcinoma cell

line HepG2

activation of the oxidative stress response pathway

NRf2-ARE

ratio of blue (450 nm) to green fluorescence emission (520 nm)

at 405 nm excitation as indicator

of the reporter enzyme β-

lactamase

p53-bla p53RE-bla HCT-116 cell line,

based on human colon

carcinoma cell line HCT-116

activation of the tumor

suppressor protein p53

same as for ARE-bla

NF-κB-bla NFκB-bla THP-1 cell cine, based

on human leukemia cell line THP-1

activation of the stress

response pathway for inflammation (NF-κB)

same as for ARE-bla

918

919

920

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

39

Table 2. Summary of effect concentrations (EC50 for Microtox and ECIR1.5 for the other assays, M

(mol/L)). Colors indicate the potency from yellow (low potency) to red (high potency). Detailed

results including coefficient of variation are given in Table S4. Effect concentration of the E. coli

assays are given in Table S5. NF-κB was inactive for all compounds. AREc32 ARE-bla p53-bla Microtox umuC -S9 umuC +S9 Ames

DBP ECIR1.5a ECIR1.5

a ECIR1.5

a EC50

a ECIR1.5

a ECIR1.5

a ECRR1.5

a

Halomethanes 1,1-dCE 2.9E-03 n.e.≤4E-2

b n.e.≤1E-2

b 5.2E-03 n.e.≤1E-2

b n.e.≤1E-2

b n.t.

c

dCM 4.1E-02 n.e.≤5E-2b n.e.≤5E-2

b 3.7E-02 4.4E-02 1.1E-01 n.t.

c

BCM 1.1E-03 4.1E-02 n.e.≤2E-2b 1.0E-02 1.2E-03 1.8E-03 n.t.

c

tCM 1.4E-02 n.e.≤4E-2b n.e.≤3E-2

b 6.8E-03 1.3E-02 7.6E-02 n.t.

c

BdCM 6.1E-03 n.e.≤4E-2b n.e.≤1E-2

b 1.8E-03 1.4E-03 6.5E-03 n.t.

c

tBM 1.4E-03 n.e.≤4E-2b n.e.≤6E-3

b 2.3E-04 5.2E-04 2.5E-03 n.t.

c

dBCM 1.9E-03 1.6E-02 n.e.≤1E-2b 1.0E-03 5.5E-04 3.5E-03 n.t.

c

dCIM 1.6E-04 2.7E-03 2.3E-03 3.2E-04 2.1E-04 7.4E-04 n.t.c

BCIM 1.2E-04 2.8E-03 2.9E-03 9.7E-05 1.8E-04 5.0E-04 n.t.c

dBIM 9.2E-05 1.8E-03 1.8E-03 8.9E-05 5.3E-05 2.0E-04 n.t.c

CdIM 2.7E-05 2.8E-04 2.6E-04 7.1E-05 5.6E-05 1.5E-04 n.t.c

BdIM 5.5E-05 5.2E-04 6.5E-05 2.2E-05 2.6E-05 7.9E-05 n.t.c

tIM 2.6E-05 2.3E-04 1.7E-05 9.4E-06 1.9E-05 8.7E-05 n.t.c

Halonitromethanes tCNM 8.1E-06 n.e.≤4E-4

b 6.9E-05 3.8E-08 1.3E-05 2.2E-05 n.t.

c

tBNM 4.9E-06 n.e.≤4E-4b 6.2E-05 3.0E-07 1.2E-06 3.5E-06 n.t.

c

Haloacetonitriles dCAN 7.7E-06 4.8E-05 2.7E-05 4.9E-05 6.0E-05 8.0E-05 n.t.

c

tCAN 1.4E-05 1.5E-04 1.4E-05 2.5E-06 1.7E-05 1.4E-05 n.t.c

BCAN 2.2E-06 1.1E-05 1.3E-05 1.3E-05 3.7E-05 9.3E-05 n.t.c

dBAN 1.5E-07 7.0E-06 1.3E-05 8.5E-06 4.1E-05 5.0E-05 n.t.c

Haloketones 1,1-dCP 6.8E-07 n.e.≤4E-3

b 3.4E-05 3.0E-04 1.1E-04 4.9E-04 n.t.

c

1,1,1-tCP 1.5E-05 n.e.≤4E-3b 7.3E-05 2.2E-04 1.7E-04 6.3E-04 n.t.

c

Haloacetic acid CAA 2.7E-04 2.5E-04 1.7E-04 3.8E-03 n.e.≤8E-3

b n.e.≤8E-3

b n.t.

c

BAA 5.2E-06 1.1E-05 9.5E-06 3.8E-05 n.e.≤2E-4b n.e.≤2E-4

b n.t.

c

IAA 3.6E-06 5.1E-06 4.7E-06 1.7E-05 n.e.≤2E-4b n.e.≤2E-4

b n.t.

c

dCAA 6.0E-03 1.6E-02 n.e.≤3E-2b 3.7E-03 n.e.≤2E-2

b n.e.≤2E-2

b n.t.

c

BCAA 1.4E-04 4.6E-04 2.3E-04 1.2E-03 3.4E-04 9.4E-04 n.t.c

dBAA 1.2E-04 2.5E-04 2.6E-04 8.5E-04 3.9E-04 6.7E-04 n.t.c

CIAA 2.2E-05 1.0E-04 1.1E-04 3.1E-05 1.9E-04 4.8E-04 n.t.c

BIAA 2.6E-05 5.3E-05 1.1E-04 1.6E-04 1.1E-04 8.7E-05 n.t.c

tCAA n.e.≤2E-2b n.e.≤2E-2

b n.e.≤2E-2

b 1.3E-02 n.e.≤2E-2

b n.e.≤2E-2

b n.t.

c

BdCAA 2.0E-03 4.0E-03 n.e.≤3E-3b 6.1E-04 1.1E-04 2.2E-03 n.t.

c

dBCAA 4.9E-03 2.2E-03 n.e.≤2E-3b 4.2E-04 1.1E-04 1.6E-03 n.t.

c

tBAA 4.4E-04 6.7E-04 n.e.≤5E-4b 1.3E-04 7.0E-06 7.2E-05 n.t.

c

Haloacetaldehyde CH 1.7E-04 6.1E-04 9.8E-03 1.5E-02 n.e.≤3E-2

b n.e.≤3E-2

b n.t.

c

Haloacetamides dCAcAm 1.2E-03 1.8E-03 n.e.≤3E-2

b 3.7E-02 n.e.≤2E-2

b n.e.≤2E-2

b n.t.

c

BCAcAm 1.4E-05 3.7E-05 3.1E-04 2.7E-03 2.7E-03 2.3E-03 n.t.c

dBAcAm 4.7E-06 2.1E-05 2.4E-05 1.8E-05 1.7E-05 7.8E-06 n.t.c

CIAcAm 5.1E-06 6.4E-05 6.0E-05 5.1E-04 3.4E-04 1.2E-04 n.t.c

BIAcAm 3.3E-06 1.6E-05 4.4E-05 2.1E-03 4.4E-04 3.8E-04 n.t.c

dIAcAm 5.4E-07 2.7E-06 4.9E-05 1.9E-03 1.5E-04 3.6E-05 n.t.c

tCAcAm 1.2E-03 4.7E-03 n.e.≤6E-3b 1.4E-04 3.2E-03 2.8E-03 n.t.

c

BdCAcAm 3.2E-06 2.0E-05 2.4E-05 4.9E-05 2.2E-05 1.2E-05 n.t.c

dBCAcAm 1.2E-06 1.3E-05 3.5E-03 4.2E-05 1.3E-05 6.1E-06 n.t.c

tBAcAm 6.6E-06 8.8E-06 2.2E-05 6.9E-06 1.0E-05 6.9E-06 n.t.c

Nitrosamines NDMA 3.3E-03 n.e.≤1E-2

b 1.3E-03 2.5E-02 1.6E-03 2.3E-03 3.9E-04

NDEA 2.4E-03 n.e.≤1E-2b 4.6E-04 7.4E-03 8.1E-04 2.9E-03 4.2E-04

N-Pip 5.1E-04 n.e.≤1E-2b n.e.≤2E-3

b 7.0E-04 1.0E-03 1.5E-03 6.3E-04

N-Morph 1.8E-03 n.e.≤9E-3b n.e.≤2E-3

b 2.2E-03 1.4E-03 2.0E-03 2.1E-03

N-But 3.0E-04 1.5E-03 9.8E-04 2.6E-04 5.5E-04 6.1E-04 5.3E-04 Furanone MX 6.5E-06 4.7E-05 5.3E-06 9.8E-07 1.3E-08 4.0E-08 n.t.

c

a arithmetic mean of all experiments (Table S4); b no effect up to the highest concentration

tested; c not tested

921

922

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

40

Table 3: Results of the linear regression analysis between effect concentrations

(Table S4) and reactivity descriptor ELUMO (Table S6).

THMs HANs HAAs HAcAms

AREc32

r2 0.98 0.27 <0.01 0.58

deviation from zero? significant not significant not significant significant

number of X values 10 4 11 10

ARE-bla

r2 0.93 0.11 0.01 0.47

deviation from zero? significant not significant not significant significant

number of X values 7 4 11 10

p53-bla

r2 0.96 0.96 0.06 0.03

deviation from zero? significant significant not significant not significant

number of X values 6 4 7 8

923

924

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

41

Figure Legends 925

926

Fig. 1 – Toxicity pathways of reactive chemicals, chemical descriptors, and molecular 927

initiating events (MIEs) of reactive chemicals as well as bioassays indicative of MIEs, 928

intermediate effects, cellular stress responses, and apical cellular effects (Table 1). 929

Abbreviations: KOW, octanol-water partition coefficient; pKa, acidity constant, ELUMO, 930

energy of the lowest unoccupied molecular orbital. 931

932

Fig. 2 – EC-values of 50 DBPs for cytotoxicity (Microtox), oxidative stress induction 933

(AREc32), and genotoxicity (p53-bla, umuC, Ames assay (nitrosamines only)). The umuC 934

assay was performed with and without metabolic activation (±S9) and only the lower ECs 935

are displayed. The symbols with dots in the middle are extrapolated data. ECs of all 936

bioassays and abbreviations are listed in Tables S1, S4 & S5. 937

938

Fig. 3 – Comparison of ECs in different bioassays: (A) AREc32 on oxidative stress vs. 939

Microtox on bacterial cytotoxicity; (B) AREc32 vs. ARE-bla assay on oxidative stress (tCAA 940

negative in both; HKs, HNMs, BdCM, tBM, tCM, 1,1-dCE, dCM, and four nitrosamines 941

negative in the ARE-bla); (C) AREc32 vs. p53 induction assay as mammalian genotoxicity 942

marker. The blue line indicates a theoretical 100% agreement and the dotted lines 943

correspond to a one order of magnitude deviation. 944

945

Fig. 4 – Reactivity descriptor ELUMO versus effect concentrations in the oxidative stress 946

induction assays (A) AREc32 and (B) ARE-bla and in the p53 induction assay (C) p53-bla. All 947

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

42

linear regressions were performed for each group individually (only if n ≥ 4) and only 948

significant regression lines (p < 0.05) are displayed. Reactivity descriptors are collated in 949

Table S6. 950

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Table 1: Bioassays used for the effect fingerprinting of 50 DBPs; OD: optical density.

bioassay test species (strain/cell line) endpoint detected signal

Microtox Aliivibrio fischeri cytotoxicity bioluminescence as indicator for

cell viability

E. coli ±GSH Escherichia coli MJF335 (GSH−)

and MJF276 (GSH+)

interaction with

proteins/peptides

OD at 600 nm as indicator for cell

density and descriptor of cell

growth

E. coli ±DNA Escherichia coli MV4108 (DNA−)

and MV1161 (DNA+)

interaction with DNA OD at 600 nm as indicator for cell

density and descriptor of cell

growth

umuC assay Salmonella typhimurium

(TA1535/pSK1002)

activation of SOS-

response, with and

without S9 for metabolic

activation

OD at 420 nm as marker for

conversion of substrate by the

reporter enzyme

Ames assay Salmonella typhimurium

(YG7108)

mutagenicity, with S9

(only for nitrosamines)

OD at 415 nm to detect the color

change of the pH indicator which

indicates growth of revertants

AREc32 AREc32 cell line, based on

human breast cancer cell line

MCF7

activation of the oxidative

stress response pathway

NRf2-ARE

bioluminescence as indicator of

the reporter enzyme luciferase

ARE-bla ARE-bla HepG2 cell line, based

on hepatocellular carcinoma cell

line HepG2

activation of the oxidative

stress response pathway

NRf2-ARE

ratio of blue (450 nm) to green

fluorescence emission (520 nm)

at 405 nm excitation as indicator

of the reporter enzyme β-

lactamase

p53-bla p53RE-bla HCT-116 cell line,

based on human colon

carcinoma cell line HCT-116

activation of the tumor

suppressor protein p53

same as for ARE-bla

NF-κB-bla NFκB-bla THP-1 cell cine, based

on human leukemia cell line

THP-1

activation of the stress

response pathway for

inflammation (NF-κB)

same as for ARE-bla

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Table 2. Summary of effect concentrations (EC50 for Microtox and ECIR1.5 for the other assays, M

(mol/L)). Colors indicate the potency from yellow (low potency) to red (high potency). Detailed

results including coefficient of variation are given in Table S4. Effect concentration of the E. coli

assays are given in Table S5. NF-κB was inactive for all compounds. AREc32 ARE-bla p53-bla Microtox umuC -S9 umuC +S9 Ames

DBP ECIR1.5a ECIR1.5

a ECIR1.5

a EC50

a ECIR1.5

a ECIR1.5

a ECRR1.5

a

Halomethanes 1,1-dCE 2.9E-03 n.e.≤4E-2

b n.e.≤1E-2

b 5.2E-03 n.e.≤1E-2

b n.e.≤1E-2

b n.t.

c

dCM 4.1E-02 n.e.≤5E-2b n.e.≤5E-2

b 3.7E-02 4.4E-02 1.1E-01 n.t.

c

BCM 1.1E-03 4.1E-02 n.e.≤2E-2b 1.0E-02 1.2E-03 1.8E-03 n.t.

c

tCM 1.4E-02 n.e.≤4E-2b n.e.≤3E-2

b 6.8E-03 1.3E-02 7.6E-02 n.t.

c

BdCM 6.1E-03 n.e.≤4E-2b n.e.≤1E-2

b 1.8E-03 1.4E-03 6.5E-03 n.t.

c

tBM 1.4E-03 n.e.≤4E-2b n.e.≤6E-3

b 2.3E-04 5.2E-04 2.5E-03 n.t.

c

dBCM 1.9E-03 1.6E-02 n.e.≤1E-2b 1.0E-03 5.5E-04 3.5E-03 n.t.

c

dCIM 1.6E-04 2.7E-03 2.3E-03 3.2E-04 2.1E-04 7.4E-04 n.t.c

BCIM 1.2E-04 2.8E-03 2.9E-03 9.7E-05 1.8E-04 5.0E-04 n.t.c

dBIM 9.2E-05 1.8E-03 1.8E-03 8.9E-05 5.3E-05 2.0E-04 n.t.c

CdIM 2.7E-05 2.8E-04 2.6E-04 7.1E-05 5.6E-05 1.5E-04 n.t.c

BdIM 5.5E-05 5.2E-04 6.5E-05 2.2E-05 2.6E-05 7.9E-05 n.t.c

tIM 2.6E-05 2.3E-04 1.7E-05 9.4E-06 1.9E-05 8.7E-05 n.t.c

Halonitromethanes tCNM 8.1E-06 n.e.≤4E-4

b 6.9E-05 3.8E-08 1.3E-05 2.2E-05 n.t.

c

tBNM 4.9E-06 n.e.≤4E-4b 6.2E-05 3.0E-07 1.2E-06 3.5E-06 n.t.

c

Haloacetonitriles dCAN 7.7E-06 4.8E-05 2.7E-05 4.9E-05 6.0E-05 8.0E-05 n.t.

c

tCAN 1.4E-05 1.5E-04 1.4E-05 2.5E-06 1.7E-05 1.4E-05 n.t.c

BCAN 2.2E-06 1.1E-05 1.3E-05 1.3E-05 3.7E-05 9.3E-05 n.t.c

dBAN 1.5E-07 7.0E-06 1.3E-05 8.5E-06 4.1E-05 5.0E-05 n.t.c

Haloketones 1,1-dCP 6.8E-07 n.e.≤4E-3

b 3.4E-05 3.0E-04 1.1E-04 4.9E-04 n.t.

c

1,1,1-tCP 1.5E-05 n.e.≤4E-3b 7.3E-05 2.2E-04 1.7E-04 6.3E-04 n.t.

c

Haloacetic acid CAA 2.7E-04 2.5E-04 1.7E-04 3.8E-03 n.e.≤8E-3

b n.e.≤8E-3

b n.t.

c

BAA 5.2E-06 1.1E-05 9.5E-06 3.8E-05 n.e.≤2E-4b n.e.≤2E-4

b n.t.

c

IAA 3.6E-06 5.1E-06 4.7E-06 1.7E-05 n.e.≤2E-4b n.e.≤2E-4

b n.t.

c

dCAA 6.0E-03 1.6E-02 n.e.≤3E-2b 3.7E-03 n.e.≤2E-2

b n.e.≤2E-2

b n.t.

c

BCAA 1.4E-04 4.6E-04 2.3E-04 1.2E-03 3.4E-04 9.4E-04 n.t.c

dBAA 1.2E-04 2.5E-04 2.6E-04 8.5E-04 3.9E-04 6.7E-04 n.t.c

CIAA 2.2E-05 1.0E-04 1.1E-04 3.1E-05 1.9E-04 4.8E-04 n.t.c

BIAA 2.6E-05 5.3E-05 1.1E-04 1.6E-04 1.1E-04 8.7E-05 n.t.c

tCAA n.e.≤2E-2b n.e.≤2E-2

b n.e.≤2E-2

b 1.3E-02 n.e.≤2E-2

b n.e.≤2E-2

b n.t.

c

BdCAA 2.0E-03 4.0E-03 n.e.≤3E-3b 6.1E-04 1.1E-04 2.2E-03 n.t.

c

dBCAA 4.9E-03 2.2E-03 n.e.≤2E-3b 4.2E-04 1.1E-04 1.6E-03 n.t.

c

tBAA 4.4E-04 6.7E-04 n.e.≤5E-4b 1.3E-04 7.0E-06 7.2E-05 n.t.

c

Haloacetaldehyde CH 1.7E-04 6.1E-04 9.8E-03 1.5E-02 n.e.≤3E-2

b n.e.≤3E-2

b n.t.

c

Haloacetamides

dCAcAm 1.2E-03 1.8E-03 n.e.≤3E-2b 3.7E-02 n.e.≤2E-2

b n.e.≤2E-2

b n.t.

c

BCAcAm 1.4E-05 3.7E-05 3.1E-04 2.7E-03 2.7E-03 2.3E-03 n.t.c

dBAcAm 4.7E-06 2.1E-05 2.4E-05 1.8E-05 1.7E-05 7.8E-06 n.t.c

CIAcAm 5.1E-06 6.4E-05 6.0E-05 5.1E-04 3.4E-04 1.2E-04 n.t.c

BIAcAm 3.3E-06 1.6E-05 4.4E-05 2.1E-03 4.4E-04 3.8E-04 n.t.c

dIAcAm 5.4E-07 2.7E-06 4.9E-05 1.9E-03 1.5E-04 3.6E-05 n.t.c

tCAcAm 1.2E-03 4.7E-03 n.e.≤6E-3b 1.4E-04 3.2E-03 2.8E-03 n.t.

c

BdCAcAm 3.2E-06 2.0E-05 2.4E-05 4.9E-05 2.2E-05 1.2E-05 n.t.c

dBCAcAm 1.2E-06 1.3E-05 3.5E-03 4.2E-05 1.3E-05 6.1E-06 n.t.c

tBAcAm 6.6E-06 8.8E-06 2.2E-05 6.9E-06 1.0E-05 6.9E-06 n.t.c

Nitrosamines NDMA 3.3E-03 n.e.≤1E-2

b 1.3E-03 2.5E-02 1.6E-03 2.3E-03 3.9E-04

NDEA 2.4E-03 n.e.≤1E-2b 4.6E-04 7.4E-03 8.1E-04 2.9E-03 4.2E-04

N-Pip 5.1E-04 n.e.≤1E-2b n.e.≤2E-3

b 7.0E-04 1.0E-03 1.5E-03 6.3E-04

N-Morph 1.8E-03 n.e.≤9E-3b n.e.≤2E-3

b 2.2E-03 1.4E-03 2.0E-03 2.1E-03

N-But 3.0E-04 1.5E-03 9.8E-04 2.6E-04 5.5E-04 6.1E-04 5.3E-04

Furanone MX 6.5E-06 4.7E-05 5.3E-06 9.8E-07 1.3E-08 4.0E-08 n.t.

c

a arithmetic mean of all experiments (Table S4);

b no effect up to the highest concentration

tested; c not tested

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Table 3: Results of the linear regression analysis between effect concentrations

(Table S4) and reactivity descriptor ELUMO (Table S6).

THMs HANs HAAs HAcAms

AREc32

r2 0.98 0.27 <0.01 0.58

deviation from zero? significant not significant not significant significant

number of X values 10 4 11 10

ARE-bla

r2 0.93 0.11 0.01 0.47

deviation from zero? significant not significant not significant significant

number of X values 7 4 11 10

p53-bla

r2 0.96 0.96 0.06 0.03

deviation from zero? significant significant not significant not significant

number of X values 6 4 7 8

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

MANUSCRIP

T

ACCEPTED

ACCEPTED MANUSCRIPT

Highlights

• 98% of 50 DBPs activated NRf2 mediated oxidative stress response

• 68% of 50 DBPs activated p53 which indicates possible genotoxic properties

• potency of trihalomethanes and haloacetamides is driven by reactivity

• potency of haloacetic acids is driven by both, reactivity and bioavailability

• indirect genotoxic mechanisms are more likely than direct DNA damage for most DBPs