Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Jul 1.
Published in final edited form as: Neurotoxicology. 2020 May 5;79:95–103. doi: 10.1016/j.neuro.2020.04.009

Application of the Hard and Soft, Acids and Bases (HSAB) Theory as a Method to Predict Cumulative Neurotoxicity

Fjodor Melnikov 1,2, Brian C Geohagen 3, Terrence Gavin 4, Richard M LoPachin 3, Paul T Anastas 5, Phillip Coish 6, David W Herr 7,*
PMCID: PMC7369154  NIHMSID: NIHMS1599056  PMID: 32380191

Abstract

Xenobiotic electrophiles can form covalent adducts that may impair protein function, damage DNA, and may lead a range of adverse effects. Cumulative neurotoxicity is one adverse effect that has been linked to covalent protein binding as a Molecular Initiating Event (MIE). This paper describes a mechanistic in silico chemical screening approach for neurotoxicity based on Hard and Soft Acids and Bases (HSAB) theory. We evaluated the applicability of HSAB-based electrophilicity screening protocol for neurotoxicity on 19 positive and 19 negative reference chemicals. These reference chemicals were identified from the literature, using available information on mechanisms of neurotoxicity whenever possible. In silico screening was based on structural alerts for protein binding motifs and electrophilicity index in the range of known neurotoxicants. The approach demonstrated both a high positive prediction rate (82–90%) and specificity (90%). The overall sensitivity was relatively lower (47%). However, when predicting the toxicity of chemicals known or suspected of acting via non-specific adduct formation mechanism, the HSAB approach identified 7/8 (sensitivity 88%) of positive control chemicals correctly. Consequently, the HSAB-based screening is a promising approach of identifying possible neurotoxins with adduct formation molecular initiating events. While the approach must be expanded over time to capture a wider range of MIEs involved in neurotoxicity, the mechanistic nature of the screen allows users to flag chemicals for possible adduct formation MIEs. Thus, the HSAB based toxicity screening is a promising strategy for toxicity assessment and chemical prioritization in neurotoxicology and other health endpoints that involve adduct formation.

Keywords: Hard and Soft Acid Base, Neurotoxicity, In Silico

1. Introduction

Each year new chemicals are introduced into the pharmaceutical, agricultural and industrial markets, with an annual growth rate estimated at 3% (Krimsky, 2017; Pool and Rusch, 2014; U.S. GAO, 2005; Wilson and Schwarzman, 2009). The existing volume of chemicals and the rapid growth in new chemicals necessitate new methods for assessing and predicting corresponding toxic potential (LoPachin and Gavin, 2014; LoPachin et al., 2019). Many available in vitro and in vivo toxicity assessment technologies cannot yet provide information on molecular mechanisms of action and Molecular Initiating Events (MIE’s) involved in toxicity. Previous reports have suggested deficiencies of several in vitro test batteries’ ability to detect neurotoxicants (Shah and Greene, 2013; Silva et al., 2015), or lower test sensitivity for specific chemical types, such as drugs (Sirenko et al., 2019).

In silico models, such as structure-activity relationship (SAR), provide an alternative for chemical screening and testing triage. However, in silico models work best when they incorporate mechanistic insights (Melnikov et al., 2016; Schultz et al., 2006). Many environmental and occupational chemicals have been shown to induce toxicity through covalent binding with endogenous macromolecules (LoPachin and DeCaprio, 2005; LoPachin and Gavin, 2012; LoPachin et al., 2012; Thomas et al., 2013). The reactions between xenobiotic electrophiles and biological nucleophiles produce covalent adducts that impair functions of enzymes, DNA, and cytoskeletal proteins, leading to defective cell processes and cytotoxicity (Barber and LoPachin, 2004; Barber et al., 2007; Enoch et al., 2008a; LoPachin et al., 2006, 2007b; Schultz et al., 2006; Schwöbel et al., 2011). For example, electrophilic α,β-unsaturated carbonyl compounds such as acrolein, acrylamide, and methylvinyl ketone cause broad organ system toxicity by forming covalent Michael-type adducts through nucleophilic addition to sulfhydryl groups on functionally-critical proteins (Barber et al., 2007; LoPachin and DeCaprio, 2005; LoPachin et al., 2007b, 2009, 2019; Martyniuk et al., 2011). Structural features can be used to screen chemicals for the potential to form covalent adducts though a series of structural alerts for common reaction mechanisms (Enoch et al., 2008a). For example, in silico screening models may identify compounds active in Direct Peptide Reactivity Assay (DPRA) assays (Alves et al., 2015; Lalko et al., 2012). However, the reactions between electrophiles and nucleophiles are not indiscriminate. Electrophiles react preferentially with different nucleophiles and the rate of interactions occur along a continuum of reactivity.

The potential for adduct formation between a toxic electrophile and a biological nucleophile can be described with the Hard and Soft Acids and Bases (HSAB) theory of Pearson (Pearson, 1963, 2005). The HSAB theory states that chemicals with soft (polarizable) electrophilic centers will react more readily with soft (polarizable) nucleophiles. Conversely, harder (non-polarizable) electrophiles will react preferentially with harder biological nucleophiles. For example, the α,β-unsaturated carbonyls have been shown to induce cytotoxicity by selectively reacting with soft nucleophiles, such as anionic thiolates on critical proteins (Enoch et al., 2008a; LoPachin et al., 2007a; Martyniuk et al., 2011; Schultz et al., 2005). Chemical reactivity can be modeled by the energies of their frontier orbitals. Specifically, Highest Occupied Molecular Orbital (HOMO) of the nucleophile and the Lowest Unoccupied Molecular Orbital (LUMO) of the electrophile will dictate the rate of adduct formation. These energies describe the transfer of electron density from the donating HOMO of the nucleophile to the recipient LUMO of the electrophile. The LUMO of electrophile can approximate chemical reactivity since common biological nucleophiles are expected to share similar HOMOs. Chemical softness (σ), hardness (η), and electrophilicity index (ω) can be calculated from HOMO and LUMO energies (see Equation 1). The electrophilicity index can be used to model chemical reactivity and has been shown to correlate with the rate constant of adduct-forming reactions, with stronger, more electrophilic compound reacting more readily with biological targets t007han their weaker, less electrophilic, counterparts (LoPachin et al., 2006; 2007b, 2009, 2016). Additionally, the electrophilicity index provides a useful tool in predicting chemical potential to form biologically-relevant adducts (LoPachin and Gavin, 2014; LoPachin et al., 2019; Schultz et al., 2005).

HSAB theory and associated descriptors has been applied to a broad range of chemical toxicity and molecular design questions (Allen et al., 2018; Coish et al., 2017; Enoch et al., 2008a; Garcia-Serna et al., 2015; Judson et al., 2016; Melnikov et al., 2016; Schwöbel et al., 2011; Shen et al., 2016a; Shen et al., 2016b; Yang et al., 2018). This work focused on HSAB applications to predict cumulative neurotoxicity. For the purposes of this manuscript, we will consider cumulative neurotoxicity as peripheral or central neuropathological changes which require repeated exposure to a xenobiotic and may involve altered protein function, possibly through the formation of adducts with the xenobiotic (or it’s metabolites). HSAB-based models for cumulative neurotoxicity have been applied successfully to small set of chemicals, such as acrolein, acrylamide, and α,β-unsaturated carbonyls (LoPachin et al., 2002, 2006, 2007b, 2009; Zhang et al., 2010). Cumulative neurotoxicity is a vital human health problem that is not easily assessed or predicted (for example n-propylbromide; http://www.nytimes.com/2013/03/31/us/osha-emphasizes-safety-health-risks-fester.html?ref=todayspaper&_r=0). However, cumulative neurotoxicity is a multifaceted concept, with diverse MIE types and toxic manifestations. Modeling all MIEs and mechanisms involved in neurotoxicity is an incredibly complicated task and is beyond the scope of this paper. However, HSAB theory can inform adduct formation MIEs that is applicable to a range of toxic outcomes, including, but not limited to, neurotoxicity (Figure 1). For example, many neurotoxic chemicals have additional toxic effects on testicular sertoli cells (Moffit et al., 2007), which contain microtubules with nucleophile-rich sites similar to neurons. The challenge is to distinguish the likeliest biological consequence based on biological knowledge and quantitative HSAB reactivity descriptors (σ, η, ω). Neurons are hypothesized to be particularly susceptible to the effects of weak electrophiles because the protein turnover at the distal neuron is slow, and neural function is dependent on intricate protein function (LoPachin and Gavin, 2015). Adducted and inactivated neuronal proteins may accumulate, resulting in disruption of biological function. While there are exceptions to every “rule” in toxicology, we hypothesize that weak, soft electrophiles that react slowly with soft nucleophilic protein targets are able to circulate in the body, reach the nervous system, and induce cumulative neurotoxicity; while highly reactive electrophiles would not remain in the general circulation long enough to accumulate in nerves (Figure 1) (Enoch et al., 2008b, 2009, 2010; LoPachin et al., 2012; LoPachin and Gavin, 2014; Schwӧbel et al., 2011).

Figure 1.

Figure 1.

Open in a new tab

Schematic representation of proposed major mechanistic aspects of electrophile-initiated toxic cascade. Soft electrophiles preferentially form adducts with soft nucleophilic residues, promoting inactivation of proteins that are constituents of specific soft electrophile responsive proteomes. Hard-hard covalent interactions cause cytotoxicity via disruption of discrete hard electrophile-sensitive cellular proteomes.

This study used the electrophilic index (ω) to predict possible neurotoxicity for a set of control chemicals for their production of neuropathy. In silico structural alerts for DPRA protein binding were used to identify chemicals with adduct forming potential, and the electrophilicity index was used as a quantitative reactivity measure to identify compounds with cumulative neurotoxic potential in this diverse set of control chemicals. The analysis expands the HSAB-based neurotoxicity models to a wide chemical set of regulatory importance. If HSAB-derived electrophilic index can effectively predict adduct formation leading to neurotoxicity for a diverse chemical set, then mechanistically-specific models based on HSAB theory can be used to screen large chemical libraries for potential chemicals that may cause cumulative neurotoxicity and possible common MIEs.

2. Methods

2.1. Chemical Selection and Prioritization

To test the hypothesis that electrophilicity may predict toxicant-induced cell injury, we selected 100 structurally diverse chemicals. The test set included 19 chemicals that are known to not produce neuropathies in vivo (negative controls) and 20 chemicals known to produce peripheral neuropathy in vivo via diverse mechanisms (positive controls). Table 1 lists the positive and negative control chemicals. Several of the chemicals are regulated under the Federal Insecticide, Fungicide, and Rodenticide Act, and have undergone extensive toxicological testing. In addition, we selected 61 structurally-diverse chemicals from the ToxCast Phase II library (U.S. Envionmental Protection Agency, 2015) as test substances. The substances were chosen based on structural alerts for protein binding and adduct formation from a larger set of approximately 1000 organic chemicals in ToxCast Phase II. The structural alerts were designed to identify chemical structures necessary for covalent adducts formation; i.e. reactions between a chemical and a biological target.

Table 1 –

Positive and Negative Control Chemicals

Chemical CAS # ω Control Evidence of HSAB mechanism Metabolite and derivatives ω (met) ω - group
1-bromopropane 000106–94–5 1.94 Positive Yes - - 2
3,3’-Thiodipropionic acid 000111–17–1 1.46 Negative - - - 3
3,5,6-trichloro2-pyridinol 006515–38–4 3.21 Negative - - - 1
5-Flurorouracil 000051–21–8 2.57 Positive Limited - - 2
Acetaminophen 000103–90–2 1.59 Negative - NAPQI 6.98 1
Acrylamide 000079–06–1 2.62 Positive Yes - - 2
Acrylonitrile 000107–13–1 3.08 Positive Limited - - 1
Bromomethane 000074–83–9 1.93 Positive N - - 2
Carbon Disulfide 000075–15–0 3.97 Positive N - - 1
Chlordane 000057–74–9 3.09 Negative - - - 1
chlorothalonil 001897–45–6 5.52 Negative - - - 1
Chlorpyrifos 002921–88–2 3.58 Negative - - - 1
Chlorpyrifos Oxon 005598–15–2 3.60 Negative - - - 1
Cisplatin 015663–27–1 - Positive N - - 3
Colchicine 000064–86–8 4.14 Positive N - - 1
Deltamethrin 052918–63–5 2.69 Negative - - - 2
Diamindiphenyl-sulphone 000080–08–0 2.15 Positive Limited - - 2
Dichloroacetic acid 000079–43–6 3.32 Positive N Dichloroacetic acid anion 1.93 2
Diethyl phthalate 000084–66–2 3.44 Negative - - - 1
Disulfiram 000097–77–8 2.59 Positive Limited - - 2
EPN 002104–64–5 5.65 Positive N - - 1
Ethyl butyrate 000105–54–4 1.82 Negative - - - 3
Fipronil 120068–37–3 3.63 Negative - - - 1
Fluoroacetic acid 000144–49–0 1.97 Positive N Fluoroacetic acid anion 3.33 1
Hexachlorophene 000070–30–4 2.76 Positive Limited - - 2
Hydrochlorothiazide 000058–93–5 2.92 Negative - - - 1
Indomethacin 000053–86–1 3.65 Negative - - - 1
Isoniazid 000054–85–3 3.21 Positive N - - 1
L-Ascorbic acid 000050–81–7 2.27 Negative - L-Ascorbate anion 1.07 3
Metronidazole 000443–48–1 5.40 Positive N - - 1
Mirex 002385–85–5 4.77 Negative - - - 1
n-Hexane 000110–54–3 0.81 Positive Yes 2,5-hexanedione 2.09 2
Paclitaxel 033069–62–4 3.29 Positive N - - 1
Perhexilline 006621–47–2 0.50 Positive N - - 3
Prednisone 000053–03–2 3.49 Negative - - - 1
Sodium hexametaphosphate 010124–56–8 0.99 Negative - - - 3
Spironolactone 000052–01-7 3.17 Negative - - - 1
Zinc Pyrithione 013463–41-7 3.04 Positive - - - 1
Zinc sulfate 007733–02–0 2.02 Negative - - - 2
Open in a new tab

The structural alerts for organic molecules from ToxCast phase II library were derived using Organization for Economic Co-operation and Development (OECD) QSAR toolbox (https://www.qsartoolbox.org). The specific assays included Direct Peptide Reactivity Assay (DPRA) Cysteine Peptide Depletion, DPRA Lysine Peptide Depletion, and Protein Binding Potency, and were used to select chemicals with the potential for reacting with cysteine or lysine residues. These assays identified chemicals capable of reacting with biological nucleophiles through acylation (Acyl), Michael addition (MA), Schiff base formation (SB), or nucleophilic substitution (SN1 and SN2) reactions. These mechanistic domains indicated reactive substructures that can act as electrophilic centers for nucleophilic attack (Enoch et al., 2008a). Thus, the substructures represent a binary alert for adduct formation without target preference or degree of reactivity. Chemicals with one or more protein binding alerts were then clustered based on structure. The resulting clusters were randomly sampled to obtain 61 chemical candidates. Random sampling eliminated structure selection bias and ensured a broad coverage of chemical space. These 61 chemicals were selected without consideration of existing knowledge related to neuropathic effects, and covered a broad range of structural motifs and reactivity patterns as described by OECD structural alerts and functional group searches (https://www.qsartoolbox.org). This selected process ensured that the “unknown” chemicals possessed a structural alert level and were structurally diverse.

2.2. Chemical Property Calculations

To assess the applicability of HSAB theory to identifying chemicals that induce neuropathology, 20 positive and 19 negative reference chemicals were identified. In addition, we selected 61 molecules with unknown neurotoxicity for further testing and prioritization. The 100 structures were screened for conjugate acids and bases, tautomers, and known metabolites. Fourteen derivative structures were identified and included in further analysis to capture biologically-relevant chemical species (Table S1). For all 114 structures in Table S1, ground state equilibrium geometries were calculated with Density Functional Becke, 3-parameter, Lee–Yang–Parr (B3LYP) 6–31G* in water starting from 6–31G* geometries (Kruse et al., 2012). The respective energies of the HOMO (EHOMO) and LUMO (ELUMO) were derived using Spartan 14 (version 1.1.8) software (Wavefunction Inc., Irvine CA). One positive control chemical (cisplatin), failed energy optimization and was excluded from the following analyses, leaving 19 positive and negative compounds. The Additional Global (whole molecule) HSAB parameters were then computed as follows (Equation 1):

  • hardness (η) = (ELUMO-EHOMO)/2

  • softness (σ; the inverse of hardness) = 1/ η

  • chemical potential (μ) = (ELUMO+EHOMO)/2

  • electrophilicity index (ω) = μ2/2 η

As discussed above, neurons may be susceptible to the effects of weak electrophiles that may evade biological scavengers and bind to nucleophilic protein residues. In the synaptic cleft where protein turnover is low, these adducts may accumulate, inhibiting vital protein functions and leading to peripheral neurotoxicity (LoPachin and Gavin, 2015). In several studies on specific chemical types, HSAB theory and electrophilicity index have been shown to associate with chemical’s ability to form protein adducts and induce peripheral neuropathies (LoPachin et al., 2002, 2006, 2007b, 2009; Zhang et al., 2010). The potential neurotoxicity of the 113 chemicals and their metabolites was then assessed based on the aforementioned principles and calculated electrophilicity indices. It is important to note that the electrophilicity index incorporates information about chemical hardness, softness, and chemical potential. These parameters are given in Supplementary Table 1. The chemicals were then divided into three classes: (1) chemicals expected to be too reactive to achieve sustained circulating levels, (2) chemicals which may produce chronic or cumulative neurotoxicity through the HSAB mechanism, and (3) chemicals with low reactivity which may produce toxicity through alternative mechanism(s) (Figure 2). Chemicals which may produce cumulative neurotoxicity (Group 2) were defined by the range of electrophilicity values for chemicals known to produce cumulative neurotoxicity, acrylamide (ω = 2.62) and 2,5-hexanedione (ω = 2.09) (LoPachin et al., 2003; Opanashuk et al., 2001; Zhang et al., 2010) and extended 5% beyond the range (e.g. 1.9 ≤ ω ≤ 2.8). Chemicals with electrophilicity above 2.8 were designated as more reactive (Group 1); while chemicals with ω below 1.9 were placed in group 3.

Figure 2:

Figure 2:

Open in a new tab

Electrophilic distribution for the 99 chemicals in the data set.

When analyzing tautomer’s, the more reactive tautomer was selected for further analysis. That is because from the small differences in energy between the pyridol/pyridine and ipropanone/ hydroxypropenone tautomers an enhanced electrophilicity can be expected. The equilibrium will be driven forward by the faster reaction of the product with a given nucleophile. For mesotrione, the equilibrium energy difference is too great to consider the trione structure to be a major species. When analyzing organic acids, the form prevalent at biological pH (7.4) was used for chemical classification. In the case of all three organic acids the anion form was expected to be dominant in biological conditions. Electrophilicity of known metabolites was used in place of parent compound, when literature supported metabolite toxicity, such as the case of acetaminophen and n-hexane. Chronic neurotoxicity associated with these chemicals is attributed to their reactive metabolites NAPQI and 2,5-hexanedione, respectively (LoPachin and Gavin, 2014; LoPachin et al., 2019).

Chemical grouping was blinded to positive and negative control designations. A goal of this project is to assess the predictive capability of chemical electrophilicity for cumulative neurotoxicity. The chemical grouping based on electrophilicity was then compared to the 38 positive and negative control chemicals selected for the study. For this analysis, chemicals in Groups 1 and 3 were combined as they were not suspected to induce cumulative neurotoxicity through HSAB mechanism. Since HSAB parameters are expected to predict toxicity of chemicals acting through non-specific adduct formation but not all neurotoxic substances, we collected the literature information on chemical mechanism of chronic neurotoxicity whenever possible. Positive control chemicals were divided into three groups: those with known evidence for HSAB-mediated neurotoxicity (n=3), chemicals with evidence of non-HSAB mediated neurotoxicity (n=11), and substances with limited information on neurotoxicity mechanism that may cause neurotoxicity through non-specific reactivity (n=5). The utility of HSAB parameters in predicting the neurotoxicity of chemicals in each of the three mechanistic groups was then evaluated.

We also estimated the predicted blood-brain barrier (BBB) permeability for the chemicals, which was calculated using Schrӧdinger QikProp implemented in Maestro suite (QikProp, 2014). Unfortunately, QikProp could not calculate BBB perbeability for six chemicals: 1-Dodecanethiol, 3,4-dichloro-1-butene, 6-hydroxy-2naphthyl disulfide, Benzal chloride, Deltamethrin, and Zinc sulfate (Supplementary Table 1).

2.3. Statistical Analysis

Data management and analyses were conducted in R statistical environment (R Core Team, 2014). Structure-based clustering was performed based using the ChemminR package with Tanimoto coefficient cutoff 0.7 (Rogers and Tanimoto, 1960). Association between predicted cumulative neurotoxicity and controls was determined with X2 test with boot strap implemented in R core stats package (Agresti, 2018). The effectiveness of HSAB-based screening for chronically neurotoxic compounds was assessed using sensitivity, specificity, and positive predictive rate (PPR) as defined in equations 1–3 respectively.

Sensitivity=#oftruepositives#oftruepositives+#offalsenegatives (Eq. 1)
Specificity=#oftruenegatives#oftruenegatives+#offalsepositives (Eq. 2)
PPR=#oftruepositives#oftruepositives+#offalsepositives (Eq. 3)

3. Results

The analysis ranked 99 chemicals in their biologically-relevant form based on electrophilicity (ω) quantum-mechanic property. The chemicals were then divided into three groups as discussed in the methods section. Group 1 contained proposed reactive chemicals. Group 2 included chemicals which may produce cumulative neurotoxicity. Chemicals with low reactivity that may not induce toxicity through HSAB mechanism were placed in Group3. Table S1 provides the details on chemical classifications.

After grouping molecules based on their electrophilicity, we evaluated the ability of this grouping to predict the cumulative neurotoxicity of the 19 positive and 19 negative control chemicals. Eleven reference chemicals were identified as potentially producing cumulative neurotoxicity based on the HSBA parameters alone (Tables 1, 2). Of the eleven, 9 were known to cause cumulative neurotoxicity, leading to a positive prediction rate (PPR) of 82%. Similarly, of the 19 negative controls, 17 were identified as not expected to induce cumulative neurotoxicity based on the HSAB parameters, equating to 90% specificity. The two negative controls suspected of chemical toxicity were examined further and found to be zinc sulfate (CAS #7733–02-0) and deltamethrin (CAS #52918–63-5). Zinc sulfate is a metal salt. The standard electrophilicity calculations based on B3LYP functional may not be appropriate to estimate metal ion reactivity parameters. More generally, metal ions are not in the applicability domain of electrophilicity model defined by electrophilicity ranges of organic compounds. When, zinc sulfate was removed from the data set, PPR increases to 90%, and specificity to 94%. Deltamethrin has an electrophilicity index (ω=2.69) on the edge of the range we defined as potentially neurotoxic. While the electrophilicity of deltamethrin is above that of known neurotoxin acrylamide (ω = 2.62), the difference is small. It is unknown how such an electrophilicity index may relate to other forms of toxicity produced by deltamethrin. It has been reported that deltamethrin, has the potential to induce oxidative stress and may be too reactive to reach the synapse and induce cumulative neurotoxicity (Kumar et al., 2015). Consequently, further research may be necessary to better define the boundaries of electrophilicity related to chronic neurotoxicity through a HSAB mechanism. Additionally, deltamethrin is known to induce neurotoxicity by blocking sodium ion channels (World Health Organization, 1990). This mechanism of action is not expected to be captured by HSAB-based model. Finally, both zinc sulfate and deltamethrin are relatively complex molecules, at least from the perspective of electron behavior. Consequently, the uncertainty in ω estimates may be larger when ω is calculated at the B3LYP level of theory and additional theory levels may be considered in the future.

Table 2:

Cumulative neurotoxicity predictions contingency tables

Known to cause cumulative neurotoxicity Not cumulatively neurotoxic
May produce chronic or cumulative neurotoxicity through HSAB mechanism (Group 2) 9 2
Not expected to produce chronic or cumulative neurotoxicity through HSAB mechanism (Groups 1 and 3) 10 17
Open in a new tab

Overall, the electrophilicity-based chemical classification into group 2 was significantly associated with chemicals known to cause chemical neurotoxicity (p=0.013 χ2 test with boot strap with 10,000 replicates). Electrophilicity index classification identified 9/19 positive control chemicals as potentially producing neurotoxicity through a mechanism consistent with HSAB theory, a sensitivity of 47%. However, the sensitivity for a random set of chemicals was not expected to approach 100%. The HSAB screening approach is based on the premise that electrophilic chemicals are toxic because they form covalent adducts with biological nucleophiles, and consequently deactivate macromolecules that are important in biological processes. The electrophilicity screening approach was intended to screen for chemicals predicted to induce chronic neurotoxicity through HSAB mechanism and was unlikely to identify substances that act through different mechanisms. Since non-specific adduct formation was not one of the selection criteria for positive controls, we did not expect the HSAB-based classification to identify all positive control chemicals and cumulatively neurotoxic.

To further investigate the relevance of electrophilicity ranking in detecting chemicals that induce chronic neurotoxicity through HSAB-related mechanisms, we reviewed the literature evidence on mechanisms of actions for the 19 positive control chemicals and categorized them into three groups as discussed in the methods section. Three chemicals (acrylamide, 1-bromopropane, and n-hexane) or their metabolites were known (or strongly suspected) to induce chronic neurotoxicity through HSAB mechanism (Table 1). All three of these chemicals were identified as suspected chronic neurotoxicants (Group 2) by the electrophilicity screen. Five chemicals (disulfiram, acrylonitrile, hexachlorophene, 5- fluorouracil, and diaminodiphenylsulphone) had limited evidence of HSAB mechanism in neurotoxicity (Table 1). Four of these five chemicals were identified as suspected chronic neurotoxicants (Group 2) by electrophilicity screen. The electrophilicity screening approach correctly identified 7/8 substances thought to induce neurotoxicity. For chemicals expected to induce neurotoxicity non-HSAB-mediated mechanisms, electrophilicity screening predicted only 2/11 to be positive. Consequently, the electrophilicity screening approach presented a robust method for identifying chemicals that may induce neurotoxicity through non-specific adduct formation. However, it could not identify all chemicals capable of inducing neurotoxicity through a diverse set of toxicity mechanisms.

The study included 61 chemicals with structural alerts for protein binding (Table S1), selected without consideration of any known in vivo toxicities. Of these, 22 were classified as potential chronic neurotoxins through HSAB-related mechanisms based on electrophilicity index (Group 2). Another 33 compounds were considered too reactive (Group 1), and the remaining six compounds were classified as low reactivity (Group 3). The overall distribution of electrophilicity for all 99 chemicals with HSAB calculations in this study is shown in Figure 2.

Finally, we evaluated the influence of BBB permeability (calculated in QikProp, 2014), on the likelihood that a chemical may induce neurotoxicity. Threshold assessment based on the known peripheral neurotoxicants was not effective because the predicted BBB permeability of acrylamide was very low (−2.323 on the log scale), the third lowest among all chemicals. The low predicted BBB permeability is in contrast with pathological and proteomic studies that show central neurotoxicity after treatment with acrylamide (Barber et al., 2007; LoPachin et al., 2003). Additionally, the calculated BBB permeability did not help to distinguish positive from negative control compounds (t-test p. value = 0.879). Since known cumulative neurotoxicants, such as acrylamide, have very low predicted BBB permeability, the applicability of this BBB model to predict neurotoxicity may be limited. Therefore, we did not make any further adjustments to the predictions based on BBB permeability.

4. Discussion

In this work, we evaluated the utility of HSAB-based classification in predicting cumulative neurotoxicity based on 19 positive and 19 negative control substances. The approach demonstrated high PPR (82–90%) and specificity (90%). However, the overall sensitivity for all positive control compounds was lower (47%). The lower sensitivity was expected as HSAB-based classifications were expected to identify chemicals acting via weak (e.g. non-covalent), non-specific adduct formation. The screening strategy was not expected to predict neurotoxicity’s resulting from other toxic mechanisms. When predicting the toxicity of chemicals known or suspected of acting via non-specific adduct formation mechanism (Table 1), the HSAB approach identified 7/8 (88%) of positive control chemicals and cumulatively neurotoxic chemicals. On the other hand, the HSAB-based classification flagged only 2/11 (18%) positive control chemicals that were expected to cause neurotoxicity through other mechanisms as potentially neurotoxic. Consequently, the HSAB-based classification approach was found useful for screening for cumulatively neurotoxic chemicals that act via non-specific protein adduct formation.

It is important to note, that the neurotoxicity predictions produced in the study must be assessed within the framework of multiple chemical parameters and technical limitations with the overall goal of achieving a better understanding of neurotoxicity MIEs. Neurotoxicity can be a function of multiple biological processes, with non-specific reactivity and adduct formation representing a single type of cumulative neurotoxicity (LoPachin and Gavin, 2015; LoPachin et al., 2019; Soffietti et al., 2014). The chemical electrophilicity index derived from the HSAB theory has been shown as an effective way to model neurotoxicity for diverse chemicals with a reactivity-based mechanism of action (LoPachin and Gavin, 2012; LoPachin et al., 2012). However, electrophilicity was not expected to be the sole parameter which determines chronic neurotoxicity for all substances. Multiple factors influence the chemical’s bioavailability, protein adduct strength, location of adduct formation, and thus, eventual neurotoxicity potential. Structural and steric factors such as functional group character (e.g., amine, alkyl halide etc.) may enable or obstruct adduct formation and chronic neurotoxicity. It must be noted that such steric factors depend on the angle of attack to the endogenous nucleophile (which is dependent on the type of chemical reaction), the distances between the adjacent centers, and the size/volume/contour of both the electrophile and the nucleophile.

In addition, some chemical structures may produce reversible adduct formation, which does not accumulate sufficiently to lead to neurotoxicity. Further research may be necessary to elucidate computable and useful predictors for these biological processes. Finally, the approach has computational limitations. Electrophilicity can be calculated at different levels of theory with varying complexity. Calculation methods should be chosen to account for the types of structures in the data set and special attention should be given to compounds containing metals and ions. More generally, chemical reactivity with biological targets can be further approximated by the second order rate constants of unhindered addition to proteins. While the reactivity is more difficult to calculate for a large chemical set, the calculations of 2nd rate constants for reactivity of environmental chemicals with model biological nucleophiles may be explored as a method to improve toxicity predictions.

Despite these limitations, the analysis highlights the utility of using structural alerts for protein binding and chemical properties together to prioritize chemical screening. Recently, multiple authors have reported using chemical structural alerts to link with various biological pathways (Allen et al., 2018; Casalegno and Sello, 2013; Garcia-Serna et al., 2015; Judson et al., 2016; Shah et al., 2016). While each of these reports uses some form of chemical alert to form groups of substances related to some form of toxicity, none of these efforts has attempted to predict neurotoxicity resulting from electrophilic reactions that likely require repeated exposures to manifest in observable neurotoxicity. The mixed approach that we utilized combines structural and property-considerations and has been shown to work for select chemical groups (LoPachin and Gavin, 2012; LoPachin et al., 2012). This project expanded the approach to a diverse chemical set and aimed to further broaden the applicability of HSAB-based neurotoxicity assessment in future work. This research successfully expanded the chemical space tested by the HSAB hypothesis. Ranking diverse chemicals by computed values of electrophilicity provides a reasonable measure for predicting the toxicity of chemicals. Gross division of electrophilic substances into strong, weak and non-electrophiles provides the initial distinction between those chemicals likely to cause systemic organ toxicity (strong electrophiles), those capable of cumulative toxicity (weak electrophiles) and those expected to be non-toxic and non-electrophilic.

In addition, the mechanistic nature of HSAB-based classification provides certain advantages in chemical screening and assessment. Chemicals identified as potentially neurotoxic via electrophilicity classification are likely to act via non-specific reactivity mechanism. The insight may be combined with adverse outcome pathway (AOP) framework to help identify key events in neurotoxicity and other toxicological pathway(s). Thus, an additional outcome of HSAB-based screening may be the demonstration of how a common MIE can be involved in multiple types of toxicity. While the study focuses on the analysis of cumulative neurotoxicity, the electrophilic classification and HSAB-based screening should be more broadly applicable to outcomes that evolve covalent reactions (e.g. pulmonary toxicity, skin sensitization) (Divkovic et al., 2005; Parkinson et al., 2017; Seed et al., 2008). Other toxicological endpoints with MIEs that involve covalent interactions between a chemical and a biological target can be modeled with HSAB theory. For example, electrophiles, such as α,β-unsaturated ketones with soft polarizable centers, can favorably bind anionic thiolates on critical proteins, deplete the cellular antioxidant defense system, and induce cytotoxicity. (Enoch et al., 2008a; LoPachin et al., 2007b; Martyniuk et al., 2011; Melnikov et al., 2019; Schultz et al., 2005; Shen et al., 2016a, 2016b). Similarly, chemical interactions with reactive biological residues have been implicated in aquatic toxicity, skin sensitization, respiratory toxicity, DNA damage, and hepatotoxicity; experimental or computationally-derived reactivity parameters have been shown to improve toxicity prediction for these endpoints (Connors et al., 2014; Melnikov et al., 2016; Schultz et al., 2006; Schultz et al., 2009; Schwöbel et al., 2011). More specific sub-cellular processes are mediated by covalent interaction MIEs because many proteins use electrophilic residues as environmental sensors that may interact with signaling molecules such as CO2 and N2 or H2O2 (Schopfer et al., 2011; Kumagai and Abiko, 2017). Well-established examples include GAPDH and NRF2 proteins that depend on reactive residues for metabolic function or transcriptional regulations (LoPachin et al., 2019; Ma and He, 2012; Mohr et al., 1994). These processes and resulting adverse outcomes can be modeled with HSAB theory if data on specific biological events are available. Together, the HSAB-based models combined with structural reactivity alerts can help identify chemicals more likely to induce toxicity through covalent binding MIEs. With sufficient in vitro data, the MIE can be further mapped to adverse effects via AOP frameworks, thus moving toxicity assessment toward more robust in vitro & in silico screening strategy for the 21st Century (Andersen and Krewski, 2009; Juberg et al., 2017; Knudsen et al., 2015; National Research Council, 2007; Thomas et al., 2019).

Our analysis further highlights the importance of considering biologically-relevant chemical forms and metabolites in electrophilicity assessment. For example, acetaminophen and n-hexane, are both known chronic toxicants with a HSAB mechanism of action but are not considered to be electrophilic. However, both molecules are metabolically activated, producing moderately electrophilic NAPQI and 2,5-hexanedione (Table 1), which are responsible for their toxicity. In addition, the considerations for acid-base equilibrium at biological pH was critical to grouping organic acids as their electrophilicity indices change substantially from ionized to protonated forms (Table 1). Consequently, it is vital to consider the electrophilicity of known metabolites and biologically-relevant chemical forms to capture chemical toxicity.

This work summarizes a starting point for an integrated, tiered neurotoxicity assessment strategy. Selected chemicals from this list will be evaluated in vitro and in vivo assays. In vitro screening techniques include neurite growth and multielectrode array electrophysiology (Johnstone et al., Submitted). In vivo methods include behavioral analysis and neurophysiological responses from peripheral nerves and the somatosensory system. The projects will commence with positive control chemicals and then expand into the 22 chemicals with unknown neurotoxicity that were identified as potentially neurotoxic based on their electrophilic index. Thus, this approach provides an integrated in silico- in vitro – in vivo testing paradigm for neurotoxicity assessment. This represents a necessary transition to a mechanistic toxicity assessment that is both cost- and time-efficient when compared to animal testing in toxicity screening. In addition, HSAB parameters can be used to model other toxic outcomes with contributions from the adduct-binding MIE and may have important implications for chemical screening and risk assessment (Schwöbel et al., 2011). Since predictions of neurotoxicity based on HSAB theory are quantitative and amendable to empirical testing in laboratory settings, the models can help prioritize chemicals for in vitro and in vivo testing. HSAB neurotoxicity predictions and deviations from predictions can help discern chemical mechanisms of action that cannot be easily identified with laboratory tests alone. While chemical bioavailability in the neural cleft is a concern, cumulative neurotoxicity is a slow process, and chemicals with low predicted BBB permeability (such as acrylamide) can be potent neurotoxicants, when exposed occurs over longer periods of time. Thus, existing BBB models which only consider “instantaneous” permeability may have limited applicability to predict cumulative neurotoxicity. However, access to neurons will be modulated by permeability factors, and a low permeability is not inconsistent with the proposed mechanism of cumulative neurotoxicity for weak electrophiles, since these chemicals produce toxic effects over a longer time period.

Finally, the neurotoxicity outcomes are expected to exhibit some variability of results in different test methods. As discussed above, differences in available binding sites and chemical bioavailability in different test systems may influence in vivo and in vitro results and the concordance of the results. Consequently, the positive and negative control chemicals selected based on in vivo data are not guaranteed to produce the same response in alternative (in vitro) test paradigms.

5. Conclusions

Cumulative neurotoxicity is a vital human toxicity endpoint that is difficult to assess or predict due to the slow physiological changes that may lead to the adverse outcomes after prolonged chemical exposure. Recent insights into the electrophile-nucleophile binding mechanisms behind cumulative neurotoxicity demonstrated HSAB theory to be a useful tool for chemical screening and triage. Here we presented a novel approach using HSAB-derived electrophilic index and structural alerts for covalent binding to assess the potential of chemicals to induce cumulative neurotoxicity. Chemicals with soft electrophilic centers of moderate reactivity may reach neural targets and impair normal function by binding targets at the synapse which cannot be repaired quickly; leading to neurotoxicity (Barber and LoPachin, 2004; Barber et al., 2007; LoPachin and Gavin, 2012; LoPachin et al., 2019). Our analysis indicated that the screening approach can identify reference chemicals known to induce neurotoxicity through electrophilic binding. However, the approach did not identify all neurotoxic chemicals, since it was designed to identify only chemicals acting through electrophilic binding MIE. Consequently, like many other models, the approach needs to be expanded to account for additional MIEs leading to neurotoxicity. At the same time, the mechanistic nature of the screen enhanced confidence in toxicity prediction and could be used to flag chemicals for potential electrophilic binding MIEs, when the MIE is not known. Furthermore, since HSAB parameters can be used to distinguish different types of adduct formation, similar mechanistic screening approaches based on HSAB theory could be applied to other toxicological endpoints that are initiated by the electrophilic binding of xenobiotics. This work expands the predictive capabilities of the screening process for multiple types of toxicity at early stages of risk evaluation and presents the first stage of an integrated, mechanistic assessment for cumulative neurotoxicity that can reduce the extent of animal testing. In this manuscript we focused on evaluating HSAB-based in silico screening protocol for neurotoxicity on a diverse set of xenobiotics. The next stages will integrate the in silico triage approach with in vitro screening data to expand the model to a broader set of chemicals. Selected xenobiotics should then be tested in vivo to verify the predictions and establish greater confidence in the general approach. While the integrated in vitro-in silico approaches build confidence in model predictions, HSAB-based screening expands and expedites chemical screening, particularly to solvents, insoluble chemicals, and other types of compounds that are not readily tested in vitro.

Supplementary Material

Supplemental Table 1

ACKNOWLEDGEMENTS

The authors would like to thank Drs. William Boyes and Witold Winnik for helpful discussions and critique of the manuscript. Discussions with Drs. Richard Judson, Virginia Moser, and Karl Jensen helped formulate the chemicals considered for investigation. The authors dedicate this manuscript to the memory of Dr. Terrence Gavin.

FUNDING

The research discussed in this manuscript was supported by EPA Contract # EP-15-D-000123 to Iona College (Terrance Gavin) and by NIH grants from the National Institutes of Environmental Health Sciences RO1 ES03830-30; RO1 ESO7912-11 (Richard LoPachin). Fjodor Melnikov would like to thank the U.S. EPA Science to Achieve Results (STAR) program for financial support (FP-91779301-0).

Non-Standard Abbreviations

H2O2

hydrogen peroxide

NAPQI

N-acetyl-p-benzoquinonimine

HSAB

Hard and Soft, Acids and Bases

ELUMO

Lowest Unoccupied Molecular Orbital energy

EHOMO

Highest Occupied Molecular Orbital energy

Footnotes

The research described in this article has been reviewed by the National Health and Environmental Effects Research Laboratory, U.S. Environmental Protection Agency, and approved for publication. Approval does not signify that the contents necessarily reflect the views and policies of the Agency, nor does mention of trade names or commercial products constitute endorsement or recommendation for use.

REFERENCES

  1. Agresti A, 2018. An introduction to categorical data analysis. Wiley. [Google Scholar]
  2. Allen TE, Goodman JM, Gutsell S, Russell PJ, 2018. Using 2D Structural Alerts to Define Chemical Categories for Molecular Initiating Events. Toxicological Sciences 165(1), 213–223 [DOI] [PubMed] [Google Scholar]
  3. Alves VM, Muratov E, Fourches D, Strickland J, Kleinstreuer N, Andrade CH, Tropsha A, 2015. Predicting chemically-induced skin reactions. Part I: QSAR models of skin sensitization and their application to identify potentially hazardous compounds. Toxicology and Applied Pharmacology 284(2), 262–272. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Andersen ME, Krewski D, 2009. Toxicity testing in the 21st century: Bringing the vision to life. Toxicological Sciences 107, 324–330. [DOI] [PubMed] [Google Scholar]
  5. Barber DS, LoPachin RM, 2004. Proteomic analysis of acrylamide-protein adduct formation in rat brain synaptosomes. Toxicology and Applied Pharmacology 201(2), 120–136. [DOI] [PubMed] [Google Scholar]
  6. Barber DS, Stevens S, LoPachin RM, 2007. Proteomic analysis of rat striatal synaptosomes during acrylamide intoxication at a low dose rate. Toxicological Sciences 100(1), 156–167. [DOI] [PubMed] [Google Scholar]
  7. Casalegno M, Sello G, 2013. Determination of toxicant mode of action by augmented top priority fragment class. Journal of Chemical Information and Modeling 53(5), 1113–1126. [DOI] [PubMed] [Google Scholar]
  8. Coish P, Brooks BW, Gallagher EP, Mills M, Kavanagh TJ, Simcox N, Lasker GA, Botta D, Schmuck SC, Voutchkova-Kostal A, 2017. The molecular design research network. Toxicological Sciences 161(2), 241–248. [DOI] [PubMed] [Google Scholar]
  9. Connors KA, Voutchkova-Kostal AM, Kostal J, Anastas P, Zimmerman JB, Brooks BW, 2014. Reducing aquatic hazards of industrial chemicals: Probabilistic assessment of sustainable molecular design guidelines. Environmental Toxicology and Chemistry 33, 1894–1902. [DOI] [PubMed] [Google Scholar]
  10. Divkovic M, Pease CK, Gerberick GF, Basketter DA, 2005. Hapten–protein binding: from theory to practical application in the in vitro prediction of skin sensitization. Contact Dermatitis 53(4), 189–200. [DOI] [PubMed] [Google Scholar]
  11. Enoch SJ, Madden JC, Cronin MTD, 2008a. Identification of mechanisms of toxic action for skin sensitisation using a SMARTS pattern based approach. SAR and QSAR in Environmental Research 19, 555–578. [DOI] [PubMed] [Google Scholar]
  12. Enoch SJ, Cronin TD, Schultz TW, Madden JC, 2008b. Quantitative and mechanistic read across for predicting the skin sensitization potential of alkenes acting via michael addition. Chemical Research in Toxicology 21(2), 513–520. [DOI] [PubMed] [Google Scholar]
  13. Enoch SJ, Roberts DW, Cronin MTD, 2009. Electrophilic reaction chemistry of low molecular weight respiratory sensitizers. Chemical Research in Toxicology 22(8), 1447–1455. [DOI] [PubMed] [Google Scholar]
  14. Enoch SJ, Roberts DW, Cronin MTD, 2010. Mechanistic category formation for the prediction of respiratory sensitization. Chemical Research in Toxicology 23(10), 1547–1555. [DOI] [PubMed] [Google Scholar]
  15. Garcia-Serna R, Vidal D, Remez N, Mestres J, 2015. Large-scale predictive drug safety: from structural alerts to biological mechanisms. Chemical Research in Toxicology 28(10), 1875–1887. [DOI] [PubMed] [Google Scholar]
  16. Juberg DR, Knudsen TB, Sander M, Beck NB, Faustman EM, Mendrick DL, Fowle JR, Hartung T, Tice RR, Lemazurier E, Becker RA, Fitzpatrick SC, Daston GP, Harrill A, Hines RN, Keller DA, Lipscomb JC, Watson D, Bahadori T, Crofton KM, 2017. FutureTox III: Bridges for Translation. Toxicological Sciences 155, 22–31. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Judson R, Houck K, Martin M, Richard AM, Knudsen TB, Shah I, Little S, Wambaugh J, Woodrow Setzer R, Kothya P, 2016. Editor’s Highlight: analysis of the effects of cell stress and cytotoxicity on in vitro assay activity across a diverse chemical and assay space. Toxicological Sciences 152(2), 323–339. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Knudsen TB, Keller D.a., Sander M, Carney EW, Doerrer NG, Eaton DL, Fitzpatrick SC, Hastings KL, Mendrick DL, Tice RR, Watkins PB, Whelan M, 2015. FutureTox II: In vitro Data and In Silico Models for Predictive Toxicology. Toxicological Sciences 143, 256–267. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Krimsky S, 2017. The unsteady state and inertia of chemical regulation under the US Toxic Substances Control Act. PLoS biology 15(12), e2002404. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Kruse H, Goerigk L, Grimme S, 2012. Why the Standard B3LYP/6–31G* Model Chemistry Should Not Be Used in DFT Calculations of Molecular Thermochemistry: Understanding and Correcting the Problem. The Journal of Organic Chemistry 77(23), 10824–10834. [DOI] [PubMed] [Google Scholar]
  21. Kumagai Y, Abiko Y, 2017. Environmental electrophiles: Protein adducts, modulation of redox signaling, and interaction with persulfides/polysulfides. Chemical Research in Toxicology 30, 203–219. [DOI] [PubMed] [Google Scholar]
  22. Kumar A, Bhaskar A, Chandra S, Sasmal D, Mukhopadhyay KM, Sharma N, 2015. Mechanism of deltamethrin induced immunotoxicity: Current and future perspectives. Receptors & Clinical Investigation 2(2). [Google Scholar]
  23. Lalko JF, Kimber I, Gerberick GF, Foertsch LM, Api AM, Dearman RJ, 2012. The direct peptide reactivity assay: selectivity of chemical respiratory allergens. Toxicological Sciences 129(2), 421–431. [DOI] [PubMed] [Google Scholar]
  24. LoPachin RM, Balaban C, Ross J, 2003. Acrylamide axonopathy revisited. Toxicology and Applied Pharmacology 188(3), 135–153. [DOI] [PubMed] [Google Scholar]
  25. LoPachin RM, Barber DS, Gavin T, 2007a. Molecular mechanisms of the conjugated α, β-unsaturated carbonyl derivatives: relevance to neurotoxicity and neurodegenerative diseases. Toxicological Sciences 104(2), 235–249. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. LoPachin RM, Barber DS, Geohagen BC, Gavin T, He D, Das S, 2006. Structure-toxicity analysis of type-2 alkenes: in vitro neurotoxicity. Toxicological Sciences 95(1), 136–146. [DOI] [PubMed] [Google Scholar]
  27. LoPachin RM, DeCaprio AP, 2005. Protein adduct formation as a molecular mechanism in neurotoxicity. Toxicological Sciences 86(2), 214–225. [DOI] [PubMed] [Google Scholar]
  28. LoPachin RM, Gavin T, 2012. Molecular mechanism of acrylamide neurotoxicity: lessons learned from organic chemistry. Environmental Health Perspectives 120(12), 1650–1657. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. LoPachin RM, Gavin T, 2014. Molecular mechanisms of aldehyde toxicity: a chemical perspective. Chemical Research in Toxicology 27(7), 1081–1091. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. LoPachin RM, Gavin T, 2015. Toxic neuropathies: Mechanistic insights based on a chemical perspective. Neuroscience Letters 596, 78–83. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. LoPachin RM, Balaban CD, Ross JF, 2003. Acrylamide axonopathy revisited. Toxicology Applied Pharmacology 188, 135–153. [DOI] [PubMed] [Google Scholar]
  32. LoPachin RM, Gavin T, DeCaprio A, Barber DS, 2012. Application of the hard and soft, acids and bases (HSAB) theory to toxicant–target interactions. Chemical Research in Toxicology 25(2), 239–251. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. LoPachin RM, Gavin T, Geohagen BC, Das S, 2007. Neurotoxic mechanisms of electrophilic type-2 alkenes: soft–soft interactions described by quantum mechanical parameters. Toxicological Sciences 98(2), 561–570. [DOI] [PubMed] [Google Scholar]
  34. LoPachin RM, Gavin T, Petersen DR, Barber DS, 2009. Molecular mechanisms of 4-hydroxy-2-nonenal and acrolein toxicity: nucleophilic targets and adduct formation. Chemical Research in Toxicology 22(9), 1499–1508. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. LoPachin RM, Geohagen BC, Nordstroem LU, 2019. Mechanisms of soft and hard electrophile toxicities. Toxicology 418, 62–69. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. LoPachin RM, Geohagen BC, Nordstrøm LU, Gavin T, 2016. Enolate-Forming Compounds as a Novel Approach to Cytoprotection. Chemical Research in Toxicology 29, 2096–2107. [DOI] [PubMed] [Google Scholar]
  37. LoPachin RM, Ross J, Reid ML, Das S, Mansukhani S, Lehning E, 2002. Neurological evaluation of toxic axonopathies in rats: acrylamide and 2, 5-hexanedione. Neurotoxicology 23(1), 95–110. [DOI] [PubMed] [Google Scholar]
  38. Ma Q, He X, 2012. Molecular Basis of Electrophilic and Oxidative Defense: Promises and Perils of Nrf2. Pharmacological Reviews 64, 1055–1081. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Martyniuk C, Fang B, Koomen J, Gavin T, LoPachin R, Barber D, 2011. Molecular mechanisms of α, β-unsaturated carbonyl toxicity: cysteine-adduct formation correlates with loss of enzyme function. Chemical Research in Toxicology 24, 2302–2311. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Melnikov F, Botta D, White CC, Schmuck SC, Winfough M, Schaupp CM, Gallagher EP, Brooks BW, Williams ES, Coish P, Anastas PT, Voutchkova-Kostal A, Kostal J, Kavanagh TJ, 2019. Kinetics of Glutathione Depletion and Antioxidant Gene Expression as Indicators of Chemical Modes of Action Assessed in Vitro in Mouse Hepatocytes with Enhanced Glutathione Synthesis. Chemical Research in Toxicology, acs.chemrestox.8b00259. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Melnikov F, Kostal J, Voutchkova-Kostal A, Zimmerman JB, Anastas T, P., 2016. Assessment of predictive models for estimating the acute aquatic toxicity of organic chemicals. Green Chemistry 18, 4432–4445. [Google Scholar]
  42. Moffit JS, Bryant BH, Hall SJ, Boekelheide K, 2007. Dose-dependent effects of sertoli cell toxicants 2, 5-hexanedione, carbendazim, and mono-(2-ethylhexyl) phthalate in adult rat testis. Toxicologic Pathology 35(5), 719–727. [DOI] [PubMed] [Google Scholar]
  43. Mohr S, Stamler JS, Brune B, 1994. Mechanism of covalent modification of glyceraldehyde-3-phosphate dehydrogenase at its active site thiol by nitric oxide, peroxynitrite and related nitrosating agents. FEBS letters 348, 223–227. [DOI] [PubMed] [Google Scholar]
  44. National Research Council, 2007. Toxicity testing in the 21st century: a vision and a strategy. National Academies Press. [Google Scholar]
  45. Opanashuk LA, He DK, Lehning EJ, LoPachin RM, 2001. Gamma-diketone peripheral neuropathy III. Neurofilament gene expression. Neurotoxicology 22, 215–220. [DOI] [PubMed] [Google Scholar]
  46. Parkinson E, Aleksic M, Cubberley R, Kaur-Atwal G, Vissers JP, Skipp P, 2017. Determination of Protein Haptenation by Chemical Sensitizers Within the Complexity of the Human Skin Proteome. Toxicological Sciences 162(2), 429–438. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Pearson RG, 1963. Hard and soft acids and bases. Journal of the American Chemical society 85(22), 3533–3539. [Google Scholar]
  48. Pearson RG, 2005. Chemical hardness and density functional theory. Journal of Chemical Sciences 117(5), 369–377. [Google Scholar]
  49. Pool R, Rusch E, 2014. Identifying and reducing environmental health risks of chemicals in our society. Workshop Summary, Identifying and reducing environmental health risks of chemicals in our society Workshop Summary. National Academies Press. [PubMed] [Google Scholar]
  50. QikProp in Maestro Varsion 12.1.013, MMshare Version 4.7.013, Release 2019–3, Schrӧdinger, LLC, New York, NY, 2014. [Google Scholar]
  51. R Core Team, 2014. R: A language and environment for statistical computing. R Foundtion for Statistical Computing, Vienna, Austria. [Google Scholar]
  52. Rogers DJ, Tanimoto TT, 1960. A Computer Program for Classifying Plants. Science 132, 1115 LP–1118. [DOI] [PubMed] [Google Scholar]
  53. Schopfer FJ, Cipollina C, Freeman BA, 2011. Fromation and signaling actions of electrophilic lipids. Chemical Reviews 111(10), 5997–6021. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Schultz T, Carlson R, Cronin M, Hermens J, Johnson R, O’Brien P, Roberts D, Siraki A, Wallace K, Veith G, 2006. A conceptual framework for predicting the toxicity of reactive chemicals: modeling soft electrophilicity. SAR and QSAR in Environmental Research 17(4), 413–428. [DOI] [PubMed] [Google Scholar]
  55. Schultz TW, Rogers K, Aptula AO, 2009. Read-across to rank skin sensitization potential: subcategories for the Michael acceptor domain. Contact Dermatitis 60, 21–31. [DOI] [PubMed] [Google Scholar]
  56. Schultz TW, Yarbrough JW, Johnson EL, 2005. Structure-activity relationships for reactivity of carbonyl-containing compounds with glutathione. SAR and QSAR in Environmental Research 16, 313–322. [DOI] [PubMed] [Google Scholar]
  57. Schwö JA, Koleva YK, Enoch SJ, Bajot F, Hewitt M, Madden JC, Roberts DW, Schultz TW, Cronin MT, 2011. Measurement and estimation of electrophilic reactivity for predictive toxicology. Chemical Reviews 111(4), 2562–2596. [DOI] [PubMed] [Google Scholar]
  58. Seed MJ, Cullinan P, Agius RM, 2008. Methods for the prediction of low-molecular-weight occupational respiratory sensitizers. Current Opinion in Allergy and Clinical Immunology 8(2), 103–109. [DOI] [PubMed] [Google Scholar]
  59. Shah F, Greene N, 2013. Analysis of Pfizer compounds in EPA’s ToxCast chemicals-assay space. Chemical Research in Toxicology 27(1), 86–98. [DOI] [PubMed] [Google Scholar]
  60. Shah I, Liu J, Judson RS, Thomas RS, Patlewicz G, 2016. Systematically evaluating read-across prediction and performance using a local validity approach characterized by chemical structure and bioactivity information. Regulatory Toxicology and Pharmacology 79, 12–24. [DOI] [PubMed] [Google Scholar]
  61. Shen LQ, Judson RS, Melnikov F, Roethle J, Gudibanda A, Zimmerman JB, Anastas PT, 2016a. Probabilistic diagram for designing chemicals with reduced potency to incur cytotoxicity. Green Chemistry 18, 4461–4467. [Google Scholar]
  62. Shen LQ, Melnikov F, Roethle J, Gudibanda A, Judson RS, Zimmerman JB, Anastas PT, 2016b. Coupled molecular design diagrams to guide safer chemical design with reduced likelihood of perturbing the NRF2-ARE antioxidant pathway and inducing cytotoxicity. Green Chemistry 18, 6387–6394. [Google Scholar]
  63. Silva M, Pham N, Lewis C, Iyer S, Kwok E, Solomon G, Zeise L, 2015. A comparison of ToxCast test results with in vivo and other in vitro endpoints for neuro, endocrine, and developmental toxicities: a case study using endosulfan and methidathion. Birth Defects Research Part B: Developmental and Reproductive Toxicology 104(2), 71–89. [DOI] [PubMed] [Google Scholar]
  64. Sirenko O, Parham F, Dea S, Sodhi N, Biesmans S, Mora-Castilla S, Ryan K, Behl M, Chandy G, Crittenden C, Vargas-Hurlston S, Guicherit O, Gordon R, Zanella F, Carromeu C, 2019. Functional and mechanistic neurotoxicity profiling using human iPSC-derived neural 3D cultures. Tixicological Sciences 167(1), 58–76. [DOI] [PMC free article] [PubMed] [Google Scholar]
  65. Soffietti R, Trevisan E, Rudà R, 2014. Neurologic complications of chemotherapy and other newer and experimental approaches, Handbook of Clinical Neurology. Elsevier, pp. 1199–1218. [DOI] [PubMed] [Google Scholar]
  66. Thomas RS, Bahadori T, Buckley TJ, Cowden J, Deisenroth C, Dionisio KL, Frithsen JB, Grulke CM, Gwinn MR, Harrill JA, Higuchi M, Houck KA, Hughes MF, Hunter ES, Isaacs KK, Judson RS, Knudsen TB, Lambert JC, Linnenbrink M, Martin TM, Newton SR, Padilla S, Patlewicz G, Paul-Friedman K, Phillips KA, Richard AM, Sams R, Shafer TJ, Setzer RW, Shah I, Simmons JE, Simmons SO, Singh A, Sobus JR, Strynar M, Swank A, Tornero-Valez R, Ulrich EM, Villeneuve DL, Wambaugh JF, Wetmore BA, Williams AJ, 2019. The Next Generation Blueprint of Computational Toxicology at the U.S. Environmental Protection Agency. Toxicological Sciences 169, 317–332. [DOI] [PMC free article] [PubMed] [Google Scholar]
  67. Thomas RS, Philbert MA, Auerbach SS, Wetmore BA, DeVito MJ, Cote I, Rowlands JC, Whelan MP, Hays SM, Andersen ME, Meek ME, Reiter LW, Lambert JC, Clewell HJI, Stephens ML, Zhao QJ, Wesselkamper SC, Flowers L, Carney EW, Pastoor TP, Petersen DD, Yauk CL, Nong A, 2013. Incorporating new technologies into toxicity testing and risk assessment: Moving from 21st century vision to a data-driven framework. Toxicological Sciences 136(1), 4–18. [DOI] [PMC free article] [PubMed] [Google Scholar]
  68. U.S. Envionmental Protection Agency, 2015. ToxCast & Tox21 Summary Files from invitrodb_v2. Retrieved from Data released October 2015. http://www2.epa.gov/chemical-research/toxicity-forecaster-toxcasttm-data (Accessed on November 25, 2015).
  69. U.S. GAO, 2005. Toxic Substances Control Act: Legislative Changes Could Make the Act More Effective. US Government Accountability Office Wash, DC. [Google Scholar]
  70. Wilson MP, Schwarzman MR, 2009. Toward a new US chemicals policy: rebuilding the foundation to advance new science, green chemistry, and environmental health. Environmental Health Perspectives 117(8), 1202–1209. [DOI] [PMC free article] [PubMed] [Google Scholar]
  71. World Health Organization, 1990. Environmental Health Criteria 97 - Deltamethrin. International Programme on Chemical Safety, Geneva, Switzerland, pp. 1–133. [Google Scholar]
  72. Yang H, Sun L, Li W, Liu G, Tang Y, 2018. Identification of Nontoxic Substructures: A New Strategy to Avoid Potential Toxicity Risk. Toxicological Sciences 165(2), 396–407. [DOI] [PubMed] [Google Scholar]
  73. Zhang L, Gavin T, DeCaprio AP, LoPachin RM, 2010. γ-Diketone axonopathy: analyses of cytoskeletal motors and highways in CNS myelinated axons. Toxicological Sciences 117(1), 180–189. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplemental Table 1
NIHMS1599056-supplement-Supplemental_Table_1.pdf (161KB, pdf)

RESOURCES