Molecular Mechanisms of the Conjugated α,β-Unsaturated Carbonyl Derivatives: Relevance to Neurotoxicity and Neurodegenerative Diseases

Abstract

Conjugated α,β-unsaturated carbonyl derivatives such acrylamide, acrolein, and 4-hydroxy-2-nonenal (HNE) are members of a large class of chemicals known as the type-2 alkenes. Human exposure through diet, occupation, and pollution is pervasive and has been linked to toxicity in most major organs. Evidence suggests that these soft electrophiles produce toxicity by a common mechanism involving the formation of Michael-type adducts with nucleophilic sulfhydryl groups. In this commentary, the adduct chemistry of the α,β-unsaturated carbonyls and possible protein targets will be reviewed. We also consider how differences in electrophilic reactivity among the type-2 alkenes impact corresponding toxicokinetics and toxicological expression. Whereas these concepts have mechanistic implications for the general toxicity of type-2 alkenes, this commentary will focus on the ability of these chemicals to produce presynaptic damage via protein adduct formation. Given the ubiquitous environmental presence of the conjugated alkenes, discussions of molecular mechanisms and possible neurotoxicological risks could be important. Understanding the neurotoxicodynamic of the type-2 alkenes might also provide mechanistic insight into neurodegenerative conditions where neuronal oxidative stress and presynaptic dysfunction are presumed initiating events. This is particularly germane to a recent proposal that lipid peroxidation and the subsequent liberation of acrolein and HNE in oxidatively stressed neurons mediate synaptotoxicity in brains of Alzheimer's disease patients. This endogenous neuropathogenic process could be accelerated by environmental type-2 alkene exposure because common nerve terminal proteins are targeted by α,β-unsaturated carbonyl derivatives. Thus, the protein adduct chemistry of the conjugated type-2 alkenes offers a mechanistic explanation for the environmental toxicity induced by these chemicals and might provide insight into the pathogenesis of certain human neurodegenerative diseases.

α,β-Unsaturated carbonyl derivatives such as acrylamide (ACR), acrolein, and methylvinyl ketone (MVK) (Fig. 1) are characterized by a conjugated structure that is formed when an electron-withdrawing group (e.g., a carbonyl or amide group) is linked to an alkene. Chemicals with this structure are classified as type-2 alkenes (Kemp and Vellaccio, 1980) and are used extensively in the manufacturing, agricultural, and polymer industries. The type-2 alkenes are also widely recognized as environmental pollutants (e.g., acrolein, MVK) and dietary contaminants (e.g., ACR, methyl acrylate; reviewed in Beauchamp et al., 1985; Bisesi, 1994; Friedman, 2003; Morgan et al., 2000). As a consequence, human exposure to conjugated alkenes is pervasive and has been associated with toxicity of most major organs including the hepatic, renal, respiratory, and nervous systems (e.g., Beauchamp et al., 1985; Bisesi, 1994; Gold and Schaumburg, 2000; Sax and Lewis, 1987; Schulz et al., 2001; Tucek et al., 2002). Although the respective toxic mechanisms have not been fully elucidated, the conjugated α,β-unsaturated carbonyl structure of these chemicals is a soft electrophile that preferentially forms Michael-type adducts with soft nucleophiles. In biological systems, sulfur atoms are the principal soft nucleophilic targets (Esterbauer et al., 1991; Friedman, 2003; LoPachin and DeCaprio, 2005; Pearson and Songstad, 1967) and proteomic analyses have shown that the type-2 alkenes form adducts with sulfhydryl groups on cysteine residues of numerous cellular proteins (Barber and LoPachin, 2004; Doorn and Petersen, 2002; Hall et al., 1993; LoPachin et al., 2006; Nerland et al., 2003).

FIG. 1.

(A) This figure illustrates the conjugated α,β-unsaturated carbonyl structure of chemicals in the type-2 alkene class. (B) This figure presents the corresponding line structures for acrolein, HNE, and other structurally related type-2 alkenes. Also shown are several examples of nonconjugated structural analogs.

Substantial evidence now suggests that the conjugated α,β-unsaturated carbonyl derivatives produce toxicity through a common mechanism involving the formation of adducts with proteins that play critical roles in cellular processes (reviewed in Friedman, 2003; Kehrer and Biswal, 2000; LoPachin and Barber, 2006; Witz, 1989). Therefore, in this commentary, we will review the adduct chemistry of the conjugated alkenes and identify their putative cellular protein targets. Our current understanding of this chemical class suggests that, although ACR and acrylonitrile are neurotoxicants, systemic exposure to other conjugated alkenes (e.g., acrolein, MVK) has been associated with peripheral organ toxicity (e.g., hepatotoxicity). Accordingly, we will consider the possibility that electrophilic reactivity is the primary determinant of the type-2 alkene tissue distribution which, in turn, determines toxic expression. Although these concepts have mechanistic implications for general toxicity, there is growing evidence that acrolein, 4-hydroxy-2-nonenal (HNE), MVK, and other α,β-unsaturated carbonyl can produce neurotoxicity by causing nerve terminal dysfunction (e.g., see Keller et al., 1997a,b; LoPachin et al., 2007a,b; Morel et al., 1999; Pocernich et al., 2001). Consequently, this commentary will focus on the newly recognized neurotoxic potential of conjugated alkenes and will explore inherent molecular and anatomical characteristics that predispose nerve terminals to electrophilic attack. The molecular mechanisms and nerve terminal sites of action discussed here could have human implications because, in addition to ACR, neurotoxicity might be an outcome of subchronic environmental exposure to other type-2 alkenes. Nerve terminal dysfunction is also an early neuropathogenic event in Alzheimer's disease (AD; e.g., Selkoe, 2002). A growing consensus suggests that AD is mediated by oxidative stress leading to lipid peroxidation and the subsequent liberation of acrolein and HNE. Therefore, in this commentary we will introduce the hypothesis that the endogenous production of type-2 alkenes in oxidatively stressed neurons is mechanistically linked to the synaptotoxicity associated with AD. Because type-2 alkenes appear to share a common nerve terminal site of action, we propose that the onset and development of AD can be accelerated by environmental exposure to these chemicals (e.g., dietary ACR, industrial methyl acrylate). We will begin this commentary with a review of adduct chemistry, because adduct formation is the fundamental mechanism of conjugated type-2 alkene toxicity.

ADDUCT CHEMISTRY OF THE CONJUGATED α,β-UNSATURATED CARBONYL DERIVATIVES

All α,β-unsaturated carbonyl and acrylic acid derivatives have a conjugated structure comprised of a functional group with electrophilic character (e.g., a carbonyl or amide group) attached to an alkene. When p orbitals overlap, as they do in a conjugated system, pi (π) molecular orbitals form. Electrons in pi orbitals are mobile and, therefore, the presence of an electron-withdrawing group can promote the polarization of these electrons. The resulting regional electron deficiency creates a “soft” electrophile as defined by the hard/soft, acid/base theory (Pearson and Songstad, 1967). Recent studies by LoPachin et al. (2007b) have demonstrated that relative softness, as calculated by the quantum mechanical parameter σ, was a primary determinant of in vitro synaptosomal dysfunction caused by the type-2 alkenes. Thus, corresponding σ values were closely correlated to the second order rate constants for the reactions of these electrophiles with sulfhydryl groups on N-acetyl-l-cysteine (NAC) (Table 1). These rates of reaction were, in turn, related to the respective in vitro neurotoxic potencies (Table 1). Soft electrophiles preferentially form covalent adducts with nucleophiles of similar softness via Michael-type addition reactions. The relative softness of a given nucleophile is determined by the polarizability of corresponding valence electrons. Sulfur has a relatively large atomic radius with highly polarizable valence electrons and is, therefore, the softest nucleophile in biological systems. Mass spectrometry and other proteomic approaches have confirmed that the type-2 alkenes preferentially form adducts with protein and nonprotein sulfhydryl groups on cysteine residues (e.g., see Barber and LoPachin, 2004; Doorn and Petersen, 2002, 2003; LoPachin et al., 2006, 2007a; Schauenstein et al., 1971; Uchida et al., 1998a,b; Uchida and Stadtman, 1993; van Iersel et al., 1997). Free sulfhydryl groups can exist in the reduced thiol-state or in the anionic thiolate-state and recent research indicates that the highly nucleophilic thiolate state is the adduct target for the type-2 alkenes (Tables 2 and 3; LoPachin et al., 2007b, see also Esterbauer et al., 1991).

TABLE 1.

Calculated Quantum Mechanical Parameters for α,β-Unsaturated Carbonyl Derivatives and Nonconjugated Analogs

	ELUMO (ev)	EHOMO (ev)	σ (ev−)	ω (ev)	log k2	log IC50
Conjugated alkenes
NEM	− 2.36	− 7.30	0.406	4.73	2.170	− 4.33
Acrolein	− 1.70	− 6.98	0.379	3.57	0.332	− 4.28
HNE	− 1.53	− 6.78	0.381	3.29	ND	ND
MVK	− 1.33	− 6.71	0.372	3.00	0.037	− 3.48
MA	− 1.01	− 7.36	0.315	2.76	− 1.893	− 0.336
ACR	− 0.69	− 6.77	0.329	2.30	− 3.651	− 0.359
Nonconjugated analogs
Propanal	− 0.33	− 6.858	0.307	1.981	—	—
Allyl alcohol	+ 0.51	− 6.933	0.269	1.386	—	—

Note. The Lowest Unoccupied Molecular Orbital (LUMO) energy (E_LUMO) and Highest Occupied Molecular Orbital (HOMO) energy (E_HOMO), were calculated using Spartan04 (version 1.0.3) software (Wavefunction Inc., Irvine, CA). Global (whole molecule) hardness (η) was calculated as η = (E_LUMO − E_HOMO)/2 and softness (σ) was calculated as the inverse of hardness or σ = 1/η. The electrophilicity index (ω) was calculated as ω = μ²/2η, where μ is chemical potential of the electrophile and was calculated as μ = (E_LUMO + E_HOMO)/2 (for details see LoPachin et al., 2007b). According to the Frontier Orbital Theory, adduct formation occurs when a soft nucleophile donates its highest energy electrons into the empty lowest energy orbital (LUMO) of a soft electrophile. Therefore, the most relevant frontier orbital for electrophiles is the LUMO, whereas the HOMO is most important for nucleophiles. Softness is defined as the ease with which electron redistribution takes place during covalent bonding and thus, the softer the electrophile (i.e., more negative E_LUMO value and higher σ value), the more readily it will form an adduct by accepting outer shell electrons from a soft nucleophile such as a sulfur atom. The electrophilicity index (ω) is a higher order parameter that combines softness with chemical potential and represents a sensitive measure of electrophilic reactivity. Based on the respective σ, ω, and E_LUMO values, NEM, and acrolein are substantially softer (more reactive) than MA and ACR, whereas HNE and MVK are of intermediate softness. These parameters were highly correlated (r² ≥ 0.93) to the second order rate constants (log k₂) for the reactions of each electrophile with cysteine (pH 7.4) and to the corresponding levels of synaptosomal neurotoxicity (r² ≥ 0.83; log IC₅₀ for inhibition of membrane dopamine transport in striatal synaptosomes). The nonconjugated analogs (propanal, allyl alcohol) did not cause in vitro neurotoxicity and had correspondingly lower values for each measured parameter (see LoPachin et al., 2007b for details).

TABLE 2.

Calculated Quantum Mechanical Parameters for Nucleophilic Amino Acids

Amino acid residue	ELUMO (ev)	EHOMO (ev)	μ (ev)
Cysteine thiol (0)	0.14	− 5.87	− 2.87
Cysteine thiolate (− 1)	4.76	− 0.35	2.21
Histidine (− 1)	4.59	− 1.12	1.74
Lysine (0)	0.39	− 5.59	− 2.60

Note. Parameters were calculated for the listed nucleophiles as described in the legend of Table 1. For each nucleophile, the respective ionization-state is presented in parentheses. As stated in the legend of Table 1, adduct formation is mediated by the transfer of high-energy electrons from the HOMO of the soft nucleophile into the LUMO of the soft electrophile. The chemical potential [μ = (E_LUMO + E_HOMO)/2] of the nucleophile is a measure of how readily electron density can be donated to the soft electrophile acceptor and is, therefore, an index of relative nucleophilicity (see LoPachin et al., 2007b for details). Values of μ are independent of pH and reflect the inherent electronic nature of the amino acid residue upon which the calculations are based. According to the respective μ values, cysteine sulfhydryl groups in the thiolate-state are substantially more nucleophilic (more positive μ value) than either the cysteine thiol-state or the ϵ amino group of lysine residues. Whereas the imidazole moiety of histidine is a better nucleophile than lysine, the μ value is smaller than that of cysteine indicating lower nucleophilicity. In addition, the μ parameter is based on electronic characteristics only and does not take into consideration other physiochemical factors such as steric hindrance of the histidine ring structure that might impact functional nucleophilicity. These data demonstrate that the cysteine thiolate-state is a better nucleophile than lysine, histidine, or the sulfhydryl thiol-state and is, therefore, the preferred target for the type-2 alkenes.

TABLE 3.

Calculated Nucleophilic Indices (ω⁻) for Type-2 Alkene Reactions with Possible Nucleophilic Targets

Electrophile	ω− Cys (− 1)	ω− Cys (0)	ω− His (− 1)	ω− Lys (0)
NEM	2.51	0.194	2.17	0.250
Acrolein	2.03	0.103	1.75	0.143
HNE	1.93	0.083	1.65	0.115
MVK	1.83	0.064	1.56	0.097
MA	1.59	0.069	1.38	0.099
ACR	1.50	0.036	1.28	0.060

Note. The nucleophilicity index (ω⁻) was calculated as ω = η_A (μ_A − μ_B)²/2(η_A − η_B)², where η = (E_LUMO − E_HOMO)/2, μ = (E_LUMO + E_HOMO)/2, A = reacting nucleophile and B = reacting electrophile (see LoPachin et al., 2007b for details). For each nucleophile, the respective ionization-state is presented in parentheses. The nucleophilicity index is a higher order parameter that considers the respective hardness and chemical potential of the electrophilic (type-2 alkene) and nucleophilic (cysteine, histidine, or lysine) reactants and is, therefore, a measure of the likelihood of subsequent adduct formation. As suggested by the respective ω⁻values, the type-2 alkenes preferentially form adducts with cysteine thiolate sites as opposed to histidine, lysine, or thiol residues. This conclusion is supported by direct evidence from numerous proteomic studies (e.g., Barber and LoPachin, 2004; LoPachin et al., 2007a; Barber et al., 2007; reviewed in LoPachin and DeCaprio, 2005).

Amino groups on lysine or histidine residues are also nucleophilic and are, therefore, potential sites for type-2 alkene adduction. However, the imidazole moiety of histidine and the ϵ-amino group of lysine are significantly harder nucleophiles than thiolates (Table 2) and, as reflected in their slower rate constants (Doorn and Petersen, 2002, 2003; van Iersel et al., 1997; reviewed in Petersen and Doorn, 2004), these residues are unfavorable targets for adduction by soft electrophiles (Table 3; reviewed in Coles, 1984–1985; Hinson and Roberts, 1992; LoPachin and Barber, 2006; LoPachin and DeCaprio, 2005; Pearson and Songstad, 1967). As a practical demonstration of sulfhydryl selectivity, neuroprotection studies have shown that NAC (500μM), but not N-acetyl-l-lysine (500μM) or β-alanyl-l-histidine (carnosine; 500μM), prevented the in vitro neurotoxicity induced by exposure of synaptosomes to graded concentrations of type-2 alkenes (LoPachin et al., 2007a; LoPachin et al., unpublished data). These data indicate that, in contrast to lysine or histidine residues, the sulfhydryl group on NAC can act as a scavenger for the type-2 alkenes. Other evidence suggests that the formation of Michael-type sulfhydryl adducts is reversible and, therefore, lacks toxicological relevance (Uchida and Stadtman, 1992; Uchida et al., 1994). However, in physiological conditions, the reverse kinetics of this carbonyl condensation reaction are extremely slow and, therefore, functionally irreversible (Esterbauer et al., 1991; Kemp and Vellaccio, 1980). This is evidenced by the survival of cysteine-conjugated alkene adducts during preparation for mass spectrometric analyses (e.g., see Barber and LoPachin, 2004; Crabb et al., 2002; Ishii et al., 2003; LoPachin et al., 2006) and by the accumulation of ACR–cysteine adducts in brain (synaptosomes) of subchronically intoxicated laboratory animals (Barber and LoPachin, 2004; Barber et al., in press). Finally, some in vitro evidence has suggested that type-2 alkene neurotoxicity is mediated by cellular oxidative stress following adduction and subsequent depletion of nonprotein, cysteine containing reducing equivalents such as glutathione (Hashimoto and Aldridge, 1970; Kruman et al., 1997; Pocernich et al., 2001). However, studies in whole animal models have shown that type-2 alkene intoxication does not affect neuronal redox-state (Johnson and Murphy, 1977; LoPachin et al., 2006). Furthermore, substantial evidence from the Toxicology literature indicates that the formation of cysteine adducts on proteins, and not GSH depletion, is the primary pathogenic step in type-2 alkene toxicity (e.g., see Biswal et al., 2002, 2003; Grafström et al., 1988; Ku and Billings, 1986; Park et al., 2002; Patel and Block, 1993; reviewed in Kehrer and Biswal, 2000).

The evidence discussed thus far suggests that α,β-unsaturated carbonyl derivatives cause toxicity by forming stable, irreversible adducts with highly nucleophilic thiolate groups on protein cysteine residues. However, most proteins contain cysteines and, therefore, target specificity is uncertain. As indicated, the highly nucleophilic anionic thiolate-state of the cysteine sulfhydryl group is the kinetically favored target for acrolein, ACR, and other conjugated type-2 alkenes (Tables 2 and 3). Based on the pH of the cellular milieu (pH ∼7.4), the thiolate concentration in biological systems is predicted to be low (∼10%) because the average pKa of cysteine sulfhydryl groups is approximately 8.5. However, thiolate groups are more prevalent than expected due to the existence of relatively low pKa sulfhydryl groups on cysteine residues located within highly specialized amino acid sequences known as catalytic triads (or diads; reviewed in LoPachin and Barber, 2006; Stamler et al., 2001). Proton shuttling between flanking or proximal basic amino acid residues (histidine, arginine, lysine) and their acidic counterparts (aspartate, glutamate) significantly lowers the pKa of cysteine sulfhydryl groups in these configurations. Catalytic triads are found in the active sites of many proteins (e.g., N-ethylmaleimide sensitive factor [NSF], glyceraldehyde-3-phosphate dehydrogenase, vacuolar-adenosine triphosphatase [v-ATPase]) and play critical roles in modulating function (e.g., see Greco et al., 2006; Leiper et al., 2002; reviewed in LoPachin and Barber, 2006; Stamler et al., 2001). Cysteine thiolates in catalytic triads, therefore, represent highly specific, mechanistically relevant targets for soft electrophilic toxicants. In contrast, the thiol-state of cysteine residues is not a preferred electrophile target, because this state has relatively low nucleophilic reactivity (Table 2).

α,β-UNSATURATED CARBONYL AND ACRYLIC ACID DERIVATIVES—NEUROTOXIC ACTIONS

Data presented in the preceding section suggest that the conjugated type-2 alkenes produce toxicity by forming Michael-type adducts with thiolate sulfhydryl groups on cysteine residues of catalytic triads. Other than the well-recognized neurotoxic effects of ACR, the majority of α,β-unsaturated carbonyl and acrylic acid derivatives are considered to be peripheral organ toxicants; for example, MVK inhalation produces pulmonary toxicity (Morgan et al., 2000). There is, however, growing evidence that this class of chemicals can also damage nerve terminals and, in this section, we review results from supporting in vivo and in vitro studies.

In Vivo Studies

Of the chemicals in the type-2 alkene class, the in vivo neurotoxicological actions of ACR and acrylonitrile have been studied in most detail. Systemic intoxication of humans and laboratory animals with ACR has been shown to produce ataxia, cognitive impairment, and skeletal muscle weakness (LoPachin et al., 2002; Spencer and Schaumburg, 1974a,b). Early neuropathological studies indicated that ACR neurotoxicity involved distal axon degeneration (axonopathy) in the peripheral nervous system and central nervous system (CNS) (Spencer and Schaumburg, 1974b). However, other morphological evidence has demonstrated that nerve terminal damage was the primary neuropathogenic event and that axonopathy was consequential (Cavanagh, 1982; DeGrandchamp and Lowndes, 1990; DeGrandchamp et al., 1990; Prineas, 1969; Tsujihata et al., 1974; reviewed in LoPachin et al., 2002, 2003). A series of seminal electrophysiological studies (De Rojas and Goldstein, 1987; Goldstein, 1985; Goldstein and Lowndes, 1979, 1981; Lowndes and Baker, 1976; Lowndes et al., 1978a,b) showed that neurotransmission was impaired at spinal cord primary afferent nerve terminals as an early consequence of experimental ACR intoxication. Defective neurotransmission was also found at peripheral neuromuscular junctions and at autonomic synapses of ACR-exposed laboratory animals (Munch et al., 1994; Tsujihata et al., 1974). More recent quantitative morphometric studies and silver stain analyses (Lehning et al., 1998, 2002a,b, 2003) have indicated that, regardless of dose-rate, ACR intoxication caused early nerve terminal degeneration. This effect preceded onset of the corresponding neurological deficits (gait abnormalities, skeletal muscle weakness) and was not secondary to glial cell injury or to changes in neuronal protein synthesis, axonal transport, energy metabolism, or nerve terminal redox status (reviewed in LoPachin, 2004; LoPachin and Lehning, 1994; LoPachin et al., 2002, 2003). Based on the preceding evidence, LoPachin et al. (2002, 2003) proposed that the nerve terminal was the primary site of ACR action and that synaptic dysfunction was a necessary and sufficient step in the production of neurotoxicity. In support of this hypothesis, in vivo and in vitro studies showed that ACR exposure was associated with reduced presynaptic neurotransmitter release. This effect involved inhibition of key proteins that regulate membrane–vesicle fusion; for example, NSF, synaptosomal associated protein of 25 kDa (Barber and LoPachin, 2004; LoPachin et al., 2004). During subsequent investigations to establish the neurotoxicological specificity of impaired release, we found that ACR also inhibited other presynaptic processes in vivo and in vitro; for example, membrane neurotransmitter uptake and vesicular storage (LoPachin et al., 2004, 2006). Proteomic analyses indicated that inhibition of these processes was associated with the formation of adducts on participating proteins; for example, the dopamine transporter and v-ATPase, respectively (Barber and LoPachin, 2004; Barber et al., 2007; LoPachin et al., 2004, 2007a). This suggested that ACR impaired neurotransmission by affecting diverse nerve terminal processes.

Acrylonitrile (ACN) is an unsaturated aliphatic nitrile of the conjugated type-2 alkene class. Oral intoxication of rats with ACN (12.5–50 mg·kg⁻¹·days⁻¹ × 60 days) produced weight loss, ataxia, and hindlimb skeletal muscle weakness similar to that caused by ACR. Muscle weakness was associated with dose-dependent reductions in both peripheral motor and sensory nerve conduction velocities (Gagnaire et al., 1998). Additional studies suggested that ACN intoxication in a rodent model produced ototoxicity mediated by cellular glutathione depletion and adduction of protein cysteine residues (Fechter et al., 2003; Pouyatos et al., 2005). Thus, both ACR and ACN produce similar neurological deficits in laboratory animals, although the specific nerve terminal effects of ACN have not been identified.

In Vitro Studies

To define the structural requirements of the sulfhydryl-dependent nerve terminal toxicity induced by ACR, we determined the structure–toxicity relationships for a series of conjugated type-2 alkenes and their nonconjugated analogs (Fig. 1B; LoPachin et al., 2007a,b; LoPachin et al., unpublished data). Results showed that conjugated α,β-unsaturated carbonyl derivatives produced parallel, concentration-dependent inhibitions of neurotransmitter (dopamine) release, membrane reuptake (Fig. 2A) and vesicular storage in exposed striatal synaptosomes(LoPachin et al., 2007a,b). Parallel mass spectral analyses and measurements of free sulfhydryl loss (Fig. 2B) confirmed that the type-2 alkene neurotoxicity involved the formation of Michael-type monoadducts with sulfhydryl groups on presynaptic proteins (Barber and LoPachin, 2004; Barber et al., in press; LoPachin et al., 2004, 2007a,b; Morel et al., 1999). Nonconjugated structural analogs such as allyl alcohol and propanal (Fig. 1B) did not affect synaptosomal sulfhydryl content and were devoid of in vitro neurotoxicity. These findings are consistent with results from earlier in vitro studies, which were designed to test the hypothesis that liberation of conjugated α,β-unsaturated carbonyls in oxidatively stressed neurons mediated nerve terminal damage in AD (e.g., Keller et al., 1997a,b; see ahead). Thus, in vitro studies showed that acrolein and HNE disrupted synaptosomal membrane protein conformation and phospholipid asymmetry (Castegna et al., 2004; Pocernich et al., 2001; Subramaniam et al., 1997), reduced glutamate uptake and GLUT3-mediated glucose transport in synaptosomes and cultured nerve cells (Keller et al., 1997a,b; Lovell et al., 2000), decreased synaptosomal mitochondrial respiration and induced oxidative stress (Humphries et al., 1998; Luo and Shi, 2005; Morel et al., 1999; Picklo et al., 1999; Picklo and Montine 2001; Raza and John, 2006), inhibited membrane Na⁺- and Ca²⁺-pump activities (Keller et al., 1997b; Mark et al., 1997), and caused loss of ion regulation in cultured nerve cells (Mark et al., 1997). HNE also has been shown to inhibit the activity of nerve terminal mitochondrial aldehyde dehydrogenase (ALDH2), which catalyzes the oxidation of 3,4-dihydroxyphenylacetaldehyde (DOPAL) to 3,4-dihydroxyphenylacetic acid (see Florang et al., 2007). DOPAL is a highly neurotoxic intermediate of DA metabolism (Burke et al., 2004) and, therefore, HNE-mediated inhibition of ALDH2 activity could result in nerve terminal accumulation of this electrophilic metabolite and the subsequent destruction of striatal synapses.

FIG. 2.

The concentration-dependent effects of α,β-unsaturated alkene derivatives on dopamine transport in striatal synaptosomes (A) and corresponding sulfhydryl content (B). Data are expressed as mean percent control ± SEM and calculated IC₅₀’s are provided in parentheses. Results show that exposure of synaptosomes to a relatively broad concentration range (1μM–10mM) of type-2 alkenes produced parallel, concentration-dependent decreases in synaptosomal transport (A). The decreases in neurotransmitter transport induced by each structural congener were highly correlated (r² ≥ 0.91) to corresponding reductions in sulfhydryl content (B). Although differences in potency were evident, all conjugated analogs exhibited comparable neurotoxic efficacy; that is, each chemical was capable of producing maximal inhibitions of the measured neurochemical parameters and correspondingly depleted sulfhydryl contents. These structure-toxicity data are consistent with previous studies and suggest that nerve terminal toxicity mediated by sulfhydryl adduct formation is a class characteristic of the type-2 alkenes (see also Castegna et al., 2004; Keller et al., 1997a,b; Morel et al., 1999; Pocernich et al., 2001). These data also indicate that the synaptosomal toxicity of the type-2 alkenes is related to their common conjugated α,β-unsaturated structure (see LoPachin et al., 2007a,b; LoPachin et al., unpublished data, for more detailed discussions).

Thus, a large body of in vivo and in vitro data indicate that ACR, a prototypical conjugated type-2 alkene, produces neurotoxicity by impairing nerve terminal function. Results of in vitro investigations also showed that other conjugated alkene species have neurotoxic potential through similar presynaptic effects. In the next section, we propose that the type-2 alkenes disrupt nerve terminal processes by inhibiting nitric oxide (NO) signaling in this neuronal region.

MOLEUCLAR MECHANISMS OF CONJUGATED TYPE-2 ALKENE NEUROTOXICITY

Recent proteomic studies have identified numerous protein targets of the type-2 alkenes (Barber and LoPachin, 2004; Barber et al., 2007; LoPachin et al., 2004, 2006; Uchida and Stadtman, 1993), of which most (e.g., NSF, the dopamine transporter, the vacuolar-ATPase) are also acceptors for NO (reviewed in LoPachin and Barber, 2006; Stamler et al., 2001). NO is a biological electrophile that reversibly forms adducts with sulfhydryl groups (S-nitrosylation) and thereby regulates the activities of many nerve terminal proteins and their respective pathways (Esplungues, 2002; Kiss, 2000; LoPachin and Barber, 2006). It now appears that the thiolates of cysteine catalytic triads in proteins are acceptors for NO (Forman et al., 2004; Stamler et al., 2001). Therefore, adduction of these NO acceptor protein sites could explain the diverse neurotoxic actions of the type-2 alkenes. The specificity and functional independence of NO-mediated actions is a product of signaling modules that act as neuronal microprocessors; for example, the regulatory subunit (NR2A) of the N-methyl-D-aspartate receptor (Lipton et al., 2002). Although controversy exists regarding the effects of NO signaling at the nerve terminal, a weight of evidence suggests a reduction in synaptic strength through binary (“on–off”) regulation of many presynaptic processes; that is, NO inhibits synaptic vesicle membrane fusion and decreases both membrane neurotransmitter uptake and vesicular storage (reviewed in LoPachin and Barber, 2006). Our studies suggest that the type-2 alkenes mimic these neurophysiological effects of NO signaling (see above). Therefore, oxidation of thiolate anionic sites in protein catalytic triads by either reversible S-nitrosylation (endogenous NO) or irreversible alkylation (type-2 alkenes) produces similar synaptic effects that differ with respect to duration of effect and outcome—transient neuromodulation versus persistent neurotoxicity, respectively. Because the type-2 alkenes form irreversible adducts with NO-targeted thiolates, we have hypothesized that NO signaling is blocked and that the loss of reversible, spatially precise neuromodulation produces synaptic toxicity (LoPachin and Barber, 2006).

NO modulates cellular processes in most major organ systems (Gaston, 1999; Hess et al., 2005) and catalytic triads are components of many neuronal and nonneuronal proteins (Mei and Zalkin, 1989; Sfakianos et al., 2002). Therefore, a hypothesis based on these generic cellular features seems inconsistent with a selective nerve terminal site of type-2 alkene action. However, several unique anatomical and functional characteristics predispose nerve terminals to damage by these toxicants. A notable vulnerability is that the multistep synaptic vesicle cycle operates at an exceptionally high rate and is extensively regulated by NO signaling (reviewed in Esplungues, 2002; Kiss, 2000; LoPachin and Barber, 2006). Disruption of presynaptic NO modulation could, therefore, have broad functional consequences for neurotransmission. Furthermore, the nerve terminal is an anatomically remote site that is separated from the cell body and is devoid of transcriptional or translational capacity. These deficiencies impose significant vulnerabilities related to reparative and protective responses. Thus, the cell body of neuronal and nonneuronal cells can mount a protective response to an electrophile mediated by, for example, the neurite outgrowth-promoting prostaglandin–Keap1/Nrf2 pathway (Satoh and Lipton, 2006; Satoh et al., 2006). Electrophilic neurotoxicants form adducts with cysteine sensors on certain cytosolic proteins (e.g., Keap1). This promotes the activation of transcription factors (e.g., Nrf2) that bind electrophile (EpRE) and antioxidant (ARE) response elements in the 5′-regulatory region of target genes. These genes encode a subset of electrophile scavenging molecules (e.g., glutathione S-transferases) and proteins that have antioxidant actions (e.g., heme oxygenase 1, thioredoxin). These coordinated responses can provide the cell body with protection or, at least, delay the onset of type-2 alkene toxicity in this neuronal region. However, in the absence of transcriptional machinery, nerve terminals cannot initiate a similar protective reaction and, as a result, are selectively vulnerable to attack by the type-2 alkenes. Because nerve terminals cannot manufacture proteins, cell body synthesis and subsequent protein delivery maintain the presynaptic proteome by anterograde axonal transport. To conserve manufacturing costs, energy, and resources, the turnover of nerve terminal proteins is exceptionally slow relative to proteins in other nerve regions or cell types (Calakos and Scheller, 1996; Katyare and Shallom, 1988; Lin and Scheller, 2000). The slower turnover rate of nerve terminal proteins has significant mechanistic implications (discussed extensively in Barber et al., 2007; LoPachin and DeCaprio, 2005; LoPachin and Barber, 2006); that is, when proteins with a long half-life are adducted and rendered dysfunctional, they are replaced slowly. With continued toxicant exposure, the pool of dysfunctional proteins increases and the related cellular pathways are progressively disabled. Given this proteome toxicodynamic, cumulative toxicity with parallel accumulation of protein adduct in target cells is the predicted outcome. Indeed, the cumulative neurotoxicity of ACR is well documented (Goldstein and Lowndes, 1979, 1981; Kuperman, 1958; LoPachin et al., 2002; Lowndes and Baker, 1976; Lowndes et al., 1978a,b) and we have recently shown that in CNS nerve terminals of intoxicated rats, ACR-cysteine adducts accumulate in parallel with the developing neurotoxicity (Barber and LoPachin, 2004; Barber et al., 2007). In contrast, adducted, dysfunctional proteins with short half-lives will not accumulate (i.e., they are rapidly replaced) and, consequently, will have minimal toxic impact.

Summary

The preceding evidence suggests that the type-2 alkenes are a class of soft electrophiles that impair protein function by forming irreversible adducts with the soft nucleophilic thiolate sites of cysteine catalytic triads. Protein turnover in nerve terminals is relatively slow and, as a result, proteins rendered dysfunctional by adduction are slowly replaced and toxicity develops in a cumulative fashion. Many of the cysteine sites of adduction are also NO acceptors and, therefore, irreversible adduction of the corresponding sulfhydryl thiolate groups will inhibit NO signaling. Because neurotransmission and other presynaptic processes are highly dependent upon NO modulation, type-2 alkene disruption of the NO pathway has broad toxic consequences for the nerve terminal. Accordingly, we have hypothesized that ACR, acrolein, HNE, and other α,β-unsaturated carbonyl derivatives can produce neurotoxicity by inhibiting NO signaling at the nerve terminal. Although a similar molecular toxicodynamic probably occurs when nonneuronal cells are exposed to conjugated type-2 alkenes (Campian et al., 2002; Kaminskas et al., 2005; Lash and Woods, 1991; McCarthy et al., 1994), relative differences in protein turnover, repair capacity, and electrophile responses in these cells likely prevent accumulation of adducts at lower dose-rates leading to reduced toxicant susceptibility.

HUMAN NEUROTOXICOLOGICAL IMPLICATIONS OF CONJUGATED TYPE-2 ALKENE EXPOSURE

Humans are pervasively exposed to type-2 alkenes through occupational sources (e.g., manufacturing, grouting), poisoning, diet, and pollution. In this section, we discuss how such exposure to these potentially neurotoxic chemicals might impact human health.

Environmental Type-2 Alkene Exposure

Conjugated α,β-unsaturated carbonyl and acrylic acid derivatives have extensive commercial and industrial applications (reviewed in Beauchamp et al., 1985; Bisesi, 1994; Friedman, 2003; Ghilarducci and Tjeerdema, 1995; Gold and Schaumburg, 2000; Kehrer and Biswal, 2000; Morgan et al., 2000; Novak, 1991). However, although neurotoxicity is a clearly defined outcome in ACR-intoxicated human cohorts; for example, as a result of poisoning or occupational exposure (see Fullerton, 1969; Garland and Paterson, 1967; He et al., 1989; Myers, 1991), similar intoxication with acrolein, MVK or other type-2 alkenes is primarily associated with systemic toxicity (Bisesi, 1994; Hales et al., 1992; Leikauf, 2002; Tucek et al., 2002). This toxicological diversity is not related to differences in molecular mechanisms among these chemicals (as discussed in Kehrer and Biswal, 2000; Lash and Woods, 1991; LoPachin et al., 2007a,b; Morgan et al., 2000), but is instead due to relative differences in electrophilic reactivity and the resulting impact on tissue distribution (Esterbauer et al., 1991; Gillette et al., 1974; Rozman and Klaassen, 2001; Uetrecht, 1992). As the data in Table 1 show, acrolein and MVK are highly electrophilic and, therefore, rapidly form adducts with sulfhydryl thiolate groups. Following systemic intoxication with reactive type-2 alkenes, the rapid formation of protein adducts essentially limits tissue distribution (Hoffmann et al., 1985) and, as a consequence, the resulting toxic manifestations are determined by the site of absorption; for example, inhalation of acrolein produces pulmonary toxicity, whereas systemic administration is associated with hepatic and vascular toxicity (Feron et al., 1978; Green and Egle, 1983; Putignano, 1954; reviewed in Beauchamp et al., 1985). In contrast to acrolein, ACR is a weak water-soluble electrophile that forms thiolate adducts slowly (Table 1). It is, therefore, less susceptible to the limiting influence of systemic “adduct buffering” and has a correspondingly larger volume of distribution that encompasses the CNS (Barber et al., 2001; Miller et al., 1982; see also Niemela, 2001; Trenga et al., 1991). These concepts are exemplified by observations that oral acrolein intoxication of rats (0.05–0.5 mg/kg) produced centrolobular liver necrosis and vascular hypertension (Green and Egle, 1983), whereas oral ACR (10–50 mg/kg) caused selective neurotoxicity in this species (LoPachin et al., 2002; Spencer and Schaumburg, 1974a,b). The large dose-rate differential between ACR and acrolein reflects differences in electrophilicity and corresponding rates of sulfhydryl adduct formation (Table 1). However, ACR forms protein thiolate adducts (albeit slowly) in both nervous and nonnervous organ systems and, therefore, neurotoxicity cannot be ascribed solely to a wider tissue distribution. Instead, as discussed above, nerve terminals are selectively predisposed to the cumulative actions of weak electrophiles due to the slow turnover of resident proteins and other liabilities. Thus, although it might seem counterintuitive, the greater threat of acquired neurotoxicity comes from exposure to weaker electrophiles (e.g., ACR, methyl acrylate (MA) acrylonitrile), whereas exposure to softer, more reactive, electrophiles (e.g., acrolein) is likely to produce systemic toxicity (i.e., hepatotoxicity).

Other than occupational or accidental/intentional intoxication (above), human exposure to the type-2 alkenes occurs at relatively low dose-rates (ng–μg/day) through diet and the ambient environment (e.g., pollution). With respect to those conjugated alkenes that are weak electrophiles (e.g., ACR, MA, acrylonitrile), the neurotoxicological significance of exposure at these lower background rates is uncertain. Data presented in this commentary suggest that the induction of neurotoxicity is directly related to the accumulation of adducts in nerve terminals and that such accumulation can occur only when the rate of formation exceeds the rate of removal by protein turnover. This indicates that a lower threshold exposure rate exists where adduct accumulation does not occur and, correspondingly, neurotoxicity does not develop. Furthermore, even though accumulation might occur at background exposure levels, the cumulative adduct levels achieved must be sufficient to impact critical presynaptic processes (e.g., synaptic vesicle cycling). Given the common presynaptic site of action for the type-2 alkenes and the ubiquitous human exposure to these chemicals, the adduct burden of nerve terminals could be significant. However, it seems unlikely that basal type-2 alkene exposure has direct human neurotoxicological relevance, because the presence of low-level hemoglobin–ACR adducts in normal animal and human cohorts (Barber et al., 2001; Calleman et al., 1994) suggests that adducts can accumulate at nontoxic rates during ambient exposure. Nonetheless, as discussed in the next subsection, the cumulative nerve terminal adduct burden resulting from baseline or higher (e.g. occupational) exposure might interact with neurodegenerative mechanisms that involve the endogenous production of highly reactive type-2 alkenes. Theoretically, the presynaptic interactions of exogenous (e.g., ACR, MA) and endogenous (e.g., acrolein) type-2 alkenes could augment the onset of neurological symptoms. Given the level of uncertainty, future research should focus on establishing the threshold adduct burden for combined type-2 alkene exposure and determining the composite Lowest Observed Adverse Effect Level.

Role of Type-2 Alkenes in Neurodegenerative Diseases

Epidemiological studies and basic molecular research have led to the supposition that exposure to chemicals in the environment contributes to the onset and development of human neurodegenerative diseases (Brown et al., 2005, 2006; Eisen, 1995; Landrigan et al., 2005). Of the many chemicals that might play a role in neurodegeneration, the type-2 alkenes are clearly relevant candidates given their ubiquitous environmental presence and the demonstrated vulnerability of nervous tissue to electrophile-induced toxicity. Not only could environmental exposure be involved, but there is also evidence that endogenous generation of highly electrophilic type-2 alkenes (e.g., acrolein, HNE) in oxidatively stressed neurons participates in the corresponding neuropathogenic mechanism. This has particular significance for neurodegenerative conditions such as AD, where nerve terminal dysfunction in neocortex, hippocampus, and other relevant brain regions precedes frank neuronal degeneration and appears to be an initial, primary pathophysiological event (e.g., see Bertoni-Freddari et al., 1996; Davies et al., 1987; DeKosky and Scheff, 1990; Scheff et al., 1990, 1996, 2006; Terry et al., 1991; reviewed in Coleman and Yao, 2003; Coleman et al., 2004; Scheff and Price, 2003; Selkoe, 2002; Walsh and Selkoe, 2004). Although the mechanism of synaptic dysfunction has not been elucidated (Honer, 2003; Klein, 2006; Mattson, 2004; Walsh and Selkoe, 2007; Yao, 2004), evidence suggests that the AD process involves several factors (genetic, age-related, and environmental) that converge to initiate neuronal oxidative stress and a subsequent pathophysiological cascade. Lipid peroxidation, a major consequence of oxidative stress, is the free-radical driven fragmentation of polyunsaturated fatty acids and is a biochemical hallmark of AD (Arlt et al., 2002; Keller and Mattson, 1998; Montine et al., 2002; Zarkovic, 2003). Peroxidation of phospholipids damages cellular membranes and generates toxic α,β-unsaturated carbonyl derivatives such as acrolein and HNE (Picklo et al., 2002; Uchida, 1999, 2003; Zarkovic, 2003). Indeed, numerous studies have reported elevated levels of HNE, acrolein, and their respective proteins adducts in relevant brain regions (e.g., amygdala, hippocampus) of AD patients and transgenic animal models (Calingasan et al., 1999; Kawaguchi-Niida et al., 2006; Lauderback et al., 2001; Lovell et al., 1997, 2001; Montine et al., 1997a,b, 1998; Sayre et al., 1997; Smith et al., 1998; Williams et al., 2005). As discussed above, acrolein and HNE can cause nerve terminal damage and, therefore, it is possible that the generation of type-2 alkenes in oxidatively stressed neurons is involved in AD synaptotoxicity. The salient features of this hypothetical scenario are illustrated in Figure 3. Whereas neurons might degenerate in response to the direct toxic effects of the type-2 alkenes (e.g., see Kruman et al., 1997), we propose that neurodegeneration in AD is secondary to both the functional (inhibited neurotransmission) and physical (nerve terminal degeneration) loss of synaptic contact (e.g., see Balice-Gordon and Lichtman, 1994; Pinter et al., 1995; reviewed in Raff et al., 2002).

FIG. 3.

(A) Several molecular features in the cell body of an oxidatively stressed neuron. Acrolein and HNE are endogenous by-products of membrane lipid peroxidation that develop as a consequence of oxidative stress (a). These type-2 alkenes can form adducts with key sulfhydryl groups of the Keap1–Nrf2 complex (b). This promotes dissociation of this complex and subsequent nuclear translocation of Nrf2 that specifically targets genes with ARE or EpRE, within their promoter regions. These genes encode a subset of drug metabolizing enzymes such as glutathione S-transferases (GST) and NAD(P)H-quinone oxidoreductase 1 (NQO1) and antioxidant molecules such as heme oxygenase 1 (HO-1) and thioredoxin. The detoxifying and antioxidative stress enzymes/proteins generated by the Keap1–Nrf2 pathway offer nerve cell protection from electrophilic attack. The relatively fast turnover of resident proteins also works in favor of the nerve cell body (c), because adducted and dysfunctional proteins can be quickly replaced and the operation of neuroprocesses maintained. (B) The possible toxicological events initiated by oxidative stress (a) in the nerve terminal. Thus, although protein adduction begins in the cell body (A) and progresses during axonal anterograde transport, adduction continues at the nerve terminal and results in a relatively large pool of adducted, dysfunctional proteins (b). These abnormal proteins participate in broad presynaptic functions; for example, neurotransmitter uptake, release and storage. Because turnover is relatively slow at the nerve terminal (b), protein adducts are replaced slowly and, therefore, accumulate. The build-up of functionally compromised proteins produces a progressive, cumulative neurotoxicity. Nerve terminals are also susceptible to electrophilic attack because, in the absence of transcriptional/translational capacity, an ARE/EpRE-type reaction cannot be initiated (c). Finally, many of the adducted cysteine sites are NO acceptors and, as a consequence, irreversible type-2 alkene adduction disrupts transient and spatially precise NO signaling. Nerve terminals are highly vulnerable to such disruption, because NO signaling modulates numerous critical presynaptic processes involved in neurotransmission; for example, the synaptic vesicle cycle (d). Decreased neurotransmission will disrupt regional neural circuits and higher order networks, which might be responsible for the difficulties in declarative memory that characterize AD (Buckner, 2004; Palop et al., 2006; Walsh and Selkoe, 2004). The peroxidation-based electrophile burdens in both nerve terminal and cell body are additive to preexisting type-2 alkene backgrounds that result from lifetime environmental exposure; for example, through food—symbolized by the French fries; through ambient/occupational exposure—symbolized by the industrial plant.

If type-2 alkene adduction of cysteine sulfhydryl groups on presynaptic proteins is involved in the synaptotoxicity associated with certain neurodegenerative diseases, then intervention methods that prevent the formation of these adducts should be effective. One strategy that has been considered is the administration of nucleophilic scavengers such as NAC (Arakawa et al., 2006; Benz et al., 1990; Lee et al., 2005). However, the results have been inconsistent, possibly because the pKa of the corresponding cysteine sulfhydryl group (e.g., NAC pKa = 9.6) and, therefore, the formation of the nucleophilic thiolate state at physiological pH, is inappropriate for effective electrophile scavenging. Alternatively, the use of nucleophiles with lower pKa's (e.g., cysteine ester pKa = 6.53) is likely to be neurotoxic due to the scavenging of endogenous electrophiles such as NO that perform critical signaling functions (LoPachin and Barber, 2006). Another neuroprotective approach is the use of hydralazine and other nucleophilic chemicals that act as “adduct traps.” Whereas studies involving in vitro type-2 alkene exposure have implicated the possible protective abilities of these chemicals (e.g., Kaminskas et al., 2005), the results from other investigations have failed to demonstrate neuroprotection (LoPachin et al., 2007a). Clearly, neuroprotection strategies that target thiolate adduct formation by the type-2 alkenes have significant therapeutic potential. However, in developing neuroprotective strategies based on nucleophilic scavengers, it is important to consider their possible toxic impact on neuronal signaling pathways that involve electrophiles.

CONCLUSIONS

In this commentary we have discussed evidence that the type-2 alkenes are a class of chemicals that produce toxicity in major organ systems through the formation of adducts with cysteine sulfhydryl groups of functionally critical proteins. The toxicological outcome (e.g., neurotoxicity, hepatotoxicity) of, for example, occupational exposure to these chemicals is dependent upon the corresponding electrophilicity and the route of intoxication. Neuronal damage induced by conjugated alkene exposure is mediated by protein adduct formation and the inherent predisposition of nerve terminals to electrophilic attack. This has considerable relevance to human health because chemicals in this class are pervasive environmental pollutants and dietary contaminants. Furthermore, there is now evidence that AD and possibly other neurodegenerative conditions (e.g., see Meissner et al., 2003; Sasaki and Iwata, 1999; Zarkovic, 2003) are associated with early nerve terminal dysfunction and liberation of the lipid peroxidation products, acrolein, and HNE. A relatively large database suggests that endogenous generation of type-2 alkenes is causally related to synaptotoxicity. Because the type-2 alkenes have common nerve terminal sites of action, it is possible that the onset and development of AD is accelerated by exogenous exposure to weak electrophiles such as ACR and methyl acrylate. Although there is substantial epidemiological data demonstrating the neurotoxicity of ACR in exposed human cohorts, comparable evidence for other conjugated alkenes with weak electrophilicity (e.g., methyl acrylate, acrylic acid) is lacking. This is because the possibility that weakly electrophilic type-2 alkenes are neurotoxic is a new development. In addition, except for specific occupational settings (e.g., grouting, polymer manufacturing), human type-2 alkene intoxication is likely to be at low subchronic rates from dietary sources, and/or environmental pollution. Whereas such exposure might be irrelevant with respect to the induction of neurotoxicity, low-level intoxication might, nonetheless, influence the age-related development of neurodegenerative diseases.

Although a large database from Neuroscience and Neurotoxicology exists, our understanding of type-2 alkene neurotoxicity is, nonetheless, limited. Therefore, it is hoped that the research discussed in this commentary will provide a framework for future studies that characterize the neurological consequences of conjugated alkene exposure and define the underlying molecular mechanisms. Such detailed information could also contribute to our understanding of certain neurodegenerative diseases and, thereby, accelerate the search for effective pharmacotherapies. Finally, it is hoped that this commentary will provide a toxicological rationale for epidemiological studies that determine the relative neurotoxicological risks of type-2 alkene exposure in human populations.

FUNDING

National Institutes of Health grants to R.M.L. from the National Institute of Environmental Health Sciences (RO1 ES03830-20, ES07912-9).