“Application of the Hard and Soft, Acids and Bases (HSAB) Theory to Toxicant–Target Interactions” LoPachin, R. M.; Gavin, T; DeCaprio, A.; Barber, D. S. Chem. Res. Toxicol. 2012, 25, 239-251. DOI: 10.1021/tx2003257
I considered posting about the Carreira group‘s work on enantioselective amination of allylic alcohols, because I think it is an awesome example of direct functionalization of hydroxylated substrates–an issue that will be of increasing importance in terms of biomass utilization. However I chose instead to stray into the less familiar territory of the bioactivity of organic molecules. I am semi-familiar with quantitative structure activity relationship (QSAR) modeling, wherein a database of known molecules and their bioactivity is used to predict the bioactivity of a molecule about which there is no bioactivity data. However, relying solely on computers leaves me wanting a more intuitive grasp on which molecules are expected to be toxic/non-toxic and why. That’s why I got excited about the recent perspective article about using the familiar Hard-Soft Acid-Base theory to predict toxicant-target interactions.
As can be seen in the above chart, most toxicants are electrophiles (e.g. acrolein, vinyl chloride, 2,5-hexanedione), and they do their damage by reacting with biological nucleophiles such as thiols/thiolates (e.g. cysteine) and amines (e.g. lysine). This technique computationally assesses the “softness” (relative polarizability) or “hardness” (relative non-polarizability) of electrophilic toxicants and nucleophilic biomolecules. From there, the concept is quite simple, electrophiles and nucleophiles of similar hardness or softness are more likely to react, leading to negative biological consequences.
As fearful as I am of anything remotely resembling math, I will briefly describe how they get their relevant values. Hardness (η) correlates directly with the energy difference between the highest occupied and lowest unoccupied molecular orbitals (HOMO-LUMO gap) of a given molecule. Softness (σ) is simply the inverse of hardness (η). For example, molecules with extended π systems or polarizable atoms such as sulfur and iodine typically have relatively small HOMO-LUMO gaps. They therefore have relatively low hardness (η) and high softness (σ) values.
The “electrophilic index” (ω) of a given electrophile can be calculated by combining its hardness (η) with its chemical potential (μ), as shown below. The authors describe chemical potential (μ) as “the propensity of a species to undergo chemical change,” and that is as good as I can do. If you have an intuitive chemical description beyond that, please comment (please!!!).
In terms of HSAB theory, the above equation reflects the fact that soft electrophiles (with low η) have high electrophilic indices (ω), and vice versa.
Let’s move on to some chemical structures! The paper shows five electrophiles for which softness (σ) and electrophilic indices (ω) were calculated and compared with in vitro nerve terminal synapse toxicity values (log IC50; more negative = more toxic).
This table shows that the softer electrophiles (acrolein and methyl vinyl ketone, entries A and B) are the most toxic, while the less soft electrophiles (methacrylate and acrylamide, entries D and E) are the least toxic. A similar toxicity correlation exists using the electrophilic indices (ω). So, remember that in this case, softness correlates directly with toxicity.
The one outlier in this model is hydroxynonenal (entry C), which has the highest values for softness and electrophilic index, yet is not the most toxic. This discrepancy is attributed to steric hindrance imposed by the long alkyl chain of hydroxynonenal that is not present in the other compounds. This issue highlights one of the primary limitations of this method, the fact that the only parameters considered are the relative electronics of the electrophile and nucleophile. The diplomatically named “intervening variables that influence interpretation” such as this were super interesting. I will describe two more below.
Hydroxynonenal is produced during the oxidative breakdown of lipids along with a number of other analogues shown in the table below. The electrophilic indices of those chemicals were calculated and their half-lives for adduct formation with the biological nucleophile surrogate N-acetyl-cysteine were measured. As you can see, the carboxylic acid/carboxylate structures in entries A & B have identical half-lives (580 s), as they were measured by placing the free acid in pH 7.4 water. At this pH the dominant species present was the carboxylate (entry B). So in fact, the half-life of the free acid (entry A) is not accurate. This is reflected in the fact that its half-life is close to that of hydroxynonenal (entry D, 732 s), but its high electrophilic index (1.70 eV) is not close to that of hydroxynonenal (1.18 eV). The electrophilic index of the carboxylate species (entry B, 1.34 eV) however, more closely matches that of hydroxynonenal (entry D). This is pretty straightforward, but serves as a reminder to keep the protonation states at physiological pH in mind when doing this kind of thing.
A more interesting “intervening variable” is that of water-solubility. The authors state that “toxicant solubility is a physiochemical determinant of tissue distribution and, therefore, target accessibility.” This can complicate analysis because solubility can be varied independently of HSAB properties. For example, the ethyl acrylate derivatives below have identical electrophilic indices (3.20 eV), but the more water soluble hydroxyethyl acrylate (lower log P value) is much more toxic to mice (lower LD50 value).
“Intervening variables” aside, let’s return to the issue of the correlation between electrophile softness and toxicity. Why do they correlate? Well, this report mainly describes Michael acceptors, which are relatively soft electrophiles. Accordingly, in biological systems they react orders of magnitude more rapidly with soft thiolate anions of cysteine side chains than they do with harder biological nucleophiles such as the amine groups of lysine, histidine, or the nucleobases.
The pKa of 8.3 of the cysteine thiol side chain means it exists primarily in the neutral, protonated state at physiological pH of 7.4. However, there are a number of pKa-lowering microenvironments in which thiolate groups can be in greater abundance. Accordingly, the authors state that “a large body of evidence indicates that α,β-unsaturated carbonyls selectively form adducts with…thiolate” anions. This is consistent with the relatively low softness value (σ) of cysteine thiol groups (0.28) and higher value of cysteine thiolate groups (0.38) due to the expanded electron cloud of the latter. This thiolate softness closely matches the softness values of acrolein (0.38) and acrylamide (0.35), so they preferentially form adducts with thiolate groups.
In order to more accurately predict whether a given nucleophile is likely to react with a given electrophile, the authors introduce the nucleophilic index (ω–), which takes into account the hardness (η) and chemical potential (μ) of both the nucleophile (A) and electrophile (B).
Using a variety of Michael acceptors, the authors show that there is a correlation between their reactivity towards cysteine thiolate groups (log kRS-) with their nucleophilic indices (ω–).
In addition, the nucleophilic indices (ω–) of those same cysteine-Michael acceptor pairs correlate directly with their neurotoxic potency (log IC50; more negative = more toxic). This provides evidence that the neurotoxicity observed is a result of the formation of thiolate-electrophile adducts. From a toxicologists’ point of view, it seems this is the most relevant insight provided by this method. Based on some relatively simple calculations and kinetics studies, you can determine what type of nucleophile is likely causing a given biological endpoint.
I will end this glossing over session by glossing over something else super cool. The cytoprotective effects of plant-derived polyphenols are typically ascribed to their antioxidant properties. However, the authors assert that part of their cytoprotective effects are likely caused by the presence of 1,3-dicarbonyl enols in their structures. These relatively soft carbon nucleophiles react with Michael-accepting toxicants and thus reduce the likelihood of those toxicants reacting with biologically important proteins.
So, what have we learned from this paper? Primarily, biology is super-complex! I may be frustrated with the non-intuitive nature of some QSAR models, but an accurate prediction of the bioactivity of a given molecule is rarely amenable to the simple intuitive descriptions that a chemist like me craves. On the other hand, we have seen here numerous examples of the softness of nucleophiles correlating directly with their toxicity, due to the prevalence of the soft thiolate side chains of cysteine residues. So, if you can’t avoid working with incredibly potent electrophiles like these, now you hopefully have a clearer picture of why exactly you should be careful when doing so.