Sciencemadness Discussion Board

Toxicity of AChE Inhibitors on a Quantitative, Molecular level

DDTea - 8-10-2009 at 14:37

I briefly mentioned in another thread (regarding 4th generation chemical agents) about my new interest in the mechanisms of action of these compounds. My goal is to determine quantitatively why one compound is more toxic than another. I suppose this may involve Structure Activity Relationships, but unfortunately those are, suffice to say, "tricky" for our abilities here. From what I can tell by reading Saunders is that they didn't do a true SAR; they took more of an approach along the lines of, "If I change this moiety, how much of the new compound will be required to kill a rat?" (Maybe something can be gleaned from a graph of LD50 vs. R group?)

So here's where I want to collaborate with people who know more about biochemistry than me. For the purpose of this discussion, I'd like to talk about essentially irreversible inhibitors.

What I'm hypothesizing that differentiates toxicity on a molecular level is the rate of inhibition of AChE. See the following diagram for an idea of what I want to talk about (for some reason, I can't get it to show up on this forum):

E = AChE, S = Acetylcholine, ES = complex, P = Choline + Acetate, and Ki = the important step I'd like to discuss

The question I'm leading up to: is the rate of Ki what determines overall toxicity (assuming there is no regeneration of AChE--again, assuming irreversible inhibition)? This is referring to the process of covalently changing the enzyme: is the rate of the leaving group/covalent change what initially determines toxicity and, subsequently, LD50?

If we expand the discussion further to include reversible inhibitors, would the rate of dissociation of the inhibitor also affect the toxicity? Thus it would reach some steady-state where some proportion of AChE is disabled, and the toxicity stems from the ability of a molecule to keep more AChE out of commission than another.

Am I on the right line of thinking? Again, this leads me to want to research some Structure-Activity Relationships on this topic. Questions that come to mind are: what promotes covalent changes to AChE? How does molecular geometry and sterics come into play--would a more sterically hindered molecule be more toxic because it is unable to leave the AChE active site, or less toxic because it cannot enter it? This might answer some questions as to why branched R groups in organophosphates have higher toxicity than linear R groups, or why the Phenyl group has significantly lower toxicity. Furthermore, if the inhibitor is able to undergo further chemical change inside the active site to bind to more amino acid residues, perhaps even to cause a change in the folding of AChE itself?

My goal is to be able to stand back like a physical chemist and predict toxicity based on molecular structure that are grounded in true chemistry rather than through conjecture. What sort of mathematical models should I use here? I'm really only familiar with Michaelis-Menten kinetics, but that doesn't represent the reaction here.

For semi-irreversible kinetics, the process [EI] --> [E] + [I] follows an exponential decay over time. From this rate of this decay, shouldn't it be possible to determine the toxicity of a compound by its "strength" of binding? That is, a more toxic compound will see a much slower decay than one that inhibits the enzyme then releases it. Therefore, the stronger inhibitor should have more [EI] than [E] in its steady-state.

EDIT 10/9/09: the picture didn't show up...
EDIT 10/19/09 17:40: bit of clarification
[Edited on 10-9-09 by DDTea]

[Edited on 10-9-09 by DDTea]

DDTea - 12-10-2009 at 20:15

I just discovered proteopedia, and specifically this article of interest:

Crystal structures of aged phosphonylated acetylcholinesterase: nerve agent reaction products at the atomic level., Millard CB, Kryger G, Ordentlich A, Greenblatt HM, Harel M, Raves ML, Segall Y, Barak D, Shafferman A, Silman I, Sussman JL, Biochemistry. 1999 Jun 1;38(22):7032-9. PMID:10353814

I haven't read through the journal article yet, but it's informative to see the "aging" process of Soman. Quoting the proteopedia page:

Organophosphorus acid anhydride (OP) nerve agents are potent inhibitors which rapidly phosphonylate AChE and then may undergo an internal dealkylation reaction (called "aging") to produce an OP-enzyme conjugate that cannot be reactivated.

So the process is basically: nucleophilic acyl subsitution by Ser200 on the phosphate producing the OP-enzyme conjugate and covalently deactivating the enzyme, followed by "aging"--or spontaneous release of the alcoholic R group, forming a carbocation. The other R group occupies the "acyl binding pocket" and is stabilized by hydrophobic protein residues Trp233, Phe288, and Phe290.

This gives information on what role the R groups play in determining toxicity. First of all, they must both be lipophilic. In the G-series, the R group directly bound to the Phosphorus "locks" the molecule in place (I'll call this R1). The second R group (P-O-R) spontaneously dissociates and forms a carbocation in a process called "aging (I'll call this R2)." The resulting negatively charged molecule is stabilized by electrostatic interactions and hydrogen bonding within the active site.

The next line of thinking, then, is that whatever stabilizes these processes should increase the reaction rate, affinity, and thus make the molecule more toxic. As always, the leaving group should be a good leaving group (and thus a poor nucleophile). R2 should be able to form a stable carbocation. This would explain why Soman is more toxic than Sarin, which is more toxic than Tabun: pinacolyl carbocation can form a tertiary, more stable, carbocation than an isopropyl carbocation, which is more stable than an ethyl carbocation (which probably wouldn't even form). Anything that would promote the formation of the carbocation, then, should make the molecule a more potent inhibitor--the clearest example is the structure of the carbocation itself. Molecules that can be stabilized by resonance when dissociated should be even better at aging, although I'm reluctant to say that because the Phenyl group confers lower toxicity than expected (however, I'm curious now about a Benzyl substituent.)

EDIT: I have obtained the Millard article, but the PDF is "damaged" and will not open. I cannot save the HTML file, so I don't think I'll be able to provide it for you. However, using the old knowledge that whenever a journal article is published, it relies on information from other research groups' publications, which in turn rely on more research groups' publications...I should be able to find some big-wig scientists in this field rather quickly. Already I'm finding Physical Organic data on this topic, which pleases me greatly!

EDIT 2: I have obtained an undamaged Millard article in addition to a few others. I will attach them in the following post--to summarize them, it is "nerve agent reactions on the atomic scale." If you enjoyed mechanisms in Organic Chem, you'll love this stuff!
[Edited on 10-13-09 by DDTea]

[Edited on 10-13-09 by DDTea]

Articles from Biochemistry

DDTea - 13-10-2009 at 04:59

Here are the articles I pulled from Biochemistry. The range of dates is from ~1990-present.

Attachment: A. Hornberg et al - Acyl Pocket Modulates Aging Reaction by Precluding Formation of Trig Bipyram Transition State.pdf (546kB)
This file has been downloaded 880 times

Attachment: CB Millard - Rxn Products of AChE and VX Reveal Mobile His in Cat Triad.pdf (83kB)
This file has been downloaded 908 times

Attachment: CB Millard et al - Nerve Agent Reaction Products at the Atomic Level.pdf (191kB)
This file has been downloaded 759 times

Attachment: D Barak et al - Evidence for P-N Bond Scission in Phosphoroamidate NA Adducts of Human AChE.pdf (115kB)
This file has been downloaded 913 times

Attachment: E Carletti - Aging of Cholinesterases Phosphylated by GA Proceeds Through O-Dealkylation.pdf (1.1MB)
This file has been downloaded 1004 times

Attachment: F Ekstroem et al - Structural Changes of Phe338 and His447 revd by Crystal Structures of GA-inhibited Murine AChE.pdf (365kB)
This file has been downloaded 804 times

DDTea - 13-10-2009 at 14:48

Well, as is clear now, there *have* been some structure-activity relationships done on this topic with lots of kinetics data. I haven't read through all of these 6 journal articles yet, but some interesting tidbits came up, especially in the Hornberg article.

"Aging" is a collective term for a variety of dealkylation processes and rearrangements that occur after the initial binding of the Organophosphate to Ser203. I was partially incorrect in my earlier post where I mentioned the pinacolyl carbocation formed when Soman ages in the AChE active site: the carbocation mechanism is one proposed mechanism of this aging, but it is *not* the only one. As such, the Isopropyl group on Sarin is able to leave, and so is the ethyl group on Tabun. Nevertheless, Soman is unique in that it "ages" very rapidly (0.1 h) in the AChE active site.

If you take a look at the diagrams showing the x-ray crystallography structures, it becomes clear how "tailored" organophosphates are for acetylcholinesterase. The initial binding simulates an intermediate in the normal cholinesterase activity, but locks it in this position (see CB Millard). From here, AChE-NA (nerve agent bound AChE) can undergo spontaneous reactivation by hydrolysis (Hornberg et. al), which is slow an unlikely, or dealkylate, which occurs faster. The phosphonyl oxygen of the dealkylated product, which I will refer to as ANA (Aged Nerve Agent), readily forms hydrogen bonds with nearby residues, namely Gly121, Gly122, and Ala204. Additional coordination is involved with the Oxyanion hole, and some rearrangement of a Histidine residue (see CB Millard - Rxn Products...) occurs depending on the agent. This looks like a stereochemical effect.

I'll take a direct quote from the Hornberg paper to finish this off (this is a lot of new information for me, and I'm trying to digest it all and summarize it in a sensible way):
By comparing the inhibition rate constant of OP
compounds with fluoride as a leaving group and with
different phosphorus substituents, Worek and co-workers
found that butylsarin, cyclosarin, sarin, and soman have
inhibition rate constants 200-4000 times higher than that
of DFP in hAChE (2). The lower inhibition rate constant of
DFP may be due to the distortion of the acyl pocket that is
necessary to accommodate the isopropoxy group of DFP.
The aging reactions of sarin, DFP, and VX proceed through
elimination of an isopropoxy (sarin and DFP) or ethoxy
group (VX), and no structural rearrangement of the remaining
groups is observed.

PainKilla - 13-10-2009 at 19:28

I'm not particularly familiar with the topic, although I did glance at the papers and the 1VXR PDB structure... However, there are some concepts that I think are worth commenting on: how are you defining toxicity - physiologically (such as an LD50) or by in vitro inhibition?

The rate of inhibition of an irreversible inhibitor also seems a little murky in this context - if 'rate of inhibition' is measured by number of binding events per successful decrease in affinity for acetylcholine by the enzyme, for a specific type of molecule for a single enzyme molecule, then in principle any analogue that is present in the 'ideal' geometric and energetic conformation for the 'rate of inhibition' to be equal to 1 (that is, one binding event always leads to inhibition) - then in this context, rate of inhibition *would* determine toxicity (assuming pharmacokinetic properties are favorable/equal across your library of compounds). However, I imagine this is almost always implicit in the inhibition constant, since a very high affinity indicates that in the low number of collisions that occur due to the very low concentration of the ligand, these few collisions STILL result in the enzyme being inhibited - thus this low number of collisions leads to the approach of the 'rate of inhibition' towards 1. I realize this isn't a 'rate' in the typical sense since you aren't defining time, but it's not really necessary to correct for this since it will be quite similarly valid (across this family of small molecules) from a thermodynamic standpoint. This is still in vitro though.

If you are considering physiological toxicity, then what you really 'want' (not sure why you would 'want' a fast-acting irreversible inhibitor, as opposed to a fast-acting competitive inhibitor (or perhaps even more to the point - selective AChR agonist), unless you aren't looking for medicinal purposes of these chemicals) is a compound with ideal pharmacodynamics and a good Ki to boot.

If you want to predict toxicity from a chemical standpoint... well, it's rather difficult. The fine details of rates aren't really clear, and it's not currently feasible to calculate them (accurately), to my knowledge. I think the place to start would be evaluating the free energies of the 'pre'-complexes using classical physics (ie, molecular dynamics). Presumably, the energies of the complex 'right before' the nucleophilic attack would be represented as a minimum in a molecular dynamics simulation, so if this is true, MD may be a viable approach. Unfortunately, I don't think QM/MM methods are computationally feasible (at least for a library, maybe doable for one or two compounds), and then obviously QM methodology would not be appropriate for a dynamic system as such. Compiling all known SAR into a library, running some MD simulations, calculating free energies, and finally examining the specific contributions on a per-residue basis or something like that might work well though. I really don't think an 'amateur' could do this because of the computational cost however (and in fact, I doubt many people in general could, as of several years ago - which is probably why the quantitative aspects are either currently being worked on and in the process of publication, or, simply have not been examined yet due to the computational costs).

[Edited on 14-10-2009 by PainKilla]

Polverone - 13-10-2009 at 21:11

In theory, molecular dynamics will eventually sample all the configuration space of interest. In practice, achieving this sort of convergence, so you can calculate energetics with reasonable confidence, is (as you have said) extremely expensive.

I have a fair amount of experience with MD, but only paper knowledge of virtual screening/docking. It's my understanding that screening/docking usually examines a constrained subset of configurations using molecular mechanics. It doesn't try to get ensemble values. This is much cheaper than trying to do use MD to find binding energies but presumably even less accurate.

The advantage of using quantum methods instead of molecular mechanics is twofold. One, it can be more accurate. Two, it enables you to sidestep the initial parameterization and parameter-validation for your molecules. These two advantages do not often outweigh the enormously increased computational cost, it seems.

In recent versions of AMBER you will find the General Amber Force Field and tools for parameterizing small molecules to use it. This is a force field intended for small molecules, usually 'druglike' compounds, that is supposed to be compatible with the biopolymer force fields AMBER is usually used/known for. In theory, you should be able to (mostly) automatically parameterize your new compound and then simulate interactions with proteins or nucleic acids. If you don't have an automatic parameterization or all-purpose force field to use for the small molecules you're testing, it can be a very tedious and lengthy process to develop the new parameterization.

I think virtual docking/screening is a pretty interesting topic, despite the relatively poor performance of common models. I'm sure you could experiment with it much more easily by using common pharmaceutical receptors and ligands that have already received thorough experimental and/or computational investigation.

PainKilla - 13-10-2009 at 21:44

The biggest problem with using MD is that without specific manipulation you can only hope that the 'pre-reaction state' is the minimal energy state; since MD doesn't treat bond breaking and looking at the crystal structure there look to be several complexed waters that are undoubtedly involved in complicated energetic mediation of the transition states, the hope that an MD simulation will find a minimum that corresponds to a state that you could use for a free-energy calculation is very much a hope indeed. So far as evaluating the free energies of a state to crudely predict affinities, there has been a lot of work (search pubmed for binding free energy, or free energy affinity etc for a number of specific examples). A commonly used approach that I have seen is by using linear interaction energies (LIE) - see: , and solvated interaction energies (SIE) - see: . Both these methods require a fair amount of simulation time in order to ensure an energy convergence... In practice, it might be possible to run on a quad-core, one compound for every 12-24 hours of simulation time (this is referring to just the simulation time, not the construction, analysis etc).

Furthermore, there is not a guarantee that the values you get will even fall into a nicely predictive pattern, especially if the binding pocket or protein is considerably mobile (well, you will probably have to run the simulation for a long period of time then)... You might also examine 'alchemical' ( transformations of your ligand (though you can only (accurately) compare off of a reference skeleton this way), as this should afford a much less computationally expensive analysis.

As Polverone mentioned, there is also the issue of parametrization, although this has become easier (and using AMBER is a suitable approach as well). I think setting up and soon perhaps even running molecular dynamics simulations will be accessible to amateurs - I think that *the* most important part however, is to understand the theory behind the methodology, as it is essentially, quite inaccurate - but still useful! Since most of the commonly used software is open source, pretty much anyone with a bit of time, ambition and intelligence can get a simulation running. In addition, you are lucky to have chosen a system which a crystal structure exists - this *greatly* simplifies the assumptive liberties taken, and also greatly reduces the work involved in generating the structure in the first place.

I don't think, aside from MD, there is any real way of predicting what you want to predict, that is quantitative and in some sense, 'intuitively' useful. A Monte Carlo approach could also work I imagine, but, this is less intuitively obvious, although it does give the same results in principle.

DDTea - 14-10-2009 at 06:56

I suppose it's time to reassess and clarify my goals with this little project.

Generally, I am defining toxicity in vitro as the ability of an agent to reduce AChE activity. That's well and dandy, but I'm also interested in the mechanistic details of how a subset of agents (namely Tabun, DFP, Sarin, Soman, and VX) bind to AChE, undergo their respective "aging" processes, and how they coordinate themselves inside the AChE active site--specifically, what favorable interactions are at play (e.g., hydrogen bonding between the phosphonyl O and Gly121, Gly122, and Ala204). I would also like to understand the reactivation chemistry and, in turn, resistance to reactivation by some OP's.

Using the old adage from Molecular Biology that "structure determines function," I want to investigate what features of OP structures give them subtly different biological effects. Through journal articles of in vitro studies, I would like to understand what on a molecular level makes OP's so frighteningly toxic.

Quantitative data such as the rate of dealkylation once bound or free energy changes would be ideal, as that would allow us to explain why, for example, Soman has a lower LD50 than Sarin when the two molecules are extremely similar in structure and their aged forms are identical. I would like to develop a physical chemical description of the binding and coordination of OP's in AChE, in the same way we can describe, in a physical way, the coiling of DNA. Lofty goals indeed!

PainKilla - 14-10-2009 at 07:36

...explain why, for example, Soman has a lower LD50 than Sarin when the two molecules are extremely similar in structure and their aged forms are identical.

I don't think the molecules are extremely similar at all; Sarin is significantly smaller, so looking for appropriate changes in the free energies (qualitatively) is a good place to start. I think the difference can be rationalized quite easily in this case (from the crystal structures).

Here is a crystal structure for Soman ( - crystallized just recently! And search the PDB databank, there seem to be ones for Sarin too (I don't have time to look at all of them) - the crystal structures are the best you could possibly get. If you haven't already, get PyMOL - I think viewing the crystal structures is the very step; there is no need after all, to do quantitative work if your qualitative prediction always works!

[Edited on 14-10-2009 by PainKilla]

DDTea - 18-10-2009 at 18:18

I'm going to reconsider my approach here. After doing quite a bit of reading, it seems quantitative measures of phosphylation/phosphonylation have been in the literature for quite some time. What is new and emerging in the literature are the crystal structures for AChE and Inhibitor-bound AChE. From studies of the enzyme-inhibitor complex via X-Ray crystallography, Mass Spectrometry, and computational methods (which I have to admit that I'm very unfamiliar with), the mechanism of poisoning has been pretty well described.

For those interested in the gory experimental details, I recommend taking a look at the papers posted above. They describe the phosphonylation and aging reactions for DFP, Soman, Sarin, Tabun, and VX better than I can with any kind of summary--although some reactions are similar, the molecules are inherently different and thus they behave differently in the active site of AChE, even if they produce the same macroscopic results on the organism.

A particularly interesting example is Tabun. Being a chiral compound, its stereochemistry determines its mechanism of "aging" once bound to the active site Serine (Carletti et. al.). Tabun orients itself such that one substituent is in the acyl binding pocket of AChE and the other is in the choline binding pocket of AChE. In the choline binding pocket is a mobile Histidine residue that acts like a sort of swinging arm in normal operation; Tabun locks it into an extreme position where it hydrogen bonds to either N (S-Tabun) or O (R-Tabun). Tabun then ages by deamination (S-Tabun) or dealkylation (R-Tabun). The end result is a salt bridge between His and P-O- along the lines of: Ser-P(O)(R)O-::::H--(Imidazolium)+. The half life of this reaction is 13 hours, and the product is irreversibly inhibited and resistant to reactivation.

I came across an interesting paper today about huprines (specifically Huprine X and Huprine Y) and their potential as "optimal" AChE inhibitors for Alzheimer's treatment; unfortunately I glazed over it and proceeded to a different search without saving it. It would be interesting, in light of recent elucidation of the binding/aging mechanisms of traditional nerve agents, to theoretically "design" an "ideal" OP inhibitor. That is, through the logic of "structure determines function," determine what substituents and leaving groups would make the "scariest" OP.

I suspect this has been done already. However, the research is often classified. Looking at who produced the publications I posted, some are from the notoriously secretive Israel Institute of Biological Research or (the less secretive, but nonetheless creepy) Edgewood Chemical and Biological Center. It's a double-edged sword, really: while knowing the mechanism of action of nerve agents could lead to better medical countermeasures, it could also lead to a refinement in nerve agent design--possibly ushering in a new "generation" of chemical agents--this time, not by accidental discovery, but by deliberate "engineering" of molecules in the same way that pharmaceutical research occurs, perhaps. Am I letting my imagination run too far?

Anyhow, I came across a journal which appears to be public and relevant to this discussion: Journal of Medical Chemical, Biological, and Radiological Defense. Here's an article I just found there: "Brain Damage from Soman-Induced Seizures Is Greatly Exacerbated by Dimethyl sulfoxide (DMSO)..."

The bibliography of articles from the U.S. Army Medical Research Institute of Chemical Defense provides a sort of "summary" into what research is going on these days: