Skip to Content

Analytical Chemistry

Right Hand, Left Hand, Either Hand

Medicinal chemists like it when one enantiomer of a compound binds to a target much more than the other one. That tells you that you’re getting real binding to a protein target, and the bigger the difference between the two, the more you can say about the actual binding site. But is this always true?

For the non-chemists, an enantiomer is a single right-handed or left-handed form of a drug; think of a glove or a shoe. Not every object (or drug) can come in those mirror-image forms, though – there’s no such thing, clearly, as a right-handed baseball bat or a left-handed drinking glass. The same with drugs – aspirin has no enantiomers, and neither does, say, Claritin (loratadine). When “handed” isomers (enantiomers) exist, though, they can have very different actions in the body, because all the proteins and carbohydrates in living cells are single-handed forms themselves. To add to the fun, some drugs that have enantiomers are sold as pure right- or left-handed forms, while others are sold as a fifty-fifty mix, which is called a racemic mixture (like ibuprofen – in its case, the body quickly turns the pure form back into the mixture anyway, so there’s no point).


Here’s a new paper that shows a particularly strong example of how enantiomers can end up performing the same way. The authors (from Sheffield and Syngenta) were looking at inhibitors of the plant enzyme imidazoleglycerolphosphate-dehydratase (IGPD) as potential new herbicides, and they describe work on triazole phosphonate compounds that mimic a key intermediate in the reaction. So far, so good – this is just the sort of thing that you do when you’re looking for enzyme inhibitors. Their lead compound was a racemate, and when they grew crystals of the IGPD protein in its presence, they obtained a bound structure down to 1.85A resolution.

Another explanation for non-chemists in the crowd: this refers to X-ray crystallography, which can determine (atom by atom) the structure of crystalline compounds. Protein crystals are notoriously tricky to grow, and you generally get somewhat fuzzier data back from them, but if you can get a structure of your target protein with your drug bound to the site of action, you can learn a great deal of information that’s otherwise very difficult to get or just unobtainable. 1.85 Angstrom is pretty decent resolution for a protein crystal, but not enough to really nail down the details of a small molecule – as will become apparent!

At that level, they could see the conformation of the protein chains well in the active site, and also got the triazole and the phosphonate ends of the molecule pretty sharply. But the middle part of the molecule, the part with the chiral center, was fuzzier and lacked electron density, which is a bit odd when the two ends are well-resolved. A different protein construct gave a much better-resolved structure (1.1A), but even this one, at first, looked wonky around the chiral center. Digging into the data, they found that a model with 60/40 occupancy for the S and R enantiomers, respectively, lands right on the electron densities: in other words, either enantiomer is capable of binding well to the target, with no change in the protein structure at all.


Sure enough, the resolved enantiomers were close to margin of error in the enzyme assay, centering around 20 nanomolar, and performed the same in greenhouse trials. Crystallizing either pure enantiomer in the enzyme gave the structures that they’d modeled from the racemate, with only a water molecule or two moving around a bit. There’s just enough room for the two enantiomers to get their phosphonates and triazoles in basically the same spot, while connecting the two differently (there’s a pseudo-mirror-plane arrangement).

When you get fuzzy electron density in a bound-inhibitor study, that can be just plain old disorder, and that’s how it’s often interpreted. But there’s disorder and there’s disorder – as the paper says, if they hadn’t gotten down to 1.1A resolution, they might not have ever figured out what was going on how (or how). Chirally promiscuous enzymes are probably more common than we think, so if you get a flattish ratio when you test your two enantiomers, don’t start to worry: there are more ways for compounds to bind than we think there are. . .

21 comments on “Right Hand, Left Hand, Either Hand”

  1. A Nonny Mouse says:

    …………………however, the metabolism between the enantiomers can vary substantially as we discovered some years ago

  2. RM says:

    You can even get equivalent binding to different *regio*isomers.

    On one compound series I’m looking at currently, you can “flip” the amide. That is, exchange which side the nitrogen is attached to versus the carbonyl. Must mean that there’s not much going on with that region, right? Nope – it forms the core of the protein-binding interface, and there’s key hydrogen bonds to the amide.

    The crystallography indicates that there’s just enough flexibility in the active site to rotate and shift the amide to get reasonable interactions to either direction of the amide. Sort of like the image that’s shown in the post. Luckily for the crystallographers, though, regioisomers are easier to make cleanly than enantiomers.

  3. anon says:

    Fluoxetine (Prozac) is another case where the different enantiomers behaved similarly at their target, though they had some differences in off-target profiles

    1. CMCguy says:

      “some differences in off-target profiles” is want really makes me nervous about dealing with chiral drug molecules as is often easier to determine and understand when examining as specific site for proposed interaction however if have encounter off target adversity it can be difficult to sort out whether due to racemic mix, either due to manufacturing or in situ, or the single isomer was the “wrong choice”. As is medchem doesn’t have enough routine headache inducers.

  4. Romas Kazlauskas says:

    In organic chemistry class, students construct plastic models of enantiomers by exchanging two substituents on the model. Much of the thinking in enantioselectivity continues in this vein – that distinguishing enantiomers involves distinguishing between two substituents.
    X-ray structures of enantiomers bound to enzymes tell a different story (Mezzetti et al., 2005). Most often enantomer binding in an umbrella-like orientation, like the example in this paper. Three substituents remain in the same region, but the fourth substituent, usually hydrogen, inverts to a new location, like an umbrella in a strong wind. In this orientation, three substitutents of both enantiomers stay in the same region. The cited paper includes a review with more than a dozen examples.

    Mezzetti, A., Schrag, J. D., Cheong, C. S., Kazlauskas, R. J. (2005). Mirror-image packing in enantiomer discrimination molecular basis for the enantioselectivity of B. cepacia lipase toward 2-methyl-3-phenyl-1-propanol. Chemistry & Biology, 12, 427–47.

  5. Curious Wavefunction says:

    I have come across my share of cases where a flipped hydroxyl similar to the one in this compound was able to form a similar hydrogen with a backbone carbonyl, so I am not too surprised. So much depends on where exactly in the protein the chiral centers are located that it’s hard to generalize.

    There’s also the interesting case of mirror image packing in catalytic sites:

    1. RING says:

      Very much agree with this comment, it’s very difficult or possibly unfair to generalize. So much depends on structure. For example, what I see in this work is a relatively small and hindered secondary hydroxyl (compared to the phosphate and triazole) which is shielded from stronger interactions within the binding site. It’d be interesting to probe longer chain lengths (ie. ethanol, propanol…), as I’d expect the observed effects to become more pronounced or influential.

      1. Curious Wavefunction says:


  6. Anon says:

    Nice work and paper. Though not particularly surprising to find such examples given that natural selection has not had a chance or need to distinguish between drug enantiomers that do not exist in nature.

  7. David Borhani says:

    Agreed on this being nice work, but not very surprising.

    We once had an chiral nido-carborane dihydrofolate reductase inhibitor, which was crystallized as the racemate. A bit surprisingly to us at the time, both enantiomers bound at ~equal occupancy, and pointed the “hole” of the carborane in the same direction. PDB entry 2C2T. Refinement at 1.5-Å resolution allowed us to clearly posit that the carbon atom on one side of the “hole” needed to be also located at equal occupancy on the other side (i.e., the enantiomer). We never tested the separated enantiomers for differences in activity.

    Reynolds et al. Novel Boron-Containing, Nonclassical Antifolates: Synthesis and Preliminary Biological and Structural Evaluation. (2007) J. Med. Chem. 50:3283.

  8. Another_Anon says:

    It is much more complicated when one enantiomer is agonist and another is antagonist, thus making racemic mixture much less active than any of them.

  9. Magrinho says:

    Molecular semantics? This is always an interesting topic.

    If a chiral center or, maybe more accurately, a methine proton, strongly drives conformational biases, i.e. the molecule’s shape, then one typically sees bigger enantioselectivities. In this case, the ground state conformation(s) are probably similar enough for the enzyme to do its thing on both enantiomers.

  10. Barry says:

    Pfeiffer’s rule is often casually invoked to say that a eudismic ratio not equal to one means that your observed effect is mediated by binding to a target protein (usually invoked in trying to elucidate what exactly inhaled anaesthetics are doing). But what Pfeiffer said was that the eudismic ratio would go up as binding to a target protein approached optimal. If that secondary alcohol isn’t making an important contact, then it’s not necessary to an optimal ligand in this protein, and discriminate binding wouldn’t be expected (even though it’s physically proximate to two tightly bound motifs)

  11. Barry says:

    to conclude that the enzyme here is actually “chirally promiscuous”, I would want to run that dehydrogenase reaction backwards. If it’s truly chirally promiscuous, the alcohol produced would be racemic or close to it. I’d bet that’s not what you’ll observe.

  12. Anon3 says:

    Nice work indeed, but it is a shame that the authors failed to recognize by citation the Blankenfeldt and Breinbauer that appeared in the same journal some years back and detailed a very similar case of both enantiomers binding to an enzyme active site at the same time:

  13. Anon says:

    Looking at the picture, the geometric difference between the two enantiomers is barely more significant than the thermodynamic vibrations of a flat and symmetric (non-chiral) molecule. I would be much more impressed if the groups on the chiral center differed by more than a single atom!

  14. Robert Burns W says:

    Tranylcypromine is an example of an old racemic drug where the two enantiomers are not that different in potency. This makes sense as the drug converts to an achiral reactive intermediate in the active site that covalently modifies the FAD cofactor.
    More intriguing is davunetide, an octapeptide that was in clinical trials for Alzheimer’s. The sequence is NAPVSIPQ = eight chiral centers. Yet, the enantiomer seemed to be equally active although I believe X-ray structures are not available to show why both bind to their target.

  15. Little Geek Girl says:

    Ofcourse it also allows new patents and designs on drugs, certrizine turning into levocertrizine, which in the original efficacy studies wasn’t even compared to certrizine. Also as everyone knows, doctors make terrible chemists and you get ones that don’t get that if someone gets no reaction from certrizine, levocertrizine doesn’t miraculously work.

  16. JH says:

    Thank you for the explanations for the non-chemists!

  17. I’m reminded of the report of “bond stretch isomers” of Mo=O in MoOCl2(PMe2Ph)3, which were reported to have different colors, blue and green. Turned out the green crystal had a bit of an impurity [I think it was MoCl3(PMe2Ph)3 …] which contributed to a different observed electron density in the crystal structure. The impurity did not disrupt the crystal packing, so it just made it look like a different Mo=O bond length. See G. Parkin Acct. Chem. Res. 1992, 25 p.455-460. (linked to my name)

Comments are closed.