Skip to Content

Analytical Chemistry

Crystallography Without Crystallizing: An Update

I wrote here about a very promising X-ray crystallography technique which produces structures of molecules that don’t even have to be crystalline. Soaking a test substance into a metal-organic-framework (MOF) lattice gave enough repeating order that x-ray diffraction was possible.
The most startling part of the paper, other than the concept itself, was the determination of the structure of the natural product miyakosyne A. That one’s not crystalline, and will never be crystalline, but the authors not only got the structure, but were able to assign its absolute stereochemistry. (The crystalline lattice is full of heavy atoms, giving you a chance for anomalous dispersion).
Unfortunately, though, this last part has now been withdrawn. A correction at Nature (as of last week) says that “previously unnoticed ambiguities” in the data, including “non-negligible disorder” in the molecular structure have led to the configuration being wrongly assigned. They say that their further work has demonstrated that they can determine the chemical structure of the compound, but cannot assign its stereochemistry.
The other structures in the paper have not been called into question. And here’s where I’d like to throw things open for discussion This paper has been the subject of a great deal of interest since it came out, and I know of several groups that have been looking into it. It is my understanding that the small molecule structures in the Nature paper can indeed be reproduced. But. . .here we move into unexplored territory. Because if you look at that paper, you’ll note that none of the structures feature basic amines or nitrogen heterocycles, just to pick two common classes of compounds that are of great interest to medicinal chemists and natural products chemists alike. And I have yet to hear of anyone getting this MOF technique to work with any such structures, although I am aware of numerous attempts to do so.
So far, then, the impression I have is that this method is certainly not as general as one might have hoped. I would very much enjoy being wrong about this, because it has great potential. It may be that other MOF structures will prove more versatile, and there are certainly a huge number of possibilities to investigate. But I think that the current method needs a lot more work to extend its usefulness. Anyone with experiences in this area that they would like to share, please add them in the comments.

58 comments on “Crystallography Without Crystallizing: An Update”

  1. Anonymous says:

    In order to get a regular lattice of asymmetric molecules that are aligned, the matrix itself needs to comprise a regular lattice of asymmetric cavities that are aligned: if the cavities are symmetric, then the asymmetric molecules will adopt multiple orientations which are not aligned, and so the x-ray structure will be degenerate. I’m not sure if this is the case with the matrix they used, but does that make sense?

  2. anon the II says:

    Darn! I hate it when this happens.
    You see something really cool and then you find out that the authors were just a little too enthusiastic about their new found abilities. So you have to dial back your enthusiasm.
    The big question is, “How much?”. The answer is that, almost always, you dial it way back past the point where it would have been if the authors had been a bit more careful and a little more conservative about their science.

  3. Anon8 says:

    @ 1..Interesting point. May be it would be equally interesting to probe, the racemic mixture.

  4. Derek Lowe says:

    #1 – that’s a good point, but I think that there’s a way out of it. If there’s some way for the different “chambers” in the lattice to talk to each other, then you could end up with a new crystalline lattice that’s now a big asymmetric unit of it own.
    The communication could be steric, or perhaps electronic – if coordination of a guest molecule somehow make that face of the scaffold less or more attractive to something on the other side, then some self-organization could take place.

  5. David Borhani says:

    That’s a bit of a pity. This is a good example of a case where sharing some amount of data with manuscript reviewers (i.e., the coordinates of the guest molecule) would likely have led to the immediate identification of this mistake.
    @1: The matrix they used has no chiral centers. I’m surprised none of us picked up on this when commenting on the original post. There were hints of something wrong, however; I had noted that “Structure of miyakosyne A: many eclipsed torsions along the alkyl chain, on either side of the central stereocenter.” I guess we now know why that was.
    @4: I don’t believe what you suggest is possible, Derek, in an achiral crystal — for every set of connecting chambers of one chirality, the inversion center in the crystal requires that there must be another set of connecting chambers of the opposite chirality. Recent example demonstrating this point: interaction of a chiral molecule — an antifreeze protein — with achiral ice (hexagonal, space group P63/mmc) is equivalent regardless of which enantiomer of the protein is used: Mirror Image Forms of Snow Flea Antifreeze Protein Prepared by Total Chemical Synthesis Have Identical Antifreeze Activities. This result is equivalent to something very familiar to us all: achiral columns don’t separate enantiomers.

  6. Gris says:

    Someone in our lab tried to reproduce this and could not. It’s the usual nature/science hype machine.

  7. Anonymous says:

    Thanks for the comments on my original post (#1), and surprised that none of the referees picked up on this in the original paper!

  8. leftscienceawhileago says:

    Careful on stereochemistry assertions.
    The MOF technique seemed conceptually similar to lipidic cubic phase…lipidic cubic phase is achiral.

  9. David Borhani says:

    @8: The lipic cubic phase, unperturbed by the addition of a chiral protein, is achiral, as you note; see this introduction to LCP. But, the protein is the controlling factor, here, unlike the small guest molecules held in the (dominant, from a crystallization point of view) achiral MOF lattice. All GPCRs, etc. crystallized using LCP crystallize (of course) in chiral space groups.

  10. leftscienceawhileago says:

    David not sure what you mean by “dominant”?
    I don’t have access right now, but did the original paper solve chiral structures in achiral spacegroups?
    I only meant to say that there is no physical reason why a chiral crystal can’t grow in an achiral lattice.

  11. Anonymous says:

    @10: The only way a chiral crystal can form within an achiral matrix is if the molecules touch each other and form a regular crystal lattice despite the matrix. But in that case, what benefit does the matrix provide? You may as well just grow crystals without the matrix.

  12. David Borhani says:

    @11: Yes, indeed. And the data below show that I was wrong regarding “dominant”: here, the guests do induce a desymmetrization of the host crystal lattice (as in LCP/protein crystals), from achiral (C2/c) to chiral (P21 or C2). The guests apparently knock the original inversion center in the crystals slightly out of kilter. (By the way, sorting out the true space group in such situations can sometimes be tricky.)
    @10: Here’s what they had to say about all this in the Nature article:
    “Surprisingly, we were able to determine the absolute structure of a trace amount (5 µg) of a chiral compound because the heavy atoms (Zn and I) in the host framework (3) showed clear anomalous scattering. As-synthesized coordination network 3 has an achiral space group (C2/c), but when treated with 5 mg of santonin (8), an anthelminthic drug bearing four chiral centres31, the space group changed to chiral P21 owing to the enclathration of the chiral guest in the pores. The refined absolute structure was validated by the Flack parameter (0.092(18); parenthetical error, s.d.) and the stereochemistry of 8 was unambiguously determined (Fig. 4).”
    Crystallographic data for 3·8:
    Monoclinic, space group, P21, a = 32.866(5), b = 14.853(2), c = 34.850(6) Å, beta = 105.848(2)º, V = 16366(5) Å^3, T = 90 K, Z = 2, 48049 unique reflections out of 60731 with I >2sigma(I), 3187 parameters, 0.75 C2 space group and revealed all the stereochemistry of 12.”
    Crystallographic data for 3·12:
    Monoclinic, space group, C2, a = 34.804(3), b = 14.8607(11), c = 31.624(2) Å, beta = 102.0830(10)º, V = 15994(2) Å^3, T = 90 K, Z = 4, 11126 unique reflections out of 14697 with I > 2 sigma(I), 1350 parameters, 1.20 3 was described by Biradha & Fujita (2002) Ang. Chem. Int. Ed. 41:3392 (in which it is compound 2). “CCDC-187829(2) … Crystal data for 2: Monoclinic, C2/c, a = 36.079(10), b = 14.978(4), c = 30.734(9) Å, beta = 102.470(2)8, V = 16 217(12) Å^3, Z = 8, Dc = 1.851 g cm^-3, 9066 reflections out of 14243 unique reflections with I>2sigma(I), 1.16

  13. David Borhani says:

    The symbols in the original text is giving the HTML conniption fits. Here’s the end of each of those lines (with the CCDC codes, in case you want to look at the structures)…
    3.8 :0.75

  14. David Borhani says:

    I give up…look at the original paper (SI) if you are interested!

  15. leftscienceawhileago says:

    OK, so for the benefit of the other readers, an achiral “host” lattice does not preclude the possibility of facilitating the crystallization of chiral guests within it (though it should facilitate both enantiomers equally).
    Nothing in the original paper is suspicious wrt to space group assignment and chirality.

  16. Anonymous says:

    @15: Except for the fact that the compound doesn’t crystallise on its own, so why should it with the matrix?

  17. Rhenium says:

    @6 Gris
    Can you tell us anything about why it didn’t work as reported? That might save others of us from barking up a bad tree…

  18. Hap says:

    16: In a crystal of X, the molecules of X only interact with other molecules of X, while as inclusions in something else they rely instead on interactions with the molecules/lattice of something else. It’s possible that they could interact more strongly with some (functional) group in the lattice than they do with other molecules of their own kind.

  19. paperclip says:

    My extremely talented coworker has been trying to get this stuff to work for weeks. Apparently there’s some issue where you have your crystals in a liquid, but then you have to switch liquids, and at this point the crystals have been falling apart. A limit that I learned from him is that your material has to be soluble in dichloromethane. (And, of course, the compound that I’m working with that I can’t crystallize or solve by NMR is polar. Grrrr…)

  20. Anonymous says:

    @18 @ others: Agreed, but this is getting quite hypothetical and unlikely. Surely it would be a lot easier to ensure the best possible results with a chiral lattice, rather than hope for some very special interactions that would be very dependent on the molecule?

  21. leftscienceawhileago says:

    No, there is no fundamental physical reason why the best results would come from a chiral host lattice. All crystals are composed of “very special interactions”, a wonder we get as much to crystallize as we do.
    Again, look at LCP and this paper (in spite of the correction).

  22. HBA says:

    Had a student trying this, but ran into the same problems as #19 mentioned.

  23. David Borhani says:

    @19,22: The 3 crystals were soaked in cyclohexane for 1 week at 45 oC, to exchange the originally included solvent, nitrobenzene. Is this where your crystals are falling apart? Then, a very small amount of the (CH2Cl2, but I would think it could be anything relatively non-polar that doesn’t disrupt the MOF) solution of guest was added. Falling apart at this stage?

  24. Anonymous says:

    “No, there is no fundamental physical reason why the best results would come from a chiral host lattice”
    Yes there is: Asymmetric molecules are more likely to align in a unique way within an asymmetric environment.

  25. paperclip says:

    @David Borhani, It was the cyclohexane step that my coworker mentioned was a problem.

  26. David Borhani says:

    @paperclip, I assume the MOF is characterized to be the correct MOF (i.e., the one they used)? The described procedure certainly seems unambiguous.

  27. paperclip says:

    @David Borhani, My understanding is that he followed the paper’s procedures on the zinc complex.

  28. leftscienceawhileago says:

    That is a bold declaration about the statistics of irregular molecular surfaces with electrostatic/quantum mechanical/relativistic attributes. I look forward to the proof!

  29. Anonymous says:

    Bold declaration? It’s pure logic!

  30. Anonymous says:

    PS. The “proof” is in the fact that it’s easier for crystals to grow once seeded, than to form in the first place, because the seed provides the asymmetric environment.

  31. Hap says:

    Since that’s true of achiral molecules as well, which wouldn’t require a chiral environment to crystallize, that doesn’t support your contention.

  32. Anonymous says:

    Would crystals of asymmetric molecules grow from seeds of symmetric molecules, or vice versa?

  33. Anonymous says:

    By the way, couldn’t the surface of a crystal (or seed) formed from symmetric molecules be asymmetric, depending on the arrangement of molecules at the surface? For example, the surface of a pure diamond may have asymmetric spots, even though the internal molecular structure is symmetric.

  34. Anonymous says:

    I attended a presentation by the PI a few months ago. He said (tongue in cheek?!?) that only one person in his (gigantic) group could make the method work, and once he left it would be over. Assuming there was a grain of reality in that comment, what does it mean? A key step accidentally left out of the published process? Exquisite sensitivity to the local conditions or reagents at that person’s bench, beyond the level of detail normally reported? Or magic? Whatever the deal, the person with the special touch will presumably be the subject of a bidding war by cashed-up MOF groups…

  35. Anonymous says:

    Oops. Here goes Fujita’s Nobel nod,

  36. Canageek says:

    I really want this to work, so that I can get crystal structures of my stuff without trying every sol event in the glove box, and even get crystal structures of the things that don’t work in them. However, I’m not sure how well this method would work for highly reactive uranium organometallics, or if we could adapt a larger version of the MOF or something like that.

  37. HBA says:

    @David Borhani: we did not consider doing the cyclohexane soaking, since out compounds do not dissolve in this solvent. We have to use lower alcohols, DMF, NMP, DMSO and the likes. We tried the mixture from which the MOFs were formed, but our compounds precipitated in those.
    For the fun of the student, we switched to sth that was suitable with the MOFs and we got some sweet inclusion X-tals, even though the outcome was not what we hoped for (read: unpublishable).

  38. Rhenium says:

    Thanks for the comment, I think I’ll leave this one in the “fuggedaboutit” lab drawer…
    Maybe “Nandeyanen” would be more appropriate.

  39. Haftime says:

    If you’re getting issues with the MOF dying, consider using a MOF that is chemically stable. It’s moderately depressing to see people trying complex chemistry on crystals known to be very water sensitive (e.g. MOF-5), and increasing the difficulty of their problem by a huge factor.
    Obviously in this case you have relatively specific requirements for pore dimensions, so you might have to try something less robust. And replication, is of course, a good in of itself.
    On the crystallisation of chiral molecules in a guest framework:
    A pure enantiomer can only crystallise in a chiral space group (it’s impossible to stack them in such a way as to introduce an improper symmetry element). This is equally true in a for an ordered guest in a host. It could be that an asymmetric host would provoke more specific binding – allowing for better ordered guests. But that’s not a given, Mr. 30.
    As 33 suggested, growth surfaces of crystals are often chiral (i.e. screw dislocations).

  40. David Borhani says:

    @37, HBA: Maybe try a trick that sometimes works for protein crystallography (for getting a compound of interest to soak into the crystals that lack any ligands). Soak the (cyclohexane’d) MOF in an solvent of medium polarity *in which your compound of interest is “insoluble”*. Some small amount of compound will, of course, dissolve, and slowly partition into the MOF. Warmer is probably better (as warm as the MOF can tolerate). And more polar a solvent, within reason, is also presumably better (maybe even nitrobenzene or nitromethane might work out, if your compound has high enough affinity for the MOF site).

  41. Anonymous says:

    @39: A symmetric (achiral) host lattice could never ensure better alignment of an asymetric (chiral) guest molecule than an asymmetric (chiral) host lattice. Period.
    Otherwise, give me just one example. Any example.

  42. leftscienceawhileago says:

    41 (assuming you are 29. and 30),
    Your position just doesn’t make any sense; I encourage you to think about it critically.
    You are trying to make a universal prediction about the rather complex phenomenon of crystallization (something we generally have a difficult time predicting things about).
    To prove your assertion you would have to (accurately (!)) simulate crystallization across all (or at least a representative sample) pairs of host lattice and guests and compare achiral vs. chiral. Certainly not a feasible experiment.
    Given the number of parameters involved in crystallization, I doubt many people would believe chiral host lattices would have a systematic effect on the very-black-magiky phenomenon known as crystallization.
    I’ve already pointed to achiral host lattices (LCP and this paper). These seem to better (in many cases) than crystallizing the (chiral) guest on its own.
    Not sure if there is much else I can do to convince you.

  43. Anonymous says:

    @42: You can theorize and rationalize all you like, but I can’t prove my theory because science is about disproving theories. That’s why I asked you to provide just one example where an achiral lattice is better than a chiral lattice at crystalising a chiral molecule. However the examples you provide do not compare chiral vs achiral lattices in their ability to seed the crystallization of chiral guest molecules.
    Perhaps this experiment has never been done, but then still my theory stands until it is disproven by example…

  44. leftscienceawhileago says:

    43. I have a unicorn in my pocket. It is very small, much smaller than what could be discovered by any tool in existence or any that you could imagine. Don’t be bothered by the fact that no one has any reason to believe my little unicorn exists.
    Would you say that my assertion is credible?

  45. Anonymous says:

    I would say that you are evading the opportunity to give an example.

  46. leftscienceawhileago says:

    No one is evading since there is no opportunity here. I sincerely doubt you’ve ever grown a crystal of something and gotten a structure. A little bit of experience (which might not be accessible to you) would help you understand things a bit better.
    You’ve gone from your assertion to being something that is “pure logic”, to something that is true since no one is providing you with counter examples.
    Your use of “better” is terribly imprecise. What makes a “better” lattice? Better chance of getting crystals? Bigger crystals? Crystals that are more stable? Crystals that grow faster? Crystals with fewer defects? Crystals that aren’t twinned?
    Do you see now how your assertion is meaningless?

  47. Hap says:

    It’s generally up to those who propose a theory to show that it fits the data – while the theory can’t be proven, it can be shown to fit the available data. Science does not have the default assumption of “you haven’t disproven this, so it must be true”, but “this theory is not helpful, but nobody’s disproven it yet, so I can’t yet stake its heart and burn its body”. Since you haven’t shown either a theoretical framework or evidence that things must crystallize better in nonracemic environments than racemic ones, “theory” is probably generous.
    Argumentum ex recto isn’t even good enough to be a logical fallacy.

  48. Anonymous says:

    @47: True, currently the theory is based only on “common sense” and reasoning, since we don’t seem to have any relevant experimental data.
    The reasoning is simple: symmetric lattices always have multiple equivalent binding sites which are not aligned, whereas asymmetric lattices have only unique binding sites that are all aligned.
    That’s the nature of symmetry. Duh.

  49. Anonymous says:

    PS. Here’s another simple argument:
    If you combine (superimpose) two crystal lattices (host lattice and guest molecule), the resulting hybrid crystal will always have the lowest order of symmetry. In other words, symmetry of the hybrid lattice can never be greater than that of the host lattice alone.
    Again, this is common sense.

  50. leftscienceawhileago says:

    49 Your arguments do nothing but tell me you don’t know anything about crystals,symmetry or physics.
    Why not design the experiment you would use to test your assertion? What molecules would you use and how would you measure the outcome?

  51. Hap says:

    Why do racemates crystallize? Do they crystallize worse than the corresponding enantiomers (lower solubility, amorphous form rather than crystal, etc)? (They may not crystallize in the same habit or take up as much of the available space, but those things aren’t required for the compound to be included in the lattice of itself or another compound).
    Perhaps you should read the epigraph to The Baroque Cycle before you comment again.

  52. David Borhani says:

    I suspect a “resolution” to this argument may lie in the copious data and thoughtful analyses provided by Jean Jacques and co-authors in their classic text Enantiomers, Racemates, and Resolutions.

  53. Anonymous says:

    “49 Your arguments do nothing but tell me you don’t know anything about crystals,symmetry or physics.”
    I got my PhD in X-ray protein crystallography from Cambridge, working with Nobel laureate Max Perutz who pioneered the field.
    But to settle the argument, please point me to any two theoreticeal symmetry groups which could be combined (superimposed) to give higher order symmetry than either of the original two groups.

  54. David Borhani says:

    @53: I believe one could combine two P21 forms (say, of one enantiomer, and of the other) to get a P21/c form of higher order symmetry. In other words, each enantiomer crystallizes in P21, but the racemate crystallizes in P21/c. I think this may be a simple example to address your question.
    I may be missing your point, however. You seem to be arguing that higher-order symmetry is equivalent to greater stability (of the entire crystal, host + guest; or perhaps of the guest only). I suspect that the two are connected, but it’s not clear to me that one (higher-order symmetry) always implies the other (greater stability).
    Furthermore, importantly, many crystals do not represent the most stable arrangement possible, and yet they do form. This common occurrence renders stability arguments a bit moot, doesn’t it?

  55. Anonymous says:

    @54: That’s actually a great example, but it is an extremely special case, not relevant here, because it depends on the molecules forming the two crystal lattices being exact mirror images of each other (i.e., enantiomers). Clearly it’s not relevant in this case, because the guest molecule is not just an enantiomer of the lattice molecule. But great example anyway, which only supports that it takes a very special requirement to get higher symmetry by combining two groups of lower symmetry.
    Binding stability is only relevant in that it ultimately translates to a real crystal lattice, so whatever rules apply to any theoretical end state (lattice) must also apply to any binding equilibrium.

  56. leftscienceawhileago says:

    55. You are not being at all clear. Your earlier messages were talking about “better lattices”, then something about seeding and now spacegroups.
    Your most recent statement is plainly true if I word it a little more precisely: Two lattices of different colors can’t be superimposed to introduce new symmetry elements.
    This isn’t what you were talking about earlier, and I’m not sure how this relates to “better” (what does that mean?) lattices and chirality.
    Proposing a simple experiment might make your point clearer , but I suspect you are merely making a simple logical error.
    Have fun!

  57. Anonymous says:

    @55: I think the argument has been quite clear:
    If you don’t want your asymmetric guest molecule to occupy alternative orientations in each site, make sure the binding cavities in your host lattice are asymmetric, so that the binding energies of the guest molecule in these alternative orientations are non-equivalent and favour only a single unique orientation.

  58. Karlos Danger says:

    Huh, I just noticed something weird! Download the cif file for miyakosyne A. Chiral molecules modeled correctly during refinement should give a Flack parameter very close to 0 with a LOW parenthetical error. Flack parameter of 1 indicates the chirality is inverted. A value of 0.5 indicates that it’s the racemate.
    For their santonin example, they got a Flack parameter of 0.092(18) or 0.092 +- 0.018, which is a good value. Their value for miyakosyne A, with parenthetical error, is 0 (10). This is the equivalent of saying 0 +- 1. So the value ranges from “chiral” to “racemate” to “inverted chirality”. So their Flack parameter for absolute stereochemical determination of the molecule is meaningless.
    I’m surprised that the authors tried to draw conclusions on miyakosyne A absolute stereochemistry in the first place IMHO… maybe they were desperate to get at something??

Comments are closed.