Skip to main content

Analytical Chemistry

X-Ray Structures Of Everything. Without Crystals. Holy Cow.

There’s an absolutely startling new paper out from Makoto Fujita and co-workers at the University of Tokyo. I’ve written a number of times here about X-ray crystallography, which can be the most powerful tool available for solving the structures of both large and small molecules – if you can get a crystal, and if that crystal is good enough. Advances in X-ray source brightness, in detectors, and in sheer computational power have all advanced the field far beyond what Sir Lawrence Bragg could have imagined. But you still need a crystal.
Maybe not any more, you don’t. This latest paper demonstrates that if you soak a solution of some small molecule in a bit of crystalline porous “molecular sponge”, you can get the x-ray structure of the whole complex, small molecules and all. If you’re not a chemist you might not feel the full effect of that statement, but so far, every chemist I’ve tried it out on has reacted with raised eyebrows, disbelief, and sometimes a four-letter exclamation for good measure. The idea that you can turn around and get a solid X-ray structure of a compound after having soaked it with a tiny piece of crystalline stuff is going to take some getting used to, but I think we’ll manage.
santonin latticesantonin[1]
The crystalline stuff in question turns out to be two complexes with tris(4-pyridyl)triazine and either cobalt isothiocyanate or zinc iodide. These form large cage-like structures in the solid state, with rather different forms, but each of them seems to be able to pick up small molecules and hold them in a repeating, defined orientation. Shown is a lattice of santonin molecules in the molecular cage, to give you the idea.
Just as impressive is the scale that this technique works on. They demonstrate that by solving the structure of a marine natural product, miyakosyne A, using a 5-microgram sample. I might add that its structure certainly does not look like something that is likely to crystallize easily on its own, and indeed, no crystal is known. By measuring the amount of absorbed material in other examples and extrapolating down to their X-ray sample size, the authors estimate that they can get a structure on as little as 80 nanograms of actual compound. Holy crap.
Not content with this, the paper goes on to show how this method can be applied to give a completely new form of analysis: LC/SCD. Yes, that means what it says – they show that you can run an HPLC separation on a mixture, dip bits of the molecular sponge in the fractions, and get (if you are so inclined) X-ray structures of everything that comes off your column. Now, this is not going to be a walk-up technique any time soon. You still need a fine source of X-rays, plenty of computational resources, and so on. But just the idea that this is possible makes me feel as if I’m reading science fiction. If this is as robust as it looks like, the entire field of natural product structure determination has just ended.
Here’s a comment in the same issue of Nature from Pierre Stallforth and Jon Clardy, whose opinions on X-ray crystallography are taken seriously by anyone who knows anything about the field. This new work is described as “breathtakingly simple”, and furthermore, that “One can even imagine that, in the near future, researchers will not bother trying to crystallize new molecules”. Indeed one can.
I would guess that there are many more refinements to be made in what sorts of host frameworks are used – different ones are likely to be effective for different classes of compounds. A number of very interesting extensions to this idea are occurring to me right now, and I’m sure that’ll be true for a lot of the people who will read it. But for now, what’s in this paper is plenty. Nobel prizes have been given for less. Sir Lawrence Bragg, were he with us, would stand up and lead the applause himself.
Update: as those of you reading up on this have discovered by now, the literature on metal-organic frameworks (MOFs) is large and growing. But I wanted to highlight this recent report of one with pore large enough for actual proteins to enter. Will they?
And here’s more on the story from Nature News.

40 comments on “X-Ray Structures Of Everything. Without Crystals. Holy Cow.”

  1. Jon says:

    Holy crap.
    This seems to be simple, elegant and powerful.
    I’d take off my hat if I had one.

  2. eugene says:

    This might start a whole new field in making crystalline templates. Things that will allow for crystallizing only a part of a mixture based on pore size and excluding other stuff (or maybe by sampling different parts of a crystal due to differing diffusion rates), or frameworks that can stabilize unstable compounds, such as something that would normally fall apart in a few hours or is only stable inside a solvent matrix.

  3. road says:

    Now we just need a molecular sponge that can soak up proteins!

  4. Boo says:

    I keep looking for the April Fools punchline!

  5. NoDrugsNoJobs says:

    Thanks Derek, I learn so many things on your blog. This is really amazing!

  6. anon says:

    Really amazing stuff!
    One has to wonder if this can be done with mixtures. If so the study of reaction mixtures with this technique would redefine the way we look at new synthetic methodologies and reaction mechanisms.

  7. anon the II says:

    This is really cool and represents the kind of perpendicular thinking that occasionally makes science nifty and fun to be a part of. I guess this is really the same experiment you do when you soak a small molecule into a protein. You reuse the same protein phasing model and solve for the difference. The real intellectual leap is to realize that you could replace the protein with a much more general and well behaved lattice and get better resolution with lower compound requirements. The really cool thing is that almost instantly, you can think of all kinds of neat experiments to do. And you wonder why someone didn’t think of this before.

  8. JAB says:

    Wow! Just wow!

  9. eugene says:

    “And you wonder why someone didn’t think of this before.”
    Oh, there have been some hints in the last few years. With a few groups working with organometallic complexes and their networks and looking at water and how it can leave under vacuum and enter back in again. They just got crystal structures of water though. Then there was an earlier Nature paper by Brookhart et al where a pincer complex crystal was exposed to different gases and the gases coordinated and displaced the previous gaseous molecule, then an X-ray of the crystal was taken again with the new gas molecule coordinated.
    I wouldn’t be surprised if the authors of the current study haven’t read any of those previous reports. Still, this is very different.

  10. bonedriven says:

    I’m a actually a little speechless. LC/SCD?! I… wow.

  11. LittleGreenPills says:

    I feel like a horse and buggy maker that just saw a car whiz by.

  12. How general and versatile is this in terms of the kinds of compounds which can be studied?

  13. Old Lab Rat says:

    Where to start? Imagine the chiral LC/SCD application. Synthesize the racemic mixture, chiral column on a small sample, (relatively) quick determination of the active stereo-isomer.
    I haven’t yet read the paper and was wondering if the authors compared the results with existing small molecule crystals.

  14. ScientistSailor says:

    Truly amazing.
    But to claim that “If this is as robust as it looks like, the entire field of natural product structure determination has just ended.” is a bit much. Remember when resin chemistry meant the end of the round-bottom flask.
    Take a look at the batzelladine family of natural products, two tricyclic guanidines separated by long aliphatic chains. Even soaked into a sponge, I bet you won’t get a good structure…

  15. Derek Lowe says:

    #12 Wavefunction – that’s the question, isn’t it? I’d say right now that no one knows yet. But the paper itself shows a variety of structures – aryl, cycloalkyl, numerous different functional groups. So even as it stands, it seems to accommodate a number of chemotypes. And that marine natural product structure – if that works, a lot of bizarre stuff will.
    But you could also imagine tuning that tris-pyridylpyrazine part of the structure and getting much different behavior. This is going to keep people busy for a long time.
    And it’s that idea in comment #3 that I imagine is also occurring to many readers of the paper as well. . .

  16. David Borhani says:

    #3 road – We would need HUGE voids, ~10x larger than what has been arduously created over the past few decades.
    #12 Wavefunction – The blind tests are interesting, and for several of the compounds the structure came only with MS data (which, of course, one would have and should use). But, it speaks toward the quality of the electron density maps — which is generally somewhat dodgy (see the SI figures). It will get better, in part through use of other hosts.
    I am curious what the temperature factors (“B-factors”) of the guests look like (need to retrieve the cif files from the CDCC, because the authors don’t discuss this seemingly important point). What is the “equivalent resolution” of these guest structures (the refinement statistics are dominated, of course, by the host, and really tell one next to nothing about the guest)? How does equivalent resolution vary with guest chemical structure? How good is good enough to determine the structure of a unknown, something a bit more complicated than miyakosyne A?

  17. SmallMoleculeCrystallographer says:

    There are a number of somewhat simpler variation of this that have been reported over the last (roughly) 10 years, although these have mostly looked at solvent and/or anion exchange in coordination polymeric materials. There have also been various attempts to get non-templating molecules crystallized inside the pores/cavities of such compounds with varying degrees of success – both by co-crystallization, and by soaking framework crystals in a solution of a molecule of interest.
    Clearly pore-size, its relation to the size of your molecules of interest, and the interactions it can form either to itself or to the framework, are all going to have influence on whether the molecule of interest is ordered and discernible in your X-ray structure, or highly disordered and not visible. We had a structure show this a couple of years ago. A large-pore framework structure would take up various dye molecules (either by co-crystallization, or by soaking a formed framework) to the extent that the crystals were visibly strongly colored, and could be cut up or crushed and the color seen to extend all the way through the crystals. However in none of the structures were the dye molecules structurally identifiable, even when the rest of the framework was. There was certainly a lot of residual electron density visible in the pores, but not anything you could identify structurally!

  18. neo says:

    Certainly looks like a game-changer, and I hope it lives up to its apparent potential.
    How easily could these results have been faked? How soon will the results be duplicated in other labs?

  19. RB Woodweird says:

    #3 and #16 – If the protein is a receptor, perhaps capture its ligand, then let the ligand bind the protein?

  20. Helical_Investor says:

    There is that crazy scene in the Sean Connery movie (Medicine Man 1992) where he inject ‘something’ into a … lets call it a chromatograph … and by clicking on a peak up pops a structure (which he immediately considers to be unsynthesizable). Chemists have mocked this scene forever.
    Now .. maybe not so much.

  21. SmallMoleculeCrystallographer says:

    #16 Various of the figures both in the main paper and the ESI show thermal ellipsoid plots at 50% probability for various of the guests. While some of these clearly show relatively large and sometimes wobbly ellipsoids, they mostly look reasonable – the structure of anything – solvent/anion/etc – within framework pores often shows significantly larger elliposids that the framework, as it simply has many more degrees of freedom of motion.

  22. David Borhani says:

    #21 SmallMoleculeCrystallographer – Yes, I agree the ORTEP images look ~OK. They give the (strong) impression of atomicity (atomic resolution). The electron density figures seem to be at variance, however.
    Having looked at one of the cif files (there is no info saying which guest is which file … grrr) — one of the flavanoids — I very slightly adjust what I said about x-ray statistics being completely dominated by the host: they will be only mostly dominated. Here, the host is about 31 “light” atoms (CHO) compared to the unique part of the host, about 99 light atoms (CN) and 18 HEAVY atoms (6 Zn and 12 I). Getting those Zn & I positions/anisotropic displacements right gives you most of the R-factor lowering, with the rest coming from the pyridyltriazene linkers.
    This particular guest has only isotropic B-factors, and it is only at 0.5 occupancy. Those B-factors range from about 10 to 35 A^2, compared to the (anisotropically refined) host atoms which are mostly in the mid single digits.

  23. Chris Swain says:

    I think these guys are trying to do something similar for proteins?

  24. jhb says:

    Comment 14
    Take a look at Figure 6, the batzelladines are long and stringy but I think that the structure of miyakosyne A gives a good idea of what can be done.

  25. Flatland says:

    It seems to be that the power of this technique will be in unambiguously assigning atomic connectivity (provided that the matrix doesn’t catalyze decomposition or rearrangement) and if the temperature factors are good (combined with MS data), the composition. The weakness will be that you will not get accurate geometry (bond angles/lengths and 3d orientations) as this will be perturbed by residing in the matrix and its interactions with the matrix. This will not kill natural product structure determination, but will be just one more tool in the toolbox. Still, very cool.
    PS Id like to see this with neutron diffraction for accurate hydrogen locations.

  26. David Borhani says:

    Structure of miyakosyne A: many eclipsed torsions along the alkyl chain, on either side of the central stereocenter.

  27. MattC says:

    Very neat idea – people have been solving included guests in structures for years now (sorry Derek!) but using it to find the identity of a guest? That’s great thinking.
    Pretty impressed that they managed to get that floppy old molecule to order, too.
    I hope that this idea is generalisable – the real problem will be finding MOFs that will have ordered guests for a wide range of species.
    Flatland – neutron diffraction, whilst it might show the protons, will kind of kill the USP here – you’ll go from needing 80ng to 80mg (at least) of material.

  28. paperclip says:

    Oh man, I work largely with natural products, and I would love to see this in general use. Rather than fearing obsolescence, I’m imagining the possibilities. For many natural product chemists these days, the structure is just the beginning. There is the genetics and biochemistry behind their formation to explore once you know what you’re working with. Every answer just brings up new questions to investigate, so bring on the answers!
    And LC/SCD would be like Christmas every day.

  29. lukas says:

    Very impressive work, no doubt, but the paper leaves you wondering just how many flabby molecules they tried to diffuse into those sponges before finding the only one that… crystallised.
    It seems that molecules that would crystallise well enough on their own do so with little trouble when exposed to the crystalline matrix. For the rest of them, while this method is a valuable addition to the toolbox, it’s still trial and error.

  30. okemist says:

    @13 I would like to comparison to known molecules to see what the influence of the sponge matrix is on the pattern if any?

  31. B says:

    What about the potential for screening neuronal aggregates and determining the structures of small oligomeric species? Only the small oligomers would penetrate the lattice and the large protein aggregates would not. Seletively isolate the “harmful” oligomers and then determine their structure. That could have huge implications for the study of AD and related neurodegenerative diseases…

  32. JBosch says:

    you said it. It’s “just” small molecules but I could see protein nanotubes developing in this direction for protein x-ray crystallography. You tether your protein specifically to the nanotubes and there you go.
    It’s still a very impressive method and great paper.

  33. Bob says:

    I would like to see something like karlotoxin in one of these “sponges”. Really interesting stuff.

  34. CoulombicExplosion says:

    Same thought as 13 & 30. I’d bet that a small molecule the size of miyakosyne A would be relatively unaffected. I’d be more suspicious of deformations of larger molecules, such as proteins. Who’s to say that these won’t adsorb and denature on the surface of these frameworks? Then again, maybe we’ll get lucky.

  35. Anonymous says:

    Agree with #7, Anon the II. The idea is indeed really cool and neat! It instantly triggers people to come up with a lot of follow-up ideas and experiments… That is the beauty. The current lattice may not be the best. But with the POC from the author, it is just a matter of time and efforts to find better ones, including the ones that could fit proteins!
    This reminded me the reason to get into science in the first place years of ago.
    Very cool!

  36. Anonymous BMS Researcher says:

    @Helical_Investor: the old movie scene that immediately popped into my head was this one (of Scotty making a 3-d rotating model of transparent aluminum on a 1980s Apple Mac):

  37. Melissa says:

    This is truly fantastic! I wonder what will come next.

  38. Pamela says:

    In 1978 I did undergrad research work under a Dr. Moore at Penn State in the Biophysics program. We were crystallizing ribosomes in a matrix of cyanide salts (seems similar to the above) on the theory that a crystalline array should orient non-crystalline elements into a pseudocrystal and then using microphotos and lasers to simulate X-ray diffractions because we didn’t have the budget or access to the x-rays. Don’t know if anything ever got published but this isn’t entirely new under the sun.

  39. AnonIII says:

    Since I am writing anonymously, I cannot honestly say that I considered a nearly identical project some years back. And, I am sure, so did a lot of people over the last three decades. The fundamental problem is that any hosts with pores big enough to get your molecule in, often does not provide defined enough interactions so that all molecules sit exactly in the same position. The situation is not much different from co-crystallized solvent, which more often than not is so disordered that you have to impose a known geometry on it.
    That it worked is cool, but it’s impact will depend if it can be generalized. If it works one time out of two great. But if those were the seven examples out of 150 attempts which actually showed localized electron density for the guest…
    And it will never replace protein cyrstallography or small molecule crystallography completely. These structures are actually quite bad (in terms of geometry results). You get connectivity, but you won’t be able to discuss structural details. There are not many cases where small molecules of this size pose a structural problem. Thus again, it remains to be seen if it can be generalized with a high success rate for larger molecules.
    The one really attractive idea is determination of absolute structure, which might attract people who do not normally grow many crystals: synthetic organic chemists.

  40. Anthony says:

    Looking at the downloaded CIF’s it is clear that the guest molecules were fixed in the refinement at occupancy 0.5 instead of the normal 1.0. Likely this occupancy number is even overestimated in view of the high B values for the atoms in the guest molecule model. This would imply that the space not taken by the guest molecule will likely still be filled with solvent molecules. The latter are expected to blur the image of the guest molecule with additional density peaks. Lots of geometry constraints & restraints based on prior information will be needed to keep the refinement of the structure stable. This will also make the identification of the chemistry of an unknown compound in terms of C,N,O etc. very tricky.

Comments are closed.