1 Limitations of x-ray crystallography

From the first crystalline structure determination of table salt in 1914; whose structure elucidation proved the existence of ionic compounds (6), single crystal x-ray diffraction (SC-XRD) has been widening our view of the hidden world of molecular structures. Today, SC-XRD continues to be the only structural analysis method that offers direct structural information at the atomic level. As such, this technique has been vital for reliably solving many structures of small molecules such as neurotransmitters, antibiotics and industrial catalysts.

SC-XRD utilises the ability of crystalline atoms to scatter or diffract a beam of incident x-ray into a series of amplified and spatially constrained beams (3). The angles and intensities of these beams can be measured and computationally processed by a crystallographer to produce a 3-dimensional image of the density of electrons in the crystal. Aside from the expertise required to process the reflection data produced, the fundamental requirement of crystals for this technique acts as major limitation, since single molecules scatter the incident x-ray to produce a weak, continuous beam that provides little useful information for analysis. While technological advances in recent decades including highly intense x-ray beams produced by synchrotrons and the development of more powerful algorithms for molecular structure imaging have allowed the size of the crystal required to be increasingly smaller, the need for a crystal has still not been eliminated. This poses a great issue as many target compounds are very difficult to crystallise, thus requiring experienced specialists; while others will simply not crystallise at all. In 2013, a new protocol, later coined the crystalline sponge method (CSM), was reported that attempted to bypasses this intrinsic limitation of the target molecule needing to be crystalline (1).

2 The journey of the crystalline sponge method

Expectations

Fujita and his team described the new method that promised to speed up SC-XRD drastically by eliminating the crystallization step of the target molecule. This was done using porous metal organic frameworks (MOFs) that act as ‘crystalline sponges’. Due to the high molecular recognition capability of their pores, these sponges can absorb target molecules from the sample solution into their pores.  In their study, Fujita and his team used two MOFs synthesised from tris(4-pyridyl)-1,3,5-triazine (TTP, 1) and the appropriate metal salt as their crystalline sponges: {[(Co(NCS)₂)₃(TTP)₄]•x(solvent)}_n (2) and {[(ZnI₂)₃(TTP)₂]•x(solvent)}_n (3). In both complexes, the void spaces showed strong binding properties for incoming guest molecules making them ideal crystalline sponges. The as-synthesized complexes 2 and 3 contained solvents in the void. By soaking the crystals of 2 and 3 in a guest solution, guest molecules slowly penetrate these ‘wet’ cavities by guest exchange, and are concentrated at the molecular-recognition pockets surrounded by TTP. A characteristic of the strong host-guest interaction in the crystals of 2 and 3 lies in panel ligand 1, which attracts various guests onto its electron-deficient Ï€-plane. The slow guest exchange allows for the process to remain under thermodynamic control, rendering the geometry of the included guests to be regularly ordered and well equilibrated, thus making it possible to analyse the accommodated guests by crystallography since the molecular structure of the absorbed guest will be displayed, along with the host framework.

Since theoretically, only one crystal is needed to perform the experiment, Fujita’s team found that even trace sample amounts of the microgram-nanogram scale can be analysed in this protocol. When the team used only 80 ng of guaiazulene guest sample with a crystal of 3 (80 Ã- 80 Ã- 80 μm³), they were surprised to see the guest was still clearly observed. Considering that the experiment was carried out using a laboratory X-ray machine, it seemed promising to accomplish crystallography with synchrotron X-ray experiments even on a mass of <10 ng.

In order to assess the scope of the method, the team carried out blind crystallographic analysis of six appropriate samples (Fig) with only ~5 μg of each sample. In conjunction with mass spectroscopic data, all six structures were correctly assigned, with three of the structures solved solely from the crystallographic data. Additionally, the protocol was successfully used to determine the absolute stereochemistry of santonin 4, an anthelminthic drug bearing four chiral centres. Unlike common absolute structure determinations, this was achieved without the chemical introduction of heavy atoms on the guest skeleton since the host framework contains heavy atoms (Zn and I) that show enhanced anomalous scattering effects. (Expand

The most impressive result of the team’s protocol however was determining the absolute structure of miyakosyne A 5, a scarce natural marine product recently isolated from a marine sponge Petrosia sp. The structure contains three chiral centres on its main alkyl chain, two of which, C3 and C26, had been previously determined to be 3R and 26R respectively. However, since the difference between the two long alkyl groups on C14 is only one methylene unit, determining the absolute configuration at C14 was ineffective by conventional spectroscopic and chemical methods. As the amount of miyakosyne A was very limited, preparation a single crystal for X-ray crystallography would propose a huge challenge. The team were able apply their method to the full characterization of miyakosyne A to determine the absolute configuration at C14 and reported success. For its appraisers, it was this result that made this new protocol transformational (4) and understandably it led to a lot of excitement in the field.

1.3 The Fall

The initial lustre of the protocol was dulled as Fujita and his team published a correction on the initial report later that year (1b). Previously unnoticed ambiguities in the crystallographic data, alongside further investigation of by the team found the initial assignment of stereochemistry at C14 of 5 to have been incorrect. Synthetic studies by the team determined the methyl’s stereochemistry was opposite to the original assignment reported. Poor data quality was concluded to be the cause of this errors.

Additionally, more problems were met as other research groups tried to use the technique in their own labs. Although success with the technique was achieved for simple molecules, in the first few months, other groups found little success with any interesting structures, particularly large molecules or molecules containing alkaline chemical groups (8b). Fujita’s team were able to aid other industrial and academic groups to master the technique in one to two weeks. Additionally, more of the issues in reproducibility were improved by the release of a more detailed report of the method (1c) that described the sponge synthesis, pore-solvent exchange and selection requirements for high quality single crystals for crystallography. However, this did not address the issue of poor data quality that led to the misassignment of 5. Since poor data quality can be attributed to all steps of the CSM, including cystal synthesis, solvent exchange, guest-soaking, data collection and crystallographic refinement of the host-guest complex molecules; in order to move the CSM from the “fascinating idea” phase into becoming the transformational and reliable new technology it was envisioned to be, much work was required to optimise all these steps.

3. {[(ZnI₂)₃(TTP)₂]•x(solvent)}_n: The most successful sponge to date

3.1 Andvantages of {[(ZnI₂)₃(1)₂]•x(solvent)}_n

In their initial paper, Fujita and his team reported using sponges 2 and 3. With further investigation, in the case of complex 2, it was observed that guest molecules absorbed in the sponge were prone to static disorder as they tend to lay on the symmetry elements of the cubic lattice (Fm3m). Additionally, complex 2 was shown to undergo unfavourable transformations when removed from solution (8c). This destabilising transformation, accompanied by a colour change from orange to green, resulted in a semiamorphous solid with a significantly altered coordination environment at the metal centre. As such, the less symmetric (C2/c) complex 3 has been employed as the primary host complex for the crystalline sponge. The versatility of 3 as a crystal sponge stems from several advantages in host-guest complexation in the pores. Firstly, the size of the pores is ideal for accommodating organic molecules of common sizes, while the hydrophobic nature of the pore cavities provides favourable binding of common organic molecules. Also, ligand 1 in the complex offers flat and electron-deficient binding site, appropriate for stacking with aromatic compounds and for CH-Ï€ interactions even with aliphatic compounds (9). Since the I atoms in the ZnI2 are good hydrogen-bond acceptors and the pyridyl protons of the ligand 1 are good hydrogen-bond donors, they provide efficient binding sites through hydrogen-bonding. Finally, the framework of sponge 3 is reatively flexible with the size of the guest not strictly limited to the pore size of the complex. Molecules larger than the portal are often accommodated by expanding the pore size. (1.3)

3.2.1 Synthesis of {[(ZnI₂)₃(1)₂]•x(solvent)}_n and solvent exchange by Fujita method and updated Clardy method

In their investigations, Fujita and co-workers prepared 3 by layering a solution of zinc iodide in methanol onto a denser solution of TTP (1) in nitrobenzene. The solution is allowed to stand for 7 days, over which crystals form at the boundary of the two solvents as they diffuse before dropping to the bottom of the test tube and being isolated by filtration. The as-synthesised crystals contain nitrobenzene molecules in the void spaces. However, since nitrobenzene has a high affinity to the pores, target guests are poorly absorbed into the as-synthesised crystal. As such, a solvent exchange step that replaces nitrobenzene with an inert, noninteractive solvent is required prior to soaking the crystal in the target guest solution.

Cyclohexane can be adopted as the inert solvent, while pentane also proves useful for guest soaking at temperatures below 0^oC. The solvent exchange step is carried out by soaking the crystal in the inert solvent for 7 days at 50^oC. The success of the process can be monitered throughout by observing the disappearance of the signal at 1346 cm^-1 in an Infrared (IR) spectrum, which can be assigned to nitrobenzene. Completion of the process is confirmed by SC-XRD by the presence of ordered cyclohexane molecules in the pores. The sponge may now be used for guest absorption. This solvent exchange process may complicate the refinement of the structure, since some nitrobenzene may reamin within the sponge structure after exchange (Vinogradova et al., 2014). This becomes an issue if the target guest molecule contains cyclohexyl or aromatic rings, as it may be difficult to distinguish the guest from residual solvent, especially if  the site occupancy is low or the data quality is poor. Accompanied with heavy use of crystallographic restraints, this increases the risk of misassignment of the desired guest molecule by using residual solvent electron density. Additionally, if the residual solvent and the guest interact similarly with the host, the likelihood of occupational disorder increases and making structure refinement much more challenging.

Clardy and co-workers later reported a simpler and less timely preparation method for the synthesis of sponge 3 using similar conditions to those reported by Fujita and his team. (5sync) Instead of conducting the layer diffusion step with TTP in nitrobenzene, TTP is dissolved in chloroform. As such, the as-synthesised crystals of sponge 3 contain chloroform in the pores. Since chloroform has a much lower affinity for the solvent pores than nitrobenzene, the solvent exchange step can be omitted and the as-synthesised crystals used immediately. As well as saving 7 days of preparation by omitting the solvent exchange step, this method is also milder as it does not require the crystal to be heated for long periods of time. This reduces the chances of introducing imperfections in the crystal.

This omission also minimises the number of solvents that the crystal is exposed to, reducing issues in structure refinement. Although some CHCl3 might remain within the sponge after guest inclusion, due to its longer C-Cl bond length (∼1.7 Å) and larger Cl electron density, CHCl₃ can still be observed. This greater electron density for CHCl3 exerts a larger influence on the structure factors relative to incorporated guest compared to nitrobenzene, however the benefits of CHCl3 usage override this issue.

In addition to the desired crystals, this preparation method has been found to simultaneously form other crystalline structures. Firstly, a crystalline compound with the formula [{(ZnI2)3(TPT)2·CHCl3}n] (2), having a much smaller pore size has been viewed. Fortunately, this crystalline structure can be easily distinguished from the desired structure from its morphology (Fig). A second undesired crystal has more recently been observed with consistently distinct unit parameter, but having indistinguishable morphology to the desired structure from its morphology (Fig). Both these crystals are believed to form due to slight changes in humidity and temperature as well as variations in mixing in the initial stages of the layering process.  desired crystal. Both these crystals are believed to form due to slight changes in humidity and temperature as well as variations in mixing in the initial stages of the layering process.