Timothy R.
Ramadhar
*a,
Shao-Liang
Zheng
b,
Yu-Sheng
Chen
c and
Jon
Clardy
*a
aDepartment of Biological Chemistry and Molecular Pharmacology, Harvard Medical School, 240 Longwood Avenue, Boston, Massachusetts 02115, USA. E-mail: timothy_ramadhar@hms.harvard.edu; jon_clardy@hms.harvard.edu; Tel: +1 617 432 3801 Tel: +1 617 432 2845
bDepartment of Chemistry and Chemical Biology, Harvard University, 12 Oxford Street, Cambridge, Massachusetts 02138, USA
cChemMatCARS, Center for Advanced Radiation Sources, The University of Chicago c/o Advanced Photon Source, Argonne National Laboratory, 9700 South Cass Avenue, Argonne, Illinois 60439, USA
First published on 14th July 2017
A strategy that leverages solvent effects to noncovalently trap solid and unstable liquid organic compounds within a crystalline sponge ({[(ZnI2)3(tris(4-pyridyl)-1,3,5-triazine)2]·x(CHCl3)}n) in a simple, mild, and efficient fashion for target molecule structure determination via X-ray diffraction is disclosed. Host–guest structures were obtained using third-generation synchrotron radiation, and new beamline hardware allowed rapid data collection in ∼5–24 minutes. This is 40–90% faster than previously reported crystalline sponge synchrotron datasets collected by us, and approximately a 150–720-fold decrease in time versus using a typical in-house diffractometer, effectively enabling the potential for high-throughput analysis. The new target molecule inclusion method using methyl tert-butyl ether (MTBE) solvent was demonstrated by trapping (E)-stilbene, vanillin, 4-(trifluoromethyl)phenyl azide, and (+)-artemisinin (an antimalarial drug). The potential of guests to maximize intermolecular interactions with the crystalline sponge framework at the expense of attenuating intramolecular interactions (e.g., π-conjugation) was observed for (E)-stilbene. Trapping of vanillin and (+)-artemisinin elicited single-crystal-to-single-crystal transformations where space group symmetry reduced from C2/c to P and C2, respectively, and the absolute configuration of (+)-artemisinin was determined through anomalous dispersion.
The guest inclusion method that we previously disclosed involves soaking 1·CHCl3 with neat liquid compound.3 However, the ability to trap amorphous and crystalline solids in a rapid and operationally simple fashion in 1·CHCl3, which is easy and fast to synthesize, would expand the scope of compounds that can be analysed. Inclusion of crystalline solids is important if the native crystals exhibit poor shape (habit), diffract poorly, are prone to problematic twinning, or one wishes to determine absolute configuration of a molecule that does not contain heavy atoms and access to a diffractometer with a CuKα radiation source is unavailable. Furthermore, inclusion of solids allows for a unique opportunity to study reaction mechanisms for those compounds in situ by turning the crystalline sponges into “molecular flasks”.5 Fujita and co-workers have reported a guest inclusion protocol for solid and liquid compounds that uses cyclohexane with up to 10% CH2Cl2, CHCl3, or 1,2-dichloroethane to solubilize microgram quantities of a target compound.6,7 The resulting solution is added to 1·cyclohexane crystals and heated at 50 °C for multiple days to concentrate the target and facilitate guest penetration. While the procedure has been demonstrated by Fujita and co-workers, it is technically challenging to execute and typically requires multiple trials, which if conducted in parallel requires a significant amount of material. With extensive disorder and low occupancies it could also be possible to confuse residual solvent as target molecules if the target contains cyclohexyl rings, which may complicate structure refinement.3 In addition, cyclohexane (polarity index: 0, water insoluble)8 is not an optimal solvent for dissolving a wide array of compounds even if a small quantity of halogenated solvent is added. A recent report by Santarsiero and co-workers describing the fastest synthesis of 1·PhNO2 to date noted the use of n-Bu2O to solvate liquid diisopropylaniline for inclusion.9 While 1·PhNO2 exhibits a limited scope for guest trapping versus1·cyclohexane and 1·CHCl3 due to the affinity of PhNO2 for the host pores, the authors demonstrated that 1·PhNO2 can be converted to 1·cyclohexane and 1·CHCl3. However, like cyclohexane, the polarity of n-Bu2O (polarity index: 1.7, water insoluble)8 is not poised to dissolve a wide range of functionalized compounds. Our attempts to include targets in CHCl3 into 1·CHCl3 have failed likely due to higher solvent polarity (polarity index: 4.4, water solubility: 0.795 g/100 mL)8 that may favour target solvation and disfavour penetration and ordering within 1·CHCl3 (however, Carmalt and co-workers10 recently included naphthalene and anthracene saturated in CHCl3 into 1·CHCl3 over 1 week, presumably facilitated through strong guest π⋯π and CH⋯π interactions with the host).
To increase the scope of our procedure, we sought to find a solvent with a polarity that could strike a balance between compound solubilisation and allowing successful inclusion and ordering within 1·CHCl3. We believe that target solubilisation is important because maximizing target concentration can aid inclusion and identification as shown by Fujita and co-workers for nobiletin.7 We found that MTBE (polarity index: 2.5, water solubility: 5.1 g/100 mL),8 which is chemically tolerant of 1·CHCl3 unlike THF (where the latter presumably interacts with Zn2+ and displaces the tpt ligand), could strike that balance, and is contrary to a point raised by Fujita and co-workers stressing the necessity of using a poor solvent.7 Herein we report the simplest method to date for trapping solid and unstable liquid organic compounds in 1·CHCl3 (a “soak it and forget it” procedure), describe the use of third-generation synchrotron radiation with recent beamline hardware advances to drastically reduce the collection time and obtain high-quality data, demonstrate the method's effectiveness by trapping (E)-stilbene (2), vanillin (3), 4-(trifluoromethyl)phenyl azide (4) and (+)-artemisinin (5), and analyse the inclusion complexes.
Data were processed in the Bruker APEX3 software suite,14 where data integration was performed in SAINT,15 and SADABS16 was used for multi-scan absorption correction. The resolution cut-off for integration was set to 0.84 Å (sin(θ)/λ = 0.6), where I/σ ≥ 3.00 for 1·2 was 0.88 Å, 1·3 was 0.91 Å, 1·4 was 0.87 Å, and 1·5 was 0.97 Å. The data was solved using intrinsic phasing (SHELXTL XT-2014)17 and least-squares refinement on F2 was performed using SHELXL.18 The riding model was used for hydrogen addition, and the Brennan and Cowan anomalous scattering coefficients (f′, f′′, and cross-section in b/atom) were calculated for the irradiation wavelength and applied during refinement. Absolute structure parameters (Flack x, Hooft y, p3(true), p3(false), p3(racemic twin)) for 1·5 were calculated in PLATON/BIJVOETPAIR.19 Molecular visualization was performed in Olex2.20
Specific procedures and guidelines for crystalline sponge system refinement as previously reported were observed.3 As specifically rationalized in our guidelines and in congruence with our previous crystalline sponge structures,3,4 and as recommended by Spek for crystalline sponge systems,21 programs to treat residual density in solvent accessible voids for the reported and deposited structures were not used. For 1·2, there was a two-part disorder of 2 over an inversion center and one molecule of 2 involved in another two-part disorder was found on another inversion center. For these molecules it was necessary to manually insert the symmetry-related atoms and refine using negative SHELXL PART instructions. For all molecules of 2 in the asymmetric unit, it was necessary to use a SHELXL SAME directive with a smaller effective standard deviation (e.s.d.) for 1,2 and 1,3 distances (0.01 and 0.02, respectively) than the default values in order to rectify convergence issues (max. shift and max. shift/e.s.d. not converging to 0) and to deal with a level B checkCIF alert on phenyl C–C distances. For 1·3, it was necessary in some cases to change SHELXL AFIX 137 and AFIX 147 to AFIX 33 and AFIX 83, respectively, to deal with max. shift and max. shift/e.s.d. not converging to 0. For 1·5, disorder and the level of guest inclusion led to the normalized structure factor amplitudes exhibiting an <|E2 − 1|> of 0.899 (centric distribution) for space group determination; thus, it was necessary to force data processing in C2 versus C2/c, C2/m, and Cc. Electron density for what appears to be a heavily disordered molecule of 5 and MTBE was observed in the asymmetric unit; however, it could not be reasonably modelled and was left unmodelled in accordance with our reported guidelines.3
Fig. 2 Asymmetric units for crystal structures of a) 1·2 (CCDC 1545812), b) 1·3 (CCDC 1545813), c) 1·4 (CCDC 1545814), and d) 1·5 (CCDC 1545815). Thermal ellipsoids are drawn at 50% probability. Guest and Zn/I host framework disorder is shown. |
The new procedure was applied to incorporate other solids and liquids into 1·CHCl3. Vanillin (3), a phenolic aldehyde that can be synthesized or isolated from processed Vanilla planifolia and Vanilla tahitensis seed pods and used as a flavouring/fragrance agent or in a popular TLC stain,24,25 was included into 1·CHCl3. Crystals of 1·CHCl3 were soaked in a 0.5 M solution of 3 in MTBE at ambient temperature for 3 days, where the crystals immediately developed a light yellow colour. Inclusion resulted in a single-crystal-to-single-crystal transformation affording a reduction in space group symmetry from C2/c to centrosymmetric triclinic P and different unit cell dimensions (1·3: a = 14.9223(8) Å, b = 18.9078(11) Å, c = 32.5910(18) Å, α = 102.7281(10)°, β = 91.7744(10)°, γ = 110.7963(9)° versus1·CHCl3:§a = 34.655(3) Å, b = 14.7307(14) Å, c = 31.081(3) Å, α = γ = 90°, β = 101.031(2)°) (Fig. 2b). Inspection of 1·3, exhibiting an R1 of 7.18%, revealed the presence of four molecules of 3 in the asymmetric unit with occupancies of 35(1)%, 41(1)%, 49(1)%, and 82(1)%. One molecule of 3 at 82(1)% occupancy had significant π⋯π stacking with a host pyridine ring (centroid⋯centroid: 3.801(8) Å), and its phenolic hydroxyl group hydrogen bonded to the aldehyde group in another molecule of 3 at 49(1)% occupancy to help anchor the latter within the sponge.
Trapping of 4-(trifluoromethyl)phenyl azide (4) in 1·CHCl3 through the new procedure was also successful. While organic azides are important precursors that participate in reactions such as Huisgen cycloadditions/click reactions, Staudinger ligations, Curtius rearrangements, Schmidt rearrangements, and [3,3]-sigmatropic shifts to afford useful molecules such as heterocycles and peptides, they are also high energy compounds that are potentially explosive.26 Azide 4 exists as an oil, however it is available for purchase as a 0.5 M solution in MTBE as a safer alternative for shipment and handling. We included 4 as a 0.5 M solution in MTBE into 1·CHCl3 upon soaking for 2 days at ambient temperature. The crystal structure of complex 1·4 was solved and refined in space group C2/c with an R1 of 5.76% (Fig. 2c). One molecule of 4 with a two-fold rotationally disordered CF3 group at roughly similar occupancy was observed in the asymmetric unit at 66(3)% occupancy, and one MTBE was present at 68(2)% occupancy. The electron density for the azide was unmistakably observed with an N–N–N angle of 170(3)°. Examination of the packing model revealed that the aryl-appended azide N forged an N⋯π contact with a pyridine (N⋯pyridine centroid: 3.86(2) Å), the central azide N formed a contact with an MTBE methoxy group (N⋯HC: 2.58(2) Å) and the terminal azide N formed contacts with the aforementioned MTBE methoxy group (N⋯HC: 2.80(3) Å) and a pyridyl C atom (N⋯C: 3.33(3) Å) (Fig. 4a). The MTBE also forged a CH⋯π interaction with the phenyl ring of 4via its tert-butyl methyl (CH⋯phenyl centroid: 2.939(9) Å), and the host framework established two additional CH⋯π interactions (CH⋯phenyl centroid: 3.06(1) Å, 3.29(1) Å). An alternating series of 4 and MTBE occurs in the channels where a F⋯HC contact between the CF3 of 4 and tert-butyl methyl of MTBE aids in its stabilization (minimum H⋯F: 2.76(5) Å) (Fig. 4b). The MTBE tert-butyl and methoxy groups can also be seen forging ancillary intermolecular contacts with the host. Overall, the success of this trapping experiment allows us to posit that liquid compounds that are unstable27 at high concentration can be included into 1·CHCl3 under mild dilute conditions using MTBE.
Finally, the new inclusion method was used to trap solid (+)-artemisinin (5) inside 1·CHCl3. Artemisinin (qinghaosu) is a tetracyclic sesquiterpene lactone endoperoxide natural product that is produced by sweet wormwood (Artemisia annua) and can be synthesized.28,29 It is used as a drug against Plasmodium falciparum malaria alongside other antimalarial agents (i.e., artemisinin-based combination therapies, ACT). Its use has saved millions of lives, and its discovery by Youyou Tu was part of the subject of the 2015 Nobel Prize in Physiology or Medicine.30 For the guest inclusion experiment, 5 was solubilized in MTBE to 0.07 M, added to 1·CHCl3, and soaking occurred at ambient temperature for 3 days. A single-crystal-to-single-crystal transformation occurred where the space group symmetry lowered from centrosymmetric C2/c to noncentrosymmetric monoclinic C2 – due to necessary destruction of inversion/c-glide symmetry upon chiral guest inclusion – to afford a chiral crystal system, and the largest change in unit cell axis was observed for the c-axis (1·5: c = 35.233(2) Å versus1·CHCl3:§c = 31.081(3) Å). An R1 of 7.00% was obtained after refinement of 1·5, and one molecule of 5 was clearly observed in the asymmetric unit with an occupancy of 93(1)% and one MTBE at 62(2)% occupancy (Fig. 2d). No anisotropic displacement parameter (ADP) restraints/constraints on 5 were required; only one soft C–C pairwise distance restraint was applied. In comparison to a single crystal structure of 5,31 the key peroxide O–O bond lengths and C–O–O–C torsion angles are similar (guest: 1.46(3) Å and 48(2)°, single crystal: 1.469(2) Å and 47.8(2)°). Some interesting interactions were observed in the packing model for 1·5 where the lactone carbonyl oxygen of 5 forged a CO⋯π interaction with a host pyridine (O⋯pyridyl centroid: 3.22(2) Å) and the endocyclic oxygen in the oxepane ring that bears the endoperoxide forged an O⋯π interaction with a host triazine (O⋯triazyl centroid: 2.93(2) Å) (Fig. 5). The endocyclic lactone oxygen appears to form two O⋯π interactions (O⋯pyridyl centroid: 3.75(2) Å, O⋯triazyl centroid: 3.69(2) Å) and an oxepane methylene hydrogen formed a CH⋯π interaction with a pyridine (CH⋯pyridine centroid: 3.246(9) Å). These and other ancillary interactions, especially the first two described O⋯π contacts, play a significant role in anchoring 5 in the sponge. The Flack x and Hooft y absolute configuration parameters32,33 (0.18(3) and 0.15(2) respectively) are larger mainly due to unresolved disorder along with the use of a high-energy beam (30 keV, ∼0.41 Å) that diminishes the magnitude of the f′ and f′′ anomalous scattering coefficients and (to a lesser extent) “bad” reflections. The dextrorotatory absolute configuration of 5 was established through analysis of a three-hypothesis probabilistic model derived through Bayesian statistical analysis on Bijvoet differences using a Student's t-distribution (p3(true) = 1, p3(false) = 0, p3(racemic twin) = 3 × 10−38, Bijvoet pair coverage: 97%, correlation coefficient: 0.999) indicating that the absolute structure should not be inverted.33
Fig. 5 Interactions of 5 with host framework 1 in complex 1·5 as observed in the packing model. Ball-and-stick representation is shown and extraneous atoms have been hidden for clarity. |
Footnotes |
† Electronic supplementary information (ESI) available: Crystallographic data tables, solvent accessible void space analysis. CCDC 1545812–1545815. For ESI and crystallographic data in CIF or other electronic format see DOI: 10.1039/c7ce00885f |
‡ Shutterless detectors do not require closure of the X-ray shutter for signal readout; it is done concurrently with irradiation. Thus, the readout step is eliminated making collection times shorter. |
§ See ref. 3 and CCDC 1007932 for 1·CHCl3. |
This journal is © The Royal Society of Chemistry 2017 |