Structure and physicochemical characterization of a naproxen–picolinamide cocrystal

The crystal structure is reported of a new 1:1 cocrystal of naproxen with picolinamide, and the pharmaceutically relevant properties are investigated. An NMR crystallography approach is used to distinguish between two crystallographically unique COOH–CONH hydrogen-bonded dimers and to confirm the location of the H atoms in the two dimers.


Supporting information
The histograms shown in Fig. S1 were calculated by in a search of the CSD for the COOH-CONH fragments shown on the right, and the following constraints: 3D coordinates determined, not disordered, no powder structures and only organics. A total of 237 structures were found to contain at least one of these dimers.
Analogous searches for the fragments corresponding to the limiting forms of "Model 1" and "Model 2", see Figure S2, yielded three structures (Jones et al., 2012, Rybarczyk-Pirek, 2012, Eberlin et al., 2013 containing Model 1 and none for Model 2.   at 130 ppm is a little ambiguous due to the overlap of 8 signals, but C14 and C51 are assigned to 130 ppm and 131.4 ppm respectively from the lack of correlation of these peaks with any low shift protons (less than 4 ppm) in the HETCOR spectrum acquired with a long contact time (Fig. S5). These carbon sites are the only sites in the 130 ppm region that are further than 3 Å away from low shift protons.
The order of the pairs of peaks was assigned using the CASTEP-calculated ordering. Other assignments were straight forward. Fig. S5 shows the 1 H -13 C HETCOR spectrum acquired with a short contact time of 100 μs so only correlations between directly bonded atoms are visible. Fig. S5 also reveals differences in 1 H chemical shift between the two crystallographically unique dimers A and B: C4/C39 correlate to H41/H391, which differ in chemical shift by less than 0.5 ppm; C23/C32 correlate to H231/H321, which differ in chemical shift by roughly 1.5 ppm. This is consistent with the packing arrangement of the NPX and PA molecules: H41/H391 are in near identical environments in the two dimers but H231 is packed close to O1 unlike H321 which is near the C52 methyl group.  b Symbols used to indicate the basis of assignment: C = CASTEP-calculated 13 C shielding, Q = 13 C peak in non-quaternary suppression spectrum, S = cross peaks in HETCOR experiment with a short contact time, L = cross peaks in HETCOR experiment with a long contact time, ? = evidence is suggestive rather than definitive.
c Distances to selected non-bonded hydrogen atoms are given up to 2.70 Å.

Figure S9
Comparison of the experimental 1 H spectrum (i) and 13 C spectrum (ii) of NPX-PA and the CASTEP-calculated spectra of NPX-PA from the structure refined from XRD data (green), Model 1 (orange) and Model 2 (red). The calculated spectra were referenced as described previously and modelled with 300 Hz and 50 Hz Lorentzian line broadening for 1 H and 13 C respectively. The predicted 1 H spectra for Model 1 and Model 2 poorly match the experimental 1 H spectrum as the amide protons H202/H291 are predicted to give rise to peaks at 11 ppm in Model 1 and 14 ppm in Model 2 that are not present experimentally.
The 1 H -1 H double quantum-single quantum (DQ/SQ) spectrum was recorded using the back-to-back (BABA) sequence with 8 rotor cycles in the recoupling period and an evolution time of 1 μs. 128 t 1 8 increments were acquired with 16 transients per increment, which were co-added using the States-TPPI method. 200 Hz Gaussian line broadening was applied prior to FT. Simulated DQSQ spectra were produced using MagresView (Sturniolo et al., 2016) and the point sizes were determined by the magnitude of the dipolar coupling, with a cut-off of 3.5 Å. The three simulated spectra were overlaid and then scaled as one to overlay with the experimental spectrum by eye.
The correlation peaks were assigned after measuring the 1 H -1 H proximities in the geometry optimised structure, see Table S4. The CASTEP-calculated DQ/SQ correlations to H31 and H361 for the XRD-refined structure are shown in blue. These predicted DQ/SQ correlations are compatible with the experimental spectrum. It is noted that two strong correlations between H31 and H41 and H251 are not visible experimentally despite the distances being relatively small at 2.56 -2.57 Å. This may be due to dipolar truncation, where strong dipolar coupling between two protons effectively swamps the weaker couplings to more distant neighbouring protons leading to potentially misleading correlation intensities (Bayro et al., 2009, Hodgkinson & Emsley, 1999. Neither Model 1 nor 2 result in a simulated DQ/SQ spectrum that is compatible with the experimental spectrum. In particular the strong correlation to H361 is predicted in Model 1 to be much closer in DQ frequency to the strongest H31 correlation than is observed experimentally. There is also no correlation predicted for H361 at a DQ frequency less than 2 ppm, but the H361/H391 correlation is clearly visible experimentally.