Crystal structure of a 2:1 piroxicam–gentisic acid co-crystal featuring neutral and zwitterionic piroxicam molecules

A new co-crystal of piroxicam and gentisic acid has been characterized, in which the ratio of piroxicam (one neutral molecule and one zwitterion) to gentisic acid is 2:1.


Chemical context
Piroxicam is a non-steroidal anti-inflammatory drug classified as a BCS Class II drug due to its low aqueous solubility (Amidon et al., 1995;Thayer, 2010). Co-crystallization of an active pharmaceutical ingredient (API) and an FDAapproved counter-ion is a common technique employed to increase the solubility of the API (Trask et al., 2005). In this work, we explored the co-crystallization of piroxicam and gentisic acid. In a previous study, piroxicam was co-crystallized with 23 carboxylic acids yielding 50 co-crystals. From this work, three co-crystals of piroxicam and gentisic acid were identified with Raman spectroscopy, but no crystal structures were reported (Childs & Hardcastle, 2007). In our prior work, we reported the crystal structure of two co-crystals of piroxicam and gentisic acid, one was a 1:1 co-crystal and the second was a solvated co-crystal that incorporated acetone into the crystal in a 1:1:1 molar ratio (Horstman et al., 2015). In this work we describe the crystal structure of a 2:1 piroxicam:gentisic acid co-crystal.

Structural commentary
The asymmetric unit of this co-crystal consists of two piroxicam molecules and one gentisic acid molecule, with all atoms residing on general positions (Fig. 1). One of the piroxicam molecules is neutral and the other is a zwitterion: the two molecules exhibit two different conformations in the crystal structure. In the neutral S2-containing molecule, intramolecular hydrogen bonding exists between the hydroxyl ISSN 2056-9890 proton H8 [HÁ Á ÁA = 1.79 (3) Å ] and the amide oxygen atom O7. In the S2 molecule, free rotation about the C-C bond (C1 and C10) allows a second zwitterionic conformation in which intramolecular hydrogen bonds exist between the amine proton H2 and the enolate oxygen atom O4 [HÁ Á ÁA = 1.85 (2) Å ] and the pyridinium proton H3 and the amide oxygen atom O3 [HÁ Á ÁA = 2.19 (2) Å ]. Further details of the hydrogen bonding are provided in Table 1.
The gentisic acid molecule shows whole molecule disorder over two orientations rotated by approximately 180 in the plane of the aromatic ring with site occupancies of 0.809 (2):0.191 (2). The major orientation participates in intramolecular hydrogen bonding between the O11 hydroxide substituent of the benzene ring and the O9 oxygen atom of the carboxylic acid [HÁ Á ÁA = 1.89 (3) Å ]. The major orientation also participates in intermolecular hydrogen bonding. The minor orientation of the gentisic acid molecule also displays intramolecular hydrogen bonding between the O11B hydroxide substituent and the O9B oxygen atom of the carboxylic acid (HÁ Á ÁA = 1.89 Å ) but does not participate in intermolecular hydrogen bonding.

Supramolecular features
In the crystal, hydrogen bonds (Table 1) between the piroxicam and gentisic acid molecules form hexameric units that propagate along the a-axis direction (Fig. 2). These units pack into layers in the ab plane; the layers stack along the c-axis direction, Fig. 3.
The repeating motif of the hexameric unit is formed by one gentisic acid molecule hydrogen bonded to two piroxicam molecules, one of each conformation (i.e. neutral and zwitterion). The non-zwitterionic form of piroxicam accepts hydrogen bonds from the gentisic acid via O-HÁ Á ÁN bonds between the carboxylic acid and the pyridine ring of piroxicam [HÁ Á ÁA = 1.74 (3) Å ] as well as N-HÁ Á ÁO bonds between the carbonyl oxygen atom of gentisic acid and the amine nitrogen atom of piroxicam [HÁ Á ÁA = 2.25 (2)  The molecular structure of the title co-crystal, showing 35% probability ellipsoids for non-H atoms and spheres of arbitrary size for H atoms. Only the major component of the disordered gentisic acid is shown. Symmetry codes: (i) Àx þ 1; Ày þ 1; Àz þ 1; (ii) x À 1; y; z.

Figure 2
Ball-and-stick model highlighting the hydrogen-bonding network in the title co-crystal. Only the major component of the disordered gentisic acid is shown.

Synthesis and crystallization
Piroxicam (>=98.0%) and gentisic acid (>=98.0%) were used as purchased from Sigma-Aldrich (St. Louis, MO, USA). Acetonitrile (>=99.9%) was used as purchased from Fisher Scientific (Fair Lawn, NJ, USA). A 1:2 molar ratio of piroxicam:gentisic acid was dissolved in acetonitrile. The concentration of piroxicam in acetonitrile was near saturation ($0.034 M). The resulting solution was introduced into a microfluidic platform. The microfluidic platform was a 6 Â 6 array of single microwells ($100 nl) (Horstman et al., 2015). After being filled, the microfluidic platform was placed inside a petri dish and then the petri dish was sealed with parafilm to slow the rate of solvent evaporation. The crystallization solution evaporated over the course of one day, after which crystals were observed via optical microscopy. Specifics of the microfluidic platform fabrication and operation have been previously reported (Horstman et al., 2015). Once crystals were observed, Raman spectroscopy was used to distinguish between crystals. Within one microfluidic chip, three different co-crystals of piroxicam and gentisic acid were observed, two of which had been previously reported (Horstman et al., 2015) and one new solid form, reported here. Once the new co-crystals had been identified, we removed the crystals from the microfluidic platform.

Refinement
Crystal data, data collection and structure refinement details are summarized in Table 2. The gentisic acid molecule shows whole molecule disorder over two sets of sites: the like C-O and C-C distances were restrained to be similar (s.u. 0.01 Å ). Similar displacement amplitudes (s.u. 0.01) were imposed on disordered sites overlapping by less than the sum of van der Waals radii. All O-H and N-H hydrogen atoms were located in the difference map except for those on the minor-disordered component of the gentisic acid. The H atoms located in the difference map were allowed to refine the O-H/N-H bond distances. These H atoms refined to good hydrogen-bonding positions (Hamilton & Ibers, 1968 Ball-and-stick packing diagram of the co-crystal, as viewed approximately down the b axis, highlighting the layers formed by the packing of the hexameric units. Color key: C gray, N blue, O red, and S yellow. H atoms have been omitted for clarity. carrier atom; remaining H-atom U iso were assigned as 1.2 Â carrier U eq . The (112) reflection was omitted from the final refinement because it was partially obscured by the shadow of the beamstop in some orientations. Data collection: APEX2 (Bruker, 2014); cell refinement: SAINT (Bruker, 2014); data reduction: SAINT (Bruker, 2014) and XPREP (Bruker, 2014); program(s) used to solve structure: SHELXS2014-4 (Sheldrick, 2008); program(s) used to refine structure: SHELXL2014-6 (Sheldrick, 2015); molecular graphics: SHELXTL (Sheldrick, 2008), CrystalMaker (CrystalMaker, 1994); software used to prepare material for publication: XCIF (Bruker, 2014).

Special details
Experimental. One distinct cell was identified using APEX2 (Bruker, 2014). Thirty frame series were integrated and filtered for statistical outliers using SAINT (Bruker, 2013) then corrected for absorption by integration using SAINT/SADABS v2014/2 (Bruker, 2014) to sort, merge, and scale the combined data. No decay correction was applied. Geometry. All esds (except the esd in the dihedral angle between two l.s. planes) are estimated using the full covariance matrix. The cell esds are taken into account individually in the estimation of esds in distances, angles and torsion angles; correlations between esds in cell parameters are only used when they are defined by crystal symmetry. An approximate (isotropic) treatment of cell esds is used for estimating esds involving l.s. planes. Refinement. Structure was phased by direct methods (Sheldrick, 2014). Systematic conditions suggested the ambiguous space group. The space group choice was confirmed by successful convergence of the full-matrix least-squares refinement on F 2 . The final map had no significant features. A final analysis of variance between observed and calculated structure factors showed little dependence on amplitude and resolution.