Screening for Cocrystallization Tendency: The Role of Intermolecular

for Science, Technology and Research), 1 Pesek Road, Jurong Island, ...... Daniels , C. R.; Charlton , A. K.; Wold , R. M.; Moreno , R. J.; Acree ...
0 downloads 0 Views 201KB Size
9890

J. Phys. Chem. B 2008, 112, 9890–9895

Screening for Cocrystallization Tendency: The Role of Intermolecular Interactions Guangwen He,*,† Chacko Jacob,† Liangfeng Guo,† Pui Shan Chow,† and Reginald B. H. Tan†,‡ Institute of Chemical & Engineering Sciences, A/STAR (Agency for Science, Technology and Research), 1 Pesek Road, Jurong Island, Singapore 627833, and Department of Chemical & Biomolecular Engineering, National UniVersity of Singapore, 4 Engineering DriVe 4, Singapore 117576 ReceiVed: April 7, 2008; ReVised Manuscript ReceiVed: May 27, 2008

Pharmaceutical cocrystals have rapidly emerged as a new class of API solids with great promise and advantages. Much work has been focused on exploring the crystal engineering and design strategies that facilitate formation of cocrystals of APIs and ligands/cocrystal formers. However, fewer attempts have been made to understand the equilibrium phase behavior and phase transition kinetics of the cocrystallizing solutions. This limited knowledge on the solution physical chemistry often leads to difficulty in screening for potential molecular pairs of API and ligand that form cocrystals effectively. In this study, the long-time self-diffusivities measured using pulsed gradient spin-echo nuclear magnetic resonance (PGSE NMR) are used to characterize the particle interactions in solutions for pharmaceutical cocrystallizing systems. For the pairs of API and ligand that produce cocrystals, the heteromeric attractions between API and ligand are found to be stronger than the homomeric attractions between API molecules and between ligand molecules, suggesting that an energetically favorable condition is induced for the formation of cocrystals. To the best of our knowledge, this is the first report of using the pair contribution of the self-diffusivity as a screening tool for cocrystal formation. I. Introduction Many active pharmaceutical ingredients (APIs) are delivered to patients as solid formulations with an appropriate dosage. Solid forms of APIs are also preferred in handling and storage of the drug products. Understanding and controlling the chemistry and physical properties of the solid APIs is an essential aspect of drug formulation and pharmaceutical manufacturing processes. Thus the development of techniques that facilitate the generation of solid forms and the relevant knowledge base with minimal time, effort and material consumption is an essential requirement in pharmaceutical industry. Among all solid forms, crystals that present the lowest free energy are probably the most well-studied and preferred states. Physicochemical properties of crystalline materials depend largely on the crystal packing that is modulated by noncovalent interactions. These intermolecular interactions are responsible for the formation of molecular systems and the association of such systems into supramolecular structures. As an attempt to modify the properties of crystalline materials, their crystal packing can be altered by changing the internal arrangement of the molecules that involves breaking and forming of noncovalent bonds. This approach, often called crystal engineering, is geared toward modifying the assembly of molecules by manipulating various noncovalent forces such as hydrogen bonding, van der Waals forces, π-stacking and electrostatic interactions.1–3 The advances in crystal engineering remarkably inspire the research and development on the molecular cocrystals. Generally, a cocrystal is broadly defined as “a multiple component crystal formed between compounds that are solid under ambient * Correspondingauthor.Telephone:(65)67963779.E-mail:he_guangwen@ ices.a-star.edu.sg. † Institute of Chemical & Engineering Sciences, A/STAR (Agency for Science, Technology and Research). ‡ Department of Chemical & Biomolecular Engineering, National University of Singapore.

conditions: at least one component is molecular and forms a supramolecular synthon with the remaining components”.4 Pharmaceutical cocrystals have rapidly emerged as a new class of API solids with great promise and advantages, in addition to polymorphs, salts, solvates/hydrates and amorphous solids.2,5 This new branch of APIs has impacted the field of pharmaceutical research by broadening the intrinsic values of APIs in terms of potential expansion of intellectual property6 and polymorphic variety,7 as well as enhancement of physicochemical properties such as solubility, dissolution profile and stability.8 Plausible industrial adaptation of this class of crystalline materials could be made possible in the near future.2 Much work has been focused on exploring the crystal engineering and design strategies that facilitate supramolecular synthesis of new crystalline complexes (cocrystals) of APIs and ligands/cocrystal formers. However, fewer attempts have been made to understand the equilibrium phase behavior and phase transition kinetics of the cocrystallizing solutions.9–13 Higuchi and co-workers studied the solution complexation and revealed different binding tendencies of small organic molecules in aqueous solution that could shed some light in screening for cocrystallization tendency.14,15 Rodrı´guez-Hornedo and coworkers10–13 constructed phase solubility diagrams of cocrystals based on solubility product and solution complexation and used these phase diagrams to explain how supersaturation is created to induce cocrystal formation. Early trials on designing and making cocrystals are mostly restricted to Edisonian approaches or high throughput surveys.7,16 This limited understanding on the solution physical chemistry often leads to difficulty in screening for potential molecular pairs of API and ligand that form cocrystals effectively. Given these limitations and the increasing interest in making cocrystals in the pharmaceutical industry, the proposed work here focuses on the characterization of intermolecular interactions of API and ligand in solution cocrystallization. The selfdiffusivities of various model systems in solutions are measured

10.1021/jp803019m CCC: $40.75  2008 American Chemical Society Published on Web 07/23/2008

Screening for Cocrystallization Tendency

J. Phys. Chem. B, Vol. 112, No. 32, 2008 9891

Figure 1. Long-time self-diffusivities of various molecules in ethanol at 294.9 K measured using PGSE NMR: (a) paracetamol in the presence of 4,4′-bipyridyl; (b) 4,4′-bipyridyl in the presence of paracetamol.

Figure 2. Normalized self-diffusivities of various molecules in ethanol at 294.9 K: (a) paracetamol; (b) 4,4′-bipyridyl; (c) ibuprofen; (d) p-aminobenzoic acid.

using pulsed gradient spin-echo nuclear magnetic resonance (PGSE NMR). The pair contribution of the self-diffusivity is used to probe the particle interactions in solutions. The relative magnitude of the pair contributions of the heteromeric (API-ligand) and homomeric (API-API, ligand-ligand) interactions is then used as a novel screening parameter in determining whether the formation of cocrystals is feasible. II. Experimental Section Materials. Two commonly used APIs paracetamol (Sigma, 98-101%, USP XXIV) and ibuprofen (Sigma, 98%), and two ligands/cocrystal formers 4,4′-bipyridyl (Fluka, g99%) and p-aminobenzoic acid (Lancaster, 99%) were selected as model systems for the current study. Solutions used were prepared by dissolving the aforementioned solute molecules, or the mixtures of them, into ethanol (BDH AnalaR, 99.7-100%). All chemicals are used without further purification. Solution Cocrystallization. The cocrystallizing solutions were made by dissolving 0.305 g (2.02 mmol) of paracetamol

Figure 3. Phase diagram of Kii as a function of φ. The lines represent the model solubility curves calculated by assuming that the molecules are interacting via square well potential for different ranges of interactions ∆ ) 0.05, 0.1, 0.2, and 0.3.23 The solid symbols correspond to experimental data.

and 0.315 g (2.02 mmol) of 4,4′-bipyridyl into 3.043 g of ethanol, and by dissolving 0.247 g (1.20 mmol) of ibuprofen

9892 J. Phys. Chem. B, Vol. 112, No. 32, 2008

He et al.

Figure 4. Normalized self-diffusivities of various molecules in ethanol at 294.9 K: (a and b) mixture of paracetamol and 4,4′-bipyridyl in the ratio of φ2/φ1 ) 1.09; (c and d) mixture of 4,4′-bipyridyl and ibuprofen in the ratio of φ3/φ2 ) 1.51; (e and (f) mixture of paracetamol and ibuprofen in the ratio of φ3/φ1 ) 1.66; (g and h) mixture of paracetamol and p-aminobenzoic acid in the ratio of φ4/φ1 ) 0.90.

and 0.094 g (0.60 mmol) of 4,4′-bipyridyl into 2.222 g of ethanol. Both paracetamol:4,4′-bipyridyl (1:1) and ibuprofen: 4,4′-bipyridyl (2:1) cocrystals were obtained through slow evaporation of the solvent. Resulting crystals were dried in oven and stored in refrigerator. Mixed solution of paracetamol and

ibuprofen, and that of paracetamol and p-aminobenzoic acid did not produce cocrystals despite several attempts through temperature ramping and solvent evaporation. Measurement of Self Diffusivity. The long-time selfdiffusivities of various molecules in ethanol were measured

Screening for Cocrystallization Tendency

J. Phys. Chem. B, Vol. 112, No. 32, 2008 9893

using 1H pulsed gradient spin-echo nuclear magnetic resonance (PGSE NMR) with a 400 MHz spectrometer (Bruker Advance 400) at 294.9 K. The sample solution was first injected into a 5 mm NMR tube, and a Wilmad stem coaxial insert containing D2O (Sigma, Standard 99.98 ( 0.01 atom % D) was inserted into the tube. This setup enables that (i) the actual selfdiffusivities of solutes in ethanol instead of deuterated ethanol are measured and (ii) the sample can be easily locked by the spectrometer for better shimming attributable to the presence of D2O. Prior to the actual experimentation, the NMR probes were calibrated using reported self-diffusivities of water.17 The self-diffusivities of individual molecules were resolved by measuring the time-dependent spatial attenuation of the characteristic peaks. III. Results and Discussion X-ray Diffraction (XRD). Single crystal XRD analysis on the resulting crystals was done using Rigaku Saturn 70 CCD area detector with Mo KR radiation ) 0.7107 Å at 50 kV and 40 mA. The cell parameters of the resulting crystals from ethanol solution of paracetamol and 4,4′-bipyridyl in this study are as follows: a ) 11.333 Å, b ) 24.157 Å, c ) 11.606 Å, R ) γ ) 90°, and β ) 96.35°, which match with those of paracetamol: 4,4′-bipyridyl (1:1) cocrystal reported in the Cambridge Structural Database (ref code: MUPQAP). Neither paracetamol nor 4,4′-bipyridyl crystals were found in the solution, suggesting that cocrystal is the preferable product in this system. Similar phenomenon was seen for ethanol solution of ibuprofen and 4,4′-bipyridyl. These observations match with those reported in the literature.18 Self Diffusivities in Solutions. The self-diffusivities of paracetamol (PAC, denoted as 1) and 4,4′-bipyridyl (BPD, 2) in mixed solutions were measured. Figure 1a shows that, for any given BPD concentration, the self-diffusivity of PAC decreases with increasing PAC concentration, suggesting that the mobility of PAC molecules is increasingly hindered when its own concentration goes up. At the fairly dilute region in the current study, such phenomenon can be correlated by a linear regression. Moreover, at any given PAC concentration, PAC molecules also display a decreasing mobility as the amount of BPD present in the solution increases. Figure 1b demonstrates similar results for BPD molecules. As one would expect, the weaker tendency of decreasing mobility shown in Figure 1b indicates a weaker correlation between concentration and mobility hindrance, as compared to that in Figure 1a. Characterization of Intermolecular Interactions. The classical theory of Brownian motion for a ternary system (two solutes and one solvent) deals with the random movement of an individual solute particle due to stochastic collisions with the solute molecules of its own and other species, and the solvent molecules of the surrounding fluid. In the dilute limit when only bimolecular interaction are considered, the long-time selfdiffusivity of the molecule of interest in a continuum can be written as19

Di ) 1 + Kiiφi + Kijφj Di0

(1a)

Dj ) 1 + Kjjφj + Kjiφi Dj0

(1b)

where Di0 (Dj0) is the Stokes-Einstein diffusivity of solute i (j) and φ is the volume fraction of solute that can be linked to the molar concentration c by φ ) c(NA/Mw)V, where NA, Mw, and V are Avogadro’s number, the molecular weight and volume,

respectively. The pair contribution of the self-diffusivity, Kij (Kji), characterizes the effects of the presence of species j (i) on the mobility of species i (j), while Kii (Kjj) characterizes such mobility hindrance in an environment that is free of impurities. Both Kii (Kjj) and Kij (Kji) are the integral measures of the hydrodynamically weighted particle interactions.19 K takes into account the effects of temperature and solvent on molecular mobility and particle interactions. A more negative value of K indicates stronger attraction between respective molecules. When species j is not present in the solutions, eqs 1a and 1b can be reduced to Di/Di0 ) 1 + Kiiφi. Values of Kii can be readily obtained from the slope of the plots shown in Figure 2. BPD and ibuprofen (IBP, denoted as 3) display less negative values of Kii as compared to PAC and p-aminobenzoic acid (PABA, denoted as 4), suggesting that pair interactions of the former two are weaker than the latter duo. This finding echoes with the fact that the solubilities of BPD (our own measurements) and IBP20 are much higher than those of PAC21 and PABA.22 In our previous work, we have developed a generalized phase diagram whose universality is testified by a variety of molecules.23,24 This phase diagram is constructed by assuming that the solute particles in the solutions are interacting through a symmetrical interaction potential (square well potential in our work), followed by development of equation of states for various phases. Last but not least, the phase boundaries such as solubility curve could be estimated by equating the chemical potentials and osmotic pressures of different phases. This phase diagram offers several advantages including direct comparison between theory and experiments (facilitated by measurement of selfdiffusivities using PGSE NMR), and its universal applicability for all molecular systems. We have shown that for short-ranged interactions (∆ e 0.3), a corresponding-states solubility behavior exists for a variety of molecules.23 Figure 3 demonstrates that the trend of the experimental solubility data (more negative K values correspond to lower solubilities) agrees with the theory reasonably well, suggesting that the molecules in this study may undergo short-ranged interactions in solutions. The small molecules studied here form hydrogen bonds in the solid state that have a range of 0.2-0.3 nm such that the range of interactions ∆ (center-to-center distance between two molecules normalized by the size of the molecules) is expected to be on the range of 0.2-0.4. This reinforces the implication that the attractions between molecules are short-ranged. The intermolecular interactions, characterized by K, are greatly influenced by the addition of another species into the solution. To facilitate comparison of particle attractions between one-solute and two-solute solutions, eqs 1a and 1b are rewritten as:

( (

) )

Di φj ) 1 + Kii + Kij φi ) 1 + Kii(ij)φi Di0 φi

(2a)

φi Dj ) 1 + Kjj + Kji φj ) 1 + Kjj(ij)φj Dj0 φj

(2b)

By carefully designing the experiments in which the ratio of the volume fractions of the two solutes (φj/φi) are kept constant, (ij) values of K(ij) ii (Kjj ) can be deduced (Figure 4) in a similar manner as shown in Figure 2. These terms are viewed as integral measures of the particle interactions between i (j) molecules by taking the effects of the presence of j (i) on particle interactions between i (j) molecules as part of the solution medium. Consequently, values of Kij and Kji can be obtained given that the Kii’s and Kjj’s are known (Table 1). In the one-solute solution, the particle interactions between the solute molecules depend on the temperature, type of solvent,

9894 J. Phys. Chem. B, Vol. 112, No. 32, 2008

He et al.

TABLE 1: Molecular Volumes of the Four Solutes, Particle Interactions, and Cocrystal Formation of the Molecular Pairs in This Studya compound

molecular volume Vi (nm3)

homomeric interactions

molecular pair

heteromeric interactions K12 K21 K23 K32 K13 K31 K14 K41

PAC (1)

0.139

K11 ) -5.26

PAC + BPD

BPD (2)

0.153

K22 ) -2.12

BPD + IBP

IBP (3)

0.231

K33 ) -2.69

PAC + IBP

PABA (4)

0.123

K44 ) -8.87

PAC + PABA

a

) ) ) ) ) ) ) )

-2.50 -5.53 -2.95 -2.27 -2.37 -4.60 -7.08 -2.74

cocrystal formation Y Y N N

Molecular volumes are obtained as described in the previous work.23

Figure 5. Comparison of normalized (by molecular volumes) values of pair interactions Kii(ij)Vi and Kjj(ij)Vj for various molecular pairs. Open (12) and closed circles correspond to K(12) 11 V1 and K22 V2; triangles represent (23) (13) (13) K(23) 22 V2 and K33 V3; squares stand for K11 V1 and K33 V3; diamonds (14) symbolize K(14) V and K V . 11 1 44 4

Figure 6. Relative magnitude of the heteromeric and homomeric pair interactions for various molecular pairs. The pair interaction index (KijKji)/(KiiKjj) > 1 corresponds to the circumstances where the heteromeric attractions are stronger than the homomeric counterparts, while