Physicochemical Properties of Pharmaceutical Co-Crystals: A Case

Sep 4, 2008 - Crystallographic and Pressure–Temperature State Diagram Approach for the Phase Behavior and Polymorphism Study of Glutaric Acid .... f...
2 downloads 11 Views 945KB Size
Physicochemical Properties of Pharmaceutical Co-Crystals: A Case Study of Ten AMG 517 Co-Crystals Mary K. Stanton and Annette Bak* Department of Pharmaceutics, Amgen Inc., Cambridge, Massachusetts 02139

CRYSTAL GROWTH & DESIGN 2008 VOL. 8, NO. 10 3856–3862

ReceiVed February 15, 2008; ReVised Manuscript ReceiVed May 1, 2008

ABSTRACT: The pharmaceutical co-crystal approach has been introduced recently to address poor physicochemical properties of new chemical entities. In this study we prepared and investigated ten co-crystals of AMG 517. Based on thermal analysis, 1H NMR and pKa values we have prepared co-crystals rather than solvates or salts. We explored co-crystal physicochemical properties such as particle size, solubility, stability, hygroscopicity, and thermal properties as well as hydrogen bonding networks. We found that all co-crystals, for which we obtained single crystal structure, had the same two hydrogen bonds between the compound and the co-crystal former. In addition, we found good correlation between the melting point of the co-crystal former and the co-crystal but less correlation between the melting point and solubility of the co-crystals. In conclusion, within this series of co-crystals some design is possible by considering features of the compound and the co-crystal former. Introduction In recent years the cost and time to develop a new drug has increased significantly, with a longer clinical development phase as the biggest cost and time factor. Therefore, if a new chemical entity (NCE) fails in this phase the economic consequences to the company can be quite significant.1,2 Changes have also occurred in the early stage of research and development. The in vitro biology screening paradigm used to discover and optimize NCEs changed in the early 1990s from an aqueous based manual approach to high-throughput screening. Furthermore, in the same time frame the popularity of combinatorial chemistry increased. Both of these techniques start with the compound solubilized in DMSO. Consequently, solubility is not incorporated in NCEs through the discovery process itself, and candidates have become less soluble.3,4 Therefore, not surprisingly solubility and dissolution limited absorption are often encountered in preclinical studies, and research has shown that some drug candidates fail in the clinical phase due to human bioavailability or formulation issues.1,2,5,6 Traditional methods to correct these problems without completely redesigning the molecule include salt selection, producing amorphous material, particle size reduction, prodrugs, and formulation approaches.7,8 New methods are frequently introduced since the traditional approaches are not suitable in all cases, and recently pharmaceutical co-crystals have been proposed as a unique crystal engineering opportunity to alter the physicochemical properties of compounds.9 The definition of the co-crystal concept is currently the focus of significant scientific debate in journal articles and presentations at conferences. The discussions are wide spanning from the use of hyphenation (i.e., co-crystal versus cocrystal) to the nature of the interactions between components of the crystal.9–16 Reaching consensus on the definition is of significance for scientific and regulatory communication, and potentially also for intellectual property purposes.17 However, since it is an evolving debate, this article will use the term co-crystal rather than cocrystal for consistency. To assess the multicomponent crystals discussed in this article we will use a recently published definition, stating that pharmaceutical co-crystals are co-crystals * Author to whom correspondence should be addressed: E-mail: [email protected].

that are formed between a drug substance and a co-crystal former, which is a solid under ambient conditions, and are not limited to two components. The components of the crystal interact by hydrogen bonding or other non-covalent and nonionic interactions.9 Introducing a guest molecule may change the crystal packing of a drug substance dramatically and therefore alter its physical and chemical properties. Consequently, producing co-crystals of pharmaceuticals has been reported to change their melting points,18 solubilityanddissolution11,19 andinvivoexposure.15,17,20–22 In a previous study we reported on using the co-crystal approach to resolve the observed solubility limited absorption for AMG 517.20 AMG 517 is a transient receptor potential vanilliod 1 antagonist (TRPV1 - see Julius and Basbaum for a review of the TRPV1 target23), that was being developed for the treatment of chronic pain.24 In this article we investigate ten co-crystals of AMG 517 with the intent to examine the effect of co-crystal former on the physicochemical properties of the co-crystals. Carboxylic acids were utilized as co-crystal formers in the study because many have a history as salt formers in drug substances, are considered pharmaceutically acceptable, or appear on the GRAS (generally recognized as safe) list.25,26 Commercially available aromatic, saturated, and unsaturated aliphatic carboxylic acids with nine or fewer carbon atoms were selected. In addition, diversity was attempted by also including acids with other hydrogen bond acceptors and donors than the acid group. The structures of AMG 517 and the ten acids used as co-crystal formers are shown in Figure 1. Ascorbic acid, aspartic acid, glucuronic acid, glutamic acid, and mucic acid were also attempted, but were unsuccessful in forming co-crystals under studied conditions. Materials and Methods Materials. Co-crystal formers were purchased from Sigma-Aldrich (L(+)-lactic acid, trans-cinnamic acid, glutaric acid, trans-2-hexenoic acid, and 2-hydroxycaproic acid), Fluka (glycolic acid), TCI (2,5dihydroxybenzoic acid), Alfa Aesar (L-(+)-tartaric acid, and benzoic acid) and EMD Chemicals (sorbic acid). Drug substance was synthesized in-house.27 Crystallization. Crystallizations were carried out by slow cooling a saturated solution. Drug substance and co-crystal former were dissolved in a 1:1.2 ratio in acetone, methanol, ethyl acetate, or acetonitrile at 50 °C, or at the solvent’s boiling point if it is less than

10.1021/cg800173d CCC: $40.75  2008 American Chemical Society Published on Web 09/04/2008

Physicochemical Properties of Pharmaceutical Co-Crystals

Figure 1. Structures of AMG 51727 and the ten acids used as co-crystal formers: (a) AMG 517, (b) benzoic acid, (c) trans-cinnamic acid, (d) 2,5-dihydroxybenzoic acid, (e) glutaric acid, (f) glycolic acid, (g) trans2-hexanoic acid, (h) 2-hydroxycaproic acid, (i) L(+)-lactic acid, (j) sorbic acid, (k) L(+)-tartaric acid. 50 °C, and then cooled at 2 °C/min in an Imperial V oven (Lab-line Instruments Inc., Melrose Park, IL). If crystallization did not occur within 48-72 h, slow evaporation was utilized until crystallization occurred. Thermal Analysis. Differential scanning calorimetry (DSC) was performed on a Q100 (TA Instruments, New Castle, DE) at 2 or 10 °C/min from 30 to 250 or 30 to 300 °C in an open, aluminum pan. Thermal gravimetric analysis (TGA) was performed on a Q500 (TA Instruments) at 2 or 10 °C/min from 30 to 300 °C in a platinum pan. X-Ray Powder Diffractometry. X-ray diffraction (XRPD) patterns were obtained on an X’Pert PRO X-ray diffraction system (PANalytical, Almelo, The Netherlands). Samples were scanned in continuous mode from 5-45° (2θ) with step size of 0.0334° on a spinning stage at 45 kV and 40 mA with Cu KR radiation (1.54 Å). The incident beam path was equipped with a 0.02rad soller slit, 15 mm mask, 4° fixed antiscatter slit and a programmable divergence slit. The diffracted beam was equipped with a 0.02rad soller slit, programmable antiscatter slit and a 0.02 mm nickel filter. Detection was accomplished with an RTMS detector (X’Cellerator). Particle Size. Particle size was determined by laser diffraction on the HELOS/BF with a CUVETTE disperser (Sympatec GmbH,

Crystal Growth & Design, Vol. 8, No. 10, 2008 3857 Clausthal-Zellerfeld). Samples were suspended in 2% hydroxypropyl methylcellulose 1% Tween 80 by vortexing. The suspension was then added dropwise to the 50 mL cuvette containing 40 mL of water until a 5-15% optical concentration was achieved. Measurements were taken for 10 s using R3 or R5 lens while mixing at 500 rpm. NMR. 1H Nuclear magnetic resonance (NMR) analysis was performed on a Bruker 400 MHz NMR (Bruker BioSpin GmbH, Germany) in DMSO-d6 or chloroform-d at 25 °C. Hygroscopicity. Hygroscopicity was determined by dynamic vapor sorption on the DVS Advantage (Surface Measurement Systems Ltd., London). Measurements were taken from 0 to 90 to 0% RH at 25 °C with 10% RH per step with equilibration set to dm/dt +0.002%/min for 5 min or 120 min/step (minimum 10 min/step). All samples reached equilibration at each step before the 120 min maximum set point was reached. Solubility. Solubility was measured from slurry (3.33 mg/mL) in Fasted Simulated Intestinal Fluid (FaSIF - 5 mM taurocholic acid sodium and 1.5 mM lecithin in pH 6.8 phosphate buffer) with samples taken at 1, 15, 30, 45, 60, 90, 120, 240, and 1440 min at 25 °C. Particle size control was not attempted nor was particle size measured during the experiment. Samples were filtered through a 0.2 µm PTFE syringe filter. Analysis was performed by HPLC-UV on an Agilent 1100 series HPLC (Agilent Technologies, Palo Alto, CA) equipped with a binary pump (G1312A), DAD detector (G1315B), auto sampler (G1329A) and a 4.5 × 150 mm, 8 nm pore size, 5 µm particle size, YMC ProC18 column (Waters Corporation, Milford, MA). Elution was achieved by a gradient method from 10 to 95% of acetonitrile 0.1% triflouroacetic acid at 1 mL/min for 8 min. Standards were prepared in 50% acetonitrile at 0.05 mg/mL and injected at 1, 5, 10, and 15 µL. The AMG 517 tartaric acid 1:1 co-crystal was not tested due to insufficient quantity. Stability. Stability samples, 10-20 mg in 4 mL clear glass vials, were placed into stability chamber (Hotpack, model 435305) at 40 °C/75% RH uncapped for 1 month. Samples were analyzed by DSC and XRPD as described above. The AMG 517 tartaric acid 1:1 cocrystal was not tested due to insufficient quantity. X-Ray Single Crystal Structure. The single crystal structure of AMG 517 and AMG 517 sorbic acid co-crystal has been published previously.20 Single crystal structures for the two co-crystals with suitable single crystals were determined by Dr. Richard J. Staples at Harvard University (Cambridge, MA). Data were collected using a Bruker SMART APEX CCD (charge coupled device) based diffractometer equipped with an Oxford Cryostream low-temperature apparatus operating at 193 K. Data were measured using omega scans of 0.3° per frame for 45 s, such that a hemisphere was collected. A total of 1221-1850 frames were collected with a maximum resolution of 0.76-0.90 Å. The first 50 frames were recollected at the end of data collection to monitor for decay. Cell parameters were retrieved using SMART software and refined using SAINT on all observed reflections. Data reduction was performed using the SAINT software which corrects for Lp and decay. The structures were solved by the direct method using the SHELXS-97

Table 1. Physicochemical Properties of Ten AMG 517 Co-Crystals calculated pKaa

co-crystal AMG AMG AMG AMG AMG AMG AMG AMG AMG AMG

517 517 517 517 517 517 517 517 517 517

-

benzoic acid trans-cinnamic acid 2,5-dihydroxybenzoic acid glutaric acidc glycolic acid trans-2-hexanoic acid 2-hydroxycaproic acid L(+)-lactic acid sorbic acid c L(+)-tartaric acid

melting onsets (°C) ∆pKa free base melt TGA molar free baseb former (base-former) former co-crystal free base observedd weight % crystalinitye ratiof 0.68 0.68 0.68 0.68 0.68 0.68 0.68 0.68 0.68 0.68

4.20 3.88 3.01 4.33 3.74 4.80 3.86 3.90 4.59 3.07

-3.52 -3.20 -2.33 -3.65 -3.06 -4.12 -3.18 -3.22 -3.91 -2.39

122 133 205 97 78 34 61 46 134 171

146 204 229 153 141 127 130 138 150 198

230 230 230 230 230 230 230 230 230 230

yes no no no yes yes yes yes yes no

21.7 25.7 24.5 24.5 14.2 19.7 24.7 16.5 20.8 23.0

crystalline crystalline crystalline crystalline crystalline crystalline crystalline crystalline crystalline crystalline

1:1 1:1 1:1 1:1 1:1 1:1 1:1 1:1 1:1 1:1

a pKa value calculated using ACD software.28 b Amino benzothiazole group pKa. c Glutaric and tartaric acid are diacids and the second CpKa values are 4.79, 5.27, and 4.35, respectively. d Determined by overlaying with the DSC of AMG 517 free base. e Crystallinity was determined by XRPD. f Molar ratio was determined by TGA and 1HNMR; NA: not relevant or available.

3858 Crystal Growth & Design, Vol. 8, No. 10, 2008

Stanton and Bak

Figure 2. XRPD, DSC, and TGA of the benzoic acid and trans-cinnamic acid co-crystal of AMG 517. Table 2. Crystallographic Data for Two AMG 517 Co-Crystals parameter formula stoichiometry formula weight crystal system space group a (Å) b (Å) c (Å) R (deg) β (deg) γ (deg) volume (Å3) calc density (g cm-3) Z θ range (deg) data/restraints/params T (K) R1 wR2

AMG 517 cinnamic

AMG 517 hexanoic

C29H21F3N4O4S 1:1 578.56 triclinic P1j 9.5364(17) 11.326(2) 13.502(2) 65.312(3) 88.844(3) 87.819(3) 1324.1(4) 1.45 2 1.66-27.91 6256/0/454 193(2) 0.0773 0.1170

C26H23F3N4O4S 1:1 544.54 monoclinic C2/c 62.498(18) 8.116(2) 19.870(6) 90.000 94.141(9) 90.000 10052(5) 1.44 16 1.31-22.50 6576/30/745 193(2) 0.0903 0.0792

Table 3. DC-O Distances for Three AMG 517 Co-Crystals compound

former

DC-01 (Å)

DC-02 (Å)

∆DC-O

AMG 517 AMG 517 AMG 517

sorbic acid cinnamic acid trans-2-hexenoic acid

1.218 1.206 1.214

1.304 1.322 1.313

0.086 0.116 0.099

program and refined by least-squares method on F2, SHELXL-97, incorporated in SHELXTL-PC V 6.10.

Results and Discussion Ten 1:1 co-crystals were prepared and their composition was confirmed by TGA and 1H NMR. DSC and XRPD were conducted on all co-crystals and assessed to be different than that of the parent free base, or the co-crystal former. The melting

points of the co-crystal former, the free base, and the resulting co-crystal as well as calculated pKa values,28 percent weight loss by TGA, crystallinity and molar ratio are shown in Table 1. The melting of the co-crystal is taken as the onset of the endotherm on DSC, which is associated with the weight loss on TGA. This weight loss is due to evaporation or sublimation of the released co-crystal former, as has previously been described in detail for the sorbic acid co-crystal.20 As summarized in Table 1, recrystallization to the free base, as judged by the presence of the characteristic free base melting point at 230 °C, is observed in the DSC of some of these co-crystals. Other endothermic or exothermic events are present for some co-crystals. These events may reflect the presence of a small amount of surface solvent when occurring prior to loss of the co-crystal former or recrystalization to other polymorphs of the free base when occurring after the loss of the co-crystal former. DSC, TGA, and XRPD of the benzoic acid and the transcinnamic acid co-crystals are shown in Figure 2. DSC, TGA, and XRPD of the remaining co-crystal are available as Supporting Information. Are the ten multicomponent crystals listed in Table 1 cocrystals? AMG 517 and all co-crystal formers are solids at room temperature (i.e., 25 °C - see Table 1); the systems therefore classify as co-crystals based on the physical state of the pure isolated solids.16 Table 1 also shows ∆pKa (pKa base - pKa acid former) ranging from -2.33 to -4.12. A recent article concludes that a ∆pKa > 3 almost exclusively results in salt formation, a ∆pKa < 0 almost exclusively results in co-crystal formation, and that 0 < ∆pKa < 3 can result in both or even complexes with partial proton transfer.12 Consequently, based on pKa values it appears that co-crystals rather than salts were formed. The single crystal structure of two co-crystals was determined. The data for AMG 517 and the sorbic acid co-crystal have been

Physicochemical Properties of Pharmaceutical Co-Crystals

Crystal Growth & Design, Vol. 8, No. 10, 2008 3859

Figure 3. Hydrogen bonding of (a) AMG 517 trans-cinnamic acid and (b) AMG 517 trans-2-hexanoic acid co-crystals. The view along the b-axis of the co-crystal is shown.

Figure 4. Melting onset of former versus co-crystal.

published previously.20 For the remaining co-crystals, we could not obtain crystals of sufficient quality for single crystal structure determination. Table 2 shows the crystallographic data. Table 3 shows the DC-O distances (i.e., carbon oxygen distances in the acid carbonyl group) obtained from the crystal structures of the three co-crystals. In a study of 2-aminopyrimidine and 2-acetaminopyridine salts and co-crystals, it was concluded that distinction between salts and co-crystals can be made based on ∆DC-O distances from single crystal structures, since a carboxyl anion possesses two similar DC-O values, whereas a neutral carboxyl group possesses two distinctively different DC-O values. For the 2-aminopyrimidine systems, it was concluded that complexes with ∆DC-O < 0.03 Å are salts, and complexes with ∆DC-O > 0.08 Å are co-crystals12,29 However, it should be noted that these values may be different for different systems. Consequently, from Table 3 it is evident that the AMG 517 sorbic acid, trans-cinnamic acid, and trans-2-hexanoic acid multicomponent co-crystals also qualify as co-crystals from a carbonyl bond distance perspective. When the hydrogen bonding in the single crystal structure of the co-crystals was studied, we found that both the co-crystals listed in Table 2 had a different network than AMG 517 free base. Furthermore, both co-crystal formers interact with the drug substance through the same two hydrogen bonds as previously published for the AMG 517 sorbic acid co-crystal.20 Briefly, we observed hydrogen bonds from the amide proton on AMG 517 to the carbonyl group on the acid and from the hydroxyl

group on the acid to the benzothiazole on AMG 517 (see Figure 3). Aside from the networks discussed here, we found no other hydrogen bonds in the crystal structures, which otherwise are held together by van der Waals forces. Hydrogen bonding networks, along with other intermolecular forces, are known to contribute to physical properties of solids such as the enthalpy of fusion. The melting point of a solid, which is an easily available physical parameter, can be determined from the ratio of the enthalpy of fusion to the entropy of fusion, since the free energy of transition is zero at the melting point.30 The melting onsets of AMG 517 and the ten co-crystals are listed in Table 1. It is evident that the melting point of all co-crystals falls between that of the former and the drug substance. Figure 4 shows a plot of co-crystal versus co-crystal former melting points with direct proportionality between the two parameters. Using linear regression, a correlation coefficient of 0.7849 was found, indicating that 78% of the variability in co-crystal melting point can be explained by variability in the melting point of the former.31 Therefore, for the series of AMG 517 co-crystals, it is possible to modify the melting point of the co-crystal by considering the melting point of the co-crystal former. Such a relationship has previously been published for a series of 2-acetaminopyridine co-crystals with various acids.29 The remaining variability in co-crystal melting point for the series described in this article may be explained by differences in structure, hydrogen bonding network, and in Van der Waals forces holding the crystals together. Solubility is a complex parameter and has been cited to depend on the enthalpy of fusion, temperature of the solvent, melting point of the solid, hydrogen bonding in the solid and solvent, and other polar and nonpolar forces in the solvent and the solute.32 Enthalpy of fusion data for the co-crystals was not readily available, since many of the co-crystals melt and convert to the free base in the same transition, as previously described for the AMG 517 sorbic acid co-crystal.20 However, in recent publications melting points were correlated directly to Log of solubility.33,34 The correlation assumes that the entropy of fusion follows Walden’s rule,35 which has been interpreted to state that the entropy of fusion for most organic compounds is approximately 13.5 cal/deg/mol. When testing solubility for AMG 517 and the nine co-crystals, we aimed at using media that would be relevant for absorption. Therefore, 0.01 N HCl (pH 2) and FaSIF (pH 6.8) were selected to reflect the conditions of the stomach and small intestine, respectively.36 The solubility of AMG 517 was determined to be higher in FaSIF than in

3860 Crystal Growth & Design, Vol. 8, No. 10, 2008

Stanton and Bak

Table 4. Solubility (Smax), Particle Size, and Potential Conversion of Nine Co-Crystalsa melting onsets (°C) co-crystal AMG AMG AMG AMG AMG AMG AMG AMG AMG

517 517 517 517 517 517 517 517 517

-

trans-cinnamic acid 2,5-dihydroxybenzoic acid 2-hydroxycaproic acid glutaric acid glycolic acid sorbic acidb trans-2-hexanoic acid L(+)-lactic acid benzoic acid

former

co-crystal

molar ratio

133 205 61 97 78 134 34 46 122

204 229 130 153 141 150 127 138 146

1:1 1:1 1:1 1:1 1:1 1:1 1:1 1:1 1:1

solubility Smax (µg/mL) free basec

co-crystal

5 5 5 5 5 5 5 5 5

1 2 3 9 9 12 17 18 21

log Smax conversion during co-crystal experimentd 0.00 0.30 0.48 0.95 0.95 1.08 1.23 1.26 1.32

no no no some FB hydratee some FB hydratee some FB hydratee some FB hydratee some FB hydratee no

particle size ×50 (µm) co-crystal

free base

144 197 510 111 82 46 76 640 447

3 3 3 3 3 3 3 3 3

a Smax: The max solubility observed in the solubility over time profile. b The solubility of the sorbic acid co-crystal was repeated to ensure consistency in these experiments. However, the values are similar to what was previously published.20 c No conversion of the parent free bases was observed during the experiment. d Co-crystal conversion was assessed by XRPD. e The solid-state characterization of the AMG 517 hydrate and related forms were previously published.38

Figure 5. Smax as a function of co-crystal melting point.

Figure 6. Solubility plot of AMG 517 and five co-crystals, with a bellshaped profile, in FaSIF.

0.01N HCl in a screening solubility assay. In addition, AMG 517 is unstable at low pH.37 Consequently, the solubility of the AMG 517 co-crystals was determined in FaSIF (except for the tartaric acid co-crystal due to insufficient material). The solubility (Smax), melting onsets, particle size, and conversion of the co-crystals during the experiment are captured in Table 4. When a correlation analysis31 was performed on the numerical parameters listed in Table 4 (data not shown), the highest interdependence was found to be between co-crystal melting point and Log Smax for the nine AMG 517 co-crystals. The correlation plot is shown in Figure 5 and indicates that 55% of the variability in Log Smax can be explained by variability in

melting point of the co-crystal for these nine co-crystals. Previously, correlations between melting points, Log P (Log octanol/water partition coefficient) values, and Log solubility have been published for nonelectrolytes, electrolytes, and pharmaceutically relevant compounds. In general the correlations described in these articles are better33,34 than what we are reporting here. It is possible that some of the remaining variability could be explained by differences in lipophilicity. However, until the association and dissociation of co-crystals in various media are better understood, Log P will be a difficult parameter to determine for multicomponent crystals. Alternative ways of measuring the lipophilicity or wetting of co-crystals are currently being explored in our laboratory. In addition, Walden’s rule may not apply for these complex multicomponent crystals. Walden’s rule was derived on a small set of simple rigid organic molecules.35 Consequently, it was suggested that the rule does not have general validity, and that enthalpy and entropy of fusion are related in a linear fashion.39,40 Therefore, it is possible that we would have obtained better correlation with enthalpy of fusion data. Finally, the Smax parameter used in this study has a kinetic or dissolution aspect, which was necessary to include, since many of the co-crystals revert back to the free base (see subsequent discussion and Table 4). This could have contributed additional variability to the solubility correlation. Six of the nine AMG 517 co-crystals (i.e, the glutaric acid, glycolic acid, sorbic acid, trans-2-hexanoic acid, lactic acid, and the benzoic acid co-crystal) reach a maximum solubility (Smax) within 1-2 h, which then decreases over time. The solubility plots for these co-crystals, except for the sorbic acid co-crystal profile, are shown in Figure 6 with a comparison to the AMG 517 free base. The bell-shaped profile can be explained by the dissolution of the co-crystal followed by conversion to the AMG 517 hydrate for all co-crystals shown in Figure 6 except for the benzoic acid co-crystal. A similar conversion has previously been reported and discussed in detail for the AMG 517 sorbic acid co-crystal.20 However, as evident from Table 4, four of the co-crystals exhibit no apparent form change when analyzed by XRPD. Three of these co-crystals (i.e., the trans-cinnamic acid, 2.5-dihydroxybenzoic acid, and 2-hydroxycaproic acid cocrystals) display Smax of less than 3 µg/mL, which is lower than AMG 517 free base solubility, and are the most insoluble cocrystals studied. Therefore, they may not have been sufficiently soluble to undergo conversion within the 24 h time frame of the study. The benzoic acid co-crystal has a significant Smax (i.e., 21 µg/mL; see Table 4), and displays a bell-shaped solubility profile (see Figure 6) indicative of conversion, although no conversion

Physicochemical Properties of Pharmaceutical Co-Crystals

Crystal Growth & Design, Vol. 8, No. 10, 2008 3861

Figure 7. DSC and XRPD of the AMG 517 benzoic acid co-crystal before and after the 24 h solubility experiment in FaSIF. Table 5. Physical Stability Data for AMG 517 and Nine Co-Crystals change in mass (%)

a

sample

80% RH

90% RH

European Pharmacopeiaa

changes in XRD after 1 month at 40 °C/75% RH

AMG 517 - benzoic acid AMG 517 - trans-cinnamic acid AMG 517 - 2,5-dihydroxybenzoic acid AMG 517 - glutaric acid AMG 517 - glycolic acid AMG 517 - trans-2-hexanoic acid AMG 517 - 2-hydroxycaproic acid AMG 517 - L(+)-lactic acid AMG 517 - sorbic acid AMG 517

-0.02 -0.28 0.15 0.10 0.04 -0.01 0.68 0.03 -0.16 0.04

-0.06 -0.27 0.23 1.01 0.05 -0.02 1.55 0.10 -0.16 0.03

not hygroscopic not hygroscopic not hygroscopic not hygroscopic not hygroscopic not hygroscopic slightly hygroscopic not hygroscopic not hygroscopic not hygroscopic

no changes no changes no changes no changes no changes no changes no changes no changes no changes no changes

See refs 43 and 44 and upcoming discussion for details.

is apparent when analyzed by XRPD. One explanation may be the sensitivity limitation of the XRPD technique and that the conversion within 24 h is not sufficient for detection. It is also possible that the patterns may not be sufficiently different to detect a conversion. Figure 7 shows the DSC and XRPD of the parent benzoic acid co-crystal overlaid with the 1440 min (24 h) solubility time point. As can be seen from Figure 7, the transition corresponding to the loss of benzoic acid is missing from the DSC indicating that the co-crystal former has left the crystal lattice. This may reflect formation of a form analogous to a desolvated solvate, which has been described to have similar XRPD pattern to the corresponding solvate.41 In support of this theory, the concentration of benzoic acid in solution during the solubility experiment also increases over time from 3 µg/mL at 15 min to 122 µg/mL at 24 h. As discussed in this article and in a previous publication,20 several of the AMG 517 co-crystals convert to a form of the free base hydrate under the conditions of the solubility experiments. It has been argued that the initial solubility advantages of the co-crystals (i.e., time before reversion to the free base or hydrate) are sufficient to give an exposure advantage in pharmacokinetic studies.17,20,42 However, the phenomena may give rise to solid-state physical stability concerns, and therefore raise questions about the suitability of these co-crystals as development forms. To investigate potential physical stability concerns, solid samples of nine co-crystals were monitored at 40 °C/75% RH for 1 month. The samples were analyzed by XRPD and DSC. In addition, the moisture sorption profiles were generated for nine co-crystals and AMG 517. This physical stability data are summarized in Table 5. As seen from Table 5 no significant changes were observed in the XRPD of the

samples, indicating that all were stable at 40 °C/75% RH for 1 month. In moisture absorption experiments, a mass change of 0.2-2% change in mass at 25 °C/80% RH has been described as slightly hygroscopic and a change of 2-15% at 25 °C/80% RH as hygroscopic.43,44 According to this definition, only one co-crystal was slightly hygroscopic, while all others were not hygroscopic. The hygroscopicity data may explain why no form conversion was seen in the solid-state stability study. Conclusion The need for new approaches to address poor physicochemical properties of drug candidates has recently become more apparent, since the cost and time of developing a new drug has increased, and physicochemical properties of NCEs have become less desirable. The research discussed in this article illustrates that the pharmaceutical co-crystal approach can be successfully and methodically used to alter physicochemical properties of NCEs while still maintaining hygroscopicity and stability suitable for drug development. We investigated ten 1:1 co-crystals of AMG 517 with commercially available acids, which are all commonly used in salt screening or on the GRAS list. It was not possible to get single crystal structures of all co-crystals, but the three obtained showed an identical hydrogen bonding network. Hydrogen bonding network is related to physicochemical properties, and we found a 78% correlation between melting point of the co-crystal former and the melting point of the co-crystal. In addition, we found a 55% correlation between the melting point and the Log solubility of the cocrystals. Therefore, we have shown that some design may be possible within a defined series of co-crystals, and consequently

3862 Crystal Growth & Design, Vol. 8, No. 10, 2008

the approach can be useful for the pharmaceutical industry. However, the pharmaceutical co-crystal approach is still in its infancy, and many of its aspects still need to be explored, defined, and adapted to drug development. All that is required, in theory, to form a co-crystal, is a hydrogen bond donor and acceptor. With such a large, diverse pool of co-crystal formers from which to choose from, and given that some design is possible, as has been illustrated in this article, it is not productive to restrict the selection of co-crystal formers to the list of acids available for salt selection. AMG 517 appears to form co-crystals readily, and therefore work is ongoing in our laboratory exploring more co-crystals of this and related compounds using a more diverse set of co-crystal formers than in the present research. This will be the topic of a future communication. Acknowledgment. We wish to acknowledge Ron Kelly, Roman Schimanovich, Drazen Ostovic, and Beata SwerydaKraweic for helpful discussions. In addition, we would like to thank Richard Staples for conducting the single crystal structure analysis, and Kristine Lee for editing assistance. Note Added after ASAP Publication. This article posted ASAP to the web on September 4, 2008 with an incorrect version of Figure 1. The current and print versions are correct. Supporting Information Available: 1H NMR, DSC, TGA, XRPD data and single crystal X-ray crystallographic information files (.cif) are available. This material is available free of charge via the Internet at http://pubs.acs.org.

References (1) FDA Challenges and Opportunity on the Critical Path to New Medical Products. http://www.fda.gov/oc/initiatives/criticalpath/whitepaper. html#innovation. (2) Tankosic, T. Drug Market DeV. 2003, 340–344. (3) Lipinski, C. Am. Pharm. ReV. 2002, 5 (3), 82–85. (4) Lipinski, C. A.; Lombardo, F.; Dominy, B. W.; Feeney, P. J. AdV. Drug DeliVery ReV. 2001, 46, 3–26. (5) Kola, I.; Landis, J. Nat. ReV. Drug DiscoVery 2004, 3, 711–715. (6) Prentis, R. A.; Lis, Y.; Walker, S. R. Br. J. Clin. Pharmacol. 1988, 25, 387. (7) Garad, S. D. Am. Pharm. ReV. 2004, 7 (2), 80–85. (8) Neervannan, S. Am. Pharm. ReV. 2004, 7 (5), 108–113. ¨ .; Zaworotko, M. J. Chem. Commun. 2004, 17, 1889– (9) Almarsson, O 1896. (10) Aakeroy, C. B.; Fasulo, M. E.; Desper, J. Mol. Pharm. 2007, 4 (3), 317–322. (11) Childs, S. L.; Chyall, L. J.; Dunlap, J. T.; Smolenskaya, V. N.; Stahly, B. C.; Stahly, G. P. J. Am. Chem. Soc. 2004, 126 (41), 13335–13342. (12) Childs, S. L.; Stahly, G. P.; Park, A. Mol. Pharm. 2007, 4 (3), 323– 338. (13) Desiraju, G. R. CrystEngComm 2003, 5 (82), 466–467. (14) Dunitz, J. D. CrystEngComm 2003, 5 (91), 506. (15) McNamara, D. P.; Childs, S. L.; Giordano, J.; Iarriccio, A.; Cassidy, J.; Shet, M. S.; Mannion, R.; O’Donnell, E.; Park, A. Pharm. Res. 2006, 23 (8), 1888–1897. (16) Vishweshwar, P.; McMahom, J. A.; Bis, J. A.; Zaworotko, M. J. J. Pharm. Sci. 2006, 95 (3), 499–516. (17) Cooke, M. W.; Stanton, M.; Shimanovich, R.; Bak, A. Am. Pharm. ReV. 2007, 10 (7), 54–59.

Stanton and Bak (18) Bailey Walsh, R. D.; Bradner, M. W.; Fleischman, S.; Morales, L. A.; Moulton, B.; Rodriguez-Hornedo, N.; Zaworotko, M. J. Chem. Commun. 2003, (2), 186–187. (19) Remenar, J. F.; Morissette, S. L.; Peterson, M. L.; Moulton, B.; MacPhee, J. M.; Guzman, H. R.; Almarsson, O. J. Am. Chem. Soc. 2003, 125 (28), 8456–8457. (20) Bak, A.; Gore, A.; Yanez, E.; Stanton, M.; Tufekcic, S.; Syed, R.; Akrami, A.; Rose, M.; Surapaneni, S.; Bostick, T.; King, A.; Neervannan, S.; Ostovic, D.; Koparkar, A. J. Pharm. Sci. 2007, DOI: 10.1002/jps.21280. (21) Chen, A. M.; Ellison, M. E.; Peresypkin, A.; Wenslow, R. M.; Variankaval, N.; Savarin, C. G.; Natishan, T. K.; Mathre, D. J.; Dormer, P. G.; Euler, D. H.; Ball, R. G.; Ye, Z.; Wang, Y.; Santos, I. Chem. Commun. 2007, (4), 419–421. (22) Variankaval, N.; Wenslow, R.; Murry, J.; Hartman, R.; Helmy, R.; Kwong, E.; Clas, S.-D.; Dalton, C.; Santos, I. Cryst. Growth Des. 2006, 6 (3), 690–700. (23) Julius, D.; Basbaum, A. I. Nature 2001, 413 (6852), 203–210. (24) Gavva, N. R.; Bannon, A. W.; Hovland Jr, D. N.; Surapaneni, S.; Immke, D. C.; Lehto, S.; Bak, A.; Davis, J.; Heyer, G.; Kuang, R.; Ernst, N.; Tamir, R.; Wang, J.; Zhu, D.; Norman, M. H.; Louis, J.-C.; Magal, E.; Treanor, J. J. S. J. Pharmacol. Exp. Ther. 2007, 323 (1), 128–137. (25) FDA Select Committee on GRAS Substances (SCOGS) Database Overview. http://www.cfsan.fda.gov˜dms/opascogs.html/. (26) Stahl, P. H.; Wermuth, C. G. In Handbook of Pharmaceutical Salts: Properties, Selection, and Use; Stahl, P. H.; Wermuth, C. G. Eds.; Verlag Helvetica Chimica Acta: Zurich, 2002; Chapter 12. (27) Doherty, E. M.; Bannon, A. W.; Bo, Y.; Chen, N.; Dominguez, C.; Falsey, J.; Fotsch, C.; Gavva, N. R.; Katon, J.; Nixey, T.; Ognyanov, V. I.; Pettus, L.; Rzasa, R.; Stec, M.; Surapaneni, S.; Tamir, R.; Zhu, J.; Treanor, J. J. S.; Norman, M. H. J. Med. Chem. 2007, 50 (15), 3515–3527. (28) ACD Advanced Chemistry Development Software Solaris V. 4.67, 1994-2005. (29) Aakeroy, C. B.; Hussain, I.; Desper, J. Cryst. Growth Des. 2006, 6 (2), 474–480. (30) Jain, A.; Yalkowsky, S. H. J. Pharm. Sci. 2006, 95 (12), 2562–2618. (31) Levine, D. M.; Berenson, M. L.; Stephan, D. In Statistics for Managers, 2nd ed.; Levine, D. M.; Berenson, M. L.; Stephan, D. Eds.; Prentice Hall: Upper Saddle River, NJ, 1999; Chapter 13.3, pp 783-794. (32) Martin, A.; Newburger, J.; Adjei, A. J. Pharm. Sci. 1980, 69 (5), 487. (33) Jain, N.; Yalkowsky, S. H. J. Pharm. Sci. 2001, 90, 234–252. (34) Ran, Y.; Jain, N.; Yalkowski, S. H. J. Chem. Inf. Comput. Sci. 2001, 41, 1208–1217. (35) Walden, V. P. Z. Elektrochem. 1908, 43, 713–728. (36) Dressman, J. B.; Amidon, G. L.; Reppas, C.; Shah, V. P. Pharm. Res. 1998, 15 (1), 11–22. (37) Yanez, E.; Daurio, D.; Buntich, S.; Bak, A. Physicochemical, solidstate, and pharmacokinetic properties of AMG 517 free base. Personal correspondence. (38) Yanez, E.; Cooke, M. W.; Bak, A. Hydrate Formation - Theory and a Case Study. In IWPCPS9, Ninth International Workshop on Physical Characterization of Pharmaceutical Solids; Natick, MA, June 2429, 2007, unpublished results. (39) Gilbert, A. S. Thermochim. Acta 1999, 339, 131–142. (40) Gilbert, A. S. Thermochim. Acta 2005, 428, 1–9. (41) Stephenson, G. A.; Groleau, E. G.; Kleemann, R. L.; Xu, W.; Rigsbee, D. R. J. Pharm. Sci. 1998, 87 (5), 536–542. (42) Sawamototo, T.; Haruta, S.; Kurosaki, Y.; Higaki, K.; Kimura, T. J. Pharm. Pharmacol. 1997, 49, 450–457. (43) European Pharmacopeia, Technical Guide; European Directorate for the Quality of Medicines: Strasbourg, France, 1999. (44) Newman, A. W.; Reutzel-Edens, S. M.; Zografi, G. J. Pharm. Sci. 2008, 97, 1047–1059.

CG800173D