Layered Structure and Swelling Behavior of a Multiple Hydrate

Layered Structure and Swelling Behavior of a Multiple Hydrate-Forming Pharmaceutical Compound Y.-H. Kiang,*,#,† Wei Xu,*,† Peter W. Stephens,‡ Richard G. Ball,§ and Nobuyoshi Yasuda| Pharmaceutical Research and DeVelopment, Merck Research Laboratories, Merck & Co., Inc., P.O. Box 4, West Point, PennsylVania 19426, Department of Physics and Astronomy, State UniVersity of New York, Stony Brook, New York 11794-3800, Medicinal Chemistry, Merck Research Laboratories, Merck & Co., Inc., 126 E. Lincoln AVe., Rahway, New Jersey 07065, and Process Research, Merck Research Laboratories, Merck & Co., Inc, 126 E. Lincoln AVe., Rahway, New Jersey 07065

CRYSTAL GROWTH & DESIGN 2009 VOL. 9, NO. 4 1833–1843

ReceiVed September 8, 2008; ReVised Manuscript ReceiVed December 16, 2008

ABSTRACT: Investigation of one anhydrous and four hydrated forms of a pharmaceutical compound (1) using both single-crystal and high-resolution powder X-ray diffraction methods revealed a two-dimensional framework which, upon exposure to moisture, absorbed water between the layers, causing the lattice to expand by as much as 20% of the axial length along a. The single-crystal structure was solved and refined for the pentahydrate form in space group C2 with unit cell parameters a ) 36.961(5) Å, b ) 7.458(2) Å, c ) 20.691(4) Å, β ) 99.461(1)°, and V ) 5626(4) Å3. In the single-crystal structure the water layers were parallel to the bc plane and sandwiched by the crystalline compound 1 framework. Upon a change of relative humidity, water goes in and out of the interlayer space with the retention of the layer structure of the development compound. Starting from the anhydrous form, each additional water of hydration increased the interlayer spacing of the pharmaceutical solid by ∼1.3 Å, half the size of a water molecule. In an exploratory formulation, this expansion of interlayer spacing caused tablets to crack upon storage at high relative humidity. Introduction In solid-state chemistry, compounds of layered structures have attracted much interest for their potential applications as catalysts, adsorbents, and agrochemical delivery systems.1-6 Most of the layered compounds are inorganic or hybrid materials such as clay, graphite, and metal oxides. Known examples of organic layered structures are few in spite of recent interest in design and synthesis of such compounds for their host-guest inclusion properties to mimic behavior of clays.7-10 For layered compounds, intercalation is one of the characteristic reactions. In intercalation reactions guest molecules or ions are reversibly inserted into the interlayer space without major rearrangement of the structural features of the twodimensional host. This host-guest intercalation property on the one hand makes many layered materials useful, while, on the other hand, causes problems in some applications. For example, clay minerals are notable for their tendency to expand and hydrate upon exposure to water. In this process, called swelling, clays absorb water into their layers, resulting in strong repulsive forces that cause the clay to expand as much as several times its original thickness. Swelling, in conjunction with ion exchange properties, makes some clays, such as smectites, useful as backfill materials in nuclear waste repositories.11,12 However, in oil and gas drilling operations, clayswelling can lead to severe problems such as borehole instability and even collapse of the well. The mechanism of clay swelling therefore has been intensively investigated using molecular modeling and other methods.13-16 Typical clay swelling in the crystalline regime occurs in discrete steps of one, two, or three water layers in the clay interlayer and can be considered as forming discrete hydrates of such clays. Hydrates of crystalline drug substances are commonly seen and have attracted increasing attention in the pharmaceutical industry * Authors to whom correspondence should be addressed. E-mail: ykiang@ amgen.com (Y.-H.K.); [email protected] (W.X.). # Present address: Amgen Inc., One Amgen Center Drive, Thousand Oaks, CA 91320. † Pharmaceutical Research and Development, Merck Research Laboratories. ‡ State University of New York. § Medicinal Chemistry, Merck Research Laboratories. | Process Research, Merck Research Laboratories.

for their potential impact on the development and performance of solid-dosage formulations.17-22 Structurally, hydrates of organic compounds can be categorized into two classes: isolated site hydrates and channel hydrates.18 In channel hydrates solvent-like water molecules can easily move in and out of the tunnels composed of crystalline drug frameworks. Continuous lattice expansion is often observed when channel hydrates are subject to increasing relative humidity. However, the extent of this lattice expansion is usually smaller than 1.5 Å, negligible compared to the layer expansion in clay swelling.23,24 To date, there are no reported cases of pharmaceutical solids that adopt layered structures and exhibit properties such as swelling and guest-molecule exchange. This lack of known structures of layered pharmaceutical solids is not difficult to understand considering the rare occurrence of layered organic materials, even with specific interest in design and synthesis of organic layered compounds based on crystal engineering.7-10,25-27 Pharmaceutical compounds are intended for their therapeutic effect and are not designed to form self-assembled clay mimics. However, when crystal packing of drugs results in an adventitious layered structure in which swelling occurs upon exposure to water, this hydration-induced lattice expansion in soliddosage formulation can be potentially as disastrous to a tablet as clay swelling to a drilling well, and calls for appropriate engineering effort.

In an exploratory tablet formulation of a development pharmaceutical compound intended for the treatment of os-

10.1021/cg801004a CCC: $40.75  2009 American Chemical Society Published on Web 02/03/2009

1834 Crystal Growth & Design, Vol. 9, No. 4, 2009

Figure 1. Moisture sorption isotherm for 1 at 25 °C. Adsorption data are represented in solid circles and desorption data are represented in open circles.

teoporosis (compound 1), cracking was observed when tablets were stored at high relative humidity. Because the cracking only occurred when compound 1 was present in the tablet, structural behavior of compound 1 when subjected to moisture is believed to cause the tablet cracking. The moisture sorption isotherm of 1 at 25 °C (Figure 1) exhibits five discrete moisture levels and the weight percentage of these five moisture levels suggests that compound 1 can exist in five different hydrated states: anhydrous, hemihydrate, dihydrate, tetrahydrate, and pentahydrate. X-ray diffraction data collected under controlled relative humidity (Figure 2) show five distinct but similar patterns indicating a lattice expansion as the relative humidity increases. The similarity of the patterns in Figure 2 is indicative that the five hydrate forms have very similar crystal structures. In this work we study the crystal structures of the multiple hydrate forms of this pharmaceutical compound using both single-crystal and high-resolution powder X-ray diffraction methods in an attempt to understand the relationship of the various hydrates at the molecular level. Experimental Section 1. General Procedures. The development compound 1 (3-{2-oxo3-[3-(5,6,7,8-tetrahydro- [1,8]naphthyridin-2-yl)propyl]-imidazolidin1-yl}-3(S)-(6-methoxy-pyridin-3-yl)propionic acid) was synthesized by Merck & Co., Inc. with at least 98% purity. Uncoated tablets were manufactured by Merck & Co., Inc. at a 2.5-kg scale by a wet granulation process. The main inactive ingredient is either Avicel PH102 or Mannitol (Pearlitol) with 67% load of compound 1. Analytical grade solvents were obtained from commercial suppliers (Aldrich, Fisher Scientific, and Mallinckrodt) and used without further purification. The glass capillaries used for X-ray diffraction, 0.01 mm in thickness, 1.5 mm and 0.7 mm in diameter were purchased from Charles Supper Company. 2. Moisture Sorption Isotherms. Moisture sorption isotherm studies were performed by measuring the increase in weight at equilibrium as a function of relative humidity. Measurements were obtained at 25 °C, 30 and 40 °C using a VTI WB-300 symmetric vapor sorption analyzer. The sample was subject to relative humidity ranging from 5% to 95% with a step of 5% following drying at 50 °C for 30 min. A point isotherm was recorded at equilibrium (the stage when less than 0.04% weight change was observed) or every 120 min. 3. X-ray Powder Diffraction. Powder X-ray diffraction patterns of compound 1 were collected under controlled relative humidity on a

Kiang et al.

Figure 2. XRD patterns of 1 at (a) 95%, (b) 75%, (c) 60%, (d) 30%, and (e) 5% relative humidity. Philips PW3040-PRO X-ray diffractometer with a Cu KR radiation (λ ) 1.54056 Å) at 45 kV, 40 mA. The Kβ radiation was eliminated by a multigraded mirror at the incident side and the diffracted beam was refocused using a secondary mirror. Each scan was recorded with a step size of 0.02° from 2 to 40° in 2θ. Samples were placed in an Anton Paar THC chamber mounted on the goniometer of the X-ray diffractometer. Relative humidity was controlled by a VTI RH-200 humidifier at 25 °C. The sample of compound 1 suspended in water for X-ray analysis was prepared as follows: A 1.0 mm special glass capillary tube was filled with water. Compound 1 was added to this capillary tube and suspended in water for 2 h before the tube was sealed. The capillary was then mounted to an INEL diffractometer equipped with a CP-120 detector. Powder X-ray data were recorded at 35 kV, 30 mA for CuKR (λ ) 1.54056 Å). During data collection the sample was rotated to reduce the effect of sample granularity. X-ray data collection for the tablets were performed on a Philips 3040-PRO diffractometer equipped with a multiplewire detector at 40 kV, 40 mA for Cu KR (λ ) 1.54056 Å). The divergence slit size was 0.5 mm. Scans were collected over a range of 2-40 2θ with a step size of 0.0167° and a counting time 50 s/step. 4. High-Resolution X-ray Diffraction. Synchrotron X-ray diffraction measurements on the tetrahydrate, dihydrate, and hemihydrate were performed on beam-line X3B1 at the National Synchrotron Light Source, Brookhaven National Laboratory; the anhydrate was measured at beamline X16C. The X-ray wavelengths specified in Table 2 were selected by double crystal Si(111) monochromator. The diffracted beam was selected using a Ge(111) analyzer and detected with a Na(Tl)I scintillation counter with a pulse-height discriminator in the counting chain. The diffracted intensity was normalized to the incident beam, monitored by an ion chamber. The powder samples were sealed in 1.5 mm thin-wall glass capillary tubes and X-ray diffraction data were recorded with a step size of 0.005°. During data collection the sample was rotated to reduce the effect of sample granularity. The peaks were fitted by a local deconvolution program and powder data were indexed by the computer program ITO.28 Possible space groups for each hydrate form were derived from the systematic absences. The data were subsequently analyzed with TOPAS software.29 5. Single-Crystal X-ray Analysis. Single-crystal X-ray data were collected on a Rigaku AFC5 4-circle diffractometer using graphite monochromated Cu KR radiation. The crystals were found to be extremely sensitive to loss of solvent and special care was taken as the crystals were mounted in capillaries in the presence of mother liquor. Diffraction data were collected at 296 K. The lattice parameters were determined from 20 reflections with 9° < θ < 14°. A structure solution was obtained using direct methods (SHELXS86) and refined using fullmatrix least-squares on F2 using SHELXL97. The non-hydrogen atoms of the organic moiety were refined anisotropically while the oxygen atoms for the included water molecules were refined isotrpopically. Crystallographic data and parameters relevant to the single-crystal data

Hydrate-Forming Pharmaceutical Compound

Crystal Growth & Design, Vol. 9, No. 4, 2009 1835

Table 1. Crystal Data and Refinement for Pentahydratea Phase of 1 formulaa formula weighta T, K wavelength, Å crystal system space group a, Å b, Å c, Å β, ° V, Å Z Fcalc, g/cm3 absorption coefficient, mm-1 no. of reflection collected no. of data/parameters absorption correction R1b wR2b goodness of fit highest peak in final difference map, e Å3

C23H37N5O8 511.580 296 1.540598 monoclinic C2 36.961(5) 7.458(2) 20.691(4) 99.46(1) 5626(4) 8 1.208 0.73 5151 5054/609 none 0.125 (I > 2σ(I)) 0.342 (I > 2σ(I)) 1.53 0.79

a Only four water molecules were refined. For pentahydrate, the formula is C23H39N5O9 and the formula weight 529.5. b R1 ) Σ|Fc| |Fo|/|Fc|, wR2 ) Σ[w(Fo2 - Fc2)]2/Σ[w(Fo2)2]]1/2.

collection and structure refinement are given in Table 1. Tables of bond distances, bond angles, and anisotropic thermal factors appear as Supporting Information. 6. Molecular Modeling. Preliminary structure determination from synchrotron X-ray powder diffraction pattern was carried out using the MC/SA program PowderSolVe, which is incorporated in the molecular simulation package Material Studio.30 Connolly surface measurement was performed using Cerius2 with a probe radius of 1.4 Å,31 which is the radius of a water molecule. All computer simulations were performed on an SGI Oxygen workstation with a R100000 processor at 250 MHz and a 1993 MHz Pentium II processor-based PC running Windows NT. 7. Sample Preparation for High-Resolution X-ray Diffraction. Synchrotron X-ray data collection sample preparation was as follows: Compound 1 was stored in a glovebox under controlled relative humidity of 30%, 60%, and 75%. At each relative humidity the sample was allowed to stay in the glovebox for 12 h to reach equilibrium. The tube was then placed in a 1.5 mm thin-walled capillary tube, moved out of the glovebox, and quickly flame-sealed. The phase purity was tested on a Philips PW3040-PRO X-ray diffractometer equipped with a multiple wire detector using CuKR radiation monochromated by a hybrid monochromator. The anhydrous capillary sample was prepared in a glovebox filled with dry nitrogen of 99% purity with less than 50 ppm water and flame-sealed quickly outside the glovebox. 8. Sample Preparation for Single-Crystal X-ray Analysis. Compound 1 (20 mg, 0.04 mmol) was dissolved with 10 mL of water in a clean vial. This solution was then moved to a fume-hood, where slow evaporation of the solvent over 5 days afforded thin, plate-shaped single crystals with typical dimensions of 0.08 × 0.25 × 0.50 mm.

Results 1. Single-Crystal Method. Single-crystal X-ray analysis has been the most powerful and reliable method to investigate structures of crystalline solids. Our ultimate goal was to obtain single-crystals of all the five forms of 1 with suitable size and quality for single-crystal X-ray analysis. This goal is, however, extremely challenging considering the metastable nature of the hydrate forms. Fortunately, with the similarity found in the X-ray diffraction patterns for the different hydrate forms, one singlecrystal structure of any hydrate form may provide enough structural information to elucidate all the other hydrate forms. Significant effort was made to grow single-crystals of compound 1 suitable for X-ray analysis. Although many methods and solvents were explored, only from water were large enough single-crystals obtained. Unfortunately even these

crystals were of only marginal quality as evidenced by the higher than normal refinement statistics. There are two crystallographically independent molecules of 1 in a unit cell. These two molecules adopt the same conformation and are geometrically similar. Figure 3 illustrates the conformation and thermal ellipsoids of one of the molecules. The tetrahydronaphthyridine moiety of the compound is almost parallel to the pyridine ring (the dihedral angle between these two rings is 12.9°) and the imidazolidine ring is perpendicular to the pyridine ring and the tetrahydronaphthyridine moiety imagine a pillar linking these two parallel planes. A view of the packing in a unit cell is illustrated in Figure 4. The tetrahydronaphthyridine groups of the two crystallographically nonequivalent molecules are stacked on top of each other, exhibiting a π-π interaction, with a plane-to-plane distance of 3.69 Å. The plane-to-plane distance of the pyridine rings is 3.94 Å with a lateral offset of the ring centroids of 2.85 Å, possibly indicating a weak π-π interaction. Thus, the two independent molecules are stacked together via π-π interaction to form a building block as illustrated in Figure 5. This unit, as shown in Figure 5, has a pseudo 2-fold rotation axis at its center and parallel to the aromatic rings. Four such building blocks are found in a unit cell, two at the top and two at the bottom, as may be seen in Figure 4. The two building blocks diagonal to each other in Figure 4 are related by a crystallographic 2-fold axis at the center of the cell along the b axis. The two building blocks next to each other, either at the top or the bottom, are related by a 2-fold screw axis located between them, also running along the b axis. In this way this building block comprising two molecules of 1 forms the framework of the crystal structure with water molecules sandwiched by the framework. The building blocks stack up along the b axis through weak π-π interactions and van der Waals forces. These one-dimensional columns of stacked building blocks along the b axis are connected, along the c axis, via hydrogen bonds between the carboxylate group and the N atoms of the tetrahydronaphthyridine group to form a two-dimensional layered framework. Figure 6 illustrates the layered structure of compound 1 in which the hydrate water molecules are sandwiched by the layers of the organic moiety. The single-crystal used for X-ray analysis, being grown from water, is likely to be the pentahydrate form, which is the highest hydrate level found. However, in the solved single-crystal structure only four oxygen atoms of the water molecules are located for each 1 molecule, not five as the pentahydrate form would possess. The Fourier difference map indicated that there was some electron density inside the water layer, but it could not be modeled effectively. Both the high R factor and the difficulty in locating the water molecules are characteristic for inclusion compounds with large channels.32-34 In order to characterize the hydrate state of the crystals grown in water, we collected powder X-ray diffraction data of compound 1 suspended in water. This powder pattern is shown in Figure 7 with the powder pattern of 1 collected at 90% relative humidity and the calculated powder pattern based on the single-crystal structure. In Figure 7b,c the powder patterns of 1 suspended in water in a capillary and 1 exposed to 90% relative humidity are essentially identical in peak positions. The powder pattern in Figure 7b showed less of a preferred orientation effect as it was collected from a spinning capillary sample. The amorphous halo observed above 22° 2θ angle is due to diffraction of water in the capillary. The calculated pattern based on the single-crystal structure shown in Figure 7a and the observed one in Figure 7b are in excellent agreement with


Kiang et al.

Table 2. Crystallographic Data from the Synchrotron Rietveld Refinements pentahydratea space group a (Å) b (Å) c (Å) β (°) volume (Å3) ∆V (%)c mol wt ∆MW (%)c Z Z′ λ (Å) 2θ range (°) reflections Rwpd (%) Rwp Pawley (%) expected Rexpe(%) χ2f refined parameters

C2 36.961 7.458 20.691 99.46 5626 23 529.5 20 8 2

tetrahydrate P22121 35.222/2 7.4564 20.6488 90 5423/2 18 511.5 16 4 1 1.14978(2) 1.5-38.95 360 9.291 7.328 2.366 15.42 59

dihydrate P21221 32.518 7.3503 20.6852 90 4945 8 475.5 8 8 2 1.15059(2) 2-33 440 9.456 6.927 2.986 10.03 78

hemihydrate b 1

P2 30.0469 c ) 7.4380 b ) 20.6617 93.479 4609 0.5 448.5 2 8 4 1.15061(1) 2-33 622 6.29 4.03 2.85 9.11 113

anhydrate P22121 29.828/2 7.4635 20.6041 90 4587/2 0 439.5 0 4 1 0.69812(1) 2-30 609 7.024 3.81 2.69 6.81 57

a Data taken from the single crystal structure listed in Table 1. b To maintain the same choice of lattice parameters as the other phases, one could equivalently describe the hemihydrate phase in space group P1121, with lattice parameters a ) 30.05 Å, b ) 7.44 Å, c ) 20.66 Å, γ ) 93.48°. c The rows ∆V (%) and ∆MW (%) contain the percent increase of unit cell volume and molecular weight, respectively, relative to the anhydrate. d The weighted profile R factor is defined as Rwp ) ((∑i wi (yicalc - yiobs)2)/(∑i wi (yiobs)2))1/2, where yicalc and yiobs are the calculated and observed intensities at the ith point in the profile, normalized to monitor intensity. The weight wi is 1/σ2 from counting statistics, with the same normalization factor. N is the number of points in the measured profile. e The expected R factor, is defined as Rexp ) ((N - P)/(∑i wi (yiobs)2))1/2, where P is the number of refined parameters. This is the R factor that would be achieved if the only difference between data and model were due to the Gaussian counting statistics. f χ2 ) (Rwp/Rexp)2.

Figure 3. Thermal ellipsoid diagram of 1 with atom numbering in the single crystal grown in water.

the exception of the first two strong reflections: reflections 200 and 400 at 4.83° and 9.60° 2θ, respectively. The intensity of these two reflections in the calculated pattern is significantly greater than in the observed one. This discrepancy in intensity is because the water molecules were not located completely in the single-crystal structure. In Figure 8, we show the (200) plane represented as a real space plane wave for the 200 reflection. It may be seen that the (200) planes place the electron density of the water molecules sandwiched by the more crystalline compound 1 framework. The relative intensity of reflection 200 corresponds to the contrast of the electron density on the plane wave’s crests and troughs, namely, the electron density of water molecules and the drug framework. As we were unable to locate all water molecules in the single-crystal structure, the calculated 200 intensity is therefore overstated when compared with the observed one. The relative intensity of 400 can be explained in a similar manner. The overall agreement between the calculated and observed patterns is evidence that the single-crystal, grown from water and selected for X-ray analysis, is indeed the same as the bulk pentahydrate form.

Figure 4. Viewing down the b axis of compound 1 single crystal grown in water. Carbon is represented in green, oxygen is in red, and nitrogen is in blue. Hydrogen atoms of water were not included in the refinement.

2. Structural Information from the Synchrotron X-ray Powder Patterns. With only the single-crystal structure of the pentahydrate form available, we started the structural investigation of the other hydrate forms by indexing their synchrotron X-ray powder patterns. These indexed cells were subsequently confirmed by structure solution and refinement, as discussed below. The cell dimensions, cell volumes, molecular weights, percentage volume, and weight change against the anhydrous form are summarized in Table 2, along with certain details of



Figure 5. Stereoview of the building block in the single crystal structure of 1. Carbon is represented in sticks, nitrogen in large spheres, and oxygen in small spheres.

Figure 7. XRD patterns of compound 1 (a) calculated from single crystal structure and observed from (b) immersing sample in water and (c) exposing sample to 90% relative humidity.

Figure 6. Layered structure of the single crystal structure of 1 viewed down the b axis. Compound 1 is represented in green and hydrated water is in red. Hydrogen atoms are removed for clarity.

the data collection and refinement statistics. As the hydrate level changes from anhydrate to pentahydrate, one axis increases from 29.863 to 36.961 Å, while the other two axes remain largely unchanged. From this one-dimensional lattice expansion, one may conclude that the hydrated water is localized in a twodimensional layer in the bc plane; as the hydration level increases, the additional water expands the lattice in the direction normal to the hydrate plane, the bc plane. In Table 2, one sees that the increase in unit cell volume is in good agreement with the weight gain from the addition of water molecules. The moisture isotherm in Figure 1 shows that except the pentahydrate form, each discrete hydrate level has a stable and constant weight over a range of relative humidity, implying that water molecules are well localized in the crystal lattice of these hydrate forms. Assuming that each individual water molecule occupies approximately the same volume of the crystal lattice, the agreement between the increase of lattice volume and the increase of molecular weight leads to a conclusion that the density, and possibly the crystal packing, of compound 1 in crystal lattices are unchanged throughout all hydrate forms. This conclusion is corroborated by the similarity exhibited in the X-ray powder diffraction patterns in Figure 2. Initial attempts to obtain approximate structures from the high-resolution powder X-ray diffraction patterns were made with the computer program PowderSolVe,35 which employs the direct space method in conjunction with a Monte Carlo/

Figure 8. View of single crystal structure of 1 down the b axis with plane waves corresponding to the 200 reflection highlighted. Bold solid lines correspond to wave crest, dotted lines to wave troughs.

Simulated Annealing (MC/SA) search algorithm. Since the single-crystal structure of the pentahydrate form was known and the packing framework was considered to remain mainly unchanged throughout the various hydration states (based on indexed cell parameters) we tried to search a restricted space as this approach has proved successful in modeling the crystal structure of a large flexible pharmaceutical solid.36 Rigid molecular fragments of 1 from the single-crystal structure of the pentahydrate form were used to start the MC/SA search with all torsional degrees of freedom fixed. In all MC/SA searches, the hydrogen atoms of water molecule were removed to further reduce the total degrees of freedom. The results from PowderSolVe were encouraging, but they were not of sufficient quality to be characterized as structure refinements.


Figure 9. Rietveld refinement of the tetrahydrate form of 1. In the upper panel, measured intensity is denoted by dots, and the line is the calculated model. Vertical lines between the panels denote allowed powder peak positions. The bottom trace is the difference observed minus computed.

Subsequently, we used TOPAS29,37 to solve and refine the crystal structures, as described below. TOPAS is particularly useful for refining organic molecular structures of the degree of complexity studied here, as it is able to define a molecule with a z matrix, whereby the molecular structure is specified with all bond lengths and angles fixed, and the position and orientation of the molecule and various torsion angles can be refined. In all of the refinements here, we took the covalent bond distances and angles from the single crystal structure of the pentahydrate. PLATON was used to check all of the refined structures for missing symmetry elements.38 2.1. Tetrahydrate Form. The 2711.5 Å3 volume of the orthorhombic unit cell of the tetrahydrate indicates that there is one molecule of 1 in the irreducible cell. The space group was identified as P22121 from systematic absences in the Pawley fit, but the 21 axes (relative to 2) cannot be regarded as decisive from the small number of absent peaks in powder X-ray diffraction data. Simulated annealing searches with one molecule of 1 and four oxygen atoms (18 parameters) were performed using TOPAS. All four possible space groups were tested; only P22121 gave a plausible result. Subsequently, 12 torsions in the molecule were refined. The resulting fit, shown in Figure 9, is generally satisfactory. The water oxygen positions, and even their number, obtained from this refinement should not be regarded as definitive, although they are all found to be in the channel between molecules as seen in Figure 13a. 2.2. Dihydrate Form. The 4945 Å3 volume of the orthorhombic unit cell of the dihydrate indicates that there are two molecules of 1 in the unit cell. The space group P21221 was suggested by systematic absences in the Pawley fit,39 and this space group is confirmed in the structure solution and refinement. Simulated annealing searches with two molecules of 1 and four oxygen atoms (24 parameters) were performed using TOPAS. Subsequently, 12 torsions in each molecule were refined. The resulting fit, shown in Figure 10, is entirely satisfactory. The conformations (set of refined torsion angles) of the two molecules of 1 were found to be quite close to one another, and to the values from the original pentahydrate solution. Nevertheless, it was found that the fit degraded by a significant amount (Rwp from 9.45% to 11.54%) if the corresponding torsion angles on the two independent molecules were constrained to be equal. However, with the large number of variable parameters relative to observations, one cannot claim

Kiang et al.

Figure 10. Rietveld refinement of the dihydrate form of 1. In the upper panel, measured intensity is denoted by dots, and the line is the calculated model. Vertical lines between the panels denote allowed powder peak positions. The bottom trace is the difference observed minus computed. Two unidentified impurity peaks are observed, at 2θ ) 22.55° (d ) 2.942 Å) and 23.54°; the latter is likely from NaCl; these peaks do not influence the refinement or the structural conclusions drawn.

Figure 11. Rietveld refinement of the hemihydrate form of 1. In the upper panel, measured intensity is denoted by dots, and the line is the calculated model. Vertical lines between the panels denote allowed powder peak positions. The bottom trace is the difference observed minus computed.

that the molecular geometry has been determined with accuracy significantly better than the difference between the two molecules. The refined structure is shown in Figure 13b. Again, water oxygen positions, and even their number, from this refinement should not be regarded as definitive, although they are all located in the channel and a reasonable distance from the molecule. However, the registry between adjacent layers of molecules firmly was established by the diffraction data. 2.3. Hemihydrate Form. The hemihydrate sample was prepared by allowing a sample of 1 to equilibrate at 30% RH. The pattern indexes to a monoclinic cell, with the 20.7 Å axis unique (perpendicular to the other two). While the synchrotron pattern appears to have absences suggesting spacegroup P21/n, that cannot be a correct assignment for a chiral molecule. In order to get the required eight molecules of 1 into the unit cell, we are left with the hypothesis that there are Z′ ) 4 independent molecules in P21. As above, searches starting with rigid


Figure 12. Rietveld refinement of the anhydrate form of 1. In the upper panel, measured intensity is denoted by dots, and the line is the calculated model. Vertical lines between the panels denote allowed powder peak positions. The bottom trace is the difference observed minus computed. This sample contains a small amount of an impurity phase, evidently NaCl (peaks at 14.22°, 20.16°, etc.) as well as an unidentified impurity peak at 13.87° (d ) 2.891 Å).

molecules in approximately the correct orientation, followed by refinements including all torsions, gave a satisfactory fit to the data (Figure 11). As in the cases previously discussed, the Rietveld fits degrade somewhat if the torsions of the four independent molecules are constrained to be equal; the number and positions of the solvent oxygens are subject to significant uncertainty, but the framework of molecules of the active pharmaceutical ingredient is well established. Figure 13c illustrates the refined hemihydrate structure. 2.4. Anhydrate Form. The anhydrate has the same spacegroup as the tetrahydrate, P22121, but it is obviously a different phase since the a axis has contracted by 15%. Again, the space group suggested by systematic absences is confirmed by the refinement, shown in Figure 12. The refined structure of the anhydrate is shown in Figure 13d. 3. Connolly Surface Measurement. The structures of the hydrate forms obtained from high-resolution powder data exhibit a layered framework for compound 1 that is similar to the framework found in the single-crystal structure of the pentahydrate form. Water molecules were found in the space between the compound 1 layers. However, in the structures obtained from Rietveld refinement, the atom coordinates of the water molecules are less certainly located compared to the compound 1 molecules. Connolly surface measurement was therefore used to examine the water accessible site inside the unit cells of compound 1 crystal structures where existing water molecules were removed from the structures.40 The probe radius of this measurement was 1.4 Å, the radius of a water molecule. The results of the Connolly surface measurement (Figure 14) show that the water accessible sites were located between the compound 1 layers in all hydrate forms. In the hemihydrate structure, due to the reduced volume of the void space between the compound 1 layers, isolated water pockets were found instead of hydrate layers. As expected, the Connolly surface measurement resulted in no water accessible site in the anhydrate form. Discussion The frameworks of compound 1 in all five hydrate states are comprised of the dimer building block illustrated in Figure 5. As discussed above in section 2.1, the building blocks form a


two-dimensional layered network along the bc plane via hydrogen-bonding and π-π interaction. These two-dimensional layers stack along the a axis and, except for the anhydrate, are separated by hydrated water layers as seen in Figure 13. In the pentahydrate structure, the layers of compound 1 are parallel to the bc plane and stack along the a axis, which is at an angle of 99.4° to the c axis. The building blocks in adjacent layers are therefore shifted in the same direction along the c axis with an offset of 3.04 Å. This packing pattern may be visualized in Figure 15a, in which three compound 1 layers are represented by a 3 × 3 array of the building blocks. As one may see in Figure 15a, the nine building blocks occupy two different heights along the b axis that are separated by half the length of the b axis. The building block at the center of the array in Figure 15a lies above the plane of the paper. The building blocks adjacent to the center one are all at the height that is beneath the plane of the paper. If one connects the centroids of the building blocks along the a axis, viewing down the c direction, one sees two zigzags with a vertical offset 2/5 of the length of the a axis (14.78 Å), shown in Figure 16a. When the centroids of the building blocks are connected along the c axis, viewing down the a direction, two zigzags also can be found as shown in Figure 16b. These two zigzags appear to be identical with a vertical offset about one-half the length of the c axis. However, the centroids, represented by circles, are not evenly spaced along the c axis in Figure 16b. Along the c axis, two adjacent centroids are spaced 0.47 and 0.53 the length of the c axis (9.72 and 10.97 Å). Neither plane of symmetry nor center of symmetry is present in the zigzag, which makes the zigzag two-dimensionally chiral. The two zigzags in Figure 16b are actually mirror images with the mirror plane perpendicular to the paper and parallel to the ac plane. Thus, the two zigzags cannot be superimposed no matter how one zigzag shifts along the c axis. The packings of the building blocks in the dihydrate and hemihydrate structures are shown in Figure 15b,c. As in the pentahydrate structure, the molecules comprising the dimers are crystallographically inequivalent. In the dihydrate structure, the building blocks stack up along the a direction such that the adjacent layers are shifted 1.17 Å along the c axis in alternating directions. The hemihydrate structure adopts the same packing with a larger offset of 3.26 Å, three times greater than that of the dihydrate form. The zigzag patterns found in the dihydrate and the hemihydrate are illustrated in Figure 16c-f. When the centroids of the building blocks are connected along the a axis, viewing down the c axis, in both the dihydrate and hemihydrate (Figure 16c,e) the two zigzags have a vertical offset 2/5 the length of the a axis (13.01 Å for the dihydrate and 12.04 Å for the hemihydrate), similar to the zigzag pattern of the pentahydrate shown in Figure 16a. Figure 16d,f shows the zigzag patterns of the connected centroids along the c axis, viewing down the a direction, for the dihydrate and hemihydrate, respectively. Unlike the pentahydrate zigzag pattern when viewed down the a axis in Figure 16b, the centroids of the building blocks in the two zigzags shown in Figure 16d,f are evenly spaced. In the dihydrate structure the lateral offset of the two zigzags is 11.38 Å, while it is 13.64 Å in the hemihydrate structure. The organic framework of the tetrahydrate and anhydrate are isostructural as seen in Figure 17. In contrast to the other three phases, the two molecules of each dimer are related by the 2-fold axis along a. In the tetrahydrate structure, the adjacent compound 1 layers register with one another in such a way that the building blocks stack one on top of each other along the a axis in an


Kiang et al.

Figure 13. Viewing down the b axis of Rietveld refined crystal structures of compound 1 (a) tetrahydrate, (b) dihydrate, (c) hemihydrate, and (d) anhydrate. Carbon: green; oxygen: red; nitrogen: blue; hydrogen: white. Hydrated water is represented by red spheres.

Figure 14. View of compound 1 (a) pentahydrate, (b) tetrahydrate, (c) dihydrate, (d) hemihydrate, and (e) anhydrate along the short axis with the Connolly surface in gray.

eclipsing fashion. The center building block in the 3 × 3 array in Figure 17a sits beneath the plane of the paper while the ones next to it along the c axis lie above the plane of the paper. When the centroids of the building blocks are connected along both the a and c directions, only viewing down the a axis does one see the zigzag pattern (Figure 18b). Two straight lines, which are half the length of the b axis apart, are observed when viewed down the c direction (Figure 18a), with a vertical offset 1/5 the length of the a axis, which is 7.05 Å for the tetrahydrate. The

schematic representation of the building block packing for the anhydrate is shown in Figure 18c,d, which are almost identical to Figure 18a,b, further showing that the anhydrate is isostructural to the tetrahydrate. The interlayer spacing of compound 1 depends on the hydrate levels. A plot of the length of the a axis projected on the normal of the bc plane against hydration state is shown in Figure 19. Linear correlation was found in Figure 19 as the data points were fitted into a linear regression with a correlation coefficient



Figure 15. Stereoview of compound 1 framework in (a) pentahydrate, (b) dihydrate, and (c) hemihydrate viewing down the b axis. Carbon is represented by sticks, nitrogen by large spheres, and oxygen by small spheres. Hydrogen atoms are omitted for clarity. For sake of structure comparison, the unique axis of hemihydrate is changed to the c axis.

0.9981. The regression equation derived is Y ) 1.37X + 29.66, where Y is the length of a projected on the normal of the bc plane and X is the hydration state. From this equation each additional hydrated water molecule per molecule of compound 1 increases the interlayer spacing by 1.37 Å, which is half the diameter of a water molecule. Despite the simple relationship between water occupancy and layer spacing, the absorbed water clearly plays a more complicated role than producing an inert filler layer. Starting with the anhydrate, and regarding the bc plane as a raft of dimers, one can imagine sliding it over a similar raft to find the best fit. If one water molecule per dimer is inserted between the layers, it takes a position that changes the best fit between adjacent rafts, and therefore the spacegroup of the crystal structure. Further addition of water molecules creates a specific sequence of advantageous connections between layers, indicated by the observed set of hydrate phases. The situation at hand is slightly more complicated, because the two-dimensional structure of the raft is different in the case of anhydrate/tetrahydrate vs hemi-/tetra-/pentahydrate. Therefore, the energy differences associated with the stacking transitions are roughly competitive with the van der Waals forces that hold the rafts together in the crystallographic b direction. The five crystallographically distinct states of hydration of compound 1 appear to be an unusually large number. To put this in context, we have searched the Cambridge Crystal

Figure 16. Schematic representation of the centroids of the building blocks in the pentahydrate (a and b), dihydrate (c and d), and hemidyrate (e and f); showing a 4 × 4 array of centroids represented by circles. Two directions of views: in (a), (c), and (d) the centroids of the building blocks are connected along the a axis, viewed down the c axis, while in (b), (d), and (f) they are connected along the c axis, viewed down the a axis.


Kiang et al.

Figure 19. Plot of the cell parameter a axis projected onto the normal of the bc plane against the hydrate state.

Figure 17. Stereoview of compound 1 framework in (a) tetrahydrate and (b) anhydrate viewing down the b axis. Carbon is represented by sticks, nitrogen by large spheres, and oxygen by small spheres. Hydrogen atoms are omitted for clarity.

Figure 20. Photograph of a tablet of 1 in stored at 40 °C/75% RH open dish after 3 weeks.

Figure 18. Schematic representation of the centroids of the building blocks in the tetrahydrate (a and b), and anhydrate (c and d); showing a 5 × 5 array of centroids represented by circles. Two directions of views: in (a) and (c) the centroides of the building blocks are connected along the a axis, viewed down the c axis, while in (b) and (d) they are connected along the c axis, viewed down the a axis.

Structure Database (CCSD) for other organic (or organometallic) compounds with multiple hydrates. Our criteria were to consider only systems with well-defined water content, and to count as distinct only those structures that had different space groups, or had significantly different lattice parameters within one space group, thereby eliminating a number of clathrates. This search was made possible by van de Streek’s compilation of solvates in the CCSD.41 That list of solvates was sorted by the number of occurrences of the word “hydrate” in the compound name field. We restricted our attention to CCSD entries for which crystal structures are given, and for which water is the only solvent (e.g., we did not include any structure with an alcohol and water simultaneously present), but we did include the desolvated phase, if present. According to our search, the largest number of distinct hydrates occurs for 18-crown-6 ether, which has structures with 0, 4, 6, 8, and 12 waters of solvation.42-44

The compound identified as pharmaceutically active with the largest number of distinct hydrates is olanzapine, with an anhydrate, three dihydrate polymorphs, and a higher hydrate.45 Compound 1 therefore ties the record in the CCSD, and significantly surpasses any other known drug compound, for greatest number of distinct states of hydration. Returning to the original motivation for this work, tablet formulations of compound 1 were developed for clinical studies. The RH of the manufacturing area was controlled so that compound 1 was in its hemihydrate during the manufacturing of the tablets. When the tablets were stored at 30C/60%RH and 40C/75%RH open dish conditions, they showed cracking within weeks of the initiation of the stability study. Conversion of compound 1 to the dihydrate or tetrahydrate was detected by powder X-ray diffraction in the tablets under those conditions. Figure 20 shows a tablet containing compound 1 stored at 40 °C/75% RH open dish for three weeks. Placebo tablets stored under those conditions did not show signs of cracking for months. This shows that the conversion between the different hydration states of compound 1 is responsible for tablet cracking. The volume expansion associated with the crystal lattice as the hemihydrate is converted to the dihydrate or tetrahydrate is the driver for the cracking. In layered materials such as clays, the two-dimensional host structure can absorb water between the layers, causing them to expand. These same properties are observed in the development


pharmaceutical compound 1. It is interesting to note that clays do not generally exhibit powder diffraction patterns indicative of such a high degree of crystallinity as what we observe here, nor as many distinct steps in the sorption isotherm. At this point, we do not have any specific understanding of what chemical or structural feature of 1 produces this unusual behavior. Acknowledgment. The authors are grateful to Ms. Laura Debusi for providing the picture of a cracked tablet containing compound 1. We thank Professor Stephen Lee of Cornell University for the use of his X-ray diffractometer. Use of the National Synchrotron Light Source, Brookhaven National Laboratory, was supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Contract No. DE-AC02-98CH10886. The SUNY X3 beamline at NSLS was previously supported by the Division of Basic Energy Sciences of the U.S. Department of Energy under Grant No. DE-FG02-86ER45231. Supporting Information Available: Tables of bond distances, bond angles, and anisotropic thermal factors. This material is available free of charge via the Internet at http://pubs.acs.org.

References (1) Zielke, R. C.; Pinnavania, T. J.; Mortland, M. M. Adsorption and Reactions of Selected Organic Molecules on Clay Mineral Surfaces; Soil Science Society of America: Madison, 1989; pp 81-97. (2) Mallouk, T. E.; Gavin, J. A. Molecular recognition in lamellar solids and thin films. Acc. Chem. Res. 1998, 31, 209–217. (3) O’hare, D. Organic and organometallic guests intercalated in layered lattices. New J. Chem. 1994, 18, 989–998. (4) Newman, S. P.; William, J. Synthesis, characterization and application of layered double hydroxides containing organic guests. New J. Chem. 1998, 22, 105–115. (5) Sposito, G. The Chemistry of Soils; Oxford University Press: New York, 1989. (6) Liebau, F. Structural Chemistry of Silicates; Springer-Verlag: Berlin, 1985. (7) Biradha, K.; Dennis, D.; MacKinnon, V.; Sharma, C.; Zaworotko, M. Supramolecular synthesis of organic laminates with affinity for aromatic guests: a new class of clay mimics. J. Am. Chem. Soc. 1998, 120, 11894–11903. (8) Coleman, A. W.; Bott, S. G.; Morley, S. D.; Means, C. M.; Robinson, K. D.; Zhang, H.; Atwood, J. L. Sodium calix[4] arenesulfonate with a layered structure. An organic clay? Angew. Chem., Int. Ed. Engl. 1988, 27, 1361–1362. (9) Matsumoto, A.; Oshita, S.; Fujioka, D. A novel organic intercalation system with layered polymer crystals as the host compounds derived from 1,3-diene carboxylic acids. J. Am. Chem. Soc. 2002, 124 (46), 13749–56. (10) Brouwer, E. B.; Udachin, K. A.; Enright, G. D.; Ripmeester, J. A.; Ooms, K. J.; Halchuk, P. A. Self-inclusion and paraffin intercalation of the p-ter-butylcalix[4]arene host: a neutral organic clay mimic. Chem. Commun. 2001, 565–566. (11) Higgos, J. J. W. Clay as a barrier to radionuclide migration. Prog. Nucl. Energy 1987, 19, 173–207. (12) Pusch, R. Use of bentonite for isolation of radioactive waste products. Clay Minerals 1992, 27, 353–361. (13) Karaborni, S.; Smit, B.; Heidug, W.; Urai, J.; van Oort, E. The swelling of clays: molecular simulations of the hydration of montmorillonite. Science 1996, 271, 1102–1104. (14) Young, D. A.; Smith, D. E. Simulation of clay mineral swelling and hydration: dependence upon interlayer ion size and charge. J. Phys. Chem. B 2000, 104, 9163–9170. (15) Mooney, R. W.; Keenan, A. G.; Wood, L. A. Adsorption of water vapor by montmorillonite. II. Effect of exchangeable ions and lattice swelling as measured by X-ray diffraction. J. Am. Soc. Chem. 1952, 74, 1371–1374. (16) Hensen, E. J.; Smit, B. Why clays swell. J. Phys. Chem. B 2002, 106, 12664–12667. (17) Morris, K. R.; Fakes, M. G.; Thakur, A. B.; Newman, A. W.; Singh, A. K.; Venit, J. J.; Spagnuolo, C. J.; Serajuddin, A. T. M. An integrated approach to the selection of optimal salt form for a new drug candidate. Int. J. Pharm. 1994, 105, 209–217.

Crystal Growth & Design, Vol. 9, No. 4, 2009 1843 (18) Morris, K. R. Structural aspects of hydrates and solvates. In Polymorphism in Pharmaceutical Solids; Brittain, H. G., Ed.; Marcel Dekker Inc.: New York, 1999; pp 125-181. (19) Stahl, P. H. The problems of drug interaction with excipients. In Towards Better Safety of Drugs and Pharmaceutical Products; Draimar, D. D., Ed.; Elsevier: Amsterdam, 1980; pp 265-280. (20) Kuhnert-Brandsta¨tter, M.; Gasser, P. Solvates and polymorphic modifications of steroid hormones. III. Microchem. J. 1971, 16, 590–601. (21) Kuhnert-Brandsta¨tter, M. Thermomicroscopy in the Analysis of Pharmaceuticals; Pergamon Press: New York, 1971. (22) Giron, D.; Goldbronn, C.; Mutz, M.; Pfeffer, S.; Piechon, P.; Schwab, P. Solid State Characterizations of Pharmaceutical Hydrates. J. Therm. Anal. Calorim. 2002, 68 (2), 453–465. (23) Cox, J. S. G.; Woodard, G. D.; McCrone, W. C. Solid-state chemistry of cromolyn sodium (disodium cromoglycate). J. Pharm. Sci. 1971, 60, 1458–1465. (24) Stephenson, G. A.; Diseroad, B. A. Structural relationship and desolvation behavior of cromolyn, cefazolin and fenoprofen sodium hydrates. Int. J. Pharm. 2000, 198 (2), 167–77. (25) Karle, I. L.; Ranganathan, D.; Haridas, V. A persistent preference for layer motifs in self-assemblies of squarates and hydrogen squarates by hydrogen bonding [X-H · · · O; X ) N, O, or C]: A crystallographic study of five organic salts. J. Am. Chem. Soc. 1996, 118, 7128–7133. (26) Gong, B.; Zheng, C.; Skrzypczak-Jankun, E.; Zhu, J. Two-dimensional molecular layers: interplay of H-bonding and van der waals interactions in the self-assembly of N,N′-dialkylsulfamides. Org. Lett. 2000 2, (21), 3273–3275. (27) Chang, Y. L.; West, M. A.; Howler, F. W.; Lauher, J. W. An approach to the design of molecular solids. Strategies for controlling the assembly of molecules into two-dimensional layerd structures. J. Am. Chem. Soc. 1993, 115, 5991–6000. (28) Visser, J. A fully automatic program for finding the unit cell from powder data. J. Appl. Crystallogr. 1969, 2, 89–95. (29) Bruker AXS TOPAS V3: General Profile and Structure Analysis Software for Powder Diffraction Data, User’s Manual; Bruker AXS: Karlsruhe, Germany, 2005. (30) Material Studio; Accelrys Inc: San Diego, 2001. (31) Cerius2; Accelrys Inc: San Diego, 2001. (32) Venkataraman, D.; Lee, S.; Zhang, J.; Moore, J. S. An organic solid with wide channels based on hydrogen bonding between macrocycles. Nature 1994, 371, 591–593. (33) Gardner, G. B.; Venkataraman, D.; Moore, J. S.; Lee, S. Spontaneous assembly of a hinged coordination network. Nature 1995, 374, 792–795. (34) Kiang, Y. H.; Gardner, G. B.; Lee, S.; Xu, Z.; Lobkovsky, E. B. Variable pore size, variable chemical functionality, and an example of reactivity within porous phenylacetylene silver salts. J. Am. Chem. Soc. 1999, 121, 8204–8215. (35) Engle, G. E.; Wilke, S.; Ko¨nig, O.; Harris, K. D. M.; Leusen, F. J. J. PowderSolve: A Complete Package for Crystal Structure Solution from Powder Diffraction Patterns. J. Appl. Cryst. 1999, 32, 1169–1179. (36) Kiang, Y. H.; Huq, A.; Stephens, P. W.; Xu, W. Structure determination of enalapril maleate form II from high-resolution X-ray powder diffraction data. J. Pharm. Sci. 2003, 92 (9), 1844–1853. (37) Coelho, A. A. Whole-profile structure solution from powder diffraction data using simulated annealing. J. Appl. Crystallogr. 2000, 33, 899– 908. (38) Spek, A. L. PLATON, A multipurpose crystallographic tool. J. Appl. Crystallogr. 2003, 36, 7–13. (39) Pawley, G. S. Unit-cell refinement from powder diffraction scans. J. Appl. Crystallogr. 1981, 14, 357–361. (40) Connolly, M. L. Analytical molecular surface calculation. J. Appl. Crystallogr. 1983, 16, 548–558. (41) van de Streek, J. All series of multiple solvates (including hydrates) from the Cambridge Structural Database. Cryst. Eng. Comm. 2007, 9, 350–352. (42) Dunitz, J. D.; Seiler, P. 1,4,7,10,13,16-Hexaoxacyclootadecane. Acta Crystallogr. B 1974, B30 (11), 2739–1741. (43) Mootz, D.; Albert, A.; Schaefgen, S.; Staeben, D. Hydrates of weak and strong bases. 12. 18-Crown-6 and water: crystal structure of a binary hydrate. J. Am. Soc. Chem. 1994, 116 (26), 12045–12046. (44) Albert, A.; Mootz, D. Hydrates of weak and strong bases. Part 17. Formation and crystal structures of the hydrates of 18-crown-6. Z. Naturforsch. B 1997, 52 (5), 615–619. (45) Reutzel-Edens, S. M.; Bush, J. K.; Magee, P. A.; Stephenson, G. A.; Byrn, S. R. Anhydrates and hydrates of olanzapine: crystallization, solid-state characterization, and structural relationships. Cryst. Growth. Des. 2003, 3, 897–907.

CG801004A

Layered Structure and Swelling Behavior of a Multiple Hydrate

Recommend Documents