CRYSTAL
`GROWTH
`& DESIGN
`2003
`VOL.3,NO.6
`897-907
`
`Articles
`
`Anhydrates and Hydrates of Olanzapine: Crystallization,
`Solid-State Characterization, and Structural
`Relationships
`
`Susan M. Reutzel-Edens,*,† Julie K. Bush,† Paula A. Magee,†
`Greg A. Stephenson,† and Stephen R. Byrn‡
`Eli Lilly and Company, Indianapolis, Indiana 46285, and Department of Medicinal
`Chemistry and Pharmacognosy, Purdue University, West Lafayette, Indiana 47907
`
`Received April 8, 2003; Revised Manuscript Received June 11, 2003
`W This paper contains enhanced objects available on the Internet at http://pubs.acs.org/crystal.
`
`ABSTRACT: Olanzapine, a novel benzodiazepine agent used in the treatment of schizophrenia and related psychoses,
`crystallizes in 25+ crystal forms, seven of which are pharmaceutically relevant: three anhydrates (I-III), three
`dihydrates (B, D, and E), and a higher hydrate. X-ray crystal structures of the thermodynamically stable anhydrous
`form (I), two dihydrates (B and D), a higher hydrate, and a Rietveld-refined structure of dihydrate E have permitted
`a detailed analysis of the conformational, hydrogen bonding, and crystal packing preferences of olanzapine. The
`symmetry and hydrogen-bonding interactions in the crystal forms have also been characterized by 13C and 15N
`CP/MAS NMR spectroscopy. Using the crystallographic and spectroscopic data, significant structural relationships
`have been identified between the crystal forms of olanzapine. The present study demonstrates the utility of integrating
`crystallography, spectroscopy, and crystal modeling in detailed structural investigations of polymorphism (and solvate
`formation) and for rationalizing crystallization outcomes. This study also shows that polymorphism and hydrate
`formation can be used to optimize the physical presentation of pharmaceutical solids.
`
`Introduction
`
`One of the primary goals of crystal engineering1,2 is
`to design and control the way molecules crystallize,
`producing materials with specific properties (e.g., second-
`harmonic generation, conductivity, thermochromism,
`photoactivity, etc.). Typical strategies direct molecular
`association through strong intermolecular interactions,
`such as hydrogen bonding, electrostatic and/or charge-
`transfer interactions, and control bulk properties by
`simply varying molecular structure.3-5 If crystals of
`pharmaceuticals could be engineered, then properties,
`such as stability, bioavailability, and processibility,
`could be optimized.6 Traditional approaches to crystal
`engineering are generally not applicable to pharmaceu-
`tical solids, however, since only limited changes to
`molecular structure can be tolerated to design a bulk
`drug material with optimal physical properties. Indeed,
`
`* Corresponding author. Tel: (317) 276-0994. Fax (317) 277-8387.
`E-mail: reutzel@lilly.com.
`† Eli Lilly and Company.
`‡ Purdue University.
`
`structural modifications to drug molecules, most com-
`monly in the form of prodrugs, are typically driven by
`bioavailability considerations.7,8
`Given the structural limitations placed on pharma-
`ceuticals, different approaches, including salt forma-
`tion,9-11 complexation,12-16 cocrystallization,17,18 solvate
`formation,19 and polymorphism, have been used to
`manipulate the supramolecular structure in pharma-
`ceutical solids. Of these methods, salt formation, com-
`plexation, cocrystallization, and solvate formation are
`limited by the toxicity of the counterions, guest mol-
`ecules, and solvents. Salt formation is also obviously
`limited to compounds with ionizable groups. Polymor-
`phism,20,21 considered to many to be the nemesis to
`crystal engineering,22 and hydrate formation are wide-
`spread phenomena that can be viewed as opportunities
`to safely manipulate the physical properties of phar-
`maceutical solids.23 In polymorphic solids, structure-
`property relationships are governed only by differences
`in the spatial arrangement of molecules in a crystal, and
`in some cases, variations in molecular conformation.24
`While true polymorphs can have significantly different
`
`10.1021/cg034055z CCC: $25.00 © 2003 American Chemical Society
`Published on Web 07/02/2003
`
`Downloaded via PENNSYLVANIA STATE UNIV on May 23, 2020 at 23:52:11 (UTC).
`
`See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
`
`Merck Exhibit 2047, Page 1
`Mylan Pharmaceuticals Inc. v. Merck Sharp & Dohme Corp.
`IPR2020-00040
`
`

`

`898 Crystal Growth & Design, Vol. 3, No. 6, 2003
`
`Reutzel-Edens et al.
`
`physical properties, the incorporation of water in a
`crystal lattice (i.e., formation of a hydrate) can have
`even more dramatic effects on the physical properties
`of a pharmaceutical solid. Hence, controlling polymor-
`phism in pharmaceutical solids must include provisions
`for hydrate formation, and vice versa.
`Olanzapine, 2-methyl-4-(4-methyl-1-piperazinyl)-10H-
`thieno[2,3-b][1,5]benzo-diazepine, is a member of a novel
`benzodiazepine class of antipsychotic drugs with dem-
`onstrated efficacy in the treatment of schizophrenia and
`related psychoses.25,26 Polymorphism and hydrate for-
`mation have proven to be particularly powerful means
`to alternate crystal forms of the drug. Olanzapine
`crystallizes in at least 25 solid forms, including three
`polymorphic anhydrates (I-III), three polymorphic di-
`hydrates (B, D, and E), and a higher hydrate. This paper
`reports the preparation and structural characteriza-
`tion of these seven pharmaceutically relevant crystal
`forms by X-ray crystallography and solid-state NMR
`spectroscopy. The molecular recognition processes re-
`sponsible for the polymorphism and hydrate formation
`of olanzapine have been examined, and the structural
`relationships between the anhydrates and the hy-
`drates have been used to rationalize its crystallization
`behavior.
`
`Experimental Procedures
`Materials. Olanzapine was provided by Lilly Research
`Laboratories.
`Form I. Olanzapine (270 g) was suspended in ethyl acetate
`(2.6 L). The stirred suspension was heated to 76 °C to dissolve
`the solids. The solution was then cooled to ambient temper-
`ature, at which time a crystal slurry formed. The solid product
`was isolated by vacuum filtration and dried in vacuo at 50 °C.
`Yield ) 197 g, TGA mass loss (exptl) 0.0%.
`Form II. A sample of mostly form II was prepared by
`desolvating olanzapine methanolate at 50 °C. TGA mass loss
`(exptl) 0.0%.
`Form III. Olanzapine (1.5 g) was suspended in CHCl3 (5
`mL). The suspension was heated to reflux to dissolve the solids.
`The solution was then cooled to ambient temperature. Hexanes
`(15 mL) were added to the stirred solution, at which time a
`crystal slurry formed. The solid precipitate (mostly form III)
`was isolated by vacuum filtration and washed with hexanes
`(10 mL). Yield ) 867 mg.
`Dihydrate B. Olanzapine (5 g) was suspended in ethyl
`acetate (50 mL) and toluene (6 mL). The suspension was
`stirred and heated to 80 °C to dissolve the solids. The solution
`was then cooled to 60 °C, and water (30 mL) was added. The
`solution was further cooled to room temperature to produce a
`crystal slurry. Yellow, rhombohedral crystals were isolated by
`vacuum filtration, washed with H2O (10 mL), and air-dried.
`Yield ) 4.16 g, TGA mass loss (exptl) 10.1%, (theory) 10.3%.
`Dihydrate D. Olanzapine form I (5 g) was suspended in
`water (50 mL) at ambient temperature and slurried for 5 days.
`The solid product was isolated by vacuum filtration, washed
`with water (20 mL), and air-dried. Yield ) 3.9 g, TGA mass
`loss (exptl) 10.2%, (theory) 10.3%.
`
`Dihydrate E. Olanzapine (3 g) was suspended in ethyl
`acetate (60 mL) and toluene (3.6 mL). The suspension was
`stirred and heated to 80 °C to dissolve the solids. The solution
`was then cooled to 65 °C, and water (6 mL) was added. The
`solution was further cooled to ambient temperature to produce
`a crystal slurry. Yellow, rhombohedral crystals were isolated
`by vacuum filtration, washed with H2O (5 mL), and air-dried.
`Yield ) 2.47 g, TGA mass loss (exptl) 10.6%, (theory) 10.3%.
`Higher Hydrate. Olanzapine (2 g) was suspended in
`CH2Cl2 (12 mL). The suspension was stirred and heated to
`reflux to dissolve the solids. Water (1.5 mL) was added as the
`solution was cooled to ambient temperature, at which time a
`crystal slurry formed. The crystal slurry was cooled to 0 °C,
`and a wetcake (2.7 g) was isolated by vacuum filtration and
`washed with CH2Cl2 ((cid:24)10 mL).
`General Methods. Thermogravimetric analyses were per-
`formed on a Seiko Simultaneous Thermo-Gravimetric Analyzer
`Model 220. Samples (3.5 mg) were run from 25 to 350 °C at a
`rate of 10 °C/min.
`XRD patterns were obtained on a Siemens D5000 X-ray
`powder diffractometer, equipped with a CuKR source ((cid:236) )
`1.54056 Å) and a Kevex solid-state detector, operating at 50
`kV and 40 mA. Each sample was scanned between 4 and 35°
`in 2ı, with a step size of 0.03° and a scan rate of 2 s/step.
`Solid-state 13C and 15N NMR spectra were collected on a
`Varian Unity spectrometer operating at a 1H resonance
`frequency of 400 MHz. All experiments were performed using
`cross polarization (CP), high power decoupling, and magic
`angle spinning (MAS ) 7 kHz). Hartmann-Hahn match
`parameters for 13C and 15N were determined using hexa-
`methylbenzene (HMB) and glycine-15N, respectively. Typical
`13C acquisition parameters include 90° pulse width 5 (cid:237)s,
`contact time 1.1 ms, relaxation delay 5 s, acquisition time 0.05
`s, and spectral width 50 kHz. Interrupted decoupling spectra
`were acquired with a 40 or 50 (cid:237)s delay without decoupling
`prior to acquisition. Chemical shifts were referenced using
`sample replacement to the methyl group of HMB, which
`resonates at 17.3 ppm. Typical 15N acquisition parameters
`include 90° pulse width 7 (cid:237)s, contact time 2.5 ms, relaxation
`delay 5 s, acquisition time 0.1 s, and spectral width 35 kHz.
`Chemical shifts were referenced using sample replacement to
`glycine-15N, which resonates at -6.39 ppm [15NH4Cl ) 0.0
`ppm].
`Crystallographic Literature Search. A search of the
`Cambridge Structural Database27 was conducted for benzo-
`diazepines. From a connectivity search of the fragments shown
`below, six uncharged structures (five unique), for which
`coordinates are available, were retrieved (see Supporting
`Information).
`
`OR
`
`X-ray Structure Determinations. Single crystals of form
`I were grown by vapor diffusion of n-pentane into a dry ethyl
`acetate solution of olanzapine. Dihydrate B crystals were
`obtained by diffusing water into a saturated toluene solution
`of olanzapine. Dihydrate D crystals were obtained by slow
`evaporation of a 3:1 acetonitrile/water solution of olanzapine.
`Single crystals of the higher hydrate were isolated from ethyl
`acetate/toluene/water.
`Crystal data for form I were collected on an Enraf-Nonius
`CAD4 diffractometer. Three standard reflections were meas-
`ured every 97 reflections; no crystal decay was detected.
`Lorentz and polarization corrections were applied to the data;
`no corrections were made for absorption. The structures were
`solved by direct methods using Shelx8628, and the remaining
`atoms were located in succeeding difference Fourier synthesis.
`The structures were refined in full-matrix least-squares using
`MolEN,29 where the function minimized was (cid:229) w(jjFoj - jFcjj)2,
`and the weight w is defined per the Killean and Lawrence
`
`Merck Exhibit 2047, Page 2
`Mylan Pharmaceuticals Inc. v. Merck Sharp & Dohme Corp.
`IPR2020-00040
`
`

`

`Olanzapine Anhydrates and Hydrates
`
`Crystal Growth & Design, Vol. 3, No. 6, 2003 899
`
`Table 1. Crystal Data, Data Collection, and/or Refinement for Olanzapine Form I and the Hydrates
`form I
`dihydrate B
`dihydrate D
`dihydrate Ea
`higher hydrate
`C17H20N4S(cid:226)2 H2O
`C17H20N4S(cid:226)2 H2O
`C17H20N4S(cid:226)2 H2O
`C17H20N4S(cid:226)2.5 H2O
`C17H20N4S
`312.44
`348.47
`348.47
`348.47
`357.47
`0.22 (cid:2) 0.13 (cid:2) 0.10
`0.11 (cid:2) 0.24 (cid:2) 0.36
`0.20 (cid:2) 0.22 (cid:2) 0.25
`0.26 (cid:2) 0.24 (cid:2) 0.21
`NA
`monoclinic
`monoclinic
`triclinic
`monoclinic
`monoclinic
`P21/c (no. 14)
`P21/c (no. 14)
`P-1(no. 2)
`C2/c (no. 15)
`C2/c (no. 15)
`10.383(1)
`9.8691(12)
`9.927(5)
`24.5195
`25.130(2)
`14.826(3)
`12.7156(15)
`10.095(5)
`12.3495
`12.2377(11)
`10.560(8)
`14.3853(16)
`10.514(6)
`15.2179
`14.9116(14)
`90
`90
`84.710(10)
`90
`90
`100.616
`92.969(2)
`62.665(8)
`125.824
`124.984(1)
`90
`90
`71.183(8)
`90
`90
`1597.8(7)
`1802.8(4)
`884.1(8)
`3736.3
`3757.2(6)
`4
`4
`2
`8
`8
`1.299
`1.284
`1.309
`1.239
`1.274
`293
`173(2)
`128(2)
`198(2)
`CuKR
`MoKR
`MoKR
`MoKR
`1.54184
`0.71073
`0.71073
`0.71073
`none
`graphite
`graphite
`graphite
`17.61
`0.197
`0.200
`0.196
`0-11
`-13 to 13
`-13 to 13
`-16 to 33
`0-16
`-11 to 16
`-13 to 13
`-12 to 15
`-11 to 11
`-19 to 11
`-13 to 13
`-18 to 16
`2-60
`2.07-28.29
`1.07-14.11
`1.94-28.27
`664.0
`744.0
`372
`1568
`2485
`4193
`3976
`4378
`2139 (I > 3(cid:243)(I))
`2417 (I > 2(cid:243)(I))
`3119 (I > 2(cid:243)(I))
`2923 (I > 3(cid:243)(I))
`279
`227
`27
`230
`0.16
`0.004
`0.01
`0.172
`0.043
`0.0663
`0.0829
`0.0730
`0.057
`0.1525
`0.2530
`0.2073
`2.007
`0.974
`1.148
`1.040
`
`experimental formula
`formula weight
`crystal dim. (mm)
`crystal system
`space group
`a (Å)
`b (Å)
`c (Å)
`R (deg)
`(cid:226) (deg)
`(cid:231) (deg)
`V (Å3)
`Z
`Fcalc (g cm-3)
`temperature (K)
`radiation
`wavelength (Å)
`monochromator
`abs. coeff. (cm-1)
`h
`k
`l
`ı range (deg)
`F000
`no. of unique data
`data used
`no. of variables
`largest shift/esd
`R
`Rw
`goodness of fit
`a Rietveld refinement.
`
`method with terms of 0.020 and 1.0.30 Atomic scattering factors
`and the values for ¢f¢ and ¢f¢¢ were taken from International
`Tables for X-ray Crystallography.31 Anomalous dispersion
`effects were included in Fc.32 Plots of (cid:229)w(jjF oj - jFcjj)2 versus
`jFoj, reflection order in data collection, sin ı/(cid:236), and various
`classes of indices showed no unusual trends. The NH hydrogen
`atoms were located and their positions and isotropic thermal
`parameters refined; the other hydrogens were located and
`added to the structure factor calculations but were not refined.
`Using the Cerius2 crystal modeling program,33 the water
`hydrogen atoms were placed in locations consistent with
`hydrogen bonding (as determined by short N(cid:226)(cid:226)(cid:226)O distances)
`for visualization purposes.
`Crystal structures of dihydrates B and D and the higher
`hydrate were determined and refined using similar procedures.
`Diffraction data were collected using a Bruker SMART system
`P4 diffractometer using MoKR radiation and CCD detection.34
`Cell refinement and data reduction were accomplished using
`the SAINT software programs.35 The structure was solved by
`direct methods using Siemens SHELXTL-PLUS.36 Non-
`hydrogen atoms were refined anisotropically. All other hydro-
`gen atoms were included in the structure factor calculations
`and placed in idealized positions (dC-H ) 0.95 Å) with assigned
`isotropic thermal parameters (B ) 1.2B of bonded atoms).
`Experimental details of the structure determinations are given
`in Table 1 (see Supporting Information).
`Rietveld Refinement of Dihydrate E. A dihydrate E trial
`crystal structure was constructed using X-ray structure data
`collected for an isostructural EtOH-water mixed solvate.37 The
`structure was interactively Rietveld-refined38 using the DBWS
`program39 until the simulated powder pattern matched the
`experimental X-ray powder pattern of dihydrate E. The results
`of the Rietveld refinement of dihydrate E are also given in
`Table 1. The R factor of 20% is somewhat high; however, there
`does seem to be reasonable agreement between the powder
`patterns. Given the quality of the experimental powder pat-
`tern, the structure was not further refined (see Supporting
`Information).
`Computational Details. Ab initio and/or density func-
`tional methods were employed to search the conformational
`
`space of olanzapine and to calculate the energy difference
`between conformations. The commercial program Spartan
`(Version 5.0) was used.40 The conformer search was restricted
`to geometry optimization (RHF/3-21G*) of conformers pro-
`duced by systematically varying the N5-C4-N1¢ -C6¢ torsion
`angle of the observed conformer in form I in 60° jumps. The
`energy difference between the two minima obtained from the
`conformer search was calculated using Hartree-Fock (3-21G*
`and 6-31G* basis sets) and SVWN (DN, DN*, and DN** basis
`set) density functional models.
`
`Results and Discussion
`
`Crystallization. Form I, the most stable nonsolvated
`crystal form of olanzapine, was directly crystallized from
`dry organic solvents, including EtOAc, THF, acetone,
`and toluene. Forms II and III are desolvates, having
`been prepared only by desolvating MeOH, CH2Cl2, or
`CHCl3 solvates of olanzapine. The desolvation of these
`olanzapine solvates proved to be difficult to control, as
`mixtures of forms I, II, and/or III were routinely
`encountered. The comparatively harsh drying conditions
`required to desolvate the MeOH solvate, for example,
`frequently resulted in form II/III materials contami-
`nated with form I. Forms II and III (free of form I) could
`be obtained by desolvating the CH2Cl2 or CHCl3 solvates
`under mild conditions; however, no conditions were
`identified that would yield pure form II or III.
`The crystalline dihydrates and the higher hydrate
`could be crystallized from pure water or mixtures of
`water and EtOAc or toluene. Dihydrate B, the kinetic
`form produced by slurrying olanzapine (form I) in water,
`could be crystallized in pure form from EtOAc-toluene-
`water at moderately high temperatures (e.g., 55 °C).
`Below 55 °C, the higher hydrate also crystallized from
`this solvent system. The higher hydrate is a metastable
`
`Merck Exhibit 2047, Page 3
`Mylan Pharmaceuticals Inc. v. Merck Sharp & Dohme Corp.
`IPR2020-00040
`
`

`

`900 Crystal Growth & Design, Vol. 3, No. 6, 2003
`
`Reutzel-Edens et al.
`
`~ '·,
`
`·'
`
`'
`
`. . ~,
`
`·.•_:.·-~
`, "."I
`'
`~'\
`i_"'.'
`.
`-· :'"
`
`.
`
`initial
`
`2 minutes
`
`. ~ .
`\ ,··
`,. ,...._ ''
`. ~· .
`
`.
`
`. ~.~ , , l,i
`
`~'·
`
`~
`
`' ,.
`' "
`Diffradion Angle (2-theta)
`Figure 1. Powder X-ray diffraction patterns of olanzapine (a)
`form I, (b) form II (+ form I contaminant), (c) form III (+ form
`II contaminant), (d) dihydrate B, (e) dihydrate D, (f) dihydrate
`E, and (g) the higher hydrate.
`
`"'
`
`a
`
`b
`
`10
`
`15
`20
`Diffraction Angle (2-theta)
`
`25
`
`30
`
`Figure 2. Similar experimental XRD patterns of (a) dihydrate
`E and (b) the EtOH-H2O mixed solvate of olanzapine reveal
`that these crystal forms are isostructural.
`crystal form that contains 2-2.5 mol of water and has
`only been observed in wetcakes of olanzapine. As the
`higher hydrate wetcake was air-dried to a flowable
`powder, the material rapidly lost the first of three
`waters of crystallization and converted to dihydrate E.
`Thus, while mixtures of dihydrate B and the higher
`hydrate crystallized from EtOAc-toluene-water be-
`tween 25 and 55 °C, dihydrates B and E were obtained
`as the solid products. Dihydrate E was isolated in pure
`form by mildly drying the higher hydrate, which was
`exclusively crystallized at or below ambient tempera-
`ture. Dihydrate D, the thermodynamically stable hy-
`drated crystal form of olanzapine, could be isolated by
`slurrying any of the crystal forms in EtOAc-H2O or
`pure water at ambient temperature for several days.
`The anhydrous and hydrated crystal forms of olan-
`zapine produced unique powder X-ray diffraction pat-
`
`5 minutes
`
`30+ minutes
`
`Figure 3. Crystals of the higher hydrate fracture as they
`desolvate to dihydrate E within minutes of isolating them from
`the crystallization solution.
`
`Figure 4. Simulated, experimental, and difference XRD
`patterns from the Rietveld refinement of olanzapine dihydrate
`E.
`
`terns, Figure 1. The powder patterns of the higher
`hydrate and its desolvation product, dihydrate E, were
`strikingly similar, suggesting no gross structural changes
`accompanying the dehydration process. The high purity
`of form I and the hydrated crystal forms could be
`confirmed by comparing the experimental powder pat-
`terns to those calculated from single-crystal X-ray
`diffraction data; however, in practice, solid-state NMR
`spectroscopy proved much more useful for establishing
`the phase purity of the olanzapine samples (vide infra).
`Isostructurality and the Rietveld Refinement of
`Dihydrate E. Powder X-ray diffraction was particularly
`useful for identifying structural relationships between
`various solvated crystal forms of olanzapine. Similarities
`
`Merck Exhibit 2047, Page 4
`Mylan Pharmaceuticals Inc. v. Merck Sharp & Dohme Corp.
`IPR2020-00040
`
`

`

`Olanzapine Anhydrates and Hydrates
`
`Crystal Growth & Design, Vol. 3, No. 6, 2003 901
`
`Table 2. Selected Torsion Angles (deg) for Olanzapine Crystal Formsa
`form I
`dihydrate B
`dihydrate D
`dihydrate E
`N5-C4-N1¢-C2¢
`12.5
`7.9
`9.7
`4.2
`C10a-C3a-C4-N5
`35.2
`37.3
`35.8
`37.7
`C3a-C4-N5-C5a
`5.4
`4.5
`5.8
`4.0
`-43.4
`-44.8
`-44.8
`-44.0
`C4-N5-C5a-C9a
`C5a-C9a-N10-C10a
`55.9
`58.2
`52.6
`58.5
`-56.2
`-59.6
`-52.6
`-59.1
`C9a-N10-C10a-C3a
`N4¢-C5¢-C6¢-N1¢
`-57.4
`-56.6
`-58.6
`-57.4
`N1¢-C2¢-C3¢-N4¢
`59.0
`58.8
`56.3
`57.9
`a Angles are reported for the same conformational enantiomer.
`
`higher hydrate
`6.2
`35.1
`5.1
`-42.7
`58.2
`-58.2
`-56.8
`58.4
`
`Table 3. Relative Conformational Energies (kJ/mol) of
`Olanzapine Conformers
`basis set
`conformer A
`
`conformer B
`
`ab initio
`3-21G*
`6-31G*
`density functional (SVWN)
`DN
`DN*
`DN**
`
`1.42
`0
`
`0
`0
`0
`
`0
`4.53
`
`1.70
`4.62
`5.38
`
`benzodiazepine substituents occupying equatorial posi-
`tions. The diazepine ring is puckered, as evidenced by
`the 127° dihedral angle between the planes of the
`thiophene and benzene rings. The piperazine and
`puckered benzodiazepine rings are nearly coplanar
`(N5-C4-N1¢-C6¢ torsion angle ) 12°). This relatively
`coplanar orientation, which has also been observed in
`dibenzodiazepines, may be attributed to the partial
`double bond between the piperazine and the diazepine
`rings. Selected torsion angles describing the molecular
`conformation of olanzapine in form I, dihydrates B, D,
`and E, and the higher hydrate are given in Table 2.
`Because all of the known crystal structures of olan-
`zapine feature the same pair of conformational enanti-
`omers, a conformational energy minimum has likely
`been realized.42 The Cambridge Crystallographic Data-
`base was searched for benzodiazepines fused to five- and
`six-membered rings to determine whether similar mo-
`lecular conformations are present in structurally similar
`molecules. Five unique, uncharged structures were
`retrieved, all of which featured a fused 6-7-6 tricyclic
`ring system and a 1-piperazinyl side chain.43 No struc-
`tures of benzodiazepines fused to five-membered rings
`were found. Like olanzapine, the five dibenzodiazepine
`analogues adopted puckered conformations, with dihe-
`dral angles between the six-membered rings ranging
`from 118 to 129°. Additionally, in each of the analogues,
`the 1-piperazinyl side chain adopts a chair conformation
`that is in a relatively coplanar orientation with respect
`to the tricyclic ring system. Like olanzapine, which
`adopts both conformational enantiomers in all of its
`crystal structures, all but one of the five structural
`analogues feature two enantiomeric puckered conforma-
`tions.
`A search for conformational minima was also con-
`ducted to assess whether the conformations of olanza-
`pine selected by crystal forces are geometrically close
`to the global minimum. The conformational search
`produced two energetic minima, A and B (Figure 5).44
`Conformer A is characterized by a N5-C4-N1¢-C6¢
`torsion angle of 8.9° and a thiophene/benzene ring
`dihedral angle of 131° and is remarkably similar to that
`observed in the olanzapine crystal structures. Con-
`former B is characterized by a N5-C4-N1¢ -C6¢ torsion
`
`Figure 5. Molecular structure and conformation of olanzapine
`observed in form I. This conformer and its enantiomer are also
`present in dihydrates B, D, and E and in the higher hydrate.
`Conformers A and B are the two energy minima produced in
`a search for conformational minima.
`W A 3D rotatable image in PDB format is available.
`
`observed between the powder patterns of dihydrate B
`and the MeOH and EtOH solvates (not shown), for
`example, revealed that these crystal forms are isostruc-
`tural, or nearly so. A particularly useful observation was
`the isostructurality of dihydrate E and several mixed
`solvates, including the EtOH-water solvate, as shown
`in Figure 2.
`Efforts to obtain single crystals of dihydrate E suit-
`able for an X-ray structure determination were unsuc-
`cessful because this crystal form could not be crystal-
`lized directly from solution, and single crystals of the
`higher hydrate, from which dihydrate E was obtained,
`fractured as the first of three waters of crystallization
`was lost from the crystal lattice, Figure 3. Since the
`X-ray structure of the isostructural EtOH-water solvate
`was available, a crystal model of dihydrate E was
`generated by simply replacing the EtOH ethyl group in
`the mixed solvate with a hydrogen to create the second
`water of crystallization; the dihydrate structure was
`then Rietveld-refined. The energy minimization pro-
`ceeded to yield a structure with a simulated powder
`pattern that closely matched that of the experimental
`pattern of dihydrate E (R ) 20.4), Figure 4.
`Conformational Analysis. Olanzapine adopts
`mirror-related conformations, which rapidly intercon-
`vert in solution by inversion of the diazepine ring.41 This
`molecular motion is sufficiently frozen in each crystal
`form, however, such that pairs of opposite enantiomers
`are observed. Importantly, the same conformers of
`olanzapine are observed in each crystal structure (for
`which X-ray data is available). One of the two mirror-
`related enantiomers present in form I is depicted in
`Figure 5. In the solid state, the piperazine ring exists
`in a chair conformation, with the N4¢-methyl and N1¢-
`
`Merck Exhibit 2047, Page 5
`Mylan Pharmaceuticals Inc. v. Merck Sharp & Dohme Corp.
`IPR2020-00040
`
`

`

`902 Crystal Growth & Design, Vol. 3, No. 6, 2003
`
`Reutzel-Edens et al.
`
`are summarized in Table 3. With the exception of the
`results obtained using the RHF/3-21G* basis set, con-
`former A was found to be the lowest energy conformer.
`Conformer B was found to be 1-5 kJ/mol less stable,
`an energy difference that could be overcome by different
`crystal environments.
`Crystal Packing and Hydrogen Bonding. The two
`enantiomers are packed in each of the crystal structures
`about inversion centers, allowing olanzapine to crystal-
`lize in centrosymmetric space groups. In form I and
`dihydrate B, a P21/c lattice is observed, while olanzapine
`crystallizes in the P1h space group in dihydrate D and
`in a C2/c lattice in dihydrate E and the higher hydrate.
`Interestingly, no specific intermolecular interactions
`(e.g., (cid:240)-stacking or H-bonding) serve to stabilize this
`dimer. Rather, packing appears to be driven by the
`spatial complementarity of the opposite enantiomers.
`Because the centrosymmetric aggregrate (Figure 6) is
`observed in each of the crystal structures, this dimer
`has been proposed to be the crystal building block from
`
`Figure 6. Two opposite enantiomers of olanzapine form a
`crystal building block.
`W A 3D rotatable image in PDB format is available.
`
`angle of 8.7° and a thiophene/benzene ring dihedral
`angle of 130° but features the N1¢-diazepine ring in an
`axial position with respect to the piperazine ring. The
`geometries of conformers A and B were optimized using
`Hartree-Fock (3-21G* and 6-31G* basis sets) and
`SVWN (DN, DN*, and DN** basis sets) density func-
`tional models; the results of the geometry optimizations
`
`'
`
`.
`
`-
`
`'·
`
`.
`
`.
`
`e
`
`ft.... ~ .. . . . . -
`~~ i
`
`'-"",'"".:,
`, ~
`
`~ ~ -
`
`. ~ --J ..
`~ --_!·:
`.
`.
`.
`·;,
`~ ··:f..
`l --
`'i-
`
`.
`.,
`
`'
`
`--l
`
`-
`
`.
`
`'
`
`'
`
`Figure 7. Two-dimensional layers formed by hydrogen-bonding interactions between olanzapine crystal building blocks in (a)
`form I, (b) dihydrate B, (c) dihydrate D, (d) dihydrate E, and (e) the higher hydrate.
`W 3D rotatable images of W (a), W (b), W (c), W (d), and W (e) in PDB format are available.
`
`Merck Exhibit 2047, Page 6
`Mylan Pharmaceuticals Inc. v. Merck Sharp & Dohme Corp.
`IPR2020-00040
`
`

`

`Olanzapine Anhydrates and Hydrates
`
`Crystal Growth & Design, Vol. 3, No. 6, 2003 903
`
`Table 4. Hydrogen-Bond Distances (Å) for Olanzapine
`Crystal Forms
`D-H(cid:226)(cid:226)(cid:226)A
`D(cid:226)(cid:226)(cid:226)A symmetry relation of A
`
`form I
`
`dihydrate B
`
`dihydrate D
`
`dihydrate E
`
`higher hydrate
`
`N10-H(cid:226)(cid:226)(cid:226)N5
`
`N10-H(cid:226)(cid:226)(cid:226)O1
`O1-Ha(cid:226)(cid:226)(cid:226)N4¢
`O2-Ha(cid:226)(cid:226)(cid:226)N5
`O1-Hb(cid:226)(cid:226)(cid:226)O1
`O2-Hb(cid:226)(cid:226)(cid:226)O1
`N10-H(cid:226)(cid:226)(cid:226)O1
`O2-Ha(cid:226)(cid:226)(cid:226)N4¢
`O2-Hb(cid:226)(cid:226)(cid:226)N5
`O1-Ha(cid:226)(cid:226)(cid:226)O2
`O1-Hb(cid:226)(cid:226)(cid:226)O2
`N10-H(cid:226)(cid:226)(cid:226)O1
`O1-Ha(cid:226)(cid:226)(cid:226)N4¢
`O2-Ha(cid:226)(cid:226)(cid:226)N5
`O1-Hb(cid:226)(cid:226)(cid:226)O2
`O1-Hb(cid:226)(cid:226)(cid:226)O1
`N10-H(cid:226)(cid:226)(cid:226)O1
`O1-Ha(cid:226)(cid:226)(cid:226)N4¢
`O2-Ha(cid:226)(cid:226)(cid:226)N5
`O1-Hb(cid:226)(cid:226)(cid:226)O2
`O1-Hb(cid:226)(cid:226)(cid:226)O1
`O2-Hb(cid:226)(cid:226)(cid:226)O3
`O3-H(cid:226)(cid:226)(cid:226)O2
`
`3.09
`
`2.91
`2.83
`2.99
`3.04
`2.94
`
`2.86
`2.82
`2.86
`2.85
`2.83
`
`2.89
`2.81
`3.06
`2.96
`2.96
`
`2.90
`2.79
`2.96
`2.78
`2.82
`2.61
`2.67
`
`x, 1/2 - y, 1/2 + z
`-x, 1/2 + y, 1/2 - z
`x, y, z
`-x, -y, -z
`-x, -y, -z
`x, y, z
`
`-x, -y, -z
`-x, -y, -z
`x, y, z
`x, y, z
`-x, -y, -z
`
`x, -y, 1/2 + z
`1/2 - x, 1/2 - y, -z
`x, y, z
`x, y, z
`1/2 - x, 1/2 - y, -z
`
`x, y, z
`1/2 - x, 1/2 + y, 1/2 - z
`x, -y, 1/2 + z
`x, y, z
`1/2 - x, 1/2 - y, -z
`x, y, z
`-x, y, 1/2 - z
`
`which other solid-state structures of olanzapine may be
`assembled.
`Olanzapine has a single hydrogen-bond donor, N10-
`H, and two good acceptors, the imine N5 and piperidine
`N4¢, which are exposed in the crystal building blocks to
`near neighbor dimers. In form I, NH(cid:226)(cid:226)(cid:226)N hydrogen-
`bonding interactions between the NH and imine N5 link
`the crystal building blocks into two-dimensional layers,
`Figure 7a. These sheets stack directly on top of one
`another in the P21/c lattice. The piperidine N4¢ does not
`participate in hydrogen bonding in this crystal form.
`The hydrogen-bonding patterns in dihydrates, B, D,
`and E and the higher hydrate are dramatically different
`from that in form I. By incorporating water into the
`crystal lattice, the donor-to-acceptor ratio is balanced,
`enabling both the imine N5 and the piperidine N4¢
`acceptors of olanzapine to participate in hydrogen
`bonding.45 Another consequence of incorporating water
`is that the NH(cid:226)(cid:226)(cid:226)N interactions present in form I are
`now disrupted by bridging water molecules (N-H(cid:226)(cid:226)(cid:226)Ow-
`Hw(cid:226)(cid:226)(cid:226)Ow-Hw(cid:226)(cid:226)(cid:226)N).46 In all of the hydrates, the water
`molecules are held by two or three hydrogen-bonding
`interactions, Table 4.
`Hydrogen bonding also appears to direct the assembly
`of the olanzapine crystal building blocks into two-
`dimensional layers in the hydrates. As in form I, the
`crystal building blocks are aligned end-on-end and stack
`directly on top of one another in dihydrate D (Figure
`7c). The waters of crystallization occupy sites between
`the olanzapine dimers in this P1h lattice. The hydrogen-
`bonded layers of olanzapine crystal building blocks in
`dihydrates B and E and the higher hydrate are signifi-
`cantly different from those observed in form I and
`dihydrate D, yet virtually identical to one another. In
`dihydrates B and E and the higher hydrate, the olan-
`zapine crystal building blocks are aligned in a herring-
`bone motif (Figure 7b,d,e). The main difference between
`B and E (and the higher hydrate) is the relative
`
`Figure 8. Parallel and perpendicular relative orientations of
`adjacent two-dimensional layers of olanzapine crystal building
`blocks in (a) dihydrate B and (b) dihydrate E. The crystal
`packing of olanzapine in the higher hydrate is the same as in
`dihydrate E.
`W 3D rotatable images of W (a) and W (b) in PDB format are
`available.
`
`orientation of adjacent two-dimensional layers. Whereas
`adjacent layers of crystal building blocks stack directly
`on top of one another in the P21/c lattice of dihydrate
`B, adjacent hydrogen-bonded layers are oriented or-
`thogonally to one another in dihydrate E and in the
`higher hydrate C2/c lattices, Figure 8. The remarkable
`structural similarity between dihydrates B and E (via
`the higher hydrate) may account for the comparative
`ease with which mixtures of these crystal forms are
`obtained.
`Solid-State 13C NMR Spectroscopy. 13C CP/MAS
`NMR spectra collected for the olanzapine anhydrates
`and hydrates featured sharp, highly resolved resonances
`commensurate with the well-defined solid-state envi-
`ronments of the 13C nuclei in these crystal forms, Figure
`9. Because the isotropic chemical shifts of olanzapine
`reflect not only the different types of carbons but also
`their solid-state environments, polymor