J AlC S
`
`COMMUNICATIONS
`
`Published on Web 08/24/2002
`
`Iterative High-Throughput Polymorphism Studies on Acetaminophen and an
`Experimentally Derived Structure for Form Ill
`
`Matthew L. Peterson,‘ Sherry L. Morissette,‘ Chris McNu|ty,‘ Andrew Go|dsweig,‘ Paul Shaw,‘
`Martin LeQuesne,‘ Julie Monaglefl Nicolas Encina,‘ Joseph Marchionna,‘ Alasdair Johnson,‘
`Javier Gonzalez-Zugasti,‘ Anthony V. Lemmo,‘ Stephen J. Ellis,‘ Michael J. Cima,2 and
`Om A|marsson*~‘
`
`TransF0rm Pharmaceuticals, Inc., 610 Lincoln Street, Waltham, Massachusetts 02451
`
`Received May 28, 2002
`
`The discovery of crystal polymorphs is often a costly and time-
`consuming process.3 There are no failsafe methods to predict the
`extent of polymorphism of a given compound.4 We have developed
`a system, CrystalMax, for high-throughput, parallel polymorph
`discovery capable of eliciting both stable and metastable forms. It
`has three main components: experimental design and execution
`software, robotic dispensing and handling hardware, and high-speed
`microanalytics. This system was used to assess the extent and nature
`of polymorphism in acetaminophen (paracetamol; p-hydroxy-
`acetanilide), a widely used NSAID.5 Three polymorphs are
`experimentally known,6 one of which (form III) has only been found
`by thermal microscopy.7 We report a series of iterative high-
`throughput experiments for generation and identification of ac-
`etaminophen polymorphs. In addition, a structural suggestion for
`form III is advanced on the basis of powder X-ray microdiffraction
`(PXRD) and prior computer predictions.
`Scheme 1 provides an overview of the CrystalMax system, which
`facilitates parallel screening of thousands of crystallization condi-
`tions. Experimental design software defines combinatorial test
`conditions on the basis of the selection of solvent properties and
`methods used to drive supersaturation.8 Experiments are executed
`in arrays of individually addressable sample containers.9 Samples
`that crystallize are removed from the original array, the solvent is
`removed by aspiration, and the residue is dried. Optical imaging
`and in situ Raman spectroscopy are used to characterize newly
`formed crystals.1° Polymorph assignments are confirmed by PXRD,
`DSC, TGA, and optical microscopy. The experiment is continuously
`tracked by a database system. Cheminformatics software aids
`automated data analysis and experiment design.
`A classification process aids the analysis of the large number of
`Raman spectra generated during an experiment. Similarity coef-
`ficients“ are calculated for all pairs of spectra. The coefficients
`are sorted, color-mapped, and displayed as an n-by-n matrix for
`easy visualization. Figure 1 shows a representative Tanimoto matrix
`generated for 120 Raman spectra of acetaminophen in which all
`three polymorphs are present.” The plot in Figure 1 illustrates that
`the Tanimoto matrix derived from spectral data is a simple visual
`way of differentiating polymorphs of acetaminophen.
`The process in Scheme 1 and the Tanimoto matrix analysis
`exemplified in Figure l were used iteratively to identify all three
`polymorphs of acetaminophen. At the end of the first iteration,
`precipitates had been observed in 9.3% (723) of 7776 crystallization
`trials, 29 of which were II and the remainder form 1.13 No form III
`or other metastable forms were identified.“ The solids comprised
`single crystals or powders.
`*To whom correspondence should be addressed. E—mail:
`transformpharmacom.
`
`almarsson@
`
`Figure 1. A representative Tanimoto matrix comparing 120 Raman spectra
`of acetaminophen polymorphs. Each pixel represents the Tanimoto value”
`for pairs of spectra, according to the color scheme on the right. The map
`displays Tanimoto values for 50, 45, and 25 spectra of forms I, H, and HI,
`respectively. A dark red pixel indicates the highest degree of similarity
`(Tanimoto R 0.0), Whereas dark blue indicates dissimilarity (Tanimoto R
`10).
`
`Scheme 1
`
`
`
`In the second iteration the conditions that gave form II in the
`initial screen were replicated in higher numbers and showed that
`only one solvent mixture, 67/33 (v/v) MeOH/toluene, consistently
`yielded form II.” The experiment highlighted the sometimes
`intractable nature of crystallization in that not all replicates produced
`crystals and that different polymorphs were obtained from seem-
`ingly identical conditions. The third iteration was designed to test
`the reliability of form II crystallization from 67/33 (v/v) MeOH/
`toluene by isothermal evaporation at 40, 54, and 65 °C.15v16 These
`experiments gave primarily forms I and II.” A few samples were
`characterized as glassy solids at the highest temperature. The largest
`proportion of form II was obtained at 54 °C, for which the Tanimoto
`matrix analysis showed that 43% of the samples were form II. Form
`III was not observed in the evaporative crystallizations.4'“‘
`Melt crystallization in crystallization containers generated form
`III in several instances.” We observed that samples of form III
`obtained in this format converted to form II within hours. The results
`
`from melt crystallization show that the confinement of the acetami-
`nophen melt between microscope slides is not a strict requirement
`for the formation of form III.4'“‘
`
`10958 I J. AM. CHEM. SOC. 2002, 124, 10958—10959
`
`10.1021Ija020751w CCC: $22.00 © 2002 American Chemical Society
`
`Janssen Ex. 2025
`
`Lupin Ltd. v. Janssen Sciences Ireland UC
`|PR2015-01030
`
`(Page 1 of 2)
`
` N
`
`100
`
`13)
`
`Ill
`Illl
`fill
`Bunndsplnmnnfllnrhor
`
`

`
`150-
`
`140 -
`
`10
`
`15
`
` 5
`
`20
`
`25
`
`30
`
`35
`
`2—Theta(’)
`
`Figure 2. PXRD patterns for forms I (top), II (middle), and III (bottom)
`of acetaminophen.
`
`We were able to prepare form III in capillaries for PXRD
`analysis.” The diffraction patterns of forms I—III are shown in
`Figure 2. It is apparent that many of the diffraction peaks of form
`III are similar to those of form II. This, however, is deceiving and
`results from accidental overlap of the diffraction peaks of structur-
`ally different polymorphs.
`We compared the PXRD pattern of form III to those calculated
`from the known structures of forms I and II and the theoretical
`
`structures recently published.“ The calculations accounted for
`crystallite size-dependent
`line broadening and global
`isotropic
`temperature factors.2° Of the theoretical structures, a monoclinic
`structure (AK6, P21/C)2l gave a calculated powder pattern that
`closely matched the observed powder pattern for form III. The
`intensity and width of the low angle peak near 5° 26) was found to
`be very sensitive to the line broadening in the a-direction. The
`apparent absence of this peak in the experimental powder pattern
`suggests that there is only short-range order along the a-axis, which
`is also the slow growth direction in the calculated morphology.“
`The proposed structure is built up of bilayers. Each bilayer is
`held together by O7H-- -O(H) hydrogen bonds in the be plane. This
`hydrogen bonding pattern is strikingly different from that observed
`in forms I and II where O7H---O(C) and N7H'"O(H) hydrogen
`bonds are observed.6 Interbilayer interactions are between acetamide
`methyl groups along the a-axis. The weak nature of methyl—methyl
`contacts is a likely reason for the difficulty in obtaining macroscopic
`crystals of form III from solution. These weak interactions also
`facilitate microtwinning along the a-axis accounting for
`the
`crystallite size-dependent line broadening suggested by the experi-
`mental and calculated PXRD patterns.
`The CrystalMax platform was used to rapidly prepare and identify
`three forms of acetaminophen. The benefit of carrying out a
`multitude of experiments under different as well as identical
`conditions to capture all metastable forms is evident. A structural
`model for the elusive third form has been proposed. Future reports
`on high-throughput crystallization and solid form discovery will
`focus on understanding how molecular interactions drive crystal-
`lization processes and their outcomes.”
`
`Acknowledgment. The authors thank Profs. Joel Bernstein,
`Stephen Byrn, Roger Davey, and Leslie Leiserowitz for critical
`discussions.
`
`Supporting Information Available: Solvents, thermal parameters,
`and concentrations used for the high-throughput crystallization experi-
`
`COMMUNICATIONS
`
`ments and representative Raman spectra and PXRD patterns (PDF).
`This material is available free of charge via the Internet at http://
`pubs.acs.org.
`
`References
`
`(I) Transform Pharmaceuticals, Inc.
`(2) Department of Materials Science and Engineering, Massachusetts Institute
`of Technology, Room I2—0I I, 77 Massachusetts Avenue, Cambridge, MA
`02 I 39.
`(3) (a) Bernstein, J. Polymorphism in Molecular Crystals; International Union
`of Crystallography Monographs on Crystallography and International
`Union of Crystallography joint publishers: Claredon Press: Oxford, 2002.
`(b) Byrn, S. R.; Pfeiffer, R. R.; Stowell, J. G. Solid State Chemistry of
`Drugs, 2nd ed; SSCI Inc.: West Lafayette, IN, I999. (c) McCrone, W.
`C. In Physics and Chemistry ofthe Organic Solid State; Fox, D., Labes,
`M. M., Weissberger, A., Eds.; Interscience: New York, I965; Vol. 2, pp
`725 7767.
`(4) (a) Beyer, T.; Day, G. M.; Price, S. L. J. Am. Chem. Soc. 2001, 123,
`508675094. (b) Verwer, P.; Leusen, F. J. J. In Reviews in Computational
`Chemistry; Lipowitz, K. B., Boyd, D. B., Eds.; Wiley/VCH: New York,
`I998; Vol. I2, Chapter 7, p 327. (c) Karfunkel, H. R.; Wu, Z. J.; Burkhard
`A.; Rihs, G.; Sinnreich, D.; Burger, H. M.; Stanek, J. Acta Crystallogr.,
`Sect. B 1996, 52, 555. (d) Filippini, G.; Gavezzotti, A. J. Am. Chem. Soc.
`1995, 117, I2299.
`(5) Nonsteroidal antiinflammatory drug.
`(6) Nichols, G.; Frampton, C. S. J. Pharm. Sci. 1998, 87(6), 6847693.
`(7) (a) Burger, A. Acta Pharm. Technol. 1982, 28(1), I720. (b) Di Martino,
`P.; Conflant, P.; Drache, M.; Huvenne, J.—P.; Guyot—Hermann, A.—M. J.
`Therm. Anal. 1997, 48, 4477458. (c) Szwlagiewicz, M.; Marcolli, C.;
`Cianferani, S.; Hard, A. P.; Vit, A.; Burkhard, A.; Von Raumer, M.;
`Hofmeier, U. Ch.; Zilian, A.; Francotte, E.; Shenker, R. J. Therm. Anal.
`1999, 57, 23743.
`(8) Proprietary Java program, see Supporting Information.
`(9) The compound and solvent(s) are placed in sealable glass tubes. In the
`case of solvent crystallization, supersaturation is generated by thermal
`cycling the tubes in aluminum blocks. Crystallization events are identified
`by an optical scanning station using automated image analysis.
`(I0) Raman spectroscopy is a suitable primary screening method for poly-
`morphs because of its sensitivity to changes in crystal form. In addition,
`spectra can be obtained from microgram amounts of material in a matter
`of seconds using standard equipment. See: Bugay, D. E. Adv. Drug
`Delivery Rev. 2001, 48, 43765.
`(I I) Tanimoto coefficient. Tm = I 7 [Nab/(Na + Nb 7 Nab)], where Nab is
`the number of peaks coincident within a user defined range in both spectra
`a and b, Na is the number of peaks in spectrum a, and Nb is the number
`of peaks in spectrum
`.
`.
`.
`(I2) The spectra were obtained from a combination of expermiental crystal-
`lization and authentic materials (the latter verified by melting point, optical
`microscopy, and/or powder X—ray diffraction).
`(I3) We conducted thermally driven solution crystallization in 7776 samples
`representing 2592 unique conditions. Each condition was run in triplicate
`to test reproducibility of crystallization and to increase the chances of
`identifying metastable forms. A diverse set of 24 solvents was used, either
`as single, binary, or ternary solvent mixtures. Crystallization solvents
`featured a broad range of chemical functionality (complete listing in
`Supporting Information). Three nominal drug concentrations were em-
`ployed with each combinatorial solvent condition in an effort to cover a
`wide range of supersaturation levels.
`(I4) Each solvent mixture that gave at least one form II crystal was replicated
`eight times in the follow—up experiment.
`(I5) Solutions of acetaminophen were prepared in arrays of 96 sealed
`crystallization containers. The arrays were heated to the desired evaporation
`temperature, and the seals were removed. The samples were allowed to
`evaporate and were visually monitored for crystallization, solids were
`rearrayed, and the crystals were analyzed by Raman spectroscopy.
`Experiments took 24 h.
`(I6) The temperatures used are just above the glass transition temperature,
`the temperature typically used in thermal microscopy experiments to
`prepare form III and just below the temperature where form III converts
`to form II, respectively.
`(I7) A few of the samples prepared at 65 °C were identified as glassy
`acetaminophen by visual inspection and Raman spectroscopy.
`(I8) The experiment was carried out at 54 °C with various nucleating surfaces
`added into the tubes. Crystallization was observed over 24 h. The Tc matrix
`indicated that out of 96 acetaminophen samples, two were a noncrystalline
`glass (verified by optical analysis), five were form III, and the remainder
`were form II.
`(I9) Microscopy and Raman are consistent with this assignment.
`(20) Detailed information is given in the Supporting Information.
`(2I) Eel] constants, nonstandard setting: a = I6.05 A, b = 5.07 A, c = 9.65
`, /7 = 79.I°.
`(22) Blagden, N.; Cross, W. I.; Davey, R. J.; Broderick, M.; Prichard, R. G.;
`Roberts, R. J. Rowe, R. C. Phys. Chem. Chem. Phys. 2001, 3, 38I973825.
`JA02075IW
`
`J. AM. CHEM. SOC. I VOL. 124, NO. 37, 2002 10959
`
`Janssen Ex. 2025
`
`Lupin Ltd. v. Janssen Sciences Ireland UC
`|PR2015-01030
`
`(Page 2 of 2)