Received 12 July 2012
Reinterpretation of the monohydrate of clarithromycin from X-ray powder diffraction data as a trihydrate
Noguchi, Fujiki, Iwao, Miura & Itai [Acta Cryst. (2012), E68, o667-o668] recently reported the crystal structure of clarithromycin monohydrate from synchrotron X-ray powder diffraction data. Voids in the crystal structure suggested the possible presence of two more water molecules. After successful location of the two additional water molecules, the Rietveld refinement still showed minor problems. These were resolved by noticing that one of the chiral centres in the molecule had been inverted. The corrected crystal structure of clarithromycin trihydrate, refined against the original data, is now reported. Dispersion-corrected density functional theory calculations were used to check the final crystal structure and to position the H atoms.
In a Cambridge Structural Database (CSD, Allen, 2002) study of methanolates, Brychczynska et al. (2008) described the remarkable isomorphism of clarithromycin methanolate (Iwasaki et al., 1993) and clarithromycin anhydrate form II (Stephenson et al., 1997) (Fig. 1). As a consequence of this isomorphism, the crystal structure of clarithromycin anhydrate features a suspiciously large void, which in the methanolate is occupied by a molecule of methanol. The anhydrate was grown from a water-saturated ethyl acetate solution, excluding the possibility of a methanolate, but a hydrate cannot be ruled out. Indeed, Tian et al. (2009) later reported the same anhydrate form II and found from single-crystal studies that the voids can be occupied by up to half a molecule of water, but confirmed that the voids can also be essentially empty. It seems puzzling that a void large enough to contain a methanol molecule would be too small to contain a full water molecule. It was this intriguing behaviour of the methanolate/anhydrate/hemihydrate system that prompted a closer look when recently the crystal structures of a second polymorph of the anhydrate (form I) and of the alleged monohydrate of clarithromycin were reported based on X-ray powder diffraction data (Noguchi, Fujiki et al., 2012; Noguchi, Miura et al., 2012).
In the monohydrate, the authors located one water molecule in the difference Fourier map. Using Mercury (Macrae et al., 2008), several water-accessible voids are readily identified in the crystal structures of both the new anhydrate form I and the monohydrate (Fig. 2). However, the presence of similar voids in the crystal structure of the previously published anhydrate form II indicates that the presence of voids alone is not sufficient proof of missing water molecules. Indeed, the voids in anhydrate form I were convincingly shown to be empty by the authors, and a Rietveld refinement using the authors' powder diffraction data with and without four O atoms manually inserted at the positions of the voids showed no significant improvement in the fit, with the occupancies of the O atoms refining to less than 0.3.
The monohydrate is different. The eight water-accessible voids were not commented on in the original paper. The experimental powder diffraction pattern had been made available in the supplementary material of the original publication and was downloaded from http://www.iucr.org (file hb6588Isup2.rtv). The 2 values and the intensities were used as provided; the estimated standard deviations were calculated as the square root of the number of counts. The space group being P212121, two crystallographically independent water molecules are needed to occupy the eight voids. Two O atoms were placed at the centres of the voids and the resulting model was Rietveld-refined using TOPAS-Academic (Coelho, 2007). The occupancies and isotropic displacement parameters of the two additional O atoms were refined individually. For comparison, the crystal structure as published was refined using the same program with the same settings. The results are compared in Table 1. After Rietveld refinement, the two additional water molecules refine to positions that allow an excellent network of hydrogen bonds, with OO distances between 2.66 (2) and 2.98 (2) Å (Fig. 3). From the chemically sensible hydrogen bonds and from the numerical results in Table 1, it is concluded that the crystal structure published by Noguchi and co-workers pertains to a trihydrate rather than a monohydrate.
One might wonder why the original authors were not able to locate the two water molecules in the residual electron-density map, in spite of the significant effect these two water molecules have on the calculated intensities. It is possible that this is caused by the way residual electron densities are calculated for powder diffraction data. For single-crystal data, the residual electron density is calculated based on the observed intensities, whereas the phases are taken as calculated from the model. For powder data, individual intensities cannot be observed directly due to peak overlap, and the observed sums of intensities must be partitioned based on the intensities calculated from the model. In other words, in residual electron-density maps from powder diffraction data, both phases and intensities are biased by, and towards, a possibly incomplete structural model. As a consequence, residual electron-density maps from powder diffraction data are less objective and therefore less reliable than those from single-crystal data.
In the present line of study, the goodness-of-fit had improved to an acceptable value and the Rietveld plot looked excellent, with no remaining features in the difference curve. Nevertheless, in spite of these substantial improvements the crystal structure remained unsatisfactory. The molecular geometry was checked using Mogul (Bruno et al., 2004) to compare all bond lengths and valence angles against distributions from single-crystal data. The relevant measure is the absolute value of the z score, which measures by how many standard deviations each value in the crystal structure differs from the mean of the distribution from the single-crystal data. Normally, all values should be smaller than about 3. For clarithromycin trihydrate, some values were still over 5, sometimes over 8. One of the major problems is that Rietveld refinement is a least-squares procedure, and any error in the model is absorbed by minor distortions throughout the rest of the model, making it difficult to locate the source of the error. Usually, dispersion-corrected density functional theory (DFT-D; Perdew et al., 1996; Grimme, 2006; Accelrys, 2011) provides a powerful method for complementing powder diffraction data, but for this crystal structure the hydrogen-bonding pattern was initially uncertain, and energy minimizations with three different hydrogen-bonding patterns led to distorted structures. Visual inspection and comparison with the energy-minimized crystal structures indicated a problem with atom C16. A disorder model was tried, resulting in a significant overall improvement, but at the expense of reversing the stereochemistry of one of the chiral centres. Comparison with the single-crystal structures of clarithromycin anhydrate form II showed that the stereochemistry had indeed been assigned incorrectly in the powder structure. According to the original paper, `The initial structure was determined by the molecular replacement method using MOLREP implemented in CCP4. The search model employed was form 0 of the [clarithromycin] crystal structure (Jin et al., 2009).' The chiral centres in the crystal structure by Jin et al. (2009) are all correct, and apparently one of them was inverted as part of the original structure solution or structure refinement process of Noguchi and co-workers.
With the corrected stereochemistry, the crystal structure refined quickly and smoothly to a final structure with excellent figures of merit (Table 1). Fig. 4 shows the final Rietveld refinements; Fig. 5 shows the final molecular structure.
The hydrogen-bonding pattern was assigned by running a short molecular dynamics simulation using the COMPASS force field (Sun, 1998), with regular local optimizations of the current structure (quenching). The hydrogen-bonding network with the lowest energy was selected. For the Rietveld refinement, C-H distances of 0.96 Å and O-H distances of 0.85 Å were used. For the final structure, the positions of the H atoms were energy-minimized using DFT-D, with the non-H atoms and unit-cell parameters kept fixed. The coordinates of the H atoms in the accompanying CIF therefore reflect nuclear positions rather than maxima in the electron density.
The correctness of the crystal structure as a whole was assessed by energy-minimizing the structure using DFT-D with the unit cell free using CASTEP (Clark et al., 2005), and calculating the root mean-square (r.m.s.) Cartesian displacement (ignoring H atoms) as described by van de Streek & Neumann (2010). R.m.s. Cartesian displacement values lower than 0.25 Å indicate a correct crystal structure; values over 0.30 Å are indicative of a problem with the crystal structure. For clarithromycin trihydrate, the r.m.s. Cartesian displacement is 0.14 Å, well within the range indicating a correct crystal structure.
In conclusion, the information content of powder diffraction patterns is high, probably much higher than most crystallographers are prepared to accept, but the least-squares characteristics of the Rietveld refinement procedure can obscure the cause of possible problems within the model by distributing the problem over large parts of the structure. It is only when the entire model describes the experimental data without discrepancies, and when the whole range of figures of merit (Table 1) and the final crystal structure (Figs. 3 and 5) and Rietveld refinement (Fig. 4) are satisfactory, that we should accept a crystal structure from powder diffraction data as `correct'.
| || Figure 1 |
Overlay of the crystal structures of the anhydrate form II (blue in the electronic version of the paper) and the methanolate (red) of clarithromycin. The structures are isomorphous. The methanol molecules in the methanolate structure (shown in space-filling style) correspond to voids in the anhydrate. H atoms have been omitted for clarity.
| || Figure 2 |
The water-accessible voids (shaded volumes) in the published crystal structures of (a) clarithromycin monohydrate and (b) clarithromycin anhydrate form I. H atoms have been omitted for clarity.
| || Figure 3 |
The new hydrogen-bonding pattern proposed here, made possible by the addition of two water molecules (shown as large spheres) occupying what are voids in Fig. 2. The introduction of the two water molecules generates six new hydrogen bonds. H atoms have been omitted for clarity. [Symmetry codes: (i) x, y, z - 1; (ii) x + , -y + , -z + 1; (iii) x + , -y + , -z.]
| || Figure 4 |
Rietveld refinements with (a) the model as published (monohydrate) and (b) with two water molecules added to give a trihydrate. Both models were fully refined using the same software with the same settings. On the y axis, the square root of the number of counts is plotted to emphasize small features in the difference curve. Note the substantial improvement caused by the addition of just two O atoms, representing 16 electrons, to a compound with the formula C38H69NO13·H2O, representing 418 electrons.
| || Figure 5 |
The molecular structure of (I), showing the atom-numbering scheme. Displacement spheres are drawn at the 30% probability level. For clarity, the water molecules have not been shown.
Data collection: local software (Osaka et al., 2010; Takata et al., 2002); cell refinement: TOPAS-Academic (Coelho, 2007); data reduction: local software (Takata et al., 2002); program(s) used to solve structure: CCP4 (Collaborative Computational Project, Number 4, 1994); program(s) used to refine structure: TOPAS-Academic; molecular graphics: Mercury (Macrae et al., 2008); software used to prepare material for publication: Crimson Editor (http://www.crimsoneditor.com ).
Supplementary data for this paper are available from the IUCr electronic archives (Reference: FA3285 ). Services for accessing these data are described at the back of the journal.
The Lundbeck Foundation (Denmark) is gratefully acknowledged for financial support (grant No. R49-A5604).
Accelrys (2011). Materials Studio. Accelrys Inc., San Diego, California, USA.
Allen, F. H. (2002). Acta Cryst. B58, 380-388.
Bruno, I. J., Cole, J. C., Kessler, M., Luo, J., Motherwell, W. D., Purkis, L. H., Smith, B. R., Taylor, R., Cooper, R. I., Harris, S. E. & Orpen, A. G. (2004). J. Chem. Inf. Comput. Sci. 44, 2133-2144.
Brychczynska, M., Davey, R. J. & Pidcock, E. (2008). New J. Chem. 32, 1754-1760.
Clark, S. J., Segall, M. D., Pickard, C. J., Hasnip, P. J., Probert, M. J., Refson, K. & Payne, M. C. (2005). Z. Kristallogr. 220, 567-570.
Coelho, A. A. (2007). TOPAS-Academic. Coelho Software, Brisbane, Australia.
Collaborative Computational Project, Number 4 (1994). Acta Cryst. D50, 760-763.
Grimme, S. (2006). J. Comput. Chem. 27, 1787-1799.
Iwasaki, H., Sugawara, Y., Adachi, T., Morimoto, S. & Watanabe, Y. (1993). Acta Cryst. C49, 1227-1230.
Jin, Z. M., Ma, L. L., Wei, W. X., Lin, C. S. & Li, W. Z. (2009). J. Struct. Chem. 50, 185-189.
Macrae, C. F., Bruno, I. J., Chisholm, J. A., Edgington, P. R., McCabe, P., Pidcock, E., Rodriguez-Monge, L., Taylor, R., van de Streek, J. & Wood, P. A. (2008). J. Appl. Cryst. 41, 466-470.
Noguchi, S., Fujiki, S., Iwao, Y., Miura, K. & Itai, S. (2012). Acta Cryst. E68, o667-o668.
Noguchi, S., Miura, K., Fujiki, S., Iwao, Y. & Itai, S. (2012). Acta Cryst. C68, o41-o44.
Osaka, K., Matsumoto, T., Miura, K., Sato, M., Hirosawa, I. & Watanabe, Y. (2010). AIP Conf. Proc. 1234, 9-12.
Perdew, J. P., Burke, K. & Ernzerhof, M. (1996). Phys. Rev. Lett. 77, 3865-3868.
Stephenson, G. A., Stowell, J. G., Toma, P. H., Pfeiffer, R. R. & Byrn, S. R. (1997). J. Pharm. Sci. 86, 1239-1244.
Streek, J. van de & Neumann, M. A. (2010). Acta Cryst. B66, 544-558.
Sun, H. (1998). J. Phys. Chem. B, 102, 7338-7364.
Takata, M., Nishibori, E., Kato, K., Kubota, Y., Kuroiwa, Y. & Sakata, M. (2002). Adv. X-ray Anal. 45, 377-384.
Tian, J., Thallapally, P. K., Dalgarno, S. J. & Atwood, J. L. (2009). J. Am. Chem. Soc. 131, 13216-13217.