Detailed Analysis of Packing Efficiency Allows Rationalization of

7 days ago - Detailed Analysis of Packing Efficiency Allows Rationalization of Solvate Formation Propensity for Selected Structurally Similar Organic ...
0 downloads 13 Views 1MB Size
Article Cite This: Cryst. Growth Des. XXXX, XXX, XXX−XXX

pubs.acs.org/crystal

Detailed Analysis of Packing Efficiency Allows Rationalization of Solvate Formation Propensity for Selected Structurally Similar Organic Molecules Agris Berziņ ̅ s,̌ * Dace Zvaniņa, and Aija Trimdale Faculty of Chemistry, University of Latvia, Jelgavas iela 1, Riga, LV-1004, Latvia S Supporting Information *

ABSTRACT: In structural study of seven bile acids it was identified that their propensity for solvate formation is directly related to the packing efficiency of the unsolvated phases: low packing index, voids, and unsatisfied hydrogen bonding lead to extensive solvate formation, whereas efficient packing leads to the opposite. This was determined to be caused by the presence of OH group attached to carbon C12. Solvate formation was determined to provide a noticeable improvement in the packing efficiency for compounds having ansolvates with inefficient packing.



INTRODUCTION Studies of the crystal form landscape are particularly important for pharmaceutical compounds,1,2 making these well-explored model compounds. While some compounds crystallize only in numerous one-component phases (e.g., ROY3 and flufenamic acid4), others are keen to form multicomponent phases by crystallizing together with different solvent molecules, by therefore forming dozens5−7 or even more than 100 solvates.8 Two main structural driving forces resulting in the incorporation of solvent in the structure are the ability of solvent to compensate unsatisfied potential intermolecular interactions between the host molecules and the ability to decrease the void space and/or lead to more efficient packing,9−11 with most solvates including contributions from both of these driving forces.9 The search for possibilities to predict propensity for solvate, or particularly hydrate,12−14 formation is still ongoing.15,16 Although several algorithms for prediction of solvate formation based on fluid-phase thermodynamics computations17,18 as well as knowledge based models15 have been proposed, they are still unable to take into account all of the factors contributing to solvate formation, and the border between compounds forming solvates and those forming only one-component phases still cannot be confidently drawn. Although it has been shown16,19,20 that the prediction of guest-free higher-energy structures in a computational structure prediction study indicates the tendency to form solvates, such an approach has not been additionally validated, and moreover, such a study is not feasible for all compounds. In this study, the propensity for solvate formation of several bile acids (Cholic acid CA, methyl cholate MC, chenodeoxycholic acid CDCA, ursodeoxycholic acid UDCA, deoxycholic acid DCA, hyodeoxycholic acid HDCA, and lithocholic acid © XXXX American Chemical Society

LCA; see Figure 1) was rationalized using detailed analysis of available crystal structures of ansolvates and solvates. The

Figure 1. Molecular structure of CA, MC, CDCA, UDCA, HDCA, DCA, and LCA with the numbering of non-hydrogen atoms and labeling of flexible dihedral angles.

propensity for solvate formation for the first four of these compounds was additionally confirmed experimentally. CA, DCA, and MC are keen solvate and other multicomponent phase formers with 107 and 41 solvates (including all unique multicomponent phases with compounds existing as liquid in ambient conditions) reported for CA and DCA, respectively, in the Cambridge Structural Database (CSD) and 27 solvates reported for MC.21 In contrast, CDCA, UDCA, HDCA, and LCA are not good solvate formers, with few solvates reported for HDCA,22 two for CDCA, and none for UDCA and LCA. Received: October 17, 2017 Revised: February 5, 2018 Published: February 26, 2018 A

DOI: 10.1021/acs.cgd.7b01457 Cryst. Growth Des. XXXX, XXX, XXX−XXX

Crystal Growth & Design

Article

CDCA is formed without involving O5 of molecule B as an acceptor and O1 of molecule A as a donor, while in HDCA instead of employing O4 as an acceptor O1 is an acceptor for two hydrogen bonds. In crystal structure of unsolvated MC, there are even more unsatisfied hydrogen bond donors and acceptors. However, it can clearly be seen that the satisfaction of hydrogen bonding capability in the unsolvated phase is not the driving force of solvate formation in the analyzed bile acids, as the propensity of CA to form solvates, as well as the poor propensity of CDCA and HDCA to form solvates, is not explained. Calculation of lattice energy, packing index, and void space was used to characterize molecular packing in unsolvated bile acids (see Table 2, where lattice energy calculated in PIXEL23 and void space calculated in PLATON24 are given, with results obtained using alternative methods given in the Supporting Information). The difference in the molecular size allows using the lattice energy for direct comparison of packing efficiency only for CDCA, UDCA, and HDCA, clearly representing the more efficient packing in UDCA and HDCA. This is expected for UDCA because of the efficient hydrogen bonding along with the highest packing index among the analyzed bile acids. Apparently, the packing in HDCA is also efficient, as three unique conventional hydrogen bonds are present, and employing O1 as an acceptor for two hydrogen bonds and O4 for none in this case can be sterically more efficient (as lower repulsion energy compensates the less negative Coulombic and dispersion energy, see Table S9, Supporting Information), resulting in very similar lattice energy despite the lower packing index and some voids in the structure. Meanwhile, hydrogen bonding in CDCA cannot lead to as low lattice energy values as for the rest of the two isomers. Interestingly, despite having a void space of 2.8%, the structure of CDCA has the second highest packing index among all analyzed bile acids. It should, however, be noted that for calculation of void space standard approach implemented in PLATON24 was chosen (using approximately 0.2 Å grid spacing with probe radius of 1.2 Å, with alternative results obtained in Mercury using a similar approach and calculation settings) for detection of larger cavities and channels, instead of detailed calculation of all empty space using, e.g., Crystal Explorer25 or very small probe radius in Mercury. Thus, in the presented results void space identify larger cavities and channels in the structure, while packing index characterizes total efficiency and compactness of the packing. As expected, introduction of an additional hydroxyl group in CA leading to possibility to form four unique hydrogen bonds results in lower lattice energy, although in comparison with UDCA and HDCA, reduction is only by 15 kJ mol−1. Moreover, the packing index of CA is only 0.651 and structure contains voids with volume of 9.5%. As expected, replacement of the carboxylic OH group with methoxy group in MC structure and thus limitation of the maximum number of unique hydrogen bonds to three per molecule increase the lattice energy with respect to the CA. However, despite the apparent additional freedom in the formation of hydrogen bonds if compared to, e.g., UDCA and HDCA, slightly higher lattice energy is achieved, which can be associated with the presence of a hydrogen bonding network not involving even all the hydrogen bond donors (see Supporting Information) and inefficient packing characterized by packing index 0.667 and void space of 3.6%.

As molecular structures of the analyzed bile acids are highly similar, the available solvate prediction algorithms15,17,18 are expected to fail in providing such drastic differences in the propensity for solvate formation.



RESULTS AND DISCUSSION Simple solvate screening using 11 common solvents confirmed that CA and MC are keen solvate formers with solvates obtained from 9 solvents for CA and MC each (see Table 1). Table 1. Outcome of the Performed Crystallization Experiments of the Selected Bile Acids solvent

CA

MC

CDCA

UDCA

H2O Methanol Ethanol Acetonitrile Nitromethane Ethyl acetate 3-pentanone DMF Dichloromethane Toluene Acetone

Sa S S S S S S N N/A S S

S S S S S N S S S S N

N N N N N S N N S N N

N N N N N N N N N N N

a

S = solvate, N = unsolvated phase, N/A = solubility was too low. See more details in Table S1, Supporting Information.

CDCA was confirmed to be able to form few solvates, whereas no solvates were obtained for UDCA. Thus, even simple solvate screening is enough to test the solvate formation propensity of these compounds, and the obtained results are fully consistent with conclusion from the available literature. Furthermore, the available crystal structures of ansolvates and solvates of CA, MC, CDCA, UDCA, HDCA, DCA, and LCA were analyzed. This was, however, slightly complicated by the absence of crystal structure information for unsolvated DCA. The analysis of hydrogen bonding in crystal structures of unsolvated CA, MC, CDCA, UDCA, HDCA, and LCA shows that all compounds except for MC have the potential to employ all hydrogen bond donors and acceptors in formation of a hydrogen bond network, but this is fulfilled only in CA, UDCA, and LCA (see Table 2). Meanwhile, hydrogen bond network in Table 2. Packing Characteristic of Crystal Structures of Unsolvated Bile Acids

CA MC CDCA UDCA HDCA LCA

solvate former

HDon/HAcc balancea

Elattice/kJ mol−1

packing indexb

void space/%b

+ + − − − −

+ − − + − +

−210.6 −191.1 −176.6 −196.6 −194.7 −154.9

0.651 0.667 0.693 0.701 0.683 0.662

9.5 3.6 2.8 0 1.3 0

“+” indicates that all hydrogen bond donors (HDon) and acceptors (HAcc, not counting O4 as acceptor) are involved in the formation of hydrogen bonds, “−“ indicates the opposite. bAll structures are determined at 295 K except for that of UDCA (FEBHUP02) determined at 123 K. However, similar results (packing index 0.717 and void space of 0.7%) are obtained for FEBHUP determined at 295 K after automatic addition of hydrogen atoms in Mercury. a

B

DOI: 10.1021/acs.cgd.7b01457 Cryst. Growth Des. XXXX, XXX, XXX−XXX

Crystal Growth & Design

Article

Figure 2. Void space (above) and relative packing index (below, except for DCA for which packing index value is shown) in structures of bile acid solvates. See more details and data in Supporting Information. For DCA values of unsolvated epimer 3β,12β-dihydroxy-5β-cholan-24-oic acid (DADLEA) are included for comparison.

Again, as expected, LCA containing only one hydroxyl group has the least negative lattice energy. Nevertheless, more detailed analysis of interaction energy suggests that intermolecular interactions should be regarded as efficient, and packing index of 0.662 also may have to be considered reasonably efficient (see Supporting Information). Further calculation of packing index and void space as well as evaluation of solvent role in the crystal structure was used to characterize molecular packing in solvates and other multicomponent phases of bile acids. The obtained data for selected solvates of CA, MC, and DCA (see Experimental Section) are presented in Figure 2, with more details and data given in the Supporting Information. Incomplete crystal structures with missing hydrogen atom coordinates were autoedited in Mercury26 before calculation of the presented packing index and void space, while those with absent coordinates for all atoms of solvent molecules were excluded from this representation. Again, in Figure 2 void space obtained using PLATON is given, while the values obtained in Mercury with both settings used (see Experimental Section for details and Table S3 − Table S6 for results) gave essentially identical conclusions. The data presented in Figure 2 (and Supporting Information) clearly show that in nearly all CA and MC solvates the packing index has increased if compared to that of

unsolvated phase, with only slightly lower or unchanged packing index observed for the remaining few solvates. Similarly, crystal structures of most solvates do not contain voids, and for the remaining solvates void space has decreased if compared to that in the unsolvated phase. Although no clearly outlying results were obtained for any of the isostructural solvate groups, G-3 solvates of CA had somewhat larger void space, while G-1 (type 1) solvates of MC had somewhat larger void space and lower packing index than observed for other solvates; see more details in the Supporting Information. The increase in the packing efficiency in MC solvates additionally appears as a more efficient hydrogen bond network employing all MC hydrogen bond donors not fulfilled in the unsolvated phase. This occurs also in solvates where the solvent is not incorporated in the hydrogen bonding network. The absence of crystal structure information for unsolvated DCA made it difficult for us to make indisputable conclusions for this compound. However, it is reported that microcrystals of unsolvated DCA give a powder pattern very similar to those of the inclusion compound with phenantrene, possibly implying that the DCA complexes are primary solid solutions of guest in host,27,28 which could imply that the purely unsolvated stable phase cannot be obtained and unsolvated DCA is obtained with empty channels, and thus expected to have quite inefficient packing (with value of 0.65 marked in Figure 2) and large void C

DOI: 10.1021/acs.cgd.7b01457 Cryst. Growth Des. XXXX, XXX, XXX−XXX

Crystal Growth & Design

Article

prediction of solvate formation propensity, as use of the packing index of unsolvated LCA alone would have led to the wrong conclusion. It should, however, be kept in mind that the identified driving force for solvate formation can be valid only in cases where the solvent mainly acts as a space filler instead of being a major part of the supramolecuar structure by compensating unsatisfied potential intermolecular interactions between the host molecules or providing a linkage between molecules unable to link efficiently to each other. Such examples are, e.g., solvates of bosutinib29 and sulfamethazine.30 Nevertheless, a number of recent examples have shown that the inability of relatively bulky molecules to pack efficiently is the main driving force for extensive solvate formation,5 with packing in solvate structures mostly being more efficient.9,21,31−33 Another useful accomplishment would be identification of factors responsible for different packing efficiency in the unsolvated phases. The molecular structures of bile acids differ with the number and position of the OH groups attached to the steroid ring system. In solvate formers CA and MC there are OH groups attached to carbons C7 and C12. CDCA, HDCA, UDCA, and DCA contain OH groups attached to C6, C7, or C12. From these, only DCA molecules, in which the OH group is attached to C12, are not able to pack efficiently, whereas CDCA and HDCA containing OH attached to C7 in axial or C6 in equatorial position, respectively, pack better, while packing in UDCA containing OH attached to C7 in the equatorial position seems to be optimal. Also, LCA in which there is no OH group attached to C6, C7, or C12 is able to pack efficiently. Thus, the presence of the OH group attached to carbon C12 seems to be the reason for the inability of the analyzed bile acids to pack efficiently.

space (see more details in Supporting Information). For comparison, packing index and void space of unsolvated DCA epimer 3β,12β-dihydroxy-5β-cholan-24-oic acid (DADLEA) having inefficient packing is also presented in Figure 2. Nevertheless, although packing index and void space for DCA solvates with small solvent molecules (two of the hydrates and acetone solvate determined at 295 K) are characteristic for inefficient packing, the packing in DADLEA is even less efficient, whereas the packing index and absence of void space classify most of the remaining DCA solvates as efficiently packed. Despite the fact that crystal structures for part of CA and DCA solvates were determined in low temperature measurements (for CA 13/40 at 173−228 K and 6/40 ≤ 123 K and for DCA 6/20 at 103−213 K), no obvious correlation between packing efficiency and temperature was observed, thus excluding the possibility that the observed increase of the packing efficiency is because structure determination of solvates is sometimes performed at lower temperatures. Thus, it can clearly be seen that the driving force of extensive solvate formation observed for CA, DCA, and MC is the inability of these compounds to form unsolvated phase with efficient packing, while incorporation of solvent molecules in either structure channels or isolated sites usually allows significant increase of the packing efficiency. Among the analyzed bile acids, particularly CA and MC, and most probably also DCA (see above), clearly have significant problems with packing efficiently in their unsolvated phases (packing index 3%). Even though MC additionally has an imbalance of hydrogen bond donors and acceptors, the increase of the packing efficiency is still the driving force for formation of MC solvates, as this imbalance is not compensated even in solvates where solvent is hydrogen bonded to the MC. CDCA and HDCA can be recognized as the next group among the analyzed bile acids having slight problems with packing efficiency in unsolvated phases (with packing index >0.68, but having some relatively large voids (98%), and MC (purity >98%) was obtained from JSC Grindeks (Riga, Latvia). Organic solvents of analytical grade were purchased from commercial sources and used without further purification. Water was deionized in the laboratory (conductivity