The solid-state structure of the β-blocker metoprolol: a combined experimental and in silico investigation

The metoprolol free base has been characterized in the solid state by X-ray diffaction (both single-crystal and variable-temperature powder diffraction) and differential scanning calorimetry. These studies are supplemented by molecular modelling and Hirshfeld surface analysis. Structural relationships with the strictly related betaxolol are discussed.


Introduction
Metoprolol, or (RS)-1-isopropylamino-3-[4-(2-methoxyethyl)phenoxy]propan-2-ol (see a in Scheme 1), is a cardioselective 1 -adrenergic blocking agent that has numerous medical applications, such as the treatment of acute myocardial infarction, heart failure, angina pectoris and hypertension (Benfield et al., 1986;Brogden et al., 1977). The drug is usually manufactured as a racemic mixture, notwithstanding the fact that the 1 -blocking activity resides in the S enatiomer (Dasbiswas et al., 2008). In addition, given its quite low melting point (323 K) (Ionescu et al., 2006), metoprolol is always marketed in salt-based formulations (i.e. tartrate, succinate and fumarate) that differ in the drug-release mechanism (Wikstrand et al., 2003). According to the Biopharmaceutics Classification Scheme, metoprolol belongs to the class I substances (Amidon et al., 1995), meaning that it has both high aqueous solubility and intestinal permeability, which makes this API (active pharmaceutical ingredient) suitable for Extended Release (ER) formulations.
Recently, we have reported on the solid-state structure and thermal behaviour of the tartrate (Paoli et al., 2016) and fumarate salts (Rossi et al., 2018) studied by in silico and experimental techniques. In both cases, comparisons with the crystal structure of the closely related succinate salt (Bartolucci et al., 2009) were made and, where possible, the results were rationalized on the basis of their respective crystal arrangements.
In this article, we have turned our attention to the metoprolol free base (MB hereafter). The interest in the solid-state investigation of MB is due to two main reasons. Firstly, metoprolol shares with a large number of -blocker drugs the 2-hydroxy-3-(isopropylamino)propoxy side arm. Therefore, it would be interesting to gain information about the molecular structure of MB, in particular, the conformational preferences due to the freedom of rotation of such a side arm, and the intermolecular interactions and hydrogen-bond patterns which originate from this side arm in order to find correlations between structural parameters and physicochemical properties, such as melting point and solubility (Datta & Grant, 2004), and possibly to extend these findings to closely related APIs, such as propanolol and betaxolol. Secondly, there is a great deal of interest by the pharmaceutical industry in the investigation of solids containing APIs with improved physicochemical properties. In this context, the assessment of the phase stability and of the thermal behaviour of compounds of pharmaceutical interest (such as temperature-related phase transformations, anisotropic lattice expansion/contraction and thermal stability) provide valuable information (Rossi et al., 2014;Paoli et al., 2016). For example, powder X-ray diffraction (PXRD) and differential scanning calorimetry (DSC) have been used to characterize time-controlled metoprolol tartrate delivery systems using acrylic resins (Eudragit RL and Eudragit RS) for the coating. Systems containing the drug salt were compared to systems containing only the neutral metoprolol or only the tartaric acid to enable a better understanding of the interactions between the metoprolol salt and the film, which can strongly affect the release of the drug (Glaessl et al., 2009).
Although the literature reports (Ionescu et al., 2006) the analysis by single-crystal X-ray diffraction of MB at 173 K (MB-173; monoclinic crystal system, space group P2 1 /n, one independent molecule in the asymmetric unit), neither the atomic coordinates nor the crystal data have been deposited in the Cambridge Structural Database (CSD; Version 5.39 of November 2017; Groom et al., 2016). For this reason, the solidstate structure of metoprolol has been redetermined by a single-crystal X-ray structure analysis.
The molecular structure of the metoprolol molecule has been compared to those of metoprolol-like molecules deposited in the CSD and the conformational space accessible to the 2-hydroxy-3-(isopropylamino)propoxy side arm has been investigated by molecular dynamics simulations and density functional theory (DFT) calculations.
The crystal structure has been analysed with the programs Mercury CSD (Macrae et al., 2008) and Crystal Explorer17 (Turner et al., 2017) in order to identify the contributions to the intermolecular contacts between the metoprolol molecules and the results have been compared to those of structurally related -blocker molecules. Finally, variable-temperature PXRD (VT-PXRD) and DSC measurements were carried out in order to study thermally-induced changes, and the results are discussed.

Experimental
Metoprolol tartrate and betaxolol hydrochloride were purchased from Sigma-Aldrich (product codes M5391-10G and B5683-50MG, respectively) and used without further purification.

Synthesis and crystallization
The metoprolol and betaxolol salts (350 and 100 mg, respectively) were dissolved in a minimal amount of Milli-Q water (0.5 and 2 ml, respectively) and passed through an anion exchange resin (Dowex Marathon 11 chloride form; Sigma-Aldrich CAS 69011-19-4) in order to obtain the free base forms of metoprolol (MB) and betaxolol (BE hereafter) directly in water. Concerning BE, water was completely removed by evaporation under reduced pressure and the resulting solid was dissolved in 3 ml of methanol-water (20:80%, v/v). Slow evaporation of the organic phase at low temperature (277-278 K) gave colourless crystals of BE suitable for single-crystal X-ray diffraction analysis after two weeks. In the case of MB, complete removal of the tartrate anion was confirmed by 1 H NMR analysis of solutions of the compound in D 2 O at pD 11.10 (pH 10.70) before and after research papers treatment with the anion-exchange resin. On purification on a column, the singlet at 4.36 ppm attributed to protons b and b 0 (Qiao et al., 2011) of the tartrate anion (Table S1 and Fig. S1 in the supporting information) disappears, indicating the complete absence of that anion in the final product. Removal of the solvent was performed by evaporation under reduced pressure to a final volume of ca 3 ml. The sample was divided into two aliquots in order to test different crystallization conditions. The first aliquot was allowed to evaporate at room temperature (298 K), resulting in the formation of a microcrystalline powder of MB, after 3 d, suitable for PXRD measurements. In the case of the second aliquot, a slower evaporation of the solvent, performed at 277-278 K, afforded the formation of colourless crystals of MB suitable for singlecrystal X-ray diffraction analysis after three weeks.

X-ray data collection and structure resolution
The crystal structure of MB was investigated by means of single-crystal X-ray diffraction. Measurements were carried out at 100 K with an Rigaku Excalibur Onyx diffractometer using Cu K radiation.
Crystal data, data collection and structure refinement details are summarised in Table 1. All H atoms were located from difference electron-density maps and their coordinates were refined freely, while their displacement parameters were linked to those of their parent atoms, i.e. U iso (H) = 1.2U eq (C,N,O), except for methyl groups, where U iso (H) = 1.5U eq (C). Table 2 lists a selection of the torsion angles. The hydrogen-bond parameters are listed in Table 3.

Variable-temperature unit-cell parameter determination
The crystal lattice parameters of MB in the 130-300 K range were determined from powder X-ray diffraction patterns measured in a vacuum using a Bruker Advance diffractometer (Cu K radiation, 40 kV Â 40 mA), equipped with a Bruker LYNXEYE-XE detector, scanning range 2 = 7-35 , 0.02 increments of 2 and a counting time of 0.8 s/step. The temperature variation rate was 10 K min À1 and, after the target temperature had been reached, the sample was kept for 10 min at that temperature before proceeding with data collection. The patterns underwent a Pawley fit with the software TOPAS (Coelho, 2012). A shifted Chebyshev polynomial with eight coefficients and a pseudo-Voigt function were used to fit the background and peak shape, respectively. The unit-cell parameters, volume and R factor for MB are summarized in Table 4. The unit-cell parameters for BE were determined by single-crystal diffraction analysis. Data were measured at 100, 130, 170, 210, 230, 260 and 300 K using an Rigaku Excalibur Onyx diffractometer. The unit-cell parameters are reported in Table S4 (Accelrys, 2018) and PARST97 (Nardelli, 1995).

Computational methods
Geometry optimizations (MM) and molecular dynamics (MD) simulations were made using the CHARMm Force Field (Brooks et al., 1983). MM calculations were performed on each species using the Smart Minimizer energy minimization procedure implemented in Discovery Studio (Version 2.1; Accelrys, 2018) and before starting the MD simulations the geometry of each compound was further optimized using the steepest descent and conjugate gradient algorithms. MD simulations were carried out at 100 and 300 K, both in vacuum and in an implicit water model; water calculations were performed mimicking the solvent by using a distance-dependent dielectric constant of 80. In the MD simulations, the time step was 1 fs for all runs, the equilibration time was 100 ps and the production time was 1000 ps, and snapshot conformations were sampled every 10 ps. The Minimization, the Standard Dynamics Cascade and Analyze Trajectory, all implemented in Discovery Studio, were the protocols used for energy minimization, MD simulations and analysis of MD trajectories, respectively.
Searching on motifs (to identify interaction motifs between molecular fragments and determine their relative abundance) and Calculating Intermolecular Energies using the UNI intermolecular potentials (Gavezzotti, 1994(Gavezzotti, , 1998 in order to identify the intermolecular interactions which are most significant from an energetic point of view, both carried out using the CSD Materials software (Macrae et al., 2008), were used to analyse the crystal packing arrangement.
CrystalExplorer17 (Turner et al., 2017) was used to compute Hirshfeld surfaces (HS) and their associated 2D (two-dimensional) fingerprint plots to further investigate the intermolecular interactions in the crystal packing of MB and of the strictly related propranolol (PR), BE and 1-[4-(cyanomethyl)-phenoxy]-2-hydroxy-3-(isopropylamino)propane (IA) molecules (details in Section 3.2). Total interaction energies for a cluster of molecules (molecules within a radius of 3.8 Å with respect to the reference molecule) of MB and BE at the B3LYP/6-31G** level of theory were also calculated. The corresponding energy frameworks were then constructed and visualized using the default values (the radii of the cylinders that make up the framework represent the relative strengths of the molecular packing in different directions). In BE, the cyclopropylmethoxy group is disordered over two positions and the model having the highest occupancy factor was used to generate the HS and for energy calculations.

Results and discussion
3.1. Molecular structure from single-crystal X-ray diffraction and modelling studies The metoprolol molecule crystallizes in the monoclinic space group P2 1 /n with one molecule in the asymmetric unit ( Fig. 1). Because the cardiac -blocking activity especially resides in the S enantiomer, the following discussion will be focused on this isomer. Bond lengths and angles are within the expected ranges (Groom et al., 2016). The side chain bearing the isopropyl group adopts an elongated conformation, with the side-chain atoms O1, C7, C8 and C9 trans-disposed (all trans or aT, Table 2), with all atoms, except for O2 and C11, being almost coplanar with the attached aromatic ring, as indicated by the torsion angles that define its orientation. By contrast, the 2-methoxyethyl group is perpendicularly oriented, as indicated by the value of the torsion angle about the C13-C14 bond.
A search of the CSD was carried out to locate structures with the molecular fragment sketched as b of Scheme 1. This moiety, which features the 2-hydroxy-3-(isopropylamino)propoxy side arm together with the phenyl ring, is quite interesting given that it is common to a large variety of -blocker drugs, such as atenolol, betaxolol, practolol and bisoprolol. The CSD survey gives six compounds [neither solvated species nor salts have been taken into account; the structure of a metoprolol analogue (refcode IQEPUP; Melgar-Fernandez et al., 2004) was not taken into account given its R configuration] which, based on the conformation adopted by the chain bearing the isopropyl group, can be classified in four different conformational families, as illustrated in Fig. 2. The superimposition of the X-ray structures of the six molecules found in the CSD highlights that three of them, identified by  Mercury (Macrae et al., 2008) view of the S enantiomer of metoprolol in MB (50% probability displacement ellipsoids). Table 4 Unit-cell parameters, volume (V) and R factor for MB at different temperatures from XRPD data.
16.103 (2) 5.459 (1) 17.858 (6) 100.588 (7)  the refcodes BEMBOK (Laguerre et al., 1981), GAPZEE (Hou et al., 2012) and KAZPOQ (Akisanya et al., 1998), adopt the aT conformation (differences about the final C-N and N-C bonds have been neglected), as found in MB and MB-173 (Ionescu et al., 2006; see Section 1); two molecules, i.e. CEZVIN (de Castro et al., 2007) and one of the crystallographically unique molecules in CIDHAZ (de Castro et al., 2007), show a trans-trans-trans-gauche(+) (tttg + ) arrangement of the C-O-C-C-C atoms, while a trans-transgauche(À)-trans (ttg À t) conformation is shown by ROKNUB (Canotilho et al.,, 2008) and, finally, a trans-trans-gauche(+)gauche(+) (ttg + g + ) conformation is shown by the second independent molecule in the crystal packing of CIDHAZ. This conformational variability is not surprising given that side chains usually have a large conformational freedom, in addition, their conformations can be biased by intermolecular interactions (vide infra). In this context, it appears interesting to study the conformational behaviour of such a chain by investigating the basic structure (BS, see c of Scheme 1) common to all the above-mentioned -adrenoreceptor antagonists by Molecular Dynamics (MD) and Quantum Chemical (QC) methods. The BS_aT, BS_TG, BS_GT and BS_GG conformational isomers representative of the four conformational families (aT, tttg + , ttg À t and ttg + g + , respectively) found in the CSD were used as the starting geometries for MD simulations at 100 and 300 K, both in a vacuum and in an implicit water model.
MD trajectories collected at 100 K, in vacuum and with the implicitly simulated water medium, show overall metoprolol geometries very close to the corresponding starting rotational isomer found in the solid state (see Figs. S2, S3, S6 and S7 in the supporting information). As expected, during MD simulations at 300 K, both in vacuum and in the simulated water medium, the side arm of the metoprolol molecule explores a wider portion of the conformational space. In particular, the starting geometry of the rotational isomer does not affect the space sampled, as shown by the distribution of the side-chain torsion-angle values, which is very similar irrespective of the starting geometry of metoprolol (see Figs. S4, S5, S8 and S9 in the supporting information). In particular, in all cases, 1 accesses the entire range of values, 3 and 4 adopt a trans conformation in vacuum, while in simulated water, 4 also populates gauche conformations. By contrast, 2 appears frozen in the starting trans conformation both in vacuum and simulated solvent, but in vacuum, at least 85% of the snapshot conformations feature an O-HÁ Á ÁO intramolecular contact (distance less than 2.5 Å ), while in simulated solvent, the percentage drops to 18%. Similarly, intramolecular N-HÁ Á ÁO contacts (distance less that 2.5 Å ) are observed in at least 86% of the sampled conformations in vacuum, while the inclusion of a distance-dependent dielectric constant makes such an interaction definitely less important (it is present in less than 24% of the snapshot conformations).
From each MD trajectory at 300 K both in vacuum and in the implicitly simulated water medium, ten snapshot conformations were extracted and their geometries optimized; the all-trans rotational isomer, i.e. the same as found in the solid  Superimposition of the X-ray structures of the neutral species found in the CSD (Groom et al., 2016). Structures are superimposed as ball-and-stick atoms. H atoms have been omitted for clarity. Upper left: aT conformation (BEMBOK = red; GAPZEE = violet; KAZPOQ = brown); upper right: tttg + conformation (CEZVIN = black; CIDHAZ = green); lower left: ttg À t conformation (ROKNUB = pink); lower right: ttg + g + conformation (CIDHAZ = pale blue). state of metoprolol, is always the most stable. An identical result comes from QC geometry optimization: the BS_aT conformational isomer which, as expected, features O-HÁ Á ÁO and N-HÁ Á ÁO contacts, has the lowest energy content, while the BS_GG isomer is the highest in energy [ÁG 298 = 13.19 (B3LYP) and 5.19 kJ mol À1 (B97-D)].
In summary, the aminohydroxy side arm appears quite flexible, being able to change its 3D arrangement in response to the environment, as provided by the X-ray (crystal environment), MD (in vacuum and simulated solvent) and QC (in vacuum) data. Modelling results identify the all-trans conformation as the most stable, irrespective of the model (MM versus QC) and of the medium (vacuum versus simulated solvent), which, consistently, is the most populated in the solid state (X-ray data of MB/MB-173, BEMBOK, GAPZEE and KAZPOQ).

Crystal structure from single-crystal X-ray diffraction and computational studies
In the crystal lattice, alternating R and S molecules of metoprolol related by an inversion centre give rise to zigzag chains extending along the b axis. A view of the crystal packing along the a axis is presented in Fig. 3. As already reported by Ionescu et al. (2006), within the chain, each molecule is held in place by two pairs of intermolecular hydrogen bonds involving the hydroxy and amine groups, which both act as hydrogenbond donors and acceptors (Table 3). For symmetry reasons, each pair of hydrogen bonds consists of two identical inversion-related O-HÁ Á ÁN/NÁ Á ÁH-O and OÁ Á ÁH-N/N-HÁ Á ÁO hydrogen bonds. When the hydroxy group acts as a donor toward the N atom of an inversion-related molecule, the resulting hydrogen bond is strong (Desiraju & Steiner, 1999) [O2-H2Á Á ÁN1 i = 1.92 (3) Å and 178 (3) ; symmetry code: (i) Àx + 1, Ày + 2, Àz + 1; Table 3]; by contrast, the hydroxy group acts as a definitely weaker hydrogen-bond acceptor toward the N-H group of an inverted neighbouring molecule [N1-H1Á Á ÁO2 ii = 2.39 (3) Å and 142 (2) ; symmetry code: (ii) Àx + 1, Ày + 1, Àz + 1; Table 3]. As a whole, these interactions give raise to two intrachain hydrogen-bond patterns of R 2 2 (10) type [R 2 2 (10)>a>a and R 2 2 (10)>b>b] (Bernstein et al., 1995), which are responsible for the formation of infinite chains of metoprolol molecules extending along the b-axis direction (Fig. S10 of the supporting information). Finally, a relatively weak interaction of the C-HÁ Á ÁO type [C5-H5Á Á ÁO3 iii = 2.56 (3) Å and 160 (2) ; symmetry code: (iii) x, y + 1, z; Fig. 3] partially oriented along the b and c axes, exists between homochiral molecules belonging to the same chain.
Since the most significant hydrogen-bond motif, i.e. R 2 2 (10) involves the molecular fragment that metropolol shares with a large number of -blocker drugs, the CSD was searched to find which hydrogen-bond motifs are formed most commonly by a pair of the molecular fragments sketched as d of Scheme 1 through O-HÁ Á ÁN(-H) interactions and the occurrence of the double R 2 2 (10) motif. In most of the retrieved hits (70.6%), at least one O-HÁ Á ÁN(-H) interaction holds the two molecular fragments together. The most common motifs are an infinite chain (C1, i.e. chain, one contact), with frequency 29.4% (calculated as the number of hits found/number of structures that feature the searched fragment), followed by R2 (i.e. ring, two contacts) (27.5%), while rings with four contacts (R4) represent about 10% of the sample (these motif descriptors are not the same as graph-set notation). Three (see e of Scheme 1) of the 14 structures featuring an R2 pattern show the same motif [R 2 2 (10)>a>a and R 2 2 (10)>b>b] as found in MB (and MB-173): two of them, propranolol [refcode PROPRA10 (Ammon et al., 1977), PR in the following] and betaxolol (ROKNUB, BE in the following) belong to the  Views of the crystal lattices of (left) propranolol (PR), (centre) betaxolol (BE) and (right) a precursor of atenolol (IA), showing chains of molecules propagating along the shortest axis direction describing an R 2 2 (10) hydrogen-bond pattern.
-blocker class of drugs, the third is a reaction intermediate in an alternative route for the synthesis of atenolol (KAZPOQ, IA in the following). As already found for MB, in all three cases, mutually inverted molecules face each other and are held together by hydrogen bonds between the hydroxy and amino groups, giving rise to chains extending along the shortest axis direction (Fig. 4 and Table S2 in the supporting information). In other words, the number, types, geometry and patterns of the intermolecular hydrogen bonds described by the OH/NH groups are practically identical. Thus, the overall packing arrangements, as well as the densities and the Kitaigorodskii packing index (KPI) (Kitaigorodskii, 1961;Spek, 1998), are very similar (Table S3 in the supporting information). As already found for MB, and also in PR, BE and IA, hydrogen bonds are definitely stronger when OH acts as a donor than when it acts as an acceptor. Accordingly, in all the crystal lattices, the most significant interaction in energetic terms, as suggested by the intermolecular potential calculated using the empirical UNI pair potential parameters (Gavezzotti 1994(Gavezzotti , 1998, is between the pair of molecules held together by the O-HÁ Á ÁN/NÁ Á ÁH-O pair of hydrogen bonds. The OÁ Á ÁH-N/N-HÁ Á ÁO pair of interactions appears less important from an energetic point of view and, in PR and BE, it even ranks third among the strongest interactions (second in MB and IA; Fig. S11 in the supporting information). In PR and BE, the relative arrangement of the pair of molecules involved in the second strongest interaction (Fig. S12 in the supporting information) suggests thatparallel-displaced interactions in PR (along b) and C-HÁ Á Á contacts in BE (along a) are at work within each chain (for geometrical details, see Table S2 in the supporting information).
The intermolecular interactions which hold together MB, IA, PR and BE in their respective solids were further investigated using Hirshfeld surface (HS) analysis. The corresponding HSs mapped with d norm highlighting the intermolecular contacts are shown in Figs. 5 and 6; in all cases, the dominant interaction is the O-HÁ Á ÁN/NÁ Á ÁH-O pair of hydrogen bonds (two large red spots); the weaker OÁ Á ÁH-N/ N-HÁ Á ÁO couple of interactions, as well as less prominent contacts, show up as pale-red regions. The corresponding fingerprint plots are given in Fig. S13 in the supporting information. All the fingerprint plots feature a pair of spikes which represent the hydrogen bonds involving the NH/OH groups (upper left OH donor, bottom left NH acceptor) and two well-defined lateral wings (except PR, see later) which account for C-HÁ Á Á contacts. Finally Fig. S14 in the supporting information shows the fingerprint plots broken down into contributions from NÁ Á ÁH and CÁ Á ÁC close contacts for the four molecules presented here (Fig. S15 shows the other contributions). From these data, it emerges that the nature and contribution of the intermolecular contacts of MB and BE, which differ with respect to the terminal group (isopropyl instead of cyclopropyl), are very similar; thus, they have almost identical roles in the corresponding crystal packing. By contrast, in PR and IA, CÁ Á ÁC and NÁ Á ÁH contacts also contribute to the crystal packing through the naphthalene group in PR and the cyano group in IA.
Due to the close similarity between MB and BE, the following discussion focuses on these two compounds. Results from intermolecular interaction energy calculations (B3LYP and HF energy models) between molecular pairs in MB and BE confirm that the O-HÁ Á ÁN/NÁ Á ÁH-O pair of hydrogen bonds are by far the most important interactions, followed by the C-HÁ Á Á interactions in BE, while in MB, all the other contacts are almost isoenergetic and definitely less important from an energetic point of view. The values of the interaction energy calculated between the closest molecules are used to construct the energy framework shown in Fig. 7. A comparison of the total energy frameworks evidences the strict similarity between MB and BE.   axes of BE (the same applies when viewing down the molecular axis).
In conclusion, an analysis of the intermolecular contacts in MB, in terms of geometry, motifs, Hirshfeld surface and intermolecular energies, highlights the close resemblance of the metoprolol crystal packing with that of another -blocker drug betaxolol. In both cases, the O-HÁ Á ÁN/NÁ Á ÁH-O hydrogen bonds appear to drive the arrangement of the molecules in the corresponding solid, giving rise to chains of alternating R and S molecules which propagate along the shortest axis direction.
3.3. Crystal structure from X-ray microcrystalline powder diffraction and differential scanning calorimetry analysis The correspondence between the crystal structure of metoprolol free base, as determined by single-crystal X-ray diffraction (MB), and that of the bulk material was checked by comparing calculated (150 K) and measured (130 K) powder diffraction patterns (Fig. S16 in the supporting information). DSC measurements performed in the 298-343 K range do not show any evidence of a thermal event (see Fig. S17 in the supporting information), except that related to the melting at around 324 K (peak 323.9 K, extrapolated peak 324.0 K) with a melting enthalpy of 188.1 J g À1 (50.3 kJ mol À1 ). Consistently, the XRPD patterns measured in the temperature range 130-300 K superimpose quite well (Fig. S18 in the supporting information); there were no differences in the overall number of peaks and in their relative intensities on heating. Thus, no phase changes occur under these experimental conditions up to the MB melting point. However, a closer inspection shows that peaks shift to a different extent as the temperature is increased; several peaks move towards lower 2 values, while the position of others remains almost unchanged, thus suggesting that an anisotropic thermal expansion takes place on raising the temperature. In particular, the shift is evident for the (h00) and (h0l) peaks; by contrast, the (0k0), (hk0) Table 5 Linear () and volume () thermal expansion coefficients (TECs) calculated for MB taking as reference the unit-cell parameter values calculated at 130 K.

Figure 7
Energy frameworks corresponding to the total interaction energy in MB and BE (views are chosen in order to highlight similarities between MB and BE).
temperature. The knowledge that MB undergoes an anisotropic expansion on heating could be relevant when the phase purity of the API, as well as its phase composition in formulations, is checked by a comparison of the powder diffraction pattern of a sample with a reference powder pattern: unexpected differences due to anisotropic lattice expansion/ contraction could lead to wrong conclusions about phase purity/composition. In particular, the lattice parameters calculated from the XRPD patterns (see Section 2) listed in Table 4 confirm this observation; the a axis expands significantly with respect to both b (which by contrast slightly contracts) and c (which remains almost unchanged). This trend is well quantified by the linear thermal expansion coefficients (TECs; Hori et al., 2007;Krishnan et al., 1979) listed in Table 5 (and it is only partially accounted for by the intermolecular hydrogen bonds involving the hydroxy and amine groups that link the metoprolol molecules along the b axis). Due to the crystal packing similarities between MB and BE, it appears interesting to assess the thermal behaviour of betaxolol. In the investigated temperature range (i.e. 100-300 K), BE does not undergo any phase transition, as shown by the single-crystal X-ray diffraction data. Polymorph I of betaxolol (Maria et al., 2013) is stable under the experimental conditions adopted and, what is more, the crystal lattice expands isotropically on raising the temperature (the linear thermal expansion coefficient for BE is reported in Table S5 of the supporting information).
Thus, notwithstanding the strict similiarity between the crystal packings of MB and BE in terms of the nature of the most significant intermolecular contacts (number, type, geometry, motifs, interaction energies) and packing efficiency (density, KPI), they respond differently to thermal stimulus; the metoprolol lattice expands anisotropically, while by contrast and quite surprisingly, an isotropic expansion is observed for betaxolol. Besides, the latter shows, as reported by Canotilho (Canotilho et al., 2008), a slightly higher melting temperature, 341 K, but a smaller melting enthalpy (45.9 kJ mol À1 ). In other words, crystal structure similarities do not imply similar macroscopic properties.
Therefore, however much has been done within the framework of structure-property/function relationships, much remains to be done, especially when APIs are concerned, due to the relevance of their solid-form properties in view of their pharmaceutical development. CrysAlis PRO (Rigaku OD, 2018); program(s) used to solve structure: SIR2004 (Burla et al., 2005); program(s) used to refine structure: SHELXL2014 (Sheldrick, 2015); molecular graphics: ORTEP-3 (Farrugia, 2012), Mercury (Macrae et al., 2008) and Discovery Studio Visualizer (Accelrys, 2018); software used to prepare material for publication: PARST97 (Nardelli, 1995). Special details Geometry. All esds (except the esd in the dihedral angle between two l.s. planes) are estimated using the full covariance matrix. The cell esds are taken into account individually in the estimation of esds in distances, angles and torsion angles; correlations between esds in cell parameters are only used when they are defined by crystal symmetry. An approximate (isotropic) treatment of cell esds is used for estimating esds involving l.s. planes.