Molecular Crystal Forms of Antitubercular Ethionamide with Dicarboxylic Acids: Solid-State Properties and a Combined Structural and Spectroscopic Study

We report on the preparation, characterization, and bioavailability properties of three new crystal forms of ethionamide, an antitubercular agent used in the treatment of drug-resistant tuberculosis. The new adducts were obtained by combining the active pharmaceutical ingredient with three dicarboxylic acids, namely glutaric, malonic and tartaric acid, in equimolar ratios. Crystal structures were obtained for all three adducts and were compared with two previously reported multicomponent systems of ethionamide with maleic and fumaric acid. The ethionamide-glutaric acid and the ethionamide-malonic acid adducts were thoroughly characterized by means of solid-state NMR (13C and 15N Cross-Polarization Magic Angle Spinning or CPMAS) to confirm the position of the carboxylic proton, and they were found to be a cocrystal and a salt, respectively; they were compared with two previously reported multicomponent systems of ethionamide with maleic and fumaric acid. Ethionamide-tartaric acid was found to be a rare example of kryptoracemic cocrystal. In vitro bioavailability enhancements up to a factor 3 compared to pure ethionamide were assessed for all obtained adducts.


Introduction
The obtainment of novel crystal forms is a well-consolidated strategy in the quest for solid molecular materials with enhanced physicochemical properties with respect to those of the pure components. The crystal engineering approach, i.e., the rational design and synthesis of new crystal forms [1], is viable for any molecule that is employed in the solid state, ranging from pigments to explosives [2], from pharmaceuticals to energy storage materials [3]. In the case of active pharmaceutical ingredients (APIs), crystal engineering has proved to be successful in modulating and improving their performances in terms of water solubility [4,5], dissolution rate [6,7], hygroscopicity [8], thermal stability [9], flow properties [10], etc. An important point of this strategy is the fact that some APIs are often doomed to obsolescence because of their poor biopharmaceutical and/or physicochemical properties [11]. Improving them represents a way to restore their value, in some cases even to reduce side effects due to a decrease in the administered doses and to extend rights on the intellectual property. This aspect gains particular importance in

Materials and Methods
FUM, GLU, MLE, MAL and all solvents were purchased from Sigma-Aldrich (Milan, MI, Italy,); ETN was purchased from Alfa Aesar (Thermo Fisher Scientific, Kendal, Germany); TAR was purchased from Schiapparelli (Carlo Erba, Cornaredo, Italy). All reagents were used without further purification.

Synthesis
ETN·GLU: A yellow microcrystalline powder was obtained by manually dry grinding 30 mg (0.18 mmol) of ETN and 24 mg (0.18 mmol) of GLU for 60 min. Crystals were obtained through seeding crystallization of the ground product in ethanol.
ETN·MAL: An orange microcrystalline powder was obtained by the slurry technique: 50 mg (0.3 mmol) of ETN and 31 mg (0.3 mmol) of MAL were stirred for 4 h with a few drops of ethanol. Crystals, suitable for SCXRD, were obtained through seeding crystallization of the slurried product in ethyl acetate.
ETN·TAR: Crystals were obtained through slow evaporation at room temperature of a methanol solution containing 30 mg (0.18 mmol) of ETN and 27 mg (0.18 mmol) of TAR. Despite many attempts, ETN·TAR could not be reproduced in pure form to undergo further analyses.
ETN·FUM: Crystals were obtained through slow evaporation at room temperature of a methanol solution containing 30 mg (0.18 mmol) of ETN and 21 mg (0.18 mmol) of FUM.
ETN·MLE: An orange microcrystalline powder was obtained by manually dry grinding 30 mg (0.18 mmol) of ETN and 21 mg (0.18 mmol) of MLE for 30 min. Crystals, suitable for SCXRD, were Scheme 1. Representation of the employed molecules, with atom numbering.

Materials and Methods
FUM, GLU, MLE, MAL and all solvents were purchased from Sigma-Aldrich (Milan, Italy,); ETN was purchased from Alfa Aesar (Thermo Fisher Scientific, Kendal, Germany); TAR was purchased from Schiapparelli (Carlo Erba, Cornaredo, Italy). All reagents were used without further purification.

Synthesis
ETN·GLU: A yellow microcrystalline powder was obtained by manually dry grinding 30 mg (0.18 mmol) of ETN and 24 mg (0.18 mmol) of GLU for 60 min. Crystals were obtained through seeding crystallization of the ground product in ethanol.
ETN·MAL: An orange microcrystalline powder was obtained by the slurry technique: 50 mg (0.3 mmol) of ETN and 31 mg (0.3 mmol) of MAL were stirred for 4 h with a few drops of ethanol. Crystals, suitable for SCXRD, were obtained through seeding crystallization of the slurried product in ethyl acetate.
ETN·TAR: Crystals were obtained through slow evaporation at room temperature of a methanol solution containing 30 mg (0. 18

Raman Spectroscopy
Raman spectra were registered with a Bruker Vertex 70 instrument (Bruker, Billerica, MA, USA), equipped with a RAM II module. An excitation source at 1064 nm was used, with a laser power between 10 and 50 mW and a number of scans between 80 and 500, depending on the analyzed sample, with a resolution of 4 cm −1 . The employed spectral range is comprised between 50 and 4500 cm −1 , using a CaF 2 beam splitter. Raman spectra are not discussed as they were used only for screening purposes, but they are reported in Figures S1-S4.

X-Ray Diffraction (SCXRD and PXRD)
Single crystals of ETN·GLU, ETN·MAL and ETN·TAR were analyzed with a Gemini R Ultra diffractometer (Rigaku Oxford Diffraction, Abingdon, Oxfordshire, UK) operating at 293(2) K, using a Mo Kα source (λ = 0.71073 Å). Data collection and reduction were performed using the CrysAlisPro software (Rigaku Oxford Diffraction, Abingdon, Oxfordshire, UK). The crystal structure was solved by direct methods and refined with the full matrix least-squares technique on F2 using the SHELXS-97 and SHELXL-97 programs (Structural Chemistry Department at the University of Göttingen, Germany). All non-hydrogen atoms were refined anisotropically; hydrogen atoms bonded to unambiguous sites were placed in geometrical positions and refined using the riding model. Hydrogen atoms between pyridinic nitrogen and carboxylic oxygen sites of nearby molecules have been detected in the Fourier maps, and their position has been further confirmed through SSNMR. See Table 1 for the crystal data and structure refinement parameters for ETN·GLU, ETN·MAL and ETN·TAR, and Tables S2-S7 for the measured crystallographic distances and angles (refer to Scheme 1 for atom numbering). ETN·TAR presents a kryptoracemic structure, and the absence of an inversion center (although quite certain from the near 0 Flack parameter) [29] or other second-type symmetry elements has been checked by pseudosymmetry search using the PSEUDO program [30] of Bilbao Crystallographic Server, and no centrosymmetric supergroup compatible with the experimental atomic positions has been found.
Powder diffractograms were obtained on the same Gemini R Ultra diffractometer (Rigaku Oxford Diffraction, Abingdon, Oxfordshire, UK), equipped with an X-ray source using Cu Kα radiation (λ = 1.54 Å). Data were collected and processed through the CrysAlisPro software.
CCDC accession codes 2019883, 2019884 and 2019885 contain the supplementary crystallographic data for ETN·MAL, ETN·GLU and ETN·TAR, respectively. These data can be obtained free of charge via www.ccdc.cam.ac.uk/data_request/cif, or by e-mailing data_request@ccdc.cam.ac.uk, or by contacting The Cambridge Crystallographic Data Centre, 12 Union Road, Cambridge CB2 1EZ, UK; fax: +44-1223-336033.  13 C and 15 N, respectively. Samples were packed in cylindrical zirconia rotors (4 mm o.d., Bruker, Billerica, MA, USA), with a sample volume of 80 µL. 13 C and 15 N spectra were acquired at room temperature with a rotation frequency of 12 and 9 kHz, respectively. All 13 C and 15 N experiments employed the RAMP-CP pulse sequence ( 1 H 90 • pulse = 3.6 µs; contact time = 4 ms) with the TPPM 1 H decoupling (rf field = 69.4 kHz) during the acquisition period. Detailed acquisition parameters (number of scans, relaxation delays, contact times) may be found in Table S7. All employed relaxation delay values were optimized on each sample by means of 1 H saturation recovery experiments and obtained by multiplying the measured T 1 1 H values by 1.27, to ensure full relaxation. 13 C and 15 N chemical shift scales were referenced with the resonance of glycine ( 13 C methylene signal at 43.5 ppm), (NH 4 ) 2 SO 4 ( 15 N signal at 24.6 ppm with respect to NH 3 ), respectively, as external standards.

Thermal Analyses
TGA measurements (TA Instruments, New Castle, UK) were performed over a temperature range of 30-350 • C under a 50 mL·min −1 N 2 flow, on a Q600 SDT TA instrument equipped with a DSC heat flow analyzer. Samples (5-10 mg of weight) were placed into the furnace inside alumina crucibles and heated with a ramp of 10 • C·min −1 . DSC curves were collected on a DSC Q200 TA Instrument (TA Instruments, New Castle, UK). Samples were accurately weighed (5-10 mg) and put into sealed aluminum pans. Calibration for temperature and heat flow was performed using high purity standards of n-decane, benzene and indium. All measurements were performed in a 30-350 • C temperature range, with heating rates of 10 • C·min −1 .

Dissolution Kinetic Tests (DKTs)
DKTs were carried out in phosphate buffer (pH = 7.4). For each measurement, 4 mg of either ETN or its adducts were added to 100 mL of the thermostatically controlled (at 37 • C) dissolution medium. Dissolution parameters were evaluated for 60 min. The solution was kept homogeneous by continued stirring at 100 rpm, and concentrations were measured using an optical fiber system (HELLMA, Milan, Italy) linked to a spectrophotometer. UV measurements (ZEISS, Wetzlar, Germany) were performed at the maximum absorption wavelength of ETN, namely 288 nm. A calibration curve ( Figure S5) was obtained with five diluted ETN solutions in phosphate buffer (the concentrations used were the following: 8, 10, 16, 20 and 40 mg/L), while pure phosphate buffer was used as the blank.

Results and Discussion
Three novel crystal forms were obtained by means of solution or mechanochemical techniques. These are a salt of ETN with malonic acid (ETN·MAL), a cocrystal between ETN and glutaric acid (ETN·GLU) and a salt cocrystal of ETN with tartaric acid (ETN·TAR). Two crystal forms of ETN were reproduced from the literature, namely a salt of ETN with maleic acid (ETN·MLE) [27] and a cocrystal between ETN and fumaric acid (ETN·FUM) [26]. Table 2 summarizes the techniques used for preparing the new crystal forms and the outcome, in terms of stoichiometry and ionization state. The crystal structures of all the adducts were obtained through SCXRD. Moreover, for each adduct, except ETN·TAR, which, despite several attempts, could not be reproduced to undergo further analysis, the XRD powder patterns calculated from crystal structures were compared to the experimental powder diffractograms obtained from bulk powders to confirm that the selected crystals were representative of the whole product (see Figures S6-S9). The HB pattern ( Figure 2) is characterized by the presence of a centrosymmetric (8) dimer, formed between the carboxylic groups of two GLU molecules (d O8′-O9′ = 2.667 (4) Å). Additionally, the COOH groups of GLU molecules, which are not involved in the formation of the dimers, interact with the NH2 groups (d N10-O6′ = 2.932 (4) Å) and with the pyridinic N5 (see above) of two different ETN molecules. The result is the formation of (22) cyclic motifs, as visible in Figure 2.  The HB pattern ( Figure 2) is characterized by the presence of a centrosymmetric R 2 2 (8) dimer, formed between the carboxylic groups of two GLU molecules (d O8 -O9 = 2.667 (4) Å). Additionally, the COOH groups of GLU molecules, which are not involved in the formation of the dimers, interact with the NH 2 groups (d N10-O6 = 2.932 (4) Å) and with the pyridinic N5 (see above) of two different ETN molecules. The result is the formation of R 4 4 (22) cyclic motifs, as visible in Figure 2. The HB pattern ( Figure 2) is characterized by the presence of a centrosymmetric (8) dimer, formed between the carboxylic groups of two GLU molecules (d O8′-O9′ = 2.667 (4) Å). Additionally, the COOH groups of GLU molecules, which are not involved in the formation of the dimers, interact with the NH2 groups (d N10-O6′ = 2.932 (4) Å) and with the pyridinic N5 (see above) of two different ETN molecules. The result is the formation of (22) cyclic motifs, as visible in Figure 2.

ETN·TAR
ETN·TAR crystallizes in a monoclinic non-centrosymmetric P21 space group. It is a kryptoracemate, i.e., a compound that crystallizes in a non-centrosymmetric space group containing only symmetry elements of the first type (Sohnke group), despite containing both the enantiomers of a molecule in the same lattice [31]. This phenomenon is still rarely detected in both organic [31] and organometallic [32] crystals (0.1% of all the structure reported in the CSD database), although some

ETN·TAR
ETN·TAR crystallizes in a monoclinic non-centrosymmetric P21 space group. It is a kryptoracemate, i.e., a compound that crystallizes in a non-centrosymmetric space group containing only symmetry elements of the first type (Sohnke group), despite containing both the enantiomers of a molecule in the same lattice [31]. This phenomenon is still rarely detected in both organic [31] and organometallic [32] crystals (0.1% of all the structure reported in the CSD database), although some attempt to rationally develop some functional material based on this peculiarity has been considered [33,34]. Its explanation is deeply debated, although it is clearly related to the existence of high Z' structures and pseudosymmetry [35]. However, the existence of the first kryptoracemic cocrystal has

ETN·TAR
ETN·TAR crystallizes in a monoclinic non-centrosymmetric P2 1 space group. It is a kryptoracemate, i.e., a compound that crystallizes in a non-centrosymmetric space group containing only symmetry elements of the first type (Sohnke group), despite containing both the enantiomers of a molecule in the same lattice [31]. This phenomenon is still rarely detected in both organic [31] and organometallic [32] crystals (0.1% of all the structure reported in the CSD database), although some attempt to rationally develop some functional material based on this peculiarity has been considered [33,34]. Its explanation is deeply debated, although it is clearly related to the existence of high Z' structures and pseudosymmetry [35]. However, the existence of the first kryptoracemic cocrystal has been reported only in 2016 [28], making our result quite peculiar. In the asymmetric unit ( Figure 5), four molecules are present: two ETN molecules and two TAR molecules (both enantiomers).
Pharmaceutics 2020, 12, x FOR PEER REVIEW 10 of 18 been reported only in 2016 [28], making our result quite peculiar. In the asymmetric unit ( Figure 5), four molecules are present: two ETN molecules and two TAR molecules (both enantiomers).  (2) Å. This introduces a degree of uncertainty in the position of the H atoms along the HB axes, which makes it complicated to assess the neutral or ionic nature of the adduct. Unfortunately, it was not possible to confirm it by means of SSNMR. As far as the X-ray analysis is concerned, the structure can be defined as a salt cocrystal as ETN is present is in both its neutral and ionic forms. In the HB pattern ( Figure 6), chains of alternated ETN and TAR molecules are present. They are linked by HB N5 + -H···O − interactions and (8) motifs involving the thioamidic (ETN) and the carboxylic (TAR) groups. Since TAR is present in both its enantiomeric forms, the two strands differ in terms of chirality, making the distances not equivalent. The bottom molecule in Figure 5 displays the following distances: d N5-O8′ = 2.571 (4) Å, d S11-O6′ = 3.144 (2) Å and d N10-O5′ = 3.025 (4) Å. The top molecule in Figure 5 presents the following distances: d N5-O8′ = 2.549 (4) Å, d S11-O6′ = 3.088 (2) Å and d N10-O5′ = 2.960 (4) Å.
The chains interact with a complex pattern of HBs involving all OH and carboxylic groups of TAR and the thioamidic group of ETN.  (2) Å. This introduces a degree of uncertainty in the position of the H atoms along the HB axes, which makes it complicated to assess the neutral or ionic nature of the adduct. Unfortunately, it was not possible to confirm it by means of SSNMR. As far as the X-ray analysis is concerned, the structure can be defined as a salt cocrystal as ETN is present is in both its neutral and ionic forms. In the HB pattern ( Figure 6), chains of alternated ETN and TAR molecules are present. They are linked by HB N5 + -H···O − interactions and R 2 2 (8) motifs involving the thioamidic (ETN) and the carboxylic (TAR) groups. Since TAR is present in both its enantiomeric forms, the two strands differ in terms of chirality, making the distances not equivalent. The bottom molecule in Figure 5 displays the following distances: d N5-O8 = 2.571 (4) Å, d S11-O6 = 3.144 (2) Å and d N10-O5 = 3.025 (4) Å. The top molecule in Figure 5 presents the following distances: d N5-O8 = 2.549 (4) Å, d S11-O6 = 3.088 (2) Å and d N10-O5 = 2.960 (4) Å.
The chains interact with a complex pattern of HBs involving all OH and carboxylic groups of TAR and the thioamidic group of ETN.
The presence of both enantiomers of TAR distinguishes the chains in the disposition of the OH groups, generating a double layer (Figure 7).
The structures of ETN·FUM and ETN·MLE are already discussed in [26] and [27]. For the sake of clarity, the asymmetric units and the HB networks are reported in Figures S10-S13.  Pharmaceutics 2020, 12, x FOR PEER REVIEW 11 of 18  The structures of ETN·FUM and ETN·MLE are already discussed in [26] and [27]. For the sake of clarity, the asymmetric units and the HB networks are reported in Figures S10-S13.

SSNMR
SSNMR was useful to verify the neutral or ionic nature of all adducts, except ETN·TAR, strengthening the X-ray evidence [36][37][38][39]. Indeed, the position of the H atoms along the HB axis was not always clearly detected from X-ray analyses. Through 1D 13 C CPMAS ( Figure 8) and 15 N CPMAS (Figure 9) experiments, the SSNMR analysis focused on the 13 C resonances of the carboxylic groups of the acids and the 15 N signals of N5 (pyridinic) and N10 (thioamidic) of ETN. Indeed, these chemical shifts are very sensitive to the protonation state of the corresponding moieties [37]. All 13 C and 15 N chemical shifts with their relative assignments are reported in Table 3. The presence of both enantiomers of TAR distinguishes the chains in the disposition of the OH groups, generating a double layer (Figure 7). The structures of ETN·FUM and ETN·MLE are already discussed in [26] and [27]. For the sake of clarity, the asymmetric units and the HB networks are reported in Figures S10-S13.

SSNMR
SSNMR was useful to verify the neutral or ionic nature of all adducts, except ETN·TAR, strengthening the X-ray evidence [36][37][38][39]. Indeed, the position of the H atoms along the HB axis was not always clearly detected from X-ray analyses. Through 1D 13 C CPMAS ( Figure 8) and 15 N CPMAS (Figure 9) experiments, the SSNMR analysis focused on the 13 C resonances of the carboxylic groups of the acids and the 15 N signals of N5 (pyridinic) and N10 (thioamidic) of ETN. Indeed, these chemical shifts are very sensitive to the protonation state of the corresponding moieties [37]. All 13 C and 15 N chemical shifts with their relative assignments are reported in Table 3.

SSNMR
SSNMR was useful to verify the neutral or ionic nature of all adducts, except ETN·TAR, strengthening the X-ray evidence [36][37][38][39]. Indeed, the position of the H atoms along the HB axis was not always clearly detected from X-ray analyses. Through 1D 13 C CPMAS ( Figure 8) and 15 N CPMAS (Figure 9) experiments, the SSNMR analysis focused on the 13 C resonances of the carboxylic groups of the acids and the 15 N signals of N5 (pyridinic) and N10 (thioamidic) of ETN. Indeed, these chemical shifts are very sensitive to the protonation state of the corresponding moieties [37]. All 13 C and 15 N chemical shifts with their relative assignments are reported in Table 3. 13 C CPMAS spectra offer the chance to assess the involvement of the carboxylic groups of the coformers in deprotonation or HB contacts. The spectrum of ETN·GLU exhibits two carboxylic resonances at 182.7 and 177.8 ppm. This can be explained by the variation in the network of interactions engaging the two COOH moieties. The former is assigned to a COOH group (182.7 ppm) forming a homodimeric R 2 2 (8) synthon with neighboring GLU molecules, as also observed in pure GLU (181.4 ppm) [40]. This translates into high-frequency chemical shifts for carboxylic groups, comparable to those typical of carboxylate moieties [37]. In ETN·GLU, the second resonance (177.8 ppm) is typical of neutral COOH groups, in this case engaging in a COOH···N HB. This nicely agrees with X-ray data as in the case of ETN·MAL. As a matter of fact, pure MAL displays carboxylic homodimers (δ = 174.3/174.8 ppm) [41], while, in ETN·MAL, we attribute the signal at 173.6 ppm to a COO − group and the remaining peak (169.8 ppm) to a COOH moiety involved in a R 2 2 (12) dimeric COOH···O HB. Homodimers are also characteristic of pure FUM [42] (172.3 ppm); on the contrary, in ETN·FUM [26], both COOH groups stay protonated and are no longer involved in homodimeric interactions, which explains the low-frequency shift of their resonances (170.0 and 167.9 ppm).
Pure MLE represents an exception, since it does not exhibit homodimeric synthons [43], which justifies the relatively low chemical shift of one of the COOH groups (169.2 ppm); the other COOH is involved in a COOH···O intramolecular interaction, bringing its chemical shift up to 172.7 ppm. In ETN·MLE [27], a single resonance can be observed, at 172.5 ppm. This can be traced back to the high symmetry of the hydrogenmaleate group, which leads the two carboxylic groups to be very similar (d C4 -O8 = 1.272 Å; d C4 -O7 = 1.241 Å and d C1 -O6 = 1.285 Å; d C1 -O5 = 1.233 Å, atom numbering in Figure S12) despite the deprotonation of one of them. Notably, the spectrum of the salt exhibits extra peaks in the aliphatic and aromatic regions with lower intensity, specifically those centered at about 17, 27, 142 and 151 ppm (highlighted with red ovals in Figure 8). These are due to disorder associated to the ethyl and pyridyl groups, as also observed in the crystal structure (see Figure S12).    In SSNMR, the 15 N chemical shift is recognized as being particularly sensitive and highly reliable on the position of neighboring protons [37]. As indicated by the drastic low-frequency shift (∆δ > 80 ppm) of the N5 signal of ETN from 308.9 ppm (pure ETN) to 215.4 ppm (ETN·MAL) and 212.7 ppm (ETN·MLE), the two adducts are confirmed to be salts, in fair agreement with X-ray measurements [27]; in ETN·FUM and ETN·GLU, the N5 signal shifts to lower frequencies as well (276.9 and 287.4 ppm, respectively), but the variation is lower than for ETN·MAL or ETN·MLE (∆δ ≈ 32 and 21 ppm), and it is consistent with the formation of a HB involving N5 rather than a proper proton transfer [26]. This indicates the neutral nature of ETN·FUM and ETN·GLU, which are to be considered cocrystals, confirming the X-ray findings.

Thermal Analyses
Thermal analyses were run to evaluate the thermal behavior of the adducts with respect to pure ETN, which melts at 165.6 • C. The corresponding curves are reported in Figures S14-S21. Table 4 reports all the obtained values. In all cases, endothermic DSC peaks, corresponding to lower melting points than for pure ETN, are observed. This behavior is recurrent for ETN as all the adducts reported in literature are characterized by lower melting points [16,[25][26][27].

Dissolution Kinetic Tests
The dissolution rate for all obtained adducts, except for ETN·TAR, was evaluated in order to assess its variation with respect to pure ETN. Dissolution tests were already performed at pH = 1.2 for ETN·FUM [23] and ETN·GLU [23]. To the best of the authors' knowledge, this is the first time they are conducted at physiological pH values (7.4). Concentrations (mg/L) were plotted against time (min), as shown in Figure 10. The dissolution rate of ETN in ETN·MLE is the highest among the obtained adducts. Nonetheless, a significant improvement in the dissolution rate of ETN is observed for all of them. The ratios between the Area Under the Curve (AUC) values of each adduct and pure ETN are reported in Table 5. This parameter allows one to assess the increase in the in vitro bioavailability of ETN in the crystal forms [44]. In all cases, a remarkable increase from two up to eight times is observed.
are conducted at physiological pH values (7.4). Concentrations (mg/L) were plotted against time (min), as shown in Figure 10. The dissolution rate of ETN in ETN·MLE is the highest among the obtained adducts. Nonetheless, a significant improvement in the dissolution rate of ETN is observed for all of them. The ratios between the Area Under the Curve (AUC) values of each adduct and pure ETN are reported in Table 5. This parameter allows one to assess the increase in the in vitro bioavailability of ETN in the crystal forms [44]. In all cases, a remarkable increase from two up to eight times is observed.

Conclusions
ETN proved promising to engineer new crystal forms with enhanced physicochemical properties. The presence in its molecular structure of a thioamidic moiety and of a heterocyclic N atom makes it easy to salify or cocrystallize with dicarboxylic acids. Three new crystal forms were obtained-namely, a salt (ETN-MAL), a cocrystal (ETN-GLU) and a salt-cocrystal (ETN-TAR). As in all cases reported in the literature, in ETN-MAL and ETN-TAR, N5 is protonated, while, in ETN-GLU and ETN-TAR, COOH···N and C=O···HN contacts are present. The salt cocrystal with TAR presents the rare characteristic to be a kryptoracemic cocrystal, a racemate that crystallizes into a Sohnke group; this behavior can be attributed to the concomitant presence of both the enantiomers in the

Conclusions
ETN proved promising to engineer new crystal forms with enhanced physicochemical properties. The presence in its molecular structure of a thioamidic moiety and of a heterocyclic N atom makes it easy to salify or cocrystallize with dicarboxylic acids. Three new crystal forms were obtained-namely, a salt (ETN-MAL), a cocrystal (ETN-GLU) and a salt-cocrystal (ETN-TAR). As in all cases reported in the literature, in ETN-MAL and ETN-TAR, N5 is protonated, while, in ETN-GLU and ETN-TAR, COOH···N and C=O···HN contacts are present. The salt cocrystal with TAR presents the rare characteristic to be a kryptoracemic cocrystal, a racemate that crystallizes into a Sohnke group; this behavior can be attributed to the concomitant presence of both the enantiomers in the asymmetric unit with some degree of distortion between each other, preserving their generation through an inversion center, a mirror plane or a glide.
The solid-state characterization of all the adducts was performed by SCXRD analyses and supported by 13 C and 15 N CPMAS SSNMR experiments. The latter are particularly informative, since they provide unambiguous results. These made it possible to assess the purity, the degree of crystallinity and the ionic/neutral nature, clarifying the exact position of protons, which were often uncertain in the obtained X-ray structures.
As for their physicochemical properties, all analyzed adducts show lower melting points than pure ETN. In this sense, by also comparing literature data (more than 10 adducts), we can affirm that cocrystallization systematically decreases its melting point. The dissolution profile for each analyzed adduct was evaluated. Their dissolution rates all proved significantly higher than for the pure API. In particular, ETN·MLE stands out as eight times more bioavailable (in vitro) than pure ETN.