Dihydrofolate Reductase Inhibitors: The Pharmacophore as a Guide for Co-Crystal Screening

In this work, co-crystal screening was carried out for two important dihydrofolate reductase (DHFR) inhibitors, trimethoprim (TMP) and pyrimethamine (PMA), and for 2,4-diaminopyrimidine (DAP), which is the pharmacophore of these active pharmaceutical ingredients (API). The isomeric pyridinecarboxamides and two xanthines, theophylline (THEO) and caffeine (CAF), were used as co-formers in the same experimental conditions, in order to evaluate the potential for the pharmacophore to be used as a guide in the screening process. In silico co-crystal screening was carried out using BIOVIA COSMOquick and experimental screening was performed by mechanochemistry and supported by (solid + liquid) binary phase diagrams, infrared spectroscopy (FTIR) and X-ray powder diffraction (XRPD). The in silico prediction of low propensities for DAP, TMP and PMA to co-crystallize with pyridinecarboxamides was confirmed: a successful outcome was only observed for DAP + nicotinamide. Successful synthesis of multicomponent solid forms was achieved for all three target molecules with theophylline, with DAP co-crystals revealing a greater variety of stoichiometries. The crystalline structures of a (1:2) TMP:THEO co-crystal and of a (1:2:1) DAP:THEO:ethyl acetate solvate were solved. This work demonstrated the possible use of the pharmacophore of DHFR inhibitors as a guide for co-crystal screening, recognizing some similar trends in the outcome of association in the solid state and in the molecular aggregation in the co-crystals, characterized by the same supramolecular synthons.

Co-crystals have emerged in recent years as attractive, alternative solid forms of active pharmaceutical ingredients (APIs), with several studies aiming at the optimization of their physicochemical properties and/or biopharmaceutical performance [12]. They are made up of the API and of one or more co-formers, all of which are solids when pure and at ambient conditions, joined together in a stoichiometric ratio in a new crystalline structure [12,13].
The co-crystal components are associated in supramolecular synthons, most commonly linked by hydrogen bond interactions [12,13].
The supramolecular synthon approach, which looks for hydrogen bonding complementarities between the target molecule and the potential co-formers, is often used as a starting point for co-crystal screening [14,15]. In this context, it is interesting to investigate if DAP can be a guide for the screening of co-crystals of DHFR inhibitors containing this pharmacophore: the same type of supramolecular synthons is expected for DAP and for the DHFR inhibitors, although the molecular complexity of the latter may also play a role in the co-crystallization trial outcome. Although several studies can be found in literature concerning the investigation of co-crystals, and mainly of salts of TMP [16][17][18][19][20][21][22][23][24][25] and PMA [19,[26][27][28][29][30][31][32][33][34] and, to a lesser extent, of DAP [35][36][37][38][39], to the best of our knowledge, an investigation comparing co-crystallization of these three related compounds with the same potential co-formers is not available.
The current work presents the results of an ongoing investigation of co-crystallization of trimethoprim, pyrimethamine and the core fragment 2,4-diaminopyrimide with different co-formers. Experimental results of co-crystallization screening, carried out in the same experimental conditions, are presented and discussed for DAP and TMP with two different types of co-formers that are capable of giving rise to different supramolecular heterosynthons, with the target molecules having, therefore, different competing/complementary effects towards the target homodimers: the three isomeric pyridinecarboxamides, picolinamide (PA), nicotinamide (NA) and isonicotinamide (INA), and the two xanthines, theophylline (THEO) and caffeine (CAF) (Figure 1). Some possible hydrogen-bonded motifs between the DAP scaffold and each of these types of co-formers are shown in Figure  1i and 1.j. The screening of co-crystals between PMA and the pyridinecarboxamides is also carried out, complementing our previous work [27] on the co-crystallization of this DHFR inhibitor with caffeine and theophylline.

Results and Discussion
COSMOquick software [40,41] was used in the search for the propensity of DAP, TMP and PMA to form co-crystals with the co-formers that are investigated in the current work. This approach takes into account the interaction between the target compound and the coformer as the excess enthalpy (∆H ex ) of the undercooled melt made up of both components in a specified proportion [41]. The software also calculates the empirical parameter f fit for a 1:1 stoichiometry which, in addition to ∆H ex , also takes into account the flexibility of the molecules described by the number of rotatable bonds of the target molecule and the co-former [41]. Negative ∆H ex values and low f fit values should indicate a propensity for co-crystal formation.
Our previous results concerning PMA + THEO and PMA + CAF confirm successful (1:1) co-crystal formation for both systems, with polymorphism of the co-crystals being observed in both cases [27].
The results obtained for the systems that were investigated in the current work are presented in Table 1 and can provide a qualitative guide to the expectation of co-crystal formation. A general comparison of the estimated values for excess enthalpy of mixing indicate that a successful outcome is probable between DAP, TMP or PMA and the two investigated xanthines, although outcomes are less favorable for TMP. The estimated ∆H ex values with the pyridinecarboxamide co-formers show a more favorable tendency for the co-crystallization of these three molecules with DAP and picolinamide with PMA. Other values are close to zero and are probably not meaningful, while indicating a smaller likelihood of the co-crystallization of TMP with the three pyridinecarboxamides. Table 1. COSMOquick parameters, ∆H ex (kcal mol −1 ) and f fit , obtained for co-crystal screening between the target compounds 2,4-diaminopyrimidine, trimethoprim and pyrimethamine, and the co-formers, picolinamide, isonicotinamide, nicotinamide, theophylline and caffeine; x is the target compound mole fraction.

(DAP/TMP/PMA) + Pyridinecarboxamides Binary Systems
Co-crystal formation of DAP, TMP and PMA with the isomeric pyridinecarboxamides was investigated experimentally using mechanochemistry as the screening method. Sample characterization was carried out using differential scanning calorimetry (DSC) together with the predicted (solid + liquid) binary phase diagrams, complemented by infrared spectroscopy (FTIR) or/and X-ray powder diffraction (XRPD) analysis.
Equation (1) was used to predict the liquidus for the binary systems made up of TMP or PMA and one of the pyridinecarboxamides and for DAP + PA and DAP + INA. Equations (1) and (2) [42] were used for DAP + NA system. 1 In these equations, T fus represents the liquidus temperature of a mixture of mole fraction x i . T * f us and ∆ f us H * m are the melting temperature and molar melting enthalpy of  Table S1 (ESI) were used in the calculations. The experimental solidus was obtained from the onset of the first observed DSC peak and the liquidus from the corrected maximum of the second one [43,44].
For all the systems, except for DAP + NA, the binary (solid + liquid) phase diagrams ( Figure S1; obtained from the DSC curves presented in Figures S2-S10) show only slight deviations from the predictions of a diagram with a simple eutectic, with the ideal liquid phase (eq. (1)) confirming that no co-crystals were obtained in the experimental conditions used. The infrared spectra shown as electronic supplementary material are, as expected, the sum of those of the target compound and of one of the polymorphs of the co-former (Figures S11 and S12-FTIR, DAP + PA and DAP + INA; Figures S13-S15-FTIR, TMP + PA, TMP + INA and TMP + NA; Figures S16-S18-FTIR and PMA + PA, PMA + INA, PMA + NA).
A (1:3) DAP:NA co-crystal was synthesized by mechanochemistry, as confirmed by the experimental binary (solid + liquid) phase diagram shown in Figure 2, and by XRP ( Figure 3) and FTIR ( Figure S19). The binary (solid + liquid) phase diagram is quite well described by eq. (1) when an excess of DAP is present, and by eq. (2) in the vicinity of the co-crystal, between the two eutectic invariants at T e1 ≈ 124.5 • C, x e1 ≈ 0.04 and T e2 ≈ 117.5 • C, x e2 ≈ 0.65. Unique reflections are observed in the co-crystal X-ray powder diffractogram, for instance, at 2θ = 9.  (Figure 3). Shifts of the wavenumbers of N-H stretching modes are observed for the cocrystal relative to both pure NA and pure DAP, as well as of the C=O stretching mode of NA ( Figure S19).
The pure co-crystal could also be obtained in a slurry of the pure components in ethanol, kept at 25 • C for 2 days (1:3 DAP + NA molar ratio).

General Remarks
The low propensity of DAP and both TMP and PMA to give rise to co-crystals with the isomeric pyridinecarboxamides, as predicted in the framework of COSMOquick, was confirmed. The DAP + NA exception is certainly a consequence of the greater simplicity of the DAP molecule, hence giving rise to fewer competing supramolecular association possibilities, and of the particular nicotinamide molecular structure. Concerning picolinamide, its lower co-crystal formation propensity when compared to their two isomers, with a different electronic structure in the ring, has been pointed out [45,46].  Figure 3). Shifts of the wavenumbers of N-H stretching modes are observed for the co-crystal relative to both pure NA and pure DAP, as well as of the C=O stretching mode of NA ( Figure  S19). The pure co-crystal could also be obtained in a slurry of the pure components in ethanol, kept at 25 °C for 2 days (1:3 DAP + NA molar ratio).

General Remarks
The low propensity of DAP and both TMP and PMA to give rise to co-crystals with the isomeric pyridinecarboxamides, as predicted in the framework of COSMOquick, was confirmed. The DAP + NA exception is certainly a consequence of the greater simplicity of the DAP molecule, hence giving rise to fewer competing supramolecular association  The representative DSC heating curves of binary mixtures of 2,4-diaminopirimidine + theophylline, prepared by ethanol-assisted grinding, are presented in Figure 4. A complex thermal behavior is observed with evidence of three invariant points, T 1 = 137 • C, T 2 = 192 • C and T 3 = 218 • C, as indicated in the Figure. This behavior suggests that two co-crystals should be identified for this system. possibilities, and of the particular nicotinamide molecular structure. Concerning picolinamide, its lower co-crystal formation propensity when compared to their two isomers, with a different electronic structure in the ring, has been pointed out [45,46].
Thermogravimetric analysis ( Figure 6) indicates that the (1:2) co-crystal is a solvated one (see also Figures S21 and S22). The same X-ray powder diffractogram was obtained for a (1:2) DAP + THEO sample grinded with ethyl acetate assistance. The mass loss of about 1.5 % observed in the TG curve is compatible with a (1:2:0.5) DAP:THEO:H 2 O multicomponent solid.    The results obtained for xDAP = 0.50 are the same when neat grinding is performed ( Figure S23, ESI). However, neat grinding experiments carried out on (1:2) DAP + THEO mixtures give rise, as expected, to a solid form different from that obtained by LAG. This form is anhydrous ( Figure 6) and has characteristic reflections in the X-ray powder diffractogram at 2 θ = 7.4, 8.2, 13.6, 14.2, 25.1 and 27.0° (Figure 7). Interestingly, it gives rise,     The results obtained for xDAP = 0.50 are the same when neat grinding is performed ( Figure S23, ESI). However, neat grinding experiments carried out on (1:2) DAP + THEO mixtures give rise, as expected, to a solid form different from that obtained by LAG. This form is anhydrous ( Figure 6) and has characteristic reflections in the X-ray powder diffractogram at 2 θ = 7.4, 8.2, 13.6, 14.2, 25.1 and 27.0° (Figure 7). Interestingly, it gives rise, upon heating, to the same solid form obtained after the dehydration of the LAG-obtained The results obtained for x DAP = 0.50 are the same when neat grinding is performed ( Figure S23, ESI). However, neat grinding experiments carried out on (1:2) DAP + THEO mixtures give rise, as expected, to a solid form different from that obtained by LAG.

Hydrate
Concerning DAP + CAF, in the experimental conditions used in the current work, cocrystal formation was not successful. This is evident from the binary (solid + liquid) phase A tentative (solid + liquid) binary phase diagram presented in Figure 8 may be proposed where two incongruently melting co-crystals are identified at (1:2) and   In conclusion, for the DAP + THEO system, a (1:1) co-crystal was identified, as well as two polymorphic (1:2) co-crystals and a (1:2) DAP:THEO co-crystal hydrate.
Concerning DAP + CAF, in the experimental conditions used in the current work, cocrystal formation was not successful. This is evident from the binary (solid + liquid) phase In conclusion, for the DAP + THEO system, a (1:1) co-crystal was identified, as well as two polymorphic (1:2) co-crystals and a (1:2) DAP:THEO co-crystal hydrate.
Concerning DAP + CAF, in the experimental conditions used in the current work, co-crystal formation was not successful. This is evident from the binary (solid + liquid) phase diagram ( Figure S24; representative DSC curves shown in Figure S25) and from FTIR spectra ( Figure S26).

Trimethoprim + theophylline and Trimethoprim + caffeine
The DSC curves shown in Figure 9a and the proposed (solid + liquid) binary phase diagram ( Figure 9b) point to (1:2) TMP:THEO co-crystal formation, which is confirmed by X-ray diffraction and FTIR investigation ( Figure 10 and Figure S27, respectively). In these Figures, the excesses of the pure components are indicated by arrows.  Molecules 2021, 26, x FOR PEER REVIEW 9 of 22 diagram ( Figure S24; representative DSC curves shown in Figure S25) and from FTIR spectra ( Figure S26).

Trimethoprim + theophylline and trimethoprim + caffeine
The DSC curves shown in Figure     The crystalline structure of a (1:2) TMP:THEO (17% H2O) was solved (CCDC 2109486) as described in Section 3.7 and discussed in Section 2.2.3. The simulated X-ray powder diffractogram for this co-crystal matches the experimental one of the co-crystal obtained by grinding ( Figure 12).  In the experimental conditions used in the current work, as it was also observed for DAP + CAF, co-crystal formation between trimethoprim and caffeine was not successful (see Figures S28 to S30). The crystalline structure of a (1:2) TMP:THEO (17% H 2 O) was solved (CCDC 2109486) as described in Section 3.7 and discussed in Section 2.2.3. The simulated X-ray powder diffractogram for this co-crystal matches the experimental one of the co-crystal obtained by grinding ( Figure 12). The crystalline structure of a (1:2) TMP:THEO (17% H2O) was solved (CCDC 2109486) as described in Section 3.7 and discussed in Section 2.2.3. The simulated X-ray powder diffractogram for this co-crystal matches the experimental one of the co-crystal obtained by grinding ( Figure 12).  In the experimental conditions used in the current work, as it was also observed for DAP + CAF, co-crystal formation between trimethoprim and caffeine was not successful (see Figures S28 to S30). In the experimental conditions used in the current work, as it was also observed for DAP + CAF, co-crystal formation between trimethoprim and caffeine was not successful (see Figures S28-S30).

Crystalline Structures and Hirshfeld Surfaces Analysis
A comparison of the propensity for co-crystal formation with a common set of coformers was performed in the preceding sections for the DAP scaffold and the derived dihydrofolate reductase inhibitors, trimethoprim and pyrimethamine. A comparison of the intermolecular association in crystals of DAP, TMP and PMA with the same set of co-formers is, naturally, also of interest.
Several attempts, using different experimental approaches, were tried in order to obtain single crystals of the multicomponent solids identified in this work that are suitable for crystal structure resolution. This is a hard task, with unpredictable successful results. Single crystals of the (1:2) TMP:THEO co-crystal could be obtained, as described in Section 3.7, and its crystalline structure solved. Additionally, when searching for DAP:THEO co-crystals, single crystals of a new (1:2:1) DAP:THEO:EtAc solvate were obtained. Crystallographic data for both structures are presented in Section 3.7. A comparison of the intermolecular association in these two new crystalline structures, with the same co-former, is presented in this section. Comparison is also made with the (1:1) PMA:THEO co-crystal structure solved by Delori et al. [28].
The ORTEP diagram, with the numbering scheme for (1:2:1) DAP:THEO:EtAc solvate, is shown in Figure 13. The unit cell and illustrative images of the crystalline arrangement and of the intermolecular hydrogen bonds are shown in Figure 14 and in Table 2.
the intermolecular association in crystals of DAP, TMP and PMA with the same set o formers is, naturally, also of interest.
Several attempts, using different experimental approaches, were tried in order to tain single crystals of the multicomponent solids identified in this work that are suit for crystal structure resolution. This is a hard task, with unpredictable successful res Single crystals of the (1:2) TMP:THEO co-crystal could be obtained, as described in Sec 3.7, and its crystalline structure solved. Additionally, when searching for DAP:THEO crystals, single crystals of a new (1:2:1) DAP:THEO:EtAc solvate were obtained. Cry lographic data for both structures are presented in section 3.7. A comparison of the i molecular association in these two new crystalline structures, with the same co-form presented in this section. Comparison is also made with the (1:1) PMA:THEO co-cr structure solved by Delori et al. [28].
The ORTEP diagram, with the numbering scheme for (1:2:1) DAP:THEO:EtAc s ate, is shown in Figure 13. The unit cell and illustrative images of the crystalline arra ment and of the intermolecular hydrogen bonds are shown in Figure 14 and in Table   Figure 13. ORTEP diagram of the (1:2:1) 2,4-diaminopyrimidine:theophylline:ethyl acetate, CC 2108308. Ellipsoids are drawn at the 50% probability level. Hydrogen atoms are drawn as sphe of arbitrary radius.   Figure 14.
The molecules crystallize in a monoclinic crystal structure, under the common P2 1 /c space group. The 2,4-diaminopyrimidine and the theophylline molecules are joined together in ribbons, with each 2,4-diaminopyrimidine molecule H-bonded to four theophylline molecules (Figure 14b and Table 2). The planar ribbons containing the DAP molecules are interconnected by skewed THEO molecules (an angle of 47.1 • is observed between the DAP ring plane and that of these theophylline molecules). The ribbons run along the (100) direction, forming channels that are occupied by the ethyl acetate solvent molecules. Concerning the solvent molecules, very weak contacts [47], only with theo- The close contacts involving the DAP molecule in the co-crystal are highlighted in its Hirshfeld surface, shown in Figure 15a, and consist of six H-bonds. In four of these, the DAP NH 2 groups act as donors to different THEO O=C groups, and in the other two, each N atom in the ring acts as acceptor from the NH groups of two THEO molecules.
The four NH···O interactions are expressed in the left spike of the fingerprint plot in Figure 15b. The tip of the spike corresponds to the stronger H bonds, towards the approximately co-planar THEO molecules in the above-mentioned ribbons. The two N···HN contacts, identified as the right spike in the plot, follow the same trend, with stronger H bonding involving the ribbon THEO molecules. It is worth mentioning the green area close to d i ≈ d e = 1.8 Å correspond to heavy atom contacts due to DAP···THEO π-π stacking interactions [48].
Single crystals of a (1:2) trimethoprim theophylline (partially solvated) co-crystal were obtained as described in Section 3.7. The ORTEP diagram, with the numbering scheme, is shown in Figure 16 and in the unit cell, and illustrative images of the crystalline arrangement and of the intermolecular hydrogen bonds are shown in Figure 17 and in Table 3.
The molecules of trimethoprim and theophylline are joined in chains by H-bonds. One of the chains is assembled by the trimethoprim and one of the two theophylline independent molecules (Figure 17b). The other theophylline independent molecules form a different chain (Figure 17c). In this planar chain, the theophylline molecules are joined by N5B-H5B···N6B hydrogen bonds, as in pure theophylline polymorph II, CCDC128707 [49], although, in pure THEO, the molecules are not in the same plane ( Figure S31). The chains pack parallel to each other (see Figure 17a; structure viewed along the b axis). An oxygen atom (water molecule) is located close to one of the amine groups of trimethoprim, but with just a 17% occupation, accounting for a weak solvation. It is possible that the 17% of occupation found in the refinement of the data coming from a single data collection is just a "snapshot" of a continuous series of possible solvent/compound ratios. There is evidence that this non-stoichiometric solvate can desolvate to the two-component phase without changing the main structural characteristics, since powder diffractograms collected before and after forced hydration do not change ( Figure S32). It is even possible that the water molecules escape the lattice at room temperature due to their small size and weak interactions within the crystal main structure.
There is only a small change in trimethoprim conformation in the co-crystal with an angle of 80 • between the planes of the two rings, slightly greater than in pure TMP crystals, 70 • (CCDC607118 [50]). In the co-crystal, as in pure trimethoprim [50], the oxygen (O2) of the middle methoxy group is hydrogen-bonded to a NH group of a different TMP molecule, N4H4 in the co-crystal and N3H3 in pure TMP. As in the (1:2:1) DAP:THEO:EtAc structure, no NH···N hydrogen bonds involving the aminopyrimidine scaffold are observed between TMP molecules. Hirshfeld fingerprint plot of the DAP molecule in the (1:2:1) DAP:THEO:EtAc co-crystal solvate. Some important pairs of contacting atoms are identified and assigned to fingerprint plot features, with dashed arrows for "in-plane" (ipTHEO) and solid arrows for "out-of-plane" (opTHEO) THEO molecules. Contacts within the (9) synthon between the DAP scaffold and THEO, shown at the side, are colored in red. di -the distance (Å) from the Hirschfeld surface to the nearest atom center in its interior; de -distance (Å) from the Hirschfeld surface to the nearest atom center in its exterior.
Single crystals of a (1:2) trimethoprim theophylline (partially solvated) co-crystal were obtained as described in Section 3.7. The ORTEP diagram, with the numbering scheme, is shown in Figure 16 and in the unit cell, and illustrative images of the crystalline arrangement and of the intermolecular hydrogen bonds are shown in Figure 17 and in Table 3. (b) Hirshfeld fingerprint plot of the DAP molecule in the (1:2:1) DAP:THEO:EtAc co-crystal solvate. Some important pairs of contacting atoms are identified and assigned to fingerprint plot features, with dashed arrows for "in-plane" (ipTHEO) and solid arrows for "out-of-plane" (opTHEO) THEO molecules. Contacts within the R 2 2 (9) synthon between the DAP scaffold and THEO, shown at the side, are colored in red. d i -the distance (Å) from the Hirschfeld surface to the nearest atom center in its interior; d e -distance (Å) from the Hirschfeld surface to the nearest atom center in its exterior.      i: x, −1 + y, z; ii: −x, 1/2 + y, 1/2 − z.
The Hirshfeld surface of the TMP molecule in TMP:THEO, presented in Figure 18a and in the fingerprint plot (Figure 18b) show similar features to DAP in the previously mentioned co-crystal. The evidence of different donating and accepting H bonds are manifested in the complex structure of their two corresponding spikes. In both co-crystals, the NH···N type between the NH group of THEO and one of the ring N atoms of the diaminopyrimidine scaffold is the strongest interaction-(d i ,d e )/Å ≈ (1.1,0.7)-despite carbonyl generally being a better acceptor of H bonding than aromatic ring nitrogen. It is worth mentioning that in both of the co-crystal structures solved in this work, the synthon that is specially highlighted in Figures 15 and 18 plays a major role. This synthon is also observed in the (1:1) PMA:THEO co-crystal structure solved by Delori et al. [28]. In this latter co-crystal, of a different stoichiometry, PMA homodimers are observed. The presence of π-π stacking is noticed in both Hirshfeld surfaces around d i ≈ d e = 1.8 Å.

General Remarks
The co-crystal screening resulted in the successful synthesis of multicomponent solid forms of all three target molecules with theophylline. DAP co-crystals revealed a greater variety of stoichiometries (1:1, 1:2 DAP:THEO) and even polymorphism in the case of the 1:2 stoichiometry. Likewise, theophylline was a viable co-former with TMP, however with a single form of 1:2 TMP:THEO stoichiometry. This concurs with previous findings, where

General Remarks
The co-crystal screening resulted in the successful synthesis of multicomponent solid forms of all three target molecules with theophylline. DAP co-crystals revealed a greater variety of stoichiometries (1:1, 1:2 DAP:THEO) and even polymorphism in the case of the 1:2 stoichiometry. Likewise, theophylline was a viable co-former with TMP, however with a single form of 1:2 TMP:THEO stoichiometry. This concurs with previous findings, where two co-crystal polymorphs were obtained for (1:1) PMA:THEO [27,28].
The crystalline structure of the (1:2) TMP:THEO was resolved in this work, as well as an ethyl acetate solvate of DAP:THEO in the same stoichiometry. Both structures lack hydrogen bonding between the aminopyrimidine scaffolds, and both share the common synthon described above in Figure 1j. and previously recognized in PMA:THEO [28]. This synthon, a R 2 2 (9) hydrogen bonding motif, identified in Figures 15 and 18, involves the NH group of theophylline as a hydrogen bonding donor. Contrary to theophylline, caffeine has no hydrogen-bonding donor groups, which could be a possible reason for its weaker co-former performance with the target compounds. It could not form co-crystals with either DAP or TMP, despite the previously reported successful synthesis of a polymorphic (1:1) co-crystal with PMA [27]. In both of these latter PMA co-crystals, the hydrogen bonds among pyrimethamine molecules involve all 2,4-diaminopyrimidine donor and acceptor groups in a ribbon pattern that is similar to that observed in pyrimethamine polymorph I [51].

Co-Crystals Screening
A preliminary co-crystal in silico screening was carried out using the BIOVIA COS-MOquick 2020 software (Vélizy-Villacoublay, France) [40,41]. COSMOquick relies on the COSMOfrag fragmentation approach which employs a database of previously computed σ-profiles [58] for a set of about 190,000 compounds and its estimation of thermodynamic properties, avoiding the use of costly quantum mechanical calculations, to implement its cocrystal screening capability. The propensity for cocrystal formation is estimated from the excess enthalpy of the virtually subcooled liquid mixture of the components. In order to account for the negative effect of conformational flexibility in cocrystal formation, a partial empirical function based on the number of rotatable torsions is introduced that punishes highly flexible compounds in a screening.
Mechanochemistry was chosen for the experimental co-crystals screening. A MM400 Retsch ball mill, with 10 mL stainless steel jars and two 7 mm diameter stainless balls per jar, was used. Typically, a total mass of~70 mg was used, and grinding was performed neat or with the assistance of 10 µL ethanol, at 15 or 30 Hz, for times ranging from 30 to 90 min.

Differential Scanning Calorimetry (DSC)
DSC experiments were carried out using a Perkin Elmer DSC7 (Norwalk, CT, USA) calorimeter, with an intracooler unit at -10 • C (ethylene glycol-water 1:1 v/v cooling mixture), and a 20 mL/min -1 nitrogen purge. Aluminum pans suitable for volatile substances (Perkin Elmer, 30 µL) were employed, with an empty pan as reference. Temperature and enthalpy calibrations were performed with certified reference materials, following the procedure described elsewhere [59]. Heating rates of 10 • C/min -1 were used. The DSC curves were analyzed with Perkin Elmer Pyris software, version 3.5.

Thermogravimetric Analysis (TGA)
A Netzsch TG 209 F3 (Selb, Germany) Tarsus equipment was employed using alumina pans. Sample masses m ≈ 10 mg were used, and temperature was scanned from 30 to 600 • C in a N 2 atmosphere (25 mL/min) at a β = 10 • C/min scanning rate. Temperature calibration was performed using indium, tin, bismuth, zinc, aluminum and silver standards. Mass and linearity of the balance was verified with standard masses of the order of milligrams. Data was analyzed by Netzsch Proteus Thermal Analysis, version 7.2.0. software.

Infrared Spectroscopy (FTIR)
Infrared spectra were obtained in a ThermoNicolet 380 (Thermo Scientific TM, CA, USA) Fourier transform infrared spectrometer, with a Smart Orbit Diamond Attenuated Total Reflection (ATR) system, 64 scans and a 2 cm −1 spectral resolution.
In the determination of crystal structures of monocrystals by X-ray diffraction analysis, a Bruker-Nonius Kappa Apex II (Bruker, Karlsruhe, DE) CCD diffractometer was used, using MoKα radiation (λ = 0.71073 Å). Structures were solved by direct methods and by conventional Fourier Synthesis (SHELXS-97) [61]. The refinement of the structures was performed by full-matrix least-squares on F 2 (SHELXL-97) [62]. Hydrogen atoms attached to carbon atoms were placed at calculated positions and refined as riding. Hydrogen atoms attached to nitrogen atoms were located in a difference Fourier synthesis and their coordinates were freely refined. For the (1:2) TMP:THEO co-crystal, at the final refinement stages, a peak in a difference Fourier map of 1.54 e/Å 3 persisted and an oxygen atom was tentatively added to the model, refining to near 17% occupancy that may correspond to a water molecule whose hydrogen atoms were not located.
A summary of the data collection and refinement details is given in Table 4.

Hirshfeld Surfaces Analysis
The CIF files were used as starting points to generate the Hirshfeld surfaces and fingerprints [63,64] using CrystalExplorer, version 17.5. (University of Western Australia), which was also used for their analysis. CrystalExplorer normalizes the C-H and N-H bonds as those obtained from neutron diffraction experiments (C-H 1.083 Å, N-H 1.009 Å). The surfaces were generated using a "very high" resolution.
In Table 5, a summary of the results of the co-crystal screening outcomes is presented. This work demonstrated the possible use of the pharmacophore of dihydrofolate reductase inhibitors as a guide for co-crystal screening, recognizing some similar trends in the outcome of association in the solid state, and in the molecular aggregation in the co-crystals characterized by the same supramolecular synthons. Nevertheless, the co-crystal outcome for the API will always be more or less influenced by the remaining molecular fragments and the result of a balance of all interactions in the crystal lattice environment.
The comparison of the experimental co-crystallization outcome with the prediction of the virtual co-crystal screening with COSMOquick confirms the low propensity of DAP, TMP and PMA to co-crystallize with pyridinecarboxamides, with a successful outcome only observed for DAP + nicotinamide. Co-crystallization attempts with theophylline were well predicted; while using caffeine as co-former, they were not successful for DAP and TMP, contrarily to what was predicted by the in silico method. This shows that the method is useful as a preliminary guide to co-crystal screening, although it is limited by misrepresentation of complete conformational space, explicit hydrogen bonding, crystal packing and kinetic factors.

DAP TMP PMA
PA of the virtual co-crystal screening with COSMOquick confirms the low propensity of DAP, TMP and PMA to co-crystallize with pyridinecarboxamides, with a successful outcome only observed for DAP + nicotinamide. Co-crystallization attempts with theophylline were well predicted; while using caffeine as co-former, they were not successful for DAP and TMP, contrarily to what was predicted by the in silico method. This shows that the method is useful as a preliminary guide to co-crystal screening, although it is limited by misrepresentation of complete conformational space, explicit hydrogen bonding, crystal packing and kinetic factors.   Figure S31: Representation of hydrogen-bonded chains in theophylline, polymorph II; Figure S32: Comparison of XRPD LAG (1:2) TMP:THEO co-crystal, LAG co-crystal exposed to a saturated water atmosphere for 15 days and simulated from CCDC2109486; Table S1: melting temperature and enthalpy of the compounds used in the current work. CIF files with the single crystal X-ray diffraction structures of (1:2)TMP:THEO (17% H2O), CCDC 2109486, and (1:2:1) DAP:THEO:EtAc, CCDC 2108308, are available from the Cambridge Crystallographic Data Center. of the virtual co-crystal screening with COSMOquick confirms the low pro TMP and PMA to co-crystallize with pyridinecarboxamides, with a suc only observed for DAP + nicotinamide. Co-crystallization attempts w were well predicted; while using caffeine as co-former, they were not suc and TMP, contrarily to what was predicted by the in silico method. Thi method is useful as a preliminary guide to co-crystal screening, although misrepresentation of complete conformational space, explicit hydrogen packing and kinetic factors.   Figure S31: Repres gen-bonded chains in theophylline, polymorph II; Figure S32: Comparison of TMP:THEO co-crystal, LAG co-crystal exposed to a saturated water atmospher simulated from CCDC2109486; Table S1: melting temperature and enthalpy of the in the current work. CIF files with the single crystal X-ray diffraction structures o (17% H2O), CCDC 2109486, and (1:2:1) DAP:THEO:EtAc, CCDC 2108308, are a Cambridge Crystallographic Data Center. of the virtual co-crystal screening with COSMOquic TMP and PMA to co-crystallize with pyridinecarb only observed for DAP + nicotinamide. Co-cryst were well predicted; while using caffeine as co-for and TMP, contrarily to what was predicted by the method is useful as a preliminary guide to co-crys misrepresentation of complete conformational spa packing and kinetic factors.   Figure S28: (Solid + liquid) binary phase di DSC heating curves and FTIR of binary mixtures TMP + gen-bonded chains in theophylline, polymorph II; Fig  TMP:THEO co-crystal, LAG co-crystal exposed to a sat simulated from CCDC2109486; Table S1: melting temper in the current work. CIF files with the single crystal X-ra (17% H2O), CCDC 2109486, and (1:2:1) DAP:THEO:EtA Cambridge Crystallographic Data Center.

INA
The comparison of the experimental co-crystallization outcome with the prediction of the virtual co-crystal screening with COSMOquick confirms the low propensity of DAP, TMP and PMA to co-crystallize with pyridinecarboxamides, with a successful outcome only observed for DAP + nicotinamide. Co-crystallization attempts with theophylline were well predicted; while using caffeine as co-former, they were not successful for DAP and TMP, contrarily to what was predicted by the in silico method. This shows that the method is useful as a preliminary guide to co-crystal screening, although it is limited by misrepresentation of complete conformational space, explicit hydrogen bonding, crystal packing and kinetic factors.   Figure S31: Representation of hydrogen-bonded chains in theophylline, polymorph II; Figure S32: Comparison of XRPD LAG (1:2) TMP:THEO co-crystal, LAG co-crystal exposed to a saturated water atmosphere for 15 days and simulated from CCDC2109486; Table S1: melting temperature and enthalpy of the compounds used in the current work. CIF files with the single crystal X-ray diffraction structures of (1:2)TMP:THEO (17% H2O), CCDC 2109486, and (1:2:1) DAP:THEO:EtAc, CCDC 2108308, are available from the Cambridge Crystallographic Data Center.
The comparison of the experimental co-crystallization outcome wit of the virtual co-crystal screening with COSMOquick confirms the low pro TMP and PMA to co-crystallize with pyridinecarboxamides, with a suc only observed for DAP + nicotinamide. Co-crystallization attempts w were well predicted; while using caffeine as co-former, they were not suc and TMP, contrarily to what was predicted by the in silico method. Thi method is useful as a preliminary guide to co-crystal screening, although misrepresentation of complete conformational space, explicit hydrogen packing and kinetic factors.   Figure S31: Repres gen-bonded chains in theophylline, polymorph II; Figure S32: Comparison of TMP:THEO co-crystal, LAG co-crystal exposed to a saturated water atmospher simulated from CCDC2109486; Table S1: melting temperature and enthalpy of the in the current work. CIF files with the single crystal X-ray diffraction structures o (17% H2O), CCDC 2109486, and (1:2:1) DAP:THEO:EtAc, CCDC 2108308, are a Cambridge Crystallographic Data Center.
The comparison of the experimental co-crysta of the virtual co-crystal screening with COSMOquic TMP and PMA to co-crystallize with pyridinecarb only observed for DAP + nicotinamide. Co-cryst were well predicted; while using caffeine as co-for and TMP, contrarily to what was predicted by the method is useful as a preliminary guide to co-crys misrepresentation of complete conformational spa packing and kinetic factors.   Figure S28: (Solid + liquid) binary phase di DSC heating curves and FTIR of binary mixtures TMP + gen-bonded chains in theophylline, polymorph II; Fig  TMP:THEO co-crystal, LAG co-crystal exposed to a sat simulated from CCDC2109486; Table S1: melting temper in the current work. CIF files with the single crystal X-ra (17% H2O), CCDC 2109486, and (1:2:1) DAP:THEO:EtA Cambridge Crystallographic Data Center.

NA
TMP and PMA to co-crystallize with pyridinecarboxamides, with a successful outcome only observed for DAP + nicotinamide. Co-crystallization attempts with theophylline were well predicted; while using caffeine as co-former, they were not successful for DAP and TMP, contrarily to what was predicted by the in silico method. This shows that the method is useful as a preliminary guide to co-crystal screening, although it is limited by misrepresentation of complete conformational space, explicit hydrogen bonding, crystal packing and kinetic factors.   Figure S31: Representation of hydrogen-bonded chains in theophylline, polymorph II; Figure S32: Comparison of XRPD LAG (1:2) TMP:THEO co-crystal, LAG co-crystal exposed to a saturated water atmosphere for 15 days and simulated from CCDC2109486; Table S1: melting temperature and enthalpy of the compounds used in the current work. CIF files with the single crystal X-ray diffraction structures of (1:2)TMP:THEO (17% H2O), CCDC 2109486, and (1:2:1) DAP:THEO:EtAc, CCDC 2108308, are available from the Cambridge Crystallographic Data Center. ments and the result of a balance of all interactions in the crystal lattice en

Co-Former
The comparison of the experimental co-crystallization outcome wit of the virtual co-crystal screening with COSMOquick confirms the low pro TMP and PMA to co-crystallize with pyridinecarboxamides, with a suc only observed for DAP + nicotinamide. Co-crystallization attempts w were well predicted; while using caffeine as co-former, they were not suc and TMP, contrarily to what was predicted by the in silico method. Thi method is useful as a preliminary guide to co-crystal screening, although misrepresentation of complete conformational space, explicit hydrogen packing and kinetic factors. Table 5. Summary of co-crystal screening results for DAP, TMP and PMA, with t PA, INA, NA, THEO and CAF.  Figure S31: Repres gen-bonded chains in theophylline, polymorph II; Figure S32: Comparison of TMP:THEO co-crystal, LAG co-crystal exposed to a saturated water atmospher simulated from CCDC2109486; Table S1: melting temperature and enthalpy of the in the current work. CIF files with the single crystal X-ray diffraction structures o (17% H2O), CCDC 2109486, and (1:2:1) DAP:THEO:EtAc, CCDC 2108308, are a Cambridge Crystallographic Data Center.

ments and the result of a balance of all interactions
The comparison of the experimental co-crysta of the virtual co-crystal screening with COSMOquic TMP and PMA to co-crystallize with pyridinecarb only observed for DAP + nicotinamide. Co-cryst were well predicted; while using caffeine as co-for and TMP, contrarily to what was predicted by the method is useful as a preliminary guide to co-crys misrepresentation of complete conformational spa packing and kinetic factors.   Figure S28: (Solid + liquid) binary phase di DSC heating curves and FTIR of binary mixtures TMP + gen-bonded chains in theophylline, polymorph II; Fig  TMP:THEO co-crystal, LAG co-crystal exposed to a sat simulated from CCDC2109486; Table S1: melting temper in the current work. CIF files with the single crystal X-ra (17% H2O), CCDC 2109486, and (1:2:1) DAP:THEO:EtA Cambridge Crystallographic Data Center.
THEO of the virtual co-crystal screening with COSMOquick confirms the low propensity of DAP, TMP and PMA to co-crystallize with pyridinecarboxamides, with a successful outcome only observed for DAP + nicotinamide. Co-crystallization attempts with theophylline were well predicted; while using caffeine as co-former, they were not successful for DAP and TMP, contrarily to what was predicted by the in silico method. This shows that the method is useful as a preliminary guide to co-crystal screening, although it is limited by misrepresentation of complete conformational space, explicit hydrogen bonding, crystal packing and kinetic factors.   Figure S31: Representation of hydrogen-bonded chains in theophylline, polymorph II; Figure S32: Comparison of XRPD LAG (1:2) TMP:THEO co-crystal, LAG co-crystal exposed to a saturated water atmosphere for 15 days and simulated from CCDC2109486; Table S1: melting temperature and enthalpy of the compounds used in the current work. CIF files with the single crystal X-ray diffraction structures of (1:2)TMP:THEO (17% H2O), CCDC 2109486, and (1:2:1) DAP:THEO:EtAc, CCDC 2108308, are available from the Cambridge Crystallographic Data Center.
of the virtual co-crystal screening with COSMOquick confirms the low propensity of DAP, TMP and PMA to co-crystallize with pyridinecarboxamides, with a successful outcome only observed for DAP + nicotinamide. Co-crystallization attempts with theophylline were well predicted; while using caffeine as co-former, they were not successful for DAP and TMP, contrarily to what was predicted by the in silico method. This shows that the method is useful as a preliminary guide to co-crystal screening, although it is limited by misrepresentation of complete conformational space, explicit hydrogen bonding, crystal packing and kinetic factors.   Figure S31: Representation of hydrogen-bonded chains in theophylline, polymorph II; Figure S32: Comparison of XRPD LAG (1:2) TMP:THEO co-crystal, LAG co-crystal exposed to a saturated water atmosphere for 15 days and simulated from CCDC2109486; Table S1: melting temperature and enthalpy of the compounds used in the current work. CIF files with the single crystal X-ray diffraction structures of (1:2)TMP:THEO (17% H2O), CCDC 2109486, and (1:2:1) DAP:THEO:EtAc, CCDC 2108308, are available from the Cambridge Crystallographic Data Center.
of the virtual co-crystal screening with COSMOquick confirms the low prop TMP and PMA to co-crystallize with pyridinecarboxamides, with a succe only observed for DAP + nicotinamide. Co-crystallization attempts with were well predicted; while using caffeine as co-former, they were not succe and TMP, contrarily to what was predicted by the in silico method. This s method is useful as a preliminary guide to co-crystal screening, although i misrepresentation of complete conformational space, explicit hydrogen bo packing and kinetic factors.   Figure S31: Represen gen-bonded chains in theophylline, polymorph II; Figure S32: Comparison of X TMP:THEO co-crystal, LAG co-crystal exposed to a saturated water atmosphere simulated from CCDC2109486; Table S1: melting temperature and enthalpy of the c in the current work. CIF files with the single crystal X-ray diffraction structures of ( (17% H2O), CCDC 2109486, and (1:2:1) DAP:THEO:EtAc, CCDC 2108308, are ava Cambridge Crystallographic Data Center. (27,28) CAF tals characterized by the same supramolecular synthons. Nevertheless, the co-crystal outcome for the API will always be more or less influenced by the remaining molecular fragments and the result of a balance of all interactions in the crystal lattice environment.
The comparison of the experimental co-crystallization outcome with the prediction of the virtual co-crystal screening with COSMOquick confirms the low propensity of DAP, TMP and PMA to co-crystallize with pyridinecarboxamides, with a successful outcome only observed for DAP + nicotinamide. Co-crystallization attempts with theophylline were well predicted; while using caffeine as co-former, they were not successful for DAP and TMP, contrarily to what was predicted by the in silico method. This shows that the method is useful as a preliminary guide to co-crystal screening, although it is limited by misrepresentation of complete conformational space, explicit hydrogen bonding, crystal packing and kinetic factors.   Figure S31: Representation of hydrogen-bonded chains in theophylline, polymorph II; Figure S32: Comparison of XRPD LAG (1:2) TMP:THEO co-crystal, LAG co-crystal exposed to a saturated water atmosphere for 15 days and simulated from CCDC2109486; Table S1: melting temperature and enthalpy of the compounds used in the current work. CIF files with the single crystal X-ray diffraction structures of (1:2)TMP:THEO (17% H2O), CCDC 2109486, and (1:2:1) DAP:THEO:EtAc, CCDC 2108308, are available from the Cambridge Crystallographic Data Center. tals characterized by the same supramolecular synthons. Nevertheless, th come for the API will always be more or less influenced by the remaining ments and the result of a balance of all interactions in the crystal lattice en The comparison of the experimental co-crystallization outcome wit of the virtual co-crystal screening with COSMOquick confirms the low pro TMP and PMA to co-crystallize with pyridinecarboxamides, with a suc only observed for DAP + nicotinamide. Co-crystallization attempts w were well predicted; while using caffeine as co-former, they were not suc and TMP, contrarily to what was predicted by the in silico method. Thi method is useful as a preliminary guide to co-crystal screening, although misrepresentation of complete conformational space, explicit hydrogen packing and kinetic factors.   Figure S31: Repres gen-bonded chains in theophylline, polymorph II; Figure S32: Comparison of TMP:THEO co-crystal, LAG co-crystal exposed to a saturated water atmospher simulated from CCDC2109486; Table S1: melting temperature and enthalpy of the in the current work. CIF files with the single crystal X-ray diffraction structures o (17% H2O), CCDC 2109486, and (1:2:1) DAP:THEO:EtAc, CCDC 2108308, are a Cambridge Crystallographic Data Center.
The comparison of the experimental co-crystallization outcome with of the virtual co-crystal screening with COSMOquick confirms the low pro TMP and PMA to co-crystallize with pyridinecarboxamides, with a succ only observed for DAP + nicotinamide. Co-crystallization attempts wi were well predicted; while using caffeine as co-former, they were not suc and TMP, contrarily to what was predicted by the in silico method. This method is useful as a preliminary guide to co-crystal screening, although misrepresentation of complete conformational space, explicit hydrogen b packing and kinetic factors.   Figure S31: Represe gen-bonded chains in theophylline, polymorph II; Figure S32: Comparison of TMP:THEO co-crystal, LAG co-crystal exposed to a saturated water atmosphere simulated from CCDC2109486; Table S1: melting temperature and enthalpy of the in the current work. CIF files with the single crystal X-ray diffraction structures of (17% H2O), CCDC 2109486, and (1:2:1) DAP:THEO:EtAc, CCDC 2108308, are a Cambridge Crystallographic Data Center. (27) which was also used for their analysis. CrystalExplorer normalizes the C-H and N-H bonds as those obtained from neutron diffraction experiments (C-H 1.083 Å, N-H 1.009 Å). The surfaces were generated using a "very high" resolution.
In Table 5, a summary of the results of the co-crystal screening outcomes is presented. This work demonstrated the possible use of the pharmacophore of dihydrofolate reductase inhibitors as a guide for co-crystal screening, recognizing some similar trends in the outcome of association in the solid state, and in the molecular aggregation in the co-crystals characterized by the same supramolecular synthons. Nevertheless, the co-crystal outcome for the API will always be more or less influenced by the remaining molecular fragments and the result of a balance of all interactions in the crystal lattice environment.
The comparison of the experimental co-crystallization outcome with the prediction of the virtual co-crystal screening with COSMOquick confirms the low propensity of DAP, TMP and PMA to co-crystallize with pyridinecarboxamides, with a successful outcome only observed for DAP + nicotinamide. Co-crystallization attempts with theophylline were well predicted; while using caffeine as co-former, they were not successful for DAP and TMP, contrarily to what was predicted by the in silico method. This shows that the method is useful as a preliminary guide to co-crystal screening, although it is limited by misrepresentation of complete conformational space, explicit hydrogen bonding, crystal packing and kinetic factors. Table 5. Summary of co-crystal screening results for DAP, TMP and PMA, with the co-formers PA, INA, NA, THEO and CAF.  Figure S31: Representation of hydrogen-bonded chains in theophylline, polymorph II; Figure S32: Comparison of XRPD LAG (1:2) TMP:THEO co-crystal, LAG co-crystal exposed to a saturated water atmosphere for 15 days and simulated from CCDC2109486; Table S1: melting temperature and enthalpy of the compounds used in the current work. CIF files with the single crystal X-ray diffraction structures of (1:2)TMP:THEO (17% H2O), CCDC 2109486, and (1:2:1) DAP:THEO:EtAc, CCDC 2108308, are available from the Cambridge Crystallographic Data Center.

Co-Former
-no co-crystal; Å). The surfaces were generated using a "very high" resolution.
In Table 5, a summary of the results of the co-crystal screening outcomes is presented. This work demonstrated the possible use of the pharmacophore of dihydrofolate reductase inhibitors as a guide for co-crystal screening, recognizing some similar trends in the outcome of association in the solid state, and in the molecular aggregation in the co-crystals characterized by the same supramolecular synthons. Nevertheless, the co-crystal outcome for the API will always be more or less influenced by the remaining molecular fragments and the result of a balance of all interactions in the crystal lattice environment.
The comparison of the experimental co-crystallization outcome with the prediction of the virtual co-crystal screening with COSMOquick confirms the low propensity of DAP, TMP and PMA to co-crystallize with pyridinecarboxamides, with a successful outcome only observed for DAP + nicotinamide. Co-crystallization attempts with theophylline were well predicted; while using caffeine as co-former, they were not successful for DAP and TMP, contrarily to what was predicted by the in silico method. This shows that the method is useful as a preliminary guide to co-crystal screening, although it is limited by misrepresentation of complete conformational space, explicit hydrogen bonding, crystal packing and kinetic factors.  Figure S31: Representation of hydrogen-bonded chains in theophylline, polymorph II; Figure S32: Comparison of XRPD LAG (1:2) TMP:THEO co-crystal, LAG co-crystal exposed to a saturated water atmosphere for 15 days and simulated from CCDC2109486; Table S1: melting temperature and enthalpy of the compounds used in the current work. CIF files with the single crystal X-ray diffraction structures of (1:2)TMP:THEO (17% H2O), CCDC 2109486, and (1:2:1) DAP:THEO:EtAc, CCDC 2108308, are available from the Cambridge Crystallographic Data Center.  Figure S22: DSC heating curves of xDAP = 0.33, LAG and NG, submitted to different thermal treatments; Figure S23; XRPD of (1:1) DAP:THEO prepared by LAG and by NG; Figure S24: (Solid + liquid) binary phase diagram of DAP + CAF; FigureS25 and S26: DSC heating curves and FTIR of binary mixtures DAP + CAF; Figure S27: FTIR of binary mixtures TMP + THEO; Figure S28: (Solid + liquid) binary phase diagram of TMP + CAF; Figures S29 and S30: DSC heating curves and FTIR of binary mixtures TMP + CAF; Figure S31: Representation of hydrogen-bonded chains in theophylline, polymorph II; Figure S32: Comparison of XRPD LAG (1:2) TMP:THEO co-crystal, LAG co-crystal exposed to a saturated water atmosphere for 15 days and simulated from CCDC2109486; Table S1: melting temperature and enthalpy of the compounds used in the current work. CIF files with the single crystal X-ray diffraction structures of (1:2)TMP:THEO (17% H2O), CCDC 2109486, and (1:2:1) DAP:THEO:EtAc, CCDC 2108308, are available from the Cambridge Crystallographic Data Center.