*2.2. Crystallography*

A colorless needle crystal of H2dap[Cd(HEDTA)(H2O)]·H2O (**1**) was mounted on a glass fiber and used for data collection. Crystal data were collected at 100(2) K, using a Bruker D8 VENTURE PHOTON III-14 di ffractometer. Graphite-monochromated MoK( α) radiation (λ = 0.71073 Å) was used throughout. The data were processed with APEX2 [33] and corrected for absorption using SADABS (transmissions factors: 1.000–0.962) [34]. The structure was solved by direct methods using the program SHELXS-2013 [35] and refined by full-matrix least-squares techniques against F<sup>2</sup> using SHELXL-2013 [35]. Positional and anisotropic atomic displacement parameters were refined for all non-hydrogen atoms. Hydrogen atoms were located in di fference maps and included as fixed contributions riding on attached atoms with isotropic thermal parameters 1.2/1.5 times those of their carrier atoms. Criteria of a satisfactory complete analysis were the ratios of 'rms' shift to standard deviation less than 0.001 and no significant features in final di fference maps. Atomic scattering factors were taken from the International Tables for Crystallography [36]. Molecular graphics were plotted with PLATON [37]. A summary of the crystal data, experimental details and refinement results are listed in Table 1. Crystallographic data for **1** has been deposited in the Cambridge Crystallographic Data Center with the CCDC number 1992206.

## *2.3. Other Physical Measurements*

Analytical data (CHN) were obtained in a Fisons–Carlo Erba EA 1108 elemental micro-analyzer. The cadmium content was cheeked as CdO by the weight of final residue in the thermogravimetric analysis (TGA) within 1% of assumed experimental error. FT-IR spectrum was recorded (KBr pellet) on a Jasco FT-IR 6300 spectrometer. TGA was carried out (r.t. to 950 ◦C) in air flow (100 mL/min) by a Shimadzu Thermobalance TGA–DTG–50H instrument and a series of 35 time-spaced FT-IR spectra of evolved gasses were recorded with a coupled FT-IR Nicolet Magna 550 spectrometer.

## *2.4. Synthesis and Relevant IR Spectrum Data*

Compound **1** was obtained in a two-step process. First, CdCO3 (1 mmol, 0.17 g) and H4EDTA (1 mmol, 0.29 g) were reacted in water (100 mL) inside an open Kitasato flask at 50–70 ◦C, with permanent stirring until a clear solution was observed. The heat was ceased and then small portions of Hdap (1 mmol, 0.15 g) were added to the Cd-H2EDTA chelate. The reaction mixture was filtered without vacuum (to remove any insoluble material) on a crystallization flask. The slow evaporation of the solution (two-three weeks at r.t.) produces needle crystals of **1**. Yield: ~70%. Elemental analysis (%): Calc. for C15 H24CdN8O10: C 30.60, H 4.11, N 19.03, Cd (as CdO) 21.81; Found: C 30.57, H 4.08, N 18.87, Cd (as CdO, final residue at 675 ◦C, in the TGA curve) 22.46. FT–IR data [cm−1]: 3500–3100 vbr <sup>ν</sup>as/νs(H2O) + <sup>ν</sup>as/νs(NH2) + νas(NH), 3411s, br, ν(OH), 2931w νas(CH2), 1674s, <sup>ν</sup>(C=O), 1596 vs δ(NH2) + δ(H2O) + νas(COO), 1400 m νs(COO), 923 w, 849 w π(C–H).


**Table 1.** Crystal data and structure refinement for H2(N3,N7)dap[Cd(HEDTA)(H2O)]·H2O.

## *2.5. Theoretical Methods*

All DFT calculations were carried out using the Gaussian-16 program [38] at the PBE1PBE-D3/ def2-TZVP level of theory and using the crystallographic coordinates. The formation energies of the assemblies have been evaluated by calculating the difference between the total energy of the assembly and the sum of the monomers that constitute the assembly, which have been maintained frozen. This methodology has been used by us [39,40] and others [41–45] to analyze supramolecular assemblies in crystal structures. The molecular electrostatic potential was computed at the same level of theory and plotted onto the 0.001 a.u. isosurface. The quantum theory of atoms-in-molecules (QTAIM) [46] analysis was carried out at the same level of theory by means of the AIMAll program [47]. The Cartesian coordinates of the theoretical models are given in the Supplementary Materials.

## **3. Results and Discussion**

## *3.1. Thermal Stability*

Under air-dry flow, the weight loss versus temperature TGA behavior consists of five steps (Figure 1). The experimental results and assignations are summarized in Table 2.

**Figure 1.** Weight loss versus temperature (in the range r.t. to 775 ◦C) in the thermogravimetric analysis of compound 1 (sample: 12.29 mg).


**Table 2.** Summary of the results and assignations in the thermogravimetric analysis of compound **1**.

 Calculated only for the loss of 2 H2O. t = trace amounts.

First of all, compound **1** overlaps the loss of uncoordinated water and aqua ligand content (with small amounts of CO2) in a consistent wide range of temperature (55–220 ◦C, experimental lost 6.056%, calculated for 2 H2O molecules 6.159%). In the second step (200–315 ◦C, with a weight loss of 12.071%) only CO2, CO and H2O were evolved, strongly suggesting that the combustion of organic ligands begins by the HEDTA3− chelator. Third and fourth steps (315–450 and 450–560 ◦C) produce (in addition to H2O, CO2 and CO) NH3 and N-oxides (N2O, NO and NO2) plus amounts of CH4. In the last fifth step (560–675 ◦C) the presence of CH4 and CO were less relevant. The weight loss during the burning steps (under an air flow) of organic material cannot be attributed to specific fragments of HEDTA3− or H2dap<sup>+</sup>. In contrast, the estimated residue (22.557% at 600 ◦C and 22.462% at 675 ◦C) reasonably agrees to the calculated weight for CdO (21.808%) within a reasonable experimental error (<1%).

## *3.2. Crystal Structure and Anion–Cation Recognition Pattern*

\*

This compound has an equimolar ratio of the tautomer H2(N3,N7)dap+ cation, the ternary anion [Cd(HEDTA)(H2O)]− and an unbounded to the metal aqua molecule (Figure 2). Table 3 shows the coordination bond distances and angles in the novel Cd(II) 'out-sphere' complex. Table 4 reports data concerning H-bonding interactions in its crystal. The first structural insight was that the assumed most basic N9 donor atom of Hdap diamino–purine in such a tautomeric form of the cation was unable to remove the aqua ligand from the seven coordinated Cd(II) chelate anion. The [Kr]4d<sup>10</sup> electronic configuration and the size of the Cd(II) center enables its rather common hepta-coordination as well as the inequality of its bond distances [2.267(1)–2.459(1) Å}. The Cd(II) coordination polyhedron in the chelate anions is best referred as a distorted mono-caped octahedron. The shortest bond is Cd-O(aqua) whereas the largest ones (<2.40 *A*) were Cd-N10<Cd-O(carboxyl)<Cd-N20. Interestingly the largest Cd-N20 bond involves de N20-HEDTA atom supporting the N-(carboxymethyl) arm of the chelating ligand. Table 3 summarizes the H-bonding interactions in compound **1**.

**Figure 2.** Asymmetric unit in the crystal of compound **1**, with relevant atom numbering scheme.

**Table 3.** Coordination bond lengths (Å) and angles (◦) in the crystal of compound **1**, H2(N3,N7) dap[Cd(HEDTA)(H2O)]·H2O. See Figure 1 for numbering scheme.


In the crystal, each anion is H-bonded to three independent neighboring cations, revealing that the anion–cation recognition of **1** is mainly featured by this kind of inter-molecular interaction (Figure 3). Deeping in this question, a H2dap+ cation links the complex anions by the H-bonds: N2-H2B···O24#1 (2.987(1) Å, 169.5◦), N3-H3···O23#1 (2.712(1) Å, 179,5◦) and (aqua)O1-H1WA···N9#1 (2.902(1) Å, 166.6◦) with #1 = −x + 1, −y + 1, −z + 1. This recognition pattern involves both O-acceptors of the same HEDTAcarboxylate group and the most basic N9 atom of the purinium(1+) ion. Another H2dap ion builds two H-bonds with O-carboxylate acceptors of the same HEDTA- carboxylate group: N6-H6B···O13#5 (2.831(1) Å, 1.67.7◦) and N7-H7···O14#5 (2.675(1) Å, 177.2◦) with #5 = x, −y, z. Figure 4 shows the way these two purinium(1+) cations were additionally related by a moderate anti-parallel <sup>π</sup>,<sup>π</sup>-stacking interaction between their 5- and 6-membered rings (inter-centroid distance dc-c 3.49 Å, interplanar distance d<sup>π</sup>-<sup>π</sup> 3.21 Å, dihedral interplanar angle 0◦, slipping angles β = γ = 25.91◦, slippage index 1.56). In this interaction the shortest interplanar distance would be related to the remarkable slippage. A third purinium(1+) ion is related with the chelate anion by the H-bond N6-H6A···O24 (2.990(1) Å, 144.6◦). Thus, O24 atom acts as twice-acceptor for H-bonding interactions.

**Table 4.** Geometric features of the hydrogen bonds in the crystal structure of H2(N3,N7)dap [Cd(HEDTA)(H2O)]·H2O (**1**). The distances were measured between the heavy atoms.


Symmetry transformations to generate equivalent atoms: #1 − x + 1, − y + 1, − z + 1, #2 x + 1, y, z, #3 − x + 1, − y + 2, − z, #4 x − 1, y, z, #5 x, y − 1, z, #6 x + 1, y − 1, z, #7 − x + 1, −y + 1, − z.

**Figure 3.** Molecular recognition pattern showing the cooperation of H-bonds between the [Cd(HEDTA)(H2O)]– chelate anion and three neighboring H2(N3,N7)dap+ ions.

In this compound all N–H and O–H bonds were involved in N-H···O or O-H···O interactions excepting for the above mentioned (aqua)O(1)-H(1WA)···N(9)#1 one (Table 3). In this manner the packing was essentially dominated by the H-bonding array that forms bilayers with Cd(II) chelate anions and unboned water molecules whereas H2(N3,N7)dap+ ions fall oriented towards both external surfaces. These 2D-frameworks lie parallel to the ab crystal plane and were H-bonded pillared along the c axis in the 3D-network (Figure 5).

**Figure 4.** Molecular recognition pattern showing the cooperation of H-bonds and <sup>π</sup>,<sup>π</sup>-stacking interactions connecting the [Cd(HEDTA)(H2O)]− chelate anion with two of the neighboring H2(N3,N7)dap+ ions.

**Figure 5.** In the 3D H-bonded crystal of compound **1**, 2,6-diaminopurinium(1+) cations oriented towards the external faces of 2D H-bonded frameworks build by complex anions and unbounded to the cadmium(II) water molecules. All H atoms and H-bonding interactions are omitted for clarity.

## *3.3. DFT Calculations*

The DFT study was focused to analyze the interesting supramolecular assemblies and H-bonding networks described above. First of all, the molecular electrostatic potential (MEP) surfaces of the anion and cation have been calculated in order to evaluate the best complementary dimer in terms of the electrostatic attraction between electron rich and electron poor regions of both molecules. The pure Coulombic attraction between the counter-ions is not directional; however weaker interaction like H-bonds or <sup>π</sup>,<sup>π</sup>-stacking interactions were able to nicely tune the final geometry of the supramolecular assembly. Evidence for the possibly structure-directing nature of these contacts was supported through an examination of MEP surfaces represented in Figure 6a. These reveal strong electropositive region (blue) at the NH groups of the Hdap+ cation and at the H-atoms of the Cd-coordinated water molecule. Moreover, the surfaces show excess of negative charge (red) at the O-atoms of the Cd-coordinated carboxylate group and at the N-atom of the five-membered ring of H2dap+ thus affording potentially favorable O-H···N and N-H···O interactions between the counter-ions. The MEP surface of Hdap+ also evidences that the N1-atom was less basic than N9, thus it was a worse H-bond acceptor. The MEP surface of the complex represented in Figure 6b shows how the charge density was significantly redistributed upon complexation.

**Figure 6.** (**a**) MEP surfaces (0.001 a.u.) for [Cd(HEDTA)(H2O)]− and H2dap+ highlighting the electropositive (blue) and electronegative (red) regions of each molecule. The dashed lines highlight a favorable electrostatic interaction between these two species. (**b**) MEP surface (0.001 a.u.) of the assembly at the PBE1PBE/def2-TZVP level of theory.

We have selected the supramolecular assembly commented above in Figure 7a–d to analyze the energetic features of the H-bonds and <sup>π</sup>,<sup>π</sup>-stacking interactions in **1**. Figure 7a shows a partial view of the solid state of **1** where these interactions are highlighted. From this quaternary assembly, we have first analyzed two H-bonded dimers (see Figure 7b,c), which present very large dimerization energies due to the strong contribution of the electrostatic attraction between counter-ions. Curiously the dimer with two H-bonds (Figure 7c) was stronger than that with three H-bonds (Figure 7b), likely due to the shorter H-bond distances. This aspect is further analyzed below. Regarding the <sup>π</sup>,<sup>π</sup>-stacked dimer, it presents a positive (repulsive) binding energy because it occurs against the Coulombic repulsion between both H2dap+ cations (ΔE3 = +44.8 kcal/mol). However, if the counter-ions were taken into consideration, the interaction becomes favorable, ΔE4 = −92.8 kcal/mol.

**Figure 7.** (**a**) Partial view of the X-ray solid state of **1** showing the self-assembled tetramer. (**b,c**) H-bonded dimers extracted from the assembly dimers. (**d**) Isolated <sup>π</sup>,<sup>π</sup>-stacked dimer of H2dap+ moieties. The distances are given in Å.

As commented above, the interaction energies were strongly dominated by the Coulombic attraction between the counter-ions and it was difficult to evaluate the real effect of the H-bonding interactions. In order to better analyze the H-bonding network, we have used the QTAIM method to estimate the contribution of each H-bond. The existence of a bond path (lines of maximum density) and bond critical point (CP) connecting two atoms is a universal indication of interaction [48]. The distribution of bond CPs and bond paths in the two H-bonded dimers of compound **1** are given in Figure 8. Each H-bond interaction was characterized by a bond CP (green sphere) and bond path interconnecting the H-atom to the N/O-atoms and confirming the interaction. The energy of each contact has been evaluated according to the approach suggested by Espinosa et al. [49] and Vener et al. [50]. The energy predictors were developed specifically for HBs and were based on the kinetic energy density (Vr) of the Lagrangian energy density (Gr). These values along with the charge density (ρr) are summarized in Table 5 for the CPs indicated in Figure 8. It can be observed that both energy predictors show that the N9-H···O H-bond (CP4) was the strongest one, even stronger than N3<sup>+</sup>-H3···O (the second strongest HB) that bears the positive charge, in line with the shortest distance (1.73 Å) and larger electron density (ρr) of CP4, see Table 5. The dissociation energies obtained for the other H-bonds were in the typical range of moderately strong H-bonds. There is an acceptable agreemen<sup>t</sup> between both energy predictors thus giving reliability to the study. It is worth mentioning that the sum of the dissociation energies of the two H-bonds of the dimer shown in Figure 8b (18.62 kcal/mol, using the Vr predictor) was larger than the sum of the three H-bonds in the dimer shown in Figure 8a (16.60 kcal/mol), in good agreemen<sup>t</sup> with the DFT energies computed for the assemblies shown in Figure 7a,b. This result confirms the fact that the H-bonds were stronger in the dimer where only two H-bonds were formed. Finally, it is interesting to note that the total energy density (Hr = Vr + Gr) was negative in CP4 thus evidencing partial covalent character for the N9-H9···O H-bond, in agreemen<sup>t</sup> with its large dissociation energy. The rest of CPs exhibit positive Hr values, evidencing their negligible covalent character.

**Figure 8.** (**a,b**). Distribution of bond and ring critical points (green and yellow spheres, respectively) and bond paths in two dimers of complex **1**. The QTAIM data at the bond CPs denoted as CP1 to CP5 are given in Table 3.

**Table 5.** Values of ρr, Vr and Gr (in a.u.) for CP1 to CP5 as indicated in Figure 7. The dissociation energy (Edis) of each H-bond based on Vr and Gr parameters are also indicated in kcal/mol.


*3.4. Structural Insides on N(heterocyclic)-Proton A*ffi*nities, H-Tautomerism and Metal Binding Patterns from Hdap and Its Cationic Forms in Salts and Their Metal Complexes.*

In a rather comprehensive review [51] we have look at the molecular recognition patterns between metal complexes and adenine or a variety of deaza- and aza-adenines (such as Hdap) on the basis of the cooperation between coordination bonds and intra-molecular interligand H-bonding interactions. This review emphasizes the relevance of the N(heterocyclic)-H tautomeric possibilities in neutral and protonated forms of such kinds of natural or synthetic closely related N-heterocyclic ligands. Recent reports from our groups extend these points of view to the guanine-synthetic acyclovir as a ligand [52–54]. Now we have the opportunity to deep into the relevance of these factors on the basis of the available crystallographic results related to cationic forms of Hdap, its salts and inner- or out-sphere metal complexes.

It is generally assumed that the proton affinity of hardly versatile ligand adenine (Hade) follows the order N9>N1>N7>N3>>N6(exocyclic amino) [51]. In a private communication to the CSD basis [55] the structure of the salt H3(N1,N7,N9)dapCl2·H2O (see reference code NULCOO in CSD Database) revealed the lesser proton affinity of the N3 atom of Hdap. That seems also agree with the depleted proton affinity found for the N3-atom of acyclovir, a well-known guanine-synthetic nucleoside [52]. The tautomers H2(N1,N3)dap<sup>+</sup>, H2(N3,N9)dap+ and H2(N7,N9)dap+ do not have received crystallographic support. The H2(N1,N7)dap+ tautomer acts as N9-donor ligand in two isomorphous compounds having all-trans octahedral complex molecules [MII(H2dap)2(hpt)2 (H2O)2]·4H2O (M = Co or Ni, htp = homophthalato(2-) ligand) [56] (see Scheme 1). The κN9-H2(N1,N7)dap<sup>+</sup> coordinating role was consistent not only with the highest proton affinity of the donor atom but also to its less steric hindrance. Interestingly in these complexes there was a cooperation of each M-N9 bond with an intra-molecular (aqua)O-H···N3 interligand interaction. The H2(N1,N9)dap+ tautomer was the counter cation of two rather distinct salts, with a dicarboxylate [57] or a dodecafluoro-*closo*-dodecaborate(2-) anion [58]. This tautomer binds metal ions by its N7-donor, in two Cd(II)-dicarboxylate coordination polymers [59,60] and a mononuclear Co(II) complex [61] displaying the appropriate cooperation between the metal-N7 coordination bond and an N6-H···O interligand interaction. This tautomeric

form was in agreemen<sup>t</sup> of the N-proton affinity assumed for the free base Hdap (N9>N1>N7>N3) which is also consistent by the crystal structure of H(N9)dap·H2O [59].

**Scheme 1.** Structure of [Cd(HEDTA)(H2O)]− and different tautomeric forms of Hdap+ with the atom numbering scheme.

The H2(N3,N7)dap+ tautomer, also here reported, is previously document in three rather distinct compounds. The out-sphere complex (H2(N3,N7)dap)2[Nd(<sup>μ</sup>2-croco) (croco)(H2O)4]2 (croco = croconate(2-) ion). This compound also builds a sophisticated H-bonded network, carefully describe by R. Baggio et al. [62] where any relevant <sup>π</sup>,<sup>π</sup>-stacking interactions appears precluded by coordination of the croco ligands. Why the H2(N3,N7)dap+ ions does not bind to Nd(III) centers can by explained on the basis of the Pearson's border-line basis of the Hdap and its cation whereas trivalent lanthanide cations were typical hard Pearson's acids. The two other compounds exhibit the κN9-H2(N3,N7)dap<sup>+</sup> ligand mode in presence of benzene-polycarboxylate anions. In the complex cation of trans-[CoII(H2O)4(H2(N3,N7)dap)2] (btec)·4H2O] (btec = bezene-1,2,4,5-tetracarboxylate) [63], aqua ligands cannot acts as H-acceptor for the N3-H bond of the H2(N3,N7)dap+ ions. Consequently, the Co-N9[H2(N3,N7)dap] coordination bond does not cooperate with an interligand N3-H···O interaction. In clear contrast the polymeric compound {[Zn(btc)(H2O) (H2(N3,N7)dap)]·4H2O}n (btc = benzene-1,2,3- tricarboxylate(3-) ion) exhibits the cooperation between the Zn-N9 coordination bond and an interligand (H2dap) N3-H···O(carboxy, btc) interaction (2.587(4) Å, 157◦) [59]. Curiously the O-carboxylate(btc) acceptor involved in such interligand H-bonding interaction implies an un-bonded to the Zn(II) O atom. This is certainly a relevant fact because of the common cooperation of metal-N(purine-like) bonds with (purine-like)N-H···O(carboxylate) intra-molecular interligand interactions was built with a metal-O(coordinated) H-acceptor atom [53].

## **4. Concluding Remarks**

In summary, the proton transfer between 2,6-diaminopurine and [Cd(H2EDTA)(H2O)] yields the outer sphere complex reported herein. The geometric features of the nucleobase in the solid state have been discussed in terms of binding pattern, protonation degree and proton tautomer as well as the hydrogen-bonding. Significantly, the solid-state structure was tuned by the synergistic formation of H-bonds and <sup>π</sup>+–π<sup>+</sup> interactions that have been described in detail. Moreover, the interaction energies of several supramolecular assemblies observed in the solid state have been evaluated and discussed by using MEP surfaces and DFT calculations. Finally, the individual H-bonding dissociation energies have been computed using two available energy predictors by means of the QTAIM method.

On the basis of our results and other above referred, it seems clear that the tautomerism plays a relevant role in the crystal having H2dap+ ions. The lack of literature concerning H2(N1,N3)dap<sup>+</sup>, H2(N3,N9)dap+ and H2(N7,N9)dap+ could be related to one of the following factors: The steric hindrance on N1, the depleted proton a ffinity of N3 and the suitability of the highest basic of N7 and N9 to metal binding. In the here reported compound, the use of the H2(N3,N7)dap+ favors the extensive H-bonding of its crystal, at the same time that precludes its coordination to the Cd(II) center instead of the aqua ligand.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4352/10/4/304/s1, Cartesian coordinates of the theoretical models shown in Figure 7.

**Author Contributions:** Conceptualization, J.N-G. and A.M.-H.; methodology, all authors; software, A.M.-H., A.F. and A.C.; investigation, J.C.B.-S. and N.R.-G.; writing—original draft preparation, A.M.-H., A.C., A.F. and J.N.-G., writing—review and editing, all authors, visualization, A.M.-H., A.C., A.F. project administration, J.N.-G.; funding acquisition, A.M.-H., A.F. and J.N.-G. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Excellence Network 'Metal Ions in Biological Systems' MetalBio CTQ2017-90802-REDT, the Research group FQM-283 (Junta de Andalucía) and MICIU/AEI of Spain (project CTQ2017-85821-R FEDER funds).

**Acknowledgments:** We thank the Centre de Tecnologies de la Informació (CTI), Universitat de les Illes Balears for computational facilities. We also thank all projects for financial support.

**Conflicts of Interest:** The authors declare no conflict of interest.
