First‐Generation Bispidine Chelators for 213BiIII Radiopharmaceutical Applications

Abstract Hepta‐ and octadentate bispidines (3,7‐diazabicyclo[3.3.1]nonane, diaza‐adamantane) with acetate, methyl‐pyridine, and methyl‐picolinate pendant groups at the amine donors of the bispidine platform have been prepared and used to investigate BiIII coordination chemistry. Crystal structure and solution spectroscopic data (NMR spectroscopy and mass spectrometry) confirm that the rigid and relatively large bispidine cavity with an axially distorted geometry is well suited for BiIII and in all cases forms nine‐coordinate complexes; this is supported by an established hole size and shape analysis. It follows that nonadentate bispidines probably will be more suited as bifunctional chelators for 213BiIII‐based radiopharmaceuticals. However, two isomeric picolinate‐/acetate‐based heptadentate ligands already show very efficient complexation kinetics with 213BiIII at ambient temperature and kinetic stability that is comparable with the standard ligands used in this field. The experimentally determined hydrophilicities (log D 7.4 values) show that the BiIII complexes reported are relatively hydrophilic and well suited for medicinal applications. We also present a very efficient and relatively accurate method to compute charge distributions and hydrophilicities, and this will help to further optimize the systems reported here.


Introduction
Interestingly, only a few years after von Hevesy's establishment of the radiotracer principle, [1] one of the early medicinal applications and the first clinical studies with radiotracers were using bismuth-214. [2] Stable chelation of bismuth(III) radioisotopes is also of importance for actinium-225 radiotherapyactinium-225 is a promising nuclide for targeted α-therapy with daughter isotopes bismuth-211 and -213. Bismuth-213 with a half-life of 46 min may be obtained from 225 Ac/ 213 Bi generators. [3] As Bi III radiochemistry has not yet been widely established, that is, only few tailor-made Bi III chelators have been reported, [4][5][6][7][8][9][10][11][12] there is interest in further work in this area.
As it is a borderline metal ion in terms of the hard-soft acidbase principle (HSAB), [13] ligands for stable Bi III complexes generally feature a combination of amine and oxygen (often carboxylate) donors, and coordination numbers from five to ten have been reported. For six-coordinate complexes, the ionic radius is 1.03 Å, for eight-coordinate it is 1.18 Å, with Bi III -N and Bi III -O distances of complexes with polyaminocarboxylate ligands around 2.40-2.65 Å and 2.25-2.60 Å, respectively. [8,9,[14][15][16] Bispidine ligands are an attractive platform in medicinal chemistry, and general applications in broad areas of coordination chemistry as well as their use in radiopharmaceutical applications have been reviewed extensively. [17][18][19][20] The advantages of the very rigid bispidine scaffold are: i) Ligands with a wide range of coordination numbers (4 to 8) and donor sets (mainly N and O) are accessible by relatively simple synthetic routes. 4-to 8-dentate bispidines with a variety of O and N donors have been reported, and others are possible and will be developed. [18,19,21] ii) There are relatively simple protocols for the preparation of bioconjugates. Various linkers have been described, and 64 Cu II tracers with bispidines conjugated to peptides or antibodies have been reported. [22][23][24][25] iii) There are specific bispidine ligands for a range of radiometal ions, combining high complex stability and inertness with relatively fast complexation kinetics. [18,21,26,27] This is due to the rigid diazaadamantyl backbone with two highly preorganized tertiary amines and two pendant pyridine donors as well as the possibility to attach rigid multidentate pendant groups such as picolinic acids at the tertiary amines to fully encapsulate specific metal ions in their preferred coordination geometry. [18,28] Here, we present the synthesis of three new heptadentate bispidines, together with their fully characterized Bi III compounds and radiolabeling experiments with 213 Bi III of two of the heptadentate and one previously known octadentate bispidine. The structural data together with the radiolabeling efficiencies [

Ligand synthesis
The new ligands L 1 , L 2 and L 3 were prepared with a methodology similar to that used for L 4 , [21] for L 2 this is shown in Scheme 1 (corresponding schemes for all ligands are given as Supporting Information). For L 1 and L 2 , the bispidine platform was built up as pentadentate ligands with three pendant pyridine groups and a secondary amine (N7 for L 1 , N3 for L 2 ), which was then alkylated with a picolinic acid precursor. For L 3 and L 4 , both pendant donors (picolinic acid and acetate) had to be introduced to N3 and/or N7 by alkylation, that is, the corresponding bispidines with secondary amines had first to be prepared. For the symmetrical octadentate bispidine L 4 this could be done in one step, [21] while for L 3 , an additional protection/deprotection step was necessary. The overall yields of the five-or six-step procedures are acceptable to good, and all four ligands were isolated as white powders and fully characterized after recrystallization. Crystal structures of the metal-free ligands and some intermediates have been deposited (see the Supporting Information) or published. [21] Bispidine-bismuth(III) complexes.  Figure 1, and selected bond distances and angles are given in Table 1.
The bispidine cavity is known to be very rigid and, with the basic hexadentate scaffold with four donors attached to C2, C4, N3 and N7, the hexadentate bispidine leads to an elastic coordination sphere, [17,18,28,29] where larger metal ions prefer relatively high coordination numbers. [18,21,30] This is also predicted from hole size and shape calculations of hexa-and octadentate bispidines using a molecular mechanics-based approach. [21,31,32] The ligand-based strain energy, the "energy penalty" imposed by the ligand, when the metal ion size  (average metal-donor distance) is smaller or larger than the optimum of about 2.5 Å is shown in Figure 2. [31,[33][34][35][36] Eightcoordinate Bi III has a nearly ideal size (see above) but, as the metal ion is not at the center of the cavity, it might require additional pendant donors to fully encapsulate the metal ion. In a first approach to evaluate the coordination geometry of bismuth(III)-bispidine complexes, we therefore prepared and characterized the structures of complexes with a series of hepta-and octadentate bispidines with mixed amine-pyridinecarboxylate donor sets.
It is unsurprising that the heptadentate ligands L 1 , L 2 and L 3 lead to unsaturated coordination spheres with Bi III that were completed by anions (bidentate nitrates, L 1 and L 3 ) or crystallize as a dinuclear structure (L 2 ) with bridging carboxylates. Interestingly, the structure of the octadentate bispidine L 4 indicates that the coordination sphere still is unsaturated with a monodentate nitrate coordinated to the complex. In solution, we anticipate that the anions are replaced by solvent molecules, and this is consistent with the solution NMR data (see the Supporting Information). We also assume that the dinuclear structure with L 2 adopts a mononuclear solution structure similar to those with L 1 and L 3 , and this also is consistent with the solution spectroscopy (see the Supporting Information). From the structural studies and solution spectroscopy it therefore emerges that nonadentate bispidines might provide an ideal coordination sphere for stable and inert Bi III complexation-coordination of decadentate bispidines might also be possible and, in terms of stability and inertness then would be preferred.
[Bi III (L 4 ')(NO 3 )] 2 (NO 3 ) 2 ( Figure 1 and Table 1) is a dinuclear complex with a partially hydrolyzed ligand, that is, the (aminomethyl)picolinic acid pendant group at N3 is hydrolyzed when the complexation reaction is performed in acidic solution. Structurally, the complex is similar to the others reported here,  (8) 129.03 (9) 129.61 (12) [a] The data for the coordinated nitrate are shown.

Figure 2.
Hole size (and shape) curve for L 4 (adapted from ref. [21]). The strain energy (MOMEC program and force field [33][34][35][36] ) is plotted as a function of the averaged metal-donor distances (MÀ D) av with the minimum of the curve set to 0 kJ mol À 1 . Apart from the metal-donor atomÀ ligand backbone angle deformation, the curve does not include any metal-ion-dependent terms. The variation of the metal-donor atom distances was asymmetric (i. e., the shape of the ligand and its variation were taken into account). The approximation adopted included full geometry optimization of the Zn II complex as a large metal ion and the Co III complex as a small metal ion and linear approximation between these structures to determine the relative changes in metal-donor atom distance for all eight bonds.
supporting the observation that the bispidine cavity is suitable for Bi III . Apart from the dinuclear structure with L 2 , all Bi III ions discussed here are nona-coordinate with N 6 O (L 1 , L 2 ), N 5 O 2 (L 3 ) or N 6 O 2 (L 4 ) donor sets of the bispidine and bi-or monodentate NO 3 À as co-ligand (except for the L 2 based dinuclear structure). The Bi III -O distances in general are approximately 0.1 Å shorter than the Bi III -N distances, and the average Bi III -donor lengths are slightly longer than 2.5 Å, that is, bispidines provide an ideal cavity for Bi III and, based on the structural data, nonadentate bispidines are predicted to be ideal ligands for 213 Bi III targeted therapy.

Bi III labeling
In previous radiolabeling studies, functionalized bispidines were found to be capable of rapidly incorporating various radionuclides into stable and inert chelates. Tetra-, penta-and, in particular, hexadentate ligands proved highly suitable for 64 Cu II , [23,25,37,38] whereas the octadentate L 4 and a similar oxinebased derivative were successfully matched to 111 In III , 177 Lu III and 225 Ac III . [21,26] Accordingly, L 1 , L 2 and L 4 show a higher labeling efficiency for 213 Bi III as compared to the current gold standards CHXÀ A''-DTPA and DOTA ( Figure 3). Also, the chelates were found to be quite stable in a transchelation challenge experiment with DTPA as competitor ( Figure 4). Whereas the transchelation velocity of the most inert complex with the octadentate ligand, 213 Bi-L 4 , is comparable to that of 213 Bi-CHXÀ A''-DTPA [5] (Figure S21), 213 Bi-L 1 and 213 Bi-L 2 , that is, the systems with heptadentate bispidines, are somewhat less inert (Figure 4, solid curves). This is because, once the metal ion is coordinated to the open and relatively large cavity of the bispidine platform, it is embraced by the donor side arms and rendered kinetically inert by encapsulation, [15] yet to a different extent for L 4 as compared to L 1 and L 2 because of the different overall efficiency of encapsulation by the ligands (see structures in Figure 1).
There is an intriguing difference in the observed resistance against transchelation for the chelate 213 Bi-L 4 , depending on whether the radiometal complexation was performed at room temperature (RT, ca. 25°C) or near boiling aqueous solutions (95°C). The dashed curves in Figure 4 indicate that the labeling product of L 4 at RT apparently comprises a very labile species, which is readily demetallated within the first minutes, and another complex that seems to be identical to the species obtained by labeling at 95°C because it is decomposing with the same velocity. Figure 5 shows that the fraction of the labile species decreases with reaction time at RT, indicating that it might be a rapidly forming encounter complex that slowly transforms into the final complex according to Figure 1. Although we can currently not provide any further details about the species involved, we hypothesize that the presence of two distant carboxylates in L 4 gives rise to a two-step complexation mechanism, similar to that observed for DOTA-type ligands. [39] First, the L 4 carboxylates replace for example two iodide ligands from the [BiI 5 ] 2À precursor obtained from the generator, resulting in a [BiI 3 L 4 ] xÀ species with some of the nitrogen donors of the bispidine still protonated. Then, a comparably slow rearrangement at RT and pH 5 delivers the same species obtained instantaneously at 95°C (Figure 1). No such observations are made for L 1 and L 2 , because a stable intermediate apparently cannot be formed with only one carboxylate.   1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57 Hydrophilicity For in-vivo applications, that is, for the biodistribution and excretion of the radiolabels, a relatively high hydrophilicity of the complexes and bioconjugates generally is of advantage. Usually, this is assessed by the distribution of the radiometal complexes between water and n-octanol at physiological pH (log D 7.4 ), and the earlier reported 64 Cu II -bispidine complexes are known to be hydrophilic and to have favorable biodistribution properties. [18,19,21,[22][23][24][25] The experimental data are given in Table 2, where computational data appear for comparison. The computed values are based on charge distributions calculated by a molecular mechanics based approach and a QSAR-type computation of the log D 7.4 values that has been described previously (modifications of the QSAR fit and parameter values used are given as Supporting Information). [40,41] Although this procedure is still in a developing stage (primarily also in terms of the scope), it allows to predict the hydrophilicities with an appreciable accuracy. The variation of hydrophilicities of isomeric species (e. g., complexes of ligands L 1 vs. L 2 , L 5 vs. L 6 ) is qualitatively correctly predicted. Two sets of computed log D 7.4 values are given in Table 2, the second one labeled all includes experimental data both from the previous study and from experimental data given here. Of some concern is that, in cases where the coordination number and sphere in solution might differ from that in the solid (i. e., in cases in which the coordination sphere in the solid is completed by a mono-or bidentate anion), the prediction is ambiguous. In these cases, modeling of the complex geometry by molecular mechanics or DFT-based approaches may need to precede the calculation of the log D 7.4 values (see the Supporting Information for computed values with various possible co-ligands and further discussion).

Conclusions
The four first-generation hepta-and octadentate bispidines prepared for efficient radiolabeling with 213 Bi III , and their ability to form thermodynamically stable and kinetically inert 213 Bi III complexes, are the first successful step toward novel agents for targeted radiotherapy with 213 Bi III . As for other bispidines used in radiopharmaceutical applications, these systems are well suited for efficient formation of various bioconjugates. [22][23][24][25] With three of the four ligands, radiolabeling is significantly faster than with the "gold standard" DOTA, and challenge experiments with an excess of a strong competing ligand (transchelation to DTPA) show that the bispidine complexes are only slightly less inert than those with DOTA. This is not unexpected due to the fact that only hepta-and octadentate bispidines were used so far, whereas all these Bi III complexes are shown to be nonacoordinate in the solid state and in solution. The structural work clearly shows that nona-or even decadentate bispidines are required to fully encapsulate Bi III , and these types of ligands with one or two carboxylate and two or three pyridine groups appended to the tertiary amines to lead to fully encapsulated mono-or dicationic Bi III complexes as well as derivatives with linkers to couple the ligands to biological vectors are relatively simple to prepare with bispidine scaffold. Structural modeling as well as the computational methods presented here to compute hydrophilicities will help to optimize these ligands.  . Two-dimensional correlation spectra (COSY, HSQC, HMBC) and 13 C-DEPT135 were used for signal assignment.

Mass spectrometry (MS). High-resolution (HR) mass spectra were recorded on a Bruker ApexQe hybrid 9.4 T FT-ICR instrument (ESI).
Reaction monitoring was conducted on a Waters Acquity + SQD2 UPLC-MS system. MALDI-TOF measurements were recorded on a Daltonic Autoflex II TOF/TOF instrument (Bruker) and 3-hydroxypicolinic acid (HPA) was used as matrix. Computational work. The charge distributions and log D 7.4 values as well as the hole size calculations were calculated with the MOMEC program and force field; [33][34][35][36]42] the methods for the computation of charge distributions and log D 7.4 values have been published before; [40,41] modifications and parameter sets are given as Supporting Information.
Supporting Information. NMR and electronic spectra, details of the labeling studies as well as experimental details of the solid state Xray analyses, including the crystallographic data tables, are given as Supporting Information.