Enhanced kinetic stability of [Pd 2 L 4 ] 4+ cages through ligand substitution †

There is considerable interest in exploiting metallosupramolecular cages as drug delivery vectors. Recently, we developed a [Pd 2 L 4 ] 4+ cage capable of binding two molecules of cisplatin. Unfortunately, this ﬁ rst generation cage was rapidly decomposed by common biologically relevant nucleophiles. In an e ﬀ ort to improve the kinetic stability of these cage architectures here we report the synthesis of two amino substituted tripyridyl 2,6-bis(pyridin-3-ylethynyl)pyridine ( tripy ) ligands (with amino groups either in the 2-( 2A-tripy ) or 3-( 3A-tripy ) positions of the terminal pyridines) and their respective [Pd 2 ( L tripy ) 4 ] 4+ cages. These systems have been characterised by 1 H, 13 C and DOSY NMR spectroscopies, high resolution electrospray mass spectrometry, elemental analysis and, in one case, by X-ray crystallography. It was established, using model palladium( II ) N-heterocyclic carbene (NHC) probe complexes, that the amino substituted compounds were stronger donor ligands than the parent system ( 2A-tripy > 3A-tripy > tripy ). Competition experiments with a range of nucleophiles showed that these substitutions lead to more kinetically robust cage architectures, with [Pd 2 ( 2A-tripy ) 4 ] 4+ proving the most stable. Biological testing on the three ligands and cages against A549 and MDA-MB-231 cell lines showed that only [Pd 2 ( 2A-tripy ) 4 ] 4+ exhibited any appreciable cytotoxicity, with a modest IC 50 of 36.4 ± 1.9 µM against the MDA-MB-231 cell line. Unfortunately, the increase in kinetic stability of the [Pd 2 ( L tripy ) 4 ] 4+ cages was accompanied by loss of cisplatin-binding ability.


Introduction
Interest in self-assembled coordination complexes of welldefined two-and three-dimensional geometries, or metallosupramolecular architectures, 1 continues to grow due to their potential in a range of applications. The molecular recognition properties of these systems have been used to develop molecular reactions flasks, 2 catalysts 3 and drug delivery agents. 4 Systems have also been used to sequester reactive species, 5 and environmental pollutants. 6 Additionally, the biological, 7 photophysical, 8 electronic 9 and redox 10 properties of these metallosupramolecular architectures have been studied.
As part of our interest in the biological properties 11 of metallosupramolecular architectures we have previously reported the synthesis of a tripyridyl (tripy) [Pd 2 L 4 ] 4+ cage capable of binding cisplatin (cis-[Pt(NH 3 ) 2 Cl 2 ]). 11d We hoped to exploit these [Pd 2 L 4 ] 4+ cages as metallosupramolecular drug delivery vectors, unfortunately the cage architecture was rapidly decomposed when exposed to chloride (Cl − ), 11a,d,12 histidine (his) 11a and cysteine (cys) 11a (common nucleophiles in biological systems). In order to use these cisplatin binding [Pd 2 L 4 ] 4+ cages as drug delivery agents a method for increasing the kinetic stability of these metallosupramolecular architectures against nucleophiles was required. 13 There are two obvious approaches to enhance the kinetic stability of the cage architectures: (1) assemble the cage using more kinetically inert metal ions or (2) sterically and electronically tune the tripy ligand framework. Exploiting kinetically inert metals for the generation of metallosupramolecular architectures can be difficult due to the formation of kinetically "trapped" intermediate structures that cannot "error-correct" into the desired system. Therefore we chose to generate more electron-donating tripyridyl ligands in an effort to obtain more kinetically robust [Pd 2 L 4 ] 4+ cages. Herein we report the synthesis of two new amino substituted tripyridyl ligands (2A-tripy and 3A-tripy) and their respective cages (Scheme 1). The addition of the amino groups to the 2-and 3-positions of the terminal ligating pyridyl units of the ligand framework was expected to increase the donor properties of the ligands and lead to more kineti-compounds 11a,d,12 and suggest that the palladium(II) cage species are stable in solution (ESI †).
After considerable effort (>50 crystallisations and data collections) the solid state structure of the [Pd 2 (2A-tripy) 4 ](BF 4 ) 4 complex was obtained using X-ray crystallography ( Fig. 2 and ESI †). Small weakly diffracting X-ray quality crystals were generated by vapour diffusion of diethyl ether into a CH 3 CN solution of the [Pd 2 (2A-tripy) 4 ](BF 4 ) 4 cage. Although the weak diffraction was, at least in part, due to the presence of multiple disordered solvent molecules and counter anions within the crystal lattice (vide infra), the cationic framework of the cage was readily identified (Fig. 2).
The solid state structure of the [Pd 2 (2A-tripy) 4 ] 4+ cage confirms that the coordination of the 2-amino pyridyl units to the palladium(II) ions was monodentate through the pyridyl nitro-gen as expected. 15 Additionally, the 2-amino units of the ligands are engaged in intra-ligand hydrogen bonding interactions (N⋯N 3.78(3) Å, N-H⋯N 2.94 Å, Fig. 2b and ESI †) consistent with 1 H NMR data described above. In contrast to the [Pd 2 (tripy) 4 ] 4+ cage, 16 the [Pd 2 (2A-tripy) 4 ] 4+ cation adopts a more twisted structure in which the ligands of the 2A-tripy cages are significantly bent out of planarity and this appears to be caused by hydrogen bonding interactions between the amino groups on the exo faces of the architecture. This is quite different to what has been previously observed in the solid state structures of unsubstituted [Pd 2 (tripy) 4 ] 4+ cations, these cages without the 2-amino groups all were found to adopt a lantern shape, with essentially planar tripy ligands. 11a,d,12 The coordinated 2A-tripy ligand distorts in two ways: a swivelling of the coordinating pyridine rings relative to the principal rotation axis of the molecule (θ = 34.44°-34.64°, compared with θ = 3.47°-9.42°for [Pd 2 (tripy) 4 ] 4+ ) and a twisting of the central pyridine out of the plane through which the ligand coordinates to the two Pd(II) centres (φ = 35.10°compared with φ = 5.61°for [Pd 2 (tripy) 4 ] 4+ ). The cavity dimensions also differ (a Pd⋯Pd distance of 11.530(9)-11.610(9) Å compared with 11.201(1) Å for [Pd 2 (tripy) 4 ] 4+ , and a core-to-core pyridyl N⋯N distance of 10.711(9)-10.732(9) Å compared with 11.07 (1) Fig. 3 and ESI †) with the endo-pyridyl unit. These interactions generate 2D supramolecular sheets of cages through the solid state structure ( Fig. 3 and ESI †).

Competition experiments, cisplatin binding and cytotoxicity studies
The relative pK a values of 2-aminopyridine (6.82), 3-aminopyridine (6.04), and pyridine (5.23) indicate that 2-aminopyridine is the most basic ligand. 17 Evidence that 2-aminopyridine was also the strongest nucleophile was obtained using the palladium(II)-N-heterocyclic carbene (NHC) probe system developed by Huynh and coworkers (Table 2 and ESI †). 18 Consistent with the pK a values, the probe complexes indicated that 2-aminopyridine (161.2 ppm) is a stronger donor than 3-aminopyridine (159.8 ppm) which is a stronger donor than pyridine (159.3 ppm). The chemical shift observed for the 2-aminopyridine ligand is very similar to that previously reported for N-methylimidazole (161.1 ppm) 18b suggesting that the donor strength of these ligands are similar. To allow direct comparison to the literature pK a values (Table 2) and for synthetic convenience 11a we have examined the probe complexes of the simpler pyridine rather than the tripyridyl ligands. However, these pyridine model systems can serve as proxies for their respective tripyridyl ligands (2A-tripy, 3A-tripy, tripy) and provide indirect experimental evidence for the donor properties of the tripy ligands because the steric and electronic changes on going from the pyridine to tripy ligands are the same across the series. Thus the model complexes provide a qualitative ranking of the substituents' effects present in the tripy ligands and strongly suggest that the donor properties of the tripyridyl ligands follow the order 2A-tripy > 3A-tripy > tripy.
The kinetic stability of the [Pd 2 (L tripy ) 4 ] 4+ architectures in the presence of common biological nucleophiles (Cl − , his and cys) was determined using 1 H NMR competition experiments ( Table 2 and ESI †). Time-course 1 H NMR competition experiments were carried out in 3 : 2 d 6 -DMSO/D 2 O where 3 mM solutions of each cage were treated with 8 equivalents of tetramethylammonium chloride or 4 equivalents of his or cys. Under these conditions the unsubstituted [Pd 2 (tripy) 4 ] 4+ cage was rapidly decomposed by all the nucleophiles. The half-life for the decomposition of the [Pd 2 (tripy) 4 ] 4+ complex with his was 18 minutes. Despite the [Pd 2 (3A-tripy) 4 ] 4+ architecture containing the slightly more electron rich 3A-tripy ligand the cage was still quickly decomposed by each of the nucleophiles. However, the t 1/2 were subtly increased against all the nucleophiles (for his t 1/2 = 25 min) suggesting that the enhanced ligand donor properties of the 3A-tripy ligand does lead to increased cage stability relative to the unsubstituted system.
The [Pd 2 (2A-tripy) 4 ] 4+ cage displayed markedly higher stability against all the nucleophiles studied. The half-lives for the 2A-tripy cage decomposition against each nucleophile were all over 2 h, whereas the corresponding t 1/2 for the other cages were all less than 30 min. 2A-tripy is only a modestly stronger donor ligand than the 3A-tripy, and thus the observed large difference in stability is presumably not predominantly due to the increase donor ability of the ligand. A more important element is likely to be the presence of the 2-amino groups on the exo-faces of the [Pd 2 (2A-tripy) 4 ] 4+ cage which sterically protect the palladium(II) ions from the incoming nucleophiles. Additionally, as observed in the X-ray structure (Fig. 3), the hydrogen bonding interactions between the eight amino groups of the four 2A-tripy ligands may further enhance the stability of the [Pd 2 (2A-tripy) 4 ] 4+ cage relative to the other tripy architectures. However, against the stronger nucleophiles (Cl − and cys) the half-lives for the decomposition of the [Pd 2 (2Atripy) 4 ] 4+ are less than 3 h suggesting that these systems would need further tuning in order to be useful in a biological setting.

Cisplatin binding
We 11d,12,19 and others 16 11.53(9)-11.610(9) Å 11.201 (1) 10.71 (9)-10.73(9) Å 11.07(1)-11.26(1) Å a θ: the swivelling of the coordinating pyridine rings relative to the principal rotation axis of the molecule. b φ: twisting of the central pyridine out of the plane through which the ligand coordinates to the two Pd(II) centres.  16 The guest molecules are rotated 180°with respect to each other; hydrogen bonds between the guests and cage (N-H⋯N Py and Cl⋯H-C Py ) as well as a metal-metal interaction between the platinum atoms of the guests were observed (Fig. 3a). 11d,16b,19b A similar 1 H NMR experiment with cisplatin and [Pd 2 (3Atripy) 4 ] 4+ indicated that the 3-amino substituted cage is also able to bind cisplatin in solution, albeit more weakly (Δδ = 0.07 ppm for H c , Fig. 4c  CH 3 CN or DMF solvents displayed no shifts relative to the free cage indicating that the 2-amino substituted cage is not able to bind cisplatin. We have previously shown that cisplatin binding is very weak 12 and that subtle changes to the size, steric profile and electronic properties of the cage cavity 21 are enough to completely turn off cisplatin binding. Presumably, the lack of cisplatin binding in this system can be ascribed to two factors. Firstly, the presence of the eight amino groups on the exo faces of the cage has caused a twisting of the architecture (as indicated in the crystal structure, Fig. 3, and discussed and listed in Table 1). This twisting subtly alters both the size of the cisplatin binding cavity and the orientations of the hydrogen-bond acceptors and donors groups within the cage cavity, weakening the interaction between the host and the cisplatin guest. Secondly, the electron donating 2-amino units push electron density back onto the terminal pyridyl rings of the tripy ligand. This would reduce the polarisation of the acidic H c protons of the pyridyl unit, weakening the hydrogen bonding interaction with chloride ligands of the cisplatin guest. These effects, in concert, appear to be enough to fully circumscribe the already weak cisplatin-cage interaction. 22

Conclusions
Two amino substituted tripyridyl 2,6-bis( pyridin-3-ylethynyl) pyridine (tripy) ligands (with amino groups either in the 2-(2Atripy) or 3-(3A-tripy) positions of the terminal pyridines) and their respective [Pd 2 (L tripy ) 4 ] 4+ cages were synthesised. These systems have been characterised by 1 H, 13 C and DOSY NMR spectroscopies, high resolution electrospray mass spectrometry, elemental analysis and, in one case, by X-ray crystallography. It was established, using palladium(II) NHC carbene probe model complexes, that the amino substituted compounds were moderately stronger donor ligands than the parent pyridyl system (2A-tripy > 3A-tripy > tripy However, while the ligand tuning resulted in more robust [Pd 2 (L tripy ) 4 ] 4+ architectures the half-lives of the systems against the stronger nucleophiles were still modest (t 1/2 = 2-3 h). Furthermore, the subtle structural changes in the most stabilised cage, [Pd 2 (2A-tripy) 4 ] 4+ were found to completely destroy the ability of the system to bind cisplatin. Thus, it appears that in order to exploit these types of metallosupramolecular cage architectures as drug delivery vectors, systems assembled from more kinetically inert metals ions such as Pt(II) 4h and Ru(II) 4e-g and Co(III) 25 will be required. Efforts to generate more robust systems, composed of kinetically inert metal ions, capable of binding drug molecules are underway.
Synthesis of 2A-tripy. In a round bottom flask, diisopropylamine (20 mL) and THF (20 mL) were degassed with N 2 , before addition of 3 (400 mg, 3.39 mmol), (2,6-dibromopyridin-4-yl) methanol (362 mg, 1.35 mmol), CuI (25 mg, 0.14 mmol), and Pd(PPh 3 ) 2 Cl 2 (38 mg, 0.050 mmol) against a positive N 2 flow. The solution was heated at 50°C for 48 hours. After removal of the solvent under vacuum, the resultant solid was taken up in 3 : 1 CHCl 3 /IPA (150 mL) and aqueous 0.1 M EDTA/NH 4 OH solution (50 mL) and stirred for 40 minutes. The organic layer was washed with brine, dried with Na 2 SO 4 , filtered and then the solvent was removed under vacuum. Purification of the resultant solid on a silica column deactivated with 3 : 97 triethylamine/CH 2 Cl 2 (0.5/4.5/95 then 1/9/90 saturated aqueous NH 4 OH solution/MeOH/CH 2 Cl 2 ) gave the product as a brown solid (364 mg, 1.07 mmol, 79%). 1 H NMR (500 MHz, d 6 -DMSO, solvent lattice were severely disordered and could not be appropriately modelled. The SQUEEZE routine within PLATON was employed to resolve this problem, resulting in ten void spaces (total of 460 electrons), variously assigned to tetrafluoroborate anions (8 in total), H 2 O (5 in total) and MeCN (3 in total) solvent molecules (total of 444 electrons), as described below. Despite repeated efforts (>50 crystallisations and data collections over a two year period) to crystallise the compound, the most suitable candidate was small and a poor diffractor. The data quality is poor, with two A alerts (a large Hirshfield difference and high MainMol U eq. compared to neighbours) and many B alerts, and we emphasise that metric data cannot be reliably extracted from the structure and should be treated with caution. However, the connectivity of the cationic framework is readily apparent. The methylene alcohol substituents from the four neighbouring cages in the lattice interpenetrate the cavity of each cage ( Fig. 3 and ESI †), interpenetrating groups shown in spacefilling mode (yellow, green, blue and pink) and preclude cisplatin encapsulation in the solid state. Around the coordinating pyridine rings, the amino groups form a hydrogen bonding network. SQUEEZE details and crystallographic parameters can be found in the ESI. †