Anticancer Activity and Cisplatin Binding Ability of Bis-Quinoline and Bis-Isoquinoline Derived [Pd2L4]4+ Metallosupramolecular Cages

New bis-quinoline (Lq) and bis-isoquinoline-based (Liq) ligands have been synthesized, along with their respective homoleptic [Pd2(Lq or Liq)4]4+ cages (Cq and Ciq). The ligands and cages were characterized by 1H, 13C and diffusion ordered (DOSY) NMR spectroscopies, high resolution electrospray ionization mass spectrometry (HR-ESIMS) and in the case of the bis-quinoline cage, X-ray crystallography. The crystal structure of the Cq architecture showed that the [Pd2(Lq)4]4+ cage formed a twisted meso isomer where the [Pd(quinoline)4]2+ units at either end of the cage architecture adopt the opposite twists (left and right handed). Conversely, Density Functional Theory (DFT) calculations on the Ciq cage architecture indicated that a lantern shaped conformation, similar to what has been observed before for related [Pd2(Ltripy)4]4+ systems (where Ltripy = 2,6-bis(pyridin-3-ylethynyl)pyridine), was generated. The different cage conformations manifest different properties for the isomeric cages. The Ciq cage is able to bind, weakly in acetonitrile, the anticancer drug cisplatin whereas the Cq architecture shows no interaction with the guest under the same conditions. The kinetic robustness of the two cages in the presence of Cl− nucleophiles was also different. The Ciq cage was completely decomposed into free Liq and [Pd(Cl)4]2− within 1 h. However, the Cq cage was more long lived and was only fully decomposed after 7 h. The new ligands (Liq and Lq) and the Pd(II) cage architectures (Ciq and Cq) were assessed for their cytotoxic properties against two cancerous cell lines (A549 lung cancer and MDA-MB-231 breast cancer) and one non-cancerous cell line (HDFa skin cells). It was found that Lq and Cq were both reasonably cytotoxic (IC50S ≈ 0.5 μM) against A549, while Ciq was slightly less active (IC50 = 7.4 μM). Liq was not soluble enough to allow the IC50 to be determined against either of the two cancerous cell lines. However, none of the molecules showed any selectivity for the cancer cells, as they were all found to have similar cytotoxicities against HDFa skin cells (IC50 values ranged from 2.6 to 3.0 μM).

The majority of the [Pd 2 (L) 4 ] 4+ cages examined to date feature pyridyl donors, as part of our efforts to improve the biological properties of these systems herein we describe the use of isoquinoline and quinoline-derived ligands for the assembly of two new [Pd 2 (L) 4 ] 4+ cages. While is well-known that isoquinoline and quinoline ligands can bind with palladium(II) and platinum(II) (Bondy et al., 2004) their use as donor systems in ligands for the generation MSAs has not been extensively explored (Bloch et al., 2016(Bloch et al., , 2017. These quinoline derived systems feature different electronic and steric properties compared to the parent pyridyl systems thus we also examine the effect these changes have on the host-guest chemistry with cisplatin, the stability of the cages in the presence of nucleophiles and the antiproliferative properties of the cages.

RESULTS AND DISCUSSION
The synthesis of the new quinoline (L q ) and isoquinoline (L iq ) based ligands was facile (Supplementary Material). Using sequential Sonogashira carbon-carbon cross coupling reactions from commercially available building blocks the ligands were generated in good yields (L q = 86% and L iq = 78%). 1 H and 13 C NMR spectroscopic data were consistent with the formation of the ligands which was supported by high-resolution electrospray mass spectrometry (HR-ESIMS) (Figure 1 and Supplementary Material). Peaks corresponding to protonated and sodiated ligand were observed at m/z = 382.1320 and 404.1132, respectively, for L iq and similar peaks were observed for L q (Supplementary Material).
With the ligands in hand, the complexation with [Pd(CH 3 CN) 4 ](BF 4 ) 2 in acetonitrile was examined (Figure 1). The cage formation was monitored using 1 H NMR spectroscopy (CD 3 CN, 298 K) and showed that mixing [Pd(CH 3 CN) 4 ](BF 4 ) 2 and the L iq ligand at room temperature (RT) in a 1:2 ratio led to the rapid (<2 min) and quantitative formation of the expected C iq cage (Figure 1), similar to what was observed with the parent L tripy system (Lewis et al., 2012). Interestingly, the behavior of L q with [Pd(CH 3 CN) 4 ](BF 4 ) 2 at room temperature was very different. After 5 min at RT the reaction mixture displayed multiple proton resonances, none of which were due to free ligand, consistent with the formation of a mixture of different cage isomers. The reaction was monitored using 1 H NMR spectroscopy for 24 h at RT however little to no changes were FIGURE 1 | (A) General scheme for formation of [Pd 2 (L) 4 ] 4+ cages (C q and C iq ) and partial 1 H NMR spectra (400 MHz, CD 3 CN, 298 K) of (B) L q and C q , and (C) L iq and C iq .
observed after the first hour and the spectrum still displayed multiple proton resonances. A 1 H DOSY experiment (CD 3 CN, 298 K) on the mixture showed that all the different proton resonances had the same diffusion co-efficient consistent with the postulate that the reaction mixture contains a series of cage isomers (Supplementary Material).
The assembly reaction between L q with [Pd(CH 3 CN) 4 ](BF 4 ) 2 was then carried out at 65 • C, in CD 3 CN and again monitored using 1 H NMR spectroscopy (Figure 2). After 5 min the same complicated series of proton resonance were observed. However, with continued heating this slowly resolved into a single series of resonances (after 7 h), consistent with the formation of a single cage isomer (Figure 2). Pleasingly, both cages (C iq and C q ) could be isolated by adding diethyl ether into the acetonitrile reaction mixtures providing the cages as colorless/tan precipitates in 88% (C iq ) or 92% (C q ) yield, respectively. 1 H NMR spectroscopy (CD 3 CN, 298 K) exhibited the expected downfield shifts of the signals pertaining to protons H a , H b and H f as well as the anticipated downfield shifts of the rest of the isoquinoline and quinoline protons resonances (Figure 1). HR-ESIMS data also supported the formation of the cages, showing ions corresponding to the loss of 2, 3 and 4 tetrafluoroborate (BF − 4 ) counterions (m/z = 956.1610 (2 + ), 608.4424 (3 + ), and 434.5832 (4 + ), Supplementary Material). 1 H DOSY experiments (CD 3 CN, 298 K) on the ligands (Diffusion coefficients (D) = 13.1 (L q ) and 15.0 (L iq ) x 10 −10 m 2 s −1 ) and cages (D = 4.1 (C q ) and 4.3 (C iq ) × 10 −10 m 2 s −1 were also consistent with the formation of the larger [Pd 2 (L) 4 ] 4+ cages (Supplementary Material).
Crystals of C q suitable for X-ray diffraction formed during the cooling of an acetonitrile solution of the cage from 65 • C to room temperature. The structure was solved in the tetragonal space group P4/mnc with the asymmetric unit containing one eighth of the cage and one quarter of a BF − 4 counterion (Figure 3 and Supplementary Material). The other BF − 4 anions and some acetonitrile molecules could not be modeled sensibly thus the SQUEEZE routine was employed to account for the diffuse electron density (Supplementary Material). The data revealed the expected [Pd 2 (L q ) 4 ] 4+ cage structure. The Pd-N bond lengths (Pd1-N2 2.045 Å) were similar to what have been previously observed for the related [Pd 2 (L tripy ) 4 ] 4+ cages where the Pd-N bond lengths range from 2.016 to 2.027 Å (Lewis et al., 2012;. The L q ligands of the cage are twisted giving a V-shaped conformation where the terminal quinoline and central pyridyl heterocyclic units are not co-planar which is quite different to what was observed with the [Pd 2 (L tripy ) 4 ] 4+ cages. In X-ray structures of the parent [Pd 2 (L tripy ) 4 ] 4+ cages the L tripy ligands were found in a linear conformation with the heterocyclic units coplanar. The twisting also alters the Pd1-Pd1 ′ distance within C q related to the [Pd 2 (L tripy ) 4 ] 4+ cages.
The Pd1-Pd1 ′ distances for the parent [Pd 2 (L tripy ) 4 ] 4+ cages range from 11.49 to 12.24 Å, whereas the Pd1-Pd1 ′ distance was found to be longer (12.506 Å) suggesting that the C q cage has a larger cavity despite featuring the same 2,6-diethynylpyridine linker units. The [Pd(quinoline) 4 ] 2+ units at the top and bottom of C q are twisted in opposite directions, the top cationic unit has a right handed twist while the bottom cationic unit has a left handed twist giving an overall meso structure (Figures 3B,C and Supplementary Material). Despite extensive efforts we were unable to obtain X-ray diffraction quality single crystals of C iq . Thus, to gain further insight into the structure of C iq we modeled the cage using Density Functional Theory (DFT) calculations (Figures 3D,E). Energy minimization of C iq (DFT, BP86 def2-SVP, acetonitrile solvation, Supplementary Material) showed that the cage adopted a lantern shape similar to what was previously observed for [Pd 2 (L tripy ) 4 ] 4+ cages (Lewis et al., 2012;. The calculated Pd -N bond distances (2.049 Å) and the Pd-Pd ′ distance (11.758 Å) match well with those observed crystallographically for the related [Pd 2 (L tripy ) 4 ] 4+ cages. The L iq ligand adopts a linear conformation with all the heterocyclic units coplanar. The DFT calculations indicated that the C iq is structurally very similar to the parent [Pd 2 (L tripy ) 4 ] 4+ cages whereas the C q is more twisted and provided a cavity of different size and shape to the parent cages and the C iq system. We and others have previously shown that other similar [Pd 2 (L tripy ) 4 ] 4+ cages can encapsulate cisplatin through hydrogen bonding interactions in CH 3 CN and DMF solvents (Lewis et al., 2012(Lewis et al., , 2013Kaiser et al., 2016;Preston et al., 2016Preston et al., , 2017Schmidt et al., 2016b). Therefore, we examined the ability of C iq and C q to interact with cisplatin in CH 3 CN using 1 H NMR spectroscopy. Addition of an excess of cisplatin to a CD 3 CN solution of the C iq cage resulted in a downfield shift and broadening ( δ = 0.03 ppm) of the internally directed cage proton H a (Figures 4A,B) indicative of cisplatin binding within the cage cavity, albeit weakly. A similar 1 H NMR experiment was carried out with the C q cage (Figures 4C,D). However, with the C q cage no shifts were observed for any of the cage proton resonances in the presence of an excess of cisplatin suggesting that the more twisted C q cage does not interact with the anticancer agent. The behavior was similar to what has been observed with a related twisted [Pd 2 (L 2Atripy ) 4 ] 4+ cage (where L 2Atripy = 2,6-bis[2-(6-amino-3-pyridinyl)ethynyl]-4pyridinemethanol) (Preston et al., 2016). The [Pd 2 (L 2Atripy ) 4 ] 4+ cage did not bind cisplatin in DMF solvent and the lack of binding was ascribed to the twisted cage cavity which was not as preorganised as those of the related lantern shaped [Pd 2 (L tripy ) 4 ] 4+ cages. Presumably the different sized cavity and different spatial arrangement of the hydrogen bond donors and acceptors caused by the twisting observed in the crystal structure of C q impedes the cisplatin-C q interaction in this case.
The kinetic robustness of the related [Pd 2 (L tripy ) 4 ] 4+ architectures in the presence of common biological nucleophiles (chloride (Cl − ), histidine and cysteine) has been determined using 1 H NMR competition experiments. When the parent [Pd 2 (L tripy ) 4 ] 4+ architectures were treated with 8 equivalents of tetrabutylammonium chloride the pyridyl substituted cages were rapidly and quantitatively decomposed (in <5 min). To examine the effect of substituting the pyridyl donor units for quinoline heterocycles time-course 1 H NMR competition experiments were carried out in d 6 -DMSO where 2 mM solutions of each cage (C q or C iq ) were treated with 8 equivalents of tetrabutylammonium chloride at 298 K (Figure 5 and Supplementary Material). Within 30 s of adding Cl − to the C iq cage, there were multiple species observed in the 1 H NMR spectrum. These were attributed to the C iq cage, [Cl⊂C iq ] 3+ , the [Pd 2 (L iq ) 2 Cl 4 ] macrocycle and free ligand based on our own previous results  and related literature. After 50 min, only uncoordinated ligand was visible in the 1 H NMR spectrum (Supplementary Material).
Under the same conditions, C q was stable for 1 h before showing signs of decomposition (Figure 5). After 3 h, there was no evidence of the C q cage, and the 1 H NMR spectrum displayed peaks corresponding to free ligand and a second metal-containing species, which based on the observed chemical shifts was most likely the neutral [Pd 2 (L q ) 2 Cl 4 ] macrocycle ( Figure 5H). This degradation behavior has been seen before with the [Pd 2 (L tripy ) 4 ] 4+ system in DMF . After 7 h, only free ligand could observed in the 1 H NMR spectrum indicating that all the ligand containing metal complexes had been completely decomposed into [Pd(Cl) 4 ] 2− (Figure 5J).
In comparison to the previously reported [Pd 2 (L tripy ) 4 ] 4+ cage (τ 1/2 = 2 min), the isoquinoline cage displayed an identical half-life (τ 1/2 = 2 min), whereas the quinoline system was considerably more robust (τ 1/2 = 2 h). Presumably the observed results reflect the different steric profiles of the two quinoline substituted cages (C q or C iq ). The C q cage has the quinoline moieties protecting the external face of the palladium, providing more impediment to nucleophilic attack from that face (Figure 6). The C iq does not feature the same steric impediment as the benzene units of the isoquinoline heterocycles do not block the top face of the C iq cage as much as they do in the quinoline C q (Figure 6).
Frontiers in Chemistry | www.frontiersin.org  To assess biological activity, the cytotoxic effect (as halfmaximal inhibitory concentrations (IC 50 )) of the ligands and cages were determined against three different cell lines: cisplatin resistant MDA-MB-231 (breast cancer) (Lehmann et al., 2011), A549 (lung cancer) and non-cancerous primary cells: adult human dermal fibroblasts (HDFa) ( Table 1 and Supplementary Material). The ligands L q and L iq exhibited limited solubility, and so data above the concentration of 1 µM was unattainable. Below this threshold, L iq displayed minimal cytotoxic activity against both cell lines, while L q was shown to be cytotoxic against A549 (IC 50 = 0.5 µM). Both cages were observed to be cytotoxic against the malignant cell lines, with C q showing the same level of toxicity as its ligand against lung cancer cells (IC 50 = 0.5 µM). C q was slightly less cytotoxic against MDA-MB-231 (IC 50 = 1.7 µM), whereas C iq was less cytotoxic than the quinoline analog, with the IC 50 values ranging from 4.0 to 7.4 µM against the cancer cells. Both quinoline cages were found to be considerably more active than the related parent [Pd 2 (L tripy ) 4 ] 4+ cage system (IC 50 = 41.4 and 56.7 µM against A549 and MDA-MB-231, respectively) (McNeill et al., 2015). The quinoline cages were also more active than cisplatin against the two cancer lines examined (cisplatin IC 50 values = 41.2 and 9.4 µM, against MDA-MB-231 and A549, respectively) McNeill et al., 2015). The quinoline cages C q and C iq were more cytotoxic than all the [Pd 2 (L tripy ) 4 ] 4+ cage systems reported in the literature (IC 50 values for the L tripy based systems ranged from 10 to 100 µM) (McNeill et al., 2015;Kaiser et al., 2016;Schmidt et al., 2016b). Additionally, C q was also more active, albeit against different cancer cell lines (HL-60, HL-60/Dox, , than the hydrophobic [Pd 2 (L anthracene ) 4 ] 4+ cages of Yoshizawa and Ahmedova (IC 50 values ranged from 0.9 to 37.4 µM) (Ahmedova et al., 2016;Anife et al., 2016).  (IC 50 ) of ligands L q and L iq , and cages C q and C iq architectures at 24 h.
C q was also more cytotoxic than a hydrophobic bis-hexyl-1,2,3-triazole substituted [Pd 2 (L hextrz ) 4 ] 4+ helicate, C hextrz , we developed previously (IC 50 values = 6.9 and 6.0 µM against A549 and MDA-MB-231, respectively) (McNeill et al., 2015). We presume that the favorable combination of high hydrophobicity and the kinetic robustness against biological nucleophiles leads to the higher observed activity of C q relative to the other [Pd 2 (L) 4 ] 4+ architectures. Disappointingly, neither of the cages (C q and C iq ) showed any selectivity for the cancer cells, they were all found to have similar cytotoxicity against HDFa skin cells (IC 50 values ranged from 2.6 to 3.0 µM). whereas the C q architecture shows no interaction with the guest under the same conditions. The kinetic robustness of the two cages in the presence of Cl − nucleophiles was also different. The C iq cage was completely decomposed into free L iq and [Pd(Cl) 4 ] 2− within 1 h. However, the C q cage was more long lived and was only fully decomposed after 7 h. The ligands (L iq and L q ) and cages (C iq and C q ) were assessed for their cytotoxic properties against two cancerous cell lines (A549 lung cancer cells and MDA-MB-231 breast cancer cells) and one non-cancerous cell line (HDFa skin cells). It was found that L q and C q were both reasonably cytotoxic against A549, while C iq was slightly less active. The higher observed cytotoxicity of C q relative to the other [Pd 2 (L) 4 ] 4+ architectures was presumed to be due the favorable combination of high hydrophobicity and the kinetic robustness against biological nucleophiles. However, none of the new molecules showed any selectivity for cancer cells, they were all found to have similar cytotoxicity against HDFa skin cells. A range of [Pd 2 (L) 4 ] 4+ cage systems have now been shown to be cytotoxic. However, in order to advance this class of MSA as anticancer agents more in depth mode of action/mechanistic studies on the origins of the cytotoxic activity are required. Studies to this effect are now underway.

AUTHOR CONTRIBUTIONS
RV and JC conceived the idea, analyzed the data and wrote the manuscript. RV and LG conducted the synthesis. RV and DP conducted stability studies. RV, DP, and JJ conducted cytotoxicity studies. GG oversaw the cytotoxicity studies and analyzed the data. All authors provided feedback on the manuscript drafts and approved the submission.