Anticancer metallohelices: nanomolar potency and high selectivity† †Electronic supplementary information (ESI) available: Experimental details, syntheses, biophysical analyses, antimicrobial, anticancer, toxicity, and mechanistic studies. See DOI: 10.1039/c5sc03677a Click here for additional data fi

New optically pure helicate-like architectures are extremely active against cancer cell lines, with IC50 values as low as 40 nM, but nearly three orders of magnitude less active against healthy cells. There is also low toxicity to microbes and amoeba.


Introduction
The main purpose of current anticancer therapies is to eradicate tumour cells without damaging overall patient health. However, side effects limit the dosage of chemotherapeutic drugs which may be safely applied, and as a result, cancer cells oen remain. This leads to poor outcomes in the clinic and the evolution of drug-resistant tumours. 1 Hence, while the potency of a drug is a very important consideration, drug selectivity towards cancer cells is key to ensuring both safety and effectiveness. 2 While we might hope that more effective cancer chemotherapies would come from drugs designed to address specic biomolecular targets, 2 this is far from uniformly the case. 3 Such drugs may be too targeted since tumours can circumvent the blockade of a specic pathway by switching to anotherso-called tumour plasticity. 4 Compounds with polypharmacology (action against multiple targets) are thus currently of considerable interest to the pharmaceutical industry. This coincides with the resurgence of phenotypic drug discovery, [5][6][7] where the targets of a drug are established aer the observation of the useful biological effect. This strategy has led to a disproportionately high number of rst-in-class drugs with novel mechanisms of action (1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008) 8,9 The accompanying challenge for synthetic chemistry is to discover, perhaps without reference to some specic biomolecular target, new classes of drug candidates which are both potent and selective.
Lehn recognized the potential of helicates in medicinal chemistry, 10 and this was borne out in early studies, particularly in the area of cancer. [11][12][13][14][15] We have argued, 16 however, that in order for helicates to be capable of translation to the clinic a number of criteria need to be addressed: optical purity and stability, solubility and chemical stability in water, availability on a practical scale, and synthetic diversity. Our recent work has attempted to address these matters 17 using a new strategy whereby the absolute congurations of individual metal centres are controlled 18 and linked together to form the prototype helicate-like architectures of Fig. 1. Of these exicates, 19 [Fe 2 L 1 3 ] 4+ contains a diamine linker while [Fe 2 L 2a 3 ] 4+ is based on a dialdehyde. 19 Promising results were reported in a number of disease areas, 16,[19][20][21] including good activity against a range of cancer cell lines. 20 Here we report the discovery of a new series of highly potent (40 nM)

Synthesis of ligands and Zn II systems
The dialdehyde units of Fig. 2(a) and (b) include various linker rigidities and orientations designed to probe structural viability and biological activity. They were synthesized via simple etherications of 5-hydoxypicolinaldehyde. 22 Treatment with Zn(ClO 4 ) 2 $6H 2 O and (R)-1-phenylethan-1amine, in appropriate proportions, led to the rapid selfassembly of the bimetallic exicates in acetonitrile solution at ambient temperature. For the majority of these new Zn II complexes NMR spectra indicated that within the limits of the experiment single diastereomers were formed (vide infra).
The sole exception was the 1,3-phenylene system L Zn -[Zn 2 L 2e 3 ][ClO 4 ] 4 $4H 2 O, which gave more complex 1 H NMR spectra [ Fig. 2(c)]. At 253 K the phenethylamine methyl group doublet region 1.4-1.7 ppm contains one more intense doublet and two broader signals in the ratio ca. 10 : 1 : 1. The proportion of the minor species increases with temperature and the resonances sharpen somewhat, such that by 313 K two of the smaller doublets corresponding to the minor species are relatively sharp and resolved while a third overlaps with the main resonance. By 353 K the smaller peaks had again broadened considerably and the ratio of the two sets of resonances was ca. 10 : 9. The imine region (8.5-7.6 ppm) behaved in a corresponding manner (253 K, three peaks in ratio 10 : 1 : 1 : 1; 353 K, ratio 10 : 3 : 3 : 3). These observations are consistent with the presence of two speciesone of high-symmetry and one lowin thermodynamic equilibrium (ratio ca. 1 : 0.3 at low temperature, increasing to almost 1 : 1 at high temperature) but with the involvement of other related conformers particularly at higher temperatures. The processes leading to the observed NMR behaviour may correspond to exchange between these conformers, or indeed between isostructural low symmetry species. While the spectra are not sufficiently well resolved to determine kinetic parameters, we sought to investigate this molecular system by computational means.

Computational studies
Following extensive searching, six conformers of L Zn -[Zn 2 L 2e 3 ] 4+ were located and minimised [ Fig. 3]. These fell into two classes: those where the three m-xylenyl groups were oriented away from the central cavity i.e. exo, and those where one such group was oriented endo. No conformers were observed in which two or three m-xylenyl groups were oriented into the cavitythis caused too much torsional and steric strain. Structure endo1 was found to be the lowest in energy, the next lowest being endo2 (ca. +5 kcal) which differs only in the fold of one of the linkers. For these structures the Zn-Zn distances are ca. 11.7 and 11.8Å respectively. The structure exo1 (+7 kcal) has a large central cavity but a similar Zn-Zn distance (11.8Å). The structure exo2 (+8 kcal) has a considerably shorter Zn-Zn distance at ca. 9.5Å with accompanying concertinaed fold. Furthermore, higher energy conformers exo3 and exo4 differed principally in how the m-xylenyl groups folded towards each metal centre. Both were found to have a short Zn-Zn distance of 9.4 and 9.5Å respectively.
While prediction of an equilibrium population from the above calculations is complicated by statistical and entropic contributions from the total number of possible structures and the differences in structural exibility, the detection of two distinct structural classes is clearly consistent with observations in solution.
We propose that the species detected by NMR displaying high symmetry (D 3 ) comprises exo conformations while the asymmetric (C 1 ) species is endo. Examination of the structures indicates that the barrier to conversion within the exo or endo manifolds would be low since it would involve relatively simple concertina-type processes, but conversion between exo and endo conformations requires the rotation of the m-xylenyl linker through a strained, high energy transition state.

Synthesis of water soluble compounds
Pairs of water-soluble Fe II exicate enantiomers [Fe 2 L n 3 ]Cl 4 (n ¼ 2b-2e) were synthesised in high yield by heating the appropriate dialdehyde linker with either (R)-or (S)-1-phenylethan-1-amine and FeCl 2 in methanol. 1 H-NMR spectra were similar though slightly broader than the analogous Zn II perchlorate complexes and are consistent, along with 13 C-NMR [ Fig. 4 and ESI †] and circular dichroism spectra (ESI †) with the presence of single, stable, non-racemising diastereomers in solution, although unsurprisingly [Fe 2 L 2e 3 ]Cl 4 exists as a similar mixture of conformers to the Zn analogue above. The complexes gave excellent electrospray mass spectrometry data e.g. L Fe -[Fe 2 L 2c 3 ] Cl 4 gave a strong peak at m/z 420.17 Da for the tetracationic ion. The formula weights of the panel of complexes, including levels of hydration, were determined by correlation of NMR, IR, thermogravimetric and elemental analyses (see ESI †). The p-xylenyl system [Fe 2 L 2f 3 ]Cl 4 displayed poor solubility in water and methanol and could not be fully characterised.

Stability in aqueous media
Absorbance spectra indicated that little decomposition of the exicates occurred in water at pH 7 over weeks, but half-lives for decomposition could readily be recorded in hydrochloric acid   Even under such conditions, rst order kinetic plots gave t 1/2 values in the region 10-20 h. This very favourable aqueous stability of exicates probably arises from the presence of extensive (hydrophobic) p-stacking. 23 Biological activity & selectivity Cytotoxicity. The activities of the new compounds and cisplatin were investigated in human tumour cell lines: (a) MDA-MB-468 (human epithelial breast adenocarcinoma); 24 (b) HCT116 p53 +/+ and (c) HCT116 p53 À/À . 25 The HCT116 p53 +/+ and HCT116 p53 À/À cancer cells are human colorectal cancer cell lines that are genetically identical (isogenic) except for the presence or absence of functional p53. 25 These were chosen to enable screening of the effects of p53 status as the loss of p53 function is common genetic event in patient tumours and is strongly associated with increased resistance to many conventional chemotherapeutic agents. 25,26 In the cisplatin-sensitive 3 ]Cl 4 and D Fe -[Fe 2 L 2c 3 ]Cl 4 were both $9-fold more active against HCT116 p53 À/À cancer cells than their genetically identical p53 +/+ counterparts [ Fig. 5

(b) and(c); see SEI].
Toxicity in healthy human cells. The most active exicates in HCT116 p53 À/À cancer cells (2a and 2c), along with cisplatin, were investigated in human non-cancer retinal pigment epithelial cells (ARPE19) 27 and normal lung broblasts (WI38) [ Fig. 5(d)-(e)]. These are healthy human cells with a stable diploid karyotype which senesce aer multiple passaging as is characteristic of non-cancer cells. 27 In Fig. 5(f) we depict an in vitro selectivity index (SI) which compares the activity of these compounds in ARPE19 and HCT116 p53 À/À cells. While for cisplatin SI was found to be signicantly less than 1, meaning that it is actually more toxic to these healthy cells than it is to the cancer cells, the exicates tested gave SI substantially higher, and for D Fe -[Fe 2 L 2c 3 ]Cl 4 SI ¼ 836 AE 280. This excellent selectivity prompted us to investigate the toxicity of the compounds against a number of organisms.
Toxicity to microbes. The compounds were screened against cultures of the gram-positive bacterium methicillin-resistant Staphylococcus aureus, USA300 JE2 (ref. 28 and 29) (MRSA) and the gram-negative Escherichia coli, TOP10 (E. coli). 30 Kanamycin 30 was used as a positive control.
The new exicates had very modest antimicrobial activity (Table 1) or did not signicantly inhibit microbial growth at concentrations well over 3 orders of magnitude higher than the IC 50 values observed in cancer cells.
Toxicity in amoebae and M. sexta larvae. We further tested the potential toxicity of these compounds using a single cell protist organism, the well-established amoeba model 3 ]Cl 4 showed comparable weight gain to controls suggesting no oral toxicity and no adverse effect on feeding behaviour (Fig. 6). Interestingly the larvae exhibited an increased mean weight gain of approximately 30% (P < 1). Also we noted that larvae that ingested the exicate solutions turned a bright purple colour over the course of the assay, suggesting that these compounds were persisting in the insect and not being rapidly metabolized or excreted.
Systemic toxicity was further tested by injection of 50 mg (0.25 mM) of the compounds directly into the hemocoel of cohorts of 5 th instar larvae (n ¼ 3). The cohorts were then allowed to continue feeding. Despite becoming purple, all larvae proceeded to develop into the pupal diapause stage as per the buffer control injections.

Mode of action
The mode or modes of action of such a new and different system will require intensive investigation and is likely to involve multiple targets and pathways. Here, we describe two preliminary studies towards this end.
Denaturation of ct-DNA. We have previously concluded that the induction of DNA damage is not involved in the mode of action of earlier exicates, despite particular examples binding in a cell free environment. 20,32,33 We investigated the effect that the new exicates had on the denaturation temperature (T m ) of ct-DNA to screen for any indications of DNA binding.
Isolated ct-DNA (0.5 mg mL À1 ) was mixed with each exicate (7.5 mM) in buffered conditions (10 mM tris, 1 mM EDTA at pH 7.0), to give 10 bases: 1 exicate complex, and the absorbance at 260 nm between 25 C and 90 C was recorded (0.4 C min À1 ). T m for each experiment was calculated from the rst derivative of a Boltzmann sigmoidal t of the plot of absorbance versus temperature.
T m of untreated ct-DNA (0.25 mg mL À1 in 10 mM tris, 1 mM EDTA at pH 7.0) was measured to be 68.3 AE 0.5 C. Most of the new exicates had no signicant effect on the denaturation of ct-DNA (Fig. 7); the small (DT ca. 1 C) reduction for L 2e enantiomers can be ascribed to an electrostatic effect. 34 We are therefore satised that DNA is unlikely to be the target of this panel of compounds.
Induction of cell death by apoptosis. The chemosensitivity observed could be due to cytostatic or cytotoxic effects, and cell death can occur by several different mechanisms. These include programmed cell death by apoptosis, inammatory necrosis, autophagy or 'self-eating', necroptosis and pyroptosis. 35 One of the hallmarks of cancers is the evasion of apoptosis, thus enabling the long-term survival and proliferation of cancer cells. 36 We thus investigated whether the most active exicates are stimulating apoptotic death in cancer cells as part their mode of action.
HCT116 p53 +/+ cancer cells (24 h post-seeding) were incubated in fresh media containing exicate or no exicate (control) and were then analysed aer 48 h for levels of apoptosis and necrosis. As cells start to undergo apoptosis, one of the rst cellular changes is the externalisation of the membrane protein phosphatidylserine (PS). This can be detected and quantied by uorescently labelled annexin V 37,38 which can selectively bind externally exposed PS but is membraneimpermeable. This enables cells in the early stages of apoptosis to be distinguished from necrotic cells and cells in the late stages of apoptosis both of which have lost membrane integrity and will therefore also stain with the membrane-impermeable DNA stain propidium idodide. 39    (Fig. 8). A signicant proportion of late apoptotic/necrotic cells were also detectable by 48 h, with levels $2.3-2.5 fold above background control levels (Fig. 8). These preliminary investigations indicate induction of apoptosis by these new exicates as part of their mode of action.

Molecular modelling
Models of a number of possible conformers of L Zn -[Zn 2 L 2e 3 ] [ClO 4 ] 4 were constructed and optimised. Starting points for geometry optimisations were taken from crystallographic data. Monometallic structures were rst optimised using the B3LYP-D3(BJ) 40 functional and the 6-31g* basis set, with convergence criteria of 0.0001 a.u. as implemented in the Firey quantum chemistry package, 41 which is partially based on the GAME-SS(US) source code. 42 Bimetallic systems were optimised using ligand eld molecular mechanics (LFMM) 43 as implemented in the DommiMOE program, 44 before being annealed at 500 K for 1 ns prior to re-optimisation. Single point energy calculations of all structures were performed using the B3LYP-D3(BJ) 40 functional and the deff2-TZVP basis set with energy convergence criteria of 0.0001 a.u. as implemented in the Firey quantum chemistry package. 41 The calculations were conducted by employing the RIJCOSX approximation with SCF convergence criteria set to 'tight', both of which are dened internally as part of the ORCA DFT quantum chemistry package. 45 Where relevant, acetonitrile solvate correction was performed using the conductor-like screening model (COSMO) 46 as implemented in ORCA. 45 Biological activity MIC values were established using a macro broth dilution method in cation-adjusted Müller-Hinton (MH) broth. 96-well plates (200 mL of 128 mg mL À1 , 64 mM) complex in MH broth, diluted 2 n mg mL À1 , inoculated with each bacterial strain (10 3 cfu mL À1 ) were sealed and growth was monitored over 20 h at 37 C with an iEMS 96-well plate reader (see ESI †).
IC 50 values were determined by incubating cells in 96-well plates (2.0 Â 10 3 cells per well) for 24 h at 37 C, 5% CO 2 prior to drug exposure. Compounds were added (100 mM to 5 nM in cell medium) for a further 96 h. 3-(4,5-Dimethylthiazol-1-yl)-2,5diphenyl tetrazolium bromide solution (0.5 mg mL À1 , 20 mL per well) was added for a nal 4 h. Upon completion all solutions were aspirated, dimethyl sulfoxide (150 ml) was added and absorbance (540 nm) was recorded with a Thermo Scientic Multiskan EX microplate photometer. Oral toxicity was established by feeding cohorts of Manduca sexta 31 one-day-old neonate larvae with each exicate (25 mg mL À1 in articial wheat germ diet) for 7 d at 28 C and weighing to assess growth rate. Systemic toxicity assays 47 were conducted by injecting an ethanol (70% v/v) swabbed region of rst day h instar M. sexta larvae with each exicate (0.5 mg mL À1 [0.25 mM] in PBS), before allowing them to continue feeding for 7 d at 28 C, using physical stimulus to assess their status.

Mode of action
Denaturation of ct-DNA was measured by mixing ct-DNA (0.5 mg mL À1 , 7.5 Â 10 À5 per base, as determined by absorbance at 200 nm) with each complex (7.5 mM) in buffered conditions (10 mM tris, 1 mM EDTA at pH 7.0) to give 10 base: 1 complex. The absorbance at 260 nm as a function of temperature (every 1 C, 25-90 C) was measured in a 1 cm masked quartz cuvette at a rate of 0.4 C min À1 and run in triplicate. T m was calculated from the rst derivative of a Boltzmann sigmoidal t of the plot of absorbance at 260 nm against temperature for each complex.
Induction of apoptosis was determined by incubating HCT116 p53 +/+ cells (5 Â 10 5 cells/ask, 10 mL complete RPMI-1640 medium) for 24 h at 37 C in 5% CO 2 , before treating with each exicate (20 mM in fresh complete media for 48 h) or fresh media containing no drug (control). The supernatant containing any non-adhered, oating cells was then collected and pooled with cells harvested by trypsinisation. This pooled single cell suspension was washed twice with PBS and incubated with propidium iodide and Annexin-V-FLUOS (Roche) to stain apoptotic cells in accordance with the manufacturer's instructions. The proportion of early stage apoptotic cells and late stage apoptotic/necrotic cells were then quantied by ow cytometry as previously described. 37,38 Conclusions Our approach to metallohelix assembly has allowed us to generate a panel of biologically-compatible enantiomers incorporating various bridging groups. This was possible because in this so-called exicate platform the stereochemistry of the metal complex units is predetermined very efficiently and largely independently of the bridges, and by a mechanism that also provides water-compatibility. 23,48 In contrast, in a conventional "helication" approach the bridging units are structure-determining, so a mechanism of stereoselection would need to be designed for each example. A further advantage of the exicate platform is beginning to emerge in that we may be able to develop asymmetric molecules from symmetric ligands via the kinds of conformational abnormalities caused by bridges that partially oppose the predetermined stereochemistry e.g. L 2e . We have already shown that asymmetric (as opposed to merely chiral) optically pure assemblies are available using directional ligands. 49 Further, this modular self-assembling system will allow us to probe the effects of peripheral functionality and lipophilicity.
The activity of these new assemblies against cancer cells is strongly dependent on structure, with a range of potencies from 30 mM to as low as 40 nM. The most active compound D Fe -[Fe 2 L 2c 3 ]Cl 4 shows a selectivity index (versus healthy cell lines) approaching 10 3 , demonstrating superiority over the clinically used anticancer drug cisplatin in vitro (SI < 1). This selectivity is substantiated in tests with various models; bacteria and amoeba exposed to high concentrations were essentially unaffected, and in Manduca sexta larvae, where the systemic stability of the drug is evidenced, there is arguably a pro-biotic effect i.e. the insects appear to thrive.
In respect of mechanism or mode of action, the lack of binding to DNA indicates that this is unlikely to be the general target in this panel. In fact only one early exicate 19 ([Fe 2 L 1 3 ] 4+ , Fig. 1) in our growing library shows signicant interactions with nucleic acids, and while there are fascinating selectivities with various motifs 19,32,33 there is no DNA damage akin to that induced by e.g. platinum drugs and alkylators. 34,49 Instead, relevant examples of protein interaction and enzyme inhibition have been characterised. 21,32 To achieve drug safety and cancer selectivity, mechanistic classes which do not involve induction of DNA damage are attractive, and this may well be the source of the excellent selectivities we describe in this manuscript. Mode of action studies indicate that these compounds can induce substantial cell death by apoptosis independent of any DNA damage. Extensive studies are now required to understand how this complex process, normally subverted in cancers, is induced by these compounds. The above observations of remarkable selectivity alongside very high potency and large enantiomeric differences are however all consistent with a subtle mechanism involving the targeting of oncogenic drivers.