Cation-mediated optical resolution and anticancer activity of chiral polyoxometalates built from entirely achiral building blocks

Homochiral polyoxometalates were optically resolved via the counter cation-mediated breaking of chiral symmetry, and showed potent cytotoxicity against various cancer cells by activating apoptotic pathways.


Introduction
Chirality plays a critical role in biology and is also important for the functions of many materials. 1-3 In biological systems, homochirality is an essential feature of basic building blocks, such as amino acids and sugars, which are distinctly le-and right-handed. 4,5 Although the origin of homochirality is rmly established in biological systems, the spontaneous assembly of enantiopure materials from achiral precursors still represents a great challenge to chemists and material scientists. 6,7 Polyoxometalates (POMs), a class of metal-oxo clusters with oxygen-rich surfaces, have been extensively explored for applications in biology, magnetism, catalysis, and material science. 8 The introduction of chirality into POMs or POMorganic hybrid compounds can endow further functionality for potential applications in nonlinear optics, enantioselective catalysis, chiral medicine, and chiral magnetism. 9-14 Chiral ligands are typically used as structure-directing agents to facilitate the synthesis of chiral POM materials. 11-13 A more versatile approach towards such a synthesis would use completely achiral precursors to construct chiral POM skeletons via bond length alteration, structure distortion, formation of lacunae, replacement with other metals, or decorating with organic ligands to remove the inversion and/ or mirror symmetry in POMs. 9,14 However, bulk materials generally contain a racemic mixture of chiral POMs or POM materials. Achieving homochirality from totally achiral precursors is still a signicant challenge in the laboratory. 15 Among numerous potential applications, POMs have been reported to possess promising antibacterial, antiviral, and anticancer activities, which may open the way toward new and affordable therapeutic strategies for various human diseases. 16 Compound 1 crystallizes in the achiral monoclinic P2 1 /c space group, whereas compound 2 crystallizes in the cubic chiral space group P2 1 3. Single-crystal X-ray diffraction analyses revealed that compounds 1 and 2 both contain a screw propeller-like cluster {CoSb 6 O 4 (H 2 O) 3 [b-Co(hmta)SbW 8 O 31 ] 3 } 15À , which is built from a trimeric capsule-type polyoxotungstate and three hmta ligands ( Fig. 1 and S1b †  S2 and S3 †). In the POM family, trivacant Keggin tungstoantimonate {SbW 9 O 33 } is a common structural motif that is used to construct other POM compounds including dimeric systems and 1-D and 2-D polymeric structures. 19 The pentavacant {SbW 7 } unit is also observed in the cryptate polyoxoanion {Sb 9 W 21 }. 20 Fig. 1 and S1b †), which exhibits a 3-fold rotational axis corresponding to the C 3 symmetry. In the crystallization process, the triangular cluster compounds form highly transparent and well-shaped single crystals, showing rhomboid and tetrahedral shapes for 1 and 2, respectively (Fig. 2). The bulk achiral and chiral crystals can be identied based on their shapes with the naked eye.

Enantioselective crystallization
The chiral molecular capsule was assembled from achiral building blocks in aqueous solution. Circular dichroism (CD) studies conrmed that both enantiomers formed in the reaction. Single crystal X-ray diffraction analysis revealed that the enantiomers displayed a complete enantioselective self-resolution with the assistance of Co 2+ ions during the crystallization process (Fig. 2). The formation of heteromeric assemblies was also observed without co-crystallization of the Co 2+ counter cation (Fig. 3 (Fig. 4). As shown in Fig. 5, the CD spectra of the chiral crystals exhibited strong Cotton effects at 314, 285, 248, and 211 nm, indicating that polyoxoanion 2 possesses optical activity in aqueous solution. This also conrmed the chiral characteristic of the anion cluster. The CD spectra of the two products were mirror images, and they possessed similar powder X-ray diffraction (PXRD) patterns ( Fig. 5a), which conclusively demonstrates that they are enantiomers.
Aerward, a single crystal was selected and dissolved in deionized water with sonication. A powder sample batch randomly picked from the bulk products was also dissolved in deionized water. Held at the same concentration, the two solutions were tested by CD and UV-vis measurements, respectively ( Fig. 5b and S7 †). As shown in Fig. 5b, the CD spectrum of the powder sample was almost identical to that of the single crystal at the same concentration, which revealed that the bulk samples of 2 crystallized in enantiopure form. Moreover, the CD spectra of mixtures of all the starting materials in    the peak positions of a simulated and experimental PXRD pattern at 20 C are in agreement with each other; (b) CD spectra for one chiral single crystal, powder samples, and one achiral single crystal in deionized water at the same concentration (1.49 mg mL À1 ); (c) and (d) CD spectra of chiral powder samples obtained by the first crystallization and second crystallization processes; (c) shows that the D-type POM was isolated in the first batch and the L-enantiomer was isolated in the second batch, and (d) shows that the L-type POM was isolated in the first batch and the Denantiomer was isolated in the second batch. the water solution and the ltrate aer the reaction did not exhibit any peaks ( Fig. 5b and S8d †). Aer collecting the rst batch of large tetrahedral crystals from the solution and allowing the water in the ltrate to slowly evaporate, another batch of small triangular crystals could be obtained. CD spectra indicated that the individual crystals obtained from the second batch exhibited the opposite chirality. However, the decrease of Cotton effects in the CD spectra of the bulk powder samples showed their reduced optical purity ( Fig. 5c and d). These results indicated that the two enantiomers both formed in the reaction but crystallized out of the solution at different crystallization rates. We can conclude that chiral symmetry breaking and chiral autocatalysis occurred during the crystallization process. 21 The nucleation of a monochiral crystal is a symmetrybreaking event on the microscale, and the nonlinear autocatalytic dynamics of secondary nucleation leads to chiral amplication and the eventual production of enantiopure crystals. In this process, the entire chirality should be inuenced by the chirality of the primary nucleus. The probability of attaining a le-or right-handed chiral form should be random in a symmetry-breaking event in the absence of any chiral inuence. To this end, we carried out 60 crystallization experiments and performed CD studies on each of the rst batch crystals. We found that 33 of these crystallizations gave D-enantiomers in their rst batches, while 27 of these crystallizations afforded Lenantiomers. Slow crystallization favored the formation of a homochiral high-nuclearity polytungstate assembly from achiral precursors. CD spectroscopy and single X-ray analysis showed that the chiral molecular capsule still possessed a highlevel of self-recognition during the fast crystallization process with a high concentration of Co 2+ ions, leading to homochiral individual crystals (Fig. S8 †). However, the decrease of the CD peaks of the bulk samples compared to that of the single crystal at the same concentration indicated that some crystals with opposite handedness were formed from the solution in a fast crystallization process, resulting in a lower optical purity. Achiral single crystal samples with a rhomboid shape (1) composed of the same polyoxoanions could be isolated by decreasing the concentration of Co 2+ in the reaction (to #46 mM). No Cotton effects were observed on the CD spectra of individual single crystals (Fig. 5b). The dened structures of the racemic and homochiral crystals thus provided a chance to study delicate interactions at the molecular-level for chiral recognition and propagation. As shown in Fig. 4, the co-crystallization of anion clusters and [Co(H 2 O) 6 ] 2+ groups resulted in homochiral crystals. Each [Co(H 2 O) 6 ] 2+ group was coordinated with three homochiral microanions through H-bonding interactions, with the typical H-bond length in the range of 2.6-2.9Å. The three homochiral microanions could be regarded as three bidentate ligands coordinated with the central [Co(H 2 O) 6 ] 2+ group to form the shape of a D or L propeller. The chiral microanion self-sorted via the chiral [Co(H 2 O) 6 ] 2+ groups into homochiral materials (Fig. 4). This illustrated that H-bonding interactions are responsible for such high-level chiral recognition, in a process similar to the supramolecular chirality frequently observed in biology. 22 In comparison, without the co-crystallization of Co 2+ ions, two enantiomers were arranged alternately, resulting in racemic crystals of compound 1 (Fig. 3).

Stability
The stability of the POMs in aqueous solutions of different pH values was investigated by UV-vis spectroscopy and cyclic voltammetry (CV). Compound 2 was dissolved in aqueous solutions of pH 5-9 and kept at room temperature. The UV-vis spectra of these solutions were taken every hour. No changes were observed in the UV-vis spectra at pH 5-8 aer a 6-hour incubation, suggesting that 2 is stable in pH 5-8 buffer solutions ( Fig. S12a-d †). In the pH ¼ 9 buffer solution, the absorbance of 2 decreased over time ( Fig. S12e †), indicating that 2 is unstable at that pH. To further conrm the stability of the POMs, the CV behaviors of 2 in pH 5-8 buffer solutions were obtained. No obvious changes in the CV characteristics were observed during 6-hour treatments (Fig. S12f-i †). We further studied the solution structure by ESI-MS. Aer compound 1 was dissolved in a pH 7 buffer solution and aged for 24 hours, ESI-MS spectra conrmed the existence of 1 ( Fig. S13 and Table  S2 †), which was in agreement with the solid state structures from X-ray diffraction experiments. These results conrm that the POMs are structurally stable in pH 5-8 aqueous solutions.

Cytotoxicity
In order to determine whether the present POMs possess antitumor activities and on which cell lines they are more effective, we studied their cytotoxicity against eight cancer cell lines, including four ovarian cancer cell lines, two colon cancer cell lines, one non-small lung cancer cell line and one breast cancer cell line, by the (3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium) (MTS) assay (Fig. S14, Tables 1, S3 and S4 †). The cytotoxicity of rac-1 was similar to both D-2 and L-2, and signicantly higher than that of {Sb 9 W 21 } on all tested cell lines. Specically, 1 and 2 showed a higher cytotoxicity on A2780, A2780cisR, and OVCAR-3 cells than on other cells, which meant that they are very effective on ovarian cancer cell lines. The IC 50 values (the concentration for a 50% growth inhibition) of 1 on these three cell lines were 0.77 AE 0.01, 4.35 AE 0.20, and 1.78 AE 0.07 mM, respectively. In addition, hmta and NH 4 Cl showed no cytotoxicity on all tested cell lines, even at very high concentrations (>500 and 900 mM, respectively). Although CoCl 2 exhibited some cytotoxicity on A2780 and A2780cisR cells, its IC 50 values (122.7 AE 3.21 and 114.6 AE 9.75 mM, respectively) were signicantly higher than those of 1 and {Sb 9 W 21 }. These results indicated that the higher cytotoxicity of 1 and 2 over {Sb 9 W 21 } was derived from POM assemblies, not the addition of other components. We also determined the cytotoxicity of POMs against normal cell line HEK-293 cells. The IC 50 value of POMs on HEK-293 cells is around 16 mM, which is $20-, 8.9-and 3.6-fold higher than that on A2780, OVCAR-3 and A2780cisR cells, respectively ( Fig. S16 and Table S6 †). These results demonstrated the potential use of POMs for the treatment of ovarian cancer.

Cellular uptake
Encouraged by the high cytotoxicity of POMs against A2780 and A2780cisR cells, the cellular uptake of POMs on both cell lines was investigated by directly measuring the metal content in the cells using inductively coupled plasma-mass spectrometry (ICP-MS, Agilent technologies, Santa Clara, CA) ( Fig. 6a and S17 †). Aer incubation for 4 hours at an equivalent concentration of 20 mM, the Sb content in A2780 cells was 0.78 AE 0.04, 0.74 AE 0.04, 0.75 AE 0.04, and 2.44 AE 0.08 nmol per 10 6 cells, and the W content was 2.02 AE 0.10, 2.08 AE 0.11, 2.03 AE 0.10, and 5.74 AE 0.29 nmol per 10 6 cells for 1, D-2, L-2, and {Sb 9 W 21 }, respectively. In addition, the Co content in A2780 cells was 0.36 AE 0.02, 0.40 AE 0.03, and 0.41 AE 0.03 nmol per 10 6 cells for 1, D-2, and L-2, respectively. These results indicated that the uptake of 1 is similar to 2, and three times less than that of {Sb 9 W 21 }. The molar ratio of internalized Sb to W was 0.38 AE 0.01, 0.36 AE 0.02, and 0.37 AE 0.01 (around 9 : 24) for 1, D-2, and L-2, and 0.43 AE 0.01 (approximately 9 : 21) for {Sb 9 W 21 }, and the molar ratio of internalized Co to Sb was 0.46 AE 0.03 (around 4 : 9) for 1 and 0.54 AE 0.01 and 0.55 AE 0.02 (roughly 5 : 9) for D-2 and L-2, respectively. These ratios did not signicantly deviate from those expected for the POMs, suggesting that the POMs and {Sb 9 W 21 } remain intact upon cellular uptake. The same molar ratio on A2780cisR cells was also observed, except that the uptake of the title compounds and {Sb 9 W 21 } on A2780cisR cells was about two times less than on A2780 cells, which may explain the lower cytotoxicity of POMs on A2780cisR cells than on A2780 cells. The cellular uptake of POMs on A2780 cells in the presence or absence of 10% FBS was also determined (Fig. S18 †). No difference in the cellular uptake was observed for the POMs incubated with or without FBS, indicating that FBS in the cell medium did not block the uptake of POMs.

Apoptosis analysis
It has been reported that POMs can exert cytotoxicity by inducing apoptosis of cancer cells. 17b,18a Thus, a ow cytometry assay was performed to see whether the present POMs can induce apoptosis on A2780 and A2780cisR cells. As shown in Fig. 6b and S19, † 1, D-2, and L-2 induced very high levels of cell apoptosis, resulting in 81.17%, 80.96%, and 80.36% apoptotic cells in A2780 cells, and 90.22%, 88.68%, and 85.16% apoptotic cells in A2780cisR cells, respectively. However, {Sb 9 W 21 } at the same concentration barely induced cell apoptosis on either cell line. These results suggest that the antitumor activity of POMs is attributable at least in part to the activation of the apoptotic pathway in cancer cells. The cell apoptosis induced by POMs was also conrmed by confocal laser scanning microscopy (CLSM) (Fig. 6c and S20 †). Most of the cells treated with POMs appeared to have apoptotic morphologies, as shown by membrane blebbing and the appearance of the membrane-associated apoptotic bodies. In addition, the presence of bright green uorescence from Annexin V in the cells treated with POMs conrmed that POMs could successfully induce cancer cell apoptosis. On the contrary, an apoptotic morphology and green uorescence were not observed in cells treated with {Sb 9 W 21 }, indicating that {Sb 9 W 21 } could not induce cell apoptosis at the same concentration. These results were consistent with the quantitative results of ow cytometry.
DNA fragmentation provided additional support for apoptosis. During apoptosis, activated nucleases degrade the higher order chromatin structure of DNA into fragments, which can be extracted from cells and visualized by gel electrophoresis followed by ethidium bromide staining. As shown in Fig. S21, † the presence of the characteristic DNA ladder in lanes 3-5 suggested that POMs can induce signicant apoptosis in both cell lines. In contrast, the absence of the DNA ladder in lane 6 indicated that {Sb 9 W 21 } could not induce apoptosis in these cell lines. The DNA fragments were further quantied by a terminal deoxynucleotidyl transferase mediated dUTP nick end labeling (TUNEL) assay ( Fig. 6d and S22 †). POMs resulted in $55% of apoptotic cells in both A2780 and A2780cisR cells. In contrast, cells treated with {Sb 9 W 21 } at the same concentration did not show any apoptotic characteristics.

Cell cycle analysis
The ability to proliferate indenitely is one of the identifying characteristics of cancer cells, and is key to tumor growth and development. The alteration of the cell cycle induced by POMs was analyzed by determining the amount of DNA in the cells. As shown in Fig. 6e and S23, † most of the blank cells are in the G 0 / G 1 phase, suggesting that the cells were preparing for DNA synthesis. The cell cycle signicantly changed aer treatment with 1, D-2, and L-2, with the ratio of the G 0 /G 1 phase decreasing by 52.2%, 46.2%, and 45.3% in A2780 cells, and 33.9%, 37.1%, and 39.3% in A2780cisR cells, respectively. Likewise, the percentage of the G 2 /M phase increased 2.57-, 2.01-, and 2.26-fold in A2780 cells, and 1.85-, 1.27-, and 1.96-fold in A2780cisR cell, respectively, aer treatment with 1, D-2, and L-2. These results indicated that POMs can inhibit cancer cell proliferation by blocking the cell cycle. However, cells treated with {Sb 9 W 21 } exhibited a cell cycle similar to the control cells, suggesting that {Sb 9 W 21 } could not change the cell cycle at a comparable concentration.

Protein binding studies
When drugs are bound to proteins, the resulting protection from metabolic degradation can prolong their activity. The proportion of drug molecules bound to the protein depends on the total drug concentration and the drugs' affinity to the protein. The distribution and metabolism of many biologically active compounds are correlated with their affinities to serum albumin, the most abundant protein in blood plasma, accounting for about 60% of its total protein. Thus, an investigation of the interaction between these compounds and serum albumin can provide important information on their potential as therapeutic drugs. 23 The interaction of POMs with bovine serum albumin (BSA) was investigated using multi-spectroscopic methods. UV-vis absorption measurements indicate the structure and the conformation of the complex. As shown in Fig. S24 and Table  S7, † BSA displayed a strong absorption peak at 204 nm, which represents the content of an a-helix structure. Aer the addition of POMs, the peak intensity decreased, which was accompanied by a bathochromic shi from 204 nm to 210 nm. The absorbance at 280 nm, mainly coming from tryptophan, tyrosine, and phenylalanine, increased in the presence of increasing amounts of POMs. The changes in the characteristic absorption indicated that POMs can interact with BSA, and this interaction may change the structure of BSA by decreasing the number of a-helix structures. Interestingly, the increase in the absorbance at 280 nm of BSA titrated with {Sb 9 W 21 } (about 18%) was much lower than when it was titrated with POMs (approximately 35%), which indicated that {Sb 9 W 21 } may form fewer complexes with BSA than POMs do.
An intrinsic uorescence study was performed to evaluate the changes in structure that were caused by the reaction of BSA with POMs. Fig. S25 † shows a conspicuous change in the uorescence emission spectra of BSA with the addition of various amounts of POMs. In the absence of POMs, the relative uorescence intensity of BSA was about 70; in the presence of POMs, the relative uorescence intensities obviously decreased to 35, 31, 34, and 48 for 1, D-2, L-2, and {Sb 9 W 21 }, respectively, which were equivalent to reductions of 51.4%, 54.5%, 53.4%, and 32.4%, respectively (Fig. 7a). These phenomena and analyses suggested that POMs could bind to BSA, and that the binding induces some changes in the microenvironment of the tryptophan residues. 24 The lower quenching effect of {Sb 9 W 21 } was consistent with the UV results, and suggested that {Sb 9 W 21 } forms fewer complexes with BSA than POMs do.
The uorescence quenching data were further analyzed by the Stern-Völmer equation: 25 where F 0 and F are the uorescence intensities without and with quenchers. K q is the quenching rate constant of the biomolecule, K sv is the Stern-Volmer dynamic quenching constant, s 0 is the average lifetime of the molecule without the quencher (10 À8 s) and [Q] is the concentration of the quencher, respectively. of POMs to BSA at variable reaction times (***P < 0.001); (e) calorimetric data for the titration of BSA with 1. The binding isotherm (heat change vs. POM/BSA molar ratio) was obtained from the integration of raw data and was fitted to a "one-site" model. Fig. 7b and Table S8, † the values of K q for 1, D-2, L-2, and {Sb 9 W 21 } were (1.02 AE 0.03) Â 10 13 , (1.18 AE 0.21) Â 10 13 , (1.11 AE 0.89) Â 10 13 , and (0.45 AE 0.06) Â 10 13 L (mol s) À1 , respectively, which were greater than the value of the maximum scatter collision quenching constant (2.0 Â 10 10 L (mol s) À1 ). Therefore, the quenching was more of a static quenching that was initiated by complex formation, rather than by dynamic collision.

As shown in
Dynamic and static quenching mechanisms can also be distinguished by changes in the uorescence lifetime. The lifetime decreases with the addition of quenchers for dynamic quenching, while no change is observed for static quenching. The static quenching mechanism was further conrmed by the lifetime determination, with no decrease observed for the lifetime of BSA aer the addition of POMs and {Sb 9 W 21 } ( Fig. 7c  and S26 †).
The in vitro binding of POMs to serum albumin was quan-tied by an ICP-MS assay (Fig. 7d). The binding of POMs to albumin occurred practically immediately, with about 45% of POMs bound to BSA aer a 30-minute incubation. The binding degree then increased to about 70% aer a 10-hour incubation and reached about 80% aer 48 hours of incubation. {Sb 9 W 21 } also showed fast binding to BSA, but only 60% of {Sb 9 W 21 } was bound to BSA aer 48 hours of incubation, conrming the higher affinity of POMs to BSA than {Sb 9 W 21 }.
The binding constants of POMs to BSA were determined by isothermal titration calorimetry (ITC). The binding constants for 1, D-2, L-2, and {Sb 9 W 21 } were determined to be (2.67 AE 0.36) Â 10 4 , (2.31 AE 0.22) Â 10 4 , (2.36 AE 0.30) Â 10 4 , and (1.68 AE 0.44) Â 10 4 M À1 , respectively. In addition, about 40% of POMs were bound to BSA, while only 13% of {Sb 9 W 21 } interacted with BSA ( Fig. 7e and S27 †). The protein binding studies demonstrated that POMs can bind to proteins, likely through electrostatic interactions, with a higher affinity than {Sb 9 W 21 }, which may explain their greater cytotoxicity. We believe that, aer internalization, POMs affect the cellular function by binding to the proteins in the cells, thereby inducing cell apoptosis and resulting in cell death. {Sb 9 W 21 } had a lower affinity to proteins than POMs, which resulted in less cell apoptosis and a lower cytotoxicity, although it showed a higher cellular uptake than POMs.

Conclusions
In summary, homochiral crystalline materials were synthesized from achiral precursors via chiral symmetry breaking and asymmetric autocatalysis. During crystallization, the enantiomers showed a high level of chiral recognition, causing them to crystallize into individual enantiopure crystals through a counterion-mediated interaction. Detailed structural analysis revealed that H-bonding interactions are responsible for such high-level chiral recognition, in a process similar to the supramolecular chirality frequently observed in biology. The successful synthesis of the POM-based materials in their homochiral forms not only provided rare examples of homochiral materials, but also suggested a new structural model for studying the chemical origins of life. In addition, the POMs synthesized here were effective against various cancer cells, especially resistant ovarian cancer cells. The POMs could block the cell cycle and complex with proteins with high affinities, thereby efficiently inducing cell apoptosis and inhibiting cell proliferation. These results suggest that they are promising antitumor agents, especially for the treatment of resistant cancers and other hard-to-treat solid tumours.

Materials and instruments
All the reagents were commercially purchased and used without further purication. The (NH 4 ) 18 [NaSb 9 W 21 O 86 ]$24H 2 O precursor was synthesized according to the literature 20 and characterized by IR spectroscopy. Elemental analyses of Sb, W, and Co were performed on a PLASMASPEC (I) ICP atomic emission spectrometer and Agilent 7700x ICP-MS and analyzed using an ICP-MS Mass Hunter version B01.03. IR spectra were recorded in the range 400-4000 cm À1 on an Alpha Centauri FTIR spectrophotometer using KBr pellets. UV-vis spectroscopy was performed with a Varian Cary 50 spectrophotometer in the range of 200-800 nm. The PXRD data were recorded on a Bruker D8 Advance diffractometer. CD spectra were measured with a Jasco model J-810 spectropolarimeter. Time-domain lifetimes were measured on a ChronosBH lifetime uorimeter (ISS, Inc.) using Time-Correlated Single Photon Counting (TCSPC) methods. The uorimeter contained Becker-Hickl SPC-130 detection electronics and a HPM-100-40 Hybrid PMT detector. Excitation was provided by a 280 nm nanosecond pulsed LED. Emission wavelengths were selected with interference lters (Semrock BrightlineFF01-341/LP). The Instrument Response Function (IRF) was measured to be approximately 120 ps FWHM in a 1% scattering solution of Ludox LS colloidal silica. Lifetimes were tted using a forward convolution method in the Vinci control and analysis soware.

Synthesis
Synthesis of 1. (NH 4 ) 18 [NaSb 9 W 21 O 86 ]$24H 2 O (0.60 g, 0.085 mmol) was dissolved in distilled water (10 mL), to which solid CoCl 2 $6H 2 O (0.10 g, 0.42 mmol) was added with strong stirring. The pH value of the mixture was adjusted to 7.40 with a 1.2 M hmta solution, and then solid NaOH was added to adjust the pH value to 8.50. The resulting mixture was heated at 100 C for 2 hours. Aer the reaction, the pH value was 7.50. The ltrate was kept at room temperature with slow evaporation for 6 days, resulting in big purple rhomboid shaped crystals of 1 (yield ca. 110 mg). Anal. found (%): Co,2.61;Sb,12.01;W,47.82;calcd: Co,2.47;Sb,11.47;W,46.20. Synthesis of 2. (NH 4 ) 18 [NaSb 9 W 21 O 86 ]$24H 2 O (0.6 g, 0.085 mmol) was dissolved in distilled water (10 mL), to which solid CoCl 2 $6H 2 O (0.15 g, 0.63 mmol) was added with strong stirring. The pH value of the mixture was carefully adjusted to 7.20 with a 1.2 M hmta solution, and then solid NaOH was added in this system to adjust the pH value to 8.0. The resulting mixture was heated at 100 C for 2 hours. Aer the reaction, the pH value was 7.30. The ltrate was kept at room temperature with slow evaporation for 28 days, resulting in big purple tetrahedral crystals of 2 (yield ca. 20 mg). Aer collecting these crystals, the ltrate was kept at room temperature for 3 days, resulting in another batch of small purple triangular shaped crystals (yield ca. 88 mg). Anal. found (%): Co,3.07;Sb,11.89;W,48.08;calcd: Co,3.17;Sb,11.77;W,47.39

X-ray crystallography
The single crystal X-ray diffraction of 1 was collected with a Bruker APEX II CCD based detector at ChemMatCARS (Sector 15), Advanced Photon Source (APS), Argonne National Laboratory. The frames were integrated with the Bruker SAINT build in the APEX II soware package using a narrow-frame integration algorithm, which also corrects for the Lorentz and polarization effects. The crystallographic data was performed on an Oxford Diffraction Gemini R CCD for 2. The data were collected at 293 K, and with graphite-monochromated Mo-Ka radiation (l ¼ 0.71073Å). These structures were solved by the direct method and rened by full-matrix least squares on F 2 using the SHELXL-97 soware. 26 During the renement of compounds 1 and 2, the command 'isor' was used to restrain the non-H atoms with ADP and NPD problems, which led to restraint values of 322 and 143 for 1 and 2, respectively. The command 'omit-3 50' was used to omit the weak reection above 50 degrees for 2. There were a number of short connections between OW(water)/O(POM) in the range of 2.50-2.90Å, suggesting extensive H-bonding interactions between the lattice water molecules and POMs. All hydrogen atoms on water molecules, protonation, and the counter NH 4 + were directly included in the molecular formula.
Carbon and the nitrogen-bound hydrogen atoms of the hmta molecules were placed in geometrically calculated positions. In the renement, 14 and 18 lattice water molecules were found from the Fourier maps for compounds 1 and 2, respectively. However, there were still accessible solvent voids in the crystal structure, indicating that more water molecules should exist in the structure. On the basis of TG results, there should be another 15 water molecules in the formula unit, which were directly included in the molecular formula. The crystal data and structural renements of compounds 1 and 2 are summarized in Table S1. † CCDC reference number 1431324 for 1 and 827684 for D-2 contains the supplementary crystallographic data for this paper.

Cytotoxicity assay
Cells seeded in 96-well plates (2 Â 10 3 cells per well) were treated with different concentrations of POMs for 72 hours, and the cell viability was then measured by MTS (Promega, Madison, WI) according to the manufacturer's instructions. At the same time, the cytotoxicity of other components was also determined and they were used as controls. IC 50 values were calculated from curves constructed by plotting cell survival (%) versus drug concentration (mM).

Cellular uptake
A2780 and A2780cisR cells seeded in 6-well plates (5 Â 10 4 cells per well) were incubated with POMs (20 mM) for 4 hours. Cells were then collected, washed with PBS, dried, and digested for metal analysis by ICP-MS. The uptake level was expressed as the amount of metal uptake associated per million cells.

Apoptosis analysis
For the ow cytometry assay, A2780 and A2780cisR cells were treated with POMs (0.8 mM and 4.5 mM, respectively) for 24 hours. Cells were then harvested, washed with PBS, stained with Alexa Fluor 488 conjugated Annexin V and PI for 15 minutes at room temperature in the dark, and analyzed by ow cytometry. For Annexin V staining, A2780 and A2780cisR were seeded on 10 mm 2 glass coverslips placed in 6-well plates. Aer treatment with POMs for 24 hours, cells were washed with PBS, xed with 4% paraformaldehyde, and stained with DAPI and Alexa Fluor 488 conjugated Annexin V. The cells were then observed using CLSM.
For the DNA ladder, the total DNA was extracted from A2780 and A2780cisR cells incubated with POMs using a DNA ladder isolation kit (life technologies, Grand Island, NY) and examined for DNA fragmentation with 2% (w/v) agarose gel electrophoresis at 35 V for 3 hours.
For the TUNEL assay, treated A2780 and A2780cisR cells were collected, washed with PBS, xed with 70% ethanol at 4 C overnight. Cells were then stained with an APO-BrdU™ TUNEL Assay Kit (Molecular Probes) according to the manufacturer's instructions, and analysed by ow cytometry.

Cell cycle analysis
A2780 and A2780cisR cells treated with POMs for 24 hours were collected, washed with PBS, xed with 70% ethanol at 4 C overnight and treated with RNase A for 45 min, followed by PI staining for 30 minutes. The alteration of the cell cycle was analysed by ow cytometry.

Protein binding studies
BSA solution (1 mM) in PBS (pH ¼ 7.2) was titrated with POMs from 0-10 mM. Aer equilibration, absorption spectra measurements were carried out on a Shimadzu UV-2401 spectrophotometer at 200-350 nm. Fluorescence spectra were recorded using a Shimadzu RF-5301 spectrouorophotometer from 300 to 400 nm at an excitation wavelength of 280 nm. Time-domain lifetimes were measured on a ChronosBH lifetime uorimeter (ISS, Inc.).
A mixture of POMs (7.5 mM) and BSA solutions (5 mg mL À1 ) was incubated at 37 C. Aliquots were continuously taken and ultraltrated through a 30 kDa cut-off lter (Millipore, Bedford, Ohio) for 15 min at 10 000 rpm. The concentration of the free POMs in the ultraltrate was measured by ICP-MS. The degree of binding was calculated as: where C 0 and C free are the total concentration and the concentration of the free POMs, respectively.
Calorimetric titrations were carried out on a MicroCal iTC200 (MicroCal Inc., Northampton, MA), and the data were analyzed using the Origin soware. Typically, 39 mL of a POM solution (500 mM) was injected into the BSA solution (30 mM in 200 mL) over 20-24 s at 240 s intervals using a 40 mL syringe rotating at 1000 rpm. The initial delay (the hold period before injections) was set at 240 s. Before use, samples were degassed at 25 C using the ThermoVac accessory (provided by MicroCal Inc.).