Conformational control of anticancer activity : the application of arene-linked dinuclear ruthenium ( II ) organometallics

All commercially purchased materials were used as received. Ruthenium trichloride hydrate was purchased from Precious Metals Online, guanosine 5'-monophosphate disodium salt hydrate and Lhistidine were purchased from Sigma, 5'-ATACATCGTACAT-3' was purchased from Microsynth and H-Asp-Ala-Glu-Phe-Arg-His-Asp-Ser-Gly-Tyr-Glu-Val-His-His-Gln-Lys-OH as its trifluoroacetate salt from Bachem. (1S,2S)-(-)-1,2-diphenylethylenediamine (min. 97%) was obtained from Strem Chemicals, (1R,2R)-(+)-1,2-diphenylethylenediamine (98+%) was obtained from Alfa Aesar and meso-1,2-diphenylethylenediamine (98%) was purchased from Aldrich. Dichloromethane and diethyl ether were purified and degassed prior to use using a PureSolv solvent purification system (Innovative Technology INC). N'N-dimethylformamide (99.8%, Extra Dry, Acroseal®) and acetone (99.8%, Extra Dry, Acroseal®) were obtained from Acros Organics and methanol (anhydrous, 99.8%) was purchased from Sigma-Aldrich. H2O was obtained from a Milli-Q Integral 5 purification system.


Introduction
Metal-based compounds offer considerable potential in medicinal chemistry where the careful choice of metal may afford compounds possessing geometrical, coordination and potentially catalytic properties not accessible through purely organic molecules. As a result, the judicious combination of a metal ion and associated ligands may result in complexes capable of unique biological activity. In the context of developing metal-based drugs for the treatment of cancer, considerable effort has been directed toward the development of novel platinum complexes, 1 stemming from the clinical success of cisplatin (and subsequent derivatives).
Apart from the platinum family of metallodrugs several other metals have been utilised in the development and identication of further classes of compounds that exhibit favourable medicinal attributes. Of these, ruthenium-based organometallic compounds constitute a rapidly developing eld that continues to yield complexes exhibiting diverse biological activity. 2 Prominent examples include the [Ru(h 6 -arene)(en)Cl] + family of organometallics that has yielded compounds with a comparable cytotoxicity to that of cisplatin in certain cell lines. 3 Structurally related compounds based on the [Ru(h 6 -arene)(L)Cl] + scaffold, with organic ligands (L) chosen because of their various biological activities, also exhibited signicant antiproliferative activity against a range of cancer cell lines. [4][5][6] An alternative strategy, based on the development of kinetically inert ruthenium half-sandwich complexes as potent inhibitors of protein kinases, has also led to complexes exhibiting high cytotoxicity against the HCT-116 cell line. 7 Ru(II)-arene complexes have also been incorporated into multinuclear systems and assessed for their anticancer activity. These have included metalla-cycles, [8][9][10][11][12] metalla-cages [13][14][15] and dendrimer-based systems. 16 Of particular interest to us are the [Ru(h 6 -arene)Cl 2 (PTA)] (RAPTA) series of organometallic ruthenium(II) compounds which display selective activity on metastatic tumours in vivo. 17,18 Recent work observed that proteins are a major intracellular target of the RAPTA compounds, 19 including the histone proteins of the nucleosome core particle. Structural studies also show that RAPTA-C binds to the protein component of the nucleosome core particle in preference to the nucleic acid component. 20 In addition, subcellular localisation of RAPTA-T in A2780 and A2780cisR cells revealed higher metal to protein ratios in the particulate fraction (including the mitochondria) relative to the cytosol and nuclear fractions. 19 Further accounts have described selective binding of RAPTA-C within protein mixtures 21 and also potent enzyme inhibition by RAPTA compounds. 22,23 These studies offer an insight into the potential of the RAPTA compounds to exert their biological activity through protein interactions rather than DNA damaging mechanisms characteristic of the vast majority of platinum metallodrugs reported so far. We reasoned that dinuclear analogues of the RAPTA series may exhibit a different spectrum of biological activity compared to mononuclear RAPTA derivatives, potentially retaining the propensity of the mononuclear RAPTA compounds to bind proteins while acting via crosslinking of target biomolecules through long-range interactions rather than short range interactions typical of mononuclear species. In the eld of platinum metallodrugs the formation of dinuclear analogues has been a successful route toward the development of compounds capable of unique binding modes and interactions with DNA. 24 This approach has led to compounds of increased potency compared to established mononuclear platinum compounds which retain high activity in cell lines resistant to cisplatin. 25,26 In contrast, dinuclear organometallic ruthenium(II) arene compounds are underexplored. 27 In the most comprehensive of these studies to date a series of dinuclear ruthenium compounds linked via maltol-derived ligands were investigated. [28][29][30] The cytotoxicity of these compounds was found to be tuned by modication of the chain length of the alkyl component of the linker ligand, with cytotoxicity correlating well to the experimentally determined lipophilicity of the resulting complexes. 28 Of notable interest is the observation that not only could these complexes crosslink two DNA duplexes but also that DNA-protein crosslinks could be formed, potentially a novel mode of action for such complexes. 30 We hypothesised that the use of linker ligands rigid enough to x the relative orientation of the metal centres in a dinuclear RAPTA complex would allow access to a series of isomeric structures of differing conformation. Such complexes would allow an assessment of the effect of conformation on biological activity to be probed, including the manner in which isomeric complexes of different conformation interact with potential target biomolecules.
Here we report the synthesis, binding studies with potential biological targets and in vitro biological evaluation of a series of dinuclear Ru(II)-arene compounds, where the conformation of the RAPTA units relative to each other is controlled by the stereochemical conguration of a 1,2-diphenylethylenediamine (DPEN) linker molecule. Compounds with a linking group possessing either an (R,R)-or (S,S)-conguration exhibit a more 'closed' conformation whereas those with a (R,S)-conguration possess an open conformation not dissimilar to that of a complex linked via a exible linker (Fig. 1). Such conformational control could signicantly inuence the interactions of each complex with biomolecular targets, ultimately yielding a different range of adducts, and potentially controlling the cytotoxicity and selectivity.
In addition to the rigid dinuclear complexes described above a further dinuclear complex, linked together using the exible ethylene diamine analogue, and a mononuclear analogue are also reported. Our approach was to develop dinuclear Ru(II)arene compounds linked together through the arene ligand in order to retain the core coordination environment around ruthenium as in the original RAPTA series to avoid signicantly perturbing the coordination mode of these dinuclear complexes with potential biological targets.

Synthesis and characterisation
The synthesis of the ruthenium dimers, and the mononuclear analogue, proceeded via amide-forming reactions between the common intermediate complex, 1, bearing a carboxylic acid substituent on the arene ligand, and the selected amines (Scheme 1). Complex 1 includes a single chelating oxalatomoiety as a protecting group in order to suppress undesired reactions with the ruthenium centre during amide bond formation. Reactions between 1 and the desired amines were performed utilising the coupling agent O-(benzotriazol-1-yl)-N,N,N 0 ,N 0 -tetramethyluronium tetrauoroborate (TBTU) with N,N-diisopropylethylamine (DIPEA) in DMF, to either yield 3a and 4a as DMF-insoluble yellow powders that were isolated by ltration, or monomer analogue 2a and dinuclear compounds 5a and 6a as crude yellow powders following precipitation with acetone. All compounds were further puried by recrystallisation until analytically pure (see ESI †). Each oxalato-protected ruthenium complex was successfully converted to its chloridoanalogue by dissolution in an anhydrous solution of HCl in methanolthe yellow solutions immediately turned red in colour followed by precipitation of the desired dinuclear compounds 3b-6b as their HCl salts, or in the case of monomer analogue 2b the product was obtained by precipitation with diethyl ether. All compounds (1-6b) are water soluble; the oxalato-protected compounds (2a-6a) possess low solubility in DMSO and other organic solvents. All compounds were characterised by 1 H, 31 P and 13 C NMR, high-resolution mass spectrometry and elemental analysis (see ESI †).
The molecular structures of 1, 2a, 4a, 5a and 6a ( Fig. 2 and S1 †) were conrmed by single crystal X-ray crystallography on  In addition, the resonance set corresponding to the methylene protons is more shielded and occurs between 2.16-2.33 ppm. These observations correlate well with the observation in the X-ray structure of 4a that the corresponding protons are shown to reside above the plane of the linker phenyl groupsa magnetic environment which would provide shielding from the applied magnetic eld.
Scheme 1 Synthesis and numbering scheme of the mononuclear oxalato-complex 2a, dinuclear oxalato-complexes 3a-6a and chloridoanalogues 2b-6b. Furthermore, in the spectrum of 4a the linker phenyl protons give rise to a single resonance at 7.4 ppm. The increased complexity in the spectra of 4a-6a relative to the spectrum of 3a reects the increased central rigidity present in the former complexes due to the DPEN linkers. In addition, the shielding observed in the spectrum of 4a provides direct evidence that the conformation observed in the solid state is also preserved in aqueous solution.

Aquation and stability studies
Many metal complexes, including cisplatin and several families of organometallic ruthenium complexes, are activated on cellular internalization (low [Cl À ]) by aquation of labile metalchlorido bonds, the aquated species being signicantly more labile toward donor groups of biological targets. 31 Early work 17 with a prototypical example of the RAPTA series, RAPTA-C, revealed that upon dissolution in pure water or 4 mM NaCl solution the complex rapidly undergoes exchange of both chlorido ligands for aqua ligands. In 100 mM saline solution this process is completely suppressed. Oxalato-and chlorido-containing ruthenium compounds synthesised in this work were assessed for their stability and reactivity in phosphate buffer (10 mM) at pH 7.2 in the presence of either 5 mM or 100 mM NaCl at 298 K. The UV-vis spectra of the chlorido analogues 2b-5b immediately change upon dissolution in phosphate buffer (10 mM) containing NaCl (5 mM), indicative of a change in coordination environment at the Ru centre. The half-lifes of the chlorido complexes (2b-5b) were estimated from the change in absorbance with time (see Fig. S5-S9 †). In phosphate buffer 3b possessed the shortest halflife (8.5 min) followed by 2b (10.5 min), 4b (13.6 min) then 5b (17.3 min), the more closed structure of the latter compound presumably slows the aquation process for this complex. For comparison, under these conditions RAPTA-C has a half-life shorter than all compounds described here (6.8 min). These results indicate that although the different structures and conformations of compounds 2b-5b affect the kinetics of ligand exchange at ruthenium, the 'active' aqua species of each complex may be readily formed on dissolution in low [Cl À ] aqueous solution. In the presence of elevated NaCl concentrations (100 mM) no aquation of these compounds was observed, in accordance with previous results. 17 The oxalato-protected compounds 2a-5a were also examined by 31 P NMR spectroscopy in the presence of HEPES-phosphate buffer (5 mM or 100 mM NaCl), and in the presence of BSA, CT-DNA and glutathione. In each case the appearance of no new 31 P-NMR signals were observed over a period of 72 h at 310 K, demonstrating that the oxalato complexes are inert towards aquation or ligand exchange under ambient conditions on the timescale of the experiment. In contrast, on incubation of 2a-5a in RPMI or DMEM media used for cell culture it was found that a new resonance at À32.4 ppm gradually appears in the 31 P NMR spectra over 48 h, with incubation of the complexes in DMEM resulting in the largest transformation ( Fig. S10-S16 †). The new resonance at À32.4 ppm may be attributed to the exchange of the oxalato ligand for a carbonato ligand as sodium bicarbonate is present in the RPMI and DMEM media used at concentrations of 23.8 mM and 44.0 mM respectively. Subsequent experiments involving the incubation of 2a with NH 4 CO 3 H (23.8 mM) in phosphate buffer (pH 7.4, 50 mM) conrmed this hypothesis, as the 31 P NMR spectrum aer 24 h exhibits the same transformation as the spectra recorded in RPMI/DMEM. The oxalato-protected complexes 2a-6a are stable toward aquation and inert in the presence of typical biological ligands, whereas media containing bicarbonate at concentration levels typically found in biological systems (10-30 mM), results in the gradual exchange of the oxalato ligand for a carbonato ligand, providing a possible route toward the intracellular activation of the oxalato-protected compounds.

Amino acid and nucleotide binding studies
In order to assess the ability of the complexes to coordinate to and crosslink biological targets, binding studies were performed with guanosine 5 0 -monophosphate, L-histidine and a range of model oligonucleotides and peptides.
To establish whether each ruthenium ion of the dinuclear compounds 3b-6b could coordinate simultaneously to target ligands, binding studies were performed initially with guanosine 5 0 -monophosphate and L-histidineboth being good ligands for RAPTA-type compounds as constituents of DNA or proteins, respectively. 20,32 For each dinuclear complex 3b-6b, following incubation in the presence of 2 equivalents of 5 0 -GMP (72 h, 310 K, pH 4.5 unbuffered), electrospray ionization mass spectrometric analysis revealed the presence of a range of 2 : 1 and 1 : 1 5 0 -GMP-metal complex adducts. Interestingly, for the complexes with the closed conformation, 5b and 6b, and the exible ethylene diamine linker, 3b, these adducts include examples where all four labile chlorido ligands are lost and replaced with a single 5 0 -GMP moleculeindicating both ruthenium centres are bridged via coordination to the same 5 0 -GMP ligand. 1 H and 31 P NMR spectroscopy indicated the most likely coordination sites are N7 and phosphate oxygen (see below). For the more open complex, 4b, the conformation of which is expected to be less favourable for intramolecular bridging coordination to occur, the corresponding peaks in the mass spectrum are of a relatively low intensity, the peaks corresponding to 2 : 1 5 0 -GMP-metal complex adducts are of higher relative intensity (Fig. S17 †). Studies with the monomer analogue 2b reveal the formation of only 1 : 1 5 0 -GMP-metal complex adducts, even in the presence of 2 equivalents 5 0 -GMP. For incubations of dinuclear compounds 3b-5b with 2 equivalents of 5 0 -GMP at pH 7.5 (310 K, 60 mM ammonium acetate) only 1 : 1 adducts were detected. Adducts include examples where all four chlorido ligands are replaced by a single 5 0 -GMP ligand, alongside similar adducts also containing a coordinated acetate ligand. It is likely that at pH 7.5 the deprotonation of the 5 0 -GMP phosphate (pK a of 6.49 reported for PO 3 H of 5 0 -GMP 33 ) facilitates the formation of intramolecularly bridged 1 : 1 adducts in preference to the formation of 2 : 1 5 0 -GMP-metal complex adducts. Complementary binding studies between 5 0 -GMP and 3b were also monitored by 1 H and 31 P NMR spectroscopy over 24 h (pD ¼ 7.5, 200 mM HEPES buffer, 310 K) (Fig. S18 †). Spectra are complex due to the formation of diastereoisomeric adducts upon coordination with 5 0 -GMP. Multiple new 5 0 -GMP H8 resonances in the region 8.00-8.85 ppm were identied by 1 H NMR spectroscopy corresponding to adducts involving coordination of N7 of 5 0 -GMP to the ruthenium ion. Concurrent analysis with 31 P NMR spectroscopy revealed the appearance of new 5 0 -GMP resonances (À2.90 to 8.45 ppm) indicative of phosphate coordination to the ruthenium ion alongside N7 coordination. This data complements the observation of intramolecular crosslinking of these dinuclear RAPTA analogues by 5 0 -GMP as observed in the mass spectrometric studies. A further distinctive proton resonance set, centred at 7.05 ppm, also gradually appeared in the aromatic region of the 1 H NMR spectra. This was deduced to be due to formation of the free arene ligand, indicating that upon coordination of 5 0 -GMP to 3b, potentially through both phosphate and N7 coordination, the arene ligand is lost from the adduct. In addition, the colour of the solution changed from yellow to black over the 24 h incubation period indicative of complex decomposition. Similar 5 0 -GMP-Ru adduct(s) were not identied during the mass spectrometric studies although the free arene ligand was detected by ESI-MS in the incubations performed at pH 7.5. The loss of the arene ligand is not unprecedented with similar loss of the arene ligand being reported previously in binding studies of RAPTA-C with a 14-mer oligonucleotide. 34 Binding experiments were also performed with 2 equivalents of L-histidine at pH 7.1 and monitored by 1 H and 31 P NMR spectroscopy and ESI-MS. As with 5 0 -GMP, ESI-MS revealed the formation of a range of 1 : 1 adducts between Lhistidine and dinuclear compounds 3b-6b. Bridging of Lhistidine between the two ruthenium ions is observed for all the dinuclear compounds; characterised by peaks corresponding to the loss of all four chlorido ligands in exchange for a single histidine, although the intensity of these peaks are relatively low compared to the intensity of other 1 : 1 adduct peaks (Fig. S19 †). Peaks corresponding to 2 : 1 L-histidine-metal complex adducts are also observed in all the spectra of the dinuclear compounds, but are of low relative intensity (<15%). For the mononuclear analogue 2b, in the presence of 2 equivalents of L-histidine, only peaks corresponding to 1 : 1 adducts are observed. 1 H NMR spectroscopy indicates that loss of the arene does not take place during the 24 h incubation, and, unlike the 5 0 -GMP incubation, the colour of the solution remains yellow.
As these small-molecule binding experiments conrm all dinuclear complexes 3b-6b readily coordinate to appropriate donor ligands simultaneously through both ruthenium atoms, binding studies were expanded to include a short model peptide sequence and a single-stranded 13-mer oligonucleotide.

Oligonucleotide and peptide binding studies
Oligonucleotide binding studies were performed on the 13-mer sequence 5 0 -ATACATCGTACAT-3 0 in unbuffered aqueous solutions (72 h, 310 K, pH ¼ 4.5, 0.2 mM complex and 2 eq. oligonucleotide). The reaction mixture was then diluted and directly analysed using ESI-MS in negative ionization mode. In higher mass regions of the spectra peaks attributable to 1 : 1 oligonucleotide-metal complex adducts are observed in which the metal complexes have lost their chlorido ligands. No higher order adducts were detected. However, the predominant species in these spectra correspond to oligonucleotide-metal complex adducts where the arene ligand is lost leaving the Ru-PTA fragment coordinated to the oligonucleotide. These results correlate with the NMR binding studies of 5 0 -GMP with 3b where loss of the arene ligand was observed.
Peptide binding studies utilised a fragment of amyloid bprotein (residues 1-16, H-Asp-Ala-Glu-Phe-Arg-His-Asp-Ser-Gly-Tyr-Glu-Val-His-His-Gln-Lys-OH). This 16-mer contains three histidine residues as well as one lysine and two glutamate residuesall of which have been observed to coordinate to RAPTA compounds in crystallographic studies. 20 Incubations were performed in unbuffered aqueous solutions (72 h, 310 K) in a 1 : 1 peptide-metal complex ratio. Using ESI-MS in all cases a range of 1 : 1 adducts were observed with no 2 : 1 or higher order complex-peptide or peptide-complex adducts detected. As with the small-molecule binding studies, 1 : 1 dinuclear metal complex-peptide adducts are detected where all four labile chlorido ligands are lost (and, in some cases, loss of PTA ligands is also observed) and substituted by a single peptide. The loss of two or three ligands at each Ru centre implies that each metal of the dinuclear compounds must be coordinated to one or more amino acid residues of the peptide. It is likely that crosslinked species where each metal centre is bound to a different amino acid residue form a proportion of these 1 : 1 adducts. ESI mass spectra of peptide adducts of 4b, 5b and 6b reveals a similar peak distribution (Fig. S20 †). To probe the nature of the binding within these 1 : 1 adducts electron-transfer dissociation (ETD) fragmentation studies were performed on selected ions.
Electron-transfer dissociation (ETD) fragmentation and ion mobility-mass spectrometry (IM-MS) studies ETD fragmentation of peptides is an established technique used to randomly fragment the N-C a bonds of a peptide backbone. This technique oen preserves post-translational modi-cations of the peptide side chains, such as phosphorylation and glycosylation, allowing their identication and localisation in a peptide sequence to be determined. 35 More recently, this technique has been used to identify drug metallation sites in peptide 36 and protein samples. 37 We used ETD to probe the binding of 2b-6b to the 16-mer peptide H-Asp-Ala-Glu-Phe-Arg-His-Asp-Ser-Gly-Tyr-Glu-Val-His-His-Gln-Lys-OH (see ESI † for a description of data analysis, see Table 1 for adducts analysed and location of metallated residues). Analysis of the ETD spectra of the 1 : 1 [peptide + 2b + 2H À 2Cl] 4+ and [peptide + 2b + 2H À 2Cl À PTA] 4+ adducts revealed complex binding at the His 6 , His 13 and His 14 sites. These observations could correspond to crosslinking of sites His 6 , His 13 or His 14 through metallation by a single Ru complex where crosslinking then breaks apart upon ETD fragmentation, or could correspond to a population of three metallated peptides within the sample, where the site of modication is solely at His 6 , His 13 or His 14 .
For dinuclear complexes 3b-6b, ETD spectra provided direct evidence that each of the complexes was able to crosslink the peptide through simultaneous coordination of each ruthenium centre to one or more different histidine residues at His 6 , His 13 and His 14 (Fig. 4). Further peptide adducts of each metal complex were analysed where either a single metallated histidine site was identied or no metallated amino acid sites were clearly identied (although metallation at the terminal amino acid residues was discounted). These spectra provide only a partial picture of the sites of metallation in these adducts but, given the 1 : 1 peptide-complex stoichiometries of the adducts, they also provide evidence towards crosslinking of the peptide through the histidine residues. It was noticeable that ETD spectra with 4b-6b are very similar for each particular adduct, exhibiting virtually identical fragment distributions in each case ( Fig. S21 and S22 †). These results suggest that although 4b and 5b/6b possess different conformations there is sufficient exibility in these peptide adducts to allow crosslinking through the same histidine positions. To probe these adducts further complementary binding studies were performed using ion mobility-mass spectrometry (IM-MS). IM-MS has been recently used to probe conformational changes in the protein ubiquitin on platination with cisplatin 38 and also in combination with fragmentation techniques to identify drug metallation sites on peptides. 36 We used IM-MS to probe the different complex-peptide adducts analysed by ETD fragmentation and comparisons of their arrival time distributions (ATDs) have been made (Table 2, Fig. 5, 6 and S23-S26 †). For the mononuclear complex-peptide adduct formed through loss of all chlorido and a PTA ligand [peptide + 2b + 2H À 2Cl À PTA] 4+ a single peak was observed in the ATD indicating this adduct exists as a single isomer in the gas phase. For the equivalent adducts of the dinuclear complexes ([peptide + X À 4Cl À PTA] 4+ (X ¼ 3b-6b)) the ATDs are more complex and exhibit two peaks (Fig. 5) indicating these adducts all exist as two isomeric species in the gas phase. For 1 : 1 peptide-complex adducts, formed through loss of all chlorido ligands from Ru whilst retaining all their PTA ligands ([peptide + X + nH À nCl] 4+ (X ¼ 2b-6b)), the ATDs in each case consist of two peaks. These results show each adduct in this series also exists as two isomeric species in the gas phase. A comparison of the dri times at which each peak in the ATDs is centred revealed that peaks for the [peptide + X + nH À nCl] 4+ adducts are all centred at equal or longer dri times than their corresponding [peptide + X + nH À nCl À PTA] 4+ adducts (X ¼ 2b-6b). This reects the larger size of the former adducts due to their retention of the PTA ligand, and highlights the important role played by the ligand set around the Ru ion in determining the shape of the adduct formed. The split distributions observed in the ATDs of all the dinuclear Ru compound-peptide adducts are likely due to crosslinking of the peptide by the dinuclear compounds between His 6 -His 13 and His 6 -His 14 to yield two adducts of different size. In contrast, the single peak observed in the ATD of the adduct [peptide + 2b + 2H À 2Cl À PTA] 4+ may be due to the metallation at the three histidine sites, as observed in the ETD analysis, yielding three adducts of identical size. An alternative interpretation is that the peptide is crosslinked by a single metal complex through sites His 6 , His 13 and His 14 to yield a single adduct; this interpretation correlates with the metal fragment of this adduct having lost two chlorido and one PTA ligand. Similar to the ETD analysis of the peptide adducts of the isomeric complexes 4b-6b no differences are observed in the ATDs of their peptide adducts, or between the ATDs of the [X À 4Cl + 2OH + 2H 2 O] 2+ (X ¼ 4b-6b, Fig. S25 †) ions themselves. Combined, these data indicate that despite the different conformations of the three complexes, as conrmed by crystallographic and 1 H NMR analysis, they have an identical size, at least in the gas phase. In addition, these data show that adducts formed between the peptide and the metal complexes are similar in terms of their size and in terms of their preferential binding sites.

Evaluation of in vitro anticancer activity
The cytotoxicity of 2a-6a and 2b-6b was assessed in human ovarian carcinoma (A2780), human ovarian carcinoma with acquired resistance to cisplatin (A2780cisR) and human Central amino acid residues (no metallation observed at terminal residues) embryonic kidney (HEK 293) cell lines using the MTT assay ( Table 3). As discussed earlier, the cytotoxicity of the original RAPTA series toward a range of cancer cell lines is low (IC 50 oen >300 mM) and these high IC 50 values are mirrored by the values determined for mononuclear compounds 2a and 2b, which are >300 mM in the three cell lines. All dinuclear  compounds 3a-6a and 3b-6b are signicantly more cytotoxic against the A2780 cell line than the mononuclear analogues. Of these, the most cytotoxic compounds are those with the closed conformation, i.e. 5a, 6a, 5b and 6b, with the chloridoanalogues (5b and 6b) being slightly more cytotoxic than the oxalato-analogues (5a and 6a). Interestingly, the diastereoisomeric complexes 4a and 4b are signicantly less active, with IC 50 values increasing over 5-fold in the case of the chloridoanalogue and over 2-fold for the oxalato-analogue. The exible dinuclear oxalato-complex, 3a, has a comparable cytotoxicity to that of 4a whereas the IC 50 value of the chlorido analogue, 3b, is signicantly greater than that of 4b.
In the A2780cisR cell line the IC 50 values of complexes 5a, 6a, 5b and 6b remain low, albeit, for complexes 6a and 6b, with the linker of (R,R)-conguration, with a slight increase in IC 50 values relative to those values obtained for the A2780 cell line. The IC 50 values of complexes 5a and 5b, with the linker of (S,S)-conguration, remain essentially unchanged. In contrast, the IC 50 values for 3a and 3b (linked by the exible ethylenediamine moiety) and 4a and 4b (with the linker of (R,S)-conguration) increase dramatically. For example, 4a and 4b possess IC 50 values 7 and 21 times, respectively, greater than those of their most active diastereoisomers (5a and 5b) whereas 3a is 9 fold less cytotoxic than the most active oxalato-compound 5a, and 3b is essentially inactive in the A2780cisR cell line (IC 50 > 300 mM). In addition, compounds 5b and 6b are signicantly more active than cisplatin in this cell line (over 3 fold in the case of 5b).
In the HEK-293 cell line, used as a model for non-tumorigenic cells, the activity of the compounds follows the pattern of activity of the compounds observed in the A2780 cell line. The cytotoxicity of 5a, 6a, 5b and 6b towards this cell line remains higher than that of 3a, 3b, 4a and 4b with no signicant differences in IC 50 values between chlorido-and oxalatoanalogues observed for each case. One exception is that in this cell line the activity of the chlorido complex 3b is greater than that observed in the A2780 and A2780cisR cell lines and comparable to the value of its oxolato analogue 3a, which in other cell lines is more active.
The differences in activity between diastereoisomeric complexes 4a-6a and 4b-6b are signicant, with an apparent conformational dependence observed for both cytotoxic activity and also for susceptibility to resistance in the A2780cisR cell line. The complexes with the "closed" conformation consistently show high cytotoxicity against all cell lines examined, and in the case of 5a and 5b, with the linker of (S,S)-conguration, virtually no loss of activity against the A2780cisR cell line. In contrast, the compounds with the more open structure (4a and 4b incorporating the linker of (R,S)-conguration), or those with a exible ethylenediamine linker that may also be expected to occupy an open conformation (3a and 3b), are signicantly less cytotoxic and also show a dramatic decrease in cytotoxicity against the A2780cisR cell line. To probe whether the origin of these observations were due to the differential intracellular uptake of these complexes, A2780 cells were incubated for 5 h with 2a-6a (300 mM) and the intracellular ruthenium content was determined using ICP-MS. It was found that the level of internalised 4a (391 AE 52 pmol Ru per 10 6 cells) is comparable to that observed for 5a (332 AE 25 pmol Ru per 10 6 cells). Unexpectedly, under identical incubation conditions the level of 6a (901 AE 97 pmol Ru per 10 6 cells) is signicantly higher than that observed for 5a despite it having a very similar cytotoxicity prole. At present, the origins of this differential uptake are unknown, though a recent report has shown the chirality of substituents on the ruthenium arene ligand can have a significant inuence on biological activity. 39 Compound 3a exhibits slightly higher cellular uptake levels (474 AE 21 pmol Ru per 10 6 cells) than 4a and 5a whereas the mononuclear analogue 2a (459 AE 89 pmol Ru per 10 6 cells) displays the highest level of internalised complex of all the compounds. These results indicate that while the levels of ruthenium cellular uptake do not correlate with the cytotoxicity proles of the compounds at 72 h, all compounds, under the experimental conditions, readily internalise in A2780 cells and at comparable levels observed for RAPTA-T. 19 In addition to these uptake studies  the log P values for complexes 2a, 3a, 4a, 5a and 6a were determined using the shake-ask method. 40 These experiments reveal the hydrophilic nature of all complexes with little variation in log P within the series. Therefore it is reasonable to conclude that the differences in the cytotoxicity, at least in the A2780 cell line, must be due to cellular events associated with the complexes and their ability to crosslink target biomolecules as demonstrated in binding studies, and is not due to signicant differences in complex hydrophilicity and cellular uptake.

Conclusions
We have developed a new route to arene-functionalised organometallic Ru(II) compounds, allowing access to exible and conformationally rigid dinuclear organometallic Ru(II) complexes. Crystallographic studies and NMR spectroscopy revealed that the stereochemical conguration of the 1,2diphenylethylenediamine (DPEN) linker controls whether the dinuclear complexes have an open or closed conformation. The range of adducts formed between these rigid dinuclear complexes, 4b-6b and the small molecules 5 0 -GMP and L-histidine appears to be governed by the conformation of the complex, and all dinuclear complexes 3b-6b are capable of crosslinking model peptide/oligonucleotide sequences. The mononuclear complexes, as observed for the original RAPTA series, are non-cytotoxic whereas dinuclear complexes are signicantly more cytotoxic. Although the different conformations of 4b vs. 5b/6b resulted in no differences in type and size of adducts formed in model peptide binding studies, an apparent conformational dependence on the cytotoxicity of these dinuclear complexes was observed. Those with the more closed conformation (5a-6b) are signicantly more cytotoxic than those with the more open conformation (4a and 4b) and are unaffected by resistance mechanisms operating in the A2780cisR cell line. In contrast, 4a and 4b were signicantly less cytotoxic toward this cell line than the A2780 cell line. However, for all dinuclear complexes, cytotoxicity toward the cancerous A2780 cell line and the non-cancerous HEK-293 cell line varied little. Further studies revealed the hydrophilicity of the compounds, as assessed by measurement of log P values, was similar within the series 2a-6a, and uptake experiments with A2780 cells revealed an appreciable level of complex association in each case. Combined, these observations lead to the conclusions that the different cytotoxicities of the mononuclear and dinuclear compounds is linked to the ability of the complex to crosslink biomolecular targets and that the cytotoxicity of the dinuclear compounds is signicantly linked to their conformation. thank Dr Luc Patiny for developing calculation tools for interpretation of ETD fragmentation spectra. We thank Baihua Ye for chiral HPLC analysis and Prof. Nicolai Cramer for access to these facilities.
Notes and references ‡ Structural data obtained from single crystals of samples of 5a and 6a is identical, with each crystal found to contain a 50 : 50 mixture of 5a and 6a resulting in the solved structures exhibiting disorder in the region of the chiral centres of the DPEN linker molecule. As the (1S,2S)-DPEN and (1R,2R)-DPEN linkers utilised in the synthesis of 5a and 6a respectively were commercial samples with a stated enantiomeric excess of 98% and 99% respectively, we postulate the racemic composition of crystals of 5a and 6a are not reective of the bulk sample, but are most likely a result of the co-crystallization of the desired and undesired enantiomers present in the samples. Chiral HPLC was used to determine the percentage of each enantiomer in bulk samples of 5a and 6a by analysis of the arene ligand following cleavage from the Ru ions (see Experimental section and Fig. S2-S4 † for details). This analysis showed that 5a was obtained with an enantiomeric excess of 99.4% and 6a was obtained with an enantiomeric excess of Table 3 In vitro anticancer activity of compounds 2a-6b in human ovarian carcinoma (A2780), human ovarian carcinoma cisplatin resistant (A2780cisR) and human embryonic kidney 293 (HEK-293) cell lines after 72 h exposure, octanol-water partition coefficients for compounds 2a-6a and cellular (A2780) uptake of ruthenium after exposure to 2a-6a for 5 h exposure at 300 mM