Monomeric Ti(iv)-based complexes incorporating luminescent nitrogen ligands: synthesis, structural characterization, emission spectroscopy and cytotoxic activities.

This manuscript describes the synthesis of a series of neutral titanium(iv) monomeric complexes constructed around a TiO4N2 core. The two nitrogen atoms that compose the coordination sphere of the metallic center belong to 2,2'-bipyrimidine ligands homo-disubstituted in the 4 and 4' positions by methyl (2a), phenylvinyl (2b), naphthylvinyl (2c) or anthrylvinyl (2d) groups. The crystal structures of these complexes named [Ti(1)2(2a)], [Ti(1)2(2b)], [Ti(1)2(2c)] and [Ti(1)2(2d)] (where 1 is a 2,2'-biphenolato ligand substituted in the 6 and 6' positions by phenyl groups) are reported. The hydrolytic stability of the four complexes is evaluated by monitoring the evolution of the free 2a-d signals by 1H NMR spectroscopy. For the conditions tested (6 mM, DMSO-d6/D2O: 8/1), a rather good stability with t1/2 ranging from 180 to 300 min is determined for the complexes. In the presence of an acid (DCl), the hydrolysis of [Ti(1)2(2a)] is faster than without an acid. The cytotoxic activity against gastric cancer cells of the titanium-based compounds and the free disubstituted 2,2'-bipyrimidine ligands is tested, showing IC50 ranging from 6.2 ± 1.2 μM to 274 ± 56 μM. The fluorescence studies of the ligands 2a-d, and the complexes [Ti(1)2(2a-d)] reveal an important fluorescence loss of the ligands 2c and 2d upon coordination with the Ti(1)2 fragment. Frontier orbitals obtained by DFT calculations permit us to explain this fluorescence quenching.


Introduction
Luminescence is a recurrent property displayed by transition metal complexes. 1 Luminescent transition metal complexes are particularly attractive for applications as chemosensors, 2 probes for biological molecules 3 or organic light emitting diodes. 4 Luminescent transition metal complexes such as zinc, 5 copper, 6 gold, 7 platinum 8 or d 6 transition metal com-plexes (especially ruthenium) 9 have also been extensively applied for fluorescence cell imaging.
Titanium(IV) complexes are very diverse and represent a major class of compounds widely used in many fields of modern molecular science. Organometallic and coordination titanium(IV) compounds are powerful reagents for catalytic enantioselective processes, 10 organic transformations 11 and as initiators for polymerization reactions. 12 Alkoxides or aryloxides titanium species are precursors to generate titaniumbased materials through sol-gel processes. 13 In metallo-supramolecular chemistry, titanium(IV) centers are intensively employed to create self-assembled architectures from oxygen ligands 14 or from multiple components. 15 Applications of titanium(IV) complexes in bioinorganic chemistry are also deeply investigated, as long as titanium(IV) is recognized for its low toxicity for living matter. Therefore, titanium is found in many biomaterials and numerous applications of Ti(IV) in a medical or biological context have been described. 16,17 Particularly, titanium(IV) complexes are promising candidates for developing new anti-cancer drugs with minor side effects. 18,19 As we were intrigued by the rarity of reports detailing the coordination chemistry of titanium(IV) complexes bearing luminescent ligands, 20 this work aimed to study the ligating properties to a titanium(IV) center of fluorescent ligands and their resulting emissions. Thus, we report the synthesis and the structural characterization of a series of TiO 4 N 2 -based monomeric octahedral complexes bearing various substituted 2,2′-bipyrimidines (Chart 1). These complexes are built from two biphenolato derivatives and one substituted 2,2′-bipyrimidine unit as depicted in Fig. 1. The choice of these nitrogen ligands is motivated by the fact that 2,2′-bipyrimidine has shown to be a well-adapted chelate to successfully generate TiO 4 N 2 -based complexes. 15,21 Also, 4,4′-dimethyl-2,2′-bipyrimidine can be easily functionalized by various groups through Knoevenagel reactions. 22 Therefore, the synthesis of a TiO 4 N 2based monomeric octahedral complex bearing 2,2′-bipyrimidines substituted by "classical" fused polycyclic aromatic emitters, as the naphthyl or anthryl group, 23 was envisaged. Additionally, we also detail in this manuscript some biological applications of these Ti(IV)-based species.

Synthesis of ligands and complexes
The ligands 2b-d were synthesized according to Scheme 1 in a one-step double Knoevenagel condensation starting from the 4,4′-dimethyl-2,2′-bipyrimidine precursor 2a. 22 Compound 2a was condensed with benzaldehyde, 2-naphthaldehyde or 9-anthraldehyde using t BuOK to afford the targeted bipyrimidine derivatives 2b-d (yields ∼ 55%). The 1 H NMR spectra for 2b-d highlighted doublets with trans vinylic coupling ( 3 J ∼ 16 Hz) confirming the formation of the desired bipyrimidine compounds. Having in hand this set of bipyrimidine derivatives, the synthesis of the targeted octahedral monomeric complexes [Ti(1) 2 (2a-d)] constructed around a TiO 4 N 2 core was undertaken. The complexes were obtained by following a well-established procedure 21 based on the exchange of the labile monodentate ligands coordinating the metal center in [Ti(1) 2 (HO i Pr) 2 ] 24 by a nitrogen bidentate ligand as proposed in eqn (1).
Complexes [Ti (1)  respectively. The coordination polyhedra found in these structures are distorted as attested by the mean angles measured around the titanium atoms (N-Ti-N = 71. 4°± 1.2, = 111. 3°± 1.6). We also notice in [Ti(1) 2 (2d)] that the two anthryl groups are rather distant from one another, as the closest C-C distance measured between the two anthryl units is found to be d(C anthryl -C anthryl ) = 6.66 Å.

Hydrolytic stability
Owing to the poor or moderate hydrolytic stability usually observed for molecular Ti(IV) species such as classical anticancer titanium-based compounds like budotitane and titanocene dichloride, 25a-c the behaviour of compounds [Ti(1) 2 (2a-d)] towards hydrolysis has been evaluated by following a methodology similar to the one already reported for titanium complexes. 25d-f These studies were conducted by 1 H NMR analysis of solutions in DMSO-d 6 of the [Ti(1) 2 (2a-d)] complexes over time, upon addition of D 2 O at room temperature. For these analyses, a large amount of D 2 O (1000 equivalents versus the complex) was added to the complex solution so that a final complex solution concentration of 6 mM was obtained. It should be noted first that when the concentration of the complex solutions was doubled, an orange precipitate formed instantaneously, reflecting the moderate solubility of these species in polar media. Secondly, when the same experiment was performed without D 2 O, a slight degradation of the initial complex was noticed in DMSO. Next, the resulting 6 mM samples were analyzed by 1 H NMR. The evolution of the mixtures composition was monitored for [Ti(1) 2 (2a)], [Ti(1) 2 (2b)] and [Ti(1) 2 (2c)] by integrating the bipyrimidine 6,6′ proton resonances for the complex and for the free diimine ligand. For compound [Ti(1) 2 (2d)], the situation was different, as the vinylic proton signals were monitored. The integration plots of these signals overtime for complexes [Ti(1) 2 (2a-d)] are shown in Fig. 2. The complete transformation of the complexes over a period of time of more than 16 hours is not observed. By analyzing the mixtures after one week of their preparation, the presence of a not negligible amount of the starting complexes is still noticed. Also, the absence of free 2,2′-biphenol derivative in solution is confirmed. Therefore, the formation of titanium(IV) oxo-aggregates incorporating 1 through an olation or oxolation reaction mechanism could be envisaged since oxo-clusters are common products when low nuclearity titanium(IV) complexes react with water. 26 The presence of novel species formed upon hydrolysis in solution is confirmed by NMR analysis. The aromatic region of the 1 H NMR spectrum indicates broad signals different from the 1 H NMR signature of 1-H 2 . ES-MS analysis of an aliquot also proves the formation of unknown hydrolytic products displaying molecular weights not compatible with the initial complex and the free ligands (see the ESI †). Plotting ln [complex] or 1/[complex] with respect to time indicated highly complex kinetics for these hydrolytic processes. The t 1/2 values found for these experiments demonstrated significant differences between each reaction ( A linear relationship was found between these t 1/2 values and the volumes of the bipyrimidine ligands were computed from the crystal structures of [Ti(1) 2 (2a-d)] with the PACHA software 27,28 as shown in Fig. 3. This relationship reveals the shield effect played by the aromatic groups substituting the  (1) (1) 2 (2c)] and [Ti (1) 2 (2d)], respectively. Dots are the integrations corresponding to the coordinated diimine ligands, squares are the integrations of the free bipyrimidine ligands. The t 1/2 values of hydrolysis calculated from these plots are compiled in Table 1.  (1) 2 (2a)] complex is nearly quantitative after 10 hours for the two DCl concentrations tested as shown in Fig. 4. With an excess of DCl (9 equivalents versus [Ti(1) 2 (2a)]), the t 1/2 value of the complex is divided by two in comparison to an experiment conducted with 1.8 equivalent of DCl vs.
[Ti (1) 2 (2a)]. Altogether, with an acid, the disappearance of the complex is faster than the same reaction performed without DCl.

Photophysical analyses
To the best of our knowledge, there are very few reports of titanium(IV) coordination complexes that would be photoluminescent in solution at room temperature, if we exclude titanium phthalocyanines, which are very specific species. 29 Yet some other group 4 metal complexes (Zr, Hf ) exhibit such properties. 30 However, the luminescence of several Ti complexes at 77 K has been investigated. 31 Besides, substituted pyrimidines are a class of organic molecules, which often display light-emitting properties in solution at ambient temperature, and which have even been considered as candidates for OLEDs. 32 Unsubstituted 2,2′-bipyrimidine has been used as a bridging ligand in photoluminescent binuclear lanthanide complexes. 33 Moreover, various substituted 2,2′-bipyr-imidines and their coordination complexes have been reported in the literature, which display interesting photophysical properties (fluorescence and non-linear optics). 34 In the present section, we describe the optical properties, viz. absorption and emission spectroscopy, of solutions in dichloromethane of the ligands 1-H 2 and 2a-d, and of the derived titanium complexes [Ti (1) Table 2. Incidentally, the UV-vis spectrum of 1-H 2 has been reported in the literature, but in a different solvent. 35 Overall, the compounds absorb UV-vis light over the 240-500 nm range. The λ max value of compound 2a in CH 2 Cl 2 (244 nm) is close to the value reported in EtOH (248 nm). 36 A significant bathochromic shift is observed for the phenylvinylsubstituted bipyrimidine 2b, in agreement with an extended Fig. 4 Plots of the integration of the bipyrimidine 6,6' proton signals related to the bound diimine ligands (dots) and the free ligands (squares) in the spectra of [Ti (1) 2 (2a)]. The red curves correspond to the measurement made for a solution of [Ti (1) 2 (2a)] in DMSO-d 6 (450 μl) where D 2 O (50 μl) was added. The green and blue curves correspond to the measurements made for a solution of [Ti (1) 2 (2a)] in DMSO-d 6 (450 μl) where D 2 O and DCl were added (50 μl in total for the volume of D 2 O and DCl). The green curves correspond to the experiment conducted with 2.5 μl of DCl solution (38 wt% in D 2 O) whereas the blue curves correspond to the experiment conducted with 0.5 μl of DCl solution (38 wt% in D 2 O). t 1/2 = 67 min ± 5% for the experiment conducted with 2.5 μl of DCl; t 1/2 = 138 min ± 5% for the experiment conducted with 0.5 μl of DCl.  (1) 2 (2a)] (blue), 2b (green) and [Ti (1) 2 (2b)] ( pink) recorded at room temperature in CH 2 Cl 2 (concentration range: 9-55 µM).  (1) 2 (2c)] (blue), 2d (green) and [Ti (1) 2 (2d)] ( pink) recorded at room temperature in CH 2 Cl 2 (concentration range: 2-18 µM). electronic delocalization. The spectrum of 2c exhibits several bands of similar intensities, ranging from 276 to 365 nm; strikingly, the UV-vis spectrum of a reported bis(naphthylvinyl)substituted pyrimidine also exhibits signals at 276 and 332 nm with comparable absorption coefficients. 22b Following the interpretation of Le Bozec et al. with regard to other related molecules, viz. tetrakis(arylvinyl)-substituted bipyrimidines, 22a these absorptions can be assigned to intramolecular charge transfer (ICT) from the phenyl or naphthyl group towards the electron-deficient bipyrimidine nucleus. This statement is supported by the coplanarity of each aryl group with a vinyl bridge and with the pyrimidine cycle, which is observed on the X-ray structures shown in Fig. 1 (assuming that coordination to titanium does not affect the conformation of the ligand); this coplanarity allows a delocalization of the π electrons. However, as far as [Ti (1) 2 (2b)] and [Ti(1) 2 (2d)] bear only two substituents instead of four, and as the aryl groups bear no donor atom, the transitions that we observe lie at a higher energy than the reported transition. 22a Finally, the absorption spectrum of compound 2d contrasts with that of 2b and 2c, with an intense blue-shifted band at 256 nm ( probably assigned to a vinylpyrimidine-centered transition) and less intense bands at 354, 372 and 392 nm. The pattern of the latter is reminiscent of the absorption spectrum of anthracene itself. 37a Therefore, the spectrum of 2d suggests the absence of electronic connection with the anthracene moieties and the rest of the molecule. Again, this is in agreement with the conformation of this crystal structure of the titanium derivative [Ti(1) 2 (2d)] (Fig. 1) where the anthryl mean planes make an angle of 67.25°with the vinyl planes.
The UV-vis spectra of the complexes [Ti(1) 2 (2a-d)] display various intense bands (4.13 < log ε < 5.33) between 244 and 433 nm that can be reasonably assigned to ligand-centered (π → π*) transitions. For complexes [Ti(1) 2 (2c)] and [Ti(1) 2 (2d)], the pattern of the free ligands 2c and 2d is roughly reproduced (Fig. 6). In addition, they exhibit a very broad and less intense band; this band is often obscured by the other transitions, but it is revealed by its tail in the visible region (roughly up to 520 nm) and is responsible for the red or orange color of the solutions. In the case of complex [Ti (1) 2 (2a)], it is fairly detectable and is cen-tered at ca. 380 nm. It can be assigned to an aryloxide-to-Ti LMCT transition, in agreement with the literature data. 30d, 38 The photoluminescence analyses were conducted at low concentrations in order to avoid self-absorption effects. Compound 1-H 2 exhibits a strong luminescence at 353 nm with a quantum yield slightly lower than 10%. In comparison, the light-induced emission properties of phenol-derived organic molecules related to compound 1-H 2 have been reported. 39 Ligand 2a is not light-emissive in solution at RT, which is not surprising because 2,2′-bipyrimidine is not either; 33a it is noteworthy that even the π-delocalized 4,4′diphenyl-6,6′-dimethyl-2,2′-bipyrimidine ligand is not photoluminescent at room temperature. 34a 2b is only slightly emissive, and the corresponding titanium complexes [Ti(1) 2 (2a)] and [Ti(1) 2 (2b)] are not. Conversely, and as we expected, the 2-naphthyl and 9-anthryl-substituted ligands 2c and 2d are characterized by distinct emission signals respectively centered at 411 and 525 nm (Table 2 and Fig. 7). It is again interesting to compare the luminescence properties of these 4,4′-di(arylvinyl)-2,2′-bipyrimidines with those of related compounds: 4,6- di(arylvinyl)pyrimidines, with aryl = 2-naphthyl or 9-anthryl, respectively, emitting at λ max = 424 and 525 nm; 22b or 4,4′,6,6′tetra(arylvinyl)-2,2′-bipyrimidines (aryl = N,N-disubstituted aminophenyl), mentioned above, whose emissions are redshifted with respect to our data. 22a With regard to quantum yields, 2c and 2d are less emissive (0.6 and 0.7%) than these species, especially the latter. We also note that under aerobic conditions, the quantum yield of anthracene itself reaches 21%, whereas it is much lower with 4,6-di(anthrylvinyl)pyrimidine (2%) or with 2d; obviously the pyrimidine ring(s) partly quench the fluorescence of anthracene.
Of the four titanium complexes [Ti (1) (1) 2 (2d)] exhibit an emission, which is very weak, so that the quantum yields cannot be accurately determined; several excitation wavelengths were tested before we could detect a significant signal. Therefore, it appears that coordination to titanium(IV) quenches the fluorescence of the ligands almost completely. The emission band of compound [Ti (1) 2 (2c)] undergoes a bathochromic shift with regard to 2c (Δλ max = 66 nm), while [Ti (1) 2 (2d)] is blue-shifted by Δλ max = 35 nm with regard to 2d. Overall, the lifetimes of the excited states are very low, within a 0.1-10 ns range (see the ESI †). In addition, we checked that degassing a solution of [Ti (1) 2 (2d)] has no influence on the emission (which remains very weak), excluding thus any interaction with dioxygen. These data are in agreement with a fluorescence process (singlet excited states). The excitation spectra corresponding to the emission maxima reported in Table 2 were recorded and were usually in agreement with absorption spectra, with regard to the wavelength maxima observed in the spectra (compounds 1-H 2 , 2c, 2d, [Ti(1) 2 (2c)]). The fit was not as good for compound [Ti(1) 2 (2d)], but many overlapping transitions were observed in the absorption spectrum; it is possible that some would be emissive and others would not.

DFT calculations, frontier orbitals and discussion
In addition to the experimental optical and emission properties evaluation, the frontier orbitals of the ligands and the complexes were computed using DFT calculations with the B3LYP hybrid functional and cc-pVTZ basis functions. Calculations were performed with the GAUSSIAN09 Revision D01 software. 40 Energies associated with these frontier orbitals are listed in Table 3. First of all, for the ligands an excellent match is found between the HOMO-LUMO gaps and the lower energy bands characterized by UV-visible absorption spectroscopy. The HOMO and LUMO representations obtained for the ligands 2a-d are shown in Fig. 8. For ligands 2b and 2c, the HOMOs and the LUMOs are localized on one half of the molecules, indicating an electronic communication between the pyrimidine unit and the cyclic aromatic hydrocarbon part of the molecules. The situation is different for ligand 2d, as the HOMO and to a lesser extent the LUMO are almost centred on the anthryl fragment. This is in good agreement with the 2d absorption spectrum in the region between 350 nm and 450 nm, which resembles strongly the spectrum of anthracene (vide supra). This originated from a tilted arrangement of the anthryl units related to the pyrimidine units in 2d, as shown in the [Ti (1) 2 (2d)] crystal structure (Fig. 1). Concerning frontier orbitals of the complexes shown in Fig. 9, similar tendencies are noticed between each complex. All HOMOs and HOMO−1 are localized on the substituted biphenolato ligand (HOMO−1 are given in the ESI †). The LUMOs are all centred both on the metal and the nitrogen ligands. Indeed, the simple inspection of HOMOs, HOMO−1 and LUMO orbitals do not permit us to give clues about the origin of the emission properties modulations of 2c and 2d upon coordination to the titanium(IV) center. However, the examination of deeper orbitals for [Ti(1) 2 (2c)] and [Ti(1) 2 (2d)] is much more enlightening (the representations of the [Ti(1) 2 (2c)] HOMO−6 and HOMO−7 are given in the ESI †). For [Ti(1) 2 (2d)] as shown in Fig. 10, the anthryl fragments start to be involved for the HOMO−2 since the HOMO and the HOMO−1 are only centered on the biphenolato ligands. Furthermore, the energies associated with these orbitals are given in Table 4. Herein, an excellent match is found between the theoretical wavelength corresponding to the ΔE(LUMO-HOMO−2) and the λ max determined experimentally in the visible region (λ (calcd) = 438 nm, λ (exp) = 433 nm). Therefore, it can be reasonably proposed that upon excitation, an excited state is gained through the electronic promotion of an electron of a 2d-centered occupied orbital to the 2d-centered LUMO orbital. As the 1-centered HOMO orbitals lie between the HOMOs and the LUMO involving 2d, the fluorescence loss of 2d upon coordination is explained by an electron transfer of one electron of the 1-centered HOMO to the hole in the 2d-centered HOMOs. Transfer of the initially excited electron to the hole in the 1-centered HOMO follows this step. This intramolecular photoinduced electron transfer (PET) mechanism 41 is shown in Fig. 11. An identical PET mechanism is proposed to explain the fluorescence quenching in [Ti(1) 2 (2c)]. For [Ti(1) 2 (2c)], an excellent match is also found between the experimental λ max (380 nm) and the wavelength λ (calcd) = 369 nm associated with the gap determined between the frontier orbitals centered on the naphthyl fragment (LUMO-HOMO−6).

Biological properties
As mentioned above, a promising issue for Ti(IV) complexes concerns some applications in biology and more particularly as antitumor agents. Therefore, the biological activity of the titanium-based compounds was tested on cell viability in the  AGS gastric cancer cells. The cells were treated for 48 hours with increased concentrations of the drugs ranging from 1 to 200 µM (Fig. 12). The compounds were first solubilized in DMSO at a concentration of 10 mM for [Ti (1)  were not possible to test due to the poor solubility of these compounds in DMSO and the toxic effect of high doses of DMSO. The concentrations able to reduce by 50% the cell viability (IC 50 ) were determined using as maximal effect wells with no cells alive due to high cisplatin doses (Fig. 13) in order to compare the relative cytotoxicity of the various drugs. Systematically, the impact of DMSO alone was tested in parallel, and its lack of toxicity checked at the concentrations of the tests. Similar IC 50 results were obtained when the drugs were incubated in a medium for 4 hours and before a fresh medium without drugs was replaced for the remaining 44 hours before MTT assay.
These results indicate that the structure of the titanium complex has a direct impact on the biological activity of the compound.
To follow the localization of the compounds in cells, we incubated for 4 hours the compounds with AGS cells and we observed the fluorescence of the compounds under a fluorescent microscope (Fig. 14a). We compared the cells treated   (1) 2 (2b)], (c) [Ti (1) 2 (2c)], (d) [Ti (1) 2 (2d)]. Cell viability was assessed by MTT assay. Graphs represent mean (n = 8) with standard deviation of one representative experiment (n = 3). *p < 0.01 as calculated by a oneway ANOVA test followed by a Tukey post-test. Lines indicate IC 50 calculated taking as maximal effect wells with cells completely dead due to the high dose of cisplatin. For each dose, the equivalent dose of DMSO without a titanium compound was tested to verify the lack of toxicity caused by DMSO.  with the [Ti(1) 2 (2d)] versus the cells treated with the fluorescent ligand alone (2d). In both conditions, we observed fluorescent aggregates, reflecting the poor solubility of the compound in the medium. At a similar concentration (50 µM), the overall fluorescence was reduced with the titanium complex compared to the ligand alone, which is in full accordance with the conclusions drawn from the fluorimetric assay. At a higher magnification, we observed the presence of fluorescence inside the cells and around the nucleus labeled with the fluorescent nucleus dye DAPI (Fig. 14b). However, at this stage, we are not able to conclude if the complex has crossed the cellular membrane and liberated the diimine ligand into the cells or if the free ligand has been generated in the extracellular medium before entering into the cells. Further investigations are planned to elucidate that point.

Conclusions and perspectives
We have reported herein the synthesis, using a well-established procedure, of a series of octahedral TiO 4 N 2 -based complexes bearing substituted 2,2′-bipyrimidine ligands starting from [Ti (1) (1) 2 (2d)] complexes relative to the free nitrogen ligands. The explanation of the emission quenching has been provided by the examination of the HOMO and LUMO levels. We have also noticed a rather good hydrolytic stability for these Ti(IV)-based species that is correlated to the size of the polyaromatic substituents decorating the nitrogen ligand backbone. Interestingly, this stability is affected by the addition of an acid. Additionally, preliminary biological studies were carried out. The toxicity of these compounds has been evaluated, highlighting that the complexes and the nitrogen ligands are moderately toxic. Also, we observed fluorescence inside the cells when they were treated with the complex [Ti (1) 2 (2d)]. The longterm applications of the behaviour of this family of complexes in a medical or biological context are envisaged since pHresponsive systems are highly attractive for developing novel drug delivery systems or analytical tools. 42 However, before envisaging such applications, the solubility of these Ti(IV)based compounds in aqueous media must be significantly improved.

General procedures
Bruker Avance-300 and Avance-500 spectrometers were used for solution NMR spectroscopy analysis. Mass spectroscopy and elemental analyses were performed at the Service Commun d'Analyses, University of Strasbourg (France). The electrospray analyses were performed on Micro-TOF (Bruker) apparatus equipped with an electrospray (ES) source. The elemental analyses were performed on a Flash 2000 (Thermo Fisher Scientific) for C, H, and N elements. The analyzed mass peaks refer to the most intense peaks (unless mentioned otherwise in the discussion). 2-Chloro-4-phenyl-6 methyl-pyrimidine, triphenylphosphine, NiCl 2 ·6H 2 O, zinc powder, and t BuOK were used as received from Alfa-Aesar or Sigma Aldrich.
For the experiments conducted with DCl, the final concentrations were also 6 mM. Complex [Ti (1) 2 (2a)

Cell cultures
Human AGS gastric cancer cells were obtained from ATCC. AGS cells were manipulated and cultured in DMEM with 10% FCS (Dominique Dutcher™) and 1% penicillin + streptomycin (Sigma) at 37°C under a 5% CO 2 atmosphere as previously described. 43 Cell survival 5000 cells were seeded per well in 96-well microplates (Falcon Multiwell), 24 hours prior to any treatment. The complexes were applied for 48 h in fresh medium. MTT assay was performed as previously described by replacing the medium with a fresh medium supplemented with 5 mg L −1 MTT (Sigma) for 1 h. 44 The cells were lysed in isopropanol with 0.04 N HCl. Measurements were performed at 550 nm. IC 50 were calculated using as maximal value wells with complete loss of cell survival due to the high dose of cisplatin (50 µM).

Microscopy
The cells were grown on cover glasses in a growth medium and incubated for 4 hours with the complexes at various dilutions. The cells were then incubated with DAPI (Thermo Fisher) for 15 min. After incubation, the cells were observed directly under an epifluorescence microscope (Zeiss Apotome V2 axiozoom) at excitation 433 nm and emission 532 nm and 461 nm (DAPI).

X-ray crystallography
The X-ray diffraction data were collected on a Bruker Smart CCD diffractometer with Mo-Kα radiation (λ = 0.71073 Å). The structures were solved and refined using the Bruker SHELXTL Software Package using SHELXS-97 (Sheldrick, 2014) and refined by full matrix least-squares on F 2 using SHELXL-97 (Sheldrick, 2014) with anisotropic thermal parameters for all non-hydrogen atoms. 45 The hydrogen atoms were introduced at calculated positions and not refined (riding model). For the structures of compounds [Ti(1) 2 (2c)] and [Ti(1) 2 (2d)], we used the squeeze command because the solvents could not be identified.

Photophysical measurements
The UV-visible absorption spectra were recorded using a Perkin Elmer Lambda 650S spectrometer at room temperature with a 1 cm path cell. The luminescence spectra were recorded at room temperature using a Fluorolog FL3-22 spectrofluorometer (Horiba Jobin Yvon) equipped with a TBX-04 or an R928 detector. All of the luminescence measurements were performed on optically dilute solutions with spectrophotometrygrade dichloromethane, with a maximum absorbance of 0.08. For the luminescence steady-state measurements, the spectrometer was equipped with a 450 W xenon lamp excitation source and excitation and emission double-grating monochromators with bandpass set at 2 nm for the experiments. The spectra were corrected for the lamp, the monochromators, and the detector responses; blanks (solvent alone) were recorded and subtracted from the raw spectra, unless otherwise stated. The luminescence quantum yields were determined in spectrophotometry-grade dichloromethane solutions using either terphenyl in cyclohexane (Φ = 0.93, for compound 1-H 2 ) or quinine sulfate in aerated 0.05 M H 2 SO 4 (Φ = 0.55, for the other compounds) as luminescence standards. 37b