Luminescent platinum(II) complexes with functionalized N-heterocyclic carbene or diphosphine selectively probe mismatched and abasic DNA

The selective targeting of mismatched DNA overexpressed in cancer cells is an appealing strategy in designing cancer diagnosis and therapy protocols. Few luminescent probes that specifically detect intracellular mismatched DNA have been reported. Here we used Pt(II) complexes with luminescence sensitive to subtle changes in the local environment and report several Pt(II) complexes that selectively bind to and identify DNA mismatches. We evaluated the complexes' DNA-binding characteristics by ultraviolet/visible absorption titration, isothermal titration calorimetry, nuclear magnetic resonance and quantum mechanics/molecular mechanics calculations. These Pt(II) complexes show up to 15-fold higher emission intensities upon binding to mismatched DNA over matched DNA and can be utilized for both detecting DNA abasic sites and identifying cancer cells and human tissue samples with different levels of mismatch repair. Our work highlights the potential of luminescent Pt(II) complexes to differentiate between normal cells and cancer cells which generally possess more aberrant DNA structures.

The d 8 pincer Pt(II) complexes have been well documented to intercalate into matched DNA 1,2,4,11,42 . With intrinsic emission properties that are highly sensitive to the microenvironment, these types of Pt(II) complexes have been developed as luminescent DNA probes and have been identified as potent therapeutic agents for cancer treatment 14,17,[43][44][45] . However, metal complexes with bare planar coordination geometries likely exhibit non-discriminative intercalative binding interactions with DNA. In the crystal structures of mismatched DNA bound to Rh(III) 18,28 or Ru(II) 19 complexes, the planar extended p-conjugated ligands (with a width of up to 11.3 Å and a length of up to 9.4 Å) favour strong p-stacking interactions with nearby base pairs at mismatched sites, concomitantly with the ejection of mismatched base pairs, whereas the other auxiliary bipyridine ligands promote a groove binding interaction. We therefore hypothesized that a pincer Pt(II) complex coordinated with an out-of-plane bulky ancillary ligand may exhibit selective binding affinity towards mismatched DNA, as discussed in the next section. The bulky ancillary ligand can be designed to hamper the intercalation of the platinum complex with well-matched DNA while exhibiting high binding affinity with mismatched DNA. This difference is possible because the mismatched sites possess larger binding pockets. We have previously described a luminescent [Pt II (C^N^N)(NHC 2C4 )] þ (HC^N^N ¼ 6-phenyl-2,2 0 -bipyridine, NHC ¼ N-heterocyclic carbene, 1b in the current report) complex with an NHC ligand almost perpendicular to the C-deprotonated C^N^N plane 46

Results
Structure and emission responses to CC mismatched DNA. We synthesized and characterized a series of [Pt II (C^N^N)(NHC)] þ complexes with different alkyl chains, aromatic groups and hydrophilic alcohol moieties on NHC and/or on C-deprotonated C^N^N ligands ( Fig. 1; details in Supplementary Information). In addition, we examined the emission responses of complexes that bind mismatched and matched DNA ( Supplementary Fig. 1). The most unstable CC mismatched DNA was chosen in the primary screen because metalloinsertors display the highest binding to this type of mismatched DNA 29 . We found that as the steric hindrance of the NHC ligand increased, the binding of the  Table 2). For example, for the [Pt(C^N^N)(NHC)] þ type complexes of 1e ( À CH 3 ), 1f ( À C 2 H 5 ), 1g ( À C 3 H 7 ), 1c ( À C 4 H 9 ), 1h ( À C 5 H 11 ), 1i ( À C 6 H 13 ) and 1j ( À CH 2 Ph), which contain the same N 3 -benzyl group but different N 1 -substituents on the NHC ligands (Table 1), an increase in the length of the N 1 -alkyl was accompanied by a marked decrease in emission enhancement towards matched DNA (I M ), from 38.9-fold to o3-fold.
Although increasing the chain length from the N 1 -methyl of 1e to the N 1 -ethyl of 1f led to a decrease in emission enhancement in the case of mismatched DNA (I MM ), the emission intensity of 1g, which bears N 1 -propyl, was much stronger than that of 1f, suggesting that N 1 -propyl favours the groove binding of the Pt(II) complex to DNA. This may enhance the tightness of the insertion binding mode, as proposed. Further lengthening the alkyl chain to n-butyl, n-pentyl and n-hexyl gradually decreased the binding with mismatched DNA. Comparing the I MM /I M ratios, 1c, containing N 1 -(n-butyl) and N 3 -benzyl substitutions on the NHC ( Fig. 1) showed the most significantly different emission responses between CC mismatched DNA and matched DNA. We characterized the structure of this complex by 1 H-1 H correlation spectroscopy (COSY) and nuclear Overhauser effect spectroscopy (NOESY) nuclear magnetic resonance (NMR; Supplementary  Fig. 2). As shown in Fig. 3a,c, the emission intensity of 1c (l max ¼ 535 nm) in a Tris-buffered solution increased by 26.4-fold in the presence of 1 equivalent of a hairpin DNA oligomer with a CC mismatch 29 (Fig. 3c) Fig. 8a,b). 1 H NMR and two-dimensional (2D) NOESY, total correlation spectroscopy (TOCSY) and COSY NMR experiments were performed using a self-complementary DNA oligonucleotide, 5 0 -C 1 G 2 G 3 A 4 C 5 T 6 C 7 C 8 G 9 -3 0 , containing CC mismatches in D 2 O (phosphate buffer, pH 6.1; Fig. 4a-d; Supplementary Figs 9 and 10) 48 . In the range of 6.8-8.0 p.p.m., the aromatic 1 H signals of C5 and A4 show substantial shifts, whereas those of the other bases remain nearly unshifted in the presence of 1c ( 1 H NMR in Fig. 4a). In addition, in the 2D H1 0 Â aromatic NOESY spectrum of 1c with DNA at a 1:1 ratio, a sequential NOESY walk (Fig. 4d), which showed a marked shift at A4 and C5 compared with DNA only (Fig. 4b), was found in combination with cytosine H6/H5 TOCSY (Supplementary Fig. 9) and H1 0 Â H2 0 -H2 00 TOCSY (Fig. 4c 1a  1b  1c  1d  1e  1f  1g  1h  1i  1j  1k  1l  1m  1n  1o  2  2a  3  4  5a  5b  5c   Complexes   1a  1b  1c  1d  1e  1f  1g  1h  1i  1j  1k  1l  1m  1n  1o  2  2a  3  4 5a 5b 5c   29 . Indeed, a favourable binding mode between 1c and CC mismatched DNA has also been identified in molecular docking experiments ( Supplementary Fig. 12), consistently with the insertion of the Pt(C^N^N) plane between base pairs and the favourable minor-groove binding interaction of the N 1 -(n-butyl) and N 3 -benzyl moieties. Complex 1d, which contains an N^C^N instead of a C^N^N ligand (Fig. 1), exhibited 4.5-fold higher emission responses towards 1 equivalent of CC mismatched DNA than matched DNA ( Supplementary Fig. 13).
The emission responses of 1c towards all other types of DNA mismatches, including CT, AC, TT, AG, TG, AA and GG mismatches, all of which are more thermodynamically stable than CC mismatches 49 , were tested (Fig. 3c). However, no significant difference in emission responses between these mismatched DNA and matched DNA was observed. We also examined the emission responses of 1c to CC mismatched DNA having 16 different adjacent base pairs (Fig. 3d). Complex 1c displayed higher emission responses to all the CC mismatched DNA sequences than the cognate matched DNA, with adjacent 5 0 -ACA-3 0 revealing the highest emission enhancement (15.0-fold higher than matched DNA). These experiments suggest that 1c is a selective probe for CC mismatched DNA.
Targeting thermodynamically more stable DNA mismatches. For recognition of the more stable DNA mismatches, an increase in the binding affinity of the Pt(II) complex with DNA is required. The dinuclear analogues of 1c, which contain an additional positive charge and a larger molecular dimension, are expected to have higher DNA-binding affinities as a result of stronger ionic bonding interactions and possibly a better geometry to fit into the pocket at the mismatched site. Therefore, we examined the emission responses of a series of doubly charged dinuclear platinum complexes with two Pt II (C^N^N) moieties linked by bridging bis(NHC) or diphosphine ligands (Fig. 1) in the presence of hairpin mismatched and matched DNA. Among the dinuclear Pt(II) complexes, [Pt II 2 (C^N^N) 2 (m-dcpm)] 2 þ 2 ( Fig. 1; the X-ray crystallographic structure is depicted in   (Fig. 2). Figure 5a and Supplementary Fig. 15 illustrate that the emission intensity of 2 at l ¼ 634 nm was elevated 50-fold upon the addition of only 0.5 equivalents of hairpin DNA containing a CC mismatch but enhanced only 8.2-fold in the presence of the same amount of matched DNA, that is, a 6.1-fold difference in emission responses. The binding affinity of 2 with DNA was examined by ITC. The interaction of 2 was much more exothermic with CC mismatched DNA than with matched DNA (Supplementary Fig. 16). Notably, the K d of the binding of 2 to CC mismatched DNA was found to be 7.0 ± 2.1 mM, which is 7.4-fold higher than that of 1c. The K d of the binding of 2 with matched DNA was negligible. In addition, the melting temperature (T m ), as determined by the change in absorbance (OD 260nm ) of DNA was increased by 6.8°C for mismatched DNA in the presence of 2 at an equimolar concentration ( Supplementary Fig. 17a). However, only a 0.5°C increase in melting temperature was found for matched DNA in the presence of 2 ( Supplementary Fig. 17b).
Complex 2 also exhibited higher emission responses to CC mismatched DNA containing 16 different adjacent sequences (Fig. 5b), with a nearby 5 0 -TCT-3 0 sequence generating the highest fold increase in emission (13.5-fold higher than matched DNA). In addition to its response to CC mismatches, 2 also showed higher emission responses to several other types of mismatched DNA over matched DNA, and the emission increased 4.9-fold for CT, 3.8-fold for AC and 3.4-fold for TT compared with matched DNA but no higher for other mismatches (AG, 1.8-fold; GT, 1.4-fold; AA, 1.4-fold; GG, 0.8-fold; Fig. 5a). Such differences in the emission responses between mismatched and matched DNA partially correlated with the thermodynamic stability of the DNA base pairs, CCrACrCToAArTToGArGToGG 49 , suggesting an insertion binding mode 50 . The emission response to AC was slightly lower than that to CT, and the response to AA was lower than that to TT; these observations are not fully consistent with the trend of thermodynamic stability of mismatched DNA. The lower emission response towards A-containing AC and AA mismatches may have been caused by the weak coordination of the N7 of purine in adenine at the mismatch site to the Pt complex, leading to decrease of emission intensity (note that the pyrimidine N3 in C and T is not accessible to the Pt(II) complexes due to steric hindrance). We performed hybrid quantum mechanics/molecular mechanics (QM/MM) calculations to study the mode of binding between 2 and mismatched DNA. As shown in Fig. 5c, 2 bound to mismatched DNA through an insertion binding mode at the minor groove (this result is also consistent with the results of a [Cu(phen) 2 ] 2 þ quenching experiment, Supplementary Fig. 18); the C^N^N planes interacted with nearby GC and AT base pairs by p-stacking interactions (inter-planar distance between the C^N^N plane and the base pairs of B3.2 Å). The QM/MM simulation also indicated that the dinuclear structure better fit the mismatched sites, which have larger pockets (the inter-planar separation between nearby base pairs at mismatched sites can be 49 Å), but not the matched site, which contains less space between base pairs.
We also tested the emission responses of 2 to dsDNA of poly(dC)-poly(dG), poly(dC-dG) 2 , poly(dA)-poly(dT) and poly(dA-dT) 2 ; single-stranded DNA of oligo(dT) 20 and oligo(dC) 20 ; dsRNA and bovine serum albumin, all of which were observed to have at least 4.5-fold lower emission intensity than that of CC mismatched DNA (Fig. 5d). Notably, the emission intensity of 2 with RNA containing a CC mismatch was 3.3-fold higher than that with matched RNA (Fig. 5d) Fig. 19). The two Pt(II) complexes were weakly emissive in buffer solutions, but their emission intensities were significantly higher in acetonitrile, even under air (11-fold for 1c, 9-fold for 2, Supplementary Fig. 20a,b; the emission intensity was found to increase with decreasing donor strength of various solvents, as shown in Supplementary Fig. 20c). Thus, the increased emission could not be accounted for by the shielding of oxygen quenching upon mismatched DNA binding to the Pt(II) complexes. In addition, variations of pH from 5 to 11 did not evoke a significant change in emission intensity ( Supplementary Fig. 21). We propose that the shielding of quenching by a buffer solution and/or a more ordered structure of the Pt(II) complex after mismatched DNA binding could account for the emission enhancement 51,52 . We used time-resolved emission spectroscopy to examine the binding between the Pt complexes and DNA. As shown in Supplementary Fig. 22, the emission of 2 in the presence of CC mismatches had a lifetime of 2.3 ms, whereas in the presence of matched DNA, the lifetime is composed of a major portion of short decay (o100 ns) and a small amount of 1.2-ms decay. The difference in emission intensity, I mismatch /I match , was observed to increase with increasing decay time, from 2.3-fold at t ¼ 0 to 11.8fold at t ¼ 10 ms (Supplementary Fig. 23). Therefore, time-resolved emission spectroscopy may be a useful strategy to achieve higher sensitivity than steady-state emission spectroscopy for the detection of mismatched DNA using luminescent Pt(II) complexes.
Targeting abasic DNA. The abasic (AP) site, in which a purine or a pyrimidine base is missing in a nucleotide, represents another type of thermodynamically unstable DNA defect 50 . In the presence of a 27-mer dsDNA oligomer containing an abasic site 50 (AP_C with a complementary base of C), the emission intensity of an equimolar amount of 1c at l max ¼ 535 nm exhibited a 32.2-fold increase ( Supplementary Fig. 24). This elevation in emission was even higher than that observed for the binding of 1c with CC mismatched DNA with similar sequences (26.4-fold). For comparison, only a 4.7-fold increase in emission was detected when 1c was incubated with normal 27-mer dsDNA with a cognate sequence. In the presence of other abasic sites of AP_A, AP_T and AP_G, the emission intensities of 1c were increased by 14.8, 21.1 and 14.4 times, respectively. Complex 2 showed 17.0-29.3-fold increases in emission intensity in the presence of different abasic DNA sequences, but only a 6.1-fold increase in emission intensity with normal DNA ( Supplementary  Fig. 25). We also measured the effects of 2 on the T m of DNA bearing an abasic site (Supplementary Fig. 26). The melting temperature of abasic DNA in the presence of 2 was increased by 5.3°C, whereas the T m of the similar DNA sequence without an abasic site was not affected, indicating a high selectivity of 2 towards abasic DNA over matched DNA. Thus, these Pt(II) complexes could be used to probe DNA abasic sites. Such a property is also found in Rh(III)-based metalloinsertors, in which the metal complex binds to thermodynamically unstable DNA structures through an insertion binding mode 50 .
Detection of genomic DNA base pair defects. The human colorectal carcinoma cell line HCT116 is known to be MMR deficient, owing to mutations in the MMR-associated hMLH1 gene, whereas HCT116 cells containing a transferred, normal chromosome 3 (denoted as HCT116N) restore MMR activity 53 . We found that 2 showed significantly stronger emission at l max ¼ 634 nm in the presence of genomic DNA isolated from HCT116 cells compared with that from HCT116N cells (Po0.01, n ¼ 3; Fig. 6a; Supplementary Fig. 27). We then measured and compared the emission responses of 2 with several types of cancer cells with different mutation rates. The human cancer cell lines DU145, HCT116, HCT116N and SW480 have mutation rates (mutations per cell per generation) in the hprt gene locus of 529 Â 10 À 7 , 77.5 Â 10 À 7 , 5.9 Â 10 À 7 and 0.75 Â 10 À 7 , respectively 54 , and the relative emission intensities of 2 (5 mM) at 5.1 mM of DNA are 7.7, 5.8, 3.7 and 2.3, respectively ( Fig. 6a shows an increase of emission intensity at 0-5.1 mM of genomic DNA from different cancer cells). Notably, the plot of the fold of emission increase versus the lg (mutation rate) yields a linear fit with R 2 ¼ 0.99, as shown in Fig. 6b, indicating a strong correlation between emission response and mutation rate. We next tested whether 2 could be used to selectively detect mismatched DNA in permeabilized cell samples by fluorescence microscopy. Because the Pt(II) complexes are lipophilic and tend to accumulate in cytoplasmic compartments, the HCT116 and HCT116N cancer cells were permeabilized by digitonin and then washed to minimize the fluorescence background due to interactions of the Pt(II) complexes with cytoplasmic contents. As shown in Fig. 6c-f, after labelling with 2, the emission signal of the HCT116 cells was significantly stronger than that of the HCT116N cells. In addition, complex 2 emitted significantly stronger signals in cancerous HCT116 cells compared with those in the immortalized normal human colon mucosal epithelial cell line NCM460 (ref. 55; Supplementary  Fig. 28); complex 2 also consistently exhibited significantly lower emission responses in the presence of the DNA isolated from NCM460 cells compared with that from HCT116 cells (Supplementary Fig. 27). The differences in emission intensity were not caused by variations in cellular uptake of Pt(II) complexes as shown by inductively coupled plasma mass spectrometer measurements (Supplementary Fig. 29). For comparison, no significant differences in the emission properties were found between the two cell lines labelled with the general DNA intercalator EB (Supplementary Fig. 30).
We further examined the ability of the Pt(II) complex to detect mismatched DNA of primary human tumours of colon cancer which is frequently characterized by MMR deficiency. DNA samples were extracted from colorectal adenocarcinoma (well to moderately differentiated) and the surrounding normal tissues of a patient (see details in the Methods section). The results of the emission responses of 2 (5 mM) to the DNA from the tumour and normal tissues are shown in Fig. 6g. In the presence of 0.3-1.9 mM of DNA extracted from normal colon tissues, weak emission responses of up to 1.3-fold of the initial emission intensity were detected; in contrast, the emission intensity increased in a concentration-dependent manner in the presence of 0-1.6 mM of DNA extracted from colon tumour tissues, showing up to a 5.5fold increase in emission at 1.9 mM. Thus, there was a difference between the emission responses of 2 to the DNA of human colon tumour tissue and the normal adjacent tissue.
In conclusion, we have identified two classes of luminescent Pt(II) complexes that can be used for the detection of DNA mismatches and abasic DNA sites. These Pt(II) complexes are able to differentiate between cells with different levels of MMR activity. Importantly, the differential emission responses of the Pt(II) complexes to DNA from human colon cancer tissues and normal colon tissue suggest that the Pt(II) complexes are promising candidates for tumour diagnosis via simple emission spectroscopy measurements of DNA extracts. The selective targeting of DNA mismatches also lends the complexes therapeutic potential for cancer treatment 21 ; and they may be used as scaffolds in the design of new anti-cancer agents for the treatment of MMR-deficient cancers.  Quantum mechanics/molecular mechanics. The structures of mismatched DNA and complex 2 were taken from the literature (PDB ID: 2O1I 56 , unit Å) and the X-ray crystal analysis (CCDC 1061095), respectively. In general, molecular docking studies were performed using ICM-pro software to obtain the initial structure of DNA-complex 2 adducts 57 ; subsequently, the binding between the DNA and complex was further optimized using the QM/MM approach and NWChem software 58 . The quantum part (QM) included complex 2 (total 133 atoms), and the rest of the system was modelled at the MM level. The QM region was treated at the density functional theory level using the M06L method 59 . The 6-31G* Pople basis set 60 was employed for C, H, N and P atoms, and for the Pt atom, we applied the LANL2DZ 61 basis set with effective core potential. The mismatched DNA was described with the AMBER parm99 force field 62 . Van der Waals parameters for the Pt atom were taken from the literature 63 . The docked structure was solvated in a 65-Å cubic box with 8,830 classical SPC/E water molecules. Twenty-two-positive (Na þ ) counterions were added to neutralize the charges in the system. The structure was optimized using the BFGS algorithm for the QM part and the steepest descent algorithm for the MM part. The optimization of these two regions (QM and MM) was alternated until self-consistency was reached.
DNA extraction from cancer cells or tumour tissues. HCT116 and HCT116N cell lines were generously provided by Thomas A. Kunkel (NIH, USA). The human tissue samples were obtained from Jinan University, Guangzhou, China, and the experimental uses were authorized by the Human Experimentation Ethics Committee of Jinan University. Genomic DNA from cells or tissues was isolated using DNAzol Reagent (Thermo Fisher ) according to the manufacturer's protocol. For DNA extraction from cell lines, 1 ml of DNAzol Reagent was added to 1-3 Â 10 7 cells grown on 10-cm 2 culture plates. Then, the cells were lysed by agitating the culture plate and gently pipetting the lysate into a microfuge tube. For the tissue sample, 1 ml of DNAzol Reagent was added to B50 mg of tissues. Then, the tissues were homogenized by applying three strokes in the homogenizer. The homogenate was sedimented for 10 min at 10,000g at 4°C. Following centrifugation, the resulting viscous supernatant was transferred to a fresh tube, mixed with 0.5 ml of 100% ethanol by inversion and stored at room temperature for 1-3 min. Then, the DNA precipitate was washed twice with 1.0 ml of 75% ethanol. The DNA was airdried and dissolved in 200 ml of 8 mM NaOH, and the pH was adjusted to 7.5 by the addition of HEPES. The emission spectrum of a solution of complex 1c or 2 (5 mM) in buffer solution (50 mM NaCl and 2 mM Tris, pH 7.5) was recorded and titrated with increasing amounts of the isolated DNA. The emission spectra were recorded after equilibration for 3 min in each addition.
Fluorescence microscopy examination. Cells (HCT116 or NCM460) were seeded in glass bottom culture dishes (35 mm) at a density of 1.0 Â 10 5 ml À 1 . After 12 h, the dishes were placed on ice, washed twice with 1 ml of cold PBS and then incubated in 1 ml of permeabilization buffer (20 mM HEPES, pH 7.5, 110 mM potassium acetate, 5 mM magnesium sulphate and 250 mM sucrose) containing 40 mg ml À 1 digitonin for 30 min. After removal of the digitonin solution, the dishes were washed three times with cold permeabilization buffer for 1, 5 and 10 min. The cells were then incubated with complex 2 (10 mM) in transport buffer (20 mM HEPES, pH 7.3, 110 mM potassium acetate, 5 mM sodium acetate, 2 mM magnesium sulphate and 250 mM sucrose) for 15 min. Cells were directly subjected to fluorescence imaging without removing the transport buffer. For co-culture experiments, HCT116 cells were first seeded in culture dishes for 12 h. Next, the dishes were treated with a non-toxic concentration of Hoechst 33342 (1 mg ml À 1 ) for 30 min and washed twice with PBS. Then, NCM460 cells were seeded in the same dishes for another 12 h. Finally, the co-cultured cells were incubated in permeabilization buffer, stained with complex 2 and subjected to fluorescence imaging as aforementioned.