Platinum(II) N-heterocyclic carbene complexes arrest metastatic tumor growth

Significance

Metastasis is a common cause of cancer mortality yet a daunting challenge in oncology. We have discovered a cytotoxic platinum(II) N-heterocyclic carbene complex that not only suppressed primary tumor growth in mice but also impeded metastatic growth of breast cancer and melanoma with long-term tolerance attributable to its eventual biotransformation into less-toxic, excretable products. It directly engages the cytoskeletal protein vimentin that plays pivotal roles in cancer stemness, survival, and invasion. It binds vimentin noncovalently via the distinct structural scaffold with a dissociation constant of 1 µM as determined by surface plasmon resonance, NMR, and molecular dynamics simulations, revealing the significance of harnessing noncovalent interactions of metal compounds with important molecular targets in the development of new anticancer metal medicines.

Abstract

Vimentin is a cytoskeletal intermediate filament protein that plays pivotal roles in tumor initiation, progression, and metastasis, and its overexpression in aggressive cancers predicted poor prognosis. Herein described is a highly effective antitumor and antimetastatic metal complex [Pt^II(C^N^N)(NHC^2Bu)]PF₆ (Pt1a; HC^N^N = 6-phenyl-2,2′-bipyridine; NHC= N-heterocyclic carbene) that engages vimentin via noncovalent binding interactions with a distinct orthogonal structural scaffold. Pt1a displays vimentin-binding affinity with a dissociation constant of 1.06 µM from surface plasmon resonance measurements and fits into a pocket between the coiled coils of the rod domain of vimentin with multiple hydrophobic interactions. It engages vimentin in cellulo, disrupts vimentin cytoskeleton, reduces vimentin expression in tumors, suppresses xenograft growth and metastasis in different mouse models, and is well tolerated, attributable to biotransformation to less toxic and renal-clearable platinum(II) species. Our studies uncovered the practical therapeutic potential of platinum(II)‒NHC complexes as effective targeted chemotherapy for combating metastatic and cisplatin-resistant cancers.

Cancer invasion and metastasis are among the major causes of recurrence and mortality (1). Cancer metastasis begins with epithelial–mesenchymal transition (EMT), where epithelial cancer cells are transformed into mesenchymal cells equipped with motility, invasion, and stem cell-like properties (2–4). Vimentin (VIM), a canonical biomarker of EMT, is an intermediate filament protein playing pivotal roles in cytoskeletal architecture, organelle positioning, cell migration, and signaling; its overexpression in aggressive epithelial cancers of lung, breast, and prostate origin, and malignant melanoma, were reported to be associated with the acquisition of migratory and invasive phenotypes (5). The enhanced cell motility and invasiveness by VIM drive cancer cells’ spreading from primary tumor to distant organs through the bloodstream or lymphatic system, resulting in a metastatic cascade (6, 7). Emerging evidence showed that high levels of VIM predicted a poor prognosis, particularly in patients with nonsmall cell lung cancer and triple-negative breast cancer (8–10). In view of its clinical significance, targeting VIM is an appealing strategy to combat advanced metastatic cancers that do not respond well to traditional chemotherapy or antihormonal therapy. Nevertheless, inhibitors/disruptors of VIM are scarce in the literature. Moreover, studies on the biomolecular interactions between VIM and its binding partner(s) are underdeveloped.

Several anticancer organic compounds have been reported to target VIM and display in vitro anticancer activities (11–14), but without mentioning their in vivo antimetastatic activity. Herein described are studies on the binding interaction of [Pt^II(C^N^N)(NHC^2Bu)]PF₆ (Pt1a, where HC^N^N = 6-phenyl-2,2′-bipyridine and NHC = N-heterocyclic carbene) with VIM, and the associated anticancer properties. This complex has a scaffold with NHC ligand being orthogonal to the pincer [Pt^II(C^N^N)] plane; such a unique structural scaffold renders Pt1a able to effectively bind to VIM noncovalently, in contrast to covalent modification by the majority of known VIM-targeting compounds. Our work features a rare example of an antitumor metal complex that engages VIM and shows promising in vivo antitumor activity in multiple mouse models bearing advanced metastatic tumors. This work also demonstrates an example of effective anticancer compound development by taking advantage of unique structural scaffolds of metal complexes for noncovalent interactions with an important molecular target of cancer as also revealed by recent work of Meggers, Che, Casini, and coworkers (15–20).

Results

Synthesis, Spectroscopic, and Self-Assembly Properties of Pincer Platinum(II) Complexes.

The structures of the pincer Pt(II) complexes (Pt1a–Pt6) are depicted in Fig. 1. Pt1a–Pt3 were reported previously (21, 22), while Pt4–Pt6 are new complexes which were synthesized by refluxing cyclometalated Pt(II) precursors with imidazolium or benzimidazolium salts in acetonitrile and obtained as yellow or orange solids. In biological buffer solution and in the presence of glutathione, these Pt(II) complexes are stable for 72 h (SI Appendix, Figs. S1 and S2). As revealed by ultra-performance liquid chromatography/mass spectrometry (UPLC/MS) analysis, no new peak was found for Pt1a in culture medium for 72 h (SI Appendix, Fig. S3). The X-ray crystal structure of Pt4 features a distorted planar molecular geometry with the NHC ligand at a dihedral angle of 68.5° to the [Pt^II(SC^N^N)] plane (Fig. 2A and SI Appendix, Tables S1–S3). The photophysical data of Pt1a–Pt6 are summarized in SI Appendix, Figs. S4 and S5 and Table S4. The complexes display cyan to orange emission in degassed dichloromethane at 298 K with lifetimes (τ) of 0.5 to 10.0 µs and quantum yields of 0.8 to 36.3%. As revealed by transmission electron microscopy (TEM) images (Fig. 2B and SI Appendix, Fig. S6A), in aqueous solutions containing 1% CH₃CN the Pt(II)–NHC complexes (Pt1a–Pt6) self-assembled into spherical-like particles with diameters of 40 to 200 nm. From size distribution of nanoparticle tracking analysis (Fig. 2C), an average hydrodynamic diameter (D_h) of nanostructures formed by Pt1a was 128.1 ± 1.7 nm. The particle size of Pt1b bearing the N-methyl substituted NHC (D_h: 147.2 ± 1.2 nm) was larger than that of Pt1a with N-butyl chains. In RPMI 1640 culture medium containing 0.5% dimethyl sulfoxide (DMSO), Pt1a formed larger nanostructures with average D_h of 173.3 ± 3.1 nm (SI Appendix, Fig. S6B) compared with that formed in aqueous solution.

Fig. 1.

Chemical structures of different pincer cyclometalated platinum(II) N-heterocyclic carbene complexes.

Fig. 2.

(A) X-ray crystal structure of Pt4 with omission of hydrogen atoms and counteranion. Thermal ellipsoids are shown at 50% probability level. (B) TEM images of self-assembled structures of Pt1a and Pt1b (1 × 10⁻⁴ M) in aqueous solution containing 1% CH₃CN. (Scale bar, 200 nm; Inset, 100 nm.) (C) Histograms of hydrodynamic diameters of particles formed by Pt1a and Pt1b in aqueous solution containing 1% CH₃CN.

In Vitro Anticancer Properties and Metabolite Analysis.

Previously we reported the low IC₅₀ values (half-maximal inhibitory concentration) of the pincer platinum(II) complexes on cancer cell proliferation (21). In this work, a more extended list of cancerous cell lines of lung, breast, colorectal, and ovarian origin together with the noncancerous or cisplatin-resistant variant are included in antiproliferation assays as measured by naphthol blue-black staining (Table 1 and SI Appendix, Fig. S7) and validated by live/dead cell assays (calcein-AM/PI staining; SI Appendix, Figs. S8 and S9). Compared to cisplatin, the Pt(II) complexes (Pt1a–Pt6), except Pt3, exhibited higher anticancer activity with IC₅₀ in a range of 0.03 to 1.49 μM. Particularly, Pt1a displayed a higher antiproliferative activity in NCI-H460 lung cancer cells compared to normal lung fibroblasts (CCD-19Lu) with around 30-fold difference in IC₅₀ values. Changing the counteranion from PF₆⁻ as in Pt1a to Cl⁻ in Pt1c showed a negligible difference in cell growth inhibition. However, Pt3, being a dicationic charged complex, showed IC₅₀ values higher by an order of magnitude in comparison to the monocationic analogs. Pt1b bearing two methyl groups was three- to fivefold less potent than Pt1a with two butyl chains for all the cancer cell lines examined. The Pt(II) complexes with either thiophene (Pt4) or selenophene (Pt5) on the C^N^N scaffold, benzimidazole-derived NHC ligand (Pt6), or N^C^N ligand (Pt2) displayed IC₅₀ values comparable to that of Pt1a. It is worth noting that complexes (Pt1a–Pt6) displayed a comparable antiproliferation effect against both cisplatin-sensitive (A2780) and -resistant (A2780cis) ovarian cancer cell lines, whereas the potency of suppression for cisplatin was attenuated by 44-fold in IC₅₀ on A2780cis compared to the parental cells.

Table 1.

In vitro antiproliferative activity (IC₅₀; micromolar, 72 h) of Pt1a–Pt6 and cisplatin toward different human cancerous and normal cell lines

Complex	NCI-H460 (lung cancer)	MDA-MB-231 (breast cancer)	HCT116 (colorectal cancer)	A2780 (ovarian cancer)	A2780cis (ovarian cancer, cisplatin- resistant)	CCD-19Lu (lung fibroblast)
Pt1a	0.16 ± 0.01	0.20 ± 0.01	0.16 ± 0.01	0.06 ± 0.01	0.10 ± 0.01	4.74 ± 0.28
Pt1b	0.51 ± 0.06	1.06 ± 0.05	0.79 ± 0.08	0.21 ± 0.02	0.32 ± 0.01	9.52 ± 0.61
Pt1c	0.17 ± 0.01	0.26 ± 0.03	0.23 ± 0.03	0.14 ± 0.04	0.15 ± 0.06	5.18 ± 0.34
Pt2	0.31 ± 0.02	0.32 ± 0.05	0.30 ± 0.03	0.28 ± 0.02	0.32 ± 0.02	4.85 ± 0.51
Pt3	2.44 ± 0.34	3.26 ± 0.45	4.01 ± 0.35	9.83 ± 0.84	10.37 ± 0.84	>100
Pt4	0.18 ± 0.01	0.21 ± 0.02	0.16 ± 0.01	0.07 ± 0.01	0.10 ± 0.01	4.94 ± 0.32
Pt5	0.17 ± 0.01	0.21 ± 0.02	0.15 ± 0.01	0.05 ± 0.01	0.08 ± 0.01	4.81 ± 0.13
Pt6	0.13 ± 0.01	0.15 ± 0.01	0.11 ± 0.01	0.03 ± 0.002	0.06 ± 0.004	3.26 ± 0.17
Cisplatin	8.07 ± 0.80	10.24 ± 0.86	9.50 ± 0.05	0.32 ± 0.01	14.24 ± 0.95	38.21 ± 1.75

The in vitro metabolism of Pt1a, Pt2, Pt3, and Pt6 was studied with rat liver microsome using UPLC/MS analysis, which is detailed in SI Appendix. Compared to Pt1a, Pt2 was shown to generate a similar number of metabolites. Dicationic Pt3 displayed no observable biotransformation and Pt6 containing benzimidazole-derived NHC ligand produced much less metabolites (SI Appendix, Fig. S10). The biotransformation of Pt1a to oxidized (M1–M5) and glucuronidated (G1–G3) platinum(II) species mainly involved the hydroxylation of the N-butyl chain of the coordinated NHC ligand (phase I reaction) and glucuronide conjugation to corresponding hydroxylated forms (phase II reaction) (SI Appendix, Figs. S11‒S15). In vivo metabolism of Pt1a in a BALB/c mouse model revealed that oxidized species such as M1 and M2 were found as the principal monohydroxylated metabolites in the mouse urine after single-dose intravenous administration of Pt1a for 24 h (SI Appendix, Fig. S15A). On the other hand, dihydroxylated M5 and glucuronidated (G1 and G3) metabolites were also found in urine, in which G1 was the major glucuronide species. These excreted metabolites can also be detected by UPLC/MS at 2 h after intravenous injection of Pt1a (SI Appendix, Fig. S15B). The proposed metabolites were synthesized independently, with chemical structures, IC₅₀ values, cellular uptake, and lipophilicity depicted in Fig. 3. The antiproliferative activity of the metabolites toward both cancer and normal cell lines are generally less effective than Pt1a (Fig. 3B and SI Appendix, Fig. S23 and Table S12), and in particular the IC₅₀ of dihydroxylated (M5) and glucuronidated metabolites (G1) are greater than 100 µM in most of the cell lines tested. These findings can be attributed to higher cellular uptake and lipophilicity (log P) (Fig. 3C) of Pt1a and M1–M4, while M5 and G1 exhibited the lowest cell penetration and log P. Additionally, Pt1a exhibited time-dependent accumulation in NCI-H460 cancer cells and displayed significantly twofold higher uptake in lung cancer cells compared to that in normal lung fibroblasts (SI Appendix, Figs. S24 and S25).

Fig. 3.

Metabolite analysis of Pt1a. (A) Chemical structures of the proposed metabolites (M1–M6 and G1) of Pt1a identified from microsomal biotransformation reactions and/or renal excretion. (B) In vitro cytotoxicity of Pt1a and synthesized M1–M6 and G1 on NCI-H460 lung, MDA-MB-231 breast cancer cells, and CCD-19Lu normal lung fibroblasts. (C) Correlation of cellular uptake (hatched bar) in NCI-H460 cells, and lipophilicity (square) of Pt1a and different synthetic metabolites.

Treatment of NCI-H460 or MDA-MB-231 cells with Pt1a (0.5 µM) for 72 h triggered over 70% cancer cell death, mainly at late apoptosis (Fig. 4A and SI Appendix, Fig. S26), assayed by annexin V-FITC/PI staining. Cell cycle analysis revealed that Pt1a-treated cells arrested in the G0/G1 phase (SI Appendix, Fig. S27). The luminescence of Pt1a was found to be mainly localized in cytoplasm and the mitochondria of NCI-H460 cells with a Pearson’s correlation coefficient of 0.77 (SI Appendix, Fig. S28) (21). Flow cytometric analysis of the mitochondrial membrane potential (Δψ_m) using JC-1 dye revealed that Pt1a treatment led to a dramatic reduction in the ratio of red (JC-1 aggregate) to green (JC-1 monomer) fluorescence intensity (SI Appendix, Fig. S29), indicating mitochondrial membrane depolarization. Pt1a treatment augmented the intracellular oxidant levels as revealed by a 7.6-fold increase in the fluorescence intensity of a 2ʹ,7ʹ-dichlorofluorescein diacetate probe compared with vehicle control (SI Appendix, Fig. S30).

Fig. 4.

In vitro anticancer activities of Pt1a. (A) Flow cytometric analysis of apoptotic cell death of NCI-H460 cells treated with Pt1a (0.5 µM) for 72 h. (B) Wound-healing assay of NCI-H460 lung cancer cells treated with Pt1a (1 µM) for 24 h. (Scale bar, 100 µm.) (C) Matrigel invasion assay of MDA-MB-231 breast cancer cells after Pt1a treatment (3 µM) for 24 h. (Scale bar, 100 µm.) (D) Representative images of secondary (self-renewing) tumor spheres derived from NCI-H460 lung or MDA-MB-231 breast cancer cells after Pt1a treatment. (Scale bar, 50 µm.) (E) Representative flow cytometry dot plot analysis of CD133 expression of NCI-H460 cells after Pt1a treatment (0.1 µM) for 72 h. Data are presented as mean ± SD (n = 2; Student’s t test, *P < 0.05, compared with DMSO vehicle control). (F) Tube-formation assay of MS-1 endothelial cells treated with Pt1a (1 µM) for 2 h. (Scale bar, 100 µm.)

In Vitro Antimigratory, Antiinvasive, and Antiangiogenic Properties and Inhibition of Tumorsphere Formation.

To determine whether Pt1a can inhibit cancer cell motility, a wound-healing scratch assay was performed. As depicted in Fig. 4B, compared with vehicle-treated NCI-H460 cells showing almost complete migration into the wound area after 24 h, the migratory ability of Pt1a-treated cells was found to be significantly impeded by 79.4%. The antiinvasive property of Pt1a was evaluated by Matrigel transwell assay using a highly invasive MDA-MB-231 breast cancer cell line. Compared to the cells exposed to DMSO vehicle (Fig. 4C), Pt1a treatment reduced the number of MDA-MB-231 cells invading through the Matrigel-coated membrane by 46%. Under stem-cell-selective culture conditions, Pt1a treatment reduced the number of self-renewed tumor spheres derived from NCI-H460 and MDA-MB-231 cells by 92% and 83%, respectively (Fig. 4D). Pt1a treatment also reduced the stemness-associated populations of NCI-H460 cells (CD133⁺) from 1.3% (vehicle) to 0.4% (Fig. 4E). In contrast, cisplatin treatment resulted in enrichment of CD133⁺ NCI-H460 cells (to 16.5%) (SI Appendix, Fig. S31). The antiangiogenic activity of Pt1a was examined by a tube-formation assay using MS-1 endothelial cells (Fig. 4F). In the vehicle control group, MS-1 cells cultured on Matrigel were interconnected to form a branched network. In the presence of Pt1a at a subcytotoxic concentration (SI Appendix, Fig. S32) the formation of tubular structure was markedly reduced.

In Vivo Tumor Growth Inhibition of Tumor Xenograft and Lung Metastatic Models.

The in vivo antitumor efficacy of Pt1a was first investigated in comparison with cisplatin using NCI-H460 nonsmall cell lung cancer xenografted nude mouse model. The mice were intravenously administrated with Pt1a (3 mg/kg) or cisplatin (1.2 mg/kg) of equivalent platinum dosage. As shown in SI Appendix, Fig. S33 A and B, both Pt1a and cisplatin treatment in a 14-d period effectively suppressed tumor growth. However, the body weight of the cisplatin-treated mice kept declining after the third injection at day 4 and its difference from vehicle control was significant at day 14 (SI Appendix, Fig. S33C). In a biodistribution study, Pt1a-treated mice displayed lower accumulation of platinum content in liver, spleen, lung, kidney, and tumor compared to that of cisplatin treatment (SI Appendix, Fig. S33D). In addition, the in vivo antitumor and antimetastatic activities of Pt1a in two independent mouse xenograft and lung metastases models of MDA-MB-231 human breast cancer were studied. The MDA-MB-231 xenograft-bearing mice receiving Pt1a (3 mg/kg) by intravenous injection demonstrated an inhibition of tumor growth by 62.5% (Fig. 5A and SI Appendix, Fig. S34A) and a decrease in tumor weight compared with those treated with vehicle (SI Appendix, Fig. S34B). No apparent difference in body weight of mice was found throughout a 31-d treatment (SI Appendix, Fig. S34C). Immunohistochemical analysis of the sectioned tumors revealed that Pt1a treatment could markedly reduce the expression of VIM compared to vehicle control (Fig. 5B); an observation relevant to the tumor differentiation into less invasive epithelial phenotype will be discussed in later sections. For the lung metastases of MDA-MB-231 human breast cancer in nude mouse model, a higher dosage (5 mg/kg) of Pt1a was used for the study. As revealed by the bioluminescence images (Fig. 5C), Pt1a treatment effectively suppressed the growth of metastatic MDA-MB-231 cancer cells in the lung by 80.3% (Fig. 5D), with negligible change in body weight in 42 d (Fig. 5E). Morphological examination and histological analysis revealed that Pt1a treatment significantly reduced the number and area of metastatic nodules in the lungs compared with vehicle control (Fig. 5F). Furthermore, no obvious histopathological lesion was observed in heart, liver, spleen, and kidney of Pt1a-treated mice (SI Appendix, Fig. S35). The weight of the lungs in Pt1a-treated mice remained comparable to the normal lungs but smaller in size relative to vehicle control (Fig. 5G). Moreover, the in vivo antimetastatic activity of Pt1a on B16-F10 melanoma, which is another highly invasive cancer cell line, was investigated. Pt1a administration for 14 d attenuated the growth of metastatic B16-F10 melanoma cells in the lung of C57BL/6N mice, as indicated by the reduced number of black nodules in the lungs of mice treated with Pt1a compared to vehicle control (Fig. 5 H and I).

Fig. 5.

In vivo antitumor activity of Pt1a on tumor xenograft and lung metastatic models. (A) Tumor volume of MDA-MB-231 xenografted mice (n = 4) after treatment with vehicle or Pt1a (3 mg/kg) for 31 d. (B) Immunohistochemical analysis of VIM expression in MDA-MB-231 tumor specimens. (Scale bar, 50 µm.) (C) Bioluminescence images of mice bearing lung metastases of MDA-MB-231 cells from vehicle and Pt1a groups (n = 6) throughout a 42-d treatment. (D) Normalized bioluminescence signals of lung metastases and (E) body weight of mice at different days. (F) Representative images and hematoxylin/eosin (H&E) histological staining of the metastases-bearing lungs of mice from vehicle control and Pt1a treatment groups at the end point (day 42; arrow: tumor nodules). (Scale bar, 200 µm.) (G) The weight of mouse lungs at day 42. (H) Photograph of excised lung tissues from mice bearing B16-F10 lung metastases (black nodules) from vehicle and Pt1a groups (n = 6) throughout a 14-d treatment. (I) The numbers of black nodules in excised lungs were counted. Data are presented as mean ± SEM (Student’s t test, *P < 0.05 and **P < 0.01, compared with vehicle control).

In Vivo Pharmacological Safety Evaluation of Pt1a.

Following the in vivo antitumor study of a NCI-H460 xenografted model in the previous part (SI Appendix, Fig. S33), blood biochemistry analysis was performed using the blood samples from each group, including vehicle control and Pt1a- and cisplatin-treated mice. As revealed by SI Appendix, Fig. S36, mice treated with Pt1a for 2 wk showed low systemic toxicity with comparable plasma levels of alanine aminotransferase (ALT) and creatinine (CREA) to vehicle control, in which ALT and CREA are liver and kidney damage biomarkers, respectively. However, the renal function of the mice upon cisplatin treatment was found to be impaired, as evidenced by significant elevation of CREA level by 1.65-fold compared to vehicle control. In vivo biosafety studies of Pt1a were also evaluated in mouse and rabbit models by a National Medical Products Administration (China)–accredited laboratory. In an acute toxicity study, single-dose intravenous administration of Pt1a (20 mg/kg) did not result in mortality, body weight loss, or signs of toxicity in Kunming mice (n = 20) after a 14-d observation period (SI Appendix, Table S13). No observable abnormality was found in various organ tissues between Pt1a and the vehicle control group after necropsy examination. In a mutagenicity test, there was no significant difference in the frequencies of bone marrow micronucleated polychromatic erythrocytes (an indicator of genotoxicity) between mice receiving Pt1a (5, 10, and 20 mg/kg) and vehicle control (SI Appendix, Table S14). Pt1a treatment (0.5 mg/kg) did not show an apparent irritation effect on blood vessels and surrounding tissues of rabbits in a 14-d period (SI Appendix, Table S15).

Target Engagement Study of Pt1a.

Our previous work using clickable photoaffinity probes has identified six cellular proteins, including VIM, heat-shock protein 60 (HSP60), nucleophosmin (NPM), nuclease-sensitive element binding protein (Y box binding protein, YB-1), nucleoside diphosphate kinase A (NDKA), and peroxiredoxin I (PRDX1), as possible molecular targets of the pincer gold(III)–NHC complex which is isostructural to Pt1a (17). Herein, we prioritized these putative protein targets in NCI-H460 lung cancer cells upon Pt1a treatment using cellular thermal shift assay (CETSA) based on the principle of small-molecule-induced thermal stabilization of target proteins (23). As shown in the immunoblot analysis and thermal stability curves (Fig. 6A and SI Appendix, Figs. S37 and S38), VIM showed the greatest increase in the melting temperature (T_m) of 3.6 °C by Pt1a treatment. Additionally, exposure of cells to Pt1a protected the VIM from denaturation at 54 °C in dose-dependent manner as evidenced by a 3.5-fold increase in protein levels compared to untreated control (Fig. 6B). The mitochondrial chaperone HSP60 could also be stabilized by Pt1a, in which the T_m shift was found to be 1.2 °C in comparison to vehicle control (SI Appendix, Fig. S37). The changes in thermal stabilities of the other four protein targets (NPM, YB-1, NDKA, and PRDX1) mediated by Pt1a treatment, however, were relatively subtle (SI Appendix, Fig. S38).

Fig. 6.

Engagement of VIM by Pt1a. (A) Thermal stabilization of VIM following treatment of NCI-H460 cells with Pt1a as determined by CETSA. (B) Isothermal dose–response stabilization of VIM by Pt1a at 54 °C. (C) TEM images of the negatively stained VIM intermediate filaments bundle network from the monomeric VIM treated with vehicle or Pt1a (3 µM) for 1 h. (Scale bar, 200 nm.) (D) Immunofluorescence staining of the VIM filamentous network and tubulin-based microtubules of NCI-H460 cells after treatment with DMSO vehicle or Pt1a (3 µM) for 24 h. (The red channel at λ_ex = 555 nm and λ_em = 560 to 700 nm; the green channel at λ_ex = 488 nm and λ_em = 510 to 530 nm; the blue channel at λ_ex = 405 nm and λ_em = 430 to 470 nm). The merged image of red, green, and blue channels is also shown. (Scale bar, 10 μm.) (E) Expression of mesenchymal protein markers in NCI-H460 cells after dose-dependent treatment of Pt1a for 24 h.

Disruption of VIM Cytoskeletal Organization.

We next investigated the in vitro effects of Pt1a on VIM by TEM. Purified VIM protein naturally assembled into filamentous structures in saline solution. In the presence of Pt1a at equivalent molar ratio, substantial aggregation occurred (Fig. 6C). Immunofluorescence staining experiment of NCI-H460 lung cancer cells demonstrated that, as compared with vehicle control, the VIM network was perturbed by Pt1a treatment at a subcytotoxic concentration (3 µM), showing a densely packed aggregation after 8 h (SI Appendix, Fig. S39). The obvious contraction of the VIM network from the peripheral leading edge was observed upon prolonged treatment till 24 h (Fig. 6D). Importantly, the structural morphology of tubulin-based microtubules remained intact in Pt1a-treated cells (Fig. 6D), indicative of specific targeting of VIM by Pt1a. From the live-cell and time-lapse imaging studies, the intermediate filaments in MDA-MB-231 breast cancer cells expressing mCherry-tagged VIM were shown to reorganize into perinuclear bundles in response to Pt1a treatment over 8 h (SI Appendix, Fig. S40). The green luminescence of Pt1a was also found to partially colocalized with the red fluorescence of the VIM network (Pearson’s correlation coefficient of 0.61; SI Appendix, Fig. S41). The expression of VIM and other EMT-associated proteins was investigated by immunoblot analysis. Upon treatment of cells with Pt1a for 24 h, the expression of mesenchymal markers, including VIM, Snail, N-cadherin, and β-catenin, whose functions are linked to cancer cell migration and invasion (3, 24), were down-regulated (Fig. 6E).

Binding Interactions with VIM.

The binding interactions between purified VIM protein and Pt1a were first studied using NMR spectroscopy. In the presence of VIM, the proton peaks of Pt1a display broadening accompanied by reduced intensity in the ¹H NMR spectrum (Fig. 7A). The biomolecular binding was further supported by saturation transfer difference (STD) NMR experiments based on saturation of the magnetization of receptor protein being transferred to a bound ligand (25). The STD intensity for the protons of Pt1a was shown to be significantly enhanced in the presence of VIM relative to the negligible STD response for Pt1a alone. By determining the binding epitopes of Pt1a to VIM with normalized STD amplification factor as a function of saturation time (Fig. 7B and SI Appendix, Fig. S42), the aromatic protons signals (H2–H6) of the pincer C^N^N moiety as well as the proton pairs of H12 and H16 on NHC ligand exhibited a relatively pronounced STD effect, implying that those ligand atoms are in close proximity to the binding site of VIM. Surface plasmon resonance (SPR) measurements revealed a dose-dependent increase in the binding responses of Pt1a to the immobilized VIM (Fig. 7C), reaching saturable steady-state binding with an equilibrium dissociation constant (K_D) of 1.06 μM. Pt1b having shortened N-alkyl chains bearing two N-methyl groups displayed about 1.8-fold reduction in binding affinity (K_D: 1.86 μM). Pt4 having thiophene instead of phenyl ring in pincer C^N^N moiety (K_D: 1.95 μM) and Pt6 consisting of benzimidazole-derived NHC (K_D: 2.11 μM) showed a modest decrease in binding affinity compared to Pt1a (Fig. 7D and SI Appendix, Fig. S43). Fluorescence titration experiment of VIM with increasing concentrations (0.5 to 9 μM) of Pt1a revealed a dose-dependent decrease in the fluorescence of tryptophan and tyrosine residues (SI Appendix, Fig. S44) and a binding constant was estimated in low micromolar concentration (K_D: 1.25 μM; tryptophan). Time-resolved lifetime measurement revealed a decrease in fluorescence lifetime (τ) of the tryptophan residue of VIM from 3.02 (in the absence of Pt1a) to 2.62 ns in the 1:2 VIM‒Pt1a ratio with biexponential decay (SI Appendix, Fig. S45 A and B). On the basis of the spectral overlap between emission spectrum of tryptophan residue (donor) and absorption spectrum of Pt1a (acceptor) (SI Appendix, Fig. S46), the distance between Pt1a and tryptophan residue of VIM is estimated based on Fӧrster resonance energy transfer (FRET) theory (26) to be 15.9 Å. Emission titration using VIM as the titrant revealed a binding constant K_D of 7.7 μM, estimated based on enhancement of emission of Pt1a at λ_em = 490‒770 nm upon increasing concentrations (1 to 5 μM) of VIM (SI Appendix, Fig. S47). Addition of VIM also resulted in increased excited state lifetime of Pt1a from 0.49 µs (free Pt1a) to 0.99 µs (Pt1a+VIM) (SI Appendix, Fig. S48). This may be attributed to restricted molecular motion of and less accessibility to platinum‒substrate interaction in the excited state of Pt1a upon VIM binding, as similarly reported in other luminescent platinum(II) complexes upon intercalation to nucleic acids and binding to proteins (27, 28).

Fig. 7.

Molecular binding studies of Pt1a with VIM. (A) ¹H and STD NMR spectra of Pt1a (100 µM) or Pt1a with VIM at a ligand:protein ratio of 10:1. (B) Proposed binding epitope mapping of Pt1a with normalized STD intensities. Deep red circles indicate the most intense STD effect. (C) Representative SPR sensorgrams for the binding interaction of Pt1a (5 to 2,000 nM) to immobilized VIM. (D) Binding affinity constants of Pt(II) complexes to VIM as determined by SPR analysis. Data are presented as mean ± SD; n = 2. (E) MD simulation of the binding interaction between Pt1a and a VIM dimer. The electrostatic potential surface of the binding pocket at the middle of the coil 2A domain is shown with red as negative and blue as positive. The hydrogen, carbon, nitrogen, and Pt atoms of Pt1a are colored in white, yellow, blue, and orange, respectively. Critical residues are labeled with the chain identity in A or B in the dimer as subscript letters. (F) Interactions of Pt1a with surrounding amino acid residues in the pocket are shown in stick model. Carbon and oxygen atoms are colored in light gray and red, respectively, whereas the rest are the same as in E. E and F were prepared using PyMOL (Schrödinger, LLC.).

Molecular Dynamics Simulations.

By employing cosolvent molecular dynamics (MD) simulations on the currently longest dimeric structural model (residues S144 to K334) of VIM rod (29), a possible binding pocket in VIM for Pt1a was identified to be centered around L284 of chain B and flapped by aromatic residues (Y276, W290, and Y291 of chain A) (Fig. 7E). The fragment (residues A247 to K334, denoted as L12C2A) spanning over the binding pocket was extracted for molecular docking and the predicted complex of L12C2A and Pt1a was subjected to MD simulations. In the simulated model, the NHC moiety is sandwiched between the helices and the N-butyl chains are surrounded by hydrophobic residues. One butyl group of the NHC extends between Y276 of chain A and A280 and L284 of chain B, while another butyl chain forms close contacts with the aromatic side chains of Y291 of chain A and W290 of chain B (Fig. 7F). For the pincer [Pt^II(C^N^N)]⁺ motif, the phenyl group and the central pyridine are located in a more nonpolar environment. The phenyl moiety is close to Y291 of chain A and the central pyridine attaches tightly to the indole side chain of W290 (chain A) with distances in the range of 6.65 to 7.60 Å (SI Appendix, Fig. S49). In contrast, the peripheral pyridine ring of the C^N^N ligand is oriented toward the backbone residues of K282 and N283 with polar side chains. These interactions are consistent with our NMR data which show that the terminal methyl groups and the imidazole ring of the NHC ligand experience more significant saturation transfer upon binding. The binding affinity K_D based on this simulated model is estimated to be 4.56 µM, which is comparable to the SPR result (1.06 μM).

Discussion

Examples of anticancer transition metal complexes showing in vivo antimetastatic activity are sparse in the literature. For the clinically used platinum-based anticancer drugs, their antimetastatic efficacy is only rarely active unless they are combined with other anticancer drugs (30, 31). Herein, the pincer [Pt^II(C^N^N)(NHC)]⁺ complexes display a broad spectrum of anticancer activities, in which Pt1a shows both effective cytotoxic effect and in vivo antimetastatic activity with favorable pharmacological safety. The self-assembly of pincer [Pt^II(C^N^N)(NHC)]⁺ complexes to form nanoparticles in aqueous solution and culture medium (Fig. 2 B and C and SI Appendix, Fig. S6B) is an appealing feature for much more Pt accumulation in cancer cells compared to normal fibroblasts (SI Appendix, Fig. S25) through endocytosis (32), which may result in preferential cytotoxicity of this class of complexes to cancer cells over normal cells (Table 1 and SI Appendix, Fig. S7).

VIM is a key cytoskeletal component in mesenchymal cancer cells, which are often endowed with tumor-initiating and metastatic properties. Increased VIM expression is identified in various tumor types and correlated with cancer aggressiveness and poor clinical outcome (3, 5). Thus, VIM-targeting compounds are advantageous for development of effective chemotherapeutic agents via inhibiting the cancer stemness and metastasis. Here, the engagement of Pt1a with VIM was evidenced by the following experiments: 1) CETSA (Fig. 6 A and B), 2) ¹H, ¹H STD NMR and epitope mapping (Fig. 7 A and B), 3) SPR (Fig. 7 C and D), and 4) MD simulations (Fig. 7 E and F). The binding interactions between Pt1a and VIM are distinct from those previously reported VIM-targeting organic compounds. In the literature, compounds bearing reactive electrophiles (e.g., Michael acceptor) such as withaferin A and 4-hydroxynonenal were found to covalently bind to VIM on Cys328 residue at the N terminus of the coil 2B domain (11, 13). A fluorescent chemical probe TiY (33) that selectively stains tumor-initiating cells was shown to bind VIM via the same binding site (11). Arylquin 1 targeted VIM to induce secretion of prostate-apoptosis-response-4 and was demonstrated through computational modeling to bind to a hydrophobic pocket between a pair of head-to-tail α-helical dimers in the tetrameric VIM (12). Herein, a model of biomolecular interaction between Pt1a and VIM was established using epitope mapping by STD NMR analysis and MD simulations. The orthogonal scaffold of Pt1a composed of cyclometalated pincer [Pt^II(C^N^N)] plane and NHC moiety with the N-butyl chains displays hydrophobic contacts with the proximal amino acid residues (e.g., tryptophan and tyrosine) at the middle of the coil 2A domain of VIM dimer (Fig. 7 E and F), resulting in geometrical distortion of the extended rod. In addition to the SPR binding analysis (Fig. 7 C and D), the binding interaction between Pt1a and VIM has also been revealed by fluorescence quenching of the aromatic tryptophan residue (W290) of VIM in the presence of Pt1a (SI Appendix, Fig. S44) as well as enhancement of emission intensity and increase in emission lifetime of Pt1a in the presence of VIM (SI Appendix, Figs. S47 and S48). While there is variation in the binding constants (K_D) estimated from different experiments, all these values are in the low micromolar regime. The observed quenching of W290 fluorescence by Pt1a can be ascribed to fluorescence energy transfer from W290 to Pt1a. The intermolecular distance between W290 of VIM (donor) and Pt1a (acceptor) calculated from FRET analysis is 15.9 Å (SI Appendix, Fig. S46B), being comparable (within a factor of 2) to that of 7.60 to 8.22 Å obtained by MD simulations based on the complex formed between Pt1a and full-rod dimeric VIM model (SI Appendix, Fig. S49). Such intermolecular distance together with the spectral overlap of fluorescence spectrum of W290 and absorption spectrum of Pt1a (SI Appendix, Fig. S46A) are supportive of fluorescence energy transfer in quenching of W290 fluorescence by Pt1a. The coil 2A domain was recently reported to be more dynamic than the N-terminal coil 1A and coil 1B domains for the proper cross-coiling mechanism (34). The MD simulations on the complex formed between Pt1a and the dimeric model of VIM rod (residues S144 to K334) further reveal that the binding of Pt1a to the coil 2A domain of VIM dimer distorts the geometry of the extended rod (SI Appendix, Fig. S50). It is envisaged that such structural distortion upon Pt1a binding can result in perturbation of the regular assembly of the VIM intermediate filament.

The structural scaffold of the pincer Pt(II)‒NHC system should be amenable to modifications on both pincer C^N^N and NHC ligands for tuning the anticancer activities. Comparable antiproliferative activity toward different cancer cell lines (Table 1 and SI Appendix, Fig. S7) and a modest decrease in the VIM binding affinity were observed when Pt1a was modified to contain thiophene-substituted C^N^N (Pt4) or benzimidazole-derived NHC ligands (Pt6) (Fig. 7D). Importantly, increasing the N-alkyl chain length (butyl for Pt1a versus methyl for Pt1b) consistently enhances the cell growth inhibition toward all cancer cell lines examined. This also resulted in slightly higher VIM binding affinity, attributable to a modulation on the nonpolar interactions between N-alkyl groups with the proximal amino acid residues (Fig. 7 E and F).

In line with the biological consequences of targeting VIM, we have been able to show that Pt1a exhibited in vitro antimigratory and antimetastatic activities (Fig. 4 B and C). Pt1a induced the aggregation of VIM filaments in vitro (Fig. 6C) and collapse of the VIM networks in cells, whereas tubulin-based microtubules remained unaffected (Fig. 6D). Pt1a treatment was also shown to reduce the expression of VIM and other mesenchymal proteins involved in tumorigenesis and metastasis, including N-cadherin and β-catenin, which form protein complexes mediating transendothelial migration (35), and Snail, which is the transcriptional factor controlling expression of proteins involved in EMT (Fig. 6E) (24). In addition, Pt1a inhibited the self-renewal growth of tumorspheres (Fig. 4D) derived from cancer stem cells which were shown to highly express VIM as a biomarker (3, 33). Pt1a treatments were demonstrated to display promising in vivo antitumor and antimetastatic activities against NCI-H460, MDA-MB-231, and B16-F10 cancers with high expression of VIM (Fig. 5 and SI Appendix, Figs. S33 and S34). The reduced VIM immunostaining in the shrunken tumors of Pt1a-treated xenografted mice (Fig. 5B) are presumably related to the specificity of VIM engagement by Pt1a.

Despite the ubiquitous use of cisplatin in chemotherapy, dose-limiting toxicities are frequently encountered (36). In this work, Pt1a could be administrated without significant adverse effects in animals. Mice having lung metastasis of human breast cancer MDA-MB-231 could tolerate 42 d of multiple administration of Pt1a treatment (5 mg/kg), displaying no body weight loss or organ damage (Fig. 5E and SI Appendix, Fig. S35). Importantly, there is a large safety window of Pt1a as indicated by 14-d acute toxicity study showing no lethality at maximally tested dosage of 20 mg/kg (SI Appendix, Table S13). The in vivo tolerability of Pt1a could be attributable to the generation of oxidized and glucuronidated metabolites, involving the hydroxylation of the N-butyl chain of the coordinated NHC ligand and glucuronide conjugation to corresponding hydroxylated species (Fig. 3 and SI Appendix, Fig. S22) that are renal-clearable (SI Appendix, Fig. S15) and display reduced in vitro potency on cell growth inhibition (SI Appendix, Table S12 and Fig. S23).

Conclusions

In summary, we have identified Pt1a as a highly effective antitumor compound targeting to VIM with high binding affinity with a K_D of 1 µM and which has been shown to be effective in suppressing tumor initiation and metastasis. Noncovalent binding interactions of Pt1a with VIM involving the [Pt^II(C^N^N)] plane and NHC motif arranged in an orthogonal scaffold is essential for the molecular targeting. The scaffold of Pt1a also allows the complex to be biotransformed into less toxic and renal-clearable metabolites. Our studies highlight the pincer-type platinum(II)–NHC complexes to be developed as a new generation of platinum-targeted chemotherapy for combating metastatic cancers.

Materials and Methods

Details of the materials, experimental procedures, and data, including synthesis and characterization, photophysical measurement, X-ray crystallography, self-assembly studies, metabolite analysis, cellular thermal shift assay, biomolecular binding (STD–NMR protein binding, SPR, and emission titration experiments), MD simulations, in vitro anticancer, and in vivo antitumor studies are described in SI Appendix.

Data Availability

X-ray crystallographic data have been deposited in The Cambridge Crystallographic Data Centre, https://www.ccdc.cam.ac.uk/ (accession nos. 2008384, 2008385, and 2008386). All study data are included in the article and/or SI Appendix.