Singlet oxygen formation vs. photodissociation for light-responsive protic ruthenium anticancer compounds: The oxygenated substituent determines which pathway dominates

Abstract

Ruthenium complexes bearing protic diimine ligands are cytotoxic to certain cancer cells upon irradiation with blue light. Previously reported complexes of the type [(N,N)₂Ru(6,6’-dhbp)]Cl₂ with 6,6’-dhbp = 6,6’-dihydroxybipyridine and N,N = 2,2’-bipyridine (bipy) (1_A), 1,10-phenanthroline (phen) (2_A), and 2,3-dihydro-[1,4]dioxino[2,3-f][1,10]phenanthroline (dop) (3_A) show EC₅₀ values as low as 4 μM (for 3_A) vs. breast cancer cells upon blue light irradiation (Inorg. Chem. 2017, 56, 7519). Herein, subscript A denotes the acidic form of the complex bearing OH groups, and B denotes the basic form bearing O⁻ groups. This photocytotoxicity was originally attributed to photodissociation, but recent results suggest that singlet oxygen formation is a more plausible cause of photocytotoxicity. In particular, bulky methoxy substituents enhance photodissociation but these complexes are non-toxic (Dalton Trans. 2018, 47, 15685). Cellular studies are presented herein that show the formation of reactive oxygen species (ROS) and apoptosis indicators upon treatment of cells with complex 3_A and blue light. Singlet oxygen sensor green (SOSG) shows the formation of ¹O₂ in cell culture for cells treated with 3_A and blue light. At physiological pH, complexes 1_A-3_A are deprotonated to form 1_B-3_B in situ. Quantum yields for ¹O₂ (ɸ_Δ ) are 0.87 and 0.48 for 2_B and 3_B respectively, and these are an order of magnitude higher than the quantum yields for 2_A and 3_A. The values for ɸ_Δ show an increase with 6,6’-dhbp derived substituents as follows: OMe < OH < O⁻. TD-DFT studies show that the presence of a low lying triplet metal-centered (³MC) state favors photodissociation and disfavors ¹O₂ formation for 2_A and 3_A (OH groups). However, upon deprotonation (O⁻ groups), the ³MLCT state is accessible and can readily lead to ¹O₂ formation, but the dissociative ³MC state is energetically inaccessible. The changes to the energy of the ³MLCT state upon deprotonation have been confirmed by steady state luminescence experiments on 1_A-3_A and their basic analogs, 1_B-3_B. This energy landscape favors ¹O₂ formation for 2_B and 3_B and leads to enhanced toxicity for these complexes under physiological conditions. The ability to convert readily from OH to O⁻ groups allowed us to investigate an electronic change that is not accompanied by steric changes in this fundamental study.

Graphical Abstract

TOC Synopsis: Which pathway dominates for protic Ru compounds: singlet oxygen formation or photodissociation? Hypothesis: the protonation state of the OH groups determines the dominant pathway.

INTRODUCTION

Platinum (Pt)-based drugs are arguably among the most successful anticancer drugs. However, the side effects of treatment are severe and even life-threatening. Pt-based anticancer agents such as cisplatin act on all rapidly dividing cells, which includes healthy cells. There is an urgent need for therapies with greater selectivity for tumors over normal tissue. Tumor-specific therapeutics could take advantage of prodrugs that are activated in the unique environment of the cancer cell or by focused light to minimize off-target effects. One example is photodynamic therapy (PDT), an FDA-approved technique that uses a photosensitizer (PS), light, and oxygen to generate reactive oxygen species (ROS), importantly cytotoxic singlet oxygen (¹O₂).^1–2 Photochemotherapy (PCT), while not FDA-approved, is an emerging light-based alternative to PDT that involves light-induced photoreactions that generate cytotoxic species (Scheme 1).³ PCT using metal complexes typically relies on light-triggered ligand dissociation to generate a ligand deficient metal center and the liberated ligand, both of which could be cytotoxic.^4–6 PDT is generally thought to be more effective because the process is inherently catalytic in the PS (Schemes 1 and 3).⁷ Therefore, a very low concentration of a PS can generate a significant quantity of cytotoxic ¹O₂. In contrast, PCT is stoichiometric, and one metal complex typically generates one equivalent of aquated metal complex and one equivalent of free ligand (Scheme 1). While requiring higher dosing levels of metal complex, the PCT mechanism (in theory) does not require oxygen and thus could have added utility for treating hypoxic tumors. In practice, however, very few PCT agents maintain activity in hypoxia,⁸ and continued efforts are underway to address this shortcoming.

Scheme 1.

Ru complexes herein can utilize both PCT and PDT pathways. This paper focuses on the factors that influence which pathway is taken.

Scheme 3.

Photodissociation products vs. singlet oxygen. Which species is responsible for the observed toxicity? What role does charge play in determining the pathway taken?

Protic Ru(II) complexes as PCT agents have proven advantageous in our preliminary work^9–12 because they are deprotonated at physiological pH, which results in improved cellular uptake due to an overall neutral charge and a more lipophilic metal complex.¹³ This paper focuses on the factors that influence which pathway (PDT versus PCT) is dominant in protic Ru(II) complexes. By understanding these factors, we can create better light-responsive agents that may generate ¹O₂ in normoxia but capitalize on PCT mechanisms in hypoxia.

Ruthenium anticancer agents are far less developed than their platinum counterparts, with no Ru-based coordination complexes FDA-approved to date (apart from the use of the ¹⁰⁶Ru isotope in radiotherapy). However, Ru-based metallotherapeutics have much to offer, specifically in the area of light-based therapeutics. The pseudo octahedral geometries of Ru(II) and Ru(III) coordination complexes accommodate six ligands, and well-established synthetic methodology for variation of these very modular architectures makes it possible to rapidly examine a diverse structural landscape for building in and fine-tuning desired properties. As with other drug-like molecules, their three-dimensional structures, redox potentials, and relative lipophilicity/hydrophilicity indices can play a role in cellular uptake, localization, and mechanism of action.¹⁴

To date only two Ru(III) complexes (NAMI-A and NKP-1339/IT-139/BOLD-100), which are not light-activated, have advanced to clinical trials (Chart 1).^15–18 NAMI-A, an antimetastatic agent, has since been abandoned after a Phase 2 study did not yield the desired efficacy (albeit in what may have been an inappropriate clinical trial design for an antimetastatic agent).¹⁸ IT-139 (formerly NKP-1339 but currently BOLD-100), was proposed to exhibit anticancer activity through modulation of ER stress, completed a Phase 1 study,¹⁷ and a Phase 2 study (Clinicaltrials.gov identifier NCT04421820) using BOLD-100 (formerly IT-139 and NKP-1339) in combination with FOLFOX for various advanced solid tumors is currently recruiting patients.

Chart 1.

Ruthenium complexes that entered clinical trials.

The only Ru(II) complex to advance to human clinical trials is TLD1433, and this light-triggered compound is being used as a PDT agent for treating non-muscle invasive bladder cancer (NMIBC).^19–20 It has an extraordinarily high ¹O₂ quantum yield (near unit efficiency) and as a consequence is extremely phototoxic toward cancer cells.²¹ TLD1433 performed well in a Phase 1b clinical trial (Clinicaltrials.gov identifier NCT03053635) and is currently in a much larger Phase 2 clinical trial (Clinicaltrials.gov identifier NCT03945162. Together, the two Ru(III) complexes NAMI-A and BOLD-100 and the one Ru(II) complex TLD1433 that have advanced to clinical trials are opening the door to many translational investigations with Ru-based complexes as anticancer therapeutics, particularly in the field of light-responsive prodrugs.²⁰

While Ru(II) complexes for PCT have not advanced to clinical trials, these agents are of special interest for their potential to act under hypoxic conditions. Glazer et al. first reported Ru(II) complexes that readily photodissociate strain-inducing ligands to bind DNA and cause cytotoxicity,^22–24 and they and others have since investigated a variety of related systems.^{8, 25–27} The premise is that steric bulk near the metal center facilitates photodissociation of a ligand with visible light. This mechanism is not limited to the bis-heteroleptic diimine systems and has also been applied to Ru(II) tris-heteroleptic complexes derived from a combination of tri-, bi-, and monodentate ligands as well.^{5–6, 28–34} Turro has shown that some of these tris-heteroleptic complexes can both photodissociate ligands and generate ¹O₂, exploiting both PCT and PDT mechanisms in a single complex in a dual-action approach,^4–6 and Glazer and McFarland have demonstrated similar dual-action capacity for bis-heteroleptic diimine complexes.^{22, 25}

The work presented herein was inspired by some of these earlier PCT examples, especially the dual action systems, but with a focus on examining the role that charged ligands may play in determining the partitioning of excited state reactivity between the PCT and PDT mechanisms. McFarland, Sadler, and others have shown that changing the charge of the metal complex (e.g. with cyclometalated C,N bound ligands) can greatly alter the photocytotoxicity,^{14, 35–41} but these examples are presumed to rely on ¹O₂ as the photocytotoxic agent. The present work investigates Ru(II) bis-heteroleptic diimine complexes with ligands that bear ionizable groups that are positioned to induce strain in the inner coordination sphere of the complex. This arrangement presents a unique opportunity to directly influence access to the dissociative triplet metal-centered (³MC) state in systems that may otherwise generate ¹O₂ from triplet metal-to-ligand charge transfer (³MLCT) states.

Our prior publications focused on a series of compounds of the type [(N,N)₂Ru(6,6’-dhbp)]²⁺ with 6,6’-dhbp = 6,6’-dihydroxybipyridine and N,N = 2,2’-bipyridine (bipy) (1_A), 1,10-phenanthroline (phen) (2_A), or 2,3-dihydro-[1,4]dioxino[2,3-f][1,10]phenanthroline (dop) (3_A) (Scheme 2a).^9–10 Herein, subscript A denotes the acidic and dicationic form of the metal complex bearing OH groups and subscript B denotes the basic and neutral form bearing O⁻ groups. The 6,6’-dhbp ligand is protic, allowing the properties of the metal complexes to be controlled by deprotonation events. The OH groups are deprotonated at physiological pH, and the uptake of the resulting neutral Ru complexes is improved significantly.¹³ Of complexes 1_A-3_A (which form predominantly 1_B-3_B at physiological pH), the most photocytotoxic complex was 3, which displayed an EC₅₀ of ~4 μM upon irradiation with blue light and a photocytotoxicity index (PI = EC_{50_dark}/EC_{50_light}) as high as 120 in breast cancer cells.¹⁰ This photocytotoxicity was initially attributed to photodissociation of the 6,6’-dhbp ligand, but in fact the evidence now supports a different mechanism described herein.

Scheme 2.

a) Deprotonation of X_A = [(N,N)₂Ru(6,6’-dhbp)]²⁺ complexes under physiological conditions changes the properties. X = 1, 2, or 3 and A denotes the acidic form bearing OH groups and B denotes the basic form bearing O⁻ groups. b) Aprotic complexes 4 and 5. 4 (N,N = phen) and 5 (N,N = dop)

The 6,6’-dhbp ligand provides steric bulk near the metal center that helps promote photodissociation of this ligand (ɸ_PD is ~10⁻³ for 1_A-3_A, Table 1). Interestingly, deprotonation reduced the quantum yields for photodissociation by 1–2 orders of magnitude (Table 1) for 2 and 3. Thus, ɸ_PD values are 10⁻⁴ to 10⁻⁵ for 2_B and 3_B under physiological pH conditions. Deprotonation is mostly an electronic change without any steric component. In contrast, the complexes [(N,N)₂Ru(6,6’-dmbp)]²⁺ with 6,6’-dmbp = 6,6’-dimethoxybipyridine and N,N = phen (4) or dop (5) were designed to be aprotic and sterically bulkier versions of 2 and 3 (Scheme 2b).¹¹ The change from OH to OMe groups resulted in increased (by a factor of 3–12) quantum yields for photodissociation (Table 1).¹¹ However, the complexes 4 and 5 were not phototoxic to breast cancer cells, which was attributed to a combination of poor uptake (as reflected in log(D_o/w) values in Table 1) for these dicationic compounds (cf. 2 and 3) and low cytotoxicity for the products of photodissociation.¹¹

Table 1.

Comparison of acidity, log(D_o/w) (previously reported)^{10–11, 13}, quantum yields for photodissociation (Φ_PD) (previously reported)^9–11, and quantum yields for singlet oxygen (Φ_Δ) (reported herein) upon irradiation at 450 nm for complexes 1–5.

compound	Structure	pKa avg	log(Do/w)a at pH 7.4	ΦPD at pH 5.0b	ΦPD at pH 7.5c	ΦΔ for XAd	ΦΔ for XBd
1A	[(bipy)2Ru(6,6’-dhbp)]2+	6.3	1.4(1)	0.0058(5)	0.0012(1)	0.041(2)	0.18(2)
2A	[(phen)2Ru(6,6’-dhbp)]2+	6.0(1)	1.6(1)	0.0020(2)	0.000036(1)	0.048(2)	0.87(9)
3A	[(dop)2Ru(6,6’-dhbp)]2+	5.9(1)	1.8(1)	0.001(1)	0.00022(3)	0.048(2)	0.48(5)
4	[(phen)2Ru(6,6’-dmbp)]2+	N/A	−1.3(2)	0.024(6)	N/A	0.01(1)e	N/A
5	[(dop)2Ru(6,6’-dmbp)]2+	N/A	−1.1(1)	0.0030(2)	N/A	0.01(1)e	N/A

^aLog(D_o/w) is the octanol-water partition coefficient measured as described in our prior publication.¹³ D_o/w is defined as the total concentration of all ruthenium complex (in various protonation states) in octanol divided by the total concentration of all ruthenium complex in water for the equilibrated biphasic mixture.

^bIn aqueous solution at pH 5.0, mostly the X_A form is present if using a protic ligand (in 1-3).

^cIn aqueous solution at pH 7.5, mostly the X_B form is present if using a protic ligand (in 1-3).

^dIsolated X_A or X_B (in 1-3) was used in CD₃OD. The aprotic complexes 4 and 5 are included in the X_A column because they carry the same charge as 1_A-3_A.

^ePhotodissociation also occured during these experiments to quantify singlet oxygen, thereby reducing the accuracy of this measurement. Incident photons could have led to either product.

Complexes 1-5 generate [(N,N)₂Ru(H₂O)₂]²⁺ upon photodissociation (N,N = bipy, phen or dop) (Scheme 3), yet the complexes with the lowest ɸ_PD values (2_B, 3_B) displayed the greatest photocytotoxicity. Bonnet recently presented evidence that for Glazer’s compound, [(bipy)₂Ru(6,6’-Me₂-bipy)]²⁺ (G, Chart 2), the methylated bipy ligand rather than the metal is the source of toxicity.^{22, 42} Bonnet stated that [(bipy)₂Ru(OH₂)₂]²⁺ (from photodissociation of G) is nontoxic by synthesizing another molecule that is expected to also form this photoproduct in cells. The photoproduct is the same one that we obtain from photodissociation of 1_A (Scheme 3). We have previously conducted experiments to probe whether the aquated metal complex or the free ligand are toxic (Scheme 3). We independently synthesized the products from the photodissociation of 1-3, namely [(N,N)₂Ru(OH₂)₂]SO₄ (N,N = bipy, phen, dop), and these were nontoxic (EC₅₀ >100 μM) under our experimental conditions.¹⁰ However, Glazer and co-workers performed extensive biological characterization on G, and demonstrated direct DNA damage in live cells through metalation of the nucleic acids at levels comparable to cisplatin, and induction of the DNA damage response as measured by immunoblotting for reporters such as H2AX.^{33, 43} They also reported that G produces a phenotype in bacteria that is virtually identical to cisplatin. Thus, the source of the biological activity of these photoejecting Ru(II) complexes is a subject of some controversy.

Chart 2.

Glazer’s compound (G) which photodissociates to generate [(bipy)₂Ru(OH₂)₂]²⁺ which binds DNA.

With regard to the free ligand being the source of cytotoxicity, some pyridone derivatives^44–46 do exhibit drug-like properties. However, 6,6’-dhbp, as the free ligand, is nontoxic (EC₅₀ >100 μM) against several breast cell lines under the assay conditions employed.¹⁰ Nevertheless, we acknowledge that there are numerous issues in attempting to quantify the cytotoxicity of free ligands, knowing that they (i) may aggregate / precipitate in the buffers and media used for making stock solutions, (ii) likely do not enter cells in the same manner as the intact metal complex, (iii) may not localize to the same part of the cell as the intact metal complex, and (iv) are expected to avidly engage with metal ions both in the buffers and media as well as the cell, among other things. With these considerations in mind, the data presented herein (Table 1) suggests that the photodissociation products, both the aquated metal complexes and their free ligands, are of low toxicity with the caveat that the considerations described above were not explored.

Herein, we discuss several of these complexes in terms of an alternative mechanism for photocytotoxicity that involves ¹O₂ generation. The protic complexes in particular allow us to investigate the role of charge, as it relates to both the ligand and metal complex, in determining which pathway dominates: photodissociation versus ¹O₂ formation.

RESULTS AND DISCUSSION

Cellular Studies Measuring Reactive Oxygen Species (ROS), ¹O₂, and Apoptosis.

Cellular studies focused on complex 3 because it had the best photocytotoxicity index (PI) toward breast cancer cell lines in our previous study. MCF7 cells were dosed with complex 3_A (which forms 3_B in situ) at a concentration of 5 μM (which is near the EC₅₀ value), and the percentage of cells showing intracellular ROS activity and late apoptosis were measured for the dark condition (solid bars) and with a sublethal dose of 450 nm blue light (striped bars) (Figure 1). This study indicates that complex 3_A, when activated by blue light, induces the formation of ROS and triggers apoptosis. Notably, no ROS were induced with 3_A in the dark (i.e., not significantly different than the negative control), but there was a moderately increased level of apoptosis, which suggests a different mechanism for the low level of dark cytotoxicity observed in our past work. ROS production for 3_A (in contrast to the positive control) was clearly light-triggered, which was also demonstrated with MDA-MB-231 cells using the same type of assay (see Figures S1 and S2 in the Supporting Information).

Figure 1.

Left: Apoptosis indicators in MCF7 as measured by Annexin WPK Right: ROS indicators in MCF7. +ROS is the pyocyanin positive control as an ROS inducer. For both plots, solid bars Indicate incubation with no drug, positive control, or 3_A for 48 h in the dark. Striped bars indicate incubation followed by irradiation with blue tight.

MDA-MB-231 cells were also tested for the presence of ¹O₂ upon treatment with 5 μM of 3_A and blue light (Figure 2; via Singlet Oxygen Sensor Green (SOSG)). Singlet oxygen was detected in only ~2% of cells from the control samples (no metal complex and 3_A in the dark), but in contrast, ¹O₂ was present in 33% of the cells for treated with 3_A and irradiated. Since SOSG is specific for detecting ¹O₂ and not hydroxyl radicals or superoxide, these results suggest that ¹O₂ may be the ROS responsible for the observed photocytotoxicity with 3_A.

Figure 2.

Left: SOSG luminescence is greatest for MDA-MB-231 cells treated with 3_A (5 μM) and Irradiated with blue light. Right: Flow cytometry shows an increase in the number of cells with SO present with blue light and 3_A.

Singlet Oxygen Quantum Yields.

The quantum yields for ¹O₂ formation (${\text{Φ}}_{\text{Δ}}$) were determined for 1-5 using the direct method based on the ¹O₂ luminescence centered at 1268 nm (Table 1). The protic complexes were studied as the isolated acidic forms (1_A-3_A) bearing OH groups and the isolated basic forms (1_B-3_B). All measurements were performed in CD₃OD as the preferred solvent given that water quenches the luminescence of ¹O₂, and acetonitrile has been shown to accelerate photodissociation.⁹ Values for ɸ_Δ were 0.041 for 1_A and 0.048 for 2_A and 3_A, but the deprotonated analogs 1_B, 2_B, and 3_B gave ɸ_Δ values of 0.18, 0.87 and 0.48, respectively. Quantum yields for ¹O₂ are enhanced 4.4- to 20-fold by deprotonation of the 6,6’-dhbp ligand, which generates the corresponding neutral Ru(II) complexes. Thus, for 3_B (the form present at physiological pH), ɸ_Δ is three orders of magnitude greater than ɸ_PD. Likewise, for 2_B, ɸ_Δ is four orders of magnitude greater than ɸ_PD. Considering that ¹O₂ generation is catalytic whereas photodissociation is stoichiometric, ¹O₂ is the most plausible source of photocytotoxicity for complexes such as 3, which was our most photocytotoxic complex (highest PI value).

The ¹O₂ quantum yields for 4 and 5 were calculated to be near 0.01 (Table 1), which shows that the change from OH (in 2_A and 3_A) to OMe groups reduces ¹O₂ production five-fold. Complexes 4 and 5 undergo a significant amount of photodissociation during the experiments designed to quantify ¹O₂ production, which results in greater error in ɸ_Δ for these complexes given that incident photons could have led to either product and also because the photoproducts themselves may then absorb incident photons. Nonetheless, it is clear that for a given amount of incident photons, the yield of ¹O₂ will be at least 87- to 48-fold higher for 2_B and 3_B (with O⁻ groups) versus for 4 and 5 (with OMe groups).

Such effects based upon changing the charge of the ligand were demonstrated by McFarland et al. where complexes of the type [(bipy)₂Ru(N,N)]²⁺ (ɸ_Δ = 0.8 to 0.9) had much higher ¹O₂ quantum yields than their cyclometallated analogs [(bipy)₂Ru(N,C)]⁺ (ɸ_Δ = 0.07 to 0.08). The change from a neutral to an anionic ligand decreased the quantum yields for ¹O₂ formation by 10-fold, which is the opposite of what we observe (with an anionic ligand increasing ɸ_Δ).³⁸ However, the key differences relative to this study were a lack of photodissociation as a competing pathway and the localization of the negative charge on a carbon atom, which is directly bound to the metal.⁴⁷

Computational Studies of the Excited States Generated Upon Irradiation.

The observation that complexes bearing O⁻ groups are less susceptible to photodissociation and have higher ¹O₂ quantum yields demonstrates that the OH (2_A, 3_A) vs. O⁻ (2_B, 3_B) groups result in electronic effects that influence the energies of the excited states (vide infra) and thus the efficiency of ¹O₂ formation versus photodissociation. The energy of the ³MLCT state relative to the ³MC state is of particular importance in determining the predominant mechanism, whereby lower ³MLCT energies relative to the ³MC energies would be expected to favor ¹O₂ production over photodissociation. The ³MLCT energies of the protic and deprotonated forms of 1-3 were estimated by their phosphorescence (Figures S15–S19 and Table S4). The acidic forms (1_A-3_A) had luminescence maxima near 615 nm (ranging from 610–630 nm depending on the compound and the solvent, ~2.02 eV) whereas the basic forms (1_B-3_B) produced emission at substantially longer wavelengths. The doubly deprotonated neutral complexes yielded emission maxima near 710 nm (ranging from 696–734 nm, ~1.71 eV), with lower-intensity transitions clearly visible near 900 nm and tailing out past 1000 nm for 1_B (Figure S15). With this is mind, time-dependent density functional theory (TD-DFT) studies were carried out on complexes 2 and 3 to probe the influence of protonation state on the relative energies of the lowest-lying ³MLCT states that were determined experimentally but also on the nonemissive ³MC states that are much more difficult to assess experimentally. These two complexes were selected for the study because they are the more photocytotoxic complexes and also show the largest enhancement in ɸ_Δ upon deprotonation.

We note that many experimental characteristics of ³MC excited states remain largely unknown in Ru polypyridyl complexes today. This includes the energy, whether it is bound or dissociative, and the existence of another potential on the way to non-radiative decay or ligand dissociation as noted by Mukata et al. and others.⁴⁸ Over decades of Ru polypyridyl photophysical literature, transient absorption measurements do not reveal the nature of the ³MC state. More recently, time-resolved infrared vibrational spectroscopy (TR-IR) was able to detect the ³MC state through a detailed analysis of the temporal evolution of vibrational bands under different experimental conditions but, of course, not the energies.^48,49 While transient absorption spectroscopy has provided experimental evidence of the ultrafast quenching of ³MLCT states by the ³MC state, these complexes are specially designed to have unusually stable ³MC states.⁵⁰ This type of experiment does not yield information on the energy of the ³MC state relative to the ³MLCT, and DFT computations are relied upon herein to assert this difference. We use caution in interpreting DFT computations because they are better at predicting relative energy differences vs. absolute energies and furthermore DFT cannot probe every possible pathway involving 2, 3, and other biomolecules in cells. Nonetheless, the DFT computations are informative when considered with the above caveats and the results herein are qualitatively supported by luminescence, photodissociation, and singlet oxygen experiments.

The OH bearing dicationic complexes 2_A and 3_A will be discussed first. Excitation with blue light at 450 nm provides 2.75 eV of energy, which populates the ¹MLCT excited states for 2_A and 3_A. After a vertical excitation with blue light, the relaxed excited state formed will be the singlet MLCT state for 2_A and 3_A (Figure 3) at 2.3–2.4 eV above the ground state. These ¹MLCT states undergo rapid intersystem crossing (ISC) to generate the ³MLCT state for 2_A and 3_A (ISC quantum yields are typically near unity for Ru(II) diimine complexes³). While the ³MLCT state could interact with ³O₂ to generate ¹O₂, the presence of a ³MC excited state for both 2_A and 3_A (at 1.7 eV) that lies below the ³MLCT state (at 2.06–2.08 eV, cf. [Ru(bipy)₃]²⁺ is at 2.1 eV above the ground state) indicates that the internal conversion to the ³MC state may be competitive. The presence of this low-lying ³MC state is not unheard of^50–51 but is a marked departure from most examples in the literature, where ³MC states are typically higher in energy than ³MLCT states.^{3, 52–54} This ³MC state has substantial antibonding character between Ru(II) and 6,6’-dhbp (Figure 4 shows the orbital diagram for ³MC on 2_A) and is expected to lead to photodissociation products. Taken together, these results suggest a mechanistic rationale for why the OH bearing dicationic complexes favor photodissociation over ¹O₂ formation.

Figure 3.

Free energy diagram for energetically accessible excited states of 2_A (in blue) and 3_A (in red) from PBE0-D3/BS1. The lowest energy, thermally accessible triplet excited state is ³MC, which leads to ligand loss and photodissoclation products. The ³MLCT state would lead to singlet oxygen, but this most likely quickly converts to ³MC for 2_A and 3_A before It comes in contact with oxygen.

Figure 4.

Orbital depiction of the ³MC excited state for 2_A shows that this state Is antibonding between Ru and 6,6’-dhbp.

The lowest-lying singlet excited states for the neutral, deprotonated complexes, 2_B and 3_B appear substantially different in both character and energy (Figure 5) compared to 2_A and 3_A. For 2_B and 3_B, the lowest-lying singlet excited states are ligand-to-ligand charge transfer (¹LLCT) at 1.3–1.4 eV above the ground state. The ¹LLCT state involves charge transfer from the deprotonated 6,6’-dhbp ligand to either the phen (2_B) or dop (3_B) ligand. This change in state is likely due to the increased π electron density donated from an O⁻ vs an OH group. These excited states are expected to undergo ISC very rapidly to the ³MLCT states. In contrast to 2_A and 3_A, for 2_B and 3_B the ³MC state is energetically uphill and inaccessible at 1.9 eV. Thus, ligand photodissociation does not occur readily for 2_B and 3_B (and this is reflected in the experimental values of ~10⁻⁴ to 10⁻⁵ for ɸ_PD at pH 7.5 in Table 1). The key reason for the difference in accessibility for ³MC is that the energy of the ³MLCT state is much lower for 2_B and 3_B at 1.21–1.26 eV (versus 2.06–2.08 eV for 2_A and 3_A), which was also confirmed experimentally from phosphorescence measurements.

Figure 5.

Free energy diagram for energetically accessible excited states of 2_B (In blue) and 3_B (in red) from PBE0-D3/BS1. The ³MLCT state is readily accessed and should lead to singlet oxygen formation. Furthermore, the ³MC (which leads to ligand loss) is not accessible and this explains the relative lack of photodissociation for 2_B and 3_B.

For 2_B and 3_B, the combination of raising the energy of ³MC and lowering the energy of the ³MLCT state makes the ³MC state inaccessible. Given that the only accessible excited state is ³MLCT for these basic forms, this state persists long enough to interact with ³O₂ and to generate ¹O₂ in much higher quantum yields of 0.48 to 0.87 (Table 1). These results explain the difference in pathways that are available for the basic versus the acidic form of these 6,6’-dhbp complexes. Namely, the basic form can undergo ¹O₂ formation (with favorable ɸ_Δ) without forming photodissociation products due to an inaccessible ³MC state with antibonding character. In contrast, the acidic forms are able to readily access the ³MC state and photodissociate the 6,6’-dhbp ligand in a pathway that competes effectively with ¹O₂ formation. To further verify these interpretations, computations were repeated with different functional/basis set combinations with similar results (see Figures S8–S10 and Tables S2–S3).

The protonation state also has a measurable effect on the ground state absorption spectra for these complexes. For both complexes 2 and 3, higher pH (more basic) results in a red shift of about 50 nm (0.3 eV)¹⁰ and significant peak broadening (Figure 6, right). Absorption spectra for both the acidic and basic forms were simulated from TD-DFT single-point computations on the ground state geometries and are also shown (Figure 6, left). The trend in both λ_max and spectral shape between the acidic and basic forms correlates nicely between the simulated spectra and experiment. For both 2 and 3 the simulated λ_max red-shift by about 50 nm and the peak broadens significantly. Additional overlays between simulated and experimental spectra are provided in Figures S11–S14.

Figure 6.

Left: Simulated UV-VIS absorption spectra for complex 2 (top) and complex 3 (bottom) from TD-PBE0-D3// PBE0-D3/BS1 Right: Experimental UV-Vis absorption spectra for complex 2 (top) and complex 3 (bottom) as a function of pH as previously reported.¹⁰

CONCLUSIONS

Many studies have investigated how synthetic changes to ligands can influence the quantum yields for ¹O₂ and photodissociation. However, very few studies have looked at the influence of ligand protonation states⁵⁵ on both the magnitude of these quantum yields and the competition between the pathways (Scheme 4). Deprotonation of complexes 1_A-3_A (bearing OH groups) to form 1_B-3_B (bearing O⁻ groups) represents an electronic change and a change in complex charge (from 2+ to neutral) without any significant steric change. We observe that deprotonation both increases the quantum yield for ¹O₂ and decreases the quantum yield of photodissociation products. Furthermore, since ¹O₂ formation is catalytic (whereas photodissociation is stoichiometric) it can result in a greater amount of toxic species generated in cells. These results are confirmed by cellular studies, which show that complex 3 generates ¹O₂ in cells. Furthermore, the combined evidence shows that the products of photodissociation appear to be of low toxicity. This explains why 3_B displays good photocytotoxicity indices despite exceedingly low quantum yields for photodissociation, and furthermore efforts to enhance quantum yields for photodissociation results in low PI values (and reduced ɸ_Δ values) for 4 and 5 (bearing OMe groups and a 2+ charge). The enhanced photocytotoxicity index of 3 in cells (vs. complexes 1 and 2) is related to enhanced cellular uptake for complex 3 due to the more lipophilic structure.^{10, 13} Our study herein did not reveal any significant differences in the quantum yields for ¹O₂ formation for 2 vs. 3. Both 2_B and 3_B are efficient at ¹O₂ formation in the deprotonated state, but those quantum yields decrease 10- to 20-fold upon protonation. Overall this study reveals that the substituent and the complex charge leads to an increase in ɸ_Δ values in this order: OMe < OH < O⁻ that is in part determined by an apparent competition between photodissociation and singlet oxygen formation (Scheme 4).

Scheme 4.

Changes to the OR group in the ligand within 1–5 controls the dominant pathway, PD = photodissociation.

TD-DFT studies suggest a rationale for why complexes 2_A and 3_A (bearing OH groups) favor photodissociation. The presence of a low lying ³MC state with substantial antibonding character between the metal and the 6,6’-dhbp ligand allows photodissociation to occur readily. Furthermore, any photodissociation that occurs will remove the possibility of subsequent ¹O₂ formation from that particular molecule because the aquated species is of low photocytotoxicity and is presumably a poor ¹O₂ generator. For 2_A and 3_A, the ³MLCT state (that can potentially lead to ¹O₂ formation) is on the pathway accessed upon light irradiation, but it likely does not persist for long enough to form significant amounts of ¹O₂. In contrast, complexes 2_B and 3_B (bearing O⁻ groups) favor ¹O₂ formation due to a low lying ³MLCT state that can readily interact with ³O₂. The presence of a much lower-lying ³MLCT state for 1_B-3_B was confirmed experimentally by luminescence experiments, with this state being substantially higher in energy for 1_A-3_A. For the deprotonated complexes, the ³MC state with antibonding character is high in energy and inaccessible, which greatly reduces access to the photodissociation pathway. While DFT results have uncertainties that are difficult to quantify without an experimental estimate of the ³MC energy levels, qualitatively the DFT results are consistent with our experimental data including photodissociation quantum yields which are much higher for 2_A and 3_A vs. 2_B and 3_B. Of course, at physiological pH complexes 2_B and 3_B are the dominant species in cellular media and inside the cells, even in the case of cancer cells displaying a Warburg phenotype.^{13, 56} Thus, the production of ¹O₂ by these complexes is consistent with the observed photocytotoxicity indices. It appears that any photodissociation does not in fact lead to photocytotoxicity. This suggests that the path forward towards enhancing photocytotoxicity in new dihydroxybipyridine derivatives would include synthetic efforts to reduce photodissociation and enhance ¹O₂ formation while improving cellular uptake.