Optochemical Control of Therapeutic Agents through Photocatalyzed Isomerization

Abstract

A Ru(bpy)₃Cl₂ photocatalyst is applied to the rapid trans to cis isomerization of a range of alkene‐containing pharmacological agents, including combretastatin A‐4 (CA‐4), a clinical candidate in oncology, and resveratrol derivatives, switching their configuration from inactive substances to potent cytotoxic agents. Selective in cellulo activation of the CA‐4 analog Res‐3M is demonstrated, along with its potent cytotoxicity and inhibition of microtubule dynamics.

Using a Ru(bpy)₃Cl₂ photocatalyst a rapid trans to cis isomerization for a range of pharmacological agents based‐on an alkene core is achieved. These include combretastatin A‐4 (CA‐4), a clinical candidate in oncology, and resveratrol derivatives, allowing their selective in cellulo activation. This provides a new platform for the rapid in situ activation of unmodified pharmacological agents.

Introduction

The ability to control the pharmacological effect of drugs using light is highly desirable for biomedical research and selective drug delivery with spatiotemporal control. This has been traditionally achieved using photolabile caging groups (Figure 1), ^[1] but more recently, a new modality leveraged on photoisomerizable functional groups has emerged. ^[2] This strategy predominantly makes use of the reversible isomerization of azobenzene to photoswitch the conformation of a drug. An important benefit of the photoisomerization approach over photocaging groups is that it is reversible, allowing for iterative cycling between on and off states. ^[3] A prominent example of the application of this concept is the development of an azobenzene version of combretastatin A4, ^[4] a high‐affinity tubulin ligand with anticancer properties. In its cis configuration, this molecule pharmacologically mimics colchicine, a clinically used antiproliferative agent, however, in its trans configuration it is inactive. ^[5] Inspired by the efficiency of photocatalysis to achieve “uphill isomerization” from trans to cis by energy transfer mechanisms, ^[6] and the growing range of applications of photocatalysis in chemical biology, ^[7] we asked if this isomerization chemistry could be harnessed as an alternative strategy to convert an inactive drug into an active one in a biologically relevant context. Herein we present, the use of a photocatalyst (Ru(bpy)₃Cl₂) to achieve the photoswitching on various known pharmacological agents. The present work offers the possibility to use known inhibitors without reengineering to incorporate a photoswitchable moiety as demonstrated with combretastatin A4 and other drugs. Furthermore, it confines the photoswitching to the vicinity of the catalyst, providing another dimension of selectivity by localization or activity of the catalyst, beyond irradiation intensity.

Figure 1.

Scope of this work in relation to the optical control of pharmacological activity.

Results and Discussion

CA‐4 is derived from the African Bushwillow tree Combretum caffrum and was shown to inhibit tubulin formation. ^[8] The clinical success of tubulin‐targeting agents such as taxol and vinca alkaloids has stimulated tremendous efforts to find other inhibitors in this class. While CA‐4 has demonstrated potent antiangiogenic, ^[9] vascular disrupting ^[10] and cytotoxic activities, ^[11] its clinical development has been hampered by limited solubility (even when employing phosphate prodrugs) and confounding isomerization. ^[12] Even when CA‐4 is administered as the active cis‐isomer, significant isomerization to the thermodynamically favored trans‐isomer, which has been reported to exhibit up to a 1000‐fold decrease in cytotoxicity,[ ⁵ , ¹³ ] occurs and thus limits therapeutic efficacy.

While this background isomerization was detrimental to previous clinical experiments, we believed it would offer a natural off‐switch in a photo‐inducible system, limiting potential off‐target effects resulting from accumulation of the active species. Indeed, CA‐4 and its analogues have been extensively studied as potential photopharmacological agents, owing to their optically active stilbene core. While it has been demonstrated that the native CA‐4 structure can undergo trans‐to‐cis isomerization under UV light and two‐photon irradiation, ^[14] the lack of biocompatibility of such conditions inspired the development of more readily isomerizable analogues, particularly those based on the azobenzene scaffolds. Trauner and Thorn‐Seshold, ^[4a] Streu ^[4b] and Hartman ^[4c] each independently explored azo‐CA‐4 analogues and demonstrated their utility. However, such systems are not compatible with historically useful therapeutics, requiring complete synthetic redesign, and the azo‐functionality has previously demonstrated metabolic liability. ^[15] Thus, we sought to develop a system in which the increased ease of photoisomerization achieved by the azo scaffold could be recapitulated without requiring synthetic modification or relying on the azo functionality.

Photocatalysis provides unique opportunities for achieving bioorthogonal transformations under mild conditions with positional control, owing to the in situ generation of the reactive species from an inert ground state catalyst.[ ⁷ , ¹⁶ ] While both organic and transition metal complexes have been reported for the trans to cis isomerization, ^[17] we reasoned that transition metal complex with a higher quantum yield for the triplet excited state would be more suitable for reaction at low concentrations (μM or below). We initiated our investigation with Ru(bpy)₃Cl₂ based on the fact that ruthenium‐based photocatalysis has been previously demonstrated in live cell and vertebrate models. ^[18] Additionally, there is intense research towards the clinical development of ruthenium polypyridyl complexes for photodynamic studies. ^[19] Furthermore, conjugation of the catalyst to pharmacological agents has been shown to confine its activity to cells expressing specific receptors or subcellular compartments. ^[20]

We sought to test whether uphill isomerization of trans to cis could be applied to the CA‐4 scaffold using Ru(bpy)₃Cl₂ as a photocatalyst. The relevant substrate was dissolved to a concentration of 3.2 mM in the corresponding deuterated solvent, followed by the addition of the photocatalyst as required. The ratio of cis to trans isomers was determined by NMR peak integration (Table 1, Figures S1–6). Pleasingly, upon treatment with 1 mol.% Ru(bpy)₃Cl₂ and irradiation at 455 nm (1 W LED, 11 cm, 6.2 mW cm⁻²) for 60 min was able to effect clean isomerization from trans‐ to cis‐CA‐4 in a ratio of 1 : 9, while no conversion was observed in the absence of either light or the photocatalyst (entry 1 vs 2 and 3). However, upon further irradiation, the reaction could not be driven to completion, indicating that a slower reverse isomerization may be occurring. This hypothesis was supported by performing the same reaction starting from the cis isomer, which yielded a comparable ratio of isomers (entry 4 and 5), suggesting this ratio is the product of an equilibrium between a forward and backward isomerization reaction. Using riboflavin as a photocatalyst instead of Ru(bpy)₃Cl₂ led to significantly poorer isomerization (entry 6). Interestingly, moving from DMF‐d₇ to a mixture of DMSO‐d₆ and D₂O significantly shifted the position of the equilibrium (entry 8), while using DMSO‐d₆ alone only resulted in a small perturbation (entry 9). This behaviour is likely related to differences in relative energy of triplet states in the different solvents, therefore leading to different relative rates of isomerization in the different solvents. However, the ability to conduct this reaction in predominantly aqueous environment does demonstrate the potential for this transformation to occur in a biocompatible solvent system, and comparable conversion was also possible in cell culture media (Figure S7). Importantly, irrespective of the solvent system used, no evidence of any other chemical transformations was observed despite the presence of oxygen, suggesting that the concentration and lifetime of ¹O₂ is insufficient to facilitate any side reactions.

Table 1.

Photocatalytic isomerization of CA‐4 using 455 nm excitation.

Entry	Substrate	Solvent	Conditions	trans [%]	cis [%]
1	trans‐CA‐4	DMF‐d7	Standard	10	90
2	trans‐CA‐4	DMF‐d7	No light	>99	<1
3	trans‐CA‐4	DMF‐d7	No catalyst	>99	<1
4	cis‐CA‐4	DMF‐d7	Standard	11	89
5	cis‐CA‐4	DMF‐d7	No catalyst	<1	>99
6	trans‐CA‐4	DMF‐d7	1 % riboflavin	84	16
7	trans‐CA‐4	DMSO‐d6	Standard	18	82
8	trans‐CA‐4	9 : 1 D2O : DMSO‐d6	Standard	49	51

We next sought to investigate the kinetics of this photocatalytic isomerization. Given our desire to apply this system to an in situ biological setting, it is vital that the reaction proceeds at a sufficiently high rate in order to overcome the inherent limitations in concentration. The two CA‐4 isomers exhibit vastly different spectral properties which we used to monitor the reaction and to identify the isomers by HPLC (Figure 2A). Using 1.25 μM of Ru(bpy)₃Cl₂ catalyst, we were able to observe the formation of the cis isomer within 30 s of irradiation (Figure 2B), while the reaction appeared to have reached completion within 6 min under the described conditions of irradiation in a H₂O with 7.5 % DMSO solvent mixture. By monitoring the rate of disappearance of the trans isomer at 329 nm (Figure 2C) we were able to extrapolate an effective pseudo‐first order rate constant (Figure 2D) of k′=2.38×10⁻³ (±0.01×−10⁻³) and a corresponding second order rate constant of 1.9×10³ M⁻¹ s⁻¹ which is in the order of the fastest reported biorthogonal reactions (tetrazine click=1.5×10³–7.3×10⁴ M⁻¹ s⁻¹). ^[21] The calculated values underestimate the true rate of the forward reaction, as it reflects the rate of the net reaction progress. It should be noted that, while the reaction could be pushed to near completion in DMF, the reaction was significantly slower, taking nearly 2 h (see Figure S8 for quantification) which is consistent with the NMR observations. The rate could be increased by the use of more intense light sources but our investigations were limited to 1 W LED (6.2 mW cm⁻²) which has been shown to be compatible with the irradiation of live cells or organisms. ^[18b]

Figure 2.

Rate of photocatalytic isomerization of CA‐4. Trans‐CA‐4 was prepared as a 3.2 mM stock in DMSO and diluted in water to a final concentration of 250 μM (7.5 % DMSO final concentration), containing 1.25 μM Ru(bpy)₃Cl₂. A) The absorption maximum was determined for each isomer; B) and C) the reaction progression (t in min) was monitored at each wavelength by triplicate injection on reversed‐phase HPLC. D) Remaining concentration of the trans isomer was calculated as a function of the absorbance of the pure cis and trans isomers at 329 nm, with the natural log taken to arrive at the pseudo‐first order rate constant. Error bars represent standard deviation.

We next sought to determine whether this conversion was unique to CA‐4 or could be applied to other alkene containing substrates. Initially we explored the related stilbene natural products trans‐resveratrol trimethyl ether (trans‐Res‐3M, also known as trimethoxystilbene (TMS)) and pterostilbene and were pleased to see that smooth conversion was achieved on a similar time scale (Figure 3, S9–S14). Buoyed by this observation we sought to test whether a broader substrate scope could be exploited. To this end, we tested a range of commercially available therapeutic agents containing a range of alkene functionalities. We were pleased to see that the conversion occurred smoothly on both electron rich and electron poor alkenes, and the presence of heterocycles was also well tolerated. Importantly, the transformation was not limited to substrates containing the stilbene core. Indeed, Mubrutinib, CNB‐001 and QNZ‐46 all converted smoothly within 30 min in spite of the presence of a conjugated heteroaromatic core (Figure S15–S24).

Figure 3.

Substrate scope for photocatalytic isomerization amongst therapeutically relevant molecules. Experiment conducted at 3.2 mM substrate. Trans to cis ratio determined by integration of NMR peaks (see Figure S9–24 for NMR quantifications and LC‐MS characterizations).

Taken together, the results indicate that isomerization of diverse pharmacophores towards the thermodynamically unfavored isomer using low concentrations (μM) and catalyst loading (1 %) is possible on a timescale that is applicable to biological setting (30 min) in a light flux dependent manner (6.2 mW cm⁻², see Figure S25 for reactions performed at lower light intensity).

We next explored whether the Ru‐mediated photocatalytic isomerization could recapitulate the phenotypic activity of a native cis‐containing molecule. In addition to CA‐4, we chose to evaluate trans‐Res‐3M, a naturally occurring resveratrol analogue that has been shown to be well tolerated at high concentrations in vivo reaching a plasma concentration of >80 μM with oral administration. ^[22] Importantly, its cis isomer (cis‐Res‐3M) has also been reported to exert similar antiproliferative activity to combretastatin (tubulin inhibition and antiproliferative effect), albeit with lower potency. ^[23] Initially, we evaluated both compounds in terms of their effect on microtubule dynamics. To do this we employed a HeLa cell line stably expressing (under antibiotic selection) GFP‐EB3, ^[24] a protein which associates with the growing end of the microtubule. ^[25] The accumulation of EB3 at dynamic end results in so‐called “comets” that can be quantified in terms of number, size and speed as a reflection of any perturbation to the native microtubule network (Figure 4A). Pleasingly, we were able to observe a significant increase in potency of microtubule disruption for each of the compounds upon photoisomerization as compared to the trans isomer.

Figure 4.

Phenotypic effects of photoisomerized stilbenes. A) representative images (microscope image, left, and comet identifying mask, right (separated by red line)) and IC₅₀ curves for disruption of microtubule dynamics for the parent trans and photoisomerized variants of B) CA‐4 and C) Res‐3M. In vitro photoisomerization of trans‐Res‐3M is also able to induce D) cytotoxicity (as demonstrated in a nuclear count assay) and E) apoptosis (measured by caspase 3/7 activity, green represents activity). Expanded version of (E) with additional controls is provided in Figure S32). For all experiments photoisomerization was conducted in DMSO at 10 mM (CA‐4) or 50 mM (Res‐3M) with 1 mol.% Ru(bpy)₃Cl₂ and irradiated prior to addition to cell at 455 nm for 30 min which provides trans:cis ratios of 18 : 82 and 10 : 90, respectively.

Briefly, GFP‐EB3 HeLa cells were treated with either the trans‐isomer or a pre‐isomerized solution of CA‐4, with both conditions containing the same amount of catalyst (Figure 3B) and Res‐3M (Figure 3C), respectively, for 15 min before determination of the comet number by automated microscopy (see Supporting Information for full experimental details, Figure S26–S30). The difference between the isomerized sample and trans isomers was clearly observable in the comet analysis (IC₅₀ of 11.9 nM for isomerized‐CA‐4 vs 7378 nM for the trans‐CA‐4; IC₅₀ of 259 nM for isomerized‐Res‐3M vs >50 μM for the trans‐Res‐3M) as well as in a cell viability assay where 20 h treatment led to cytotoxicity at 50 nM for the isomerized‐Res‐3M, whereas no cytotoxicity is observed at 50 μM for the trans‐Res‐3M (Figure 4D, Figure S31 for in situ isomerization). Using an assay for caspase‐3/7 activity (fluorogenic substrate) as a marker of apoptosis, we also demonstrated that the cytotoxicity of the isomerized sample is indeed the result of apoptosis (Figure 4E) whereas the trans isomer in combination with the catalyst did not induce any measurable apoptosis relative to a DMSO control (Figure 4E, see Figure S32 for additional controls with trans isomer, photocatalyst and irradiation). Given the lower toxicity of the trans isomer of Res‐3M compared to CA‐4, the preponderance of resveratrol derivatives in food and the established tolerability of trans‐Res‐3M, we chose to continue our studies with this analogue for in cellulo experiments.

Following our demonstration that phenotypic differences could be observed between the trans‐ and photoisomerized Res‐3M, we next sought to evaluate whether we could achieve in situ isomerization and activation of Res‐3M. We chose to use the EB3‐GFP labelled system given the real‐time readout and sensitivity to small perturbations in microtubule dynamics. HeLa cells stably expressing EB3‐GFP (under antibiotic selection) were incubated with 5 μM Res‐3M and 0.05 μM Ru(bpy)₃Cl₂ and irradiated for 180 seconds at 455 nm (Figure 5). Pleasingly, a stark decrease in the number of comets was observed in the treatment group, while all other combinations of the treatment conditions showed no effects.

Figure 5.

In situ photoisomerization of trans‐Res‐3M disrupts microtubule dynamics. E3B‐GFP HeLa cells were incubated with 5 μM trans‐ Res‐3M, 0.05 μM Ru(bpy)₃Cl₂ and irradiated with an LED light for 180 s at 455 nm. Red boxes indicate the areas corresponding to the expansions shown below each image.

A key limitation of blue light‐mediated photopharmacology platforms is the inherent poor tissue penetration (typically≤0.2 cm) ^[26] which hampers potential clinical application. One key way in which this can be overcome is through the use of two‐photon irradiation. This affords equivalent energy transitions through rapid, sequential absorption of two lower energy (and thus more penetrating) photons by the same molecule. Thus, we sought to explore whether two‐photon excitation could also be applied to our system. Once again, EB3‐GFP HeLa cells were treated with 5 μM Res‐3M (with or without the presence of 10 μM Ru(bpy)₃Cl₂) and irradiated with a 916 nm two‐photon laser over a period of 3 minutes (Figure 6, S33–S35). Within 50 s of irradiation a significant decrease was once again observed in the number of comets (with complete disappearance observed within 125 s) while the treatment group lacking the photocatalyst remained unchanged throughout the course of the experiment. Significantly, these conditions did not lead to the formation of detectable levels of singlet oxygen from the Ru(bpy)₃Cl₂ photocatalyst using a fluorescent dye (Si‐DMA) ^[27] responsive to ¹O₂. While Ru(bpy)₃Cl₂ is known to sensitize the formation of ¹O₂, the results indicate that at the catalyst loading and irradiation power, the cellular redox buffer is sufficient to quench the ¹O₂ generated (see Supporting Information Figures S36 and S37).

Figure 6.

Perturbation of microtubule dynamics occurs within 50 s using two‐photon irradiation at 916 nm to photoisomerize trans‐ Res‐3M in situ. E3B‐GFP HeLa cells were incubated 10 μM of Ru(bpy)₃Cl₂ for 3 hours. After extensive washing cells were treated with 5 μM trans‐ Res‐3M and irradiated for 180 s at 916 nm. A) Raw time course images for cells (and expansions corresponding to the areas indicated in red boxes, below) in the presence or absence of Ru(bpy)₃Cl₂ and quantification of B) time course and C) endpoint comet number (see Figure S25 for images corresponding to catalyst alone or pure negative). P values≤0.05 denoted as *, P values≤0.0001 denoted as ****. See videos 1 and 2 for a time‐lapse of shown conditions (trans‐ Res‐3M+Ru(bpy)₃Cl₂ or trans‐ Res‐3M alone respectively). Expanded version of A) containing all relevant controls is provided in the Supporting Information Figures S33.

Conclusion

In summary, we have demonstrated a mild and biocompatible means of undertaking photocatalyzed geometric isomerization of a range of therapeutically relevant compounds. Not only does this provide exciting opportunities for in situ activation of therapeutic agents with enhanced selectivity due to the spatio‐ and temporal control afforded by irradiation, but also the potential to enhance this through photocatalyst localization to target cell subtypes. For CA‐4, the rate of photocatalyzed isomerization was 1.9×10³ M⁻¹ s⁻¹ at moderate irradiation intensity (1 W LED lamp, 6.2 mW cm⁻²), enabling the reaction to proceed at low substrate concentration and low catalyst loading. Given the simple and rapid nature of the reaction, in addition to its functional group tolerance, one may envisage incorporating such isomerization reactions into the biological screening of alkene containing compound collection. The paucity of drugs with a cis‐stilbene motif is not surprising considering the isomeric instability of the cis‐alkene. In fact, this isomerization has been an issue in the development of combretastatin. Furthermore, the trans to cis isomerization reaction described herein could also be used in designer molecule where the stilbene moiety does not necessarily participate directly into the binding but merely acts as a steric discrimination, enabling drug binding in the cis configuration and precluding drug binding in the trans. This strategy would allow the stilbene moiety to be tuned electronically to the most suitable isomerization configurations. The reaction could also be used to alter the conformation of stapled peptides (i and i+7). While the screening of isomerization of the different bioactive compounds indicates a broad scope, the cis/trans ratio obtained at equilibrium varied. It is possible that tuning the photocatalyst's triplet energy level in order to achieve a more selective energy transfer to the trans vs. cis isomer could be used to optimize the ratio obtained. However, this point is less important if the bioactive compound is the thermodynamically unfavored isomer and there is a broad therapeutic window between the cis and trans isomers, as is the case for CA‐4 and Res‐3M. It should be noted that azobenzene photoswitch also displays a variable cis‐trans isomerization ratio. The salient feature of this chemistry is the fast kinetic of the photocatalyzed reaction at biologically tolerated irradiation. This is by virtue of the efficiency of the energy transfer mechanism. While the photoredox properties of transition metals are increasingly used in chemical biology, there is only one example of a Dexter energy transfer reported to activate a diazirine cross linker. ^[28] The application of an energy transfer to effect an isomerization is yet another example of the application of photocatalysis in chemical biology and therapeutic ends.

Experimental Section

Experimental procedures, characterization data, spectroscopic data and microscopy data can be found in the attached Supporting Information file. The raw data have been deposited on Zenodo https://doi.org/10.5281/zenodo.6542168

Conflict of interest

The authors declare no conflict of interest.

Supporting information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supporting Information

Supporting Information

Supporting Information

E. E. Watson, F. Russo, D. Moreau, N. Winssinger, Angew. Chem. Int. Ed. 2022, 61, e202203390; Angew. Chem. 2022, 134, e202203390.

Data Availability Statement

The data that support the findings of this study are available in the Supporting Information of this article.