Azobenzene Photoswitches in Proteolysis Targeting Chimeras: Photochemical Control Strategies and Therapeutic Benefits

The photoswitchable-PROTACs have photoswitches that work as an actual ON/OFF switch for any particular bioactivity. In reported studies, the ON/OFF activity correlates with the photostationary state and quantum yield of a specific isomer under a definite wavelength of light, providing a spatiotemporal control over protein degradation (an example of photo-chemistry-based precision medicine). The studies reported the role of azobenzenes as photoswitches ( ortho -tetrameth-oxybenzenes, ortho-tetraflurobenzenes as push-pull and push-push) to develop light-controlled PROTACs. This paper dis-cussed the photophysical properties of photoswitchable-PROTACs (such as photoisomerization, quantum yield, and thermal relaxation states) and correlated their photochemical-based structural bioactivity relationships. The proteins studied under photoswitchable-PROTACs are kinase proteins: BCR-ABL fusion protein, ABL; Bromodomain proteins: BRD2, 3, 4; Immunophilins: FKBP12.

The photoswitchable-PROTACs have photoswitches that work as an actual ON/OFF switch for any particular bioactivity. In reported studies, the ON/OFF activity correlates with the photostationary state and quantum yield of a specific isomer under a definite wavelength of light, providing a spatiotemporal control over protein degradation (an example of photochemistry-based precision medicine). The studies reported the role of azobenzenes as photoswitches (ortho-tetramethoxybenzenes, ortho-tetraflurobenzenes as push-pull and push-push) to develop light-controlled PROTACs. This paper discussed the photophysical properties of photoswitchable-PRO-TACs (such as photoisomerization, quantum yield, and thermal relaxation states) and correlated their photochemical-based structural bioactivity relationships. The proteins studied under photoswitchable-PROTACs are kinase proteins: BCR-ABL fusion protein, ABL; Bromodomain proteins: BRD2, 3, 4; Immunophilins: FKBP12.

Introduction: Current medicinal chemistry approaches and underlying shortcomings
Classical medicinal approaches rely on orthosteric and allosteric inhibitors for direct target inhibition. As these approaches mainly focus on the development of small-molecule inhibitors (SMIs), therefore suffer from typical issues of occupancy-driven pharmacology (shown in Figure 1. A), such as (a) frequent occurrence of resistance after prolonged use and (b) require a higher degree of potency to achieve complete inhibition of protein of interest (POI). These issues encouraged the eventdriven pharmacology, as shown in Figure 1. B, to design new type molecules with the ability to degrade the POI (are known as protein degradation) and therefore diminish the related protein functions.

Status of Protein degradation Approaches
As intracellular proteins are responsible for specific cellular functions; therefore, their relative rate of synthesis and degradation is vital for normal cell functioning. Controlling cellular functions also requires a highly organized signaling system that comprises multi-partner proteins, which communicate with each other in a spatiotemporal manner. Therefore, any aberrant change in the intracellular protein expression eventually results in the progression of diseases. One strategy is to degrade those aberrant proteins. Therefore, two stages of the cell cycle [1] are broadly targeted (a) at the genomic level, where modification of DNA knockout the POI, [2] and (b) at the mRNA level, where RNA interference (RNAi) abolishes the expression of a POI. [3] However, DNA knockout and RNAi take longer to deplete the POI, which encouraged researchers to develop event-driven pharmacology-based protein-targeting technologies. Some of the key examples of such proteintargeting technologies are PROTACs (Proteolysis targeting chimeras), [4] HyTs (Low-molecular-weight hydrophobic tags), [5] TRIM-Away. [6] These techniques harness the cell's protein degradation machinery to remove specific POI. Unfortunately, these also have limited applicability because of their submaximal potencies, incompatible cellular updates, and irreversible control over their targeted protein knockdown. [7] Among these, the strategy of PROTACs has enjoyed a wider success against protein targets (EGFR, [8] CDK9, [9] TRIM24, [10] Bcl-2 family proteins (Bcl-2, [11] Bcl-x L, [12] Mcl-1 [11] ) BRD4, [13,14] c-Met, [8] ALK, [15] ERRα, [16] Tau. [17] PROTACs exploit the endogenous ubiquitin-proteasome system to induce protein degradations. As illustrated in Figure 2. A, PROTACs are bifunctional compounds, chemically composed of three parts: Part-A (yellow), a molecule that binds to the protein of interest (POI); Part-B (green), as a linker that joins part-A and -C; Part-C (blue), a molecule which has E3 ligase binding activity (called "E3 ligase ligand", for examples Von Hippel-Lindau (VHL) or cereblon (CRBN)). Unlike conventional inhibitor with higher binding affinity, PROTAC "only" need to hook the POI to bring into proximity to E3 ligase to activate the endogenous ubiquitin-proteasome system (as shown in Figure 2. A). E3 ligase is a protein that recognizes the marked POI and facilitates its protein degradation. As shown in Figure 2. A, one side of PROTAC binds to POI, and the other side binds to E3 ligase protein, forming a ternary complex (POI: PROTAC:E3 ligase), resulting in polyubiquitination and protein degradation. Also, PROTACs have two major advantages over classical SMIs: (a) as submicromolar-to-nanomolar range potency for POI is not required, therefore PROTACs have less-dose dependent toxicity and improved therapeutic indexing, (b) their ability to degrade the POI than POI inhibition, lowers the rate of emergence of resistance. Despite of these advantages, PROTACs share similar design flaws like medium-to-large-sized heterocycles because of their larger molecular size. Also, their systemic application on normal tissue usually leads to the undesired side-effects, for example the use of ARV-771 (a potent BET protein PROTAC), achieved exceptional regression of castrate-resistant prostate cancer in mouse model, but systemic cytotoxicity and deterioration of skin at the injection site was observed. This example exemplifies a need for those structural features which improves specificity and selectivity of PROTACs. Therefore, tools of photopharmacology were investigated. Depending on the chemical nature of photoelements, two major PROTACs strategies were developed: (a) a photo-  After a bifunctional molecule binds to the target protein from one end, the other end binds to E3 ligase protein, which induces protein hydrolysis of a target protein. (C) Hook effect: The effectivity of bifunctional molecules to bind two protein targets (or two active sites) proportionally depends on their bindingcooperativity. If the binding-cooperativity is positive, binding to one protein facilitates subsequent binding to another protein. However, most bifunctional molecules commonly exhibit negative cooperativity; therefore, bifunctional molecules (even at high concentrations) form multiple binary complexes instead of the desired ternary complexes. cleavable protecting group (PPG) (caging or photocaging) [18] as shown in Figure 2. C, and (b) reversible photoswitches [19] as shown in Figure 2. A.

Photoswitches in PROTACs
Irrespective of the effectiveness of photocaged PROTACs in implementing the photochemical elements of photopharmacology, their irreversible activation of the pharmacophore from PPG, requires cellular metabolism for their systemic clearance. Therefore, the incorporation of photoswitches (as shown in Figure 3) was studied as it provides additional reversible photochemical control over the PROTAC-mediated targeting. [20] Based on previously studied photochemical properties of azobenzenes led to their selection as reversible photoswitches for PROTACs (Except in one case of PHOTAC-I-13, where diazocines selected as photoswitch for PROTACs, as shown in Figure 6): (a) resistance towards fatigue, (b) (E)/(Z)-geometrical changes, (c ) facile tuning of photothermal properties, (d) relatively smaller size that does not lead to the molecular obesity of the final molecule. [21,22,23] As azobenzene incorporation leads to two isomers, where active photoisomeric form can be either as the thermodynamically more stable (trans or E)-isomer (azo-PROTAC4C, photoPROTAC-1) or as the metastable (Cis-Z)-isomer (PHOTACs). Structural activity relationship (SAR) studies of azobenzene-photoswitch depend on the following key points: (A) Where is the azo photoswitch incorporated in the PROTAC structure: (a) if directly tethered to E3 ligase ligand, that doesn't directly affect the affinity of PROTACs for respective POI but indirectly can vary the protein degradation activity (for example, similar degradations achieved for PHOTAC-I-1 to 8, but no degradation observed for PHOTAC-I-9, as shown in Figure 6). (b) if incorporated in the middle of POI and E3 ligand and share tethering via aliphatic or aromatic structures, highly exploited to optimize the linker's length, chemical nature,  Figure 7, PHOTAC-II-6 as shown in Figure 8). (c) if placed vicinal to POI ligand, it can directly affect the affinity of the photoswitchable-PROTAC towards its POI. However, synthetic complexity in tethering an azo-switch to the POI ligand; and the risk of complete loss of the overall affinity of PROTAC towards POI, no such attempts were made).

Photoswitchable PROTACs in Kinase protein
The design of Azo-PROTACs by Jin et al., from the School of Pharmacy at China Pharmaceutical University, Nanjing (China) [21] utilized the protein binding of lenalidomide as an E3 ligase ligand (PDB id: 4TZ4) (as can be shown in Figure 4. A). The co-crystal studies illustrated that the lenalidomide binding pocket is relatively smaller and possesses a steric effect at the solvent boundary interface, which led authors to incorporate an azo-unit at the 3-position of lenalidomide. To target POI (BCR-ABL fusion and ABL protein), dasatinib was selected, as it is an II nd -generation ATP-competitive BCR/ABL tyrosine kinase inhibitor used in chronic myelogenous leukemia (CML), acute lymphoblastic leukemia (ALL), and to an extent in CML patients with acquired resistance of imatinib. The co-crystal (Figure 4. B) and molecular modeling of Azo-PROTACs with BCR-ABL fusion protein exhibited only trans-configuration was combinatory due to steric effect after binding of dasatinib ( Figure 4. C [21] ), implying that the cis-and trans-configurations of lenalidomideazo could bring significant degradation activity differences. Later, the authors performed screening to find the suitable length linkers and tested their ABL and BCR-ABL protein degradation abilities on myelogenous leukemia cell line (K562, which expresses the ABL, BCR-ABL, and E3 ligase CRBN). Among all Azo-PROTACs-2C to 6C, Azo-PROTAC-4C was found with an appropriate linker length and higher potency in degrading the BCR-ABL fusion protein, as shown in Figure 5. A.
Also, the authors found substantially different degradation activities with both configurations of Azo-PROTAC-4C in terms of their ABL and BCR-ABL degradation activities. For Azo-PROTAC-4C-trans at 100 nM, noticeable BCR-ABL and ABL protein degradation was observed, while no reasonable Azo-PROTAC-4C-cis protein degradation was even observed at 500 nM, which authors could mention clearly with the "hookeffect" (a common issue with these strategies, explained in Figure 1. C) or based on the presence of photostationary state of Azo-PROTAC 4C-cis in the solution. [20] Time-course analysis for Azo-PROTAC-4C-trans shows ABL protein degradation started after 4 hours (hr), and, substantial BCR-ABL and ABL proteins degradation occurred after 10 hr, and > 90 % BCR-ABL degradation after 32 hr was observed. Contrary to the case of Azo-PROTAC-4C-cis, no reasonable BCR-ABL reduction was observed until 32 hr of treatment, which is surprising as the reported half-life (620 min at 25°C) and, therefore a large proportion of trans-form could be present at earlier time points. [20] To assess the reversible photocontrol on degradation activity, the K562 cells were treated with Azo-PROTAC-4C-trans for 24 hr and then irradiated with UVÀ C light every 4 hr. ABL and BCR-ABL levels were recovered in the UV-irradiated group, but at low levels in the visible light group. Furthermore, preincubated K562 cells with Azo-PROTAC-4C-cis were shielded from light in K562 cells, when transferred into a fresh medium and exposed to visible light BCR-ABL was degraded with time. It would be worth observing the cell viability of Azo-PROTAC-4C-cis, which the authors did not assess. [20]

Photoswitchable PROTACs in Bromodomain proteins
The work of Trauner research group (Department of Chemistry, New York University, New York, USA) in collaboration of Pagano research group (Department of Biochemistry and molecular pharmacology, New York School of Medicine) led to the incorporation of Azo-benzene as photoswitches (as PHOTAC-I) tethered to (+)-JQ1 (a high-affinity inhibitor of BET proteins BRD2-4) and E3-ligase ligands (with, lenalidomide PHOTAC-I-1 to 8 & PHOTAC-II-1 to 4; pomalidomide PHOTAC-I-9 to 13 & PHOTAC-II-6) [23] as shown in Figure 6. Among all azobenzenephotoswitch PROTACs as well as diazocine-photoswitch PRO-TACs, PHOTAC-I-3 was found to be the most effective. The PHOTAC-I-3 exhibited optimal wavelength (390 nm) to switch to the cis isomer, and comparable photostationary states (PSSs) can be attained in the range of 380 and 400 nm.  To access the optical control of BRD2-4 with PHOTAC-I-1 to 13, cell viability assays were used using RS4;11 cell lines (human acute leukemia cells) were treated with 390 nm light pulses (100 ms/10 seconds) for 72 hr or incubated with PHOTACs in dark as a control. This study led to the identification of a promising compound (PHOTAC-I-3) with a 7.1-fold EC 50 difference. The activity difference was found EC 50 = 88.5 nM (when irradiated with 390-nm light) and, EC 50 = 631 nM (in the dark). Others (PHOTAC-I-1,2,4 to 8,10) also showed similar difference activity patterns. Interestingly, PHO-TAC-I-9, 11 to 13, and (+)-JQ1 (control used for these studies) didn't show light-dependent activity differences, which might be related to the role of E3-ligand as PHOTAC-I-1 to 8 (contains lenalidomide) and PHOTAC-I-9 to 13 (pomalidomide features). The light-dependent targeted BET protein BRD2-4 degradation of PHOTAC-I-3 showed a noticeable decrease in BRD3 and BRD4 levels (in range of 100 nM to 3 mM, with 390-nm light but not in the dark). Further protein degradation at above 3 mM of PHOTAC-I-3 (specifically at 10 mM), was not observed because of the "hook-effect". [16,25 , 26] Whereas a milder BRD2 Furthermore, to evaluate the reversible photocontrol of PHOTAC-I-3, initial irradiation for 1 min PHOTAC-I-3 (activating wavelength = 390 nm) and then pulse irradiation (deactivating wavelength = 525 nm) resulted in a drop of cellular BRD2 levels initially but, recovered quickly, even when it was left in the dark.
The color of the incident light can control the extent of protein degradation, as it estimates the ratio of inactive versus active isomeric forms in the PSS (color-dosing, a photocontrol ability to measure the color intensity of incident light to the concentration of the active isomer). The left-shifted curves towards 390 nm (from cell viability assays) and highest protein degradation at 390 nm (western blotting) and, later, steady increase in BRD4 levels with the longer wavelength of the incident light demonstrated an appropriate photochemical reversible control over BRD4 protein degradation.
However, a collaboration of Crews (from Yale University, United States of America) and Carreira (Laboratory of Organic Chemistry, ETH Zürich, Switzerland) coworkers utilized the molecular frame of ARV-771 by replacing the polyether linker region with an azobenzene photoswitch. [22] The authors were prompted by previous studies, which stated that "the length and chemical nature of linker region [27] has a subtle effect on the sensitivity of ternary complex, [27] but requires a minimum length for activity [28] to rotate the conformation to bind POI and E3 ligase protein". Other reports indicated the difference between active and inactive linker's length is 3 Å, which corresponds to the topological distance between cis-and trans-azobenzene photoswitch (in range of 3-4 Å). [29,30] Furthermore, the authors also considered the stability of azobenzene photoswitches, to attain long-lived photostationary states (PSS). Therefore, orthotetrafluoroazobenzenes (o-F 4 -azobenzenes) were introduced as azobenzene-photoswitches, [31] which have (thermal τ 1/2 's ∼ 2 years than few hours when compared to their parent cisazobenzenes [32] ). The author used the ARV-771 skeleton (which is itself is a PROTAC, as shown in Figure 7. A) and replaced its polyether linker region with diamide photoswitch linkers, one with o-F 4 -azobenzene diacid (1 as a representation of pull-pull azoswitch shown in Figure 7. B) and another with o-F 4azobenzene amino-acid (2 as a representation of push-pull azoswitch as shown in Figure 7. B). The photoswitching of o-F 4azobenzene amino-acid (push-pull azoswitch) was found rapid under irradiation as cis-form isomerized to thermodynamically stable trans-isomer with τ1/2's ∼ 2 hours, and also in agreement with previous studies. [33,34] The photoswitching with pull-pull azoswitch type in DMSO found PSS (68 % of cis-photoPROTAC-1 at 530 nm irradiation) and (95 % of trans-photoPROTAC-1 at 415 nm irradiation). The calculated quantum yields for both isomers ([φEZ (530 nm) = 0.28, φZE (415 nm) = 0.65]) were found within the agreement with previously reported for underivatized o-F 4 -azobenzenes. [32] Furthermore, no reverse thermal isomerization of cis-form (even in acetonitrile as well as in aqueous buffer, for several days at 37°C), indicated the suitability and stability of pull-pull azoswitch as photoswitches.
The biological evaluation of trans-photoPROTAC-1 led to BRD2 degradation in Ramos cells. However, no protein degradation was observed until the irradiation with 415 nm,   indicating the relative higher affinity and presence of cis-isomer for either BRD2 or VHL proteins than trans-isomer. Both isomers of photoPROTAC-1 were harvested at different time points with the Ramos cells (1 to 24 hrs) to evaluate their timedependent degradation. For 1-3 hrs, no reasonable activity was found while inbetween 3-7 hrs a high degradation was observed which remained the same until 17 hrs. A concomitant targeting of photoPROTAC-1 with MLN-4924 exhibited proteasomal degradation of BRD2. However, compared to ARV-771 degrader potency, moderate activity for photoPROTAC-1 for BRD2 and loss of activity for BRD4 were based on its distinctive structural features than ARV-771: (a) reversed amide bond between JQ-1 and o-F 4 -azobenzene moiety, (b) intrinsic rigidity of azobenzene. [35] Furthermore, the Ramos cells were treated with either single-irradiated (trans-photoPROTAC-1 at 415 nm; or cis-photoPROTAC-1 at 530 nm) or double-irradiated (trans-photoPROTAC-1 at 530/415 nm; or cis-photoPROTAC-1 at 415/530 nm). Similar BRD2/4 degradation activity was observed for singly irradiated cis-and trans-form of photoPROTAC-1, while doubly irradiated treatments showed reversal of protein degradation, confirming the reverse photoswitchablity of photoPROTAC. To test the spatiotemporal control, the Ramos cells were treated with trans-form in two sets: one set was incubated in the dark while the other was incubated at 530 nm. As expected, cis-form was recovered from the later set. The cells were incubated with cis-form and irradiated at 415 nm to confirm the findings, which led to a significant BRD2 degradation activity. These experiments successfully incorporated o-F 4 -azobenzenes as photoswitches in the PROTAC linker region with spatiotemporal control to induce protein degradation.

Photoswitchable PROTACs in Immunophilins
To generalize the application of optical control in protein degradation, Trauner and Pagano research groups developed prolyl cis-trans isomerase (FKBP12 protein [36] ) based PHOTACs. [23] These six chemical structures (PHOTAC-II-1-6, Figure 8) consist of an E3 ligase ligand (lenalidomide), an azobenzene photoswitch in varied positions, a linker, and SLF (a synthetic ligand of FKBP). Among six PHOTACs, PHOTAC-II-5 and PHOTAC-II-6 were found better tolerable for biological studies. The strong effect on FKBP12 levels upon pulse irradiation by PHOTAC-II-5 after irradiating with 390 nm but not in the dark. The profound activity of PHOTAC-II-5 (for FKBP12 protein) and PHOTAC-I-1-8 (for bromodomain proteins) with similar structural features can be related to their dependence on positioning on azoswitch tethering to the E3 ligase ligand. In the case of PHOTAC-II-6, photoswitch was placed into the linker region and showed similar strong light-dependent degradation like PHOTAC-II-5 with a mild activity in the dark (at 24 hr), which might be the presence of the (E)-isomer remained in the solution. However, the PHOTAC-II-5 and PHOTAC-II-6 were found inactive with MLN4924 exhibiting their FKBP12 protein degradation is CRBN-based, and a noticeable hook effect was also observed during these experiments.

Summary and Outlook
The azobenzene-photoswitches provide a reversible control to PROTACs over specific protein degradation, where one form can thermally relax into an inactive state while the other form gets photochemically isomerized into an active form. It exemplifies an application of azo-switch for selective drug targeting, as compiled in Table 1. Successful incorporation of azobenzene photoswitches in PROTACs is reported by Jin et al.
(Azo-PROTAC-2C-6C for kinase protein degradation [21] ), Reynders et al. (PHOTAC-I-1-13 for bromodomain protein degradation, [22] and PHOTAC-II-1-6 for immunophilins [22] ). Furthermore, photochemical control activates or deactivates any specific protein of interest in a spatiotemporal manner, thereby avoiding the potential systemic toxicities typical to protein degradation methods and serving as a tool for precision medicine. One of the potential shortcomings of these photoswitchable PROTACs is their molecular obesity (around 1000 Da), which results in poor cell permeability and prevents passive tissue diffusion. However, selecting appropriate intermembrane transporters [37] or specific-cell receptors (such as FOLR1 and HER2) for the guided delivery can address these issues. [38,39] Additionally, their conjugations with coenzymes (such as folic acid, pyridoxine) and antibodies could improve their cell selectivity and uptake. [40] Also, the route of administration can play a vital role in enhancing the degradation efficiency, as was observed with other types of PROTACs when administered through intravenous and intraperitoneal injections. [40] The UV light used in these experiments has low tissue penetration and restricts light-controlled PROTACs to topical use, such as leukemia and skin cancers. Therefore, investigating other photoswitches and light sources will mitigate these issues. Additionally, red-shifted azobenzenes could use the Near Infrared region or Infrared region for optimal tissue penetration, [41,42] and two-photon excitation processes may be incorporated for effective photoswitching. [43,44] However, recent development in implantable localized irradiation and optofluidic systems for drug delivery certainly help in expanding the photoswitchable PROTACs' applicability as one of the promising photomedicine strategies.