Discovery of a CNS active GSK3 degrader using orthogonally reactive linker screening

Abstract

Bifunctional targeted protein degraders, also known as Proteolysis Targeting Chimeras (PROTACs), are an emerging drug modality that may offer a new approach for treating neurodegenerative diseases. Identifying chemical starting points for PROTACs remains a largely empirical process and the design rules for identifying Central Nervous System (CNS) active PROTACs have yet to be established. Here we demonstrate a concept of using orthogonally reactive linker reagents, that allow the construction of screening libraries whereby the E3 ligase binder, the target protein binder and the linker can be simultaneously varied and tested directly in cellular assays. This approach enabled the discovery of Glycogen Synthase Kinase 3 (GSK3) PROTACs which are CNS in vivo active in female mice. Our findings provide opportunities to investigate the role of GSK3 paralogs in cellular and in vivo disease models and for the rapid discovery of in vivo quality bifunctional chemical probes for CNS disease concepts.

Fig. 1. An orthogonally reactive linker synthesis and screening platform concept.

a Proposed approach for simultaneous variation of all components of a bifunctional molecule in a library setting with direct-to-biology cellular screening against a protein-of-interest (POI). Created in BioRender. Farnaby, W. (2025) https://BioRender.com/ncfcg8e. b Selected conjugation chemistries, followed by strong cation exchange (SCX) yields direct-to-biology (D2B) libraries. HBD hydrogen bond donors.

Fig. 2. Synthesis of a PROTAC screening library using orthogonally reactive linkers and S_N2/CuAAC chemistry.

a Selected reaction conditions for the D2B S_N2/CuAAC chemistry b design of orthogonally reactive linker reagents to cover a broad physicochemical property space, including aromatic, heteroaromatic, rigid, PEG-based and aliphatic linkers. c E3-ligase binder building blocks. d Selected GSK3-targeting binders containing azide functionalities for CuAAC conjugation. e Library Plate 1: PROTAC library based on binder 19, systematically arranged by linkers in columns and E3-binders in rows, with heat-maps based on UV-purity data. f Library Plate 2: PROTAC library using binder 20, with a similar reaction layout and organisation as in Library Plate 1.

Fig. 3. Identifying GSK3 degraders from a direct-to-biology screen.

a–d HiBiT lytic assay-based high-throughput screening for GSK3β abundance with 24 h PROTAC treatment in GSK3β-HiBiT HEK293 cells. Dose response curves were generated with non-linear 4 parameter fit correction, and representative graphs are combined results of n = 2 biological replicates, each from n = 2 technical replicates with error bars representing mean and ± SD (a shades of blue correspond to individual compounds from Library Plate 1; b Shades of green correspond to individual compounds from Library Plate 2, with the red curve indicating the positive control PT-65). Heatmaps were generated using DC₅₀ values. Compounds with D_max < 30% were considered as having DC₅₀ > 1 µM. c, d Shading intensity reflects compounds DC₅₀ values, with white indicating >1 µM and deep orange indicating <100 nM. e–j Compounds 21 (orange), 22 (pink), 23 (purple), 24 (green), 25 (teal) and 26 (blue) were selected for resynthesis and further characterisation, which were originated from wells C7, F2, F12, D2, D7 and D12 of library plate 2, with the GSK3 inhibitor 29 (black) included as a negative control for comparison. Potency of these purified compounds were compared with crude mixtures (dashed lines) for GSK3β degradation upon 24 h treatment in GSK3β-HiBiT HEK293 cells (Dose response curves are means of n = 3 biological replicates ± SD, 4-parameter non-linear curve fitting (GraphPad)). k, l Kinetic live cell degradation of GSK3β with selected compounds 24 (green) and 26 (blue), respectively, monitored in LgBit overexpressing GSK3β-HiBiT KI HEK293 cells where shading intensity reflects compound concentration, with darkest shading indicating highest tested concentration (3 µM) and lightest shading indicating lowest tested concentration (0.01 µM) (n = 3 biological replicates, and data representatives of one biological replicate per compound with error bars representing mean of technical replicates and ± SD).

Fig. 4. Characterisation of cellular GSK3 degrader probes.

a Endogenous GSK3 paralog degradation with compounds 24 and 26 and negative controls 27 and 28 in HEK293 cells (representative blots of n = 3 biological replicates, 2–4 h). The samples derive from the same experiment, gels and blots were processed in parallel. b, c Immunoblots shown in panel A were quantified, normalised to relevant internal controls and plotted for changes in GSK3β and GSK3α levels. Compound 27 in light grey, 24 in green, 28 in dark grey, and 26 in blue, with shading intensity corresponding to concentration (lightest = 10 nM, darkest = 30 nM) (n = 3 biological replicates with error bars representing mean and ± SD, p values are calculated as <0.0001, two-tailed unpaired t-test (GraphPad)). d Mechanism of action studies in HEK293 cells for compounds 24 and 26 showing rescue with 2 h pre-treatment of proteasome inhibitor (1 µM, MG132), neddylation inhibitor (1 µM, MLN4924) and GSK3 inhibitor (1 µM, 29). Sample processing controls run on different gels in parallel (Representative blot of n = 3 biological replicates). e, f Immunoblots shown in (d) were quantified, normalised to relevant internal controls and plotted for changes in GSK3β and GSK3α levels, 24 in green and 26 in blue (n = 3 biological replicates with error bars representing mean and ± SD). g Evaluation of changes in phosphorylation levels of GSK3β following compound 24, 27 or GSK3 inhibitor (29) treatments for 30 and 120 min in HEK293 cells (n = 3 biological replicates). The samples derive from the same experiment, gels and blots were processed in parallel. h Dose dependent effect of compound 24 on GSK3 substrates in differentiated SH-SY5Y cells (representative of n = 3 biological replicates, at indicated concentrations, 24 h). The samples derive from the same experiment, gels and blots were processed in parallel.

Fig. 5. Global and phospho-proteomic analysis of KH1 (24).

Proteome wide compound selectivity for compound 24 (green) vs 27 (grey) normalised to DMSO, following 10 nM treatment for 2 h in (a) and 24 h in (b) in HEK293 cells (n = 3 biological replicates). Phosphoproteomic analysis of compound 24 (purple) and 27 (grey) normalised to DMSO following 10 nM treatment for 2 h in (c) and 24 h in (d) in HEK293 cells (n = 3 biological replicates). Two-tailed t-test performed with 250 randomisations to assess the statistical significance between groups using Perseus software.

Fig. 6. In vivo PK/PD profiling of KH1 (24).

a Pharmacokinetic profile of KH1 (24) in female Balb/c mice following a single intravenous (i.v.) bolus dose of 0.37 mg/kg (black dot) or oral (p.o.) gavage dose of 3.0 mg/kg (green cross). Mean of samples from three mice per dose route showing whole blood concentration-time profile. b Serum shift assay. Degradation of GSK3β-HiBiT by KH1 (24) in GSK3β-HiBiT KI HEK293 cells in presence of 10% foetal bovine serum (FBS) (green) or mouse serum (MS) (black) (24 h, n = 3 biological replicates, with error bars representing mean and ± SD, 4-parameter non-linear curve fitting (GraphPad)). c Immunoblotting based endogenous degradation profile of KH1 (24) in mouse embryonic fibroblast (MEF) cells for 4 h with indicated concentrations (Data representative of n = 3 biological replicates). d Dose dependent GSK3β degradation of KH1 (24) in MEF cells (4 h, n = 3 biological replicates, with error bars representing mean, ± SD, 4-parameter non-linear curve fitting (GraphPad)). e Assessment of GSK3β in liver and brain following single i.v. dose of KH1 (24) 5 mg/kg in female Balb/c mice after 4 h, data shows samples from three mice compared to control (CNT). f Relative quantification of GSK3β levels in brain and liver from (e) following a single i.v. bolus dose of  5 mg/kg of KH1 (24) (green) compared to control (CNT) (black) (n = 3 mice, error bars representing mean and ± SD, two-tailed unpaired t-test, p values calculated as p = 0.0159 for brain samples and p < 0.0001 for liver samples (GraphPad)). g Whole Proteome (green) and h Phospho-proteome (purple) analysis of liver tissue samples (n = 3 mice) collected from the pharmacodynamic study as detailed in (e). Two-tailed t-test performed with 250 randomisations to assess the statistical significance between groups using Perseus software.