Cereblon versus VHL: Hijacking E3 ligases against each other using PROTACs

Graphical abstract


Introduction
Targeting proteins for degradation by hijacking the ubiquitin-proteasome system with small molecules is a powerful modality of intervention into biology, and an emerging therapeutic strategy. [1][2][3][4] A primary approach to targeted protein degradation involves the design of PROTACs (PROteolysis-Targeting Chimeras). PROTACs are bifunctional compounds that form a ternary complex with a target protein of interest and an E3 ubiquitin ligase, such that the target protein is ubiquitinated by the hijacked E3 ligase and subsequently degraded by the proteasome. 5,6 PROTACs are defined by a catalytic, sub-stoichiometric mode of action that can allow for rapid, profound and selective target depletion inside cells, and an extended duration of action, also in vivo. [7][8][9][10] Because their mode of action differs from that of conventional inhibitors, the concentrations at which PROTACs exert degradation activity are often much lower than expected based on their dissociation constants with the target protein. [11][12][13][14] Furthermore, PROTAC's selectivity can be greater than the binding selectivity of the ligands alone, allowing to discriminate between highly similar proteins or isoforms in ways that are not possible with occupancy-based inhibitors. 8,11,12,[15][16][17] Within the past four years, potent and selective PROTACs have been designed to hijack either the von Hippel-Lindau (VHL) or cereblon (CRBN) E3 ligase against a target protein of interest. 18,19 Targets that have been shown to be degraded by PROTACs include members of bromodomain-containing proteins such as the BET proteins (Brd2, Brd3 and Brd4), [7][8][9]14,15,17,20,21 amongst other epigenetic protein classes; [22][23][24][25][26] protein kinases; 10,12,[27][28][29][30][31] as well as non-bromodomain and non-kinase target proteins. [32][33][34][35] Recent progress in understanding principles of PROTAC mode of action, and demonstration of applicability across different target classes, suggest that PROTACs have the potential to target new protein families, including proteins that are difficult to block using current approaches. Clinical validation of small molecules inducing protein degradation is provided by recent discoveries on the molecular mechanism of thalidomide and related clinical anticancer immunomodulatory drugs (IMiDs) such as lenalidomide and pomalidomide, which induce the proteasomal-dependent degradation of cancer-driving proteins. 36,37 More recently, a PROTAC compound (ARV-110) that targets the androgen receptor for degradation has been announced as a clinical candidate. 38 E3 ubiquitin ligases are key players in the ubiquitin-proteasome pathway because they catalyse ubiquitination of substrate proteins. [39][40][41] As important regulators of cellular ubiquitination, E3 ligases are emerging as attractive drug targets, particularly in cancer. [42][43][44] However, E3 ligases have proven difficult to target using small molecule inhibitors. So far only few high-quality inhibitors have been developed, mainly against the ligases MDM2, 45 VHL, 46 and IAPs. 47 E3 T ligases lack deep binding sites to accommodate endogenous small-molecule cofactors or substrates, as is the case for ATP in protein kinases. 48 Targeting E3 ligases therefore requires disruption (or modulation) of protein-protein interactions. 49 E3 ligase inhibitors face particular challenges: first, the difficulty to compete with high-affinity endogenous substrates, which increase in level as a result of E3 blockade; 50 and second, the observation that small molecules that bind to E3 ligases may modulate the surface of the targeted E3 in such a way that new substrate proteins are recruited for degradation, as shown for the E3 ligases CRBN, 37,51,52 and DCAF15. 53,54 We hypothesized that the E3 ligases themselves might be hijacked against one another using a PROTAC approach, thus inducing E3 ligase degradation as opposed to E3 blockade. In 2017, we disclosed the first report of a small molecule dimerizer of an E3 ligase as a means to induce its own degradation, an approach that we called "homo-PROTAC". 11 We designed bifunctional molecules made up of the same ligand for the ubiquitously expressed VHL protein, connected via a linker, that would induce VHL dimerization as the key step to trigger VHL ubiquitination and subsequent degradation. The best degrader, the symmetric homo-PROTAC CM11 (Figure 1), dimerized VHL in vitro with high avidity (cooperativity) of ∼20-fold, leading to potent, complete and prolonged degradation of VHL in different cell lines. With CM11, we confirmed the hypothesized mechanism and qualified a novel chemical probe degrader for VHL. 11 Subsequently, the same idea was applied by Krönke, Gütschow and co-workers, who reported homo-PROTACs for the CRBN ligase, and showed compound 15a (CC15a in Figure 1) to be the most active compound. 55 As an extension of our homo-PROTAC approach, we envisaged that two different E3 ligases could be brought together using hetero-bifunctional PROTACs made of a ligand handle for one ligase and another handle for a different ligase. 56 We hypothesized that with such compounds the two E3 ligases might be hijacked against one another, leading to two potential scenarios: 1) both ligases being degraded in cell; 2) one of the two being preferentially degraded -resulting in one ligase 'winning' over the other one. In the present study, we describe the design, synthesis and cellular activity of VHL-CRBN heterodimerizing PROTACs, and interrogate the outcome of hijacking these two E3 ligases against each other.

Design of a library of CRBN-VHL PROTACs
In order to better explore potentially different relative orientations between the two E3 ligases, we began by designing three series of heterodimerizers characterized by different attachment points on the VHL ligase handle ( Figure 2): 1) out of the terminal acetyl group of VHL ligand VH032. 50,57 Amidation of a terminal tert-Leu of the VHL ligand (compound 1, Figure 3) is a widely-explored conjugation strategy for   PROTACs, including our homo-PROTAC CM11; 11 2) via a phenolic substituent out of VH101, a more potent VHL ligand in which the cyano-cyclopropyl group of chemical probe VH298 is replaced with a fluoro-cyclopropyl group, as shown in our published SAR of VHL ligands. 46 Successful conjugation of this optimized VHL ligand (compound 2, Figure 3) was recently reported by our laboratory in Brd9 degrader VZ185; 23 3) via a thioether linkage out of the tert-butyl group of the VHL ligand, in which the tert-Leu group is replaced with a penicillamine group, as we previously incorporated in Brd4-selective degrader AT1. 15 Unlike AT1, in which the VHL ligand handle bears a terminal acetyl group, here we decided to keep the terminal fluorocyclopropyl group as in VH101. As CRBN handle, we chose pomalidomide because of its greater cellular stability compared to other IMiDs. 58 To derivatize pomalidomide we appended an ethylenediamine spacer out of the phthalimide ring (compound 3, Figure 3), to provide a synthetically convenient attachment point for amide conjugation of a linker. 23,29 The linker plays a crucial role in PROTAC design and activity. Small changes in both length and physicochemical nature e.g. alkylic versus polyethylene glycol (PEG) as well as mixtures thereof, are known to impact degradation activity and selectivity in often unpredictable ways. 21,22,31 We therefore decided to explore different linkers, focusing on varying lengths and ratio between carbon and oxygen atoms, as we and others have found that these modifications can have a profound impact on PROTAC structure-activity relationships. [21][22][23]31 As a result, the designed compounds explore diversity in the derivatization point, linker length and chemical properties.

First series of PROTACs.
The first series of VHL-CRBN PROTACs ( Figure 2) comprises compounds 7a,b and 14a-e. Compounds 6a and 6b, bearing respectively a 2 and 4 PEG unit linkers, were synthesized as previously reported. 11 Briefly, triethylene or pentaethylene glycol were first converted to monobenzyl ethers and then reacted with tert-butyl bromoacetate under biphasic conditions to yield linkers 4a-b in good yields (SI Scheme 1). After deprotection of the benzyl group by catalytic hydrogenation, the primary alcohol was oxidised to carboxylic acid and subsequently coupled with VHL ligand 1, as described, 11 to afford compounds 6a-b (SI Scheme1). Deprotection of the tert-butyl group in acidic condition followed by coupling with CRBN ligand 3 afforded the final PROTACs 7a-b in 95% and 84% yield, respectively (Scheme 1).
For the synthesis of PROTACs 14a-e, symmetric linkers 12a-e bearing two terminal carboxylate groups were designed with different length and composition. Compounds 12a-e were prepared starting from the corresponding diols 10a-e. Diol 10b-e were commercially available, instead 10a was synthesized in house by adapting a previously reported method. 31 Briefly, 10a was obtained after a nucleophilic substitution reaction between the tosyl derivative of a monobenzyl protected 1,5 pentadiol and ethylene glycol in a 2:1 ratio, followed by deprotection of the benzyl group by catalytic hydrogenation (SI Scheme 2). Nucleophilic substitution in phase transfer catalysis of diols 10a-e followed by deprotection in acidic conditions, based on our previously reported synthetic route, 11 delivered compounds 12a-e (SI Scheme 3).
Subsequently, mono N-hydroxysuccinamide ester derivatives of 12a-e, obtained via reaction with N-hydroxysuccinimide (NHS) and N,N'-dicyclohexylcarbodiimide (DCC), were reacted with CRBN ligand 3 in a 2:1 ratio to afford 13a-e (Scheme 2). The NHS activation of the linkers was required in order to better control the reaction and to reduce the formation of 2:1 conjugates between 3 and linkers. After removal of dicyclohexylurea (DCU) side product by filtration, the 1:1 conjugates 13a-e were subsequently coupled with 1 to obtain the final PROTACs 14a-e in 42-62% yields (Scheme 2).

Second series of PROTACs.
Linkers for the second PROTAC series ( Figure 2) were designed to contain a carboxylic group protected as tert-butyl ester on one side and a leaving group on the other side, which could be coupled with the phenol group on VHL ligand 2. Linkers 15a-b were synthesized as previously reported, 31 and their alkyl iodide derivatives 16a-b were prepared by reaction of the alcohol group with Ph 3 P·I 2 reagent prepared in situ (Scheme 3). Ligand 2 was reacted with compounds 16a-b and commercially available methyl 5-bromobutanoate (16c) in the presence of K 2 CO 3 to afford 17a-c, respectively, in good yields. Final PROTACs 18a-c were obtained upon deprotection of either the tert-butyl group, in case of 17a-b, or the methyl group for 17c, and subsequent amide coupling with CRBN ligand 3, using (1-cyano-2-ethoxy-2-oxoethylidenaminooxy)dimethylamino-morpholino-carbenium hexafluorophosphate (COMU) as coupling reagent and N,N-diisopropylethylamine (DIPEA) as base (Scheme 3).

Third series of PROTACs.
For the synthesis of this series of PROTACs, VHL ligand 20 was synthesized in two steps: a first coupling reaction of previously reported compound 19 (ref. 15 ) with 1-fluorocyclopropane-1-carboxylic acid, followed by deprotection of the thiol moiety (Scheme 4). Linkers 16a-c were connected to 20 via a sulphur alkylation reaction in the presence of DBU as the base. Deprotection of the tert-butyl ester group of 21a-c, and subsequent coupling with 3 under the same conditions described above, delivered the final compounds 22a-c in good yields (Scheme 4).

Evaluation of PROTAC cellular activity.
To profile the degradation activity of our panel of PROTACs, VHL and CBRN protein levels in HeLa cells were quantified by western blot analysis following a 4 h treatment with 1 µM compounds, using CM11 and CC15a as positive controls for VHL and CRBN degradation, respectively ( Figure 4). Interestingly, we observed significant degradation of CRBN with a few compounds, while no significant degradation of VHL was observed with any of the compounds tested. The most profound CRBN degradation was observed with PROTAC 14a (64% protein degradation, as quantified by western blot), followed by compound 18b which induced CRBN degradation to a lower extent (54% degradation). The same screen was conducted in a different cell line (HEK293), confirming 14a as the most potent compound at inducing CRBN degradation (data not shown). To provide a more stringent screen the same experiment was conducted by testing compounds at 10 nM in HeLa cells ( Figure S1). The purpose of this experiment was to exclude the possibility of dismissing any potent compound as a false negative potentially due to the "hook effect" characteristic of bivalent molecules: whereby unproductive binary complexes preferentially form at high PROTAC concentration, which compete with and eventually suppress the formation of a productive ternary complex. 9 PROTAC 14a induced less CRBN degradation (19% protein degradation) at 10 nM compared to 1 µM, as expected. Importantly, CRBN protein levels remaining after treatment at 10 nM were not significantly lower than the levels remaining after the same treatment at 1 µM (cf. Figure S1 with Figure 4), making it unlikely that any of the compounds might be false negatives due to a hook effect. Interestingly, at this lower concentration some compounds appeared to induce up to 50% degradation of pVHL30 ( Figure S1). This suggests that depending on the concentration being used this class of compounds could preferentially induced the depletion of one ligase over the other. Encouraged by the promising and consistent degradation of CRBN observed with PROTAC 14a, we selected this compound for further characterization. We next profiled the concentration-and time-dependent activity of 14a in both HeLa ( Figure 5) and HEK293 cells ( Figure  S2). Compound 14a degraded CRBN with a half-degrading concentration DC 50 (i.e. the concentration causing 50% reduction of protein level relative to vehicle) of 200 nM, and reached a maximal degradation (D max ) of 75% after 4 h treatment with 1 µM. A hook effect was observed above 1 µM, indicating that 14a preferentially forms the 1:   50 and D max were found for 14a in HEK293 ( Figure S2). Again, some concentration-dependent depletion of pVHL30 was seen at the lower end of the concentration range (5-50 nM) in HeLa ( Figure 5). However interestingly this effect was not observed in HEK293 ( Figure S2). From the time-course data, compound 14a was able to induce rapid degradation, with > 50% CRBN levels relative to control depleted already after 1 h; maximal degradation > 80% was attained after 8 h ( Figure 5). The 14a-induced degradation of CRBN was found to be even faster in HEK293, with > 80% protein already depleted after 1 h, and 98% degradation achieved after 8 h ( Figure S2). Once again the compound displayed selectivity for CRBN, as there was no appreciable VHL degradation at 1 µM over the time points tested in either cell line ( Figure 5 and Figure S2).

Discussion
We described dually targeting CRBN-VHL PROTACs, developed with the aim of investigating the relative ability of CRBN and VHL E3 ligase to induce degradation of one other. Among the three series of compounds developed, we observed preferential degradation of one ligase i.e. CRBN over the other one (VHL) with some of the compounds from two of the series. The most potent PROTAC, compound 14a, induced CRBN degradation with high potency (DC 50 of 200 nM) and to profound levels (D max of up to 98%) and rapidly (within 1 h of treatment). Further structure-activity relationships could help to better understand and improve the already high potency and efficiency of CRBN degradation achieved with 14a.
Our data thus suggests that VHL can 'win the battle' with CRBN when the two ligases are brought together by a PROTAC. Future mechanistic studies are warranted to attempt to elucidate the contributors for this preferential unilateral outcome of our 'double-hijacking' approach. We also cannot exclude that different combinations of conjugation patterns (via different attachment points for example) and linker lengths and structures of CRBN-VHL PROTACs might be able to discriminate different relative orientation of the ternary complex in such a way that the outcome might become reverse, i.e. VHL being preferentially degraded over CRBN -a hypothesis that will be tested in future work. In this regard, it is interesting to note that minor concentration-dependent depletion of pVHL30 was observed at the lower end of the concentration range (5-50 nM) in HeLa ( Figure 5) as well as in the screen at lower compound concentration (10 nM, Figure S2). pVHL30 is the VHL isoform that is preferentially degraded by the homo-PROTAC CM11. 11 No observable PROTAC-induced degradation of pVHL19 was instead observed with any of our compounds, consistent with the cellular outcome observed with CM11. These observations together suggest an enticing possibility that differential ligase degradation might apply at distinct ranges of concentration of CRBN-VHL dimerizers. Differential absolute concentration between the two E3  ligases, and/or differential binding affinities of each end of the bivalent molecule for its respective ligase, are likely to be amongst the contributing factors that could effectively skew the hook effect towards one ligase versus the other one depending on the PROTAC concentration, ultimately imparting differential protein degradation outcomes. Such an effect could be of relevance in a broader context for other E3 ligase pairs. It is noteworthy that a recent study reported MDM2 PROTAC degraders, designed by linking an MDM2 inhibitor via either a thalidomide-based CRBN ligand or a VHL ligand. 59 Potent and selective PROTAC-induced degradation of MDM2 was observed for the CRBN-MDM2 heterodimers. However notably, protein level of the hijacked CRBN or VHL ligases were not monitored. 59 Hetero-bifunctional VHL-CRBN PROTACs were also disclosed in a study recently published by Steinebach et al. 60 Preferential degradation of CRBN over VHL was also observed by Steinebach et al., with their most potent compound (CRBN-6-5-5-VHL) being a conjugate of pomalidomide and VHL ligand via the terminal acetyl group, as with 14a, albeit with a different linker structure. 60 Our study provides proof of principle for dimerizing two different E3 ligases as a novel approach to inducing one ligase to degrade the other one. The outcome of 'ligase versus ligase' PROTAC-mediated activity might be unpredictable a priori, but could reveal a new mechanism for proximity-mediated hijacking between E3 ligases. Future work is warranted to interrogate many more combinations of E3 ligases and hetero-dimerizer compounds to bring E3 ligases together as a mechanism to induce their intracellular degradation. Given the number of E3 ligases predicted to function in cells (up to 600) this approach could speed up our ability to chemically intervene on E3 ligase themselves using targeted protein degradation, with both biological and therapeutic benefits.
All reactions were carried out using anhydrous solvents. Analytical thin-layer chromatography (TLC) was performed on precoated TLC plates (layer 0.20 mm silica gel 60 with fluorescent indicator (UV 254: Merck)). The TLC plates were air-dried and revealed under UV lamp (254/365 nm) or permanganate stain. Flash column chromatography was performed using prepacked silica gel cartridges (230-400 mesh, 40-63 mm; SiliCycle) using a Teledyne ISCO Combiflash Companion or Combiflash Retrieve using the solvent mixtures stated for each synthesis as mobile phase. Preparative HPLC was performed on a Gilson preparative HPLC with a Waters X-Bridge C18 column (100 mm × 19 mm; 5 μm particle size, flow rate 25 mL/min). Liquid chromatography-mass spectrometry (LC-MS) analyses were performed with either an Agilent HPLC 1100 series connected to a Bruker Daltonics MicroTOF or an Agilent Technologies 1200 series HPLC connected to an Agilent Technologies 6130 quadrupole spectrometer. For LC-MS the analytical cololum used was a Waters X-bridge C18 column (50 mm × 2.1 mm × 3.5 mm particle size); flow rate, 0.5 mL/min with a mobile phase of water/MeCN + 0.01% NH 4 OH (basic analytical method) or water/MeCN + 0.01% HCOOH (acidic analytical method); 95/5 water/MeCN was initially held for 0.5 min followed by a linear gradient from 95/5 to 5/95 water/MeCN over 3.5 min which was then held for 2 min. The purity of all the compounds was evaluated using the analytical LC-MS system described before, and purity was > 95%. 1 H NMR and 13 C NMR spectra were recorded on a Bruker Avance II 500 spectrometer ( 1 H at 500.1 MHz, 13 C at 125.8 MHz) or on a Bruker DPX-400 spectrometer ( 1 H at 400.1 MHz, 13 C at 101 MHz). Chemical shifts (δ) are expressed in ppm reported using residual solvent as the internal reference in all cases. Signal splitting patterns are described as singlet (s), doublet (d), triplet (t), multiplet (m), or a combination thereof. Coupling constants (J) are quoted to the nearest 0.1 Hz.

General method to obtain di-tert-butyl protected carboxylate (A):
To a solution of diol (1 eq.) in dichloromethane (DCM) (4 mL per mmol), tert-butyl bromoacetate (8 eq.), TBABr (1.1 eq.) and 37% w/w aqueous NaOH (4 mL per mmol) were added. The biphasic reaction was vigorously stirred at room temperature (r.t.) overnight. The organic phase was separated from the aqueous layer and then the aqueous phase was extracted with DCM (x3). Organic layers were collected, dried over MgSO 4 and evaporated under reduced pressure. The crude was purified by flash chromatography (using a gradient from 10 to 100% of ethyl acetate in heptane).

General method B:
A solution of the starting material in a 50% v/v trifluoroacetic acid (TFA) in DCM (6 mL per mmol) was stirred at r.t. for 2 h. TLC analysis (10% methanol in DCM) showed complete conversion of the starting material. Then, the reaction mixture was concentrated under reduced pressure and the crude was freeze-dried to obtain the desired product.

General method C:
Potassium tert-butoxide (1 eq.) was added to polyethylene glycol (8 eq.) in anhydrous THF (0.2 mL/mmol) at 0°C. The resulting mixture was stirred at 60°C for 0.5 h, then it was cooled to r.t. A solution of tertbutyl-bromoacetate (1.0 eq.) in anhydrous THF (0.1 mL/mmol) was added to the reaction mixture at r.t. The resulting mixture was stirred at r.t. for 24 h. The reaction was quenched with brine and the aqueous phase was extracted with ethyl acetate. The combined organic phase was evaporated to dryness. The crude material was purified by column chromatography (from 0 to 8% of methanol in DCM) to afford the desired compound.

General method D:
Iodine (1.3 eq.) was added to triphenylphosiphine (1.3 eq.) and imidazole (1.3 eq.) in DCM (7 mL/mmol) at 0°C. The resulting mixture was stirred at r.t. for 5 min, then was cooled to 0°C. A solution of alcohol (1.0 eq.) in DCM (3 mL/mmol) was added to the reaction mixture at 0°C and the resulting mixture was stirred at r.t. for 3 h. TLC analysis (50% ethyl acetate in heptane) showed complete conversion of the starting material. The reaction was quenched with saturated NaHCO 3 solution and saturated Na 2 SO 3 solution and the aqueous phase was extracted with ethyl acetate. The combined organic phase was evaporated to dryness. The crude material was purified by column chromatography (from 20 to 75% of ethyl acetate in heptane) to afford the desired compound.

General method E:
The dicarboxylic acid linker (1 eq.) and NHS (1.1 eq.) were dissolved in dry DCM (∼10 mL per mmol). DCC (1.2 eq.) was added and the reaction was left to stir overnight. The DCU was filtered off, the solution was evaporated and the residue dissolved in dry DMF. Compound 3 (0.5 eq.) and DIPEA (3 eq.) were added. The reaction mixture was left to stir at r.t. for 2 h, quenched with ice, dried under high vacuum and purified by HPLC using a gradient from 10% to 80% v/v acetonitrile with 0.01% v/v aqueous solution of formic acid over 15 min to yield the desired compound.

General method F:
To a solution of carboxylic compound (1 eq.) in dry DMF (∼50 mL per mmol), COMU (1 eq.), compound 1 (1.1 eq.) and DIPEA (3 eq.) were added. The reaction mixture was left to stir for 1 h and monitored by LC-MS (acidic method). When completed, ice was added to quench the reaction, the volatiles were evaporated under reduced pressure and the residue purified by HPLC with a gradient from 5% to 90% v/v acetonitrile with 0.01% v/v aqueous solution of formic acid over 15 min to yield the desired compound.

General method G:
To a solution of 20 (1 eq.) and the linker (1.1 eq.) in dry DMF (∼14 mL per mmol), DBU (1.1 eq.) was added at 0°C under a nitrogen atmosphere. The reaction mixture was stirred at r.t. for 4 h and monitored by LC-MS (acidic method). The reaction was quenched with a 5% v/v aqueous solution of citric acid and the solvent was evaporated under high vacuum. The crude was purified by HPLC using a gradient from 5% to 90% v/v acetonitrile with 0.01% v/v aqueous solution of formic acid over 15 min to yield the desired compound.

General method H:
To a solution of the carboxylic compound (1 eq.) in dry DMF (∼100 mL per mmol), COMU (1 eq.), compound 3 (1.1 eq.) and DIPEA (3 eq.) were added. The reaction mixture was left to stir for 1 h and monitored by LC-MS (acidic method). Then, ice was added to quench the reaction, the volatiles were evaporated under reduced pressure and the residue purified by HPLC using a gradient from 5% to 90% v/v acetonitrile with 0.01% v/v aqueous solution of formic acid over 15 min to yield the desired compound.

General method I:
Compound 2 (1 eq.), K 2 CO 3 (3 eq.) and the halogenated linker (1.5 eq.) was dissolved in DMF (∼50 mL per mmol) and heated at 70°C overnight. Complete conversion of the starting material was observed by LC-MS (acidic method). The reaction mixture was taken up with water and extracted with DCM (x3). Organic layers were collected, dried over MgSO 4 , evaporated under reduced pressure and purified by HPLC using a gradient from 5% to 95% acetonitrile with 0.01% v/v aqueous solution of formic acid over 10 min to yield the desired compound.