Structure-Guided Design and Optimization of Covalent VHL-Targeted Sulfonyl Fluoride PROTACs

Proteolysis-targeting chimeras (PROTACs) are heterobifunctional molecules that have emerged as a therapeutic modality to induce targeted protein degradation (TPD) by harnessing cellular proteolytic degradation machinery. PROTACs which ligand the E3 ligase in a covalent manner have attracted intense interest; however, covalent PROTACs with a broad protein of interest (POI) scope have proven challenging to discover by design. Here, we report the structure-guided design and optimization of Von Hippel-Lindau (VHL) protein-targeted sulfonyl fluorides which covalently bind Ser110 in the HIF1α binding site. We demonstrate that their incorporation in bifunctional degraders induces targeted protein degradation of BRD4 or the androgen receptor without further linker optimization. Our study discloses the first covalent VHL ligands which can be implemented directly in bifunctional degrader design, expanding the substrate scope of covalent E3 ligase PROTACs.


■ INTRODUCTION
The Von Hippel-Lindau (VHL) protein is among the most widely recruited E3 ligases in the PROTAC field.−8 This motif is considered essential for VHL recognition and presents a challenge for the optimization of VHL PROTAC potency and cell uptake.
−15 Recently reported covalent CRBN E3 ligase binders bearing a sulfonyl fluoride show intriguing molecular glue activity, although they have yet to be incorporated in a covalent PROTAC. 16erein, we document the design and optimization of the first rationally designed covalent VHL ligands and their incorporation in PROTACs for TPD applications.We demonstrate that the hydroxyproline motif of a known VHL binder can be replaced by a sulfonyl fluoride moiety, and through structure-guided optimization, we generated a ligand which covalently modifies Ser110 of VHL in the HIF1α binding site.We systematically assess VHL occupancy in recombinant proteins and live cells and the capacity of covalent PROTACs derived from this ligand to induce degradation of BRD4 and androgen receptor (AR).We suggest that this novel covalent VHL PROTAC paradigm will prove valuable for future studies of target engagement and optimization of the pharmacokinetic and pharmacodynamic properties of VHLrecruiting PROTACs.
■ RESULTS AND DISCUSSION First-Generation Sulfonyl Fluoride Covalent VHL Ligand.The binding mode of VH032, the most widely exploited VHL ligand in PROTACs to date, features a critical hydrogen bond to Ser110 through the (R)-hydroxyproline motif (Figure 1B). 8In our initial ligand designs, we sought to replace the hydroxy group with a sulfonyl fluoride, an electrophilic warhead featuring balanced reactivity, resistance to hydrolysis under physiological conditions, and the capacity to covalently modify proteins at varied nucleophilic residues beyond cysteine, including serine. 17,18We investigated how the hydroxyl replacement would perturb VHL binding by docking the prototype covalent ligand VHL-SF1 (Figure 1C) in VHL derived from a VH032/VHL complex (PDB: 4W9H) (Figures 1D and S1). 2 These models suggest that VHL-SF1 has the potential to covalently bind Ser110, thereby maintaining some of the critical interactions observed in the hydroxyproline motif despite a degree of displacement of the remainder of the molecule (Figure S1), which may compromise the noncovalent interactions exhibited by VH032.We considered VHL-SF1 a reasonable starting point to probe covalent VHL modification.
Synthesis of VHL-SF1 commenced with the generation of (S)-hydroxyproline through a short sequence of reactions (Scheme S1).Displacement of mesylate 1 with thioacetate afforded compound 2, which was converted to sulfonyl fluoride 4 via sulfonyl chloride 3 (Scheme 1).While this reaction generated the desired sulfonyl fluoride, epimerization of the proline ring occurred, resulting in a mixture of diastereomers detected by LC−MS analysis.To circumvent this, we developed novel reaction conditions to synthesize the desired sulfonyl fluoride in one step directly from thioacetate 2, resulting exclusively in the desired stereoisomer 5.A proposed mechanism for the novel sulfonyl fluoride transformation based on known analogous reactions is shown in Scheme S2. 19 Following Boc-deprotection, the amine was acetylated to afford VHL-SF1 or biotinylated to generate VHL-SF1-Biotin.
To assess the ability of VHL-SF1 to covalently modify VHL, we initially developed a streptavidin shift assay in which we exposed recombinant human VCB, a stable complex of VHL protein with elongin C and elongin B, to VHL-SF1-Biotin.VHL biotinylation could then be directly quantified by the apparent shift in molecular weight observed when the sample was mixed with streptavidin and analyzed by an anti-VHL Western blot (Figure S2). 20Through this assay, we concluded that 10 μM VHL-SF1-Biotin modified 32% VHL following 2 h of incubation at room temperature, which was further confirmed to be concentration-dependent with respect to the probe (Figure S3).
Consistent with this modest reactivity, VHL-SF1 at concentrations up to 100 μM was unable to displace a fluorophore-labeled HIF1α peptide, known to occupy both the VH032 binding site and a second VHL domain, in a competitive fluorescence polarization (FP) assay (Figure S4).We next focused on developing a second generation covalent VHL binder with enhanced potency and occupancy.
Second-Generation Sulfonyl Fluoride VHL Covalent Ligand.In order to improve covalent ligand potency, we reasoned that optimization of the groups peripheral to the hydroxyproline motif could provide enhanced affinity for VHL.Drawing inspiration from structure−activity studies of previously reported VHL binders, 4 we examined docked poses for analogues of VHL-SF1, including VHL-SF2, in which the tert-leucine moiety was swapped for a methyl isoxazole and the ligation vector moved to the benzylic position (Figure 2A).In contrast to VHL-SF1, this analysis suggests that VHL-SF2 may covalently bind Ser110 while maintaining many of the noncovalent interactions seen with VH032 (Figure 2B).VHL-SF2 was synthesized in 11 steps (Scheme S3).
Initially, we subjected VHL-SF2-Biotin to the streptavidin shift assay and observed a greater extent of VHL modification (44%) than for VHL-SF1-Biotin under the same conditions (Figure S2).Furthermore, VHL-SF2 was able to displace the labeled peptide in the FP assay, with an apparent IC 50 of 35 μM at 2 h, consistent with the predicted covalent occupancy of the VHL HIF1α binding site (Figure 2C).Intact LC−MS analysis confirmed single labeling of VHL by VHL-SF2, with 65% conversion at 24 h (Figure S5).Although we attempted several site ID identification experiments by digesting the labeled recombinant VHL, the Ser110-containing peptide was not detected in the samples treated with VHL-SF2.This observation is consistent with a change in peptide properties that is incompatible with mass spectrometry detection compared to untreated samples (Table S2).However, there was no evidence of modification of the protein at high sequence coverage (>80% on average) at any other site, offering indirect evidence of binding at the predicted site.Encouraged by this evidence for biochemical engagement of VHL, we explored engagement of VHL by VHL-SF2 in live cells through a competition pull-down assay from HEK293T cells (Figure 2E).HEK293T cells were incubated with varying concentrations of VHL-SF2 for 2 h at 37 °C, lysed, and treated with VHL-SF2-Biotin at 50 μM for 2 h, followed by pull-down on streptavidin beads.Elution under strongly denaturing conditions (5% β-mercaptoethanol in Laemmli Buffer for 10 min at 95 °C), SDS-PAGE, and Western blot analysis confirmed the pull-down of VHL by VHL-SF2-Biotin, consistent with expected covalent engagement, which could Target engagement and cellular VHL binding potency were further confirmed through a NanoBRET target engagement assay, assessing inhibition of the interaction between VHL-NanoLuc and a cell-permeable fluorescent VHL tracer ligand in HEK293 cells (Figure 2D). 21In agreement with the FP assay, VHL-SF2 inhibited the BRET signal with an IC 50 of 35 μM, compared to >100 μM for VHL-SF1 and 0.5 μM for VH032, consistent with intracellular HIF1α binding site occupancy.The reduced stability of the sulfonyl fluoride warhead at physiological pH at 40 °C (Table S1) could partially account for the significant reduction in potency observed in cell-based assays.
BRD4 was not depleted in the presence of BRD-SF2 when treated with inhibitors of either proteasome activity or NEDDylation (epoxomicin and MLN4924, respectively), supporting a proteasome-and Cullin E3 ligase-dependent mechanism consistent with recruitment of VHL (Figure 3D,E).
Washout Experiments Reveal an Advantage for Covalent VHL PROTACs.To further investigate the potential advantage of a covalent VHL ligand, we performed a head-tohead comparison of BRD-SF2 and MZ-1 in a washout experiment.In this assay, we incubated HEK293 HiBiT-BRD4   4).(E) HEK293 HiBiT-BRD4 Cln3 cells were treated with MLN4924 (1 μM) for 3 h, followed by the respective PROTAC or VHL-SF2 dose-Cln3 cells with varying concentrations of each PROTAC for 5 h, followed by 24 h recovery after treatment washout.Despite the significant disparity in degradation efficiency between the two compounds, we observed a similar relative decrease in BRD4 degradation (Figure S8).To prevent extended degradation due to residual PROTAC not being washed out, we introduced an excess of the VHL ligand VH032, which binds potently to VHL and blocks degradation by displacing any residual PROTAC which is not covalently bound. 25ndeed, in this experiment, BRD-SF2 showed a reduced relative decrease in degradation activity compared to MZ-1 (Figure 4A,B), quantified at 27% vs 48%, respectively (Figure S9C).The BRD4 HiBiT assay in the presence of competing concentrations of VH032 orthogonally confirmed that BRD4 degradation by MZ-1 was significantly more affected by the presence of the competitor at 6 h (Figure 4E,F).These results further support the covalent mechanism of action of BRD-SF2 and highlight the potential to prolong degradation activity following the removal of free PROTAC or in the presence of competing binders.

■ CONCLUSIONS
To our knowledge, this is the first report of a VHL ligand that lacks the hydroxyproline motif and engages covalently with VHL via a sulfonyl fluoride moiety, and the first covalent E3 ligase binder developed by design rather than screening.When incorporated into bifunctional degraders, the resulting sulfonyl fluoride-based PROTACs are capable of inducing proteasomeand ubiquitin ligase-dependent TPD of both BRD4 and AR.However, the observed degradation efficiency does not currently match optimized noncovalent PROTACs (e.g., MZ-1), and these compounds would require further medicinal chemistry optimization, including to improve stability under physiological conditions.Nonetheless, it is encouraging that functional degradation is observed for these first-generation, unoptimized prototypes, with the potential to expand the substrate scope of E3 ligase-covalent PROTACs to the >20 target proteins previously reported to be degradable with VHLrecruiting PROTACs.Interestingly, the degradation efficiency of these PROTACs was uncoupled from the low occupancy of VHL, and we obtained submicromolar degraders of AR without further optimization of the warhead.Our washout experiments provide insight into the potential pharmacody-

Journal of Medicinal Chemistry
namic advantages of covalent VHL PROTACs, which show relatively sustained degradation in the presence of a VHL binder (VH032) compared to more potent PROTACs (e.g., MZ-1), consistent with the persistence of covalent VHLadducts.This finding points to the advantages of covalent E3 ligase recruiters in terms of catalytic efficiency and prolonged efficacy of a predicted binary complex compared to a standard ternary complex for noncovalent PROTACs.In the future, the effects of this adduct population on the natural substrates of VHL should be further investigated by whole proteomics to determine whether minimal occupancy could be sufficient to promote TPD without altering VHL function.
In summary, our work paves the way for additional structural modifications at the hydroxyproline center, which could ultimately result in improved pharmacokinetic and pharmacodynamic properties in PROTACs targeting VHL.This work benchmarks the covalent recruitment of VHL for TPD and provides the basis for medicinal chemistry campaigns to further optimize the covalent VHL ligand to enhance potency and stability.We also report novel bifunctional degraders of BRD4 and AR which could be used as starting points for the development of PROTACs with improved pharmacokinetic and pharmacodynamic properties.

Figure 1 .
Figure 1.(A) Potential advantage of covalent PROTACs allowing transformation of the ternary complex into a simple binary interaction between modified E3 and substrate.(B) VH032/VHL cocrystal structure, illustrating the critical Ser110 interaction between VH032 (purple) and VHL (PDB: 4W9H).(C) Structure of VH032 and the replacement of the hydroxyproline moiety with a sulfonyl fluoride to generate VHL-SF1.(D) Docking of VHL-SF1 (blue) within a VHL crystal structure (PDB: 4W9H).

Scheme 1 .
Scheme 1. Development of a Novel Synthetic Route for the Introduction of the Sulfonyl Fluoride Moiety in VHL-SF1 and VHL-SF1-Biotin Journal of Medicinal Chemistry

Figure 2 .
Figure 2. (A) Structure of VHL-SF2.(B) Docking of VHL-SF2 (orange) within a crystal structure of VHL (PDB: 4W9H) superimposed onto VH032 (purple).(C) Dose−response of VHL-SF2 inhibition of the VCB and FAM-conjugated HIF1α-derived peptide interaction assessed by FP, following 2 h incubation between VCB and VHL-SF2.The data show the mean ± SEM (n = 3).(D) Cellular potency of VHL-SF2 and VH032 measured by a NanoBRET target engagement assay, following a 5 min incubation between VHL-NanoLuc HEK293 cells, VHL-SF2, NanoBRET VHL tracer ligand, and digitonin.The data show the mean ± SEM (n = 3).(E) Competition pull-down assay from live cells between VHL-SF2-Biotin and VHL-SF2.HEK293T cells were pretreated with DMSO or varying concentrations of VHL-SF2 for 2 h at 37 °C.The cells were lysed and treated with DMSO or VHL-SF2-Biotin (50 μM) for 2 h at room temperature.A pull-down with streptavidin beads was conducted, after which the protein was resolved on SDS/PAGE.VHL and loading control GAPDH levels were visualized by Western blotting.