Chemoproteomics-Enabled De Novo Proteolysis Targeting Chimera Discovery Platform Identifies a Metallothionein Degrader to Probe Its Role in Cancer

Abstract

Proteolysis targeting chimeras (PROTACs) represent powerful tools to modulate the activity of classically “undruggable” proteins, but their application has been limited to known ligands and a few select protein classes. Herein, we present our chemoproteomic strategy for simultaneous de novo discovery of novel degraders and ligands for challenging and previously “undruggable” targets. Using comparative PROTAC versus ligand global proteomics analyses, we rapidly identify proteins selectively downregulated by several “untargeted” PROTACs containing a VHL E3 ligase recruiter and various covalent and non-covalent ligands. We showcase our approach by identifying a first-in-class PROTAC for metallothionein 2A (MT2A) – a small, cysteine-rich, metal-binding protein implicated in heavy metal detoxification, zinc homeostasis, and cellular invasion. Notably, isoform-specific MT overexpression has been shown to augment cellular migration and invasion across several cancer cell lines, though precise mechanisms are unknown due to insufficient tools to study MTs. We show that optimized PROTAC AA-BR-157 covalently binds conserved C44, degrades overexpressed MT2A with nanomolar potency, and reduces migration and invasion of MDA-MB-231 cells. We further demonstrate a time-dependent increase in intracellular zinc levels following MT2A degradation as well as downregulation of protein diaphanous homolog 3 (DIAPH3), a positive regulator of actin and cell motility. Super-resolution imaging of MDA-MB-231 cells shows that downregulation of MT2A and DIAPH3 inhibits cell polarization and thereby migration, suggesting that MT2A may regulate motility via DIAPH3-dependent cytoskeletal remodeling. In summary, our strategy enables the de novo discovery of PROTACs and ligands for novel disease-related targets and lays the groundwork for expansion of the druggable proteome.

Graphical Abstract

INTRODUCTION

Targeted protein degradation (TPD) has emerged as a formidable approach to modulate the function of classically “undruggable” proteins, such as intrinsically disordered proteins.^1–3 This method leverages event-driven pharmacology to achieve protein degradation and complete abolishment of protein activity. TPD offers significant advantages over direct inhibition, including a broader target range and prolonged duration of action.^3–6 A key strategy in TPD utilizes heterobifunctional proteolysis targeting chimeras (PROTACs) to recruit E3 ubiquitin ligases to proteins-of-interest (POIs) for subsequent poly-ubiquitination and proteasomal degradation.⁷ The efficacy of this approach hinges on efficient ternary complex formation between the PROTAC, POI, and E3 ligase.^{8, 9} This not only requires sufficiently high affinity binders for both the POI and E3 ligase but also optimized linker geometry; all of which is difficult to simultaneously optimize in classical “ligand-first” PROTAC synthesis campaigns.^{1, 5, 10} Thus, despite the promise of TPD, the development of novel PROTACs remains a significant challenge and has been met with limited success. Only a handful of protein classes with well-established POI ligands, such as kinases^{11, 12} and BET^{13, 14} proteins, have been successfully degraded.^{9, 10, 15–18}

Mass spectrometry (MS)-based chemoproteomics has become an invaluable tool in chemical biology, wherein chemical probes enable the functional characterization of individual proteins within complex proteomes.¹⁹ Key advances have established efficient methods for proteome-wide mapping and ligand screening in native biological systems,^20–28 aiding in the development of novel inhibitors and E3 ligase recruiters^{29, 30}. However, these methods have limited utility in PROTAC discovery, as mere PROTAC binding to the POI does not guarantee proteasomal degradation. To address this limitation and expand the range of druggable proteins, we propose a strategy for de novo PROTAC discovery that features the synthesis of collections of target-agnostic covalent and non-covalent degraders and systematic, proteome-wide identification of their protein targets (target ID) via global proteomics. This PROTAC-centric approach, while inspired by fragment-based ligand discovery (FBLD),^{20, 28} differs in one key aspect – the ability to identify both a ligand and active degrader simultaneously. We believe that this overall approach is more effective than first identifying targets of ligands and then attempting to convert them into PROTAC probes,³¹ which may not ultimately yield a working degrader.

Global proteomics offers a streamlined protocol for observation of proteome-wide changes in abundance levels, which has been leveraged to identify degraded proteins.^{5, 32, 33} It also enables visualization of diverse classes of proteins – including low molecular weight proteins incompatible with precipitation steps in enrichment-based proteomic workflows. Moreover, degraders also have the potential to further enhance target diversity towards difficult therapeutic targets, resulting from the combined advantages of both traditional PROTACs and covalent inhibition.^{1, 34} Although extensive changes following compound treatment can be difficult to interpret, our two-part comparative approach identifies proteins selectively downregulated by our PROTAC versus the warhead and active linker-E3 ligase ligands alone. This allows us to filter out indirectly dysregulated proteins, thereby enabling extended treatment times and improved identification of slowly degraded proteins. We demonstrate how our approach led to the discovery of several covalent and non-covalent premier hit degraders for “undruggable” proteins with high therapeutic relevance. We then further showcase our method by developing a potent and selective acrylamide-based degrader and chemical tool for metallothionein 2A (MT2A).

MTs are small, cysteine-rich proteins that chelate up to seven divalent cations to maintain numerous biological processes including zinc homeostasis and metal detoxification.³⁵ Their unique metal-binding properties also play critical roles in various pathophysiological processes, including cell proliferation,³⁶ invasion,³⁷ and cell cycle arrest³⁸.³⁹ Due to the lack of noteworthy chemical tools for MTs, research has primarily focused on the biophysical and structural characterization of metalation^{40, 41} and correlative phenotypic studies using genetic knockdown^{37, 38, 42–45} (Figure 1, top). Using our first-in-class MT2A PROTAC, we demonstrate that MT2A degradation may inhibit cancer cell migration and invasion through DIAPH3-dependent cytoskeletal remodeling (Figure 1, bottom).

Figure 1.

Previous studies into metallothionein family members (top) as well as our current work identifying a MT2A targeted degrader (bottom). Green dots represent divalent metal ions. Abbreviations: MT = metallothionein, MS = mass spectrometry, CD = circular dichroism, MD = molecular dynamics, qPCR = quantitative polymerase chain reaction, OE = overexpression, KD = knockdown, DC₅₀ = half-maximal degradation concentration.

In conclusion, the herein presented chemoproteomic method offers a rapid and efficient approach for de novo discovery of PROTACs with minimal experimental optimization. We successfully applied our method to discover several covalent and non-covalent degraders, one of which we further advanced to a valuable chemical tool that enabled mechanistic investigations into native MT2A-regulated cellular processes.

RESULTS & DISCUSSION

Cysteines play diverse structural and functional roles in biology,¹⁹ owing in part to the enhanced nucleophilicity of their side chain thiol. As a result, the cysteine thiol serves as an attractive target in inhibitor discovery^{21, 46–50} and, more recently, in PROTAC development.^{10, 29, 51} We first synthesized a small collection of target-agnostic, cysteine-reactive PROTACs by joining the von Hippel Lindau (VHL) ligand with various acrylamides via a polyethylene glycol (PEG) linker. We herein describe our two-component chemoproteomic method to rapidly identify the cellular targets of our de novo PROTAC probes (Figure 2A). Briefly, cells are treated with the PROTAC, or its constituent linker-E3 ligase recruiter or POI ligands alone and then subjected to MS-based global proteomics analysis. Treatment with the latter two control probes allows us to better filter out protein abundance changes due to the warhead or E3 ligase recruiter binding alone, thus increasing confidence in validating direct cellular targets of our degraders. We then highlight how our method led to the discovery and validation of several first-in-class covalent and non-covalent degraders and ligands for “undruggable” proteins with high therapeutic interest, including metallothionein 2A (MT2A), RAS GTPase activating protein 3 (RASA3), and transmembrane protein 189 (TMEM189).

Figure 2.

Identification of a targeted degrader for metallothionein-2A. A) Schematic workflow of our LC-MS/MS-based PROTAC discovery approach. B) Synthesis of PROTAC 10. Reaction conditions: (i) 1 (1.2 equiv), 2 (1 equiv), EDC (1.1 equiv), HOBt (1.1 equiv), and DIPEA (3.0 equiv) in DCM, rt, 12 h; (ii) TFA (14.5 equiv) in DCM, rt, 2 h; (iii) CuO (1.0 equiv), dppf (0.2 equiv), and Pd₂(dba)₃ (0.1 equiv) in degassed DMF, 60°C, 12 h; (iv) 3 (1.5 equiv), Pd(dppf)Cl₂ (0.1 eq), and Cs₂CO₃ (4.5 eq) in degassed 1,4-dioxane/H₂O, 100°C, 12 h; (v) acryloyl chloride (1.2 equiv) and TEA (3.0 equiv) in DCM, 0°C to rt, 3 h; (vi) TFA (14.5 equiv) in DCM, rt, 2 h; (vii) 8 (1.2 equiv), HATU (1.5 equiv), and DIPEA in DMF, rt, 2 h. C) Volcano plots showing proteome-wide changes in protein levels identified using the general workflow presented in (A) comparing treatment of MDA-MB-231 cells with 10 μM 10 versus 10 μM 9 (top) and 10 μM 11 (bottom) for 24 h (n = 6, 2 biological x 3 technical). Blue regions highlight the proteins selectively downregulated by PROTAC 10. Dotted lines represent the following thresholds: −Log₁₀(P-value) = 2 and Log₂(fold change) = −0.6 and 0.6. D) Chemical structure of POI ligand acryloylpiperazine 11. E) Table of calculated fold change ratios for identified MT2A peptide. Shown as average fold change ± SDs. F) Immunofluorescence microscopy images and G) quantification of MT staining in MDA-MB-231 cells treated with either DMSO, or 10 μM 9, 10, or 11 for 24 h. White bars indicate 30 μm. Shown as fold changes to DMSO-treated cells ± SDs (n = 9). ****p < 0.0001 by unpaired Student’s t-test; ns indicates p > 0.05.

Synthesis of PROTAC 10 and Target Identification.

We initiated our synthetic studies with the preparation of acrylamide warhead 7. Briefly, piperazine 3 and benzothiazole 6 were independently generated via amide coupling and palladium-catalyzed cyclocondensation, respectively (Figure 2B). These fragments were then subjected to Suzuki-Miyaura cross-coupling followed by coupling of the piperazine with acryloyl chloride to obtain electrophilic acrylamide 7. Treatment of 7 with TFA yielded the unprotected amine 8, which was then used to access PROTAC 10 via a HATU-mediated amide coupling with PEG6-appended VHL E3 ligase ligand 9. The cellular targets of PROTAC 10 were then identified using our de novo discovery platform. Briefly, triple negative breast cancer (TNBC) MDA-MB-231 cells were treated with 10 μM 10, 9, or 11 for 24 h followed by lysis, reduction, alkylation, digestion, and liquid chromatography-tandem mass spectrometry (LC-MS/MS)-based global proteomics. Comparative analyses revealed a small subset of proteins that exhibited significant downregulation (fold change ratio < 0.66) upon treatment with PROTAC 10 but not with the linker-appended VHL ligand 9 or warhead 11 (Figure 2C, D and Table S1). A few proteins were also selectively upregulated by PROTAC 10, possibly due to a downstream cellular response to the degraded protein or a compensatory stress response. Owing to our interest in zinc-cysteine interactions,⁵² MTs emerged as intriguing potential targets of 10 for their prominent but understudied roles in zinc homeostasis and regulation of vital cellular processes (e.g., cell proliferation,³⁸ invasion³⁷, and apoptosis^{53, 54}).³⁹ Human MTs are categorized into four sub-families (MT1–4), of which only MT1s and MT2 are ubiquitously expressed across different tissue types.³⁹ Among the detected MT isoforms, only MT2A exhibited fold change ratios below our set thresholds for both 10 versus 9 (0.63) and 10 versus 11 (0.46) (Figure 2E).

We were particularly enthusiastic about MT2A considering reports indicating that this protein is the most abundant and biologically significant MT isoform across numerous cell lines with key roles in the regulation of TNBC migration and invasion, among others.^{37, 44, 55–57} To preliminarily assess whether only PROTAC 10 downregulates endogenous MTs, we treated MDA-MB-231 cells with DMSO, 9, 10, or 11 and visualized MT levels by immunofluorescence microscopy using a non-specific MT antibody due to the limited availability of commercial isoform-specific antibodies. As anticipated, no changes in MT levels were observed upon treatment with DMSO, 9, and 11, whereas an ~2-fold decrease was observed upon treatment with 10 (Figures 2F, G and Figure S1).

PROTAC 10 Binds and Degrades MT2A.

Following identification of MT2A, we then proceeded to validate this protein as a direct target of 10 using a competitive pull-down assay and protein adduct analysis by proteomics. Overexpressed MT2A-FLAG lysate was treated with increasing concentrations of 10 for 2 h at 37°C, followed by treatment with 100 μM probe 12, a desthiobiotin-functionalized derivative of ligand 11 (Figure 3A). The proteins were then precipitated, enriched over streptavidin beads, eluted, and visualized by Western blot. Excitingly, we observed not only enrichment of MT2A by 12 but also competition by PROTAC 10 in a concentration-dependent manner, thus establishing target engagement with low micromolar binding affinity (Figure 3B and Figure S2A). To further investigate the interaction between ligand 11 and MT2A, we pre-treated purified His₆-MT2A with either PBS or excess ZnCl₂ for metal saturation prior to the addition of 11 and LC-MS/MS analysis of the cysteine binding site. The cysteine-11 adduct was observed on conserved C44 within the α-domain for both conditions, suggesting our PROTAC binds MT2A regardless of its metalation status (Figure 3C and Figure S3A). This was further validated via a pull-down assay using desthiobiotin probe 12, wherein no enrichment of C44S-mutated MT2A-FLAG was observed (Figure S3C). Subsequent Schrödinger CovDoc analyses indicated that ligand 11 fits along the metal binding site of the α-domain and forms a hydrogen bond with G40 in the β-domain (Figure 3D). Importantly, our docking results indicate that ligand binding disrupts metal coordination within both domains through interactions with key metal coordinating residues C24, C29, and C33.

Figure 3.

Validating 10 as a targeted degrader of MT2A. A) Chemical structure of desthiobiotin-functionalized probe 12. B) Chemical competitive pull-down of overexpressed MT2A-FLAG from HEK293T lysates pre-treated with various concentrations of PROTAC 10 for 2 h at 37°C followed by treatment with 100 μM desthiobiotin probe 12 for 2 h at 37°C. Proteins were enriched via streptavidin agarose beads, eluted, and visualized by Western blot with FLAG antibody. Eluted proteins are shown on top, and the corresponding input controls are below. C) Annotated MS/MS spectrum of modified C44 on His₆-MT2A with POI ligand acryloylpiperazine 11. * indicates cysteine carbamidomethylation. Grey peaks correspond to POI ligand fragmentation. See Figure S3B for more details. D) Covalent docking of POI ligand acryloylpiperazine 11 onto MT2A (AF-P02795-F1) at C44. Residues critical for ligand binding are highlighted in black. Western blotting analysis of MT2A-V5 expression in MT2A OE cells treated for 24 h with E) DMSO, or 10 μM 9, 10, or 11; or F) DMSO, 10 μM 10, 1 μM (S)-MG132 (18 h), or pre-treatment with 10 μM 10 (6 h) followed by co-treatment with 1 μM (S)-MG132 (18 h). Western blot membranes were probed for V5. G) Chemical structure of inactive epimeric PROTAC 13 and H) subsequent Western blotting analysis of MT2A-V5 expression in MT2A OE cells treated for 24 h with DMSO, or 10 μM inactive PROTAC 13 or active PROTAC 10.

We next sought to determine whether 10 downregulates MT2A through bona fide targeted proteasomal degradation. We first generated stable MT2A-V5 overexpressing MDA-MB-231 (MT2A OE) cells via lentiviral transduction and re-confirmed degrader-induced MT2A downregulation via Western blot (Figure 3E and Figure S2B, S4). We then investigated proteasomal dependence of this downregulation by pre-treating MT2A OE cells with 10 μM 10 for 6 h and then co-treating with 1 μM proteasomal inhibitor (S)-MG132 for 18 h. Indeed, MT2A levels were rescued in co-treated cells, indicating that the 80% decrease observed upon treatment with 10 occurs through proteasomal degradation (Figure 3F). To ensure the observed decrease in MT2A was dependent on VHL E3 ubiquitin ligase recruitment, we synthesized the inactive epimer of 10 featuring inverted stereochemistry of the hydroxyproline moiety of the VHL ligand (R ’ S, inactive PROTAC 13) (Figure 3G). As expected, no effect on MT2A levels was observed upon treatment with 10 μM 13 (Figure 3H). Collectively, our results confirm that 10 is a premier targeted degrader of MT2A with low micromolar in vitro binding affinity and degradation efficacy. Encouraged by these results, we then proceeded to quantify MT2A degradation by 10. To determine the optimal treatment duration, MT2A OE cells were incubated with 10 μM 10 for various amounts of time prior to measuring MT2A levels by Western blot. We observed a half-maximal degradation (D₅₀) at 17.3 h and maximum degradation after 24 h (Figure S5A, B). We then treated MT2A OE cells with various concentrations of 10 for 24 h to determine the potency of degradation. Indeed, we observed a dose-dependent decrease in MT2A expression with a half-maximal degradation concentration (DC₅₀) of 6.5 μM and maximal level of degradation (D_max) of ~70% (Figures 4A and Figure S5C). Some recovery of protein signal was observed at 30 μM due to unproductive ternary complex formation, a phenomenon that has been reported for other covalent PROTACs at high concentrations (i.e., the Hook effect).¹⁰

Figure 4.

De novo discovery platform identifies several degraders for “undruggable” proteins. A) Quantification of MT2A degradation upon treatment with 10, corresponding to the Western blot in Figure S5C. Degradation at 30 μM 10 was not included in the fitting due to the Hook effect. Shown as normalized values ± SDs (n = 3). Western blot membrane probed for V5. B) Chemical structure of covalent degrader 14. C) Quantification of RASA3 degradation upon treatment with 14, corresponding to the Western blot in Figure S6B. Degradation at 10 and 30 μM 14 was not included in the fitting due to the Hook effect. Shown as normalized values ± SDs (n = 3). Western blot membrane probed for RASA2/3. Western blot of RASA3 expression levels upon D) 24 h treatment of either VHL knockdown or wildtype MDA-MB-231 cells with DMSO or 3 μM 14, or E) 2 h treatment of MDA-MB-231 lysate with DMSO or 3 μM 14 37°C that were subjected to CETSA at the indicated temperatures. Western blot membranes probed for RASA2/3 and VHL. F) Chemical structure of noncovalent degrader 15. G) Quantification of TMEM189 degradation upon treatment with 15, corresponding to the Western blot in Figure S8B. Shown as normalized values ± SDs (n = 3). Western blot membrane probed for TMEM189. Western blot of TMEM189 expression levels upon H) 24 h treatment of either VHL knockdown or wildtype MDA-MB-231 cells with DMSO or 1 μM 15, or I) 2 h treatment of MDA-MB-231 lysate with DMSO or 1 μM 15 37°C that were subjected to CETSA at the indicated temperatures. Western blot membranes probed for TMEM189 and VHL.

Expanding our Discovery Method Towards Additional PROTAC Probes and Protein Targets.

Having validated MT2A as a cellular target of hit 10, we next sought to investigate whether our method could robustly discover covalent degraders for other novel targets. We synthesized additional degraders for our collection, one of which we report herein (PROTACs 14; Figure 4B and Figure S6). To this end, PROTAC 14 was synthesized in three steps and subjected to our de novo discovery platform using TNBC MDA-MB-231 cells. Again, our analyses revealed a small subset of proteins selectively downregulated by 14 compared to both linker-VHL ligand 9 and warhead 14a (Figure S6A and Table S2). Among these was RAS GTPase activating protein 3 (RASA3) – a negative regulator of R-RAS and focal adhesion stability that promotes breast cancer metastasis.⁵⁸ To assess whether only PROTAC 14 downregulates endogenous RASA3, we treated MDA-MB-231 cells with DMSO, 9 (10 μM), 14a (10 μM), or various concentrations of 14 and visualized RASA3 expression levels by Western blot (Figure S6B). Gratifyingly, we observed not only a dose-dependent decrease in RASA3 levels upon treatment with 14 (DC₅₀ = 3 μM; D_max ~ 60%) but also no effect with either 9 or 14a (Figure 4C). Like hit 10, some recovery of protein signal was also observed at 30 μM.

We then proceeded to determine whether 14 downregulates RASA3 via PROTAC-mediated proteasomal degradation. We first investigated the VHL E3 ubiquitin ligase dependence of this downregulation by knocking down VHL with siRNA and measuring degradation levels upon 24 h treatment with 3 μM 14 via Western blot. Indeed, RASA3 degradation was completely attenuated in VHL knockdown cells compared to wildtype cells (Figure 4D). To further ensure the observed degradation was directly dependent on 14, we performed subsequent cellular thermal shift binding assays (CETSA). MDA-MB-231 lysates were treated with DMSO or 3 μM 14 for 2 h at 37°C, subjected to increasing temperatures, and centrifuged. Soluble proteins were then visualized via Western blot, which indicated significant thermal stabilization of RASA3 upon PROTAC treatment with a ΔT_m of 6.5°C (Figure 4E and Figure S6C). Collectively, our results not only confirm that 14 is a targeted degrader for RASA3 but also that our method reliably identifies low micromolar hit degraders for “undruggable” POI.

Moreover, encouraged by the success of our covalent degraders, we hypothesized that our method has broad utility in identifying targets of any PROTAC, regardless of its binding mechanism. To assess this, we first synthesized non-covalent PROTAC 15 and again subjected it to our de novo discovery method. All compounds were tested at 1 μM instead of 10 μM due to the catalytic nature of non-covalent PROTACs (Figure 4F, Figure S7A, and Table S3). Transmembrane protein 189 (TMEM189; also known as plasmanylethanolamine desaturase 1, PEDS1) emerged as an interesting target due to its roles in plasmalogen biosynthesis,⁵⁹ ferroptosis,⁶⁰ and autophagy⁶¹. Indeed, several recent reports^{62, 63} also demonstrate its prominent role in progressing growth, proliferation, and migration of breast cancer cells, among other cancers^{61, 64}. We then validated degrader-induced downregulation via Western blot, as previously described for RASA3 (Figure S7B). Excitingly, unoptimized hit 15 exhibited excellent sub-micromolar degradation of TMEM189 with a DC₅₀ of 390 nM and D_max of ~70% (Figure 4G). Subsequent investigations also confirmed VHL-dependent degradation upon VHL knockdown (Figure 4H) and PROTAC binding upon TMEM189 thermal stabilization during CETSA (ΔT_m of 5.4°C with 1 μM 15; Figure 4I and Figure S7C). Altogether, our results indicate that 15 is a first-in-class targeted degrader of TMEM189.

In summary, we discovered several hit targeted degraders for previously “undruggable” POI with diverse structures and binding mechanisms to emphasize the robust and versatile nature of our de novo discovery platform. These degraders represent an excellent starting point for the development of more potent and selective degraders and chemical tools for discovery biology of diverse protein classes (Figure S8).

Optimization towards Lead MT2A Degrader AA-BR-157.

In light of our longstanding interest in zinc biology, we next sought to improve the degradation potency of hit 10 towards MT2A by optimizing the linker length and composition.^{65, 66} We synthesized three derivatives with varied PEG linker lengths (PEG2, 4, and 9; compounds 16–18); the corresponding alkyl derivatives of PROTACs 10, 16, and 17 (C12, 16, and 20; compounds 19-21); and a shorter C8 alkyl derivative (compound 22). MT2A OE cells were then treated with increasing concentrations of 16-22 for 24 h prior to measuring MT2A levels via Western blot. We observed a dose-dependent decrease in MT2A levels with all PROTACs, but interestingly, there was no obvious trend between linker length and degradation (Figure 5A, B, C and Figure S9). There was, however, significantly higher degradation observed with all the PEG-containing PROTACs compared to those with alkyl linkers. Gratifyingly, the PEG4-based PROTAC (17, AA-BR-157) exhibited an ~35-fold higher potency than 10 with a DC₅₀ value of 200 nM and sustained D_max of 90% starting at ~1 μM (D₅₀ = 14.6 h; Figure 5A, B, C and Figure S10). No protein recovery was observed at higher concentrations, suggesting a broad therapeutic window for maximal degradation. Irreversible covalent PROTACs are historically criticized for their non-catalytic mechanism of action, which was thought to reduce potency.^{67, 68} While difficult to compare directly due to differences in protein targets, the potency of AA-BR-157 is on the same order of magnitude as that of many previously reported high-efficiency PROTACs.^{10, 69, 70}

Figure 5.

Optimization of hit 10 towards potent PROTAC AA-BR-157 (17). A) Western blot of concentration-dependent MT2A degradation in MT2A OE cells treated with 17 for 24 h to determine its half maximal degradation concentration (DC₅₀). B) Quantification of MT2A degradation upon treatment 10 or 17, corresponding to the Western blots in Figure S5C and (A), respectively. Degradation at 30 μM 10 was not included in the fitting due to the Hook effect. Shown as normalized values ± SDs (n = 3). Western blot membrane probed for V5. C) Table of DC₅₀ and D_max values for MT2A OE cells treated with PEG- or alkyl- linker containing PROTACs 16-22 for 24 h. Degradation at 30 μM 16 was not included in fitting due to Hook effect. Shown as normalized values ± SDs (n = 3). D) Western blot of concentration-dependent C44S-mutated MT2A-FLAG degradation in Hek293T cells treated with 17 for 24 h.

Moreover, AA-BR-157 also exhibits remarkable selectivity towards MT2A C44 in complex proteomes (Figure S11A and Table S4). First, no degradation was observed upon dose-dependent treatment of C44S-mutated MT2A cells, as expected (Figure 5D). Furthermore, to determine its proteome-wide degradation selectivity, MDA-MB-231 cells were treated with AA-BR-157, the PEG4-linked VHL E3 ligase ligand 17a, or the warhead 11 (all at 10 μM) for 24 h, followed by lysis, reduction, alkylation, digestion, LC-MS/MS analysis, and data processing using MaxQuant. Indeed, among the 4,220 quantified proteins, MT2A was the only protein downregulated by AA-BR-157 but not by either control compound (Figure S11B and Table S5). In summary, our results highlight the potential of AA-BR-157 as a first-in-class degrader and only reported small molecule binder of MT2A to date.

AA-BR-157 Inhibits Cellular Migration and Effects DIAPH3 Expression and Cytoskeletal Remodeling.

With our optimized PROTAC in hand, we then sought to investigate its potential as a research tool to better understand the role that MTs play in driving cancer cell migration. Isoform-specific MT overexpression has been shown to augment cellular migration and invasion across several cancer cell lines, though precise mechanisms are unknown.^{37, 55, 71} To this end, we first performed transwell migration assays using metastatic TNBC MDA-MB-231 and glioblastoma U-87 MG cells. Cells were treated with DMSO, MT siRNA as a genetic control, or increasing concentrations of AA-BR-157 and allowed to migrate across the transwell chamber for 24 h. Excitingly, we observed a concentration-dependent decrease in cellular migration, with a maximal decrease greater than 60% following treatment with 10 μM AA-BR-157 and MT siRNA across cell lines (Figure 6A, Figure S12, and Table S6). We then excluded the possibility that latent PROTAC cytotoxicity was responsible for this decrease, as no cell death was observed upon exposure to AA-BR-157 in a WST-1 assay (Figure 6B). Further, we confirmed that our PROTAC also inhibits MDA-MB-231 cellular invasion through Matrigel using the transwell assay (Figures 6C, 6D).

Figure 6.

AA-BR-157-mediated degradation of MT2A inhibits cellular invasion and increases intracellular zinc levels. A) Quantification of TNBC MDA-MB-231 and glioblastoma U-87 MG cells that migrated during a 24 h transwell assay following treatment with DMSO, the indicated concentration of AA-BR-157, or MT siRNA (KD; treated for 48 h) for 24 h, corresponding to images in Figure S14. Shown as normalized percentages ± SDs (n = 3). B) WST-1 assay measuring the toxicity of AA-BR-157 (24 h) and MT siRNA knockdown (48 h) in MDA-MB-231 (black) and U-87 MG (grey) cells. Cells were incubated with WST-1 (5%) for 1 h. Shown as normalized values ± SDs (n = 3). C) Crystal violet images and D) quantification of MDA-MB-231 cells that invaded during a 24 h transwell assay following treatment with DMSO, 10 μM AA-BR-157, or MT siRNA (KD; treated for 48 h) for 24 h. Shown as normalized percentages ± SDs (n = 3). E) Fluorescence spectroscopy quantification of intracellular zinc levels in MDA-MB-231 cells using 2 μM FZ-3AM. Cells were pre-treated with DMSO (24 h), 10 μM AA-BR-157, or MT siRNA (KD; 48 h) for the indicated time lengths, prior to incubation with FZ-3AM for 30 min and then PBS for 1 h to allow for hydrolysis and activation of FZ-3AM. Fluorescence intensity was normalized using a WST-1 cytotoxicity assay. Shown as normalized fluorescence ± SDs (n = 6). *p < 0.005, ***p < 0.0005, ****p < 0.0001 by unpaired Student’s t-test; ns indicates p > 0.05.

To explore the mechanistic link between MT2A degradation and suppression of MDA-MB-231 cell migration, we first investigated the effect on zinc homeostasis. All cells regulate a small pool of labile zinc (Zn²⁺), which readily interacts with cysteines, as well as histidines, aspartic acids, and glutamic acids, to affect cellular processes, such as cell cycle progression.^{52, 72} Since each MT maximally coordinates seven Zn²⁺ that are excluded from the labile zinc pool, we hypothesized that PROTAC-mediated degradation of MT2A would cause a measurable increase in intracellular Zn²⁺ and broadly influence cellular functions.^{56, 72–75} To probe our hypothesis, we first quantified intracellular Zn²⁺ levels upon time-dependent treatments with AA-BR-157 and MT siRNA using zinc-specific FluoZin-3AM dye and fluorescence spectroscopy. As expected, Zn²⁺ levels were markedly increased following 12 and 18 h treatment with AA-BR-157 but tapered at 24 h, likely due to gradual compensation through other zinc-binding proteins (Figure 6E). Conversely, only a modest increase in Zn²⁺ levels was observed upon MT siRNA knockdown, which we attribute to the longer treatment time required to observe knockdown (48 h) versus degradation.

Intrigued by these results, we employed comparative global proteomics to investigate the proteome-wide effects of elevated Zn²⁺ levels caused by the decrease in MT2A levels. MDA-MB-231 cells were treated with DMSO, 10 μM AA-BR-157, or MT siRNA and subjected to LC-MS/MS analysis to identify commonly up- or downregulated proteins (Figure 7A and Table S7). Among these commonly dysregulated proteins, protein diaphanous homolog 3 (DIAPH3) emerged as an interesting target due to its established pro-cancerous role in cellular migration and cytoskeletal architecture (Figure 7B).^{76, 77} DIAPH3 is an actin nucleation and elongation factor required for assembly of actin cables and stress fibers.⁷⁶ Furthermore, Jang and coworkers disclosed a link between MT1E, matrix metalloproteinase (MMP)-9 expression, and/or cytoskeletal remodeling during glioma invasion.⁷¹ Suspecting a similar relationship, we validated the downregulation of DIAPH3 in MDA-MB-231 cells following treatment with various concentrations of AA-BR-157 and MT siRNA via Western blot. An ~50% decrease in DIAPH3 levels was observed following treatment with both AA-BR-157 and MT siRNA (Figure 7C).

Figure 7.

AA-BR-157-mediated degradation of MT2A may inhibit cellular migration through DIAPH3-dependent cytoskeletal remodeling. A) Volcano plots showing global protein expression changes in MDA-MB-231 cells treated with either 10 μM AA-BR-157 versus DMSO for 24 h (top), or MT siRNA versus Mock for 48 h (bottom). Colored dots represent proteins commonly down- or up-regulated by both AA-BR-157 and MT siRNA treatments (n = 6, 2 biological x 3 technical). Dotted lines represent the following thresholds: −Log₁₀(P-value) = 1.3 and Log₂(fold change) = −0.6 and 0.6. B) Table of downregulation ratios for MT2A and DIAPH3, a protein linked to cellular migration. Shown as average fold change ± SDs. C) Confirmation by Western blot analysis of DIAPH3 expression in MDA-MB-231 cells following treatment with DMSO, the indicated concentrations of AA-BR-157, or MT siRNA (48 h) for 24 h, unless otherwise noted. D) DeepSIM super-resolution microscopy images of the cytoskeleton components showing actin (red), vimentin (green), tubulin (purple), and DAPI (blue) immunofluorescence staining of MDA-MB-231 cells treated with either DMSO or 10 μM AA-BR-157 for 24 h; or MT siRNA or DIAPH3 siRNA for 48 h. White bars indicate 10 μm.

We then leveraged Deep Structured Illumination Microscopy (DeepSIM) super-resolution imaging to investigate the effects of MT and DIAPH3 downregulation on MDA-MB-231 cytoskeletal morphology. Cell motility requires dynamic polarization in the direction of movement, which is easily observed in the DMSO-treated cells (Figure 7D and Figure S13).⁷⁸ Cells treated with 10 μM AA-BR-157 and both MT and DIAPH3 siRNAs, however, exhibited a strongly rounded (unpolarized) shape, indicating a loss of actin cables and migratory potential compared to the DMSO-treated cells. As expected, a similar effect was also observed with the remaining cytoskeletal components – intermediate filaments (vimentin) and microtubules (tubulin). Vimentin reorganized from a strong perinuclear localization with tail-like projections in migrating DMSO-treated cells to a dispersed cytosolic staining in the non-migrating cells,^{79, 80} whereas tubulin was broadly dispersed throughout the cytosol. Importantly, DIAPH3 knockdown control cells recapitulated the morphology of MT-downregulated cells (Figure S14), suggesting that MTs may regulate cellular migration through DIAPH3 downregulation and cytoskeletal remodeling.

Collectively, these results demonstrate that AA-BR-157 is an excellent chemical tool to study MT2A at the cellular level, thus addressing a longstanding challenge associated with the limited availability of research tools to probe native MT biology.

CONCLUSIONS

In summary, we presented our chemoproteomics-enabled PROTAC discovery method, which facilitates the identification of ligands and degraders of disease-implicated proteins currently considered undruggable. Classical enrichment-based target ID methods are highly beneficial for inhibitor discovery, but binding alone is insufficient for identifying functional targeted degraders. As a result, PROTAC development has predominantly relied on the use of previously established high affinity ligands and requires extensive optimizations to degrade the intended target, which is not only time consuming but fails to expand the druggable proteome. We show that our global proteomics method offers an excellent alternative for novel PROTAC discovery. The output of this method illuminates changes in protein levels across the proteome to identify functional degraders with minimal optimization required. Furthermore, our method features a PROTAC versus ligand comparison step that allows for stringent and rapid identification of proteins that are degraded rather than just downregulated following ligand binding. Without this crucial step, global proteomics-based target ID may require extensive pathway analyses using predetermined phenotypes to deconvolute truly degraded hit proteins from unrelated transcriptional downregulation.

We showed that our PROTACs provide a strong foundation for selective degradation of challenging therapeutic targets for which no high-affinity inhibitors are available. Herein, we presented the rapid development of several novel degraders and an in-depth evaluation of our acrylamide- and VHL-based PROTAC AA-BR-157 as a targeted degrader for MT2A, a protein with strong therapeutic potential but no reported small molecule modulators. This is likely due to its lack of clear binding pockets and small size (~6 kDa), enabling it to elude detection in enrichment-based ligand screening. Altogether, identification of an MT2A targeted degrader emphasizes the ability of our method to identify degraders of challenging therapeutic targets, such as small proteins with minimal structural domains for inhibitor binding. Importantly, unlike many chemoproteomic fragment discovery methods,^{28, 81} which are often limited to one particular class of small molecule ligands, our method should, in principle, be compatible with any class of small molecules. We showcase this by identifying targets of two other degraders with vast diversity in structures and binding mechanisms (PROTACs 14 and 15). Among these targets are proteins that, despite having clear therapeutic importance, have no reported small molecule binders, underscoring how our method robustly and broadly contributes to expanding the druggable proteome for diverse protein classes. Moreover, it may also help inform the synthesis of molecular glues over time as more PROTACs are identified using our method by providing a foundation to study how proteins interact with E3 ligases and undergo degradation.

We further demonstrated that our lead PROTAC 10 degrades MT2A through engagement with C44 via the bona fide proteasomal degradation mechanism. With only minimal optimization, we obtained a more potent degrader AA-BR-157 (DC₅₀ = 200 nM; D_max = 90%). Collectively, our discovery of an MT2A-targeting PROTAC offers a pioneering chemical tool to natively study MT-regulated processes, whereas current methods are limited to genetic knockdown^{44, 45, 82, 83} and/or correlative gene expression or purified protein studies.^{53, 84–86}

Aberrant MT overexpression has been linked to the increased metastatic potential of several cancers, including breast cancer and glioblastoma.^{37, 55, 71} Previous reports by Jang and coworkers suggested that MTs may regulate cellular migration through inhibition of MMP-9 and/or through cytoskeletal remodeling, though no cytoskeletal remodeling proteins were identified.⁷¹ Our subsequent biochemical experiments show AA-BR-157 not only decreases cellular migration and invasion of aggressive cancer cell lines, but also downregulates DIAPH3, a positive regulator of actin and cell motility,^{76, 77} and strongly inhibits cell polarization.

Further work is needed to better understand the regulatory relationship between MT2A and DIAPH3, but it is plausible that MT2A regulates a zinc-binding transcription factor responsible for DIAPH3 synthesis.³⁵ MTs have already been shown to interact with a myriad of proteins, such as transcription factors (e.g., NF-κB⁸⁷ and TFIIIA⁷⁵), metalloproteins (e.g., MMP-9³⁷ and Cu/Zn SOD⁸⁸), and endocytic receptors (e.g., low-density lipoprotein receptors⁸⁹), many of which interact during direct zinc exchange.³⁵ Given that MT overexpression is implicated in several pro-cancerous pathways (e.g., migration and invasion^{37, 55, 71}, drug resistance^{56, 83}) across several cancer types, we believe that AA-BR-157 represents an excellent starting point for the development of (pre)clinical MT2A-targeting degraders with broad anti-cancer potential. Our efforts are currently being devoted to furthering our mechanistic understanding of the biochemical consequences of MT2A degradation and to investigating the impact of MT2A degradation on cancer metastasis in vivo.

Supplementary Material

Supporting Information. The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.4c17827.

Experimental procedures, supporting figures, chemical synthesis, and spectra (PDF).

Supporting Proteomic Tables (Excel).