sign in sign up
<<
Home , CUL3, DARPin, Fusion protein, MATH domain, SPOP

bioPROTACs as versatile modulators of intracellular therapeutic targets including proliferating nuclear (PCNA)

Shuhui Lima, Regina Khooa, Khong Ming Peha, Jinkai Teob, Shih Chieh Changa, Simon Nga, Greg L. Beilhartzc, Roman A. Melnykc,d, Charles W. Johannese, Christopher J. Browne, David P. Lanee, Brian Henrya, and Anthony W. Partridgea,1

aQuantitative Biosciences, MSD International, Singapore 138665; bPacific Translational Biomarkers, MSD International, Singapore 138665; cMolecular Medicine, Hospital for Sick Children, Toronto, ON, Canada M5G 0A4; dDepartment of Biochemistry, University of Toronto, Toronto, ON, Canada M5S 1A8; and ep53Lab, Agency for Science, Technology and Research (A*STAR), Singapore 138648

Edited by Vishva M. Dixit, Genentech, San Francisco, CA, and approved February 7, 2020 (received for review November 22, 2019) Targeted degradation approaches such as targeting localization, cell-cycle dependent regulation or tissue- or disease- chimeras (PROTACs) offer new ways to address disease through specific expression adds an additional layer of selectivity that can tackling challenging targets and with greater potency, efficacy, be leveraged with targeted degradation strategies. and specificity over traditional approaches. However, identifica- While small molecule targeted degradation approaches offer tion of high-affinity ligands to serve as PROTAC starting points compelling advantages, the discovery of corresponding clinical remains challenging. As a complementary approach, we describe a candidates is not without its challenges. For molecular glues, class of molecules termed biological PROTACs (bioPROTACs)— limited examples exist and PROTACs have thus far been only engineered intracellular consisting of a target-binding applied to targets with available small molecules inhibitors (10). domain directly fused to an E3 ligase. Using GFP-tagged Classically “undruggable” proteins remain challenging although proteins as model substrates, we show that there is considerable opportunities exist to repurpose previously identified small flexibility in both the choice of substrate binders (binding positions, molecule ligands that did not block function for a deg-

scaffold-class) and the E3 ligases. We then identified a highly ef- radation strategy. Also, of more than 600 E3 ligases encoded BIOCHEMISTRY fective bioPROTAC against an target, proliferating cell by the , only 4 are routinely used in PROTAC nuclear antigen (PCNA) to elicit rapid and robust PCNA degrada- design—CRBN, VHL, MDM2, and cIAP (11). The choice of the tion and associated effects on DNA synthesis and cell cycle pro- E3 determines degradation efficiencies and the current selection gression. Overall, bioPROTACs are powerful tools for interrogating lacks the diversity needed to harness the full potential of the degradation approaches, target biology, and potentially for mak- ubiquitin- system (UPS). Even if small molecule li- ing therapeutic impacts. gands to the POI were available, considerable time and effort (without assurance of success) have to be invested for testing bioPROTAC | targeted degradation | PCNA various combinations of linker lengths and E3s recruited. There is currently a poor understanding of the rules that govern stable argeted degradation approaches function by inducing the Tassembly of the ubiquitination complex in close proximity to Significance a protein of interest (POI) to catalyze its selective ubiquitin- tagging and subsequent proteasome-mediated degradation (1). Several such approaches exist including molecular glues, which Intracellular proteins interact with each other to perform — functions that are critical to normal and disease states. At- remodel the surface of an E3 ligase to induce binding to and – degradation of—neo-substrates (e.g., lenolidimide, an approved tempts at altering pathological protein protein interactions with traditional approaches have largely failed. Here, we ex- therapeutic). PROTACs (proteolysis targeting chimeras), the “ ”— other major targeted degradation class, are bispecific molecules plore an emerging approach we call bioPROTACs engi- that induce substrate degradation by simultaneously binding a neered fusion proteins that consist of a target binding domain POI and an E3 ligase (e.g., ARV-110, a degrader of the andro- and an E3 ligase, an arrangement that results in the specific gen and the first-in-class PROTAC to enter clinical degradation of the therapeutic target. Our systematic study trials). Pharmacologically, small molecule degraders offer several shows bioPROTAC design requirements are highly flexible in advantages over traditional inhibitor-based therapeutics. First, terms of both the binding domain and E3 ligase components. degradation can be induced via interaction sites across the POI Resulting molecules can be used as powerful tools for uncov- surface, regardless of whether the binding site is of functional ering biology, informing on the design of small molecule target consequence (2), thus expanding the chemical space for tackling degraders (e.g. PROTACs), and, if delivery issues can be otherwise intractable targets (3). Second, molecules can be addressed, potential therapeutics. recycled for multiple rounds of degradation, a substoichiometric Author contributions: S.L., S.C.C., S.N., G.L.B., R.A.M., C.W.J., C.J.B., D.P.L., B.H., and A.W.P. property which is especially useful for high-abundance targets designed research; S.L., R.K., K.M.P., and J.T. performed research; S.L. contributed new compared to stoichiometric inhibitors may become limited by the reagents/analytic tools; S.L. and A.W.P. analyzed data; S.L. and A.W.P. wrote the paper; high systemic doses required and corresponding polypharmacology- and S.L., R.K., K.M.P., J.T., S.C.C., S.N., G.L.B., R.A.M., C.W.J., C.J.B., D.P.L., B.H., and A.W.P. based toxicities (4). Third, superior pharmacological inhibition contributed to discussion of the results and next steps. can be achieved as degradation attenuates all biological activities The authors declare no competing interest. (enzymatic, transactivation, scaffolding) and inhibition is sus- This article is a PNAS Direct Submission. tained pending protein resynthesis (5). Hence, reduced drug Published under the PNAS license. dosing frequencies can potentially be realized. Finally, enhanced 1To whom correspondence may be addressed. Email: [email protected]. specificity can be attained through differences in substrate This article contains supporting information online at https://www.pnas.org/lookup/suppl/ degradability, E3 suitability. and ternary complex stability (6–9). doi:10.1073/pnas.1920251117/-/DCSupplemental. Also, the ability to engage E3s with differences in subcellular First published March 2, 2020.

www.pnas.org/cgi/doi/10.1073/pnas.1920251117 PNAS | March 17, 2020 | vol. 117 | no. 11 | 5791–5800 Downloaded by guest on September 28, 2021 ternary complex formation between substrate, PROTAC and E3, was reduced in cells expressing vhhGFP4-SPOP, as marked by making informed decisions on PROTAC design difficult. Lastly, the mCherry-positive signal. Interestingly, as vhhGFP4-SPOP ex- as PROTACs are typically composed of two ligands connected by pression increased, the degradation of H2B-GFP was attenuated. a linker, these molecules usually violate Lipinski’s rule of five and, This is consistent with the well documented “hook effect” seen thus, often suffer from permeability and metabolic liabilities (12). with small molecule-based PROTACS. Specifically, beyond a As a complementary approach to small molecule-based de- threshold concentration of a PROTAC molecule, reduced deg- graders, we sought to develop a biologic equivalent to serve both radation occurs, due to the decreased likelihood of forming the as a biological tool and as a potential therapeutic approach. prerequisite substrate:PROTAC:E3 ternary complexes in favor of Specifically, instead of using small molecules to bridge the sub- substrate:PROTAC and PROTAC:E3 binary complexes (28). strate and the E3 ligase, we have reengineered the E3 ligase by Control constructs were engineered where either one or both directly replacing its natural substrate recognition domain with a modular components were mutated. Specifically, vhhGFP4mut or a miniprotein that binds a POI. These fusion proteins, lacks the complementarity determining region 3 (CDR3) and no which we term bioPROTACs (biological PROTACs), were longer recognizes GFP, whereas SPOPmut lacks the three-box expressed in cells to drive targeted degradation of POIs. Al- motif responsible for recruiting CUL3 and, thus, cannot assem- though bioPROTACs are not novel entities, the work described ble the ubiquitination machinery. In all controls, H2B-GFP levels herein represent a systematic exploration of this approach. were maintained (Fig. 1B). The selective depletion of H2B-GFP Historically, effective degraders have been described for the in mCherry-positive cells expressing vhhGFP4-SPOP but not its classically “undruggable” proteins such as β-catenin (13–15), controls was also recapitulated with confocal imaging (Fig. 1C) KRAS (16), and c-Myc (17) using domains engrafted from their and Western blot analysis of cells sorted into mCherry-positive endogenous interacting partners on to E3 ligases beyond those and mCherry-negative populations (Fig. 1D). Finally, to dem- that can be recruited by small-molecule PROTAC (e.g., βTrCP onstrate that H2B-GFP down-regulation was indeed mediated and CHIP). Other proteins also reported to be successfully de- through proteasomal degradation, cells expressing vhhGFP4- graded include cyclin A/CDK2 (18), pRB (19, 20), maltose-binding SPOP were treated with MG132, a proteasome inhibitor. With protein (MBP) (21), β-galactosidase (21), and GFP-tagged proteins increasing concentrations of MG132, the turnover of H2B-GFP (22–24). As a starting point, we further expanded on the published was blocked while the levels of FLAG-tagged vhhGFP4-SPOP or bioPROTAC that degrades GFP-tagged proteins to showcase mCherry (Fig. 1 E and F) remained unaffected. This suggested the remarkable versatility of this system. Building on the insights that H2B-GFP was selectively targeted by vhhGFP4-SPOP for gained, we used proliferating cell nuclear antigen (PCNA) as an degradation in a proteasome-dependent manner. illustrative example to describe and apply a systematic paradigm for the development of potent bioPROTACs against potentially The Extensive Flexibility of bioPROTACs. Having validated the any POI. PROTAC activity of vhhGFP4-SPOP, we next explored how amenable it was to changes in either the GFP binder or the E3 Results ligase. From existing literature, we shortlisted a variety of GFP Validation of vhhGFP4-SPOP for the Degradation of H2B-GFP. To binders that were based on different protein scaffolds, including establish a model system for identifying active bioPROTAC DARPins (29), αReps (30), and (31). They were molecules, we focused on validating published bioPROTACs also further diversified by their GFP binding interfaces (SI Ap- directed at GFP-tagged proteins since successful turnover can be pendix, Fig. S2) and reported binding affinities (Fig. 2). Each was readily measured through multiple and convenient readouts. In fused to MATH domain deleted SPOP and tested for the ability the first report (22), the region containing the F-box domain to degrade H2B-GFP. Surprisingly, despite the drastic differ- from Slmb, a Drosophila melanogaster E3, was fused N-terminally ences in size, structure, binding position, and affinity, all except to a high-affinity anti-GFP nanobody called vhhGFP4 (25, 26). the two weak binders of GFP (the monobodies GL6 and GL8) When expressed in various Drosophila lines bearing GFP-fusion were able to deplete GFP signal once expressed (mCherry pos- protein knock-ins, the authors reported effective depletion of itive) in HEK293 Tet-On 3G cells (Fig. 2). This contrasts with GFP signal intensities with the NSlmb-vhhGFP4 chimera. How- small molecule PROTACs where there was not always a clear ever, we did not achieve knockdown when NSlmb-vhhGFP4 was correlation between binding affinities and degradation expressed in mammalian HEK293 cells with stable integration of efficiencies. It is important to note that small molecule PROTACs 2B (H2B)-GFP (SI Appendix, Fig. S1). Since Slmb are sandwiched between the substrate and the E3, inducing ex- functions as part of the -RING E3 (CRL) tensive new protein–protein and protein–ligand contacts that complex, species-related differences could affect complex for- further stabilizes the ternary complex beyond affinities of the in- mation and ubiquitination efficiencies. dividual ligands (6, 7). This does not happen in bioPROTACs, and To improve on NSlmb-vhhGFP4, Shin et al. swapped NSlmb the extent of degradation should be directly proportional to the with other mammalian E3 adaptors from the CRL family and substrate binding affinity, which helps in simplifying rational de- identified a combination, vhhGFP4-SPOP, that successfully de- sign and lead optimization of bioPROTACs. graded H2B-GFP and other GFP-fusion proteins in the nucleus Through direct fusion of the substrate binder to the E3 ligase, (23). SPOP (speckle type POZ protein) (27) is an E3 adaptor one could potentially recruit any E3-of-interest with a bioPROTAC protein that functions in complex with cullin-3 (CUL3). The 374- approach. CRLs are the largest and best-studied family of E3 residue protein is comprised of two modular domains, the substrate- ubiquitin ligases. They function as multisubunit complexes that binding MATH domain and the CUL3-binding BTB domain sep- include a cullin scaffold, a RING-H2 finger protein, a receptor arated by a flexible loop. To change the substrate specificity of responsible for substrate recognition, and with the exception of SPOP and enable the targeting of GFP-tagged proteins, its MATH CUL3-based CRLs, an adaptor subunit that links the substrate domain was replaced by vhhGFP4 (Fig. 1A). We adopted the receptor to the complex (32). We picked representative E3 re- same doxycycline-inducible bidirectional system used in the orig- ceptors from each of the five major categories of (CUL1 inal publication (23) to drive coexpression of the bioPROTAC to CUL5). Selection was made based on the availability of struc- vhhGFP4-SPOP and an mCherry reporter of transfection/ tural information to guide truncations. For each E3 receptor, we expression. After transient transfection and doxycycline induction replaced the substrate recognition domain with vhhGFP4 and in HEK293 Tet-On 3G cells with stable integration of H2B-GFP, retained the portion that binds the remaining E3 complex plus the GFP and mCherry fluorescence were measured by flow cytom- flexible loop that naturally links the two modular domains (Fig. etry (Fig. 1B). Consistent with the previous report, GFP intensity 3A). We also fused vhhGFP4 to the U-box E3 CHIP as it has been

5792 | www.pnas.org/cgi/doi/10.1073/pnas.1920251117 Lim et al. Downloaded by guest on September 28, 2021 BIOCHEMISTRY

Fig. 1. vhhGFP4-SPOP induces the ubiquitin-mediated proteasomal degradation of H2B-GFP. (A) Design of the chimeric protein vhhGFP4-SPOP167–374 for the degradation of GFP-tagged proteins. The substrate-binding MATH domain of the E3 adaptor SPOP (amino acids 1–166) was replaced by vhhGFP4 (25, 26), a single-domain fragment that binds GFP. This will enable the ubiquitin tagging of GFP fusion proteins such as H2B-GFP by the CUL3-based CRL complex. (PDB) structures are shown for SPOP (3HQI) and GFP:GFP-nanobody complex (3OGO). (B) Flow cytometric analysis of H2B-GFP/ HEK293 Tet-On 3G cells transiently transfected with various bidirectional, Tet-responsive plasmids. Doxycycline (100 ng/mL) was added to induce the si-

multaneous expression of mCherry and vhhGFP4-SPOP167–374 (or its controls). vhhGFP4mut lacks the complementarity determining region 3 (CDR3) and cannot bind GFP. SPOPmut lacks the three-box motif and cannot bind CUL3. GFP and mCherry fluorescence intensities were measured 24 h after doxycycline induction. (C) Confocal imaging analysis of the same set of cells as in B. Plasma membrane (pseudocolored white) was labeled using the CellMask Deep Red plasma membrane stain. Yellow arrow denotes an example of a transfected cell (mCherry positive) that have lost H2B-GFP. (D) Western blot analysis of H2B-GFP/ HEK293 Tet-On 3G cells treated as in B and sorted according to the levels of mCherry using FACS. Gating was set such that mCherry (−) cells have the same

signal intensities as untreated cells in the mCherry channel, and anything above this basal level was assigned mCherry (+). Expression of vhhGFP4-SPOP167–374 (or its controls) was detected using an anti-FLAG-tag antibody (Left Lower, red bands, n = 3 for vhhGFP4-SPOP167–374). The substrate H2B-GFP was detected using an anti-GFP antibody (Left Lower row, green bands), and band intensities were quantified and normalized to the levels of the loading control Hsp90

(Right). (E) Western blot analysis of H2B-GFP/HEK293 Tet-On 3G cells transiently transfected with the mCherry/vhhGFP4-SPOP167–374 bidirectional inducible plasmid and treated with the indicated concentrations of doxycycline and MG132 (a proteasome inhibitor) for 16 h. FACS sorting was conducted as in D. Band

intensities of H2B-GFP and FLAG-tagged vhhGFP4-SPOP167–374 were quantified and plotted in Right.SeeSI Appendix, Fig. S5 for uncropped blots and expected molecular weight of each protein. (F) Flow cytometric analysis of the same set of cells as in E.

Lim et al. PNAS | March 17, 2020 | vol. 117 | no. 11 | 5793 Downloaded by guest on September 28, 2021 Fig. 2. Flexibility in the type of binder used for generating bioPROTACs. Flow cytometric analysis of H2B-GFP/HEK293 Tet-On 3G cells transiently transfected with various bidirectional, Tet-responsive plasmids. Doxycycline (100 ng/mL) was added for 24 h to induce the simultaneous expression of mCherry andthe

different GFP binders fused to SPOP167–374. A total of seven GFP binders were tested: one nanobody (vhhGFP4; refs. 25 and 26), one DARPin (3G86.32; ref. 29), two αReps (bGFP-A and bGFP-C; ref. 30), and three monobodies (GS2, GL6, and GL8; ref. 31). The values (in green) on the scatter plots indicate the percentage of GFP-negative cells in the mCherry-positive transfected population, which corresponds to successful H2B-GFP depletion by the respective SPOP-based anti- GFP bioPROTAC. The table summarizes the molecular mass of the GFP binders and their reported binding affinities to GFP. Representative PDB structures are shown for each scaffold (3OGO, 2QYJ, 4XVP, 1TTG), alpha helixes are colored blue, and beta strands are colored red.

proven to be effective in bioPROTAC approaches (16, 17, 21). missing the flexible sequence that facilitates unfolding and trans- Strikingly, 8 of 10 of our bioPROTAC combinations were again fer into the catalytic core of the proteasome. When H2B was at- successful in degrading H2B-GFP (Fig. 3B), with 5 of them tached to GFP, GFP signal intensity started to decline upon the yielding more than 70% clearance. Both CUL4-based CRBN- expression of various anti-GFP bioPROTACs (SI Appendix,Fig. vhhGFP4 and DDB2-vhhGFP4 failed, suggesting that CUL4 S3 and Fig. 4, Middle), suggesting that the properties which trig- CRLs may be less active in HEK293 cells, or that the protein gered successful proteasomal degradation were imparted by H2B. truncations were not designed optimally, resulting in loss of Therefore, by tagging any POI to GFP, we are now able to recruit ubiquitin ligase activity. Nevertheless, our results highlighted the a representative pool of E3 ligases to evaluate if the POI possesses versatility of bioPROTACs and the ease at which novel active the necessary traits that enable its targeted proteolysis—ubiquitin- molecules can be discovered. acceptor surface lysine(s) located within or proximal to structur- ally disordered degradation initiation sites (33). bioPROTACs Inform on Substrate Degradability and E3 Selection. The To extend our work to a potential therapeutic target, we sought panel of bioPROTACs that degrade GFP-tagged proteins form to determine whether PCNA is a good substrate for ubiquitin- the basis of our platform to interrogate the degradability of novel mediated proteasomal degradation. PCNA is an essential protein substrates as they span representative members in the largest E3 expressed in the nuclei of all proliferating cells where it forms a ubiquitin ligase family. For a protein to be degraded through the homotrimeric ring structure encircling DNA (35). PCNA serves as UPS, a tripartite model has been proposed (33): 1) a primary a sliding DNA clamp to recruit a myriad DNA replication and degron (short, linear peptide motif) that specifies substrate damage repair proteins to the chromatin. Expression of PCNA is recognition by cognate E3 ubiquitin ligases, 2) a secondary site elevated in rapidly dividing tumor cells and, in most cases, is as- comprising of surface lysine(s) that favor (poly)ubiquitin conju- sociated with poor prognosis, making it an attractive target for gation, and 3) a structurally disordered segment within or proxi- (36). Considering its scaffolding function and the mal to the secondary site such that ubiquitin chain recognition is multiple protein–protein and protein–DNA interactions that simultaneously coupled to substrate unfolding at the 26S protea- PCNA is involved with, PROTAC strategies are ideally poised to some. GFP by itself is a poor substrate for ubiquitin-mediated abolish all its activities concurrently and achieve therapeutic ef- proteasomal degradation (SI Appendix,Fig.S3and Fig. 4, Top) ficacy. Prior to our study, it was not clear whether PCNA levels (22). Upon the expression of vhhGFP4-SPOP, the localization of could be modulated through proteolysis, although numerous re- GFP switched from uniform cellular distribution to nuclear ports have described the extensive regulatory monoubiquitination speckles where SPOP complexes function (34) (SI Appendix,Fig. and polyubiquitination of PCNA in response to genotoxic stress S3). However, overall fluorescence intensity was unaffected, sug- (37, 38). These nonproteolytic signaling events help to coordinate gesting that GFP was not efficiently turned over although it was the dynamic engagement of PCNA with its vast array of inter- bound by vhhGFP4-SPOP. Even if GFP was concentrated to the acting partners. Since surface lysines are available for ubiquitin nucleus through the addition of a nuclear localization signal conjugation, the appropriate E3s could be recruited to extend the (NLS), the speckles persisted (SI Appendix, Fig. S3). It is unlikely right linkages that target PCNA to the proteasome. To test this that GFP lacked available lysines for polyubiquitination, as 19 hypothesis, we tagged PCNA with GFP and applied our panel of lysine residues are evenly distributed across its surface (22). In- anti-GFP bioPROTACs. Five of the 10 tested—vhhGFP4 fused to stead, due to its compact and well-folded nature, GFP might be FBW7 (CUL1), VHL (CUL2), SPOP (CUL3), SOCS2 (CUL5)

5794 | www.pnas.org/cgi/doi/10.1073/pnas.1920251117 Lim et al. Downloaded by guest on September 28, 2021 BIOCHEMISTRY

Fig. 3. Flexibility in the type of E3 ubiquitin ligase used for generating bioPROTACs. (A) Table of 10 different substrate recognition subunit (SRS) used in bioPROTAC designs to explore alternative E3 ligases. Information regarding each SRS are listed as follows: the E3 ubiquitin ligase complex they function in, PDB structural information, National Center for Biotechnology Information protein accession number, region of the protein fused to vhhGFP4, molecular mass of the region fused to vhhGFP4 and previous records of their use in PROTAC strategies. Truncations were designed to replace the original substrate- binding domain with the GFP-binding nanobody vhhGFP4. (B) Flow cytometric analysis of H2B-GFP/HEK293 Tet-On 3G cells transiently transfected with various bidirectional, Tet-responsive plasmids. Doxycycline (100 ng/mL) was added for 24 h to induce the simultaneous expression of mCherry and the dif- ferent truncated SRSs fused to vhhGFP4. A total of 10 truncated SRSs was tested, grouped according to the cullin E3 scaffold they recruit. The values (in green) on the scatter plots indicate the percentage of GFP-negative cells in the mCherry-positive transfected population, which corresponds to successful H2B-GFP depletion by the respective vhhGFP4-based anti-GFP bioPROTAC. PDB structures are shown for each SRS, alpha helixes are colored blue, and beta strandsare colored red. The dotted area represents the portion fused to vhhGFP4.

and CHIP (U-box)—were able to deplete PCNA-GFP (Fig. 4, whether the POI can be targeted for proteasomal degrada- Bottom). Since the GFP tag on its own was not efficiently de- tion and also to identify suitable E3 ligases to recruit for its graded, this result indicated that PCNA possesses the necessary polyubiquitination. traits that enabled its targeted degradation. Among the active E3 ligases identified, SPOP was of specific interest as it shares Con1-SPOP Induced Robust Degradation of PCNA. To verify if the the same subcellular localization as PCNA—the nucleus. In- findings from the anti-GFP bioPROTAC platform can be deed, the silencing of PCNA-GFP by vhhGFP4-SPOP but not translated to the degradation of endogenous PCNA, we selected its controls was corroborated with imaging (SI Appendix,Fig. a published PCNA-binding peptide termed Con1 to be incor- S4). More importantly, SPOP-based bioPROTACs displayed porated into our anti-PCNA bioPROTAC design (Fig. 5A). The immense flexibility as swapping the nanobody vhhGFP4 with 16-residue Con1 peptide binds PCNA with a reported Kd of either the DARPin 3G86.32, the GS2, or the αReps 100 nM and contains the conserved PIP (PCNA-interacting pro- bGFP-A/C all generated effective bioPROTACs that down- tein) box motif common to PCNA binding partners (39). Con1 was regulated PCNA-GFP (SI Appendix,Fig.S4). Hence, by tag- fusedtoSPOP167–374 (Fig. 5A), the E3 ligase identified through our ging POIs such as PCNA to GFP and applying the anti-GFP anti-GFP bioPROTAC screen. Similar to vhhGFP4-SPOP (with bioPROTAC platform that we have built, we are able to probe specificity for GFP), Con1-SPOP (with specificity for PCNA) also

Lim et al. PNAS | March 17, 2020 | vol. 117 | no. 11 | 5795 Downloaded by guest on September 28, 2021 the three-box motif responsible for recruiting CUL3, Con1- SPOPmut also failed to alter the levels of PCNA-GFP, sug- gesting that proper assembly of the ubiquitination complex was needed to drive the down-regulation of PCNA-GFP (Fig. 5B). We next investigated the ability of Con1-SPOP to degrade endogenous PCNA. In HEK293 cells, as the expression of FLAG-tagged Con1-SPOP increases (inferred by increases in mCherry signal), the protein levels of PCNA dropped corre- spondingly (Fig. 5C). In the control where Con1 is able to bind PCNA but does not degrade it, PCNA levels were maintained as expected (Fig. 5C). Using a doxycycline-inducible line with stable integration of Con1-SPOP, PCNA down-regulation was observed as early as 4 h following the addition of doxycycline and, by 24 h, PCNA protein was barely detectable (Fig. 5D). This effect was again lost when either SPOP or Con1 were mutated (Fig. 5D). The rapid silencing of PCNA demonstrated in this study using bioPROTACs contrasts with previous studies using siRNAs, where more than 72 h was needed to achieve knock-down (40– 43), an observation that is in line with PCNA’s reported protein half-life of 78.5 h (44). Indeed, a key distinguishing feature of targeted degradation approaches is that the protein target is di- rectly depleted, whereas RNA interference (RNAi) approaches depend on the natural turnover of the existing pool of proteins while preventing de novo protein synthesis. In this study, we present a bioPROTAC approach to delete PCNA at the protein level and achieve superior degradation kinetics over RNAi. Thus, we envision that bioPROTACs can become a valuable research tool for studying the function of long-lived proteins that are known to be refractory to RNAi. Since PCNA is indispensable for DNA replication (35), the degradation of PCNA in cells expressing high levels of Con1- SPOP (mCherryhigh) resulted in complete S-phase withdrawal (Fig. 5 E, Top row), as indicated by the lack of cells that stain positive for EdU (a nucleoside analog of thymidine that gets incorporated into newly synthesized DNA). Based on previous reports, the Con1 peptide alone was active as a stoichiometric inhibitor of PCNA since it was able to disrupt the binding of PCNA effector proteins including p21 (39, 45, 46). Indeed, Con1-SPOPmut (that binds PCNA but was unable to induce its degradation) prevented DNA synthesis but was only effective at high concentrations (Fig. 5 E, Top). In cells with lower expres- mid sion of Con1-SPOPmut (mCherry ), effects on the cell cycle were lost although the PCNA degrader Con1-SPOP continued to show robust inhibition (Fig. 5 E, Middle). This result highlighted the substoichiometric efficacy of PROTAC strategies and the ability to achieve functional effects at reduced doses compared to conventional “occupancy-driven” inhibitors (11). The failure to undergo DNA replication also translated into robust growth inhibition in cells where the expression of Con1-SPOP was reg- ulated by doxycycline (Fig. 5F). Doxycycline concentrations (1– 100 ng/mL) were effective in maintaining complete growth arrest Fig. 4. bioPROTACs inform on substrate degradability and E3 selection. of HEK293 cells over 10 d, while the lower concentrations re- Flow cytometric analysis of HEK293 Tet-On 3G cells with stable integration of duced proliferation rates compared to the nonbinding control GFP, H2B-GFP, or PCNA-GFP. Each of the three stable cell lines was tran- Con1mut-SPOP. At all doxycycline concentrations tested, com- siently transfected with the same panel of bidirectional, Tet-responsive plete growth arrest could not be achieved with the stoichiometric plasmids. Doxycycline (100 ng/mL) was added for 24 h to induce the simul- inhibitor Con1-SPOPmut although growth was impaired at the taneous expression of mCherry and the different anti-GFP bioPROTAC. Cells higher concentrations. This result highlights that for certain in Q1 represent successful H2B-GFP depletion by the respective anti-GFP bioPROTAC. (Bottom) Confocal imaging analysis of PCNA-GFP/HEK293 Tet- targets (which may be high in abundance, prone to compensatory On 3G cells transiently transfected with various bidirectional, Tet-responsive feedback mechanisms, or have multiple functions that cannot be plasmids. Doxycycline (100 ng/mL) was added for 24 h to induce the simul- inhibited through a single binding site), a degradation approach taneous expression of mCherry and the different anti-GFP bioPROTACs. may be needed to achieve the desired functional outcome. Discussion degraded PCNA-GFP and, furthermore, appeared to do so more The advent of small molecule-based targeted degradation ap- effectively (53.4% versus 85.8% GFP-negative cells at 24 h after proaches has ignited a paradigm shift in drug discovery and cre- expression, Fig. 5B). By replacing the three conserved residues ated unique opportunities to tackle historically intractable targets. critical for binding to PCNA in Con1 by alanine, Con1mut-SPOP To capitalize on this general approach, we are using an orthogonal was no longer able to bind and degrade PCNA-GFP. By deleting system termed bioPROTACs, where a polypeptide-based binder

5796 | www.pnas.org/cgi/doi/10.1073/pnas.1920251117 Lim et al. Downloaded by guest on September 28, 2021 to the POI is fused directly to an E3 ubiquitin ligase. Unlike their small molecule counterparts, bioPROTAC discovery is not limited by the ligandability of the POI and the E3. In the present study, we showed that bioPROTACs are highly modular in nature and can readily accommodate changes to either the binder or the E3 ligase. This remarkable flexibility enabled us to tap into the rich diversity offered by the UPS. Indeed, 8 of 10 different mammalian E3 ligases tested in this study gave significant degradation ac- tivities. We also demonstrated a capacity to engage substrates (GFP-fusion proteins and PCNA) to determine their degradability through a bioPROTAC approach. Although we were able to generate first-generation active bioPROTACs through rationale design, we anticipate that improved versions (catalytic efficiency, protein half-life, optimal subcellular localization) can be obtained through further /maturation and the optimi- zation of binding affinity/specificity, linker length, stability, and E3 selection. While we are optimistic of the potential of bioPROTACs as a therapeutic modality, they can also serve to enable the development of small molecule-based degraders. In particular, bioPROTACs are poised to address key questions that should be understood prior to initiating small molecule-based programs, including 1) is the POI a good substrate for polyubiquitination and proteasomal degradation (i.e., exposed lysines, structurally disordered segment that initiates unfolding at the 26S proteasome; ref. 33), 2) which E3 ligases are the most effective at inducing its degradation (correct cellular ex- pression of the entire ubiquitination machinery, prolific activity under the relevant diseased state), and 3) what are the functional consequences associated with its degradation. Of significant ad- BIOCHEMISTRY vantage is the ability of the bioPROTAC approach to rapidly pro- vide these insights to evaluate the feasibility of a small molecule- based campaign before embarking on resource-intensive medicinal chemistry. For target validation purposes, bioPROTACs offer a way to deplete a target at the protein level—one that is complementary but offers distinct benefits to the more established RNAi and CRISPR approaches (Fig. 6). Key advantages of bioPROTAC- mediated silencing include 1) insights can be gained quickly as

interacting residues and cannot bind PCNA. The values (in green) on the scatter plots indicate the percentage of GFP-negative cells in the mCherry- positive transfected population, which corresponds to successful PCNA-GFP depletion by the respective SPOP-based bioPROTAC. (C) Western blot anal- ysis of HEK293 Tet-On 3G cells transiently transfected and induced with doxycycline as in B and sorted according to the levels of mCherry using FACS. Gating was set such that mCherrylow cells have the same signal intensities as untreated cells in the mCherry channel, whereas mCherrymid and mCherryhigh cells have increasing levels of mCherry fluorescence. Expression of Con1-

SPOP167–374 (or its control) was detected using an anti-FLAG-tag antibody (Left Lower, red bands) and the expected molecular mass of each chimeric protein is indicated in kilodaltons. The substrate PCNA was detected using an anti-PCNA antibody (Left Lower, green bands). Band intensities of FLAG-

tagged vhhGFP4-SPOP167–374/SPOPmut and endogenous PCNA were quanti- fied and normalized to the levels of the loading control Hsp90 (Right). (D) Western blot analysis of T-REx-293 cells with stable integration of Con1-

SPOP167–374 (or its controls) under the control of a Tet-responsive promoter. Various concentrations of doxycycline were added to the culture media for Fig. 5. Rapid and robust degradation of PCNA with Con1-SPOP. (A) Design the indicated length of time and lysates were collected. Proteins were de-

of the chimeric protein Con1-SPOP167–374 for the degradation of PCNA. The tected as in C. PCNA levels were quantified and expressed as fold of untreated substrate-binding MATH domain of the E3 adaptor SPOP (amino acids 1–166) cells. (E) EdU labeling and flow cytometric analysis of HEK293 Tet-On 3G cells was replaced by Con1, a high-affinity peptide ligand of PCNA. This will en- transiently transfected and induced with doxycycline as in B. Cells undergoing able the ubiquitin tagging of PCNA by the CUL3-based CRL complex. PDB DNA synthesis were labeled with 10 μM EdU for 2 h and the percentage of structures are shown for SPOP (3HQI) and PCNA (1AXC). (B) Flow cytometric EdU-positive S-phase cells (in purple) was expressed according to the level of analysis of PCNA-GFP/HEK293 Tet-On 3G cells transiently transfected with mCherry signal intensities. Gating for mCherry expression was performed as in various bidirectional, Tet-responsive plasmids. Doxycycline (100 ng/mL) was C.(F) Incucyte confluency measurements of T-REx-293 cells with stable in-

added for 24 h to induce the simultaneous expression of mCherry and the tegration of Con1-SPOP167–374 (or its controls) under the control of a Tet- different chimeric proteins. vhhGFP4mut lacks the complementarity determining responsive promoter. Various concentrations of doxycycline were added to region 3 (CDR3) and cannot bind GFP. SPOPmut lacks the three-box motif and the culture media and the percentage confluency of the cells was tracked cannot bind CUL3. Con1mut bears point in the three critical PCNA- continuously over 10 d.

Lim et al. PNAS | March 17, 2020 | vol. 117 | no. 11 | 5797 Downloaded by guest on September 28, 2021 itself was a poor substrate for polyubiquitin-mediated degradation with these constructs (Fig. 4), PCNA is likely amenable for targeted deg- radation strategies. We were thus motivated to engineer bioPROTACs against endogenous PCNA. This was accomplished by leveraging Con1, a high-affinity peptide ligand of PCNA. Indeed, Con1-SPOP proved to give rapid and robust degradation of endogenous PCNA. More importantly, we achieved the expected functional effects in terms of cell cycle arrest and inhibition of cellular proliferation with the active bioPROTAC but not the nonbinder control, indicating that on-target biological effects could be seen with the potential to advance this molecule toward the clinic. Finally, by comparing the active bioPROTAC (Con1-SPOP) with the stoichiometric inhibitor (Con1-SPOPmut), superior pharmacology was demonstrated with the inhibit-and-degrade approach. To realize the potential of bioPROTACs as a therapeutic modality, the fundamental challenge of delivery will need to be addressed. Although significant challenges lie ahead, delivery of mRNAs encoding for a bioPROTAC seems particularly attrac- tive as human clinical proof of concept has been achieved with several RNA candidates presently in late stage clinical trials (47). The optimization of the RNA delivery vehicle as well as the Fig. 6. Summary of the different modalities that can be used for the in- modification of RNA to increase stability and reduce immune hibition of a POI. The availability of ligands, the mode of inhibition (stoi- chiometric versus degradation), and the ease of delivery are parameters that activation were key drivers of these latest developments. We can influence the tractability of each approach. By working at the protein anticipate that this will be a viable route for the intracellular level, bioPROTACs reduces risks associated with genetic manipulation using delivery of bioPROTACs encoded by mRNA. Key challenges CRISPR and off-target effects commonly seen with siRNA. Compared to small ahead include achieving sufficient protein expression and tissue molecules and , ligands used in bioPROTAC approaches are easier to targeting while avoiding toxicities related to delivery vehicles and discover. However, delivery challenges need to be addressed and options in vitro transcribed mRNA (48, 49). include the delivery of bioPROTAC mRNA (refer to Discussion). During the final stages of manuscript preparation, an or- thogonal study by DeLisa and coworkers was published (50). Although that work focused on leveraging IpaH9.8, a bacterial knockdown is achieved over a shorter time period and is not de- E3 mimic, for achieving targeted degradation, some interesting pendent on the natural turnover of the existing pool of proteins parallels were noted. First, using IpaH9.8 as the E3 of choice, a (RNAi) nor does it require genetic manipulation (CRISPR), 2) variety of GFP and GFP-derived fusion proteins could be de- specificity is defined by the substrate binder which can be optimized graded. In addition, they also noted some flexibility in terms of through yeast- and phage-display technologies, 3) specific post- the substrate binder that could be used. Together with the work translational modifications (e.g., phosphorylation) or protein state presented here, the combined results reinforce the modular na- (e.g., aggregated) can be selectively targeted, 4) effect of knock- ture of the bioPROTAC approach and one that can be used as a — down is reversible once the bioPROTAC is gone, new proteins biological tool and potential therapeutic. can be made to replenish the deleted pool, 5) knockdown can be controlled temporally and/or spatially in animal models through Concluding Remarks: Advancing the Field with the . For example, knock-in mice where the bioPROTAC Platform bioPROTAC expression is under the control of a doxycycline- In this paper we describe a platform that can be applied sys- inducible promoter combined with a tissue-specific Tet regulator tematically to identify “bioPROTACs”—engineered fusion pro- ’ protein will allow one to delete the bioPROTAC s substrate in a teins that degrade a POI. Our survey of a panel of E3 ligases and specific tissue at a specific time in the mouse development. The binders advance the state-of-the-art in several ways. First, substrate can also be reaccumulated by withdrawing doxycycline through our large panel of binder-E3 fusions, we demonstrated and discontinuing the expression of the bioPROTAC. With the that the discovery of active bioPROTACs, while not routine, is continual advancements in yeast- and phage-display technologies, it relatively facile, especially compared to the equivalent task for is not difficult to identify a peptide or a miniprotein binder against a small molecules. Indeed, well over 50% of constructs tested POI. We showed that as long as the binding affinity is high, it against GFP-H2B and GPF-PCNA showed activity. This high can be combined with the wealth of E3 ligases to generate active success rate was relevant not only to the choice of binder (i.e., bioPROTACs with robust silencing activity. This ease of develop- scaffold type, size, binding position and affinity) but also in terms ment should encourage its broad utility as a general proteome editing of the choice of truncated E3 used. This inherent flexibility and tool. Moreover, unlike small molecule PROTAC, bioPROTACs can ease-of-discovery should encourage researchers to explore the be integrated into the genome, making the precise modulation of utility of the bioPROTAC paradigm, both as biological tools with protein levels for the study of protein function possible. unique advantages as well as the exciting possibility of applying We have also put forth a systematic workflow to guide the this drug modality to so called “intractable” therapeutic targets. selection of binder-E3 ligase pairs. As an illustrative example, we Implicit in our results are also guidelines as to where the E3s generated a highly effective bioPROTAC against an endogenous should be truncated to arrive at highly efficient bioPROTACs. protein PCNA, which can be used as a research tool and/or Second, the ability to inform on the degradability of a target potentially as a cancer therapeutic. The workflow started with protein provides highly valuable information for small molecule the testing of our library of vhhGFP4-based anti-GFP bioPROTACs PROTAC approaches. Indeed, systems such as HaloPROTACs as degraders of PCNA-GFP. The advantage of this approach is that (51) and dTAG (52) have been designed to that end. While these early insights with respect to a target’s degradability can be gained systems require the POI to be tagged with either a HaloTag or since a specific binder to the POI is not required at this stage. In- FKBP12F36V, respectively, the bioPROTAC strategy has the deed, we observed significant degradation of the PCNA-GFP fusion added advantage that no modification of the target protein is protein with 5 of 10 vhhGFP4-E3 combinations (Fig. 4). Since GFP required, making the study of biological effects associated with

5798 | www.pnas.org/cgi/doi/10.1073/pnas.1920251117 Lim et al. Downloaded by guest on September 28, 2021 endogenous protein degradation in multiple cell lines more straight- Generation of stable cell lines is described in SI Appendix. All cells were forward and physiologically relevant. Finally, the bioPROTAC maintained at 37 °C, 5% CO2, and 90% relative humidity. platform coupled to endogenous or model substrates (e.g., H2B-GFP) offers a facile way to test the potential of a spectrum Flow Cytometric Analysis, Cell Sorting, and EdU Labeling. Twenty-four hours after dox induction, harvested cells were analyzed on a BD LSRFortessa X- of E3 ligases that could be leveraged for targeted degradation. 20. To sort cells according to mCherry or GFP expression, cells were har- Indeed, the platform provides a means to address questions like vested in complete media after transfection and dox induction. Four-way 1) does the subcellular location of the employed E3 ligase need sorting on mCherry- or GFP-expressing cells was facilitated by a BD FACSAria to match that of the target protein, and 2) which of the E3 li- Fusion instrument and sorted populations were processed for Western blot gases beyond those employed to date for small molecule ap- analysis or for further expansion in culture. For EdU labeling, cells were pulse proaches can be successfully leveraged for targeted degradation labeled for 2 h with 10 μM EdU and stained using the Click-iT EdU Flow strategies? For the former question, this could be addressed by Cytometry Assay Kit (Thermo Fisher Scientific). FxCycle Far Red Stain (Thermo testing a panel of bioPROTACs that uses E3 ligases with dif- Fisher Scientific) was included to determine DNA content. Samples were an- alyzed on BD FACSAria Fusion. ferent subcellular localizations. For the latter question, the current work serves as a starting point. Indeed, some of the E3s Imaging. One day after cell seeding, transfection and dox induction were used in this study (FBW7, SKP2, SOCS2, and ASB1) had not performed. Images of live cells were acquired on the Opera Phenix High previously been employed for targeted degradation. While Content Confocal Screening System. these serve as important illustrative examples, the excitement lies in the potential of the bioPROTAC platform to identify Western Blot Analysis. Twenty to 50 micrograms of protein extract from cell disease-relevant and/or tissue-specific E3 ligases that can be lysate was separated by SDS/PAGE and transferred onto nitrocellulose mem- leveraged for therapeutic application. branes using the Trans-Blot Turbo semidry system (Bio-Rad). Blocked mem- branes were probed with the appropriate primary followed by Materials and Methods incubation with the secondary antibodies IRDye 680RD donkey anti-mouse IgG and IRDye 800CW donkey anti-rabbit IgG (Li-Cor) for 1 h at room tem- Refer to SI Appendix for full methods. perature. Fluorescent signals were imaged and quantified using Odyssey CLx.

Cell Culture. HEK 293 Tet-On 3G cells were cultured in Minimum Essential Data Availability. Please refer to SI Appendix for uncropped Western blots Medium (MEM) GlutaMAX (Gibco) supplemented with 10% Tet system- and DNA sequences for all constructs used in this paper. approved fetal bovine serum (FBS) (Clontech) and 100 μg/mL geneticin. T-REx-293 cells were cultured in MEM GlutaMAX supplemented with 10% ACKNOWLEDGMENTS. We thank Tomi K. Sawyer, Chandra Verma, Sriniva- μ BIOCHEMISTRY Tet system-approved FBS (Clontech) and 5 g/mL blasticidin. Cells were saraghavan Kannan, Tsz Ying Yuen, Cynthia R. Coffill, Farid J. Ghadessy, and seeded and transfected the following day. Doxycycline was used to induce all members of the Quantitative Biosciences, MSD International team for expression from pTRE3G-BI-mCherry plasmids at 24 h after transfection. helpful discussions and comments on the manuscript.

1. M. Toure, C. M. Crews, Small-molecule PROTACS: New approaches to protein deg- 19. P. Zhou, R. Bogacki, L. McReynolds, P. M. Howley, Harnessing the ubiquitination radation. Angew. Chem. Int. Ed. Engl. 55, 1966–1973 (2016). machinery to target the degradation of specific cellular proteins. Mol. Cell 6, 751–756 2. L. N. Gechijian et al., Functional TRIM24 degrader via conjugation of ineffectual (2000). bromodomain and VHL ligands. Nat. Chem. Biol. 14, 405–412 (2018). 20. J. Zhang, N. Zheng, P. Zhou, Exploring the functional complexity of cellular proteins 3. X. Huang, V. M. Dixit, Drugging the undruggables: Exploring the ubiquitin system for by protein knockout. Proc. Natl. Acad. Sci. U.S.A. 100, 14127–14132 (2003). drug development. Cell Res. 26, 484–498 (2016). 21. A. D. Portnoff, E. A. Stephens, J. D. Varner, M. P. DeLisa, Ubiquibodies, synthetic E3 4. I. Churcher, Protac-induced protein degradation in drug discovery: Breaking the rules ubiquitin ligases endowed with unnatural substrate specificity for targeted protein or just making new ones? J. Med. Chem. 61, 444–452 (2018). silencing. J. Biol. Chem. 289, 7844–7855 (2014). 5. G. M. Burslem et al., The Advantages of targeted protein degradation over inhibition: 22. E. Caussinus, O. Kanca, M. Affolter, Fluorescent fusion protein knockout mediated by An RTK case study. Cell Chem. Biol. 25,67–77.e3 (2018). anti-GFP nanobody. Nat. Struct. Mol. Biol. 19, 117–121 (2011). 6. D. P. Bondeson et al., Lessons in PROTAC design from selective degradation with a 23. Y. J. Shin et al., Nanobody-targeted E3-ubiquitin ligase complex degrades nuclear promiscuous warhead. Cell Chem. Biol. 25,78–87.e5 (2018). proteins. Sci. Rep. 5, 14269 (2015). 7. M. S. Gadd et al., Structural basis of PROTAC cooperative recognition for selective 24. L. J. Fulcher et al., An affinity-directed protein missile system for targeted proteolysis. protein degradation. Nat. Chem. Biol. 13, 514–521 (2017). Open Biol. 6, 160255 (2016). 8. H. T. Huang et al., A chemoproteomic approach to query the degradable kinome 25. D. Saerens et al., Identification of a universal VHH framework to graft non-canonical – using a multi-kinase degrader. Cell Chem. Biol. 25,88–99.e6 (2018). antigen-binding loops of camel single-domain antibodies. J. Mol. Biol. 352, 597 607 9. B. E. Smith et al., Differential PROTAC substrate specificity dictated by orientation of (2005). recruited E3 ligase. Nat. Commun. 10, 131 (2019). 26. U. Rothbauer et al., Targeting and tracing in live cells with fluorescent – 10. I. Collins, H. Wang, J. J. Caldwell, R. Chopra, Chemical approaches to targeted protein nanobodies. Nat. Methods 3, 887 889 (2006). 27. Y. Nagai et al., Identification of a novel nuclear speckle-type protein, SPOP. FEBS Lett. degradation through modulation of the ubiquitin-proteasome pathway. Biochem. J. 418,23–26 (1997). 474, 1127–1147 (2017). 28. M. J. Long, J. R. Poganik, Y. Aye, On-demand targeting: Investigating biology with 11. A. C. Lai, C. M. Crews, Induced protein degradation: An emerging drug discovery proximity-directed chemistry. J. Am. Chem. Soc. 138, 3610–3622 (2016). paradigm. Nat. Rev. Drug Discov. 16, 101–114 (2017). 29. M. Brauchle et al., Protein interference applications in cellular and developmental 12. S. D. Edmondson, B. Yang, C. Fallan, Proteolysis targeting chimeras (PROTACs) in biology using DARPins that recognize GFP and mCherry. Biol. Open 3, 1252–1261 ‘beyond rule-of-five’ chemical space: Recent progress and future challenges. Bioorg. (2014). Med. Chem. Lett. 29, 1555–1564 (2019). 30. A. Chevrel et al., Specific GFP-binding artificial proteins (αRep): A new tool for in vitro 13. F. Cong, J. Zhang, W. Pao, P. Zhou, H. Varmus, A protein knockdown strategy to study to live cell applications. Biosci. Rep. 35, e00223 (2015). the function of beta-catenin in tumorigenesis. BMC Mol. Biol. 4 , 10 (2003). 31. A. Koide, J. Wojcik, R. N. Gilbreth, R. J. Hoey, S. Koide, Teaching an old scaffold new 14. Y. Su, S. Ishikawa, M. Kojima, B. Liu, Eradication of pathogenic beta-catenin by Skp1/ tricks: Monobodies constructed using alternative surfaces of the FN3 scaffold. J. Mol. Cullin/F box ubiquitination machinery. Proc. Natl. Acad. Sci. U.S.A. 100, 12729–12734 Biol. 415, 393–405 (2012). (2003). 32. D. R. Bosu, E. T. Kipreos, Cullin-RING ubiquitin ligases: Global regulation and activa- 15. J. Liu, J. Stevens, N. Matsunami, R. L. White, Targeted degradation of beta-catenin by tion cycles. Cell Div. 3, 7 (2008). – chimeric F-box fusion proteins. Biochem. Biophys. Res. Commun. 313, 1023 1029 33. M. Guharoy, P. Bhowmick, M. Sallam, P. Tompa, Tripartite degrons confer diversity (2004). and specificity on regulated protein degradation in the ubiquitin-proteasome system. 16. Y. Ma et al., Targeted degradation of KRAS by an engineered ubiquitin ligase sup- Nat. Commun. 7, 10239 (2016). – presses pancreatic cancer in vitro and in vivo. Mol. Cancer Ther. 12, 286 34. M. R. Marzahn et al., Higher-order oligomerization promotes localization of SPOP to 294 (2013). liquid nuclear speckles. EMBO J. 35, 1254–1275 (2016). 17. S. Hatakeyama, M. Watanabe, Y. Fujii, K. I. Nakayama, Targeted destruction of c-Myc 35. G. L. Moldovan, B. Pfander, S. Jentsch, PCNA, the maestro of the replication fork. Cell by an engineered ubiquitin ligase suppresses cell transformation and tumor forma- 129, 665–679 (2007). tion. Cancer Res. 65, 7874–7879 (2005). 36. S. C. Wang, PCNA: A silent housekeeper or a potential therapeutic target? Trends 18. W. Chen, J. Lee, S. Y. Cho, H. A. Fine, Proteasome-mediated destruction of the cyclin a/ Pharmacol. Sci. 35, 178–186 (2014). cyclin-dependent kinase 2 complex suppresses tumor cell growth in vitro and in vivo. 37. J. Chen, W. Bozza, Z. Zhuang, Ubiquitination of PCNA and its essential role in eu- Cancer Res. 64, 3949–3957 (2004). karyotic translesion synthesis. Cell Biochem. Biophys. 60,47–60 (2011).

Lim et al. PNAS | March 17, 2020 | vol. 117 | no. 11 | 5799 Downloaded by guest on September 28, 2021 38. N. Mailand, I. Gibbs-Seymour, S. Bekker-Jensen, Regulation of PCNA-protein inter- 45. G. Kontopidis et al., Structural and biochemical studies of human proliferating cell actions for genome stability. Nat. Rev. Mol. Cell Biol. 14, 269–282 (2013). nuclear antigen complexes provide a rationale for cyclin association and inhibitor 39. D. I. Zheleva et al., A quantitative study of the in vitro binding of the C-terminal design. Proc. Natl. Acad. Sci. U.S.A. 102, 1871–1876 (2005). domain of p21 to PCNA: Affinity, stoichiometry, and thermodynamics. Biochemistry 46. E. Warbrick, A functional analysis of PCNA-binding peptides derived from protein se- – 39, 7388–7397 (2000). quence, interaction screening and rational design. 25,2850 2859 (2006). 40. T. Senga et al., PCNA is a cofactor for Cdt1 degradation by CUL4/DDB1-mediated N- 47. J. C. Kaczmarek, P. S. Kowalski, D. G. Anderson, Advances in the delivery of RNA therapeutics: From concept to clinical reality. Genome Med. 9, 60 (2017). terminal ubiquitination. J. Biol. Chem. 281, 6246–6252 (2006). 48. P. S. Kowalski, A. Rudra, L. Miao, D. G. Anderson, Delivering the messenger: Ad- 41. Y. Yu et al., Proliferating cell nuclear antigen is protected from degradation by vances in technologies for therapeutic mRNA delivery. Mol. Ther. 27,710–728 forming a complex with MutT Homolog2. J. Biol. Chem. 284, 19310–19320 (2009). (2019). 42. B. Xu et al., Proliferating cell nuclear antigen (PCNA) regulates primordial follicle 49. D. Weissman, mRNA transcript therapy. Expert Rev. Vaccines 14, 265–281 (2015). assembly by promoting of oocytes in fetal and neonatal mouse ovaries. 50. M. B. Ludwicki et al., Broad-spectrum proteome editing with an engineered bacterial PLoS One 6, e16046 (2011). ubiquitin ligase mimic. ACS Cent. Sci. 5, 852 –866 (2019). 43. A. Niimi et al., Regulation of proliferating cell nuclear antigen ubiquitination in 51. D. L. Buckley et al., HaloPROTACS: Use of small molecule PROTACs to induce degra- mammalian cells. Proc. Natl. Acad. Sci. U.S.A. 105, 16125–16130 (2008). dation of HaloTag fusion proteins. ACS Chem. Biol. 10, 1831–1837 (2015). 44. B. Schwanhäusser et al., Global quantification of mammalian expression control. 52. B. Nabet et al., The dTAG system for immediate and target-specific protein degra- Nature 473, 337–342 (2011). dation. Nat. Chem. Biol. 14, 431–441 (2018).

5800 | www.pnas.org/cgi/doi/10.1073/pnas.1920251117 Lim et al. Downloaded by guest on September 28, 2021