RNA‐PROTACs: Degraders of RNA‐Binding Proteins

Abstract Defects in the functions of RNA binding proteins (RBPs) are at the origin of many diseases; however, targeting RBPs with conventional drugs has proven difficult. PROTACs are a new class of drugs that mediate selective degradation of a target protein through a cell's ubiquitination machinery. PROTACs comprise a moiety that binds the selected protein, conjugated to a ligand of an E3 ligase. Herein, we introduce RNA‐PROTACs as a new concept in the targeting of RBPs. These chimeric structures employ small RNA mimics as targeting groups that dock the RNA‐binding site of the RBP, whereupon a conjugated E3‐recruiting peptide derived from the HIF‐1α protein directs the RBP for proteasomal degradation. We performed a proof‐of‐concept demonstration with the degradation of two RBPs—a stem cell factor LIN28 and a splicing factor RBFOX1—and showed their use in cancer cell lines. The RNA‐PROTAC approach opens the way to rapid, selective targeting of RBPs in a rational and general fashion.


Introduction
RNAb inding proteins (RBPs) constitute al arge fraction of ac ellsp roteome. [1] Over 1500 RBPs are known and their genes are evolutionally conserved and transcribed into splicing variants with unique functions.R BPs bind to RNAs in adynamic,coordinative and sequence-selective manner to form ribonucleoprotein (RNP) complexes that play key roles in RNA-dependent processes. [2] Several diseases are caused by genetic alterations in RBPs,w hich affect their binding to RNA. [3] Here we introduce anew concept in targeting RBPs using an ovel class of chimeric structure,w hich we have termed RNA-PROTACs.C onventional PROTACs (PROteolysis-TArgeting Chimeras) comprise three components:atargetbinding ligand, au biquitin E3 ligase-recruiting ligand and ac onnecting linker group. [4] Thef unction of the E3 binding moiety is to recruit E3-dependent factors,whereby formation of at ernary complex between the PROTACa nd the protein of interest (POI) via the targeting ligand, activates ubiquitination of the latter,m arking it for degradation through the proteasome pathway. [5] This results in lowered levels of the target, and ac orresponding loss of its function. Following ar apid phase of development, the first PROTACsh ave now entered clinical trials. [6] Thetargeting ligand of aconventional PROTAC is adruglike small molecule that binds selectively to the POI. Ligands have been reported for av ariety of proteins,i ncluding enzymes and receptors. [4] They are typically identified by high-throughput screening or rational design, based upon the natural ligands of the target proteins. [7] Intuitively,t he RNAb inding site of an RBP represents aviable target site for drugs,since it is functionally important and structural data for RBPs is often available.However,with the exception of af ew natural products inhibiting splicing, (ref. [8] and refs therein) examples of small-molecule ligands that target RBPs are rare. [9] This may be due to several reasons:RBP binding pockets are intrinsically disordered and change conformation upon RNAb inding; [10] compoundscreening assays are challenging to implement; [11] and RBPs share homologous domains that recognize short, degenerate sequence motifs. [12] We employed short oligonucleotides that are iso-sequential with the RNAc onsensus binding element (RBE) of an RBP for the first time as the targeting moiety for the PROTAC( Figure 1A). Theo ligonucleotide competes with native RNAs for binding to the RBP in cells.Wedescribe here aproof-of-concept for this approach with the design, synthesis and characterization of an RNA-PROTAC targeting LIN28 (Lin28A), asmall RBP comprising aC-terminal CCHC-type zinc knuckle domain (ZKD). [13] To target Lin28, we used 5'-AGGAGAU-3' (L28 RBE ), aconserved sequence present in microRNAs,towhich Lin28 binds ( Figure 1B). We modified L28 RBE with 2'-O-methoxy- ethyl (MOE) ribonucleotides and conjugated it to aE 3recruiting peptide derived from the HIF-1a protein. This RNA-PROTACb inds selectively to Lin28 in vitro,a nd suppressed levels of Lin28A in two cancer cell lines in ubiquitin-dependent fashion. Taken together, these findings pave the way to an ew class of rationally designed inhibitors for RBPs,aprotein family that until now has been difficult to address with conventional drugs.

Results and Discussion
As hort oligoribonucleotide corresponding to the consensus RBE was ac onceptually natural choice for the targeting element of an RNAP ROTAC. Short oligonucleotides have been investigated previously as antisense [14] and anti-miRNAr eagents, [15] but have not been pursued into clinical development to our knowledge.Oligoribonucleotides require structural modifications to protect them against ubiquitous nucleases in vivo.Therefore,weadopted asimilar strategy employed for recently approved oligonucleotide therapeutics:r eplacement of the phosphodiester backbone by diastereoisomeric phosphorothioate (PS) linkages [16] and alkylation of the 2'-hydroxyl group.P S-linkages endow oligonucleotides with three essential properties for use in vivo:e nhanced nuclease resistance,a na bility to enter cells and aw eak non-specific binding to proteins that slows renal excretion and allows systemic circulation. [17] Especially,t he alkylation of the ribose 2'-OH with 2'-O-methoxyethyl (MOE) substituents [18] brings further nuclease stability, [19] and adds ah ydration layer to the structure that improves biodistribution and tolerability in vivo. [20] Theoligonucleotide element of an RNA-PROTACbinds its target RBP via interactions at the riboses and the backbone,a sw ell as the nucleobases.T herefore,i tw as important to show first that the MOE and PS groups would not disturb these interactions.W es howed previously using NMR spectroscopy that Lin28_ZKD recognizes the Watson-Crick faces of G 2 and G 5 in 5'-A 1 G 2 G 3 A 4 G 5 A 6 U 7 -3' (Figure 1B), and that the two zinc finger domains are necessary and sufficient to recognize selectively the GNNG sequence of its partner RNAs. [13b, 21] Thus,w eu sed the same method to examine how various ribose modifications of L28 RBE (Table 1, Figure S1) might affect binding of an RNA-PROTAC to Lin28_ZKD.P leasingly,w eo bserved similar chemical shift perturbations for these analogs as for wild-type RNA ORN1 (Figure 2A). This suggested that the modified oligonucleotides adopted asimilar mode of binding to Lin28_ZKD.Inthe case of ORN2, ORN3 and ORN7 ( Figure 2B,F igure S2) some of the NMR signals broadened or disappeared (precipitation at higher concentrations was also observed). This may have been at least partly caused by the presence of PS diastereoisomers.
In cells,R NA-PROTACs must compete with native RNAs for binding the target RBP.I ng eneral, RNA-RBP binding interactions are complex (sometimes multimeric) and often extend outside the RBE. [12] RNAs econdary structure may also facilitate the interactions.T herefore,i no rder to determine whether short, modified derivatives of L28 RBE could compete with native RNAs for binding to Lin28 in cells,w ea dapted ac ompetition binding assay that was originally developed to screen for small-molecule inhibitors of Lin28 ( Figure S3). [13c,22] PreE-let-7f-1 RNAi sa30-nucleotide (nt) stretch of structured sequence present in let-7 precursors to which Lin28_ZKD binds with nanomolar affinity. [13c, 22] Thus,p reElet-7f-1 was synthesized with a6 -fluoroscein (FAM) label at its 5' terminus (let7 FAM ;F igure S1, Figure 2C). It was then incubated with graded concentrations of human recombinant Lin28_ZKD [13b] in order to determine an equilibrium dissociation constant for the Lin28_ZKD/let7 FAM interaction. This yielded a K D of 27 nM, in agreement with previously recorded values [22] ( Figure 2D). Based on this,40nMconcentrations of Lin28_ZKD were subsequently used for testing L28 RBE analogs.
Incubation of Lin28_ZKD/let7 FAM with unlabeled preElet-7f-1 decreased the fluorescence polarization signal with an IC 50 of 117 nM ( Figure 2E), consistent with displacement of let7 FAM from the protein. In contrast, the 7-mer ORN1 (unmodified L28 RBE RNA) was inactive ( Figure 2F), consistent with observations of Wang et al. [22] Thes tark difference between the activities of the 7-mer and the 30-mer RNAs in this assay may indicate that Lin28_ZKD can make contacts with elements outside of the 7-nt consensus sequence in native RNAs.S urprisingly,t he fully phosphorothioated analog ORN2 yielded an IC 50 of 58 nM in the assay (Figure 2F). ORN4 and ORN3,w hich have PS-OMe and PS-MOE modifications,r espectively,y ielded IC 50 values of 56 and 192 nM, respectively.Arandomized sequence (negative control) of L28 RBE (ORN5)a nd a5 -nt analog of ORN3 (ORN8)s howed about 20-fold and 7-fold weaker activities than ORN4 and ORN3,r espectively.
To provide insight on why the short PS-oligonucleotides displaced let7 FAM so readily from Lin28_ZKD,w em easured binding affinities of selected PO-and PS-L28 RBE analogs to Lin28_ZKD by surface plasmon resonance spectroscopy (SPR). Lin28_ZKD was immobilized to achip and incubated with graded concentrations of selected ORNs,a sw ell as positive and negative controls.

Research Articles
Alet-7 precursor,pre-let-7 gbound Lin28_ZKD with the highest affinity (K d approx. 10 nM; Figure S4). ORN1, ORN2 and ORN3 bound Lin28_ZKD with K d so f2 5.8 mM, 550 nM and 580 nM, respectively.W er e-examined the NMR-structure of Lin28_ZKD/L28RBE in an effort to explain the enhanced binding/activity of the PS-oligonucleotides.I nterestingly,t he phosphodiester group between A 4 /G 5 of ORN1 contacts Lys 150 of Lin28_ZKD ( Figure 2G). Thus,aphosphorothioate group at this position might positively enhance this interaction and partly explain the superior activity of the PSanalogs.Based on the results from these independent assays, three L28 RBE analogs (ORN3, ORN4 and ORN7), together with the sequence-randomized control (ORN5)were selected for further study.
Va rious E3 systems have been employed as PROTACsto mediate degradation of POIs. [5c, 23] Empirical testing is generally used to identify that combination of linker and E3 system that yields the most potent PROTAC reagent. [24] One of the earliest PROTACs ystems was an ALAPYIP-containing peptide derived from the HIF-1a (hypoxia-inducible factor 1a)t ranscription factor.W hen its central proline is hydroxylated, [25] this peptide recruits VHL (von Hippel-Lindau) into the VBC-Cul2 E3 ubiquitin ligase complex. [26] A shortened segment of the peptide (comprising the hydroxylated proline;LA[Hyp]YI) still retains ubiquitinating activity and was selected for this PoCstudy,although recent work has uncovered non peptide-based E3-ligands that might also be used. [27] LA[Hyp]YI is cell-permeable, [28] although it seemed likely that the uptake of RNA-PROTACs into cells would be dominated by the properties of the oligonucleotide.
Single-stranded PS oligonucleotides are internalized after prolonged incubation into many types of cultured cells via gymnosis.U ptake leads to endosomal-lysosomal accumulation, with as low leakage into the cytosol, from where the oligonucleotides distribute to other parts of the cell, including the nucleus. [29] TheP S-groups are crucial for the uptake,a nd the most efficient uptake is typically seen with longer oligonucleotides. [30] We examined the uptake of FAM-conjugated oligonucleotides ORN3 (7-nt) and ORN7 (11-nt) into human immortalized myelogenous leukemia line K562 cells using fluorescence microscopy.Both ORN3 and ORN7 were visible in the cytosol and the nuclei of cells,with the uptake of the longer ORN7 being higher than that of ORN3 (Figure S5). Taken together,t he data suggested that an RNA-PROTACc omprising a7 -mer PS-MOE sequence would reach the cell cytoplasm and therefore be able to elicit its effect.
L28 RBE analogs (ORN3, ORN4, ORN5, ORN7)w ere conjugated in solution to LA[Hyp]YI via the peptide Nterminus ( Figure 3A,F igures S1, S6). TheP S-MOE derivative ORN3 was prioritized for study since PS-MOE oligonucleotide drugs distribute widely after administration in vivo, [17b] are metabolically stable and are safe for patients. [31] Masked maleimide-oligonucleotides were prepared on solid support. They were activated prior to coupling with peptide P 1 (LA[Hyp]YI) or acontrol sequence P con ,inwhich norleucine was exchanged for the hydroxyproline ( Figure 3A). In all, ten oligonucleotide-peptide conjugates were synthesized and purified (Table 2, Figure S1). Several of the conjugates were tested in the Lin28_ZKD competition binding assays to ensure that the addition of the linker/peptide did not adversely affect binding of the RNA-PROTACt oL in28 ( Figure 3B,F igure S7). ORN3P1 and ORN7P1 competed slightly more strongly for binding to Lin28_ZKD than their parent unconjugated oligonucleotides,p resumably due to contacts between LA[Hyp]YI and the protein ( Figure 3B, Figure S7). Consistent with this,t he sequence control ORN5P1 showed aw eak effect on Lin28_ZKD ( Figure S7).
After 48 hofincubation with ORN3P 1 and ORN7P 1 ,the tagged Lin28 (and the control QKI) was immunoprecipitated using an anti-Myc antibody.T he isolated protein was then analyzed by western blotting using the anti-HA and anti-c-Myc antibodies.C onsistent with other previously described PROTACreagents, [32,33] ORN3P 1 -and ORN7P 1 -treated cells showed as mear on the gel towards higher molecular weight ( Figure 3C,D, Figure S8), consistent with conjugation of poly-Ub to the myc-Lin28A fusion protein. Vehicle-treated cells, or cells transfected with ap lasmid that expresses an alternative RBP as negative control (Quaking homolog, QKI; pDEST-Myc-QKI) instead of pCMV-myc-Lin28A, showed no such effects.
Humans express two paralogs of the protein, LIN28 (Lin28A) and LIN28B (Lin28B), that play key roles in development, metabolism, and pluripotency. [34] Lin28 proteins are abundantly expressed during embryonic development, as well as in several tumors and tumor-derived cell lines. [35] They bind with nanomolar affinity to let-7 precursors and the 3' untranslated regions (UTRs) of mRNAs involved in cell proliferation. In order to confirm that RNA-PROTACs bind to endogenous Lin28 in cells,weperformed apull-down RNAE LISA assay that we have previously described. [36] Thus,weconjugated abiotin group to the 3' ends of ORN3P 1 and ORN7P 1 and added these reagents separately on to myelogenous leukemia K562 cells.W eincubated lysates from treated cells on streptavidin-coated plates so as to capture the RNA-PROTACs present in the cell lysates.W ethen assayed for the presence of Lin28A in the plates using an anti-Lin28A antibody and an egative control anti-FUS antibody.L ysates from both ORN3P 1 -and ORN7P 1 -treatments yielded astrong signal for Lin28A protein, in comparison to lysates from mock-treated cells,o rc ompared to measurement using the FUS antibody ( Figure 4B,F igure S9). Thed ata confirmed that RNA-PROTACs entered cells and engaged their intended target.  [a] Ahx = 6-(Fmoc-amino)hexanoic acid, Nle = Norleucine, Hyp = l-hydroxyproline(mutated negative control peptide).

Research Articles
Finally,w et ested RNA-PROTACs for their ability to degrade their target in two cell lines:the embryonic NT2/D1 cell line,which expresses high levels of Lin28A, [13a] and K562 cells which express Lin28A and Lin28B.NT2/D1 cells treated with 2 mMconcentrations of ORN3P 1 or ORN7P 1 showed an approximate 50 %reduction of Lin28A, whereas the negative control peptide ORN3P con was inactive ( Figure S10). Suppression of the target was not quantitative,a si so ften observed from testing PROTACr eagents without an optimization of the targeting ligand, the linker and the E3 system. Furthermore,there was no statistical difference in the activity of the 7-mer and 11-mer reagents.I nK 562 cells, ORN3P 1 suppressed 50 %o fL in28A at the highest concentration, whereas the negative control ORN5P 1 was inactive,confirming as equence-selective action of the RNA-PROTAC ( Figure 4C). Ther eagents had no discernible effect on cell toxicity ( Figure S11). Lin28A and Lin28B have identical zinc finger domains, [13c,21] however ORN3P 1 had little effect on Lin28B ( Figure 4C), despite its presence in the nucleus.T his may have been in part, because Lin28B is mainly localized in the nucleus where proteasomal degradation might be less effective or require additional steps or factors ( Figure 4A, Figure S12). [35b,37] Finally,w ec o-treated K562 cells with ORN3P 1 and MG132, ar eversible inhibitor of the 26S proteasome.T his largely attenuated Lin28A suppression by the RNA-PROTAC, strongly suggesting that ORN3P 1 elicited its effects through the proteasome,a si ntended (Figure 4C). In order to explore the versatility of RNA-PRO-TACs with non peptide-based VHL ligands,w ec onjugated small-molecule VH032 [38] to ORN3 via an amino linker to yield ORN3VH 032 ( Figure S1). In K562 cells, ORN3VH 032 also reduced levels of Lin28A after 24 hi naw ay that could be rescued by addition of MG132 ( Figure S13).
In order to demonstrate that RNA-PROTACs can be generated against other RBPs using their respective consensus RNAbinding elements as atargeting ligand, we turned to RBFOX1. RBFOX1( A2BP1) [39] is an alternative splicing factor expressed in neuronal tissues, [40] muscle and heart. [41] It binds with nanomolar affinity to 5'-UGCAUGU-3' through its RNAr ecognition motif (RRM). We synthesized aP S-MOE-modified variant of the RBFOX1 RBE (FOX RBE )and conjugated it to C[Ahx]LA[Hyp]YI (ORN9P 1 )u sing the same protocol as for the Lin28 RNA-PROTACs. In parallel, we prepared two negative control reagents: ORN10P 1 comprises ar andom oligonucleotide sequence,w hereas ORN9P con has the FOX RBE conjugated to the mutated peptide (Table 2).
Human embryonic kidney cells (HEK293T) cells were treated with the three reagents at increasing concentrations, under similar conditions for the Lin28 RNA-PROTACs. Western blotting of protein isolated from treated cells showed an approximate 50 %reduction of RBFOX1protein at 2 mM ( Figure 4D). ORN9P con and ORN10P 1 did not affect levels of RBFOX1protein, suggesting that ORN9P 1 is selective for its target. In order to demonstrate further the selectivity of ORN9P 1 for RBFOX1, the blot was probed with an antibody against heterogeneous nuclear ribonucleoproteins HNRNPC. HNRNPC contains an amino-terminal sequence that is known to bind strongly uridine (U)-rich sequences,a nd therefore provided as trict test of selectivity for the FOX RBE analogs. [42] Happily,n one of the treatments with ORN9P 1 , ORN10P 1 and ORN9P con affected significantly the level of endogenous HNRNPC protein ( Figure 4D).

Conclusion
Conventional drugs can be classified into distinct structural types,such as small-molecules,therapeutic proteins and oligonucleotides.S ingle-stranded oligonucleotide drugs have cellular RNAa st heir common target, with which they hybridize according to Watson-Crick rules.A mongst them, the most prevalent groups are the antisense-and spliceswitching-oligonucleotides that target mRNAs.H owever, new targeting mechanisms mediated by anti-miRNAa nd decoy oligonucleotides are also under investigation (see refs. [43] and [44] for examples).
Here we describe RNA-PROTACs,anew class of chimeric oligonucleotides that are rationally designed to target RBPs,t hat is,p roteins.M any groups have conjugated oligonucleotides to peptides,usually however,for the peptide to transport the oligonucleotide to its target RNAi nv itro/ vivo. [45] Here,the roles of the peptide and the oligonucleotide are reversed;t he oligonucleotide delivers the peptide to its target site.R NA-PROTACsd ock into the RNAb inding site of an RBP via the structurally modified oligoribonucleotide that is sequence-identical with the native RNA-binding element of the RBP.T he first RNA-PROTACt argets the Lin28 protein, as tem cell factor and oncoprotein of high interest as ap otential drug target for several diseases.U sing as tructure-based approach, we designed a7 -nt PS-MOE oligonucleotide analogue that binds tightly to the zinc finger domain of Lin28A. AV HL-recruiting peptide that is conjugated to the 5'-end of the oligonucleotide then mediates degradation of the target in cells via the ubiquitination pathway.T his proof-of-concept represents an ew means of degrading and thereby inhibiting RNA-binding proteins, at arget class that until now has proven difficult to address pharmacologically.