Recently, we have reported the development of DARPins that target all folded domains of the transcription factor p63 [29] as well as the heterotetramer consisting of a p63 dimer and a p73 dimer [30]. Here we report on the development and characterization of DARPins selective for the p73 SAM and the oligomerization domains (OD), thus completing the arsenal of DARPins for the individual domains of p63 and p73, and we demonstrate that these DARPins can be fused to E3 ligases or E3 ligase adaptor proteins to create highly selective and affine bioPROTAC degraders.

We used a similar selection strategy to find highly selective binders for the OD and the SAM domain of p73 as we had used to develop DARPins for all folded domains of p63 [29] and the hetero-tetramer of p63 and p73 [30] (see Materials and Methods for details). Binders for each domain were selected using ribosome display of the DARPin library, and individual clones were subsequently screened by homogenous time-resolved fluorescence (HTRF) (see Methods). Out of this initial screen, DARPin 1800 was selected as binder for the p73 OD and DARPin B9 as binder for the p73 SAM domain. The previously described non-binding DARPin E3_5 (control DARPin) [31] was used as a negative control for all experiments.

To investigate the selectivity of the DARPins we performed pulldown experiments with full-length, Myc-tagged p63 and p73 isoforms overexpressed in HeLa cells using biotinylated DARPins immobilized on Streptavidin beads. The pulldown demonstrated that both lead DARPins selectively bind to p73 but not to p63 (Fig. 1A, B, Supplementary Fig. S1), while no binding was detectable for the control DARPin. Next, we performed isothermal titration calorimetry (ITC) to determine the binding affinity of both DARPins to their respective target domain. For this purpose, the OD of p73 (aa 351-398 of TAp73α) was titrated with DARPin 1800, yielding a dissociation constant of 94.5 nM (Fig. 1C, Supplementary Table S1). Titration of DARPin 1800 with the ODs of p53 and p63 as well as with the p63/p73 heteroOD [32, 33] did not reveal any binding (Supplementary Fig. S2A). Titration of the SAM domain of p73 (aa 489-550 of TAp73α) with DARPin B9 yielded a K of 58.8 nM (Fig. 1D, Supplementary Table S1), while titration with the SAM domain of p63 did not show binding, confirming specificity for p73(Supplementary Fig. S2B). We also titrated the ODs of all p53 family members, the p63/p73 heteroOD and the SAM domains of p63 and p73 with the control DARPin without detecting any binding (Supplementary Fig. S2C, D).

As p73 is expressed in many different isoforms, we wanted to determine the specificity of the DARPins by performing pulldown experiments with Myc-tagged p73 isoforms expressed in Rabbit reticulocyte lysate. While DARPin 1800 detected all C-terminal isoforms, DARPin B9 is a highly specific binder for the α-isoform, being the only one containing the entire SAM domain (Fig. 1E). The control DARPin showed no binding to any isoform. The same pulldown assay was performed with Myc-tagged p73 isoforms transiently overexpressed in H1299 cells, a human non-small cell lung carcinoma cell line. Unfortunately, the input signals of some isoforms (e.g. TAp73γ) are barely visible on the Western Blot, as these transcriptionally active isoforms are rapidly degraded inside cells. Nonetheless, the initial result obtained with Myc-tagged p73 isoforms expressed in RRL has been confirmed (Supplementary Fig. 1E).

For further structural understanding of the interaction of the DARPins with their target domains, we solved the X-ray crystal structure of both DARPins in complex with their respective target domains. The structure of DARPin 1800 in complex with the p73 OD revealed a 2:1 binding stoichiometry (couting the entire OD as one unit; Fig. 1F). Furthermore, the structure revealed that DARPin 1800 is binding to the hinge region of the OD which connects helix 1 with helix 2 [33, 34]. A detailed analysis of this interaction demonstrated that the interface of DARPin 1800 includes stretches of helix 1 and helix 2 forming multiple hydrophobic contacts as well as some hydrogen bonds (Supplementary Fig. S3 A, B). The high selectivity of DARPin 1800 towards the p73 OD, despite the high sequence identity with the other family members, can be explained by direct contact with non-conserved residues within the p73 OD including L380, P384 and L385.

We also solved a high-resolution crystal structure of DARPin B9 in complex with the p73 SAM domain, which showed a 1:1 stoichiometry, as is also indicated by the ITC measurements (Fig. 1F). The interaction interface is mainly located within helix 1 and helix 2 of the p73 SAM domain. Additional contacts are formed between some N-terminal residues as well as helix 5 of the p73 SAM domain and the DARPin (Supplementary Fig. S3C, D). The selectivity relative to the p63 SAM domain is mediated by hydrophobic contacts to a variety of non-conserved residues throughout the SAM domain, including P491, G499, P503, Y537 and T540.

Previously, we have shown that the functional affinity of the DARPins to their target domain can be boosted by dimerizing the DARPin using a leucine zipper [29] as has also been shown for other DARPins before [35, 36]. We applied the same approach for the DARPins binding to p73 domains. Additionally, we dimerized the DARPins by creating linear fusion constructs using a (GS) linker which was attached to the C-terminus of one DARPin and the N-terminus of the second DARPin. To investigate if the affinity of the DARPins towards p73 was increased, we performed ITC measurements with a p73 construct containing the DBD, the OD and the SAM domain (DBD-OD-SAM, amino acids 112-550 of TAp73α) and the respective DARPins. As a reference, the monomeric DARPins were included in the experiments. For DARPin 1800 we were able to increase the affinity from 67 nM of the monomeric DARPin to 4 nM for the DARPin dimerized via the leucine zipper domain (1800 LZ) and to 13 nM for the linear fusion construct 1800-1800 (Fig. 2A, Supplementary Table S2). In the case of DARPin B9 the affinity was increased from 309 nM (which is within this multi-domain construct significantly higher than the K measured for the isolated SAM domain) to 64 nM for the leucine zipper construct DARPin B9 LZ and to 46 nM for the linear fusion construct B9-B9 (Fig. 2B, Supplementary Table S2). Dimerized versions of the non-binding control DARPin did not show any binding to p73 DBD-OD-SAM (Supplementary Fig. S4).

The results so far demonstrated that the DARPins bind their domains with high selectivity and affinity and can be used as a tool for affinity precipitation experiments. Next, we wanted to develop the DARPins as a tool for the detection of p73 isoforms in cells. The DARPins were modified with an N-terminal HA-tag for this purpose. Stably expressing U-2 OS cell lines that either express Myc-tagged TAp73α, ∆Np73α or ∆Np63α were fixed using PFA-fixation and incubated with 100 nM DARPin solution overnight. Subsequently, the cells were stained for the Myc-tag of the p73 isoforms and the HA-tag of the DARPins. The results in Fig. 3A, B showed no HA-tag signal for the non-dimerized DARPin B9, while weak signals for the dimerized version B9 LZ are detectable for both TAp73α and ∆Np73α. The linear fusion construct B9-B9 showed a weak staining of TAp73α and no detectable staining of ∆Np73α. For DARPin 1800 a weak HA-tag signal is detectable for the monomeric DARPin, and a much stronger signal for the dimerized versions 1800 LZ and 1800-1800. All cells showing an HA-tag signal also stained for the anti-Myc antibody, indicating that the DARPins stained p73 with low background and high specificity. The control DARPin as well as the dimerized versions of the control DARPin did not show any HA-tag signals. To further validate the high specificity of the DARPins we performed staining with U-2 OS cells expressing ∆Np63α but no HA-tag signal was detectable for any of the DARPins again demonstrating the high specificity of the DARPins (Supplementary Fig. S5).

So far, all experiments were performed with purified proteins or in cell culture. To investigate whether our DARPins can also be used to detect p73 in primary tissue we performed pulldown experiments with lysates generated from mouse skin and mouse brain. We used mice as a source of primary tissue because human and mouse p73 have a very high sequence identity [37]. Homogenous extracts from mouse skin were incubated with biotinylated DARPins immobilized on streptavidin magnetic beads. Additionally, we included the previously described DARPin C14 which is known to bind to the DNA binding domain of p63 and p73 [29]. Furthermore, we included the dimeric version of all DARPins that have been generated by the addition of a leucine zipper (LZ) domain to the C-terminus of the DARPins. All DARPins were effective in pulldown experiments showing pulldown signals for p73 from mouse skin (Fig. 4A, Supplementary Fig. S1). However, only a relatively weak pulldown was observable for DARPin B9, while DARPin B9 LZ showed a pulldown efficiency comparable to the pulldown efficiencies of DARPins 1800 and C14. For DARPin 1800 and C14 no difference in pulldown efficiency was observable between the monomeric and the dimerized versions. We performed the same pulldown experiment with a homogenous extract derived from the mouse brain, but only very weak pulldown signals were observable for all DARPins (Fig. 4B, Supplementary Fig. S1). DARPin B9 and DARPin C14 showed nearly no pulldown of p73. Stronger pulldown signals were detectable for DARPin 1800, 1800 LZ, B9 LZ, and C14 LZ.

So far transcription factors like p73 have been classified as "difficult to drug" proteins as they lack binding pockets for small organic compounds that can be used as inhibitors. A recently developed alternative to classical inhibitor-based drug development is the Proteolysis Targeting Chimeras (PROTAC) approach in which a target protein is brought in close contact with an E3 ligase to get degraded. In the bioPROTACs approach, binding to the target protein is achieved by small protein-based binding modules that are directly linked to E3 ligases or E3 ligase adaptors [8,9,10,11]. With the DARPins as highly affine and selective binders, a very attractive class of binding modules for the development of bioPROTACs is available. We have developed DARPin-based bioPROTACs by replacing the substrate binding domain of several E3-ligases or E3-adaptor proteins with the respective DARPins. In particular, we replaced the substrate recognition domain of the E3-ligases MDM2, CHIP, ITCH, VHL, and TRIM, as well as the substrate binding domain of the cullin 3 adaptor SPOP and the bacterial E3-ligase mimic IpaH9.8 with our DARPins (Fig. 5A). In addition to the DARPins 1800 and B9 described in this study, we built the same constructs with DARPin C14 that has been previously identified as a high-affinity inhibitor of p63 and p73 [29]. DARPin C14 binds to the DNA binding interface of the DBD of p63 or p73 thereby blocking DNA binding and inhibiting transcription of p63 or p73 target genes.

Transcription factors such as the p53 family members are located in the nucleus and as a consequence, our degraders have to be located in the nucleus as well. We first examined the cellular localization of these bioPROTACs through transient transfection of H1299 cells followed by IF staining (Supplementary Fig. S6A-F). While certain DARPin-E3 chimeras were found in the nucleus (e.g. DARPin-MDM2 and DARPin-SPOP), others exhibited a more cytosolic localization (e.g. DARPin-Stub and DARPin-VHL). To attain nuclear localization for all bioPROTACs, we included the nuclear localization signal of SV40 (PKKKRKV) in all DARPin-E3 ligase chimeras.

Next, we employed the HiBit Dual Luciferase system (Promega) to assess the efficacy of the different DARPin-E3 fusion constructs for the degradation of the target proteins. For this purpose, we co-transfected H1299 cells with the DARPin-E3 fusion constructs and the pBit4.1-N[HiBit-IRES-luc2/CMV/Blast] vector in which TAp73α was introduced via restriction-dependent cloning using XhoI and XbaI. In this system a small part of the nanoLUC luciferase is N-terminally fused to the protein of interest, in this case TAp73α. After cell lysis the larger part of the nanoLUC luciferase is added for complementation together with substrate. For normalization, firefly luciferase is constitutively expressed from the same plasmid using an IRES. Twenty-four hours after transfection, firefly and nanoLuc luciferase activities were measured, and the relative protein levels were determined by calculating the ratio between both luciferase activities (Fig. 5B). As a reference, we co-transfected the respective DARPins with an N-terminal HA-tag and the SV40-NLS but without the E3 ligase fusion part.

The assay results revealed that none of the three DARPins induced degradation of TAp73α when fused to MDM2 or IpaH9.8. Significant degradation of TAp73α was observed for C14-CHIP, C14-ITCH and C14-VHL. However, the fusion of DARPin 1800 or B9 with these E3 ligases did not induce degradation of TAp73α. Among the tested bioPROTACs, the fusion of any of the DARPins with TRIM21 or SPOP demonstrated the most potent degradation, with the SPOP fusion constructs exhibiting a slightly higher efficiency. Consequently, the C14-SPOP, 1800-SPOP and B9-SPOP fusions were selected for further experiments. Notably, for all E3-ligases, the control DARPin-fusions did not induce any degradation of TAp73α.

We also explored the specificity of our bioPROTACs by co-transfecting the corresponding DARPin-SPOP fusion with the p63 isoforms TAp63α and ∆Np63α. Once again, the HA-tagged DARPins served as a reference (Supplementary Fig. S7A, B). This assay demonstrated the high specificity of the 1800-SPOP and B9-SPOP fusions, as no significant degradation of either p63 isoform was observed. DARPin C14, which is known to bind to p63 and p73, efficiently degraded both p63 isoforms when fused with SPOP. The control DARPin-SPOP fusion, utilized as a negative control, demonstrated no degradation of p63. To explore whether the degradation efficiency relies on the transfected quantity of degrader DNA and to identify the optimal DNA amount, we co-transfected increasing amounts of the respective DARPin-SPOP fusion with pBit4.1-TAp73α. The degradation efficiency was calculated by assessing the p73 protein level through the dual luciferase assay (Supplementary Fig. S7C). This assay revealed that degradation efficiency is indeed dependent on the transfected bioPROTAC amount, reaching maximum efficiency at approximately 5-10 ng of transfected DNA. As observed in prior assays, the control DARPin-SPOP fusion exhibited no effect.

To further explore the degradation of TAp73α mediated by DARPin-E3 constructs and its impact on the transcriptional activity of p73, we conducted a transactivation assay using firefly luciferase under the control of the pBDS-2 promotor, which consists of three copies of the p53 binding site from the 14-3-3σ promotor, as reporter. Notably, DARPins 1800 and B9 enhanced p73 activity when co-transfected with TAp73α, while the control DARPin showed no such effect (Fig. 5C). However, this increased activity did not correspond to an elevation in protein levels, as observed in the corresponding Western Blots (Fig. 5D). As previously documented, DARPin C14 inhibits TAp73α in a concentration-dependent manner [29]. When using the C14-SPOP fusion, this inhibition is more efficient than DARPin C14 itself. This is due to a combined inhibition by binding of the DARPin to the DNA binding interface and degradation. with TAp73α being undetectable on the Western blot even at the lowest applied concentration of C14-SPOP (Fig. 5C, D, Supplementary Fig. S1). The other two DARPin-SPOP fusion constructs inhibited TAp73α as well through degradation, albeit with a lower efficiency as compared to the C14-SPOP fusion (Fig. 5C).

It is well known that p73 isoforms that lack the TA domain such as ∆Np73α are transcriptionally inactive and can act as dominant negative inhibitors of transcriptionally active p53 family members [38]. Moreover, there are reports suggesting an upregulation of these isoforms in certain cancers, leading to the inhibition of p53 and TAp73α [39]. The DARPin-E3 chimeras could represent an innovative tool to counteract the inhibitory effect of ∆Np73α on p53, as depicted in Fig. 6A. To test this hypothesis, we co-transfected Myc-tagged p53 with Myc-tagged ∆Np73α and the DARPin or the corresponding DARPin-SPOP fusion, conducting a luciferase reporter assay to assess p53 activity. Co-transfection of Myc-tagged p53 with Myc-tagged ∆Np73α resulted in a nearly complete depletion of p53 activity (Fig. 6B). However, co-transfection of p53, ∆Np73α and the 1800-SPOP fusion restored p53 activity. The Western Blot shown in Fig. 6C indicated that this reactivation was based on the degradation of ∆Np73α. A similar effect was observed for the B9-SPOP fusion. The inhibitory DARPin C14 alone was sufficient to reactivate p53 because this DARPin prevents binding of ΔNp73α to DNA. However, the assay revealed that the C14-SPOP fusion was more efficient than DARPin C14 alone, as p53 activity was completely restored even at low concentrations. Compared to the 1800-SPOP and the B9-SPOP fusion, the efficacy of C14-SPOP was higher as well. Interestingly, the DARPin 1800 seems to have a stabilizing effect on ΔNp73α as the transfection of this DARPin resulted in a higher cellular concentration of ΔNp73α (Fig. 6C, Supplementary Fig. S1). Presumably this effect is due to stabilizing the tetrameric state of the p73, making its proteasomal degradation by endogenous ligases less efficient.

To demonstrate that the DARPin-E3-induced degradation of TAp73α relies on the proteasomal pathway and does not occur through alternative pathways, we transfected a cell line stably expressing Myc-tagged TAp73α with the corresponding DARPin-E3 fusion. Subsequently, we treated the cells with either DMSO as a negative control or the proteasome inhibitor bortezomib (Fig. 6D, Supplementary Fig. S1). The Western Blot analysis showed that in case of the SPOP-fusions an increased p73 protein level was observed in samples treated with bortezomib demonstrating, that the degradation of p73 is indeed dependent on the proteasome.