Computational Modeling of TP63–TP53 Interaction and Rational Design of Inhibitors: Implications for Therapeutics

TP63 protein plays an important role in epidermal morphogenesis and limb development (1). There are multiple TP63 isoforms with widely different transactivation potentials that can have pro- or antitumorigenic function (2). These isoforms are named depending on their ability to transactivate TP53-responsive element (RE); TA isoforms have N-terminal transactivation (TA) domain (TAD), whereas ΔN isoforms (ΔNp63) are terminally truncated and lack the TAD (Fig. 1A). Alternative splicing at the C-terminus generates three additional isoforms α, β, and γ. Therefore, six isoforms are available for TP63; three with the TAD (TAp63α, TAp63β, and TAp63γ) and three with the truncated N-terminus (ΔNp63α, ΔNp63β, and ΔNp63γ; refs. 2–4). Because each structural domain has its own role, the function and activity of these TP63 isoforms differ from each other (5).

TAD can bind to DNA through TP53-RE causing cell-cycle arrest and activating genes to induce apoptosis. DNA-binding domain (DBD) and oligomerization domain (OD) are common to all TP63 isoforms and they facilitate DNA binding and oligomerization, respectively. C-terminal transactivation inhibitory domain (TID) is specific to α isoforms and plays a role in controlling TP63 activity by inhibiting TAD (Fig. 1B; ref. 6). TAp63 isoforms are expressed in epithelial cells and oocytes at elevated concentrations. In non-stressed cells, TAp63 is found in an inactive dimeric conformation to regulate its apoptotic activity (6–8). TAp63 activation leads to tetramer formation and a 20-fold increase in DNA-binding affinity of TAp63α (8). It has been shown that expression of ΔNp63 is sufficient to promote squamous differentiation in human pancreatic ductal adenocarcinoma (PDAC) cells as well as enhanced cell motility and invasion. This process resulted in addiction of squamous PDAC cells to ΔNp63 expression (4, 9, 10). Recent transcriptome studies showed that more than 25% of PDAC are squamous like and they are associated with poor clinical outcomes (4, 9). ΔN isoforms are overexpressed in various cancer types, including lung, head and neck, cervical, skin, and pancreas (11, 12). Therefore, TP63 proteins can play a pro- or antitumorigenic role depending on the TP63 isoform activity.

TP53 is an important tumor suppressor, which is mutated in over 50% of all cancers (13, 14). Mutant TP53 proteins can lose their tumor-suppressor activity and contribute to tumor progression. Mutations in TP53 frequently result in single amino acid variations in the DBD (15). Although wild-type (WT) TP53 is unable to interact with TP63, some TP53 conformational mutants (CM; R175H, Y220C, and I254R, hereafter referred to as TP53^CM) can interact with some TP63 isoforms (16, 17). TP53^CM are common CM alleles resulting in the misfolding of the TP53 protein. TP53^CM can bind strongly to TAp63α (17). This interaction can inhibit transcriptional activity of TAp63α (18). Moreover, TP53^CM enhances pro-tumorigenic activity of ΔNp63. It has been observed that TP53^CM and ΔNp73 (another TP53-family protein) can promote ΔNp63 expression (19).

Despite the functional importance of TP63 isoforms and their interaction with TP53^CM, there is no structural model of the TP63–TP53^CM-binding interface to guide our mechanistic understanding of this interaction. Previous studies have provided some valuable insights on the TP63–TP53^CM interaction (Fig. 1C). First, both TP63β and TP63γ isoforms lacking the TID are significantly impaired for TP53^CM-binding compared with TP63α (17, 20). TP53^CM preferably binds tetrameric TAp63α in which TID is not occupied (8, 17, 20). These two findings indicate the importance of TP63 TID in TP53^CM interactions. Second, there is evidence that the removal of the DBD of TAp63α led to increased binding to TP53^CM. This interaction indicates that TP63 DBD is not required for the TP53^CM binding. Third, deletion of amino acids 251–312 in TP53^CM DBD specifically reduced the interaction of TP53^CM with TP63 (21). These findings strongly suggest that the 251–312 region of TP53 is crucial for TP63 interaction. Overall, the previous studies on TP63–TP53^CM provide strong evidence in this interaction (21); however, there is no available structure or computational models.

Over the last 40 years, TP53 has been extensively studied in cancer with very little clinical success in preventing mutant TP53 oncogenic activity, warranting the need for new therapeutic strategies. Previous attempts have focused on designing drugs or peptides to inhibit TAp73–TP53^CM interaction that restores TAp73 apoptotic function (22–24) and reduces invasion. However, no therapeutics that directly and specifically interfere with the TAp63–TP53^CM interaction have been reported. As previously discussed, releasing TAp63 from TAp63–TP53^CM complex can both induce apoptosis and prevent invasion (17, 25, 26). Moreover, increasing free and active TAp63 isoforms in TP53^CM cells relative to ΔNp63 could allow the TAp63 isoforms to compete against pro-tumorigenic ΔNp63 activity (3). Therefore, molecular modeling of the TAp63–TP53^CM interaction would provide structural insights into the factors governing TP63 isoform binding and lay the groundwork for rational design of inhibitors with antitumorigenic effects.

Here, we provide a detailed model of TAp63–TP53^CM complex combining computational homology modeling and molecular dynamics (MD) simulations techniques with experimental approaches. Then, we studied different strategies to induce antitumorigenic activity of TAp63 as well as to prevent protumorigenic function of ΔNp63 isoforms. Our data showed that there is potential in developing therapeutics to interfere in the TAp63–TP53^CM interaction.

The structure of TP63 TID was modeled using I-TASSER Suite 5.1 package (27); the last 71 residues from the FASTA sequence (UniProt ID: Q9H3D4–1) of TP63 protein were used as template (607-LHEFSSPSHLLRTPSSASTVSVGSSETRGERVIDAVRFTLRQTISFPPRDEWNDFNFDMDARRNKQQ RIKEE-678). I-TASSER confidence score (C-score) were first used to estimate the global accuracies of the proposed models. C-score should be in the range of −5 to 2, where a higher score indicates a high confidence. The curated models were further analyzed and evaluated by performing atomistic MD simulations.

TP63 OD and TP53 DBD structures were adapted from their crystal structures [PDB ID: 4A9Z (ref. 28) and 1TUP, respectively]. Chains A and B of 4A9Z were used to study TP63 OD dimerization. Chain A in 1TUP covers TP53 residues from 94 to 289, which engulfs DBD almost completely. I-TASSER was used to model whole-length TP53 DBD and also to include the residues important in TP63 interaction (21). Therefore, our final TP53 DBD structure consisted of TP53 residues from 94 to 312. The structural zinc atom attached to TP53 DBD was kept because it is an important factor in TP53 interaction. Conformational mutations were introduced to TP53 using PyMOL (The PyMOL Molecular Graphics System, Version 2.3.2 Schrödinger, LLC) protein mutagenesis tool.

The modeled structures such as TP63 TID and TP53^CM DBD were first minimized (see energy minimization below in atomistic MD simulations section) to eliminate any steric clashes or inappropriate geometry. To obtain TP63 TID and TP53^CM DBD complex, the minimized models were docked using the prediction interface of the HADDOCK-docking program (29, 30). The active site for TP53^CM DBD was defined as residues between 251 and 312, whereas the whole 71 residues of TP63 TID were chosen as docking site (21). HADDOCK predictions were ranked according to a scoring function called HADDOCK score.

To further evaluate the predicted complex structures, HotRegion web server was used to identify the interaction interfaces as well as hot spots of the structures (31).

After obtaining initial complex structures of TP63 TID and TP53^CM DBD using HADDOCK, we performed unbiased docking to (i) further validate our interaction interface and (ii) identify short peptide sequences crucial for the interaction to be used as template for the peptide design. In this method, the preferential interactions of the peptides to various parts of the target protein in all-atom explicit solvent MD simulations were measured (32).

We applied this method to both TP63 TID–TP53^CM DBD (predicted by HADDOCK) and TP63 OD–TP63 OD (adapted from 4A9Z; ref. 28) complexes. We designed our peptides based on the unbiased docking results. BeAtMuSiC web server was used to predict binding-free energy changes upon any mutations to the interface residues (33).

Our initial TP63 TID models, TP53 DBD and TP53^CM DBD structures, TP63 TID–TP53^CM DBD, and TP63 OD–TP63 OD complexes from unbiased docking, refined TP63 TID–TP53^CM DBD complex model and finally the peptides we designed were subject to the MD simulations in an aqueous environment. These initial configurations were simulated for 100 ns using the GROningen MAchine for Computer Simulations (GROMACS-2018) on an exacloud cluster at Oregon Health & Science University (Portland, OR). MD simulations were carried out in an aqueous environment. The TIP3P water model was used to solvate the system in an aqueous environment with proper number of counterions (Na⁺ or Cl⁻) to ensure charge neutrality. A 3D periodic box was used to center the complex with at least 1.0 nm from the edge, accounting for >2 nm of solvent buffer. Force fields for the proteins were generated using AMBER-type force-field (FF99SB-ILDN; ref. 34). Energy minimization was performed for 10,000 steps where we applied a steepest descent algorithm. The equilibration and production runs were run in NPT ensemble, where the temperature was maintained at 300K using V-rescale thermostat with a temperature coupling constant of 0.2 ps (35) starting from a random distribution of velocities generated consistent with 300K. The pressure was maintained at 1 bar using Parrinello–Rahman barostat with a 5 ps period of pressure fluctuations at equilibrium (36). The all bonds were constrained in the system. The MD simulations incorporated leap-frog algorithm with a 2 fs time-step to integrate the equations of motion. The long-ranged electrostatic interactions were calculated using particle mesh Ewald (37) algorithm with a real space cutoff value of 1.2 nm. Lennard-Jones (LJ) interactions were also truncated at 1.2 nm. Coordinates of the protein molecule were stored every 1 ps for further analysis. Zinc AMBER Force Field (ZAFF) was used to develop parameters for zinc ion attached to TP53 DBD (38). A bonded model was used where zinc ion was covalently bonded to Cys176, His179, Cys238, and Cys242 (39).

To monitor the systems to reach equilibration, root–mean–square deviation (RMSD) and radius of gyration (Rg) were calculated as a function of time. Root–mean–square fluctuation (RMSF) calculations were calculated for each residue in the systems.

All peptides were synthesized by GenScript ≥98% pure. Tat peptide–bearing sequence YGRKKRRQRRR was used as negative control for cell viability assay. Some peptides were synthesized with a C-terminal-(Lys)-biotin tag for Bio-layer interferometry experiments and are noted in Table 1. Peptide stocks of 10 mg/mL were freshly prepared in PBS for in vitro experiments.

Recombinant protein of human TP63 was purchased from OriGene Technologies (Cat# TP710041) and of human TP53 was purchased from Millipore Sigma-Aldrich (Cat# 23–034).

Human PDAC cell lines BxPC-3 (ATCC CRL-1687) and AsPC-1 (ATCC CRL-1682) were cultured in RPMI-1640 media supplemented with 10% FBS and 1X penicillin–streptomycin (PS) and grown at 37°C with 5% CO₂. All cell lines were tested for Mycoplasma that was negative and were authenticated using STR profiling at OHSU DNA Services Core.

Promega Cell-titer 96 Aqueous One Solution Cell Proliferation Assay kit (MTS Assay; Cat# G3580) was used to measure cell viability in MTS [3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium)] assay. Briefly, 100 μL of BxPC-3 (TP63 high; TAp63 low, ΔNp63 high with Y220C conformational mutation in TP53) and AsPC-1 (TP63 low; TAp63 low and ΔNp63 null) cells were seeded in 96-well plates at 1 × 10³ cells/well in RPMI-1640 media containing 10% FBS and 1X PS and grown overnight at 37°C with 5% CO₂. The following day cells were treated with or without different concentrations of test peptide. Peptide concentrations used for treatment included 150, 100, 75, 50, 37.5, 25, 18.8, 12.5, 9.4, 6.3, 4.7, and 3.1 μmol/L in 100 μL media. 24 hours following treatment a second fresh treatment of the same dose was given. 48 hours from the first treatment dose, 10 μL of Cell-titer 96 aqueous one reagent was added to each well (100 μL) and incubated for 4 hours at 37°C. The absorbance at 490 nm was recorded on a Tecan Spark 20 M multimode plate reader. For each individual experiment, six wells of each peptide concentration, was read and values averaged. Each individual experiment was repeated three times. Cell survival graphs were plotted and expressed as a fraction of the number of cells survived relative to untreated control cells for each peptide concentration. IC₅₀ values for each peptide, which refers to the concentration required to inhibit 50% of cell proliferation, were calculated using non-linear regression analysis.

BxPC-3 or AsPC-1 cells were seeded into 96-well plates (5,000 cells per well), cultured overnight, and then treated with the IC₅₀ value of 20 μmol/L for Peptide OD-e. Each treatment condition was repeated in triplicate. Promega Real-Time-Glo Annexin V Apoptosis and Necrosis Assay Kit (Cat #JA1011) was used to measure loss of membrane integrity for necrosis according to the manufacturer's instructions. Fluorescence signal was sequentially collected throughout peptide exposure for 0, 15, 30, 60, 120, 180, 240, 300, 360 minutes using the Tecan Spark 20 M multimode plate reader.

For absolute quantification of TAp63 expression in BxPC-3 and AsPC-1 cells RT-ddPCR was performed with Forward primer 5′-TGTATCCGCATGCAGGACT-3′, Reverse primer 5′-CTGTGTTATAGGGACTGGTGGAC-3′ and Probe 5′ 6-FAM/ZEN-TCCTGAACAGCATGGACCAGCA 3′ IBFQ as previously described (40). ddPCR for apoptotic gene expression was carried out with primers and probes for Cyclin-dependent kinase inhibitor 1a CDKN1A/p21 (Bio-Rad Cat #10031256, Assay ID dMmuCPE5096496), PMAIP1/NOXA (Bio-Rad Cat# 10031252, Assay ID dHsa5037552) and BBC3/PUMA (Bio-Rad Cat# 10031255, Assay ID dHsaCPE5034069) genes following treatment of BxPC-3 and AsPC-1 cells with the IC₅₀ value of 20 μmol/L Peptide OD-e for 30 minutes.

RNA was extracted using the Qiagen RNeasy Kit (Cat# 74004). 5 ng of RNA was used for 22 μL PCR reaction mixtures using the One-Step RT-ddPCR Advanced Kit for Probes (Bio-Rad Cat #1864022) following the manufacturer's protocol. ddPCR droplets were generated using the Bio-Rad Automated Droplet Generator. Plate containing ddPCR droplets was sealed and RT-PCR was performed using the C-1000 Thermal Cycler (Bio-Rad). ddPCR was performed in the QX200 Droplet Digital PCR System (Bio-Rad). Analysis of the ddPCR data was performed using QX200 analysis software.

Bio-layer Interferometry (BLI) measurements were performed on a ForteBio Octet RED384 instrument (Molecular Devices) using ForteBio Data Acquisition software 9.0 and ForteBio biosensors (Molecular Devices). Kinetic assays were performed with 1X Kinetics running buffer (Molecular Devices) at 30°C using settings of Standard Kinetics Acquisition rate (5.0 Hz, averaging by 20) at a sample plate shake speed of 1,000 rpm. Streptavidin sensors (SA; Molecular Devices) equilibrated in PBS were mock loaded (no ligand reference sensor) or loaded with 25 nmol/L of C-terminal biotinylated peptide (GenScript) ligands with an approximate response of 1 nm. To establish a baseline, ligand loaded SA sensors were equilibrated in running buffer and subsequently dipped into wells containing 2-fold dilutions of TP53 (Millipore Sigma-Aldrich; Cat #23–034), or TP63 (OriGene Technologies; Cat #TP710041) analytes in running buffer, or buffer alone (reference sample). Seven analyte concentrations ranging from 4.7 to 300 nmol/L were allowed to associate for 600 seconds, followed by dissociation step where the TP53/TP63-bound peptide sensors were washed with running buffer for 600 sec. Binding curves were analyzed using ForteBio Data Analysis HT 10.0 evaluation software. To control for background signal and non-specific binding, raw experimental data were processed by subtracting reference biosensor and reference sample (no ligand). Processed data from at least six different concentrations of analyte binding were globally fit to a 1:1 Langmuir binding model to calculate the association rate constant k_a, the dissociation rate constant k_d to achieve R² > 0.90. Equilibrium dissociation constants K_D that defines the strength of the interaction or affinity was calculated as the kinetic dissociation rate constant divided by the kinetic association rate constant.

The data generated in this study are available upon request from the corresponding author.

Because the interaction of TP63 TID and TP53 DBD domains are crucial for the formation of the TP63–TP53 complex (8, 17, 20), we first modeled TP53 DBD and TP63 TID structures. We studied the changes in TP53^CM alleles: V143A, R175H, Y220C, R249S, and I254R (Supplementary Fig. S1). Our findings suggested that these mutations mostly affected the conformations of the regions between residues 172 and 195 (the first region) as well as residues 256 and 267 (the second region; Supplementary Fig. S1A and S1B). Our RMSF calculations showed that the mutations increased the fluctuations in the first region, whereas the conformation of the second region was more stabilized compared with WT TP53. This difference is more obvious in our TP53^CM (Supplementary Fig. S1A), which suggests that conformational changes in the first and second regions of TP53^CM may facilitate interaction with TP63. Parallel with our findings, a study by Li and colleagues (41) also concluded that the R175H mutation increases the flexibility of the residues between 163 and 194 (corresponding to the first region).

After obtaining the TP53^CM DBD structure and gaining more insight about the conformational effects of the mutation, we focused our effort to TP63 TID modeling. We computationally modeled TP63 TID structure. Initially, we obtained five models using I-TASSER protein structure prediction approach (27), and then we continued with extensive MD simulation calculations to refine our model (Supplementary Fig. S2). On the basis of C-score provided by I-TASSER, RMSD, and Rg calculations, model 1 was chosen as the model with the highest quality.

Next, we used TP63 TID (Supplementary Fig. S2, Model S1) and TP53^CM DBD models to predict the structure of TP63 TID–TP53^CM DBD complex using HADDOCK flexible docking web server (29). A total of 20 clusters with 110 structures were generated by the server. The top five clusters (consisting of 40 complex structures) with the lowest HADDOCK score were further evaluated. Interaction interfaces and hot spots of those complex structures were identified using HotRegion web server (31). TP63 TID residues 610–617 and 644–686 and the TP53^CM DBD residues 271–297 most frequently contributed to the interactions in all five clusters (Supplementary Table S1). By using homology modeling, MD simulations and flexible docking approaches, we obtained preliminary TP63 TID–TP53^CM DBD complex structure models for use in more complex simulations.

To further refine the top models of TP63 TID–TP53^CM DBD complex structure, we applied an iterative unbiased docking approach. We first extracted the TP63 TID and TP53^CM DBD regions that are most crucial in the interaction based on identified interface and hot spot residues (Supplementary Table S1). These extracted regions corresponded to 645-PRDEWNDFNFDMDARRNKQ-672 for TP63 TID and 273-RVCACPGRDRRTEEENLR-290 for TP53^CM DBD. Because, this region on TP53 does not correspond to any conformational mutations, the peptides we designed to target TP63 TID have the potential to work for different cell lines with different TP53 conformational mutations. Next, we randomly distributed five TP63 TID and five TP53^CM DBD-derived peptides around full-length TP53^CM DBD and full-length TP63 TID, respectively (Fig. 2). We then simulated the systems where any initial interaction between the peptides and proteins was avoided (Fig. 2). Our aim here was to validate and refine our predicted interaction interfaces with an unbiased docking approach. Two of the TP63 TID-derived peptides interacted with predicted TP53^CM DBD interface (residues 271–297, sphere representation; Fig. 2A). TP53^CM DBD residues 243, 248, 250, 271, 273, 280, 283–285, 287, 293–296 were involved in this interaction with TP63 TID derived peptides. However, one of the peptides docked to a different interface covering residues 150–154, 199–202, 219–225 and 230–233 (labeled with the ellipse). This potential secondary binding site did not correlate well with previous experimental data and the primary interface, to which the other two peptides interacted. However, this potential secondary binding site on TP53^CM DBD was not disregarded but instead was used to derive TP53^CM DBD peptides. Of note, two remaining peptides out of five did not interact anywhere on TP53^CM DBD.

Unbiased docking results of TP53^CM DBD-derived peptides targeting full-length TP63 TID were shown in Fig. 2B. This time four out of five TP53^CM DBD-derived peptides (one was not docked anywhere on TP63 TID) docked to the previously predicted binding site (residues 610–617 and 644–686). These results support that TP63 TID residues 617–618, 648, 652–655, 658–659, 668–669, 672 and 680 are involved in the interaction with TP53^CM DBD-derived peptides.

As a result, a final model of TP63 TID–TP53^CM DBD complex structure was further refined after unbiased docking (Supplementary Fig. S3A). We simulated the complexes for 100 ns to reach an equilibrium and study the conformational dynamics as well as the interactions. The conformational mutations R175H, Y220C, and I254R (CM) on TP53 are on the different sites of our predicted TP63 TID–TP53 DBD interfaces, indicating that they are not directly involving in the interaction (Supplementary Fig. S3A). Moreover, these mutations are promoting the conformational changes on the same regions (Supplementary Fig. S1A, the first and the second region). This suggests that our predicted model for TP63 TID–TP53 DBD does not significantly change depending on CM. To investigate the potential effects of CM to our peptide design, we also calculated binding free energy of TP63–TP53^CM interactions (Supplementary Fig. S3B). We observed that binding affinities (especially for TP63–TP53^R175H and TP63–TP53^Y220C) are not considerable different than each other upon different mutations. Therefore, this indicates that the peptides we designed on the basis of TP63 interaction interface as well as the ones we designed on the basis of TP53 interaction interface do not depend on CM.

Identification of the interaction sites of proteins is fundamental for a better understanding of their function and many biological processes and drug discovery (42). As we mentioned, there are no reported agents to disrupt the TP63–TP53^CM complex and restore TP63 tumor-suppressor activity. Therefore, we used the TP63 TID–TP53^CM DBD complex model to design peptides based on the interaction interfaces of the complex. These peptides were designed to compete with the proteins they were derived from for binding to their target and inhibit the TP63–TP53^CM interaction. The peptides derived from the TP63 TID interface were designed to bind to TP53^CM DBD and the peptides derived from the TP53^CM DBD interface were designed to target TP63 TID to the release of TP63. Because blocking TID should inhibit inactive TAp63 dimer formation, which has a 20-fold more affinity for TP53-RE compared with the monomeric form (8), this design had more potential to induce TAp63-mediated apoptosis.

In addition, we wanted to extend our peptide design efforts to cover ΔNp63 biology. Because ΔNp63 isoforms can induce squamous-like lineage, we concentrated on promoting TAp63-mediated apoptosis and inhibiting ΔNp63-mediated squamous transdifferentiation as a comprehensive solution for eliminating cancer cells. To inhibit ΔNp63 activity, we designed peptides derived from its OD to prevent ΔNp63 oligomerization. The crystal structure presenting the tetramerization of ODs was already available (PDB ID: 4A9Z), therefore, we used that structure to rationally design our peptides.

The rationally designed peptides were named after their target proteins (Table 1). For example, Peptide TID-a, was designed on the basis of the TP53^CM DBD interface to target TP63 TID. To promote the uptake of peptides in cells, we tried two different cationic cell penetration strategies. The first strategy involved the addition of the naturally occurring HIV-1 Tat protein transduction domain (YGRKKRRQRRR) to the N-terminus of some peptides to facilitate crossing the cell membrane barrier (see the bold residues in Peptide TI-a/d, Peptide DBD-a, and Peptide OD-a/c/d/e in Table 1). As a second cell penetration strategy, we introduced cationic Arg residues into the peptides by mutating some non-hot spot residues (see the red residues in Peptide TID-c, Peptide DBD-a/b, and Peptide OD-a/b). The TP53^CM DBD interface residues between 273 and 290 (RVCACPGRDRRTEEENLR) were chosen to target TP63 TID. This region included the most hotspot residues at the interface (Supplementary Table S1). The details about the design of individual peptides are discussed in Supplementary Methods. These designed peptides will allow us to initiate the antitumorigenic activity and inhibit the protumorigenic activity of TP63 isoforms.

We screened each of our rationally designed peptides based upon either the TP63–TP53 interaction (Peptides TID and DBD) or the TP63 self-interaction (Peptide OD) in two different PDAC cell lines BxPC-3 and AsPC-1, to assess their ability to alter cell viability and induce cell death. BxPC-3 cells have previously been reported to express high levels of TP63 and carry mutant TP53^Y220C, whereas AsPC-1 cells have previously been reported to have low levels of TP63 expression and null for TP53 (43). Of the six peptides designed to the TP63–TP53 interaction (Peptide TID-a/b/c/d and Peptide DBD-a/b), only three peptides Peptide TID-a, TID-d, and DBD-a decreased cell viability in the PDAC cell lines (Table 1, Fig. 3B, Supplementary Fig. S4B). These peptides all contained the Tat domain suggesting this was necessary for cellular uptake (Table 1). By contrast, peptides like Peptide TID-c, which has cationic Arg residues introduced for cell penetration, or lacking the Tat domain (Peptide TID-b) had no effect on cell viability (Table 1; Supplementary Fig. S5), suggesting that these peptides perhaps did not work due to a lack of cellular entry.

Peptides designed on the basis of TP53^CM DBD interface to bind TP63 TID such as Peptide TID-a, and TID-d reduced cell viability in both PDAC cell lines (Table 1; Fig. 3B). However, unlike Peptide TID-a, which had similar cell viability inhibition (IC₅₀ = 120 μmol/L) in both cell lines (Fig. 3B, left), Peptide TID-d showed preferential reduction of cell viability in TP63 high BxPC-3 (IC₅₀ = 78 μmol/L) cells compared with AsPC-1 cells (IC₅₀ = 107 μmol/L; Fig. 3B, right). Moreover, Peptide TID-d that was based on the secondary TP53^CM DBD interface identified through unbiased docking had a 1.5-fold improvement in IC₅₀ value over Peptide TID-a; therefore, we tested the affinity of Peptide TID-d to TP63 using Bio-layer interferometry. Our data showed that Peptide TID-d bound TP63 with a K_D = 0.08 nmol/L suggesting a strong affinity (Fig. 3C; Supplementary Table S2). Similar to the TP63 TID target peptides, Peptide DBD-a had a moderate affinity for its target TP53 (Supplementary Fig. S4) but did not reduce viability in BxPC-3 cells compared with AsPC-1 cells, hence we decided not to further study this peptide. Furthermore, when we tested the Tat control peptide in our cell viability assay (Supplementary Fig. S6), we did not see an effect on cell viability in both PDAC cell lines. Taken together, these data suggest that the nmol/L affinity of Peptide TID-d for its target TP63, combined with preferential reduction of cell viability in TP63 high cells makes it a candidate peptide for future refinement and potential therapeutic applications in cancer (Fig. 3B right, and C, Supplementary Table S2).

Next, a series of TP63 self-interaction (OD-derived peptides) that were designed to inhibit ΔNp63 oligomerization were evaluated for their ability to inhibit PDAC cell viability (Fig. 4). Treatment with four Peptides OD-a/c/d/e containing an N-terminal Tat domain showed a dose-dependent decrease in cell viability of BxPC-3 and AsPC-1 cells (Fig. 4B). As in the above experiments, Peptide OD-b lacking the Tat domain had no effect on cell viability (Supplementary Fig. S6). Treatment of BxPC-3 cells with Peptide OD-c resulted in reduced cell viability at a lower dose (IC₅₀ = 65 μmol/L) compared with Peptide OD-a (IC₅₀ = 145 μmol/L). Because both peptides have similar sequences and target the primary TP63 OD interface, the different dosages required to achieve cell growth inhibition in TP63 high cells can be attributed to the cationic Arg residues introduced in Peptide OD-a. Furthermore, Peptide OD-e that was rationally optimized from Peptide OD-a, had the highest growth inhibitory effects in BxPC-3 cells (IC₅₀ = 18 μmol/L). Finally, the rationally optimized Peptide OD-d targeting a secondary TP63 interface also inhibited cell viability with an IC₅₀ = 104 μmol/L in BxPC-3 cells. From these results we have found two lead candidates to inhibit TP63 pro-tumorigenic activity, the native sequence Peptide OD-c and the rationally optimized Peptide OD-e, both of which showed specific growth inhibition of TP63 high cells compared with TP63 low cells. Furthermore, these peptides promoted cell death in ΔNp63 high BxPC-3 cell line, suggesting that the release of TP63 from TP53^CM is not promoting ΔNp63 oligomerization but facilitating cell death.

To further evaluate the binding of inhibitory peptides Peptide OD-c/d/e for TP63, we performed BLI experiments (Fig. 4C; Supplementary Table S2). Briefly peptides were immobilized on streptavidin sensors and kinetic parameters of TP63 binding was assessed in real time using a 1:1 binding model. Affinities (K_D) of Peptide OD-c/d/e for TP63 were 8.2, 5.1, and 191 nmol/L, respectively. Interestingly Peptide OD-e, which required a low dose (IC₅₀ = 18 μmol/L) for cell growth inhibition in our PDAC cell viability assay, demonstrated fast kinetics in our BLI experiments with a half-life of 38 seconds, but it had lower affinity compared with the other inhibitory peptides. Although the binding affinity of Peptide OD-e is less than the other peptides, the greatly reduced cell viability in BxPC-3 cells suggested that the weaker association of Peptide OD-e is better at interfering with the TP63 self-interaction and worth further studies.

Because Peptide OD-e was one of the most promising TP63 therapeutics, we performed a series of tests to understand whether the peptide is cytotoxic and by what pathway. First, we measured the levels of TP63 in BxPC-3 and AsPC-1 cells lines and found a 5-fold increase in TP63 expression in BxPC-3 cell line, as expected (Fig. 4D). Second, we measured real-time cytotoxicity after treatment with Peptide OD-e and found a rapid loss of membrane integrity in the BxPC-3 cell line compared with AsPC-1 cell line (Fig. 4E). Finally, we showed that Peptide OD-e rapidly induced p21 and NOXA expression after treatment, but not PUMA (Fig. 4F). Because p21 and NOXA are both regulated by TP63 and TP53, whereas PUMA is regulated by TP53, this result suggests that Peptide OD-e is working through TP63 or a combination of TP63- and TP53-mediated cell death pathways.

Overall, our experimental data showed that our rationally designed peptides could release TAp63 to carry its antitumorigenic activity and prevent ΔNp63 protumorigenic activity in TP63 enriched PDAC cells for therapeutic applications in cancer. Furthermore, it is tempting to speculate that peptides targeting two different aspects of TP63 different interactions, when given in combination could potentially improve treatment efficacy if the peptide do not interfere with each other. To test this, we developed a computational model for using a combination of the peptides (Supplementary Fig. S7). We modeled TP63 OD and TID together and introduced Peptide TID-d and Peptide OD-e, simultaneously. The MD simulation of the model showed that none of the interacting residues on TP63 OD - Peptide OD-e and TP63 TID-Peptide TID-d interfaces are overlapping, suggesting that using the combination of our peptides will not interfere with each other. Taken together, we have demonstrated that rationally designed peptides have the potential to be used for cancer treatment.

TP63 has both antitumorigenic and protumorigenic activity depending on the isoform (2). Like TP53, in normal cells, antitumorigenic TAp63 isoforms can transactivate TP53-responsive genes and cause cell-cycle arrest and apoptosis (6–8). On the other hand, ΔNp63 isoforms are expressed at low levels, cannot promote transactivation of TP53 target genes, but can regulate the expression of their own set of squamous lineage genes (Fig. 5, left). About half of cancers harbor TP53 mutations that change the dynamics of these interactions (13, 14). First, TP53^CM interacts with TAp63 and inhibits its transcriptional activity resulting in impairment of both TP53-mediated tumor suppression, and TP63-induced apoptosis (Fig. 5, middle). Second, TP53^CM also helps ΔNp63 to enhance its protumorigenic activity (20) and ΔNp63 continues to drive squamous lineage (11, 12). In addition, ΔNp63 comprise as many as 25% of pancreatic cancers, and the majority of cancers originating from lung, head and neck, urothelial, cervical, and skin (4, 9, 10). ΔNp63 was also shown to enhance cell motility and invasion in human PDAC cells (9). Together, this illustrates the collective importance of the interactions of TP63 isoforms with TP53 mutants especially in the context of cancer. Understanding the TP63–TP53^CM interactions, therefore, can facilitate novel therapeutic strategies to induce apoptosis aimed at targeting TP63 rather than TP53^CM; however, there is no available structure of the TP63–TP53^CM complex.

Proteins usually carry out their functions by interacting with other proteins. Although it is challenging to come up with new drugs and drug targets, one strategy for addressing this problem is to target protein–protein interactions (PPI; refs. 44, 45). Therefore, it is crucial to study PPIs to learn more about their functions as well as potentially target them. Here, we have taken a computational approach using existing experimental evidence, available structures of TP53 and TP63, molecular modeling, MD simulations and molecular docking to develop a model of the TP63–TP53 complex. We started by identifying details about the conformational changes in TP53 structure that occurred as a result of key mutations. Then, we developed a detailed model of the TP63–TP53^CM complex where we identified residues involved in the PPI interface, hot spots within this interface and intermolecular interactions between two proteins (Supplementary Fig. S3A). This computational model is the first view of the TP63 TID–TP53 DBD complex. We further tested the validity of our model and its implications designing peptide-based therapeutics. We developed two strategies to validate PPI interfaces of our computational models and assess their potential in drug targeting.

Using our computational model of TP63–TP53^CM as well as available TP63 OD tetramer structure, we rationally designed inhibitory peptides (Table 1) to promote cell death in TP63 enriched cancer cells by releasing TAp63 from mutant TP53 and inhibiting ΔNp63 oligomerization (Fig. 5, right). We demonstrated five of the eleven peptides could preferentially inhibit cell growth in a TP63 high PDAC cell line (Table 1; Figs. 3 and 4). We showed a lead candidate peptide, Peptide OD-e, which was rationally designed to bind the primary TP63 OD interface bound TP63 with a modest affinity (K_D = 191 nmol/L, Supplementary Table S2; Fig. 4C) and preferentially decreased cell viability (IC₅₀ = 18 μmol/L) in TP63 high PDAC cell line (Fig. 4B). Notably, optimization of Peptide OD-e from Peptide OD-a resulted in an 8-fold increase in cell growth inhibition in our PDAC cell viability model with an increase in cellular cytotoxicity most likely working through TP63 or a combination of TP63 and TP53-mediated cell death.

The results from our experimental studies provide proof that rationally designed inhibitory peptides can bind their TP63 targets and have clinical significance in the context of cancer where >25% of tumors are addicted to TP63 (9). Taken together, this study opens new avenues for rational design and delivery of peptide-based drugs in cancer. Previous studies using CRISPR–Cas9 and sgRNA-targeting TP63 showed that knocking down TP63 can kill TP63-addicted cancer cells (9). However, genetic manipulation has limited clinical feasibility, thus we focused our efforts on peptide-based therapeutics. Peptide-based drugs have the potential to modulate anticancer activity (46) and are promising with approximately 140 peptide-based drugs being evaluated in the clinic (47). It is possible that these inhibitory peptides can induce apoptosis and prevent cell invasion. Future studies should be directed at optimizing these peptides and performing a more detailed investigation of their effects in the context of cancer.

No disclosures were reported by the authors.

E.S. Ozdemir: Writing–review and editing, conceptualization, data curation, software, formal analysis, investigation, visualization, methodology, writing–original draft. M.M. Gomes: Writing–review and editing, conceptualization, formal analysis, investigation, visualization, methodology, writing–original draft. J.M. Fischer: Writing–original draft, writing–review and editing, conceptualization, supervision.

We thank Dr. P.A.J. Muller for helpful discussion on the current gaps in the field of TP63 biology and TP63–TP53 interaction. This work was supported by the Cancer Early Detection Advanced Research Center at Oregon Health and Science University (Project ID: Full6911219 and Exploratory 2022–1503).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.