Disruption of focal adhesion kinase and p53 interaction with small molecule compound R2 reactivated p53 and blocked tumor growth

The HCT116p53^-/^- and HCT116p53⁺/⁺ colon cancer cells were obtained from Dr. Bert Vogelstein (Johns Hopkins University) and maintained in McCoy’s5A medium with 10% FBS and 1 μg/ml penicillin/streptomycin. The HCT116 cell lines were authenticated by Western blotting with p53 antibody and passaged less than 6 month after resuscitation of frozen aliquots. MCF-7, PANC-1, and SW620 cells were obtained from ATCC and cultured according to the manufacturer’s protocol. The cell lines were passaged less than 6 month after resuscitation of frozen aliquots.

The FAK monoclonal FAK (4.47) antibody was purchased from Upstate Biotechnology, phospho-Y397-FAK antibody was obtained from Biosource Inc. Monoclonal anti-β-actin antibody was obtained from Sigma. Anti-p53 antibody (Ab-6, clone DO-1) was obtained from Oncogene Research Inc. p21, Mdm-2 and Bax antibodies were obtained from Santa Cruz.

The p21-pGL3, BAX-pGL3 and Mdm-2-pGL2 promoter luciferase constructs, were described previously [18]. The recombinant baculoviral FAK [18] was used for pull-down assay. The FAK-NT (1–422 aa) fragment was subcloned into the pET200 vector (Invitrogen) and the His-tagged FAK-NT protein was isolated according to the instructions of the Ni-NTA Purification system kit (Invitrogen). The recombinant p53 was obtained from BD Pharmingen. The R2 compound (1-benzyl-15,3,5,7-tetraazatricyclo [3.3.1.1~3,7~] decane) was kindly provided by Drs. Ethirajan Manivannan and Ravindra Pandey. A18 compound (1,4-bis(diethylamino)-5,8-dihydroxy anthraquinon) [32] and M13 compound (5′-O-Tritylthymidine) [33] were obtained from NCI and Sigma, respectively.

We used a structure-based approach combining docking of FAK and p53 peptide interaction and molecular docking of small molecule compounds with functional testing, as described [33]. Initially, we predicted the three dimensional structure of the p53 region involved in interaction with FAK in the N-terminal domain of p53 by the PHYRE (Protein Homology/analog Y Recognition Engine) server (http://​www.​sbg.​bio.​ic.​ac.​uk/​phyre) [34]. PHYRE is an efficient protein structure prediction method by sequence homology to existing structures [34]. While the portion of the p53 region described [35] was successfully modeled by the PHYRE server, the region, which involved in interaction with FAK-NT [20] was predicted as disordered. We therefore isolated the disordered seven-amino-acid peptide (RMPEAAP) known to be involved in interaction with FAK [20] from the model, assigned residue charges and add hydrogen atoms with the UCSF CHIMERA program and performed flexible docking to the FAK-FERM domain by DOCK 6.0 software to find the highest scoring complex of FAK and p53 peptide. The crystal structure of FAK, N-terminal FERM domain (PDB ID:2AL6), reported [31] was used for docking and computer modeling of the FAK and p53 peptide interaction. To model the FAK-NT-p53 peptide interaction, the DOCK 6.0 software analyzed >10,000 possible orientations of this interaction, based on the scores of the resulting interfaces using electrostatics (ES) and van der Waals (vWS) energies. The model with the highest scoring of FAK-NT and p53 peptide interaction has been generated and compared with the FAK lobes amino acids reported recently to interact with FAK [19], and FAK-NT region [20]. All binding poses were evaluated using the DOCK grid-based scoring, involving the non-bonded terms of the AMBER molecular mechanics force field (vDW+ES).

More than 200,000 small-molecule compounds from National Cancer Institute Development Therapeutics Program NCIDTP library (http://​dtp.​nci.​nih.​gov) [36] and compounds from ZINC UCSF (University of California, San Franscisco) database (http://​zinc.​docking.​org/​catalogs/​ncip (version 12) [37] following the Lipinski rules were docked into the pocket of the N-terminal domain of FAK and p53 interaction in 100 different orientations using the DOCK5.1 program. The spheres describing the target pocket of FAK-p53 were created using the DOCK 5.1 suite program SPHGEN. Docking calculations were performed on the University of Florida High Performance Computing supercomputing cluster (http://​hpc.​ufl.​edu). Scores were based on a grid spaced five angstroms from the spheres selected for molecular docking. Each compound was docked in 100 orientations, and grid-based energy scores were generated for the highest scoring orientations. Scores approximate delta G values based on the sum of polar electrostatic interactions and van der Waals energies. Small molecule partial atomic charges were calculated using the SYBDB program, as described [38, 39].

The top compounds that were detected by the DOCK5.1 program to best fit into FAK-p53 pocket were ordered from the NCI/DTP database free of charge. Each of the compounds (called Roslin compounds) was solubilized in water or DMSO at a concentration of 25 mM. The R2 compound was chemically synthesized for biochemical analyses in vitro and for mice studies in vivo.

The 1000 cells were plated on 6 well plates and incubated with or without tested compound for 1–2 weeks. Then cells were fixed in 25% methanol and stained with Crystal Violet, and colonies were visualized and counted.

The cells (1×10 4 cells per well) were plated on a 96 well plate and after 24 hours treated with compounds at different concentrations for 24 hours. The 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium compound from Promega Viability kit (Madison, IL) was added, and the cells were incubated at 37C for 1–2 hours. The optical density at 490 nm on 96-plate was analyzed with a microplate reader to determine cell viability.

Western blotting, immunoprecipitation, immunostaining and immunohistochemical staining using were performed, as described [40].

For the pull-down assay we used recombinant baculoviral FAK, GST and GST-p53 proteins, as described [18] and performed pull-down assay, as described [20].

The binding was performed by ForteBio Inc. company (http://​www.​fortebio.​com). The human FAK-N-terminal domain protein was biotinylated using NHS-PEO4-biotin (Pierce). Superstreptavidin (SSA) biosensors (FortéBio Inc., Menlo Park, CA) were coated in a solution containing 1 μM of biotinylated protein. A duplicate set of sensors was incubated in an assay buffer (1× kinetics buffer of ForteBio Inc.) with 5% DMSO without protein for use as a background binding control. Both sets of sensors were blocked with a solution of 10 mg/ml Biocytin for 5 minutes at 25°C. A negative control of 5% DMSO was used. The binding of samples (500 μM) to coated and uncoated reference sensors was measured over 120 seconds. Data analysis on the FortéBio Octet RED instrument was performed using a double reference subtraction (sample and sensor references) in the FortéBio data analysis software.

The dual-luciferase was performed, as described (18). In brief, 2×10⁵cells were plated on 6-well plates, and co-transfected with the p21, Mdm-2 or Bax promoters in the pGL2 or pGL3-luciferase containing plasmids (1 μg/well) and pPRL-TK plasmid containing the herpes simplex virus thymidine kinase promoter encoding Renilla luciferase (0.1 μg/well) using Lipofectamine (Invitrogen) transfection agent according to the manufacture’s protocol. HCT116 p53^-/^- cells were co-transfected with the above plasmids and p53 in the presence or absence of FAK plasmids and tested either without or with 25 microM R2 compound for 24 h.

Flow cytometry analysis was performed by the standard protocol at Roswell Park Flow Cytometry Core Facility. The percentage of G1, G2, S phase-arrested and/or apoptotic cells was calculated.

Female nude mice, 6 weeks old, were obtained from Harlan Laboratory. The mice experiments were performed in compliance with IACUC protocol approved by the Roswell Park Cancer Institute Animal Care Committee. HCT116 p53⁺/⁺ and p53 ^-/^- cells (3.7×10⁶ cells/injection) were injected subcutaneously into the right and left leg side of the same mice, respectively. Three days after injection, the R2 compound was introduced by IP injection at 60 mg/kg dose daily 5 days/week. Tumor diameters were measured with calipers and tumor volume was calculated using this formula = (width)²×Length/2).

Student’s t test was performed to determine significance. The difference between treated and untreated samples with P<0.05 was considered significant.

We detected the 7 amino-acid region in p53 involved in the interaction with FAK [20], and because the crystal structure of this N-terminal region of p53 remained unsolved, we performed computer modeling with a PHYRE program (Protein Homology/analog Y Recognition Engine) that allowed us to predict its three-dimensional structure, based on protein homology and an analogy recognition engine [18]. The region containing 43 to73 amino acids of the N-terminal proline-rich domain of p53 had an alpha-helical conformation and contained the 7 amino-acid peptide involved in the interaction with FAK) (Figure 1A). We performed docking of the 7 amino acid p53 peptide (65–71 amino acids) involved in interaction with FAK into the N-terminal domain of FAK and found the best complex of FAK and p53 peptide (Figure 1B). The model with the highest scoring of the FAK N-terminal domain (FAK-NT) and p53 peptide interaction was created, which included amino-acids from the F1 (33–127 aa) and F2 lobes (128–253 aa) of FAK reported to interact with p53 [19] (Figure 1C).

To find small molecule compounds targeting the FAK and p53 interaction we screened more than 200,000 small-molecule compounds from the National Cancer Institute database and docked them into the region of FAK-p53 interaction (Figure 1D). We identified a series of small molecule compounds that we called Roslins that effectively docked into the FAK-p53 interaction region (Figure 1D). The p53 peptide (blue color) and small molecules (purple color) which target the region of FAK and p53 interaction are shown in Figure 1D.

We selected 19 compounds targeting the FAK and p53 interaction, R1 to R19 (Table 1), and tested them for p53-dependent decrease of cell viability in HCT116p53⁺/⁺ and HCT116p53^-/^- cells (Figure 2A). The R2, R4-R11, R13, R17 and R18 compounds decreased HCT116 p53⁺/⁺ cell viability more efficiently than in HCT116 p53−/− cells (Figure 2A). Most of these compounds also decreased viability in a A375 melanoma cancer cell line with wild type p53 (Additional file 1: Figure S1). Then we tested R compounds that decreased viability in a p53-dependent manner for disruption of FAK and p53 interaction by immunoprecipitation of FAK and p53. Among these compounds R2, R5-R10 and R13 effectively disrupted FAK and p53 interaction. To test specificity for the FAK and p53 pathway we used control p53-null MEF FAK^-/^- cells and PANC-1 with mutant p53, as negative controls (Additional file 1: Figure S1). As expected, most of these compounds did not affect viability of control FAK^-/^-p53^-/^- cells, except for R9, R10 and R13 or PANC-1 cells with mutant p53, except for R9,R10 and R13 (Additional file 1: Figure S1). Thus, among all R compounds, R2, R5, R6, R7, and R8 were the most specific compounds in targeting the FAK and p53 interaction and pathway.

To test these compounds for long-term effects, we performed clonogenicity assays in HCT116 p53⁺/⁺ and HCT116p53^-/^- cells. Among the R compounds targeting FAK and p53, R2 compound (Table 1, marked in bold) maximally decreased clonogenicity in HCT116p53⁺/⁺ (Additional file 2: Figure S2). The R2 compound decreased clonogenicity in a p53- and dose-dependent manner (Figure 2B). The structure of R2 is shown on Figure 2C. R2 also decreased viability of HCT116 cell in a p53- and dose-dependent manner (Figure 2D). Thus, the small molecule compound R2 was selected for further study because it decreased viability and clonogenicity in a dose and p53-dependent manner in HCT116 cells.

We tested the effect of the R2 compound on viability of the MCF-7 breast cancer cell line with wild type p53. R2 decreased viability in the MCF-7 cells in a dose-dependent manner (Figure 2E). In the MDA-231 breast cancer cell line with mutant p53, R2 also decreased viability, but the significantly decreased viability was observed at higher dose than in cells with wild type p53: 50 μM in MDA-231 (Figure 2F) versus 20 μM in MCF-7 cells (Figure 2E). R2 did not significantly affect viability of Miapaca-2 pancreatic cancer cells with mutant p53 (Figure 2G). In normal fibroblasts, WI38-hTERT cells, R2 also did not significantly affect viability (Figure 2H). Thus, the lead compound R2 significantly decreased the viability of cancer cells with wild type p53, without a significant decrease of viability in normal human fibroblasts and in cancer cells with mutant p53.

We performed computer modeling of the R2 compound docked into the FAK-NT region involved in interaction with p53 protein (Figure 3A). The R2 compound effectively docked into the FAK-NT domain (Figure 3A upper panels; zoomed image, lower panel).

To detect direct binding of R2 to the N-terminal domain of FAK, we isolated human N-terminal domain of FAK and performed Real-time binding assays with R2 compound using ForteBioOctet Red384 system (Figure 3B). The assay demonstrated that R2 directly bound to the FAK-NT protein, but not to the negative control (Materials and Methods) (Figure 3B). In addition, we performed by Octet assay kinetic analysis of association and dissociation of FAK and p53 proteins, either without R2 or with three different doses of R2 (Additional file 3: Table S1). The increased doses of R2 increased dissociation constant K_D of FAK and p53 protein interaction, supporting disruption of FAK and p53 complex by R2 in a dose-dependent manner.

To test disruption of FAK and p53 binding by R2 in cells, we performed immunoprecipitation (IP) of FAK and p53 proteins in HCT116 p53⁺/⁺ cells without R2 and with R2 (Figure 3C). While we detected the complex of FAK and p53 by IP in untreated cells (Figure 3C), we did not detect this complex in R2-treated cells. Thus, the R2 compound disrupted the interaction of FAK and p53 in HCT116 cells (Figure 3C). To test that R2 directly disrupted the binding of FAK and p53 proteins, we performed pull-down assays using purified recombinant baculoviral FAK, GST and GST-p53 proteins (Figure 3D, left panel). The pull-down assay clearly showed that FAK bound to p53 without R2, but there was no binding in the presence of R2 (Figure 3D, right panel). R2 disrupted the binding of FAK and p53 in a dose-dependent manner, while the negative control compound (A18),[32] which did not bind the FAK-p53 region did not disrupt the binding of FAK and p53 (Figure 3E). Thus, R2 bound FAK-NT and directly disrupted the binding of FAK and p53 proteins in vitro and in vivo.

To study the effect of R2 compound on p53-dependent signaling, we tested the effect of R2 on p53-regulated transcriptional targets, such as p21, Mdm-2, and Bax. We have shown before that overexpression of FAK plasmid blocked the transcriptional activity of p53 through interaction with p53 protein [18]. To test if disruption of the FAK and p53 interaction by R2 de-repressed p53 transcriptional activity, we co-transfected HCT116 p53^-/^- cells with p53 plasmid and p21 promoter luciferase plasmid in the presence of R2 compound either without FAK plasmid or with the FAK plasmid. After 24 hours we added R2 at 25 μM and compared its effect with untreated cells. FAK blocked p53-induced p21 activity (Figure 4A), while treatment with R2 compound reversed this inhibition and re-activated p53-activity of the p21 target (Figure 4A). The same reactivation of p53 was demonstrated by R2 with Mdm-2 target (Figure 4B) and Bax target (Figure 4C). This effect was specific and not observed with the negative control compound M13, that targeted the FAK-MDM-2 interaction [33], but not the FAK and p53 interaction (Additional file 4: Figure S4). Thus, R2 specifically targeted FAK and p53 interaction and re-activated p53 targets: p21, Mdm-2, and Bax promoters.

To study the effect of R2 on p53 and p53-regulated targets, we performed Western blotting on HCT116 p53 treated cells with different doses of R2. We treated cells with different doses of R2 from 1 to 50 μM for 24 hours and performed expression analysis of p53 and its targets: p21, Mdm-2, and Bax (Figure 5A). R2 increased expression of p53 targets: p21, Mdm-2 and Bax in a dose-dependent manner in HCT116 cells (Figure 5A, left panel). In addition, we treated wild type p53 breast cancer MCF-7 cells with R2 (Figure 5A, right panel). R2 also increased p21 and Mdm-2 levels and at higher doses caused PARP-1 cleavage and caspase-8 activation in MCF-7 cells. In contrast to cancer cells with wild type p53, there was no up-regulation of Mdm-2 and p21 in SW620 colon cancer cells with mutant p53 (not shown). Thus, R2 increased the expression of p53 and its targets in a dose-dependent manner in cancer cells with wild type p53.

To detect the effect of R2 on p21 and p53 localization and activation, we performed immunostaining of p21 and p53 in HCT116p53⁺/⁺ and HCTp53^-/^- cells that were either untreated or were treated with R2. We detected activation of p21 and increased nuclear localization by immunostaining of p21 in R2-treated HCT116p53⁺/⁺ (Figure 5B, upper panel). The activation and nuclear localization of p21 was observed in HCT116p53⁺/⁺ cells, but not in p53-negative cells, indicating p53-dependent activation of p21 by R2. Increased nuclear localization of p53 was observed in HCT116p53⁺/⁺ cells treated with R2, but was not detected in the negative control HCT116p53^-/^- cells (Figure 5B, lower panel). Thus, R2 activated p53-targets in a p53-dependent manner.

We also performed cell cycle analysis of R2-treated and untreated HCT116 p53^-/^- and p53⁺/⁺ cells by FACS (Figure 5C). We treated HCT116 cells with 10, 20, and 100 μM of R2 for 24 hours and then performed analysis of the cell cycle. We detected a significant dose-dependent increase of G1-arrest in R2-treated HCT116 p53⁺/⁺ cells from 46% in untreated cells to 56% at 100 μM of R2 (p<0.05). We also observed a decrease of G2-phase in these cells from 16% in untreated to 6% in R2-treated, but not in HCT116 p53^-/^- cells (Figure 5C). Thus R2 activated the p53-target, p21, and increased G1 arrest in HCT116 cells in a p53-dependent manner.

To test the effect of R2 on tumor growth in vivo, we subcutaneously injected isogenic HCT116 p53⁺/⁺ and HCTp53^-/^- cells in the same mice into their right and left sides, respectively, and then treated them, with R2 and measured xenograft tumor growth (Figure 6A, upper panels). R2 significantly decreased tumor volume in HCT116 p53 ⁺/⁺ mice xenografts (Figure 6A, left upper panel), while it did not significantly decrease tumor growth in HCTp53^-/^- xenografts (Figure 6A, right upper panel). We analyzed tumors from HCT116 p53⁺/⁺ xenografts and detected up-regulated expression of p21 in a p53-dependent manner: R2 increased p21 in HCT116 p53⁺/⁺ xenografts but not in p53^-/^- xenografts, while it did not affect FAK and p53 protein levels (Figure 6A, lower panels). We also observed activation of caspase-3 in HCT116 p53⁺/⁺, but not in p53^-/^- xenografts, consistent with a significant decrease of HCT116 p53⁺/⁺ xenograft tumor growth. Western blotting demonstrated increased p21 in HCT116 p53⁺/⁺ xenograft tumors but not in p53^-/^- xenograft tumors (not shown). In addition, we performed immunoprecipitation of p53 and FAK in untreated and R2-treated HCT116p53⁺/⁺ xenografts and detected disruption of FAK and p53 complex in the HCT116 p53⁺/⁺ xenografts (Figure 6B). Thus, R2 blocked tumor growth, disrupted FAK and p53 and re-activated p53 by up-regulating p21 in HCT116 p53⁺/⁺ xenografts in vivo. Furthermore, the p53 specificity of R2 was confirmed with the lack of effect in the control p53 negative xenografts in each animal.

To test the effect of R2 on cancer cell viability in combination with chemotherapy, we treated HCT116 p53⁺/⁺ and HCT53^-/^- cells with R2 alone, doxorubicin alone, or with R2 and doxorubicin together (Figure 7A). R2 sensitized HCT116 p53⁺/⁺ cells to doxorubicin (Figure 7A, upper panel) but not HCT116 p53^-/^- cells (Figure 7A, lower panel). Western blotting detected increased p53, p21 and Mdm-2 expression in the case of a combination of R2 and doxorubicin compared with each agent alone in HCT116p53⁺/⁺ cells (Figure 7B, left), but not in HCT116p53^-/^- cells (Figure 7B, right). Thus, R2 sensitized colon cancer cells to doxorubicin in a p53-dependent manner. The same sensitizing effect as in the case of R2 and doxorubicin was observed in the combination of R2 and 5-fluorouracil, where FACS analysis demonstrated significantly increased apoptosis in the case of the combination of R2 and 5-fluorouracil in HCT116p53⁺/⁺ cells, but not in p53^-/^- cells (Figure 7C). Thus, R2 sensitized cancer cells to different chemotherapy drugs, which can be important for developing FAK-p53 combination therapy approaches.