Live Cell, Image-Based High-throughput Screen to Quantitate p53 Stabilization and Viability in Human Papillomavirus Positive Cancer Cells

Abstract

Approximately 5% of cancers are caused by high-risk human papillomaviruses. Although very effective preventive vaccines will reduce this cancer burden significantly over the next several decades, they have no therapeutic effect for those already infected and remaining at risk for malignant progression of hrHPV lesions. HPV-associated cancers are dependent upon the expression of the viral E6 and E7 oncogenes. The oncogenic function of hrHPV E6 relies partially on its ability to induce p53 degradation. Since p53 is generally wildtype in hrHPV-associated cancers, p53 stabilization arrests proliferation, induces apoptosis and/or results in senescence. Here we describe a live cell, image-based high-throughput screen to identify compounds that stabilize p53 and/or affect viability in HPV-positive cancer HeLa cells. We validate the robustness and potential of this screening assay by assessing the activities of approximately 6,500 known bioactive compounds, illustrating its capability to function as a platform to identify novel therapeutics for hrHPV.

Introduction:

Papillomaviruses are small non-enveloped DNA viruses that generally cause benign proliferative squamous epithelial lesions. There are over 200 different human papillomaviruses (HPVs) that are grouped by their genome sequence similarities into 5 distinct genera (1). A subset of the alpha genus HPVs associated with human cervical cancer, other anogenital cancers and oropharyngeal cancer, referred to as the high-risk human papillomaviruses (hrHPVs), are responsible for 5% of human cancers, most notably cervical cancer. The development of a preventive vaccine for these hrHPVs using virus-like-particles (VLPs) consisting of the major L1 capsid protein has been a major advance in cancer prevention, but the VLP vaccines have no therapeutic value (2). Thus, there is a major unmet need for the development of targeted therapeutic options for HPV-associated cancers as well as HPV-associated precancerous lesions.

There are several viral proteins and viral-host protein interactions that could serve as potential targets for hrHPV-specific antiviral therapies, including the viral E1 helicase, viral E2 regulatory protein, as well as the E6 and E7 oncoproteins. HPV-associated cancers are dependent upon the continued expression of E6 and E7 to target cellular host proteins and pathways. Among the targets of E6 and E7 are the p53 tumor suppressor protein and retinoblastoma tumor suppressor protein pRB105 (1). The oncogenic function of hrHPV E6 is tied in part to its induction of p53 proteolysis via hijacking the cellular ubiquitin ligase E6-associated protein (E6AP) to form a complex that ubiquitylates p53 (3). Inhibition of E6 or E6AP in HPV-positive cells leads to p53 stabilization and subsequent apoptosis due to E7-induced oncogenic stress (4). Since E6 effectively degrades p53 in HPV-infected cells, TP53 remains wildtype, as there is no selective pressure for mutation. Thus, restoration of wildtype p53 expression in HPV-positive cancer cells induces the expression of p53 transcriptional targets leading to cell cycle arrest, apoptosis and senescence (5–8).

Our laboratory has developed several cell lines that enable quantification of p53 protein levels in HPV-positive cervical cancer cells in an automated, high-throughput manner. In this manuscript, we describe a quantitative, live cell assay employing the hrHPV18 positive HeLa cervical cancer cell line stably expressing a fluorescent p53 reporter construct that is a substrate of the E6/E6AP ubiquitin-ligase activity. There have been a number of in vitro screens for inhibitors of E6/E6AP binding that have validated that in vitro, compounds that selectively inhibit E6 binding to E6AP, interfere with p53 proteolysis (9–13). The high-throughput assay described here has several advantages over those previously conducted. These include live cell imaging that enables time course analysis, a low per well cost as a consequence of minimal reagent use, the quantitative and flexible nature of this high content assay, and rapid image acquisition/analysis.

Materials and methods

Cell lines and culture conditions

The reporter cell line HeLa RIG3 p53(R273C) (HeLa/mRuby-p53) has previously been described (14). Briefly, HeLa cells were transduced with the lentiviral vector PHAGE-N CMVt N-RIG3 p53(R273C) expressing a bicistronic mRNA encoding the fusion proteins mRuby-p53(R273C) (referred to here as mRuby-p53) and H2B-SGFP2 (p7709) (14). Following transduction, cells were selected using neomycin and fluorescent activated cell sorting to acquire cells with the strongest expression of H2B-SGFP2 and no red fluorescence, indicative of efficient E6/E6AP degradation of mRuby-p53.

All cells used in this study were grown in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10% fetal bovine serum and maintained at 37°C in a 5% CO₂ atmosphere.

Transfections and siRNAs

siRNA reverse transfections were performed using Lipofectamine RNAiMAX (Thermo Fisher Scientific) according to manufacturer’s instructions. In all cases, the final siRNA concentration was 20 nM. After 72 h incubation, cells were harvested for further analysis. The siRNA duplexes used in these studies were: siE6AP2 (5’-CCAGAUUGCUCUCUAAUGA-3’, Dharmacon D-005137–02); siE6AP4 (5’-GCAGUUGAAUCCAUAUUUG −3’, D-005137–04); C911 version siE6AP2C (5′-CCAGAUUGGAGUCUAAUGA-UU-3′, custom synthesis from Dharmacon to include same chemical modifications); C911 version siE6AP4C 5′-GCAGUUGAUAGCAUAUUUG-UU-3′, custom synthesis from Dharmacon to include same chemical modifications) (15).

Western blots

At time of harvest, cells were washed with 1X PBS and lysed directly in wells using SDS lysis buffer (62.5 mM Tris-HCl, pH 6.8, 2% SDS). Lysates were harvested and sonicated, and protein concentration determined using the Pierce^™ BCA Protein Assay Kit (Thermo Fisher, 23225). 10 μg of lysates were separated on 10–20% NuPage BisTris midi gels using 1X MES buffer (Thermo Fisher Scientific) and transferred to polyvinylidene difluoride (PVDF) membranes using 1X transfer buffer (25 mM Tris-HCl pH 7.6, 192 mM glycine, 10% methanol). After blocking in 5% nonfat dried milk in TBST (Tris-buffered saline [pH 7.4], 0.05% Tween 20), membranes were incubated with the following primary antibodies: mouse monoclonal anti-β-Actin-HRP (1:50000; Millipore Sigma, A3854); goat polyclonal anti-p53-HRP (1:3000; R&D Systems, HAF1355); mouse monoclonal anti-HPV18E6 (G-7) (1:500; Santa Cruz Biotechnology sc-365089); mouse monoclonal anti-HPV18E7 (F-7) (1:500; Santa Cruz Biotechnology sc-365035); and mouse monoclonal anti-E6AP (E-4) (1:1000; Santa Cruz Biotechnology, sc-166689). For detection of the anti-E6AP antibody, blots were incubated with a secondary anti-mouse-HRP antibody (1:1000; Kindle Biosciences, R1005). All membranes were washed in TBST after each antibody incubation. Blots were developed using KwikQuant Ultra Digital-ECLTM Substrate Solution and images were acquired using KwikQuantTM Imager (Kindle Bioscience).

Dose response experiments

HeLa mRuby-p53 cells were plated in 384-well plates (Corning #3764) using a Multidrop^™ Combi Reagent Dispenser (Thermo Fisher Scientific, MA, USA), at a density of 750 cells and a final volume of 30 μL/well in culture medium. Plates were spun for 1 minute at 1000 rpm and incubated overnight at 37°C, 5% CO₂. The following day, DMSO-dissolved compounds were dispensed using a Hewlett Packard (HP) D300e digital dispenser. The compounds were tested across a 12-point titrations series, with final concentrations varying by compound within the range of 1 nM to 50 μM and DMSO concentration kept at 0.5% for all wells. Velcade (300 nM) and DMSO (0.5%) were used as positive and negative controls respectively. All compounds were tested in biological triplicate, each with technical triplicates. Analysis was conducted as described below for the HTS assay, but the values obtained were not normalized. Chemicals used in this assay: DMSO (Sigma, D8418); actinomycin D (Sigma, A1410); Velcade (EMD Millipore Corp, 5.04314.0001); roscovitine (MedChemExpress, HY-30237); RITA (MedChemExpress, HY-13424); curcumin (MedChemExpress, HY-N0005), and ellagic acid (MedChemExpress, HY-B0183).

High-throughput screen (HTS)

HeLa mRuby-p53 cells were plated in 384-well plates (Corning #3764) using a Multidrop^™ Combi Reagent Dispenser, at a density of 750 cells and a final volume of 30 μL/well in culture medium. Plates were spun for 45 seconds at 500 rpm and incubated overnight at 37°C, 5% CO₂. The following day, 33 nL of DMSO-dissolved compound was transferred to assay wells using a custom pin-transfer workstation at the ICCB-Longwood Screening Facility (Harvard Medical School). Approximately 6,000 compounds from the Known Bioactives Collection at ICCB-Longwood were screened at final concentrations ranging between 0.055 nM to 11 μM. Velcade (Adipogen Life sciences, AG-CR13602-M005) at a concentration of 330 nM served as the positive control, while 0.1% DMSO (Sigma, D8418) was the negative control. The screen was conducted in duplicate, with replicate wells on distinct assay plates. Following compound addition, cells were incubated in a humidified incubator at 37°C, 5% CO₂.

At 24 h and 48 h post-compound addition, total cell number and mRuby-p53 levels were quantitated using an Acumen® ex3 Laser Scanning Imaging Cytometer (TTP Labtech). A 488 nm laser and 561 nm laser (both set at 6 mW output, 600 voltage gain, and sensitivity threshold at 1) were used for the excitation of SGFP2 and mRuby respectively. Scans were run using a resolution of 1 × 7 μm, with a scanned area of 2.3 × 1.694 mm/well to avoid the periphery of the well. Nuclei were defined using a size filter of 5 to 50 μm.

Data analysis

Total nuclei (nuclei showing H2B-SGFP2 expression) and percentage of nuclei showing mRuby-p53 stabilization were calculated for all wells at 24 h and 48 h using the Cellista software (TTP Labtech). The total nuclei and percent p53 positive nuclei in each well were normalized on a per plate basis to the average of the positive and negative control wells. This was done in order to compare results throughout the entire screen. The following were calculated for each well:

$$p53(\%)=\frac{TNred}{TN\mathit{\text{green}}}$$

$$p53(\%)n=\frac{p53(\%)-Avp53(\%)NC}{Avp53(\%)PC-Avp53(\%)NC}\times 100$$

$$TN(\%)n=\frac{TN\mathit{\text{green}}}{AvTN\mathit{\text{green}}NC}\times 100$$

Where “p53(%)” is the percentage of nuclei showing stabilization of mRuby-p53 in a well, “TN red” is the total number of nuclei showing p53 stabilization (red fluorescence) in a well, “TN green” is the total number of nuclei showing H2B-SGFP2 expression (green fluorescence) in a well, “p53(%)n” is the percentage of nuclei showing stabilization of mRuby-p53 in a well normalized to the plate positive and negative controls, “Av p53(%) NC” is the average of the p53(%) values obtained from all the negative control wells in the plate, “Av p53(%) PC” is the average of the p53(%) values obtained from all the positive control wells in the plate, “Av TN green NC” is the average of the TN green values obtained from all the negative controls in the plate, and “TN(%)n” is the number of nuclei observed in a well expressed as a percentage of the Av TN green NC value in the corresponding plate.

After normalization, the values of replicate experimental wells were averaged to obtain a single value/well. A compound was considered a potential hit in this screen when the corresponding p53(%)n ≥ 50% for either or both time points. Those compounds were selected for confirmation. Since HeLa cells are mononucleated, the number of nuclei corresponds very well with the number of cells, therefore we used the TN(%)n value to compare the effect of different compounds on cellular viability/proliferation relative to the negative controls.

Validation and titration experiments

Cells were seeded into 384-well plates using the same conditions described above for the HTS assay. Approximately 24 h after seeding, compounds and DMSO were dispensed into assay wells using a HP D300e digital dispenser. The compounds were tested across a 15-point titrations series, with final concentrations ranging from 5 nM to 20 μM and DMSO concentration kept at 0.2% for all wells. Velcade (330 nM) and DMSO were used as positive and negative controls respectively. All compounds were tested in technical triplicates. Analysis was conducted as described above for the HTS assay, but the values obtained were not normalized.

Software

Figures were assembled using Affinity Photo and Affinity Designer (Affinity). Graphs in Figure 3 were done in Excel (Microsoft). Graphs in Figs 2, 4, and 5 were prepared using Prism 9 (GraphPad).

Figure 3. Overview of HTS results.

Scatterplots displaying the average values of replicate wells for negative controls (DMSO) in blue, positive controls (Velcade) in red and experimental wells in gray. The x-axis represents plate and well location while the y-axis displays the percentage of p53 positive cells relative to the controls in the corresponding plate at 24 (A) and 48 h (B) or the total nuclei relative to the negative controls in the same plate at 24 (C) and 48 h (D).

Figure 2. Response of HeLa/mRuby-p53 reporter cells to known inducers of p53 accumulation in HeLa cells.

Five compounds documented to induce p53 in HeLa cells were assayed at 12 different concentrations (0.0009, 0.0013, 0.0040, 0.0067, 0.0120, 0.0227, 0.0426, 0.0813, 0.1519, 0.2932, 0.5331, and 1.0128 μM for Velcade and actinomycin D; 1.002, 1.459, 1.990, 2.919, 4.113, 5.970, 8.491, 12.073, 17.247, 24.543, 35.024, and 50.015 μM for roscovitine, RITA, curcumin, and ellagic acid; and 1.002, 1.327, 1.857, 2.521, 3.449, 4.643, 6.368, 8.756, 11.807, 16.185, 22.023, and 29.983 μM for curcumin in combination with ellagic acid), 24h and 48 h after addition of the corresponding compounds for their effect on the percentage of cells showing increased expression of mRuby-p53 levels (p53(%)) and the total number of cells in the well (TN green). Each experiment was performed in triplicate, with each replicate on a separate assay plate. Symbols indicate the average of three independent experiments and the error bars indicate standard deviations. X axes are in log scale.

Figure 4. Effect of HDAC, CDK, Topoisomerase s and Aurora kinase inhibitors on HeLa/mRuby-p53 cells at 48 h.

Effect of HDAC inhibitors (A), CDK inhibitors (B), topoisomerase inhibitors (C), and Aurora kinase inhibitors (D) on p53(%)n and TN(%)n in the HTS at 48 h post-treatment. All Aurora kinase inhibitors are displayed as grey circles. For clarity, in panels A and B, only compounds that produced values of p53(%)n > 25% under at least one of the conditions used in the screen are displayed. Individual compounds may be represented by multiple points because they were assayed at different concentrations and/or represented in multiple libraries. The CDK inhibitor roscovitine is also displayed in panel B because of its known effect on p53 protein levels.

Figure 5. Confirmation/titration experiments of selected potential hits from HTS after 48 h incubation.

Twelve potential hits from the primary screen were assayed at 15 different concentrations after 48 h incubation. Each compound is represented in a separate graph. The percentage of cells with increased levels of p53 (p53(%)) are indicated by red inverted triangles and effect on total cell number (TN green) are shown by blue circles. Experiments were performed in HeLa/mRuby-p53 cells and conducted in triplicate, with each replicate on a separate assay plate. Error bars indicate standard deviations. The vertical green line indicates the compound concentration in the HTS that was considered a potential positive (p53(%)n ≥ 50). If more than one concentration reached that value, the green line signals the lowest of those concentrations. The two horizontal gray lines are placed at p53(%) values of 0 and 100.

Results

Cell line development and reporter system validation

To quantitate p53 protein levels in vivo in a robust and high-throughput manner over time, we engineered HPV18-positive HeLa cells to express a p53 fluorescent reporter. HeLa cells were transduced with a vector expressing a bicistronic mRNA that encodes a transcriptionally inactive mutant p53 (R273C) fused to the monomeric red fluorescent protein mRuby (mRuby-p53) as well as histone 2B (H2BC11) fused to the green fluorescent protein SGFP2 (H2B-SGFP2) (Figure 1A) (14). We chose to utilize the R273C transcriptionally inactive p53 mutant because it acts as a dominant negative by being incorporated into a p53 tetramer also containing endogenous wildtype (wt) p53, precluding its binding to DNA and therefore interfering with cell cycle arrest and cell death as a consequence of p53 transcriptional activity (16). This mutation might also cause a deficiency in p53 mitochondrial pro-apoptotic activities. A similar p53 mutation, R273H, as well as other mutations in the p53 DNA binding domain, have been shown to be deficient in p53 mitochondrial pro-apoptotic activities by failing to bind to Bcl2 and induce outer mitochondrial membrane permeabilization (17). This dominant negative effect allows the assay to be conducted over time, attenuating the deleterious effects of wt p53 accumulation. In these reporter cells, nuclei fluoresce green as a consequence of H2B-SGFP2 expression. Since the majority of the HeLa cells are mononucleated, this enables straightforward and rapid quantitation of cell number without the need of an additional staining step (e.g. Hoechst dye for DNA stain). Under normal conditions, there is minimal red fluorescence because of the efficient degradation of mRuby-p53 by E6/E6AP. When p53 is stabilized, the nuclei remain green as a consequence of H2B-SGFP2 expression and there is also accumulation of red fluorescence from mRuby-p53, primarily in the nuclei. If cell viability or proliferation is compromised, a decrease in the number of green fluorescent nuclei will be observed relative to negative control samples.

Figure 1. Cell-based reporter system to quantitate p53 stabilization in HPV positive cells.

A. Schematic diagram of the reporter cassette used in the assay. The arrows on top represent the ORFs of the reporter proteins. CMVp: CMV promoter, IRES: encephalomyocarditis virus internal ribosomal entry site, H2B: histone 2B (H2BC11), SG linker: serine –glycine linker. B-C. Western blots showing the effect of siRNA-mediated knock down of E6AP (B) or proteasomal inhibition by Velcade (C) on the protein levels of E6AP, p53 and mRuby-p53 in HeLa and HeLa/mRuby-p53. D. Stabilization of mRuby-p53 by siRNA-mediated knock down of E6AP or proteasome inhibition visualized by fluorescent microscopy in the HeLa/mRuby-p53 cells. The suffix C in the name of a siRNA indicates it is the C911 variant of the corresponding siRNA.

To validate the ability of mRuby-p53 to function as a reporter for p53 stabilization in the HeLa/mRuby-p53 reporter cells, these and wt HeLa cells were transfected with two different siRNAs targeting E6AP (siE6AP) and their respective C911 negative controls (15). As shown in Figure 1, both endogenous p53 and mRuby-p53 accumulated when E6AP was knocked down by either siE6AP. This was observed via western blot (Figure 1B) and immunofluorescence (Figure 1D). This stabilization was specific to the knock down of E6AP, as no p53 accumulation was observed in non-transfected cells or when cells were transfected with the respective C911 negative control siRNA duplexes. An increase in p53 was also observed when cells were treated with the proteasome inhibitor Velcade (Figure 1C and 1D). These results confirmed that wt p53 and mRuby-p53 are similarly regulated by E6/E6AP-mediated ubiquitylation and proteasomal degradation.

To further characterize the behavior of the HeLa /mRuby-p53 reporter cells, they were treated with several compounds, namely Velcade (18), actinomycin D (19), roscovitine (20, 21), RITA (22, 23), and curcumin (24), that have been reported to stabilize endogenous p53 in HeLa cells. For this experiment, cells were seeded in 384-well plates and incubated for 24 h prior to compound addition. The number of green and red fluorescent nuclei in each well were quantitated 24 h and 48 h post-compound addition using a laser scanning cytometer. To estimate the impact of each compound on cell viability/proliferation, the total nuclei (TN) green value obtained for each individual well was normalized to the average TN green value of six wells localized in the same row of that plate containing cells treated only with DMSO, which was considered 100%. The results of this experiment are shown in Figure 2 and Table S1. As expected, Velcade stabilized mRuby-p53 at sub micromolar concentrations. Since 293 nM was the lowest concentration tested that produced a maximum effect on mRuby-p53 accumulation at 24 h and 48 h, we chose a similar concentration, 330 nM, to be used as the positive control during the screen. Actinomycin D, roscovitine, and curcumin produced accumulation of mRuby-p53 at similar concentrations to those reported to stabilize endogenous p53 in HeLa cells (20, 21, 24). However, actinomycin D only induced mRuby-p53 accumulation after 48 h; and in the case of roscovitine, the increase in mRuby-p53 levels at 48 h was more than double the increase observed after 24 h, highlighting the importance of quantitating p53 accumulation over time. The IC50 obtained from this experiment for the effect of roscovitine on HeLa/mRuby-p53 cell viability/proliferation at 24 h is 24.34 μM, very similar to the 28 μM reported in the literature (20, 21). Interestingly, RITA had only a small effect at the highest concentrations used in this experiment. However, a similar observation was reported by Messa et al. when they described a similar assay performed in live cells using Renilla luciferase fused to p53 as reporter for p53 stability (25). Kumar et al. reported that curcumin and ellagic acid synergistically induced p53 accumulation and apoptosis in HeLa cells (26). Consistent with those observations, a synergistic inhibitory effect of curcumin and ellagic acid (which alone has no noticeable effect on either p53 accumulation nor on viability/proliferation in HeLa/mRuby-p53 cells) was observed at both 24 and 48 h. The IC50s of curcumin with 0 μM, 10 μM, and 20 μM of ellagic acid were 9.11 μM, 6.98 μM, and 7.35 μM at 24 h, and 10.09 μM, 7.54 μM, and 5.37 μM at 48 h respectively. We also observed synergy between curcumin and ellagic acid to induce mRuby-p53 accumulation. This was more evident at 48 h. The curcumin EC50s in the presence of 0 μM, 10 μM, and 20 μM of Ellagic acid were 19.71 μM, 15.91 μM, and 15.75 μM at 24 h, and 16.57 μM, 11.97 μM, and 9.52 μM at 48 h respectively.

Altogether, this data indicates that the mRuby-p53 reporter is a reliable indicator of the in vivo status of p53 protein levels in HeLa cells and provides accurate information about the effect of compounds on HeLa cell viability/proliferation.

Hight-throughput assay development and optimization

After validation of the HeLa/mRuby-p53 reporter cells, a high-throughput assay was developed and optimized for conducting a small molecule screen. Several components were examined to minimize variability and maximize the dynamic range. Cells were plated in each well of a 384-well plate the day prior to experimental compound treatment. Small molecules were then added to duplicate assay plates via a custom pin transfer workstation resulting in a final DMSO concentration of 0.1%. Following 24 h and 48 h incubation, the number of nuclei expressing H2B-SGFP2 and mRuby-p53 were quantitated using a laser scanning cytometer. Acquired images were simultaneously analyzed with STP Labtech’s Cellista software. An automated Z’ factor experiment, mimicking all steps in the screen including automation, was conducted utilizing DMSO (0.1% final concentration) as the negative control and Velcade (330 nM final concentration) as the positive control. A Z’ factor of 0.91 and 0.94 was obtained at 24 h and 48 h respectively, indicating assay robustness (27).

High-throughput small molecule screen to identify compounds that stabilize p53

A total of 11,805 experimental wells were screened across 10 different known bioactive small molecule libraries. Since some libraries were tested at multiple concentrations and there was compound overlap between libraries, a total of 6,449 unique compounds were tested. The data obtained were normalized on a per plate basis relative to the controls as described in the Materials and Methods. Experimental compounds were considered to be potential hits when more than 50% of cells scored as p53-positive in the average of the two replicates at either or both 24 h and 48 h post-treatment. Following this criterion, a total of 75 unique compounds were classified as potential hits, resulting in a hit rate of 1.15% (Table S2). All data from the screen, both raw and analyzed, have been deposited in PubChem BioAssay, AID: 1508601 (https://pubchem.ncbi.nlm.nih.gov/bioassay/1508601?viewcode=D732F17F-028C-4E57-9782-9E4AC2897DCA).

Throughout the screen, the assay had a large dynamic range, with clear separation between positive and negative controls (Figures 3A and 3B). This is reflected in the Z’ factors for p53 stabilization that were calculated for each assay plate. Throughout the entire screen, the Z’ factor was 0.869 +/− 0.037. Not surprisingly, the majority of experimental compounds had a minimal impact on p53 expression. This was consistent at both 24 h and 48 h post-treatment. As shown in Figures 3C and 3D, there was more variability in the overall cell number. This was expected, as the majority of known bioactive compounds have some impact on mammalian cell viability and/or proliferation. Treatment with Velcade led to ~40% decline in overall cell number at 24 h and ~60% at 48 h post-treatment. The variability in total cell number observed did not impact analysis of p53 stability, as the p53 percent positivity was not affected by the total number of cells in that well.

Compound effect increased with incubation time. This was most apparent when examining cell viability/proliferation. At 24 h post-treatment, a total of 10 compounds reduced the number of cells to less than 25% of the negative controls (Figures 3C and S1A). However, at 48 h post-treatment the number of compounds with this effect increased to 110 (Figures 3D and S1A). Lengthening compound treatment also impacted p53 stabilization. There were 48 compounds that resulted in more than 50% of cells showing p53 stabilization at 24 h, while this increased to 70 compounds after 48 h of treatment (Figure S1B). Of the 48 compounds that led to p53 stabilization at 24 h, 43 also displayed stabilization at 48 h. There were 5 compounds that only scored positive at 24 h and 27 compounds that met the criteria to be considered potential hits only after 48 h. We did not find a clear correlation between the compounds that only increased p53 stabilization at 24 h and their impact on cell viability or proliferation at 48 h, therefore we cannot attribute the decrease in p53 stabilization at 48 h to toxic effects of these compounds. While a mutant form of p53 was utilized in the screen to interfere with p53-mediated cell cycle arrest and apoptosis, compounds that increased the mRuby-p53 signal frequently decreased the total number of cells relative to the negative controls wells at both 24 h and 48 h (Figures S1C and S1D). However, there were numerous other compounds that markedly decreased the overall cell number but had no impact on the amount of mRuby-p53 detected (Figure S1C and S1D).

A very straightforward way to evaluate the performance of a high-throughput screen is to analyze the effect of compounds previously reported to have a specific effect on the phenotype evaluated in the assay. As anticipated, the proteasome inhibitors Velcade (bortezomib), carfilzomib, CEP-18770, epoxomicin, MG-132, MLN2238, MLN9708, ONX-0914, and oprozomib all scored as potential hits in the primary screen (Figures S2A and S2B, Table S3). Two proteasome inhibitors present in the screened libraries, LDN-57444 and PI-1840, failed to increase the levels of mRuby-p53. In the case of LDN-57444, this could be a consequence of the concentration assayed since this compound has been described to inhibit proteasome activity in human neuroblastoma SK-N-SH cells at concentrations of 25 μM and above, which is higher than the concentration of 3.67 μM tested in the screen (28).

Other compounds that have been shown to impact HPV-positive cancer cell viability are histone-deacetylase (HDAC) inhibitors and cyclin-dependent kinase (CDK) inhibitors. HDAC inhibitors have been reported to induce apoptosis in HeLa cells in a p53-independent manner (29). In addition, HDAC inhibitors have been shown to arrest HeLa cells in G1 to S transition, concomitant with upregulation of the CDK inhibitors p21 and p27 (30). Therefore, we analyzed the performance in the screen of 30 HDAC inhibitors and 31 CDK inhibitors (Table S3). As expected, several HDAC inhibitors decreased the number of cells; surprisingly they also increased mRuby-p53 levels, with 13 of them producing values of p53(%)n ≥ 50% at some of the concentrations tested. In addition, there seems to be a correlation between mRuby-p53 levels and total cell number. This is most evident at 48 h (Figures 4A, and S3A, Table S3). Several CDK inhibitors also decreased the number of cells, with a concomitant increase in levels of mRuby-p53 and five CDK inhibitors increased p53(%)n values ≥ 50%. However, for several CDK inhibitors (e.g. CGP80474, flavopiridol, and dinaciclib), at the same range of concentrations that similarly impacted the total cell numbers, their effect on mRuby-p53 stabilization was variable, ranging from a strong increase to little or no effect (Figures 4B and S3B, Table S3). This points to a lack of correlation between the two phenotypes for these compounds.

Topoisomerase inhibitors are well documented to increase p53 levels and activity, and induce apoptosis in HPV-positive cancer cells (31–34). From 13 topoisomerase inhibitors analyzed in the primary screen (Figures 4C and S3C, Table S3), four anthracyclines, daunorubicin, doxorubicin, idarubicin, and pyrromycin, stabilized mRuby-p53 with p53(%)n values higher than 50%, while the anthracycline-related compounds mitoxantrone and epirubicin, as well as camptothecin, produced lower increases in mRuby-p53 protein levels. This is consistent with the results from a previous study in which several chemotherapeutic drugs, including these compounds, were tested for their ability to reactivate p53 in HPV-positive cancer cells (31). All of the compounds mentioned above considerably affected the cell viability/proliferation (TN(%)n) of the HeLa/mRuby-p53 cells. In contrast, the Aurora kinase inhibitor alisertib has been shown to induce apoptosis in HPV-positive cancer cell lines, including HeLa cells, without stabilizing p53 (35). We analyzed the effect of 22 Aurora kinase inhibitors (Table S3), including alisertib, on p53(%)n and TN(%)n (Figures 4D and S3D). Consistent with what was reported for alisertib, these compounds greatly decreased the number of cells but had no significant effect on mRuby-p53 levels.

Altogether, these results demonstrate that this high-throughput assay system accurately detects compounds expected to stabilize p53 in HPV-positive cancer cells as well as compounds that impact the viability/proliferation independent of p53 protein levels.

Confirmation of potential hits

Based on the results of the primary screen, 74 potential hits were selected for confirmation experiments with the objective of determining the reproducibility of the responses observed in the primary screen. These experiments were conducted in 384-well microplate format utilizing the same reporter cell line as the primary screen. Each compound was assayed across 15 concentrations ranging from 5 nM to 20 μM in technical triplicates, as this would enable us to determine adequate concentrations and incubation times to use in subsequent validation experiments. Images were acquired 24 h and 48 h post compound addition and analyzed. Of these 74 compounds, 69 stabilized p53 in a range of concentrations consistent with the results in the primary screen, providing a 92% confirmation rate (Table S4). Figure 5 displays the results after 48 h incubation (similar results for 24 h incubation are shown in Figure S4) of the confirmation/titration experiments on p53 stabilization and cellular proliferation/viability in HeLa/mRuby-p53 cells for several compounds that were subsequently assayed to determine their impact on endogenous p53 levels in HeLa cells. Two proteasome inhibitors that scored as potential hits in the primary screen, Velcade (bortezomib), which also served as the positive control for the screen, and CEP-18770 (delanzomib), were also included. Treatment with auranofin, chaetocin, and obatoclax resulted in different titration-based response patterns for p53(%) compared to the other compounds tested. Surprisingly obatoclax produced p53(%) numbers that considerably exceeded 100%. In the case of auranofin and chaetocin, the percentage of cells displaying mRuby-p53 stabilization reached a plateau and then abruptly dropped at higher concentrations. This is consistent with what was observed in the primary screen, in which the auranofin and chaetocin approached 100% p53(%)n and then decreased below 50% at the higher concentrations of 11 μM for auranofin and 3.67 μM for chaetocin (Table S3). Even with the mutant p53 in HeLa/mRuby-p53 cells, for several compounds there appears to be a correlation between the concentrations that stabilized mRuby-p53 and those that affected cell proliferation/viability. Overall, there was a strong correlation between the data obtained in the primary screen and the confirmation experiments (Figures 5 and S4).

Validation with endogenous p53

Based on the results of the primary screen and confirmation experiments, we selected eight compounds to assess whether they also stabilized endogenous p53 in HeLa cells. The selected compounds included two topoisomerase II inhibitors, doxorubicin (Adriamycin) and idarubicin, which are anthracyclines widely used in the treatment of cancer (36). Both compounds were identified as strong positives in the primary screen (Figures 4B and S3B), and were expected to activate p53 in HPV-positive cancer cells (31). To our knowledge, the other six selected compounds, chaetocin, celastrol, auranofin, CCF642, 18650, and obatoclax, have not been previously associated with increased p53 levels in HPV-positive cancer cells. Since 13 HDAC inhibitors were confirmed hits in the screen (Figures 4C and S3C, Tables S2 and S3), they were also included in these experiments to determine if they stabilized endogenous p53 in HeLa cells. HeLa cells were treated with the selected compounds for 24 h and cell lysates were harvested. Endogenous p53 and actin protein levels were visualized by western blot (Figures 6A and 6B). All compounds were tested at three concentrations based on the results of the titration experiments in the HeLa/mRuby-p53 reporter cells for 24 h (Figure S4). As expected, both doxorubicin and idarubicin led to an increase in endogenous p53 protein levels. When comparing the intensity of the p53 bands to the corresponding actin loading controls, it was apparent CCF642, 18650, obatoclax, and the HDAC inhibitors failed to stabilize endogenous p53 in HeLa cells and therefore were considered false positives. In contrast, auranofin, celastrol, and chaetocin stabilized endogenous p53 protein in HeLa cells. The compound concentrations that stabilized endogenous p53 in Hela cells were consistent with those that resulted in stabilization of mRuby-p53 in HeLa/mRuby-p53 cells, providing additional evidence that this reporter system is suitable as an indicator of p53 protein levels in vivo.

Figure 6. Effect of compounds on p53 stabilization in HeLa cells.

A. Eight potential hits from the screen were tested for their ability to stabilize p53 in wt HeLa cells by western blotting. B. Western blot showing the effect of several HDAC inhibitors on p53 levels in HeLa cells. C. Western blot showing the effect of the selected compounds on p53 and actin protein levels in different cell lines. Drug concentrations are shown in μM. kDa: kilodaltons, NT: no treatment, NC: negative control (DMSO). Western blots were done after 24 h incubation with the corresponding compounds.

To investigate if the selected compounds can also stabilize p53 in other hrHPV-positive cancer cell lines, SiHa (HPV16), CaSki (HPV16), and ME-180 (HPV68) cells were treated with the selected compounds for 24 h at the concentrations previously used for HeLa cells and their effect on p53 and actin protein levels quantitated by western blot analysis (Figure 6C). The HPV-negative human U2OS cells, an osteosarcoma-derived cell line expressing wt p53 under the control of Mdm2, was also included in this experiment to address the question of whether the stabilization of p53 by these compounds is dependent on the expression of hrHPV oncoproteins. Auranofin stabilized p53 only in hrHPV18 and 16 positive cancer cell lines at the conditions used in this experiment, displaying some HPV type specificity. In contrast, treatment with celastrol, chaetocin, doxorubicin, and idarubicin induced p53 accumulation in all cell lines tested, indicating that the effect of these compounds on p53 protein levels is not restricted to hrHPV-infected cells. It was not possible to obtain data from ME-180 cells incubated with 5 μM celastrol because this treatment was 100% lethal.

The selected compounds affect the expression of E6 and E7 in HeLa cells

We next asked whether the selected compounds affected expression of the E6 and E7 oncoproteins or E6AP in HeLa cells. To answer this question, cells were incubated with the compounds for 24 h, cell lysates harvested and protein levels visualized by western blot (Figure 7). All compounds decreased E6AP protein levels. It is unclear whether this change in expression would contribute to p53 stabilization, since low levels of E6AP can still efficiently target p53 for degradation in HeLa cells (14), and there is no apparent negative correlation between the protein levels of E6AP and p53. Auranofin, chaetocin, doxorubicin, and idarubicin all greatly reduced E6 protein levels. This was particularly evident with the latter two compounds that decreased E6 levels below the level of detection. These compounds also affected E7 expression, suggesting their effect may be prior to or during transcription since both proteins are generated by alternative splicing of mRNAs controlled by the same promoter. Surprisingly, celastrol treatment increased the expression levels of E6 and E7. This indicates that the effect of celastrol on E6 and E7, and perhaps its stabilization of p53, is via a different mechanism(s) when compared to the other compounds studied in this experiment. However, the finding that all five compounds affected E6 and E7 oncoprotein expression suggests this could be important for the stabilization of p53 induced by these compounds in HeLa cells.

Figure 7. Select compounds affect E6 and E7 expression in HeLa cells.

Western blot showing the effect of the indicated compounds on E6, E7, E6AP, p53 and actin protein levels in HeLa cells. kDa: kilodaltons, NT: non treatment, NC: negative control (DMSO). Western blots were done after 24 h incubation with the corresponding compounds.

Discussion

Although currently approved HPV VLP vaccines are extraordinarily effective in preventing infection by hrHPV types, there are no targeted therapies available for those already infected with hrHPV and therefore at risk for HPV-associated cancers. A therapy that targeted E6/E6AP-mediated degradation of p53, which is exclusive to hrHPV-positive cells, would have potential clinical value, since reactivation of p53 in HPV-infected cells results in senescence and/or apoptosis. A compound that interferes with E6 binding to E6AP or the binding of E6/E6AP to p53 would lead to the stabilization of p53 only in hrHPV-positive cells and likely display less toxicity than general inducers of p53 such as DNA damaging agents (37–39). Furthermore, new compounds that specifically decrease the proliferation/viability of hrHPV-associated cancer cells, even if partially or completely independent of p53, could be a valuable addition to currently used chemotherapeutic agents that non-specifically target rapidly proliferating cells. Therefore, a high-throughput assay that simultaneously measures p53 stabilization and cell viability/proliferation could enable the rapid screening of tens of thousands of compounds to identify novel potential therapeutic compounds for hrHPV-associated cancers.

While a number of publications have described strategies to identify inhibitors of E6/E6AP binding, most focus on in vitro strategies in cell-free systems (11–13, 40), including in silico screening, enzyme-linked immunosorbent assays, and pull-down assays. Cell-based luminescent assays have also been developed to quantitate E6-mediated degradation of p53 using a fusion reporter linking p53 to renilla luciferase (25, 41). While luciferase reporters are frequently used in screening assays because of their sensitivity, large dynamic range, and ease of analysis, they are not ideal because of the high reagent cost, sensitivity to edge effects and temperature gradients, and the number of compounds that interfere with luciferase activity (42–44). Such false positives can easily be identified in validation experiments.

Here, we describe a live cell, image-based high-throughput screen that simultaneously quantitates p53 stabilization in hrHPV-positive cancer cells and total cell number. This strategy has several advantages over those previously described. Because no fixation or staining are required, p53 stabilization and cell number can be monitored over time. This allows for the identification of compounds that have a more direct and rapid effect on p53 levels, as well as those that impact p53 protein levels indirectly and/or after longer periods of time. Since the percentage of cells expressing mRuby-p53 is quantitated rather than just absolute numbers, this parameter is not impacted by the number of cells in the well. By examining the same wells across time, inherent well-to-well variability is eliminated. The fact that no manipulation (e.g. fixing, staining) is required for assay readout also saves time, reduces cost and decreases the variability intrinsic to each step of an assay.

We chose to establish the assay in HeLa cells, because HPV18 is one of the most prevalent HPV types linked to HPV-associated cervical cancer. Conducting the screen in HeLa cells also enables the effect of compounds on the viability/proliferation of hrHPV positive cancer cells to be quantitated. Using hrHPV-infected cells derived from cervical cancer allows the screen to be performed in a cellular environment most similar to that in which the E6-mediated degradation of p53 takes place in HPV-associated cancers. This permits the identification of compounds that not just disrupt binding of E6 to E6AP or E6/E6AP binding to p53, but may lead to p53 stabilization more indirectly, for example affecting the activity of E6AP and/or the viral oncogene promoter. The expression of H2B-SGFP2 from the bicistronic reporter mRNA enables rapid quantitation of the number of cells in each well, providing proliferation/viability data for each compound and time point. This allows us to simultaneously identify compounds that stabilize p53 and/or affect viability/proliferation of HPV-associated cancer cells. Since p53 stabilization in hrHPV-positive cells leads to apoptosis and/or senescence, we utilized a reporter construct that expressed the mutant p53(R273C) protein that is susceptible to E6/E6AP-mediated proteasomal degradation, but unable to bind DNA. This mutant functions in a dominant negative manner by blocking the transcriptional activity of wt p53 (reviewed in (45, 46)) and possibly affecting the p53 pro-apoptotic mitochondrial pathways (17). Under conditions leading to p53 stabilization in HeLa/mRuby-p53 cells, both wt p53 and mRuby-p53 are expressed. Expression of the mutant interferes with the deleterious consequences of wt p53 stabilization in hrHPV-positive cells, enabling the assay to be performed over a longer period of time. Since the dominant negative effects of the mutant p53(R273C) may be less efficient in blocking the cell cycle arrest activity of wt p53 (47), stabilization of p53 and mRuby-p53 was still be expected to affect the cell number. Not surprisingly, we observed a correlation between the effect of the compounds on p53 stability and cellular proliferation/viability during the primary screen as well as in the confirmation/titration experiments.

As with any high-throughput screening approach, potential hits can be false positives. In this assay, compounds with intrinsic fluorescence in the image acquisition channel could be scored as false positives. In addition, since there are no wash steps in this assay, non-internalized and aggregated fluorescent compounds could be visualized. The majority of such compounds were excluded during the image analysis step via the size criteria established for object counting. However, if a fluorescent compound aggregates to a size within the object criteria or stain the nuclei of the cells, they will pass the size filter and score as a potential positive. A possible example of this is obatoclax. Treatment with obatoclax led to high levels of p53 stabilization during the primary and secondary screens, likely due to its intrinsic fluorescence (48). However, such false positives are quickly identified in validation experiments when they fail to stabilize the endogenous p53 in HeLa cells, as observed with obatoclax in the western blot experiments in Figure 6. False positives may also occur if a compound impacts promoter activity of the reporter. Since all HDAC inhibitors identified as potential hits in this screen failed to increase endogenous p53 protein levels in HeLa cells, we hypothesize that the HDAC inhibitors led to mRuby-p53 stabilization by increasing the activity of the CMV promoter driving reporter cassette transcription (49, 50), and consequently overriding the capacity of E6/E6AP to target mRuby-p53 for proteasomal degradation. Despite not affecting endogenous p53 protein levels, these HDAC inhibitors considerably decreased the number of cells in the treated wells. This is consistent with the report that HDAC inhibitors induce apoptosis in HeLa cells in a p53-independent manner (29). In addition, HDAC inhibitors have also been shown to arrest HeLa cells in G1 to S transition, concomitant with upregulation of cyclin-dependent kinase inhibitors (30) and have been proposed as potential therapeutics for the treatment of cervical cancer, either alone or in combination with proteasome inhibitors (30, 51). Interestingly, we found several CDK inhibitors, including CGP60474, R547, and riviciclib, that strongly decreased cell viability/proliferation and stabilized mRuby-p53 in the HeLa/mRuby-p53 cells. However, we hypothesize that the two effects are independent because cell viability/proliferation was similarly impacted at concentrations that had no effect on the levels of mRuby-p53. Other CDK inhibitors, for example AT7519 and abemaciclib, greatly decreased cell number while having less of an impact on p53 protein levels. Taken together, these data demonstrate that this reporter system has the ability to identify compounds that impact multiple aspects of cervical cancer cells, including those that affect cellular viability/proliferation independent of p53 stabilization.

Using this system, we screened several libraries composed of small molecules with known biological activities/targets and identified 75 compounds capable of stabilizing mRuby-p53 in the reporter HeLa cells. Most of the potential hits confirmed their ability to increase mRuby-p53 levels when assayed in the confirmation/titration assays. Included in the positive hits were several compounds expected to stabilize mRuby-p53, including proteosome inhibitors, progesterone (52, 53), and some topoisomerase inhibitors (31–34). Interestingly, roscovitine, curcumin, and cctinomycin D did not score as hits for p53 stabilization in the primary screen. In the case of roscovitine and curcumin this was due to the fact that both compounds induce p53 accumulation in HeLa cells at concentrations greater than 11 μM, the highest concentration used in this screen. These results are consistent with previous reports and our data from the dose-response experiments used to characterize the reporter system. The data obtained for actinomycin D in the primary screen corresponded to what it was observed in the dose-response experiment, with p53 stabilization occurring at submicromolar concentrations after 48 h. However, since the stabilization of p53 was less than 50% of the observed for the positive controls, it did not pass the cutoff to be considered a hit. Of the compounds assayed for their impact on endogenous p53, five were shown to stabilize endogenous p53 in HeLa cells by western blot. Two of them, doxorubicin and idarubicin, are both anthracyclines. These molecules intercalate between DNA bases and inhibit Topoisomerase II, producing double-stranded DNA breaks and eventually inducing apoptosis (54–57). In addition, anthracyclines can produce oxidative stress that contributes to their apoptotic effects (58). Doxorubicin has previously been shown to increase the protein levels of wt p53 as well as mutant p53 in different cell lines (59–62) including cervical cancer cells (63–65). Two of the other validated compounds, auranofin, approved in 1985 for the treatment of rheumatoid arthritis and under research for its potential therapeutic application in other diseases including cancer (66, 67), and celastrol, a pharmacologically active compound present in thunder god vine root extracts used as a remedy of inflammatory and autoimmune diseases and under active research for its potential use in the treatment of chronic diseases and cancer (68–70), have been reported to efficiently inhibit the proteasome at similar concentrations to the ones observed to produce accumulation of mRuby-p53 and wt p53 in this study (71, 72). When tested in HeLa cells, auranofin resulted in a decrease in the protein levels of E6AP, E6, and E7, excluding the possibility this compound is acting as a general proteasome inhibitor under these conditions. In contrast, treatment with celastrol increased E6 and E7 protein levels, suggesting inhibition of the proteasome may indeed be involved in the stabilization of p53 by this compound. However, celastrol also led to a decrease in E6AP protein levels, indicating that additional mechanisms or one other than general proteasomal inhibition, may contribute the observed effects of this compound in HeLa cells. The remaining validated compound is chaetocin, a non-specific inhibitor of histone lysine methyltransferases (73, 74) that has been reported to have anticancer activity through different mechanisms depending on the type of cancer cells analyzed (75–78). Interestingly, treatment of Glioblastoma Multiforme cells with 100 nM chaetocin for 24 h has been shown to produce reactive oxygen species that resulted in DNA damage and p53 accumulation (79). We observed mRuby-p53 stabilization and wt p53 accumulation in HeLa cells under analogous conditions suggesting similar mechanisms may be involved. Additional research is needed to explore these possibilities. When tested in different cancer cell lines, only auranofin showed specificity towards HPV16 and HPV18-positive cells, the two types associated with the majority of hrHPV-associated cancers. Celastrol, chaetocin, doxorubicin, and idarubicin lacked HPV specificity and induced p53 stabilization in all the cancer cell lines tested, perhaps reflecting their documented capabilities to induce cellular stress. These results highlight the need for screening novel chemical libraries to search for compounds that may display HPV or HPV type specificity in their effects on p53 stabilization and hrHPV viability/proliferation.

Altogether, we have presented a cell-based reporter assay that allows for monitoring both p53 stabilization and cell viability/proliferation in hrHPV-positive cancer cells over time. Through utilization of fluorescent reporters, this system is amenable to high-throughput screening in a miniaturized format and the per well reagent cost is minimal. It is consistently robust, with high reproducibility and a large dynamic range. The positive compounds identified in this limited screen illustrate the capability of this system to identify relevant compounds in a larger screening campaign that may become potential therapeutics for the treatment of hrHPV infections, cervical cancer, and other hrHPV-associated neoplasias.

Highlights.

• Cell-based reporter assay for p53 stabilization in hrHPV-positive cancer cells.

• Simultaneous detection of changes in cell viability/proliferation.

• Fluorescent reporters allow multiple measurements over time.

• Suitable for high throughput screening.