Phyllanthusmin Derivatives Induce Apoptosis and Reduce Tumor Burden in High-Grade Serous Ovarian Cancer by Late-Stage Autophagy Inhibition

The most lethal gynecologic malignancy is ovarian cancer, of which high-grade serous ovarian cancer (HGSOC) is the most common and deadly subtype (1, 2). The overall 5-year survival rate (46%) of HGSOC has not noticeably changed in the last 30 years (1, 3). Treatment for HGSOC generally involves cytoreductive surgery and a combination of platinum-based (e.g., cisplatin) and taxane (e.g., paclitaxel) chemotherapy. Although first-line treatment is initially effective, relapse with chemoresistance occurs rapidly in up to 80% of cases (4). Though targeted therapies, such as PARP inhibitors and antiangiogenesis agents, have recently been approved for recurrent disease, gains in overall survival have been lagging and new drug development for this disease is warranted (4–7).

Natural products, like paclitaxel, and their derivatives provide a rich source of new chemotherapeutic agents. Phyllanthusmins (PHYs), which were isolated from various Phyllanthus species, are an example of a promising class of antiproliferative natural product compounds (8, 9). In 2014, bioassay-guided fractionation of Phyllanthus poilanei extracts afforded PHYD, a novel and particularly potent member of this class, which exhibited nanomolar activity against a human colon cancer cell line (HT-29). Preliminary in vivo assays indicated that PHYD was able to inhibit the growth of HT-29 embedded intraperitoneally in fibers in athymic nude mice without displaying gross toxicity. The arylnaphthalene lignan lactone substructure of PHYs makes them similar to the aryltetralin, etoposide, a known DNA topoisomerase II inhibitor. However, PHYD was not able to inhibit this enzyme, indicating a novel mechanism of action (9).

Compounds with similar structures to PHYs, especially those with a diphyllin core, have been described as V (vacuolar)-ATPase inhibitors that affect lysosomal acidification in various cell types (10–12). Acidification of the lysosome plays a vital role in autophagy, the process by which cells recycle their intracellular components, organelles, and macromolecules (13). During times of cell stress, autophagy may act as a survival mechanism, whereas too much intracellular recycling can induce cell death (14). Cancer cells may become dependent on autophagy to make intracellular building blocks available for their own energetic and proliferative needs or to eliminate dysfunctional or damaged organelles over time (15, 16). Inhibiting or amplifying autophagy to lethal levels in these cells may be a viable option for treating disease (14, 15, 17, 18). Cancer, and ovarian cancer in particular, appears to be reliant on the process of autophagy as evidenced by a recent analysis of haploinsufficiency networks, which found autophagy and compensatory proteostasis pathways to be the most disrupted pathways in ovarian cancer as compared with 20 other cancer types (19). A number of recent studies are beginning to uncover a role for autophagy in combating HGSOC chemoresistance (20–27) and are investigating potential autophagy-targeting compounds preclinically for efficacy against ovarian cancer (28–32), including V-ATPase inhibitors (33–35).

Building on previous results (9), we synthesized a number of analogues, selecting for potency against ovarian cancer cell lines. In this study, we report on the in vitro and in vivo activity of the three most potent analogues, PHY25, 30, and 34. These compounds inhibit late-stage autophagy, followed by apoptotic cell death in HGSOC. Our data exemplify how iterative medicinal chemistry optimization of a natural product can generate potent late-stage autophagy inhibitors that trigger apoptosis with nanomolar cytotoxicity in vitro and induce significant reduction of tumor burden in vivo.

PHY analogues were synthesized with functionalized carbohydrates to the aglycone core and characterized with ¹H- and ¹³C-nuclear magnetic resonance (NMR) (Supplementary Fig. S1). Prior to biological testing, purity was determined by high-performance liquid chromatography (Supplementary Fig. S2A for PHY25, S2B for PHY30, and S2C for PHY34). Briefly, diphyllin glycoside starting materials (β-D-glycosyl diphyllin, α-l-arabinosyl diphyllin, and β-D-galactosyl diphyllin) were prepared according to a recent publication (36). Following the selective tert-butyldimethylsilyl protection of β-D-glycosyl diphyllin, per-methylation was cleanly accomplished (37). Fluoride-mediated desilylation then afforded PHY25 (Supplementary Fig. S1A). PHY30 and PHY34 (Supplementary Fig. S1B and S1C, respectively) were prepared from their respective glycosyl starting materials via an acid-catalyzed acetal formation (38). The experimental data for PHY30 are in agreement with the same compound that was recently disclosed as in intermediate in the first total synthesis of phyllanthusmin D (39).

Purchased compounds included two chemotherapeutics [paclitaxel (Sigma; #T7402) and etoposide (Sigma; #E1383)], the autophagy inducer [rapamycin (Sigma; #R8781)], an early-stage autophagy inhibitor [PIKIII (Cayman Chem; #17002)], and two late-stage autophagy inhibitors [bafilomycin A1 (Sigma #SML1661 and LC Labs #B-1080 in Fig. 4) and chloroquine (Sigma; #C6628)] (40). Structures of non–FDA-approved small molecules may be found here: PIKIII (41) and bafilomycin A1 (42). All compounds were suspended in dimethyl sulfoxide (DMSO), and final vehicle concentrations in all experiments did not exceed 0.1% except where noted.

The ovarian cancer (OVCAR3 and OVCAR8), colon cancer (HT-29), melanoma (MDA-MB-435), and breast cancer (MDA-MB-231) cell lines were purchased from the American Type Culture Collection. OVCAR8-RFP cells expressing red fluorescent protein were a gift from Sharon Stack at the University of Notre Dame, immortalized human ovarian surface epithelial cells (IOSE80) were a gift from Nelly Auersperg at the University of Vancouver, and eGFP-LC3 HeLa and mCherry-eGFP-LC3 HeLa cell lines were a gift from Ramnik Xavier at Massachusetts General Hospital. OVCAR3 was grown in RPMI 1640 supplemented with 20% FBS, 1% penicillin/streptomycin (P/S), and 10 μg/mL insulin. OVCAR8 and OVCAR8-RFP were grown in DMEM with 10% FBS and 1% P/S. HT-29 and MDA-MB-435 cells were grown in RPMI 1640 with 10% FBS and 1% P/S. MDA-MB-231 cells were grown in DMEM with 5% FBS, 1% l-glutamine, 1% P/S, and 0.02 μg/mL insulin. IOSE80 was maintained in v/v 50% Medium 199 and v/v 50% MCBD with 15% FBS, 1% P/S, 1% l-glutamine, and 11 ng/mL epithelial growth factor. HeLa cells were cultured in DMEM with 8.8% FBS, 1.8% l-glutamine, and 1% P/S. Cultured cells were maintained in a humidified incubator at 37°C in 5% CO₂. Cells were passaged a maximum of 20 times. Cell lines were validated by short tandem repeat analysis and tested mycoplasma-free in 2017.

Cells were seeded in 96-well, clear, flat-bottomed plates at 2,500 to 5,000 cells per well, depending on the cell line, and allowed to attach overnight. Compounds suspended in DMSO were diluted to final concentrations and added to the cells. The final vehicle concentration was 0.25% to achieve a wide dose range. Cells were incubated for 24, 48, or 72 hours. Cellular protein content was measured using a sulforhodamine B assay as a measure of cell survival (43). Treatment measurements were normalized to vehicle, and dose–response curves with corresponding IC₅₀ values were generated using Graphpad Prism Software.

OVCAR3 and OVCAR8 were plated at a concentration of 400 or 200 cells per 60 mm dish, respectively, allowed to attach overnight, and treated as indicated for 72 hours. Foci were allowed to grow in nontreated medium with media changes every 3 days. When two-dimensional (2D) foci became visible (day 20 for OVCAR3 and day 14 for OVCAR8), they were fixed with 4% paraformaldehyde, stained with 0.05% crystal violet, and imaged with the FluorChem E system (ProteinSimple). Foci counts were quantified with ImageJ software (NIH) and normalized to the vehicle control. Significance was determined using one-way ANOVA with Dunnett multiple comparisons, compared with the vehicle control.

After culture and treatment, cells were lysed in RIPA lysis buffer (50 mmol/L Tris, pH 7.6, 150 mmol/L NaCl, 1% Triton X-100, 0.1% SDS) with protease (Roche Applied Science) and phosphatase (Sigma) inhibitors, incubated at −80°C, and centrifuged. Protein concentration was obtained using the bicinchoninic acid assay (Pierce), and 15 μg protein per sample was separated by SDS-PAGE and transferred to nitrocellulose or activated 0.45 μm PVDF membranes (Thermo Fisher). After 5% milk block, the following primary antibodies were applied to blots at a 1:1,000 concentration: GAPDH as a loading control (Cell Signaling Technology; #2118), PARP (Cell Signaling Technology; #9542), or LC3B (Cell Signaling Technology; #2775). Membranes were incubated with 1:1,000 anti-rabbit secondary antibody (Cell Signaling Technology; #7074) prior to visualization of signal with SuperSignal West Femto substrate (Thermo Fisher) and imaging on a FluorChem E system (ProteinSimple).

Cells were seeded into T25 flasks, allowed to attach overnight, and treated as indicated for 3 days. Cells were subjected to an Annexin V–FITC/Propidium Iodide Apoptosis Assay (Nexcelom Biosciences) according to instructions provided by the manufacturer, except with 10 μL Annexin V–FITC reagent per sample. Fluorescence was detected using a K2 Cellometer (Nexcelom Biosciences). Data were analyzed with the FCS Express program (De Novo Software). Each replicate contained at least 2,000 cells for quantification of fluorescent signal. Statistics were generated with one-way ANOVA with Dunnett's multiple comparisons to vehicle control within each group (Fig. 2B) or Tukey's multiple comparisons (Fig. 6C).

HeLa cells stably expressing eGFP-LC3 (LC3 puncta assay) or stably expressing mCherry-eGFP LC3 (autophagic flux assay) were seeded in 384-well plates at 3,000 cells per well (44). After attachment overnight, test compounds and controls were moved into assay plates using a 96-well pin tool (V&P Scientific) and a liquid handler, Biomek NXP Lab Automation Workstation (Beckman Coulter). Assay plates were incubated at 37°C for 4 (LC3 puncta assay) or 24 (autophagic flux assay) hours. Cells were then fixed with w/v 4% formaldehyde (Thermo Scientific) in PBS for 12 minutes, washed with PBS, DNA-stained with 2 μg/mL Hoechst 33342 (Molecular Probes) in PBS for 12 minutes, and diluted with PBS. Plates were sealed using PlateMax Semi-Automatic Plate Sealer (Axygen). Plates were imaged at 10X by an automated fluorescence microscope, ImageXpress Micro XLS (Molecular Devices), using DAPI and FITC filters (both assays) and Texas Red filter for the autophagic flux assay. Four sites were imaged per well. The number of puncta per cell was quantified using the MetaXpress High-Content Image Analysis Software with Transfluor Application Module (Molecular Devices). The average number of autophagosomes (eGFP⁺/mCherry⁺; yellow puncta) and autolysosomes (eGFP⁻/mCherry⁺; red puncta) per cell was quantified using CellProfiler Software (Broad Institute). Treatment measurements were normalized to vehicle, and dose–response curves with corresponding EC₅₀ values were generated using GraphPad Prism Software.

Cells were plated according to the manufacturer's instructions on 8-well, culture-treated μ-slides (Ibidi USA Inc.) for live cell confocal imaging. Cells were treated as indicated and incubated in either 1 μg/mL acridine orange (Sigma) in PBS for 15 minutes or 50 nmol/L Lysotracker Red DND-99 (Invitrogen) in PBS for 30 minutes before imaging on a Zeiss Laser Scanning 710 confocal microscope. Images were obtained with Zen software (Zeiss).

All animals were treated in accordance with the NIH Guidelines for the Care and Use of Laboratory Animals and the established Animal Care and Use Committees at the Ohio State University (OSU, protocol #2009A0196-R2) of the University of Illinois at Chicago (UIC, protocol #16-035). In vivo studies conducted at OSU (MTD, bioavailability) utilized ICR male mice 6 weeks in age and around 20 grams in weight (Harlan Laboratories), and studies at UIC (hollow fiber assay, intraperitoneal xenografts) utilized NCr nu/nu athymic female mice 6 to 8 weeks in age (Taconic). Mice used in all in vivo assays were housed in a temperature- and light-controlled environment under 12:12 hour light:dark cycles and provided food and water ad libitum.

PHY34 was prepared at a starting concentration of 120 mg/mL in DMSO, 50 mg/mL in PEG300 and 30% HPβCD, or 1 mg/mL in PBS and centrifuged at 13,500 rpm for 10 minutes. The soluble fraction was serially diluted, and 100 μL was analyzed by absorption spectrometry using a BioTek Synergy HT Multi-Mode Microplate Reader. The dilution of the original formulation within the linear range of the calibration curve was used to assess the concentration of PHY compounds in the soluble fraction. Data were acquired by using Software Gen5 2.05 (BioTek).

Mice were administered PHY34 in the formulation of 50% PEG300:50% saline intravenously at dose levels of 3, 1, 0.6, and 0.3 mg/kg until the maximum tolerated dose (MTD) was obtained. MTD is defined as the highest dose that can be administered to a mouse without causing unacceptable toxicity. MTDs were obtained starting at 10x the intravenous MTD by the intraperitoneal route or 50x the intravenous MTD by the oral route and decreasing doses down to 3x or 1x the intravenous MTD. Five mice were used at each dosing level for each route of administration.

A sterile dosing solution of PHY34 was prepared in 50% PEG300:50% saline. Mice were dosed at the MTD via intravenous (0.6 mg/kg) or intraperitoneal (1.8 mg/kg) injections or 75 mg/kg for oral administration. The injection volumes were 100 μL for intravenous and 200 μL for intraperitoneal and oral route per mouse. Around 30 mice for each dosing route were used to perform plasma pharmacokinetic (PK) study. At each time point after dosing, 3 mice were euthanized by CO₂ asphyxiation, and blood was immediately collected via cardiac puncture then transferred to heparinized tubes. PHY34 was quantified using LC-MS/MS (LC, liquid chromatography; MS, mass spectrometry) with a Dionex RSLC nano LC system and a TSQ Quantiva mass spectrometer. Briefly, 200 μL internal standard solution (200 ng/mL PHY1, analogue of PHY34, in DMSO/isopropal alcohol/acetonitrile, 5/10/10 v/v/v) was added to 100 μL mouse plasma followed by centrifugation at 4°C and 13,500 rpm for 5 minutes. After nitrogen drying, 100 μL water was added to reconstitute the analytes, which were loaded in an autosampler vial for LC-MS/MS analysis. Analytes were separated on an Extend C-18 column (50 × 2.1 mm, 3.5 μm) using a gradient of water and isopropanol, each with 0.1% formic acid. Mass transitions monitored were 583.3 > 381.2 (PHY34) and 711.3 > 109.1 [M+H]⁺. The calibration curve was linear from 1 to 2,000 nmol/L in mouse plasma. Plasma concentration–time data were analyzed by noncompartmental methods using default settings and the NCA analysis object in Phoenix WinNonlin v6.3 (Certara).

OVCAR3, OVCAR8, and HT-29 cells were grown and embedded in biocompatible hollow fibers for xenograft mice intraperitoneally on day 0 as previously described (45, 46). Paclitaxel (5 mg/kg in 20% DMSO:40% PEG300:40% H₂O), PHY analogues (0.75 mg/kg in 10% DMSO:40% PEG300:50% H₂O), or vehicle (2% DMSO:48% PEG300:50% H₂O) were dosed intraperitoneally once per day for 4 days (days 3–6). Treatment group sizes were as follows: 3 mice received paclitaxel, 7 received each PHY analogue, and 6 received vehicle. Mice were sacrificed on day 7, and fibers retrieved. The viable cells were evaluated with a modified 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay (Promega). Statistics were generated with one-way ANOVA with Holm's multiple comparisons to vehicle control for each cell line.

OVCAR8-RFP cells were xenografted intraperitoneally (5 × 10⁶ cells per mouse), and tumor growth was monitored with the Xenogen IVIS Spectrum In Vivo Imaging System (PerkinElmer) as previously described (47). When tumors became detectable via IVIS imaging, mice were separated into treatment groups of 5 to 6 mice and dosed once daily 3 times per week with paclitaxel (5 mg/kg), PHY34 (0.75 mg/kg), or vehicle (10% DMSO:40% PEG300:50% H₂O) for 3 weeks. Mice were weighed 3 times a week and imaged once weekly with an exposure of 2 seconds, an Fstop of 2, and excitation and emission wavelengths of 535 and 620 nm, respectively. Average abdominal radiant efficiency was quantified with the system's Living Image 4.0 software and normalized to day 0. After week 3, mice were sacrificed and tumors were collected for immunohistochemistry (IHC). Statistics were generated with two-way ANOVA with Dunnett's multiple comparisons to vehicle control.

Tumors obtained from the i.p. xenograft experiment were fixed with 4% paraformaldehyde before dehydration and paraffin sectioning. Sectioning and immunohistochemistry (IHC) were performed using standard histologic procedures. Sections were incubated in anti-Ki67 (Abcam; #ab15580), followed by anti-rabbit IgG (Leica Biosystems; #DS9800). Immunoreactivity was visualized with 3,3-diaminobenzidine chromogen and counterstained with Mayer's hematoxylin. Tissues without primary antibody treatment were used as a negative control. Images were acquired on a Nikon Eclipse E600 microscope using a DS-Ri1 digital camera and NIS Elements software (Nikon Instruments).

Data presented are the mean ± SEM and represent at least three independent experiments. Statistical analysis was carried out using GraphPad Prism software and SAS 9.4 for the hollow fiber assay. Statistical significance was determined by ANOVA with Dunnett's, Holm's, or Tukey's adjustment for multiple comparisons, as noted in methods/legends. Adjusted P < 0.05 is considered significant. Symbols used to signify significance include: *, P < 0.05; †, P < 0.01; ‡, P < 0.001; §, P < 0.0001.

PHYs originally isolated from Phyllanthus poilanei Beille with demonstrated activity against cancer cell lines in vivo were used as inspiration to generate synthetic analogues with greater potency based on alterations to the sugar moiety (9). In order to determine if synthetic analogues were more potent than the natural products, a series of cell lines was tested with increasing concentrations (0.001 nmol/L to 50 μmol/L) of PHY25, 30, and 34 (Fig. 1A). HGSOC cell lines, OVCAR8 and OVCAR3, were selected based on their designation as reliable models of HGSOC (48). Colon cancer cell line, HT-29, was also included for continuity as it was the cell line used to discover the original natural product PHY derivatives (9). IOSE80 served as the nontumorigenic cell control (49). Paclitaxel and etoposide were included as positive controls. Dose–response curves (Fig. 1B) with corresponding IC₅₀ values (Fig. 1C) were generated. PHY34 was the most cytotoxic compound in both HGSOC cell lines with an IC₅₀ of 4 nmol/L, more potent than that of paclitaxel (8–10 nmol/L). PHY30 was cytotoxic in both HGSOC cell lines (IC₅₀ = ∼12 nmol/L), whereas PHY25 was selectively potent against OVCAR8 (IC₅₀ = 17 nmol/L) over OVCAR3 (IC₅₀ = 1.9 μmol/L). PHY30 and 34 were not cytotoxic (IC₅₀ > 50 μmol/L) to the nontumorigenic cell line control, IOSE80. PHY34 was subjected to the NIH NCI's 60 Tumor Cell Line Screen (50, 51), which showed PHY34′s capability to significantly inhibit the growth of 8 of the 9 types of cancer represented, including ovarian, colon, melanoma, and leukemia, after 48-hour treatment (Supplementary Fig. S3A and S3B). PHY34 showed greater ability to inhibit the growth of OVCAR8 as compared with OVCAR3 in this screen. We additionally tested PHY25, 30, and 34 against two cell lines from the NCI screen, melanoma (MDA-MB-435) and triple-negative breast cancer (MDA-MB-231), for 72 hours (Supplementary Fig. S3C). PHYs, especially PHY34, had nanomolar IC₅₀s in both these cell lines and were more cytotoxic against MDA-MB-231 (PHY34 IC₅₀ in MDA-MB-435 = 23 nmol/L, PHY34 IC₅₀ in MDA-MB-231 = 5.2 nmol/L).

To further characterize if the PHY compounds were effective due to cytostatic versus cytotoxic action, 2D foci assays were performed (Fig. 1D), where cells were exposed to drug for 72 hours and then foci were allowed to grow in drug-free medium. PHYs displayed greater potency in OVCAR8 than in OVCAR3 with all analogues being cytotoxic at 10 nmol/L. In OVCAR3, PHY25 and 30 were cytotoxic at 1 μmol/L, and PHY34 was cytotoxic at 100 nmol/L. The low concentration (100 nmol/L) of PHY25 and 30 was cytostatic in OVCAR3, showing decreased 2D foci formation compared with vehicle (DMSO) control.

In order to determine the mechanism of cell death in HGSOC cell lines, Western blot analyses were performed for cleaved PARP (cPARP) as a biomarker for apoptosis. cPARP levels were increased in all PHY analogue treatments in OVCAR8 and most prominently with PHY34 in OVCAR3 (Fig. 2A). Apoptosis induction was confirmed with Annexin V–FITC (AV) and propidium iodide (PI) co-staining (Fig. 2B). PHY analogue treatment at 10 nmol/L resulted in a significant increase in the number of cells in early and late apoptosis (AV⁺,PI⁻ and AV⁺,PI⁺, respectively) in OVCAR8 as compared with the vehicle control, with a corresponding decrease in nonapoptotic cells (AV⁻,PI⁻). Only PHY34 treatment (100 nmol/L) resulted in significant increases in early and late apoptotic cells in OVCAR3. Because PHY34 was the most potent compound in both HGSOC cell lines and induced apoptosis to the highest extent, it was selected for use in remaining studies.

Prior to conducting in vivo experiments, the solubility and MTD of PHY34 were assessed. PHY34 was resuspended in various solvents at increasing concentrations to elucidate the maximum soluble concentration displayed in mg/mL and mmol/L in Supplementary Fig. S4A. PEG300 was selected for in vivo experiments. Observed MTDs of PHY34 by various routes of administration are as follows: 0.6 mg/kg by intravenous administration, 1.8 mg/kg by intraperitoneal administration, and >30 mg/kg by oral administration. No toxicities were seen by oral administration. An intraperitoneal dose of 0.75 mg/kg, less than half the MTD, was selected for in vivo studies.

To elucidate the bioavailability of PHY34 administered by various routes, plasma from mice dosed with 0.6 mg/kg intravenously, 1.8 mg/kg intraperitoneally, or 75 mg/kg orally was subjected to LC-MS/MS analysis to determine the PHY34 plasma concentration over time (Fig. 3A). The plasma concentration–time data were analyzed by noncompartmental methods to determine PK parameters listed in Supplementary Fig. S4B. The observed intraperitoneal bioavailability (F) for PHY34 was 56.6% and for oral route was 2.5%. Systemic clearance (Cl) following oral dose was 194.1 L/hr/kg, roughly 40 times the mouse liver blood flow, suggesting significant extrahepatic clearance. The T_max of PHY34 (0.25 hour) suggests very fast absorption of this compound orally.

To evaluate the ability of PHYs to induce toxicity in an animal model, a hollow fiber assay was performed (Fig. 3B). Cancer cell lines (OVCAR3, OVCAR8, and HT-29) embedded in biocompatible hollow fibers were implanted intraperitoneally into female nude mice. All analogues significantly decreased HT-29 tumor cell survival in the hollow fiber. PHY34 significantly reduced cell viability of OVCAR8, but not OVCAR3, and was selected for further in vivo study.

OVCAR8 cells fluorescently labeled with RFP (OVCAR8-RFP) were xenografted intraperitoneally into female nude mice. Tumor formation was monitored by IVIS imaging, and once tumors were established, they were dosed 3 times a week for 3 weeks. PHY34 (0.75 mg/kg) and the positive control, paclitaxel (5 mg/kg), significantly decreased tumor burden based on average abdominal radiant efficiency as compared with vehicle at week 3 with paclitaxel also achieving significant tumor burden reduction at week 2 (Fig. 3C). No gross toxicities were observed, and mice maintained similar body weights to vehicle animals throughout the study in contrast to the paclitaxel control group, which saw a significant reduction in body weight (Supplementary Fig. S4C). The proliferative marker, Ki67, was reduced in tumors treated with PHY34, as illustrated by immunohistochemical analysis (Fig. 3D) of tumors from 3 mice per group.

Compounds with similar diphyllin core structures to PHYs have been shown to modulate autophagy (10–12). PHY25, 30, and 34 were subjected to two screening assays for autophagy modulators (44). HeLa cells expressing eGFP-LC3B were treated with PHYs for 4 hours, fixed, and fluorescent LC3B puncta were counted over a range of doses (9.31 fmol/L to 20 μmol/L; Fig. 4A–C). PHYs, especially PHY34, sustained high levels of LC3B puncta at low doses (EC₅₀ = 2 nmol/L). In comparison, an early natural product analogue, PHYC (9), had a much higher EC₅₀ (979 nmol/L).

Because an increase in LC3B puncta can be caused by autophagy inducers or late-stage autophagy inhibitors, an autophagic flux assay was performed (Fig. 4D–F), wherein an LC3B protein tagged with both mCherry and eGFP was visualized to count autophagosomes (eGFP⁺/mCherry⁺) and autolysosomes (eGFP⁻/mCherry⁺). Flux was inferred by the acidic stability of the tags: the eGFP signal was attenuated in the acidic lysosomal environment, whereas the mCherry signal remained stable (52). An accumulation of autophagosomes, or higher ratio of autophagosomes/autolysosomes, is indicative of a late-stage inhibitor. Bafilomycin A1 (BAF) and chloroquine (CQ), both late-stage inhibitors, had significantly higher autophagosome/autolysosome ratios. All PHY compounds performed similarly to BAF and CQ as late-stage inhibitors of autophagy, and PHY34 had the greatest ability to inhibit autophagy of all compounds tested (EC₅₀ = 3.9 nmol/L). The original natural product, PHYC, required a much higher dose to achieve the same effect (EC₅₀ = 1.1 μmol/L). An increased ability to modulate autophagy correlated with the increased cytotoxicity seen in these synthetic analogues over their natural product predecessor.

Because similar compounds have been hypothesized to modulate autophagy (10–12), we next visualized lysosomes and their acidity. OVCAR8 and OVCAR3 cells were treated with PHY34 for 24 hours at 100 nmol/L or 1 μmol/L, respectively, and stained with either Lysotracker Red or acridine orange to monitor changes in the lysosomes due to autophagy inhibition using live confocal microscopy. A reduction in the fluorescence from the lysosomes was detected after treatment with PHY34 in both ovarian cancer cell lines (Fig. 5A). BAF, a late-stage autophagy inhibitor, almost completely eliminated lysosome detection at 50 nmol/L. PIKIII (an early-stage autophagy inhibitor), which should block autophagy before the changes to lysosomal expression, was visually similar to DMSO (40). Acridine orange, a green dye that emits red/orange light when sequestered in acidic cellular compartments like the lysosome, similarly displayed a dramatic reduction in fluorescence after treatment with PHY34 or BAF (Fig. 5B), suggesting a reduction in lysosomal acidification.

In order to determine if autophagy preceded apoptosis, a series of time courses was evaluated. PHY34 induced cytotoxicity earlier and with greater potency in OVCAR8 than OVCAR3 (Fig. 6A; Supplementary Fig. S5 for PHY25 and 30). Western blot analyses (Fig. 6B) supported an earlier induction of apoptosis in OVCAR8 based on enhanced cPARP levels after 48 hours, whereas it was not apparent until 72 hours in OVCAR3. Next, PHY34 was combined with a late-stage autophagy inhibitor, BAF, and demonstrated an additive effect on apoptosis induction. To determine if autophagy inhibition was necessary to induce apoptosis, the combination of a late-stage autophagy inhibitor (PHY34 or BAF) was tested with an autophagy inducer, RAP (40). The presence of RAP reversed the conversion of PARP to cPARP. A rescue of apoptosis was also observed with Annexin V–FITC and PI costaining (Fig. 6C) where the addition of RAP to PHY34 treatment significantly decreased the number of apoptotic cells as compared with PHY34 alone in both OVCAR8 at 48-hour treatment and OVCAR3 at 72-hour treatment. In contrast, PIKIII, an early-stage autophagy inhibitor, did not contribute to apoptosis-mediated cell death. To confirm that autophagy was directly affected by PHYs and the various controls, the conversion of LC3B-I to LC3B-II was monitored. Correspondingly, an increase in LC3B-II level can be seen in all PHY34- and BAF-treated cell lysates. The increase in LC3B-II can be seen as early as 24 hours in OVCAR8 and 48 hours in OVCAR3, suggesting that the late-stage inhibition of autophagy precedes the induction of apoptosis.

This work demonstrates that the synthetic derivatives, PHY25, 30, and 34, act as effective late-stage autophagy inhibitors able to interfere with tumor growth in vivo. Inhibition of late-stage autophagy by PHY34 may be required for the resulting apoptotic cell death. Importantly, late-stage autophagy inhibition by PHY34 preceded apoptosis induction, and this chain of events occurred at least 24 hours earlier in the more PHY-sensitive cell line, OVCAR8. Combination treatment with an autophagy inducer, RAP, rescued the PHY-treated cells from apoptosis. This implies that the apoptotic cell death seen after PHY34 treatment may rely on its ability to block autophagy. Bafilomycin A1 and RAP co-treatment displayed a similar rescue, further supporting PHY34′s role as a late-stage autophagy inhibitor. PHY34 was shown to modulate autophagy by its ability to increase both LC3B puncta and the autophagy marker, LC3B-II. An autophagic flux assay revealed its action as a late-stage inhibitor of the pathway, blocking the final breakdown of the autolysosomes. Overall, these PHY compounds, PHY25, 30, and 34, demonstrated nanomolar cytotoxicity in HGSOC cell lines and were able to reduce tumor burden in vivo.

For this study, derivatization of the PHY class of natural products focused on the modification and optimization of the glycone portion of the molecules and involved the introduction of variously functionalized carbohydrate scaffolds to diphyllin. These efforts provided PHY25 and 30, each of which possesses a different chemical moiety thought to be an important pharmacophore. A hybrid of the two compounds, incorporating the key pharmacophore from each, was designed and synthesized, providing PHY34. This technique resulted in the aforementioned improvement of PHY34′s in vitro and in vivo potencies as compared with the parent compounds (PHY25 and 30).

PHYs exhibit qualities that suggest they are good chemotherapeutic drug leads. Of the three compounds, PHY34 was effective in vivo against both ovarian and colon cancer models without gross toxicity, and in vitro was not toxic to the normal cell control. The dose of PHY34 required to eliminate tumors in vivo was lower than that of paclitaxel (0.75 as compared with 5 mg/kg), and the animals displayed almost no change in body weight with PHY34 treatment. In the NCI 60 Tumor Cell Line Screen, PHYs demonstrated promising growth inhibition in 8 of the 9 tumor types investigated at 48-hour treatment and had nanomolar cytotoxicity against two cell lines from this screen, a melanoma and a triple-negative breast cancer cell line, after 72-hour treatment. Of the HGSOC cell lines, OVCAR3 was less sensitive to PHY34 treatment than OVCAR8, and mechanistic effects of PHY treatment, such as autophagy inhibition and apoptosis induction, occurred 24 hours later in OVCAR3. However, both OVCAR3 and OVCAR8′s IC₅₀s were in the nanomolar range, and the delay in cytotoxicity may be explained by the slower growth of OVCAR3 in cell culture. Originally, the structure of PHYD was compared with etoposide, and both previous in vitro assays studying topoisomerase inhibition and DNA coil relaxation confirmed that these PHYs have a different mechanism (9). Our data further demonstrate that PHYs act differentially based on their far more potent IC₅₀ values in vitro and their strong ability to eliminate colonies after cessation of treatment. These compounds were bioavailable through multiple routes of administration, including oral. Notably, despite the relatively low oral bioavailability of 2.5%, oral doses achieved plasma concentration greater than in vitro IC₅₀ values, which suggests oral dosing is feasible.

The development of autophagy inhibitors for the treatment of HGSOC is currently an active area of interest and research. Indeed, an increasing number of compounds are being developed to target autophagy (28–35). Late-stage autophagy inhibitors, like bafilomycin A1 and PHYs, are able to induce apoptotic cell death, whereas the early-stage inhibitor used in this study, PIKIII, requires additional cellular stress or nutrient deprivation (40). Large-scale studies have demonstrated the particular importance of the autophagic pathway in ovarian cancer (19). One analysis revealed greater V-ATPase immunoreactivity in epithelial ovarian cancer patient samples (n = 59) and decreased overall survival in patients expressing higher levels of V-ATPase mRNA by analysis of the Cancer Genome Atlas data (33), suggesting that V-ATPase expression could serve as a viable biomarker to stratify patient response to its inhibition. Biomarkers for autophagy, such as Beclin1 and LC3A, have also been shown to correlate with poor prognosis in epithelial ovarian (53) and clear-cell ovarian cancers (54), respectively. Future studies should continue to assess the prognostic value of autophagy markers in HGSOC and work to detail the mechanisms that govern the potency of these preclinical compounds.

Furthermore, combination treatment with autophagy modulators may combat disease recurrence and chemoresistance (20–27), an issue plaguing approximately 80% of HGSOC patients (4). Of over 20 current clinical trials involving autophagy modulators, almost all involve chloroquine or hydroxychloroquine, and none are being conducted in ovarian cancer patients. Specific inhibition of V-ATPase subunits by shRNA and inhibitory monoclonal antibodies increases cisplatin sensitivity and reverses platinum resistance in ovarian cancer in vitro (55). Specific attention should be paid to elucidate the best therapeutic strategy of autophagy modulation and combination therapies, including which patients would best respond to its induction or inhibition, and the specific means by which perturbation of the autophagic pathway elicits cytotoxicity via apoptosis. Taken together, this study supports further investigation of autophagy inhibitors for the treatment of HGSOC due to the variety of potentially clinically useful biomarkers inherent in their mechanism of action and the promise of extending patient survival by delaying chemoresistance with strategic combination therapy.

No potential conflicts of interest were disclosed.

Conception and design: A.N. Young, A.D. Kinghorn, J.R. Fuchs, J.E. Burdette

Development of methodology: A.N. Young, J.E. Burdette

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.N. Young, D. Herrera, A.C. Huntsman, M.A. Korkmaz, D.D. Lantvit, S. Kolli, S. King, H. Wang, M.A. Phelps, J.E. Burdette

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.N. Young, M.A. Korkmaz, S. Kolli, C.C. Coss, X. Zhang, M.A. Phelps, L.N. Aldrich, J.E. Burdette

Writing, review, and/or revision of the manuscript: A.N. Young, D. Herrera, A.C. Huntsman, C.C. Coss, S.M. Swanson, A.D. Kinghorn, X. Zhang, M.A. Phelps, L.N. Aldrich, J.R. Fuchs, J.E. Burdette

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.C. Huntsman, C.C. Coss, S.M. Swanson, J.E. Burdette

Study supervision: L.N. Aldrich, J.R. Fuchs, J.E. Burdette

Other (design, synthesis, and characterization of chemical constituents utilized in the study): A.C. Huntsman

Other (performed the animal studies for PHY compounds and compiling/analysis of data for the solubility studies): S. Mazumder

The authors gratefully acknowledge the support of grants P01CA125066 and 1F30CA217079 from the NCI of the NIH (Bethesda, MD). We are grateful to the University of Illinois at Chicago Research Resource Center's Center for Cardiovascular Research and Physiology Core for training and use of the Xenogen IVIS Spectrum Imager, the Histology Core for their immunohistochemical processing of tissues obtained from in vivo experiments, and the Core Imaging Facility for training and use of the Zeiss LSM 710 Confocal Microscope.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.