Ovarian clear cell adenocarcinoma (OCCA) is characterized by a particularly poor response to conventional chemotherapy and a short overall survival time in women with established disease. The development of targeted treatments for OCCA relies on a better understanding of its molecular characteristics. IL6 is strongly expressed in OCCA and may therefore provide a novel therapeutic target. Here we use CRISPR/Cas9 and conditional short hairpin interfering RNA to perform loss-of-function studies in human OCCA cell lines to explore the requirement for IL6 in vitro and in vivo. While reduction of IL6 expression exerted limited effects in vitro, its attenuation significantly impaired tumor growth and neovascularization in vivo. In contrast to typical signaling via STAT3, IL6 in OCCA signaled via a noncanonical pathway involving gp130, Src, and the Hippo pathway protein YAP. A high-throughput combination drug screen identified agents that enhanced cell killing following reduction of IL6 signaling. Intersection of screen hits obtained from two cell lines and orthogonal approaches to attenuation of IL6 yielded AKT and EGFR inhibitors as enhancers of the inhibitory monoclonal IL6 receptor antibody tocilizumab. This study defines for the first time the requirements for, and mechanisms of, signaling by IL6 in human OCCA cell lines and identifies potential combinatory therapeutic approaches. Given the molecular diversity of OCCA, further in vitro and in vivo studies are warranted to determine whether such approaches will overcome the limited efficacy of tocilizumab observed in ovarian cancer to date.
This study defines the requirements for and mechanisms of noncanonical signaling by IL6 in human ovarian clear cell adenocarcinoma cell lines and identifies combinatory therapeutic approaches to be explored clinically.
Ovarian clear cell adenocarcinoma (OCCA) accounts for approximately 10% of invasive epithelial ovarian cancers in most Western nations but is more common in Asian countries, including Japan, where it is associated with almost 30% of ovarian cancers (1, 2). It is generally refractory to platinum-based therapy, with a response rate of only 11%–15% (3). OCCA is associated with coexistent endometriosis (4, 5), and a definitive molecular link between these diseases has been demonstrated through cooccurring somatic mutations in the chromatin-modifying gene ARID1A (6).
The development of novel cancer therapies rests on first gaining insight into the biology of the disease. In contrast with serous cancers, TP53 mutation is uncommon in OCCA (7). OCCA cell lines show a gene expression pattern associated with oxidative stress, glycogenesis, MAPK pathway, and cytokine activation (8). Approximately 50% of OCCAs carry somatic mutations in ARID1A, which encodes BAF250a, a member of the SWI/SNF chromatin-modifying proteins. OCCA is characterized by very high levels of circulating IL6 arising from the transformed epithelial part of the tumor (8–10). It was recently shown that coexistent ARID1A and PIK3CA mutations could only promote OCCA carcinogenesis in a mouse model through sustained IL6 overproduction (11). IL6 therefore appears to be central to OCCA biology and as such provides a novel therapeutic target in OCCA.
IL6 is a proinflammatory cytokine that has tumor-promoting actions including induction of cell proliferation, survival, migration, invasion, and angiogenesis, and promotes inflammatory responses in both malignant and stromal cells (12). IL6 activates pathways involving JAK–STAT3, SHP2–Ras–ERK, and PI(3)K–AKT–mTORC1 via the common coreceptor gp130 (13, 14). Recent studies have identified a noncanonical IL6 pathway involving gp130 signaling via the transcriptional coactivator Src–Yes-associated protein (YAP), a member of the Hippo growth control pathway (15). This is of interest, as we previously showed that YAP is specifically overexpressed and required in OCCA (16).
Here we provide a comprehensive study of the role of IL6 signaling in human OCCA using CRISPR/Cas9-mediated knockout (KO) and conditional short hairpin RNA (shRNA)-mediated knockdown (KD) models in OCCA cell lines, both in vitro and in vivo, to better understand its value as a therapeutic target. We reveal a role for the noncanonical IL6 signaling mechanism pathway in OCCA and demonstrate potential approaches in targeting IL6 signaling pathway directly and indirectly.
Materials and Methods
All animal experiments were approved by the Peter MacCallum Cancer Centre Animal Experimentation Ethics Committee and conducted in accordance with the National Health and Medical Research Council Australian Code of Practice for the Care and Use of Animals for Scientific Purposes.
OCCA cell lines were obtained from the NCI Repository, actively passaged for less than 6 months. Cell lines were confirmed to be Mycoplasma negative (MycoAlert, Mycoplasma Detection Kit, Lonza) in August 2013 and fingerprinted using short tandem repeat (GenePrint 10 System, Promega) markers to confirm their identity against the Cancer Genome Project database (Wellcome Trust Sanger Institute, Cambridgeshire, United Kingdom). Cell were maintained at 37°C and 5% CO2 [volume for volume (v/v)] and cultured in RPMI1640 media (Gibco) containing 10% (v/v), FCS (Gibco), and 1% penicillin/streptomycin (Gibco) as described previously (17).
Gene expression microarray analysis
Clustering analysis of IL6/STAT3 signaling gene signatures was performed on a large panel of HGSC and OCCA cell lines using our previously published microarray expression dataset (9) and the gplots R package (version 3.0.1). IL6/STAT3 signature genes were selected on the basis of the Human IL6/STAT3 Signaling Pathway Plus PCR Array Kit (Qiagen).
CRISPR/Cas9-mediated IL6 KO in OCCA cell lines
IL6 KO OCCA cell lines were generated using a lentiviral CRISPR/Cas9 system as described previously (18). Briefly, a dual lentiviral vector expression system consisting of (i) a constitutive Cas9 endonuclease (FUCas9) and (ii) inducible single-guide RNA (sgRNA; FGT1-UTG) vectors linked to a mCherry and GFP reporter. Two sgRNA sequences were designed to target exon 2 and 3 of IL6 using MIT online tool (http://crispr.mit.edu). JHOC5 and TOV21G cells were transduced with FUCas9 (MOI<5) and FACS sorted for mCherry-positive cells followed by transduction with FGT1-UTG and FACS sorted for GFP+ cells. Cells were seeded in 6-well plates and gRNA expression was induced 24 hours postseeding with doxycycline (μg/mL; Sigma-Aldrich, D9891) for 72 hours before single-cell sort by FACS to generate clonal cell lines. IL6 status was verified by IL6 ELISA (R&D Systems) and INDELS were confirmed by MiSeq sequencing.
Conditional shRNA–mediated IL6 KD OCCA cell lines
Conditional IL6 shRNA-mediated KD in vitro and in vivo models were generated using a lentiviral shRNA vector system as described previously (19, 20). Briefly, optimized mirE backbone is a dual reporter construct that delivers both a Tet-responsive element promoter–driven shRNA and rtTA in a single lentiviral vector (T3G-mCherry-miRE/shRNA-PGK-tGFP-IRES rtTA3 (LT3CEPGIR). JHOC5 and TOV21G were transduced with LT3 GEPCIR_shRNA IL6 and LT3 CEPGIR-shRNA NS (nonspecific). MCherry-positive cells were FACS sorted for each line and treated with doxycycline for 72 hours to confirm that all positive cells express shRNA. We screened for 10 unique shRNA sequences and we selected only two highly potent shRNAs against IL6 (IL6_shRNA5 and IL6_shRNA10) for each line for further studies. shRNA expression, and KD efficiency was validated by RT-PCR and Western blot analysis.
To monitor target gene KD, the total RNA was isolated and qRT-PCR was performed as described previously (21). Primer sequences used for qRT-PCR are provided in Supplementary Table S1.
CellTiter 96 AQueous One Solution Cell Proliferation Assay (Promega) was performed according to the manufacturer's protocol to measure the proliferative rate of each cell line. Briefly, 3,000 cells were seeded in a 96-well plate in triplicates and incubated for 72 hours (37°C, 5% CO2). Cells were then incubated for further 3 hours with AQueous One Solution Reagent, and absorbance was measured at 490 nm using an ELISA plate reader.
Clonogenic survival assay
For clonogenic assays, cells were trypsinized to form a single-cell suspension and were counted using Coulter Counter (Beckman Coulter Life Sciences). Cells were then seeded at low density (500–2,000 cells) in wells of a 6-well plate in triplicates and left to form colonies for up to 10 days. At 7–10 days, cell colonies were fixed and stained with 20% (v/v) methanol and 0.1% (w/v) crystal violet. Cells were rinsed in water, air dried, and discrete colonies were counted using MetaMorph (Molecular Devices; ref. 22).
Cell migration assay
Migration assays were performed as described previously (23). Briefly, cells were plated into a 96-well cell culture plate and were allowed to grow in 10% FBS-containing RPMI medium to form a monolayer. Cells were wounded by generating a longitudinal scratch using a robotically driven (Seiko) stainless-steel pin machined to deliver a scratch of 1.5 × 4 mm2. After wounding, cells were washed once with growth medium and incubated for further 8 and 16 hours at 37°C. Wound closure was analyzed at 8 hours and 16 hours. Cells were fixed (2% paraformaldehyde, 200 mmol/L KCL, 20 mmol/L Pipes pH 6.8, 14% sucrose, 4 mmol/L MgCl2, and 2 mmol/L EGTA) for 10 minutes then stained with rhodamine-conjugated phalloidin (Molecular Probes) and DAPI in the presence of 0.2% Triton-X-100 in Tris (pH 7.5) for 20 minutes. The wound was imaged as individual panels and combined into a montage for visualization of the entire wound using an Applied Precision CellWorx microscope with a fixed × 10 objective and 1 × 1 binning. Motility was quantified using ImageJ (MRI wound-healing tool) to determine the area remaining after closure. Average closure area was calculated for each time point and normalized as percentage of wound at T0 for each treatment.
Cell invasion assay
Invasion assays were carried out using 24-well format transwell chambers with 8 μm pore membranes (24). Cells were synchronized by serum starvation 24 hours prior plating to prevent proliferation during the assay. Cells were combined with matrigel were plated into the top of the transwells (Corning). Cells were allowed to migrate for 24 hours at 37°C and 5% CO2 (v/v). At this time, the cells remaining on the top surface of the membrane were wiped off with a cotton swab. Cells that invaded the membrane were fixed in methanol and stained with DAPI. Membranes were then mounted on slides and quantitated by microscopy. Each experiment was performed twice in triplicates, and five microscopic fields were counted per well.
OCCA xenograft studies
OCCA cells were transplanted into the right flanks of immunodeficient Balb/c nude mice (Nu/nu). Cell lines were grown in vitro, washed twice with PBS and resuspended in 50% Matrigel (BD Biosciences) in PBS. Mice were injected subcutaneously with 4.5 × 106 and 1.5 × 106 JHOC-5 and TOV21G cells, respectively, in 100 μL cell suspension and monitored at least twice weekly. Tumor volume was calculated using the equation: Volume = (width)2 × length/2. shRNA expression was induced by addition of doxycycline to drinking water (2 mg/mL in 2% sucrose). Survival studies measured the number of days for animals to reach the ethical tumor size limit and presented in the form of Kaplan–Meier plots. One week following doxycycline-mediated induction, mice were imaged using the Maestro2 (Cambridge Research & Instrumentation) automated imaging system to detect green and red fluorescent protein expression. Tumors were harvested once the Control group reached the ethical limit (1,000 mm3) for biomarker analysis. Samples were either snap frozen in liquid nitrogen or fixed buffered formalin and were paraffin embedded.
Sections from formalin-fixed paraffin-embedded tissues/fresh frozen, were stained according to standard protocols. Antibodies against Ki67 (Abcam), IL6 (Abcam), and CD31 (Novus biological) were used. Images were captured using an Olympus BX61 microscope.
Reverse phase protein array
The MD Anderson Cancer Center (Houston, TX) reverse phase protein array (RRPA) platform (25) was used to perform analysis on cell pellets from JHOC-5 and TOV21G cell lines, and a detailed procedure is provided in the Supplementary Methods section.
Western blot analysis
Whole-cell protein lysates were boiled, resolved by SDS-PAGE using 12%–15% (w/v) acrylamide gels, and then transferred to polyvinylidene difluoride membranes. Blots were blocked in 5% (w/v) nonfat milk powder or 5% (w/v) BSA in PBS-T (0.1% Tween in PBS) and probed overnight at 4°C in primary antibody. Membranes were washed in PBS-T and incubated with peroxidase-conjugated secondary antibody for 1 hour at room temperature, washed and developed by chemiluminescence before being exposed to radiographic film. Blots were reprobed with α-tubulin or β-actin antibodies to assess protein loading. A complete list of all primary and secondary antibodies is provided in (Supplementary Table S2).
Chemical library screen
The chemical library consisted of 463 compounds including 73 small-molecule inhibitors of oncogenes, 89 epigenetic modifiers, and 276 kinase inhibitors (Supplementary Table S3; ref. 26). A primary screen was conducted on single clonal wild-type (WT) and KO IL6 cell lines for both JHOC-5 and TOV21G using four-point titrations. Briefly, early passage cells were seeded into 384-well microtiter plates at 1,000 cells (JHOC-5) or 1,200 cells (TOV21G) per well using a multidrop dispenser (Thermo Fisher Scientific) in 40 μL of complete media. Cells were allowed to adhere overnight. All compounds (Compound Australia) were received in 100% DMSO and diluted to concentrations from 0.01 μmol/L to 12.5 μmol/L. Compounds were added directly to assay plates using a 384, hydrophobic slotted pin tool (V&P Scientific). Cells were exposed to the drugs for 48 hours, and viability measured using the CellTiter-Glo 2.0 luminescent assay (Promega) and EnVision multilabel plate reader (PerkinElmer). Average viability was normalized to DMSO control wells, and IC50 dose was approximated by fitting a four-parameter dose-response curve with R and Prism 6 (GraphPad). Primary screen hits were selected on the basis of a minimum of a 2-fold difference in IC50 between paired WT and IL6 KO cell lines. A 10-point titration secondary screen was performed, involving 56 hits evaluated against three independent WT and paired KO cell lines of both JHOC-5 and TOV21G. To identify compounds that showed efficacy with anti-IL6 therapy, an additional combination screen was performed involving the 56 hit compounds with a constant concentration of the anti-IL6 receptor monoclonal tocilizumab (100 μg/mL) in two independent clones, each of WT JHOC-5 and TOV21G cell lines.
Validation of secondary screen findings
A 10-point titration validation assay was performed with erlotinib (Sigma-Aldrich) and MK-2206 (SelleckChemicals) in three independent WT and paired KO cell lines of both JHOC-5 and TOV21G. Three independent WT for JHOC5 and TOV21G, as well as a panel of OCCA cell lines, were utilized for combination treatment with tocilizumab (Roche). Drug sensitivity to individual agents or combinations was determined using the CellTiter-Glo 2.0 Cell Proliferation Assay (Promega) as described previously (22). The IC50 dose was approximated by fitting a four-parameter dose-response curve using Prism 6 (GraphPad).
Data were analyzed with Student t test. Survival differences were compared using Kaplan–Meier log-rank analysis. Correlation between two groups was evaluated by the Pearson test. Statistical analyses were performed using Prism 6 (GraphPad) or R [version 3.5., R Core Team (2018)]. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria (https://www.R-project.org/) with P < 0.05 considered statistically significant.
In vitro growth is not altered following IL6 KO in OCCA cell lines
We chose JHOC-5 and TOV21G OCCA cell lines to study the requirement for IL6 in OCCA, as these have higher levels of IL6 expression and downstream targets compared with other lines (Fig. 1A). In addition, these lines have been nominated as useful OCCA models based on extensive biomarker characterization (17). We began by inhibiting IL6 signaling either using inducible shRNA of IL6; tocilizumab, a humanized monoclonal IL6 receptor antibody; or Pyridine 6, a small-molecule inhibitor of the JAK/STAT3 signaling.
We used a lentiviral vector system that allowed expression of shRNA under the control of a tetracycline-inducible transactivation/promoter system (Supplementary Fig. S1A; refs. 19, 20). Viral integrants were selected by FACS for mCherry-positive cells. Transduced and FACS-sorted clones were expanded, and we confirmed induction of shRNA expression (Supplementary Fig. S1B). After testing 10 shRNA constructs, we obtained two that moderately (shRNA-10) or strongly (shRNA-5) inhibited IL6 RNA and protein expression (Supplementary Fig. S1C–S1F). We also confirmed that tocilizumab attenuated pSTAT3Y705 phosphorylation moderately, a marker of pathway activation (Supplementary Fig. S2A), as did Pyridine 6, a small-molecule inhibitor of the JAK/STAT3 signaling in a dose-dependent manner (Supplementary Fig. S2B–S2E). We found in each of these models that in vitro proliferation was not significantly changed despite clear attenuation of IL6 levels and downstream signaling (Supplementary Fig. S2F and S2G).
IL6 is expressed in very high levels in JHOC-5 and TOV21G, and we reasoned that even low levels of protein or partial attenuation of IL6-R or STAT3 activation might be permissive for in vitro growth. We therefore inactivated both alleles of IL6 in each cell line by CRISPR/Cas9 (Fig. 1B; Supplementary Fig. S3A and S3B) and confirmed the IL6 KO by ELISA and next-generation sequencing (Fig. 1C–E). To control for random clonal variation influencing the interpretation of functional assays, at least five unique TOV21G and JHOC-5 IL6 KO clonal cell lines were analyzed (Supplementary Fig. S3C–S3E), as well as clonal WT control lines. Consistent with shRNA, antibody, and small-molecule inhibitor data, growth in vitro was not significantly altered in OCCA lines in which IL6 production was completely abrogated (Fig. 1F and G).
A partial requirement for IL6 in mediating migration and invasion in vitro
IL6 has a number of tumor-promoting properties including cell proliferation, survival, migration, invasion, promotion of angiogenesis, and inflammation. These processes are all hallmarks of cancer development and progression, affecting malignant epithelial and stromal cells in cancers such as ovarian cancer (27–30). Among these effects, cell migration and invasion can be measured in vitro using surrogate assays of wounding of monolayers and invasion through transwells. Cell migration was assessed in JHOC-5 cells using a scratch assay of monolayers (Fig. 2A). We monitored closure of the deficit over the next 16 hours, finding that migration was significantly impaired in IL6 KO clones compared with controls. Migration was partially restored by the addition of recombinant IL6 (recIL6) to KO cultures (Fig. 2B). Because TOV21G cells do not form uniform monolayers suitable for the scratch assay, we also performed transwell invasion assays in both cell lines (Fig. 2C and D). Consistent with our findings in the scratch assay, we found that JHOC-5 IL6 KO clones showed reduced transwell migration. In contrast, invasion was unaltered in TOV21G KO clones, suggesting that OCCA may not respond uniformly to IL6 signaling.
Conditional KD of IL6 in the OCCA xenograft model delays tumor growth
To assess IL6 dependencies of JHOC-5 and TOV21G in vivo, we transduced parental cells with our most robustly active conditional IL6 shRNAs, shRNA-5, shRNA-10. We selected bulk populations based on mCherry expression and expanded these rather than individual clones to avoid possible clonal variation confounding our results. We injected 4.5 × 106 and 1.5 × 106 JHOC-5 and TOV21G cells, respectively, into nude mice subcutaneously and then monitored the presence of the shRNA transgenes (mCherry) and their induction (tGFP) by fluorescence imaging one-week posttetracycline administration (Fig. 3A; Supplementary Fig. S4A). We compared the effect of reduced IL6 following induction immediately after implantation (D0) or when tumors were established (100 mm3) to noninduced controls. Tumor growth of JHOC-5 and TOV21G KD lines was significantly delayed with both IL6 shRNA constructs when induced at D0, whereas the NS shRNA construct had no delay in tumor growth. Growth was only significantly reduced in established tumors upon induction of the more potent shRNA-5 (Fig. 3B–D; Supplementary Fig. S4B–S4D). Animals demonstrated significant survival (number of days for animals to reach ethical tumor size limit) advantage with both shRNAs when IL6 KD was induced at D0, but only following potent shRNA KD in established tumors (Supplementary Fig. S4E–S4G). We performed a more detailed analysis of survival and tumor morphology with JHOC-5 KD clones. IHC with anti-IL6 antibodies confirmed substantial but not complete attenuation of IL6 expression in JHOC-5 lines in vivo following induction (Fig. 3E). Staining for the proliferative marker Ki67 also showed a reduction in actively cycling cells following induction (Fig. 3F). IL6 is known to promote tumor blood vessel formation in solid cancers, including ovarian cancer (14). Staining with the vascular marker CD31 showed a significant reduction in tumor vascular density in KD cells, particularly in tumors where IL6 KD was induced at implantation (Fig. 3G and H). Therefore, while OCCA tumor growth is dependent on IL6 in vivo, a therapeutic benefit of targeting it in established disease is only likely to be achieved with a substantial reduction of IL6 levels.
IL6 mediates the YAP signaling pathway in OCCA
To further investigate how IL6 influences OCCA growth, we analyzed downstream signaling of receptor engagement. The canonical IL6 signaling pathway involves IL6-mediated dimerization of its receptor and gp130, leading to JAK activation, phosphorylation of STAT3, and its translocation to the nucleus where it modulates gene expression (30). We analyzed protein lysates from cells following CRISPR/Cas9-mediated IL6 KO to determine whether baseline STAT3 phosphorylation was reduced. Our Western blot analysis showed a reduction in pSTAT3Y705 phosphorylation in TOV21G but not JHOC-5 clones (Supplementary Fig. S5A). Given our finding that JHOC-5 cells have an obvious requirement for IL6, we considered other signaling pathways that may mediate the effects of IL6.
We first performed RPPA capable of monitoring the levels of 220 proteins and 63 phosphoproteins (Supplementary Table S4; ref. 25). We compared JHOC-5 and TOV21G cells, including parental and multiple IL6 KO clones, and focused on proteins where there was at least a 0.25-fold change in signal intensity (Fig. 4A and B). Overall, we did not observe an alteration of proteins associated with IL6 canonical signaling pathway in either TOV21G or JHOC-5 KO cell line compared with their matching parental clones. Indeed, only a single protein, YB1S102, was differentially regulated in IL6 KO clones of TOV21G and JHOC-5 relative to the parental lines, suggesting that IL6 may have a different mechanism of action in different OCCA cell line models (Supplementary Table S5).
Recently, a noncanonical STAT3-independent IL6 signaling pathway has been described in intestinal epithelial cells in which receptor binding by IL6 engages gp130 and interacts with and activates Src and Yes (15). Src and Yes bind YAP through their conserved SH3 domain and activate YAP through Y357 phosphorylation, thereby stabilizing the protein and increasing its nuclear concentration (15, 31, 32). We hypothesized that protumorigenic actions of IL6 in OCCA may be mediated through the noncanonical STAT3-independent pathway. In the RPPA data, we also observed a trend toward the downregulation of pSrc in both KO cell lines compared with parental, particularly in JHOC-5 (Supplementary Fig. S5B and S5C). We further tested this hypothesis by Western blot analysis, analyzing the regulation of Src and YAP in the IL6 WT and KO clones (Fig. 4C).
We extended our RPPA and Western blot data by performing a phosphokinase array on WT and IL6 KO JHOC-5 clonal lines. We found that the activation status of YAP pathway proteins were consistent with the RPPA data (Supplementary Fig. S5D and S5E), and extended to include downregulation of pYES426 in IL6 KO cells (15). We additionally validated our findings by demonstrating a dose-dependent reduction in phosphorylation of these proteins in WT cells following treatment with tocilizumab (Fig. 4D) and phosphorylation of YAP and Src following the addition of recIL6 to JHOC-5 KO cells (Fig. 4E).
We have previously shown that YAP expression correlates with poor patient prognosis in epithelial ovarian cancer samples, particularly OCCA, where it is a prognostic indicator independent of stage (16). Upon activation, YAP translocates to the nucleus to regulate the expression of target genes (32). Given our previous studies and current work, we considered whether IL6 expression and YAP subcellular localization were related in patients with OCCA. Using IHC and IL6 mRNA expression data from the Australian Ovarian Cancer Study (9, 16), we found a trend toward greater YAP nuclear localization with higher IL6 levels in patients with OCCA (Supplementary Fig. S6).
To investigate a functional role for YAP in JHOC-5 cells, we generated clones transduced with a vector constitutively expressing shRNA to YAP and validated their ability to reduce YAP protein expression (Fig. 5A). Consistent with our IL6 KO data, suppression of YAP protein in JHOC-5 did not impact cell proliferation (Fig. 5B) but did significantly impair migration, as assessed in scratch and transwell assays (Fig. 5C–E). These findings and our previous data suggest that IL6 mediates its protumorigenic actions via a noncanonical IL6 signaling pathway involving gp130/pSrc/YAP.
Targeting the IL6 pathway indirectly is a potential therapeutic strategy in patients with OCCA
Although IL6 KD had demonstrable effects on the growth of JHOC-5 tumors, we reasoned that targeting IL6 in combination with chemotherapy agents or downstream signaling molecules would be required in patients to enhance therapies directed to IL6. This view is consistent with recent limited clinical activity of the anti-IL6 monoclonal siltuximab in an ovarian cancer clinical trial that included one patient with OCCA (14). We performed pilot experiments in JHOC-5 IL6 KO cell lines (n = 3) and TOV21G IL6 KO cell lines (n = 3) with cisplatin or paclitaxel but did not see any increased sensitivity with depletion of IL6. We therefore conducted a high-throughput compound library screen of JHOC-5 and TOV21G lines, seeking molecules that specifically increased killing in IL6 KO versus WT clones. The library consisted of 463 compounds (Supplementary Table S3), including kinase inhibitors (N = 276), epigenetic regulators (N = 89), and other agents to oncogenic proteins (N = 73). The screen, depicted schematically in (Fig. 6A), involved an initial four-point titration assay for each compound on IL6 KO and WT JHOC-5 and TOV21G clones. Data were normalized to DMSO. A total of 81 primary screen hits were identified: 39 and 42 for TOV21G and JHOC-5, respectively (Supplementary Table S6 and S7). These were reduced to 56 compounds (Supplementary Table S8) for a secondary screen, based on their strength and specificity of effect, consistency between cell lines, RPPA data, and/or the current status of compounds with respect to Federal Drug Agency approval.
Compounds were assayed in a 10-point titration curve to obtain more detailed dose-response data, and we increased the number of IL6 KO and WT clones for each cell line (Fig. 6A). In addition, to more closely resemble an approach in patients, we also screened the 56 compounds in WT JHOC-5 and TOV21G cells in the presence of 100 μg/mL tocilizumab. There was considerable variation in response depending on cell line or the method of IL6 inhibition (Fig. 6B and C). Five compounds appeared to enhance the effects of IL6 inhibition on JHOC-5 cells, independent of the means of IL6 inhibition: BEZ-235 and GDC-0941, which target PI3 kinase, MK-2206, which inhibits AKT, and gefitinib and vandetinib that target EGFR pathway and other protein tyrosine kinases (Fig. 6D). The data obtained in TOV21G KO and anti-IL6 antibody-treated cells were highly consistent and involved a number of histone deacetylase and HSP90 inhibitors (Fig. 6E). Importantly, common hits with both TOV21G and JHOC-5 cell lines included AKT and EGFR inhibitors. These data are consistent with fact that the TOV21G cell line has a PI3 kinase–activating mutation (33), hence its increased sensitivity to AKT inhibitors, particularly in combination with the inhibition of IL6. In addition, IL6 KO in JHOC-5 cell lines leads to upregulation of AKT phosphorylation, as shown by the RPPA data (Fig. 4B; Supplementary Fig. S7). Previously, we and others have shown that attenuation of IL6 signaling or downregulation of YAP expression sensitize ovarian cancer cells to EGFR inhibitors (34, 35). Therefore, we mainly focused on two inhibitors, MK-2206 and erlotinib, directed against AKT and EGFR, respectively, to further validate our secondary screen findings. Validation studies were performed on a larger number of clonal IL6 KO and parental clones for both cell lines (Fig. 6F–I). In addition, to demonstrate that these effects are not specific to these cell lines, we performed the combination treatment of these inhibitors and tocilizumab (100 μg/mL) in additional OCCA cell lines. The increased sensitivity of JHOC-5 to a combination of tocilizumab and erlotinib was recapitulated in ES-2 with trends to greater sensitivity in TOV21G and JHOC-7, and to tocilizumab and MK-2206 in JHOC-7, and trend in TOV21G (Fig. 6J and K).
Among ovarian cancer histotypes, IL6 production is particularly prominent in OCCA (8–10), and IL6 is associated with nonneoplastic systemic effects that accompany the disease, including thrombocytosis and vascular thrombosis (36). In mice, ARID1A and PIK3CA mutations cooperate to form highly penetrant and aggressive tumors that closely resemble human OCCA and whose growth is supported by IL6 overexpression (11). Here we sought to validate IL6 as a therapeutic target in human OCCA by characterizing the effects of attenuation, its mode of signaling, and exploring possible drug combinations.
Our studies made use of TOV21G and JHOC-5 as examples of OCCA (17). TOV21G has several pathogenic mutations commonly observed in OCCA patient samples (PIK3CA, PTEN, KRAS, CTNNB1, ARID1A; refs. 17, 33). In comparison, JHOC-5 lacks these driver mutations but has a high level of MET amplification and copy-number gain for HNF1B. Amplification of MET is a feature of OCCA compared with other ovarian cancer histotypes (9, 37, 38). In addition to variation in driver mutations and copy-number change, OCCA have been subtyped on the basis of mutational signatures (39). Despite these advances, it remains the case that to date only small numbers of OCCA have been characterized with high-resolution genomic techniques compared with high-grade serous ovarian cancer, where hundreds of tumors have been analyzed (40). The range of OCCA cell lines required to represent tumor subtypes therefore remains to be determined.
Given the very high levels of IL6 produced by OCCA tumors and cell lines, we used orthogonal approaches to attenuate IL6 production/signaling, including conditional KD, an inhibitory mAb, a small-molecule inhibitor of the JAK/STAT3 signaling, and CRISPR/Cas9-mediated gene KO. Our in vitro findings indicate that while OCCA cell lines are vulnerable to a reduction in IL6, effects appear to be mediated mainly through reduced migration rather than a direct effect on cell proliferation. These findings are consistent with previous studies, where partial inhibition of IL6 signaling either by siRNA or antibodies against IL6/IL6-R have been shown to impair migration and invasion but with no impact on cell growth (14, 41). A major caveat of these studies has been that the partial attenuation of IL6 signaling might not be sufficient to alter cell growth in vitro. However, our in vitro finding with CRISPR/Cas9-mediated KO of IL6 removes that caveat and supports the notion that the effect of IL6 on tumor growth in vivo is indirect. Our finding of reduced tumor growth of OCCA lines with impaired IL6 expression is consistent with observations made using anti-IL6, siltuximab (14). Previous studies have shown that IL6 plays a critical role in the tumor microenvironment by promoting angiogenesis in many solid cancers, including ovarian, and influence tumor growth (14, 27, 42). We also observe that IL6 is required for vascular recruitment and show that targeting of IL6 is most likely to be effective in a low volume rather than in bulky, established disease. Therefore, targeting IL6 therapeutically in OCCA may require early, prolonged, and substantive inhibition of IL6.
We performed RPPA analysis to identify more effective ways of targeting IL6 dependencies in OCCA, characterize downstream signaling from IL6, and identify adaptive/compensatory pathway activation when IL6 is abolished. Our data demonstrate that elevated IL6 levels might be common in all OCCA cell line models, but the protumorigenic effects and its mechanism of action may differ between cell lines, and by extension, between different patients with OCCA. Interestingly, our data demonstrate that IL6 has different mechanisms of action in the two OCCA cell lines. Knocking out IL6 in TOV21G moderately downregulated the canonical STAT3-dependent IL6 signaling as shown previously (14). In contrast, in JHOC-5, IL6 KO downregulated a noncanonical STAT3-independent IL6 signaling pathway. These findings are consistent with a recent study in intestinal epithelial cells, describing receptor binding by IL6 engages gp130 and interacts with and activates Src and YAP (15). Involvement of YAP in IL6 signaling in OCCA is consistent with our earlier observation of nuclear YAP levels being associated with poor outcomes, particularly in OCCA (16). We extended those findings by documenting a partial association between nuclear YAP and the level of IL6 production in the AOCS patient cohort (16). The observation that the noncanonical pathway was affected by IL6 KO in JHOC-5 but not TOV-21G additionally illustrates the intrahistotype heterogeneity of OCCA, and this should be addressed if targeting the IL6/gp130/YAP axis is to be considered.
To date, targeting the IL6 signaling pathway directly has not been very effective in preclinical or clinical settings. Previously, the IL6 inhibitory mAb siltuximab provided limited efficacy in a phase II trial involving 20 patients with ovarian cancer, including one with OCCA (14). These observations accord with the limited efficacy of IL6 attenuation in our in vivo models. More recently, upregulation of EGFR signaling has been demonstrated in high-grade serous ovarian cancer in response to neutralizing IL6 antibodies (35). Coordinated upregulation of the EGFR/PI3K/AKT and IL6 signaling pathways have also been observed in lung cancer following the loss of the candidate tumor suppressor ING, with an increase in epithelial–mesenchymal markers (43). In addition, KD of YAP has been shown to sensitize cancer cells to various anticancer agents, such as cisplatin and the EGFR tyrosine kinase inhibitor erlotinib (36). On the basis of this body of work and our findings, we performed high-throughput drug screening in a background of IL6 attenuation. Similar to our in vitro and biochemical studies, there are considerable variations in the list of compounds that attenuate TOV21G and JHOC5 cell lines in combination with IL6 inhibition. However, we find that blocking IL6 signaling sensitizes both cell lines to EGFR and AKT inhibition. These data are consistent with fact that the TOV21G cell line has a PI3 kinase–activating mutation (34). In addition, IL6 KO in JHOC-5 cell lines leads to upregulation of AKT phosphorylation, as shown by the RPPA data. While we have extended our combination studies (IL6 attenuation with EGFR and AKT inhibition) to three additional OCCA lines, further in vivo studies are required to validate and support these findings.
In summary, our findings validate IL6 as a therapeutic target in OCCA, identify a noncanonical IL6 signaling pathway, and suggest that targeting IL6 signaling indirectly in OCCA may be the most efficacious therapeutic option, particularly in well-established OCCA disease. Further preclinical experiments are needed to explore the consistency of the findings across OCCA lines and the efficacy in vivo.
Disclosure of Potential Conflicts of Interest
W.J. Azar reports grants from National Health and Medical Research Council of Australia (NHMRC APP1044447 and APP631701) during the conduct of the study. E.l. Christie reports grants from National Health and Medical Research Council of Australia and grants from Victorian Cancer Agency during the conduct of the study. G. Au-Yeung reports grants from AstraZeneca and Roche/Genentech outside the submitted work. D.D.L. Bowtell reports grants from AstraZeneca, Genentech Roche, Beigene and personal fees from Exo Therapeutics outside the submitted work. No potential conflicts of interest were disclosed by the other authors.
W.J. Azar: Conceptualization, formal analysis, validation, investigation, visualization, writing-original draft. E.L. Christie: Formal analysis, visualization, writing-review and editing. C. Mitchell: Investigation. D.S. Liu: Conceptualization. G. Au-Yeung: Conceptualization. D.D.L. Bowtell: Conceptualization, supervision, funding acquisition, writing-original draft, writing-review and editing.
This work was supported by the National Health and Medical Research Council of Australia (NHMRC APP1044447, APP1117044, and APP631701). The Australian Ovarian Cancer Study was supported by the U.S. Army Medical Research and Materiel Command under DAMD17-01-1-0729, The Cancer Council Victoria, Queensland Cancer Fund, The Cancer Council New South Wales, The Cancer Council South Australia, The Cancer Foundation of Western Australia, The Cancer Council Tasmania and the National Health and Medical Research Council of Australia (NHMRC; ID400413, ID400281). The AOCS gratefully acknowledges additional support from Ovarian Cancer Australia and the Peter MacCallum Cancer Centre Foundation. We gratefully acknowledge additional support from Mrs. Margaret Rose AM and the Rose family, The WeirAnderson Foundation, Border Ovarian Cancer Awareness Group, Wendy Taylor and Arthur Coombs and family. E. Christie was supported by the Victorian Cancer Agency and the National Health and Medical Research Council of Australia.
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.