Abstract
Acquired and intrinsic resistance to receptor tyrosine kinase inhibitors (RTKi) represents a major hurdle in improving the management of clear cell renal cell carcinoma (ccRCC). Recent reports suggest that drug resistance is driven by tumor adaptation via epigenetic mechanisms that activate alternative survival pathways. The histone methyl transferase EZH2 is frequently altered in many cancers, including ccRCC. To evaluate its role in ccRCC resistance to RTKi, we established and characterized a spontaneously metastatic, patient-derived xenograft model that is intrinsically resistant to the RTKi sunitinib, but not to the VEGF therapeutic antibody bevacizumab. Sunitinib maintained its antiangiogenic and antimetastatic activity but lost its direct antitumor effects due to kinome reprogramming, which resulted in suppression of proapoptotic and cell-cycle–regulatory target genes. Modulating EZH2 expression or activity suppressed phosphorylation of certain RTKs, restoring the antitumor effects of sunitinib in models of acquired or intrinsically resistant ccRCC. Overall, our results highlight EZH2 as a rational target for therapeutic intervention in sunitinib-resistant ccRCC as well as a predictive marker for RTKi response in this disease. Cancer Res; 77(23); 6651–66. ©2017 AACR.
Introduction
Receptor tyrosine kinase inhibitors (RTKi), such as sunitinib, are presently used as standard of care in patients with recurrent metastatic renal cell carcinoma (RCC), but durable responses are not common (1, 2). The clinical benefit of RTKi in RCC is thought to be mediated primarily by their antiangiogenic effect, especially in VEGF-driven tumors such as ccRCC. The mechanism(s) by which sunitinib harnesses angiogenesis is by targeting several RTKs, including VEGF receptor 2 (VEGFR2) and platelet-derived growth factor receptor (PDGFR). However, recent reports from our group and others have suggested that the antitumor effects of RTKi, such as sunitinib, may be in part independent of their antiangiogenic effect (3–5).
Induction of multiple kinases, such as FAK, SCR, MET, FGFR2, EGFR, IGF-1R, and ERBB2, has been reported to be associated with resistance to RTKi due to either transcriptional activation or posttranslational upregulation (6–8). Induced kinome reprogramming can occur via deregulation of feedback loops and cross-talk of regulatory nodes, which in turn fosters the ability of cells to bypass the drug-inhibitory effect leading to resistance (9). The kinome network can be affected by several factors including acquired mutations, stromal interactions, and epigenetic modifications (10). However, there are no reports on the potential role that specific epigenetic modifications may play in reprogramming the kinome following RTKi.
Epigenetic modifications have been implicated in cancer progression and are potential drivers of drug resistance (11–16). The overexpression of EZH2 has been reported in numerous cancer types, including advanced RCC (17–19), suggesting its role in modulating several cellular processes involved in cell survival and drug resistance (20, 21). Inhibition of EZH2 has resulted in the attenuation of drug resistance in tumor and stem cells by suppressing its repressive function on target tumor suppressor genes (22). However, little is known about the epigenetic mechanisms of resistance to RTKi in RCC. We have recently reported that sunitinib resistance may be transient and be reversed by dose escalation both in selected patients with RCC and tumor models, suggesting a dynamic tumor adaptation to RTKi that is likely driven by epigenetic modifications (3). The acquired resistance in the preclinical models was associated with the dynamic/reversible changes of EZH2 expression at the time of response and resistance to sunitinib. Thus, the onset of acquired resistance to sunitinib and progression may be in part due to alterations of EZH2 profiles of these tumors that provide a selective advantage to evade targeted kinase inhibition.
The approval of RTKi for advanced RCC has revolutionized the clinical treatment of this disease, although therapeutic responses are in general short-lived. Understanding the mechanisms of resistance through the establishment of patient-derived xenograft (PDX) models has improved our understanding of target predictions and drug response (23–28). Herein, we report the role of EZH2 in adaptive sunitinib-induced kinome reprogramming in a novel, clinically relevant, metastatic PDX model of clear cell RCC (ccRCC), RP-R-02LM. Also utilizing a human RCC cell line (acquired resistance), we report that sunitinib resistance is associated with increased EZH2 expression and induces global phosphorylation of kinases in both serine and tyrosine residues. Furthermore, molecular and pharmacologic inhibition of EZH2 in both cells lines and PDX, respectively, attenuated the global kinase phosphorylation, increased activation of tumor suppressors and consequently reestablished sensitivity to sunitinib. Taken together, these results suggest that pharmacologic targeting of EZH2 is a promising strategy to overcome RTKi resistance in RCC.
Materials and Methods
Xenograft studies
Xenograft models.
RP-R-01 and RP-R-02 are PDXs of ccRCC models previously described (3, 29). RP-R-02LM is a metastatic ccRCC that spontaneously metastasizes to the lungs from primary tumors implanted either subcutaneously under the skin or orthotopically in the kidney subcapsule.
Tumor implantation.
All in vivo experiments were approved and performed in strict accordance with the guidelines of the Institutional Animal Care and Use Committee (IACUC) at Roswell Park Cancer Institute (RPCI; Buffalo, NY) and IACUC Indiana University (IU; Indianapolis, IN). Six-week-old homozygous ICR SCID female mice were housed in a sterile, pathogen-free facility and maintained in a temperature-controlled room under a 12-hour light/dark schedule with water and food ad libitum. To generate the metastatic model, RP-R-02 tumors were implanted ectopically into the prostate of mice to select for a metastatic population. Lung metastasis developed were collected and reimplanted either subcutaneously under the skin flank or orthotopically in the kidney subcapsule to select for pure metastatic population. For drug treatment studies, RP-R-02LM viable tumors were selected and dissected into approximately 1-mm2 tumor pieces and implanted either subcutaneously or orthotopically into mice. All mice were operated under sedation with oxygen, isoflurane, and buprenorphine. When tumors were established and reached 50 mm2 (subcutaneous implantation), were palpable (orthotopic implantation), or metastases were detected by imaging (micro-CT), mice were randomly grouped and placed in either control group or treatment groups (n = 5–20).
Drug treatment schedule.
For the endpoint studies using sunitinib, bevacizumab, or axitinib, mice implanted with RP-R-02LM subcutaneously (n = 5–10/group) were randomly grouped into either control or treatment groups. Mice in the sunitinib treatment arm were treated with 40 mg/kg 5 days on/2 days off, mice in the bevacizumab treatment arm were treated with 10 mg/kg two times a week, and mice in the axitinib treatment arm were treated with 36 mg/kg two times a week. For survival studies using sunitinib, we implanted RP-R-02LM tumor pieces (∼1 mm2) into the kidney subcapsule of 20 mice. One month after implantation, mice bearing tumors were randomly grouped into control and sunitinib treatment groups. Mice in treatment groups were treated with a dose of 40 mg/kg (sunitinib-free base; LA Laboratories) 5 days on/2 days off by oral gavage. In a second set of experiments, mice implanted with RP-R-02LM tumor pieces were grouped into control, sunitinib-treated group, EPZ011989 (Epizyme)-treated group, or a combination group (n = 5/group). Mice in the sunitinib group were treated with 40 mg/kg sunitinib 5 days on/2 days off; mice in the EPZ011989-treated group received 500 mg/kg EPZ011989 2×/day, 5 days on/2 days off; and the combination group received both drugs at the same dose and same treatment time.
Tumor and body weight assessment.
Tumor size and body weights were assessed and recorded once a week. For tumors implanted subcutaneously, tumor size was measured once a week by caliper measurement of 2 diameters of the tumor (L × W = mm2) and reported as tumor volume [(L × W2)/2 = mm3]. Body weights were assessed using a weighing scale and recorded in grams. Endpoint tumor weights were assessed using a weigh scale and recorded in grams.
Blood and tissue collection
Tissue and blood were collected under aseptic conditions. One milliliter of blood was collected by cardiac bleeds (terminal) at the end of the experiment. Serum and plasma were separated, and aliquots were stored at −80°C for further analysis. Tumor tissues were excised, weighed, and cut into sections. Sections were snap-frozen and stored in −80°C, fixed in 10% buffered formalin or zinc for histopathology, and saved in TRIzol for RNA analysis.
Imaging studies
MRI.
Experimental MRI examinations were performed as described previously (30). In brief, experiments were conducted in a 4.7T/33-cm horizontal bore magnet (GE NMR Instruments) equipped with AVANCE digital electronics (Bruker Biospec, Paravision 3.0.2; Bruker Medical Inc.), a removable gradient coil insert (G060) generating a maximum field strength of 950 mT/m and a custom-designed 35-mm RF transmit receive coil. Animals were secured in a form-fitted, MR-compatible sled (Dazai Research Instruments) equipped with temperature and respiratory monitoring sensors. Animal body temperature was maintained at 37°C during imaging using an air heater system (SA Instruments), and automatic temperature feedback was initiated through thermocouples in the sled. A phantom containing 0.15 mmol/L gadopentetate dimeglumine (Gd-DTPA; Magnevist; Berlex Laboratories) was used for monitoring changes in noise and system performance. Preliminary localizer images were acquired for subsequent slice prescription. Tumor volumes were calculated from multislice T2-weighted (T2W) spin echo images with the following parameters: field of view (FOV) = 3.20 × 3.20 cm, matrix (MTX) = 256 × 192, slice thickness = 1 mm, NEX = 4, TR = 2500 ms, TEeff = 41.0 ms, and RARE/Echoes = 8/8. Following image acquisition, raw image sets were transferred to a processing workstation and converted into Analyze format (AnalyzeDirect, version 7.0). All postprocessing of imaging data was carried out in Analyze and MATLAB. A region of interest (ROI) was manually traced around the entire tumor area on each tumor slice. Tumor volume was calculated by measuring the cross-sectional area on each slice and multiplying their sum by the slice thickness.
Micro-CT imaging.
Respiratory gated micro-CT (μCT) was performed at 33.63 μm on a SkyScan 1176 (Bruker) using the following scan parameters: 80 kV tube voltage; 313 mA tube current; 0.5 mm Al filter; 0.9° rotation step; no frame averaging, 4 × 4 binning, and 32 horizontal pixel overlaps. Animals were induced and maintained with 4%–5% and 1%–3% isoflurane gas, balance medical oxygen, delivered at 2 mL/minute, respectively. Animals were transferred to a heated carbon fiber scanning bed and landmarked, and scout scans were performed to permit scan prescription of thoracic region in two beds. Images were reconstructed using a vendor-supplied filtered back-projection algorithm, using a Hamming filter (α = 0.54) with a Nyquist relative cutoff frequency of 100. Postacquisition, DICOM images were imported into Analyze (AnalyzeDirect, version 12.0) and registered using normalized entropy methods (31), and intensity normalized over the interval (0.0–1.0). Images were then segmented using constrained region growing approaches, where lung air spaces and soft tissues were defined to have values ranges of (0.0–0.1) and (0.11–1.0), respectively. To minimize the impact of tracheal tissue contamination on total lung volume, trachea objects were digitally removed, and object maps quantified according to the following general equation:
Where , i, j, k, l, a, b, c, |{O_{Map}}$|, and |\mu $| are the volume of the “i'th” region, “j'th” object map index, “k'th” object map index, “l'th” object map index, “a” total number of row voxels, “b” total number of column voxels, “c” total number of slices, object map, and voxel dimension in millimeters. Determination of lung nodules was then assessed as the reduction in lung volume for each subject with time.
In vitro assays
Cell lines.
The 786-0 and UMRC2 cell lines from ATCC were maintained and cultured in the appropriate media, supplemented with 10% FBS and 1% penicillin and streptomycin. The 786-0R (resistant cell line) was generated from the 786-0 parental line, adopting a drug-escalation scheme. The 786-0 cells were exposed to an initial concentration of 2 μmol/L that was the concentration at which 50% of cell death was observed. When cells became resistant at 2 μmol/L, drug concentration was gradually increased by 0.5 μmol/L increments with every resistant cycle. Cells were considered resistant at ≥5 μmol/L sunitinib. UMRC2R cells were established using the same method. The period for generated resistant cells from the parental line was approximately 4 months. The IC50 at 48-hour treatment of the 786-0 and UMRC2-resistant cells was 19.9 μmol/L and 20 μmol/L, respectively. All cells are routinely tested and checked for the absence of Mycoplasma. Cells were also checked to ensure that they maintained human origin by detecting the Alu sequence via PCR.
MTT and crystal violet assay.
786-0 and 786-0R cells were plated in 6-well plates for 24 hours and treated with either DMSO (vehicle control), varying concentrations of sunitinib, or a single concentration of sunitinib in combination with EPZ011989 (EZH2 inhibitor). The MTT assay (Cayman Chemical) or Crystal violet assay (Sigma) was used to evaluate the cells' growth after 24, 48, 72, and 96 hours of treatment, and absorbance was read using a spectrometer (xMark spectrometer; Bio-Rad Laboratories).
IHC and immunofluorescence staining
Tissue specimens were fixed for 24 hours, paraffin embedded, and sectioned (4 μm). Sections were deparaffinized and rehydrated through graded alcohol washes. Antigen unmasking was achieved by boiling slides in either sodium citrate buffer (pH = 6.0) or EDTA. For IHC staining, sections were further incubated in hydrogen peroxide to reduce endogenous activity. To examine the expressions of our proteins of interest, tissue sections were blocked with 2.5% horse serum (Vector Laboratories) and incubated overnight in primary antibodies against Ki67 (1:500; Thermo Fisher Scientific), E-cadherin (1:1,000; Cell Signaling Technology), EZH2 (1:1,000; Cell Signaling Technology), H3K27me3 (1:1,000; Cell Signaling Technology) and CD31 (1:100; Dianova). Following primary incubation, tissue sections were incubated in horseradish-conjugated anti-rabbit or anti-rat antibody according to manufacturer's protocol (Vector Laboratories), followed by enzymatic development in diaminobenzidine (DAB) and counterstaining in hematoxylin. Sections were dehydrated and mounted with cytoseal 60 (Thermo Scientific). For immunofluorescence staining (IF), sections were blocked with 5% BSA (Sigma), costained with EZH2 and either E-cadherin (1:500; Cell Signaling Technology), pTyrosine (1:400; BioVision), pSerine (1:100), pAKT (1:400; Cell Signaling Technology), or pFAK (1:500; Cell Signaling Technology), and incubated overnight at 4°C. Following primary incubation, sections were incubated with either Alexa Fluor or FITC fluorochrome–conjugated anti-rabbit (Thermo Fisher) or anti-mouse (1:400; Thermo Fisher) antibody at room temperature in a humid light-tight box. Afterward, slides were counterstained with DABI and mounted with Vectashield mounting medium (Vector Laboratories). Stained sections were analyzed either under bright field (IHC) or under appropriate fluorescence wavelength (IF) using the EVOS FL cell imaging microscope (Life Technologies) and Leica confocal microscope. The number of positive cells was determined in a blinded fashion by analyzing four random 20× fields per tissue and quantified using ImageJ software.
Chromatin immunoprecipitation-sequencing
Chromatin immunoprecipitation (ChIP) was performed as reported previously (32) using 1% formaldehyde fixing and antibodies to EZH2 (5246, lot 7; Cell Signaling Technology) or H3K27me3 (9733, lot 8; Cell Signaling Technology). For sequencing, purified ChIP DNA pooled from at least three replicates was resheared to <150 bp in a Diagenode Bioruptor. Libraries were prepared by Truseq sample preparation (Illumina). Libraries were sequenced with NextSeq 500 (Illumina). Sequencing reads from ChIP and input controls were aligned to the human genome (hg19) and bound regions identified using Useq (33).
RNA-sequencing
For RNA sequencing (RNA-seq) and analysis, tumor tissues and cells were collected to extract total RNA including small RNA using AllPrep DNA/RNA/miRNA Universal Kit (Qiagen) according to the manufacturer's protocol. RNA-seq data were demultiplexed and aligned against the human genome (hg19) using the TopHat alignment tool. The resulting file was aligned to BAM-formatted sequence alignment map utilized by the Cufflinks program to estimate transcript abundance. Cufflinks generates FPKM (fragments per kilobase of exon per million fragments mapped) values for every transcript, allowing comparisons between samples. Differential expressed transcripts were then identified between untreated and treated samples with Cuffdiff. Gene set enrichment analysis using Hallmark and Oncogenic Signatures was performed with a preranked gene list from the RNA-seq differential analysis. The genes were ranked on the basis of the square root of the sum of squares for the log2 (fold change) and the magnitude of the difference between the fold changes under the two analyzed conditions. A weighting factor was multiplied with the ranking score if the treatment returned the gene toward the expression under the normal condition. For miRNA sequencing analysis, Illumina sequencing reads were demultiplexed, trimmed for adaptor sequence, and aligned to the human genome with Bowtie. The aligned reads were then mapped to miRBase and annotated miRNAs tabulated. Differentially expressed miRNAs was determined using the DESeq R package with parametric normalization at FDR < 10%.
Quantitative RT-PCR
Quantitative RT-PCR (qRT-PCR) was performed utilizing EZH2, E-cadherin, SETD2, ALDH1A, HIF2a, E2F1, N-cadherin, ZEB, VASH1, SNAIL, and GAPDH human-specific primers (Integrated DNA Technologies; sequence in Supplementary Table S1). The denaturation step was carried out at 95°C for 10 seconds, the annealing step was carried out at 58°C for 30 seconds, and the extension step was carried out at 72°C for 1 minute using the Applied Biosystems 7900HT Fast Real-Time PCR System.
Time-correlated single photon counting–fluorescence lifetime imaging of peptide biosensor phosphorylation
For peptide synthesis, the experiments were performed as described previously (34). 786-0 and 786-0R cells were plated on coverslips in 12-well plates for 12 hours and treated with sunitinib or DMSO for 24 hours. Afterward, wells were washed gently three times with PBS and incubated with the peptide biosensors of interest, p-VEGF-R, p-AKT, p-SCR, and p-FAK. After incubation of live cells with multiplex peptide sensors, cells were imaged using time-correlated single photon counting (TCSPC)-fluorescence lifetime imaging (FLIM) to detect multiple kinase phosphorylation activity. Quantitative analysis was performed using photon decay collected from TCSPC-FLIM and fitted with a multiexponential model employing the Levenberg–Marquardt algorithm to construct a lifetime histogram and FLIM map.
Immunoblotting
Whole-cell protein extracts from tissue and cell were denatured, separated on SDS-PAGE gels, and transferred to nitrocellulose membranes. After blocking in 5% enhanced blocking agent (GE) in Tris-buffered saline–Tween, membranes were probed overnight at 4°C with the following primary antibodies: EZH2 (1:1,000 dilution; Cell Signaling Technology), E-cadherin, and β-Actin (1:1,000; Santa Cruz Biotechnology). After incubation with the appropriate secondary antibody, results were detected using Western Lightning Chemiluminescence Reagent Plus according to the manufacturer's instructions (Thermo Fisher Scientific) and captured on film. Quantitative measurements of Western blot analysis were performed using ImageJ and GraphPad software (Prism 7).
Phospho-RTK array
Phosphorylation levels of multiple RTKs were examined using the Proteome Profiler Human Phospho-RTK Array Kit (product number ARY001B; R&D Systems) following the manufacturer's guidelines.
Reverse-phase protein array
Reverse-phase protein array (RPPA) was performed for 12 RP-R-01 PDX tumors in the following order: control (n = 3), sunitinib-sensitive period (n = 3), and sunitinib-resistant period (n = 3). Slides were stained using antibodies as listed (https://www.mdanderson.org/education-and-research/resources-forprofessionals/scientific-resources/core-facilities-and-services/functional-proteomics-rppa-core/index.html), and relative protein levels were determined by fitting each dilution curve with a logistic model (SuperCurve fitting; http://bioinformatics.mdanderson.org/OOMPA), heatmap, and statistical analysis [generated using Partek Genomics Suite software (Partek Inc.)].
Phosphoproteomic mass spectroscopy
Cells from sunitinib-sensitive and -resistant cell lines were lysate and protein denatured in 0.1% RapiGest (Waters) and further reduced with 5 mmol/L dithiothreitol for 30 minutes at 50°C. Afterward, denatured proteins were alkylated in 15 mmol/L iodoacetamide for 1 hour in the dark and then digested with proteomics grade trypsin at a 1:100 ratio overnight at 37°C. The pH was adjusted below 3, and the sample was incubated for 45 minutes at 37°C. Samples were centrifuged at 16,000 × g to remove RapiGest, and the supernatant was collected. Enrichment of phosphopeptide, mass spectrometric data acquisition, and data analysis are described previously (35).
Clinical specimens and tissue microarray
Tumor specimens from 101 patients with ccRCC were embedded in paraffin, and two established tissue microarrays (TMA) were provided by the clinical data network RPCI and the IU Simon Cancer Center Tissue Bank (Indianapolis, IN). Treatments after nephrectomy include tyrosine kinase inhibitor, VEGF blocker, or mTORC inhibitors. For the overall survival (OS), 76 metastatic (RPCI) and 25 primary (IU) tumors were analyzed. For the RTKi-specific survival, 14 metastatic (RPCI) and 19 primary (IU) tumors from patients who received either sunitinib or pazopanib were analyzed. Slides stained for EZH2 (1:500; Cell Signaling Technology) were analyzed by two pathologists, and generated H-scores were used to define EZH2 status.
Statistical analysis
Data analyses are expressed as the mean ± SEM. Statistical significance where appropriate was evaluated using a two-tailed Student t test when comparing two groups or by one-way ANOVA using the Student–Newman Keuls post test for multiple comparison. A P < 0.05 was considered significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant. Statistical analyses were done by GraphPad software.
Results
Establishment and characterization of a novel metastatic ccRCC PDX model
To establish a metastatic ccRCC PDX model, we ectopically implanted tumor pieces from the PDX model RP-R-02 (3, 34) into the anterior prostate of ICR SCID mice based on our prior experience with prostate cancer PDX that metastasize to the lungs at high frequency following intraprostatic implantation (data not shown). All mice implanted with the tumors developed lung metastases four months after tumor implantation. Thus, we passaged the tumors from the lungs into a new set of mice. The newly established metastatic PDX model was named RP-R-02LM. Surprisingly, RP-R-02LM formed spontaneous lung metastases (Fig. 1A and B; Supplementary Fig. S1A) irrespective of the site of tumor implantation (subcutaneous or renal capsule). RP-R02LM tumors at the primary site, as well as the metastatic site, maintained the clear cell phenotype after several passages, with a high incidence (>95%) of spontaneous lung metastases at every passage (Fig. 1C and D). In addition, PCR confirmed the presence of human Alu sequence, indicating that the tumors were human, even though they are passaged in mice (Fig. 1E). The growth rate of RP-R-02LM was faster (100 mm2 at day 31) than the parental RP-R-02 (100 mm2 at day 49; Supplementary Fig. S1B). Noninvasive MRI showed that metastatic burden was independent of primary tumor volume in mice bearing orthotopic RP-R-02LM (Supplementary Fig. S1C), suggesting that tumor shedding may occur early. In this model, we observed a significant increase in EZH2, SETD2, ALDH1A1, E2F-1, and Snail1 expression. Evidence of epithelial-to-mesenchymal transition (EMT) was also observed, with an increase in N-cadherin, ZEB1, and HIF2a levels and a decrease in VASH1 and E-cadherin expression in RP-R-02LM compared with RP-R-02 (Fig. 1F). Immunofluorescence costaining analysis of E-cadherin–EZH2 indicated a decrease in E-cadherin and an increase in EZH2 in the metastatic model RP-R-02LM compared with the parental line RP-R-02 (Fig. 1G; Supplementary Fig. S1D).
Characteristics of RP-R-02LM. A, Schematic of the development of the metastatic ccRCC model, RP-R-02LM. VHL, von Hippel-Lindau. B, Coronal T2-weighted magnetic resonance images showing primary tumor (outlined in red dotted lines) and lung metastases (red arrows) in mice with orthotopic implanted RP-R-02LM tumor. C, Representative photographic images and H&E-stained lungs of mice bearing RP-R-02LM indicate lung metastasis (mets) compared with the parental model. D, Representative H&E-stained sections of RP-R-02 and RP-R-02LM primary PDX (subcutaneous) indicate that both models still maintain the clear cell morphology after several passages in mice (400 μm). E, PCR data indicate the presence of human Alu sequence in both the parental and the metastatic models. F, Quantitative RT-PCR of genes associated with metastases shows an increase in ZEB, HIF1α, N-cadherin, STED2, EZH2, E2F1, Snail, and ALH1A1 expression and a decrease in E-cadherin and VASH1 expression in RP-R-02LM compared with RP-R-02 (*, P < 0.01; **P, < 0.001; ns, not significant). G, IHC staining and quantification analysis of E-cadherin and EZH2 in RP-R-02 and RP-R-02LM tumors.
Characteristics of RP-R-02LM. A, Schematic of the development of the metastatic ccRCC model, RP-R-02LM. VHL, von Hippel-Lindau. B, Coronal T2-weighted magnetic resonance images showing primary tumor (outlined in red dotted lines) and lung metastases (red arrows) in mice with orthotopic implanted RP-R-02LM tumor. C, Representative photographic images and H&E-stained lungs of mice bearing RP-R-02LM indicate lung metastasis (mets) compared with the parental model. D, Representative H&E-stained sections of RP-R-02 and RP-R-02LM primary PDX (subcutaneous) indicate that both models still maintain the clear cell morphology after several passages in mice (400 μm). E, PCR data indicate the presence of human Alu sequence in both the parental and the metastatic models. F, Quantitative RT-PCR of genes associated with metastases shows an increase in ZEB, HIF1α, N-cadherin, STED2, EZH2, E2F1, Snail, and ALH1A1 expression and a decrease in E-cadherin and VASH1 expression in RP-R-02LM compared with RP-R-02 (*, P < 0.01; **P, < 0.001; ns, not significant). G, IHC staining and quantification analysis of E-cadherin and EZH2 in RP-R-02 and RP-R-02LM tumors.
RTKi-induced resistance is associated with loss of antitumor effect but persistent antiangiogenic and antimetastatic activity
We have previously reported that acquired resistance to sunitinib can be transient and reversible both in preclinical and clinical studies of ccRCC (3). Thus, we wanted to assess the antitumor effect of sunitinib in our metastatic model. To our surprise, RP-R-02LM grown subcutaneously was intrinsically resistant to sunitinib (Fig. 2A) in contrast with the significant efficacy of sunitinib in the RP-R-02 parental model (3). However, when we investigated the effect of sunitinib on OS, mice treated with sunitinib had a survival benefit compared with control mice (Fig. 2B). Macroscopic and microscopic evaluations both indicated that mice following sunitinib treatment had a significant reduction of lung metastatic burden compared with the control group (Supplementary Fig. S2A). Interestingly, a close examination of the tumor vasculature showed a significant decrease in blood vessel density following sunitinib treatment (Supplementary Fig. S2B), although the percentage of proliferating cells did not change compared with the control. Similar results were observed in mice treated with axitinib, another RTKi approved for patients with RCC (Fig. 2C; Supplementary Fig. S3A–S3C). We further investigated whether the resistance was specific to RTKi only or also to a VEGF blocker such as bevacizumab. We first confirmed that bevacizumab has an antitumor effect, in a nonmetastatic PDX model, RP-R-01 (Supplementary Fig. S4A–S4C). Then, we treated mice bearing subcutaneous RP-R-02LM tumors with sunitinib, bevacizumab, or vehicle and found that bevacizumab, but not sunitinib, had a significant antitumor effect, as measured by growth inhibition (Fig. 2D and E). Both bevacizumab and sunitinib treatment groups showed decreased blood vessel density (Fig. 2F and G), but a reduction in proliferating cells occurred only in the bevacizumab-treated group (Fig. 2F and H). A close observation of tumor metastases in the RP-R-02LM indicated an antimetastatic effect in both sunitinib- and bevacizumab-treated groups, although the effect was greater in the sunitinib-treated group despite the lack of control of the primary tumor (Fig. 2I and J). This observation suggests that the antimetastatic effect of sunitinib may be due to EMT inhibition, as indicated by an increase in E-cadherin expression in tumor cells (Supplementary Fig. S4D).
Sunitinib treatment fails to inhibit tumor growth in RP-R-02LM but inhibits metastases. A, Tumor growth curves of mice treated with sunitinib had no significant difference in the growth rate compared with the control, indicating that the tumors are intrinsically resistant to sunitinib. B, Kaplan–Meier curves showing a significant survival benefit with sunitinib treatment compared with untreated control animals. C, Tumor growth curve for mice untreated or treated with either sunitinib or axitinib. D and E, Tumor growth of mice untreated or treated with either sunitinib or bevacizumab indicates a significant decrease in the bevacizumab-treated group compared with sunitinib and control. F, IHC indicates decreased vessel density (CD31) with sunitinib and bevacizumab treatment but a decrease in the percentage of proliferating cells (Ki67) in only bevacizumab-treated groups. G and H, Graphs indicate quantitative analysis of CD31 and Ki67 in the primary tumors within each treatment group. I, Photographic evidence of decreased metastatic burden with both sunitinib and bevacizumab treatment. J, Graph indicates quantitative analysis of percentage of tumor burden in the lungs within each treatment group. Results are presented as mean ± SEM. *, P < 0.05; **, P < 0.005; ***, P < 0.001; ns, not significant.
Sunitinib treatment fails to inhibit tumor growth in RP-R-02LM but inhibits metastases. A, Tumor growth curves of mice treated with sunitinib had no significant difference in the growth rate compared with the control, indicating that the tumors are intrinsically resistant to sunitinib. B, Kaplan–Meier curves showing a significant survival benefit with sunitinib treatment compared with untreated control animals. C, Tumor growth curve for mice untreated or treated with either sunitinib or axitinib. D and E, Tumor growth of mice untreated or treated with either sunitinib or bevacizumab indicates a significant decrease in the bevacizumab-treated group compared with sunitinib and control. F, IHC indicates decreased vessel density (CD31) with sunitinib and bevacizumab treatment but a decrease in the percentage of proliferating cells (Ki67) in only bevacizumab-treated groups. G and H, Graphs indicate quantitative analysis of CD31 and Ki67 in the primary tumors within each treatment group. I, Photographic evidence of decreased metastatic burden with both sunitinib and bevacizumab treatment. J, Graph indicates quantitative analysis of percentage of tumor burden in the lungs within each treatment group. Results are presented as mean ± SEM. *, P < 0.05; **, P < 0.005; ***, P < 0.001; ns, not significant.
Sunitinib-induced resistance is associated with increased kinase activation and EZH2-mediated global kinome programming
To determine what regulatory proteins were altered with sunitinib resistance, we ran a RPPA on RP-R-01 PDX treated with sunitinib. The analysis identified dynamic changes in the phosphorylation of several kinases and antiapoptosis-associated proteins, including AKT, EGFR, FAK, CDK1, PI3K, JAK2, and mTORC, as tumors progressed from a sensitive to a resistant state (P > 0.05; Fig. 3A; Supplementary Fig. S5A). To identify potential mechanisms of sunitinib resistance, we first compared sensitive (786-0) and resistant (786-0R) ccRCC cell lines. 786-0R cells were generated by chronic in vitro exposure to increasing concentrations of sunitinib. We compared the level of phosphorylation activity of several kinases using specific peptide-based kinase biosensors in 786-0 and 786-0R cells. Fluorescence capture of specific kinases indicated a significant increase in phosphorylated Src (P > 0.001), AKT (P > 0.001), and FAK (P > 0.001) and a decrease in phosphorylated VEGFR in 786-0R cells, compared with the parental sensitive cell line 786-0, indicating a potential kinome reprogramming in the resistant cells (Supplementary Fig. S5B). Then, we compared the effect of sunitinib treatment on kinase activity in 786-0 and 786-0R cells. Surprisingly, sunitinib treatment had opposite effects in these cell lines causing loss of kinase activity in 786-0 cells but inducing kinase activity such as AKT in 786-0R cells (Fig. 3B). Simultaneous detection of relative phosphorylated kinases using human phospho-RTK array indicated increase in phosphorylated kinases in resistant cells with sunitinib treatment (Supplementary Fig. S5C), indicating that sunitinib induces activation of alternative kinases.
Sunitinib-induced resistance is associated with an increase in EZH2 and alternative kinases. A, Heatmap of selected phosphorylated kinases and proteins in RP-R-01 PDX models at the point of sensitivity and resistance to sunitinib. B, Representative images of peptide-based kinase biosensor in sunitinib-resistant generated cells 786-0R and 786-0. Bar graphs indicating significant increase in phosphorylating activity of AKT with resistance. C, Generated heatmap from phosphoproteomic mass spectrometry analysis indicates a shift in the differential enrichment of phosphosites as cells move from sensitivity to resistance in ccRCC cell lines. D, Immunofluorescence staining of 786-0, 786-0R, and 786-0RshEZH2 cells indicate increased global phosphotyrosine and phosphoserine in 786-0R cells with sunitinib treatment that was attenuated with knockdown of EZH2. E, Quantitative analysis on RP-R-02LM tumor pieces indicates a significant increase in global tyrosine and global serine phosphorylation with sunitinib treatment that is abrogated with pharmacologic inhibition of EZH2 and combination with sunitinib. Results are presented as means + SEM. ****, P < 0.0001.
Sunitinib-induced resistance is associated with an increase in EZH2 and alternative kinases. A, Heatmap of selected phosphorylated kinases and proteins in RP-R-01 PDX models at the point of sensitivity and resistance to sunitinib. B, Representative images of peptide-based kinase biosensor in sunitinib-resistant generated cells 786-0R and 786-0. Bar graphs indicating significant increase in phosphorylating activity of AKT with resistance. C, Generated heatmap from phosphoproteomic mass spectrometry analysis indicates a shift in the differential enrichment of phosphosites as cells move from sensitivity to resistance in ccRCC cell lines. D, Immunofluorescence staining of 786-0, 786-0R, and 786-0RshEZH2 cells indicate increased global phosphotyrosine and phosphoserine in 786-0R cells with sunitinib treatment that was attenuated with knockdown of EZH2. E, Quantitative analysis on RP-R-02LM tumor pieces indicates a significant increase in global tyrosine and global serine phosphorylation with sunitinib treatment that is abrogated with pharmacologic inhibition of EZH2 and combination with sunitinib. Results are presented as means + SEM. ****, P < 0.0001.
To confirm the association of EZH2 expression with sunitinib exposure, we established a UMRC2 sunitinib–resistant cell line, and we subjected the resistant cells to drug washout before rechallenging them with sunitinib. Our immunofluorescence data indicate an increase in the intensity of EZH2 protein expression, as the resistant cells were either acutely (after a 4-week washout) or chronically exposed to sunitinib treatment (Supplementary Fig. S6). Similarly, we observed an increased global tyrosine and serine phosphorylation as well as increased phosphorylated FAK (Supplementary Fig. S6).
Phosphoproteomic mass spectrometry analysis in 786-0–sensitive and 786-0R–resistant cells identified 3,041 phosphosites, with 429 phosphosites enriched in the resistant cells and 427 enriched in the sensitive cells by volcano plots (Supplementary Fig. S7A and S7B). The generated heatmap indicated changes in enriched phosphosites as cells moved from sunitinib sensitivity to drug resistance (Fig. 3C). Further analysis to determine motif enrichment using motif-x and pLOGO indicated the presence of more acidic motifs in the resistant cells as compared with the sensitive cells (Supplementary Fig. S7C). DAVID pathway analysis for these enriched phosphosites in the resistant cells revealed that the associated kinases were involved in downstream VEGF signaling pathways, even though VEGF-R expression was decreased (Supplementary Fig. S8). We also identified other signaling pathways prominent with resistance (Supplementary Figs. S9 and S10).
We have previously reported an association of EZH2 induction with sunitinib acquired resistance (3). To test whether the increased levels of EZH2 observed in sunitinib-resistant ccRCC cells was responsible for the kinome reprogramming associated with resistance, we generated EZH2 knockdown cell lines using 786-0R cells (Supplementary Fig. S11A) and used immunofluorescence techniques to assess the kinome status. As expected, there was an increase in global phosphoserine and global phospho-tyrosine in the resistant cells as compared with the parental control (Fig. 3D and E). Sunitinib treatment decreased global phosphorylation in sensitive cells but significantly increased it in the resistant cells (P = 0.0014). Similar effects on global serine phosphorylation were observed with other RTKi (Supplementary Fig. S11B). This opposite effect of sunitinib was consistent with our analysis of individual kinases and indicates a reprogramming induced by sunitinib that is unique to the resistant cells. Similar results were observed in our intrinsic sunitinib-resistant in vivo model RP-R-02LM, where immunofluorescence staining detecting global phosphorylated tyrosine and serine containing proteins revealed an increase that was abrogated by the EZH2 inhibitor EPZ011989 (36), and further diminished in the combination treatment group (Fig. 3F). Taken together, these data suggest that EZH2 is responsible for the kinome reprogramming that occurs in sunitinib-resistant ccRCC cells upon sunitinib treatment.
Sunitinib-resistant cells depend on alternative kinase activation for growth that is attenuated by EZH2 inhibition
Given the dynamic changes in global phosphorylation and EZH2 role, we further determined whether these induced kinases activation led to cell survival. Using a 3D spheroid model, we treated 786-0, 786-0R, and 786-0R shEZH2 cells with sunitinib at different concentrations and measured cell death by propidium iodide (PI) uptake. In the parental 786-0 cells, PI uptake increased (P = 0.0017), indicating sunitinib-mediated cell death. In contrast, PI uptake in the 786-0R cells decreased upon sunitinib treatment consistently with sunitinib resistance, and potentially indicating an antiapoptotic effect (Fig. 4A). Interestingly, 786-0R cells with the EZH2 knockdown responded similarly to 786-0 cells, indicating a role for EZH2 in sunitinib resistance. Similar results were observed using another RTKi, axitinib, in a conventional 2D assay, where decreased cell viability was observed with EZH2 inhibition in the resistant cell line (Supplementary Fig. S12). Thus, we compared the effect of sunitinib treatment on global phosphorylated tyrosine kinase activity in 786-0R and 786-0RshEZH2 cells by phosphoproteomic mass spectroscopy. The generated heatmap indicated dynamic changes in phosphorylated tyrosine kinases in 786-0R and 786-0RshEZH2 in the presence or absence of sunitinib (Fig. 4B). Multiple tyrosine kinases phosphorylation at specific sites, including FGFR2, EGFR, EPHA4/5/3, AXL, and JAK1, were increased in sunitinib-treated 786-0R cells. Interestingly, most of the tyrosine-enriched phospho-kinases were decreased in EZH2 knockdown cells and further significantly decreased in the presence of sunitinib (P > 0.05) also. In addition, FLIM peptide biosensor activity of p-AKT and p-VEGF-R2 and immunofluorescence intensity of p-FAK again revealed increased kinase activity with sunitinib treatment in resistant cells, which was abrogated with inhibition of EZH2 (Fig. 4C). These observations were consistent with the results in our RP-R-02LM model (Supplementary Fig. S13A and S13B). Interestingly, we also observed a reversal of E-cadherin with treatment (Supplementary Fig. S13C). This observation suggests that resistant cells in the presence of sunitinib are dependent on alternative activated kinases for growth and evasion of cell death, and kinome reprogramming may enable tumor cells to escape the antitumor effect of sunitinib.
EZH2 inhibition diminishes induced global phosphorylation, leading to sunitinib sensitivity in resistant models. A, 3D spheroid assay indicating fold change in PI uptake in 786-0, 786-0R, and 786-0RshEZH2 cells with increasing concentration of sunitinib. B, Heatmap generated from pY enrichment analysis indicating dynamic changes in tyrosine phosphorylated kinases in 786-0R and 786-0RshEZH2 cells in the presence or absence of sunitinib. Table indicates prominent kinases that were increased with sunitinib treatment to be decreased with EZH2 inhibition. C, Quantitative analysis of immunofluorescence intensity for p-FAK and p-AKT in 786-0, 786-0R, and 786-0RshEZH2 cells revealed an increased intensity and activity with sunitinib treatment in 786-0R cells that was significantly decreased with knockdown of EZH2. Results are presented as means + SEM. ****, P < 0.0001.
EZH2 inhibition diminishes induced global phosphorylation, leading to sunitinib sensitivity in resistant models. A, 3D spheroid assay indicating fold change in PI uptake in 786-0, 786-0R, and 786-0RshEZH2 cells with increasing concentration of sunitinib. B, Heatmap generated from pY enrichment analysis indicating dynamic changes in tyrosine phosphorylated kinases in 786-0R and 786-0RshEZH2 cells in the presence or absence of sunitinib. Table indicates prominent kinases that were increased with sunitinib treatment to be decreased with EZH2 inhibition. C, Quantitative analysis of immunofluorescence intensity for p-FAK and p-AKT in 786-0, 786-0R, and 786-0RshEZH2 cells revealed an increased intensity and activity with sunitinib treatment in 786-0R cells that was significantly decreased with knockdown of EZH2. Results are presented as means + SEM. ****, P < 0.0001.
Sunitinib-resistant phenotype is associated with a genomic EZH2 signature
To elucidate the molecular mechanism underlying EZH2 involvement in kinome reprogramming, we mapped EZH2-binding sites by ChIP next-generation sequencing (ChIP-seq) in 786-0, 786-0R, and 786-0RsEZH2 cells. First, we compared EZH2 occupancy across the genome in both parental (786-0) and sunitinib-resistant (786-0R) cells. Using a P value cutoff of <0.01, we identified 2,344 regions bound by EZH2 in 786-0 cells and 3544 regions in the 786-0R. This significant difference indicates that the higher EZH2 expression in 786-0R cells resulted in increased genomic occupancy. Bound regions in both 786-0 and 786-0R lines strongly correlated with known EZH2/SUZ binding sites in other cell types from the ENCODE ChIP-seq data (Supplementary Fig. S14A), confirming the quality of these data. Interestingly, our results showed little overlap between regions bound by EZH2 in the 786-0 and the 786-0R cells, indicating a redistribution of EZH2 binding upon resistance (Fig. 5A; Supplementary Fig. S14B). EZH2 ChIP-seq was repeated in 786-0R cells with EZH2 knockdown (786-0RshEZH2). This resulted in decreased occupancy of the regions bound by EZH2 in the 786-0R cells (Fig. 5A) but a slight increase at regions bound in 786-0 cells (Supplementary Fig. S14B), consistent with a reversal of the EZH2 redistribution. ChIP-seq of H3K27me3 indicated that this histone mark was present at regions bound by EZH2 and increased in 786-0R cells (Fig. 5A), consistent with the canonical role of EZH2 as an H3K27me3 writer. To test whether the change in EZH2 chromatin occupancy was associated with an altered gene expression program that promotes sunitinib resistance, we compared RNA-seq of 786-0, 786-0R, and 786-0RshEZH2. The 786-0R cells had 3,174 genes that were up- or downregulated 2-fold or more compared with the 786-0 cells. Strikingly, knockdown of EZH2 reversed these gene expression changes, with 87% of the upregulated genes going back down, and 87% of the downregulated genes going back up (Fig. 5B; Supplementary Table S2). Direct EZH2 target genes showed a similar reversal (Supplementary Fig. S14C). These findings indicate that EZH2 expression is a major regulator of the gene expression changes associated with sunitinib resistance in 786-0 cells. To test whether a similar phenomenon occurs in the PDX model, we performed RNA-seq of the parental RP-R-02 tumors, the metastatic RP-R-02LM tumors, and RP-R-02LM tumors from mice treated with the EZH2 inhibitor EPZ011989. In the PDX model the “return to normal” upon EZH2 inhibition was not as striking as it was when EZH2 was knocked down in the cell line model (Fig. 5B; Supplementary Table S2). This is consistent with the observation that there are multiple differences between the RP-R-02 and the RP-R-02LM tumors (i.e., both sunitinib resistance and metastasis), and thus, the gene expression program mediated by EZH2 might be one of multiple changes. To compare the gene expression programs regulated by EZH2 in both cell line and tumor models, we examined the genes that change expression upon resistance and “return to normal” upon EZH2 knockdown or inhibition. These EZH2-regulated genes were associated with very similar functional categories in both the cell line and tumor system including KRAS signaling in the kidney and proteins of the apical surface of epithelial cells (Supplementary Fig. S14D). We then generated a list of 1,847 genes that were differentially expressed in the sunitinib-resistant cells and “return to normal” upon EZH2 inhibition or knockdown in both the cell line and tumor systems (Fig. 5C; Supplementary Table S2). Pathway analysis of this combined gene list showed a strong enrichment for pathways related to cancer and signaling pathways associated with kinases we identified as modified upon sunitinib resistance (Fig. 5C; Supplementary Table S2).
EZH2 expression mediates a gene expression program associated with kinome reprogramming and epithelial cell function. A, Heatmap comparison of EZH2 or H3K27me3 ChIP-seq read counts from indicated cell lines at the 3,544 regions bound by EZH2 in 786-0R cells. B, Heatmap showing all up- and downregulated genes rank-ordered by z-score change in 786-0R cells compared with 786-0 cells and showing how expression of these genes changes upon EZH2 shRNA in 786-0R cells (left). Similarly, genes altered in RP-R-02LM tumors compared with RP-R-02 tumors are shown followed by “return to normal” upon EZH2 inhibition (right). C, Heatmap displays 1,847 genes that increase or decrease in expression upon resistance in both cell line and tumor system and return to normal upon EZH2 shRNA in both systems. Pathway analysis (33) of these 1,847 genes indicates strong enrichment for pathways related to cancer and the kinases, which were modified upon sunitinib resistance. D, Heatmap depicts EZH2-bound sites near kinase and phosphatase encoding genes in 786-0, 786-0R, and 786-0RshEZH2. E, ChIP-seq tracks and bar charts for DAB2IP, PTPN3, and PTK2 show gain or loss of EZH2 binding that correlates with gene expression changes measured by RNA-seq (n = 2, error bars show 95% confidence interval).
EZH2 expression mediates a gene expression program associated with kinome reprogramming and epithelial cell function. A, Heatmap comparison of EZH2 or H3K27me3 ChIP-seq read counts from indicated cell lines at the 3,544 regions bound by EZH2 in 786-0R cells. B, Heatmap showing all up- and downregulated genes rank-ordered by z-score change in 786-0R cells compared with 786-0 cells and showing how expression of these genes changes upon EZH2 shRNA in 786-0R cells (left). Similarly, genes altered in RP-R-02LM tumors compared with RP-R-02 tumors are shown followed by “return to normal” upon EZH2 inhibition (right). C, Heatmap displays 1,847 genes that increase or decrease in expression upon resistance in both cell line and tumor system and return to normal upon EZH2 shRNA in both systems. Pathway analysis (33) of these 1,847 genes indicates strong enrichment for pathways related to cancer and the kinases, which were modified upon sunitinib resistance. D, Heatmap depicts EZH2-bound sites near kinase and phosphatase encoding genes in 786-0, 786-0R, and 786-0RshEZH2. E, ChIP-seq tracks and bar charts for DAB2IP, PTPN3, and PTK2 show gain or loss of EZH2 binding that correlates with gene expression changes measured by RNA-seq (n = 2, error bars show 95% confidence interval).
To identify EZH2 target genes that might mediate kinome reprogramming, we specifically examined genes encoding kinases and phosphatases. Comparing all EZH2-binding sites near a kinase or phosphatase gene, we observed a dramatic redistribution of EZH2 in the resistant cells that is partially restored upon EZH2 knockdown (Fig. 5D; Supplementary Table S2). Examples of genes associated with increased EZH2 occupancy include the Ras-GAP DAB2IP and the phosphatase PTPN3 (Fig. 5E). Both DAB2IP and PTPN3 decrease in expression in resistant cells but increase upon EZH2 knockdown (Fig. 5E). Both DAB2IP and PTPN3 are tumor suppressors that inhibit oncogenic signaling pathways such as RAS/MAPK and PI3K/AKT (37–40). In contrast, the PTK2 gene, which encodes the oncogenic kinase FAK, has decreased EZH2 binding in resistant cells that increases again upon EZH2 depletion (Fig. 5E). Consistently, PTK2 expression increases in resistant cells and decreases upon EZH2 depletion (Fig. 5E). Together, these data suggest that gene expression programs associated with kinome reprogramming and EMT in both the sunitinib-resistant cell line and metastatic PDX model are caused by redistribution of EZH2 across the genome and can be reversed by its inhibition.
Combination of an EZH2 inhibitor and sunitinib induces regression of established lung metastases
To assess the efficacy of a combination treatment of EZH2 inhibition and sunitinib in both primary tumor growth and metastases in vivo, we treated mice bearing RP-R-02LM tumors prior to tumor dissemination to the metastatic site with an EZH2 inhibitor, EPZ011989, alone and in combination with sunitinib (Fig. 6A and B). EZH2 inhibition resulted in a modest inhibition of tumor growth. Again, we did not observe a significant inhibitory effect of sunitinib on tumor growth, although there was a decrease in tumor vasculature and metastases. The greatest reduction in primary tumor burden was observed with the combination treatment. We also observed decreased metastases in the combination, although the inhibition was not greater than with single-agent sunitinib (Fig. 6B and C). Then, we further investigated the effect of this combination strategy in mice with established metastases. Using micro-CT imaging, we observed that single-agent sunitinib did not have a significant effect on established lung metastases. Treatment with EPZ011989 resulted in smaller lung nodules. However, we observed a significant regression of the metastatic burden in the mice treated with the drug combination (Fig. 6D and E). Taken together, our data provide evidence that EZH2 plays a significant role in resistance to RTKi such as sunitinib and inhibition of its activity reestablishes drug sensitivity.
Combination of EZH2 inhibitor and sunitinib overcomes sunitinib resistance. A, Schema representation of in vivo study and time of treatment prior to development of metastases is depicted. B, Tumor growth curve assessed by blinded measurements using calipers, endpoint tumor weight, and decreased percentage of vessel density and proliferating cells (4 random fields/tissue; magnification, ×20) indicates an overall decrease in tumor burden in the combination group. C, Representative photographic evidence and H&E images of lungs in each group indicate a significant decrease in metastatic burden in sunitinib-treated group and the group treated with combination of both drugs. D, Schema representation of in vivo studies treated after development of metastases. E, Representative μCT images (first and second panel), photographic images (third panel), and H&E staining (fourth panel) of lungs with or without treatment indicate a significant decrease of metastases in combination treatment compared with controls and single agent (right). E, Quantification of tumor nodules in the lungs indicates a significant decrease in the combination group (left). Results are presented as means ± SEM, *, P < 0.05; **, P < 0.005; ***, P < 0.001; ns, not significant.
Combination of EZH2 inhibitor and sunitinib overcomes sunitinib resistance. A, Schema representation of in vivo study and time of treatment prior to development of metastases is depicted. B, Tumor growth curve assessed by blinded measurements using calipers, endpoint tumor weight, and decreased percentage of vessel density and proliferating cells (4 random fields/tissue; magnification, ×20) indicates an overall decrease in tumor burden in the combination group. C, Representative photographic evidence and H&E images of lungs in each group indicate a significant decrease in metastatic burden in sunitinib-treated group and the group treated with combination of both drugs. D, Schema representation of in vivo studies treated after development of metastases. E, Representative μCT images (first and second panel), photographic images (third panel), and H&E staining (fourth panel) of lungs with or without treatment indicate a significant decrease of metastases in combination treatment compared with controls and single agent (right). E, Quantification of tumor nodules in the lungs indicates a significant decrease in the combination group (left). Results are presented as means ± SEM, *, P < 0.05; **, P < 0.005; ***, P < 0.001; ns, not significant.
EZH2 expression correlates with metastatic disease and decreased sunitinib response in patients with ccRCC
To assess whether EZH2 expression is associated with intrinsic sunitinib resistance in patients with ccRCC, we utilized a TMA made with 101 samples of primary tumors from nephrectomy specimens. We performed IHC staining on the tissues to detect EZH2 and first grouped the samples into EZH2 high (H-score > median) and EZH2 low (H-score < median; Fig. 7A). Kaplan–Meier survival curves generated indicated a longer OS in patients with low EZH2 expression (median OS, 60 months) compared with patients with high EZH2 expression (median OS, 19 months; P = 0.05; Fig. 7B). In a subset of 33 patients treated with RTKi, we found that low EZH2 expression was associated with increased OS (median OS, 122 months vs. 23 months; P = 0.005; Fig. 7C). In addition, when we stratified into low, intermediate, and high, we again observed poor survival benefit for patients with high EZH2 expression (Supplementary Fig. S15A and S15B).
EZH2 expression is associated with poor survival in ccRCC. A, IHC analysis for EZH2 and H-score on 3 cores of primary and metastatic ccRCC. Representative pictures of either high (H-score > median) or low (H-score < median) EZH2 expression are shown. B, The Kaplan–Meier survival curve indicates that the median OS for EZH2 expression “below median” is 60 months compared with “above median,” 19 months. The HR for above versus below (below as baseline) is 1.6, with 95% confidence interval as (1.0–2.5) and P value as 0.05. C, Similarly, the OS of a subset of patients that received RTKi indicates median OS for “below median” to be 122 months versus “above median,” 23 months. The HR for above versus below (below as baseline) is 5.6, with 95% confidence interval as (1.7–18.7) and P value as 0.005.
EZH2 expression is associated with poor survival in ccRCC. A, IHC analysis for EZH2 and H-score on 3 cores of primary and metastatic ccRCC. Representative pictures of either high (H-score > median) or low (H-score < median) EZH2 expression are shown. B, The Kaplan–Meier survival curve indicates that the median OS for EZH2 expression “below median” is 60 months compared with “above median,” 19 months. The HR for above versus below (below as baseline) is 1.6, with 95% confidence interval as (1.0–2.5) and P value as 0.05. C, Similarly, the OS of a subset of patients that received RTKi indicates median OS for “below median” to be 122 months versus “above median,” 23 months. The HR for above versus below (below as baseline) is 5.6, with 95% confidence interval as (1.7–18.7) and P value as 0.005.
Discussion
Our findings suggest that EZH2 expression is associated with sunitinib intrinsic/acquired resistance in ccRCC via an adaptive kinome reprograming and modulates the direct antitumor effect of RTKi. EZH2 inhibition was associated with decreased global protein phosphorylation and specific kinase activation, and restoration of sunitinib directs antitumor effect both in vitro and in vivo. These findings provide a rationale for the clinical testing of combination strategies with EZH2 and RTKi in ccRCC.
E-cadherin plays a critical role in cell–cell adhesion in epithelial tissues, some of which includes cell transformation, tumor development, and progression. The loss of function of E-cadherin is generally associated with increased expression of N-cadherin, ZEB, and SNAIL. ZEB and SNAIL mediate EMT, allowing cells to detach from the tumor and migrate to the secondary site (41). Our qRT-PCR data revealed an increase in N-cadherin, ZEB, and SNAIL in the metastatic model. A significant increase in HIF2α was observed in RP-R-02LM. Increase in HIF2α has been reported to correlate with gene signatures of EMT in lung cancer models (42). We also observed an increase in expression of metastasis-associated genes ALDH1A, E2F-1, SETD2, and EZH2 (42–45). To our knowledge, RP-R-02LM is the first ccRCC PDX model with high incidence of spontaneous lung metastases to be reported in the literature and, therefore, represents a valuable tool to study the molecular mechanisms involved in disease progression and for drug development.
The human kinome contains protein kinases that comprise 1.7% of the human genome (46). Protein phosphorylation modulates enzyme localization and activity involved in several processes including cell cycle, angiogenesis, proliferation, and differentiation (47). Among the protein kinases, tyrosine kinases, specifically RTKs such as VEGFR, HER/EGFR, PDGFR, and FGFR, have been associated with cancer progression. Several mechanisms have been proposed for the acquired/intrinsic resistance to anti-VEGF agents in ccRCC, including alternative pathway activation (i.e., c-MET and AXL; ref. 7) and EMT (48). Targeting specific kinases and signaling networks critical for tumor growth may evoke adaptive kinome response that may bypass the primary drug target and activate or upregulate alternative kinase signaling networks. Our reverse-phase protein microarray data highlighted several proteins with increased expression in PDX tumors at the time of resistance to sunitinib. Interestingly, the majority of these kinases and receptor kinases analyzed were increased, although, as expected, the expression of sunitinib-targeted kinases, such as VEGF-R2 tyrosine kinase, was decreased. Other prosurvival proteins or downstream signaling proteins were expressed with resistance. Following the results from our RPPA on the activation of alternative kinases and RTKs, we performed phosphoproteomic analysis of phosphosites by mass spectrometry, and assessed the phosphorylating activity of selected kinases in sensitive and resistant cell lines by FLIM. This analysis confirmed the increased expression levels of alternative kinases as a result of sunitinib-induced resistance. The data from FLIM of kinase peptide biosensors, which measures the phosphorylating activity of kinases, suggest that alternative kinases such as AKT, FAK, and Src had an increased phosphorylating activity in the resistant cell line. After a complete sunitinib washout phase in the resistant cells, followed by reexposure to sunitinib, we observed an increase in p-AKT in the reexposed sunitinib-treated cells compared with the control, confirming an activation of an alternative kinase with resistance to sunitinib. In addition, from all 3,041 phosphosites identified by phosphoproteomic mass spectroscopy, 429 phosphosites were enriched in the resistant cells and 427 were enriched in the sensitive cells. Although the differences between sensitive and resistant cells were not significant, further motif analysis indicated an increase in the presence of acidic motifs in the resistant cells. Acidic motifs have been identified as potential downstream substrates of RTK signaling pathways (49). Thus, our data suggest that induced resistance to RTKi, such as sunitinib, may be associated with an adaptive kinome reprogramming where alternative kinases are activated to compensate for the inhibition of the target receptor kinases, a phenomenon that has been reported in other cancers resistant to tyrosine kinase inhibitors (6, 50). Taken together, the activation of kinases in resistant 786-0R cells following reexposure to sunitinib suggests that, in the setting of acquired resistance, RTKi may indeed act as an activator rather than inhibitor.
Recent reports have suggested that cells in an “epigenetic poised” state may be associated with drug resistance (51). This state has the potential to revert back to nontolerance if the epigenetic effect is inhibited. Our models of acquired (786-0R) and intrinsic (RP-R-02LM) resistance were associated with an increased expression of EZH2 suggesting a possible epigenetic mechanism. Indeed, increased global phosphorylation at tyrosine and serine residues was associated with increased EZH2 expression in the resistant phenotype. More interestingly, sunitinib treatment further increased global phosphorylation and colocalization with EZH2, suggesting that EZH2 may be interacting posttranslationally with kinases leading to the stability of their expression and activity. Recent studies have reported that inhibition of epigenetic modulators can hinder the adaptation to an induced kinase reprogramming (6, 52). We found that the higher EZH2 levels in sunitinib-resistant cells correlated with novel EZH2 chromatin occupancy and increased H3K27 trimethylation. Furthermore, knockdown of EZH2 largely reversed gene expression changes associated with resistance. To elucidate the molecular mechanism of EZH2 involvement in kinome reprogramming, we mapped EZH2-binding sites in 786-0, 786-0R, and 786-0RshEZH2 cells and specifically examined genes encoding kinases and phosphatases. Our data indicated a dramatic shift in EZH2-enriched sites as cells moved from sensitivity to resistance. In the resistant cells, we saw increased EZH2 occupancy at some genes, such as DAB2IP and PTPN3. However, we observed loss at others, indicating redistribution, rather than simply more EZH2 binding. In particular, EZH2 binding decreased in resistant cells at the gene (PTK2) encoding focal adhesion kinase (FAK). In the past, several studies have reported FAK to play a crucial role in mediating tyrosine phosphorylation of proteins—in some cases, functioning as a scaffold in various signaling events—and its interaction with EZH has been shown to play a crucial role in disease progression (53). Interestingly, our RPPA data also indicated an increase in FAK expression in the resistant models. Upon knockdown of EZH2, we observed a significant decrease in FAK expression at the gene level. In addition, immunofluorescence analysis indicated a significant decrease in pFAK in both 786-0RshEZH2 and RP-R-02LM treated with EZH2 inhibitor. FAK has also been shown to interact with PI3K, activating the PI3K/AKT pathway (54). Although our ChIP data showed no enrichment for AKT, protein expression analysis indicated an increase in pAKT, which was attenuated upon inhibition of EZH2i, which indicates the possible role of FAK mediating PI3K/AKT signaling pathway. Interestingly, pathway analysis identified PI3K/AKT signaling pathway as one of the top altered pathways. Our data indicate that a possible mechanism by which EZH2 affects kinome reprogramming is through modulation of kinases such as FAK at the transcriptional and posttranslational levels. Targeting EZH2 provides the mechanism by which inhibition of EZH2 may modulate kinome reprogramming.
EZH2 has been reported to be associated with aggressive disease and poor survival in patients with RCC (18, 55). Our data confirm the clinical relevance of EZH2 by showing a shorter OS in ccRCC patients with high EZH2-expressing tumors. Interestingly, a subset of patients with ccRCC who have received RTKi had an even greater difference in OS if EZH2 was overexpressed in the primary tumors as compared with patients with low levels of EZH2. Although we recognize that these results are limited due to the small sample size, and validation in a larger cohort of patients is needed. However, all together, these clinical data suggest that EZH2 not only represents a potential target for therapeutic interventions but also a possible predictive marker to identify those patients who may not benefit from RTKi. Prospective clinical studies will be required to confirm this hypothesis.
In conclusion, our results suggest that sunitinib resistance in ccRCC may be due in part to a compromised direct antitumor effect that appears to be modulated by EZH2. Adaptive kinase reprogramming resulting from resistance to small-molecule RTKi such as sunitinib may be associated with epigenetic modulators that upregulate alternative kinases as an escape mechanism to tumor survival. We anticipate that our findings will be of immediate clinical relevance, as EZH2 inhibitors are already in advanced clinical development and may be combined with RTKi to overcome/delay drug resistance in ccRCC.
Disclosure of Potential Conflicts of Interest
H. Keilhack has ownership interest (including patents) in Epizyme Inc. No potential conflicts of interest were disclosed by the other authors.
Authors' Contributions
Conception and design: R. Adelaiye-Ogala, S. Chintala, M. Seshadri, R. Pili
Development of methodology: R. Adelaiye-Ogala, N.P. Damayanti, S. Chintala, K.M. Miles, E. Ciamporcero, M. Seshadri, M. Radovich, J. Irudayaraj, R. Pili
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R. Adelaiye-Ogala, N.P. Damayanti, J. Arrington, M.W. Ferris, C.-C. Hsu, M. Elbanna, S. Arisa, P. Pettazzoni, G.F Draetta, M. Seshadri, B.A. Hancock, M. Radovich, M. Buck, B.P. McCarthy, P.R. Territo, J. Irudayaraj, A.W. Tao, R. Pili
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R. Adelaiye-Ogala, J. Budka, N.P. Damayanti, C.-C. Hsu, A.R. Orillion, M. Elbanna, E. Ciamporcero, G.F Draetta, M. Seshadri, M. Radovich, M. Buck, S.A. Persohn, P.R. Territo, Y. Zang, A.W. Tao, P. Hollenhorst, R. Pili
Writing, review, and/or revision of the manuscript: R. Adelaiye-Ogala, J. Budka, N.P. Damayanti, S. Chintala, M. Seshadri, M. Radovich, J. Kota, M. Buck, P.R. Territo, J. Irudayaraj, P. Hollenhorst, R. Pili
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Adelaiye-Ogala, A.R. Orillion, L. Shen, S. Arisa, R. Pili
Study supervision: R. Adelaiye-Ogala, P. Hollenhorst, R. Pili
Acknowledgments
We want to thank Epizyme Inc. for providing the EZH2 inhibitor EPZ011929. We would like to acknowledge Brian Gillard, Dr. Swathi Ramakrishnan, and Mr. Sheng-Yu Ku for assisting with maintenance of the metastatic model. We would like to acknowledge Dr. George Sandusky, Mr. Mitchel Nettleton, Mr. Max Jacobsen, and the Indiana University pathology core. We are grateful for the following shared resource facilities: Mouse Tumor Modeling Resource (MTMR) and Translational Imaging Shared Resource at Roswell Park Cancer Institute, Buffalo, NY; RPPA Core Facility, The University of Texas MD Anderson Cancer Center, Houston, TX; Imaging Core Facility, Indiana University, Indiana; Bindley Bioscience Imaging Facility, Purdue University West Lafayette, IN; Genomics Core Facility, Indiana University; Indianapolis, IN; Indiana Center for Biological Microscopy, Indiana University, Indianapolis, IN; Next-Generation Sequencing & Expression Analysis Core, University at Buffalo; and the Clinical Data Network (CDN), Roswell Park Cancer Institute, Buffalo, NY.
Grant Support
This research was funded by Roswell Park Cancer Institute's Cancer Center Support Grant from the National Cancer Institute, NIH P30CA016056 (to R. Pili) and a generous donation by Richard and Deidre Turner (to R. Pili). This investigation was conducted in part in a facility constructed with support from Research Facilities Improvement Program grant number C06 RR020128-01 from the National Center for Research Resources, NIH.
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.