Abstract
Patients harboring germline breast cancer susceptibility genes 1 and 2 (BRCA1/2) mutations are predisposed to developing breast, pancreatic, and ovarian cancers. BRCA2 plays a critical role in homologous recombination (HR) DNA repair and deleterious mutations in BRCA2 confer sensitivity to PARP inhibition. Recently, the PARP inhibitors olaparib and rucaparib were FDA approved for the treatment of metastatic breast cancer and patients with recurrent ovarian cancer with mutations in BRCA1/2. Despite their initial antitumor activity, the development of resistance limits the clinical utility of PARP inhibitor therapy. Multiple resistance mechanisms have been described, including reversion mutations that restore the reading frame of the BRCA2 gene. In this study, we generated olaparib- and rucaparib-resistant BRCA2-mutant Capan1 cell lines. We did not detect secondary reversion mutations in the olaparib- or rucaparib-resistant clones. Several of the resistant clones had gene duplication and amplification of the mutant BRCA2 allele, with a corresponding increase in expression of a truncated BRCA2 protein. In addition, HR-mediated DNA repair was rescued, as evidenced by the restoration of RAD51 foci formation. Using mass spectrometry, we identified Disruptor Of Telomeric silencing 1-Like (DOT1L), as an interacting partner of truncated BRCA2. RNAi-mediated knockdown of BRCA2 or DOT1L was sufficient to resensitize cells to olaparib. The results demonstrate that independent of a BRCA2 reversion, mutation amplification of a mutant-carrying BRCA2 contributes to PARP inhibitor resistance.
Introduction
Germline mutations in the BRCA1 or BRCA2 genes significantly increase an individual's lifetime risk of developing breast, prostate, and ovarian cancer (reviewed in ref. 1). The BRCA1 and BRCA2 proteins play essential roles in homologous recombination (HR)–mediated repair of DNA double-strand breaks (DSB). BRCA1 has been implicated in DNA end resection as well as RAD51 loading, whereas BRCA2 is considered essential for RAD51 loading onto resected single-stranded DNA (reviewed in ref. 2). In addition, both proteins prevent excessive MRE11-mediated degradation of DNA replication forks (3). Thus, BRCA1/2 play critical roles in maintaining genome stability.
Cancers with mutations that disrupt BRCA1/2 protein activity are highly sensitive to treatment with inhibitors of PARP. The PARP inhibitors (PARPi) olaparib and rucaparib are now approved for the treatment of BRCA1/2 wild-type and mutated ovarian cancers and BRCA1/2-mutated breast cancers (reviewed in ref. 4; ref. 5). Although not yet approved for pancreatic cancer, a clinical trial reported that olaparib maintenance significantly extended progression-free survival (6). However, PARPi resistance poses a significant clinical challenge and is understudied in the context of pancreatic cancer (7). Previously described mechanisms of PARPi resistance include secondary mutations in the BRCA2 gene that restore the open-reading frame (ORF), increased p-glycoprotein expression, elevated expression of mutated BRCA1 proteins, stabilization of the replication fork, and loss of DNA end resection inhibitory proteins such as 53BP1 (8–12). Notably, most of these mechanisms have only been demonstrated in vitro. Notwithstanding these observations, there are other routes to PARPi resistance.
The Capan1 cell line was derived from a pancreatic adenocarcinoma and harbors a single-base pair deletion in exon 11 (r.6174del) of the BRCA2 gene, which is found at an elevated frequency in the Ashkenazi Jewish population and is associated with an increased risk of breast, ovarian, and pancreatic cancers (13–15). Mutations within the region of exon 11 of the BRCA2 gene result in a premature stop codon. If mRNA is successfully translated, the BRCA2 protein generated would be predicted to lack the C-terminal DNA-binding domain, but retain 7 of the 8 BRC motifs that are needed for RAD51 loading onto DNA (16). Capan1 cells have been characterized as being HR-deficient with low basal levels of RAD51 foci (17). Previous studies have generated cisplatin- and olaparib-resistant Capan1 derivatives and showed that cells acquire secondary reversion mutations that restored the BRCA2 reading frame and were responsible for generating a functional BRCA2 protein, capable of promoting HR and therapy resistance (18, 19).
In this report, we utilized two independent PARPi to generate multiple Capan1-resistant derivatives. Secondary mutations were not detected in the BRCA2 gene. Rather, there was a gain in gene copy number of the mutation-carrying BRCA2 allele that correlated with an increase of a truncated BRCA2 protein. Knockdown of BRCA2 resensitized resistant Capan1 cells to PARPi. Resistant cells with BRCA2 amplification had an increase in histone H3 lysine 79 methylation (H3K79Me) and subsequent BRCA2 knockdown reduced Disruptor of Telomeric silencing 1-like (DOT1L). The findings of this report offer a novel mechanism of PARPi resistance that is mediated through the amplification mutated BRCA2.
Materials and Methods
Cell culture and PARP inhibitor resistance establishment
Capan-1 and MDA-MD-231 (positive control for wild-type BRCA2) were purchased from the ATCC. TOV-21G (positive control for wild-type BRCA2) from Japanese Cancer Research Resources Bank, and DLD-1 from Horizon (#HD 105-007). Cells were cultured in RPMI1640 with 10% FBS and 1% penicillin/streptomycin. Capan-1 olaparib- and rucaparib-resistant cell line was established by a stepwise exposure to increasing concentrations of olaparib from 15 nmol/L to 128 μmol/L. The established resistant cells were maintained in 2 μmol/L olaparib. Cells lines were cultured for a maximum of 8 weeks and monthly tested for Mycoplasma using LookOut Mycoplasma PCR based kit (Sigma, cat. # MP0035). Capan-1 and DLD-1 cells were most recently tested on July 2, 2019 and July 22, 2019, respectively. Cell lines were authenticated at University of Arizona Genetics Core via small tandem repeat.
Reagents and antibodies
Olaparib (AZD2281), rucaparib, and pinometostat (EPZ5676) were obtained from Selleckchem. The following antibodies and reagents were obtained from the indicated suppliers: anti-BRCA2 (Bethyl Laboratories, cat. no. A311-267A, 1:5,000), anti-BRCA2 ab-1 (Millipore, cat. no. OP95, 1:1,000), anti-BRCA2 ab-2 (Millipore, cat. no. CA1033), anti-β-actin (Sigma, cat. no. A1978), anti-GAPDH (Millipore, cat. no. MAB374, 1:20,000), anti-DOT1L (Cell Signaling Technology, cat. no. 77087), anti-RAD51 (Millipore, cat. no. ABE257 1:500), and anti-BRCA1 (Calbiochem, cat. no. OP92, 1:500), anti-phospho-γH2AX (Ser139; Millipore, cat. no. 05-636, 1:400), anti-Vinculin (Cell Signaling Technology, cat. no. 13901, 1:1,000), anti-H3K79Me (Abcam, cat. no. Ab177185 1:1,000), and anti-BrdUrd (BD Biosciences, cat. no. 347583).
Lentivirus
Lentiviral constructs were packaged using the Virapower Kit and as described previously (20, 21). pLKO.1, pLKO.1-shBRCA2 #1 (TRCN0000040194), and pLKO.1-shBRCA2 #2 (TRCN0000040197) were obtained from at the Wistar Institute. shDOT1L #1 (TRCN0000236343) and shDOT1L #2 (TRCN0000236345) were obtained from Functional Genomics Facility at the University of Colorado Denver (Aurora, CO). Dr. Neil Johnson's laboratory developed pLenti-IRES-GFP-mCherry, pLenti-IRES-GFP-Full Length BRCA2, and pLenti-IRES-GFP-Truncated BRCA2 plasmids. Following pLenti-IRES-GFP transduction, GFP-positive cells were sorted twice to establish stable cell lines.
Reverse-transcriptase quantitative PCR and RNA sequencing
RNA was isolated from cells with RNeasy Mini Kit (Qiagen) followed by on-column RNase-free DNase I treatment (Qiagen). BRCA2 expression (FWD, 5′-GGGAAGCTTCATAAGTCAGTC-3′, and REV, 5′-TTTGTAATGAAGCATCTGATACC-3′) was determined using SYBR Green 1-step iScript Kit (Bio-Rad) on a Bio-Rad Chromo4 machine. β-2-microglobulin (B2M) was used as an internal control (FWD, 5′-GGCATTCCTGAAGCTGACA-3′, and REV, 5′-CTTCAATGTCGGATGGATGAAAC-3′). DOT1L expression (FWD, 5′- CACCAGACTGACCAACTCGC -3′ and REV, 5′- TCCTAGTTACCTCCAACTGTGC-3′) was determined using Luna universal One-Step RT-qPCR Kit (New England Biolabs). Isolated RNA was utilized for global next-generation sequencing [RNA sequencing (RNA-seq)] at The Wistar Genomics Facility. RNA-seq data (GSE86394) were aligned with bowtie2 (22) algorithm and RSEM v1.2.12 software (23) was used to estimate read counts and FPKM values on transcript level using Ensemble transcript information. DESeq algorithm (24) was used to compare two conditions and differences of at least 2-fold that passed FDR < 15% threshold were considered significant. SNPs were called using VarScan2 software and annotated using SnpEff tool (25, 26). Results that had P < 0.001 by Fisher exact test differences and FDR < 15% between resistant and parental cells were considered significant.
DNA sequencing
Genomic DNA from cell line rucaparib-resistant subclones were sequenced with the BROCA-HRv7 targeted sequencing assay as described previously (27). BROCA-HR includes 81 DNA repair genes. The known BRCA2 mutation site was sequenced with Sanger sequencing.
Colony formation assay
Cell lines were transduced with pLKO.1-shBRCA2 (#1 or #2), pLKO.1-shDOT1L (#1 or #2), or pLKO.1-control followed by puromycin selection (1 μg/mL). Cells were seeded in 24-well plates. Indicated doses of compounds were added and cell medium with compounds was refreshed every three days. Cells were cultured for 12 days. Colonies were fixed with 10% methanol/10% acetic acid. Colonies were stained with 0.5% crystal violet. Colonies were counted using ImageJ Software or dissolved into fixation buffer (10% methanol, 10% acetic acid, PBS) and optical density (570 nm) measured.
Copy number variation assay
Total genomic DNA was isolated from cells with Quick-gDNA Universal Kit (Zymo Research). Copy number variation was determined using predesigned BRCA2 primers (Life Technologies, cat. no. 440291) and TaqMan Genotyping MasterMix (Life Technologies). TaqMan Copy Number Reference Assay, Human RNase P (Life Technologies, cat. No. 4403326) was used as an internal control.
Immunoblotting
Cells were collected with sample buffer (2% SDS, 10% glycerol, 62.5 mmol/L Tris, pH 8.0, 0.25% Bromophenol Blue, and 100 mmol/L DTT) or RIPA buffer (150 mmol/L NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, and 50 mmol/L Tris-HCl pH 8.0) supplemented with Complete protease inhibitor cocktail (Roche). Protein concentrations were measured via BCA Kit (Pierce, cat#23225). Protein samples were separated on SDS-PAGE and transferred on PVDF membrane. Membranes were incubated in primary antibody, followed by either incubating with HRP linked anti-IgG antibodies and detected with chemiluminescent substrate (Thermo Fisher Scientific), or IRDye conjugated anti-IgG antibodies (Li-Cor) and detected with Odyssey digital fluorescence system (Li-Cor).
Histone extraction
Histones were extracted from cells using the Abcam Extraction Kit (ab113476) per the manufacturer's protocol.
Immunoprecipitation
Cell pellets were suspended into 500 μL of Buffer A [10 mmol/L HEPES, 1.5 mmol/L MgCl2, 10 mmol/L KCl, 0.5 mmol/L DTT, 0.05% NP40, Complete protease inhibitor (Roche) and phosphatase inhibitors (10 mmol/L NaF and 0.2 mmol/L Na3VO4), pH 7.9] at 4°C. Using 1 mL syringe, we passed cell suspension through a 26G needle 10×. We incubated lysate at 4°C for 15 minutes and centrifuged at 720 × g for 5 minutes at 4°C. The cell pellet was suspended into 374 μl of Buffer B [5 mmol/L HEPES, 1.5 mmol/L MgCl2, 0.2 mmol/L EDTA, 0.5 mmol/L DTT, 26% glycerol (v/v), Complete protease inhibitor and phosphatase inhibitors, pH 7.9] and 26 μL of 4.6 mol/L NaCl. The pellet was homogenized with disposable pestle and incubated at 4°C for 30 minutes, and centrifuged at maximum speed (13,000 × g) for 20 minutes at 4°C. Supernatant was collected, sonicated for 10 seconds (Branson Sonifier 250, 20% Duty Cycle, 2 output control), and used for the immunoprecipitation (IP). Protein concentration was measured using a BCA Kit (Pierce, Cat#23225), and 50 to 75 μg of protein was used for IP.
N-terminus targeting anti-BRCA2 (Bethyl Laboratories, cat. no. A303-434A) or rabbit IgG (Santa Cruz Biotechnology, cat. no. SC-2027) was added to protein samples with Dynabeads protein A (Thermo Fisher Scientific). Samples were incubated on a rotator and the beads were washed with NETN buffer (100 mmol/L NaCl, 20 mmol/L Tris-Cl, pH 8.0, 0.5 mmol/L EDTA, 0.5% NP-40). Proteins were eluted with 1× sample buffer. Eluted samples were separated on SDS-PAGE and transferred on PVDF membrane. Each IP set of samples was incubated with isotype control (IgG), N-terminus targeting anti-BRCA2 antibody (Ab-1, Bethyl Laboratories, cat. no. A303-434A, 1:5,000), C-terminus targeting anti-BRCA2 antibody (Ab-2, Bethyl Laboratories, cat. no. A303-435A, 1:5,000), or anti-DOT1L antibody (Cell Signaling Technology, 1:1,000).
Mass spectrometry
Capan-1 and Capan-1-OR cells were irradiated at 5 Gy, collected 4 hours after irradiation, and cells were lysed in ice-cold IP lysed buffer [50 mmol/L Tris-HCl, pH 8.0, 1 mmol/L EDTA, 0.5% Triton X-100, 150 mmol/L NaCl, Complete protease inhibitor (Roche), 10 mmol/L NaF, 0.2 mmol/L Na3VO4]. Cells were sheared using a 26G needle. Lysates were incubated on ice for 30 minutes and were centrifuged at 17,000 × g for 20 minutes at 4°C. The concentration of the supernatant was determined with a BCA protein and 3.2 mg from Capan-1 or 16 mg from Capan-1-OR cells of total protein were used for IP. After being precleared with 75 μL of protein A beads (NEB, Cat# 91425) that had been preequilibrated with IP lysis buffer for 30 minutes at 4°C, lysates incubated with either 15 μL of 1 mg/mL anti-BRCA2 antibody (Bethyl Laboratories Inc., Cat# A303-434A) or control normal rabbit IgG (1 mg/mL, R&D Systems, Cat# AB-105-C) overnight at 4°C. Protein A magnetic beads were added and incubated for 1 hour at 4°C. Beads were washed with IP lysis buffer 3× for 10 minutes each at 4°C, and proteins were eluted with 2× Laemmli sample buffer (Bio-Rad) and boiled for 10 minutes. Three microliters of the elution was used for immunoblot. Thirty microliters was used for 4% to 20% SDS-PAGE (Bio-Rad) and stained with Imperial Protein Stain (Thermo Fisher Scientific). The stained gel was processed by the University of Colorado Biological Mass Spectrometry Core Facility.
Immunofluorescence
Cells were treated with 5 Gy irradiation and incubated for 4 hours or 3 μmol/L olaparib for 48 hours prior to immunofluorescence staining. Cells were fixed in 4% paraformaldehyde and permeabilized in 0.5% NP-40. Samples were incubated with antibodies against BRCA1, phospho-γH2AX (Ser139) and RAD51 followed by secondary antibodies (Invitrogen). Slides were mounted with prolong anti-fade reagent (Invitrogen), and images were captured on a Nikon Fluorescence microscope and analyzed using NIS-Elements software (Nikon).
FISH
Capan1 and Capan1-RR cells were arrested in metaphase by colcemid treatment (0.1 μg/mL) for 2 hours. Cells in metaphase were subsequently collected and incubated in 0.075 mol/L KCl for 10 minutes, followed by 3 washes in Carnoy's fixative (3:1 methanol: glacial acetic acid). Interphase nuclei and cells in metaphase were dropped onto humidified glass slides, aged overnight, and FISH was performed by immersing slides in ascending ethanol series (70%, 80%, and 100%) for 2 minutes. BRCA2 FISH probe (Empire Genomics) was applied to slides, and the cellular DNA and FISH probes were codenatured at 75°C for 3 minutes. Hybridization was then carried out overnight at 37°C, followed by washing in 0.4× saline-sodium citrate buffer (SSC) and 2× SSC/0.1% Tween-20. SSC diluted from a 20× stock solution (3 mol/L NaCl, 0.3 mol/L sodium citrate, pH 7). Anti-fade mounting medium with DAPI was applied to the slide, and cells in metaphase were imaged with an Olympus BX43 fluorescent microscope equipped with a QIClick camera.
Statistical analysis
Statistical analysis was performed using GraphPad Prism (Prism 8). Two-tailed student t test or ANOVA were utilized to calculate P values. All quantitative data are graphed as mean with SEM. A calculated P value of less than 0.05 was considered statistically significant.
Results
Derivation of PARP inhibitor–resistant cell lines
To identify mechanisms of PARPi resistance in a pancreatic cell model, we examined Capan1 cells. Capan1 cells have a single-base pair deletion in exon 11 (r.del6174 or 6174delT) of the BRCA2 gene, which is predicted to produce a protein containing the RAD51-binding sites (BRC repeats) and lacking the DNA-binding domain (Fig. 1A). Examination of The Cancer Genome Atlas detected 35 tumor types and 269 individual tumors with a BRCA2 mutation beyond the BRC repeats (Table 1). Furthermore, 20 tumor types and 11.8% of tumors with BRCA2 mutations are predicted to result in a truncated protein lacking the DNA-binding domain, which recapitulates the 6174delT mutation (Table 1). To establish resistance, Capan1 cells were incubated with increasing concentrations of either olaparib or rucaparib. Initially, we examined a heterogeneous population of olaparib-resistant cells (Capan1-OR-Het). To confirm resistance, colony formation assays on Capan1 and Capan1-OR-Het cells were performed using increasing concentrations of olaparib. We observed the Capan1-OR-Het cells were an average of 743-fold (P < 0.0001) more resistant compared with the Capan1 parental population (Fig. 1B and C). Capan1-OR-Het cells were also an average of 1,000-fold resistant to rucaparib (Supplementary Fig. S1A). Colony formation assays revealed changes in proliferation rates between Capan1 and Capan1-OR-Het cells, which was confirmed via cell counting and BrdUrd incorporation assays (Supplementary Fig. S1B–S1D).
PARP inhibitor resistant cells are cross-resistant to DNA-damaging agent. A, The Capan1 pancreatic adenocarcinoma cell line has a deletion at position 6174. B, Capan1 and heterogeneous population of olaparib-resistant Capan1 (Capan1-OR-Het) cells were plated on 24-well plates, treated with indicated doses of olaparib, and subjected to colony formation after 12 days. Representative images of colony formation are shown. C, Dose–response curve was calculated with Capan1 (gray) and Capan1-OR-Het (black). Calculated IC50 values with 95% confidence intervals (CI) are indicated. D, Same as B, but examined the sensitivity to a DNA-damaging agent, cisplatin, in Capan1 and Capan1-OR-Het cells. Cells were treated with indicated doses. Representative images of foci formation assay are shown. E, Same as C, but a cisplatin dose–response curve for Capan1 (gray) and Capan1-OR-Het (black) was calculated. F, Same as B, but sensitivity to a microtubule stabilization agent, docetaxel, was assessed in Capan1and Capan1-OR-Het cells. Cells were treated with indicated doses. Representative images of foci formation assay are shown. G, Same as C, but a docetaxel dose–response curve for Capan1 (gray) and Capan1-OR-Het (black) was calculated. H, Same as C, but clonal olaparib-resistant populations [Capan1-OR-1 (squares) and -2 (triangles)] were isolated and examined for olaparib sensitivity compared with Capan1 cells (circles). Colonies were counted with ImageJ software. Data are representative of three independent experiments. Error bars, SEM.
Examination of BRCA2 mutations in The Cancer Genome Atlas.
PARPi resistance can be associated with cross-resistance to additional DNA-damaging agents, including cisplatin (18). We observed that Capan1-OR-Het cells were 110-fold (P = 0.0036) more resistant to cisplatin compared with Capan1 (Fig. 1D and E). In contrast, we examined the response of Capan1-OR-Het cells to a microtubule stabilization agent, docetaxel, and found Capan1 and Capan1-OR-Het cells displayed similar sensitivities (Fig. 1F and G). To limit genetic heterogeneity, we established clonal populations from Capan1-OR-Het cells, referred to as Capan1-OR-1 and -2. Utilizing colony formation assays, we observed Capan1-OR-1 and 2 cells were an average of 387 and 279-fold (P < 0.0001 and P < 0.0001) resistant to olaparib compared with Capan1 cells (Fig. 1H). Similar results were obtained with Capan1 cells cultured in the presence of an independent PARPi, rucaparib, and seven individual resistant (Capan1-RR) subclones were developed. All of the subclones were observed to be highly resistant to rucaparib, olaparib, and cisplatin (Supplementary Fig. S1E–S1G). These data demonstrate that the Capan1 rucaparib- and olaparib-resistant cells are cross-resistant to other PARPi and platinum-based agents.
Examination of resistant clones for BRCA2 secondary mutations
Previous studies showed that olaparib and cisplatin resistance results from the upregulation of p-glycoprotein or secondary mutations that restore the ORF of the BRCA2 gene (11, 18, 19). We examined changes in functional efflux activity in Capan1 versus resistant clones. We did not observe a significant difference between Capan1 and resistant clones for the rate of drug efflux based on doxorubicin release assays (Supplementary Fig. S2A). To further investigate the mechanism of PARPi resistance, we performed next-generation sequencing of total RNA (RNA-seq) isolated from parental and Capan1-OR-Het cells (GEO: GSE86394). p-glycoproteins were not differentially upregulated in the Capan1-OR-Het cells compared with the parental cells (Supplementary Table S1). Examination of BRCA2 SNPs allele frequencies between Capan1-OR-Het and Capan1 cells did not detect any changes suggesting there was not a reversion event (Supplementary Table S2). Furthermore, Sanger sequencing of DNA from Capan1-OR-1 and -2 for BRCA2 reversion mutations did not detect reversions in the region of the 6174delT (Supplementary Fig. S2B). In Capan1-RR cells, we assessed BRCA2 mutational status by examination of genomic DNA via BROCA-HR testing (27). We detected the original BRCA2 (r.del6174) mutation but no other BRCA2 mutations. Furthermore, there were no new mutations in DNA repair genes in the resistant clones compared to parental cells including TP53BP1, CHD4, or other genes that might confer resistance (Supplementary Fig. S2C). The lack of BRCA2 reversion mutations and efflux activity indicate these mechanisms are not playing a role in PARPi resistance.
IP analysis of BRCA2
To further ensure a BRCA2 reversion event had not occurred, we evaluated total BRCA2 protein levels. We immunoprecipitated BRCA2 using antibodies that recognize the epitopes before (Ab-1) and after (Ab-2) the predicted truncation (Supplementary Fig. S3A). As a control for antibody specificity, we transduced Capan1-OR-2 with two short hairpins specific for BRCA2 (shBRCA2). In the Capan1-OR-1 IP, we did not detect full-length BRCA2, but observed an increase in a truncated form (Fig. 2A and B; Supplementary Fig. S3B). In the rucaparib-resistant subclones 8 and 13, Ab-1 was utilized for BRCA2 IP. We observed that the truncated form of BRCA2 was immunoprecipitated from the two subclones compared with Capan1, and was only detected by Ab-1, not by Ab-2 (Fig. 2C). In addition, we detected an interaction of the truncated form of BRCA2 with two mediators of HR, PALB2, and RAD51 (Fig. 2C). These data demonstrate that BRCA2 reversion was not observed in Capan1 PARPi-resistant clones, but there was an increase in truncated BRCA2 expression, and BRCA2 interacted with HR proteins.
PARP-resistant cells express truncated BRCA2. A, Capan1-OR-1 cells were transduced with shControl and shBRCA2 (#1 and #2). Protein was extracted from Capan1, Capan1-OR-1 (shControl, shBRCA2 #1 and #2), and BRCA2 wild-type control (TOV-21G) cells. Protein was used for IP against isotype control (IgG) and BRCA2 (N-terminus, Ab-1). immunoprecipitated protein was used for immunoblot against BRCA2. B, Same as A, but IP against BRCA2 (C-terminus, Ab-2). C, Nuclear protein was extracted from Capan1, Capan1-RR clones 8 and 13, and a BRCA2 wild-type control. Protein was used for IP against BRCA2 (N-terminus, Ab-1). Immunoprecipitated protein was used for immunoblot against BRCA2 (Ab-1 and Ab-2), RAD51, and PALB2.
Amplification of mutated BRCA2
RNA-seq analysis revealed a significant increase in the number of aligned reads to the BRCA2 gene in the resistant cells compared with Capan1 cells (7,585 vs. 1,381, Fig. 3A; Supplementary Table S1). mRNA overexpression was confirmed by quantitative PCR of BRCA2 in Capan1 parental and resistant (Capan1-OR-Het, -1, and -2) populations (Fig. 3B). In both Capan1-OR and –RR cells, we observed that the clones overexpress the truncated BRCA2 protein (Fig. 3C; Supplementary Fig. S3C). Not all of the Capan1-RR cells had an increase in BRCA2 expression, but clones 8 and 13 had the highest truncated BRCA2 expression. Genomic localization of differentially expressed genes showed an enrichment of upregulated genes on chromosome 13q (Fig. 3D), suggesting a chromosomal aberration. Given this observation and increase in both mRNA and protein expression, using a PCR-based approach, we examined changes in the BRCA2 gene copy number. In the Capan1-OR cells, BRCA2 gene copy number was significantly increased in Capan1-OR-1 and -2 compared with parental Capan1 cells [9.9 (P = 0.0004) and 13.32 (P < 0.0001); Fig. 3E]. BRCA2 copy number gain was also confirmed in Capan1-RR-8 and RR-13 [4.21 (P = 0.0024) and 6.05 (P = 0.0002); Fig. 3F]. Furthermore, FISH on Capan1 and Capan1-RR clones (# 8 and 13) demonstrated multiple BRCA2-containing chromosomal rearrangements including, amplification and gene duplication (white arrows, Fig. 3G). These data indicate the truncated BRCA2 gene is amplified and its resulting protein are overexpressed in PARPi-resistant populations.
Truncated BRCA2 is overexpressed in CAPAN1 PARP inhibitor–resistant cells. A, Total RNA was isolated from Capan1 and Capan1-OR-Het cells and subsequently used for next-generation sequencing (RNA-seq). Aligned-reads for the BRCA2 gene are shown for Capan1 (parental, blue) and Capan1-OR-Het (resistant, red) cells. B, RNA was isolated from Capan1, Capan1-OR-Het, and Capan1-OR-1 and 2 and used for qPCR against BRCA2 (B2M = internal control). ANOVA, *, P = 0.0068; **, P = 0.007 and ***, P = 0.0027. C, Protein was extracted from Capan1, Capan1-OR-Het, Capan1-OR-1 and 2, and BRCA2 wild-type control (TOV-21G) cells. Protein was immunoblotted against BRCA2 (N-terminus). Arrows indicate full-length and truncated BRCA2. β-actin = loading control. D, RNA-seq analysis of Capan1 versus Capan1-OR-Het detected 411 differentially expressed genes (|fold change| > 2 and FDR < 0.15). Differentially regulated genes were mapped based on genomic location. E, Genomic DNA was isolated from Capan1, CAPAN1-OR-1, and 2 and used for qPCR on BRCA2 intron–exon junction to determine changes in gene copy number. RNase P = internal control. ANOVA, *, P = 0.0004 and **, P < 0.0001. F, Genomic DNA was isolated from Capan1, Capan1-RR-8, and 13 and utilized for qPCR BRCA2 to determine changes in gene copy number. RNase P = internal control. ANOVA, **, P = 0.0024 and ***, P = 0.0002. G, FISH against BRCA2 (green) in Capan1 and Capan1-RR-8 and -13 cells. Chromosomes = DAPI/blue. White arrows = positive BRCA2 regions. ? = possible extrachromosomal DNA. Data are representative of three independent experiments. Error bars, SEM.
HR repair is functional in PARPi-resistant cells
PARPi resistance has been attributed to a restoration in the HR DNA repair pathway (18). We therefore evaluated the HR repair pathway response by inducing DNA DSBs through irradiating (IR; 5 Gy) Capan1 cells and Capan1-OR-1 and -2. To assess the HR repair pathway activity on a single-cell level, we performed immunofluorescence analysis following IR to examine the presence of BRCA1 and RAD51 foci. We observed that IR induced a significant increase in BRCA1 foci–positive cells in Capan1-OR-1, -2 clonal and heterogeneous populations compared with Capan1 cells (44.8% vs. 81.5%, P = 0.0317 and 44.8% vs. 79.3%, P = 0.0307; Fig. 4A and B; Supplementary Fig. S4A and S4B). RAD51 is recruited to DSBs through BRCA2′s BRC motifs to facilitate strand invasion during HR repair (28). The formation of RAD51 foci after DNA damage is an output of functional HR repair (29). IR-induced DNA damage led to a significant increase in RAD51 foci–positive cells in Capan1-OR-1, -2 clonal populations versus Capan1 cells (0% vs. 22.0%, P = 0.0099 and 0% vs. 23.6%, P = 0.0067; Fig. 4A and B). Similar results were observed in rucaparib-resistant Capan-1 cells (Supplementary Fig. S4C and S4D). These data suggest that the HR DNA repair pathway is functional in PARPi resistant Capan1 cells.
Irradiated PARP-resistant cells demonstrate restored homologous recombination DNA repair. A, Capan1 (parental) and Capan1-OR-1 cells were irradiated (5 Gy) and used for immunofluorescence against BRCA1 (red) and RAD51 (green). White arrowheads indicate BRCA1 and RAD51 foci. B, Quantified BRCA1 and RAD51 foci in 200 Capan1, Capan1-OR 1, or 2 cells treated without (−IR) or with (+IR) 5 Gy and graphed as a percentage. Statistical test = ANOVA; *, P < 0.05. Data are representative of at least three independent experiments. Error bars, SEM.
Modulation of truncated BRCA2
To directly address the role truncated BRCA2 in mediating HR repair, we performed loss-of and gain-of-function studies. In PARPi-resistant Capan1 cells, we observed increased expression of truncated BRCA2. Therefore, we assessed the sensitivity of DNA-damaging agents after knocking down truncated BRCA2. We transduced shControl and two independent shBRCA2 (#1 and #2) into Capan1-OR-1 and observed a varying degree of BRCA2 knockdown (Fig. 5A and B). shControl and shBRCA2-expressing cells were utilized for colony formation assays with increasing doses of olaparib or cisplatin. We observed that knockdown of BRCA2 in Capan1-OR-1 cells significantly restored sensitivity to olaparib by 1.7 and 20-fold (P = 0.0027 and P < 0.0001; Fig. 5C). We also confirmed these findings in the Capan1-OR-Het cells (Supplementary Fig. S5A and S5B). Next, transduction of two shBRCA2s into Capan1-RR-8 significantly restored rucaparib sensitivity by 7.5- and 44-fold (P = 0.0023 and P = 0.0015; Supplementary Fig. S5C). Knocking down truncated BRCA2 significantly sensitized Capan1-OR-1 to cisplatin by 3.9- and 11.7-fold (P = 0.0004 and P < 0.0001; Fig. 5D). There was a concordance of the level of BRCA2 knockdown and degree of resensitization.
BRCA2 knockdown in CAPAN1-PR cells restores sensitivity to DNA-damaging agents. A, Capan1-OR-1 cells were transduced with shControl or shBRCA2 (#1 or #2). After drug selection, RNA was collected and followed by quantitative PCR for BRCA2. ANOVA, *, P < 0.001. B, Same as A, but protein was extracted and immunoblotted for BRCA2. GAPDH = loading control. C, Capan1, Capan1-OR-1 shCtrl, or Capan1-OR-1 shBRCA2 (#1 and #2) cells were plated in 24-well plates, treated with increasing doses of olaparib, and cultured for 12 days. Cells were then fixed and used for colony formation assays. Dose–response curve shown with indicated IC50 values. D, Same as C, but cells were treated with cisplatin. Dose–response curve shown with indicated IC50 values. E, DLD-1 BRCA2−/− cells were transduced with mCherry or GFP/HA-tagged full-length (BRCA2-HA) or truncated BRCA2 (BRCA2(del6174T)-HA). RNA was isolated and BRCA2 expression was measured via qPCR. ANOVA, ***, P < 0.001. F, Same as E, but cells were used for a colony formation assay with increasing doses of olaparib. FL = full length. Colonies were counted with ImageJ software. Data are representative of three independent experiments. Error bars, SEM.
For the gain-of-function studies, we were unable to stably transduce full-length or truncated BRCA2 in Capan1 cells. Therefore, we stably overexpressed GFP/HA-tagged full-length or truncated BRCA2 in the DLD-1 BRCA2−/− cell line (30). DLD-1 is a colorectal adenocarcinoma cell line that is BRCA2 wild type; however, cells were genetically engineered with exon 11 deleted, which results in a BRCA2−/− cell line (30). Following transduction into DLD-1 BRCA2−/− cells, we confirmed expression of both full-length and truncated BRCA2 (Fig. 5E). As expected, the DLD-1 BRCA2−/− expressing the full-length BRCA2 led to olaparib resistance compared with the DLD-1 BRCA2−/− mCherry control cells (765 nmol/L vs. 0.462 nmol/L, respectively; Fig. 5F). The DLD-1 BRCA2−/− expressing the truncated form, however, failed to promote olaparib resistance compared with the DLD-1 BRCA2−/− mCherry control cells (2.04 nmol/L vs. 0.462 nmol/L, respectively; Fig. 5F). In contrast, knockdown of truncated BRCA2 in PARPi-resistant Capan1 cells led to resensitization, which suggests truncated BRCA2 in the Capan1 cells is promoting PARPi resistance through a secondary adaptation.
Truncated BRCA2 interactome
To investigate the possibility that an additional adaptation is required to support the ability of truncated BRCA2 to promote resistance in Capan1 and Capan1-OR-2 cells, we irradiated cells and immunoprecipitated BRCA2. Subsequently, BRCA2-interacting proteins were identified via mass spectrometry. As a loading control for the IP-mass spectrometry, nonspecific tubulin and heat shock proteins were observed at similar levels between Capan1 and Capan1-OR (Supplementary Table S3). Notably, BRCA2 was effectively immunoprecipitated and peptide mapping of BRCA2-associated peptides failed to identify any C-terminal peptides confirming the absence of a reversion mutation (Supplementary Fig. S6A). IP-mass spectrometry data were filtered on the basis of peptide counts from the IgG nonspecific pulldown and by the CRAPome prevalence (<10% of all proteomic experiments; ref. 31). Thirty-seven proteins differentially interact with truncated BRCA2 in Capan1-OR-2 compared with Capan1 cells (Supplementary Table S3). Similar to Capan1-RR (Fig. 2C), the BRCA2 pulled down in Capan1-OR cells showed an enrichment of HR proteins, PALB2 and RAD51 (Fig. 6A). Notably, the histone methyltransferase, DOT1L, and its cofactor, MLLT10, were significantly enriched in the BRCA2 IP in Capan1-OR-2 cells compared with Capan1 cells (Fig. 6A).
BRCA2 interacts with DOT1L and DOT1L contributes to olaparib resistance. A, Protein from irradiated (5 Gy, 4 hours) Capan1-OR 1 cells was used for an IP against BRCA2 (Ab-1) and isotype control (Rb IgG). Immunoprecipitated protein was separated on a SDS-PAGE and used for mass spectrometry. Peptide count table of Capan1-OR-1 and Capan1 cells. B, Capan1 and Capan1-OR-Het cells used for IP against BRCA2 (Ab-1) and isotype control (Rb IgG). Immunoprecipitated protein was separated on a SDS-PAGE and immunoblotted for DOT1L and BRCA2. C, RNA extracted from Capan1 and Capan1-OR-Het was used for qPCR against DOT1L. Internal control = B2M. Two-sided t-test, ***, P < 0.001. D, Protein from Capan1, Capan1-OR-Het shControl, or shBRCA2 (#1 and #2) was separated on a SDS-PAGE and immunoblotted for DOT1L. Loading control = β-actin. E, Capan1-OR-Het cells were transduced with a control shRNA (shCtrl) and two DOT1L-specific shRNAs (#1 and #2). RNA was extracted from cells and used for qPCR against DOT1L. Internal control = B2M. ANOVA, ***, P < 0.001. F, Total histones were extracted and immunoblotted for H3K79Me. Loading control = Histone H3 (H3). G, Capan1-OR-Het shCtrl and shDOTL1 (#1 and #2) cells were used for an olaparib dose–response colony formation. IC50 were calculated and graphed. ANOVA, *, P < 0.05. H, Immunofluorescence (IF) on nonirradiated (−IR) and irradiated (5 Gy, +IR) Capan1-OR-Het cells against γH2Ax (green) and RAD51 (red). White arrowheads = RAD51-foci positive cell. I, Same as I, but examined Capan1-OR-Het shDOT1L #1 and #2 cells. J, Quantification of RAD51 foci–positive cells of H and I. At least 200 cells were counted in triplicate. ANOVA, ***, P < 0.001. K, Capan1-OR-Het shCtrl and shDOTL1 (#1 and #2) cells incubated with olaparib (3 μmol/L for 48 hours) used for IF against RAD51. At least 200 cells were counted in triplicate. ANOVA, ***, P < 0.001. L, Capan1-OR-Het cells treated with vehicle control (Ctrl) or 2 μmol/L pinometostat (Pino). Histones were extracted and immunoblotted for H3K79Me. Loading control = Histone H3. M, Capan1-OR-Het shCtrl cells were treated with vehicle control or 2 μmol/L pinometostat and cells were used for an olaparib dose response colony formation. IC50 were calculated and graphed. P value calculated with a two-sided t test. N, Nonirradiated (−IR) and irradiated (5 Gy, +IR) Capan1-OR-Het cells treated with pinometostat (2 μmol/L). IF against γH2Ax (green) and RAD51 (red). White arrowheads = RAD51 foci–positive cell. O, Quantification of RAD51 foci–positive cells. At least 200 cells were counted in triplicate. ANOVA, **, P < 0.01. P, Capan1-OR-Het shCtrl cells were treated with vehicle control or 2 μmol/L pinometostat followed by olaparib (3 μmol/L for 48 hours) used for IF against RAD51. Quantification of RAD51 foci–positive cells. At least 200 cells were counted in triplicate. ANOVA, ***, P < 0.001. Data are representative of three independent experiments. Colony formation was quantified by dissolving crystal violet. Error bars, SEM.
DOT1L functions in a complex with its cofactor, MLLT10, and is the only enzyme known to methylate histone H3 lysine 79 (H3K79Me1/2/3) and is linked to DNA damage repair (reviewed in ref. 32). In mixed lineage leukemia (MLL), aberrant DOT1L and MLLT10 activity is responsible for chromosomal rearrangements (33, 34). We validated the BRCA2 and DOT1L IP in Capan1-OR-Het cells (Fig. 6B). In Capan1-OR-Het cells, DOT1L mRNA expression was only modestly upregulated by 1.3-fold (Fig. 6C). In contrast, DOT1L protein was increased in Capan1-OR-Het cells compared with Capan1 cells, and knockdown of BRCA2 promoted the loss of DOT1L protein (Fig. 6D). In the DLD-1 model, we evaluated DOT1L and were unable to detect protein expression in either the BRCA2-null cells or in truncated BRCA2 overexpressing cells. However, we subsequently overexpressed DOT1L in these cells and DOT1L protein overexpression was only observed in cells expressing truncated BRCA2 (Supplementary Fig. S6B). These data provide a possible explanation as to why overexpressing truncated BRCA2 in the DLD-1 cell line failed to induce olaparib resistance.
We next wanted to assess the role of DOT1L in Capan1 PARPi-resistant cells. A control and two independent shRNAs specific for DOT1L were transduced into Capan1-OR-Het cells. Varying levels of DOT1L knockdown were achieved (Fig. 6E) and corresponded with the loss of H3K79Me (Fig. 6F). Notably, DOT1L's enzymatic product was also decreased following BRCA2 knockdown (Fig. 6F). In the Capan1-OR-Het cells, DOT1L #1 and #2 shRNA-mediated knockdown resensitized cells to olaparib by 2.7- and 6.1-fold, respectively (Fig. 6G). Taken together these data suggest that DOT1L and truncated BRCA2 are potentially cooperating to promote PARPi resistance.
We next determined whether the loss of DOT1L expression (shRNA) or its enzymatic activity [DOT1L inhibitor – pinometostat (EPZ5676; ref. 35)] attenuated HR repair via RAD51 loading. DOT1L knockdown or inhibited cells were irradiated with 5 Gy, incubated for 4 hours, and RAD51 foci–positive cells were quantified. Similarly, DOT1L knockdown or inhibited cells were treated with olaparib and RAD51 foci–positive cells were quantified. As previously observed in Capan1-OR cells, about 20% of cells were positive for IR-induced RAD51 foci; however, in either IR or olaparib-treated cells, DOT1L knockdown significantly reduced RAD51 foci (Fig. 6H–K). Treatment with a DOT1L inhibitor reduced H3K79Me in a dose-dependent fashion (Supplementary Fig. S6C) and H3K79Me was significantly downregulated by 2 μmol/L pinometostat (Fig. 6L). Doses of pinometostat used are consistent with previous literature (36–38). Treatment with 2 μmol/L pinometostat in combination with olaparib slightly reduced olaparib's IC50 in Capan1-OR-Het cells (Fig. 6M) and attenuated RAD51 foci formation by an average of 38%, which is a lesser extent than the shRNA knockdowns, 75% reduction in RAD51 foci formation (Fig. 6N–P). These data suggest DOT1L potentially contributes to olaparib resistance by increasing DNA repair and that it could be partially through a methyltransferase-independent function.
Although PARPi are FDA approved for ovarian cancer, they are not currently approved for BRCA-mutated pancreatic cancer. We examined DOT1L expression using the ovarian cancer TCGA data. Increased expression DOT1L conveyed a worse progression-free survival and was associated with resistance to the DNA damaging platinum-based chemotherapy (Supplementary Fig. S6D and S6E). Moreover, in a previously published BRCA2-mutated PARPi resistant model of ovarian cancer (39), DOT1L knockdown significantly inhibited colony formation (Supplementary Fig. S6F–S6H). These data highlight that targeting DOT1L in PARPi-resistant tumors could be more broadly applicable.
Discussion
PARPi (olaparib and rucaparib) have entered the clinic for the treatment of BRCA1/2-mutated cancers; however, the development of resistance remains a significant clinical challenge and elucidating resistance mechanisms is critical. Therefore, we established olaparib- and rucaparib-resistant cells in the context of a mutated form of BRCA2. PARPi-resistant cells had an amplification of the truncated form of BRCA2, which led to an increase in mRNA and protein expression. Truncated BRCA2 was observed to interact with DNA damage repair effectors, RAD51 and DOT1L. Subsequently, knockdown of truncated BRCA2 restored sensitivity to PARPi.
In contrast to previous reports, we did not detect BRCA2 reversion mutations, but similarly found that HR DNA repair had been restored in resistant cells (18, 19). In the original report describing olaparib-resistant Capan1 cells, the authors utilized two methods for generation of resistant cells: a step-wise increase in olaparib (1 nmol/L–50 μmol/L) or a constant concentration (100 nmol/L). Although reversion mutations were observed under both conditions, the authors noted that only with the step-wise method did they observe amplified BRCA2 (18). Comparatively, in our study, we used the step-wise method for olaparib with a higher maximum dose, suggesting that the degree of selective pressure likely mediates an alternative resistance mechanism.
In the clinical setting, several studies have detected secondary reversion mutations of BRCA1/2. A small study examining 16 germline BRCA1/2-matched primary and recurrent high-grade serous ovarian cancer patients detected five recurrent tumors with reversion mutations. All five patients were treated with platinum-based chemotherapy and three of the five patients had been treated with PARP inhibitors (40). The therapeutic agent (platinum-based chemotherapy or PARP inhibitors) driving the observed reversion mutations is not clear. These data suggest that although reversion mutations occur, other mechanisms are also likely playing a role in the development of systemic resistance.
In Capan1 PARPi-resistant cells, BRCA2 knockdown led to olaparib and rucaparib resensitization. Given the significant increase in BRCA2 copy number and significant upregulation of truncated BRCA2, we further evaluated expression of truncated BRCA2 in an independent BRCA2−/− cell line. Overexpression of truncated BRCA2 was not sufficient to promote resistance, suggesting that in the context of amplified truncated BRCA2, Capan1 cells have potentially acquired a second adaptation to facilitate the restoration of HR repair.
We demonstrated that the HR DNA repair pathway was restored in the PARP inhibitor–resistant cells measured via BRCA1 and RAD51 foci formation. These findings are consistent with previous reports showing the BRCA2 protein–containing BRC repeats and lacking the C-terminal domain is sufficient to interact with RAD51 (41). Through BRCA2 pull-down and mass spectrometry, we identified DOT1L as a BRCA2-associated protein. The histone methyltransferase, DOT1L, directly interacts with DNA and promotes methylation on H3K79 and is the only known methyltransferase to catalyze H3K79. DOT1L is an established mediator of cell-cycle regulation and DNA DSB repair (reviewed in ref. 32). In MLL, DOT1L contributes to chromosomal rearrangements and fusions, which suggests that DOT1L could be playing an active role in the chromosomal instability in the Capan-1 PARPi-resistant cells (Fig. 2H). To date, the only known “reader” of the DOT1L-dependent H3K79 methylation is the tumor suppressor p53-binding protein 1, which has established roles in dictating specific DNA repair pathways (42). The DOT1L cofactor, MLLT10, was also observed in the BRCA2 pull-down. Interestingly, knockdown of DOT1L promoted PARPi resensitization to a greater degree than inhibiting DOT1L methyltransferase activity. We plan to determine whether DOT1L has methyltransferase-independent functions that mediated PARPi resistance. In our study, we predict that truncated BRCA2 and DOT1L are interacting to increase RAD51 loading, HR-mediated DNA damage repair, and PARPi resistance. Future work will investigate the relationship between truncated BRCA2 and DOT1L interaction in DNA damage response and potentially extend our findings into BRCA wild-type tumors.
Disclosure of Potential Conflicts of Interest
K.M. Turner is the principal scientist at Boundless Bio, Inc. P.S. Mischel is the co-founder (chair of scientific advisory board) at and has ownership interest (including patents) in Boundless Bio, Inc. No potential conflicts of interest were disclosed by the other authors.
Authors' Contributions
Conception and design: T.M. Yamamoto, N. Johnson, B.G. Bitler
Development of methodology: T.M. Yamamoto, H. Li, B.G. Bitler
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): P.H. Park, T.M. Yamamoto, A.L. Alcivar, Y. Wang, A.J. Bernhardy, K.M. Turner, Z.L. Watson, S. Casadei, E.M. Swisher, N. Johnson, B.G. Bitler
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): P.H. Park, T.M. Yamamoto, Y. Wang, A.V. Kossenkov, E.M. Swisher, P.S. Mischel, N. Johnson, B.G. Bitler
Writing, review, and/or revision of the manuscript: T.M. Yamamoto, K.M. Turner, A.V. Kossenkov, K. Behbakht, E.M. Swisher, P.S. Mischel, N. Johnson, B.G. Bitler
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.L. Alcivar, B.G. Bitler
Study supervision: B. Xia, B.G. Bitler
Acknowledgments
We acknowledge Dr. Rugang Zhang for his invaluable support. We acknowledge Lindsay J. Wheeler and Scott H. Kaufman for their critical discussion. This work was supported by grants from the NIH/NCI to N. Johnson (R01CA214799) and B.G. Bitler (R00CA194318), and Department of Defense Awards to N. Johnson (OC140040/OC130212) and B.G. Bitler (OC170228). Research supported by a Stand Up To Cancer-Ovarian Cancer Research Fund Alliance-National Ovarian Cancer Coalition Dream Team Translational Cancer Research Grant to E.M. Swisher (SU2C-AACR-DT16-15). Stand Up To Cancer is a division of the Entertainment Industry Foundation. Research grants are administered by the American Association for Cancer Research, the Scientific Partner of SU2C. Support of Core Facilities was provided by The Wistar Institute Cancer Center Support Grant CA010815 and the University of Colorado Cancer Center Support Grant (P30CA046934).
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.
Footnotes
Note: Supplementary data for this article are available at Molecular Cancer Therapeutics Online (http://mct.aacrjournals.org/).
Mol Cancer Ther 2020;19:602–13
- Received February 28, 2019.
- Revision received August 4, 2019.
- Accepted September 26, 2019.
- Published first October 1, 2019.
- ©2019 American Association for Cancer Research.
References
Volume 19, Issue 2
