Background
An attractive strategy to improve current treatment options is to inflict cytotoxic DNA damage with chemotherapy, and then impede DNA repair by molecular targeting. Poly-ADP-ribosyl-transferase-1 (PARP1) is a key sensor of DNA damage and initiates recruitment of the DNA-repair machinery to the site of damage [
1‐
3]. The development of PARP1 inhibitors has drawn closer the goal of combining these compounds with current therapies [
2,
4‐
9]. Unfortunately, despite several preclinical data confirming an increased antitumor activity by combining PARP1 inhibitors with either chemotherapy or radiotherapy [
10‐
16], dose escalation in phase 1 combination studies has been greatly hampered by the observed hematologic toxicities. These adverse events limited the possibility to exploit these combinations into clinical practice [
4,
5,
9,
11,
17‐
27]. As a consequence, PARP1 inhibitors are today registered as monotherapy in cancers bearing DNA-repair deficiencies [
28‐
34] and the strong rational to combine selected cytotoxics (especially alkylators) with PARP1 inhibitors has to face the risk of myelotoxicity.
Among chemotherapeutics, trabectedin has some peculiarities that point out this drug as an ideal candidate to be combined with PARP1 inhibitors: its favorable safety profile on the hematologic side and its unique mechanism of action [
35,
36]. In particular, trabectedin traps enzymes belonging to the transcription-coupled nucleotide excision repair (TC-NER) system that in the attempt to remove trabectedin adducts, generates DNA single- and double-strand breaks. Understandably, trabectedin displays a greater clinical benefit in BRCA1/2-deficient tumors than in proficient ones [
37,
38]. These characteristics prompted us to explore the combination of trabectedin and PARP1 inhibitors taking advantage of a large set of bone and soft tissue sarcoma (BSTS) cell lines. Despite a common mesenchymal origin, BSTS are characterized by a great degree of heterogeneity and, indeed, trabectedin displays a spectrum of activity in various sarcoma histotypes [
39‐
41]. This heterogeneity allowed us to test cells with different intrinsic sensitivity to trabectedin, in order to explore if trabectedin could efficiently activate PARP1, and if PARP1-specific inhibition could be exploited sufficiently to cause irreversible DNA damage, and eventually tumor cell death. The observed results were subsequently validated in other tumor types and by combining PARP1 inhibitors with other cytotoxics characterized by different mechanisms of action.
Methods
Cell line characterization, cell viability and western blot
Cell lines characteristics are depicted in Additional file
1: Table S1 and S2. Short tandem repeat (STR) profile was checked and the genomic status of DNA-repair key components (BRCA1, BRCA2, ATM, CHEK2, PTEN) was analyzed by Multiplex ligation-dependent probe amplification (MLPA) and Denaturing high performance liquid chromatography (DHPLC, Wave 3500HT DNA Fragment Analysis System, Transgenomic Inc.) followed by direct sequencing (ABI PRISM3100 DNA Sequencer, Applied Biosystem, Forster City, CA). PARP1, RAD51, and BRCA1 copy number variations was confirmed by real-time polymerase chain reaction (PCR) on genomic DNA (TaqMan Assay, ABI PRISM 7900HT System, Applied Biosystem).
In order to evaluate the proliferation rate of each cell line, cells were seeded at a density of 2 × 10
4 cells/cm
2, and the doubling-time (DT) of the harvested cells was calculated during the exponential growth phase (48 h of culture) by using the algorithm provided by
http://www.doubling-time.com/compute.php: DT = t x lg
2/(lgNt - lgN
0) where N
0 is the initial concentration of cells, Nt is final concentration of cells and t is the culture time in hours. Cell viability was determined with Cell Titer-Glo (Promega) after 72-h treatment with scalar doses (2–0.125 nM) of trabectedin (PharmaMar), as single agent or in constant combination with olaparib (20–1.25 μM) or veliparib (80–5 μM) (Sequoia research products). These ranges of concentrations were chosen on the results of previous studies testing the sensitivity to each single agent in sarcoma and non-sarcoma cell lines [
42‐
45]. Protein extracts were obtained after 24-h treatment and resolved by western blot using primary antibodies from Cell Signaling Technology except for anti- poly(ADP-ribose) (PAR, Trevigen), phospho-histone H2AX (Ser139) (Millipore).
Quantitative real-time polymerase chain reaction (PCR), gene expression profiling and Gene Signature enrichment analysis (GSEA)
2.5 × 10
6 cells were plated in 150 mm diameter and grown for 24 h in complete medium. Cells were then treated for additional 24 h with trabectedin (0.125 nM) and olaparib (1.25 μM) as single agent and in combination. In each condition, 500.000 cells were fixed in 70% ethanol for cell cycle analysis and the remaining fraction was lysed with Qiazol reagent (Qiagen) for RNA extraction by means of RNeasy Kit (Qiagen) following manufacturer’s instructions. Gene expression profiling was performed with Human HT-12 v4.0 Expression BeadChip Kit (Illumina) Real-time PCR was performed by TaqMan Gene Expression Assay using ABI PRISM 7900HT System (Applied Biosystem). Fluorescence data were automatically converted into C
T (cycle threshold) values. To export data, the threshold was 0.20. Raw data were analyzed by Microsoft Office Excel. Additional file
2: Figure S1 outlined the design of the expression profile project analyzed by GSEA (
http://www.broadinstitute.org/gsea/msigdb/index.jsp) [
46]. To highlight gene patterns associated and involved with trabectedin and olaparib synergism, we designed a specific expression profile project comparing cell lines displaying high synergism (HS-C: TC-106, 402.91, DMR) and cell lines displaying low/no synergism (LS-C: SJSA-1, HT1080, SW684). First, we compared gene expression profiles of HS-C and LS-C by GSEA to identify gene sets associated with high synergism. Then we selected the cores of the emerging gene sets matching the criterion of false discovery rate (FDR) < 0.05 and we challenged them with GSEA comparing each treated HS-C against untreated. GSEA analysis was performed with default parameters. Probes were collapsed into gene symbols. Gene set size for inclusion was set between 15 and 500, and the genes were permutated 1,000 times. Gene sets that met the FDR < 0.05 were considered significant. The catalog of gene sets was downloaded from Molecular Signature Database (C2, C5, C6, Hallmarks, MSigDB, version 5.0,
http://www.broadinstitute.org/gsea/msigdb/index.jsp) for a total of 4965 curated gene sets.
Array comparative genomic hybridization (aCGH) analysis
Comparative genomic hybridization using aCGH microarrays was carried out using the enzymatic labeling method. Digestion, labeling, hybridization, washing and slide scanning were performed following the manufacturer protocols (Agilent Technologies). Briefly, 750 ng of samples and control DNA in a total volume of 10.1 μL were digested with restriction enzymes and labeled with Cy3 and Cy5. Subsequently, labeled DNA was cleaned up and hybridized using the Agilent-030587 CCMC CGH plus SNP 180 k Microarray platform. Each test DNA sample (Cy5) was hybridized with a reference DNA (Cy3) from Homo sapiens (Agilent). After hybridization for 24 h at 20 rpm and 65 °C, slides were washed following Agilent procedure and scanned with the dual-laser microarray scanner version C (G2505C, Agilent Technologies). Images were analyzed using Feature Extraction software version 10.7 (Agilent Technologies). Raw data were processed using the Agilent Genomic Workbench version 7. Aberrant regions were detected using ADM-2 algorithm with threshold = 6. To avoid false positive calls, the minimum number of consecutive probes for amplifications/deletions was set to 3, together with a minimum average absolute Log Ratio for aberrations > = 0.25.
Gene ontology (GO) enrichment analysis
Genes mapped in the regions with differential aberrations between HS-C and LS-C were uploaded into DAVID bioinformatics tool for functional enrichment analysis and GO biological processes (level 5). Enrichment p-values less than 0.01 were considered as statistically significant.
Mice xenograft models
Nonobese diabetic/severe combined immunodeficient (NOD/SCID) mice (Charles River) were injected with: a) 1–2.5 × 106 SJSA-1 subcutaneously (s.c.) into the right flank or 105 SJSA-1 intravenously (i.v.) into the tail vein to originate subcutaneous primary tumors and lung metastases, respectively, or b) 106 DMR s.c. or 4 × 105 DMR orthotopically into the uterine wall by laparotomy. Six mice per group (DMR s.c. model) were randomized to receive: a) 25 or b) 50 mg/kg/day intraperitoneal (i.p) injection of olaparib (5 days/week); c) 0.150 or d) 0.1 or e) 0.05 mg/kg weekly i.v. injection of trabectedin; f) combination of 25 mg/kg/day olaparib and 0.050 mg/kg trabectedin or g) left untreated for 21 days and then sacrificed for histological and molecular assays. SJSA-1 s.c. models were treated with a) 0.05 mg/kg or b) 0.025 mg/kg weekly i.v. injection of trabectedin; c) 25 mg/kg olaparib i.p; d) combination of 25 mg/kg/day olaparib and 0.050 mg/kg trabectedin; e) combination of 25 mg/kg/day olaparib and 0.025 mg/kg trabectedin or g) left untreated for 17 days. DMR orthotopic model was treated with 0.050 mg/kg trabectedin and 25 mg/kg/day olaparib as single agent, in combination, or left untreated. These in vivo experiments were conducted in accordance with the protocol approved by the Institutional Ethics Review Board (IRB) and by the Italian Ministry of Health (Aut. Min. 178/201S-PR). For in vivo imaging, each mouse received an intra-peritoneal dose of 150 mg luciferin/kg body weight, and ventral and dorsal images were obtained by IVIS Lumina II and quantified using Living Image software (PerkinElmer).
Flow cytometry assays: cell cycle, P-H2AX, and apoptosis analysis
The effects of trabectedin and olaparib on the cell cycle and the expression of the DNA-damage marker P-H2AX were determined by staining DNA content with propidium iodide (PI, Sigma Aldrich) and FITC-conjugated primary antibody anti-P-H2AX (Ser139) (Millipore), respectively, after treatment with trabectedin (0.125 nM) as single agent or in combination with olaparib (1.25 μM) for 24 h.
For apoptosis evaluation, after 96-h treatment APC-labelled Annexin V (eBioscence) and PI (0.5 μg/ml) staining was done. All samples were acquired by Cyan ADP Flow cytometer (Beckman Coulter) and analyzed by the Summit v4.3 (Dako) and FlowJo (Tree Star) softwares.
Comet assay (Single Cell Gel Electrophoresis assay)
For determination of DNA single- and double-strand breaks, a single-cell gel electrophoresis assay was used (Comet Assay TM, Trevigen) per the manufacturer’s instructions after 48-h treatment with trabectedin (0.125 nM) and olaparib (1.25 μM), as single agents or in combination. Quantization of the DNA in the tails of the comets was performed with Quantity One software (Bio-Rad Laboratories) and ImageJ software version 1.49 (
http://rsbweb.nih.gov/ij/index.html).
Lentiviral vector production
pCCL.sin.cPPT.polyA.CTE.eGFP.minhCMV.hPGK.Luciferase.Wprep or lentiviral plasmid vector PLKO-puro (Sigma-Aldrich), the Precision LentiORF PARP1 (Id:PLOHS_100004266, Thermo Scientific) and packaging vectors pMDLg/pRRE pRSV-REV and pMD2.VSVG were used for lentiviral preparation. Efficiency of transduction was confirmed by western blot analysis and flow cytometry analysis by Cyan ADP Flow cytometer (Beckman Coulter). The luciferase activity was tested by in vitro bioluminescent assay (Caliper Life Sciences, Inc.) and measured by IVIS Lumina II (PerkinElmer).
Archival tumor samples, immunostaining, and terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay
Archival formalin-fixed, paraffin-embedded (FFPE) tumor samples were collected in accordance with an IRB approved protocol (202/2014). Four μm slices were cut and stained with primary antibodies against PARP1, BRCA1 and RAD51 (Abcam) following standard immunohistochemistry procedure. The same antibodies and one antibody against PAR (Calbiochem) were used for immunocytochemistry on cells grown as monolayer in chamber slides. TUNEL assays were used to evaluate the number of apoptotic cells in tumors explanted from mice, using the ApopTag kit (Millipore). Immunostaining on mice xenografts was performed according to standard protocols with primary antibodies from Sigma Aldrich (proliferating cell nuclear antigen, PCNA), Abcam (p-H2AX Serine 139). Nuclei were counterstained with hematoxylin. Immunofluorescence was done with FITC-conjugated antibody against P-H2AX (Millipore). Nuclei were counterstained with DAPI. Fluorescence in situ hybridization (FISH) analysis of PARP1 and centromeres was performed following standard procedures.
Visible images were acquired with a Leica DM1000 microscope equipped with a color 3.1 M Pixel CMOS digital camera. Fluorescent images were acquired with a Leica confocal laser-scanning microscope (TCS SP5 AOBS). For signal quantification, 5 images/sample were acquired by maintaining constant parameters. Image quantification was performed using ImageJ software and Leica Confocal Software.
Statistical and pharmacological combination analyses
All in vitro experiments were performed at least three times. Differences between treatment groups were analyzed by the two-tailed Student’s t test and the two-way ANOVA with post hoc Bonferroni’s correction for multiple tests using GraphPad Prism 5 (GraphPad Software Inc.). The concentration inhibiting 50% of the cell growth (IC50) with its 95% confidence intervals (95% CI), and drug synergism, expressed as combination index (CI) calculated at IC50 with its estimated standard deviation were obtained by using CalcuSyn software (Biosoft). The correlation analyses between ΔCT values (mRNA levels) or protein expression levels and combination indexes were performed by calculating Pearson correlation coefficient, t-distribution, and P values by Microsoft Excel.
Discussion
We identified and demonstrated that PARP1 basal expression is crucial in the mechanism behind the synergy between PARP1 inhibitors and trabectedin independent of BRCA1/2 status, in several robust models across different histotypes. The chemo-sensitization observed in high-PARP1-expressing cells suggests that PARP1 inhibitors may extend and improve the clinical application of DNA-repair targeting when combined with other cytotoxics also.
Our findings support a reappraisal of PARP1 inhibitors as chemo-sensitizers. In general, after chemotherapy exposure, tumor cell fate is highly dependent on the efficiency of the same DNA-damage response pathways that restore normal cell DNA integrity. This observation fueled the idea to target key points in DNA-repair machinery to increase cytotoxicity [
2,
4] and brought about the development of several compounds, among which PARP1 inhibitors eventually demonstrated clinical efficacy [
4]. However, so far, their clinical use is limited to tumors harboring hereditary or acquired genetic defects in homologous recombination (HR) exploiting an intrinsic cell weakness through so-called “synthetic lethality” [
28‐
34]. Interestingly, patients affected by HR-defective tumors have the greatest benefit from treatment with DNA-damaging agents, such as platinum compounds and trabectedin [
37,
38,
47].
We pursued to induce a sort of “chemical synthetic lethality” also in HR-proficient tumors, by combining trabectedin with PARP1inhibitors. Our hypothesis stemmed from the observation that trabectedin not only induced a peculiar DNA damage but also caused impairment in the DNA-repair machinery that might become lethal in presence of PARP1 inhibitors [
35].
Demonstration that trabectedin strongly activates PARP1 enzyme strengthened our idea to combine this chemotherapy with PARP1 inhibitors. However, after trabectedin treatment, we observed significantly different PARylation across various histotypes. For example, single nucleotide polymorphism in the PARP1 gene (Val762Ala) is known to result in a less efficient PARP1 variant [
48,
49], which may partly explain differences in the degree of PARylation. To understand the impact of PARylation differences, we turned to PARP1 basal expression and proved that PARP1 protein basal level was not only directly proportional to chemotherapy-induced PARP1 activation, but also that it dictated the synergism with trabectedin and several other chemotherapy compounds as well. Indeed, PARP1-depleted cells were tolerant to PARP1 inhibitors [
50,
51], and therefore, PARP1 activity is the prerequisite to induce a significant amount of complexes formed by PARP1, damaged DNA and PARP1 inhibitor that are plausibly more cytotoxic than unrepaired single-strand breaks alone [
52‐
54]. Clearly, trabectedin synergized with both tested PARP1 inhibitors, but we found olaparib was significantly more potent than veliparib. Our result reflects the fact that veliparib is a pure catalytic inhibitor of PARP1 activation, whereas olaparib is a poisoning drug that blocks activated PARP1 enzyme at the site of damage (PARP1-trapping activity) impeding the subsequent recruitment of repairing machinery [
11,
52,
55]. We studied the synergism of PARP1 inhibitors and trabectedin in a large set of different tumor cell lines and in in vivo models, demonstrating that PARP1 inhibition improved the antitumor activity of trabectedin in a cell-line dependent intensity, as previously shown in breast cancer and in Ewing’s sarcoma cell lines [
44,
55]. Among the tested BSTS histotypes, Ewing’s sarcoma cells were the most sensitive to the combination, a result that might be explained by the known exquisite sensitivity of the pathognomonic fusion protein EWS/FLI1-expressing cells to both trabectedin and PARP1 inhibition [
10,
56‐
59].
Another important datum emerged from study of the MES-SA-DX5 model. This leiomyosarcoma cell line had been previously made resistant to doxorubicin [
60], and interestingly, we found it also shared cross-resistance to trabectedin. Of note is that the addition of PARP1 inhibitors restored sensitivity to clinically achievable concentration of trabectedin. The increased activity in MES-SA-DX5 can be explained by comparing the genomic and protein status of this cell line with its parental counterpart. Indeed in MES-SA-DX5, the PARP1 gene was amplified and consequently caused higher PARP1 expression and activity that confirmed the mechanistic role played by PARP1 in determining synergy between trabectedin and PARP1 inhibitors. These consistent observations further prove the key role played by PARP1 expression.
Given the redundant complexity of DNA repair machinery, we took advantage of gene array analyses to delve deeper into the mechanisms behind trabectedin + olaparib synergism. We identified specific gene sets involved in the combination synergism. These genes belong to cell cycle control such as normally activated G2/M cell cycle checkpoints and DNA-damage response and repair pathways (i.e., Timeless, AURKA, MCM2, RAD54L, POLD1, MRE11A) [
61‐
64]. Moreover, specific genomic aberrations of these genes differentially characterized HS-C and LS-C. We showed that enzymes involved in DNA-damage response and repair were actually expressed more in cell lines displaying higher PARP1 basal expression and high trabectedin + olaparib synergism. The cell cycle phase is crucial for trabectedin and olaparib response and these findings confirmed our data on cell cycle perturbation. Undeniably, trabectedin-induced DNA damage activates the DNA-damage response leading to G2/M cell cycle arrest [
35]. However, in this phase, repairing enzymes may still fix the damage caused by either single agent, such that tumor cells survive after initial cell cycle arrest. On the contrary, when the combination was used, DNA repair and cell survival were impeded because of irreparable DNA fragmentation and apoptosis.
Again the intensity of this event directly correlated with PARP1 expression. We carefully validated the role of identified candidate genes in two independent panels of 20 and 11 different cell lines. We selected key enzymes of DNA-damage response and repair pathways and investigated their expression by real-time PCR in the first 20 BSTS cell-line panel. We confirmed a statistically significant direct correlation between PARP1 expression and trabectedin + olaparib synergism. Interestingly, we also observed a significant correlation between drug synergism and RAD51 and BRCA1 expression, which might be explained by the fact that in HR-proficient cells, PARP1 expression drives the transcription of BRCA1 and RAD51 genes co-activating E2F transcription factor [
65‐
69]. If so, then the statistical correlation between combination synergism and RAD51 or BRCA1 expression might simply reflect the level of PARP1 expression. These data are consistent with previous observations of correlation between DNA repair protein expression and sensitivity to PARP1 inhibition [
70]. Nonetheless, in HR-defective cells PARP1 activity was found increased [
54]. However, we did not find any correlation between the genomic status of key DNA repair genes and the synergism of the combination. To strengthen our observation and generalize our hypothesis on the role of PARP1 expression also in non-sarcoma cells, we replicated our experiments in a second independent panel (11 cell lines of different tumor origin) and confirmed the correlation between trabectedin + olaparib synergy and PARP1 expression also at the protein level. The functional role of PARP1 was further validated in silencing and overexpression experiments that confirmed the data reported above. Noteworthy, the modulation of PARP1 expression or the inhibition of its activity might modulate the nuclear factor Kappa-b pathway that is a crucial determinant of drug resistance [
71‐
73]. This intriguing aspect warrants further investigations.
Acknowledgements
The authors would like to thank Joan C. Leonard for language editing; Roberta Porporato, Daniela Cantarella, Barbara Martinoglio and Stefania Giove for technical support; Claudio Isella for helpful discussion. Trabectedin was kindly provided by PharmaMar (Madrid, Spain).