Background
Colorectal cancer (CRC) is the third most frequent malignancy and the second leading cause of cancer death worldwide, accounting for more than 1.8 million new cases and 880,000 deaths in 2018 [
1].
Despite the increase in therapeutic options for the treatment of metastatic CRC (mCRC), cytotoxic chemotherapy using fluoropyrimidines, oxaliplatin, and irinotecan still constitutes the mainstay of the treatment [
2]. These genotoxic drugs function by inducing direct or indirect DNA damage, which lead to cell cycle block and cell death. Nevertheless, DNA damage is recognized by different DNA repair pathways such as mismatch repair (MMR), nucleotide excision repair (NER), and homologous recombination repair (HRR) [
3]. All these mechanisms of DNA damage response (DDR) are crucial for the survival of normal cells and are often deregulated in cancer, allowing for the accumulation of mutations that are ultimately associated to cancer progression and therapeutic resistance [
4]. For these reasons, blocking DDR has been widely investigated in the context of cancer treatments as a strategy to potentiate radiotherapy/chemotherapy-induced DNA damage and overcome drug resistance. Indeed, several compounds that interfere with the different mechanisms of DNA repair are already available or in advanced clinical testing [
5]. Among these compounds, Poly-ADP-Ribose Polymerase inhibitors (PARPi) were the first to be approved in clinical practice and exhibit a strong activity in ovarian and in other cancers characterized by a functional impairment of HRR known as homologous recombination deficiency (HRD) [
6‐
8]. In the last years, several biomarkers were used to predict HRD and sensitivity to PARP inhibitors, most of them focused on finding genetic alteration in HRR-associated genes (mainly BRCA1/2, but also ATM, PALB2 and others) and/or on analyzing the
genomic scars associated to HRD (loss of heterozygosity, large scale transitions, subchromosomal allelic imbalance) [
9]. Moreover, mutational signature single base substitution 3 (SBS3) has been associated to HRD and correlates with PARPi sensitivity [
10].
Initial preclinical data reported PARPi efficacy also in CRC cell lines, apparently increased in case of microsatellite instability (MSI). Notably, in the largest analyzed preclinical cohort, including more than 100 human CRC cell lines, no clear association was evidenced between olaparib sensitivity (13% of MSS cell lines screened) and either mutations in HR genes or mutational signatures [
11]. However, a phase II clinical trial with olaparib did not show any benefit in chemorefractory mCRC patients [
12,
13]. In addition, several preclinical reports evidenced promising synergism of PARP inhibitors in combination with oxaliplatin or irinotecan in CRC, independently from microsatellite status [
14‐
16]. However, similarly to previous cases, a phase Ib clinical trial failed to identify any benefits for the combination of the PARP inhibitor olaparib and irinotecan in chemorefractory mCRC [
17]. Very disappointedly, this study evidenced a high-grade hematological toxicity that led to a significant dose reduction, which probably hampered the efficacy in this unselected population [
17]. These results might be explained by the low prevalence of biallelic loss in genes involved in homologous recombination in CRC, which is less than 3% versus more than 50% in ovarian cancer [
18]. Nevertheless, 26% of mCRC patients from the large MSK IMPACT database exhibit at least one mutation in HRR genes, the most frequent being ATM (8%) and BRCA2 (8%) [
19]. Taking this into account, in this study we hypothesized that HRR genes mutations might represent a vulnerability to the combination of PARPi/chemotherapy in CRC. Thus, we have systematically evaluated the synergism between the PARPi niraparib and three genotoxic agents approved for mCRC in a panel of 12 human CRC cell lines representative of the main molecular subtypes, showing that irinotecan is the best candidate for combination therapy. Interestingly, in our work we have identified ATM mutations as a common genetic background associated with niraparib/ irinotecan combination efficacy. We further confirmed that a low dose combination is very effective in an in vivo model and in primary 3D cultures obtained from fresh CRC surgical specimens, which are mutated in ATM or in its downstream pathway. Moreover, we showed that a malfunction of HRR due to heterozygous mutations in ATM is responsible for such effects.
Methods
Research resource identifiers (RRIDs)
In order to support rigor and transparency in this publication, key resources such as Antibodies, Model Organisms, Cell Lines, and Softwares have been matched to their unique identified from the RRID portal (
https://scicrunch.org/resources).
Drugs and chemicals
Niraparib (Cat # S2741) and SN38 (Cat # S4908) were purchased from Selleckchem. Both drugs were dissolved in sterile DMSO at 10 mmol/L stock solution concentration and stored in aliquots at − 20 °C. Methylcellulose (Methocel, Cat #64632) was purchased from Sigma-Aldrich. Irinotecan, oxaliplatin and 5-fluorouracil were kindly provided by the hospital pharmacy service of the Oncology Unit of University of Campania. Working concentrations were diluted in culture medium just before each experiment. SN38 was used for in vitro experiments as the active metabolite of irinotecan, while irinotecan was used for mice xenograft experiments. Niraparib for animal studies was resuspended in 0.5% w/v methylcellulose, while irinotecan was resuspended in sterile saline for infusions.
Cell line cultures and mutational profiles of cell lines
Human HCT15 (RRID:CVCL_0292; ATCC Cat# CCL-225), LOVO (RRID:CVCL_0399; ATCC Cat# CCL-229), SW1116 (RRID:CVCL_0544; ATCC Cat# CCL-233), LS1034 (RRID:CVCL_1382; ATCC Cat# CRL-2158), SW403 (RRID:CVCL_0545; ATCC Cat# CCL-230), SW948 (RRID:CVCL_0632; ATCC Cat# CCL-237), CACO2 (RRID:CVCL_0025; ATCC Cat# HTB-37) and WIDR (RRID:CVCL_2760; ATCC Cat# CCL-218) authenticated colorectal cancer cell lines were obtained from the American Type Culture Collection (ATCC). The human SW48 (RRID:CVCL_1724; Cat# HTL99020), HCT116 (RRID:CVCL_0291; Cat# HTL95025), SW480 (RRID:CVCL_0546; Cat# HTL99017) cell lines were obtained from Istituto di Ricovero e Cura a Carattere Scientifico (IRCCS) “Azienda Ospedaliera Universitaria San Martino-Istituto Nazionale per la Ricerca sul Cancro, Genova,” Italy. LIM1215 (RRID:CVCL_2574) CRC cell line was obtained from Dr. F. Di Nicolantonio (Candiolo National Cancer Institute, Candiolo, Italy) and authenticated by IRCCS “Azienda Ospedaliera Universitaria San Martino-IST Istituto Nazionale per la Ricerca sul Cancro, Genova,” Italy. Cells were grown in RPMI- 1640, DMEM/F12, EMEM or McCoy medium (Lonza), supplemented with 10% FBS and 1% penicillin/streptomycin, in a humidified incubator with 5% of carbon dioxide (CO
2) and 95% air at 37 °C and were routinely screened for the presence of mycoplasma (Mycoplasma Detection Kit; Roche Diagnostics). Microsatellite status and transcriptional profiling according to the consensus molecular subtypes (CMS) were obtained from the work of Sveen and colleagues [
20]. Mutational profiles in 29 relevant homologous recombination repair genes [
18] were obtained from cBio-portal [
21,
22] using the Cancer Cell Line Encyclopedia (CCLE) dataset [
23,
24] [last accessed June 20th 2020]. Functional prediction for ATM and BRCA2 mutations were obtained using FATHMM algorithm integrated in COSMIC [
25] or Leiden Open Variants Database (LOVD) [
26] [last accessed July 1st 2020].
Proliferation and colony assays
Cell proliferation was analyzed by the MTT assay (Sigma-Aldrich), according to manufacturer’s instructions. Briefly, for each cell line 2–10 × 10
3 cells/well were plated in 48 multiwell plates. After 24 h, cells were treated with different concentrations of niraparib, 5-fluorouracil, oxaliplatin, SN38 or niraparib/chemotherapy combination for 96 h. The IC
50 values were determined by using the CompuSyn 1.0 and plotted in dose response curves using Graphpad Prism 8.0 (RRID: SCR_002798). Results represent the median of the three experiments, each performed in triplicate. Combinations were performed according to IC
50 ratio, as described by the Chou-Talalay model [
27], and combination index was obtained using CompuSyn 1.0 (Combosyn Inc.).
Colony formation assay was performed to evaluate the long-term proliferative potential of cell lines following treatments with niraparib, SN38 or their combination at different concentration ratios (100:1; 50:1) and dose levels. For each experiment, 3–15 × 103 cells/well were seeded in 6-well plates and incubated with the drugs in serum-containing medium for 24 h. The medium was then replaced with fresh culture medium every 3 days. After 14 days, cells were fixed with 4% paraformaldehyde at room temperature (RT) for 15 min, stained with 0.1% crystal violet and counted using ImageJ (RRID:SCR_003070). Results represent the median of at least two separate experiments, each performed in duplicate.
Mice xenografts
Four- to six-week-old female athymic nude mice (NU-Foxn1nu, IMSR Cat# CRL:194, RRID:IMSR_CRL:194) were purchased from the Charles River Laboratories. LS1034, CACO2, HCT116, WIDR and SW48 human colorectal cancer cell lines were used. A total of 3–5 × 106 cells was resuspended in 200 μL of Matrigel (BD Biosciences) and PBS (1:1) and implanted subcutaneously into the right flank of 20 mice for each cell line. Once tumors reached a volume of 75–100 mm3, mice were randomized to 4 arms each of 5 mice: control arm to receive vehicle alone (0.5% methylcellulose per os using oral gavage, 5 days a week + PBS intraperitoneally, 2 days a week), irinotecan arm (10 mg/kg intraperitoneally, 2 days a week), niraparib arm (50 mg/kg per os using oral gavage, 5 consecutive days a week), and their combination. Treatment was continued for a total of 4 weekly cycles. Tumor measurements were performed twice a week using a caliper, tumor volumes were calculated using the formula: V = (W2 × L)/2. Relative tumor volume (RTV) was calculated for each tumor relative to day 1 of treatment. After treatment end, mice were followed up for survival analysis up to 100 days. Mice were euthanized in case of tumor volume > 2000 mm3, tumor ulceration or onset of distress.
Immunofluorescence
25 × 103 cells/well were seeded on a coverslip in a 12-multiwell and treated for 24 h with 100 nmol/L niraparib, 1 nmol/L SN38 or their combination. After 24 h, cells were washed in PBS and fresh culture medium without drugs was added in order to allow DNA damage recovery. After additional 24 h, cells were washed in PBS, fixed with 4% paraformaldehyde (PFA) solution and permeabilized with 0.1% TRITON -X/PBS solution, then blocking was performed in 1% BSA for 1 h at RT. Cells were incubated with rabbit primary anti-RAD51 antibody (Abcam Cat# ab133534, RRID:AB_2722613) and mouse anti-phospho-Histone-H2AX antibody (Millipore Cat# 05–636-I, RRID:AB_2755003) in PBS for 90 min. Secondary goat anti-rabbit TRITC-conjugated (Abcam Cat# ab6718, RRID: AB_955551) and donkey anti-mouse FITC-conjugated (Abcam Cat# ab150105, RRID:AB_2732856) antibodies were added after a PBS wash in the same conditions. Cells were incubated in a 1:500 solution of 10 mg/mL Hoechst (Invitrogen) in PBS for 10 min in the dark. Images were collected under a fluorescence microscope (EVOS FL Cell Imaging System, Thermo Scientific, Rockford, USA). Each experiment was performed in quadruplicate and at least 100 nuclei were considered in each replicate. ImageJ (Fiji plugin, RRID:SCR_002285) was used to generate merged images and quantify colocalization puncta.
Cell cycle analysis
50 × 103 cells/well were seeded in six-well plates using FBS-containing medium. After 24 h cells were serum starved for 24 h and then treated for 24 h in serum-containing medium with 100 nmol/L niraparib, 1 nmol/L SN38 or their combination. Cell pellets were harvested in phosphate-buffered saline (PBS) containing 2 mM EDTA, washed once with PBS, fixed in iced ethanol 70%, washed with PBS and incubated with 25 μg/ml PI plus RNase (Invitrogen) 1 mg/ml for 120 min at 4 °C in the dark. Stained nuclei were analysed with FACS Aria III (Becton and Dickinson, Mountain View, CA, USA), and data analysed using ModFit 2.0 cell cycle analysis software (ModFit LT, Verity Software House, Topsham, UK. RRID:SCR_016106).
3D primary cell cultures
Fresh tissue specimens, derived from primary or metastatic colorectal cancers from patients enrolled in the I-CURE project (Regione Campania), were transported to the laboratory within 2 h from surgical harvesting. The tissues were weighed, washed twice with phosphate buffered solution (PBS) and cut in fragments. Briefly, tumor fragments were incubated with a shake with digestion medium (DMEM F-12, Sigma-Aldrich) supplemented with 2% Penicillin/Streptomicin, 10X Amphotericin, 2X Collagenase and Hyaluronidase for up 6 h in a 37 °C. All undigested fragments and debris were filtered through a cell strainer (BD-Falcon) after digestion followed by centrifugation for 5 min at 300 rcf. The supernatant was removed, and the pellet was washed with PBS and then centrifuged as described above. The pellet was further re-suspended in ice-cold 1:1 mixture of growth medium and Matrigel (BD-Falcon) and then seeded in 24 well-plates (Corning). The matrigel droplets were polymerized for 30 min a 37 °C and growth medium was added after polymerization. All patient-derived tumor spheroids originated from primary colonic cancer surgical samples from untreated (chemo-naïve) patients, with the exception of IC-001 that was generated from an abdominal wall metastasis from a patient previously treated with 5FU, oxaliplatin and irinotecan. For drug screening experiments, 10–20 × 103 tumor spheroids/well were plated as described in 24 multiwell plates; after 48 h, spheroids were treated with 1000 nM niraparib, 10 nM SN38 or their combinations. Growth inhibition was performed using MTT assay (Sigma-Aldrich), normalized on untreated control, after 14 days of treatment. Matrigel was degraded using Cell Recovery Solution (BD-Falcon) according to the manufacturer’s procedures and pellets collected for absorbance detection. These experiments were performed in triplicates.
Genomic profiling of patients’ tumor specimens
FoundationOne® (F1CDx) was performed in a single site at Foundation Medicine, using adequate tissue specimens from FFPE blocks to provide a minimum yield of 55 ng of genomic DNA to ensure enough DNA for quality control (QC) and to proceed with library construction. In total, the assay detects alterations in 324 genes. Using the Illumina HiSeq 4000 platform, hybrid capture-selected libraries are sequenced to high uniform depth (targeting >500X median coverage with > 99% of exons at coverage > 100X). Additionally, genomic signatures including MSI and TMB are reported. To determine MSI status, 95 intronic homopolymer repeat loci (10–20 bp long in the human reference genome) with adequate coverage on F1CDx Assays are analysed for length variability and compiled into an overall MSI score via principal components analysis. Each sample is assigned a qualitative status of microsatellite instable (MSI) or microsatellite stable (MSS). Tumor Mutational Burden (TMB) is a quantitative index of the number of mutations present in the cancer genome. TMB is measured by counting all synonymous and non-synonymous variants present at 5% allele frequency or greater and filtering out potential germline variants according to published databases of known germline polymorphisms including Single Nucleotide Polymorphism database and Exome Aggregation Consortium. The resulting mutation number is then divided by the coding region corresponding to the number of total variants counted or 793 kb. The derived number is communicated as mutations per Mb unit (mut/Mb): low TMB for 1–5 mut/Mb, intermediate TMB for 6–19 mut/Mb, high TMB for ≥20 mut/Mb). Approved results are annotated by automated software with CDx relevant information and are merged with patient demographic information.
Statistical analyses
All statistical analyses were performed using Graphpad Prism 8.0 software (RRID: SCR_002798). Distribution of IC50 or combination index values according to molecular features was calculated using Wilcoxon-Mann-Whitney test. Quantitative in vitro and in vivo data were reported as mean ± standard deviation (SD). Results were compared by analysis of variance (ANOVA), and a p value < 0.05 was considered statistically significant. Survival analysis was carried out visually by means of Kaplan-Meier curves, while the difference in survival in each arm was calculated using the log-rank test.
Discussion
In our work, we have analyzed the activity of the combination between standard-of-care chemotherapeutics used in CRC and niraparib in a panel of 12 cell lines that encompasses the 4 consensus molecular subtypes, the different microsatellite status, and the different KRAS and BRAF mutational profiles. Niraparib was selected for this study due to its optimal pharmacodynamic and pharmacokinetic properties in patients, showing a strong PARP trapping activity (stronger than olaparib), but also a long half-life (> 24 h) and a wide therapeutic interval [
29,
33]. Niraparib in vitro sensitivity was in line with previous reports in CRC cell lines, with mean IC
50 values above the trough plasma concentration of this drug at the commonly used dose of 300 mg/die in monotherapy trials (Table
1, Fig.
1a), showing no association with microsatellite status (Supplementary figure
1) [
6,
16,
29]. Moreover, transcriptional classification using the CMS classifiers did not evidence any correlation with sensitivity to the PARPi (Supplementary figure
1). SN38, the active metabolite of irinotecan, was shown to be the best partner for combination with niraparib, and the sensitivity to this combination was independent from microsatellite status or CMS (Fig.
1b, Supplementary Table
1, Fig.
2). Mutational profiles in 29 relevant genes in the HRR pathway were obtained for the cell lines included in the panel from public databases and were correlated with sensitivity to niraparib or the niraparib/SN38 combination. Notably, ATM and BRCA2, that constitute the most frequently mutated HR genes in large mCRC patients datasets such as MSKCC or COSMIC [
19,
34], are also recurrently mutated in our panel (Table
1 and Supplementary Table 2). The functional prediction of these genetic alterations showed that in our cell panel all ATM mutations (5/5) were functionally relevant, compared with only 2/5 BRCA2-mutated cell lines carrying mutations with a pathogenic significance (Supplementary Table 3). However, in our panel, non-synonymous mutations in ATM or BRCA2 genes were not associated to a significant higher sensitivity to niraparib used as single agent (Supplementary figure
1). This may seem counterintuitive, as both ATM and BRCA2 inactivation are associated to homologous recombination deficiency and synthetic lethality with PARP inhibition [
7], and a previous work demonstrated that ATM depletion using shRNAs induces sensitivity to PARP inhibitors [
35]. However, it must be noted that a mutation in ATM does not correspond to ATM loss, though a decreased protein level was found in case of heterozygous ATM mutations in CRC cell lines compared to the wild type counterpart [
35]. Nevertheless, we evidenced a significant association between the presence of ATM mutations and synergism with the niraparib-SN38 combination, while no correlation is present with microsatellite status, CMS or BRCA2 mutations (Fig.
2).
One possible explanation for the increased sensitivity to the combination in ATM-mutated cell lines lies in the relative insufficiency of timely DNA repair in these cells in case of double strand break damage overload. This dysfunction is unveiled by the combination treatment but not by single agent SN38, since most SN38-induced single strand breaks can be repaired before precipitating into DSBs [
3]. Notably, both DSB recognition and HRR are intact in ATM-mutant cell lines, as evidenced by γH2Ax and Rad51 immunostaining (Fig.
4a-b), though persistent HR engagement upon combination treatment reflects a delay in the resolution of DSBs 24 h after treatment washout in case of heterozygous ATM mutations (Fig.
4c-d), possibly unveiling a haploinsufficient phenotype. The delay in repairing DSBs and the following cell cycle arrest are also evident at the cell cycle analysis in HCT116 compared to LS1034 cell lines (Supplementary Figure
3).
In this framework, the inhibition of both PARP and topoisomerase 1 is fundamental to achieve the effect in ATM-mutated CRCs, even if myelotoxicity characterizes a major concern that has hampered the development of other PARP inhibitors/chemotherapy combinations [
17]. For this reason, we investigated whether the molecular subgroup characterized by the presence of pathogenic mutations in ATM exhibits an increased sensitivity to the niraparib/irinotecan combination at lower doses. Indeed, low drug concentrations of niraparib and SN38 induced a clonogenic arrest in vitro in synergistic cell lines (HCT15, LOVO, LIM1215, SW948, HCT116, SW480) (Fig.
3a-b). In mice models, low doses of niraparib and irinotecan were effective and synergistic in ATM-mutated HCT116 xenograft, in which they induced complete tumor regressions in 60% of treated mice (Fig.
3e). The low doses used for the mice experiments are predicted to achieve plasma concentrations of 100 nM for niraparib and less than 25 nM for SN38, concentrations that are way below the desirable ones in monotherapy, achieved with oral doses of 40 mg/die for niraparib and less than 100 mg/iv for irinotecan [
29‐
32,
36].
Finally, we have also confirmed our findings in a relevant translational model using patient-derived 3D spheroids, in which the presence of a pathogenic mutation in ATM or in its downstream effector CHEK2 confer increased sensitivity to the combination of niraparib and SN38 (Fig.
5).
Recently, several papers have investigated the role of ATM mutations in cancer susceptibility and prognosis, suggesting that such mutations induce an attenuated cancer-predisposing phenotype in heterozygote carriers, that represent up to 1–2% of the general population [
37], while presenting a favorable prognostic effect in CRCs bearing such heterozygous alterations [
38]. Taken together, these evidences underpin a possible role for ATM heterozygous mutations in cancer therapy [
39,
40]. Recently, a phase I study investigating different schedules for the combination of irinotecan and the PARP inhibitor rucaparib in refractory cancers bearing mutations in HRR genes was presented, showing how pulse schedules are clinically feasible and that the patient population with ATM-mutated cancers exhibits the greatest benefit from the combination [
41].
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.