Abstract
Therapeutic strategies targeting DNA repair pathway defects have been widely explored, but often only benefit small numbers of patients. Here we characterized potential predictive biomarkers for treatment with AsiDNA, a novel first-in-class DNA repair inhibitor. We evaluated genetic instability and DNA repair defects by direct and indirect assays in 12 breast cancer cell lines to estimate the spontaneous occurrence of single-strand and double-strand breaks (DSB). For each cell line, we monitored constitutive PARP activation, spontaneous DNA damage by alkaline comet assay, basal micronuclei levels, the number of large-scale chromosomal rearrangements (LST), and the status of several DNA repair pathways by transcriptome and genome analysis. Sensitivity to AsiDNA was associated with a high spontaneous frequency of cells with micronuclei and LST and specific alterations in DNA repair pathways that essentially monitor DSB repair defects. A high basal level of micronuclei as a predictive biomarker for AsiDNA treatment was validated in 43 tumor cell lines from various tissues and 15 models of cell- and patient-derived xenografts. Micronuclei quantification was also possible in patient biopsies. Overall, this study identified genetic instability as a predictive biomarker for sensitivity to AsiDNA treatment. That micronuclei frequency can be measured in biopsies and does not reveal the same genetic instability as conventional genome assays opens new perspectives for refining the classification of tumors with genetic instability. Cancer Res; 77(16); 4207–16. ©2017 AACR.
Introduction
Alterations in the DNA damage response (DDR) and repair machinery enhance genetic instability, a hallmark of tumor cells. These features are widely studied and exploited to treat cancer, as they provide the opportunity to specifically kill tumors. Normal cells display functional DDR machinery, with active cell-cycle checkpoints, allowing adequate DNA damage repair to ensure their genome stability. However, cancer cells sustain their hyper proliferative status by having only some of these checkpoints, raising the possibility to specifically kill tumors by targeting the remaining functional checkpoints and DNA repair pathways (1). The importance of an efficient DDR for maintaining genome integrity and counteracting cancer is highlighted by the fact that many hereditary defects in this machinery predispose individuals to cancer (2), establishing a clear link between genomic instability and tumor development.
Cells are exposed daily to endogenous and exogenous genotoxic agents that induce various types of DNA lesions. The most frequent are oxidative damage and single-strand breaks (SSB). If this damage is not repaired, it can lead to collapsed replication forks, giving rise to double-strand breaks (DSB), the most lethal type of DNA damage (3). DSBs are mostly repaired by two distinct DNA repair pathways, homologous recombination (HR) and nonhomologous-end-joining (NHEJ). The repair process of DSBs is initiated by a sensing and signaling step, essential for the recruitment of downstream repair proteins (4). This step leads to the phosphorylation of H2AX histone proteins into γH2AX, a very sensitive surrogate biomarker for DSBs, which remains phosphorylated if the DSBs are not repaired due to defects in the DSB repair machinery. This DSB biomarker however has several limitations. One of the most important is that it does not efficiently mark heterochromatin DSBs (5, 6), the most dangerous lesions as they are slowly and less efficiently repaired than euchromatin DSBs (7, 8). Thus, the identification of a more reliable biomarker of DSBs and their repair defects has drawn increasing interest during the last decade. Micronuclei are small pieces of DNA resulting from unrepaired DSBs or mitotic spindle damage that appear near the nucleus following cell division (9). They are commonly used to quantify the genotoxicity of chemical and physical compounds (10). Various techniques have also been developed to measure the activity of DSB repair pathways for the detection of eventual alterations in these important cellular mechanisms. More recently, transcriptomic signatures of HR deficiency (HRD; refs. 11, 12), and SNP array-based quantification of large genomic rearrangements, also known as large-scale state transitions (LST; refs. 13, 14), have allowed broader and wide-scale understanding of DSB repair defects. Furthermore, LSTs provide information about the HRD status of tumors.
Genetic instability is not only a boon for cancer development, but also the Achilles' heel of tumors. The development of PARP inhibitors to treat patients with HRD tumors (harboring BRCA1 or BRCA2 mutations) is the first example of exploiting DNA repair defects (15–17). Despite the good safety and efficacy of these specific inhibitors, their limitation is that they target only a small population with a specific genetic defect, and tumors can easily acquire resistance, after a short period of treatment, by restoring HR repair, hyperactivating NHEJ repair, or overexpressing Pgp drug efflux pumps (18). The hurdles faced by these well-tailored therapies necessitate the development of a new family of DNA repair inhibitors with a wider spectrum of action to avoid the emergence of resistance. We have thus developed an original class of DNA repair pathway inhibitor, Dbait (DNA Bait), a double-stranded DNA molecule consisting of 32 base-pairs with a linker on one side, to stabilize the double-strand structure, and a free end on the other to mimic a DSB (19). AsiDNA is a molecule of the Dbait family with a cholesterol vector linked to its 5′-end to allow spontaneous delivery into cells (Supplementary Fig. S1). AsiDNA molecules are recognized by the DNA-PK proteins as DSBs. This induces hyperactivation of DNA-PK and its downstream targets, such as H2AX proteins, leading to a pan-nuclear γH2AX signal (20). This false damage signaling inhibits the recognition of genomic DSBs and the recruitment of repair proteins for efficient repair (19, 20). AsiDNA also acts as a lure for PARP1 proteins, which are also hyperactivated, leading to the sequestration of proteins acting in the same pathway (XRCC1, PCNA; ref. 21). This strategy sensitizes tumors to chemo- and radiotherapy (19, 22–24). The first phase I clinical trial combining AsiDNA to radiotherapy to treat patients with skin metastases of melanoma showed encouraging results, with a 30% complete response rate (25).
We have recently demonstrated that AsiDNA monotherapy shows efficacy in cells and tumors (26). All cell lines deficient in HR via inactivation of BRCA function were highly sensitive to AsiDNA, but this sensitivity did not appear to be restricted to HR defects, as some of the BRCA+/+ cell lines were also highly sensitive. It may thus be informative to screen tumor cells for specific defects that can predict sensitivity to AsiDNA. Indeed, genetically unstable cell lines and tumors are highly dependent on DNA repair pathways, and inhibition of these pathways can have a dramatic effect on their viability (27). We analyzed genetic instability using various tests to identify predictive biomarkers to AsiDNA treatment as it acts by inhibiting DNA repair pathways.
Materials and Methods
Cell culture
Cell cultures were maintained for four BRCA-deficient breast cancer cell lines [BC227, BRCA2-deficient, from the Institut Curie (28), and HCC1937, HCC38, and MDA-MB-436, BRCA1 deficient], eight BRCA-proficient breast cancer cell lines [BC173 from Institut Curie (28), and BT20, HCC1143, HCC1187, HCC70, MCF7, MDA-MB-231, and MDA-MB-468 purchased between 2006 and 2008], and three nontumor mammary cell lines (184B5, MCF10A, and MCF12A purchased between 2006 and 2008 from the ATCC); three human cervical cancer HeLa cell lines silenced for BRCA1 (HelaBRCA1SX, purchased in 2014 from Tebu-Bio, ref. 00301-00041), BRCA2 (HelaBRCA2SX, purchased in 2014 from Tebu-Bio, ref. 00301-00028), and control (HeLaCTLSX, purchased in 2014 from Tebu-Bio, ref. 01-00001); seven human glioblastoma cell lines (MO59K, MO59J, SF767, SF763, FOG, U87 MG, and T98G, given between 2008 and 2010 by P. Verrelle, Clermont-Ferrand); three human melanoma cell lines (SK28, SK28LshCTL, and SK28 LshDNA-PKcs; ref. 29); seven human colorectal cancer cell lines (HCT116, HT-29, CACO-2, SW480, SW620, CT329 × 12, and CR-LRB018, Institut Curie collection, obtained between 2009 and 2012); seven human hepatocellular carcinoma cell lines (HepG2, SNU-423, HLE, PLP-PRF-5, SNU-449, Huh-7, and Huh-6, purchased before 2012 from ATCC by J. Zucman-Rossi, Hôpital St Louis, Paris) and the human head and neck cancer cell line Hep-2 (purchased in 2014 from the ATCC). Cell lines were authenticated by short tandem repeat profiling (Geneprint 10, Promega) at 10 different loci (TH01, D21S11, D5S818, D13S317, D7S820, D16S539, CSF1PO, AMEL, vWA, TPOX). Cell lines were verified to be negative for Mycoplasma contamination using the VenorGeM Avance Kit (Biovalley). Cells were grown according to the manufacturer's instructions. Cell lines were maintained at 37°C in a humidified atmosphere at 5% CO2.
AsiDNA molecule
AsiDNA is a new chemical entity, a 64-oligodesoxyribonucleotide (nt) consisting of two 32-nt strands of complementary sequence connected through a 1,19-bis(phospho)-8-hydraza-2-hydroxy-4-oxa-9-oxo-nonadecane linker with a cholesterol at the 5′-end and three phosphorothioate internucleotide linkages at each of the 5′ and the 3′ ends (Agilent). The sequence is: 5′- X GsCsTs GTG CCC ACA ACC CAG CAA ACA AGC CTA GA L - CLTCT AGG CTT GTT TGC TGG GTT GTG GGC AC sAsGsC -3′, where L is an amino linker, X a cholesteryl tetraethyleneglycol, CL a carboxylic (hydroxyundecanoic) acid linker, and is a phosphorothioate linkage (Supplementary Fig. S1).
Measurement of cellular sensitivity to AsiDNA
AsiDNA cytotoxicity was measured by quantification of relative survival and cell death. Adherent cells were seeded in 24-well culture plates at appropriate densities and incubated for 24 hours at 37°C before AsiDNA addition. Cells and supernatant were harvested on day 10 of treatment, stained with 0.4% Trypan blue (Sigma-Aldrich), and counted with a Burker chamber. Cell survival was calculated as the ratio of living treated cells to living mock-treated cells. Cell death was calculated as the number of dead cells over the total number of counted cells.
Micronuclei detection
Micronuclei result from chromosomal breakage or spindle damage. They arise in the nuclei of daughter cells following cell division and form single or multiple micronuclei in the cytoplasm. For in vitro analysis, cells were grown on cover slips in a Petri dish. Cells were then fixed with PFA (4%), permeabilized with Triton (0.5%), and stained with DAPI (0.5 μg/mL). The frequency of micronuclei was estimated as the percentage of cells with micronuclei over the total number of cells. At least 1,000 cells were analyzed for each cell line. Micronuclei assessment in tumors was performed as follows: tumors were fixed in formalin and then embedded in paraffin. Sections were cut and stained with hematoxylin, eosin, and saffron. The percentage of micronuclei was estimated in the nonnecrotic and proliferative area by quantifying the cells presenting micronuclei in the cytoplasm, according to their described characteristics in at least 1,000 cells.
SNP array processing and evaluation of the number of LSTs
SNP arrays were processed using GAP methodology to obtain absolute copy number (CN) and allelic content profiles (30). LSTs were defined as a chromosomal breakpoint (change in CN or allelic content) between adjacent regions, each of at least 10 megabases (Mb). The number of LSTs, representing the number of breakpoints between large chromosome fragments, was calculated as previously described (13).
High-throughput data analysis
mRNA expression analysis.
mRNA expression data for the breast cancer cell lines were produced using Human Exon 1.0 ST Affymetrix microarrays. Raw data were RMA normalized and summarized with FAST DB annotation (version 2013_1; refs. 31, 32). Gene expression data were log2 transformed and the mean centered over all the cell line samples, which were then assigned into the two groups (MNHigh, MNLow). Genes associated with cell cycle and DNA repair pathways were retrieved from Atlas of Cancer Signalling Networks (https://acsn.curie.fr). The mean expression of these genes in MNHigh and MNLow samples was represented in heatmaps showing the distance to mean expression for each gene. The clustering has been performed using Euclidean distance and Ward agglomeration method.
Copy number data analysis.
The CN values for each gene for the cell lines was assessed by GAP analysis of the data generated on the Affymetrix Genome Wide SNP Array 6.0 (30) and corrected for ploidy. Two CN were considered to be “normal,” less than two CN as a loss and more than two CN as a gain. Each gene was then given a score using the average CN across samples of the cell lines in the same group (MNHigh, MNLow).
In vivo experiments
MDA-MB-231 and MDA-MB-468 cell-derived xenografts (CDX) were obtained by injecting 107 breast cancer tumor cells into the mammary fat pad of 6- to 8-week-old adult female nude NMRI-nu Rj:NMRI-Foxn1nu/Foxn1nu mice (Janvier). BC227 and BC173 breast cancer tumor models are patient-derived xenografts (PDX) and were established at the Curie Institute (France) as described in ref. 28. The uveal melanoma models were also PDXs (MM26, MP34, MP41, and MP55). Glioblastoma models were either PDXs (GBM14-RAV, TG1-HAM, ODA4_GEN) or CDXs (CB193, T98G, SF763, and SF767). The colorectal cancer (HT29) and skin melanoma models (SK28) were CDXs. For PDX engraftment, fragments of 30 to 60 mm3 were grafted into the mammary fat pad (BC227 and BC173) or right flank of 6- to 8-week-old female nude mice (Janvier). The animals were housed at least 1 week before tumor engraftment, under controlled conditions of light and dark (12-hour light/dark cycle), relative humidity (55%), and temperature (21°C). When engrafted tumors reached 80–250 mm3, mice were individually randomized into groups of 8–12 to different treatment groups. AsiDNA was injected locally (intratumoral and peritumoral subcutaneous administration). Tumor growth was evaluated three times a week using a caliper and calculated using the following formula: (length × width × width/2). Mice were followed for up to 6 months, and ethically killed when the tumor volume reached 2,000 mm3. The Local Animal Experimentation Ethics Committee approved all experiments.
Statistical analysis
Two-sided unpaired t tests were used for comparisons between micronuclei formation and tumor growth delay (TGD). TGD was calculated by subtracting the mean tumor volume quadrupling time of the control group from the tumor volume quadrupling times of individual mice in each treatment group. The mean TGD was calculated for each treatment group from the individual measurements. Statistical analyses were performed using GraphPadPrism (GraphPad Software, Inc.) and statEL software (Ad Science).
Results
Genetic instability is a predictive biomarker of response to AsiDNA in breast cancer cell lines
We assessed whether the level of spontaneous DNA damage could determine sensitivity to AsiDNA, as it is a DNA repair pathway inhibitor. We measured endogenous SSBs and base damage to reveal differences between cell lines using the alkaline COMET assay (which essentially measures SSBs) and ELISA to detect PAR levels (which measures the spontaneous activation of the PARP enzyme; Supplementary Fig. S2A and S2B). There was no correlation between comet tail moments and PAR content of the cells, although PARP activation, detected by PAR polymer formation, is thought to be mainly triggered by SSBs. This indicates that other factors, such as PARP and poly-ADP-ribose glycohydrolase (PARG) levels or other signaling enzymes, may affect the PARP-dependent signal (Supplementary Fig. S2E). Statistical analyses showed no correlation between survival in the presence of AsiDNA with either comet tail moments or PAR levels (Supplementary Fig. S2C and S2D), suggesting that the level of SSBs and oxidative damage to chromosomes is not predictive of sensitivity to AsiDNA.
DSBs are the most harmful type of DNA damage. Dbait inhibits both HR and NHEJ repair pathways (19, 20), and should thus be toxic to cells with a high basal content of DSBs. We tested the persistence of broken chromosomes and abnormal chromosome segregation in breast cancer cell cultures by investigating the presence of micronuclei after DAPI staining. Micronuclei were identified as small nuclear bodies containing DNA and chromatin located in the vicinity of the nucleus (33). Micronuclei assays have been developed in human lymphocytes to measure both whole chromosome loss and chromosome breaks upon genotoxic exposure. Micronuclei are also frequently found in some solid tumors (34). We analyzed the frequency of cells with micronuclei in growing cultures. Breast cancer cell lines showed micronuclei frequencies from 1% to 12% (Fig. 1B). The frequency of micronuclei in cell cultures highly correlated with sensitivity to AsiDNA (Fig. 1C; Spearman r = −0.77, P = 0.0008). Micronuclei appear after cell division. Importantly, there was no correlation between sensitivity of breast cancer cell lines to AsiDNA or micronuclei frequency and their rate of proliferation (Supplementary Fig. S3), indicating that sensitivity is due to defects in DSB repair (revealed by micronuclei) and not a high rate of cell proliferation.
Effects of endogenous DSB repair defects on sensitivity to AsiDNA. A, Effect of AsiDNA (4.8 μmol/L) on 15 malignant (six BRCA deficient, gray; nine BRCA proficient, black) and three nonmalignant (white) cell lines. Survival is defined as the percentage of living cells relative to nontreated cells (NT). B, Analysis of cell lines for LST or micronuclei (MN). LSTs, which indicate large genomic rearrangements, were assessed for breast cancer cells by SNP-array profiling (ND, not determined). Micronuclei were quantified in 1,000 cells cultured on cover slips after DAPI staining. Results are given as the percentage of cells harboring one or two micronuclei. C–E, Spearman correlation analysis between sensitivity to AsiDNA and basal levels of micronuclei (C), sensitivity to AsiDNA and number of LSTs (D), basal levels of micronuclei and number of LSTs (E). a.u., arbitrary unit.
Effects of endogenous DSB repair defects on sensitivity to AsiDNA. A, Effect of AsiDNA (4.8 μmol/L) on 15 malignant (six BRCA deficient, gray; nine BRCA proficient, black) and three nonmalignant (white) cell lines. Survival is defined as the percentage of living cells relative to nontreated cells (NT). B, Analysis of cell lines for LST or micronuclei (MN). LSTs, which indicate large genomic rearrangements, were assessed for breast cancer cells by SNP-array profiling (ND, not determined). Micronuclei were quantified in 1,000 cells cultured on cover slips after DAPI staining. Results are given as the percentage of cells harboring one or two micronuclei. C–E, Spearman correlation analysis between sensitivity to AsiDNA and basal levels of micronuclei (C), sensitivity to AsiDNA and number of LSTs (D), basal levels of micronuclei and number of LSTs (E). a.u., arbitrary unit.
We confirmed the micronuclei data by performing genome analysis to estimate the number of chromosomal breakpoints within a tumor cell population genome by measuring the level of LSTs, corresponding mainly to CN alterations (13). The breast cancer cell lines showed different levels of LSTs (Fig. 1B), reflecting large differences in genetic instability. The number of LSTs correlated with sensitivity to AsiDNA (Fig. 1D; Spearman r = −0.7, P = 0.0078). The frequency of cells with micronuclei highly correlated with the level of LSTs (Fig. 1E; Spearman r = 0.88, P < 0.0001). This result shows that the genetic instability in most of the cell lines, as measured by LSTs, which reflect the history of the accumulated events in the tumor cell genome during tumor development, correlates with their recent genomic instability and defects in DSB repair, revealed by the presence of micronuclei.
Mechanisms underlying micronuclei formation and sensitivity to AsiDNA
We investigated the mechanisms at the origin of micronuclei formation, a prerequisite for sensitivity to AsiDNA, by evaluating the performance of the DNA repair pathways and cell-cycle checkpoints of the breast cancer cell lines. For these analyses, we integrated mRNA expression and CN variations, of DNA repair pathways and cell-cycle checkpoints of the breast cancer cell lines. Gene expression analysis showed that cells that were MNLow and resistant to AsiDNA upregulated all DNA repair pathways and cell-cycle checkpoints, whereas these pathways were downregulated in the MNHigh cells (Fig. 2A). Moreover, we observed many genetic alterations in the DNA repair pathways and cell-cycle checkpoints of the MNHigh group, where CN losses were abundant (48% of genes with copy losses). There were very few alterations in the MNLow group, but they included CN gains, which could increase the performance of these pathways, and only 9% of genes with copy losses. Interestingly, 13 genes among these pathways allowed the breast cancer cell lines to cluster according to their micronuclei status (Fig. 2B). These genes were significantly differentially expressed between the MNHigh and the MNLow groups (fold change ≥ 1.5; P < 0.05), and belong essentially to the DSB repair pathways HR and NHEJ (BRCA1, POLD3, MRE11, DCLRE) and to the BER pathway (SMUG, MBD4). Defects in the DSB repair pathways may explain the high frequency of micronuclei and LSTs, and ultimately the sensitivity of these cell lines to AsiDNA.
DNA repair and cell-cycle pathway profiles of breast cancer cells with a MNHigh or MNLow status. A, Pathway-based heatmap showing the distance to mean expression for each pathway in MNHigh and MNLow breast cancer cell line samples. B, Hierarchical clustering of the breast cancer cell lines based on gene expression values of 13 DNA repair and cell-cycle genes. Clustering was done using Euclidean distance and Ward linkage. The color intensity indicates the level of change relative to the mean value of the breast cancer cell line group. CP, checkpoint; NER, nucleotide excision repair; BER, base excision repair; MMR, mismatch repair; SSA, single-strand annealing; A_NHEJ, alternative nonhomologous end-joining; C_NHEJ, classical nonhomologous end-joining; MMEJ, microhomology-mediated end-joining.
DNA repair and cell-cycle pathway profiles of breast cancer cells with a MNHigh or MNLow status. A, Pathway-based heatmap showing the distance to mean expression for each pathway in MNHigh and MNLow breast cancer cell line samples. B, Hierarchical clustering of the breast cancer cell lines based on gene expression values of 13 DNA repair and cell-cycle genes. Clustering was done using Euclidean distance and Ward linkage. The color intensity indicates the level of change relative to the mean value of the breast cancer cell line group. CP, checkpoint; NER, nucleotide excision repair; BER, base excision repair; MMR, mismatch repair; SSA, single-strand annealing; A_NHEJ, alternative nonhomologous end-joining; C_NHEJ, classical nonhomologous end-joining; MMEJ, microhomology-mediated end-joining.
Micronuclei are predictive biomarkers of AsiDNA sensitivity in all tumor type cell lines
Analyses of the genomes of multiple tumor types have shown that genomic alterations vary between tumors from different tissues of origin. We tested whether a high level of micronuclei could also predict the response to AsiDNA in cancer cell models other than breast cancer. We assessed 43 different cell lines of various cancer type (hepatocellular carcinoma, colorectal cancer, glioblastoma, melanoma, and head and neck cancer) for their sensitivity to AsiDNA and their basal level of micronuclei (Fig. 3A). All cell lines with a high level of micronuclei (>3.6% of cells with MN: MNHigh) showed higher sensitivity to AsiDNA treatment than those with low levels of micronuclei (<3.6% of cells with MN: MNLow), confirming the predictive value of micronuclei levels (P < 0.0001) in all the tested models (Fig. 3B).
Micronuclei (MN) are predictive biomarkers of AsiDNA treatment. A, Forty-three tumor cell lines were characterized for the frequency of micronuclei (black) and survival (gray) in the presence of AsiDNA (4.8 μmol/L) as described in Material and Methods. B, Statistical analysis of the correlation between survival to AsiDNA and the percentage of cells with micronuclei. Cells were grouped in two groups: MNLow (% cells with MN < 3.6%) and MNHigh (% cells with MN >3.6%). t test significance, ****, P < 0.0001.
Micronuclei (MN) are predictive biomarkers of AsiDNA treatment. A, Forty-three tumor cell lines were characterized for the frequency of micronuclei (black) and survival (gray) in the presence of AsiDNA (4.8 μmol/L) as described in Material and Methods. B, Statistical analysis of the correlation between survival to AsiDNA and the percentage of cells with micronuclei. Cells were grouped in two groups: MNLow (% cells with MN < 3.6%) and MNHigh (% cells with MN >3.6%). t test significance, ****, P < 0.0001.
Micronuclei are detected in xenografted tumors and predict response to AsiDNA
We validated the predictive value of micronuclei levels in tumors using four breast cancer xenograft models derived from breast cancer cell lines already studied in vitro. Two models were patient-derived xenografts BC227 and BC173, which were used to generate the BC227 and BC173 cell lines, studied in vitro. The two other models were obtained by grafting the MDA-MB-231 and the MDA-MB-468 cells. The cell-line analysis classified MDA-MB-231 as MNLow and the three others as MNHigh (Fig. 2A). Histologic samples of the tumors were stained with hematoxylin/eosin and analyzed by microscopy for the frequency of tumor cells with micronuclei. The micronuclei levels were generally lower in the xenografts than in the corresponding cell cultures (Table 1). The three MNHigh cell lines formed tumors with detectable micronuclei, whereas the MNLow MDA-MB-231 cell line gave rise to tumors with no detectable micronuclei. Tumor sensitivity to AsiDNA correlated with the micronuclei frequency in the xenografts. We observed complete tumor growth arrest for up to 60 days after treatment in the MNHigh xenografts, whereas the MNLow MDA-MB-231 tumor escaped treatment immediately after completion (3 weeks; Fig. 4A). We observed no partial (>20% decrease in the initial size of the tumor) or complete responses in the MNLow model. In contrast, all MNHigh tumors showed partial or complete responses (Table 2). We confirmed the general predictive value of micronuclei by evaluating the presence of micronuclei in histologic samples of all tumor models used in previous studies (19, 22–24) and comparing them to their response to AsiDNA monotherapy. They included models for breast cancer, colorectal cancer, skin melanoma, uveal melanoma, and glioblastoma. As observed for breast cancer, the frequency of micronuclei did not depend on the cancer type, but rather the tumor model studied. The xenografts were classified MNLow if they presented <2.8% cells with micronuclei and MNHigh if they showed >2.8% cells with micronuclei, similar to the threshold used for the breast cancer model. Xenografts with a MNHigh status showed a greater TGD induced by AsiDNA treatment than xenografts with MNLow status (P < 0.05; Fig. 4B and Table 2). All MNLow models were poor responders and did not show a significant difference in growth after AsiDNA treatment, with a maximal TGD of approximately 160%, meaning that it took only 60% more time for the tumors to achieve a four-fold increase in size relative to vehicle-treated tumors. In contrast, most tumor types (5/8) with a MNHigh status showed an at least three-fold increase (300%) in TGD after AsiDNA treatment. The three remaining MNHigh tumors behaved like the MNLow models (Fig. 4B). Taken together with the in vitro analysis, our results suggest that the absence of micronuclei could be a predictive biomarker for resistance to AsiDNA monotherapy, whereas the presence of micronuclei is necessary, but not sufficient, to provide sensitivity.
Basal levels of micronuclei in xenograft models
Tumor type . | BRCA status . | Structure (HES) . | Micronuclei in cell culture (%) . | Micronuclei in tumors (%) . |
---|---|---|---|---|
BC227 | BRCA2−/− | / | 9 | 6.7 |
BC173 | wt | Lobular structure | 13 | 5.6 |
MDA-MB-468 | wt | Necrotic | 5 | 2.8 |
MDA-MB-231 | wt | Necrotic | <3 | 0 |
Tumor type . | BRCA status . | Structure (HES) . | Micronuclei in cell culture (%) . | Micronuclei in tumors (%) . |
---|---|---|---|---|
BC227 | BRCA2−/− | / | 9 | 6.7 |
BC173 | wt | Lobular structure | 13 | 5.6 |
MDA-MB-468 | wt | Necrotic | 5 | 2.8 |
MDA-MB-231 | wt | Necrotic | <3 | 0 |
NOTE: BC227 and BC173 are PDXs; MDA-MB-468 and MDA-MB-231 are CDXs. At least 1,000 cells were analyzed per sample. The frequency of micronuclei is the percentage of cells with micronuclei in tumor samples.
Effect of AsiDNA monotherapy on xenografts. A, Relative tumor growth compared to the initial tumor volume. Mice bearing PDXs or CDXs were treated (black) or not (gray) with AsiDNA for 3 weeks. Vt, tumor volume at a given time; Vi, initial tumor volume before treatment. B, Effect of AsiDNA in xenografts showing a low (MNLow) or high (MNHigh) basal level of micronuclei, as measured by TGD. TGD values are the mean of at least six xenografted mice per cancer type. Black circles, glioblastoma; squares, uveal melanoma; diamonds, breast cancer; gray circle, colon cancer; black triangles, skin melanoma. *, P < 0.05.
Effect of AsiDNA monotherapy on xenografts. A, Relative tumor growth compared to the initial tumor volume. Mice bearing PDXs or CDXs were treated (black) or not (gray) with AsiDNA for 3 weeks. Vt, tumor volume at a given time; Vi, initial tumor volume before treatment. B, Effect of AsiDNA in xenografts showing a low (MNLow) or high (MNHigh) basal level of micronuclei, as measured by TGD. TGD values are the mean of at least six xenografted mice per cancer type. Black circles, glioblastoma; squares, uveal melanoma; diamonds, breast cancer; gray circle, colon cancer; black triangles, skin melanoma. *, P < 0.05.
Effect of AsiDNA monotherapy on xenografted tumors
Xenograft . | Tumor type . | Micronuclei (%) . | Treatment . | Number of mice . | Median survival (days) . | Relative risk (P value) . | TGD (mean %) . | PR (%) . | CR (%) . |
---|---|---|---|---|---|---|---|---|---|
BC227 | BC | 6.7 | NT | 15 | 67 | 100 | 0 | 0 | |
PDX | AsiDNA | 9 | >150 | 0.04 (1 × 10−5) | 387 | 11 | 56 | ||
BC173 | BC | 5.6 | NT | 10 | 87 | 100 | 0 | 0 | |
PDX | AsiDNA | 10 | >101 | NAa (1, 7 × 10−5) | 318 | 20 | 0 | ||
MDA-MB-468 | BC | 2.8 | NT | 11 | 106 | 100 | 0 | 0 | |
CDX | AsiDNA | 11 | 235 | 0.14 (1 × 10−5) | 300 | 45 | 0 | ||
MDA-MB-231 | BC | 0 | NT | 9 | 35 | 100 | 0 | 0 | |
CDX | AsiDNA | 9 | 66 | 0.24 (4 × 10−4) | 223 | 0 | 0 | ||
CB193 | GB | 0 | NT | 9 | 77 | 100 | 0 | 0 | |
CDX | AsiDNA | 12 | 117 | 0.24 (8 × 10−4) | 209 | 0 | 0 | ||
GBM14-RAV | GB | 2.7 | NT | 14 | 21 | 100 | 0 | 0 | |
PDX | AsiDNA | 10 | 30 | 0.24 (8 × 10−4) | 194 | 0 | 0 | ||
T98G | GB | 1.7 | NT | 17 | 19 | 100 | 0 | 0 | |
CDX | AsiDNA | 10 | 30 | 0.69 (0.32) | 186 | 0 | 0 | ||
TG1_HAM | GB | 0 | NT | 16 | 14 | 100 | 0 | 0 | |
PDX | AsiDNA | 10 | 21 | 0.55 (P < 0.088) | 122 | 0 | 0 | ||
SF763 | GB | 5.4 | NT | 18 | 39 | 100 | 0 | 0 | |
CDX | AsiDNA | 13 | 80 | 0.22 (6.9 × 10−5) | 288 | 0 | 0 | ||
ODA4_GEN | GB | 3.7 | NT | 6 | 23 | 100 | 0 | 0 | |
PDX | AsiDNA | 7 | 42 | 0.22 (P < 2.1 × 10−4) | 352 | 0 | 0 | ||
MM26 | UM | 0 | NT | 10 | 49 | 100 | 0 | 0 | |
PDX | AsiDNA | 9 | 52 | 0.32 (0.68) | 121 | 0 | 0 | ||
MP55 | UM | 0 | NT | 10 | 113 | 100 | 0 | 0 | |
PDX | AsiDNA | 10 | 85 | 2.65 (6.8 × 10−3) | 116 | 0 | 0 | ||
MP34 | UM | 5.4 | NT | 10 | 30 | 100 | 0 | 0 | |
PDX | AsiDNA | 10 | 37 | 0.38 (5.3 × 10−3) | 147 | 0 | 0 | ||
HT29 | CC | 3.1 | NT | 11 | 28 | 100 | 0 | 0 | |
CDX | AsiDNA | 12 | 37 | 0.46 (0.059) | 132 | 0 | 0 | ||
SK28 | SM | 5.1 | NT | 21 | 57 | 100 | 0 | 0 | |
CDX | AsiDNA | 10 | 81 | 0.48 (3 × 10−2) | 149 | 0 | 0 |
Xenograft . | Tumor type . | Micronuclei (%) . | Treatment . | Number of mice . | Median survival (days) . | Relative risk (P value) . | TGD (mean %) . | PR (%) . | CR (%) . |
---|---|---|---|---|---|---|---|---|---|
BC227 | BC | 6.7 | NT | 15 | 67 | 100 | 0 | 0 | |
PDX | AsiDNA | 9 | >150 | 0.04 (1 × 10−5) | 387 | 11 | 56 | ||
BC173 | BC | 5.6 | NT | 10 | 87 | 100 | 0 | 0 | |
PDX | AsiDNA | 10 | >101 | NAa (1, 7 × 10−5) | 318 | 20 | 0 | ||
MDA-MB-468 | BC | 2.8 | NT | 11 | 106 | 100 | 0 | 0 | |
CDX | AsiDNA | 11 | 235 | 0.14 (1 × 10−5) | 300 | 45 | 0 | ||
MDA-MB-231 | BC | 0 | NT | 9 | 35 | 100 | 0 | 0 | |
CDX | AsiDNA | 9 | 66 | 0.24 (4 × 10−4) | 223 | 0 | 0 | ||
CB193 | GB | 0 | NT | 9 | 77 | 100 | 0 | 0 | |
CDX | AsiDNA | 12 | 117 | 0.24 (8 × 10−4) | 209 | 0 | 0 | ||
GBM14-RAV | GB | 2.7 | NT | 14 | 21 | 100 | 0 | 0 | |
PDX | AsiDNA | 10 | 30 | 0.24 (8 × 10−4) | 194 | 0 | 0 | ||
T98G | GB | 1.7 | NT | 17 | 19 | 100 | 0 | 0 | |
CDX | AsiDNA | 10 | 30 | 0.69 (0.32) | 186 | 0 | 0 | ||
TG1_HAM | GB | 0 | NT | 16 | 14 | 100 | 0 | 0 | |
PDX | AsiDNA | 10 | 21 | 0.55 (P < 0.088) | 122 | 0 | 0 | ||
SF763 | GB | 5.4 | NT | 18 | 39 | 100 | 0 | 0 | |
CDX | AsiDNA | 13 | 80 | 0.22 (6.9 × 10−5) | 288 | 0 | 0 | ||
ODA4_GEN | GB | 3.7 | NT | 6 | 23 | 100 | 0 | 0 | |
PDX | AsiDNA | 7 | 42 | 0.22 (P < 2.1 × 10−4) | 352 | 0 | 0 | ||
MM26 | UM | 0 | NT | 10 | 49 | 100 | 0 | 0 | |
PDX | AsiDNA | 9 | 52 | 0.32 (0.68) | 121 | 0 | 0 | ||
MP55 | UM | 0 | NT | 10 | 113 | 100 | 0 | 0 | |
PDX | AsiDNA | 10 | 85 | 2.65 (6.8 × 10−3) | 116 | 0 | 0 | ||
MP34 | UM | 5.4 | NT | 10 | 30 | 100 | 0 | 0 | |
PDX | AsiDNA | 10 | 37 | 0.38 (5.3 × 10−3) | 147 | 0 | 0 | ||
HT29 | CC | 3.1 | NT | 11 | 28 | 100 | 0 | 0 | |
CDX | AsiDNA | 12 | 37 | 0.46 (0.059) | 132 | 0 | 0 | ||
SK28 | SM | 5.1 | NT | 21 | 57 | 100 | 0 | 0 | |
CDX | AsiDNA | 10 | 81 | 0.48 (3 × 10−2) | 149 | 0 | 0 |
NOTE: Micronuclei column indicates the percentage of cells with micronuclei.
Abbreviations: PR, partial response; CR, complete response; GB, glioblastoma; UM, uveal melanoma; CC, colorectal cancer; SM, skin melanoma.
aNA, not applicable because no mice died during the experiment.
Micronuclei frequency in patient biopsies
We analyzed whether micronuclei can be observed in patient biopsies. We quantified the micronuclei in hematoxylin/eosin stained tumor sections of biopsies from unselected breast cancer samples from the Curie Hospital Pathology Department (Fig. 5). Half of the 12 analyzed breast cancer biopsies displayed a MNHigh status (% cells with MN > 3%), whereas only few or no micronuclei were detectable in the others (MNLow; % cells with MN < 1%). There was no evident link between micronuclei status and the type or grade of the tumors (Supplementary Table S1). The lack of correlation between genetic instability and tumor grade or type has already been reported in a large study of 5,371 tumors using three different methods to estimate genetic instability (35).
Assessment of micronuclei (MN) in breast cancer patient biopsies. A, Micronuclei quantification in 12 breast cancer patient' biopsies. Biopsies showing a percentage of cells with micronuclei < 1% are considered MNLow and biopsies containing a percentage of cells with micronuclei > 1% are considered MNHigh. B, Representative images of tumor sections after hematoxylin & eosin staining and micronuclei (arrows) assessment in breast cancer patient biopsies. Scale bar, 100 μm.
Assessment of micronuclei (MN) in breast cancer patient biopsies. A, Micronuclei quantification in 12 breast cancer patient' biopsies. Biopsies showing a percentage of cells with micronuclei < 1% are considered MNLow and biopsies containing a percentage of cells with micronuclei > 1% are considered MNHigh. B, Representative images of tumor sections after hematoxylin & eosin staining and micronuclei (arrows) assessment in breast cancer patient biopsies. Scale bar, 100 μm.
Discussion
AsiDNA is a DNA repair inhibitor that prevents the recruitment of repair enzymes to damaged sites by jamming DNA damage signaling. It acts on enzymes involved in the HR, NHEJ, BER, and SSBR pathways (19, 21, 26) and thus broadly inhibits DNA repair. Its antitumor activity has been demonstrated in preclinical studies in association with DNA damaging treatments, such as radiotherapy and chemotherapy in many tumor models, and in the clinic in association with radiotherapy. However, its activity as a monotherapy is more selective as tumors must be dependent on DNA repair for their survival to be sensitive to its inhibition. In this study, we determined the sensitivity of a set of tumor cell lines to AsiDNA and estimated the presence of DNA damage by various direct and indirect methods. Sensitivity to AsiDNA was associated with the presence of unrepaired or misrepaired DSBs, revealed by micronuclei formation, and large genome rearrangements (measured by LST analysis). SSBs and spontaneous PARP activation were not essential, confirming that DSBs are probably the main factor for cell survival. Transcriptomic and genomic analysis revealed downregulation of DNA repair pathways and several alterations, especially in DSB repair pathways of sensitive breast cancer cells. These DSB repair defects may explain the high frequency of micronuclei and LST in these cells, and therefore their sensitivity to AsiDNA treatment. We analyzed the sensitivity of breast cancer cell lines to Olaparib (a PARP inhibitor) in a previous study (26). We did not observe a correlation between the frequency of micronuclei and LST and sensitivity to Olaparib in these cells (data not shown). Thus, these predictive biomarkers appear to be specific to AsiDNA and are not applicable to other DNA repair inhibitors.
The occurrence of spontaneous DSBs is difficult to quantify as cells do not survive, except if they are repaired in a faithful (HR) or unfaithful (NHEJ) manner. Unfaithful repair leads to an accumulation of breaks and ligations that can be detected in the tumor cell genomes as discontinuous or amplified/deleted large sequences (LSTs). However, these scars accumulate during tumor development and do not represent the temporal status of the genetic events at the time of treatment. Therefore, the detection of micronuclei is an additional piece of information that reflects the real-time occurrence of spontaneous damage as they do persist for more than one or two divisions (36). Moreover, they are specific to the proliferating population, as they are produced at mitosis. Here, micronuclei and LST frequencies correlated in cell line cultures. Thus, it may be informative to validate this correlation in the next clinical trial of AsiDNA to determine how to use this dual information on instability “scars” (LST) and proliferation-linked instability (micronuclei) for predicting sensitivity to AsiDNA.
Micronuclei detection has been used for decades to assess the genotoxicity of chemical agents and irradiation exposure for the purpose of public health and safety, especially in blood and epithelial tissues as they are in immediate contact with these agents (37). More recently, several studies have linked the tumor biology and micronuclei biogenesis (34), and proposed using micronuclei analysis as a biomarker for screening and risk prediction. Indeed, prospective studies have shown that individuals with a high frequency of micronucleated blood cells are more prone to develop cancer (38). Moreover, several studies have suggested that irradiation-induced micronuclei in blood cells could be used to screen for cancer predisposing genetic defects, such as BRCA mutations in hereditary breast cancer, as this protein is implicated in radiation-induced DNA damage repair (39). Micronuclei frequency has also been implicated in the monitoring of cancer treatments. Many studies have shown that an increase in micronuclei frequency during radiotherapy correlated with a better response (40, 41). Responses to many DNA damaging chemotherapies have also been evaluated by micronuclei analysis (42). Reports have shown that a high frequency of micronuclei is associated with more aggressive cancer and a poor clinical outcome (43). A striking difference in median survival between patients harboring cancer bone marrow cells with a high (> 0.5%, 34 days) or low (<0.5%, 148 days) frequency of micronuclei has also been reported (44). Although other studies have correlated the presence of micronuclei with the pathogenesis of many malignancies, suggesting its use for cancer grade assessment, the work described here suggests, for the first time, that micronuclei could be used as a predictive biomarker for response to AsiDNA. Detection of basal micronuclei frequencies in biopsies of different tumor types, before AsiDNA treatment, could help to predict good and poor responders. This tool would allow us to stratify patients in the next clinical trials for tailored treatment and appropriate dosing. In general, our finding that low basal levels of LSTs and micronuclei could be biomarkers of resistance to AsiDNA suggests that aggressive tumors with high genetic instability (frequently with a poor prognosis) may be the preferential indication for AsiDNA treatment.
Disclosure of Potential Conflicts of Interest
T. Popova and M.-H. Stern have ownership interest (including patents) in Myriad. M. Dutreix is a consultant/advisory board member of Onxeo. No potential conflicts of interest were disclosed by the other authors.
Authors' Contributions
Conception and design: M. Dutreix
Development of methodology: W. Jdey, S. Thierry
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): W. Jdey, S. Thierry, M.-H. Stern
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T. Popova, M.-H. Stern, M. Dutreix
Writing, review, and/or revision of the manuscript: M. Dutreix
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T. Popova
Study supervision: W. Jdey, S. Thierry, M. Dutreix
Acknowledgments
We thank Amélie Croset, Nathalie Berthault, and Marie-Christine Lienafa for their participation in this project. We thank the RadExp platform for performing the comet assays and the LIP for providing the BC227 and BC173 cell lines. We thank Pierre De Lagrange and Ariane Jolly from Genosplice for their help on transcritptomic analysis. We also acknowledge the local CICS platform facilities.
Grant Support
This work was supported by the SIRIC-Curie, the program PIC SysBio of the Institut Curie, the Centre National de la Recherche Scientifique, the Institut National de la Santé Et de la Recherche Médicale, the Agence Nationale de la Recherche, and the Conseil général de l'Essonne. W. Jdey was supported by fellowship CIFFRE-ANRT (2013/0907). S. Thierry was supported by the Institut National du Cancer (TRANSLA13-081).
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.