Pediatric sarcomas represent a heterogeneous group of malignancies that exhibit variable response to DNA-damaging chemotherapy. Schlafen family member 11 protein (SLFN11) increases sensitivity to replicative stress and has been implicated as a potential biomarker to predict sensitivity to DNA-damaging agents (DDA). SLFN11 expression was quantified in 220 children with solid tumors using IHC. Sensitivity to the PARP inhibitor talazoparib (TAL) and the topoisomerase I inhibitor irinotecan (IRN) was assessed in sarcoma cell lines, including SLFN11 knock-out (KO) and overexpression models, and a patient-derived orthotopic xenograft model (PDOX). SLFN11 was expressed in 69% of pediatric sarcoma sampled, including 90% and 100% of Ewing sarcoma and desmoplastic small round-cell tumors, respectively, although the magnitude of expression varied widely. In sarcoma cell lines, protein expression strongly correlated with response to TAL and IRN, with SLFN11 KO resulting in significant loss of sensitivity in vitro and in vivo. Surprisingly, retrospective analysis of children with sarcoma found no association between SLFN11 levels and favorable outcome. Subsequently, high SLFN11 expression was confirmed in a PDOX model derived from a patient with recurrent Ewing sarcoma who failed to respond to treatment with TAL + IRN. Selective inhibition of BCL-xL increased sensitivity to TAL + IRN in SLFN11-positive resistant tumor cells. Although SLFN11 appears to drive sensitivity to replicative stress in pediatric sarcomas, its potential to act as a biomarker may be limited to certain tumor backgrounds or contexts. Impaired apoptotic response may be one mechanism of resistance to DDA-induced replicative stress.

Pediatric sarcomas are a heterogeneous group of malignancies disproportionately affecting adolescents and young adults. Multimodal therapy with chemotherapy, surgery, and radiatiotherapy has improved outcomes for patients with localized disease. However, progress has stalled for patients with metastatic and recurrent disease, who continue to have survival rates of less than 30% for the most common subtypes (1–5). Therefore, novel therapeutic strategies and biomarkers that predict sensitivity to therapy are needed.

Previously, we reported that combining the PARP inhibitor (PARPi) talazoparib (TAL) with the topoisomerase I inhibitor (Topo1i) irinotecan (IRN) and temozolomide (TMZ) resulted in high rates of complete response (CR) in a murine model of Ewing sarcoma (6). This work motivated a clinical trial testing of TAL plus IRN with and without TMZ in children with refractory/recurrent solid tumors (NCT02392793). Results were encouraging: of 41 evaluable patients, 1 with Ewing sarcoma had a CR, 5 others had a partial response (PR), and 18 had disease stabilization (7). However, the major driver of sensitivity to this combination is still unknown and we currently lack the ability to predict which tumors may respond to this type of therapy.

Although BRCA mutations are rare in Ewing sarcoma, several mechanisms have been proposed to explain the PARPi sensitivity in Ewing sarcoma, most notably functional BRCA deficiency and Schlafen family member 11 (SLFN11; refs. 8–11). Fusion proteins involving the peptide encoded by EWSR1 at the N-terminus are the oncogenic drivers in Ewing sarcoma (which has EWSR1-FLI1/ERG fusion) and in a subset of aggressive sarcomas, such as desmoplastic small round-cell tumors (DSRCT; which have EWSR1-WT1 fusion) and clear-cell sarcomas (which have EWSR1-ATF1 fusion). Gorthi and colleagues (11) reported that the EWSR1 fusion protein increases R-Loop formation, which sequesters BRCA1, rendering the tumor cell BRCA-deficient and susceptible to replicative stress. They speculated that BRCA deficiency might create a liability in all tumors possessing an EWSR1-translocation. Tumors deficient in mediators of homologous recombination such as BRCA1 and BRCA2 are more susceptible to DNA single-strand breaks: consequently, PARPis are selectively lethal in these cells. However, in contrast to PARPi treatment of adult tumors with BRCA1 or BRCA2 mutations, single-agent PARPi treatment in patients with relapsed/refractory ES elicited no significant responses or durable disease control (12).

Ewing sarcomas express high levels of SLFN11, a putative DNA/RNA helicase whose expression has been associated with the response to DNA-damaging agents (DDA), thymocyte maturation, viral immunity, and IFN production (13). SLFN11 augments sensitivity to replicative stress by stalling replication forks and impairing the DNA-repair checkpoint response (14, 15). PARPis cause replicative stress through PARP trapping (9), whereby the PARP protein becomes physically associated with DNA. This is similar to the mechanism of action of Topo1is, which are well-known inducers of replicative stress (16). Additionally, PARP inhibition augments Topo1i toxicity by preventing the recruitment of repair enzymes to the site of damage, and TMZ enhances PARPi-mediated replicative stress by augmenting PARP trapping (17). Therefore, the strategy of combining a PARPi, a Topo1i, and TMZ is a rational means of exploiting replicative stress in cancer cells. In retrospective studies, patients with ovarian cancer (18), breast cancer (19), prostate cancer (20), and Ewing sarcoma-family of tumors (10) who were treated with DDAs had a better prognosis if their tumors had high SLFN11 expression. Patients with small-cell lung cancer expressing SLFN11 showed improved progression-free survival (PFS) and overall survival (OS) when treated with the PARPi veliparib and TMZ (21). SLFN11 mutations are rare, and epigenetic regulation has been suggested as a mediator of resistance (16, 22).

Determining whether SLFN11 or the EWSR1 fusion drives sensitivity to PARPi combinations in pediatric sarcomas is crucial for identifying the patients who might benefit the most from such treatment. Outside of Ewing sarcoma, the expression pattern of SLFN11 in pediatric sarcomas is largely unknown. Furthermore, in-depth exploration of Ewing sarcoma tumors, which possess an EWSR1 fusion combined with high expression of SLFN11, might enable the elucidation of mechanisms of resistance to TAL + IRN, as response to this drug combination in this tumor type is not universal (7). In this work, we show that SLFN11 is widely expressed in common pediatric sarcoma subtypes and that the SLFN11 protein drives sensitivity to TAL and IRN both in vitro and in vivo, expanding the relevance of this combination beyond EWSR1-translocated sarcoma. Despite this finding, we found that SLFN11 expression does not portend a better prognosis in these patients. Importantly, we also show that impairment of intrinsic apoptosis, not loss of SLFN11 expression, is one means of resistance to PARPi combination therapy in pediatric sarcoma, and that sensitivity to TAL + IRN can be increased by selective inhibition of BCL-xL. Our work supports the use of combinations involving strong-trapping PARPis and Topoi1s as targeted therapy for SLFN11-positive pediatric sarcomas, and it offers novel strategies to combat tumors resistant to replicative stress.

Genomics of drug sensitivity in cancer correlations

Drug sensitivity (v17.3) and expression data were downloaded from the Genomics of drug sensitivity in cancer (GDSC) website (https://www.cancerrxgene.org/gdsc1000/GDSC1000_WebResources/Home.html) in May 2018 and June 2018, respectively. Catalogue of Somatic Mutations in Cancer (COSMIC) mutation data (CosmicMutantExport.tsv.gz) was downloaded from https://cancer.sanger.ac.uk/cosmic/download in March 2020. We used the area-over-the-curve (AOC) to quantify drug activity, wherein the curve in question is the fitted proportional survival relative to negative controls. This metric gives intuitive results across a wide range of dose-response behaviors, yielding more robust results than those obtained with metrics such as EC50 (which can be poorly defined in low-activity systems) or IC50 (which is fully undefined for systems that do not reach 50% killing). Fitting proportional survival is done using standard nonlinear least-squares regression, fitting a four-parameter log-logistic equation to logarithmically transformed proportional cell survival. Fitting to the logarithm of cell survival allows for more reliable fitting of drugs that achieve high levels of killing (e.g., 1,000-fold or 10,000-fold killing relative to vehicle). All fits were performed in the R Statistical Computing Environment (23). However, because the AOC in logarithmic survival space is theoretically unbounded, we transformed the fitted survival curve into linear survival space to estimate the AOC to maintain a bounded range of AOC values.

Mutational signatures

The set of 30 mutational signatures (MS), a 96 × 30 matrix Z, were obtained from COSMIC (https://cancer.sanger.ac.uk/cosmic/signatures). To obtain the weight of each signature in a sample, we started with the list of all simple somatic mutations. For a given sample, each single-nucleotide mutation was classified as one of the 96 possible mutation-types, resulting in a 96 × 1 probability vector f. The vector p representing the weights of the 30 COSMIC signatures, was chosen such that the sum |\sum\nolimits_{i\ = \ 1}^{96} {{{( {Zp - f} )}^2}} $| was minimized. The optimization procedure was implemented in MATLAB using the built-in quadratic programming function quadprog.

IHC staining of pediatric tumor samples

IHC was performed with the Dako Omnis instrument (Agilent) on 4‐μmol/L‐thick formalin‐fixed paraffin‐embedded whole‐tissue sections, using a rabbit anti-SLFN11 (anti-SLFN11) polyclonal antibody (Sigma-Aldrich catalog no. # HPA023030, RRID:AB_1856613; 1:25 dilution, 60-minute incubation), Dako Low pH Target Retrieval Solution, and the Dako EnVision Flex Detection Kit. Immunoreactivity was scored using H scores with the percentage of cells with positive staining being estimated at one of 3 levels of intensity (weak, moderate, or strong). Cells with no staining were given a score of 0+. The resultant H score was calculated with the following formula [H score = (1 ×%weak)+(2 ×%moderate)+(3 ×%strong)], with the overall score ranging from 0 (negative) to 300 (100% strong staining).

Cell lines

The name and source of the cell lines used in this study are: ES8 [St. Jude Children's Research Hospital (SJCRH), RRID:CVCL_1204], A673-shEF (Francisco J Alonso (24), RRID:CVCL_JM58), JN-DSCRT (SJCRH, RRID:CVCL_9W68), EW8 (SJCRH, RRID:CVCL_V618), CADO-ES1 [Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures (DSMZ), RRID:CVCL_1103], RH30 (SJCRH, RRID:CVCL_0041), ES1 (SJCRH, RRID:CVCL_1198), H-EMC-SS (European Collection of Authenticated Cell Cultures, RRID:CVCL_1238), SaOS2 (SJCRH, RRID: CVCL_0548), CY143B (SJCRH, RRID:CVCL_2270), SJSA1 (SJCRH, RRID:CVCL_1697), RH3 (SJCRH, RRID:CVCL_L415), U2OS (SJCRH, RRID:CVCL_0042), SU-CCS-1 (ATCC, RRID:CVCL_B470), CHLA-258 (Children's Oncology Group Cell Culture/Xenograft Repository, RRID:CVCL_A058), EW18 [Dr. Zisis Kozlakidis and International Agency for Research on Cancer/World Health Organization (IARC), RRID:CVCL_1213], EW13 (IARC, RRID:CVCL_1211), EW11 (IARC, RRID:CVCL_1209). Cell lines were authenticated using short tandem repeat analysis via PowerPlex (Promega) and tested for mycoplasma using MycoAlert (Lonza). Translocation status was confirmed using PCR and FISH.

Cell viability assay

Cell viability was measured using CellTiter-Glo (Promega). The luminescent signal was read with an EnVision Multimode Plate Reader (PerkinElmer). The results of screening experiments were processed and visualized using 2 programs developed in-house: Robust Investigation of Screening Experiments (RISE) and AssayExplorer. Cells were plated in 96-well plates at variable densities 12 to 24 hours in advance of drug to ensure that they continued in the log-growth phase for the duration of the experiment. The drug plates containing stocks of compounds and the control plate containing the negative control (DMSO) and the positive control (staurosporine) were generated by the Compound Management Center in the Department of Chemical Biology and Therapeutics at our institution. Dose-response experiments were performed with 10-point, 3-fold dilution (19,683-fold concentration range). Before compound transfer, the assay and compound plates were centrifuged at 201 × g (1,000 rpm) for 1 minute in an Eppendorf 5810 centrifuge equipped with an A-4–62 swing-bucket rotor (Eppendorf AG). Approximately 102 nL/well of compound was transferred to a 100 μL volume in the corresponding wells of the assay plates with a 100SS 96-well pin-tool (V&P Scientific) using a Biomek FX Liquid Handler (Beckman Coulter). The final DMSO concentration in each well was ≤0.2%.

Raw luminescence relative light unit (RLU) values for each compound at each concentration were log2 transformed; normalized to obtain the percentage of activity by using the following equation: percentage of activity = 100 × [{mean(negctr) – compound} / {mean(negctr) – mean(posctrl)}]; then pooled from replicate experiments before fitting. Here, negctrl and posctrl refer to the negative (DMSO) and positive (20–30 μmol/L staurosporine) controls on each plate. Dose-response curves were fit using the drc (25) package in R (23). Both a 3-parameter model (with y0, the response without drug, set to zero) and a 4-parameter model (with y0 allowed to vary) were fit using the sigmoidal function LL2.4. The hill slope was constrained to be between −10 and 0, and the EC50 was constrained to be between 10−11 and 10−4 (which roughly equated to the drug-concentration range tested in these experiments). For the 3-parameter model, yFin, the maximum response of the dose-response curve, was constrained to be between 0 and the maximum of the median activities calculated at each concentration over all pooled measurements. For the 4-parameter model, y0 and yFin were both constrained to be between the minimum and the maximum of the median activities calculated at each concentration over all pooled measurements. The model with the lowest corrected Akaike information criterion (AICc) was selected as the best fit model.

The AUC was calculated from the fitted curve by using the trapezoid rule in the concentration range 10−11 to 10−4 molar. In the event of a failure to fit a sigmoidal dose response curve, the smooth.spline option in R was used to fit a curve that could be used to determine the AUC.

Western blot analysis

The following antibodies were used: SLFN11 (Sigma-Aldrich catalog no. # HPA023030, RRID:AB_1856613), BAK (clone D4E4, Cell Signaling Technology catalog no. # 12105, RRID:AB_2716685), BAX (Cell Signaling Technology catalog no. # 2772, RRID:AB_10695870), β-Actin (clone 13E5, Cell Signaling Technology catalog no. # 4970, RRID:AB_2223172), α-Tubulin (clone DM1A, Novus catalog no. # NB100–690, RRID:AB_521686), IRDye 680LT Goat anti-Rabbit IgG antibody (LI-COR Biosciences catalog no. # 926–68021, RRID:AB_10706309), and IRDye 800CW Donkey antimouse IgG antibody (LI-COR Biosciences catalog no. # 926–32212, RRID:AB_621847). Cell pellets were flash frozen and stored at −80° C. Cells were lysed using radio-immunoprecipitation assay lysis buffer containing cOmplete Mini Protease Inhibitor (Roche) and 1× phosphatase inhibitor, then sonicated at 50% for 20 seconds and centrifuged at 4°C for 20 minutes at 20,000 × g. The supernatant was collected, and the protein concentration was measured using a bicinchoninic acid assay kit (Life Technologies). Proteins were separated by electrophoresis on 10% NuPage Bis-Tris gels and transferred to PVDF membranes (ThermoFisher Scientific), which was then blocked with Odyssey blocking buffer for 1 hour. The membranes were incubated with primary antibody and 0.1% Tween overnight at 4°C. Membranes were then washed with PBS and 0.1% Tween (PBST) and secondary antibodies were added with blocking solution and Tween for 40 minutes. Membranes were then washed with PBST + 0.02% SDS and imaged using an Odyssey CLx Infrared Imaging System (LI-COR). Protein was quantified with LI-COR Biosciences software.

qRT-PCR analysis

Total mRNA was extracted using the Maxwell RSC simplyRNA Tissue Kit and reverse transcribed using iScript Reverse Transcription Supermix (Bio-Rad). TaqMan primers were used for real-time PCR. β-Actin was used as the internal control.

Cell engineering

Generation of over-expression models

SLFN11 cDNA (OriGene Technologies catalog no. # RC226247L4) and pVector control vector (OriGene Technologies catalog no. # PS100093) were transiently transfected into U2OS cells at 50% to 70% confluence by using TurboFectin 8.0 Transfection Reagent (OriGene Technologies) according to the manufacturer's protocol. Twenty-four hours after transfection, cells were passaged into fresh growth medium containing puromycin (selective medium). A mock transfection well in a 6-well plate was used in parallel as a control. The selective medium was changed every 2 to 3 days until all cells in the control well were dead. The selective medium was then changed into fresh growth medium. The surviving cells were maintained in culture and collected for further validation. SLFN11 cDNA (OriGene Technologies catalog no. # RC226247L4) and pVector control vector (OriGene Technologies catalog no. # PS100093) were nucleofected into ES8-SLFN11-KO cells using a 4D-Nucleofector (Lonza) according to the manufacturer's protocol. Cells were passaged until they reached 75% confluency in a 25 cm2 flask then the GFP positive cells were selected for using flow cytometry. The resulting GFP+ cells were maintained in culture and collected for further validation.

Generation of CRISPR knock-out cell lines

ES8-SLFN11-KO, A673-SLFN11-KO, and JN-DSRCT-SLFN11-KO cells were generated using CRISPR-Cas9 technology. Briefly, 400,000 cells were transiently cotransfected with 500 ng of guide RNA (gRNA) expression plasmid (cloned into Addgene plasmid #43860), 1 μg of Cas9 expression plasmid (Addgene plasmid #43945), and 200 ng of pMaxGFP via nucleofection (Lonza, 4D-Nucleofector X Unit), using solution P3 and program DS150 for ES8 cells, EO-100 for A673-shEF cells, and EH-100 for JN-DSRCT cells in small (20 μL) cuvettes in accordance with the manufacturer's recommended protocol. Cells were single-cell sorted by FACS to enrich for GFP+ (transfected) cells, clonally selected, and verified for the desired targeted modification via targeted deep sequencing. Two clones were identified for each modification and assessed in relevant assays. The sequences for the single-guide RNA (sgRNA) and the relevant primers are listed in the Supplemental Data.

Annexin-V and cell-cycle analysis

Cells were plated at appropriate densities to obtain 5 × 105–1 × 106 cells at the time of collection and allowed to adhere 24 hours at 37°C prior to addition of drug. After drug treatment, the supernatant was collected and combined with gently trypsinized cells that were then resuspended in fresh medium. Cells were then divided equally and placed on ice for cell-cycle analysis and quantification of apoptosis. Half of the cells were stained with annexin-V and DAPI to determine their apoptosis levels. The other half were stained with propidium iodide to determine their cell-cycle status. Cells were interrogated using an LSR Fortessa Flow Cytometer (BD Bioscience). Apoptosis statistics were obtained with DiVa Software (BD Bioscience), and cell-cycle statistics were obtained with ModFit software (Verity Software House).

Alkaline comet assay

Alkaline single-cell electrophoresis was performed using the CometAssay Reagent Kit (Trevigen catalog no. # 4250–050) in accordance with the manufacturer's instructions. Briefly, cells were plated at a density of 50,000 cells per well in a 96-well plate, incubated for 24 hours, then treated with 0.2% DMSO, 1 μmol/L TAL, 1 μmol/L SN-38, or a combination of 1 μmol/L TAL and 1 μmol/L SN-38. After being treated for 2.5 hours, the cells were trypsinized then washed with ice-cold 1× PBS. They were then combined with low-melting-point agarose (LMAgarose, Trevigen catalog no. # 4250–500–02) at a ratio of 1:10 [volume for volume (v/v)] and immediately plated onto CometSlides (Trevigen catalog no. # 4253–096–03). The slides were placed at 4°C in the dark for 30 minutes to improve cell adherence then immersed in lysis solution for 60 minutes. The slides were then drained and incubated with Alkaline Unwinding Solution for 20 minutes at room temperature. Electrophoresis was then performed using the CometAssay Electrophoresis System II unit (Trevigen catalog no. # 4250–050-ES) with 21 volts being applied for 30 minutes for ES8 and ES8-SLFN11-KO cells and for 45 minutes for CHLA-258. Samples were dried for 30 minutes at 37°C and washed with dH2O × 2 and 70% ethanol. Samples were then dried for an additional 30 minutes at 37°C then stained with SYBR Gold solution for 30 minutes. Comets were imaged by using the LionHeart FX automated microscope (Biotek) and the Gen5 Image Prime software to construct image montages that were analyzed using TriTek CometScore 2.0.0.38.

R-loop immunofluorescence

Cells were seeded on Ibidi slides (uClear) and allowed to attach overnight. Next day, the cells were fixed using cold, dry methanol for 30 minutes at 4°C followed by a wash step with cold PBS. The cells were then permeabilized with 10% normal goat serum (NGS) and 0.1% Triton-X in PBS for 20 minutes at room temperature (RT). They were then washed with PBS and incubated with S9.6 primary antibody (Kerafast catalog no. # ENH001, RRID:AB_2687463) in 1% NGS in PBS overnight. After washing with PBS, the cells were incubated with Alexa Fluor, Goat anti-Mouse, 647 antibody (Molecular Probes) in 1% NGS in PBS at radiotherapy for 2 hours then stained with Hoechst stain (Invitrogen). Cells were imaged with a Leica microscope using 40× and 63× objectives. Images were collected using the Photon Counting 3D Nyquist technique. To quantify R-loop levels, nuclear outlines were traced by hand, using the freeform selection tool in ImageJ, on images of dimensions 1084 pixels × 1084 pixels, representing regions of 150 μmol/L × 150 μmol/L. To calculate the background-subtracted R9.6 signal, the R9.6 channel of each image was blurred with a Gaussian filter with a standard deviation of 2 pixels (to smooth out dead or outlier pixels) then passed through a local minimum filter with a radius of 20 pixels. This gave the minimum nearby value for each pixel and was subtracted from the original R9.6 channel. The mean image intensity and area were calculated on the background-subtracted R9.6 channel by using the Measurement tool in ImageJ. The total nucleolar luminous intensity was estimated by multiplying the average background-subtracted R9.6 intensity (ranging from 0 to 1) by the area of the nucleus (scaled to square micrometers).

Expression analysis

Expression profiles were generated from biological triplicates of ES8 and ES8-SLFN11-KO cells. Cells were treated with 0 or 2 Gy of gamma irradiation then harvested at 4 hours or 24 hours postirradiation. Total RNA (100 ng) was purified from treated cells with a RNeasy Mini Kit (Qiagen catalog no. # 74104) and analyzed using the Affymetrix Clariom S Human assay (ThermoFisher Scientific catalog no. # 902927). Probe signals were normalized and transformed into log2 transcript expression values by the robust multi-array average (RMA) algorithm, using the Affymetrix Expression Console software v1.1. Differentially expressed transcripts were identified by an ANOVA model in which parental lineage, genotype, radiation dose, and irradiation timepoint were used as the factors (Partek Genomics Suite software v6.6; Partek, Inc.). The FDR was estimated by the method of Benjamini and Hochberg, with an FDR threshold of less than 0.05 being applied to identify differentially expressed transcripts. The expression data is available from the Gene Expression Omnibus (Accession ID GSE181094).

Analysis of SLFN11 in SJEWS049193

Consensus RNA sequencing (RNA-seq) SLFN11 sequences (chr17:33675329–33702720) from SJEWS049193_D1 (primary site tumor, site #1), SJEWS049193_D2 (primary site tumor, site #2), SJEWS049193_X1 (metastatic tumor, site #1), and SJEWS049193_X2 (metastatic tumor, site #2) were aligned to the coding region (exons 4, 5, 6, and 7) of hSLFN11 (NM_001104587). Sequence alignments are reported in the Supplemental Data.

Patient samples

The St. Jude electronic database was surveyed for patients with solid tumors enrolled on St. Jude trials from 2000 to 2018. Samples were assessed for viability and availability. Once staining was performed, a retrospective review of the electronic medical record was conducted. Patient data were matched with the IHC samples, and the results were analyzed for correlation by an independent statistician. Data collection resulted in a dataset of 353 pathology samples from 220 patients. Samples obtained at the same time from the same specimen were summarized by their median value, which reduced the number of sample measurements to 335. We defined a patient as “SLFN11 positive” if any sample from that patient had an H score greater than 0 at any time, and “SLFN11 negative” if every sample from that patient had an H score equal to 0. We restricted the survival analysis to patients who had a sample at diagnosis or treatment. The observation selected for inclusion in this data subset was the first of the following: (i) human sample at diagnosis, (ii) xenograft sample at diagnosis, (iii) human sample at treatment, and (iv) xenograft sample at treatment. If a patient did not have any observation from the above list, the patient was excluded from the data subset. This data subset contained 143 patients (one measurement per patient).

In vivo experiments

Athymic nude immunodeficient mice were purchased from Charles River (strain code 553). This study was carried out in strict accordance with the recommendations in the Guide to Care and Use of Laboratory Animals of the NIH. Drug dosing and efficacy studies were performed as described previously (6).

Briefly, orthotopic xenografts were created by injecting luciferase-labeled cells into athymic nude mice, using the following techniques: bone-marrow injection for ES8 and ES8-SLFN11-KO cells; intraperitoneal injection for JN-DSRCT and JN-DSRCT-SLFN11-KO cells; and subcutaneous injection for SU-CCS-1 cells. SJEWS049193_X1 was derived from an 11-year-old male patient with a history of recurrent metastatic Ewing sarcoma under the MAST protocol (NCT01050296). Patient-derived xenografts were created using SJEWS049193_X1 tumor cells obtained from the Childhood Solid Tumor Network (http://www.stjude.org/CSTN/). These cells were dissociated and passaged as described previously (26).

Mice were screened weekly by Xenogen and the bioluminescence was measured (see below). Mice were enrolled in the study after a target bioluminescence signal of 107 photons/s/cm2 or a palpable tumor was obtained, and chemotherapy was started on the following Monday. Mice received a maximum of 4 courses of chemotherapy (3 weeks per course) and bioluminescence was monitored weekly and at the end of therapy. Mice were monitored daily while receiving chemotherapy.

Xenogen imaging and quantification

Mice were given intraperitoneal injections of Firefly D-Luciferin (Caliper Life Sciences; 3 mg/mouse). Bioluminescent images were acquired 5 minutes later with the IVIS 200 imaging system. Anesthesia (isoflurane 1.5% in O2 delivered at 2 L/min) was administered throughout image acquisition. Living Image 4.3 software (Caliper Life Sciences) was used to generate a standard region of interest (ROI) encompassing the largest tumor at the maximal bioluminescence signal. The identical ROI was used to determine the average radiance (in photons/s/cm2) for all xenografts. Disease response was classified according to the bioluminescence signal as follows: CR (≤105 photons/s/cm2, similar to background); PR (105–107 photons/s/cm2); stable disease (SD; 107–108 photons/s/cm2, similar to the enrollment signal); and progressive disease (PD; >108 photons/s/cm2). Mice with tumor burden greater than 20% of body weight at any time were also classified as having PD.

SLFN11 expression is highly correlated with sensitivity to SN-38 and TAL

To determine the extent to which SLFN11 expression is correlated with sensitivity to DDA, we analyzed the GDSC database which contains over 1,000 cell lines assayed for cell viability 72 hours after exposure to hundreds of drugs (27). The efficacies of several DDAs, as measured by the AUC, were highly correlated with SLFN11 expression, with those of the strong-trapping PARPi TAL, and SN-38, the active metabolite of IRN, showing the highest statistical significance and the largest effect sizes (Fig. 1A; Supplementary Table S1A). In contrast, the microtubule inhibitors vinorelbine and vinblastine and the weak-trapping PARPi olaparib showed poor associations. The Pearson correlation between the mean AUC of SN-38 and TAL was 0.51 (P < 0.001), and there was a clear trend between this average and the increasing quintiles of SLFN11 expression (P < 0.001, 1-way ANOVA; Fig. 1B; Supplementary Table S1B). We observed no association between drug response and genetic lesions known to impair homologous recombination and sensitize tumor cells to both PARPis and Topo1is, such as BRCA1, BRCA2, and ATM (Fig. 1C; Supplementary Fig. S1A and S1B; Supplementary Table S1C–S1E; refs. 28–30). The short timescale of the GDSC viability assay suggests that SLFN11 induces rapid cytotoxicity, and this phenotype appears to be distinct from that induced by HR defects.

Figure 1.

SLFN11 was highly correlated with the efficacy of SN-38 and TAL, and was widely expressed in pediatric sarcoma. Correlation between SLFN11 expression and the single-agent AUC (A) or mean (B) of the AUC for SN-38 and TAL after 72-hour drug exposure as reported in the GDSC database. Olaparib and SN-38 appeared as 2 different batches in the database. C, Correlation between the mean of the AUC for SN-38 and TAL from the GDSC and BRCA2 mutational status as annotated in the COSMIC database. Dotted line equals the median activity of the highest quintile of SLFN11 expression (“Q5”) from (B). D, Distribution of the mean of the AUC for SN-38 and TAL from the GDSC in ES, RMS, OST, NB, and glioma cell lines. Cells marked “SLFN11 High” (salmon) expressed the highest quintile of SLFN11 expression (“Q5”) from (B), while all others were defined as “SLFN11 Low” (teal). Dotted line equals the median activity of the highest quintile of SLFN11 expression (“Q5”) from (B). E and F, Mutational signatures and total number of mutations calculated from ES, RMS, and OST tumor samples from pediatric patients. BRCA-wt and BRCA-deficient samples were included as controls for MS3. Melanoma was included as a positive control for MS7. G, SLFN11 status as assessed by IHC in 353 samples from 220 unique patients treated on solid-tumor protocols at our institution. “SLFN11 Negative” was defined as H score = 0, and “SLFN11 Positive” as H score > 0. H, SLFN11 H score at diagnosis for select pediatric sarcoma. The mean H score is reported for each tumor type.

Figure 1.

SLFN11 was highly correlated with the efficacy of SN-38 and TAL, and was widely expressed in pediatric sarcoma. Correlation between SLFN11 expression and the single-agent AUC (A) or mean (B) of the AUC for SN-38 and TAL after 72-hour drug exposure as reported in the GDSC database. Olaparib and SN-38 appeared as 2 different batches in the database. C, Correlation between the mean of the AUC for SN-38 and TAL from the GDSC and BRCA2 mutational status as annotated in the COSMIC database. Dotted line equals the median activity of the highest quintile of SLFN11 expression (“Q5”) from (B). D, Distribution of the mean of the AUC for SN-38 and TAL from the GDSC in ES, RMS, OST, NB, and glioma cell lines. Cells marked “SLFN11 High” (salmon) expressed the highest quintile of SLFN11 expression (“Q5”) from (B), while all others were defined as “SLFN11 Low” (teal). Dotted line equals the median activity of the highest quintile of SLFN11 expression (“Q5”) from (B). E and F, Mutational signatures and total number of mutations calculated from ES, RMS, and OST tumor samples from pediatric patients. BRCA-wt and BRCA-deficient samples were included as controls for MS3. Melanoma was included as a positive control for MS7. G, SLFN11 status as assessed by IHC in 353 samples from 220 unique patients treated on solid-tumor protocols at our institution. “SLFN11 Negative” was defined as H score = 0, and “SLFN11 Positive” as H score > 0. H, SLFN11 H score at diagnosis for select pediatric sarcoma. The mean H score is reported for each tumor type.

Close modal

We then defined the cell lines in the GDSC database as “SLFN11 High” (within the top quintile of expression) and all others as “SLFN11 Low” and examined the distribution of AUCs by tumor subtype (Fig. 1D). The correlation between SLFN11 expression and the average of SN-38 and TAL activity was evident in Ewing sarcoma, rhabdomyosarcoma (RMS), and osteosarcoma (OST) cell lines, although sampling was low for the latter 2 tumor types. In contrast, all but 1 neuroblastoma model expressed low SLFN11, despite 11 of 24 cell lines (46%) showing a drug response comparable with that of the highest quintile of SLFN11 expressors. Moreover, there was little difference in the drug response of high and low SLFN11 expressors in glioma cell lines, suggesting that the potential for SLFN11 to act as a biomarker predicting sensitivity to SN-38 and TAL varies across tumor types.

To better understand the relation between SLFN11 levels and DNA repair defects, we calculated MSs in tumors from pediatric patients with Ewing sarcoma, RMS, and OST (Fig. 1E and F; Supplementary Fig. S1C; Supplementary Table S1F), using data from BRCA-deficient and BRCA–wild-type (WT) cohorts as control (31, 32). As expected, expression of MS3, a signature associated with homologous recombination repair defects, was highest in the BRCA-deficient group. Consistent with recent reports, OST also showed elevated expression of this signature (33). In contrast, Ewing sarcoma tumors had MS3 levels comparable with those in BRCA-WT tumors, a finding inconsistent with reports that suggest translocations involving EWSR1 induce functional BRCA deficiency. Consistent with previously reported whole-genome sequencing studies (34, 35), Ewing sarcoma tumors have low genomic instability as assessed by the total number of mutations—an observation incompatible with the presence of HR deficiency.

Given the high level of SLFN11 expression in Ewing sarcoma and the scarcity of data on its expression in other pediatric sarcoma, we developed an IHC protocol using a commercially available antibody to assess protein levels. We assayed 353 samples from 220 different patients with non–central nervous system solid tumors who had sufficient material for staining at St. Jude Children's Research Hospital (Fig. 1G; Supplementary Table S1G and S1H). The patient demographics and diagnosis groups are shown in Table 1. SLFN11 had variable expression, but was nearly universal in Ewing sarcoma and DSRCT, with 90% and 100% of those tumors showing SLFN11 positivity (H score > 0), respectively. SLFN11 was detected in 75% of the samples from patients with OST or embryonal RMS (eRMS). Quantification of SLFN11 in the other tumor types was limited by small sample sizes. Using H score at diagnosis, we found the highest SLFN11 expression in Ewing sarcoma, followed by eRMS, OST, and DSRCT (Fig. 1H), with a few samples of the latter 3 tumor types having high expression levels similar to those observed in ES. Overall, SLFN11 was expressed in 69% of pediatric sarcoma sampled, and 76% of the most common pediatric sarcomas—a significantly higher percentage than has been implicated in adult tumors (36).

Table 1.

Demographics of the pediatric cohort.

All, N = 220 (%)Survival studies, N = 143 (%)
Age (years) 
N 185 135 
 Mean ± SD 9.2 ± 6.1 9.0 ± 6.4 
 Median (min–max) 9.6 (0.3–23.5) 8.8 (0.3–23.5) 
Sex 
 Female 82 (44.3) 63 (46.7) 
 Male 103 (55.7) 72 (53.3) 
 NA/missing 35 
Race 
 White 129 (74.1) 91 (71.7) 
 Black 34 (19.5) 26 (20.5) 
 Other 11 (6.3) 10 (7.9) 
 NA/missing 46 16 
Ethnicity 
 Non-Hispanic 123 (87.2) 97 (88.2) 
 Hispanic 18 (12.8) 13 (11.8) 
 NA/missing 79 33 
Metastatic Disease 
 No 68 (40.2) 43 (34.1) 
 Yes 101 (59.8) 83 (65.9) 
 NA/missing 51 17 
Chemotherapy 
 No 3 (1.7) 2 (1.6) 
 Yes 171 (98.3) 124 (98.4) 
 NA/missing 46 17 
Radiation 
 No 64 (36.8) 42 (33.3) 
 Yes 110 (63.2) 84 (66.7) 
 NA/missing 46 17 
Surgery 
 No 32 (18.4) 20 (15.9) 
 Yes 142 (81.6) 106 (84.1) 
 NA/missing 46 17 
Transplantation 
 No 127 (73.8) 84 (67.2) 
 Yes 45 (26.2) 41 (32.8) 
 NA/missing 48 18 
Diagnosis group 
 DSRCT 7 (3.2) 5 (3.5) 
 aRMS 8 (3.6) 5 (3.5) 
 oRMS 1 (0.5) 0 (0) 
 eRMS 12 (5.5) 5 (3.5) 
 OST 55 (25.0) 20 (14.0) 
 NB 44 (20.0) 42 (29.4) 
 ES 48 (21.8) 34 (23.8) 
 NRSTS 19 (8.6) 13 (9.1) 
 sNOS 26 (11.8) 19 (13.3) 
All, N = 220 (%)Survival studies, N = 143 (%)
Age (years) 
N 185 135 
 Mean ± SD 9.2 ± 6.1 9.0 ± 6.4 
 Median (min–max) 9.6 (0.3–23.5) 8.8 (0.3–23.5) 
Sex 
 Female 82 (44.3) 63 (46.7) 
 Male 103 (55.7) 72 (53.3) 
 NA/missing 35 
Race 
 White 129 (74.1) 91 (71.7) 
 Black 34 (19.5) 26 (20.5) 
 Other 11 (6.3) 10 (7.9) 
 NA/missing 46 16 
Ethnicity 
 Non-Hispanic 123 (87.2) 97 (88.2) 
 Hispanic 18 (12.8) 13 (11.8) 
 NA/missing 79 33 
Metastatic Disease 
 No 68 (40.2) 43 (34.1) 
 Yes 101 (59.8) 83 (65.9) 
 NA/missing 51 17 
Chemotherapy 
 No 3 (1.7) 2 (1.6) 
 Yes 171 (98.3) 124 (98.4) 
 NA/missing 46 17 
Radiation 
 No 64 (36.8) 42 (33.3) 
 Yes 110 (63.2) 84 (66.7) 
 NA/missing 46 17 
Surgery 
 No 32 (18.4) 20 (15.9) 
 Yes 142 (81.6) 106 (84.1) 
 NA/missing 46 17 
Transplantation 
 No 127 (73.8) 84 (67.2) 
 Yes 45 (26.2) 41 (32.8) 
 NA/missing 48 18 
Diagnosis group 
 DSRCT 7 (3.2) 5 (3.5) 
 aRMS 8 (3.6) 5 (3.5) 
 oRMS 1 (0.5) 0 (0) 
 eRMS 12 (5.5) 5 (3.5) 
 OST 55 (25.0) 20 (14.0) 
 NB 44 (20.0) 42 (29.4) 
 ES 48 (21.8) 34 (23.8) 
 NRSTS 19 (8.6) 13 (9.1) 
 sNOS 26 (11.8) 19 (13.3) 

Abbreviations: DSRCT, desmoplastic small round cell tumor; aRMS, alveolar rhabdomyosarcoma; oRMS, other rhabdomyosarcoma; eRMS, embryonal rhabdomyosarcoma; OST, osteosarcoma; NB, neuroblastoma; ES, Ewing sarcoma; NRSTS, non-rhabdomyosarcoma soft tissue sarcoma; sNOS, sarcoma not otherwise specified.

SLFN11 drives sensitivity to SN-38 and TAL in vitro

To further assess SLFN11 and EWSR1 translocation as drivers of sensitivity to TAL and SN-38, we profiled 14 sarcoma cell lines that varied by translocation type, p53 status, and histology, then we assessed SLFN11 expression by IHC, Western blot, and qPCR analysis (Fig. 2A, Supplementary Table S2A and S2B). The Pearson correlation between SLFN11 protein and mRNA levels was 0.64 (P = 0.018; Supplementary Fig. S2A). Protein levels correlated with sensitivity to single-agent SN-38 and TAL, with Pearson correlations of 0.72 (P = 0.003) and 0.74 (P = 0.003), respectively (Supplementary Fig. S2B); and with the mean AUC of the 2 compounds, with a Pearson correlation of 0.77 (P = 0.001; Fig. 2B). We found no association between sensitivity to these drugs and p53 status (P = 0.12, t test; Supplementary Fig. S2C). Although most EWSR1-translocated cell lines were more sensitive to drug treatment when compared with nontranslocated cell lines, they also tended to express the highest levels of SLFN11. The exception was SU-CCS-1, a SLFN11-negative (no protein detected by Western; IHC H-score = 0) EWSR1-ATF1-translocated clear-cell sarcoma, suggesting that the EWSR1 translocation alone was insufficient to drive drug sensitivity.

Figure 2.

SLFN11 predicted sensitivity to SN-38 and TAL in vitro. A, SLFN11 protein levels and CellTiter-Glo (CTG) results for 14 sarcoma cell lines that varied by translocation status, p53 status, and histology. Protein levels were normalized to ES8. The AUC for the dose-response curve was calculated at 24, 48, and 72 hours in the concentration range 10–11 to 10–4 Molar. Each value was then normalized to 700 – the maximum observed AUC for 100% efficacy at all concentrations in the range – yielding a number from 0 to 1. Although extraskeletal myxoid chondrosarcoma (EMC) cancers such as H-EMC-SS typically have EWSR1-NR4A3 fusions, we did not detect a EWSR1-translocation in this line. n ≥ 2. B, Correlation between SLFN11 protein levels and the mean of the AUC for SN-38 and TAL for the cell panel in (A). C, CTG dose-response following 72-hour drug exposure of SN-38 and vincristine in ES8, ES8-SLFN11-KO, and ES8-SLFN11-KO+OE cells. n ≥ 2. D, Difference in normalized CTG AUC following 72-hour drug exposure of TAL, SN-38, and vincristine in KO and overexpression models. n ≥ 2. E, Flow cytometry assessment of cytotoxicity in ES8 and ES8-SLFN11-KO cells following 24-hour exposure to ‘Low’ (10 nmol/L SN-38 + 10 nmol/L TAL) and ‘High’ (1 μmol/L SN-38 and 1 μmol/L TAL) drug combinations. The percent live cells is reported in blue. EA, early apoptosis; LA, late apoptosis. n ≥ 2. F, Cell-cycle analysis of ES8-SLFN11-KO cells following 24-hour exposure to ‘Low’ SN-38 + TAL. Arrows highlight the build-up of S-phase cells induced by the drug combination. G, Volcano plot showing the difference in expression of reported EWS-FLI1 downregulated (salmon) and upregulated (teal) genes between ES8 and ES8-SLFN11-KO cells. Expression was assessed by microarray at 4 hours and 24 hours following exposure to 0 and 2 Gy. n = 3. H, Exemplar image and quantification of the alkaline comet-tail assay of ES8 and ES8-SLFN11-KO cells following 2.5-hour exposure to the ‘High’ concentration of SN-38 and TAL. Mean percent comet tail DNA is reported in blue. I, Immunofluorescence quantification of R-loops in untreated ES8 and ES8-SLFN11-KO cells. Nuclei were stained with DAPI (blue) and R-loops were stained with S9.6 antibody (red). Mean (SEM) S9.6 nuclear intensity is reported.

Figure 2.

SLFN11 predicted sensitivity to SN-38 and TAL in vitro. A, SLFN11 protein levels and CellTiter-Glo (CTG) results for 14 sarcoma cell lines that varied by translocation status, p53 status, and histology. Protein levels were normalized to ES8. The AUC for the dose-response curve was calculated at 24, 48, and 72 hours in the concentration range 10–11 to 10–4 Molar. Each value was then normalized to 700 – the maximum observed AUC for 100% efficacy at all concentrations in the range – yielding a number from 0 to 1. Although extraskeletal myxoid chondrosarcoma (EMC) cancers such as H-EMC-SS typically have EWSR1-NR4A3 fusions, we did not detect a EWSR1-translocation in this line. n ≥ 2. B, Correlation between SLFN11 protein levels and the mean of the AUC for SN-38 and TAL for the cell panel in (A). C, CTG dose-response following 72-hour drug exposure of SN-38 and vincristine in ES8, ES8-SLFN11-KO, and ES8-SLFN11-KO+OE cells. n ≥ 2. D, Difference in normalized CTG AUC following 72-hour drug exposure of TAL, SN-38, and vincristine in KO and overexpression models. n ≥ 2. E, Flow cytometry assessment of cytotoxicity in ES8 and ES8-SLFN11-KO cells following 24-hour exposure to ‘Low’ (10 nmol/L SN-38 + 10 nmol/L TAL) and ‘High’ (1 μmol/L SN-38 and 1 μmol/L TAL) drug combinations. The percent live cells is reported in blue. EA, early apoptosis; LA, late apoptosis. n ≥ 2. F, Cell-cycle analysis of ES8-SLFN11-KO cells following 24-hour exposure to ‘Low’ SN-38 + TAL. Arrows highlight the build-up of S-phase cells induced by the drug combination. G, Volcano plot showing the difference in expression of reported EWS-FLI1 downregulated (salmon) and upregulated (teal) genes between ES8 and ES8-SLFN11-KO cells. Expression was assessed by microarray at 4 hours and 24 hours following exposure to 0 and 2 Gy. n = 3. H, Exemplar image and quantification of the alkaline comet-tail assay of ES8 and ES8-SLFN11-KO cells following 2.5-hour exposure to the ‘High’ concentration of SN-38 and TAL. Mean percent comet tail DNA is reported in blue. I, Immunofluorescence quantification of R-loops in untreated ES8 and ES8-SLFN11-KO cells. Nuclei were stained with DAPI (blue) and R-loops were stained with S9.6 antibody (red). Mean (SEM) S9.6 nuclear intensity is reported.

Close modal

To confirm SLFN11 as the primary driver of sensitivity to these agents, we knocked out the gene by using CRISPR/Cas9 in our 3 highest-expressing models, generating the isogenic pairs: ES8/ES8-SLFN11-knockout (KO), A673/A673-SLFN11-KO, and JN-DSRCT/JN-DSRCT-SLFN11-KO. We also overexpressed the protein in U2OS cells, which showed little baseline expression, to create U2OS+OE cells (61% protein relative to total ES8 by Western blot); and in ES8-SLFN11-KO cells, to create ES8-SLFN11-KO+OE cells (37% protein relative to total ES8). KO and overexpression were confirmed by Western blot analysis and IHC (Supplementary Fig. S2D and S2E; Supplementary Table S2A and S2B). Loss of SLFN11 protein significantly reduced sensitivity to both SN-38 and TAL in all three KO lines, whereas overexpression in U2OS and ES8-SLFN11-KO cells increased drug sensitivity (Fig. 2C and D). Consistent with our GDSC analysis, SLFN11 loss had little effect on vincristine sensitivity. Despite expressing a lower level of the same engineered SLFN11 protein construct, ES ES8-SLFN11-KO+OE cells had higher AUC values for SN-38 and TAL compared with OST U2OS+OE cells (0.40 vs. 0.25 for SN-38 and 0.26 vs. 0.10 for TAL), consistent with the hypothesis that the magnitude of sensitization induced by SLFN11 varies between tumor types (Supplementary Fig. S2F). Ionizing radiation is another means to induce replicative stress (37). In agreement with our SN-38 and TAL experiments, ES8-SLFN11-KO and JN-DSRCT-SLFN11-KO were more viable, and U2OS+OE was less viable, compared with their WT counterparts at 72 hours after exposure to 4 Gy radiation (Supplementary Fig. S2G).

To further study the effect of SLFN11 KO in our models, we used flow cytometry to compare cell-cycle effects and the degree of cell death induced by “Low” (10 nmol/L SN-38 + 10 nmol/L TAL) and “High” (1,000 nmol/L SN-38 + 1,000 nmol/L TAL) dose combinations after 24 hours of exposure. Based on our previous pharmacokinetic assessment, “Low” approximates to the upper bound of clinically relevant concentrations for both drugs, whereas “High” is physiologically unobtainable but useful for studying mechanism and resistance (6). WT ES8 cells showed near-complete loss of viability at both “Low” and “High” doses of the combination, whereas the ability of SN-38 + TAL to induce cell death in ES8-SLFN11-KO cells was significantly diminished (Fig. 2E). A similar decrease in cell viability was observed in JN-DSRCT compared with JN-DSRCT-SLFN11-KO cells, and the combination was also less cytotoxic in SLFN11-negative SU-CCS-1 cells (Supplementary Fig. S2H). SLFN11 selectively induces death in cells arrested in S-phase because of replicative stress (15). Consistent with this finding, we observed a significant build-up of S-phase–arrested ES8-SLFN11-KO cells (Fig. 2F).

Previous studies have shown that SLFN11 is a transcriptional target of EWS-FLI1 (10). To determine whether SLFN11 itself contributes to the regulation of EWSR1-FLI1 target genes, we assessed gene expression in ES8 and ES8-SLFN11-KO cells at 4 hours and 24 hours following exposure to 0 or 2 Gy of ionizing radiation (baseline and stress conditions). EWSR1-FLI known targets mapped equally between upregulated and downregulated genes, and we found no enrichment of directional activation or inhibition of those EWSR1-FLI1 target genes when using Gene Set Enrichment Analysis (GSEA, FDR > 0.05; Fig. 2G; Supplementary Table S2C; ref. 38). Therefore, although SLFN11 is regulated by EWS-FLI1, it appears to perturb gene expression independently of the fusion protein.

To determine whether SLFN11 influenced the extent of DNA damage induced by SN-38 and TAL, we performed an alkaline comet-tail assay after exposing ES8 and ES8-SLFN11-KO cells to DMSO and the “High” concentration of SN-38 and TAL for 2 hours (Fig. 2H). The amount of DNA damage was similar in both cell lines after drug treatment, indicating that SLFN11 does not enhance the degree of damage but rather increases the probability of cell death following drug insult. Finally, given the findings of high levels of R-loops in EWSR1-translocated tumors (11), we quantified R-loop expression in ES8 and ES8-SLFN11-KO cells. Despite a remarkable difference in their response to SN-38 and TAL, we found no significant difference in R-loop levels in the WT and SLFN11 KO models (Fig. 2I).

SLFN11 drives sensitivity to TAL and IRN in vivo

To confirm our in vitro finding indicating that SLFN11 was an important driver of drug response in EWSR1-translocated tumors, we conducted in vivo efficacy studies, as described previously (6), using luciferase-labeled xenografts of ES8, ES8-SLFN11-KO, JN-DSRCT, JN-DSRCT-SLFN11-KO, and SU-CCS-1 cells. Mice were screened weekly by Xenogen imaging and enrolled once a target bioluminescence signal of 107 photons/s/cm2 or a palpable tumor was obtained. We used clinically relevant doses and schedules for all treatment groups tested (Fig. 3A; ref. 6) and administered 4 courses of therapy (21 days/course).

Figure 3.

SLFN11 was required for sensitivity to SN-38 and TAL in vivo. A, Drug schedule selected for each combination treatment regimen. B, Line plot of tumor burden over time as measured by bioluminescence for ES8 xenografts. Each line is a different mouse. C, Representative images from the study in (B) are shown with PD in the placebo control group and a CR in the TAL + TMZ + IRN group. D, Line plot of tumor burden over time for ES8-SLFN11-KO xenografts. E, Representative images from the study in (D) are shown with PD in both the placebo and TAL + TMZ + IRN groups. F, Survival curves and response at end of treatment with TAL + TMZ + IRN for ES8 and ES8-SLFN11-KO models. G, Line plot of tumor burden over time for JN-DSRCT xenografts. H, Representative images from the study in (G) are shown with PD in the placebo group and a CR in the TAL + TMZ + IRN group. I, Line plot of tumor burden over time for SU-CCS-1 xenografts. J, Representative images of a mouse treated over time in the TAL + TMZ + IRN group from (I) showing a PR during treatment and regrowth of tumor upon stopping therapy.

Figure 3.

SLFN11 was required for sensitivity to SN-38 and TAL in vivo. A, Drug schedule selected for each combination treatment regimen. B, Line plot of tumor burden over time as measured by bioluminescence for ES8 xenografts. Each line is a different mouse. C, Representative images from the study in (B) are shown with PD in the placebo control group and a CR in the TAL + TMZ + IRN group. D, Line plot of tumor burden over time for ES8-SLFN11-KO xenografts. E, Representative images from the study in (D) are shown with PD in both the placebo and TAL + TMZ + IRN groups. F, Survival curves and response at end of treatment with TAL + TMZ + IRN for ES8 and ES8-SLFN11-KO models. G, Line plot of tumor burden over time for JN-DSRCT xenografts. H, Representative images from the study in (G) are shown with PD in the placebo group and a CR in the TAL + TMZ + IRN group. I, Line plot of tumor burden over time for SU-CCS-1 xenografts. J, Representative images of a mouse treated over time in the TAL + TMZ + IRN group from (I) showing a PR during treatment and regrowth of tumor upon stopping therapy.

Close modal

In ES8 (high SLFN11) xenografts, mice treated with TAL + IRN + TMZ had the best response, with 100% surviving the 84 days of therapy, and 75% experiencing a CR or PR as determined by the bioluminescence signal (Fig. 3B and C; Supplementary Table S3). ES8 mice treated with TAL + IRN survived an average of 60.5 days, with 50% surviving all 4 courses of therapy. In sharp contrast, mice with ES8-SLFN11-KO xenografts treated with TAL + IRN and TAL + IRN + TMZ survived an average of 7.5 days and 23.1 days, respectively, with 100% having PD (Fig. 3D and E; Supplementary Table S3). No mice in any ES8-SLFN11-KO treatment cohort survived all 4 courses of therapy, although 1 control mouse appeared to have lost its engraftment signal after 2 weeks and survived the study. Compared with ES8 mice, survival in ES8-SLFN11-KO mice was significantly lower when treated with either TAL + IRN (P < 0.001) or TAL + IRN + TMZ (P < 0.001; ref. Fig. 3F).

Testing JN-DSRCT (high SLFN11) xenografts in vivo was challenging, as these tumors grew slower and had lower engraftment rates by comparison with ES8 xenografts. Five WT mice were enrolled and treated with either TAL + IRN or TAL + IRN + TMZ, and all experienced a CR by 84 days (Fig. 3G and H; Supplementary Table S3). One untreated mouse was also enrolled and maintained SD throughout the 4 courses of therapy. Interestingly, JN-DSRCT-SLFN11-KO cells lost the ability to engraft and were unable to be tested. In SU-CCS-1 (no SLFN11) xenografts, 90% of mice treated with TAL + IRN + TMZ had SD or a PR at the end of therapy (Fig. 3I). However, all mice regrew tumors within a few weeks of stopping therapy (Fig. 3J). Together, these findings confirm the importance of SLFN11 in driving in vivo sensitivity to combinations involving SN-38 and TAL in EWSR1-translocated tumors.

SLFN11 positivity is not associated with better outcomes in children with sarcoma

Motivated by the strong evidence that SLFN11-sensitized pediatric sarcomas to PARPi combination therapy in vitro and in vivo, we performed a retrospective analysis of the patient cohort profiled in our IHC study to determine how SLFN11 status changed throughout therapy and whether protein levels predicted clinical outcome. Only patients who had a sample available prior to recurrence or progression were included in the survival analysis (N = 143, Table 1). 98.4% of evaluable patients were treated with at least one DDA and 66.7% received radiation at some point in their therapy. This population was more refractory than would be expected historically, with a 5-year OS and event-free survival (EFS) of less than 50% in patients with NRSTS, ES, and eRMS (Fig. 4A; Supplementary Fig. S3A).

Figure 4.

SLFN11 expression was not associated with better outcomes in children with solid tumors. A, Five-year OS rates by diagnosis for the patients in our IHC study. B, SLFN11 status for 18 patients with at least 2 SLFN11 IHC measurements spanning different points in treatment. “Negative” and “Positive” SLFN11 status were defined as H score equal to zero and H-score greater than 0, respectively. C, ROC curve for 5-year OS of 143 patients with solid tumors as a function of H score. D, ROC curve for 5-year EFS of 143 patients with solid tumors as a function of H score. E, ROC curve for 5-year EFS of ES patients as a function of H score. F, ROC curve for 5-year EFS of NRSTS patients as a function of H score. G, ROC curve for 5-year EFS of OST patients as a function of H score. H, Adjusted OS as a function of SLFN11 status after controlling for age, metastatic status, and disease. I, Adjusted EFS as a function of SLFN11 status after controlling for age, metastatic status, and disease.

Figure 4.

SLFN11 expression was not associated with better outcomes in children with solid tumors. A, Five-year OS rates by diagnosis for the patients in our IHC study. B, SLFN11 status for 18 patients with at least 2 SLFN11 IHC measurements spanning different points in treatment. “Negative” and “Positive” SLFN11 status were defined as H score equal to zero and H-score greater than 0, respectively. C, ROC curve for 5-year OS of 143 patients with solid tumors as a function of H score. D, ROC curve for 5-year EFS of 143 patients with solid tumors as a function of H score. E, ROC curve for 5-year EFS of ES patients as a function of H score. F, ROC curve for 5-year EFS of NRSTS patients as a function of H score. G, ROC curve for 5-year EFS of OST patients as a function of H score. H, Adjusted OS as a function of SLFN11 status after controlling for age, metastatic status, and disease. I, Adjusted EFS as a function of SLFN11 status after controlling for age, metastatic status, and disease.

Close modal

Loss of SLFN11 expression has been proposed as a mechanism of DDA resistance in adult tumors (16). To explore the trend of SLFN11 levels in our cohort, we identified 18 patients with at least 2 IHC measurements spanning different points in treatment (Fig. 4B; Supplementary Fig. S3B). We found no evidence that SLFN11 was silenced over time: 10 of 18 of patients remained positive throughout, 3 of 18 went from being SLFN11 negative to SLFN11 positive, 3 of 18 went from being SLFN11 positive to SLFN11 negative, and 2 of 18 were initially SLFN11 positive then had a negative sample followed by another positive sample. When including all patients in our IHC study, 37.2% of diagnostic samples were positive, compared with 46.2% of progression samples and 53.7% of relapse samples (Supplementary Table S4A). We observed no clear trend in the mean values for SLFN11 H score in samples obtained at diagnosis, relapse, or autopsy.

Using ROC analysis, SLFN11 expression as measured by H score failed to discriminate patients by EFS, OS, recurrence-free survival (RFS), or PFS across all tumor types (Fig. 4C and D; Supplementary Fig. S3C and S3D), or within individual disease cohorts (Fig. 4EG). Surprisingly, SLFN11 positivity was a statistically significant predictor of worse outcome in terms of RFS (P = 0.045, hazard ratio = 1.76; 1.01–3.05) in univariate analysis (Supplementary Table S4B). The OS (P = 0.072, hazard ratio = 1.82; 0.95–3.49), EFS (P = 0.10, hazard ratio = 1.58; 0.92–2.73), and PFS (P = 0.078, hazard ratio = 1.72; 0.94–3.13) were nonsignificantly associated with poorer outcomes in patients with SLFN11-expressing tumors. In fact, although statistical significance was not achieved for any metric in multivariate analysis controlling for age, metastatic status, and disease, the HR for positivity remained at or greater than unity (Fig. 4H and I; Supplementary Fig. S3E and S3F). ROC analysis using the H scores failed to discriminate patients with metastatic disease (AUC = 0.605). Moreover, neither SLFN11 H score nor positivity were statistically significant predictors of metastasis in multivariate analysis that controlled for diagnosis (P = 0.58 and 0.35, respectively). In summary, our clinical findings indicated that in pediatric sarcomas: (i) SLFN11 positivity is common; (ii) SLFN11 is not frequently silenced at recurrence or disease progression; and (iii) SLFN11 positivity does not predict improved outcome after treatment with DDA.

Mechanisms of resistance to SN-38 and TAL in SLFN11-positive pediatric sarcomas

The lack of SLFN11 silencing in the pediatric sarcoma population compelled us to search for alternative mechanisms of resistance to DDA. Therefore, we mined the GDSC database to identify the sarcoma cell lines with the highest levels of SLFN11 expression that were also refractory (having less than the median activity observed in the GDSC) to both SN-38 and TAL. We identified 3 lines of interest: EW-13, EW-18, and EW-24 (Supplementary Fig. S4A). These lines showed broad resistance to oncology drugs – including vincristine – when compared to sensitive ES models with high SLFN11 expression (Fig. 5A; Supplementary Table S5). We obtained EW-13 and EW-18, as well as EW-11 which showed intermediate sensitivity in the database. Separately, we acquired CHLA258, which was reported elsewhere to be less sensitive to Topo1is (39). We confirmed drug resistance with CTG assays and flow cytometry (Fig. 5B; Supplementary Fig. S4B). Western blot analysis indicated strong SLFN11 expression in EW-11, EW-13, and CHLA258; however, protein expression in EW-18 was weaker and the band migration corresponded to a lower molecular weight (Fig. 5C; Supplementary Fig. S4C). Cross-referencing with the COSMIC database (https://cancer.sanger.ac.uk/cosmic) revealed that the SLFN11 gene in EW-18 had a frameshift mutation, c.1928_1929insA, which resulted in a C-terminal truncation that was predicted to limit nuclear localization and, therefore, reduce DDA sensitization (14). IHC confirmed that SLFN11 was predominantly expressed in the cytoplasm, and consequently, the cell line was assigned an H score of 0 (Supplementary Fig. S4D and S4E). It is important to note that, as far we are aware, all commercially available antibodies target the N-terminus of SLFN11 and could fail to detect C-terminal truncations unless a distinction is made between nuclear and cytoplasmic staining as we did in this study.

Figure 5.

Identification and characterization of SLFN11-positive ES models that are resistant to SN-38 and TAL. A, Boxplot comparing drug sensitivity in resistant and sensitive ES cell lines from the GDSC. B, Heatmap of AUC values in 4 resistant ES cell lines normalized to ES8 (CTG assay, 72-hour drug exposure). n ≥ 2. C, Western blot from 2 biological replicates of EW-18 confirming expression of a truncated SLFN11 protein. D, Dose-response curves for SN-38 (CTG, 72 hours) in ES8 (black), SLFN11-expressing resistant ES cell lines (grays), ES8-SLFN11-KO (red), and EW-18 (dark red). n ≥ 2. E, Dose-response curves for SN-38 and TAL (CTG, 72 hours) in ES8 (black), ES8-SLFN11-KO (red), 2 ES8-BAK-KO models (blue), 2 ES8-BAX-KO models (green), and 2 ES8-BAK-BAX double KO models (purple). n ≥ 2. F, Heatmap of AUC values in ES8-SLFN11-KO and the BAK, BAX, and BAK-BAX-DKO models normalized to ES8 (CTG assay, 72-hour drug exposure). n ≥ 2. G, RNA-seq expression profile comparing the ES PDOX model SJEWS049193_X1 and the matched primary tumor (SJEWS049193_D1, Pearson r = 0.85), and hematoxylin and eosin (H&E) stain of SJEWS049193_X1. H, SLFN11 IHC and Western blot from a SJEWS049193_X1 tumor sample. I, Survival curves for SJEWS049193_X1 and J, representative images from the efficacy study showing PD in both the placebo and TAL+TMZ+IRN groups.

Figure 5.

Identification and characterization of SLFN11-positive ES models that are resistant to SN-38 and TAL. A, Boxplot comparing drug sensitivity in resistant and sensitive ES cell lines from the GDSC. B, Heatmap of AUC values in 4 resistant ES cell lines normalized to ES8 (CTG assay, 72-hour drug exposure). n ≥ 2. C, Western blot from 2 biological replicates of EW-18 confirming expression of a truncated SLFN11 protein. D, Dose-response curves for SN-38 (CTG, 72 hours) in ES8 (black), SLFN11-expressing resistant ES cell lines (grays), ES8-SLFN11-KO (red), and EW-18 (dark red). n ≥ 2. E, Dose-response curves for SN-38 and TAL (CTG, 72 hours) in ES8 (black), ES8-SLFN11-KO (red), 2 ES8-BAK-KO models (blue), 2 ES8-BAX-KO models (green), and 2 ES8-BAK-BAX double KO models (purple). n ≥ 2. F, Heatmap of AUC values in ES8-SLFN11-KO and the BAK, BAX, and BAK-BAX-DKO models normalized to ES8 (CTG assay, 72-hour drug exposure). n ≥ 2. G, RNA-seq expression profile comparing the ES PDOX model SJEWS049193_X1 and the matched primary tumor (SJEWS049193_D1, Pearson r = 0.85), and hematoxylin and eosin (H&E) stain of SJEWS049193_X1. H, SLFN11 IHC and Western blot from a SJEWS049193_X1 tumor sample. I, Survival curves for SJEWS049193_X1 and J, representative images from the efficacy study showing PD in both the placebo and TAL+TMZ+IRN groups.

Close modal

Although the SLFN11 mutation in EW-18 could explain its weak response to DDAs, our analysis of the GDSC and COSMIC databases indicated that the frequency of SLFN11 mutation in the resistant population was low, with 18/22 cell lines (82%) having a wild-type sequence. We chose SLFN11-wild-type CHLA258 for further investigation. We confirmed strong nuclear expression of the protein by IHC (H score = 225; Supplementary Fig. S4E). Consistent with the other tumor lines, SN-38 + TAL induced a high level of DNA damage as assessed by the alkaline comet-tail assay (Supplementary Fig. S4F). The “Low” combination of SN-38 and TAL did not substantially increase apoptosis but instead induced a build-up of S-phase cells similar to that observed with ES8-SLFN11-KO cells (Supplementary Fig. S4G). However, the S-phase population was depleted by the “High” concentration.

Overlaying the SN-38 dose-response curves for the ES8, ES8-SLFN11-KO, and resistant cell lines revealed 2 distinct phenotypes (Fig. 5D). The dose-response curves for EW-11, EW-13, CHLA258, and A673 (the least sensitive Ewing sarcoma cell line in our original panel) showed an inflection at concentrations of less than 10 nmol/L, as seen with ES8 cells, but did not show the same level of maximum efficacy (percentage inhibition of viability). Consistent with the lack of SLFN11 nuclear localization, EW-18 behaved more like the ES8-SLFN11-KO cells. The efficacy of TAL was substantially less in all cell lines, and its potency was 2- to 6-fold within that in ES8 cells. The exceptions to this were EW-13, in which TAL was more potent than in ES8 cells, and EW-18, in which TAL was inactive (Supplementary Fig. S5A).

To probe the mechanism of cytotoxicity induced by TAL and SN-38, we treated ES8 cells exposed to each drug with the caspase inhibitor Z-VAD-FMK, and discovered a substantial reduction in efficacy suggesting that apoptosis was the primary mode of cell death with this drug combination (Supplementary Fig. S5B). To investigate the role of intrinsic apoptosis in mediating drug sensitivity, we knocked out BAX, BAK, and both BAX and BAK in ES8 cells (Supplementary Fig. S5C). The potency of SN-38 and TAL in BAX and BAX + BAK double KO cells remained within 2-fold of that in WT ES8 cells, but their efficacy was significantly reduced—a phenotype similar to that observed in EW-11, EW-13, CHLA258, and A673 (Fig. 5E). BAX deletion alone reduced sensitivity to the DDAs doxorubicin and etoposide to levels comparable with those in ES8-SLFN11-KO cells, but the deletion also reduced sensitivity to vincristine—a phenotype distinct from that of SLFN11 KO cells and comparable with that observed in the resistant Ewing sarcoma cell lines described earlier (Fig. 5F).

During this project, it was reported that resistance to the PARPi olaparib in Ewing sarcoma cell lines could be overcome with the pan-BCL2 inhibitor navitoclax (40). Motivated by that work and by our own findings, we screened small-molecules inhibitor that were selective for individual BCL2 family members: venetoclax (BCL2), S63845 (MCL), and A-1331852 (BCL-xL). BCL-xL inhibition alone sensitized sarcoma cells to the combination of SN-38 and TAL (Supplementary Fig. S5D). Addition of A-1331852 decreased cell viability in A673 and CHLA-258 by a factor of 5.2- and 7.9, respectively, relative to that with the “Low” concentration alone. Although BCL-xL inhibition enhanced the efficacy of the combination in SLFN11-negative U2OS cells, the change in U2OS+OE cells was significantly greater (a 7.5-fold increase vs. a 2.8-fold increase). Taken together, these studies suggest that impairment of the intrinsic apoptotic pathway constitutes one means of resistance to TAL and SN-38 in pediatric sarcomas, and that selective inhibition of BCL-xL can increase sensitivity to the drug combination in SLFN11-positive resistant tumors.

Finally, to explore the relationship between SLFN11 and sensitivity to TAL and IRN in a more physiologically-relevant model of pediatric sarcoma, we developed and characterized a patient-derived orthotopic xenograft (PDOX) model, SJEWS049193_X1, using a tumor obtained at autopsy from a child with metastatic Ewing sarcoma treated with multiple salvage regimens, including treatment on the TAL + IRN + TMZ clinical trial NCT02392793. The model recapitulated the expression profile and histopathology of the parent tumor from which it was derived, SJEWS049193_D1 (Fig. 5G). IHC confirmed a high level of nuclear SLFN11 expression in this tumor (H score = 285), and Western blot analysis and RNA-seq indicated no discernible alteration or mutation (Fig. 5H; Supplemental Data). Consistent with its clinical response, the PDOX model also failed to respond to TAL + IRN + TMZ in vivo, and 100% of mice showed PD (Fig. 5I and J; Supplementary Table S3). To our knowledge, this is the first reported PDOX model that expressed high levels of SLFN11 but failed to respond to combination therapy involving DDA.

The heterogeneous and aggressive nature of pediatric sarcomas makes it imperative to identify biomarkers for drug response and new therapeutic targets. Using IHC, we found that SLFN11 was widely expressed in our cohort of pediatric sarcomas, with H scores reaching near maximum in some samples. The sensitivity to TAL and SN-38 in sarcoma cell lines correlated with protein levels both in vitro and in vivo and was independent of the EWSR1 translocation. However, despite the strong link between protein expression and DDA sensitivity in our preclinical studies, we found no association between SLFN11 status and improved outcome in a retrospective analysis of our patient cohort. Moreover, we found no evidence that SLFN11 was silenced in recurrent disease. These findings contrast with recent reports indicating that certain SLFN11-positive tumors have better outcomes with DDA therapy and that gene silencing is a principal route to DDA resistance.

Strategies such as combining ATR inhibitors with Topo1i have been reported to sensitize SLFN11-negative tumors to DNA-damage induced replicative stress (15). Our work advances the field by identifying sarcoma models expressing high levels of SLFN11 that were resistant to TAL and SN-38. Further, except for EW-18 in which the SLFN11 gene was truncated, mutations in SLFN11, the extent of DNA damage, or change in R-loop levels could not account for the resistance. Some of these resistant models appear to compensate for the DDA vulnerability induced by SLFN11 expression by attenuating intrinsic apoptosis. Selective inhibition of BCL-xL increased the cytotoxicity of the combination of TAL and SN-38 in resistant Ewing sarcoma cell lines and resulted in enhanced drug efficacy in OST U2OS cells engineered to overexpress SLFN11.

Our retrospective analysis of SLFN11 and outcome in pediatric sarcoma was limited. Although most patients received at least one DDA, the treatment modalities and disease types varied widely. Moreover, our population appeared to be more refractory to therapy than what would be expected, likely owing to the categories of patients treated at our institution. This characteristic could potentially skew our analysis, as our work suggests that mechanisms of resistance besides SLFN11 silencing may be responsible for resistance in this population. Further, the importance of histology needs to be explored in more depth. While the KO of SLFN11 reduced sensitivity to the combination of TAL and SN-38 in Ewing sarcoma and DSRCT cells, and the overexpression of SLFN11 increased sensitivity in an OST cell line, the sensitivity to these drugs observed in largely SLFN11 negative neuroblastoma cell lines and the lack of correlation to SLFN11 levels in glioma cell lines indicates that that there are multiple factors driving response to replicative stress. Consequently, our work strongly supports the prospective evaluation of SLFN11 as a biomarker predicting the response to PARPi and Topo1i in pediatric sarcoma that appear to express varying levels of the protein, such as ES, DSRCT, OS, and eRMS, along with concurrent interrogation of BCL-2 family members such as BAX and BCL-xL.

The apparent lack of silencing of SLFN11 in the pediatric sarcoma population suggests the strategies to overcome resistance to these agents in adults may not be applicable to a significant portion of the pediatric population, and therefore, alternative avenues should be explored. The ability to augment therapeutic response by selective BCL-xL inhibition warrants further investigation of drug combinations that target both replicative stress and the intrinsic apoptotic pathway. Ultimately, our work provides a framework to develop rational, targeted combination therapy approaches for both treatment naïve and refractory SLFN11-positive pediatric sarcoma.

No disclosures were reported.

J. Gartrell: Conceptualization, data curation, formal analysis, investigation, writing–original draft, writing–review and editing. M. Mellado-Lagarde: Conceptualization, formal analysis, investigation, writing–original draft, writing–review and editing. M.R. Clay: Data curation, formal analysis, investigation, writing–original draft. A. Bahrami: Data curation, formal analysis, supervision, investigation, writing–original draft. N.A. Sahr: Data curation, software, formal analysis, writing–original draft. A. Sykes: Data curation, software, formal analysis. K. Blankenship: Investigation. L. Hoffman: Investigation. J. Xie: Investigation, writing–original draft. H.P. Cho: Investigation, methodology. N. Twarog: Software, formal analysis, investigation, visualization, writing–original draft. M. Connelly: Investigation, writing–original draft. K.K. Yan: Formal analysis, investigation, writing–original draft. J. Yu: Supervision, investigation. S.N. Porter: Investigation. S.M. Pruett-Miller: Supervision, investigation, writing–original draft. G. Neale: Data curation, formal analysis, investigation, writing–original draft. C.L. Tinkle: Conceptualization, resources, data curation, supervision, writing–original draft, project administration, writing–review and editing. S.M. Federico: Conceptualization, resources, supervision, writing–original draft. E.A. Stewart: Conceptualization, resources, formal analysis, supervision, funding acquisition, investigation, writing–original draft, project administration, writing–review and editing. A.A. Shelat: Conceptualization, resources, formal analysis, supervision, funding acquisition, investigation, writing–original draft, project administration, writing–review and editing.

The authors thank Nancy E. Martinez for assistance with drug screening and characterization of the BAX and BAK knockout cell lines. We thank Keith A. Laycock, PhD, ELS, for scientific editing of the manuscript. We thank Dr. Richard Ashmun and the Flow Cytometry and Cell Sorting Shared Resource at St. Jude for expert guidance, and the Childhood Solid Tumor Network (http://www.stjude.org/CSTN/) for providing xenograft models. Preclinical imaging was performed with help from the Center for In Vivo Therapeutics at St. Jude. Cell images were acquired at the Cell and Tissue Imaging Center, cytogenetics was performed by the Cytogenetic Shared Resource, and microarray data were generated at the Hartwell Center for Bioinformatics and Biotechnology, all of which are supported by St. Jude and by the NCI grant P30 CA021765. Additionally, this work was supported by the American Lebanese Syrian Associated Charities (A. Sykes, E.A. Stewart). E.A. Stewart was supported by the National Comprehensive Cancer Network and is a St. Baldrick's Scholar with generous support from the Invictus Fund. A.A. Shelat was supported by the Sarcoma Foundation of America and the St. Baldrick's Foundation. The content of this paper is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Pappo
AS
,
Dirksen
U
. 
Rhabdomyosarcoma, Ewing sarcoma, and other round cell sarcomas
.
J Clin Oncol
2018
;
36
:
168
79
.
2.
Womer
RB
,
West
DC
,
Krailo
MD
,
Dickman
PS
,
Pawel
BR
,
Grier
HE
, et al
Randomized controlled trial of interval-compressed chemotherapy for the treatment of localized Ewing sarcoma: a report from the Children's Oncology Group
.
J Clin Oncol
2012
;
30
:
4148
54
.
3.
Skapek
SX
,
Ferrari
A
,
Gupta
AA
,
Lupo
PJ
,
Butler
E
,
Shipley
J
, et al
Rhabdomyosarcoma
.
Nat Rev Dis Primers
2019
;
5
:
1
.
4.
Spunt
SL
,
Million
L
,
Chi
YY
,
Anderson
J
,
Tian
J
,
Hibbitts
E
, et al
A risk-based treatment strategy for non-rhabdomyosarcoma soft-tissue sarcomas in patients younger than 30 years (ARST0332): a Children's Oncology Group prospective study
.
Lancet Oncol
2020
;
21
:
145
61
.
5.
Kager
L
,
Tamamyan
G
,
Bielack
S
. 
Novel insights and therapeutic interventions for pediatric osteosarcoma
.
Future Oncol
2017
;
13
:
357
68
.
6.
Stewart
E
,
Goshorn
R
,
Bradley
C
,
Griffiths
LM
,
Benavente
C
,
Twarog
NR
, et al
Targeting the DNA repair pathway in Ewing sarcoma
.
Cell Rep
2014
;
9
:
829
41
.
7.
Federico
SM
,
Pappo
AS
,
Sahr
N
,
Sykes
A
,
Campagne
O
,
Stewart
CF
, et al
A phase I trial of talazoparib and irinotecan with and without temozolomide in children and young adults with recurrent or refractory solid malignancies
.
Eur J Cancer
2020
;
137
:
204
13
.
8.
Brenner
JC
,
Feng
FY
,
Han
S
,
Patel
S
,
Goyal
SV
,
Bou-Maroun
LM
, et al
PARP-1 inhibition as a targeted strategy to treat Ewing's sarcoma
.
Cancer Res
2012
;
72
:
1608
13
.
9.
Murai
J
,
Huang
SY
,
Das
BB
,
Renaud
A
,
Zhang
Y
,
Doroshow
JH
, et al
Trapping of PARP1 and PARP2 by clinical PARP inhibitors
.
Cancer Res
2012
;
72
:
5588
99
.
10.
Tang
SW
,
Bilke
S
,
Cao
L
,
Murai
J
,
Sousa
FG
,
Yamade
M
, et al
SLFN11 is a transcriptional target of EWS-FLI1 and a determinant of drug response in Ewing sarcoma
.
Clin Cancer Res
2015
;
21
:
4184
93
.
11.
Gorthi
A
,
Romero
JC
,
Loranc
E
,
Cao
L
,
Lawrence
LA
,
Goodale
E
, et al
EWS-FLI1 increases transcription to cause R-loops and block BRCA1 repair in Ewing sarcoma
.
Nature
2018
;
555
:
387
91
.
12.
Choy
E
,
Butrynski
JE
,
Harmon
DC
,
Morgan
JA
,
George
S
,
Wagner
AJ
, et al
Phase II study of olaparib in patients with refractory Ewing sarcoma following failure of standard chemotherapy
.
BMC Cancer
2014
;
14
:
813
.
13.
Ballestrero
A
,
Bedognetti
D
,
Ferraioli
D
,
Franceschelli
P
,
Labidi-Galy
SI
,
Leo
E
, et al
Report on the first SLFN11 monothematic workshop: from function to role as a biomarker in cancer
.
J Transl Med
2017
;
15
:
199
.
14.
Mu
Y
,
Lou
J
,
Srivastava
M
,
Zhao
B
,
Feng
XH
,
Liu
T
, et al
SLFN11 inhibits checkpoint maintenance and homologous recombination repair
.
EMBO Rep
2016
;
17
:
94
109
.
15.
Murai
J
,
Tang
SW
,
Leo
E
,
Baechler
SA
,
Redon
CE
,
Zhang
H
, et al
SLFN11 blocks stressed replication forks independently of ATR
.
Mol Cell
2018
;
69
:
371
84
.
16.
Thomas
A
,
Pommier
Y
. 
Targeting topoisomerase I in the era of precision medicine
.
Clin Cancer Res
2019
;
25
:
6581
9
.
17.
Murai
J
,
Zhang
Y
,
Morris
J
,
Ji
J
,
Takeda
S
,
Doroshow
JH
, et al
Rationale for poly(ADP-ribose) polymerase (PARP) inhibitors in combination therapy with camptothecins or temozolomide based on PARP trapping versus catalytic inhibition
.
J Pharmacol Exp Ther
2014
;
349
:
408
16
.
18.
Zoppoli
G
,
Regairaz
M
,
Leo
E
,
Reinhold
WC
,
Varma
S
,
Ballestrero
A
, et al
Putative DNA/RNA helicase Schlafen-11 (SLFN11) sensitizes cancer cells to DNA-damaging agents
.
Proc Natl Acad Sci U S A
2012
;
109
:
15030
5
.
19.
Coussy
F
,
El-Botty
R
,
Chateau-Joubert
S
,
Dahmani
A
,
Montaudon
E
,
Leboucher
S
, et al
BRCAness, SLFN11, and RB1 loss predict response to topoisomerase I inhibitors in triple-negative breast cancers
.
Sci Transl Med
2020
;
12
:
eaax2625
.
20.
Conteduca
V
,
Ku
SY
,
Puca
L
,
Slade
M
,
Fernandez
L
,
Hess
J
, et al
SLFN11 expression in advanced prostate cancer and response to platinum-based chemotherapy
.
Mol Cancer Ther
2020
;
19
:
1157
64
.
21.
Pietanza
MC
,
Waqar
SN
,
Krug
LM
,
Dowlati
A
,
Hann
CL
,
Chiappori
A
, et al
Randomized, double-blind, phase II study of temozolomide in combination with either veliparib or placebo in patients with relapsed-sensitive or refractory small-cell lung cancer
.
J Clin Oncol
2018
;
36
:
2386
94
.
22.
Gardner
EE
,
Lok
BH
,
Schneeberger
VE
,
Desmeules
P
,
Miles
LA
,
Arnold
PK
, et al
Chemosensitive relapse in small cell lung cancer proceeds through an EZH2-SLFN11 axis
.
Cancer Cell
2017
;
31
:
286
99
.
23.
r-project.org [homepage on the internet]
.
Vienna, Austria
:
The R Project for Statistical Computing
; 
2012
.
Available from
: http://www.R-project.org/.
24.
Carrillo
J
,
Garcia-Aragoncillo
E
,
Azorin
D
,
Agra
N
,
Sastre
A
,
Gonzalez-Mediero
I
, et al
Cholecystokinin down-regulation by RNA interference impairs Ewing tumor growth
.
Clin Cancer Res
2007
;
13
:
2429
40
.
25.
Ritz
C
,
Streibig
JC
. 
Bioassay analysis using R
.
J Stat Softw
2005
;
12
:
22
.
26.
Stewart
E
,
Federico
S
,
Karlstrom
A
,
Shelat
A
,
Sablauer
A
,
Pappo
A
, et al
The Childhood Solid Tumor Network: a new resource for the developmental biology and oncology research communities
.
Dev Biol
2016
;
411
:
287
93
.
27.
Garnett
MJ
,
Edelman
EJ
,
Heidorn
SJ
,
Greenman
CD
,
Dastur
A
,
Lau
KW
, et al
Systematic identification of genomic markers of drug sensitivity in cancer cells
.
Nature
2012
;
483
:
570
5
.
28.
Maede
Y
,
Shimizu
H
,
Fukushima
T
,
Kogame
T
,
Nakamura
T
,
Miki
T
, et al
Differential and common DNA repair pathways for topoisomerase I- and II-targeted drugs in a genetic DT40 repair cell screen panel
.
Mol Cancer Ther
2014
;
13
:
214
20
.
29.
Bryant
HE
,
Schultz
N
,
Thomas
HD
,
Parker
KM
,
Flower
D
,
Lopez
E
, et al
Specific killing of BRCA2-deficient tumours with inhibitors of poly(ADP-ribose) polymerase
.
Nature
2005
;
434
:
913
7
.
30.
Keung
MYT
,
Wu
Y
,
Vadgama
JV
. 
PARP inhibitors as a therapeutic agent for homologous recombination deficiency in breast cancers
.
J Clin Med
2019
;
8
:
435
.
31.
Davies
H
,
Glodzik
D
,
Morganella
S
,
Yates
LR
,
Staaf
J
,
Zou
X
, et al
HRDetect is a predictor of BRCA1 and BRCA2 deficiency based on mutational signatures
.
Nat Med
2017
;
23
:
517
25
.
32.
Alexandrov
LB
,
Nik-Zainal
S
,
Wedge
DC
,
Aparicio
SA
,
Behjati
S
,
Biankin
AV
, et al
Signatures of mutational processes in human cancer
.
Nature
2013
;
500
:
415
21
.
33.
Kovac
M
,
Blattmann
C
,
Ribi
S
,
Smida
J
,
Mueller
NS
,
Engert
F
, et al
Exome sequencing of osteosarcoma reveals mutation signatures reminiscent of BRCA deficiency
.
Nat Commun
2015
;
6
:
8940
.
34.
Tirode
F
,
Surdez
D
,
Ma
X
,
Parker
M
,
Le Deley
MC
,
Bahrami
A
, et al
Genomic landscape of Ewing sarcoma defines an aggressive subtype with co-association of STAG2 and TP53 mutations
.
Cancer Discov
2014
;
4
:
1342
53
.
35.
Crompton
BD
,
Stewart
C
,
Taylor-Weiner
A
,
Alexe
G
,
Kurek
KC
,
Calicchio
ML
, et al
The genomic landscape of pediatric Ewing sarcoma
.
Cancer Discov
2014
;
4
:
1326
41
.
36.
Murai
J
,
Thomas
A
,
Miettinen
M
,
Pommier
Y
. 
Schlafen 11 (SLFN11), a restriction factor for replicative stress induced by DNA-targeting anti-cancer therapies
.
Pharmacol Ther
2019
;
201
:
94
102
.
37.
Zhang
J
,
Dai
Q
,
Park
D
,
Deng
X
. 
Targeting DNA replication stress for cancer therapy
.
Genes
2016
;
7
:
51
.
38.
Subramanian
A
,
Tamayo
P
,
Mootha
VK
,
Mukherjee
S
,
Ebert
BL
,
Gillette
MA
, et al
Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles
.
Proc Natl Acad Sci U S A
2005
;
102
:
15545
50
.
39.
May
WA
,
Grigoryan
RS
,
Keshelava
N
,
Cabral
DJ
,
Christensen
LL
,
Jenabi
J
, et al
Characterization and drug resistance patterns of Ewing's sarcoma family tumor cell lines
.
PLoS One
2013
;
8
:
e80060
.
40.
Heisey
DAR
,
Lochmann
TL
,
Floros
KV
,
Coon
CM
,
Powell
KM
,
Jacob
S
, et al
The Ewing family of tumors relies on BCL-2 and BCL-XL to escape PARP inhibitor toxicity
.
Clin Cancer Res
2019
;
25
:
1664
75
.