Enhancing transcription–replication conflict targets ecDNA-positive cancers

Antibodies and reagents

Antibodies

Antibodies were procured from the following: H3K36me3 (Abcam, catalogue no. ab9050), γH2AX (Millipore, catalogue no. 05-636 for IF), γH2AX (Cell Signaling Technology, catalogue no. CST9718 for western blot), pRPA2S33 (Novus Biological, catalogue no. NB100-544), pCHK1S345 (Invitrogen, catalogue no. PA5-34625), 53BP1 (Novus Biological, catalogue no. NB100-304), cyclin A (BD Biosciences, catalogue no. 611268), pRNAPII S2/S4 (Abcam, catalogue no. ab252855), pCHK1-S345 (Cell Signaling Technology, catalogue no. CST2348), CHK1 (Abcam, catalogue no. ab32531), pRPA32/RPA2-Ser8 (Cell Signaling Technology, catalogue no. 54762 S), Vinculin (Cell Signaling Technology, catalogue no. CST13901), pFGFR2-Tyr653/654 (Cell Signaling Technology, catalogue no. CST3476S) and FGFR2 (Cell Signaling Technology, catalogue no. CST11835S).

Chemicals

Chemicals were procured from the following: CHIR-124 (Selleckchem, catalogue no. S2683), XL413 (Selleckchem, catalogue no. S7547) and triptolide (Millipore, catalogue no. 645900-5MG).

Cell culture

GBM39ec, GBM39HSR and HK296 were patient-derived neurosphere cell lines and were established as previously described^2,7. The parental PC3 line was obtained from ATCC. PC3 DM and PC3 HSR lines were isolated by the Mischel Lab through single-cell expansions of the parental PC3 line and are available from the Mischel Lab upon request. All the other cell lines were purchased from ATCC. Human prostate cancer cell line PC3 DM, PC3 HSR; colorectal cancer cell line COLO320DM, COLO320HSR; gastric cancer cell line SNU16; lung cancer cell line PC9 and hTERT-immortalized retinal pigment epithelial cell line RPE1 were cultured in 4.5 g l^−1 glucose-formulated Dulbecco’s Modified Eagle’s Medium (Corning) supplemented with 10% fetal bovine serum (FBS; Gibco). For GRO-seq and ChIP–seq, COLO320DM and COLO320HSR were grown in Roswell Park Memorial Institute 1640 with GlutaMAX (Gibco) with 10% FBS. GBM39ec, GBM39HSR and HK296 cell lines were cultured in Dulbecco’s Modified Eagle’s Medium/F12 (Gibco, catalogue no. 11320-033) supplemented with 1× B27 (Gibco, catalogue no. 17504-01), 20 ng ml^−1 epidermal growth factor (Sigma, catalogue no. E9644), 20 ng ml^−1 fibroblast growth factor (Peprotech, catalogue no. AF-100-18B), 1–5 µg ml^−1 heparin (Sigma, catalogue no. H3149) and 1× GlutaMAX (Gibco, catalogue no. 35050-061). GBM39 cells used in sequencing assays were cultured without additional GlutaMAX. All the cells were maintained at 37 °C in a humidified incubator with 5% CO₂. Cell lines routinely tested negative for mycoplasma contamination.

GRO-seq

COLO320DM and COLO320HSR RNA was prepared by washing cells with ice-cold phosphate-buffered saline (PBS), then adding ice-cold LB (10 mM Tris-HCl pH 7.4, 2 mM MgCl2, 3 mM CaCl2, 0.5% IGEPAL-CA630, 10% glycerol, 1 mM DTT, protease inhibitors (Roche, catalogue no. 11836170001), RNase inhibitor (Ambion, catalogue no. AM2696)) and scraping cells into a 15 ml conical tube. Cells were spun at 1,000g for 10 min at 4 °C. Supernatant was removed and pellet was thoroughly resuspended in 1 ml LB using a wide bore tip. An additional 9 ml LB was added and then cells were spun at 1,000g for 10 min at 4 °C. Cells were resuspended in LB and spun down. Pellets were resuspended in ice-cold freezing buffer (50 mM Tris-HCl pH 8.3, 5 mM MgCl₂, 40% glycerol, 0.1 mM EDTA, 0.2 μl RNase inhibitor per ml of freezing buffer) and spun at 2,000g for 2 min at 4 °C. Nuclei were resuspended in 100 μl freezing buffer per 5 million cells. A nuclear run-on master mixed was prepared (10 mM Tris-HCl pH 8.0, 5 mM MgCl2, 1 mM DTT, 300 mM KCl, 0.5 mM ATP, 0.5 mM GTP, 0.003 mM CTP (unlabelled ribonucleotide triphosphates from Roche, catalogue no. 11277057001), 0.5 mM Bromo-UTP (Sigma, catalogue no. B7166), 1% Na-laurylsarcosine, 1 μl RNase inhibitor per 100 μl) and preheated to 30 °C. An equal volume of master mix was added to aliquoted nuclei (5 million nuclei per replicate) and incubated at 30 °C for 5 min with gentle shaking. DNase digestion was performed using RQ1 DNase I and RQ1 buffer (Promega, catalogue no. M610A) for 30 min at 37 °C; the reaction was stopped with the addition of stop buffer to a final concentration of 10 mM Tris-HCl pH 7.4, 1% sodium dodecyl sulfate (SDS), 5 mM EDTA, 1 mg ml^−1 proteinase K. Samples were incubated for 1 h at 55 °C. NaCl was added to final concentration of 225 mM. Two phenolchloroform extractions were done, followed by one extraction with chloroform. RNA was precipitated in 75% EtOH with 1 μl glycoblue (Ambion, catalogue no. 9516) overnight at −20 °C.

For GBM39ec and GBMHSR, cells were washed with ice-cold PBS and then spun for 5 min at 500g at 4 °C. Cells were then resuspended in ice-cold 10 ml swelling buffer (10 mM Tris-HCl pH 7.5, 2 mM MgCl2, 3 mM CaCl2, protease inhibitor, RNase inhibitor) and incubated on ice for 5 min. Cell were spun at 400g for 10 min at 4 °C and resuspended in 10 ml ice-cold glycerol swelling buffer (0.9× swelling buffer, 10% glycerol). While agitating the tube, 10 ml ice-cold lysis buffer (glycerol swelling buffer, 1% IGEPAL-CA630) was slowly added. Samples were incubated on ice for 5 min, then another 25 ml lysis buffer was added and samples were spun for 5 min at 600g at 4 °C. Samples were resuspended in ice-cold freezing buffer (50 mM Tris-HCl pH 8.0, 5 mM MgCl2, 40% glycerol, 0.1 mM EDTA, RNase inhibitor) and spun at 900g for 6 min at 4 °C. An equal volume of pre-warmed nuclear run-on master mix was added to aliquoted nuclei (10 million nuclei per replicate) and incubated at 30 °C for 7 min with gentle shaking. Samples were then mixed thoroughly with 600 μl Trizol LS and incubated at room temperature for 5 min. Next, 160 μl chloroform was added to each sample, shaken vigorously, then incubated at room temperature for 3 min and centrifuged at 12,000g at 4 °C for 30 min. NaCl was added to the aqueous phase to a final concentration of 300 mM and RNA was precipitated in 75% EtOH with 1 μl glycoblue overnight at −20 °C.

For all cell types, after overnight RNA precipitation, RNA was spun for 20 min at 21,130g at 4 °C. RNA pellets were washed in fresh 75% EtOH, briefly air-dried and then resuspended in 20 μl water. Base hydrolysis was performed using 5 μl 1 N NaOH for 10 min and then neutralized with 25 μl 1 M Tris-HCl pH 6.8. Buffer exchange was performed using P30 Micro columns (Bio-Rad, catalogue no. 7326250), then treated with RQ1 DNase I and RQ1 buffer and incubated at 37 °C (10 min for COLO320 and 30 min for GBM39). Buffer exchange was performed again. Samples were treated with 3 μl T4 polynucleotide kinase (PNK; New England Biolabs, catalogue no. M0201), 1× PNK buffer, 2 μl 10 mM ATP and 2 μl RNase inhibitor and incubated for 1 h at 37 °C. Another 2 μl PNK was added per sample and incubation was continued for 30–60 min. RNA decapping was performed by adding ammonium chloride (final concentration 50 mM), poloaxamer 188 (final concentration 0.1%), 2 μl messenger RNA decapping enzyme (New England Biolabs, catalogue no. M0608S) and 1 μl RNase inhibitor and incubated at 37 °C for 30 min. EDTA was then added to the final concentration of 25 mM and samples were incubated at 75 °C for 5 min. Samples were then incubated on ice for at least 2 min. Sample volume was then brought to 100 μl with binding buffer (0.25× SSPE, 1 mM EDTA, 0.05% Tween 20, 37.5 mM NaCl, RNase inhibitor). During T4 PNK treatment, 60 μl anti-BrdU agarose beads (Santa Cruz Biotechnology, catalogue no. sc-32323ac) per sample were equilibrated in 500 μl binding buffer by rotating for 5 min at room temperature, spun and washed again in binding buffer. Beads were then blocked in blocking buffer (1× binding buffer, 0.1% polyvinylpyrrolidone, 1 ug ml^−1 ultrapure bovine serum albumin (BSA), RNase inhibitor) by rotating for 1 h at room temperature. Beads were then washed twice in binding buffer and resuspended in 400 μl binding buffer. Decapped RNA was then added to the blocked beads and rotated for 1 h at room temperature. Beads were then washed once in binding buffer, once in low-salt buffer (0.2× SSPE, 1 mM EDTA, 0.05% Tween 20, RNase inhibitor), once in high-salt buffer (0.2× SSPE, 1 mM EDTA, 0.05% Tween 20, 137.5 mM NaCl, RNase inhibitor) with 3 min of rotation, and twice in Tris-EDTA-Tween20 buffer (10 mM Tris-HCl pH 8.0, 1 mM EDTA, 0.05% Tween 20, RNase inhibitor). All spins with agarose beads were performed for 2 min at 1000g at room temperature and all washes were performed in 500 μl buffer rotating for 5 min at room temperature unless otherwise noted. Samples were then eluted in elution buffer (50 mM Tris-HCl pH 7.5, 20 mM DTT, 1 mM EDTA, 150 mM NaCl, 0.1% SDS, RNase inhibitor) pre-warmed to 42 °C; four 10-min elutions were performed at 42 °C with periodic vortexing. The eluates for each replicate were pooled and RNA was then purified by phenolchloroform and chloroform with EtOH precipitation (COLO320) or by column purification using New England Biolabs Monarch RNA Cleanup Kit T2030 (GBM39). Sequencing libraries were prepared using the NEBNext Small RNA Library Prep Kit (New England Biolabs, catalogue no. E7330) and sequenced by Novaseq PE150. The sequence data were mapped to human reference genome (hg38) using STAR, v.2.7.10b (ref. ¹⁷). HOMER (v.4.11.1) was used for de novo transcript identification on each strand separately using the default GRO-seq setting. Reads with MAPQ values less than 10 were filtered using SAMtools (v.1.8). Duplicate reads were removed using picard-tools. GRO-seq signal was converted to the bigwig format for visualization using deepTools bamCoverage¹⁸ (v.3.3.1) with the following parameters: –binSize 10 –normalizeUsing CPM –effectiveGenomeSize 3209286105 –exactScaling.

Total RNA library preparation

Total RNA from each sample was isolated with Quick-RNA Miniprep Kit (Zymo Research, catalogue no. R1054) with input of 1–2 million cells. RNA libraries were constructed using TruSeq Stranded Total RNA Library Prep Kit with Ribo-Zero (Illumina, catalogue no. 20020596). Nextseq 550 sequencing system (Illumina) produced 20–30 million of ×2, 75 bp paired-end reads per sample. The sequence data were mapped to human reference genome hg38 using STAR, v.2.7.10b (ref. ¹⁷), following the ENCODE RNA-seq pipeline. Reads with MAPQ values less than ten were filtered using SAMtools (v.1.8). Ribo-Zero signal was converted to the bigwig format for visualization using deepTools bamCoverage¹⁸ (v.3.3.1) with the following parameters: –binSize 10 –normalizeUsing CPM –effectiveGenomeSize 3209286105 –exactScaling.

KAS-seq library preparation

KAS-seq experiments were carried out as previously described¹² with modifications¹³. Briefly, cell culture media was supplemented with 5 mM N₃-kethoxal (final concentration), and cells were incubated for 10 min at 37 °C in a six-well plate. Genomic DNA was then extracted using the Monarch gDNA Purification Kit (NEB T3010S) following the standard protocol but with elution using 50 µl 25 mM K₃BO₃ at pH 7.0. Click reaction was carried out by mixing 87.5 µl purified DNA, 2.5 µl 20 mM DBCO-PEG4-biotin (dimethylsulfoxide (DMSO) solution, Sigma, catalogue no. 760749) and 10 µl 10× PBS in a final volume of 100 µl. The reaction was then incubated at 37 °C for 90 min. DNA was purified using AMPure XP beads by adding 50 µl beads per 100 µl reaction, washing beads on a magnetic stand twice with 80% EtOH and eluting in 130 µl 25 mM K₃BO₃. Purified DNA was then sheared using a Covaris E220 instrument down to around 200–400 bp size. Pulldown of biotin-labelled DNA was initiated by separating 10 µl of 10 mg ml^−1 Dynabeads MyOne Streptavidin T1 beads (Life Technologies, catalogue no. 65602) on a magnetic stand, then washing with 180 µl of 1× Tween Washing Buffer (TWB; 5 mM Tris-HCl pH 7.5; 0.5 mM EDTA; 1 M NaCl; 0.05% Tween 20). Beads were then resuspended in 300 µl of 2× binding buffer (10 mM Tris-HCl pH 7.5, 1 mM EDTA, 2 M NaCl), sonicated DNA was added (diluted to a final volume of 300 µl if necessary) and beads were incubated for at least 15 min at room temperature on a rotator. Beads were separated on a magnetic stand and washed with 300 µl of 1× TWB and heated at 55 °C in a Thermomixer with shaking at 1,000 rpm for 2 min. The supernatant was removed on a magnetic stand and the TWB wash and 55 °C incubation were repeated.

Libraries were prepared on beads using the NEBNext Ultra II DNA Library Prep Kit (NEB, catalogue no. E7645). First, end repair was carried out by incubating beads for 30 min at 20 °C in a Thermomixer with shaking at 1,000 rpm in 50 µl 1× EB buffer plus 3 µl NEB Ultra End Repair Enzyme and 7 µl NEB Ultra End Repair Enzyme. This was followed by incubation at 65 °C for 30 min. Second, adaptors were ligated by adding 2.5 µl NEB adaptor, 1 µl ligation enhancer and 30 µl blunt ligation mix, incubating at 20 °C for 20 min, then adding 3 µl USER enzyme and incubating at 37 °C for 15 min (in a Thermomixer, with shaking at 1,000 rpm). Beads were separated on a magnetic stand and washed with 180 µl TWB for 2 min at 55 °C and 1,000 rpm in a Thermomixer. After magnetic separation, beads were washed in 100 µl 0.1× TE buffer, resuspended in 15 µl 0.1× TE buffer and heated at 98 °C for 10 min. PCR was carried out by adding 5 µl of each of the i5 and i7 NEBNext sequencing adaptors together with 25 µl 2× NEB Ultra PCR Mater Mix, with a 98 °C incubation for 30 s and 15 cycles of 98 °C for 10 s, 65 °C for 30 s and 72 °C for 1 min, followed by incubation at 72 °C for 5 min. Beads were separated on a magnetic stand and the supernatant was cleaned up using 1.8× AMPure XP beads.

Libraries were sequenced in a paired-end format on an Illumina NextSeq instrument using NextSeq 550 High-Output Kits (2 × 36 cycles). The sequence data were mapped to the hg38 assembly of the human genome using Bowtie^19,20 with the following settings: -v 2-k 2-m 1–best–strata-X 1000. Duplicate reads were removed using picard-tools (v.1.99). MACS2 (ref. ²¹) (v.2.1.1) was used for peak-calling with the following parameters: –broad -g hs –broad-cutoff 0.01 -q 0.01. Browser tracks are generated after normalizing to input using bamCompare default setting.

ChIP–seq library preparation

Three million cells per replicate were fixed in 1% formaldehyde for 15 min at room temperature with rotation and then quenched with 0.125 M glycine for 10 min at room temperature with rotation. Fixed cells were pelleted at 1,300g for 5 min at 4 °C and washed twice with cold PBS before storing at −80 °C. Membrane lysis was performed in 5 ml LB1 (50 mM HEPES pH 7.5, 140 mM NaCl, 1 mM EDTA, 10% glycerol, 0.5% IPEGAL-CA630, 0.25% Triton X-100, Roche protease inhibitors 11836170001) for 10 min at 4 °C with rotation. Nuclei were pelleted at 1,400g for 5 min at 4 °C and lysed in 5 ml LB2 (10 mM Tris-Cl pH 8.0, 200 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, Roche protease inhibitors) for 10 min at room temperature with rotation. Chromatin was pelleted at 1,400g for 5 min at 4 °C and resuspended in 1 ml of TE buffer plus 0.1% SDS before sonication on a Covaris E220 with the following settings: 140 W, 10% duly, 200 cycles per burst, 600 s per sample. Samples were clarified by spinning at 16,000g for 10 min at 4 °C. Supernatant was transferred to a new tube and diluted with two volumes of IP dilution buffer (10 mM Tris pH 8.0, 1 mM EDTA, 200 mM NaCl, 1 mM EGTA. 0.2% Na-DOC, 1% Na-laurylsarcosine, 2% Triton X-100). Then, 50 µl of sheared chromatin was reserved as input and ChIP was performed overnight at 4 °C with rotation with 7.5 µg of H3K36me3 antibody (ab9050) (1:300 dilution). Per sample, 100 μl protein A dynabeads were washed three times with 1 ml chilled block buffer (0.5% BSA in PBS) and then added to the chromatin after overnight incubation with antibody and rotated for 4 h at 4 °C. Samples were washed five times in 1 ml pre-chilled wash buffer (50 mM HEPES pH 7.5, 500 mM LiCl, 1 mM EDTA, 1% IPEGAL-CA630, 0.7% Na-DOC) and then 1 ml pre-chilled TE + 50 mM NaCl. Samples were eluted in elution buffer (50 mM Tris pH 8.0, 10 mM EDTA, 1% SDS) at 65 °C. NaCl was added to a final concentration of 455 mM. Samples were incubated with 0.2 mg ml^−1 proteinase K for 1 h at 55 °C and then decross-linked overnight at 65 °C. Samples were treated with 0.2 mg ml^−1 RNAase for 2 h at 37 °C and then purified with the Zymo ChIP DNA Clean & Concentrator Kit (D2505). Libraries were prepared using the NEBNext Ultra II DNA Library Prep Kit (E7645) and sequenced by NovaSeq PE150. The sequence data were trimmed by Trimmomatic²² (v.0.36) to remove adaptor and then mapped to the hg38 assembly of the human genome using Bowtie2 (refs. ^19,20) with the following settings: –local –very-sensitive –phred33 -X 1000. Reads with MAPQ values less than ten were filtered using SAMtools (v.1.8). Duplicate reads were removed using picard-tools. CHIP–seq signal was converted to the bigwig format for visualization using deepTools bamCoverage¹⁸ (v.3.3.1) with the following parameters: –binSize 10 –normalizeUsing CPM –effectiveGenomeSize 3209286105 –exactScaling.

IF and DNA FISH staining

Coverslips were coated with 100 µg ml^−1 poly-l-lysine overnight or 10 µg ml^−1 laminin for 1 h at 37 °C before seeding cells. Asynchronized cells were seeded onto slides and subject to different treatment. Where indicated, EdU was added to each well at 10 µg ml^−1 30 min before collecting samples. IF and dual-IF DNA FISH staining were performed as described before. Briefly, slides were fixed with ice-cold 4% paraformaldehyde (PFA) for 15 min, followed by permeabilization with 0.5% Triton X-100 in PBS for 15 min at room temperature. Samples were blocked with 3% BSA in PBS for 1 h at room temperature before incubation with primary antibody diluted in blocking buffer at 4 °C overnight. Dilution ratio for first antibodies was as follows: γH2Ax, 1:500; pRPA2-S33, 1:1,000; pCHK1S345, 1:250; 53BP1, 1:500; cyclin A, 1:100; pRNAPII S2/S4, 1:1,000. After washing with PBS a total of three times for 5 min each, slides were incubated with secondary antibody diluted in blocking buffer at room temperature for 1 h. Samples were fixed with ice-cold 4% PFA for 20 min after washing with PBS. If combined with DNA FISH staining, fixed samples were further permeabilized with ice-cold 0.7% Triton X-100 per 0.1 M HCl (diluted in PBS) for 10 min on ice. DNA was denatured by 1.5 M HCl for 30 min at room temperature, followed by dehydration in ascending ethanol concentration. Diluted FISH probes (Empire Genomics) were pre-denatured at 75 °C for 3 min and added onto air-dried slides. After incubation at 37 °C overnight, slides were washed with 2× SSC to get rid of non-specific binding, followed by DAPI staining. Where indicated, EdU staining was performed with the Click-iT Plua EdU Alexa Fluor 647 Imaging Kit (Invitrogen, catalogue no. C10640).

Validation of PC3-DM and PC3-HSR cell lines

Genomic DNA was extracted from a confluent six-well dish using the QIAamp DNA Mini Kit (Qiagen) according to the manufacturer’s protocol. Briefly, single cells were collected and resuspended in 200 µl 1× PBS, followed by the addition of 20 µl QIAGEN proteinase K and 200 µl buffer AL. The mixture was pulse-vortexed for 15 s and incubated at 56 °C for 10 min. A volume of 200 µl absolute ethanol was added to the sample and pulse-vortexed for 15 s. The entire mixture was pipetted into a QIAamp Mini spin column and centrifuged at 6,000g for 1 min. Filtrate was discarded and 500 µl buffer AW1 was added to the column. After centrifugation at 6,000g for 1 min, the column was subjected to another round of wash with 500 µl buffer AW2. The filtrate was discarded after centrifugation at full speed for 3 min. The column was then placed in a clean 1.5 ml microfuge tube and 50 µl of buffer AE was added to reconstitute genomic DNA after centrifugation at 6,000g for 1 min.

WGS library preparation was performed with the FS DNA Library Prep Kit from NEB according to the manufacturer’s protocol, with these parameters in place: (1) 250 ng gDNA was used as input; (2) fragmentation was done with an incubation time of 18 min to yield 200–450 bp fragments; (3) the final library size distribution was between 320–470 bp (that is, first bead selection was done with a bead volume of 30 µl and second bead selection was done with a bead volume of 15 µl); (4) the final PCR amplification was performed for four cycles. PE150 sequencing was performed on NovaSeq to yield at least 10× coverage at Novogene. Adaptor sequences were removed from raw fastq files using Trim Galore at default settings, followed by alignment to the hg38 reference genome using Map with BWA-MEM to generate the BAM files. BAM files were then uploaded to the GenePattern Notebook for AmpliconArchitect analysis under default settings.

ecDNA structure analysis

We utilized the AmpliconSuite-pipeline (v.1.2.2, https://github.com/AmpliconSuite/AmpliconSuite-pipeline), which invoked CNVKit (v.0.9.9)⁴³, AmpliconArchitect⁴⁴ (AA; v.1.3.r8) and AmpliconClassifier³ (AC; v.1.1.2). In brief, the analysis pipeline first identifies seed regions of focal amplification from whole-genome copy number calls, then among the seed regions AA analyses copy number and structural variation jointly to construct a local genome graph encoding structural rearrangements and copy numbers. AA then extracts genome paths and cycles from the genome graph that explain the observed changes in copy number and structural variation. The outputs of AA are passed to AC, which applies a rule-based method to match the patterns of copy number, structural variation and structures extracted from the genome graph to known types of focal amplifications, such as ecDNA. To minimize sequencing artefacts derived from insert size distribution variance, we set the AmpliconSuite-pipeline argument –AA_insert_sdevs 9. For PC3 samples, –downsample 1 was also set to reduce additional sequencing artefacts. Default parameters were used otherwise.

For COLO320DM/HSR, we utilized the general ecDNA regions and the candidate ecDNA structure from ref. ¹⁰, after lifting over coordinates to hg38. For GBM39ec/HSR and PC3-DM/HSR, ecDNA regions were derived from AA output files. From the DM samples, regions with copy number greater than ten in the AA amplicon containing the ecDNA of interest were defined as the ecDNA regions. In GBM39ec/HSR, we also included the vIII deletion in the ecDNA region. Candidate ecDNA structures were derived from the AA cycle with highest assigned copy count containing the oncogene of interest (GBM39: EGFR and PC3: MYC). For GBM39, the ecDNA structure was consistent with a previously published reconstruction¹¹. Circular ecDNA visualizations were generated with CycleViz (https://github.com/AmpliconSuite/CycleViz). Gene and focal amplification copy numbers were derived from the AA graph file and the AC feature basic properties file, respectively. Structural similarity scores of the focal amplifications were computed using the feature_similarity.py script in AC, which computes a similarity score based on the overlapping genomic boundaries and shared structural variants between two focal amplifications. For the PC3 samples, we utilized the related amplicon_similarity.py script to obtain similarity scores, as the exact boundaries of the ecDNA could not be easily resolved with AC.

Replication combing assay

Replication fork speed in ecDNA was evaluated using the molecular combing assay. COLO320DM and COLO320HSR cells were seeded into plates and allowed to grow into log phase, nascent DNA synthesize was pulse labelled with thymidine analogues: CldU and IdU sequentially for equal amount of time. Following pulse labelling, cells were harvested and embedded into agarose plugs using the Genomic Vision FiberPrep Kit (Genomic Vision). DNA extraction, combing and immunostaining was performed according to the EasyComb service procedures (Genomic Vision). Coverslips were scanned with a FiberVision scanner and images were analysed using FiberStudio software (Genomic Vision). Fork speed was calculated using replication signals with contiguous CldU–IdU tracks. Only intact signals, flanked by counterstaining of the DNA fibre, were selected for analysis.

Locus-replication combing assay

DNA replication activity at the MYC loci was assessed using molecular combing assay. COLO320DM and COLO320HSR cells were seeded into plates and allowed to grow into log phase, nascent DNA synthesize was pulse labelled with thymidine analogues: CldU and IdU for equal amount of time. Following pulse labelling, cells were harvested and embedded into agarose plugs using the Genomic Vision FiberPrep kit (Genomic Vision). DNA extraction and combing was performed according to the EasyComb service procedures (Genomic Vision). DNA-labelled FiberProbes (Genomic Vision) targeting MYC loci were produced and hybridized to combed DNA. Correspondence between theoretical and experimental probe coverage patterns was validated by measuring hybridized probe length in control samples. After immunostaining of replication signals and DNA probes, coverslips were scanned with a FiberVision scanner. Image analysis and measurements were performed using FiberStudio software (Genomic Vision). Fork speed was calculated using replication signals with contiguous CldU–IdU tracks.

Comet-FISH

Alkaline comet-FISH assays were performed according to the literature, with minor modifications^45,46. Cells were harvested by trypsinization, washed with PBS and placed on ice. Cells were diluted in 37 °C low melting point (LMP) agarose (IBI Scientific) in PBS to a final concentration of 0.7% and spread on precoated glass slides with a coverslip. Overnight lysis was performed at 4 °C in alkaline lysis solution (2.5 M NaCl, 100 mM EDTA, 10 mM Tris pH 10, 1% Triton X-100, 10% DMSO) protected from light. The following day, slides were equilibrated for 30 min in alkaline electrophoresis buffer (200 mM NaOH, 1 mM EDTA, pH less than 13) in a Coplin jar and subsequently electrophoresed at 25 V for 30 min. Slides were then neutralized with Tris, dehydrated in 70% ethanol and dried at room temperature.

To detect ecDNA through FISH, Cy5-labelled probes were generated from RP11-440N18 BAC DNA sonicated to 150 bp and labelled using a DNA labelling kit (Mirus Bio). Slides were denatured with 0.5 M NaOH for 30 min at room temperature, dehydrated in an ethanol series (70%, 85%, 95%) and allowed to dry at room temperature. The hybridization mixture containing probe DNA (200 ng per slide) and Cot-1 DNA (8 μg per slide) was denatured separately at 75 °C for 10 min and pre-annealed at 37 °C for 1 h. Probe was added to the slides and spread with a glass coverslip and incubated at 37 °C overnight in a humidified chamber. The following day, slides were washed four times in 2× SSC, 50% formamide at 42 °C and subsequently washed twice in 2× SSC at 42 °C. Slides were dipped briefly in 70% ethanol and air-dried. Slides were mounted with Everbrite (Biotium) containing SYBR Gold (Invitrogen) diluted 1:10,000 and sealed with nail polish. Images were collected on a Nikon Eclipse TE2000-E using a ×60 oil objective.

Cell viability assay

Cell viability assay was performed using CellTiter-Glo (Promega, catalgoue no. G8461) as previously described⁴⁷. Briefly, cells were seeded into a 384-well plate one day before adding inhibitors. Equal volumes of vehicles or drugs diluted at indicated concentration were added into each well the next day, and the cells were incubated for three days. On the third day, after equilibrating plate and CellTiter-Glo reagent at room temperature for 30 min, reagent was added into each well and incubated for 15 min at room temperature. Luminescence was measured using a Synergy 2 microplate reader. Four biological replicates were performed for each condition. Data analysis was performed with GraphPad Prism (v.9.1.0).

TUNEL

TUNEL assay (Invitrogen, catalogue no. C10617) was performed to detect DNA fragmentation during apoptosis. COLO320DM, COLO320HSR and SNU16 cells were treated with 1 µM CHIR-124 for indicated times. All cells including floating cells were collected and spun down onto slides using a cytospin (Thermo Scientific Cytospin 4 Centrifuge). Slides were fixed with 4% PFA and permeabilized with 0.25% Triton X-100, followed by labelling of free double strand end with EdUTP by reaction catalysed by TdT enzyme in a humidified chamber at 37 °C for 60 min. Incorporated EdUTP was detected through Click-iT Plus TUNEL reaction according to the manufacturer’s manual at 37 °C for 30 min. Slides were counterstained with DAPI and mounted with ProLong Diamond Antifade.

Annexin V staining

Cell apoptosis was detected through flow cytometry using a FITC Annexin V Apoptosis Detection Kit (BD Biosciences, catalogue no. 556547). Cells were treated with 1 µM of CHIR-124 for the indicated time, and all the cells including floating cells were collected. After washing with PBS twice and cell number counting, cells were resuspended in 1× binding buffer provided by the kit at a concentration of 1 × 10^6 cells per millilitre. One hundred microlitres of the cell suspension was transferred to a FACS tube and stained with FITC Annexin V and propidium iodide. After incubation at room temperature for 15 min, all the samples were analysed with BD LSR II flow cytometry (BD Biosciences) within 1 h. Flow cytometry data were analysed through Beckman Coulter Kaluza software (v.2.1).

Microscope and image analysis

Images were taken by conventional fluorescence microscopy or confocal microscopy. Conventional fluorescence microscopy was performed on a Leica DMi8 widefield microscope by Las X software (v.3.8.2.27713) using a ×63 oil objective. Confocal microscopy was performed on a ZEISS LSM 880 inverted confocal microscope using ZEN (black v.2.3) (Stanford CSIF Facility). Z-stacks were taken for each field of view and a best-in-focus stack was identified for downstream image analysis, except for Fig. 3a, where a max projection was performed by ImageJ (v.1.53t).

Image analysis and quantification were performed using the open-source software CellProfiler (v.4.2.1). For foci number analysis, DAPI staining, IF staining and DNA FISH channel were analysed through automatic thresholding and segmentation to cell nuclei, pRPA2S33/γH2AX foci and DNA FISH foci respectively. Colocalization was performed using an object-based colocalization method. For fluorescence intensity measurement, nuclei were called based on DAPI channel through automatic thresholding and segmentation; mean fluorescence intensity was retrieved by measuring mean fluorescence intensity within each nucleus.

RS score computation

RS score 1

The gene set variation analysis⁴⁸ was utilized to assess the enrichment of the DNA RS response (RSR) gene set²⁰ in TCGA samples using RNA-seq data⁴⁹. The RSR gene set was curated based on genes affected by defects in the DNA RS response. RNA-seq transcripts per kilobase million values for TCGA samples were retrieved from the GDC data portal⁴⁹. Gene set variation analysis generated enrichment scores for both up- and down-regulated RSR genes. The final RSR score was determined as the difference between the up and down enrichment scores.

RS score 2

The RS signature score of each sample from TCGA was retrieved from the literature from ref. ²¹, which was transformed linearly between zero and one by subtracting the minimum score and dividing by the maximum score. TCGA sample ecDNA status classification was performed as stated in a previous publication¹.

Both methods

Briefly, 1,921 TCGA samples were grouped into five subtypes by AC (https://github.com/AmpliconSuite): ecDNA, breakage–fusion–bridge, complex non-cyclic, linear and no amplification. Samples with a break–fusion–bridge or complex non-cyclic status were removed from the analysis due to the challenges of detecting ecDNA from short-read data. Samples with linear amplification and no amplification were classified as ecDNA^−. After removing metastasis sample and ecDNA^− samples without matching ecDNA⁺ samples of the same tissue origin, a total of 232 ecDNA⁺ and 582 ecDNA^− samples were included in the analysis.

CRISPR experiment

sgRNA template oligos targeting the gene encoding CHK1 was synthesized (Integrated DNA Technologies) and was ligated into a CRISPR expression vector with red fluorescent protein (RFP) (Cellecta-pRSG16-U6-sg-HTS6C-UbiC-TagRFP-2A-Puro). Non-targeting green fluorescent protein (GFP) (sgNT-GFP) plasmid was purchased.

ecDNA⁺ and ecDNA^− Hela cells were transduced with sgCHK1-RFP or sgNT-GFP virus, and puromycin (Sigma) was added at 2.5 µg ml^−1 for selection for 48 h. After 48 h of puromycin selection (day 0), an equal number of cells expressing either sgCHK1-RFP or sgNT-GFP were mixed to obtain the RFP to GFP population ratio. In the following days, flow cytometry analysis was performed to determine the sgCHK1-RFP to sgNT-GFP ratio. The mixed cell population cultures were maintained at subconfluency. The sgRNA sequences targeting CHK1 were as follows:

No. 17: CCTGACAGCTGTCACTGGGT

No. 18: GCTGTCAGGAGTATTCTGAC

Western blotting

Samples were lysed in radioimmunoprecipitation assay buffer (Boston BioProducts, catalogue no. BP-115) supplemented with protease/phosphatase inhibitors (Fisher Scientific, catalogue no. 78444). Protein concentration was quantified with bicinchoninic acid assay (Fisher Scientific, catalogue no. 23225) and samples were prepared in 4× sample buffer (Bio-Rad, catalogue no. 1610747). Samples were loaded and run on 4–12% Bis-Tris Gradient Gel (Fisher Scientific, catalogue no. WG1403BOX) and transferred onto a nitrocellulose membrane (Bio-Rad, catalogue no. 1704271). The membrane was blocked with 5% BSA in Tris-buffered saline with Tween (Fisher Scientific, catalogue no. 28360) for an hour, and then primary antibody (1:1,000 dilution) was added and incubated overnight at 4 °C. Following primary antibody incubation, the membrane was washed with Tris-buffered saline with Tween and incubated with secondary antibody for 1 h. The membrane was then incubated with enhanced chemiluminescence reagent (Fisher Scientific, catalogue no. 32106) and image acquisition was performed on ProteinSimple FluorChemE.

Detection of phosphorylated CHK1 Ser345 using the AlphaLisa SureFire assay

Compound activity in cells was measured using an AlphaLISA SureFire Ultra p-CHK1 (Ser345) assay (Perkin Elmer, catalogue no. ALSU-PCHK1-A10K). HT29 cells were cultured in McCoy 5 A medium with 10% FBS and 1% penicillin-streptomycin and seeded to 96-well plates (Corning, catalogue no. 3599). Compounds were serially diluted in DMSO over a 10-point dose range with 3-fold dilution, and compound solution was added to each well containing cells. Plates were centrifuged at 1,000 rpm for 30 s. Plates were incubated at 37 °C for 16 h. Supernatant was removed by flicking the plate against a paper towel. Wells were washed once with PBS solution. To each well was added freshly prepared lysis buffer and plates were agitated on a plate shaker at 400 rpm for 30 min. The 96-well cell plates were centrifuged at 1,500 rpm for 1 min. From each well was transferred 10 µl of the lysates to a 384-well Optiplate (Perkin Elmer, catalogue no. 6007290). To each well was added Acceptor Mix (5 µl) and the plates were sealed and wrapped in foil. Plates were agitated on a plate shaker for 2 min, then incubated at room temperature for 1 h. To each well was added Donor Mix (5 µl) and the plates were sealed and wrapped in foil. Plates were agitated on a plate shaker for 2 min, then incubated at room temperature for 1 h. AlphaLisa signal was read on an EnVision multimode plate reader (Perkin Elmer). Data were fitted to dose–response curves using XLfit (IDBS) or GraphPad Prism (GraphPad software) to calculate IC₅₀ values for each compound tested.

Kinase HTRF biochemical assay

CHK1 enzyme activity was measured using a homogeneous time resolved fluorescence (HTRF) KinEASE assay (Cisbio, catalogue no. 62ST1PEC). Full-length human CHK1 protein (GenBank accession number NP_001265.1) was obtained from Carna Biosciences, Inc. (catalogue no. 02-117). The enzyme reaction was carried out in assay buffer containing (final concentrations): CHK1 enzyme (0.012 ng µl^−1), MgCl2 (5 mM) and DTT (1 mM). To determine compound dose response, DMSO stock solutions were serially diluted in a ten-point concentration series in duplicate. Compound solution (50 nl) was added to 384-well assay plates (Greiner, catalogue no. 784075). To each well containing compound solution was added assay buffer solution (5 µl). Plates were centrifuged at 1,000 rpm for 1 min, then incubated at room temperature for 10 min. The reaction was started by addition of substrate buffer (5 µl per well) containing (final concentrations): STK substrate 1-biotin (120 nM) and ATP (1 mM). Assay plates were centrifuged at 1,000 rpm for 1 min, then incubated at room temperature for 60 min. The reaction was stopped by the addition of detection buffer (Cisbio, 10 µl) containing (final concentrations): STK antibody cryptate (0.25 nM) and streptavidin-XL665 (7.5 nM). Plates were centrifuged at 1,000 rpm for 1 min, then incubated at 25 °C for 2 h. HTRF signal was read on an EnVision multimode plate reader (Cisbio) in HTRF mode. Data were fit to dose–response curves using XLfit (IDBS) or Prism (GraphPad Software) to calculate IC₅₀ values for each compound tested.

Phospho-RPA32 S8 IF high content imaging

Optical-bottom 96-well plates (Thermo Scientific, catalogue no. 165305) were coated with 50 µl of 1:1 poly-l-lysine (R&D Systems, catalogue no. 3438-100-01) and poly-d-lysine (R&D Systems, catalogue no. 3439-100-01) for 3 h at room temperature. The wells were washed once with 100 µl of PBS (Gibco, catalogue no. 10010-023) and all liquid was removed from the wells and allowed to dry fully at room temperature. COLO320 ecDNA⁺ cells were seeded at 15,000 cells per well in 100 µl of Roswell Park Memorial Institute media (Thermo Fisher, catalogue no. 22400089) supplemented with 10% FBS (Omega Scientific, catalogue no. FB-01). Cells were left to attach in a 37 °C incubator with 5% CO₂ overnight. The following day, cells were treated with BBI-825 for 16 h. Following treatment, all culture media was removed, and cells were fixed with 4% PFA (Boston BioProducts, catalogue no. BM-155) for 15 min at room temperature. After fixation, the 4% PFA was removed and wells were washed twice with 100 µl of PBS. The cells were then permeabilized with 100 µl of 0.5% Triton X-100 (Sigma-Aldrich, catalogue no. T8787) in PBS for 15 min at room temperature. After permeabilization, wells were washed twice with 100 µl of PBS and then blocked with 5% goat serum (Abcam, catalogue no. ab7481) and 1 mg ml^−1 of BSA (GeminiBio, catalogue no. 700-100 P) for 1 h at room temperature. The primary antibody (phospho-RPA32 (S8); Cell Signaling, catalogue no. 54762) was diluted at 1:200 in blocking buffer and 50 µl was added to all wells and incubated at 4 °C overnight. Plates were then washed three times with 100 µl of PBS and then incubated with 1:1,000 dilution of secondary antibody (Goat anti-Rabbit IgG Alexa Fluor Plus 594; Thermo Fisher, catalogue no. A32740s) and 1:1,000 dilution of Hoechst 33342 (Biotium, catalogue no. 40046) in blocking buffer for 1 h at room temperature. Plates were then washed three times with 100 µl of PBS; 100 µl of PBS was left in the wells following the final wash. The plate was imaged using a CellInsight CX7 LZR Pro High Content imager (Thermo Fisher Scientific) and data analysed using the Spot Detector BioApplication module on the HCS Studio Cell Analysis software (Thermo Fisher Scientific). Puncta were detected using a pixel thresholding method within a nucleus, and cells that contained three or more puncta of phosphorylated RPA32 Ser8 staining were considered as a positive signal.

Xenograft

Animal experiments were performed in accordance with protocols approved by the Charles River Accelerator and Development Lab (CRADL) Institutional Animal Care and Use Committee (protocol no. EB17-010-066). Mice were socially housed in individually ventilated cages on a 12/12 h light/dark cycle with temperatures between 65 and 75 °F and 30–50% humidity. The SNU16 gastric cancer cell line was purchased from ATCC (catalogue no. CRL5974) and maintained in Roswell Park Memorial Institute growth medium (Gibco, catalogue no. 22400-089) supplemented with 10% FBS (Omega Scientific, catalogue no. FB-02). To establish tumours, 1 × 10^6 SNU16 cells in 200 µl of a 1:1 mixture of PBS and Matrigel (Corning, catalogue no. 354234) were given by subcutaneous injection into the right flank of 9-week-old female severe combined immunodeficient beige mice (Envigo, strain code 186). Tumour measurements were taken two to three times per week and body weights were taken daily. Tumour volume measurements were obtained using digital calipers and tumour volumes (mm³) were determined using the formula: tumour volume = (L × W2)/2, where L is the length/largest tumour diameter and W is the width/shortest tumour diameter, with all tumours collected before reaching 1,500 mm³. Animals (eight mice per group, which historically allowed for significance determination between vehicle and infigratinib) were randomly assigned to unblinded treatment with vehicle, infigratinib (15 mg kg^−1 oral (PO) once-daily (QD)), BBI-2779 (30 mg kg^−1 PO every other day (Q2D)) or the combination of BBI-2779 and infigratinib once average tumour volume was 285 (±10)/mean (±s.e.m.) mm³. One vehicle tumour was taken down on day 22; the mouse was sacrificed due to large tumour volume. Infigratinib was formulated in a 1:1 mixture of sodium acetate buffer, pH 4.6 and polyethylene glycol 300. BBI-2779 was formulated in 0.5% methylcellulose (Sigma-Aldrich, catalogue no. M0512) and 0.2% Tween 80 (AG Scientific, catalogue no. T-2835) in HyPure Molecular Biology Grade Water (HyClone, catalogue no. SH30538.02). Dose holidays were provided to individual animals that demonstrated greater than −10% body-weight change from baseline, and Nutra-Gel was provided to the entire treatment group. Animals were sacrificed 6 h, 24 h or 36 h after the last dose, and tumours were collected for western blot or copy number analysis.

Copy number analysis from xenograft samples

For copy number analysis, tumours were cut into 10–20 mg pieces and flash-frozen in liquid nitrogen. DNA was extracted using the QIAcube DNA Extraction Kit (Qiagen, no. 51331). Briefly, a mixture of buffer ATL and proteinase K was added to the frozen tumour pieces, and they were set out to equilibrate to room temperature. Tumours were then vortexed for 30 s and placed into an incubator at 56 °C to digest overnight. The next morning, an additional 150 μl of buffer ATL was added and samples vortexed for an additional 30 s to reduce the viscosity of the samples before transfer to the S block. Qiagen protocol for the 96 QIAcube HT was followed for the remainder of the DNA isolation. Purified DNA was quantified for the presence of double-stranded DNA on the QIAxpert (Qiagen, catalogue no. 9002340). The DNA was diluted to 5 ng µl^−1 (5× working stock) in RNase/DNase free water (Thermo Fisher Scientific, catalogue no. 10977015) and 2 µl was loaded into a 384-well plate. Master mix recipe (Master Mix (2×), 5.5 µl; CNA (Target Gene) 20×, 0.55 µl; CNR telomerase reverse transcriptase (TERT) 20×, 0.55 µl; nuclease-free water, 2.2 µl) was made containing TaqPath Pro Master Mix 2× (Thermo Fisher Scientific, catalogue no. A30866) human female genomic DNA (Promega, catalogue no. G1521) as a reference, internal controls (human TERT) and FGFR2 or MYC target gene probe (Thermo Fisher Scientific, catalogue no. 4400292). Reactions were run on the QuantStudio 6 or 7 (Thermo Fisher Scientific) using the qPCR reaction settings as follows: denature/enzyme activation: 95 °C, 10 min; 40 cycles of denature 95 °C, 15 s; anneal/extend 60 °C, 60 s.

Quantifications and statistical analysis

All statistical methods and sample size have been stated in figure legends or the Methods section. No statistical methods were used to predetermine the sample size. The default test type was a two-sided statistic test, unless indicated in the text. The investigators were not blinded to allocation during experiments and outcome assessment.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.