Chromothripsis drives the evolution of gene amplification in cancer

Abstract

Focal chromosomal amplification contributes to the initiation of cancer by mediating overexpression of oncogenes^1-3, and to the development of cancer therapy resistance by increasing the expression of genes whose action diminishes the efficacy of anti-cancer drugs. Here we used whole-genome sequencing of clonal cell isolates that developed chemotherapeutic resistance to show that chromothripsis is a major driver of circular extrachromosomal DNA (ecDNA) amplification (also known as double minutes) through mechanisms that depend on poly(ADP-ribose) polymerases (PARP) and the catalytic subunit of DNA-dependent protein kinase (DNA-PKcs). Longitudinal analyses revealed that a further increase in drug tolerance is achieved by structural evolution of ecDNAs through additional rounds of chromothripsis. In situ Hi-C sequencing showed that ecDNAs preferentially tether near chromosome ends, where they re-integrate when DNA damage is present. Intrachromosomal amplifications that formed initially under low-level drug selection underwent continuing breakage-fusion-bridge cycles, generating amplicons more than 100 megabases in length that became trapped within interphase bridges and then shattered, thereby producing micronuclei whose encapsulated ecDNAs are substrates for chromothripsis. We identified similar genome rearrangement profiles linked to localized gene amplification in human cancers with acquired drug resistance or oncogene amplifications. We propose that chromothripsis is a primary mechanism that accelerates genomic DNA rearrangement and amplification into ecDNA and enables rapid acquisition of tolerance to altered growth conditions.

Extended Data Figure 1 |. Genomics and transcriptomics before and after methotrexate resistance

(a) Representative DNA-FISH images showing parental HeLa karyotype of chromosome 5 found in the five primary clones used in the study. The parental der(3p5q) is shown in the top left image. (b) Colony assay showing methotrexate sensitivity of the parental HeLa and five derivative clones. (c) Representative DNA-FISH images (of the indicated independent experiments n displayed below each image) of surviving cells and colonies of naïve HeLa cells treated with methotrexate at indicated concentrations for the indicated times. Increased DHFR signals over wild-type (>3 signals) were frequently observed, and DHFR aggregates indicative of HSR formation (with chromosome 5 interphase bridge detected in some cases – outlined with white dashed line in top right image) or DHFR+ DMs (dispersed signal) were found in surviving colonies of 100–200 cells at day 17. (d) RNA expression and DNA copy number levels plotted on the linear maps of chromosome 5 of six methotrexate resistant clones with no DHFR amplification. Representative DNA-FISH images are displayed (of at least 10 different chromosome spreads from each clone). (e) Linear regression comparing DNA copy number and RNA expression levels of DHFR in methotrexate resistant HeLa clones. (f) Principal component analysis (PCA) of naïve and resistant HeLa clones (with DMs or without DHFR amplification).

Extended Data Figure 2 |. Formation of DMs through chromothripsis in methotrexate resistant cells

(a) Representative DNA-FISH of metaphase spread prepared from clone PD29424h showing DHFR amplified in DMs (marked with white circles). Alignment of the normal chromosome 5 (DHFR+) and shorter chromothriptic chromosome 5 (DHFR-) is presented. Representative DNA-FISH from a metaphase spreads hybridized with chromosome 5 paint probe (green) and BAC probe (RP11–958F12, red, ~30Mb away from DHFR) found on both the normal chromosome 5 (single location) and shorter chromothriptic chromosome 5 (multiple dispersed locations) is also presented (b, I, k, m-n) Copy number, allelic ratio, and structural variation profiles of indicated samples. (c) Representative FISH images from clone PD29425d showing a normal chromosome 5 (DHFR+, RP11–958F12 positive), a shorter chromothriptic chromosome 5 (DHFR-, RP11–958F12 positive), and a DM (DHFR+, RP11–958F12 negative). See panel b for BAC probe target. (d) Representative DNA-FISH image of metaphase spreads prepared from clone PD29427k and hybridized with BAC RP11–314L7 (labels fragment #1 of the DM as shown in Figure 1d) and DHFR locus probes. Insets show the co-localization of the non-contiguous genomic locations within the DMs. (e) Representative FACS analysis using propidium iodide staining showing cell cycle distribution in asynchronized and mitotic clones prepared for Hi-C experiments. For gating strategy see Supplementary Information Figure 2. (f) In situ Hi-C map of chromosome 5 (65–110Mb) of asynchronized and mitotic cells from methotrexate resistant clone PD29427k. Interactions between distant chromosome locations (preserved also in the mitotic sample where topologically associated domains are erased) are circled in black (above the diagonal red line) and presented natively (below the line). (g) Reconstruction of the circular map of the DM present in clone PD29427k. Numbers represent the four DM segments appearing in panel f (notice that segment 3 is rearranged). (h) List of the genes found on the DM (with RNA expression relative to naïve HeLa cells). Numbers represent the four DM segments appearing in panel f. (j) Representative DNA-FISH from clone PD29425l showing co-localization of DHFR (green) and BAC 33C19 probes signals. See panel i for BAC probe target. (l) Representative DNA-FISH image of metaphase spreads prepared from clone PD29425g hybridized with chromosome 5 paint (green) and DHFR locus (red) probes. Inset shows the DHFR positive DM. Notice only one chromosome 5 and one der(3p5q) are present. (o-p) Representative DNA-FISH image of metaphase spreads prepared from clones PD45714a (o) and PD45725b (p) hybridized with chromosome 5 paint (green) and DHFR locus (red) probes showing DHFR amplification in DMs. (q) List of DM+ HeLa clones showing size, copy number, and number of non-contiguous fragments. (r) Colony assay of naïve HeLa cells treated with DNA repair inhibitors (control – untreated). Images are representative of two independent experiments.

Extended Data Figure 3 |. BRAF^V600E amplification in drug resistant colorectal cancer patients

(a, c) Treatment and biopsies collection timeline of colorectal cancer patient 1 and 2. (b, d) Representative FISH images (of entire section stained) showing BRAF amplification in biopsies from post-treatment biopsies from patients 1 and 2. Probes used are listed in the methods section.

Extended Data Figure 4 |. DMs are numerically and structurally unstable

(a) Strategy used to isolate DM or HSR positive clones from a heterogeneous population of methotrexate resistant HeLa cells. (b) Left: Average number of DHFR+ DMs in three clones derived from a methotrexate resistant HeLa population was determined using DNA-FISH. Analyses with indicated p-values above plots were performed using Student t-test. *p<0.05, **p<0.01, ***p<0.001 are p-values calculated using one-way ANOVA. Error bars represent mean±s.e.m., number of spreads examined for each condition is written below each graph. Right: Representative DNA-FISH image of metaphase spreads from clone PD29429f (640nM) containing 46 DHFR+ DMs as determined by DNA-FISH. Inset shows a representative DHFR positive DM. (c) Average intensity of DHFR signal in DMs relative to the intensity of the endogenous DHFR signal on chromosome 5 from the same spread, as determined by DNA-FISH using a probe (RP11–90A9) specific for the DHFR locus. Analyses with indicated p-values above plots were performed using Student t-test. **p<0.01 and ***p<0.001 are p-values calculated using one-way ANOVA. Error bars represent mean ± s.e.m. Representative DNA-FISH image of chromosome spread from which insets are shown in Figure 2b is presented. (d) Copy number, allelic ratio, and structural variation profiles of indicated samples. (e) Read counts in structural variation breakpoints in sample PD29429i showing potential different DM subspecies forming through chromothripsis (f) Breakpoint PCR using primers specific for three different rearrangements detected using WGS. For gel source data, see Supplementary Information Figure 1. (g) Percentage of DM-positive cells exposed to low or high methotrexate concentrations with large micronuclei, as determined by DNA-FISH. Results represent an average of 3 clones per methotrexate concentration (as seen in Extended Data Fig. 4a), error bars represent mean ± SD. Representative DNA-FISH image of a cell with a large micronucleus is provided. 40nM: n = 147, 130, 98; 640nM: n = 74, 216, 133. (h) Analysis of DM inheritance to two daughter cells after one cell division. Cells were seeded at low dilution and 24 hours later DNA-FISH on interphase daughter cells from PD29429h (40nM) or PD29429i (640nM) was performed using a DHFR probe. DM numbers in each daughter cell were counted and plotted on an x-y plot (n = 54 daughters per condition). (i) Schematic explaining the steps of a simulation testing the effect of harboring DMs with more than one DHFR gene. (j-l) In-silico simulation showing that cells containing a 2-copy DHFR DM have a selection advantage (i), with increased DM content (j), and that 2-copy DHFR DMs are selected over DMs with a single copy of DHFR (k). Plotted is the median and first and third quartile (grey ribbon). For simulation pseudocode see Supplementary Information Figure 3.

Extended Data Figure 5 |. DMs integrate into ectopic chromosomes following DNA damage

(a-c) Percentage of DM integration as detected using DNA-FISH with probes against chromosome 5 (green) and DHFR (red) in clone PD29429i after ionizing irradiation (a), doxorubicin (b), or transfection with nucleases specific for a region near the DHFR locus (c). Representative images of DNA-FISH of DMs integrated into ectopic chromosomes are presented below each graph. (d) Percentage of ectopic HSRs detected in DM-positive clones treated with increased methotrexate concentration (x2.5 fold higher) with or without (control, 4 clones) the addition of ABT-888 (PARPi, 7 clones) and NU7026 (NHEJi, 4 clones) for 3 weeks. Means ± s.e.m. are presented. control: n = 107, 96, 46, 164; PARPi: n = 67, 247, 94, 128, 17, 130; NHEJi: n = 51, 45, 28, 38. Representative FISH images of ectopic HSRs at the end of chromosome (78% of the cases) or in the middle of chromosomes (22% of the cases) are presented. (e) Percentage of GFP+ HSRs detected in colo320DM-GFP cell line treated with DMSO (control) or with ABT-888 (PARPi) or olaparib (PARPi) for 2 weeks. Means ± s.e.m. of three independent experiments are shown. p-values *p<0.05 and **p<0.001 calculated using one-way ANOVA are presented. control: n = 26, 47, 33; ABT-888: n = 42, 44, 37; Olaparib: n = 29, 32, 34. Representative fluorescent images of GFP+ DMs (from control) and GFP+ HSRs (from PARPi treated cells) are presented. (f) Percentage of chromosome spreads in which large DHFR+ DMs were detected using DNA-FISH when PD29429i was exposed to 1500nM methotrexate only (control) or with addition of 15μM ABT-888 (PARPi) or 10μM NU7026 (DNA-PKi) for 3 weeks. Means ±s.e.m of three independent experiments are presented. * p-value < 0.001 calculated using one-way ANOVA and Tukey’s Multiple Comparison Test. Number of spreads scored per condition/experiment: 1500nM MTX (52, 53, 49), 1500nM MTX + ABT-888 (50, 76, 55), 1500nM MTX + NU7026 (57, 16, 42). Representative image (of three independent experiments as shown in the plot to the left) of a chromosome spread containing multiple normal sized and large DMs is presented. (g) Percentage of chromosome spreads in which large myc+ DMs were detected using DNA-FISH in colo320DM-GFP cells with indicated types of treatment for 2 weeks (DMSO served as control, PARP inhibitors used were 15μM ABT-888 and 10μM Olaparib, hydroxyurea used at 100μM). Representative image of a chromosome spread containing multiple normal sized and large DMs (white arrows) is presented. Means ± s.e.m of three independent experiments with *p<0.05 calculated using one-way ANOVA are presented. control: n = 22, 22, 21; ABT-888: n = 26, 19, 21; Olaparib: n = 20, 22, 21; hydroxyurea: n = 20, 22, 20; hydroxyurea+ABT-888: n = 19, 22, 24; hydroxyurea+Olaparib: n = 20, 22, 22. (h) Quantification of the interaction frequency between DM or HSR sequences and the rest of the genome in mitotic or asynchronous cells. Linear chromosome maps (chromosomes 1-X, excluding chromosome 5) are presented from beginning to end (p arm to q arm direction), and the normalized interaction frequency is presented. (i) Interaction frequency as a function of distance from the centromere (R - pearson correlation). (j) Fold-change (relative to cells with no amplification) of DHFR sequences interaction in DM+ and HSR+ with center of chromosomes (50% of sequences in chromosome center) and with chromosome ends (25% of sequences in each chromosome end). p-values calculated using two-sided paired t-test.

Extended Data Figure 6 |. Increased selective pressure drives the transition of intra- to extra- chromosome amplification through chromothripsis

(a) Experimental outline of a parental HSR+ clone treated with increasing methotrexate concentrations. (b) Representative FISH image of 23 metaphase spreads prepared from the methotrexate resistant HSR+ clone PD29426a hybridized with chromosome 5 paint (green) and DHFR locus (red) probes. Inset shows the DHFR repeat located on the 5q arm of the derivative 3p5q chromosome. (c) DNA copy number, rearrangement, and allelic ratio profile of clone PD29426a showing two copy number jumps flanked by head-to-head and tail-to-tail inversions corresponding to two BFB cycles. (d) Schematic depicting the order of events during the two break-fusion-bridge (BFB) cycles that led to the formation of the HSR in clone PD29426a. (e) RNA expression presented on the linear chromosome 5 map showing increased gene expression from the HSR region. (f) In situ Hi-C map of chromosome 5 (65–110Mb) of mitotic cells from methotrexate resistant clone PD29426a, showing the HSR in which there is higher interaction within the regions with copy number jumps. (g) In Situ Hi-C map of chromosome 5 (65–110Mb) of asynchronized cells from HSR+ clone PD29426a. TADs are visible throughout this region and increased interactions are observed in the HSR region (increased red coloration, top left quadrant). (h) Cell cycle analysis using propidium iodide staining showing cell cycle distribution in asynchronized and mitotic cells of clone PD29426a. For gating strategy see Supplementary Information Figure 2. (i) Representative FISH image (of two independent clones, total of 26 spreads imaged) of metaphase spreads prepared from a DM subclone (PD29426h) derived from the HSR+ clone PD29426a. Spreads were hybridized with chromosome 5 paint (green) and DHFR locus (red) probes. Inset shows the DHFR positive DMs. (j) DNA copy number, rearrangement, and allelic ratio profile of clone PD29426h. The DM is composed of fragments with varying copy number states derived from the original HSR region, connected by multiple rearrangements.

Extended Data Figure 7 |. Chromothripsis and kataegis in gene amplification

(a) Representative FISH image (of 12 spreads imaged) of metaphase spreads prepared from clone PD29427p and hybridized with chromosome 5 paint (green) and DHFR locus probes, revealing the existence of a DHFR+ HSR. (b) Top: Copy number profile of a methotrexate resistant clone (PD29427p) presenting an HSR profile resulting from 3 BFB cycles with overlaid complex rearrangements. Rearrangements are presented on top: TD – tandem duplication, D – deletion, HH – head-to-head inversion, TT – tail-to-tail inversion. Inter-mutation distance (i.m.d) and allelic ratio (bottom, blue/red dots) are also presented. Bottom: RNA expression profile of clone PD29427p, relative to control naïve HeLa cells, presented on the linear map of chromosome 5 and matching the DNA copy number plotted above. (c) Representative FISH image (of 6 spreads imaged) of metaphase spreads prepared from clone PD29427p and hybridized with chromosome 5 paint (green) and BAC probe 314L7 (see chromosomal location indicated in panel b) probes. Inset shows probe 314L7 signal is flanking the DHFR signal within the HSR. (d) Rainfall plots showing the inter-mutation distance (i.m.d.) within chromosome 5 of each DM clone. (e) Structural variations and inter-mutation distance (i.m.d.) of three kataegis positive DM clones.

Extended Data Figure 8 |. Characterization of steps leading to HSR fragmentation and DM formation

(a) Representative FISH images (of 464 spreads imaged from 8 different timepoints as shown in Figure 4a) showing chromosome 5 (green) and DHFR (red) probes in HSR+ clone PD29428e at basal methotrexate concentration (60nM, top), and following treatment with increased methotrexate (500nM, middle and bottom) in which HSR fragments of variable sizes and DMs can be detected. (b) Colony assay of clone PD29428e (HSR+) resistant to 60nM and treated with 500nM methotrexate. Numbers of cells seeded are indicated and colonies were fixed and stained at day 18. (c) Number of cells counted (automatically) using the CQ1 microscope in 10 minutes interval during 48 hours of filming of PD29428e cells treated with 60nM or 500nM or without (dFBS) methotrexate. Cell numbers were normalized to the initial number present at time point 0. P-value calculated using one-way ANOVA. (d) Measurement of the HSR length in clone PD29428e after ten days of exposure to higher methotrexate concentration. The length of the HSR was normalized to the length of the normal chromosome 5 from the same spread. Mean ± s.e.m. of n=15 per group and p-value calculated using two-tailed t-test are presented. Insets showing length of the HSR at day 10 after exposure to basal (left) or increased (right) methotrexate concentrations (scale is 180MB, equivalent to chromosome 5 size). (e) Measurement of the HSR length in clone PD29428e during two-week exposure to higher methotrexate concentration. The length of the HSR was normalized to the length of the normal chromosome 5 from the same spread. Means ±s.e.m of 15 HSR lengths per each timepoint (except day 14 – 16 HSRs scored). p-values calculated using one-way ANOVA and * denotes significant (range of p<0.05–0.001) change in HSR length between each point of 8–14 days and each point of 0–6 days. (f) Percentage of spreads containing dicentric HSR+ chromosomes in clone PD29428e exposed to basal or increased methotrexate concentration for 10 days. Mean ± SD of three independent experiments (60nM: n = 42, 30, 33; 500nM: 23, 33, 29) and p-value calculated using two-tailed t-test are presented. (g) Percentage of spreads containing dicentric HSR+ chromosomes in clone PD29428e after increasing methotrexate concentration for the indicated times. day 0: n = 98; day 2: n = 41; day 4: n = 39; day 6: n = 63; day 8: n = 53; day 10: n = 55; day 12: n = 66; day 14: n = 123. Inset shows representative image (from 538 spreads imaged) of a chromatid-type dicentric chromosome with one fused end, and chromosome type fusion with two distinct centromeres. Percentages indicate frequency range of each dicentric type throughout the two-week experiment. (h) Abnormalities identified in PD29428e cells treated without or with increased methotrexate concentration (60nM and 500nM, respectively) for 11 days during 48 hours filming. Mean ±S.D. of three independent experiments and p-values (*p<0.01 and *p<0.001) calculated using two-way ANOVA are presented. Division scored: 60nM: n = 115, 108, 95; 500nM: n = 51, 50, 42. Bridges (anaphase bridges), micronuclei (MN), bridges+MN, death, and multipolar divisions were scored. Insets from Supplementary Video 1 showing anaphase bridges, micronuclei, and also interphase bridges are presented. (i) Analysis as performed in (h) looking at events occurring during the division of the daughter cells generated from the parental cells scored in (h). Division scored: 60nM: n = 100, 90, 66; 500nM: n = 85, 76, 66 (j) Top: Percentage of nuclei presenting DHFR+ nuclear bulges indicative of bridge rupture in PD29428e HSR+ clone treated with increased methotrexate (500nM) over the course of 12 days. Mean ± SD of three independent experiments (60nM: n = 144, 221, 204; 500nM: n = 163, 218, 214) and p-value calculated using two-tailed t-test are presented. Bottom: Representative image of DHFR+ (red) labeled nuclear bulge. (k) Representative image of the data in Fig. 4d. (l) Copy number profile of a methotrexate resistant clone (PD45727b) presenting an HSR profile resulting from multiple BFB cycles (marked with red arrows) with overlaid complex rearrangements. A representative DNA-FISH image (of 20 spreads imaged) showing dicentric HSRs is presented. DMs were observed in low frequency under this condition. (m) Copy number profile of clone PD45727b subjected to increased methotrexate concentration (sample PD45728b) presenting additional rearrangements and formation of multiple DMs (as seen in the representative DNA-FISH image from 10 spreads imaged, white circles). TD – tandem duplication, D – deletion, HH – head-to-head inversion, TT – tail-to-tail inversion.

Extended Data Figure 9 |. Chromothripsis drives resolution of Kif15 intra-chromosomal amplification through telomere capture

(a) Copy number profile of chromosome 3 showing a rearranged 3p region with Kif15 amplification (copy number = 15) and a translocation of a telomeric fragment from a lost chromosome 1. (b) Representative DNA FISH images (of 33 spreads imaged) from STLC resistant clones with (PD45759a) or without (PD45760a) Kif15 amplification showing the translocation of the telomeric chromosome 1 fragment to the end of a chromosome 3 arm. (c) Representative DNA FISH images (of 20 spreads imaged) for chromosome 3 (green) and Kif15 (RP11–659N22, red) showing normal Kif15 in PD45760a (STLC resistant without Kif15 amplification, 3 Kif15 copies labeled with “a”, “b”, and “c”) and in PD45759a (STLC resistant with Kif15 amplification found on two rearranged chromosomes marked with “*” and originating from chromosome “c” as seen in PD45760a). (d) Representative DNA FISH image (from 25 spreads imaged) showing from PD45759a showing the rearranged chromosome containing Kif15 amplification and telomeric region of chromosome 1 at the tip of the chromosome. (e) Proposed order of events leading to Kif15 amplification and resolution of the dicentric state through a chromothriptic event.

Extended Data Figure 10 |. Structure of the interphase chromosome bridge

(a) Snapshots from live cell imaging (Supplementary Video 2) of PD29428e cells exposed to increased methotrexate concentration (500nM) for 11 days, labeled with SiR-DNA label, and filmed for 48 hours (of 944 cell divisions filmed, as shown in Extended Data Figure 8h–i). Frames are at 10-minute intervals. An initial DNA+ bridge forms from which a DNA fragments is maintained as a micronucleus following bridge rupture. (b) Representative image of PD29428e cells exposed to increased methotrexate concentration (500nM) for 10 days and analyzed using DNA FISH probes for chromosome 5 centromere (green) and DHFR (red), showing micronuclei within interphase DNA bridges (representative of 34 bridges) (c) Immunofluorescent image analysis of zeocin treated cancer cell lines showing formation of micronuclei within interphase DNA bridges (representative images of 85 bridges with micronuclei from the four cell lines tested. Number of cells scored: 293T control: n = 99; 293T zeocin: n = 116; DLD1 control: n = 88; DLD1 zeocin: n = 107; HeLa control: n = 126; HeLa zeocin: n = 412; HT29 zeocin: n = 25; T47d zeocin: n = 73). (d-e), correlative light and electron microscopy (CLEM) images (of a total of 5 images taken) of an anaphase bridge (d, red arrow marks aberrant nuclear envelope at midbody region) and an interphase bridge (e, red arrow marks constriction in DNA bridge at midbody site). NE – nuclear envelope. (f) Representative immunofluorescence images (of 43 different bridges) showing AuroraB localization in interphase DNA bridges. (g) Correlative light and electron microscopy (CLEM) of an abnormally shaped micronucleus found in PD29428e cells exposed to increased methotrexate concentration (500nM) for 10 days in which micronuclei form following interphase bridge rupture (representative of 2 micronuclei imaged).

Extended Data Figure 11 |. Similarities between DHFR and oncogene amplifications

(a-b) Summary of the overall structural variant events found across all samples analyzed in the methotrexate resistant HeLa cohort (a, 18 clones) and in glioblastoma cancer patients from the TCGA database (b, 41 patients). (c) Amplifications (>= 10 copy numbers) of either of 11 indicated genes in patients found in the TCGA database. (d) Distribution of rearrangements in chromosomes containing specific amplifications. Near equal distribution of rearrangements suggest events are random as expected in chromothripsis. TD – tandem duplication, D – deletion, HH – head-to-head inversion, TT – tail-to-tail inversion. (e) Copy number plots showing the mean copy number gain presented on the linear chromosome map in methotrexate resistant HeLa cells (chromosome 5) and TCGA samples presenting oncogene amplifications. The percentage of each type of rearrangement found in chromosomes with amplifications is presented. Red – tandem duplications, Blue – head-head inversions, Green – tail-tail inversions, Purple – deletions. (f) Microhomology (red), non-templated (teal) sequences, or direct end-joining (yellow), found at breakpoints in chromosomes with specific oncogene amplifications from TCGA patients. The number of homologous or non-homologous bases at breakpoints are shown on the x-axis. The number of breakpoints showing each kind of junction is shown on the y-axis. (g) Schematic illustrating a roadmap to gene amplification in cancer cells.

Figure 1 |. Chromothripsis drives ecDNA formation and amplification as DMs early during drug resistance development in a PARP and NHEJ dependent manner

(a) Strategy used to generate unique and independent methotrexate resistant HeLa clones. Pie charts show the distribution of chromosome 5 abnormalities. Representative DNA-FISH images (from the 57 clones) of each abnormality are presented. (b) Summary of the genomic mechanisms (determined by whole-genome paired-end sequencing) leading to each type of karyotypic outcome (determined by FISH). (c-g, k-l) Copy number and structural variation profiles of (c-g) DHFR+ DM+ methotrexate resistant clones and (k-l) vemurafenib resistant patients. Allelic ratios are shown for c-e. TD – tandem duplication, D – deletion, HH – head-to-head inversion, TT – tail-to-tail inversion. (h) Sequence analysis of structural variation breakpoints in chromothriptic DM HeLa clones. (i) Colony assay of methotrexate treated naïve HeLa cells with or without DNA repair inhibition. Images represent three independent experiments. (j) Quantification of cells with DHFR+ DMs as determined by FISH under indicated conditions. Average of three independent experiments. *p<0.05 (repeated measures one-way ANOVA with post-hoc Tukey test, mean ± s.e.m).

Figure 2 |. DM structure is optimized by chromothripsis during adaptation to increased selection pressure

(a) Quantification of DM numbers using DNA-FISH. Mean ± SD of three independent clones (see Extended Data Fig. 4b) and p-value calculated using two-tailed t-test are presented. (b) Left: Average intensity of DHFR signal in DMs relative to the intensity of the endogenous DHFR signal on chromosome 5 from the same spread, as determined by DNA-FISH. Mean ± SD of three independent DM clones is presented (see Extended Data Fig. 4c). p-value was calculated using two-tailed t-test. Right: DNA-FISH image from a chromosome spread of clone PD29429i (see Extended Data Fig. 4c) showing a DM with increased signal intensity. (c-d,f-i) Copy number and structural variation profiles of indicated samples. Allelic ratios are shown for HeLa clone PD29429h and a derivative population following adaptation to 640nM methotrexate (c-d). A population of HT-29 cells with DHFR+ DMs resistant to initial 300nM methotrexate (f, PD45712a), and after increasing methotrexate concentration to 4000nM (g, PD45713a) are presented. Results of a clone derived from the population (PD45712a) is presented in h (PD45717a), also after increasing methotrexate concentration to 1000nM (i, sample PD45718a, notice massive rearrangements are different from PD45713a shown in g, highlighting the random nature of the rearrangements) are also presented. (e) Quantification of DHFR+ DM sized (small) micronuclei using DNA-FISH (representative image presented). Mean ± SD of three independent clones and p-value calculated using two-tailed t-test are presented. 40nM: n = 147, 130, 98; 640nM: n = 74, 216, 133.

Figure 3 |. Double minutes preferentially integrate in broken DNA sites near chromosome ends and form ectopic HSRs

(a) Quantification of DM integration into ectopic sites on chromosome 3 following double strand break induction using Cas9. DNA-FISH images of chromosomes with DM integration from each experiment are provided (b-e) Copy number, structural variation, and translocation profiles of four independent subclones derived from clone PD29429i treated with 1500nM methotrexate and 15μM ABT-888 (PARP inhibitor). DNA-FISH images of chromosomes with DM integration in each clone are provided (integrations observed in all spreads from each clone, n = 15 per clone).

Figure 4 |. HSRs fragment in interphase bridges and form micronuclei and DMs through chromothripsis and NHEJ repair

(a) Location of DHFR amplification as determined using DNA-FISH in clone PD29428e over the course of two weeks when exposed to increased methotrexate concentration (500nM) (b) Quantification of DHFR+ interphase bridges in clone PD29428e using DNA-FISH. Mean ± SD of three independent experiments (number of cells scored - 60nM: 144, 221, 204; 12 days in 500nM: 163, 218, 214) and p-value calculated using two-tailed t-test are presented. (c) Representative DNA-FISH images of DHFR+ interphase bridges (from data shown in panel b). (d) Quantification of DHFR+ over the course of two weeks in clone PD29428e exposed to increased methotrexate concentration as determined by DNA-FISH (representative image from the data is shown in Extended Data Fig. 8k). Day 4 (timepoint 1) shows increased DHFR- micronuclei (due to the random DNA damage from increased methotrexate). Day 12 (timepoint 2) shows increased DHFR+ micronuclei (product of HSR fragmentation), followed by reduction of DHFR- micronuclei two days later (timepoint 3). (e-f) Copy number and structural variation profiles of indicated samples. (g) Quantification of amplification patterns in clone PD29428e exposed to increased methotrexate concentration for 10 days, with or without addition of the DNA repair inhibitors. Bars represent mean ± SD of three independent experiments (spreads counted - 60nM:42, 32, 33; 500nM: 36, 50, 38; 500nM+10μM NU7026: 37, 35, 45; 500nM+15μM ABT-888: 35, 34, 32). *p<0.05, ***p<0.001 (compared to 60nM condition). #p<0.001 (compared to 500nM condition). p-values calculated using two-way ANOVA. (h) Correlative light and electron microscopy (CLEM) showing an interphase bridge (of 17 bridges imaged) with micronuclei on both sides (blue arrowheads) of the midbody (red arrowhead). (i) Summary of the events underlying BFB-chromothripsis.