Chromothripsis and ecDNA initiated by N4BP2 nuclease fragmenting cytoplasm exposed chromosomes

Abstract

Genome instability, including chromothripsis, is a hallmark of cancer. Cancer cells frequently contain micronuclei—small, nucleus-like structures formed by chromosome missegregation—that are susceptible to rupture, exposing chromatin to cytoplasmic nucleases. Through an unbiased imaging-based siRNA screen targeting all 204 known and putative human nucleases, we identified a previously uncharacterized cytoplasmic endonuclease, NEDD4-binding protein 2 (N4BP2), that enters ruptured micronuclei and initiates DNA damage, leading to chromosome fragmentation. N4BP2 promoted formation of extrachromosomal DNA (ecDNA) in drug-induced gene amplification. N4BP2 promoted DNA fragmentation, tumorigenesis, and tumor cell proliferation in an induced model of human high-grade human glioma. Analysis of over 10,000 human cancer genomes revealed elevated N4BP2 expression to be predictive of chromothripsis and copy number amplifications, including ecDNA.

One-Sentence Summary:

The cytoplasmic nuclease N4BP2 was identified to fragment chromosomes within ruptured micronuclei, thereby initiating genome rearrangements, extrachromosomal DNA (ecDNA) formation, and tumorigenesis, with elevated expression of it in human cancers predictive of chromothripsis and copy number amplifications, including ecDNA.

Chromothripsis, the most common type of chromoanagenesis (1) and a major driver of genome instability in cancer (1–3), involves extensive chromosome shattering and random reassembly of fragments, initiated by rupture of aberrantly formed micronuclei (1–4). Such ruptures expose chromatin to cytoplasmic nucleases, potentially triggering chromosomal fragmentation and rearrangement (1, 5–10). We performed an unbiased screen covering all – both reported and putative – human nucleases to identify nucleases whose activity promotes fragmentation of chromosomes within ruptured micronuclei.

Imaging-based screen for human nucleases that fragment micronucleated chromosomes

We first generated a list of all known and putative human nucleases (n=204, based on the protein function resource https://www.uniprot.org/, combined with relevant literature, Table S1), and developed a quantitative imaging-based approach to monitor DNA damage specifically in micronuclei, using phosphorylation of histone H2AX on Ser139 (γH2AX) as a marker of DNA double strand breaks (11) (Fig. 1A; Fig. S1A; Fig. S2A). To systematically induce micronuclei, we employed the Dld1-Y human colorectal adenocarcinoma cell model (“Dld1-Y^MN”) (9), where selective inactivation of the Y chromosome centromere induced controlled missegregation and incorporation of the Y chromosome into micronuclei (Fig. S2B), which showed pronounced DNA damage upon micronuclei envelope rupture (Fig. 1C) (9, 12).

Figure 1. Imaging-based screen for human nucleases that fragment micronucleated chromosomes.

(A, B) Schematic of siRNA-mediated screen for human nucleases inducing DNA damage and chromosome fragmentation in micronuclei (MN). Dld1-Y^MN cells allow for inducible micronucleation of Y chromosome (see also Fig. S1A). The MN of MN-induced (for 72 hours) Dld1-Y^MN cells transfected with siRNAs targeting 204 nucleases (see Table S1) were scored for the reduction of DNA damage (using a DNA double strand breaks marker γH2AX) compared to non-silencing (NS) siRNA (negative control, high γH2AX) and γH2AX inhibitor wortmannin (positive control, low γH2AX) (γH2AX immunofluorescence, primary assay shown in [A]). Next, mitotic spreads from MN-induced Dld1-Y^MN cells, transfected with the siRNAs targeting hits identified in a primary assay, were analyzed for the presence of fragmented Y chromosomes using DNA FISH probes against Y chromosome (Y chromosome DNA FISH, secondary assay, shown in [B]). (C) DNA damage in Y chromosome-containing micronuclei with abnormal nuclear envelope in MN-induced Dld1-Y^MN cells. Y chromosome paint probe (green), γH2AX (red), Lamin B 1 (magenta). Scale bar, 5 μm. (D) Distribution of γH2AX signal in micronuclei normalized to non-silencing control (NS, dashed line) for each siRNA of siRNA library (blue rhombuses). Reduction of γH2AX upon wortmannin treatment was used as a baseline. Pink box highlights candidate genes with the strongest γH2AX reduction (zoomed in on top), N4BP2 nuclease shown as red rhombus; TREX1 and APE1 are shown as black rhombuses. Mean±SEM, n=3 independent experiments. (E, F) Representative images (E) and quantification (F) of DNA damage accumulation (detected with γH2AX, red) in micronuclei with ruptured nuclear envelope (indicated by GFP-NLS efflux from the micronucleus). Scale bar, 2 μm. Mean±SEM, n=3 independent experiments. (G) Immunoblot of N4BP2 using lysates from Dld1-Y^MN cells of indicated genotypes. Short and long exposures are shown. Tubulin was used as a loading control. (H) Quantification of metaphase spreads from MN-induced Dld1-Y^MN cells of indicated genotypes containing fragmented Y chromosome. n=3 independent clones per each genotype. (I) Representative images of metaphase spreads in Dld1-Y^MN, N4BP2 knockout (KO) clones, TREX1 KO clones in Dld1-Y^MN, after induction of MN for 72 hours. Y chromosome paint probe (green), Y centromere probe (red). Yellow arrows point at the intact Y chromosome, red arrows: Y centromere fragments, white arrows: Y chromosome fragments. Scale bars, 20 μm, 5 μm. (C, E, I) DNA was stained with 4′,6-diamidino-2-phenylindole (DAPI) (blue).

Use of Dld1-Y^MN allowed nuclease assessment on two levels: (i) DNA double strand breaks (visualized by γH2AX signals) within Y chromosome containing micronuclei (Fig. 1A; Fig. S1A) and (ii) Y chromosome fragmentation scored on metaphase spreads by DNA fluorescence in situ hybridization (FISH) with Y chromosome-specific probes (Fig. 1B; Fig. S2C). Using γH2AX in micronuclei as the primary read out—and benchmarking siRNA-mediated reductions against wortmannin, an inhibitor of γH2AX formation (11, 13)—we screened all 204 nucleases (Fig. 1A, D), excluding hits that altered γH2AX in the main nucleus (Fig. S2D). Neither apurinic/apyrimidinic endonuclease 1 (APE1), previously implicated in promoting DNA breaks in a model of micronuclei-associated inhibition of transcription followed by aberrant DNA repair (14), nor three-prime repair exonuclease 1 (TREX1), previously reported to facilitate chromothripsis following induced telomere fusion (15, 16), scored high in the screen (Fig. 1D).

In a secondary screen, we tested whether any of the top 12 candidates from the primary screen influenced the frequency of Y chromosome fragmentation after micronucleation and identified NEDD4-binding protein 2 (N4BP2)—a poorly characterized nuclease reported to nick supercoiled DNA (17) —as the top candidate. N4BP2 knockdown markedly reduced γH2AX in micronuclei and the fraction of spreads with fragmented Y chromosomes (Fig. 1B; Fig. S2E), whereas similar reduction of TREX1 or APE1 had no effect (Fig. S2F). Knockdown of N4BP2 with a different set of siRNAs similarly lowered γH2AX in ruptured micronuclei (GFP-NLS negative) and decreased Y fragmentation (Fig. 1E, F; Fig. S2G, H). Moreover, CRISPR/Cas9 mediated homozygous disruption of N4BP2 abolished Y chromosome fragmentation after micronucleation, while TREX1 deletion did not affect either Y chromosome fragmentation or γH2AX level in micronuclei (Fig. 1G-I; Fig. S2I,J). These data establish N4BP2 as a key nuclease that drives DNA damage and chromosome fragmentation within ruptured micronuclei. The residual γH2AX in N4BP2 knockout (KO) micronuclei might reflect the action of other nucleases (Fig. 1D; Fig. S2E) and/or DNA damage caused by replication, transcription, or repair stress (18–22).

Fragmentation of chromosomes exposed to N4BP2

We next examined N4BP2 localization and function. Both endogenous N4BP2 (detected by the N4BP2 antibody) and GFP-tagged N4BP2 (N4BP2-GFP) were predominantly cytoplasmic in cancer-derived (Dld1-Y^MN, U2OS, HeLa) and near-diploid RPE-1 cells (Fig. 2A; Fig. S3A; Fig. S8A). Untagged, endogenous N4BP2 formed bright foci inside micronuclei, co-localizing with—or adjacent to—γH2AX foci (Fig. 2A). In rare (<1%) examples, beyond micronuclear puncta, endogenous N4BP2 accumulated within a main nucleus in large, rounded, droplet-like foci, which partially co-localized with bright, abundant γH2AX (Fig. S3B), consistent with N4BP2 recruitment after spontaneous rupture of the main nucleus. Furthermore, use of live-cell imaging identified that nuclear entry of N4BP2-GFP into the main nucleus was followed by cell death marked by membrane blebbing, DNA compaction, and cell detachment (Fig. S3C-E).

Figure 2. Fragmentation of chromosomes exposed to N4BP2.

(A) Representative immunofluorescence (IF) images of MN-induced Dld1-Y^MN cells showing localization of endogenous N4BP2 nuclease (green) forming droplet-like foci in MN with extensive DNA damage (γH2AX, red). Scale bars, 10 μm and 2 μm, as indicated. (B) Percentage of γH2AX-positive ruptured MN in GFP-NLS-expressing Dld1-Y^MN cells upon indicated siRNA treatments. Mean±SEM, n=3 independent experiments. (C) Percentage of γH2AX-positive ruptured MN in GFP-NLS-expressing Dld1-Y^MN cells of the indicated genotypes. See (S4A) for a scheme of experiment. Mean±SEM, n=3 independent experiments. (D) Percentage of γH2AX-positive intact MN in GFP-NLS-expressing Dld1-Y^MN cells of the indicated genotypes. See (S4A) for a scheme of experiment.Mean±SEM, n=3 independent experiments. (E) Representative IF images of GFP-NLS-expressing U2OS cells, showing localization of endogenous N4BP2 nuclease within ruptured micronuclei (identified by the absence of GFP-NLS, orange arrows) with extensive DNA damage (γH2AX, red). MN were induced by nocodazole. White arrows indicate intact, GFP-NLS-positive micronuclei. Scale bars, 10 μm and 2 μm, as indicated. (F) Representative IF images of Dld1-Y^MN cells with induced MN, transfected as indicated, showing localization of N4BP2-GFP (green) inside γH2AX (red)–positive MN. Scale bars, 5 μm and 1 μm, as indicated. (G) Live-cell imaging of U2OS cells expressing N4BP2-GFP and RFP-NLS, MN were induced with nocodazole. N4BP2-GFP enters ruptured MN (identified by the loss of RFP-NLS, occurs at 15 min time point) and gradually accumulates in MN. Scale bar, 10 μm and 2 μm, as indicated. (H) (Top) Schematic of the approach used to target N4BP2-GFP into intact nuclei and micronuclei in N4BP2 KO Dld1-Y^MN cells. (Bottom) Representative images of DNA damage accumulation (γH2AX, red) upon targeting N4BP2-GFP-NLS into intact nuclei vs cytoplasmic N4BP2 (N4BP2-GFP). Scale bar, 5 μm. (I) (Top) Schematic of the approach used to target N4BP2-GFP into intact nuclei and micronuclei in U2OS cells. (Bottom) Representative images of mitotic spreads upon N4BP2-GFP-NLS expression for 0h, 24h and 48h. Chromosome fragments are indicated by green arrowheads. Numbers indicate each individual fragment in the zoomed-in area. Scale bar, 5 μm. Quantifications are shown in (J). (J) Quantifications of fragmented mitotic spreads (percentage of all spreads) in U2OS cells upon N4BP2-GFP-NLS expression for 0h, 24h and 48h. n=3 independent experiments per each condition, minimum 300 cells per each experiment. See schematic of experiment and images of representative mitotic spreads in (I). (K) Quantifications of percentage of cells with MN in WT U2OS cells (white column) vs 24h and 48h of N4BP2-GFP overexpression (red columns). Pie charts below each corresponding condition show percentage of ruptured (blue, numbers) and intact (white) MN. N=3 independent experiments, 219 cells. See (S6G) for GFP control. (L) Percentage of Y chromosome-positive nuclei in long-term cultured (40 passages) wild type (WT) and N4BP2 knockout (KO) Dld1-Y^MN cells. Schematic of the experiment is shown in (S7A,B). n=3 independent clones per each genotype, ≥200 cells per clone. See also (S7).

Both siRNA knockdown and CRISPR-Cas9 knockout of N4BP2 significantly reduced the fraction of γH2AX-positive ruptured micronuclei and the γH2AX signal intensity within them, without affecting intact micronuclei (Fig. 2B-D; Fig. S4A, Fig. S8C-E). Almost all (95% in Dld-Y^MN and 85% in U2OS) of N4BP2-positive micronuclei were ruptured and γH2AX-positive (Fig. S4B,C), indicating that envelope rupture is required for N4BP2-mediated DNA damage. Indeed, live cell imaging (Fig. 2G; Fig. S5A,B) of N4BP2-GFP expressing cells, as well as fixed cell imaging of endogenous N4BP2 or N4BP2-GFP, revealed rapid entry of N4BP2 into ruptured micronuclei which was accompanied by generation of γH2AX foci (Fig. 2E,F; Fig. S4B,C; Fig. S8B). Altering N4BP2 levels did not change the proportion of ruptured micronuclei (Fig. S4D-F), indicating that N4BP2 acts on micronuclei after rupture rather than inducing rupture itself. Ruptured micronuclei that accumulated endogenous N4BP2 or N4BP2-GFP displayed a peripheral “rim” of DNA (DAPI or SiR-DNA) surrounding an apparently DNA-free core (Fig. 2A, F, G; Fig. S4A-B; Fig. S5A, B), a morphology confirmed by correlative light and electron microscopy (CLEM) (Fig. S6A, B).

To determine whether N4BP2 was sufficient to induce DNA damage, we generated a doxycycline-inducible lentiviral construct encoding N4BP2-GFP fused to a nuclear-localization signal (N4BP2-GFP-NLS) (Fig. S6C). Induction in N4BP2-knockout Dld1-Y^MN cells targeted N4BP2-GFP-NLS to otherwise intact nuclei and micronuclei (Fig. 2H; Fig. S6D, E), triggering robust γH2AX accumulation within 24–48 h in both asynchronous cultures and cells arrested in G1 with palbociclib (Fig. 2H; Fig. S6D, E). Forced nuclear targeting of N4BP2 also caused a rise in chromosome fragmentation and increased fragment number per mitotic spread (Fig. 2I-K; Fig. S6F), evidence that N4BP2 is sufficient to drive DNA damage and chromosomal shattering when present in intact nuclei.

Consistent with N4BP2-mediated DNA damage and micronuclear chromosome fragmentation, 24–48 h overexpression of N4BP2-GFP removed DNA from ruptured micronuclei and accelerated their loss (Fig. 2K; Fig. S6G). Entry of cytoplasmic nucleases including N4BP2 might initiate chromothripsis by shattering a micronucleated chromosome; any residual fragments, if incorporated into a daughter cell nucleus, can be re-ligated into an aberrant chromosome but without reacquiring a centromere and telomeres the outcome would be chromosome loss. The human Y chromosome, which is dispensable for viability and with the weakest human centromere (23–25), provided an ideal read-out: its loss could have resulted either from (1) missegregation of an intact micronucleated Y chromosome that was transmitted to only one of two daughter cells or (2) from nuclease-driven fragmentation and degradation (Fig. S7A). In clonal populations of Dld1-Y^MN cells expressing endogenous N4PB2 and passaged for >40 generations, the Y chromosome was completely lost (Fig. 2L; Fig. S7B, D, E), consistent with previous reports of frequent, spontaneous loss of the Y chromosome (23–25). In contrast, half of cells in N4BP2-knockout clones retained the Y chromosome (Fig. 2L). Moreover, after 10 days of induced Y centromere inactivation (confirmed in Fig. S7G), N4BP2 KO cells kept the Y chromosome at up to three-fold higher frequency than wild-type (WT) cells (Fig. S7B-E), reinforcing N4BP2’s role in eliminating micronucleated chromosomes.

N4BP2 promotes ecDNA generation

Chromosome bridges, formed during anaphase and persisting into interphase, have been implicated in initiating chromothripsis in cancer, as well as formation of extrachromosomal DNA (ecDNA) following bridge fragmentation (1, 2, 4, 10, 16, 26–28). These structures are frequently enclosed by aberrant nuclear envelopes (10, 15, 26, 29, 30), thus permitting entry of cytoplasmic nucleases. To test whether N4BP2 contributes to ecDNA formation, we used a model of ecDNA-mediated resistance to chemotherapy (26). High-dose methotrexate (MTX, 640 nM), an inhibitor of dihydrofolate reductase (DHFR) and long-used anticancer drug (31), induces DHFR amplification through two distinct mechanisms: 1) formation of DHFR-containing ecDNAs via fragmentation of chromosome bridges with amplified DHFR genes (26) or 2) amplification of DHFR genes within homogenously stained regions (HSRs) via breakage–fusion–bridge cycles (32) (Fig. 3A, Fig. S9A).

Figure 3. N4BP2 promotes ecDNA generation.

(A) Generation of HeLa cells that carry DHFR-encoding double minute/ecDNA providing resistance to high methotrexate (high MTX) concentrations. MTX-resistant HeLa clone with amplified DHFR gene within homogenously staining region (HSR, orange) was treated with high MTX concentration (640 nM) for 10 days, which provided selection pressure and promoted double minutes/ecDNA formation (shown as orange circles). Cells were transfected with indicated siRNAs at day 1 and day 5. N4BP2 is depicted as yellow circle, ruptured nuclear envelope as green dashed line. (B) Representative image of N4BP2-GFP localizing on a chromatin bridge (yellow arrows) in HeLa clone with amplified DHFR gene, zoomed-in view shown below. Scale bars, 10 μm and 2 μm, as indicated. (C) Representative image of HeLa cells from (A) showing co-localization of N4BP2-GFP with γH2AX (red) in the chromatin bridge. Scale bars, 10 μm and 2 μm, as indicated. (D) Immunoblot of N4BP2 using lysates of HeLa cells from (A), transfected with indicated siRNAs. Short and long exposures are shown. Tubulin was used as a loading control. (E) Percentage of HeLa cells from (A) with ecDNA, after induction of ecDNA formation with high (640 nM) concentration of MTX for 14 days, treated with indicated siRNAs. n=605 cells from 3 independent experiments. (F) Representative DNA-FISH images demonstrating ecDNA-positive mitotic spreads (marked with yellow arrows) upon treatment with NS siRNA vs ecDNA-negative mitotic spreads in N4BP2 siRNA condition. Scale bars 5 μm and 20 μm.

We showed that N4BP2 localized to chromosome bridges (Fig. 3B, C), where it colocalized with γH2AX (Fig. 3C). siRNA-mediated knockdown of N4BP2 (Fig. 3D) in cells exposed to MTX for 10 days led to more than a two-fold reduction in ecDNA-positive cells compared to those treated with non-silencing (NS) siRNA (Fig. 3E, F). Given that chromosome bridges remain physically connected to the main nucleus, N4BP2 may act not only on DHFR genes amplified in the bridge but also on nearby nuclear DNA by diffusing through the rupture site, potentially compromising cell viability. Supporting this view, cells treated with siN4BP2 showed a two-fold increase in cell number compared to NS siRNA or siTREX1-treated cells (Fig. S9B). Indeed, despite reduced levels of DHFR-containing ecDNAs, these cells retained a selective advantage by amplifying DHFR within DHFR-containing HSRs on chromosome 5 (Fig. S9C-E).

N4BP2 initiates genome rearrangements and chromothripsis

To directly test the role of N4BP2 in inducing chromosomal rearrangements, we transiently expressed low levels of N4BP2-GFP or GFP alone, concurrently with a 72-hour induction of Y chromosome missegregation in Dld1-Y^MN N4BP2 knockout cells (Fig. 4A). N4BP2-GFP, but not GFP alone, was expected to induce DNA damage within micronuclei, leading to Y chromosome shattering and, in rare cases, reassembly of chromosome fragments into a chromothriptic chromosome. Following transient expression of N4BP2-GFP or GFP, we isolated and expanded clonal cell lines from Dld1-Y^MN N4BP2 knockout cells and assessed genome integrity in each clone (n=88) (Fig. 4A-G; Fig. S10A-E) using FISH (Fig. 4C), spectral karyotyping (SKY) (Fig. 4D), whole genome sequencing (WGS) (Fig. 4E-G), and optical genome mapping (OGM) (Fig. S10C). In agreement with a role for N4BP2 in promoting genome instability, such rearrangements including translocations (Fig. 4C-E; Fig. S10B,C), tandem duplications (Fig. S10D,E) and chromothripsis (Fig. 4G) were observed exclusively in clonal isolates of N4BP2 knockout cells in which N4BP2-GFP had been transiently expressed (“N4BP2-GFP-derived clones”), but not in clones expressing GFP alone (“GFP-derived clones”) (Fig. 4B). FISH analysis of metaphase spreads using a Y chromosome-specific probe revealed abnormal signal patterns in N4BP2-GFP-derived clones, but not in GFP-derived clones (Fig. 4C; Fig. S10A). These findings were further corroborated by SKY (Fig. 4D; Fig. S10B, top panel), OGM (Fig. S10C), and WGS (Fig. 4E-G; Fig. S10D-E).

Figure 4. N4BP2 initiates genome rearrangements and chromothripsis.

(A) Schematic of a strategy used to assess genome rearrangements induced by N4BP2. Low levels of N4BP2-GFP or GFP were transiently expressed in Dld1-Y^MN, N4BP2 KO cells, micronuclei were induced by adding DOX/IAA for 72h. After single cell sorting (based on green GFP signal) and clonal expansion, clonal isolates were analyzed by a set of methods (shown below). FISH, Fluorescence In Situ Hybridization; SKY, Spectral Karyotyping; WGS, Whole Genome Sequencing; OGM, Optical Genome Mapping. (B) Y chromosome status and chromosomal rearrangements observed in clonal isolates produced in (A). (C-E) Representative FISH images (C) of a normal Y chromosome without visible defects on metaphase spreads of Dld1-Y^MN N4BP2 KO clonal isolate following transient GFP expression (left) and an example of derivative Y chromosomes (also shown with SKY analysis (D) and long read WGS (E) from Dld1-Y^MN, N4BP2 KO clonal isolate obtained after transient expression of N4BP2-GFP as shown in (A) (right panels). Y chromosome paint probes are green and Y centromere probes are red. DAPI is shown in blue. (F, G) WGS-enabled reconstruction of a pre-existing fusion [in a parental clone, (F)] between a nearly full length (196.8 Mb) chromosome 3 (Chr3) and an extra copy of the first 41.7 Mb of the p-arm of chromosome 1 (Chr1), producing a large, 238.5 Mb fusion chromosome. (G) N4BP2-GFP-derived clone, in which the first 41.7 Mb of the p arm of Chr1 had undergone focal chromothripsis with insertions at 3 different locations of pieces of Chr3 mapped adjacent to the initial Chr1/Chr3 fusion site. (H-K) Assessment of the role of N4BP2 in genome/chromosomal rearrangements in cancer using pan-cancer genomics. (H) Independent effect size (x-axis) of chromothripsis and copy number segmentation (y-axis) on N4BP2 expression in a multivariate regression model. (I) Chromothripsis is enriched in tumor samples with N4BP2 copy number gains, while compared with the association of TP53 mutation status (J) (WT=no mutation, MUT=LOH, homozygous deletion or driver point mutation/indel) and chromothripsis status across TCGA (two-sided Fisher’s exact test). (K) Example of structural variant profile from samples with high expression of N4BP2, exhibiting characteristic signs of ecDNA. Arcs indicate rearrangements between two regions of the genome, TLOC=translocation (green), INV=inversion (brown), DUP=tandem duplication (purple), DEL=deletion (red).

Furthermore, WGS of the parental Dld1-Y^MN clone revealed a pre-existing fusion between a nearly full-length chromosome 3 (196.8 Mb) and an extra copy of the first 41.7 Mb of the p-arm of chromosome 1, resulting in a 238.5 Mb fusion chromosome (Fig. 4F, “parental chromosome”; Fig. S10B, bottom panel). As larger chromosomes are known to missegregate and be incorporated into micronuclei at high frequency (33, 34), the presence of an immobile, centromereless Y chromosome after induction of micronuclei formation is expected to further hinder the segregation of such large chromosomes. Consistent with this, we identified an N4BP2-GFP-derived clone which, following induced inactivation of the Y centromere, had not only lost the Y chromosome but also exhibited ‘focal’ chromothripsis of the first 41.7 Mb of the Chr1 p-arm (Fig. 4G, “derivative clone with chromothripsis”).

N4BP2 is associated with chromosome rearrangements in cancer

To further investigate the link between N4BP2 and chromosome rearrangements, we analyzed publicly available cancer genome datasets, including 9,691 samples from The Cancer Genome Atlas (TCGA) (35) and 667 samples from the Pan-Cancer Analysis of Whole Genomes (PCAWG) Consortium (4). Across multiple cancer types, N4BP2 copy number gains were more frequent than loss-of-function mutations (Fig. S11A). Notably, tumors with high N4BP2 expression in the TCGA dataset exhibited a significantly increased prevalence of chromothripsis (Fig. 4H; Fig. S11B), with N4BP2 copy number gains associated with a 3.7-fold increase in chromothripsis (p=3.8*10^-8; Fig. 4I). This association was stronger than that observed for TP53 disruption, the hallmark tumor suppressor gene, which showed a lower odds ratio of 2 (Fig. 4J).

Furthermore, we identified a strong association between elevated N4BP2 expression and an increased number of genomic segments, indicating that tumors with higher N4BP2 levels harbor more complex genomes. This association remained significant after adjusting for both N4BP2 copy number and overall genome aneuploidy (Fig. S11C). In addition, tumors with high N4BP2 expression showed a significant increase in the proportion of translocations across the genome (median fold change=1.26, p=0.02, two-sided Mann-Whitney test; Fig. S11D). Finally, elevated N4BP2 expression was associated with a higher prevalence of genomic amplicons, including ecDNA (odds ratio=1.38; p=0.016) (Fig. S11E). Tumors with high N4BP2 levels often contained multiple ecDNAs characterized by focal, high-level amplifications (up to 30 copies) across multiple chromosomes. For example, ecDNAs encoding SOX2 (derived from chromosome 3) were prominent in a lung squamous cell carcinoma with high N4BP2 levels (Fig. 4K, see also Fig. S11F-G).

In addition, we performed a comprehensive analysis using the Hartwig Medical Foundation dataset (36), a large, well-curated resource containing whole-genome sequencing and matched RNA-sequencing data from 3,000 metastatic and primary tumors across diverse cancer types. To test the specificity of N4BP2's association with chromothripsis, we constructed a multivariable model including other genes previously implicated in chromothripsis, such as TREX1, APEX1 (encoding APE1), and TP53. We used both structural equation modeling (SEM) with the weighted least squares mean and variance adjusted estimator (WLSMV), as well as a uniform modeling framework based on generalized linear models (GLMs). (Fig. S12, Fig. S13). In addition, we analysed the TCGA cohort, which is whole-exome sequenced, using GLMs (Fig. S14). N4BP2 expression remained significantly associated with chromothripsis in both Hartwig (Fig. S12, Fig. S13) and TCGA (Fig. S14), independent of TP53 status, overall genome complexity, background genomic instability (BGI) score, and tumor type. These results support the hypothesis that elevated N4BP2 promotes chromosome fragmentation beyond normal expression levels, aligning closely with our experimental data showing that increased N4BP2 activity drives DNA damage, chromosome shattering, and ecDNA formation.

N4BP2-induced DNA damage and ecDNA formation in induced gliomas

Genome rearrangements and ecDNA formation are common features of many human cancers, including glioblastomas (GBM), where they are associated with drug resistance and tumor evolution (26, 37–39). To test whether N4BP2 contributes to DNA damage, chromosome fragmentation, and ecDNA formation in vivo, we used an induced high-grade glioma model (37) which carries deletion of the TP53 gene encoding the p53 tumor suppressor and an activating mutation in the gene encoding the platelet-derived growth factor receptor alpha (PDGFRA) (deletion of exons 8 and 9a, referred to as PDGFRA^Δ8–9). This model closely mimics the aggressive, rapidly growing, and invasive behavior characteristic of human glioblastoma (40).

We generated N4BP2 KO, TREX1 KO, and WT human neural/glial progenitors (“iNPCs”) from pluripotent stem cell clones containing TP53 KO/PDGFRA^Δ8–9 driver mutations (Fig. 5A, Fig. S15C). These progenitors were engrafted into the brains of Nod-scid mice, resulting in the formation of primary glioma tumors (“primary tumor spheres”) (37). Both iNPCs and primary tumor spheres formed micronuclei spontaneously, with an increased frequency following transient exposure to nocodazole. These micronuclei exhibited ruptured nuclear envelopes, which were indicated by abnormal Lamin B1 staining (Fig. S15A) (41), developed extensive γH2AX-positive DNA damage (Fig. S15A, B), and accumulated N4BP2 (Fig. S15B). Loss of N4BP2, but not TREX1, significantly reduced both the production of micronuclei and the occurrence of DNA damage in ruptured micronuclei in iNPCs and primary tumor spheres, while decreasing the presence of extrachromosomal DNA (ecDNA) (Fig. 5B-F; Fig. S15D-F).

Figure 5. N4BP2-induced DNA damage and ecDNA formation in induced gliomas.

(A) Schematic of induced human high-grade glioma generation. Induced pluripotent stem cells (iPSCs) with a combination of driver mutations commonly altered in human glioblastomas were genetically modified (WT, N4BP2 KO, TREX1 KO) and differentiated into neural/glial progenitor cells, to be intracranially injected into mouse brain to form primary tumors. (B) Schematic of the approach used to induce MN in iNPCs and primary tumor spheres [generated as shown in (A)], followed by immunofluorescence analysis of fixed cells (C, D; S12D, E) and DNA-FISH analysis of metaphase spreads (E; S12F). (C) Percentage of γH2AX-positive MN in the primary tumor spheres in the absence of N4BP2 or TREX1. n=3 independent experiments per each genotype, 970 cells. 3 independent experiments per each condition. (D) Quantification of the fraction of γH2AX-positive ruptured micronuclei (aberrant Lamin B1 staining) in primary tumor spheres in the presence and absence of N4BP2 or TREX1. 2 independent experiments per each genotype, n=300 micronuclei (control), n=350 micronuclei (N4BP2 KO) and n=320 micronuclei (TREX KO). 3 independent experiments per each condition. (E) Quantification of metaphase spreads containing ecDNA in primary tumor spheres in the presence and absence of N4BP2 or TREX1. 3 independent experiments per each genotype, n=77 spreads (control), n=406 spreads (N4BP2 KO) and n=197 spreads (TREX KO). 3 independent experiments per each condition. (F) Representative images of metaphase spreads in iHGG model showing ecDNA (yellow arrowheads), genotypes as indicated. Scale bar, 10 μm. (B, C) DNA was stained with DAPI (blue). (G) Quantification of tumor area on mouse brain sections in the presence and absence of N4BP2 or TREX1 (n = 5 mice for control, n = 5 mice for N4BP2 KO, and n = 4 mice for TREX1 KO) (H) Number of dividing tumor cells (Ki67-positive, normalized to wild type) in the presence or absence of N4BP2 or TREX1. Sample sizes: n = 3 mice (control), n = 4 mice (N4BP2 KO), and n = 6 mice (TREX1 KO). (I) Representative composite images of brain sections showing iHGG tumors, with human nuclei (HuNuc) stained in purple and tumor boundaries indicated by dashed lines. Genotypes as indicated.

To directly evaluate the impact of N4BP2 on tumor growth and aggressiveness, we analyzed tissue sections from tumors formed by WT, N4BP2 KO, and TREX1 KO cells evaluating both tumor size and the proliferation index of human cancer cells within each tumor. These analyses demonstrated that loss of N4BP2 resulted in less aggressive tumors, characterized by reduced tumor size and fewer proliferating cells compared to tumors formed from WT or TREX1 KO cells (Fig. 5G-I).

Discussion

Genome rearrangements are hallmarks of human cancers (1, 42, 43). Mechanistically, rearrangements can be initiated by DNA double stranded breaks, leading to chromosome translocations, chromothripsis (the major type of chromoanagenesis (44)), lost chromosomes and, in case of circularization of the small remaining fragments, generation of ecDNA. Even as simple an event as a single break, if unrepaired prior to mitosis, can be sufficient to initiate a chromothriptic cascade in which an acentric chromosome fragment is missegregated into a micronucleus (45, 46). Spontaneous rupture of the micronuclear envelope exposes the previously encapsulated chromatin to cytoplasmic nucleases and chromosome shattering. Here, using an unbiased cellular screen and multiple cancer cell models, we have identified the largely uncharacterized N4BP2 nuclease to accumulate in foci within ruptured micronuclei (or the main nucleus following envelope breakage in interphase) and induce shattering of cytoplasm exposed chromosomes, thus initiating multiple rearrangements that can include loss of whole chromosomes, focal chromothripsis, and production of ecDNA. Although N4BP2’s structure and in vitro activity have been partially characterized (47, 48), future studies are needed to determine its precise nuclease and/or ATPase activity(ies).

Conceptually, the identification of N4BP2 entry and enhancement of chromatin fragmentation within ruptured micronuclei adds mechanistic support for an ‘all-at-once’ chromosome shattering (as initially proposed (2)). In the broader landscape of human cancer genomes, N4BP2 amplifications were found in 19 cancer types, with high N4BP2 expression associated with increased prevalence of chromothripsis, translocations, and amplicons including ecDNA. Loss of N4BP2 in human induced high-grade glioma model led to decreased tumor size and reduced number of micronuclei. Identification of N4BP2 adds a new target for potential therapeutic interventions in cancer, with N4BP2 inhibition likely to disrupt initiation of genome rearrangements, including production of ecDNA, with enhancement of its activity expected to selectively target cancer cells with unstable genomes.

Data and materials availability:

All sources of data, codes and materials are provided in corresponding sections of Methods. All materials are available under materials transfer agreement for noncommercial replication or extension of this work, upon request to corresponding authors. All other information is available in the manuscript or the supplementary materials.