Trabectedin derails transcription-coupled nucleotide excision repair to induce DNA breaks in highly transcribed genes

Abstract

Most genotoxic anticancer agents fail in tumors with intact DNA repair. Therefore, trabectedin, anagent more toxic to cells with active DNA repair, specifically transcription-coupled nucleotide excision repair (TC-NER), provides therapeutic opportunities. To unlock the potential of trabectedin and inform its application in precision oncology, an understanding of the mechanism of the drug's TC-NER-dependent toxicity is needed. Here, we determine that abortive TC-NER of trabectedin-DNA adducts forms persistent single-strand breaks (SSBs) as the adducts block the second of the two sequential NER incisions. We map the 3'-hydroxyl groups of SSBs originating from the first NER incision at trabectedin lesions, recording TC-NER on a genome-wide scale. Trabectedin-induced SSBs primarily occur in transcribed strands of active genes and peak near transcription start sites. Frequent SSBs are also found outside gene bodies, connecting TC-NER to divergent transcription from promoters. This work advances the use of trabectedin for precision oncology and for studying TC-NER and transcription.

Fig. 1. Trabectedin induces TC-NER-dependent DNA strand breaks in G1 cells.

a Trabectedin and its DNA adduct structure (10.5452/ma-c4e6e) rendered using PyMol. b HAP1 or U2OS WT, XPC-, XPA-, and CSB-KO cells were treated with trabectedin or DMSO for 2 h, and colony counted after 8 days. Mean ± SEM of 3 and 2 biological replicates for HAP1 and U2OS, respecively (3 technical replicates per experiment). P-values of ordinary two-way ANOVA with Dunnett’s multiple comparisons test (between WT and mutants at each concentration) are provided. c Scheme for assessing NER incision activity following DNA damage by alkaline COMET chip assays. d HAP1 WT and e XPA-KO cells were arrested in G1 with palbociclib and treated with trabectedin (50 nM, 2 h) and allowed to recover for up to 6 h with or without repair synthesis inhibitors (0.5 mM HU, 5 μM AraC). ssDNA breaks were analyzed by alkaline COMET chip assays. f U2OS WT, g XPC-, h CSB- and i XPA-KO cells were arrested in G1 with palbociclib (1 µM, 24 h) and treated with trabectedin (50 nM, 2 h), and allowed to recover for up to 4 h with or without repair synthesis inhibitors (1 mM HU, 10 μM AraC). ssDNA breaks were analyzed by alkaline COMET chip assays. d–i Each dot represents DNA in tail (%) of a comet analyzed. Each box represents the mean value of DNA in tail (%) from all comets used in all experiments. The number of comets used is provided above each box. An error bar represents SD. j Summary and statistical analysis of COMET chip experiments in HAP1 WT and XPA-KO cells. Mean ± SEM of 4 biological replicates. P-values of two-tailed paired t-tests (between WT and XPA-KO at each recovery time) are provided. Mean values of individual experiments (shown as boxes in panels d, e) served as the input data. k Summary and statistical analysis of COMET chip experiments in U2OS cells. Mean ± SEM of 4 (WT, XPC-KO), 3 (CSB-KO) biological replicates. No error bars for DDB2-KO and XPA-KO (n = 1). P-values of ordinary two-way ANOVA with Dunnett’s multiple comparisons test (between WT and XPC- or CSB-KO at each recovery time) are provided. Mean values of individual experiments (shown as boxes in panels f–i) served as the input data. Source data are provided as a Source Data file.

Fig. 2. Trabectedin-induced DNA break formation and toxicity depend on the catalytic activity of XPF but not that of XPG.

a HAP1 WT, ERCC1-KO, and XPF-D687A cells were treated with trabectedin for 2 h and colonies counted after 7 days. Mean ± SEM of 5 (WT, XPF-D687A) and 3 (ERCC1-KO) biological replicates (3 technical replicates per experiment). Provided p-values are derived using ordinary two-way ANOVA with Dunnett’s multiple comparisons test (between WT and ERCC1-KO or XPF-D687A at each concentration). b HAP1 WT, ERCC1-KO, and XPF-D687A cells were arrested in G1 with palbociclib and treated with trabectedin (50 nM, 2 h) and incubated for up to 4 h with or without repair synthesis inhibitors (0.5 mM HU, 5 μM AraC). ssDNA breaks were analyzed by alkaline COMET chip assays. Mean ± SEM of 3 biological replicates (Supplementary Fig. 2c). Provided P-values are derived using two-tailed paired t-test (between WT and ERCC1-KO or XPF-D687A at each recovery time). c XP2YO patient cells were treated with trabectedin for 2 h and colonies were counted after 8 days. Mean ± SEM of 2 biological replicates (3 technical replicates per experiment). d U2OS WT and XPF-KO cells were arrested in G1 with palbociclib and treated with trabectedin (50 nM, 2 h) and allowed to recover for up to 4 h with or without repair synthesis inhibitors (1 mM HU, 10 μM AraC). ssDNA breaks were analyzed by alkaline COMET chip assays. Mean ± SEM of 4 biological replicates. P values were derived using ordinary two-way ANOVA with uncorrected Fisher’s LSD (between WT and XPF-KO at each recovery time). e HAP1 WT, XPG-KO, and XPG-E791A cells were treated with trabectedin for 2 h and colonies counted after 7 days. Mean ± SEM of 4 biological replicates (3 technical replicates per experiment). Provided P-values are derived using ordinary two-way ANOVA with Dunnett’s multiple comparisons test (between WT and XPG-KO or XPG-E791A at each concentration). f HAP1 WT, XPG-KO, and XPG-E791A cells were arrested in G1 with palbociclib (2 µM, 24 h) and treated with trabectedin (50 nM, 2 h). Cells were kept in G1 with or without repair synthesis inhibitors (0.5 mM HU, 5 μM AraC) and incubated for up to 4 h. ssDNA breaks were analyzed by alkaline COMET chip assays. Mean ± SEM of 5 biological replicates (Supplementary Fig. 3d–e). Provided P-values are derived using ordinary two-way ANOVA with Tukey’s multiple comparisons test (between WT and XPG-KO or XPG-E791A at each recovery time). g A simplified schematic of assessing NER incision activity on trabectedin-induced DNA adducts in HAP1 WT, XPF-D687A, ERCC1-KO, XPG-E791A, and XPG-KO cells using alkaline COMET chip assays. Source data are provided as a Source Data file.

Fig. 3. Trabectedin-induced DNA breaks are mapped genome-wide with upgraded GLOE-Seq.

a Principle of GLOE-Seq that maps DNA breaks in a genome-wide and strand-specific fashion and provides an estimate of the frequency of individual breaks in a population of cells. The distal adapter with a unique molecular identifier (UMI) is an upgrade from the original GLOE-Seq protocol. Pos: position, chr: chromosome, OH: free 3’ hydroxyl. b DNA break count along chromosome 19 in 4 cell lines after 2h exposure to trabectedin or vehicle (DMSO) and subsequent 2 h recovery. Solid lines: individual biological replicates, 2 for 50 and 0 nM drug in WT, 50 nM in CSB-KO; 1 for 0 nM in CSB-KO; 3 for 50 and 0 nM in XPC-KO and XPA-KO. Vertical bar heatmap: gene expression level in unexposed U2OS WT. All data are shown per 100-Kb bin. We summed DNA-break counts within this chromosome and provided the min-max range of this value across replicates for either treatment condition (shown also in Supplementary Fig. 4d). P-value: Mann-Whitney U test with the one-sided alternative hypothesis that the trabectedin-treatment-related distribution is stochastically greater than the control distribution (see more details in Supplementary Fig. 4e). TPM, transcripts per million transcripts. Arb. unit, arbitrary unit: Methods describe DNA break count normalization. c Genome-wide correlation of DNA break count with the abundance of DNase I hypersensitivity (HS) sites (transcriptional activity), H3K4me3 (active gene promoters), H3K4me1 (active enhancers) and H3K27ac (active promoters and enhancer) as well as gene expression. Bars: mean, markers: biological replicates (n = 2) of break mapping. N = 28,513 genomic bins to compute the correlation. Source data are provided as a Source Data file.

Fig. 4. Trabectedin-induced DNA break counts correlate with gene expression levels.

DNA break count on each strand of protein-coding genes in U2OS WT (a), CSB-KO (b), XPC-KO (c) and XPA-KO (d) after 2 h exposure to trabectedin and subsequent 2 h recovery versus gene expression level in unexposed U2OS WT. ρ: Spearman correlation coefficient calculated for continuous (not tiered) data of gene expression and DNA break count (correlation analysis for all replicates in Supplementary Fig. 5h), n = 16,740. Heatmaps: the color of a hexagonal bin reflects the gene count. TPM, transcripts per million transcripts. Box plots: gene expression was categorized in tiers via percentiles shown by vertical lines in the heatmaps; boxes are interquartile ranges, internal horizontal lines are medians, circular markers are means, whiskers extend to the furthest datapoint within 1.5x interquartile range, datapoints beyond are shown as small markers. Number of genes per tier and number of genes beyond the maximal y-axis value are provided in Methods. Presented data: one biological replicate of drug exposure per cell line. Mean DNA break count of protein-coding genes in U2OS WT (e), CSB-KO (f), XPC-KO (g) and XPA-KO (h) after 2h exposure to trabectedin or vehicle and subsequent 2h recovery versus gene expression level in unexposed U2OS WT. Presented data: multiple biological replicates. Gene expression tiers are described in a–d. Number of genes per tier are provided in Methods. The internal plots zoom in on the y-axis. a-h: arb. unit, arbitrary unit: Methods describe DNA break count normalization. Source data are provided as a Source Data file.

Fig. 5. DNA-break profiles of genes with highest trabectedin-induced damage.

Strand-specific profiles of individual genes in the indicated cell lines after 2h exposure to trabectedin or DMSO and subsequent 2h recovery. Genomic coordinates were adjusted such that TSS is zero and gene directionality is rightward. Bin length: 1 kilobase. Bar: mean across biological replicates, n = 2 for WT 50 nM drug, WT DMSO, CSB-KO 50 nM, n = 3 for XPC-KO 50 nM and XPA-KO 50 nM. Kb: kilobase. We present protein-coding genes with highest trabectedin-induced damage within 5 Kb downstream of TSS in U2OS WT. PLEKHM1 is in top 50% expressed genes of U2OS WT, the other genes are in top 30%. Source data are provided as a Source Data file.

Fig. 6. TRABI-seq detects divergent transcription and provides evidence for XPF sequence preference.

Strand-specific profile of the mean DNA break count and its 95% c.i. (shade) throughout the gene body and adjacent regions in U2OS WT (a), CSB-KO (b), XPC-KO (c) and XPA-KO (d) after 2h exposure to trabectedin and subsequent 2h recovery. n = 4425 protein-coding genes (top 30% expressed in unexposed U2OS WT) are considered to compute the means and c.i. Solid, dashed, and dotted curves: means of different biological replicates. Strand- and gene-length-specific profile of the mean DNA break count and its 95% c.i. (shade) in the ±5 kilobase (Kb) proximity of TSS in TC-NER proficient cell lines U2OS WT (e) and XPC-KO (f), zooming out (left panel) and in (right). The same gene set as in a–d. Methods provide gene numbers per gene-length group. DNA break count in two branches of divergent transcription in U2OS WT (g) and XPC-KO (h) versus gene expression. +5 Kb: within 5 Kb downstream of; −5 Kb: within 5 Kb upstream of. The plots are built analogously to Fig. 4a–d (lower panels). Supplementary Fig. 6g–h presents respective correlation analysis for all replicates. Gray band: endogenous DNA breaks not caused by trabectedin treatment (upper quartile of DNA break count in unexpressed genes); this threshold shows that around 25% (lower boundary of boxes) of highly expressed genes may not have trabectedin-induced breaks upstream of the TSS. Sequence logos around DNA breaks in U2OS WT (i) and XPC-KO (j). We considered DNA breaks located in the indicated regions of the gene set used in a–f. The percentage of G at position 1 (+2 relative to the break) is shown. Data: all biological replicates united per cell line. Supplementary Fig. 6i–j: analogous analysis for TC-NER-deficient cell lines. a–f: bin sizes are absolute (a base number) or relative (a percentage of gene length; the corresponding average base number indicated in parentheses). a–h: arb. unit: Methods describe DNA break count normalization. a–j TSS and TES: transcription start and end sites; Kb: kilobase; b: base. Source data are provided as a Source Data file.

Fig. 7. Summary of the mechanism of trabectedin-induced TC-NER-mediated SSB formation and toxicity, and the development of TRABI-Seq.

Trabectedin-DNA adducts are exclusively recognized by TC-NER, and not by GG-NER. The adducts then block the incision of the XPG endonuclease, causing persistent XPF-mediated breaks. Mapping those breaks in a genome-wide fashion displays where TC-NER is active: predominantly in highly transcribed genes and their upstream regions due to a link to divergent transcription.