CDK12 controls transcription at damaged genes and prevents MYC-induced transcription-replication conflicts

Curti, Laura; Rohban, Sara; Bianchi, Nicola; Croci, Ottavio; Andronache, Adrian; Barozzi, Sara; Mattioli, Michela; Ricci, Fernanda; Pastori, Elena; Sberna, Silvia; Bellotti, Simone; Accialini, Anna; Ballarino, Roberto; Crosetto, Nicola; Wade, Mark; Parazzoli, Dario; Campaner, Stefano

doi:10.1038/s41467-024-51229-5

Download PDF

Article
Open access
Published: 18 August 2024

CDK12 controls transcription at damaged genes and prevents MYC-induced transcription-replication conflicts

Nature Communications volume 15, Article number: 7100 (2024) Cite this article

7873 Accesses
13 Citations
1 Altmetric
Metrics details

Subjects

Abstract

The identification of genes involved in replicative stress is key to understanding cancer evolution and to identify therapeutic targets. Here, we show that CDK12 prevents transcription-replication conflicts (TRCs) and the activation of cytotoxic replicative stress upon deregulation of the MYC oncogene. CDK12 was recruited at damaged genes by PARP-dependent DDR-signaling and elongation-competent RNAPII, to repress transcription. Either loss or chemical inhibition of CDK12 led to DDR-resistant transcription of damaged genes. Loss of CDK12 exacerbated TRCs in MYC-overexpressing cells and led to the accumulation of double-strand DNA breaks, occurring between co-directional early-replicating regions and transcribed genes. Overall, our data demonstrate that CDK12 protects genome integrity by repressing transcription of damaged genes, which is required for proper resolution of DSBs at oncogene-induced TRCs. This provides a rationale that explains both how CDK12 deficiency can promote tandem duplications of early-replicated regions during tumor evolution, and how CDK12 targeting can exacerbate replicative-stress in tumors.

An mRNA expression-based signature for oncogene-induced replication-stress

Article 29 January 2022

Deep profiling of gene expression across 18 human cancers

Article 17 December 2024

Enhancing transcription–replication conflict targets ecDNA-positive cancers

Article Open access 06 November 2024

Introduction

Replication stress (RS) is a major hallmark of cancer cells and a source of genomic instability¹. During tumor progression, mutations in oncogenes and tumor suppressor genes can induce RS by imposing a state of high and deregulated DNA replication, which increases the chances for fork collapse and generation of DNA breaks². In this scenario, fork collapse has been linked to multiple causes, including the depletion of nucleotides, mis-incorporation of ribonucleotides into nascent DNA, inefficient activation of protective fork processing pathways, and conflicts arising from the encounter of replication and transcription complexes^3,4,5.

RS is considered to have a dual role in cancer evolution: on one hand, it can act as a driving factor by providing a mutation-prone background that favors further selection of driver mutations; on the other, it can trigger tumor suppressive responses such as apoptosis or cellular senescence⁶. In line with the latter, the exacerbation of RS in cancer cells can offer therapeutic opportunities, either in combinatorial treatments, to enhance the effectiveness of current chemotherapeutic drugs, or in targeted therapies exploiting synthetic-lethal dependencies in genetically defined tumors^7,8,9. Thus, identifying factors and processes that regulate RS in cancer cells is a subject of active studies and has high translational potential.

The c-MYC proto-oncogene is frequently overexpressed in cancer cells, due to genomic alterations (e.g., translocation, amplification) or oncogenic mutations in upstream regulatory pathways^10,11. Its product, MYC, is a transcription factor belonging to the bHLH-LZ family. It dimerizes with another bHLH-LZ protein, MAX, to bind DNA in a sequence-specific manner and to control the transcription of genes linked to cell proliferation and metabolism¹¹. Oncogenic levels of MYC lead to increased binding to proximal and distal regulatory elements within the genome, a phenomenon dubbed “enhancer-promoter invasion”, that pairs with broadened transcriptional regulation^12,13,14,15. The essentiality of MYC in cancer cells and its temporal dispensability in normal tissues¹⁶, makes it an attractive and potentially universal therapeutic target¹⁷, whether through direct inhibition of MYC activity¹⁸, or through exploitation of MYC-induced dependences¹⁸.

While still under investigation, evidence shows that MYC can trigger RS by promoting the processivity of DNA synthesis, increasing origin firing, anticipating S-phase entry and possibly by inducing transcription-replication conflicts (TRCs)^19,20,21. Several lines of evidence indicate that MYC-overexpressing cells are addicted to genetic dependencies that curb RS to avoid rampant genomic instability and cell death. These include the ATR/CHK1 kinases^22,23,24,25, fork remodelers and regulators^26,27,28, and factors preventing transcriptional interference^{25,29,30,31,32,33}. Altogether, these pathways may cooperate to prevent the collapse and/or excessive processing of stalled replication forks.

To identify factors regulating genomic and genetic stability in MYC overexpressing cells we conducted a loss of function genetic screen. Among other genes, we identified CDK12 as being selectively required to prevent RS and cytotoxic DDR. CDK12 is a member of the cyclin dependent kinase family that associates with and is activated by Cyclin K (CCNK)³⁴. Like other transcriptional CDKs, CDK12 - and its paralogue CDK13 - can phosphorylate and activate RNA polymerase II (RNAPII)³⁵. In addition, CDK12 regulates the expression and processing of genes linked to cell cycle progression and DNA damage signaling and repair^36,37,38. In line with this, inhibition of CDK12 sensitizes cancer cells to DNA-damaging agents³⁴. CDK12 is also recurrently mutated in several cancers, and an actionable therapeutic target^34,39.

Here, we report that CDK12 is recruited onto transcribed and damaged genes to repress transcription. This depends on PARP, engaged by DDR signaling, and the elongation factor SPT5. Inhibition or silencing of CDK12 unleashes transcription at damaged genes and exacerbates TRCs leading to DSBs between early replicating regions and the promoters of genes transcribed co-directionally to the replication fork. Our work unveils a new role for CDK12 in controlling transcription and genomic stability and reveals how management of TRCs is a major liability in MYC-driven tumors.

Results

Identification of genes controlling genome stability and viability in MYC-overexpressing cells

To identify modulators of genomic stability and cell viability in MYC-overexpressing cells, we devised a high-content siRNA screen in Rosa26-MycER mouse embryo fibroblasts, where conditional activation of MycER mimics the oncogenic properties of c-MYC²⁷. These cells were generated from homozygous knock-in embryos carrying a gene encoding a MYC-estrogen receptor chimera (MycER) that can be activated by 4-hydroxytamoxifen (OHT)⁴⁰. We screened this cell line with a custom siRNA library (DDR-library) targeting 1196 genes previously implicated in genome integrity⁴¹, and a commercial library (druggable-library) targeting 1400 druggable genes. For the screen, cells were retro-transfected in 384-well plates, with each well containing a siRNA mix targeting one gene. Forty-eight hours post-transfection, cells were analyzed by immunofluorescence for a parallel quantification of DDR signaling (based on the integration of single-cell γH2AX intensities) and cell growth (by cell counting), both in mock (no OHT) and MycER-activated cells (plus OHT).

For each gene, we calculated the average normalized differential values (Z-score) for both DDR and cell growth (Fig. 1a, Supplementary Fig. 1a, b, d, e and Supplementary dataset 1) and identified as hits those genes that, when silenced, would preferentially affect DDR (DDR hits) and/or cell growth (viability hits). We scored a total of 629 genes: 386 DDR-hits and 335 viability-hits; of these 92 were both DDR and viability hits (Fig. 1b and Supplementary Fig. 1c, f). A subset of the primary hits was cross-validated in a secondary screen (Supplementary Fig. 1b, c, f and Supplementary dataset 1). Gene ontology and protein-protein interaction network analyses with Metascape⁴², revealed a prevalence of genes involved in RNA processing, translation, protein degradation, cell cycle control and signaling (Fig. 1c, Supplementary Fig. 2 and Supplementary dataset 2). This confirmed previously identified MYC-dependencies, such as its reliance on transcriptional regulators⁴³, the splicing machinery^44,45, regulators of mitosis^46,47, nucleotide biosynthesis^48,49, and proteasome activity⁵⁰.

**Fig. 1: CDK12 is synthetic lethal with Myc activation.**

Among the hits confirmed in the secondary screen (Supplementary Fig. 1b), we identified CDK12, a kinase implicated in transcriptional control and genome stability, which has recently drawn attention as a potential therapeutic target in cancer^34,39. Most noteworthy, CDK12 also scored in a previous list for MYC-synthetic lethal candidates, but without further characterization⁴⁷. In addition, re-analysis of a recently published screen⁵¹ revealed that CDK12, its paralogue CDK13, and its regulatory CCNK are synthetic lethal with MYC amplification in medulloblastoma (Supplementary Fig. 3).

Genetic or chemical targeting of CDK12 impairs cell viability and triggers a DNA damage response in MYC-overexpressing cells

For further validation and mechanistic analysis, we engineered U2OS cells expressing the MycER chimera²⁷ (U2OS-MycER cells) with doxycycline-inducible shRNAs targeting CDK12 (shCDK12 #1 and #2) or a non-targeting shRNA (shNT) (Supplementary Fig. 4a). CDK12 knock-down (CDK12-KD) impaired cell growth, with a stronger inhibition when MycER was activated (Supplementary Fig. 4b, c). On the other hand, cell viability was reduced only when cells experienced both CDK12-silencing and MycER activation (Fig. 1d), indicating a strong synthetic-lethal phenotype. This was paralleled by enhanced DNA damage response (DDR), engaging both ATM and ATR signaling, as indicated by WB analysis (Fig. 1e). Synthetic lethality and synthetic DDR activation were also confirmed by siRNA mediated silencing (Supplementary Fig. 5) and upon CDK12 inhibition by THZ531 or dinaciclib (Fig. 1f and Supplementary Fig. 6). Similar to CDK12, silencing of CDK13 or CCNK lead to synthetic lethality and DDR upon MycER activation (Supplementary Fig. 7). Next, we evaluated whether CDK12 inhibition would induce DDR in Burkitt’s lymphoma and multiple myeloma cell lines in which expression of MYC is deregulated by chromosomal translocations. Most of the cell lines tested were sensitive to THZ531, with IC50 values in the nanomolar range (Supplementary Fig. 8a). Similar to what was observed in U2OS-MycER cells, THZ531 induced the activation of a DDR characterized by a recurrent increase of phospho-KAP1 and γH2AX signals (Fig. 1g). The only exception was the RAMOS cell line, which had the highest IC50 for THZ531 and did not display an increase in phospho-KAP1 or γH2AX following CDK12 inhibition, but instead showed increased phosphorylation of CHK1 (Supplementary Fig. 9). Other DDR markers had inconsistent fluctuations when these cell lines were treated with THZ531 (Supplementary Fig. 9), possibly suggesting that the engagement of select DDR pathways may depend on the genetic background or other context-specific properties of the cancer cells. In all the cell lines tested CDK12 inhibition also led to an altered cell cycle distribution characterized by a decrease in BrdU-labeled cells, indicating reduced DNA synthesis and altered cell cycle progression (Supplementary Fig. 8b).

Genome-wide transcriptional alterations upon CDK12 silencing

CDK12 has been implicated in transcriptional regulation as it can phosphorylate RNAPII on serine 2 of the carboxyl-terminal repeat domain (CTD), thus stimulating elongation³⁵. It was also reported to prevent early transcriptional termination at cryptic poly-A sites and to regulate alternative splicing^37,38,52. Thus, we assessed how CDK12 silencing would affect the transcriptome in our cells.

RNA-seq analysis showed that CDK12-KD led to both up- and downregulation of distinct sets of genes (DEG-up and -down, Supplementary Fig. 10a). These alterations in mRNA expression were coherent with phospho-Ser-2 RNAPII ChIP-seq data, with consistent increases and decreases in the elongating form of RNAPII on DEG-up and DEG-down genes, respectively (Supplementary Fig. 10b), while the global level of Ser2-Pi RNAPII was not affected by CDK12 silencing (Supplementary Fig. 10c), confirming previous observations^35,37,38. Most noticeably, DEG-down genes tended to be more expressed in unchallenged cells and had a shorter size, while the opposite was true for DEG-up genes (Supplementary Fig. 10d, e and Supplementary dataset 3). We also evaluated whether CDK12 silencing would broadly alter the expression of DDR genes. The top 20 GO enrichment terms of DEG did not include terms associated with genome stability or DNA repair (Supplementary Fig. 10f, g and Supplementary dataset 3). The analysis of a custom collection of DDR-associated genes did not reveal global repression of these genes (Supplementary Fig. 11a-c and Supplementary dataset 3). Exon level analysis of the mRNA processing of long-DDR genes failed to show any consistent or overt alterations in splicing or premature termination (Supplementary Fig. 11d and 12). Overall, the lack of pervasive regulation of DDR genes following CDK12 silencing suggests that the increased DDR observed upon MYC activation in CDK12-KD cells may not be solely ascribed to altered expression or processing of some DDR genes. In addition, CDK12-KD did not alter MYC-dependent transcription, since MYC targets were still efficiently activated when CDK12 was silenced (Supplementary Fig. 13 and Supplementary dataset 3).

CDK12 represses RNA synthesis at damaged genes

Given the reported role of CDK12 in regulating RNAPII activity, we asked whether silencing of CDK12 might induce a global change in RNA synthesis rate, not necessarily detected by steady state expression analyses (RNA-seq). To this end, we pulsed labeled cells and evaluated RNA synthesis by quantitative immunofluorescence (IF) analysis of 5-ethynyluridine (EU) incorporation. We assessed RNA synthesis in unchallenged cells or following DNA damaging radiations (UV or IR), which are known to repress transcription. Silencing of CDK12 led to an increase in EU incorporation suggesting a raise in global RNA synthesis or, alternatively, an imbalance between genes repressed and genes activated by CDK12 (i.e. DEG-up genes would contribute more to global RNA synthesis than DEG-down). Surprisingly, while IR or UV irradiation caused a global decrease in RNA synthesis, the increase caused by CDK12-KD persisted in irradiated cells, as shown in multiple cell lines (Fig. 2a, b and Supplementary Fig. 14a,c). Inhibition of CDK12 by THZ531 yielded similar results (Supplementary Fig. 14b). As increased RNA synthesis caused by CDK12 depletion persisted in IR- or UV-irradiated cells, we hypothesized that damaged genes may not be repressed in these conditions. A role for CDK12 in the repression of transcription at damaged loci would have important implications, as this DNA-damage-induced transcriptional shutdown is critical for preserving genomic integrity⁵³. To further address this, we took advantage of the U2OS-TRE-I-SceI-19 reporter cell line that allows the introduction of DNA double-strand breaks (DSBs) upstream of the promoter of a Tet-regulated MS2-based transcriptional reporter⁵⁴. This system enables simultaneous locus-specific detection of nascent RNA and DSBs, at the single-cell level. Upon activation of the mCherry-tTA-ER transcription factor, the transcribed locus can be visualized by the colocalization of mCherry-tTA-ER and the nascent RNA, which is bound by the MS2-YFP RNA binding protein. At the same time, DSBs generated by the expression of the I-SceI restriction enzyme can be visualized by γH2AX IF-staining. As expected, upon expression of I-SceI, DSBs were generated at the locus (visualized by the colocalization of γH2AX foci with mCherry-tTA-ER) (Fig. 2d), thus leading to loss of nascent RNA (i.e. loss of the nuclear MS2-YFP dot, Fig. 2d). Indeed, less than 10% of mock silenced cells (siLuc) showed an MS2-YFP signal colocalized with γH2AX (Fig. 2e), thus confirming that upon DNA damage the activity of the transcriptional reporter was strongly suppressed. Instead, upon CDK12 silencing, transcription was rescued in 50% of the cells showing a damaged site (Fig. 2d, e). A similar rescue was observed upon inhibition of CDK12 by THZ531 (Fig. 2f, g). In both cases, loss of CDK12 activity does not affect the efficiency of DSBs induced by I-SceI (Supplementary Fig. 15a, b). In the absence of I-SceI, mCherry-tTA-ER and MS2-YFP colocalized and were not affected by CDK12 silencing or inhibition (Fig. 2d, e, f, g). Altogether, these data imply that CDK12 activity is required for locus-specific transcriptional repression following DNA damage. Silencing of either CCNK or CDK13, also led to a partial rescue of transcription at the damaged TRE-I-SceI-19 reporter, while simultaneous silencing of CDK12, CDK13 and CCNK further increased transcriptional rescue of the locus, suggesting partial redundancy (Supplementary Fig. 16a,b).

**Fig. 2: CDK12 silencing or inhibition rescues transcription at DNA-damaged genes.**

CDK12 is recruited to damaged and transcribed genes

Next, we asked whether CDK12 could localize to sites of DNA damage. Cells were irradiated and proximity ligation assay (PLA) was used to assess the proximity of CDK12 to γH2AX foci. While signals were negligible in unirradiated cells, there was a marked increase in PLA-signals (foci) 30 minutes post-irradiation (Fig. 3a, b); by 4 hours the signal had returned to baseline level, possibly coinciding with ongoing DNA repair (Fig. 3a, b). The same result was confirmed with a different CDK12 antibody (Supplementary Fig. 17). Silencing of CDK12 lowered the PLA signal in irradiated cells, confirming the selectivity of the assay (Fig. 3a, b and Supplementary Fig. 17). Hence, CDK12 can be detected in close proximity to DDR-foci. To assess whether DNA damage and/or the resulting DDR would trigger the recruitment of CDK12, we transiently expressed an mCherry-tagged CDK12 and performed live imaging analyses on laser micro-irradiated cells. In unchallenged cells, CDK12 showed a diffused/dotted nuclear signal, while upon irradiation CDK12 promptly localized at damaged sites (Fig. 3c, Supplementary Fig. 18a, b and Supplementary Video 1). Similarly, mCherry-CDK12 was recruited onto the reporter locus only when this was cut by the I-SceI nuclease (Fig. 3d and Supplementary Fig. 19a).

**Fig. 3: CDK12 is recruited on DNA damaged sites proximal to transcribed genes.**

Further analysis with chemical inhibitors revealed two important features of the CDK12 relocalization. First, CDK12 inhibition by THZ531 did not prevent its recruitment to laser-damaged sites, thus indicating that this was independent from CDK12 catalytic activity (Supplementary Fig. 18c), yet this required CCNK (Supplementary Fig. 16c). Second, the recruitment of CDK12 onto damaged sites in both microirradiated and I-SceI transfected cells was suppressed by the transcriptional inhibitors 5,6-Dichloro-1-beta-D-ribofuranosylbenzimidazole (DRB) or triptolide, indicating that it depended on RNAPII transcriptional activity (Fig. 3c, d). Similarly, lack of activation of rtTA and the consequent recruitment of the transcriptional machinery prevented the localization of CDK12 on the damaged reporter in I-SceI transfected cells (Supplementary fig. 19b). To verify CDK12 recruitment on endogenous DNA damaged loci, we took advantage of the CRISPR/Cas9 genome editing system and designed sgRNAs to generate DSBs either at the promoter of expressed genes (MCM2, IFDR2 and POLR1B) or at non-transcribed distal regions. While sgRNA- and Cas9-mediated induction of DSBs proximal to promoters induced two γH2AX foci that co-localized with CDK12, the latter was absent at DSBs induced at the non-transcribed distal sites (Fig. 3e, f). At transcribed loci, CDK12 localization depended on the activity of RNAPII, since it was reduced to background level by DRB (Fig. 3f). Also, CDK12 recruitment was low and near background levels when DSBs were induced by sgRNAs targeting the transcriptional termination sites of transcribed genes (Supplementary Fig. 19c). Overall, these results suggest that CDK12 is recruited to DSBs that occur proximal to the promoter of transcribed loci or at gene bodies.

PARP, but not ATM- or ATR-dependent DDR, controls the recruitment of CDK12 to damaged genes

Next, we asked which components of the DDR would be required for the recruitment of CDK12 to DNA-damaged sites. Chemical inhibitors of either ATM or ATR did not prevent mCherry-CDK12 localization to laser-irradiated DNA (Fig. 4a, b). In addition, recruitment of CDK12 was preserved in H2AX-KO cells (Fig. 4c and Supplementary Fig. 20a,b). On the other hand, inhibition of PARP, which has also been implicated in the control of DDR and transcription at damaged sites⁵⁵, prevented CDK12 recruitment to DDR-sites (Fig. 4d, e). PARP1 recruitment to DDR sites was independent of either ATR or ATM activity, suggesting that PARP1 and ATM/ATR may control DDR signaling independently (Supplementary Fig. 20c). We next evaluated whether ATM, PARP or CDK12 inhibition would affect RNA synthesis at the damaged MS2-reporter in I-SceI cells. Inhibition of either PARP or CDK12 rescued transcription, while their combined inhibition led to a marginal increase of the percentage of cells showing transcriptional rescue, consistent with the evidence that PARP and CDK12 are on the same signaling axis (Fig. 4f). As previously reported⁵⁶, ATM inhibition rescued transcription (Fig. 4f). Combination of ATM and CDK12 inhibition did not significantly increase the rescue of transcription, compared to the single inhibitors (Fig. 4f). This suggests that while CDK12 recruitment to damaged sites is ATM-independent, the ATM and PARP/CDK12 signaling may converge downstream to control transcription at damaged sites.

**Fig. 4: PARP-dependent DDR-signaling is necessary for CDK12 recruitment to DNA-damaged sites.**

CDK12 prevents the recruitment of CDK9 to damaged genes

Our experiments showed that CDK12 recruitment to damaged promoter-proximal regions depended upon the recruitment of active transcriptional complexes (tTA-ER) and the presence of a transcribing RNA polymerase. This possibly implies that on these loci CDK12 may repress the activity of the elongating RNAPII. Given the prominent role of CDK9 in elongation, we tested whether CDK12 might modulate CDK9 recruitment to damaged DNA. Coherently with transcriptional inhibition, CDK9 did not localize at laser-irradiated DNA in mock-silenced cells. Instead, silencing of CDK12 allowed the recruitment of CDK9 on damaged DNA (Fig. 5a, Supplementary Fig. 21a and Supplementary Video 4, 5). Similarly, I-SceI-induced DSBs reduced the co-localization of CDK9 and mCherry-tTA-ER; conversely, silencing of CDK12 engendered CDK9 positive spots proximal to DDR foci (γH2AX) and transcriptional foci (mCherry-tTA-ER) (Fig. 5b, c and Supplementary Fig. 21b). This suggested that CDK12 blocks the transcription of damaged genes by preventing CDK9 recruitment.

**Fig. 5: CDK12 is recruited by SPT5 to prevent CDK9 loading to DNA-damaged sites.**

Two major complexes, NELFc and DSIF, regulate pause-release and elongation by controlling CDK9 recruitment to RNAPII. The NELF-E subunit of NELFc can repress transcription at damaged promoters⁵⁷, thus raising the question of whether NELFc may mediate CDK12 association to damaged loci. However, silencing of NELF-E or other NELFc subunits did not affect CDK12 recruitment on micro-irradiated nuclear regions (Fig. 5d and Supplementary Fig. 22a-c). Moreover, silencing of CDK12 did not prevent NELF-E recruitment, thus indicating that both proteins are recruited independently to DSBs (Fig. 5e). We next addressed the DSIF complex, a heterodimer composed of SPT4/5. While silencing SPT4 did not alter CDK12 recruitment to DSBs (Supplementary Fig. 22d,e), knock-down of SPT5 prevented it (Fig. 5f and Supplementary Video 6). This was also confirmed in the MS2-reporter line, where silencing of SPT5 decreased the colocalization of CDK12 with DDR foci (Fig. 5g, h). Overall, these results indicate that upon DNA breaks proximal to promoter regions, CDK12 is recruited by the DSIF complex, thus preventing CDK9 loading and processive elongation by RNAPII.

Loss of CDK12 exacerbates replicative stress and transcription-replication conflicts in MYC-overexpressing cells

Replicative stress induced by MYC has been associated with premature initiation of DNA synthesis and the collision of the replicative apparatus with loci transcribed during early S-phase^19,21. For this reason, we evaluated whether silencing of CDK12 would exacerbate MYC-induced RS. Cell cycle analysis of asynchronous cultures showed that silencing of CDK12 in MYC-overexpressing cells led to an increase in the S-phase fraction (Supplementary Fig. 4c), which, associated with the increased DDR (Fig. 1e), was suggestive of RS. This was confirmed in experiments where cells were released from a mitotic block to allow their synchronous initiation of DNA replication. While MYC stimulated S-phase entry, silencing of CDK12 alone had little effect on cell cycle progression (Fig. 6a). On the other hand, the combination of MYC activation and CDK12 silencing delayed progression through S-phase, as evidenced by EdU incorporation profiles that were still predominantly composed of early S-phase cells (i.e., EdU positive cells with DNA content closer to 2 N), both at 12 and 18 hours post-release (Fig. 6a and Supplementary Fig. 23a). Overall, this suggested that loss of CDK12 increased MYC-induced RS. This was supported by quantitative analysis of DNA replication foci by PLA: combined MYC activation and CDK12 loss increased the number of PCNA/EdU foci, indicating an increased compensatory firing of replication origins, which is typically observed upon RS (Supplementary Fig. 23b, c). Consistent with the onset of RS, the increased γH2AX observed upon loss of CDK12 and MYC activation was associated with newly replicated DNA (Fig. 6b, c). Given the predominant role of transcription-replication conflicts (TRCs) as a potential source of DSBs positioned near transcribed genes, and considering that our data implicates CDK12 in the control of transcription at damaged genes, we asked whether loss of CDK12 would promote the formation of RNA-DNA hybrids. In particular, we reasoned that the loss of CDK12 might increase the chances of forming RNA-DNA hybrids (and R-loops), due to the lack of transcription inhibition at TRC loci and increased annealing of the nascent RNA on to the template DNA. Indeed, in cells depleted of CDK12, we did observe a robust formation of RNA-DNA hybrids following laser-induced DNA-damage, assessed by enhanced recruitment of RNaseH1-GFP (Fig. 6d, e). For a quantitative assessment of TRCs induced by MYC activation, we employed PLA for the detection of RNA-DNA hybrids (detected by the S9.6 antibody) and newly synthesized DNA (EdU-labeled). Knockdown of CDK12 marginally increased the number of S9.6/EdU foci, while MYC activation increased the formation of S9.6/EdU foci (Fig. 6f, g), as expected given its ability to induce TRCs¹⁹. Importantly, combined activation of MYC and silencing of CDK12 led to a stronger increase of S9.6/EdU foci, thus indicating that loss of CDK12 may indeed exacerbate TRCs (Fig. 6f, g). While these data suggest increased TRCs, whether or not these RNA-DNA hybrids are causal and or linked to increased R-loop formation will need further assessment. Enhanced Myc-induced TRCs were also confirmed by PLA analysis of PCNA and RNAPII, which showed increased signals upon CDK12 silencing (Supplementary Fig. 23e, f). In line with previous reports³³, activation of Myc led to decreased PCNA/RNAPII signals, whose significance will need further investigation. The evidence that the silencing of CDK12 triggered MYC-induced TRCs suggested a protective role of CDK12, which might be recruited to TRCs loci, once DNA is damaged. In line with this hypothesis, PLA experiments showed that MycER activation enhanced colocalization of CDK12 with DDR foci (Fig. 6h and Supplementary Fig. 23d).

**Fig. 6: Loss of CDK12 enhances MYC-induced replicative stress.**

Loss of CDK12 exacerbates DSBs due to co-directional TRCs at early-replicating regions

The data above suggested TRCs as a cause of DNA damage proximal to newly synthetized DNA. To seek direct evidence, we mapped early replicating regions by EdU-HU-seq¹⁹ and DSBs by BLISS^58,59. Since BLISS signals are inherently sparse and “digital”, due to their single-base resolution, we devised an algorithm to identify genomic regions more prone to undergo DSBs (i.e. enriched in BLISS signals), henceforth dubbed as BLISS+ regions.

Of note, while BLISS signals could be detected in any of the conditions considered, BLISS+ regions could only be identified in cells when MycER was activated, possibly reflecting the stochastic nature of DSBs arising in either wild-type or shCDK12 cells. Activation of MycER in shCDK12 cells triggered a four-fold increase in the number of BLISS+ regions (5473 vs 1637, Fig. 7a). Of these, only a minority (12%) was detected also in shNT-MycER cells (Fig. 7a). In addition, BLISS+ regions detected in shCDK12 cells were more conserved among replicates than those detected in wild-type cells (Supplementary Fig. 24a), suggesting that these were recurrent hotspots for DSBs. Also, while BLISS+ regions detected upon MycER activation in shNT-cells were predominantly distributed in either intergenic regions, in shCDK12 cells, BLISS+ regions were preferentially mapped at either promoters or intragenic regions (74%, Fig. 7b). This suggested that transcription was a causal factor in the generation of DSBs when MYC was activated and CDK12 silenced. Given the role of early DNA replication as a potential source for DNA-damage in MYC-overexpressing cells¹⁹, we used EdU-HU-seq to identify early replicating regions (ERRs, genomic regions enriched in early replicating origins). In shCDK12-MycER cells, half of the BLISS+ regions overlapped with ERR (47%) (Fig. 7c and Supplementary Fig. 24b), either at boundaries or within-ERRs (Fig. 7d and Supplementary Fig. 25a), while the remaining were proximal to sparse EdU signals (Supplementary Fig. 26). ERR-associated BLISS+ regions had two peculiar topological features: (i) they were often positioned at the boundary of the EdU-HU-seq signal, (either on the left or on the right side) and (ii) they were adjacent to the promoter of an expressed gene (Fig. 7e; Supplementary Fig. 25b and 26). Directionality analysis of both DNA and RNA synthesis, indicated that these events were predominantly co-directional, so that almost invariably, BLISS+ regions were positioned between the front of an incoming early DNA replicating locus and the promoter of an expressed gene (Fig. 7e, f and Supplementary Fig. 26). In line with the above, the majority of the BLISS+ regions outside ERR (about 56%) were also associated with sparse EdU-HU-seq signals (due to DNA synthesis from interspersed early replicating origins) and transcribed genes, mostly in a co-directional head to tail orientation of the upcoming DNA replication front and the transcribed gene (Supplementary Fig. 27). On the other hand, in shNT-MycER cells, only 26% of the BLISS+ regions were near ERR (Fig. 7c), and, when proximal to early replicated DNA, they were less frequently associated with transcribed genes (56% of the BLISS+ regions did not overlap with expressed genes) (Fig. 7b). In addition, these BLISS+ regions were neither strongly associated with promoters, nor with the other topological features found in shCDK12-MycER cells, since only a minority were positioned between codirectional ERR and transcribed genes (Supplementary Fig. 28, 29). The observation that the DSB-prone regions in shCDK12-MycER cells were associated with expressed genes strongly suggested that loss of CDK12 exacerbated MYC-driven TRCs, leading to accumulation of unresolved DSBs. We next asked whether, at these loci, DNA synthesis and/or RNA expression were modulated by either MYC activation or silencing of CDK12. EdU-HU-seq indicated that DSBs-prone regions in shCDK12-MycER cells (BLISS+ regions) were the replicating regions the most stimulated by MycER activation (Fig. 7g, Supplementary Fig. 30). We also noticed that silencing of CDK12 led to a general reduction of MYC-induced DNA replication since the EdU incorporation levels were lower than those assessed upon MycER activation in mock silenced cells (Fig. 7g, h, Supplementary Fig. 30). This potentially reflected the activity, in trans, of the replication checkpoint. Silencing of CDK12 (alone) did not affect early-replication (Fig. 7g, Supplementary Fig. 30). In addition, while transcripts near BLISS+ regions were not preferentially regulated by MycER activation or by CDK12 silencing, they were synthesized more rapidly than all the other expressed genes (Fig. 7i). Thus, these DSBs were generally associated with regions of strong MYC-induced DNA synthesis and high rates of RNA synthesis, suggesting that (i) loss of CDK12 exacerbated TRCs in regions of intense RNA and DNA synthesis and (ii) these conflicts were triggered by MYC due to its ability to anticipate and boost DNA synthesis (Supplementary Fig. 31).

**Fig. 7: Genome-wide mapping of transcription replication conflicts.**

Discussion

Transcriptional silencing upon DNA damage is a process regulated by signaling pathways activated by the DDR⁶⁰. This process is triggered by multiple DNA lesions, including DSBs. By inhibiting the activity of transcriptional complexes, cells avoid unnecessary synthesis of potentially faulty mRNAs (a source of transcriptional dependent mutagenesis)⁶¹, prevent molecular crowding by transcriptional complexes and repair factors, and create a molecular scaffold (e.g., R-loops) to facilitate DNA damage resolution. Current models highlight the prominent roles of apical regulators of the DDR, like ATM and PARP1, which control chromatin remodeling, deposition of repressive histone marks and accessibility at damaged sites^56,62, thus leading to repression of gene transcription. Our data shows for the first time that CDK12 is recruited to transcribed loci upon DNA damage to repress gene’s expression (Supplementary Fig. 32). This implies an active mechanism that allows selective localization of CDK12 at damaged sites. Genetic dissection of potential upstream regulators revealed that CDK12 localization at these sites is independent from the DDR signaling controlled by ATM, and from γH2AX but instead relies on PARP activity. This may suggest a branched control of transcription at damaged sites, whereby ATM regulates the epigenetic state and accessibility⁵⁶ while PARP may have the added function of regulating RNAPII activity by recruiting CDK12. Since several RNAPII-associated factors are PARylated following DNA damage⁵⁵, it is possible that these may also promote CDK12 localization, possibly contributing to transcriptional inhibition at damaged loci. Selective recruitment of CDK12 at damaged DNA also required an elongation-competent RNAPII, as it was impaired by inhibition of initiation, by triptolide, or inhibition of elongation by DRB.

RNAPII escape from promoters is dynamically regulated by the association with the NELF complex and the DSIF complex, which, by binding to “opposite” sides of the RNAPII holoenzyme control its pausing and pause-release. Two subunits of the NELF complex, NELF-A and NELF-E, are recruited to RNAPII upon DNA damage to repress transcription⁵⁷. Our data suggest that this is independent of CDK12 since (i) silencing of CDK12 or its inhibition did not prevent NELF-E localization to damaged DNA and (ii) silencing of NELF-E or the full complex did not affect CDK12 localization to damaged DNA. Instead, CDK12 recruitment required SPT5, one of the two subunits of the DSIF complex. This suggests an independent control of DNA damage-induced pausing by the NELF complex and by the CDK12/DSIF complex.

DSIF, once phosphorylated by pTEFb, is converted into a positive elongation factor stably associated with RNAPII. Therefore, our observation that CDK9 inhibition by DRB suppresses localization CDK12 at damaged and transcribed loci suggests that the “elongating” form of SPT5 favors CDK12 recruitment. This also implies that, at least in principle, CDK12 could also support transcriptional inhibition at intragenic DNA damaged sites, and not only at promoter proximal sites.

CDK12 is recurrently mutated in several cancers: in ovarian and prostate cancers^63,64 homozygous (loss of function) mutations are associated with genomic instability characterized by tandem focal duplication at gene-dense regions located at early and late replicated domains^63,65,66.

While CDK12 was reported to regulate processing and expression of genes responsible for DDR signaling and homologous recombination, these genes are not altered in CDK12 mutant tumors⁶³. In addition, CDK12 loss of function tumors display genomic instability, mutational profiles and therapeutic responses that are different from homology-repair deficient (HRD) tumors^63,64,66,67, thus indicating that CDK12 mutant tumors are distinct from HRD-tumors and that loss of CDK12 affects genome integrity also by mechanisms distinct from homology directed repair.

In light of our findings, we propose that CDK12, besides regulating processing and expression of HR-genes, may protect genome integrity by repressing gene transcription and thus facilitating the resolution of DNA damage due to TRCs⁶⁸.

CDK12 is also recurrently amplified in tumors of different origin, suggesting its potential duality in cancer as either an oncogene or a tumor suppressor, depending on the context^39,69. This is similar to other DDR genes, which, depending on the genetic background, may support or suppress tumor growth. Coherently with a protective function of CDK12 in tumor cells, we and others have found that CDK12 is required to prevent cytotoxic DNA damage and favor cell survival upon oncogenic activation of MYC^47,51. In particular, we report here that loss of CDK12 predisposes cells to TRCs and RS. Replicative stress has emerged as a hallmark of oncogene driven proliferation which may act as a tumor suppressive barrier or a therapeutic liability⁶. Several lines of evidence indicated that MYC-induced RS is restrained by safeguarding mechanisms that, once activated, lead to accumulation of cytotoxic DNA damage^24,70. Still to be fully established are the potential causes of RS and the engaged safeguarding pathways⁷⁰. Recent evidence suggests that anticipated S-phase entry driven by MYC may predispose TRCs, thus leading to fork stalling and accumulation of DNA breaks¹⁹. In line with this, RS triggered by CDK12 loss led to the accumulation of DNA breaks preferentially located between early replicated regions and genes transcribed in a co-directional orientation. Such regions were characterized by strong MYC-driven DNA replication, but were not selective for MYC-regulated genes. This suggests that MYC-induced DNA replication, rather than transcription, is precipitating these TRCs. Based on the evidence presented here, it is likely that loss of CDK12 (which leads to DDR resistant gene expression) may affect how efficiently DNA damage is resolved at these sites. It thus follows that the increased detection of DSBs might be more likely due to defective repair than to increased TRC frequencies. Other possibilities exist: for instance, we cannot exclude that DDR activation at stalled forks (in the absence of DSBs) might activate CDK12 to stop transcription and avoid fork collapse and DSBs.

Several factors have been implicated as causal in RS-induced genomic instability, here the inactivation of CDK12 allowed to map the precise location on DSBs. This suggests that the management of DNA damage at these sites of potential TRCs is a major liability in MYC-driven cancers. On a broader scale, this may imply a prominent role for TRCs as a source of genome instability that could be exploited therapeutically to induce cytotoxic DNA damage.

We posit that future advancements in the mechanistic understanding of processes controlling transcription at damaged genes have potential for the identification of valuable therapeutic targets in cancer treatment. While loss of function analyses have allowed the identification of factors that regulate CDK12 recruitment to damaged genes and the dissection of how CDK12 modulate RNAPII activity, lack of biochemical data and structure-and-function analyses prevents a precise definition of how CDK12 is recruited to damaged genes or a detailed description of how CDK12 may control CDK9 recruitment. Further studies will be needed to fully define this pathway.

Methods

Cell lines

HeLa, U2OS, U2OS-TRE-I-SceI-19 (kind gift from Dr. Yasui Akira, (Tohoku University, Japan) and U2OS-H2A.X KO cells (kindly provided by Stephen P. Jackson, Wellcome Trust/CRUK Gurdon Institute and Department of Biochemistry, University of Cambridge, UK) were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 1% penicillin/streptomycin in 5% CO₂ incubator. For CDK12 stable knockdown, U2OS-MycER cells (bleomycin resistant)²⁷ were transduced with doxycycline inducible human shRNAs (Transomic) were cultured in DMEM containing 10% Tetracycline-free FBS. iCut-RPE-1 hTERT cells⁷¹ (kindly provided by René H. Medema, Oncode Institute, Division of Cell Biology, The Netherlands Cancer Institute, Amsterdam) were maintained in DMEM/F12 GlutaMAX medium supplemented with 10% Tetracycline-free FBS, 1% penicillin/streptomycin and Puromycin (20 ug/ml).

siRNAs, Plasmids and Transfection

Human siRNAs against CDK12 and siRLuc were purchased from Ambion (silencer select s28621, s28623). For all siRNA transfection, cells were reverse-transfected using Lipofectamine RNAiMAX reagent (Invitrogen by Life Technologies, Monza, Italy).

All plasmids were transfected using Lipofectamine 3000 reagents (Invitrogen by Life Technologies, Monza, Italy) according to the manufacturer’s instruction.

Cloning

The mCherry-CDK12 expression plasmid was generated by PCR using the pRSET-mCherry and PTX002-3XFLAG-CDK12 plasmids as templates. Briefly, a first PCR product (T1) was generated by using a forward primer designed to introduce the sequence of the NheI restriction enzyme at the 5’ end of the mCherry cDNA and a reverse primer designed to add the 5’ end of the CDK12 cDNA to the 3’ end of the mCherry cDNA. A second PCR product (T2) was generated using a forward primer designed to add the 3’ end of the mCherry cDNA to the 5’ end of the CDK12 cDNA and a reverse primer designed to overlap the sequence of the XmaI restriction enzyme present in the 5’ portion of CDK12 cDNA. Next, the two PCR products were used as template of a third PCR aimed to generate the final fragment, then inserted in the PTX002 plasmid by NheI (NEB) and XmaI (NEB) digestion and ligation. The following primers were used: T1 FWD CATCATGCTAGCCACCATGGTGAGCAAGGGCGAGGAGGA; T1 REV CCCATGTCTCTCTGAATTGGGCTTGTACAGCTCGTCCATGC; T2 FWD GCATGGACGAGCTGTACAAGCCCAATTCAGAGAGACATGGG and T2 REV CTTTTTCGGTCTGTTTAGCTTTTAGTAAG.

Western Blotting

Cells were lysed in lysis buffer containing 20 mM Hepes, 0,5 mM NaCl, 5 mM EDTA, 10% Glycerol, 1% Triton-X with protease and phosphatase inhibitors (Roche), kept 15 min on ice and then subjected to 10 cycles of sonication (10 s on, 10 s off) with the Bioruptor Plus (Diagenode UCD-300).

Nascent RNA labeling (EU labeling)

U2OS cells were reverse transfected and grown on glass coverslips for 48 h. Then, cells were irradiated at 10 Gy and fixed 30 minutes after IR. Alternatively, cells were irradiated with UV (15 J/m². Cells were pulse-labeled with 1 mM EU for 20 minutes before collection and fixation with 4% PFA. EU incorporation was detected by Click-iT technology (Invitrogen) with an Andy fluor 488 azide, according to the manufacturer’s protocol. Image analysis was performed measuring nuclear EU signal intensity with a custom script.

(https://github.com/0774v10/CDK12-MYC-transcription-replication/compare/v1.0.0…main).

Laser micro-irradiation

Cells were grown on glass coverslips and treated with 10 µM BrdU for 24 h to sensitize DNA to laser induced damage. To generate localized DNA damage, selected nuclear regions were microirradiated with 10 iterations of a 405-nm laser with 100% power using confocal Spinning Disk microscope (Olympus) equipped with IX83 inverted microscope provided with an IXON 897 Ultra camera (Andor) and a FRAP module furnished with a 405 nm laser microscope. Images were acquired every 388 ms for a total duration of 5 minutes. The signal intensity at DNA damaged sites was measured using the ImageJ software with a custom script from the IFOM imaging facility. For silencing experiments, cells were seeded on glass coverslips in reverse transfection with siRNAs directed to the indicated targets for 24 h and subsequently transfected with mCherry/EGFP plasmid of interest for an additional 24 h prior BrdU treatment.

Alt-R crRNA design and transfection

Alt-R crRNA were designed with Cas9 crRNA Design tool (Integrated DNA Technologies). On-target and off-target scores were determined using IDT specificity score and CRISPOR tool⁷². tracrRNA:crRNA duplex was generated and transfected according to the manufacturer’s protocol. In brief, iCut-RPE-1 hTERT cells were treated with doxycycline (Sigma, 1 mM), SHIELD-1 (Aobious, 1 μM) for 17 h in order to activate Cas9. The tracrRNA:crRNA duplex (10 nM final concentration, from a 1 µM stock) was transfected using Lipofectamine RNAiMAX (Invitrogen by Life Technologies, Monza, Italy) for 8 h.

Cell cycle synchronization

Asynchronous U2OS-MycER cells were seeded and treated with 1 µg/mL doxycycline and 400 nM OHT for 24 h and then treated with Nocodazole (50 ng/ml) for 18 h. Cells were then detached by mitotic shake off, washed with PBS, seeded and collected at the indicated time points. Time 0 represents cells in G₂/M, collected at the end of the Nocodazole treatment.

Monitoring MS2 transcription in U2OS-TRE-I-Sce-I reporter cell line

U2OS-TRE-I-Sce-I reporter cells were seeded on glass coverslips and transfected with the mCherry-tTA-ER plasmid, the pYFP-MS2 plasmid and the pCMV-NLS-I-SceI plasmid (to induce DSBs). The mCherry-tTA-ER chimera was activated by 1 µM OHT for two hours.

For silencing experiments, cells were first seeded on glass coverslips and reverse transfected with siRNAs. After 24 h, cells were transfected with the mCherry-tTA-ER, pYFP-MS2 and pCMV-NLS-I-SceI plasmids. 24 h post-plasmid transfection, mCherry-tTA-ER was activated by 1 µM OHT for 2 hours. Cells were then fixed with 4% PFA and processed for γH2AX staining and imaging acquisition.

Cell cycle analysis by FACS

For cell cycle analysis using BrdU incorporation, asynchronous growing cells were pulse-labeled with BrdU (100 μM) for 30 min prior to harvesting. After fixation with cold 70% EtOH, cells were treated with 1 ml of 2 N HCl for 20 min to denature labeled DNA. HCl was then neutralized with addition of 3 ml of sodium borate (0.1 M, pH 8.5) for 2 min. Cells were then incubated with anti-BrdU antibody (BD Biosciences). Following incubation with secondary fluorescence labeled antibody, cells were resuspended in PBS with propidium iodide (25 μg/ml) and RNase A (250 μg/ml). Cells were stored at 4 °C until FACS analysis.

Immunofluorescence analysis

Cells were grown on coverslips, fixed with 4% paraformaldehyde (PFA) for 15 min and permeabilized with 0.2% Triton-X100 for 10 min before blocking with 2% BSA in PBS for 30 min at RT. Cells were incubated 1 h at RT with the primary antibody in 2% BSA/PBS, washed three times with PBS and incubated 1 h at RT with the secondary antibody in 2% BSA in PBS. DAPI staining was performed for 5 min before mounting. Slides were imaged using a wide field microscopy (Leica DM6 B Multifluo) with 60X oil objective.

Proximity Ligation Assay (PLA) and PLA at replication forks

For PLA, we used the Duolink^R PLA Fluorescence Protocol Kit (SIGMA) according to the manufacturer’s instructions.

For PLA at nascent DNA⁷³, cells were pulse-labeled with EdU (10 µM) for 10 min and permeabilized with 0.5% Triton in PBS for 10 min at 4 °C. After washing with PBS, cells were fixed with 3% formaldehyde/2% sucrose in PBS for 10 min and blocked in 3% BSA in PBS for 30 min. Then, Click-IT reaction with biotin-azide was performed for 30 min using the following reagents: Biotin-azide 10 µM (Invitrogen B10184), Sodium Ascorbate 10 mM (Sigma Aldrich A4034), CuS0₄ 10 mM (Sigma Aldrich C8027) in PBS. Primary antibodies incubation was performed in PBS with 1% BSA and 0.1% saponin overnight at 4 °C. The following day, samples were processed according to the manufacturer’s protocol starting from the “PLUS and MINUS probes” incubation step.

RNA-seq

RNA was then extracted and purified using the Quick-RNA^TM Miniprep Kit (Zymo Research, cat. #R1055) according to the manufacturer’s instructions. Libraries for RNA-Seq were prepared with the TruSeq RNA Sample Preparation Kits v2 (Illumina) following manufacturer instructions.

ChIP-seq

Cells were fixed with 1% formaldehyde for 10 min and quenched with 0.125 m glycine for 5 min at room temperature. Cells were washed twice with PBS and resuspended in LB1 buffer (50 mM HEPES pH 7.5, 140 mM NaCl, 1 mM EDTA, 10% Glycerol, 0.5% NP-40, 0.25% Triton X-100) for 10 min on ice. After centrifugation, nuclei were extracted resuspending cells at room temperature for 10 min in LB2 buffer (10 mM Tris-HCl pH 8, 200 mM NaCl, 1 mM EDTA, 0.5 mM EGTA). The extracted nuclei were finally resuspended in LB3 buffer (10 mM Tris-HCl pH 8, 100 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 0.1% Na-Deoxycholate, 0.5% N-lauroylsarcosine) and sonicated in order to obtain DNA fragments of 300–100 bp. The lysate from 50 × 10⁶ cells was incubated with 10 μg of RNA PolIIS2p antibody (ChromoTek, E310) previously bound to protein G Dynabeads (Invitrogen) in PBS + 0.5% bovine serum albumin. After the incubation with the antibody, beads were collected using the DynaMag magnet, washed six times with 1 ml of RIPA buffer (50 mM HEPES pH 7.5, 500 mM LiCl, 1 mM EDTA, 1% NP-40, 0.7% Na-Deoxycholate) and once with 1 ml of TE 1X with 50 mM NaCl. De-crosslinking was performed overnight at 65 °C with 150 μl of TE 1X and 2% SDS. DNA was purified with PCR Qiaquick columns (Qiagen) and quantified using PicoGreen (Invitrogen) or QUBIT (Invitrogen).

EdU-HU-seq

U2OS-MycER cells were cultured with 1 µg/mL doxycycline, with or without 300 nM OHT, for 40 h. Cells were then treated with 100 ng/ml nocodazole (Sigma, Cat. No. SML-1665) for 8 h to induce mitotic arrest. G2/M arrested cells were isolated by shake-off, washed with PBS and released in warm medium containing 25 μM EdU (Invitrogen, Cat. No. A10044) and 2 mM hydroxyurea (Sigma, Cat. No. H8627-5G). After 12 or 18 h, cells were collected and fixed in 90% methanol. Next, cells were permeabilized with 0.2% Triton-X in PBS and EdU was coupled to a cleavable biotin-azide linker (Azide-PEG(3 + 3)-S-S-biotin) (kind gift from Thanos Halazonetis) using the Click-IT chemistry. DNA was then purified by phenol-chloroform extraction and ethanol precipitation and subjected to EdU-labeled DNA isolation according to the published protocol⁷⁴. The eluted DNA was directly used for library preparation using the HT-Chip-seq library preparation⁷⁵. Samples were spiked-in with 0.03% of NIH3T3 cells’ EdU-labeled DNA. 50 base pair paired-end read sequencing reactions were then performed on an Illumina Nova-Seq 6000. An aliquot of cells was saved for flow cytometry analysis in order to assess the percentage of cells that entered S-phase in each condition. In this case, after fixation, cells were permeabilized with 0.2% Triton-X in PBS, EdU was coupled to a fluor biotin-azide (Alexa Fluor 647 azide, Thermo Fisher Scientific, cat. no. A10277) and genomic DNA was stained with propidium iodide (Biotium, Cat. No. 40017) in combination with RNAseA (Roche, Cat. No. 10109169001) according to the published protocol⁷⁴.

EU-seq

U2OS-MycER cells were cultured with 1 µg/mL doxycycline, with or without 300 nM OHT, for 40 h. Cells were then treated with 100 ng/ml nocodazole (Sigma, Cat. No. SML-1665) for 8 h to induce mitotic arrest. G2/M arrested cells were isolated by shake-off, washed with PBS, released in warm medium for 12 h and then collected and fixed in 90% methanol. 0.5 mM EU (5-ethynyl-uridine, Jena Biosciences, Cat. No. CLK-N002-10) was added to the cells 2 h before cells’ collection. RNA was then extracted and purified using the Quick-RNA^TM Miniprep Kit (Zymo Research, cat. #R1055) according to the manufacturer’s instructions. Nascent RNA was biotinylated and purified starting from 50 μg of total RNA. EU was coupled to a cleavable biotin-azide linker (Azide-PEG(3 + 3)-S-S-biotin) (kind gift from Thanos Halazonetis) using the Click-IT chemistry as shown below: 50 μg of RNA (40 μg of human EU-labeled RNA and 10 μg of murine EU-labeled RNA) were mixed with the same volume of 2X Click-IT reaction cocktail (171 mM Tris-HCl pH 8, 8 mM CuSO4, 200 mM sodium-L-ascorbate, 0,1 mM biotin–azide) and incubated for 30 minutes at RT with gentle shaking. RNA was precipitated and the EU-labeled RNA was then isolated using 50 μl of Dynabeads MyOne streptavidin C1 (Invitrogen, Cat. No. 65001) each sample, that were washed before use two times with Binding and Washing Buffer 1X (5 mM Tris-HCl pH 7.5, 0.5 mM EDTA, 1 M NaCl, 0.5% Tween-20), followed by two washes with solution A (0.1 M NaOH, 0.05 M NaCl) and two washes with solution B (0.1 M NaCl). After washing, the beads were resuspended to twice the original volume with Binding and Washing Buffer 2X. RNase OUT (1:100, Invitrogen, Cat. No. 10777-019) was added to RNA and then samples were heated at 70 °C for 5 minutes and placed back on ice to remove secondary structures. Then EU-labeled RNA was mixed with an equal volume of washed beads (100 µL) and incubated for 30 minutes on a rotating wheel at room temperature. The RNA-beads complex was then washed three times with Binding and Washing Buffer 1X and once with RNase free-water. Finally, the EU-labeled RNA was eluted by incubating the streptavidin beads with 55 μl of 2% β -mercaptoethanol (Calbiochem, Cat. No. 444203) in 10 mM Tris-HCl pH. 8 for 1 h at room temperature. 5 μl of each sample were used for Qubit quantification (RNA HS assay kit; Thermofisher, Cat. No. Q32852). Sequencing libraries were prepared with TruSeq Stranded Total RNA Gold (Illumina, Cat. No. 20020598).

BLISS

For BLISS experiments, U2OS-MycER cells were seeded in medium with 1 µg/mL doxycycline, with or without 300 nM OHT for 24 h. Cells were then treated with 50 ng/ml nocodazole (Sigma, Cat. No. SML-1665) for 18 h to induce mitotic arrest. G2/M arrested cells were isolated by shake-off, washed with PBS and released in warm medium for 18 h and fixed according to BLISS protocol⁵⁹. In brief: cells were harvested in order to have a single cell suspension and then resuspended in pre-warmed PBS/10% FCS in order to have 10⁶ cells/ml (8–10 × 10⁶ cells for each condition). Fixation was performed by adding 2% PFA (Thermofisher, Cat. No. 28908) for 10 min at room temperature, and 125 mM glycine for 5 minutes gently rotating and then 5 minutes on ice. Cell pellets were washed twice with ice-cold PBS and processed as described⁵⁹. Briefly, cell pellets were permeabilized in buffer consisting of 10 mM Tris-HCl/150 mM NaCl/1 mM EDTA/0.3% SDS pH 8 and DSBs were blunted with the Quick Blunting Kit (New England Biolabs). sBLISS adapters were ligated to the blunted DSB ends with 25 U of T4 DNA Ligase (ThermoFisher Scientific) at 16 °C for 20–24 hours in a 100 μL ligation reaction mix containing 3 μL of 50 mg/mL BSA (ThermoFisher Scientific). gDNA was extracted by proteinase K digestion followed by Phenol-Chloroform purification. DNA was sonicated to obtain fragments with sizes ranging from 300 to 800 bp, with a peak around 400–600 bp. 90–300 ng of sonicated gDNA was In vitro transcribed with the MEGAscript T7 Transcription Kit (Thermo Fisher Scientific), The resulting RNA was first ligated to the Illumina RA3 adapter by T4 RNA Ligase 2 (New England Biolabs), for 2 hours at 25 °C, and then reverse transcribed using the Illumina RTP primer the SuperScript IV Reverse Transcriptase for 50 min. Libraries were amplified with NEBNext Ultra II Q5 Master Mix (New England Biolabs).

siRNA based screen

Automated high-throughput siRNA transfection

The murine siGenome SMARTpool druggable library was obtained from Dharmacon (Thermo Fisher Scientific) in 384-well format. Each well in the library contains a pool of four distinct siRNA oligos targeting different sequences of the target transcript. Five library plates were analyzed targeting a total of 1400 druggable genes plus positive controls for monitoring transfection efficiency (SMARTpool siGENOME Plk1 siRNA), assay specific positive control (SMARTpool siGENOME Rad21 siRNA) and a siRNA targeting the Renilla Luciferase gene (siGENOME Rluc siRNA) as negative control for data normalization.

The esiRNA custom libraries were purchased from Sigma (MISSION® esiRNA, Sigma) in 384-well format. esiRNAs are endoribonuclease-prepared siRNA pools composed of a mixture of siRNAs that all target the same gene. Control esiRNAs targeting mouse Plk1, Rad21 and Rluc were purchased in individual tubes and included in the final 384-well plates.

For validation, silencer pre-designed siRNAs were obtained from Ambion (Silencer®, Ambion) in a 384-well plate format. Three different individual silencer siRNAs were chosen for each gene. Silencer siRNAs against murine Rad21 (#150458) and was used as positive control.

For screening, 800 Bz1 R26-MycER immortalized MEF cells/well were reverse-transfected in 384-well black, gelatin-coated optical plates (Corning) using the Freedom EVO automated liquid handler system (Tecan) with 0.13 µL/well of DharmaFECT1 transfection reagent (Dharmacon, Thermo Fisher Scientific) and 25 nM Dharmacon SMARTpool or Ambion silencer siRNA or 50 nM of Sigma esiRNA. After transfection, cells were cultured for 48 hours under standard conditions (37 °C in humidified atmosphere, with 5% CO2), in the SteriStore automated incubator (HighRes Biosolutions). A total of 6 replicated plates (3 ethanol- and 3 OHT-treated) were transfected for each of the library plates.

Fixing and immunofluorescence staining

Forty-eight hours post siRNA transfection, cells were fixed by adding 40 µl of 4% Paraformaldehyde (PFA) directly to the medium to reach a final concentration of 2%. The plates were then incubated at room temperature for 15 minutes. After that, the medium was removed, and the plates were washed twice with PBS. Cells were permeabilized by adding 20 µl of 0.1% Triton X-100 (in PBS) to each well. After 10 minutes, cells were washed with PBS and blocked with 20 µl of blocking buffer (PBS + 2% BSA + 10% goat serum) for 30 minutes. Cells were then rinsed twice with PBS and incubated with 25 µl of anti-phospho-Histone H2AX (Ser139) antibody (Merck Millipore, Cat#05-636) diluted 1:1500 in 2% BSA PBS, for 1 hour at room temperature. Afterwards, cells were washed two times with PBS and then incubated with 25 µl of secondary antibody dilution (1:400 Alexa Fluor® 488 goat anti-mouse antibody, Invitrogen Cat#A11001) and DAPI (1:4500 in PBS) for 1 hour. Following 3 washes with PBS, 70 µl of PBS was left in the plates and the plates were sealed and stored at 4 °C until imaged.

Image acquisition and processing

Cells were imaged on the Olympus ScanR wide-field microscope. γH2AX-Alexa 488 fluorescence was acquired using BP470-495 excitation filter and BP510-550 emission filter (U-MNIBA3; Olympus) while DAPI was acquired by BP360-370 excitation filter and BP420-460 emission filter (U-MNIBA3; Olympus). Image acquisition was done with a 10X objective, and six different fields of view were recorded for each well, thus covering most of the entire area of a 384-well format.

The acquired images were analyzed using a custom image analysis algorithm (by Adrian Andronache, available upon request), developed and executed in the Acapella software development/run-time environment (Perkin Elmer). The algorithm used some Perkin Elmer proprietary procedures, such as the nuclei detection on the DAPI channel. The image analysis procedure was run on every single field of view acquired on two channels, and eventually merged the results for each individual well by summing or averaging single parameters. In brief:

Step 1: Background removal and uneven illumination correction

Background removal and uneven illumination correction was performed using a sliding parabola transform which divides the input image into two parts: background and foreground signals. The curvature of the parabola was set to 1/500 pixels.

Step 2: Nuclei segmentation and local background definition

First, using Perkin Elmer (PE) proprietary algorithms, we identified all the individual nuclei from the DAPI channel. Then we defined the local background of every nucleus using the Voronoi tessellation. Voronoi tessellation divides the image space by defining region boundaries equally distanced from the neighboring given centers (i.e. nuclei). A maximum border distance of 50 pixels away from the nuclei was imposed. Once the nuclei were segmented and their Voronoi regions were estimated, we defined individual foreground and background masks for every nucleus, these masks were further used to extract phenotype descriptors for every cell.

Step 3: Preliminary features extraction

For all the detected objects (nuclei having both a foreground and an associated local background), a series of basic parameters were directly extracted or derived from the different images to describe the morphology and phenotype of individual cells. First, from the nuclei segmentation we estimated the radius, the area, and the shape of every single nucleus. Second, the nuclear DAPI intensity was defined as the difference between the average intensities of each individual foreground and background nuclei. The DNA content was defined as the product of the DAPI intensity and the nuclear area. Third, from the Alexa 488 channel we estimated all the γH2AX related features as described in the following paragraph.

Step 4: γH2AX foci and pan-nuclei definition

The γH2AX foci were defined using a local adaptive thresholding approach. Using the previously defined nuclear masks, we calculated the average Alexa 488 intensities of every nucleus in the population of negative controls (siRluc). This resulted in the threshold that differentiate the background from γH2AX signal. Further on, this threshold was used to segment the γH2AX foci from the Alexa 488 channel on every single nucleus (within the foreground masks) on all the screened wells. A local adaptation of the threshold was performed for every single nucleus, by correcting for the Alexa 488 background level extracted from the individual local background mask. All nuclei with γH2AX foci covering more than 70% of their surface were declared (labeled) as “pan-nuclear γH2AX”.

Step 5: Phenotype descriptive parameters

For every single well, the information extracted from all the 6 field of view was finally integrated and the following main phenotype description parameters reported:

N°_Of_Nuclei: the total number of nuclei from all the 6 fields of view, remained after the different filtering described previously;

N°_Of_PanPosNuclei: the total number of pan-nuclear γH2AX nuclei from all 6 fields of view;

Percent_Of_PanPos: the percent of pan-nuclear γH2AX nuclei with respect to the total number of nuclei from all 6 fields of view;

Dapi_Intensity: the average DAPI intensity of all detected nuclei;

DNAContent: the average DNA content of all detected nuclei;

NuclearArea: the average nuclear area of all detected nuclei;

γH2AX_Intensity: the average Alexa488 intensity of all detected nuclei;

γH2AX_AreaCoverage: the average γH2AX foci area coverage of all detected nuclei.

Each well of an assay plate was associated with a phenotype descriptive vector (an array of the 8 previously main parameters) and eventually tabulated and associated the biological and technical annotation according to the plate-layout design and siRNA-database definition.

Statistical data processing and analysis

The statistical data processing and analysis was performed independently on two of the main phenotype descriptor parameters extracted by the image analysis: the viability (i.e. the N°_Of_Nuclei) and the DNA damage response (DDR) (the percentage of pan-nuclear γH2AX nuclei: Percent_Of_PanPos). The data of each screened siRNA was available in triplicates sample values for each of the two chemical treatments ethanol (EtOH) and OHT. Each assay plate had a series of control wells (with negative and positive controls for transfection and assay-specific positive control) and multiple experimental samples. From here on, the data of a single well will denote the specific parameter value of any well, independent from the nature of the transfection reagent class (negative or positive controls or experimental samples). The statistical data processing and analysis is organized in the following 5 main steps: intra plate normalization of raw data (on each individual assay plate); intra-treatment data dimensionality reduction; inter-treatment data dimensionality reduction; hit metric estimation; hit calling. The intra-plate normalization of raw data ensures the removal of inter-plate variability. This operation is done on every assay plate with respect to the median data of all the negative controls present on the analyzed plate (siRluc). The normalization procedure was performed only on the viability parameter, as the percent of pan-nuclear γH2AX nuclei is internally normalized to the number of nuclei of a given well. All the samples presenting a very low viability (i.e. less than 100 nuclei within a well) were excluded from the downstream statistical data processing and analysis of both viability and percent of pan-nuclear γH2AX parameters. The intra-treatment data dimensionality reduction merges the replicates (triplicates) values of the wells on the assay plates. This operation estimates the median of the transfection of sample replicates in order to assign one single data value for each well (siRNA) within each experimental condition (EtOH and OHT treatments). The inter-treatment data dimensionality reduction is the operation of combining and confronting the parameter values of single wells in between the different experimental conditions. As such, the viability-ratio was estimated for all the wells by dividing normalized viability in OHT-treated to EtOH-treated samples. At the same time, the DDR-gain was estimated as the difference between the percent of pan-nuclear γH2AX nuclei in OHT-treated and the EtOH-treated samples. Both viability-ratio and DDR-gain previously estimated underwent the classical Z-score hit metric estimation. In the primary screens the entire population of experimental samples was used as reference, under the assumption that most of the siRNAs have a null effect. In the deconvolution/validation screens the population of negative samples present on the analyzed plate was used as reference. Viability Hits and DDR-Hits were called based on Z-score thresholds defined using negative and positive controls as a reference.

Bioinformatic analyses

RNA-seq analysis

Paired-end reads from Illumina HiSeq 2000 were aligned using tophat -r 170 -p 8 --no-novel-juncs --no-novel-indels --library-type fr-unstranded --transcriptome-index transcriptome_index reffile fastq_file_R1 fastq_file_R2 using hg19 genome assembly as reference. Duplicated reads were removed using samtools rmdup -sS bamfile. The number of counts for each gene were computed using featureCounts -T 2 -p -P -a <gtffile > . Differentially expressed genes (DEGs) between experimental conditions (OHT, siCDK12, OHT+siCDK12) and not treated samples were calculated using DESeq2 R Bioconductor package⁷⁶. All the genes with an adjusted p-value < 0.01 and with a log2FC > 0.5 or < −0.5 were defined as DEG up or DEG down, respectively.

ChIP-seq analysis

Paired-end reads from Illumina HiSeq 2000 were aligned using the BWA v.0.6.2 tool⁷⁷ with default settings, using hg19 genome assembly. Peaks were called using MACS v2 1.4⁷⁸, using a p-value threshold of 10-5.

BLISS analysis and identification of BLISS positive regions

Raw sequencing data were demultiplexed on Illumina’s BaseSpace to generate FASTQ files, which were processed using the sBLISS pipeline previously described in detail⁵⁹. The pipeline first identifies reads containing the expected sBLISS prefix consisting of an 8 nt UMI followed by an 8 nt sample barcode, using SAMtools (v1.10)⁷⁹ and Scan for Matches⁸⁰, allowing at most one mismatch in the barcode. Prefixes are clipped off and stored, while the trimmed reads are aligned to the human reference genome (GRCh37/hg19) using BWA-MEM22 (v0.7.17-r1188) with default options⁷⁷. Reads with mapping quality score lower than 30 and PCR duplicates are removed by searching for adjacently mapped reads (at most 10 bp apart along the reference genome) with at most one mismatch in their UMI sequence. The output of the pipeline is a list of unique DSB locations and corresponding number of unique UMIs identified at each position.

We devised a custom R script to identify genomic regions enriched for BLISS signals (BLISS+ regions). Briefly, the hg19 genome was divided in 2 kb bins. For each bin, we calculated the coverage (number of BLISS UMI for each bin) using the GRbaseCoverage2 and the makeMatrixFrombaseCoverage functions from the ChroKit framework⁸¹. Then, based on the distribution of the BLISS signal coverage across bins, we set an arbitrary threshold on the BLISS signal (threshold of the bin coverage = 2^4.6) above which bins were called as BLISS + . BLISS+ bins were then collapsed into BLISS+ positive regions by merging BLISS+ bins that were separated by a maximum gap of 2 kb. For genomic analyses we used the union of the BLISS+ regions identified in two replicates for each condition (shCDK12+OHT, +OHT alone).

EdU-seq-HU analysis

Paired-end reads from NovaSeq 6000 were aligned using bwa mem software (hg19 genome assembly and UCSC annotation). Murine spike-in reads (genomic DNA from EdU labeled NIH3T3 cells) were aligned the same way, but using the mm10 genome assembly and ENSEMBL annotation. Bam files were normalized by their spike-in using the following procedure: (i) For each sample, a coefficient (C1) was calculated: C1_sample = (1000000/ n of reads) X ratio; where “ratio” is the ratio in ng of spike-in gDNA/ ng sample gDNA. (ii) C1 coefficients were normalized to the C1 coefficient of highest value, thus obtaining the normalized coefficient C2 (C2_sample = C1_sample/max(C1)). (iii) C2 was used to downsample the reads and obtain the spike-in normalized bam files using the samtools program⁷⁹: samtools view -b bam -s C2_sample > downsampled_bam.

The bam files of the replicates for each experimental condition were then merged together (samtools merge) and bigWig were generated using the bamCoverage function of deeptools v. 3.5.1. EdU-Seq HU peaks were called using epic2 peak caller⁸². Briefly, bed paired-end files (.bedpe) were obtained from the spike-in normalized bam files (after step 3) using bedtools:

bedtools bamtobed -i reads.bam -bedpe > reads.bedpe.

Then, the following command produced the final bed files:

epic2 -t reads.bedpe -gn hg19 --gaps-allowed 10 --bin-size 400 --output EdU_regions.bed

Heatmaps and figures were generated using ChroKit framework⁸¹.

EU-seq analysis

Paired-end reads from NovaSeq 6000 were aligned using STAR with the following command: STAR --runThreadN 8 --runMode alignReads --outSAMtype BAM SortedByCoordinate --genomeDir <index > --readFilesIn reads_R1.fq reads_R2.fq. The number of counts for each gene were computed using featureCounts -T 2 -p -P -a gtffile. Differentially expressed genes between experimental conditions (OHT, siCDK12, OHT+siCDK12) and not treated samples were calculated using DESEq2 R Bioconductor package. To evaluate expression data, we calculated TPMs for each gene. To identify TSSs associated to ERR and DSBs, we took BLISS+ regions identified in U2OS cells upon activation of MycER and CDK12 silencing, and selected those overlapping with the left-boundary of an ERR. We then extracted BLISS+ regions having a TSS mapping on the minus strand. We repeated this procedure selecting the right boundaries of the ERR regions and considering the TSS mapping on the plus strand of the BLISS+ region. We then derived the list of gene symbols of all the TSSs associated with these regions.

Additional information concerning antibodies, plasmids, primers, cell lines, siRNA/shRNA sequences, software and other reagents can be found in the supplementary data 4.

Reporting summary

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

Data availability

The genomic data generated in this study have been deposited in the GEO database under accession code GSE236553. Source data are provided with this paper.

Code availability

Source code can be found at: https://github.com/0774v10/CDK12-MYC-transcription-replication⁸³.

References

Macheret, M. & Halazonetis, T. D. DNA replication stress as a hallmark of cancer. Annu Rev. Pathol. 10, 425–448 (2015).
Article PubMed CAS Google Scholar
Burrell, R. A. et al. Replication stress links structural and numerical cancer chromosomal instability. Nature 494, 492–496 (2013).
Article ADS PubMed PubMed Central CAS Google Scholar
Gaillard, H., Garcia-Muse, T. & Aguilera, A. Replication stress and cancer. Nat. Rev. Cancer 15, 276–289 (2015).
Article PubMed CAS Google Scholar
Wilhelm, T., Said, M. & Naim, V. DNA Replication Stress and Chromosomal Instability: Dangerous Liaisons. Genes (Basel) 11, 642 (2020).
Article PubMed CAS Google Scholar
Bowry, A., Kelly, R. D. W. & Petermann, E. Hypertranscription and replication stress in cancer. Trends Cancer 7, 863–877 (2021).
Article PubMed CAS Google Scholar
Halazonetis, T. D., Gorgoulis, V. G. & Bartek, J. An oncogene-induced DNA damage model for cancer development. Science 319, 1352–1355 (2008).
Article ADS PubMed CAS Google Scholar
Baillie, K. E. & Stirling, P. C. Beyond Kinases: Targeting Replication Stress Proteins in Cancer Therapy. Trends Cancer 7, 430–446 (2021).
Article PubMed CAS Google Scholar
da Costa, A., Chowdhury, D., Shapiro, G. I., D’Andrea, A. D. & Konstantinopoulos, P. A. Targeting replication stress in cancer therapy. Nat. Rev. Drug Discov. 22, 38–58 (2023).
Article PubMed Google Scholar
Keller, K. M. et al. Target Actionability Review: a systematic evaluation of replication stress as a therapeutic target for paediatric solid malignancies. Eur. J. Cancer 162, 107–117 (2022).
Article ADS PubMed CAS Google Scholar
Schaub, F. X. et al. Pan-cancer Alterations of the MYC Oncogene and Its Proximal Network across the Cancer Genome Atlas. Cell Syst. 6, 282–300.e282 (2018).
Article PubMed PubMed Central CAS Google Scholar
Kress, T. R., Sabo, A. & Amati, B. MYC: connecting selective transcriptional control to global RNA production. Nat. Rev. Cancer 15, 593–607 (2015).
Article PubMed CAS Google Scholar
Sabo, A. et al. Selective transcriptional regulation by Myc in cellular growth control and lymphomagenesis. Nature 511, 488–492 (2014).
Article ADS PubMed PubMed Central CAS Google Scholar
Wolf, E., Lin, C. Y., Eilers, M. & Levens, D. L. Taming of the beast: shaping Myc-dependent amplification. Trends Cell Biol. 25, 241–248 (2015).
Article PubMed CAS Google Scholar
Lin, C. Y. et al. Transcriptional amplification in tumor cells with elevated c-Myc. Cell 151, 56–67 (2012).
Article PubMed PubMed Central CAS Google Scholar
Nie, Z. et al. c-Myc is a universal amplifier of expressed genes in lymphocytes and embryonic stem cells. Cell 151, 68–79 (2012).
Article PubMed PubMed Central CAS Google Scholar
Soucek, L. et al. Modelling Myc inhibition as a cancer therapy. Nature 455, 679–683 (2008).
Article ADS PubMed PubMed Central CAS Google Scholar
Donati, G. & Amati, B. M. Y. C. and therapy resistance in cancer: risks and opportunities. Mol. Oncol. 16, 3828–3854 (2022).
Article PubMed PubMed Central CAS Google Scholar
Whitfield, J. R., Beaulieu, M. E. & Soucek, L. Strategies to Inhibit Myc and Their Clinical Applicability. Front Cell Dev. Biol. 5, 10 (2017).
Article PubMed PubMed Central Google Scholar
Macheret, M. & Halazonetis, T. D. Intragenic origins due to short G1 phases underlie oncogene-induced DNA replication stress. Nature 555, 112–116 (2018).
Article ADS PubMed PubMed Central CAS Google Scholar
Dominguez-Sola, D. et al. Non-transcriptional control of DNA replication by c-Myc. Nature 448, 445–451 (2007).
Article ADS PubMed CAS Google Scholar
Robinson, K., Asawachaicharn, N., Galloway, D. A. & Grandori, C. c-Myc accelerates S-phase and requires WRN to avoid replication stress. PLoS One 4, e5951 (2009).
Article ADS PubMed PubMed Central Google Scholar
Murga, M. et al. Exploiting oncogene-induced replicative stress for the selective killing of Myc-driven tumors. Nat. Struct. Mol. Biol. 18, 1331–1335 (2011).
Article PubMed PubMed Central CAS Google Scholar
Lecona, E. & Fernandez-Capetillo, O. Targeting ATR in cancer. Nat. Rev. Cancer 18, 586–595 (2018).
Article PubMed CAS Google Scholar
Curti, L. & Campaner, S. MYC-Induced Replicative Stress: A Double-Edged Sword for Cancer Development and Treatment. Int J. Mol. Sci. 22, 6168 (2021).
Article PubMed PubMed Central CAS Google Scholar
Wood, M. A., McMahon, S. B. & Cole, M. D. An ATPase/helicase complex is an essential cofactor for oncogenic transformation by c-Myc. Mol. Cell 5, 321–330 (2000).
Article PubMed CAS Google Scholar
Puccetti, M. V., Adams, C. M., Kushinsky, S. & Eischen, C. M. Smarcal1 and Zranb3 Protect Replication Forks from Myc-Induced DNA Replication Stress. Cancer Res 79, 1612–1623 (2019).
Article PubMed PubMed Central CAS Google Scholar
Rohban, S., Cerutti, A., Morelli, M. J., d’Adda di Fagagna, F. & Campaner, S. The cohesin complex prevents Myc-induced replication stress. Cell Death Dis. 8, e2956 (2017).
Article PubMed PubMed Central CAS Google Scholar
Grandori, C. et al. Werner syndrome protein limits MYC-induced cellular senescence. Genes Dev. 17, 1569–1574 (2003).
Article PubMed PubMed Central CAS Google Scholar
Sato, M. et al. The UVSSA complex alleviates MYC-driven transcription stress. J. Cell Biol. 220, e201807163 (2021).
Article PubMed PubMed Central CAS Google Scholar
Roeschert, I. et al. Combined inhibition of Aurora-A and ATR kinases results in regression of MYCN-amplified neuroblastoma. Nat. Cancer 2, 312–326 (2021).
Article PubMed PubMed Central CAS Google Scholar
Buchel, G. et al. Association with Aurora-A Controls N-MYC-Dependent Promoter Escape and Pause Release of RNA Polymerase II during the Cell Cycle. Cell Rep. 21, 3483–3497 (2017).
Article PubMed PubMed Central Google Scholar
Herold, S. et al. Recruitment of BRCA1 limits MYCN-driven accumulation of stalled RNA polymerase. Nature 567, 545–549 (2019).
Article ADS PubMed PubMed Central CAS Google Scholar
Papadopoulos, D. et al. MYCN recruits the nuclear exosome complex to RNA polymerase II to prevent transcription-replication conflicts. Mol. Cell 82, 159–176 e112 (2022).
Article PubMed CAS Google Scholar
Liu, H., Liu, K. & Dong, Z. Targeting CDK12 for Cancer Therapy: Function, Mechanism, and Drug Discovery. Cancer Res 81, 18–26 (2021).
Article PubMed CAS Google Scholar
Fan, Z. et al. CDK13 cooperates with CDK12 to control global RNA polymerase II processivity. Sci. Adv. 6, eaaz5041 (2020).
Article ADS PubMed PubMed Central CAS Google Scholar
Blazek, D. et al. The Cyclin K/Cdk12 complex maintains genomic stability via regulation of expression of DNA damage response genes. Genes Dev. 25, 2158–2172 (2011).
Article PubMed PubMed Central CAS Google Scholar
Dubbury, S. J., Boutz, P. L. & Sharp, P. A. CDK12 regulates DNA repair genes by suppressing intronic polyadenylation. Nature 564, 141–145 (2018).
Article ADS PubMed PubMed Central CAS Google Scholar
Krajewska, M. et al. CDK12 loss in cancer cells affects DNA damage response genes through premature cleavage and polyadenylation. Nat. Commun. 10, 1757 (2019).
Article ADS PubMed PubMed Central Google Scholar
Lui, G. Y. L., Grandori, C. & Kemp, C. J. CDK12: an emerging therapeutic target for cancer. J. Clin. Pathol. 71, 957–962 (2018).
Article PubMed CAS Google Scholar
Murphy, D. J. et al. Distinct thresholds govern Myc’s biological output in vivo. Cancer Cell 14, 447–457 (2008).
Article PubMed CAS Google Scholar
Paulsen, R. D. et al. A genome-wide siRNA screen reveals diverse cellular processes and pathways that mediate genome stability. Mol. Cell 35, 228–239 (2009).
Article PubMed PubMed Central CAS Google Scholar
Zhou, Y. et al. Metascape provides a biologist-oriented resource for the analysis of systems-level datasets. Nat. Commun. 10, 1523 (2019).
Article ADS PubMed PubMed Central Google Scholar
Chipumuro, E. et al. CDK7 inhibition suppresses super-enhancer-linked oncogenic transcription in MYCN-driven cancer. Cell 159, 1126–1139 (2014).
Article PubMed PubMed Central CAS Google Scholar
Koh, C. M. et al. MYC regulates the core pre-mRNA splicing machinery as an essential step in lymphomagenesis. Nature 523, 96–100 (2015).
Article ADS PubMed CAS Google Scholar
Hsu, T. Y. et al. The spliceosome is a therapeutic vulnerability in MYC-driven cancer. Nature 525, 384–388 (2015).
Article ADS PubMed PubMed Central CAS Google Scholar
Kessler, J. D. et al. A SUMOylation-dependent transcriptional subprogram is required for Myc-driven tumorigenesis. Science 335, 348–353 (2012).
Article ADS PubMed CAS Google Scholar
Toyoshima, M. et al. Functional genomics identifies therapeutic targets for MYC-driven cancer. Proc. Natl Acad. Sci. USA 109, 9545–9550 (2012).
Article ADS PubMed PubMed Central CAS Google Scholar
Sun, Z., Zhang, Z., Wang, Q. Q. & Liu, J. L. Combined Inactivation of CTPS1 and ATR Is Synthetically Lethal to MYC-Overexpressing Cancer Cells. Cancer Res 82, 1013–1024 (2022).
Article PubMed PubMed Central CAS Google Scholar
Cunningham, J. T., Moreno, M. V., Lodi, A., Ronen, S. M. & Ruggero, D. Protein and nucleotide biosynthesis are coupled by a single rate-limiting enzyme, PRPS2, to drive cancer. Cell 157, 1088–1103 (2014).
Article PubMed PubMed Central CAS Google Scholar
Lankes, K. et al. Targeting the ubiquitin-proteasome system in a pancreatic cancer subtype with hyperactive MYC. Mol. Oncol. 14, 3048–3064 (2020).
Article PubMed PubMed Central CAS Google Scholar
Gwynne, W. D. et al. Cancer-selective metabolic vulnerabilities in MYC-amplified medulloblastoma. Cancer Cell 40, 1488–1502 e1487 (2022).
Article PubMed CAS Google Scholar
Magnuson, B. et al. CDK12 regulates co-transcriptional splicing and RNA turnover in human cells. iScience 25, 105030 (2022).
Article ADS PubMed PubMed Central CAS Google Scholar
Heine, G. F., Horwitz, A. A. & Parvin, J. D. Multiple mechanisms contribute to inhibit transcription in response to DNA damage. J. Biol. Chem. 283, 9555–9561 (2008).
Article PubMed PubMed Central CAS Google Scholar
Janicki, S. M. et al. From silencing to gene expression: real-time analysis in single cells. Cell 116, 683–698 (2004).
Article PubMed PubMed Central CAS Google Scholar
Ray Chaudhuri, A. & Nussenzweig, A. The multifaceted roles of PARP1 in DNA repair and chromatin remodelling. Nat. Rev. Mol. Cell Biol. 18, 610–621 (2017).
Article PubMed PubMed Central CAS Google Scholar
Shanbhag, N. M., Rafalska-Metcalf, I. U., Balane-Bolivar, C., Janicki, S. M. & Greenberg, R. A. ATM-dependent chromatin changes silence transcription in cis to DNA double-strand breaks. Cell 141, 970–981 (2010).
Article PubMed PubMed Central CAS Google Scholar
Awwad, S. W., Abu-Zhayia, E. R., Guttmann-Raviv, N. & Ayoub, N. N. E. L. F.- E is recruited to DNA double-strand break sites to promote transcriptional repression and repair. EMBO Rep. 18, 745–764 (2017).
Article PubMed PubMed Central CAS Google Scholar
Yan, W. X. et al. BLISS is a versatile and quantitative method for genome-wide profiling of DNA double-strand breaks. Nat. Commun. 8, 15058 (2017).
Article ADS PubMed PubMed Central CAS Google Scholar
Bouwman, B. A. M. et al. Genome-wide detection of DNA double-strand breaks by in-suspension BLISS. Nat. Protoc. 15, 3894–3941 (2020).
Article PubMed CAS Google Scholar
Min, S., Ji, J. H., Heo, Y. & Cho, H. Transcriptional regulation and chromatin dynamics at DNA double-strand breaks. Exp. Mol. Med 54, 1705–1712 (2022).
Article PubMed PubMed Central CAS Google Scholar
Saxowsky, T. T. & Doetsch, P. W. RNA polymerase encounters with DNA damage: transcription-coupled repair or transcriptional mutagenesis? Chem. Rev. 106, 474–488 (2006).
Article PubMed CAS Google Scholar
Shanbhag, N. M. & Greenberg, R. A. Neighborly DISCourse: DNA double strand breaks silence transcription. Cell Cycle 9, 3635–3636 (2010).
Article PubMed CAS Google Scholar
Wu, Y. M. et al. Inactivation of CDK12 Delineates a Distinct Immunogenic Class of Advanced Prostate Cancer. Cell 173, 1770–1782 e1714 (2018).
Article PubMed CAS Google Scholar
Popova, T. et al. Ovarian Cancers Harboring Inactivating Mutations in CDK12 Display a Distinct Genomic Instability Pattern Characterized by Large Tandem Duplications. Cancer Res 76, 1882–1891 (2016).
Article PubMed CAS Google Scholar
Sørensen, S. G. et al. Pan-cancer association of DNA repair deficiencies with whole-genome mutational patterns. bioRxiv, 2022.2001.2031.478445 (2022).
Menghi, F. et al. The Tandem Duplicator Phenotype Is a Prevalent Genome-Wide Cancer Configuration Driven by Distinct Gene Mutations. Cancer Cell 34, 197–210.e195 (2018).
Article PubMed PubMed Central CAS Google Scholar
Reimers, M. A. et al. Clinical Outcomes in Cyclin-dependent Kinase 12 Mutant Advanced Prostate Cancer. Eur. Urol. 77, 333–341 (2020).
Article PubMed CAS Google Scholar
Yang, Y. et al. Large tandem duplications in cancer result from transcription and DNA replication collision. medRxiv, https://doi.org/10.1101/2023.05.17.23290140 (2023).
Filippone, M. G. et al. CDK12 promotes tumorigenesis but induces vulnerability to therapies inhibiting folate one-carbon metabolism in breast cancer. Nat. Commun. 13, 2642 (2022).
Article ADS PubMed PubMed Central CAS Google Scholar
Rohban, S. & Campaner, S. Myc induced replicative stress response: How to cope with it and exploit it. Biochim Biophys. Acta 1849, 517–524 (2015).
Article PubMed CAS Google Scholar
van den Berg, J. et al. A limited number of double-strand DNA breaks is sufficient to delay cell cycle progression. Nucleic Acids Res 46, 10132–10144 (2018).
Article PubMed PubMed Central Google Scholar
Haeussler, M. et al. Evaluation of off-target and on-target scoring algorithms and integration into the guide RNA selection tool CRISPOR. Genome Biol. 17, 148 (2016).
Article PubMed PubMed Central Google Scholar
Taglialatela, A. et al. Restoration of Replication Fork Stability in BRCA1- and BRCA2-Deficient Cells by Inactivation of SNF2-Family Fork Remodelers. Mol. Cell 68, 414–430 e418 (2017).
Article PubMed PubMed Central CAS Google Scholar
Macheret, M. & Halazonetis, T. D. Monitoring early S-phase origin firing and replication fork movement by sequencing nascent DNA from synchronized cells. Nat. Protoc. 14, 51–67 (2019).
Article PubMed CAS Google Scholar
Blecher-Gonen, R. et al. High-throughput chromatin immunoprecipitation for genome-wide mapping of in vivo protein-DNA interactions and epigenomic states. Nat. Protoc. 8, 539–554 (2013).
Article PubMed Google Scholar
Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014).
Article PubMed PubMed Central Google Scholar
Li, H. & Durbin, R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 25, 1754–1760 (2009).
Article PubMed PubMed Central CAS Google Scholar
Zhang, Y. et al. Model-based analysis of ChIP-Seq (MACS). Genome Biol. 9, R137 (2008).
Article PubMed PubMed Central Google Scholar
Danecek, P. et al. Twelve years of SAMtools and BCFtools. Gigascience 10, giab008 (2021).
Article PubMed PubMed Central Google Scholar
Dsouza, M., Larsen, N. & Overbeek, R. Searching for patterns in genomic data. Trends Genet.: TIG 13, 497–498 (1997).
Article PubMed CAS Google Scholar
Croci, O. & Campaner, S. ChroKit: a Shiny-based framework for interactive analysis, visualization and integration of genomic data. Nucleic Acids Res 51, W83–W92 (2023).
Article PubMed PubMed Central CAS Google Scholar
Stovner, E. B. & Saetrom, P. epic2 efficiently finds diffuse domains in ChIP-seq data. Bioinformatics 35, 4392–4393 (2019).
Article PubMed CAS Google Scholar
Croci, O.; “CDK12 controls transcription at damaged genes and prevents MYC-induced transcription-replication conflicts”; ‘0774v10/cdk12-myc-transcription-replication: Cdk12-myc-transcription-replication V1’. Zenodo, 24 June 2024. https://doi.org/10.5281/zenodo.12515564.

Download references

Acknowledgements

This work was in part supported by the AIRC investigator grant IG-13135 to S.C. R.B. was funded by the Karolinska Institutet KID Funding Program supporting doctoral students. All BLISS experiments were funded through grants by the Swedish Research Council (grant no. 2018-02950) and the Ragnar Söderberg Foundation (Fellows in Medicine 2016) to N.C. We thank Ramiro Vazquez, Bruna Caridi, Gaia Sambruni, Emanuele Martini, Simona Rodighiero e Chiara Soriani (Imaging Unit, IEO), Simona Ronzoni (FACS facility, IEO), Giuseppina Giardina (Screening Unit, IEO), Luca Rotta, Thelma Capra e Salvatore Bianchi (IEO-IIT Genomic Unit) for technical assistance; Bruno Amati, Yilli Docksani, Fabrizio D’adda di Fagagna e Giovanni Tonon for critical discussions and suggestions. We thank Y. Akira, S.P. Jackson, R.H. Medema, B. Majello, N. Ayoub, A. Shibata, G.B. Morin and T. Ikura for sharing reagents.

Author information

Roberto Ballarino
Present address: Department of Cell and Molecular Biology, Karolinska Institutet, Stockholm, Sweden
Mark Wade
Present address: Astex Pharmaceuticals, 436 Cambridge Science Park, CB4 0QA, Cambridge, UK

Authors and Affiliations

Center for Genomic Science of IIT, CGS@SEMM (Istituto Italiano di Tecnologia at European School of Molecular Medicine), Fondazione Istituto Italiano di Tecnologia (IIT), 20139, Milan, Italy
Laura Curti, Sara Rohban, Nicola Bianchi, Ottavio Croci, Adrian Andronache, Michela Mattioli, Fernanda Ricci, Elena Pastori, Silvia Sberna, Simone Bellotti, Anna Accialini, Mark Wade & Stefano Campaner
IFOM ETS, The AIRC Institute of Molecular Oncology, Milan, Italy
Sara Barozzi & Dario Parazzoli
Department of Microbiology, Tumor and Cell Biology, Karolinska Institutet, SE-17165, Stockholm, Sweden
Roberto Ballarino & Nicola Crosetto
Science for Life Laboratory, Tomtebodavägen 23A, SE-17165, Solna, Sweden
Roberto Ballarino & Nicola Crosetto
Human Technopole, Viale Rita Levi-Montalcini 1, 20157, Milan, Italy
Nicola Crosetto

Authors

Laura Curti
View author publications
Search author on:PubMed Google Scholar
Sara Rohban
View author publications
Search author on:PubMed Google Scholar
Nicola Bianchi
View author publications
Search author on:PubMed Google Scholar
Ottavio Croci
View author publications
Search author on:PubMed Google Scholar
Adrian Andronache
View author publications
Search author on:PubMed Google Scholar
Sara Barozzi
View author publications
Search author on:PubMed Google Scholar
Michela Mattioli
View author publications
Search author on:PubMed Google Scholar
Fernanda Ricci
View author publications
Search author on:PubMed Google Scholar
Elena Pastori
View author publications
Search author on:PubMed Google Scholar
Silvia Sberna
View author publications
Search author on:PubMed Google Scholar
Simone Bellotti
View author publications
Search author on:PubMed Google Scholar
Anna Accialini
View author publications
Search author on:PubMed Google Scholar
Roberto Ballarino
View author publications
Search author on:PubMed Google Scholar
Nicola Crosetto
View author publications
Search author on:PubMed Google Scholar
Mark Wade
View author publications
Search author on:PubMed Google Scholar
Dario Parazzoli
View author publications
Search author on:PubMed Google Scholar
Stefano Campaner
View author publications
Search author on:PubMed Google Scholar

Contributions

L.C. project supervision, experimental design, experimental work and data analyses. S.R. designed and carried out the siRNA screen set-up, genomic experiments and molecular analyses. N.B. designed and carried out experiments, data analyses. S. Bellotti, A. Accialini., S.S. and E.P. for experimental work. O.C. carried out bioinformatic and genomic analyses. F.R., M.M. performed the high-throughput siRNA screen, A.A. bioinformatic, statistical and image analyses of the siRNA screen data, M.W. supervised the high-throughput siRNA screen. S.Barozzi designed and carried out live-microscopy experiments, D.P. designed and supervised live-microscopy experiments, R.B. (experiments) and N.C. (supervision) for sBLISS experiments. L.C., N.B. and O.C. contributed to manuscript preparation. S.C. study conceptualization, project supervision, data analyses and manuscript preparation. All the Authors revised the manuscript.

Corresponding author

Correspondence to Stefano Campaner.

Ethics declarations

Competing interests

S.C. and L.C. declare a patent application (IT102021000017414, PCT/IB2022/056112, grant date 03/08/2023). The remaining Authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks Oscar Fernandez-Capetillo, Marcos Malumbres, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Peer Review File

Description of Additional Supplementary Files

Supplementary Data 1

Supplementary Data 2

Supplementary Data 3

Supplementary Data 4

Supplementary Movie 1

Supplementary Movie 2

Supplementary Movie 3

Supplementary Movie 4

Supplementary Movie 5

Supplementary Movie 6

Reporting Summary

Source data

Source Data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Cite this article

Curti, L., Rohban, S., Bianchi, N. et al. CDK12 controls transcription at damaged genes and prevents MYC-induced transcription-replication conflicts. Nat Commun 15, 7100 (2024). https://doi.org/10.1038/s41467-024-51229-5

Download citation

Received: 15 November 2023
Accepted: 01 August 2024
Published: 18 August 2024
DOI: https://doi.org/10.1038/s41467-024-51229-5

This article is cited by

Senataxin prevents replicative stress induced by the Myc oncogene
- Silvia Sberna
- Marco Filipuzzi
- Stefano Campaner
Cell Death & Disease (2025)
Transcriptional and epigenetic rewiring by the NUP98::KDM5A fusion oncoprotein directly activates CDK12
- Selina Troester
- Thomas Eder
- Florian Grebien
Nature Communications (2025)