DNA-PKcs-mediated transcriptional regulation of TOP2B drives chemoresistance in acute myeloid leukemia

Although 60-70% of adults with acute myeloid leukemia (AML) respond to the chemotherapy regimen of cytarabine and anthracycline drugs, more than 80% of patients relapse due to acquired resistance (Döhner et al., 2017). The primary mechanism of action of anthracycline is the inhibition of topoisomerase II (TOP2) enzyme activity, leading to irreparable replication stress, DNA damage and cell death (Gewirtz, 1999). Multiple mechanisms, such as altered metabolic preferences (Farge et al., 2017; Salunkhe et al., 2020), reduced drug accumulation (Durmus et al., 2014; Graydon Harker et al., 1989; van der Kolk et al., 2000), enhanced DNA repair (Salunkhe et al., 2018; Stefanski et al., 2019), activation of survival pathways (Chen et al., 2018; Piya et al., 2016) and alterations in the drug target TOP2, either due to decreased levels of TOP2B or the presence of truncated non-functional TOP2A (Harker et al., 1991; Kanagasabai et al., 2018), have been ascribed to anthracycline resistance. We have also shown that aggressively resistant AML cells evolve to deregulate TOP2, rendering them unresponsive to anthracyclines (Salunkhe et al., 2018). However, the mechanisms underlying this topoisomerase dysregulation remain unclear.

DNA-dependent protein kinase catalytic subunit (DNA-PKcs) plays an essential role in maintaining genomic stability through non-homologous end joining (NHEJ) (Yoo and Dynan, 1999). Moreover, decreased therapeutic response and poor prognosis (Beskow et al., 2009; Bouchaert et al., 2012; Chen et al., 2021) have been shown to correlate with DNA-PKcs expression in multiple cancers, even in the absence of DNA-damage-inducing therapies (Evert et al., 2013; Willmore et al., 2008), suggesting a DNA-damage response (DDR)-independent role in cancer. DNA-PKcs has also been reported to alter multiple signaling pathways other than DDR, such as metabolism, immune response, hypoxia-induced signaling and transcriptional regulation (Goodwin and Knudsen, 2014; Goodwin et al., 2015). Interestingly, DNA-PKcs was initially characterized as part of the SP1 transcriptional complex (Jackson et al., 1990), a regulatory component of RNA polymerase II (RNAPII) (Dvir et al., 1992), which is an interactor of the basal transcriptional machinery (Maldonado et al., 1996), and was recruited to the active sites of transcription (Ju et al., 2006). These findings illustrate the significance of DNA-PKcs as a transcriptional modulator. However, its role in anthracycline resistance regulation remains unclear.

Using the anthracycline-resistant cellular model developed in our laboratory and samples from individuals with AML, we demonstrate that DNA-PKcs is a therapeutic target for anthracycline-resistant AML and delineate its non-canonical function in mediating acquired resistance to anthracycline.

To evaluate the mechanism of acquired resistance to anthracyclines, we used two leukemic cell lines (HL-60 and THP-1) to generate cellular models of anthracycline-acquired resistance (Fig. 1A). The resistant subline developed previously in our lab, as described by Salunkhe et al., (2018), from THP-1 cells was named THP-1/LDRP (late-drug resistant population). Doxorubicin IC₉₀ for THP-1 cells increased from 1.437 µM to 2.81 µM for THP-1/LDRP cells. Another anthracycline (mitoxantrone)-resistant sub-cell line (HL-60/MX2) developed by Graydon Harker et al. (1989) from HL-60 cells using a similar strategy was procured from the ATCC. These resistant sublines showed multidrug-resistant (MDR) phenotypes and were cross-resistant to doxorubicin, daunorubicin, mitoxantrone and cytarabine (Salunkhe et al., 2018, 2020) (Fig. S1A,B). Furthermore, THP-1/LDRP and HL-60/MX2 cells were more tumorigenic in vitro (Salunkhe et al., 2018) and in vivo in NOD-SCID mice (Fig. 1B). As seen from the Kaplan–Meier (KM) survival plot in Fig. 1B, mice injected with 2.5 million resistant cells reached a humane endpoint early, indicating significantly lower overall survival (OS) than those with tumors from an equal number of sensitive parent cells.

Importantly, mitoxantrone, which induces apoptosis in cancer cells by triggering DNA double-strand breaks (DSBs) in the genome and acts as a TOP2 poison (Minotti et al., 2004; Shandilya et al., 2020; Yaqub, 2013), did not induce DNA damage in both resistant cells (HL-60/MX2 and THP-1/LDRP), as seen in the neutral comet assay after treating the cells with the lowest IC₅₀ across all four cell lines, which was 50 nM mitoxantrone. There was a significant (P<0.001) increase in the tail moment for both HL-60 (2.98-fold)- and THP-1 (8.39-fold)-sensitive cells 24 h post-treatment. However, there was no significant change in the tail moment of resistant cells (HL-60/MX2 and THP-1/LDRP) after mitoxantrone treatment (Fig. 1C). Moreover, mitoxantrone induced G2/M- and S-phase cell cycle arrest only in sensitive cells, but not in resistant cells (Fig. 1D), despite similar or higher mitoxantrone uptake and retention in resistant cells, as indicated by the median fluorescence intensity (MFI) at 561/695 nm after 2 and 24 h of treatment (Fig. 1E). This further confirmed mitoxantrone resistance in HL-60/MX2 and THP-1/LDRP cells. As the resistant sub-cell lines recapitulated the clinical AML relapse characteristics, they appeared to be a valuable tool for identifying dysregulated pathways, making them suitable for further mechanistic studies of anthracycline resistance.

We hypothesized that resistant cells have an efficient DSB repair system owing to the higher expression of DDR proteins; thus, mitoxantrone could not induce DSBs. Hence, we analyzed the expression of genes associated with DDR in RNA-seq data from HL-60 and HL-60/MX2 cells, as well as two independent public datasets GSE83533 [dbGaP28 of RNA-seq data from 19 paired AML patients (diagnosis and relapse) representing acquired resistance; Li et al., 2016] and GSE103424 [RNA-seq from 18 AML patients who responded to therapy and achieved complete remission (CR) compared with 18 resistant (refractory) patients (non-CR), representing inherent resistance; Chiu et al., 2019]. We observed that the majority of DDR genes were not differentially expressed in either the dataset or the cell lines. Interestingly, PRKDC and XRCC6 showed 1.66-fold (P<0.001) and 1.21-fold (P<0.001) increased expression, respectively, in resistant HL-60/MX2 cells compared with sensitive HL-60 cells (Fig. 2A; Fig. S2A,B); however, in relapsed and non-CR AML patients, only PRKDC showed 1.32-fold (P<0.01) and 1.18-fold (P=0.25) higher expression, respectively (Fig. 2B;,Fig. S2C–E). Consistent with the RNA-seq data analysis in both the resistant cellular model and patient datasets, validation of mRNA expression of major DDR proteins using quantitative PCR (qPCR) showed that of all the DDR genes tested, PRKDC was the only gene that was significantly upregulated, i.e. 1.42-fold and 2.68-fold in both HL-60/MX2 and THP-1/LDRP cells, respectively, compared with their respective sensitive cells (Fig. 2C).

Moreover, of all the DDR proteins analyzed by immunoblotting, only DNA-PKcs (which is encoded by PRKDC) expression was significantly higher in both HL-60/MX2 (1.74-fold) and THP-1/LDRP cells (1.51-fold) (Fig. 2D). This was further confirmed by immunofluorescence, which also showed significantly high DNA-PKcs protein levels in resistant AML cells (Fig. 3A,B). Importantly, we found significantly (P<0.001) higher expression of DNA-PKcs by immunohistochemistry-immunofluorescence (IHC-IF) in nucleated cells present in bone marrow biopsies from individuals with AML (n=20) than in baseline samples (n=20) (Fig. 3C). Furthermore, IHC-IF of γH2AX in the relapse and baseline patient biopsies (n=20) did not show any differential γH2AX expression (Fig. 3D), confirming the higher expression of DNA-PKcs in resistant cells and in relapsed AML patients independently of DNA damage.

To investigate the functional relevance of DNA-PKcs overexpression in mitoxantrone resistance, we performed siRNA-mediated knockdown of PRKDC in resistant HL-60/MX2 and THP-1/LDRP cells. We did not observe any changes in the viability of the DNA-PKcs knockdown cells. However, when these cells were treated with 50 nM and 100 nM mitoxantrone for 48 h post-siPRKDC, a significant decrease in the viability of HL-60/MX2 (50%) and THP-1/LDRP (80%) cells was observed (Fig. 4A). These data suggest a role for DNA-PKcs in rendering mitoxantrone resistance in resistant AML cells.

As DNA-PKcs is a kinase, we used NU7026, a competitive inhibitor of DNA-PKcs kinase activity, and treated both resistant and sensitive cells with 30 µM NU7026. This concentration significantly decreased S2056 phosphorylation of DNA-PKcs (Fig. S3A). Notably, NU7026 alone did not affect the viability of the sensitive or resistant cells (Fig. S3B). However, in combination with mitoxantrone, both sensitive and resistant cell viability were affected, as indicated by the increase in the percentage of propidium iodide (PI)-positive cells. Similar to DNA-PKcs knockdown, inhibition of DNA-PKcs kinase activity resulted in HL-60/MX2 and THP-1/LDRP cells susceptible to 50 nM mitoxantrone, reducing cell viability by >50% and apoptosis by 25% post-96 h and 48 h, respectively (Fig. 4B,C; Fig. S3C). Additionally, we observed that after treatment with DMSO, mitoxantrone or NU7026, resistant cells formed subcutaneous tumors when injected with an equal number of viable cells in NOD/SCID mice. These mice were sacrificed on the 25th day post-injection due to an overgrown tumor. However, combining NU7026 (30 µM) and mitoxantrone (50 nM) prevented the formation of subcutaneous tumors in NOD-SCID mice (Fig. 4D), indicating decreased tumorigenicity of resistant cells with the combination treatment.

Importantly, we subsequently wanted to confirm whether NU7027 and mitoxantrone would have similar combinatorial effects in primary cultures of AML patient samples. For this, we first treated baseline and relapse patient samples (n=3) for 72 h with a logarithmic dilution series ranging from 6.25 nM to 3200 nM concentrations of mitoxantrone, both alone and in combination with 30 µM NU7026. Indeed, cells of relapse patients were resistant (33% cell death) even to 1600 nM mitoxantrone compared with baseline patient cells (Fig. S3D), as determined by the MTT assay. Notably, treatment of AML primary cultures (n=6; Fig. S3E) with 30 µM NU7026 significantly increased the sensitivity of relapse cells to as low as 12.5 nM mitoxantrone, where 12.5 nM mitoxantrone alone showed only 4.36% cell death, which increased to tenfold (46.8%) in combination with 30 µM NU7026 (Fig. 4E). Together, these data demonstrate that inhibition or absence of DNA-PKcs kinase activity renders resistant and relapsed AML cells susceptible to mitoxantrone treatment.

Next, we sought to understand the mechanism of DNA-PKcs inhibition-mediated re-sensitization of resistant cells to mitoxantrone. Given that mitoxantrone induces cell death via DSBs induction, we speculated that, in the absence of DNA-PKcs, mitoxantrone induces DSBs in resistant cells. This was confirmed using a neutral comet assay after post-treatment with 30 µM NU7026, 50 nM mitoxantrone, or a combination of 30 µM NU7026 and 50 nM mitoxantrone for 48 h. A significant increase (P<0.001) in the tail moment of resistant HL-60/MX2 cells treated with the combination of 30 µM NU7026 and 50 nM mitoxantrone, when compared with NU7026 or mitoxantrone-alone treatment (Fig. 5A), showed that inhibition of DNA-PKcs activity potentiates mitoxantrone-mediated induction of DSBs in resistant cells, thereby leading to cell death, even at the IC₅₀ of sensitive parent cells.

We then studied how mitoxantrone induces DSBs in the absence of DNA-PKcs. Anthracyclines primarily cause DNA DSBs by stabilizing the DNA–TOP2 complex, leading to irreversible DNA damage (Shandilya et al., 2020; Yaqub, 2013). AML cells acquire anthracycline resistance by adapting to either structural changes or expression of the drug target topoisomerase enzyme (Beck et al., 1993; Burgess et al., 2008; Harker et al., 1995; Kanagasabai et al., 2017, 2018; Yu et al., 1997), in addition to other mechanisms. To begin with, deregulation of TOP2 was deemed the most relevant to anthracycline resistance. Therefore, we first assessed the status of TOP2A and TOP2B in resistant cells and observed a significant reduction (>80%) in TOP2B at both the transcript and protein levels in resistant cells compared with their sensitive counterparts (Fig. 5B,C), but not in TOP2A.

It was intriguing to realize that mitoxantrone could not induce DNA breaks in these cells, despite the presence of the TOP2A protein in resistant cells. Reports suggest that the retention of intron 33 in TOP2A cDNA results in mutant nonfunctional TOP2A (Harker et al., 1995; Kanagasabai et al., 2017; Mirski et al., 2000). Therefore, we reasoned that TOP2A in resistant AML cells might be a mutant non-functional enzyme. However, we did not find an exon 33-intron 33 amplicon in the cDNA of the resistant cells (Fig. S4A) using primers specifically designed to amplify intron 33 (Harker et al., 1991). Moreover, using the method developed by Shiraishi et al. (2022) to detect intron retention using transcriptome data in an unbiased manner, we analyzed the HL-60 and HL-60/MX2 RNA-seq data, and found no intron retention in TOP2A (Fig. S4B).

Interestingly, TOP2A formed topoisomerase–DNA covalent complexes in resistant HL-60/MX2 cells, albeit to a lesser extent than in sensitive cells, as seen from the trapped in agarose immunostaining (TARDIS) assay (Fig. S4C). However, the nuclear lysates of resistant cells did not show in vitro decatenation activity when incubated with catenated kinetoplast DNA, unlike the nuclear lysates of sensitive cells (Fig. S4D), suggesting that TOP2A can form covalent complexes with DNA, but may not be functional.

We then investigated TOP2B protein levels in bone marrow biopsies from individuals with acute myeloid leukemia (AML) at diagnosis and relapse (n=20). Using IHC-IF, we found that TOP2B expression was significantly downregulated in patients with relapse compared with the baseline levels (Fig. 5D). The mechanism underlying TOP2B downregulation, which causes anthracycline resistance, remains unknown. Therefore, we investigated the possibility of hypermethylation or inhibitory mutations in the TOP2B promoter region leading to its low expression in resistant cells; however, we found no effects of a demethylating agent or mutations through Sanger sequencing (Fig. S5A,B). These findings suggest that changes in the promoter region are not responsible for the low TOP2B expression in resistant cells.

We ruled out reactive oxygen species (ROS) accumulation causing DSBs and apoptosis upon DNA-PKcs inhibition, because DCFDA (2′,7′-dichlorofluorescein diacetate) assays showed no change in ROS levels in both sensitive and resistant cells (Fig. 6A). Therefore, we investigated whether drug target expression was altered after DNA-PKcs knockdown in order to re-sensitize anthracycline-resistant cells. We performed siRNA-mediated knockdown of DNA-PKcs in HL-60/MX2 and THP-1/LDRP cells, and analyzed the TOP2A and TOP2B expression levels. There was no change in TOP2A transcript levels (Fig. 6B). Surprisingly, TOP2B expression was 2- and 1.6-fold higher in HL-60/MX2 and THP-1/LDRP cells, respectively, than that in DNA-PKcs wild-type cells. TOP2B expression also increased by 1.3-fold after NU7026 treatment of HL-60/MX2 cells (Fig. 6C;  Fig. S5D). Importantly, TOP2B mRNA expression, but not TOP2A, increased upon NU7026 treatment of primary cultures from six relapse patients (Fig. 6D). Protein levels of TOP2A also remained unchanged, whereas TOP2B expression was significantly higher in the absence of DNA-PKcs in resistant cell lines (Fig. 6E–G; Fig. S5C). The data showed that, in resistant relapse cells, DNA-PKcs kinase activity suppressed TOP2B expression. Furthermore, rescue experiments confirmed that re-sensitization of resistant cells to mitoxantrone was due to the presence of re-expressed TOP2B, as knockdown of re-expressed TOP2B reversed the DNA-PKcs knockdown phenotype, making the cells significantly more resistant to mitoxantrone (P<0.05) (Fig. 6H,I).

Our data demonstrated that AML cells acquire anthracycline resistance by overexpressing DNA-PKcs, which represses TOP2B expression. To begin with, we analyzed the ChIP sequencing data for phosphoDNA-PK from fibroblast cells (Chandler et al., 2014). We identified enrichment of the pDNA-PK peak over the cis-regulatory region (chr3:25,663,611-25,669,075) of TOP2B (Fig. S6A). Furthermore, we performed ChIP-qPCR to determine whether DNA-PKcs directly binds to the TOP2B promoter in our cellular model. We pulled down chromatin from HL-60 and HL-60/MX2 cells with DNA-PKcs or IgG antibodies, and analyzed them via qPCR with ten primer sets covering the TOP2B promoter region [−1252 bp to +169 bp relative to the transcription start site (TSS)]. Primers for the PLA1 and TMPRSS2 promoter regions were used as positive controls (Goodwin et al., 2015) for DNA-PKC binding. We observed significant DNA-PKcs binding enrichment (2.5% of input) in the specific region chr3:25665843-25665374, i.e. −936 bp to −467 bp from the TSS of TOP2B in HL-60/MX2 cells, unlike in HL-60 cells (Fig. 7A; Fig. S6F). The results showed that DNA-PKcs occupy the TOP2B promoter upstream sequence. We examined whether DNA-PKcs inhibition affected the TOP2B promoter activity by cloning the TOP2B promoter region into the pGL3.basic vector and performing a luciferase promoter assay in transfected 293FT cells. After treatment with 30 µM NU7026 for 48 h, luciferase activity increased 1.94-fold compared with that of the control, indicating that DNA-PKcs regulates TOP2B promoter activity (Fig. 7B).

We determined the minimal region upstream of the TOP2B gene where DNA-PKcs binds to regulate its expression. We serially deleted the five sequences from the F1_TOP2B clone and created F2_TOP2B, F3_TOP2B and F4_TOP2B constructs (Fig. 7C). These were cloned upstream of the luciferase gene into the pGL3-basic vector and transfected into 293FT cells for promoter assay. As shown in Fig. 7C, F1_TOP2B, F2_TOP2B and F3_TOP2B treatment with a DNA-PKcs inhibitor led to an over twofold increase in luciferase activity. However, no change was noted in F4_TOP2B compared with DMSO-treated cells, indicating that DNA-PKcs binds to the 223 bp sequence (−547 to −324 of the TSS) region between F3 and F4 to repress TOP2B gene expression.

To verify DNA-PKcs binding to this region, we conducted an electrophoretic mobility supershift assay (EMSA) with a 5′-biotin-labeled sequence probe corresponding to the 223 bp TOP2B promoter region and nuclear lysates from DNA-PKcs inhibitor-treated and DMSO-treated resistant cells. The DNA-PKcs antibody (lanes 4 and 8) caused an upward shift in the bands compared with lanes 3 and 7, confirming that DNA-PKcs binds to the 223 bp sequence (Fig. 7D; Fig. S6B,C). Additionally, the commercial DNA-PK enzyme bound to the TOP2B promoter sequence (Fig. S6D) but only in the presence of ATP. No DNA-PK binding to the probe was observed in the absence of ATP in the binding buffer, suggesting kinase activity-dependent DNA-PKcs binding but not KU70/80 (Fig. S6E). As EMSA is a qualitative assay to quantify DNA-PKcs binding to the TOP2B promoter sequence, we performed ChIP with DNA-PKcs from HL-60 and HL-60/MX2 cells treated with NU7026 and DMSO. DNA-PKcs binding to the 223 bp TOP2B promoter region (−547 bp to −324 bp) decreased by more than 50% in the presence of NU7026 (30 µM) (Fig. 7E; Fig. S6G), indicating that DNA-PKcs binds to this region and represses TOP2B transcription, which is moderated by its kinase activity.

To understand the mechanism of DNA-PKcs overexpression in resistant cells, we investigated the possibility of copy number gain and epigenetic changes (Gibney and Nolan, 2010; Santarius et al., 2010) as reasons for DNA-PKcs overexpression. First, we analyzed the copy number of PRKDC in HL-60/MX2 and THP-1/LDRP cells compared with their sensitive counterparts, HL-60 and THP-1, respectively. We performed qPCR on four randomly chosen PRKDC exons (exons 85, 66, 33 and 82) and three exon-intron junctions (exon-intron junctions 1 and 18) encompassing the entire PRKDC gene. The change in the CT value (Δ CT) in resistant cells was not greater than 2.5 (Fig. S7A), suggesting that there was no change in the copy number of PRKDC in resistant cells.

Analysis of epigenetic modifiers (Marakulina et al., 2023) from the RNA-seq data of HL-60 and HL-60/MX2 cells did not show differential expression (Fig. S7B). Further validation of genes belonging to histone acetyltransferases, histone deacetylases, histone methyltransferases, demethylases, high-mobility group proteins, methylcytosine dioxygenases and DNA methyltransferase families by qPCR showed increased expression of only KAT2A (GCN5), a histone acetyltransferase enzyme, in both resistant cell lines (Fig. 8A). In addition, acetylation of lysine residues 27, 18 and 9 in H3 increased in the chromatin fraction of HL-60/MX2-resistant cells. Except for low H3K79-di-methylation, other modifications were non-differential in resistant cells (Fig. 8B). Moreover, GCN5 was overexpressed at the protein level in both resistant cells and AML relapse patients (Fig. 8C,D). With increased GCN5 and H3K27ac levels, we investigated whether GCN5-induced histone acetylation increased DNA-PKcs expression in resistant cells.

GCN5 is a transcriptional activator (Haque et al., 2021; Kapoor-Vazirani et al., 2008; Shen et al., 2014). Therefore, we examined the role of GCN5 in DNA-PKcs expression using siRNA-mediated GCN5 knockdown in resistant cells, which significantly reduced DNA-PKcs expression in both resistant cell lines (Fig. 8E). We then assessed whether GCN5 acetyltransferase activity influenced DNA-PKcs expression. After treating resistant AML cells with the GCN5 acetyltransferase inhibitor butyrolactone 3 (BL3) (Salunkhe et al., 2018), a 30% decrease in DNA-PKcs expression was observed (Fig. 8F).

To investigate the direct binding of GCN5 to the PRKDC promoter, we analyzed the public GCN5 ChIP-seq data (Guo et al., 2020). We found GCN5 enrichment in the upstream region of the PRKDC promoter (chr8:47,956,427-47,972,423) (Fig. S7C). ChIP-qPCR analysis using HL-60 and HL-60/MX2 chromatin confirmed this, showing a 1.5-fold enrichment of H3K27Ac at −710 to −212 bp to PRKDC TSS and GCN5 at −399 to −212 bp to PRKDC TSS (Fig. 8G; Fig S7D) in HL-60/MX2. To measure the effect of GCN5 on PRKDC gene expression, we performed a luciferase activity assay in GCN5 knockdown 293FT cells and found decreased promoter activity. Co-transfection with the GCN5 overexpression vector (Lerin et al., 2006) increased promoter activity twofold (Fig. 8H). Overexpression of catalytically inactive GCN5 (Lerin et al., 2006) had no effect on promoter activity, and 100 µM BL3 treatment reduced promoter activity (Fig. 8H), suggesting that GCN5 activates PRKDC transcription and that this activation depends on GCN5 acetyltransferase activity.

Refractory AML relapse is difficult to treat and is associated with poor survival. To target this, we must comprehend the molecular mechanisms underlying chemoresistance. In the present study, we developed an evolutionary cellular model with which to study anthracycline resistance in AML cell lines, using the well-established strategy of long-term exposure and drug escalation protocol adopted to better understand the resistance mechanism (Ricciardi et al., 2015; Sales Amaral et al., 2019; Szakács et al., 2004). These studies have contributed significantly to clinical therapeutic strategies (Binkhathlan and Lavasanifar, 2013; Donawho et al., 2007; Lehmann et al., 2012; Pepper et al., 2008; Underhill et al., 2011). Therefore, despite limitations, these cellular models remain clinically relevant and important for mechanistic studies. Moreover, we validated the clinical relevance of findings from resistant cellular model in the AML bone marrow samples, primary cells from relapse AML patients and in the patient cohort data from independent studies.

Our model shows that AML-acquired resistance can be divided into early and late resistance (Salunkhe et al., 2018), with late-resistant cells recapitulating clinical AML relapse. Using a clinically relevant late anthracycline-resistant cell line and patient-derived primary cultures, we found that chemotherapy failure in relapsed cells was caused by the inability of anthracyclines to induce DSBs. Different mechanisms (Zhang et al., 2019) have been proposed to explain the lack of DNA damage (Salunkhe et al., 2018; Song et al., 2009) in AML resistance. The major causes have been reported to be changes in the TOP2 enzyme (Harker et al., 1991, 1995; Kanagasabai et al., 2018) due to TOP2B downregulation and the presence of the non-functional TOP2A subtype caused by the retention of intron 33 with exon 33. In concordance with these reports, we also found that TOP2B was downregulated. We did not detect any mutant TOP2A subtype. However, for unknown reasons, we observed that, although TOP2A could form a covalent complex with DNA in resistant cells, it lacked the decatenation activity. Additionally, the mechanism by which resistant cells compensate for the lack of the TOP2B enzyme, and possibly non-functional TOP2A, poses an exciting question for further exploration.

Despite the absence of DSBs in resistant cells after anthracycline exposure, DNA-PKcs was overexpressed at both transcript and protein levels. KU80, BRCA2, CHEK1 and NHEJ1 transcripts were higher in HL-60/MX2 cells and ATR protein levels were higher in THP-1/LDRP cells. To identify the mechanisms common to both resistant cell lines, DNA-PKcs expression was validated in relapsed refractory AML patients. DNA-PKcs is a key kinase in NHEJ repair (Chan et al., 2002). Overexpression of DNA-PKcs is associated with poor therapeutic response (Bouchaert et al., 2012; Goodwin and Knudsen, 2014; Goodwin et al., 2013; Willmore et al., 2008) in multiple cancers. Moreover, we observed significantly poor survival in individuals with high PRKDC and low TOP2B expression in the LAML-TCGA dataset (Fig. S2F). However, these studies were association studies. We demonstrated that pharmacological or genetic inhibition of DNA-PKcs re-sensitizes resistant AML cell lines and relapse primary patient samples to anthracycline. Mikusová et al. (2011) also reported an increase in the susceptibility of HL-60/MX2 cells to mitoxantrone upon inhibition of DNA-PKcs kinase activity; however, the mechanism underlying this susceptibility remains unclear. Willmore et al. (2004) also reported inhibition of the NHEJ repair pathway and G2/M arrest as the mechanism for NU7026 potentiation of anthracyclines in chronic myelogenous leukemia (CML) cell line K562. Here, we discovered that TOP2B gene repression mediated by DNA-PKcs directly binding to its promoter upstream region to be the mitoxantrone resistance mechanism in AML relapse. Thus, DNA-PKcs perturbation makes resistant relapse cells susceptible. Although most studies have focused on the role of DNA-PKcs in DNA DSBs, earlier biochemical studies have speculated that DNA-PK acts as a transcription modulator and have demonstrated that DNA-PKcs interacts and regulates the network of known transcription factors and binds to the transcriptional sites (Goodwin et al., 2013; Jackson et al., 1990). In addition to the previously unreported regulation of TOP2B transcription, as detailed in this study, we cannot rule out the possibility that DNA-PKcs regulates the expression of other genes to impart resistance phenotypes in relapsed AML cells.

In summary, this study demonstrates that, in AML-resistant cells, GCN5 (KAT2A) mediates the transcriptional upregulation of DNA-PKcs, which consequently functions as a transcriptional repressor of the anthracycline target protein TOP2B. Therefore, inhibition of DNA-PKcs alleviate this repression by re-sensitizing resistant AML cells to anthracycline (Fig. 8I). These data not only provide new mechanistic insights into the functions of DNA-PKcs but also provide clinically relevant information that merits further evaluation of DNA-PKcs inhibition in a larger cohort of AML patient samples that represent different degrees of heterogeneity and subtypes of AML.

All specimens used to establish primary cell cultures were obtained from Tata Memorial Hospital (Mumbai, India) after approval by the institutional ethics committee. All patients gave their written informed consent. 5-10 ml peripheral blood of AML patients was collected in EDTA vials. Age and sex information from baseline patients are B1 [45, female (F)], B2 [58, Male (M)], B3 (32, M), B4 (41, F), B5 (22, M) and B6 (18, M); relapse patients are R1 (45, M), R2 (23, F), R3 (39, M), R4 (44, F), R5 (31, M) and R6 (46, M).

The obtained samples were processed to isolate mononuclear cells (PBMCs) using the Ficoll-Paque density gradient method. In a 15 ml centrifuge tube, blood was diluted using twice the volume of pre-warmed 1×PBS and mixed by inverting the tube several times. 3 ml Ficoll-Paque media was added to another 15 ml centrifuge tube, followed by careful layering of the 6 ml diluted blood sample onto Ficoll, ensuring the diluted blood and Ficoll did not mix. To separate the content based on density, the Ficoll-blood layer was centrifuged at 400 g for 30 min at 18°C to 20°C without any brakes. The interface layer containing mononuclear cells was transferred to a sterile tube and washed twice gently using 1×PBS at 400 g for 5 min at 18°C to 20°C. The cell pellet was immediately cryo-preserved in FBS-based cryoprotectant media at −196°C for further experiments. For the drug cytotoxicity assay, cell pellets were thawed at 37°C and cultured for 24 h before starting the treatment and further MTT assay.

HL-60 and THP-1 cell lines were obtained from NCCS Pune, and the HL-60/MX2 cell line was obtained from ATCC (CRL 2257). Cells were maintained in the humidified CO₂ incubator at 37°C temperature, ≥80% relative humidity and 5% CO₂ in the RPMI-1640 media (Gibco, 31800022) with 10% FBS (Gibco, 10270-106) and 1×antibiotic-antimycotic solution (Himedia-A002). 293FT cells (a kind gift from Dr Amit Dutt, TMC-ACTREC, India) were cultured under similar conditions but in DMEM media (Gibco, 12800-017) with 10% FBS. Short tandem repeat (STR) markers profiles were determined for all cell lines using PROMEGA STR profiling 10 markers kit and were authenticated in the ATCC STR database (see Table S3).

NOD-SCID mice have been obtained by the in-house Laboratory Animal Facility (LAF) from The Jackson Laboratory, USA. LAF breeding and maintenance was carried out as per the guidelines approved by the Committee for the Purpose of Control and Supervision of Experiments on Animals, Government of India (CPCSEA). The Institute's animal ethics committee approved the use of NOD-SCID mice and all procedures followed for this study.

2.5 million viable cells were untreated or treated with 30 µM DNA-PK inhibitor NU7026 (Sigma, N1537), 50 nM mitoxantrone (Sun Pharma), a combination of both, and DMSO as vehicle control for 48 h resuspended in 100 µl 1×PBS were injected subcutaneously into 4- to 6-week-old NOD-SCID mice (n=9 for each treatment group). After injection, tumor growth was monitored for at least 25 days. Tumor size was measured using Vernier Caliper.

Twenty bone marrow biopsy FFPE blocks of AML confirmed baseline and relapse patients were acquired from the Pathology Department at Tata Memorial Hospital (Mumbai, India) after the approval of the institutional ethics committee (TMC-IEC III). The committee waived written informed consent for retrospective FFPE block collections.

Briefly, 1×104 cells resuspended in 1×PBS were diluted in warm LMPA (Invitrogen, 16520-050) at 42°C such that 25 µl resuspension has ∼100 cells. The volume of 25 µl was evenly spread with the help of a coverslip on frosted glass slides coated with 1%agarose as a thin layer. Slides were left on ice for 30 min so that LM agarose-cell suspension is solidified properly. Coverslips were then gently removed, such that the agarose layer is not disturbed. Slides were then incubated in pre-cooled lysis buffer [pH 8.0; 2.5 M NaCl,100 mM EDTA, 10 mM TrisCl, 0.05 mg/ml Proteinase K, 1% (v/v) Triton-X and 10% (v/v) DMSO] at 4°C for 1 h. Slides were rinsed three times with ultrapure water and followed by electrophoresis in fresh 1×TAE buffer (pH 8.5) for 10-15 min at a constant voltage of 15 V. Slides were then immediately dehydrated in 100% ethanol for 20 min, allowed to dry at room temperature and stained with 3.35 μg/ml propidium iodide (PI) for 5 min. Slides were then rinsed once using fresh ultrapure water before image acquisition. Images were captured using a fluorescent Carl Zeiss Axio Imager Z1 upright microscope with a 20× objective and quantified using Open Comet Score software. Comets were scored using the OpenComet software in terms of the tail moment, which takes both the migration of DNA (tail length) as well as the amount of DNA in the comet (intensity of comet) into consideration. At least 50 cells were counted, and the average was plotted for the respective time points.

Cells (HL-60, HL-60/MX2, THP-1 and THP-1/LDRP) treated with 50 nM mitoxantrone for 24 h were harvested and resuspended in ice-cold 300 µl 1×PBS. 700 µl chilled ethanol (final concentration 70% v/v) was added dropwise and incubated on ice for 2 h with intermittent mixing (by tube tapping) for fixation. Pellets of fixed cells were washed once in 1×PBS and then resuspended in staining buffer (PBS with 100 μg/ml RNase A, 50 μg/ml propidium iodide and 0.1% Triton-X100) at 4°C overnight before acquisition by flow cytometry (Thermo Fischer-Attune). The acquired file was analyzed using Modfit LT v5 software.

For mitoxantrone drug uptake, treated cells (50 nM mitoxantrone treatment for 2 h and 24 h) and untreated cells (controls) were harvested and resuspended in 1×PBS. The cell suspension was then analyzed for mitoxantrone median fluorescence intensity using the yellow-green laser at wavelength 561 nm excitation and 695±40 nm bandpass emission in Thermo Fischer-Attune flow cytometer.

From the cell pellet resuspended in TRIzol Reagent (Thermo Fisher), RNA was isolated using the chloroform-isopropanol method (Sambrook and Russell, 2006). 1 µg RNA was used for cDNA preparation using PrimeScript RT Reagent kit (Takara,RR037A) as per the manufacturer's instructions. For qPCR, 1 µl of cDNA (10 ng/µl) was used for each 5 µl (2.5 µl SYBR-green, 0.5 µM primer pairs and 1 µl PCR-grade water) reaction in triplicate. qPCR was carried out in the Roche LifeCycler480 qPCR machine using the following parameters: 3 min initial denaturation at 95°C, followed by amplification for 40 cycles of 95°C for 10 s, 58°C for 25 s (standardized for respective primer sets) and 72°C for 20 s. GAPDH or RPL19 was used as the internal control. Oligonucleotide details are provided in Table S2.

Primers specific to exon-intron junctions were used to amplify the seven random junctions encompassing the whole gene length. Genomic DNA was used as the template for SYBRgreen-based qPCR of the exon-intron junction. Relative quantification using GAPDH as an internal control was carried out. ΔΔCt values were multiplied by 2 to obtain the copy number per diploid cell; values greater than 4 were called gain in copy; values under 1.5 were called loss in copy (Ma and Chung, 2014). Oligonucleotide details are provided in Table S2.

Equal numbers of cells were harvested and washed once with 1× ice-cold PBS. For approximately one-million cells, 100 µl 1X SDS-Laemmli lysis buffer was added and mixed vigorously (vortex for 1 min). Cells were resuspended in SDS-Laemmli lysis buffer and boiled at 100°C for 10 min followed by vortex mix (10-15 s) and short spin. Cell lysates prepared in SDS-Laemmli lysis buffer were loaded, resolved on SDS-PAGE and transferred overnight to the nitrocellulose membrane. These membranes were probed with different primary antibodies: DNA-PKcs, p(S2056)-DNA-PKcs, TOP2B and TOP2A (1:3000); ATM, GCN5, pChk1 (Ser345), Chk1, pChk2 (Thr68), Chk2, XRCC4, pRPA2 (Ser8) and RPA2 (1:2000); RAD50, KU80, Mre11, all histone modifications (H3K27ac, H3K18ac, H3K9ac, H3K14ac, H3K56ac, H4K20me2, H4R3me2, H3K4me2, H3K79me2, H3K36me2, H3K27me2 and H3K9me2) and phospho (ser139)-γH2AX (1:1000); actin, vinculin and tubulin (1:5000); and H3 and H4 (1:4000). For all proteins, primary incubation was carried out at 4°C overnight followed by a 1 h incubation at 18-25°C for secondary HRP-tagged antibody. Both incubations were followed by three washes with Tween-20-based buffer. Blots were developed using ECL reagent (BioRad) in BioRad Chemidoc. For representation, the gamma setting of every blot was ensured to be ≤1 and bands were quantified as mentioned in respective figure legends using ImageJ software. Antibody details are provided in Table S1.

The protocol was followed as published by Salunkhe et al. (2018). Briefly, ∼0.5×10⁶ cells were fixed with chilled 100% methanol for 2 h at −20°C before being spread on poly-L-lysine-coated coverslips and kept in one well of a six-well tissue culture plate. Except for antibody incubation, all steps were carried out in the plates themselves. Cells were air dried briefly (20-30 s) and then permeabilized using 500 µl 0.5% Triton X-100 for 30 min at 40°C. After one wash with 1×PBS, cells were blocked using 1 ml blocking agent (3% BSA with 0.1%Tween-20 in PBS). Upon blocking, coverslips were incubated with 15 µl of their respective primary antibody overnight in a moist chamber at 4°C. After primary antibody incubation, coverslips were washed with 1×PBS three times by gentle shaking for 10 min each to remove non-specifically bound antibodies. After washing, cells were incubated with 15 µl secondary antibody either with Alexa Fluor 488 or AF633, and then washed again. Finally, coverslips were mounted with Vectashield mounting medium containing 0.01 µg/µl DAPI and sealed with nail polish. Imaging was carried out using a Zeiss LSM 780 confocal microscope (magnification 63×), and its processing and intensity quantification was carried out using ImageJ software. Antibody details are provided in Table S1.

FFPE blocks of bone marrow biopsies were obtained after ethics committee approval. Sections (4 µm) were cut using microtome and fixed on poly-l-lysine-coated glass slides. Slides were kept at 60°C for 30 min and rehydrated in xylene thrice for 10 min each, followed by sequential 10 min incubation in 100%, 90% and 70% ethanol for proper deparaffinization. Slides were then kept in running tap water for 10-15 min before antigen retrieval. Beyond this step, at no time were slides allowed to dry. Antigen retrieval was carried out by boiling the slides in Tris-EDTA buffer [10 mM Tris base and 1 mM EDTA (pH=9)] in the pressure cooker for 15 min (approximately two whistles). Glass slides were then allowed to cool down in the pressure cooker itself slowly and were then washed using TBS-Tween buffer and blocked with 1% horse serum (Invitrogen, 31874; 60 mg/ml) in 1×PBS for 30 min at room temperature. Sections were incubated overnight with anti-DNA-PK, anti-TOP2B and anti-GCN5 antibodies at 1:100 dilution for each. After overnight primary antibody incubation at 4°C and three PBS washes for 5 min each, cells were incubated with secondary antibody tagged either with Alexa Fluor 488 or AF633. Cells were washed similarly with PBS again and then mounted with Vectashield mounting medium containing 0.01 µg/µl DAPI. Slides were kept at −20°C until imaging. Imaging was carried out using a Zeiss LSM 780 confocal microscope, and mean fluorescence intensity quantification was carried out using ImageJ software. Antibody details are provided in Table S1.

1×10⁴ cells were treated with doxorubicin (Pfizer Laboratories), mitoxantrone (Sun Pharmaceuticals), NU7026 (Sigma) and butyrolactone 3 (Cayman Chemicals, 12095) at different concentrations (see text). Treated cells were counted after Trypan Blue staining or an MTT [3-(4,5-dimethylthiazol-2-yl)-2,5 diphenyl tetrazolium bromide] assay was performed using the manufacturer's protocol. Briefly, for the MTT assay, an equal numbers of cells were seeded in 96-well plates in triplicate wells. Cells were then treated with respective drugs at different concentrations and times, followed by a 4 h incubation with 5 mg/ml MTT (Himedia, TC191-1G; 0.25 mg/ml final concentration). Formazan crystals were then solubilized using acidified isopropanol-tween20 (10% v/v) solution, and then absorbance was acquired at a 570 nm wavelength using a spectrophotometer (Biotek, Epoch2).

Drug-treated cells were stained with Annexin V tagged with FITC fluorochrome and propidium iodide (PI) according to the protocol mentioned in the Annexin V-FITC Early Apoptosis Detection Kit (Cell Signaling, 6592). Briefly, 0.1 million cells harvested in 1× PBS were stained with 2 µl annexin V FITC stain and 12.5 µl propidium iodide (PI) for 20 min on ice and analyzed by flow cytometry (Thermo Fischer-Attune Nxt) using gates designed to exclude debris, and include FITC and PI positive stain compared with unstained controls.

PRKDC and TOP2B: 0.2×10⁶ cells/ml were seeded per well in a 6-well plate and transfected the next day with siRNA at 80 pmols PRKDC (siRNA select-ThermoFisher, 4390824) and 50 pmols TOP2B (pooled siRNA-Origene, SR304899) per well concentration using the standard 6-well protocol for Lipofectamine3000 reagent (Invitrogen, L3000015). Cells were collected 48 h post-transfection and tested for knockdown of respective proteins by qPCR and immunoblotting. Knockdown cells were seeded in triplicate wells of 96-well plates and treated with mitoxantrone (50 nM and 100 nM) for a drug toxicity assay using Trypan Blue staining for 48 h. For sequential knockdown of PRKDC and TOP2B, post 48 h of PRKDC transfection, cells were harvested in three parts: one for knockdown validation, a second for 48 h drug toxicity assay post-PRKDC knockdown and a third for re-transfection of TOP2B siRNA. 48 h after TOP2B transfection, cells were harvested similarly and processed for a drug toxicity assay and knockdown validation by immunoblotting.

GCN5: 0.2×10⁶ cells/ml were seeded per well in a 6-well plate and transfected the next day with siRNA (Invitrogen, 4390824) at 30 pmols per well using the standard 6-well protocol for Lipofectamine3000 reagent. Cells were harvested 48 h post-transfection and processed for knockdown validation by immunoblotting.

The control siRNA used (siControl) was Trilencer-27 Universal scrambled negative control siRNA duplex (Origene, SR30004).

Chromatin preparation from HL-60 and HL-60/MX2 cells was carried out using the sonication method (23 cycles, 20 s ON and 30 s OFF, High power in Bioruptor Plus-Diagenode) for DNA-PKcs ChIP and MNase enzymatic digestion method (0.06 µl per 10⁶ HL-60 and 0.05 µl per 10⁶ HL-60/MX2 cells) for GCN5 and H3K27Ac using a SimpleChIP Enzymatic Chromatin IP Kit (Cell Signaling, 9003). At least 20 million cells per chromatin preparation were used, and both methods were standardized for each cell line such that chromatin band size ranges between 150 and 900 bp. DNA-PKcs chromatin immunoprecipitation was carried out using a Hi-ChIP- IT Kit (Active Motif, 53040) and chromatin immunoprecipitation for other proteins was carried out using an Enzymatic Chromatin IP Kit (SimpleChIP, 9003). The protocol was followed according to the manufacturer's instructions. Immunoprecipitated chromatin was reverse cross-linked and purified to recover DNA for qPCR quantification using specific primers (see Table S2).

10,000 cells were cultured in triplicate wells of a 96-well plate with mitoxantrone (50 nm) or NU7026 (30 µM), or both, for 48 h then stained with 5 µM 2′,7′-dichlorofluorescein diacetate (DCFDA) at 37°C in the dark. Fluorescence was measured (Ex/Em=485/535 nm) in a microplate reader. Cells treated with 1 mM H₂O₂ were used as a positive control.

Constructs for the TOP2B and PRKDC promoter upstream regions with specific primers containing Kpn1 and Nhe1 restriction sites (see Table S2) were amplified using HL-60/MX2 genomic DNA as a template. This was then ligated (5 h at 16°C) into Kpn1 (NEB, R3142) and Nhe1 (NEB, R3131), and digested (2 h, 37°C) with a pGL3.Basic vector upstream to luc gene. The cloned fragments were then confirmed by insert-specific PCR and Sanger sequencing. To generate TOP2B promoter constructs with 5′ serially deleted sequences, the first construct (F1R1) was used as a template for the next promoter construct (F2R1) using Kpn1 and Nhe1 restriction enzyme sites. The F2R1 reporter construct, in turn, was used for F3R1, which was subsequently used to generate F4R1 reporter constructs of the TOP2B promoter.

Transient transfections of promoter constructs were carried out in 293FT cells. The cells were seeded in a 96-well Blackwell plate for each treatment group in technical triplicate at a density of 5000 cells per well. Along with co-transfection of 160 ng promoter constructs and 40 ng of pRL-TK control vector, pGL3 basic vector (negative control), pRL-TK (positive control) and 200 ng tdRED-empty vector (transfection control) were also transfected using Lipofectamine3000 transfection reagent according to the manufacturer's protocol. Briefly, 0.5 µl L3000 diluted in 5 µl serum-free DMEM media was mixed with plasmid DNA diluted in serum-free media with P3000 reagent. This (1:1) DNA–lipid complex mixture was incubated for 20 min at room temperature and then added to the cells. For the treatment group, 24 h post-transfection cells were treated with 30 μM NU7026. 48 h after transfection, a dual-luciferase assay was carried out as per the manufacturer's instructions (Promega). Briefly, after the transfection of cells, the growth medium was removed, and PBS was added gently to remove detached cells and residual growth medium. After completely removing the rinse solution, cells were lysed using 100 µl passive lysis buffer (Luciferase assay reagent II, LARII) for 15 min followed by luminescence measurement in the luminometer (Biotek, Cytation5) using parameters: a 2 s pre-measurement delay, followed by 10 s measurement. Stop & Glo solution (20 µl; Promega, E1910) was then added to measure Renilla luciferase activity. For the DNA-PKcs promoter assay, GCN5 wild-type overexpression construct (pAd-Track Flag GCN5) and acetyltransferase mutant (pAd-Track Flag GCN5 Y621A/P622A) plasmids were purchased from Addgene (Addgene 14106 and Addgene 14425, respectively). Both plasmids were deposited at Addgene by Pere Puigserver (Harvard Medical School, Harvard University, Cambridge, USA). GCN5 overexpression constructs and shRNA were co-transfected with reporter constructs to evaluate DNA-PKcs promoter activity. The GCN5 shRNA was purchased from Sigma Aldrich (TRCN0000307319). The control shRNA was pLKO.1 Puro shRNA Scramble (Addgene 162011; deposited by David Bryant, Institute of Cancer Sciences, University of Glasgow, Glasgow, UK).

Nuclear lysates were prepared using NE-PER Nuclear and Cytoplasmic Extraction Reagent (ThermoFisher Scientific, 78833) as per the manufacturers’ instructions. Biotin end-labeled probes (detailed in Table S2) were prepared by Sigma Aldrich. EMSA was performed using a LightShift Chemiluminescent EMSA Kit (Pierce) according to the manufacturer's instructions. DNA binding reactions were performed in a 20 µl reaction volume containing 1×binding buffer [10 mM Tris, 50 mM KCl, 1 mM DTT (pH 7.5)], labeled probe (20 fmol), unlabeled probe for competition (4 pmol) and polydI-dC (50 ng/µl). For EMSA using nuclear lysates, extra components (i.e. 2.5% glycerol, 50 mM KCl and 2 mM DTT) were added to the reaction mixture. Nuclear lysate (5 µg), DNA-PKcs antibody (2 µg, rabbit polyclonal DNA-PKcs; ThermoFisher, PA1-84583) and 2 µg IgG isotype antibody (Cell Signaling, 2729) were added to the respective reaction mix and incubated on ice without disturbing for 45 min. Reaction products were then resolved by electrophoresis in 0.5×TBE buffer on 5% native polyacrylamide gel. After this, the protein–DNA complexes were transferred onto a positively charged Biodyne B nylon membrane (ThermoFisher Scientific,77016). UV crosslinked membrane was then blocked and incubated with streptavidin-horseradish peroxidase conjugate for chemiluminescence detection in a BioRad Chemidoc machine. All buffers (blocking, conjugate, wash buffer and substrate equilibration buffer) were provided with their respective kits and used according to the manufacturers’ protocols. For purified DNA-PK enzyme EMSA, a similar protocol was followed to the one mentioned above, except the binding reaction had 0.2 mM ATP, 1 mM EDTA and 0.2 mg/ml molecular grade bovine serum albumin (BSA) as extra components without DNA-PKcs and IgG isotype antibody.

Raw FASTQ files for RNA-Sequencing datasets GSE103424 and GSE83533 were downloaded from NCBI GEO and NCBI dbGaP (https://dbgap.ncbi.nlm.nih.gov/) after due permission from the authors.

SRR files were downloaded from NCBI Sequence Read Archive (SRA) ftp through NCBI dbGaP for the GSE83533 dataset and converted to FASTQ format using fastq-dump command in the SRA-toolkit. These FASTQ files and the raw FASTQ files from GSE103424 dataset were first processed for quality control using the fastQC tool. If necessary, adapter trimming was carried out using Trimmomatic. Another round of QC was carried out after trimming. If the data managed to pass QC, they were aligned with the latest reference human genome downloaded from GENCODE website (GRCh38.p13) and annotated with the respective GTF file (gencode v38) using STAR aligner 2.7.0c. The bam files generated by STAR aligner were quantified, and subsequent raw read-counts files were generated using HT-Seq tool. Finally, the read-counts were used for generating counts per million (CPM) normalize expression values with the EdgeR package in R software. Genes that are involved in DDR, NHEJ or HR pathways were taken from Kaur et al. (2020) and the CPM values were plotted as a heatmap using Graphpad prism 8.

Genomic DNA of HL-60, HL-60/MX2, THP-1 and THP-1/LDRP cells were isolated using the QIAamp DNA mini kit (Qiagen, 51304) exactly according to the instruction in manual.

Genomic DNA isolated from HL-60, HL-60/MX2, THP-1 and THP-1/LDRP cells was amplified using 1 U One Taq Polymerase enzyme (NEB, M0480), 200 µM dNTPs and 5% DMSO. 0.25 µM forward and reverse primers. Thermocycling conditions were used for Touchdown PCR wherein, initial denaturation at 95°C for 5 min was followed by 20 cycles of 95°C for 30 s, 65°C for 30 s (with a 0.8°C decrease in temperature per cycle) and 68°C for 4 min; 20 cycles of 95°C for 30 s, 55°C for 30 s and 68°C for 4 min); and final extension at 68°C for 1 min. PCR-amplified DNA was purified using the NucleoSpin Gel and PCR clean-up kit (MN, 740609) exactly as described in the manual. The purified PCR product was then sent to an in-house Sanger sequencing facility.

cDNA from sensitive and resistant cells was prepared as mentioned above. Genomic DNA was isolated and used as positive control for the exon-intron junction. Both cDNA and gDNA were amplified using sense primer (5′-GCACCCAGCCTTAAACTTAATTCAAT-3′) in exon 33, and the antisense primer (5′-TTGAAGCTGATGATGTTAAGGGCA-3′) in intron 33. Thermocycling conditions were: initial denaturation at 94°C for 45 s, followed by 30 cycles of 94°C for 30 s, 55°C for 30 s and 68°C for 5 min; and final extension for 5 min at 68°C. GAPDH was used as PCR control, and it give two amplicons in case of gDNA and a single amplicon in case of cDNA; thus, it served as control for possible gDNA contamination in cDNA.

The protocol for the TARDIS (trapped in agarose immunostaining) assay was followed as described previously (Cowell and Austin, 2018). Briefly, mitoxantrone (50 nM) treated and untreated HL-60 and HL-60/MX2 cells were harvested, and 50 µl resuspension was mixed with 50 µl 2% molten agarose. 100 µl mix was spread in a thin layer of agarose onto glass microscope slides. As the slides were prepared, each slide was placed on top of the ice-cooled glass plate to set the agarose. The slides were submerged in lysis and extraction buffers [1% SDS (w/v); 20 mM sodium phosphate, 10 mM EDTA (pH 6.5) and 1×protease inhibitor cocktail]. Incubation of this at room temperature for 30 min leaves an adduct of genomic DNA and a covalently linked protein molecule that is too large to escape and thus remains trapped in the agarose. The covalently linked molecules such as TOP2 are then detected by immunofluorescence. In this example, 200 µl of 1:100 diluted TOP2B and TOP2A primary antibody in 1% (v/v) BSA in PBS blocking buffer was used followed by secondary antibody incubation using a FITC-linked secondary antibody. Slides were counterstained with DAPI (100 µg/ml). Immediately, imaging was carried out using a fluorescence Carl Zeiss Axio Imager Z1 upright microscope. FITC signal was determined for each object (nuclei) that was positive for DAPI staining (see antibody details in Table S1).

TOP2 activity of HL-60, HL-60/MX2, THP-1 and THP-1/LDRP cells was determined by a decatenation assay of kinetoplast DNA by their nuclear lysates. The method followed the instructions for the Human Topoisomerase II Assay Kit (TopoGEN, TG-1001-1A). Briefly, nuclear lysates were incubated with 200 ng kinetoplast DNA for 30 min at 37°C in the TOP2 reaction buffer [50 mM Tris-HCl (pH 8.0), 10 mM MgCl₂, 2 mM ATP, 0.5 mM DTT, 30 μg/ml BSA and 2 mM ATP]. The reaction was stopped by the addition of stop buffer (5% Sarkosyl, 0.125% Bromophenol Blue and 25% glycerol). Samples were cleaned up by proteinase K (200 μg/ml) digestion for 15 min at 37°C followed by purification by an equal volume of chloroform: isopropanol (24:1) and run on a 1% agarose gel containing ethidium bromide (Thermo Fisher, 15585011). The image was detected under the ChemiDoc System (Bio-Rad).

Intron retention-associated variants were identified from HL60 and MX2 RNA-seq data following the GitHub tutorial (Shiraishi et al., 2022). Briefly, the iravnet-0.1.0b15 package was downloaded from GitHub, along with all its dependencies and required singularity images from GitHub. After this, we ran the singularity images, which internally used Star aligner to align the reads and Samtools for sorting according to the coordinates; then the iravnet tool was used to identify and annotate the intron retention-associated variants.

Details of all reagents and resources are provided as supplementary tables: antibodies details are in Table S1; oligonucleotides details in Table S2; and other reagents and resources, such as chemicals, cell lines, plasmids and software used in the study, are listed in Table S3.

All statistical tests were performed using GraphPad Prism version 8 using the tests described for each experiment. Data from three independent biological replicates were considered unless stated otherwise for the statistical test. Significant results (*P<0.05; **P<0.01; ***P<0.001; ****P<0.0001) are considered if P<0.05. Detailed information about statistical tests is provided in figure legends.

We thank Dr Ashwin Butle, a post-doctoral fellow in Dr Amit Dutt's lab (TMC-ACTREC), for assisting with subcutaneous injections. We acknowledge expert service and support by the ACTREC common instrument facility, flow cytometry facilty, digital imaging facility and animal house facility.