PARP inhibition leads to synthetic lethality with key splicing-factor mutations in myelodysplastic syndromes

Abstract

Background

Splicing factors are frequently mutated in patients with myelodysplastic syndromes and acute myeloid leukaemia. Recent studies have revealed convergent molecular defects caused by splicing factor mutations, among which R-loop dysregulation and resultant genome instability are suggested as contributing factors to disease progression. On the other hand, understanding how mutant cells survive upon aberrant R-loop formation and genome instability is essential for developing novel therapeutics.

Methods

The immunoprecipitation was performed to identify R-loops in association with PARP1/poly-ADP-ribosylation. The western blot, immunofluorescence, and flow cytometry assays were used to test the cell viability, cell cycle arrest, apoptosis, and ATM activation in mutant cells following the treatment of the PARP inhibitor. The Srsf2(P95H) knock-in murine hematopoietic cells and MLL-AF9 transformed leukaemia model were generated to investigate the potential of the PARP inhibitor as a therapy for haematological malignancies.

Results

The disease-causing mutations in SRSF2 activate PARP and elevate the overall poly-ADP-ribosylation levels of proteins in response to R-loop dysregulation. In accordance, mutant cells are more vulnerable to the PARP inhibitors in comparison to the wild-type counterpart. Notably, the synthetic lethality was further validated in the Srsf2(P95H) knock-in murine hematopoietic cell and MLL-AF9 leukaemia model.

Conclusions

Our findings suggest that mutant cells antagonise the genome threat caused by R-loop disruption by PARP activation, thus making PARP targeting a promising therapeutic strategy for myeloid cancers with mutations in SRSF2.

Introduction

Myelodysplastic syndromes (MDS) are a group of myeloid neoplasms, which are generally characterised by inefficient hematopoiesis and a high risk of progression to acute myeloid leukaemia (AML) [1]. Recent efforts in genomic sequencing of MDS specimens have led to the discovery that more than half of patients contain a mutation in at least one RNA splicing factor (SF), most frequently SF3B1, SRSF2 and U2AF1 [2–8]. Further genetic evidence showed all major mutations caused the alteration of numerous splicing events and defects in self-renewal/differentiation of hematopoietic stem cells, suggesting these mutations are causal to disease progression and thus disclosing a new research direction in RNA processing for mechanistic understanding of MDS pathogenesis [9–17].

Meanwhile, researchers were prompted to evaluate the vulnerability of mutant cells to splicing modulators as a new therapeutic strategy and found small molecules targeting the SF3B1 complex, such as Pladienolide B, E7107 and H3B-8800, were selectively cytotoxic in both mutant cell and animal models [2, 18–20]. However, concerns were also raised regarding the potential side effects [19, 21], since the normal functionality of the splicing machinery is essential for maintaining cell fitness and survival [22]. Indeed, several following clinical trials claimed the failure of splicing modulator treatment due to lack of remission or merely partial responses (NCT02841540) [19, 21], and to vision loss in two cases that resulted in study discontinuation (NCT00499499) [23]. Therefore, it remains to be determined how spliceosome mutations and mis-spliced transcripts contribute to the development of MDS and other myeloid cancers, and whether more specific splicing events instead of the general splicing machinery need to be targeted.

Accordingly, although most SF mutations in MDS are involved in 3’-splice site recognition, the altered splicing events are largely non-overlapping between mutations, raising a possibility of non-splicing mechanisms that exist to contribute to disease progression [2, 3, 15, 24, 25]. Notably, many SFs, including SRSF2 and SF3B1, play essential roles in transcriptional elongation by directly regulating the transition of RNA polymerase II (RNAPII) from pausing to release for elongation [22, 26, 27]. Once such transition is interfered with, the paused RNAPII may not only lead to disruption of the gene expression but also induce aberrant formation of the R-loop, a three-stranded nucleic acid structure that consists of a DNA: RNA hybrid and a displaced strand of DNA. While involved in many physiological processes in cells, unscheduled R-loops cause transcription-replication collision and various types of DNA damage [28–31], exerting selective pressure on preleukemic/Clonal hematopoiesis of indeterminate potential (CHIP) clones for cancer progression [22, 31, 32]. Indeed, recent efforts provoked a series of discoveries that all causal mutations elicited non-negligible insults to the genome through aberrant R-loops [22, 32–34].

On the other hand, survived mutant cells exhibited elevated activity of ataxia telangiectasia mutated and Rad3 related (ATR)/checkpoint kinase 1 (CHK1) pathway in response to R-loop induced genome instability, likely due to the replication stress that is not severe enough to cause double-strand breaks (DSB) and activation of the ataxia telangiectasia mutated (ATM)/checkpoint kinase 2 (CHK2) pathway [22, 34, 35]. Accordingly, a CRISPR screening has demonstrated that factors in the cell cycle and DNA damage response pathways are among the top candidates causing synthetic lethality with Srsf2(P95H) mutation [36]. Later studies confirmed that pharmacological inhibition of ATR induces selective toxicity to mutant cells [33, 34]. Taken together, these findings have not only provided paradigm-shifting insights for understanding disease progression but also defined an alternative direction to explore new molecular targets involved in DNA damage responses for eliminating MDS/AML [33]. Indeed, the ATR inhibitor-AZD6738 is currently in a Phase I trial (NCT03770429) for the treatment of MDS or CMML patients containing SF mutations [37]. While promising, like many other small molecule therapies, the potential side effects of ATR/CHK1 inhibitors need to be considered. Moreover, the resistance to ATR/CHK1 inhibitors was also reported in various types of cancer cells [38–42]. Since cells harbour divergent mechanisms to maintain genome stability [43–45], an expanded investigation of DNA damage responses to SF mutations is a pressing need.

In this study, we show that mutations in SRSF2 elevate the level of poly-ADP-ribosylation (PARylation). Treatment of mutant cells with the PARP1 inhibitor (PARPi) Talazoparib results in remarkable viability inhibition, cell cycle arrest, and ATM activation. We further confirm these observations in the Srsf2(P95H) knock-in murine hematopoietic cells and MLL-AF9 transformed leukaemia model. Mechanistically, we show that enhanced PARP activity in mutant malignant cells is triggered by R-loop dysregulation. In sum, our findings indicate PARP is a potential therapeutic target for MDS/AML with SF mutations.

Results

SRSF2 mutant cells exhibited elevated PARylation level

We have previously documented that ATR/CHK1, but not ATM/CHK2 pathway, was activated by major SF mutations and suggest R-loop induced replication stress plays an indispensable role in mediating DNA damage responses and cellular defects [22]. Of note, recent studies have revealed the interaction of R-loops with many DNA damage repair-related proteins, one of which is the PARP family [46–50]. Due to the essential function of the PARP family proteins in maintaining cancer cell survival, we hypothesise that the PARP is involved in the adaptive responses to SF mutation-induced genome instability. To test this possibility, we examined the overall PARylation signal in MDS-L and 293T cells carrying the doxycycline (DOX)-inducible construct that expresses the wild-type (WT) or mutant (P95H and Δ8AA, deletion of eight amino acids from Pro 95 to Ala 102) open reading frame of SRSF2, as well as shRNA targeting the 3’ UTR of endogenous SRSF2 mRNA [22, 51]. In line with the previous study, around one- to two fold of exogenous SRSF2 mRNA were induced, while about 50% of endogenous mRNA were repressed following DOX treatment in both cell lines (Supplementary Fig. S1A, S1B), well mimicking the expression condition of hematopoietic cells with heterogenous SF mutations [22].

The dot blot analysis showed the PARylation level remained unchanged when the expression of exogenous WT SRSF2 was induced (Supplementary Fig. S1C, S1D), suggesting mild induction of SRSF2 has little effect on PARP activity, in line with the lack of phenotypic alterations reported previously [22]. In contrast, the PAR signal was markedly upregulated in both P95H and Δ8AA cells following DOX treatment, which was reversed by the treatment of the PARP1 inhibitor Talazoparib (Supplementary Fig. S1C, S1D). The results were reproducible in 293T cell lines as well (Supplementary Fig. S1E, S1F). We further confirmed the increase of PARylation modification in proteins by Western blot (Fig. 1a, b and Supplementary Fig. S1G–S1K), and the nuclear localisation by immunostaining (Fig. 1c, d and Supplementary Fig. S1L–S1S). Altogether, these results suggest that the expression of mutant SRSF2 elevates PARP activity in the nucleus.

Fig. 1. Effects of SRSF2 mutation on PARylation levels.

a Levels of PARylation in SRSF2(WT and P95H) MDS-L cell. MDS-L cells were treated with 25 nM Talazoparib and 1 μg/mL DOX for 2 Days. Cells were treated with 10 μM PARG inhibitor for 10 min before harvesting. b Quantitative analyses of PARylation in SRSF2 (WT and P95H) MDS-L cell of (a). Data are presented as mean (±SD) of three technical replicates. The relative abundance of PARylation was normalised with ACTIN signals. Statistical comparisons were performed using unpaired two-tailed Student’s t tests. *P < 0.05, **P < 0.01, ***P < 0.001. c PARylation signals were detected by immunocytochemistry in SRSF2(WT and P95H) MDS-L cells by DOX for 2 days. d Quantitative analyses of PARylation signals for MDS-L cells with WT or mutant SRSF2(P95H) of (c). More than 100 cells per treatment condition per cell type were analysed; Purple, PARylation; Green, EdU; blue, DAPI; Scale bar, 10 μm. Statistical comparisons were performed using unpaired two-tailed Student’s t tests. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

The PARP is activated by R-loops in mutant cells

To dissect the mechanism of PARylation induction by SF mutations, we first performed the cell fractionation to examine the association of PARP proteins to chromatin in MDS-L cells expressing WT or mutant SRSF2. The results showed that the abundance of PARP1 in the chromatin fraction, the representative member of PARP family proteins, was significantly higher in the mutant than in WT cells (Fig. 2a, b and Supplementary Fig. S2A). Meanwhile, the immunoprecipitation experiment with the PARP1 antibody showed that the presence of mutant SRSF2 markedly induces the association of PARP1 with RNase H1 and RNase H2, two endonucleases for specifically removing R-loops in mammalian cells, and an RNA binding protein HNRNPLL (Fig. 2c). Such associations were unaffected by DNase I treatment, indicating they are direct protein–protein interactions (Fig. 2c and Supplementary Fig. S2B). Altogether, our observations indicate that mutant SRSF2 induces the physical association of PARP1 with R-loop sites on the chromatin.

Fig. 2. PARP1 was closely linked to R-loop by PARylation in SRSF2 mutations.

a The chromatin fractionation assay for PARP1 in SRSF2(WT) and SRSF2(P95H) MDS-L cell, with DOX-induced for 2 days. b Quantitative analyses of chromatin fractionation assay for PARP1 in SRSF2(WT) and SRSF2(P95H) mutation MDS-L cell of (a). The relative abundance of PARP1 was normalised with H3 signals. Data are presented as mean (±SD) of three technical replicates. Statistical comparisons were performed using unpaired two-tailed Student’s t tests. *P < 0.05, **P < 0.01, ***P < 0.001. c Immunoprecipitation of PARP1 in SRSF2(WT) and SRSF2(P95H) mutation 293T cell, with 1 μg/mL DOX treatment for 3 days. Appropriate Dnase I (10U) was used to degrade the DNA of the samples at 37 °C for 2 h. d Suppression of PARylation by overexpressed RNASEH1 in MDS-L cells upon induced expression of SRSF2(P95H) by DOX for 2 days. e Quantitative analyses in suppression of PARylation by overexpressed RNASEH1 in MDS-L cells upon induced expression of SRSF2(P95H) by DOX for 2 days of (d). The relative abundance of Parylation was normalised with ACTIN signals. Data are presented as mean (±SD) of three independent experiments. Statistical comparisons were performed using unpaired two-tailed Student’s t tests. *P < 0.05, **P < 0.01, ***P < 0.001. f The pulled-down DNA-RNA hybrids were detected by the S9.6 antibody. The DNA-RNA hybrids were cross-linked with proteins and pulled down by the anti-PAR antibody. RNase H was treated with 5 U for 2 h.

We were next prompted to investigate whether R-loops play a causal role in PARP activity by overexpressing RNase H1 to resolve R-loops in the cell (Supplementary Fig. S2C, S2D). The results showed that the induced PARylation level in mutant cells was remarkably reversed by overexpressed RNase H1 (Fig. 2d, e and Supplementary Fig. S2E). To further verify if PARP can directly recognise the R-loop structure and add PAR modifications to associated proteins, we performed PAR-immunoprecipitation followed by the dot blot assay with the S9.6 antibody, which indeed showed stronger R-loop signals in mutant cells (Fig. 2f and Supplementary Fig. S2F).

The PARP inhibitor suppresses the viability of SRSF2 mutant cells

Given the importance of PARylation in DNA damage repair, it has been shown that targeting the PARP family may offer a selective tumour-killing effect on multiple types of cancers [44, 52]. To evaluate the sensitivity of SRSF2 mutant cells to the PARP inhibitor, we first tested different concentrations of Talazoparib in MDS-L cells with or without the expression of WT, or mutant (P95H) form of SRSF2, revealing much higher sensitivity of the mutant cells to Talazoparib (Supplementary Fig. S3A). Similarly, the selective growth inhibition of mutant (SRSF2 P95H and Δ8AA) MDS-L and 293T cells to Talazoparib was shown (Fig. 3a, b and Supplementary Fig. S3B, S3C).

Fig. 3. The viability and growth of SRSF2 mutant cells by PARPi.

a, b The proliferation of MDS-L cells (n = 3 biological replicates) with or without expressing mutant SRSF2(P95H) (a) and (Δ8aa) (b) upon 25 nM Talazoparib and 1 μg/mL DOX treatment for different days. c, d Cell cycle analysis of MDS-L cells with or without expressing SRSF2(P95H) at 2 days. e, f Apoptosis of MDS-L cells with or without expressing SRSF2(P95H) with 25 nM Talazoparib and DOX treatment for 2 days were analysed, and the percentage of cells in ANNEXIN V + PI− and ANNEXIN V + PI+ were shown. Statistical comparisons were performed using unpaired two-tailed Student’s t tests. *P < 0.05, **P < 0.01, ***P < 0.001.

The lower growth rate of mutant cells was attributed to cell cycle arrest while only marginal apoptosis was seen [22]. In accordance, the SRSF2(P95H and Δ8aa) mutant cells had a significantly higher G2/M ratio, which was further elevated by Talazoparib treatment (Fig. 3c, d and Supplementary Fig. S3D–S3K). In comparison, mutant cells with or without Talazoparib treatment did not show a clear trend of alteration in the S-phase (Fig. 3c, d and Supplementary Fig. S3D–S3K). These results suggest that while the R-loop may induce genome instability by promoting the conflict between transcription and replication during S-phase, which is exacerbated by PARP inhibition, insulted cells are prone to be trapped later in the G2/M phase. Interestingly, Talazoparib treatment resulted in a significantly higher ratio of apoptosis for the SRSF2(P95H) MDS-L and U937 cells, an AML cell line (Fig. 3e, f and Supplementary Fig. S3L, S3M), suggesting the requirement of PARP activity for mutant cell survival, as well as a therapeutic potential of PARPi in MDS and AML with SF mutations.

PARPi enhances both ATR and ATM activity in SRSF2 mutant cells

We further dissected the molecular responses of mutant cells to PARPi and observed that Talazoparib treatment marginally induced the level of γH2AX, a DNA damage marker, in DOX untreated mutant cells (P95H and Δ8aa) (Fig. 4a–d), likely reflecting a mild non-specific effect of the drug [53]. In the presence of DOX, the level of γH2AX was increased in P95H and Δ8AA cells, and further elevated by Talazoparib treatment, corroborating with the exacerbated cell growth defect and apoptosis phenotype (Fig. 4a–d and Supplementary Fig. S4A–S4F).

Fig. 4. The activity of ATR and ATM of SRSF2 mutant cells with PARPi.

a, b Western blot analysis of γ-H2AX(Ser139) and phospho-Chk1(S345) levels in MDS-L cells with or without expressing of SRSF2(P95H) (a) or SRSF2(Δ8aa) (b) at 25 nM Talazoparib and DOX treatment for 0, 1 and 2 days. c γ-H2AX(Ser139) signals detected by immunocytochemistry in MDS-L cells upon induced expression of WT or mutant SRSF2(P95H) (left) and (Δ8aa) (right) at 25 nM Talazoparib and 1 μg/mL DOX treatment for 2 days. Red, γ-H2AX(Ser139); Green, EdU; blue, DAPI; Scale bar, 10 µm. d Quantitative analyses of γ-H2AX(Ser139) signals for MDS-L cells with WT or mutant SRSF2(P95H) (up) and (Δ8aa) (down). More than 100 cells per treatment condition per cell type were analysed. Statistical comparisons were performed using unpaired two-tailed Student’s t tests. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. e, f Western blot analysis of phospho-Chk2(Thr68) and cleaved caspase 3 in MDS-L cells with or without expressing SRSF2(P95H) (e), SRSF2(Δ8aa) (f) mutation. g Levels of PARylation in MDS-L cell with or without expressing of SRSF2(P95H) under treatment with 100 nM, 300 nM ATR inhibitor-AZD6738 and 1 μg/mL DOX for 2 days. Cells were treated with 10 μM PARG inhibitor for 10 min before harvesting. The relative abundance of Parylation was normalised with ACTIN signals.

ATR and ATM mediate two major signalling pathways in response to DNA damage. ATR/CHK1, but not ATM/CHK2, is activated by SF mutation-induced R-loops [22]. Consistent with the result of γH2AX induction, Talazoparib treatment significantly promoted the phosphorylation of CHK1 at S345 in mutant cells (Fig. 4a, b and Supplementary Fig. S4A, S4B), suggesting more severe DNA damage. More interestingly, while CHK2 remained inactive in mutant cells, the phosphorylation of CHK2 at T68 was elicited by Talazoparib treatment in SRSF2(P95H and Δ8AA) MDS-L cells, and SRSF2(P95H) and U2AF1 (S34F) U937 cells (Fig. 4e, f and Supplementary Fig. S4G–S4I). Corroboratively, the cleaved portion of Caspase 3, a marker of apoptosis, was significantly increased in mutant cells with Talazoparib treatment (Fig. 4e, f and Supplementary Fig. S4G–S4I). To confirm that these results were attributed to a specific loss of PARP activity, we transfected the P95H mutant cells with siRNAs targeting PARP1 (Supplementary Fig. S4J). Loss of PARP1 expression resulted in a decreased PAR signal and elevated DNA damage similar to Talazoparib treatment (Supplementary Fig. S4J, S4K). Altogether, our findings imply that PARP inhibitor blocks the cell’s ability to resolve mutation-induced R-loop and replication stress, which eventually results in the accumulation of DSBs and cell death.

We then examined the relationship between PARP and ATR, the major pathway responding to SF mutations [33, 54]. Interestingly, treatment with the ATR inhibitor (ATRi, AZD6738) resulted in a higher PARylation level in mutant cells, and so did the inhibitor of WEE1 (AZD1775) (Fig. 4g and Supplementary Fig. S4L). Collectively, these results suggest that both ATR/CHK1 and PARP are activated by dysregulated R-loops in mutant cells, and the functional loss of ATR may further promote the PARylation activity at the R-loops sites.

Selective inhibition of leukaemia cell growth in vivo by PARPi

Given the selective role of PARPi in promoting DNA damage and viability inhibition in cell lines expressing mutant SRSF2, we were next prompted to examine if these observations can be verified in vivo and if blocking PARP activity may exert preferential therapy potential on leukaemia cells with SRSF2(P95H) mutations. LSK bone marrow cells (BM) from Srsf2^P95H/+ mice were sorted, followed by transduction with the retrovirus expressing the MLL-AF9 (MA9) oncogene and transplantation into mice for leukaemia development [18]. Then, fully transformed MA9 leukaemia cells (WT or Srsf2^P95H/+) were collected from sick mice. Specifically, mutant cells were more sensitive to Talazoparib treatment (IC₅₀: 155.9 nM vs 23.11 nM) (Supplementary Fig. S5A), and barely formed colonies compared to WT MA9 cells (Fig. 5b, c). Similar results were obtained in Srsf2^P95H/+ and Srsf2^WT ckit-positive cells (Supplementary Fig. S5B, S5C), suggesting both mutant primary hematopoietic and tumour cells depend on PARP activity for survival. Moreover, MA9 Srsf2^P95H/+ cells displayed a higher overall R-loop level, which was further exacerbated by Talazoparib treatment (Fig. 5d, e). We also confirmed a more potent DNA damage response as evidenced by elevated levels of γH2AX, ATR/CHK1, ATM/CHK2, and an enhanced apoptosis level in mutant cells relative to WT cells upon PARPi treatment (Fig. 5d–f and Supplementary Fig. S5D, S5E).

Fig. 5. The effects of pharmacological PARPi on AML in vivo.

a Schematic of secondary transplantation experiments of mouse MLL-AF9 Srsf2^P95H/+ primary leukaemias background to test the effects of PARPi treatment in vitro and in vivo. Following transplantation mouse model onset on the 14th day, mice were randomly divided into two groups, with 6 6–8 weeks-old female mice in each group and followed by treatment with vehicle (1% DMSO + PBS) and Talazoparib (0.33 mg/kg; i.g.; once daily; for 28 days). b, c Colony formation (b) and quantification of colony numbers (c) (n = 3 biological replicates) with Srsf2^P95H/+ MA9 cells at 25 nM Talazoparib for 10 days. Statistical comparisons were performed using unpaired two-tailed Student’s t tests. *P < 0.05, **P < 0.01, ***P < 0.001. d R-loop signals were detected by immunostaining with S9.6 in mice MA9^WT and Srsf2^P95H/+ MA9 cells at 150 nM Talazoparib for 24 h. Red, S9.6; blue, DAPI; Green, Nucleolin. Scale bar, 10 µm. e Quantitative analyses of S9.6 in mice MA9^WT and Srsf2^P95H/+ MA9 cells of (d). More than 100 cells per treatment condition per cell type were analysed. Statistical comparisons were performed using unpaired two-tailed Student’s t tests. *P < 0.05, **P < 0.01, ***P < 0.001. f Western blot analysis of phospho-Chk1(S345), phospho-Chk2(Thr68) and γ-H2AX(Ser139) in mice MA9^WT and Srsf2^P95H/+ MA9 cells at 150 nM Talazoparib for 24 h. g The percentage of peripheral blood leukaemia GFP+ cells after Talazoparib treatment. h The percentage of bone marrow leukaemia GFP+ cells after Talazoparib treatment. i The splenic change after Talazoparib treatment. j The percentage of leukaemia GFP+ cells after Talazoparib treatment. Statistical comparisons were performed using unpaired two-tailed Student’s t tests. *P < 0.05, **P < 0.01, ***P < 0.001.

We next evaluated the anti-AML effect of PARPi in the mouse model. MA9 leukaemic cells expressing the Srsf2^P95H/+ mutant genes were transplanted into a cohort of recipient mice to generate leukaemia. Following leukaemia development, mice were randomised to two groups, receiving either vehicle or Talazoparib (Tala, 0.33 mg/kg/d orally for 28 d) treatment, the dose of which was non-toxic for normal mice as verified by the body weight comparable to untreated mice (Fig. 5a and Supplementary Fig. S5F). Remarkably, Talazoparib treatment resulted in reduced disease burden by nearly completely eradicating leukaemia cells in the peripheral blood and bone marrow (Fig. 5g, h). Moreover, splenomegaly, as well as leukaemia cells engrafted in the spleen, were also decreased by PARPi treatment in the Srsf2^P95H/+ MA9 mice transplants (Fig. 5i, j).

Finally, we were intrigued to evaluate the combination effect of PARPi and ATRi on preventing leukaemia progression since PARP and ATR appear to sense and recover the R-loop-induced genome instability cooperatively (Fig. 4g and Supplementary Fig. S4L). As expected, the Srsf2^P95H/+ MA9 cells were more sensitive to ATRi than the WT counterpart (Supplementary Fig. S5G). In the following colony formation assay, we reduced the concentration of PARPi (from 25 to 10 nM) and ATRi (from IC₅₀: 102 to 50 nM) to minimise the side effects of the two drugs. Interestingly, the results showed that while the single molecule of either PARPi or ATRi significantly reduced the number of colonies with the Srsf2^P95H/+ MA9 cells, the combination treatment markedly synergised the inhibition effect, while only a marginal effect was observed for the WT cells (Supplementary Fig. S5H, S5I).

Discussion

Due to the prevalence of SF mutations in myeloid neoplasms and their adverse impacts on disease progression reported recently, extensive attention has been given to this new direction of leukaemia research. Interests are focused on elucidating the disease mechanisms from the perspective of splicing disruption [55, 56]. Indeed, most mutations elicit alterations of massive splicing events, but whether those splicing defects are required to cause disease phenotypes and worth targeting remains to be determined so far [21, 23]. On the contrary, we and others have discovered a striking phenomenon of genome instability and ATR/Chk1 activation in mutant cells induced by aberrant R-loop formation [22, 33, 34, 57], therefore suggesting an alternative disease mechanism. Here, we continued to expand our understanding of the DNA damage repair network by showing mutations, including SRSF2(P95H) and U2AF1(S34F), activate the PARP pathway in an R-loop-dependent manner. Interestingly, a recent report has shown that the SF3B1 mutant was sensitive to PARP inhibition through similar mechanisms [58]. Taken together, these findings warrant a further investigation of PARPi in treating MDS/AML with SF mutations.

In mammalian cells, the PARP superfamily plays essential roles in DNA damage responses, transcription, chromatin structure stabilisation, and cell metabolism [59]. As a representative member, PARP1 is widely involved in repairing transcription-coupled single- and double-strand breaks (SSBs and DSBs) [60–63]. It rapidly recognises and binds to DNA-damaged sites, PARylating itself and other proteins to recruit additional DNA repair factors [62, 64]. Our finding that R-loops mediate the recruitment of PARP1 to chromatin brings an intriguing question as to where on this three-stranded structure the PARP1 binds. Previous proteomics studies identified PARP1 in the R-loop-associated protein–protein interaction network [46, 47]. In parallel, proximity proteomics of PARP1 identified many proteins in RNA biogenesis, including R-loop regulators DHX9 and RNase H2 [65]. More recently, PARP1 was shown to directly bind R-loop structure in vitro, most likely at the displaced DNA strand [50], where PARP1 mediates PARylation and recruitment of R-loop resolving factors, such as TonEBP [66], DHX9, and DDX18 [47, 49, 67]. Indeed, we observed both RNase H1 and RNase H2 were associated with PARP1 in mutant cells. It therefore indicates that aberrantly formed R-loops in mutant cells may carry certain docking sites, probably SSBs in the displaced DNA strand, to recruit the PARP proteins [28]. In the future, several interesting questions remain to be addressed, such as what the key targets of the PARP proteins are at R-loop sites and how the PARP pathway mediates the resolution of R-loops.

It also inspires us to evaluate the selective sensitivity of MDS/AML with SF mutation to PARPi. One advantage to pursuing this possibility is that divergent PARPi have been widely tested in clinical trials to treat a variety of tumours, including MDS/AML, used as a single agent or in combination with chemotherapy, antiangiogenic agents, or ionising radiation, therefore providing tremendous experience and lessons to support our study [68–71]. It is known that PARPi acts on the catalytic activity of PARP enzymes and the accumulation of SSBs is the primary genotoxic lesion promoting PARPi sensitivity [72]. If unrepaired, SSBs would be converted to more toxic DSBs and cell death [71], which is indeed consistent with our observations of induced CHK2 activation and apoptosis in PARPi-treated mutant cells. Collectively, the selective sensitivity of MDS/AML cells with SFs mutations to PARPi paves a promising way to further characterise the clonal and subclonal genetic background of leukaemia patients for seeking preferred benefits following PARPi-based treatment.

Like any other chemotherapy drug, the potential side effects of PARPi, especially drug resistance and toxicity, need to be taken into consideration. It is known that the mechanisms of PARPi resistance include restoration of homologous recombination, stabilisation of the replication forks, and suppression of single-stranded DNA gap accumulation [72], all of which are at least partially attributed to ATR/CHK1 activation [73]. In line with these findings, simultaneously inhibition of the PARP and ATR/CHK1 pathway significantly enhances the efficacy of PARPi and circumvents the issue of drug resistance [74, 75]. In this study, higher activity of ATR/CHK1 in mutant MDS/AML cells treated with PARPi, and ATRi further induced PARP enzyme activity, indicating two mechanisms are co-activated by the same source of DNA damage. Altogether, these results have integrated individually activated DNA repair pathways into a more interrelated molecular network, encouraging further decoding of the overall pictures of cellular responses to SF mutations (Fig. 6).

Fig. 6. A model of PARPi for the treatment of MDS and AML with SF mutations.

The SF mutations in MDS and AML cells induce R-loop accumulation and cause replication stress. (I) The activation of the ATR-CHK1 pathway may recruit p-CHK1, and PARP1 (with PARylation) is also recruited to chromatin near to R-loop formation site which can contribute to repair of the R-loop induced DNA damage, but cause cell cycle arrest. (II) With PARPi inhibition of PARylation which impaired DNA damage response and the DNA damage repair pathway, stronger p-CHK1 and p-CHK2 signals were detected, and the elevated damage caused leukaemia cell death.

Regarding the potential toxicity, recent reports showed that PARPi may increase the risk of MDS and AML [76]. However, as most patients treated with PARPi have also taken chemotherapy drugs known to be associated with MDS/AML, it is difficult to attribute this entirely to the inhibition of PARP [77]. Several strategies that reduce toxicity and limit or suppress resistance, including continuous development of more efficacious PARPi [78–80], and the design of drug combinations with optimised timing and dosing regimens, could tackle the toxicity issue [76, 81]. Several clinical trials with similar strategies are now conducted (NCT00588991, NCT04207190, NCT02878785, NCT03974217, NCT03289910, and NCT03953898) [82, 83]. In sum, further understanding of the pathological basis of MDS and the potential interactions between splicing factor mutations and PARP is critical for designing better and more effective MDS/AML therapeutic strategies.

Methods

Animals

Rag2^-/- (B6(Cg) -Rag2^tm1.1Cgn/J, JAX#: 008449), NSGS (NOD.Cg-Prkdc^scid Il2rg^tm1WjlTg (CMV-IL-3, CSF2, KITLG) 1Eav/MloySzJ, JAX#: 013062), Srsf2^P95H/+ (B6J.B6NTac (SJL) -Srsf2^tm1.1Oaw/J, JAX#: 028376) mice were purchased from the Jackson Laboratory. Mice purchased were acclimated to housing conditions for at least one week at the COH Animal Resource Center before experiments. Colonies for each mouse strain were maintained in the same animal facility. Mouse care and experimental procedures complied with established institutional guidelines and protocols approved by the Institutional Animal Care and Use Committee at COH National Medical Center.

Retrovirus production and transduction

Virus production was described previously [84]. Briefly, HEK293T cells were transfected with pMD2.G and psPAX2 packaging vectors plus retrovectors designed to knock in genes using the calcium-phosphate co-precipitation method. Supernatants containing virus particles were filtered and concentrated. After titration, murine LSK cells were exposed to the virus (MOI = 10) in the presence of 1 µg/µl retronectin via spinoculation.

Retroviral transduction and transplantation of primary hematopoietic cells

Freshly dissected femora and tibiae were isolated from Vav-Cre⁺Srsf2^P95H/+ CD45.2⁺ female mice. BM was flushed with a 3-cc insulin syringe into cold PBS (without Ca²⁺ and Mg²⁺) with 2% BSA. BM cells were pelleted by centrifugation at 1500 rpm for 5 min and red blood cells (RBCs) were lysed in ammonium chloride-potassium bicarbonate lysis (ACK) buffer for 5 min on ice. After centrifugation, cells were resuspended in PBS/2% BSA, passed through a 40-μm cell strainer, and counted. Mouse Bone Marrow lineage-negative cells were selected by microbeads (Miltenyi® Direct Lineage Cell Depletion Kit). Lineage-negative (Lin-) murine cells were sorted by flow cytometry with staining antibodies (lineage: CD3, CD4, CD8, CD11b, CD11c, CD19, CD41, Ter119, B220, IgM, NK1.1, Gr-1 and IL7Ra; Sca1, c-Kit). LSK murine cells were transduced with viral supernatants containing pMIG-FLAG-MLL-AF9-GFP in IMDM/20% foetal bovine serum (FBS) supplemented with mSCF (25 ng/mL), mIL-3 (10 ng/mL) and mIL-6 (10 ng/mL), followed by injection of ~50,000 cells per recipient mouse via tail vein injection into sublethally irradiated (200 cGy) NSGS mice.

Secondary transplantation experimentation and treatment

Sublethally irradiated mice were transplanted with 1 million of MLL-AF9 Srsf2^P95H/+ primary leukaemias, after onset on day 14 followed by treatment with vehicle (1% DMSO + PBS) and Talazoparib (0.33 mg/kg; i.g.; once daily; for 28 days).

Analysis of cell viability and colony formation

Cell viability was assessed using the Cell Titer-Glo Luminescent Cell Viability Assay Kit (Promega) and Cell Counting Kit-8 (CCK-8) (Yeason) according to the manufacturer’s instructions. Colony formation capacity based on colony number was determined in methylcellulose progenitor assays. Cells were suspended in IMDM medium with 20% FBS and 1% penicillin–streptomycin supplemented with murine growth factors (mIL-3 10 ng/mL, mIL-6 10 ng/ml, mSCF 100 ng/ml). In total, 4000 cells were plated in triplicate into each 24-well plate; colonies were counted after 7–10 days and examined under Widefield Zeiss Observer 7 [85, 86].

Cell lines and cell culture conditions

HEK293T and cells were obtained from the China Center for Type Culture Collection (CCTCC; Wuhan, China). The human MDS cell line MDS-L was provided by Dr. Kaoru Tohyama [87]. HEK293T cells were grown in DMEM supplemented with 10% FBS and 100 U/ml penicillin–streptomycin. K562 cells and U937 cells were grown in RPMI1640 supplemented with 10% FBS and 100 U/ml penicillin–streptomycin. MDS-L cells were grown in RPMI1640 with 10% FBS, 10 ng/ml recombinant human IL-3 (MCE), and 100 U/ml penicillin–streptomycin. Murine MA9-transformed cells were cultured in IMDM medium with 20% FBS and 100 U/ml penicillin–streptomycin supplemented with murine growth factors (mIL-3, 10 ng/mL) (Peprotech). All mediums, penicillin–streptomycin, and FBS were from GIBCO. All cells were cultured at 37 °C in a 5% of the CO₂ incubator.

Inhibitors and antibodies

Cells were treated with inhibitors including Talazoparib (HY-16106), Diacerein (HY-N0283), Ceralasertib (AZD6738) (HY-19323), Adavosertib (HY-10993), and PDD 00017273 (HY-108360) was purchased from MedChemExpress (MCE). Antibodies used in this study include β-actin (AC006), β-Tubulin (AC021), GADPH (AC001), HA tag (AE036), Rnase H1 (A9116), RnaseH2B (A4972), RnaseH2C (A13884), HNRNPLL (A10360), CHK2 (A19543), phospho-Chk2 (Thr68) (AP0590), Histone H3 (A2348), Caspase 3 (A11319), and DNase I (RK20549) was purchased from ABclonal Technology; PARP1 (39561) was purchased from Active motif; γ-H2AX (Ser139) (#2577), phospho-Chk1 (Ser345) (#2341) was purchased from Cell Signaling Technology; Anti-PAR(10H) antibody (sc-56198) were from Santa Cruz Biotechnology; Anti-DNA-RNA Hybrid S9.6 antibody (ENH001) was purchased from Karafast; Anti-nucleolin antibody (ab129200) was purchased from Abcam.

Vector construction, viral infection, transfection and siRNA knockdown

All plasmids used, and the related protocols were described previously [22]. Briefly, the human SRSF2 cDNA sequence and U2AF1 cDNA sequence were used to generate the construct of vectors. The vectors for SRSF2 and U2AF1 were with shRNA sequence targeting the 3’UTR region of SRSF2 (5’-ATATCGGCAAGCAGTGTAAACG-3’) and U2AF1 (5’-CGAAAGTGTTGTAGTTGATTGA-3’) inserted after the CDS region and with or without a HA tag sequence was added to the 3’ into pTRIPZ. Site-directed mutagenesis was used to generate the P95H and aa 95-102 deletion (Δ8aa) alleles for SRSF2 and S34F alleles for U2AF1. These vectors were transfected into 293T cells with the packaging vector psPAX2 and envelope vector pMD2.G and viral supernatant was collected 3 days later. Cells were infected with lentiviral supernatant and puromycin selection (2 mg/ml) for 2 days. Cells were infected with lentiviral supernatant, and with puromycin selection (2 µg/ml) 2 days later to generate stable cell lines.

Human RNASEH1 CDS sequence tagged with the nuclear localisation signal (NLS) at the N-terminus and V5 sequence at the C-terminus was cloned into the ppyCAG expression vector, which was kindly provided by Dr. Juan Calos Izpisua Belmonte. To generate stable RNASEH1-V5 overexpressing cell lines, cells were transfected with the ppyCAGRNASEH1-V5 vector followed by hygromycin selection (200 mg/ml) 2 days later to generate stable cell lines. All cell lines with the expression of designated exogenous genes were confirmed by Western blot (Supplementary Fig. S6A–S6E).

For siRNA knockdown experiments, the siRNA was synthesised by Sangon Biotech (Shanghai), and RNAiMax transfection reagents were purchased from Thermofisher. The siRNA sequences are listed in Supplementary Table S2, and all experiments were carried out according to the same knockdown procedures [49].

Western blotting

Cells were collected by centrifugation for 1000 rpm with 10 min at 4 °C, and then the pellet was suspended in the cell lysis buffer (0.02% SDS, 0.5% Triton-X, 300 mM NaCl, 20 mM Tris-HCl pH 7.5, 1 mM EDTA, 1 mM DTT, 10 mM NaF) with containing 1×protease inhibitors (MCE) and incubated on ice for 10 min. Proteins were loaded on a 12% polyacrylamide gel and transferred to the PVDF membrane (Millipore). The membrane was incubated for 1 h at room temperature in blocking buffer (TBS-T containing 5% skimmed milk), and incubated with primary antibody (2 h at 4 °C). After three washes with TBS-T, the membrane was incubated with HRP-conjugated secondary antibody (1 h at room temperature). Antibody binding was visualised using the ECL Prime western blotting detection system (Bio-Rad).

Dot blot assay

The protein samples were twofold serially diluted with SDS loading buffer and blotted onto the nitrocellulose membrane (Millipore). The membrane was then blocked with the blocking buffer (TBS-T containing 5% skimmed milk) for 1 h at room temperature. Three washes with PBS-T were performed after the primary antibody was incubated for 1 h at 4 °C. After another three washes with PBS-T, the membrane was incubated with the HRP-conjugated secondary antibody for 1 h at 4 °C. The ECL Prime western blotting detection system (Bio-Rad) was used for imaging.

Genomic DNA for the dot blots with S9.6 antibodies was prepared using the QIAamp DNA mini kit (Qiagen), except that the amount of RNase A added was reduced by half. The DNA samples were twofold serially diluted with DEPC-H₂O and blotted onto the hydrophilic nylon membrane (Millipore). The DNA was cross-linked to the membrane with a UV Crosslinker for 120.0 mJ/cm² in 254 nm. Other operations are performed as normal dot blot assay above.

Analyses of cell cycle and apoptosis

For cell cycle analysis, cells were washed with cold PBS and collected for fixation with 75% ice-cold ethanol for 2 h at 4 °C. Resuspended Cell pellets with PBS added 10 μg/ml Rnase A (Thermofisher) and 20 μg/ml propidium iodide (Thermofisher), and samples were incubated in the dark for 30 min at 37 °C. Apoptosis was assessed by Annexin V/DAPI or Annexin V/PI staining. Then, the cells were analysed with flow cytometry (Beckman, Cytoflex, USA).

Immunofluorescence

Cells were fixed with 4% paraformaldehyde/PBS for 10 min at 37 °C, permeabilized with PBS-T (PBS/1% BSA/0.1% Triton-X), blocked with blocking buffer (1×PBS, 0.2% Triton X-100, 5% BSA) for 1 h at RT, and incubated with primary antibodies DNA-RNA Hybrid S9.6 antibody (1:500, Karafast ENH001) and anti-nucleolin (1:1000, Abcam 129200), or anti-PAR(10H) antibody (1:500, sc-56198) or γ-H2AX (Ser139) (1:500, #2577) antibody overnight at 4 °C. On the next day, Cells were washed three times with PBS-T at RT, and were incubated with Abflo® 594-conjugated goat anti-mouse IgG (H + L) (Abclonal, AS054) or Abflo® 488-conjugated goat anti-rabbit IgG (H + L) (Abclonal, AS053) secondary antibodies (1:100 dilution) for 2 h at RT in the dark. Cells were washed three times with PBS-T at RT stained with DAPI for 30 min and mounted with Fluoromount-G (Thermofisher) before imaging. EdU imaging experiments were conducted using the Yefluor 488 EdU Imaging Kit (YEASEN, 40275ES60). S9.6 and nucleolin imaging experiments were carried out as described previously [88, 89], and the nucleolar signal was subtracted before quantifying the nuclear S9.6 signal. All imaging was under Leica TCS SP8 Confocal Microscope.

Immunoprecipitation

HEK293T cell (1 × 10⁷ per sample) incubated in hypotonic buffer (10 mM Tris pH 7.9, 20% glycerol, 1.5 mM MgCl₂, 0.1 mM EDTA, 10 mM KCl, 1×protease inhibitor cocktail, 10 µM PARG inhibitor PDD 00017273), and then lysed in extraction buffer (20 mM HEPES pH 7.9, 25% glycerol, 0.42 M NaCl, 1.5 mM MgCl₂, 0.2 mM EDTA, and 1× protease inhibitor cocktail, 10 µM PARG inhibitor PDD 00017273). The sample was sonicated to the appropriate size by using a Diagenode Bioruptor at 4 °C. Bioruptor setting: High, 30 cycles of 30-sec sonication and 30-sec pause). For Dnase I treatment, the sample was incubated with Dnase I (10U) for 2 h at 37 °C. The lysates were incubated for 12 h with an anti-PAR(10H) antibody (Santa Cruz, sc-56198), or anti-PARP1 antibody (Active motif, 39561), or rabbit IgG (Cell Signaling Technology, #2729) and then incubated for 12 h with either Pierce Protein A/G Magnetic Beads (Thermofisher, 88803). Immunoprecipitated proteins and DNA-RNA hybrids were analysed by western blotting or dot blotting.

RNA extraction, cDNA synthesis and quantitative real-time PCR (qRT-PCR)

RNA was extracted by using Rneasy Plus Mini Kit (Qiagen) and was reverse transcribed using SuperScript III from Invitrogen. qPCR was performed by using TaqMan Fast Advanced Master Mix (Thermofisher). Relative expression levels were calculated using the 2^-∆∆CT, method, and the PCR primers used for targeting SRSF2-exon and 3’ UTR were designed as previously [22] and are listed in Supplementary Table S1.

Quantification and statistical analysis

All graphs represent mean values ± SD or ±SEM. *P < 0.05; **P < 0.01, ***for p < 0.001, ****P < 0.0001 as calculated by unpaired T test.

Author contributions

FZ, LL and LC. conceived and designed the experiments. FZ performed most of the experiments. JS. and LZ contributed to mouse-related assays. YW performed immunocytochemistry- and flow cytometry-related experiments. RL, HG and CS. contributed to immunocytochemistry. LL. provided Srsf2^P95H/+ MA9 mice and MDS-L cells. All authors were involved in interpreting the data. FZ and LC wrote the original manuscript with the contribution of all other authors.

Funding

The work was supported by the National Natural Science Foundation of China (32171289, 31970619), the National Key R&D Program of China (2021YFA1100500), the Innovative Research Group Program of Hubei Province (2020CFA017), and the Fundamental Research Funds for the Central Universities (2042022rc0008) to L.C. This work was also supported in part by NIH grant R01 HL141336, R01 CA248149, ACS RSG-19-036001-LIB, When Everyone Survives Leukaemia Research Award, CURE Childhood Cancer Award, Vera Joseph Dresner Foundation Established Investigator Award, Tower Cancer Research Foundation Scientific Innovations Award for Senior Investigator and Gehr Family Center for Leukaemia Research support to LL.

Data availability

All data generated or analysed during this study are included in this published article (and its supplementary information files).

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Supplementary information

The online version contains supplementary material available at 10.1038/s41416-024-02729-0.