Reconsidering the mechanisms of action of PARP inhibitors based on clinical outcomes

Abstract PARP inhibitors (PARPis) were initially developed as DNA repair inhibitors that inhibit the catalytic activity of PARP1 and PARP2 and are expected to induce synthetic lethality in BRCA‐ or homologous recombination (HR)‐deficient tumors. However, the clinical indications for PARPis are not necessarily limited to BRCA mutations or HR deficiency; BRCA wild‐type and HR‐proficient cancers can also derive some benefit from PARPis. These facts are interpretable by an additional primary antitumor mechanism of PARPis named PARP trapping, resulting from the stabilization of PARP‐DNA complexes. Favorable response to platinum derivatives (cisplatin and carboplatin) in preceding treatment is used as a clinical biomarker for some PARPis, implying that sensitivity factors for platinum derivatives and PARPis are mainly common. Such common sensitivity factors include not only HR defects (HRD) but also additional factors. One of them is Schlafen 11 (SLFN11), a putative DNA/RNA helicase, that sensitizes cancer cells to a broad type of DNA‐damaging agents, including platinum and topoisomerase inhibitors. Mechanistically, SLFN11 induces a lethal replication block in response to replication stress (ie, DNA damage). As SLFN11 acts upon replication stress, trapping PARPis can activate SLFN11. Preclinical models show the importance of SLFN11 in PARPi sensitivity. However, the relevance of SLFN11 in PARPi response is less evident in clinical data compared with the significance of SLFN11 for platinum sensitivity. In this review, we consider the reasons for variable indications of PARPis resulting from clinical outcomes and review the mechanisms of action for PARPis as anticancer agents.


| INTRODUC TI ON
Among the PARP family members (PARP1-PARP17), PARP1 and PARP2 act as DNA repair enzymes for DNA single-strand breaks.
Hence, catalytic PARP inhibition by PARP inhibitors (PARPis) prevents the repair of DNA single-strand breaks, and PARPis act as DNA repair inhibitors. Since the discovery of synthetic lethality of PARPis in BRCA mutant cells that impair the repair of DNA doublestrand breaks (DSBs), 1,2 clinical PARPis with comparably high catalytic inhibition potency (olaparib, niraparib, talazoparib, rucaparib, and veliparib) have been developed. According to the original concept, PARPis should be selectively toxic to BRCA mutation or HRD cancer cells. However, clinical trials revealed the significant benefit of PARPis in BRCA wild-type or HR-proficient cancers, while BRCA mutation or HR-deficient cancers received superior benefit. [3][4][5][6] Hence, current indications of PARPis have been expanded and are not restricted to cancers with BRCA mutations ( Table 1). Former favorable response to platinum derivatives (cisplatin and carboplatin) is used as a clinical biomarker for PARPi sensitivity regardless of BRCA status. These facts are not interpretable by the initial synthetic lethal model of PARPis.
A decade ago, we reported an additional primary mechanism of action of PARPis, named PARP trapping. 7,8 In the presence of PARPis, PARP1 and PARP2 (hereafter, which we described as PARP if not explicitly mentioned) bind the 5′-deoxyribose phosphate group-containing DNA ends noncovalently, 9 generating highly toxic PARP-DNA complexes ( Figure 1). As PARPis turn the PARP protein toxic, they act as "PARP poisons," which explains that the antitumor effects of PARPis completely disappear in PARPdeficient cells. 8 Table 1). Structural studies revealed an allosteric folding change of a helical domain of PARP1, leading to different retention potency of PARP1 on single-strand breaks. 11 Overall, these differences are reflected in the drug dosing; for example, the daily dose of talazoparib is 1 mg, while the daily dose of other PARPis is hundreds mg, indicating that PARP-trapping potency is the limiting factor to decide the clinical dose. The variety of indications and usages of PARPis based on the results of clinical trials deepen the understanding of PARPis and let us reconsider the most relevant mechanisms of action of PARPis in the human body. In this review, we first summarize recent topics about PARP trapping and then consider the reasons for variable indications of PARPis resulting from clinical outcomes. Next, we introduce Schlafen 11 (SLFN11) as a cause of cross-sensitivity with platinum derivatives and propose the "hyper synthetic lethal strategy" using SLFN11 protein expression and BRCA mutation as biomarkers for PARPis.

| HOW DO C AN CER CELL S PRO CE SS PARP-TR APPING LE S IONS?
Replication is often challenged by proteins covalently bound to DNA, also known as DNA-protein crosslinks (DPCs). DPCs originate when proteins become crosslinked to DNA after exposure to UV light or aldehydes or due to faulty enzymatic reactions. 12 A representative example of enzymatic DPC is a topoisomerase 1 (TOP1)-DNA cleavage complex (TOP1cc) generated through the TOP1-mediated covalent bond between 3'-DNA ends and the catalytic tyrosyl residue of TOP1. 13 Failure in the self-resealing of TOP1ccs results in stabilized TOP1-DPCs, which are trapped by TOP1 poisons, such as camptothecin (CPT), and its clinical derivatives irinotecan and topotecan. 14 Because TOP1-DPCs are products of a physiological reaction, eukaryote cells possess multiple pathways to dissolve the TOP1-DPCs by excising and ligating the associated breaks. Tyrosyl-DNA phosphodiesterase 1 (TDP1) cleaves the tyrosyl-DNA bonds, whereas a structure-specific endonuclease MRE11 removes the TOP1-DPC along with the adjacent DNA segment. 14 A metalloprotease Spartan (SPRTN) debulks TOP1-DPCs to make the peptide-DNA bonds accessible to the repair factors. 14 Similar repair pathways exist for TOP2-DPC with TDP2 and MRE11 for their excision. 15 Getting back to the subject of PARP trapping, the PARP-DNA complex, a noncovalent bond at the 5'-DNA ends, is an unnatural product that is uniquely formed in the presence of PARPis. One possible exception can be the case happening in XRCC1-deficient condition, where PARP1 occupies DNA ends and blocks base excision repair. 16 However, XRCC1-deficient cell is not found at the tran- The metalloprotease SPRTN involved in the debulking of TOP1-DPCs is recruited to trapped PARP1 in S-phase to assist in the excision and replication bypass of PARP1-DNA complexes. 18 Hence, SPRTN-deficient cells are hypersensitive to talazoparib and olaparib but not to veliparib. 18 The serine protease FAM111A, a PCNAinteracting protein, also plays a vital role in mitigating the effects of protein obstacles on replication forks. FAM111A protects replication forks from stalling at PARP1-trapping lesions, thereby promoting cell survival after PARPi treatment. 19  interactomes identified an interaction between trapped PARP1 and the ubiquitin-regulated p97 ATPase/segregase. 22 Trapped PARP1 has been shown to be SUMOylated by PIAS4 and subsequently ubiquitylated by the SUMO-targeted E3 ubiquitin ligase RNF4, promoting the recruitment of p97 and removal of trapped PARP1 from chromatin ( Figure 1). 22 Notably, this pathway appears rather general as it is also involved in the repair of trapped TOP cleavage complexes. 23

| D IFFEREN CE S B E T WEEN PARPIS AND CONVENTIONAL DNA-DAMAG ING ANTIC AN CER DRUG S
PARPis generate lesions leading to replication-dependent DSBs with trapped PARP. In terms of expectation of replication-dependent cell death, platinum drugs and TOP inhibitors, which are conventional DNA-damaging anticancer agents, also have similar mechanisms of action in that they ultimately induce DSBs. 24 Hence, HR genes are common critical repair factors for PARPis and DNA-damaging agents.
Here, we point out the differences between PARPis and the conventional DNA-damaging agents that generate bulky DNA adducts.
Platinum drugs covalently crosslink DNA. TOP inhibitors trap covalent TOP1-and TOP2-DPC. Therefore, even if the drug concentration is reduced, DNA lesions, once generated, will not be restored unless they are repaired. In contrast, the PARP-DNA complex is not a covalent bond, so the trapped PARP can be quickly released from DNA when the concentration of PARPis becomes lower ( Figure 2). 8 We previously showed that when the PARPi was removed from the cell culture medium, PARP-DNA complexes began to be released after 5 minutes and were wholly released after 30 minutes with the recovery of PARylation. 8 Once trapped PARP is released, the remaining single-strand breaks can be rapidly repaired by the reactivated PARP regardless of HR status ( Figure 2). However, if the PARP-DNA complex has already generated collisions with replication forks, DSBs with clean (ie, protein-unbound) DNA ends remain.
The clean DSBs can be repaired in HR-proficient cells, while still highly toxic in HR-deficient cells, which is attributed to the original synthetic lethality model ( Figure 2).  (Figure 2). By contrast, maintaining PARP trapping should be a key for anticancer acting in HR-proficient cancer cells ( Figure 2

| WHI CH G ENE S CONTRIBUTE TO THE CROSS -S EN S ITIVIT Y B E T WEEN PARPIS AND PL ATINUM DERIVATIVE S?
Although some repair factors are uniquely crucial for platinum derivatives or PARPis, 10 the utility of platinum sensitivity as a clinical biomarker for PARPis implies that PARPis and platinum agents share similar sensitivity and resistance factors. The clinical outcomes are readily recapitulated across ~900 cell lines, revealing the extremely high sensitivity correlation between PARPis (talazoparib, olaparib, and veliparib) and cisplatin regardless of tumor types ( Figure 3).
Notably, the P value of each PARPi correlates with their PARPtrapping potency ( Figure 3 and Table 1 However, we may miss unlisted repair factors involved in the crosssensitivity. Hence, we need to find a way to identify the patients who do not carry HR gene mutations but are yet sensitive to PARPis and platinum agents.

| S LFN11 IS A COMMON S EN S ITIZER TO PL ATIN UM DERIVATIVE S AND PARPIS IN C AN CER MODEL S
We raise the issue of why each cell line dot in Figure 3 is along the re- independent laboratories identified that SLFN11 expression is highly correlated with sensitivity of the TOP1 inhibitor topotecan. 36,37 The high correlation also applies to cisplatin and DNA replication inhibitors such as cytarabine. 36 Although the initial actions are different, these drugs commonly induce replication blocks, activate S-phase checkpoint, and generate abnormal (stressed) replication forks.
Replication stress 38   The correlation between SLFN11 expression and drug sensitivity is also applicable to PARPis. We previously showed that SLFN11 expression is significantly correlated with sensitivity to talazoparib in the NCI-60. 59 The significant correlation is also validated in the GDSC database for talazoparib, olaparib, and veliparib ( Figure 4).
Again, the P values were correlated with the PARP-trapping potency. We showed that SLFN11 enhances sensitivity to olaparib Together, the observations and conclusions listed above explain why SLFN11 can be a primary common sensitivity factor for platinum drugs and PARPis.

| DOE S S LFN11 NEED B RC A M UTATI O N TO ENHAN CE OL APARIB S EN S ITIVIT Y IN THE CLINI C AL S E T TING?
Recently, the group of AstraZeneca examined the effect of SLFN11 on olaparib sensitivity. 56 They first showed that pa-  The single-strand binding protein replication protein A (RPA) protects the replication gaps, but exhaustion of RPA results in genome instability, leading to synthetic lethality 62 (left). In SLFN11-positive cells, SLFN11 likely binds the DNA-bound RPA, possibly leading to SLFN11-mediated cell death named "hyper synthetic lethal." (right, the smaller cartoon indicates more shrunk cancer). Note that the sub-micromolar concentration of olaparib is not supposed to generate a PARP-DNA complex based on our previous report but is enough to inhibit the catalytic activity of PARP. 8 (B) Continuous PARP trapping is likely a key for the sensitivity of SLFN11-positive cancer cells to PARP trappers. Continuous PARP trapping induces replication stress that activates SLFN11 toward cell death