Topoisomerases and cancer chemotherapy: recent advances and unanswered questions

Type IB topoisomerases: maintenance of nuclear and mitochondrial genome stability

In eukaryotes, nuclear TOP1 catalyzes the relaxation of local domains of positive and negative supercoils during DNA replication, recombination, transcription, and possibly chromosome condensation^4,5,10. Stabilization of TOP1ccs by camptothecins during replication is an effective strategy for treating solid tumors and hematologic malignances, as discussed below. During transcription, the phosphorylated C-terminal domain of the catalytic subunit of RNA polymerase II binds and activates TOP1, effectively tethering TOP1 to the transcriptional machinery^11,12. TOP1 then relaxes positive supercoils, which are generated ahead of the transcription complex and could otherwise impede its progress, as well as negative supercoils behind the transcription complex.

In the absence of TOP1, local accumulation of negative supercoils facilitates the formation of R-loops, stable hybrid RNA–DNA duplexes of the nascent RNA transcripts and template strands. R-loops also allow the formation of secondary structures, such as G-quadraplexes and hairpins, in the single-stranded non-template strand. RNase H1 and H2 can degrade RNA in these RNA–DNA heteroduplexes. While genome-wide R-loop mapping indicates context-dependent gains and losses in R-loops when TOP1 is depleted¹³, it is the increased levels of R-loops and G-quadraplexes that are associated with dysregulation of transcription and replication as well as genome instability.

The misincorporation of ribonucleotides into DNA, at rates approaching 10⁶ ribonucleotides per genome per replication cycle¹⁷, can also lead to replication stress, single- and double-strand breaks, and small deletions^18–20. Ordinarily these ribonucleotides are removed by the concerted action of RNAse H2, DNA polymerase δ, FLAP endonuclease, and DNA ligase 1¹⁷. However, if ribonucleotides are not removed, TOP1 cleavage of the strand immediately 3’ to the ribonucleotide results in the nucleophilic attack of the 2’OH of the ribonucleotide on the TOP1cc to generate a 2’,3’ cyclic phosphate at the 3’ DNA end and release of TOP1. A second, upstream TOP1 cleavage event can then liberate a short oligo with the modified 3’ end, trapping the TOP1cc across a gap. If DNA strand realignments juxtapose the free 5’OH from the first cleavage and the TOP1cc, enzyme-mediated ligation can produce short deletions. In highly transcribed genes, this TOP1-mediated mutagenesis can be exacerbated by the tethering of TOP1 to RNA polII²¹.

A recent genome-wide CRISPR screen showed that interruption of genes encoding the three subunits of RNase H2 enhances human cell line sensitivity to the poly(ADP-ribose) polymerase (PARP) inhibitor olaparib²². Further studies attribute this olaparib hypersensitivity to increased ribonucleotide-dependent stabilization of TOP1ccs, which can serve as PARP1 substrates. These observations provide a compelling rationale for inhibiting PARP in order to trigger TOP1cc-initiated killing in cancers with deleted or mutated RNASEH2B.

In vertebrates, a second nuclear-encoded type IB topoisomerase (TOP1MT) selectively localizes to mitochondria and catalyzes the relaxation of circular mitochondrial DNA²³. Despite its similarity to nuclear TOP1, TOP1MT does not contribute to camptothecin-induced toxicity. Instead, TOP1MT physically associates with mitochondrial ribosome subunits to promote mitochondrial translation, which is critical for hepatocellular carcinoma cell growth²⁴. These findings suggest that inhibition of TOP1MT activity, rather than stabilization of TOP1MTccs, might be an effective strategy for targeting this enzyme to treat some cancers.

Evolving understanding of eukaryal topoisomerase IIIα and β (TOP3α and β)

Distinct from the swivelase activity ascribed to type IB enzymes, type IA topoisomerases exhibit a mechanism of enzyme-bridged strand passage (Figure 1)²⁵. As with bacterial TOPA, eukaryotic TOP3α enzymes (including yeast TOP3) preferentially relax highly negatively supercoiled DNA and decatenate duplex DNA molecules tethered by single-stranded DNA interlinks or hemicatenanes²⁶. Differential splicing produces nuclear and mitochondrial isoforms of vertebrate TOP3α. Nuclear TOP3α forms a complex with the BLM helicase and RMI1 and RMI2 proteins²⁷ to resolve double Holliday junctions during recombination²⁸. In contrast, mitochondrial TOP3α decatenates newly replicated mtDNA circles, which are linked by a hemicatenane formed at the origin of replication, to allow segregation of replicated mitochondrial genomes²⁹. Accordingly, TOP3α dysregulation results in human mitochondrial disease.

TOP3β, another type IA topoisomerase encoded by the TOP3B gene, binds mRNA and functions during neurodevelopment³⁰. Recent studies, made possible by the development of circular double-stranded and knotted single-stranded RNA substrates, suggest that TOP3β can catalyze RNA topoisomerization^31–33. In multicellular organisms, an association with Tudor domain-containing protein 3 (TDRD3) localizes TOP3β to transcriptionally active chromatin and polyribosomes^34,35. Although type IA enzymes with RNA topoisomerase activity have been detected in all domains of life³⁴, the biological significance of RNA topoisomerization requires further study.

Contribution of topoisomerase II to chromosome architecture and genomic stability

In eukaryotes, TOP2 is a homodimeric enzyme that relaxes positively or negatively supercoiled DNA and catenates or decatenates duplex DNA via transient breakage of both DNA strands (Figure 1). Yeast encode a single TOP2, while human cells express TOP2α and TOP2β enzymes, encoded by the TOP2A and TOP2B genes, respectively. Although human TOP2 enzymes exhibit structural and mechanistic similarities, TOP2α decatenates sister chromatids during chromosome segregation, whereas TOP2β has been implicated in transcription. Several recent studies further define distinct roles of these enzymes in chromosome dynamics.

TOP2β plays a surprising and important role in interphase chromatin organization. High-resolution whole-genome chromatin conformation capture (Hi-C), or in situ Hi-C with DNA–DNA proximity ligation, allows chromatin fragments in close proximity to be identified. These techniques have determined that chromosomes are organized into topologically associated domains (TADs) of ~200 kb to 1 Mb, typically bound by chromatin enriched in transcriptionally active genes. According to current models, DNA is actively extruded through one or paired cohesin rings to generate TADs until DNA bound by the CCCTC binding factor (CTCF) is encountered^36,37. Recent studies suggest that CTCF becomes associated with loop anchors and unidirectionally halts DNA extrusion. TOP2β is then recruited to loop anchors to alleviate the positive supercoils induced by cohesin-derived DNA extrusion. The resulting TOP2β-induced breaks are transcription independent but correlate with cohesin³⁸. At a low frequency, unresolved TOP2βccs at these loop anchors can also lead to DNA breakage and translocations³⁹. Thus, TOP2β involvement in topological dynamics associated with chromosome organization contributes somewhat unexpectedly to chromosome breakage and rearrangements.

During chromosome segregation, intertwined DNA duplexes (catenanes) are resolved or decatenated by TOP2 in yeast and TOP2α in human cells. TOP2 enzymes can also readily catenate DNA duplexes in close proximity. Yet increased positive supercoiling drives decatenation, based in part on an intrinsic enzyme bias towards decatenation. A persistent question, then, has been the source of this positive supercoiling to drive decatenation. In yeast, condensin-mediated positive DNA supercoiling increases as cells enter mitosis⁴⁰. In human cells, we now know that this positive supercoiling reflects the action of TOP3α, which (as part of the TRR complex with RMI1 and RMI2) associates with the Plk1-interacting checkpoint helicase (PICH) to produce extremely high-density positive supercoils⁴¹. Subsequent relaxation of negative supercoils by TOP3α results in the accumulation of positive supercoils, which drives decatenation by TOP2α. These studies provide the first evidence for topoisomerase-induced stable domains of positive supercoils in eukaryotic cells and illustrate how DNA extrusion can be locally harnessed to drive chromosome disjunction.

TOP1cc removal: multiple pathways and unanswered questions

DNA–protein crosslinks (DPCs) include not only TOP1ccs, but also crosslinks induced by aldehyde products of demethylation reactions, cisplatin, UV light or ionizing radiation, and trapping of DNA methyltransferases covalently bound to 5-aza-cytosine (reviewed in 51–53). Distinct repair pathways have evolved to resolve these DPCs; however, TOPccs present unique challenges because they also involve protein-linked DNA breaks. Recent studies have provided new insight into the action of tyrosyl-DNA phosphodiesterases 1 and 2 (TDP1 and TDP2, respectively) and DNA-dependent proteases such as SPARTAN (also known as SPRTN) that recognize and reverse persistent TOP1ccs.

Several lines of evidence implicate TDP1 in TOP1cc removal.TDP1 can de-esterify peptidic tyrosine-phosphoesters^54,55, and TDP1 knockdown results in increased foci containing the TOP1 active site peptide covalently bound to DNA⁵⁶. Earlier studies suggested that TPD1 efficiently removes short TOP1 peptides from DNA but is less efficient at removing longer peptides or full-length TOP1⁵⁷. However, recent studies of TDP1 mutants suggest that full-length TOP1 can, in fact, be released from chromatin-bound TOP1ccs in yeast and human cells^58,59.

The observation that TDP1 knockdown or knockout has little impact on yeast or mammalian cell sensitivity to camptothecins^55,60 suggested early on that there must be redundant or overlapping repair pathways. In the absence of TDP1, the 5’-tyrosyl phosphodiesterase TDP2^73,74 and a pathway involving the repair proteins XPF and ERCC1⁷⁵ participate in TOP1cc removal. An additional pathway involves cleavage of the adducted DNA by the nuclease MUS81 followed by polymerization and ligation across the resulting gap⁷⁶.

Conditions that promote the use of one pathway over another are still being elucidated. Poly(ADP-ribosyl)ation of TDP1 appears to influence this choice⁷⁵. In addition, the deubiquitylase UCHL3 was recently shown to regulate TDP1 proteostasis⁷⁷, implicating ubiquitin-dependent regulation of TDP1 in the repair of camptothecin-induced TOP1ccs.

Emerging results also suggest a role for proteases in the removal of TOP1ccs. Although early studies implicated the proteasome in this process^78–81, the observation that proteasome-mediated TOP1 degradation occurs only at micromolar camptothecin concentrations and not at more clinically relevant low nanomolar concentrations⁵⁶ calls this model into question. Instead, the nuclear metalloproteinase SPARTAN, which contains a ubiquitin-binding domain and a single-stranded DNA-binding motif⁸², has recently been shown to reverse TOP1ccs trapped by normal DNA metabolism or nanomolar camptothecin concentrations^56,82–84. In Saccharomyces cerevisiae, the SPARTAN homolog Wss1 is critical for survival after camptothecin treatment, and the recombinant protease is able to cleave TOP1ccs⁸⁵. Likewise, Sprtn downregulation increases TOP1ccs in murine fibroblasts⁵⁶ and enhances camptothecin sensitivity in vitro^56,82,83. Mice bearing a hypomorphic Sprtn allele contain increased hepatocyte TOP1ccs and develop hepatic neoplasms⁵⁶, which recapitulates Ruijs-Aalfs syndrome, a disorder characterized by germline SPRTN mutations, genomic instability, and early onset hepatocellular carcinoma^86–88. This hepatocyte-specific pathology is, at present, poorly understood. Higher TOP1 protein levels⁵⁶ might contribute to preferential trapping of TOP1ccs in Spartan-deficient hepatocytes, but the possibility that alternative proteases facilitate the removal of TOP1ccs in other tissues also merits investigation. Additional unresolved issues include i) the coupling between proteases and phosphodiesterases or nucleases and ii) the relative contributions of protease-dependent versus protease-independent pathways in TOP1cc removal.

Recognition of trapped TOP1ccs: a plethora of modifications

A particularly perplexing question is how do trapped TOP1ccs come to be marked for repair or proteolytic degradation? Post-translational modifications of TOP1 and TOP1ccs by ubiquitin^78–81, ubiquitin-like modifiers^89,90, and phosphorylation^91–93 have been reported, but the physiological relevance of these modifications to TOP1cc resolution is complicated by the use of high camptothecin concentrations.

In this context, studies implicating the small ubiquitin-like modifier (SUMO) in TOP1 action might be pertinent. TOP1 is modified by SUMOylation in CPT-treated yeast and mammalian cells^89,90,94. In addition, downregulation or mutation of the sole SUMO E2 ligase Ubc9 is associated with TOP1cc stabilization and enhanced camptothecin toxicity^89,90,95,96. However, recent studies ascribe these effects to a change in Ubc9 substrate specificity⁹⁷, consistent with more global changes in SUMOylation of other proteins involved in the DNA damage response and not a direct effect on TOP1.

It is also worth noting that TOP1cc degradation by Wss1 (yeast SPARTAN) occurs in a SUMO-dependent fashion⁹⁸, while SPARTAN preferentially binds ubiquitin through a UBZ domain, and its activity is regulated by deubiquitinylation^82,99. These differences in the SUMO- versus ubiquitin-mediated regulation of Wss1 and SPARTAN, and the inability of SPARTAN to complement wss1Δ yeast cells, led Mailand and colleagues to examine SUMO-dependent responses to various DPCs¹⁰⁰. Their studies implicate SprT metalloproteases of the ACRC/GCNA-1 family in SUMO-dependent resolution of DPCs. While it remains to be determined if GCNA-1 family proteases impact sensitivity to drug-stabilized TOP1ccs, these observations support the notion that other, as-yet-uncharacterized metalloproteases may regulate cellular responses to topoisomerase-mediated DNA damage via distinct ubiquitin-like protein modifications. The potential therapeutic implications of these recently recognized repair pathways remain to be more fully investigated.

Extending the paradigm to TOP2cc

The machinery responsible for removing trapped TOP2ccs is even less clearly defined. Proteasomal degradation of TOP2 after teniposide treatment has been reported¹⁰¹, contributing to a model in which collisions between advancing transcription complexes and TOP2ccs result in irreversibly trapped TOP2–DNA complexes, which are marked by ubiquitylation and degraded by the proteasome.

More recent studies have identified several alternatives to this model. First, SPARTAN knockdown results in slightly increased levels of TOP2ccs and etoposide sensitivity⁸³, suggesting SPARTAN might degrade TOP2 before removal of the active site peptide from DNA. Contrary to this model, however, increased TOP2ccs were not observed in MEFs conditionally deleted for Spartan, and MEFs harboring a hypomorphic Spartan allele were not hypersensitive to etoposide⁵⁶. Thus, the role of SPARTAN and other nuclear metalloproteinases in the removal of trapped TOP2cc requires further clarification.

Trapped TOP2ccs may also be removed without TOP2 proteolysis. The MRE11 nuclease has been implicated in the removal of TOP2 from DPCs^102,103. Moreover, TDP1^104,105 and TDP2^73,106 have both been reported to release the TOP2 active site peptide when it is linked to 5’OH of the DNA backbone. In particular, TDP2 can reverse covalent binding of TOP2α or TOP2β to a suicide DNA substrate, and this activity increases up to 1000-fold in the presence of the SUMO E3 ligase ZNF451 owing to increased binding of TDP2 to SUMOylated TOP2¹⁰⁷.

Two recent studies further suggest that coordination of SUMO- and Ub-dependent TOP2 modifications may be critical for genomic stability. In etoposide-treated fission yeast, the DNA translocase Rrp2 binds to SUMOylated TOP2ccs and prevents recruitment of the SUMO-dependent E3 ubiquitin ligase STUbL, thereby preventing STUbL-mediated TOP2 ubiquitinylation and degradation¹⁰⁸. Instead, Rrp2 facilitates the eviction of intact TOP2 from the DNA and concomitant DNA resealing, thereby increasing genomic stability and etoposide resistance. In other studies, the Smurf2 E3 ubiquitin ligase was shown to switch the pattern of TOP2α modification from K48 polyubiquitylation that promotes proteasomal degradation to monoubiquitylation, which leads to increased TOP2α protein levels, suppression of anaphase bridge formation, and etoposide resistance¹⁰⁹.

In summary, although multiple pathways have been implicated in the reversal of trapped TOP2ccs, other studies suggest that protecting TOP2ccs from proteolytic degradation is also critical for maintaining genome stability. Further studies are required to assess whether distinct pathways are called into play in response to different levels of DNA damage, as appears to be the case with TOP1, or whether current inconsistencies reflect differential expression of pathway components in different cell types.

Can the delivery and efficacy of topoisomerase poisons be improved?

Consistent with observations that TOP1 and TOP2 poisons are preferentially toxic during S phase, classic studies demonstrated that the administration of irinotecan on a five-times-daily schedule for 2 weeks is more active against human cancer xenografts than less-protracted schedules¹¹⁰. Etoposide administered every other day for three doses is likewise more effective against L1210 murine leukemia than a higher dose administered once. In the clinical setting, these observations have been translated into protracted schedules of both irinotecan¹¹¹ and etoposide¹¹². Because these prolonged schedules can be inconvenient and toxic¹¹¹, there has been an ongoing search for alternatives, including new topoisomerase poisons, drug formulations that extend the half-life of TOPccs, and strategies to increase tumor-selective drug delivery (Table 3)¹¹³.

Among the new classes of TOP1 or TOP2 poisons, TOP1-directed indenoisoquinolines^114,115 are furthest along in development. These agents, which lack a lactone ring and, in contrast to camptothecin derivatives, do not exist in equilibrium between an active agent and inactive derivative¹¹⁵, exhibit promising activity against canine lymphomas¹¹⁶. Assessments of their activity in humans are awaited with interest.

An alternative approach involves new formulations that extend tumor exposure. MM398, a nanoliposomal irinotecan formulation¹¹⁷, gained FDA approval in combination with 5-fluorouracil and leucovorin for gemcitabine-resistant pancreatic cancer^64,65. In contrast, NKTR-102, a PEGylated irinotecan, exhibited disappointing activity in breast and ovarian cancer^118,119. Whether the different outcomes for these two sustained-release irinotecan formulations reflect differences in pharmacokinetics, intratumoral accumulation, or simply choice of tumors studied is not clear.

Santi and coworkers developed an ultra-long-acting Prolynx PEG~SN-38 that accumulates in tumors and delivers active SN-38 rather than the prodrug irinotecan¹²⁰. Liposomal topotecan formulations are also being developed¹²¹. Whether the promising preclinical activity seen in experimental tumors, which is thought to reflect enhanced permeability and retention of nanoformulations¹²², can be translated into increased clinical efficacy remains to be determined.

Antibody–drug conjugates (Table 3) also hold the promise of more selectively delivering TOP1 poisons to tumor cells. DS-8201^123,124, a conjugate of the TOP1 poison deruxtecan with the anti-HER2 antibody trastuzumab, is currently undergoing extensive preclinical and early clinical testing (www.ClinicalTrials.gov). Promising clinical activity has been observed in trastuzumab-resistant breast and gastric cancers^125,126. An immunoconjugate of SN-38 and antibody to human trophoblast cell surface antigen 2 (TROP2), a glycoprotein found on several solid tumors^135–137, likewise exhibits promising activity in breast^138,139 and lung cancers^140,141 (Table 3).

Should topoisomerase poisons and DNA damage response modulators be combined?

Because TOP1 and TOP2 poisons lead to DNA damage, there has been substantial interest over the past few years in combining these drugs with several different DNA damage response modulators.

PARP inhibitors. PARP inhibitors (PARPis), which inhibit PARP1 as well as other PARP family members¹⁵³, are FDA approved for high-grade serous ovarian cancer, germline BRCA1/2-mutated breast cancer, and BRCA1/2-mutated castration-resistant prostate cancer^154–158. Additional studies identified a role for PARP1 in stabilizing^159–161 and restarting^162,163 stalled replication forks, including forks stalled by TOP1ccs. Consistent with these studies, Curtin et al. demonstrated that PARPis increase killing by TOP1 but not TOP2 poisons¹⁶⁴. This TOP1 poison/PARPi synergy likely results from trapping of inhibited PARP^165,166 at sites of TOP1ccs or TOP1cc-induced DNA damage¹⁶⁷, perhaps in concert with diminished recruitment of TDP1 to TOP1ccs⁷⁵.

Building on xenograft studies^168,169, several clinical trials have evaluated TOP1 poison/PARPi combinations (Table 4). Most started with myelosuppressive topotecan or irinotecan regimens¹⁷⁰. Because PARPis also suppress bone marrow function¹⁷¹, it is not surprising that profound myelosuppression occurs with these combinations, limiting drug doses that can be safely administered together (Table 4). In contrast, by starting with a less myelosuppressive weekly topotecan regimen¹⁷² and only administering PARPi for 72 hours around each topotecan dose to maximize the synergy, Wahner Hendrickson and coworkers were able to escalate topotecan and veliparib to three-quarters of the single-agent MTDs¹⁷³. Whether the approach of i) using a less myelosuppressive TOP1-directed regimen and/or ii) giving intermittent PARPi timed to coincide with maximal TOP1cc stabilization will be an effective way forward with TOP1 poison/PARPi combinations remains to be further assessed.

Combinations with ATR and CHK1 inhibitors. Stalled replication forks activate the replication checkpoint, a biochemical pathway involving the DNA damage-activated kinases ATR and CHK1 that inhibits new origin firing, stabilizes stalled forks, and increases DNA repair^174–176. Consistent with a role for this pathway in cellular recovery from TOP1cc-induced damage^177,178, inhibition of CHK1^179–181 or ATR^182,183 sensitizes cancer cells to TOP1 poisons in vitro and in xenografts. Earlier development of a CHK1 inhibitor/TOP1 poison combination¹⁸⁴ was abandoned because of off-target cardiac toxicities of the CHK1 inhibitor¹⁸⁵. More recent studies have examined ATR inhibitors (e.g. M6620 and AZD6738)^186,187 with TOP1 poisons. Reportedly, an M6620/topotecan combination was well tolerated, except for myelosuppression, and induced partial responses in two out of 21 (9.5%) patients¹⁸⁸. A phase II trial of this combination in small cell lung cancer (ClinicalTrials.gov identifier: NCT02487095) and a phase I trial of an irinotecan/M6620 combination (NCT02595931) are ongoing.

TOP1 poison/immune checkpoint inhibitor combinations. While immune checkpoint blockade is highly active in certain solid tumors^189,190, many common cancers respond poorly. However, recent studies suggest that DNA damage can stimulate immune responses through multiple mechanisms. First, release of DNA to the cytosol after DNA damage¹⁹¹ activates the stimulator of interferon genes (STING) pathway^192,193, leading to the production of pro-inflammatory cytokines. Second, DNA damage-induced release of tumor cell microvesicles can increase immune activation^194,195. Third, DNA damage increases antigen presentation on tumor cell MHC class I molecules, leading to enhanced dendritic cell activation and T cell responses¹⁹⁶. Importantly, these changes have been observed after treatment with TOP1 poisons, potentially contributing to the synergy observed when irinotecan or DS-8201a is combined with anti-PD-1 in vivo^133,197,198. Clinical trials are also evaluating TOP2 poisons in combination with PD-1 antibodies (www.ClinicalTrials.gov).

Predicting response to topoisomerase poisons

Given the toxicities of topoisomerase poisons, the ability to predict responses and avoid treatment of patients unlikely to benefit would represent a major advance. In isogenic yeast^199,200 or mammalian cells^201,202, elevated TOP1 or TOP2 levels are associated with increased killing by topoisomerase poisons (Figure 2). Additional studies indicate that high TOP1 expression correlates with improved colorectal cancer response to irinotecan^203–206 and TOP2 gene amplification is associated with improved breast cancer response to doxorubicin^207,208. However, expression and response are not so tightly correlated that outcomes of individual patients can be predicted from expression data alone.

The frequent occurrence of transport-mediated resistance raises the possibility that responses might be better predicted by assaying TOPccs after the first dose of therapy. While earlier techniques for measuring TOPccs were labor intensive and nonspecific, a recently described antibody to TOP1ccs²⁰⁹ opens the possibility of specific, quantitative assays to address the relationship between TOP1ccs and response to TOP1 poisons. Unfortunately, similar reagents to assess TOP2ccs are not currently available.

It is possible that factors other than TOPccs will need to be assessed to predict drug responses. Homologous recombination (HR) defects convey heightened sensitivity to TOP1 and TOP2 poisons in yeast²¹⁰ and mammalian cells^{167,211–213}. Moreover, BRCA1- or BRCA2-mutant ovarian cancers have a higher response rate to liposomal doxorubicin^214,215. Likewise, breast cancers deficient in BRCA1 or HR activity respond better to anthracycline-based neoadjuvant therapy^216,217. In contrast, BRCA1/2 mutation status was not correlated with response to the TOP1 poison topotecan administered alone²¹⁴ or in combination with PARPi¹⁷³. Thus, HR status might need to be considered in predictive algorithms, but the impact of HR status might also vary by drug class.