The use of poly(ADP-ribose) polymerase inhibitors (PARPis; such as olaparib) to induce synthetic lethality has been limited primarily to cancers that have homologous recombination (HR) defects or BRCA mutations. Esposito et al. report the mechanisms of PARPi sensitivity in cells expressing certain leukaemia fusion oncoproteins, and they provide preclinical proof-of-concept for the use of PARPis as a novel approach for the treatment of acute myeloid leukaemia (AML), a cancer for which the same toxic and largely ineffective therapies have been used for more than 50 years.

Credit: Lara Crow/NPG

The authors first showed that PARPis suppressed the in vitro colony-forming ability of primary mouse haematopoietic cells transformed with the leukaemia-associated transcription factors AML1–ETO and PML–RARα, but not those transformed with MLL–AF9 or E2A–PBX. These findings were replicated in olaparib-treated patient-derived leukaemia cell lines and primary AML samples carrying AML1–ETO, PML–RARα or MLL–AF9. Furthermore, in immunocompromised mice transplanted with patient-derived cell lines, olaparib treatment significantly delayed the onset of disease driven by AML1–ETO and PML–RARα but not that driven by MLL–AF9, suggesting that PARPis are a promising therapy for leukaemia driven by certain fusions. But what are the mechanisms that underpin this differential PARPi sensitivity?

Transformed primary mouse haematopoietic cells expressing AML1–ETO, PML–RARα and MLL–AF9 (but not those expressing E2A–PBX) showed significant levels of γH2AX DNA damage foci, which were further increased by PARPi treatment. These data were in line with the generally accepted view of a defective DNA damage response (DDR) underlying PARPi sensitivity, but the effect was not specific to cells expressing PARPi-sensitive fusions. Notably, HR efficiency and expression of HR-associated genes was shown to be defective in transformed mouse cells, patient-derived cell lines and primary AML samples expressing the PARPi-sensitive fusions AML1–ETO and PML–RARα. Compared with MLL–AF9- and E2A–PBX-transformed mouse cells, those expressing AML1–ETO and PML–RARα showed defective DNA repair and an increased error rate, a phenotype consistent with known mechanisms of PARPi sensitivity.

The authors then focused on MLL–AF9-expressing cells, which had been shown to be PARPi resistant in both in vitro and in vivo assays. Consistent with its known role as a crucial downstream target gene of MLL fusion proteins, high levels of HOXA9 expression were observed in cells expressing MLL–AF9. In support of the hypothesized role of HOXA9 in PARPi resistance, Hoxa9 deficiency conferred PARPi sensitivity on both MLL–AF9-transformed (but not control E2A–PBX-transformed) mouse cells in vitro and MLL–AF9-expressing primary mouse leukaemia cells in vivo. Conversely, Hoxa9 overexpression in mouse cells transformed with PARPi-sensitive fusions (AML1–ETO and PML–RARα) rendered these cells PARPi resistant, a phenotype characterized by high expression of HR-associated genes and an efficient DDR.

Having implicated HOXA9 in PARPi resistance, the authors investigated the effects of pharmacological inhibition of HOXA9; in the absence of a direct inhibitor, they used inhibitors of the essential HOXA9 cofactor glycogen synthase kinase 3 (GSK3is). In vitro GSK3i treatment conferred PARPi sensitivity to MLL–AF9-expressing mouse cells, human cell lines and primary human samples. Furthermore, compared with PARPi or GSK3i monotherapy, combined PARPi and GSK3i treatment significantly suppressed the development of MLL–AF9-driven leukaemia in both a syngeneic model and a primary xenograft model involving the transplantation of fully developed leukaemia cells.

the use of PARPis ... as a novel alternative to traditional therapies for AML

In summary, this study provides insights into mechanisms of PARPi sensitivity in cells expressing AML fusion proteins and highlights the use of PARPis — as either single agent or combination therapy — as a potential novel alternative to traditional therapies for AML.