Therapy Response and Outcome Explained by Leukemia Cell of Origin

Cancer cells show considerable phenotypic and functional heterogeneity that contributes to therapeutic response and clinical outcome (1). Although genetic and epigenetic alterations are well established to determine cancer cell behavior, there is increasing evidence to support the notion that cancers of distinct subtypes may arise from different cells of origin within a tissue (2). Conceptually, cancer-initiating cells can arise from normal stem cells by gaining proliferative potential or from committed progenitor cells by gaining self-renewal capacities. These two distinct modes of oncogenic transformation have been described in diverse cancers including solid tumors and acute leukemias (3–5). The inherited transcriptional and epigenetic programs from the original target cells may contribute to the biological properties of the fully developed tumor cells, but the underlying mechanisms have remained elusive. Thus, a long-standing question in cancer biology is whether and how differences in cell of origin influence therapy response and clinical outcome in otherwise molecularly defined cancer subtypes.

In acute myeloid leukemia (AML), both normal hematopoietic stem cells (HSC) and committed myeloid progenitors including granulocyte-macrophage progenitors (GMP) can be transformed by mixed lineage leukemia (MLL)–rearranged fusion proteins such as MLL–AF9 and MLL–ENL (3–5). Although leukemia stem cells derived from HSCs or GMPs had a similar immunophenotype, they showed differences in gene-expression profiles inherited from the cells of origin, including a high level of EVI1 expression in HSC-derived AML (3, 4). Moreover, MLL-rearranged leukemias derived from HSCs were more likely to initiate AML in mice and more resistant to chemotherapy than GMP-derived leukemias (4). On the other hand, heterogeneity in therapy response and clinical outcome has been observed in patients with MLL-rearranged AML, raising the possibility that distinct cell of origin may contribute to the observed clinical heterogeneity. Of note, aberrant expression of EVI1 is a hallmark of patients with AML with 3q26 alterations, and it can also be seen in patients without cytogenetic abnormalities, with high EVI1 expression being associated with poor prognosis, suggesting a broader role for EVI1 in driving malignant phenotypes in AML. Despite these findings, it remains largely unknown how distinct cell of origin influences EVI1 function in AML and how EVI1 contributes to leukemia pathobiology, therapy response, and clinical outcome.

In this issue of Cancer Discovery, Cai and colleagues describe a new mechanism linking cell of origin, the oncogenic transcription factor EVI1, p53-mediated apoptotic priming, and therapeutic susceptibility in MLL-rearranged AML (6). Building on previous findings that MLL–AF9 AML derived from HSCs and GMPs had a similar immunophenotype, but distinct biological properties (4), the authors first investigated whether the cell of origin of leukemic transformation may influence sensitivity to inhibition of LSD1 (or KDM1A), a histone H3-Lys4 demethylase involved in p53-mediated DNA damage responses (7). Although MLL–AF9-induced leukemias from mouse Lin⁻Sca1⁺Kit⁺ (LSK) HSCs or GMPs had no difference in baseline growth kinetics, they exhibited differential sensitivity to LSD1 inhibitors by more than 100-fold differences in IC₅₀ values. HSC-derived MLL–AF9 leukemias showed a markedly higher expression of EVI1 than GMP-derived leukemias, which was recapitulated in human HSC- and GMP-derived MLL–AF9 leukemias. Differential EVI1 expression was also seen in normal HSCs and GMPs in mice and humans, suggesting that the observed difference in EVI1 expression in leukemia cells was intrinsic to the cells of origin. Importantly, the EVI1 gene was epigenetically repressed by Polycomb protein-mediated H3-Lys27 trimethylation (H3K27me3) in normal GMPs and GMP-derived leukemias, but not HSCs, suggesting that the cell of origin influences the biological properties of the developed leukemias through preexisting transcriptional and epigenetic programs. Consistent with this notion, shRNA-mediated EVI1 knockdown sensitized HSC-derived leukemias to a variety of cytotoxic and proapoptotic stimuli, such as doxorubicin, ionizing radiation, the BH3 mimetic venetoclax, and LSD1 inhibitors. These results demonstrate that the cell of origin of leukemia initiation determines sensitivity to LSD1 inhibition and chemotherapy through differential EVI1 expression (Fig. 1).

To establish the physiologic basis for the differential sensitivity to LSD1 inhibition, Cai and colleagues measured cell apoptosis and mitochondrial transmembrane potential in LSD1 inhibitor–treated MLL–AF9 leukemia cells. Significantly increased apoptosis and loss of mitochondrial potential were observed in GMP-derived but not HSC-derived leukemias, consistent with differential apoptotic priming in leukemias from distinct cells of origin. Similarly, profiling of BH3 domains of proapoptotic proteins showed increased apoptotic priming in murine and human GMP-derived relative to HSC-derived leukemias, whereas overexpression of the antiapoptotic protein BCL2 in GMP-derived leukemias induced resistance to LSD1 inhibition.

Having established differential apoptotic priming as the basis for susceptibility to LSD1 inhibition, the authors next explored the underlying mechanism. In a chromatin regulator-focused shRNA screen in HSC-derived MLL–AF9 leukemia, p53 emerged as one of the top targets mediating resistance to LSD1 inhibition, together with other p53 pathway components including Gadd45a, Csnk2a, Smyd2, and Bmi1. These findings were corroborated by RNA-sequencing analyses showing that p53 target genes were more highly expressed in GMP- than HSC-derived MLL–AF9 leukemias, indicating that GMP-derived leukemias have higher basal p53 activity. Importantly, analysis of The Cancer Genome Atlas (TCGA) AML samples based on EVI1 expression also recapitulated the differential p53-associated gene-expression profiles observed in murine leukemias. TCGA samples consist of a variety of AML subtypes; thus, these findings indicate that the p53-mediated program is directly modulated by EVI1 irrespective of MLL gene rearrangement, consistent with a broader role for EVI1 in conferring adverse outcomes in AML. Along these lines, the authors made interesting discoveries that the level of p53 protein, but not mRNA, was significantly higher in murine and human GMP-derived than HSC-derived leukemias, accompanied by enhanced induction of p53 targets and DNA damage responses upon LSD1 inhibition. Of note, modulating EVI1 expression influenced p53 protein abundance in a proteasome-independent manner. Taken together, these findings support a model that EVI1 functions as an upstream modulator of p53 protein abundance and that leukemia cell of origin influences p53 function through EVI1-dependent mechanisms (Fig. 1).

Given that HSC-derived and/or EVI1^hi AML cells display more aggressive properties, higher rates of relapse, and resistance to chemotherapy and LSD1 inhibition, the next key question is whether modulating p53 activity and apoptotic priming would sensitize this group of high-risk AML to genotoxic stresses and/or LSD1 inhibitors. To this end, the authors noted that p53 knockdown or knockout in GMP-derived leukemias induced resistance to LSD1 inhibitors and abrogated the proapoptotic, but not the prodifferentiation, effects of LSD1 inhibition. More importantly, combining venetoclax, a BH3 mimetic and small-molecule inhibitor of BCL2, and LSD1 inhibition enhanced apoptotic priming and antileukemic activity in HSC-derived MLL–AF9 leukemias in animal models. Hence, by systematic analysis of leukemia cells derived from distinct cells of origin, a new mechanism-based therapeutic approach was proposed by leveraging the intrinsic differences in apoptotic priming to enhance the antileukemic activity of LSD1 inhibition in EVI1^hi and/or HSC-derived leukemias.

LSD1 inhibitors are currently being investigated in phase I or I/II clinical trials to treat patients with myeloid malignancies. Most of these early-phase studies were completed in small numbers of patients with relapsed or refractory leukemia; thus, there are no clear predictive features of patients with AML to suggest who might respond to LSD1 inhibition. In myeloproliferative neoplasms including myelofibrosis, LSD1 inhibition by IMG-7289 showed promise in early-phase studies by affecting the proinflammatory cytokines. Given that LSD1 inhibitors are generally well tolerated, the use of LSD1 inhibitors alone or in combination with chemotherapeutic agents is particularly attractive for patients who are otherwise ineligible for intensive therapy. The findings of Cai and colleagues suggest that patients with EVI1^hi expression may define a subgroup of treatment-naïve patients amenable to LSD1 inhibition in combination with venetoclax or standard-of-care treatment. In addition, the combination of venetoclax with hypomethylating agents or low-dose cytarabine has recently been FDA-approved for the treatment of elderly patients with AML (8). This work may predict that patients with EVI1^hi expression, who have very low rates of response to conventional chemotherapy, may benefit from these venetoclax combination therapies. Finally, such patients may also benefit from strategies to reactivate p53, such as MDM2 inhibitors currently in clinical trials for AML (https://clinicaltrials.gov/ct2/show/NCT04029688).

Therefore, this study represents an important step toward a better understanding of the functional and mechanistic roles of cell of origin in determining cancer heterogeneity. More importantly, elucidation of the underlying mechanisms provides opportunities for developing new approaches to overcome drug resistance by therapeutic combinations designed to precisely target cell of origin–dependent programs. The concepts and strategies developed in this study may be extended to other AML subtypes or other cancers originated from distinct cell types. In future studies, interrogating the mechanisms by which EVI1 regulates p53 protein abundance may provide new insights into this key cell of origin–dependent pathway and reveal actionable targets for modulating EVI1 and/or p53 activity to improve therapy response. The analysis of the cell of origin–EVI1–p53 axis in other AML subtypes with different genetic lesions or other tumor types may uncover a generalizable role for EVI1 in p53 wild-type cancers. In addition, this study demonstrates the in vivo efficacy of combinatorial therapy using venetoclax and LSD1 inhibitor in mouse models, thus raising the prospect of developing improved strategies to circumvent drug resistance in high-risk, chemoresistant leukemias. Finally, a better understanding of the mechanisms by which genetic mutations found in clonal hematopoiesis (e.g., DNMT3A or TET2 mutations) exploit cell of origin–dependent programs to drive clonal expansion and leukemic transformation will pave the way for developing more accurate risk stratification and preventive strategies (9, 10). Hence, elucidating the molecular pathways underlying leukemia cell of origin will not only enhance our understanding of cancer pathophysiology, but also enable the development of improved diagnostic, prognostic, and therapeutic approaches for the heterogeneous blood cancers.

No potential conflicts of interest were disclosed.