Oncogenic Drivers and Development

Genomic rearrangement can generate novel fusion oncogenes. These fusion genes either dysregulate protein expression or produce chimeric proteins with abnormal functionality. Interestingly, some fusion oncogenes are exclusively acquired in utero, and the disease is limited to childhood (1–3). These fusion oncogenes are associated with diverse cancer phenotypes including pediatric acute myeloid leukemia (AML; ref. 4). Approximately 5% of pediatric AML cases are characterized by a cryptic inversion of chromosome 16 (inv(16)(p13.3q24.3)) that confers a poor prognosis (5, 6). This chromosome 16 inversion produces a chimeric protein where the nuclear corepressor CBFA2T3 (also known as ETO2) is fused with the zinc finger DNA-binding transcription factor GLIS2. Lopez and colleagues show that the phenotype of CBFA2T3–GLIS2-associated acute leukemia varies with age (7). In infants and young children, it is more likely to present as an acute megakaryoblastic leukemia (AMKL; mean age of presentation 1.6 years of age), whereas in older children, it presents as an AML (mean age 8.5 years). Rarely, it is associated with both AMKL and AML phenotypes in the same patient.

To understand the cellular and molecular basis of this phenotype variation, the authors created a novel doxycycline-inducible ETO2–GLIS2 (iEG)-expressing transgenic mouse model. EG expression produced transplantable megakaryoblastic and myeloid leukemias, thus phenocopying the human disease. The authors then hypothesized that the developmental stage of the hematopoietic cell transformed by iEG helps to determine the phenotype of the resulting leukemia. To test this hypothesis, they transplanted fetal liver (FL) and adult bone marrow (ABM) cells into primary recipients. Importantly, iEG-transformed FL cells primarily gave rise to CD41⁺ AMKL, whereas iEG-transformed ABM cells primarily gave rise to a Gr1⁺/CD11b⁺ AML phenotype. Furthermore, FL-derived leukemia was markedly more aggressive with a shorter latency than ABM-derived disease. Thus, the developmental stage of the transformed cells modifies the phenotype of iEG-driven leukemia.

The authors then further dissected whether the differentiation stage of the iEG-transformed cells altered the lineage phenotype. At the apex of the hemopoietic cellular hierarchy reside hemopoietic stem cells (HSC) that give rise to all blood cells. Downstream of HSCs are a series of lineage-biased multipotential progenitor (MPP) cells known as MPP2–MPP4 (8). In vitro, MPP2 cells give rise predominantly to erythroid–megakaryocytic cells; MPP3 cells to granulocyte–monocyte myeloid cells; and finally MPP4 cells to lymphoid cells. FL iEG-transduced HSCs and MPP2 cells gave rise to cells expressing CD41⁺ (a megakaryocyte marker), whereas iEG-transduced FL MPP3 cells gave rise to both CD41⁺ and Gr1⁺ (granulocyte marker) cells. In ABM, the situation was a little different. Like FL, ABM HSCs gave rise to mainly CD41⁺ cells both when tested as a population and as single cells. ABM MPP2, MPP3, and MMP4 cells all gave rise to mixed CD41⁺ and Gr1⁺ cells with progressively more Gr1⁺ cells in MPP3 and MPP4 compared with MPP2 cells. Collectively, this suggests that differentiation stage of the transformed cells also modifies the phenotype of the cells produced.

To begin to understand the mechanism of phenotype differences, the authors performed single-cell RNA sequencing in FL and ABM HSCs and MPP4 cells. They also performed Assay for Transposase-Accessible Chromatin using sequencing on fetal cell populations after EG induction in vitro. As expected, prior to EG expression the baseline transcriptional programs differed in development between FL and ABM; and during differentiation stage, between HSCs and MPP. There was increased expression of the megakaryocyte transcription factor Gata1 and the stem cell transcription factor Erg in FL cells and the myeloid transcription factor Cebpa in ABM. These data suggest a hypothesis that immortalization of FL cells, with higher Gata1 expression, skews the leukemia phenotype to AMKL, whereas immortalization of ABM cells, with higher Cebpa expression, skews the leukemia phenotype to AML. Supportive of this suggestion, patient CBFA2T3–GLIS2 AMKL cells have higher GATA1 expression and patient CBFA2T3–GLIS2 AML cells have higher CEBPA levels. Direct experimental support of this hypothesis was provided by the observation that ectopic expression of GATA1 and EG in ABM cells promoted cells with a megakaryocyte phenotype rather than a myeloid phenotype. Finally, analysis of mRNA showed that induction of EG expression resulted in more pronounced expression of Erg in HSCs compared with MPP cells consistent with the notion that EG promoted a stem-cell program, and this is easier to induce in cells with a preexisting stem-cell transcriptional program (see Fig. 1).

Taken together, this suggests that both the developmental- and differentiation-stage ground state of transcriptional factor–driven transcriptional networks cooperate with EG to determine the leukemic phenotype. The authors also present data supportive of this concept with another fusion oncogene, MLL–AF9. Others have also recently published findings supportive of ontogeny-dependent leukemic phenotype with NUP98–HOXA9 (9) and KMT2A–ENL (10) fusion oncogenes.

Finally, the authors also showed that sustained expression of the chimeric EG protein was required to maintain the transplantable leukemia. Withdrawal of doxycycline led to differentiation of leukemic blasts, and mice survived. This suggests that CBFA2T3–GLIS2, and the pathway downstream of it, could be a therapeutic target.

These experiments raise the question why fetal-derived and young adult stem progenitor cells are exquisitely sensitive to EG transformation. Why does EG not transform cells in older adults? Similarly, why are gene mutations that transform cells in older adults not detected in infants and young children? Is it due to cancer cell–autonomous differences (e.g., differences in transcriptional ground states between cells in children/young adults, or differences in DNA-repair mechanisms during aging?). Alternatively, are there differences in the microenvironment or the inflammatory/immune state of the younger versus older host? Given the power of gene editing and the variety of experimental models now available, these questions can be directly studied.

No potential conflicts of interest were disclosed.