In AML, MLL fusions do not co-occur readily with cohesin mutations, but instead co-occur with Ras family mutations (5, 6, 8–10). In contrast, cohesin mutations tend to co-occur with NPM1 mutations but not with Ras family mutations (7, 12, 13, 18). Further, NPM1 mutations are not found in MLL-fusion AMLs (10). Based on these different mutational landscapes, we hypothesized that the addition of cohesin haploinsufficiency to MLL-AF9-expressing HSPCs would be detrimental. Contrary to this hypothesis, we found that both a complete loss of Stag2 and Smc3 haploinsufficiency were well tolerated by MLL-AF9-expressing cells. While Stag2 loss increased the self-renewal capacity of MLL-AF9-expressing cells at later passages (Fig. 1A), minimal effects on self-renewal were observed upon excision of one allele of Smc3 (Fig. 2A). Thus, despite their predicted synthetic lethality and differing co-mutational spectra in AML, MLL-AF9 and cohesin mutations are permissible for HSPC growth and self-renewal.
Interestingly, both cohesin haploinsufficient and MLL-AF9 HSPCs and AMLs have elevated expression of HOXA cluster genes, particularly HOXA7 and 9, which are critical for enhanced self-renewal (5, 28, 29). HOXA9 is also important for AML transformation and maintenance (28, 30, 31). Knockdown of HOXA9 reduces proliferation in primary human MLL-fusion AMLs through cell cycle arrest, increased differentiation, and increased apoptosis (28). Over-expression of HOXA9 in hematopoietic stem and progenitor cells results in stem cell expansion and leukemic development, suggesting that both MLL-AF9 and cohesin mutations co-opt key regulators of HOXA gene expression (28).
In support of this, both MLL-AF9 and cohesin-mutated cells influence gene expression through interactions with the DOT1L and PRC2 histone methyltransferase complexes, respectively (5, 29, 32). MLL-AF9 recruits the DOT1L methyltransferase to target genes such as HOXA9 (5). DOT1L deposits the activating mark H3K79me2, which drives gene expression. In hematopoietic cells, HOXA7/9 are repressed by the PRC2 complex, which deposits the repressing mark H3K27me3 at promoter regions. The cohesin subunits Rad21 and Smc3 have been shown to interact with components of the PRC2 complex, with cohesin loss resulting in decreased PRC2 recruitment to murine Hoxa7/9 promoters (29). This allows for Dot1L recruitment and enhanced gene activation. Interestingly, DOT1L is critical for both the initiation and maintenance of MLL-AF9 AML, and small molecule inhibitors of DOT1L are particularly effective against mouse and human MLL-rearranged leukemias in vitro and in xenograft models (31–34). Furthermore, the abnormal self-renewal observed in murine HSPCs depleted of cohesin were restored by Dot1l inhibition (35). Collectively, these data suggest that DOT1L inhibitors may be particularly effective against MLL-rearranged or cohesin-mutated AMLs.
Because MLL-AF9 is a strong driver of HOXA gene expression and self-renewal and, unlike Rad21-depletion, up-regulates the HOXA9 co-factor MEIS1 (28, 29), there is likely little selective pressure for the acquisition of cohesin mutations in MLL-AF9-expressing AMLs, potentially explaining why they are not commonly observed together. Interestingly, NPM1 mutations are more commonly observed with cohesin mutations and, like MLL-AF9, NPM1 also drives HoxA expression. This suggests that additional mechanisms beyond HOXA-induced self-renewal may contribute to NPM1-driven leukemias, and such mechanisms are supported by cohesin haploinsufficiency. We have shown one such mechanism is through Rac-regulated growth and apoptosis (4). While it is important to consider that differences may exist in in vitro vs. in vivo assays as well as between mice and humans, our experiments suggest that the absence of co-occurrence, while often interpreted as synthetic lethality, may instead reflect significant overlap in the molecular mechanisms driving the observed disease biology. Intriguingly, both cohesin mutant and MLL-AF9 AMLs are sensitive to DOT1L inhibitors. Thus, a lack of co-occurrence may point to a common, targetable mechanism behind two initially dissimilar appearing leukemias, underlining the importance of genetically testing mutual exclusivity.
Limitations
Assays were performed on HSPCs isolated from a single Stag2fl/fl or Smc3fl/+ mouse. Each Stag2fl/fl and Smc3fl/+ parental cell line, once infected with MLL-AF9 and treated with 4-OHT, was followed over serial re-passaging, and while some assays were repeated (serial replating), others were not (qPCR, genotyping, western blot, or GFP + assays). Thus, differences may exist between individual mice or with long-term culture, which would not be revealed by our assays. However, the consistency between our results in Stag2−/− and Smc3−/+ cells and between passages does not support this. In addition, Fig. 2B and C show that Smc3 protein levels did slightly increase between passage 1 and 4, suggesting a selection for cells with higher Smc3 expression may occur in the presence of MLL-AF9. However, the level of Smc3 expression did not rise above that of an Smc3−/+ control. Further experiments would need to be performed to determine if higher Smc3 expression is selected for in the presence of MLL-AF9. Additionally, as mentioned in our discussion section, all of our assays are performed in vitro, and different results may be observed in vivo. Lastly, our experiments focus primarily on leukemogenesis. Thus, loss of Stag2 or Smc3 may contribute to aspects of leukemia maintenance that were not addressed in this manuscript.