Neomorphic missense mutations in histone H3 genes (most notably K27M and K36M) are frequent in solid cancers such as diffuse intrinsic pontine gliomas (DIPG), supratentorial glioblastoma multiforme (GBM) or chondroblastomas. Despite high incidences of these mutations and their profound impact on global histone H3 lysine methylation, their relevance has not been studied in hematologic cancers.Here, we report the first identification of histone H3 mutations in AML. We find that two of 415 AML patients in the Leucegene cohort and one of 200 patients in the TCGA cohort carry either K27M (n=2) or K27I (n=1) mutations in one of twelve canonical H3 genes. No K36M or other lysine-to-methionine mutations on any of the H3 genes, and no K27M mutation on any of the replication-independent H3.3 variant genes were detected in the same cohorts. H3K27 mutations invariably occurred in the context of either truncating RUNX1 mutations or t(8;21)(q22;q22), suggesting an association with altered RUNX1 function (p < 0.003; Fisher exact test). Characterization of two of the identified H3K27 mutant patient specimens revealed that both samples contained major sub-clones with markedly reduced H3K27me2/3 levels. The H3K27 mutants but not co-occurring mutant ASXL2 or USP7 alleles were only detectable in the H3K27me2/3-low sub-clones, demonstrating that the H3K27 mutations had the most profound impact on H3K27me2/3 in these patients. Moreover, although a heterozygous TET2S1203R mutation was equally present in both H3K27me2/3-high and low sub-fractions, a homozygous RUNX1L98Sfs*22 mutation was restricted to the H3K27me2/3-low sub-clone, further strengthening the possibility that impaired RUNX1 function and reduced H3K27me2/3 associate with each other.We tested the possibility that H3K27 mutants may specifically collaborate with the product of the t(8;21) translocation (AML1-ETO-9a ; AE9a) in transduced mouse hematopoietic stem and progenitor cells (HSPCs). Selectivity of this collaboration was assessed using the combination of the H3K27 mutant gene with the MLL-AF9 fusion product. Whereas expression of H3 WT resulted in a mild competitive proliferation disadvantage in both AE9a and MLL-AF9 cells, expression of H3.3K27I/M conferred a proliferative advantage in AE9a cells in three of four H3K27I -infected and in all H3K27M- infected cultures. This was not observed in MLL-AF9 cells, where H3K27I/M expressing cells were progressively outcompeted in the same culture period. H3K27I/M -infected AE9a cells exhibited significant reductions in H3K27me2/3 and elevated levels of H3K27 acetylation (H3K27ac) compared to H3WT -infected control cells despite undiminished Ezh1, Ezh2 and Suz12 protein levels. In the context of AE9a driven AML in vivo, expression of H3K27M caused a striking disease acceleration leading to all mice reaching end stage disease with a median of 88 days, in contrast to recipients of AE9a ; empty vector or AE9a ; H3WT cells, of which none or only two of nine transplanted mice succumbed to AML within 330 days, respectively (p<0.0001, Log-rank test). Seven of nine AE9a ; H3.3K27I transplanted mice developed AML with a median latency of 208 days (p=0.0004, Log-rank test, compared to controls). Importantly, combined expression of H3K27I or H3K27M with MLL-AF9 did not change disease latencies compared to control groups, indicating selectivity.We next tested the influence of the H3K27M mutation on the differentiation and proliferation of human cord blood CD34+ HSPCs ex vivo . Relative to controls, H3.3K27M expression led to an 8-fold net expansion of a phenotypically primitive CD34+/CD45RA-/CD90+ cell population, which surprisingly lacked expression of the CD133 stem cell marker. Interestingly, granulocytic/granulo-monocytic colony-forming cells (CFU-G/GM) but not multipotent colony-forming cells (granulo/erythroid/monocytic/megakaryocytic; CFU-GEMM) were expanded in 14d cultures initiated with H3K27M transduced CD34+ CB cells when compared to controls. Xenografts of transduced CD34+ CB cells indicated a marginal loss in in vivo reconstitution activity by K27M-transduced cells, mostly at 6 and 12-week time-points.Taken together, our results represent the first description of neomorphic H3K27M/I mutations in AML. We demonstrate that these mutations have a profound impact on H3K27me2/3 and provide evidence that they selectively collaborate with impaired RUNX1 function. DisclosuresNo relevant conflicts of interest to declare.