Loss of H3K27 Methylation Identifies Poor Outcome in Adult-Onset Acute Myeloid Leukemia

Background: Acute myeloid leukemia (AML) is an epigenetically heterogeneous disease. The intensity of treatment is currently guided by cytogenetic and molecular genetic risk classifications; however these incompletely predict outcomes, requiring additional information for more accurate predictions. We aimed to identify potential prognostic implications of epigenetic modification of histone proteins, with a focus of H3K27 methylation in relation to mutations in chromatin, splicing and transcriptional regulators.

Material and methods: Histone methylation mark expressions were evaluated in a cohort of 241 AML bone marrow (BM) and peripheral blood (PB) samples from patients admitted at the MD Anderson Cancer Center relative to their expression in CD34⁺ BM derived samples from healthy donors. Simultaneous analysis of 230 proteins was performed using the reverse phase protein array – a high-throughput, quantitative proteomic platform that enables identification of aberrant expressed proteins and the pathways they act in. Additional mutational analysis was performed on 65 BM samples.

Results: H3K27Me3 was significantly lower in both BM and PB leukemic-derived samples compared to their expression in normal BM (figure 1A). A greater loss of H3K27Me3 associated with increased proliferative potential and shorter overall survival (OS) in the whole patient population (n=241, HR=0.64, 95% CI=0.47-0.87, p<0.01), as well as in subsets, e.g. cytogenetically normal AML (n=110, HR=0.62, 95% CI=0.40-0.97, p=0.03). To study the prognostic impact of H3K27Me3 in the context of cytogenetic aberrations and mutations, multivariate cox regression analysis was performed which identified H3K27Me3 level as an independent favorable prognostic factor in all (HR=0.74, 95%CI=0.57-0.95, p=0.02) as well as in P53 mutated AML (n=54, HR=0.48, 95%CI=0.26-0.87, p=0.02).

A total of 78 AML patients had molecular data available for the major methylation affecting genes, i.e. IDH1, IDH2, DNMT3A and TET2. The level of H3K27Me3 was not prognostic in patients without any DNA methylation affecting mutation present, but patients with at least one mutation in any of these had better outcome when H3K27Me3 levels were high (highest tertile, figure 1A) compared to those with lower levels (median OS 7.1 vs. 24.1 months, HR=0.42, 95% CI=0.21-0.83, p=0.01, figure 1B).

Mutations in U2AF1 and SRSF2 affect the spliceosome and are frequently found in antecedent hematological disorders (AHD), as well as are mutations in chromatin regulating genes ASXL1 and BCOR. We observed significant decreased H3K27Me3 in patients with these mutations corresponding with observed lower levels of H3K27Me3 in patients with AHD than those without (p=0.035). BCOR, SRSF2, U2AF1 and ASXL1 mutations confer poor prognosis in myeloid malignancies, however, in our cohort of 65 sequenced AML patients; not individual or a combination of these mutations were independent prognostic factors, but the degree of H3K27Me3 in these patients (HR= 0.49, 95% CI=0.25-0.95, p=0.03).

To recognize dysregulated pathways in AML patients with the identified loss of H3K27Me3, we examined correlations of H3K27Me3 with the other 229 proteins on the array. H3K27Me3 is catalyzed by the polycomb group protein EZH2 and is linked to transcriptional repression via the formation of heterochromatin regions. To identify upregulated proteins and pathways upon the loss of H3K27Me3, we focused on significant negatively correlated proteins with H3K27Me3 leading us to the identification of 20 total and 6 phospho-proteins that showed increased expression upon decreased H3K27Me3. Functional enrichment analysis of this protein set revealed an upregulated anti-apoptotic phenotype.

Conclusion: This study shows that proteomic profiling of epigenetic modifications on the histone level have clinical implications in AML and support the idea that epigenetic patterns contribute to a more accurate picture of the leukemic state complementing cytogenetic and molecular genetic subgrouping.

Figure 1. A) Lower H3K27Me3 in BM and PB derived AML samples compared to normal CD34+. **** represents p<0.0001, ns = not significant. B) Overall survival probability in AML patients with any DNA methylation affecting mutation (i.e. IDH1/2, DNMT3A, TET2, n=53) according to H3K27Me3 low (blue) and high (orange) status.