EZH2 inactivation in RAS-driven myeloid neoplasms hyperactivates RAS-signaling and increases MEK inhibitor sensitivity

To the Editor

Mutations modifying RAS (RAS^mut) occur frequently in myeloid neoplasms (MN) and play a key role in myeloid leukemogenesis [1, 2]. The most commonly observed RAS^mut in MN comprise aberrations in NRAS and KRAS, as well as in three genes that modulate the levels of active RAS-GTP (NF1, PTPN11, and CBL) [3]. Mechanistically, RAS^mut activate a multitude of downstream signaling cascades, with the MAPK/ERK module being considered one of the major RAS-effector pathways [4]. Consequently, pharmacologic MAPK/ERK inhibition—i.e., by MEK inhibitors—is an appealing therapeutic approach. Indeed, the development of MN in Ras^mut mice can be effectively attenuated by treatment with these substances [2]. Unfortunately, these promising results could not be translated into human MN, with disappointing results in clinical trials [5]. One potential reason is the fact that RAS^mut do not exist as solitary events within these tumors [1, 6]. The existence of co-occurring mutational and non-mutational aberrations has the potential to further influence the activating effects of RAS^mut, which ultimately aggravates or inhibits RAS^mut-driven leukemogenesis and thereby changes the dependency on activated RAS-signaling [6, 7]. Consequently, these co-occurring events might also change the sensitivity to MEK inhibitors, as recently shown for the co-existence of mutations in NRAS and TET2 [7]. Enhancer of zeste homolog 2 (EZH2) is the core component of the Polycomb Repressive Complex 2 (PRC2). It regulates the expression of a broad range of genes and thereby controls a variety of basic cellular functions [8]. In more detail, EZH2 serves as histone methyltransferase that catalyzes trimethylation of lysine 27 of histone H3 (H3K27me3), which in turn causes the transcriptional repression of its target genes. Inactivation of EZH2 (EZH2^inact)-either by mutation, deletion or a decrease in EZH2 expression-can be observed in a series of MN [8, 9]. Recently, EZH2^inact has been linked to RAS-signaling as Ezh2 deletion aggravated the development of Nras^mut-driven MN in mice. Moreover, Ezh2 deletion in Kras^mut-induced lung cancers hyperactivated Kras^mut-driven MAPK/ERK-signaling [10]. These findings suggest that the dependence on activated RAS-signaling in RAS^mut tumors might be altered by the additional occurrence of EZH2^inact.

In this study, we aimed to investigate this hypothesis in the context of myeloid leukemogenesis. By studying almost 450 primary patient specimens with chronic myelomonocytic leukemia (CMML) and acute myeloid leukemia (AML), we show that EZH2^inact and RAS^mut co-exist in MN, and that this co-occurrence is associated with a poor prognosis in affected patients. Importantly, however, we further demonstrate that concomitant EZH2^inact and RAS^mut increases the dependence on RAS-signaling and, consequently, the sensitivity to pharmacologic MEK inhibition in myeloid leukemia cells.

Initially, we were interested whether EZH2^inact and RAS^mut indeed co-exist in MN. Therefore, we re-analyzed previously published Next-Generation Sequencing (NGS) data of 260 chronic myelomonocytic leukemia (CMML) patients within the Austrian Biodatabase for CMML [11]. We chose this entity, since CMML is often driven by mutations modifying the RAS genes [1, 3, 6]. Within this cohort, 112/260 (43.1%) patients exhibited at least one RAS^mut, EZH2 was mutated in 50/260 (19.2%) cases and 32/260 (12.3%) presented with both genetic aberrations together. 32/112 (28.6%) RAS^mut patients exhibited additional EZH2 mutations, whereas 32/50 (64%) cases with EZH2 mutations presented with an additional RAS^mut (Fig. 1A; Supplementary Table 1). Importantly, the frequency of patients with EZH2 mutations was increased in cases with one or more RAS^mut (28.6% in RAS^mut, vs. 12.2% in RAS^wt; P = 0.001; Fig. 1B). From a clinical point of view, RAS^mut and EZH2 aberration co-occurrence was associated with a shortened overall survival (median 14 vs 29 months, P = 0.005; Fig. 1C). To delineate whether these findings are of relevance for other MN as well, we then performed a database retrieval of 187 AML patients via The Cancer Genome Atlas (TCGA) (see supplementary methods for details) [12]. In addition to clinical parameters, this database comprises information about mutations, gene expression and DNA copy number variations [12]. Out of the 187 patients within this cohort, 33 (17.6%) exhibited at least one RAS^mut, 25/187 (13.4%) exhibited inactivation of EZH2 and 9/187 (4.8%) presented with both genetic aberrations together. 9/33 (27.3%) RAS^mut patients exhibited with additional EZH2^inact, whereas 9/25 (36%) cases with EZH2^inact presented with an additional RAS^mut (Fig. 1A; Supplementary Table 2). Moreover, in line with our data from CMML, EZH2^inact was significantly more common in RAS^mut cases (27.3% in RAS^mut vs. 10.4% in RAS^wt; P = 0.020; Fig. 1B; EZH2^inact defined as EZH2 mutations and/or copy number losses). As in CMML, this genetic co-existence was associated with a dismal outcome (median survival 7 vs 19 months, P = 0.039, Fig. 1C). Accordingly, the mRNA expression of EZH2 was significantly decreased in AML cases carrying one or more RAS^mut (Supplementary Fig. 1). Taken together, these data indicate that RAS^mut and EZH2 aberrations indeed co-exist in human MN and that this co-occurrence seems to be associated with a poor prognosis. Hence, novel therapeutic approaches are desperately needed for these patients, particularly as RAS^mut has been described as difficult to target so far.

Fig. 1: Association between mutations modifying RAS and EZH2 aberrations in MN.

A Next-Generation Sequencing (NGS) results of 260 chronic myelomonocytic leukemia (CMML) patients studied within the Austrian Biodatabase for CMML [15] showing the distribution of mutations modifying RAS (RAS^mut; defined as mutations in KRAS and NRAS, as well as in the RAS-GTP modulators NF1, PTPN11 and CBL) and EZH2. In summary, 112/260 (43.1%) and 50/260 (19.2%) CMML patients had one or more RAS^mut or EZH2 mutation(s) (EZH2^mut), respectively. Below are the results of the database retrieval of 187 acute myeloid leukemia (AML) patients via The Cancer Genome Atlas (TCGA) [12] showing the distribution of RAS^mut and EZH2 inactivation (EZH2^inact; defined as EZH2 mutations and/or copy number losses). Every column describes one CMML or AML patient specimen. Colored fields indicate the presence of at least one mutation (for RAS^mut) or EZH2^inact, respectively. In summary, 33/187 (17.6%) and 25/187 (13.4%) AML patients had one or more RAS^mut mutation(s) or inactivation of EZH2, respectively. B Within both cohorts, EZH2 aberrations were significantly more common in patients harboring one or more RAS^mut compared to those without: 28.6%, vs. 12.2% (P = 0.001) for the CMML cohort (left), and 27.3%, vs. 10.4% (P = 0.020) for the AML cohort (right). Fisher’s exact test was employed for the statistical analysis. C Survival curves of the patients belonging to the CMML cohort (left), and the TCGA AML cohort (right). In both cohorts, RAS^mut and EZH2 aberration co-occurrence was associated with a shortened overall survival (median 14 vs 29 months and 7 vs 19 months for the CMML and AML patients, respectively). Censored events are indicated by a vertical line. A log-rank test was used for these comparisons.

Next, we investigated whether EZH2^inact influences the RAS^mut-driven MAPK/ERK activation in myeloid leukemia cells (for details on materials and methods see supplementary data). For this purpose, we chose two myeloid cell lines (HL-60 and THP-1). Both carry an activating RAS^mut, show normal EZH2 mRNA expression and lack other EZH2 aberrations (Supplementary Fig. 2 and Supplementary Table 3). We treated these cells with the two EZH2 inhibitors GSK-126 and 3-Deazaneplanocin A (DZNep), respectively. While GSK-126 is an enzymatic inhibitor, which does not affect EZH2 protein expression itself, DZNep induces EZH2 protein degradation [9]. Both drugs successfully inhibited EZH2 activity, as assessed by reduced H3K27me3 levels. Importantly, however, both inhibitors caused hyperactivation of RAS-MAPK/ERK-signaling, as evidenced by increased phosphorylation of ERK (pERK; Fig. 2A, B; Supplementary Fig. 3A, B). To exclude potential unspecific off-target effects of the EZH2 inhibitors used, we established a puromycin-selected stable short hairpin RNA (shRNA)-mediated EZH2-knockdown (EZH2-KD) after lentiviral transduction in both cell lines. Empty vector-transduced cells served as controls. Again, EZH2-KD reduced H3K27me3 levels and simultaneously increased pERK (Fig. 2C, Supplementary Fig. 4), which indicates that EZH2^inact amplifies MAPK/ERK activation in RAS^mut myeloid cells. Next, we explored whether EZH2^inact increases the sensitivity to MEK inhibitors in RAS^mut myeloid cells. Therefore, we treated HL-60 and THP-1 cells with and without EZH2-KD with the MEK inhibitor U0126. U0126 efficiently inhibited pERK in all conditions tested. Most importantly, however, the U0126-induced apoptosis was significantly increased in cells with additional EZH2-KD (Fig. 2D; Supplementary Fig. 5), which indicates that these cells are hypersensitive to pharmacologic inhibition of the MAPK/ERK pathway. These findings could be corroborated in 7-AAD/BrdU cell cycle/proliferation assays. Again, the U0126-mediated decrease in proliferation was enhanced in cells with additional EZH2-KD (Supplementary Fig. 6). We then aimed to shed more light on the mechanisms behind EZH2^inact-induced MAPK-hyperactivation in RAS^mut myeloid cells. As EZH2 regulates a multitude of cellular gene expression profiles via H3K27me3–induced transcriptional repression, we reasoned that EZH2^inact causes the upregulation of genes involved in RAS-MAPK/ERK-signaling. Such a scenario has been identified in a murine in-vivo model of Kras^mut/Ezh2-deleted lung cancer previously [10]. To test this assumption, we performed RNA-sequencing in HL-60 cells with and without EZH2-KD and performed gene set enrichment analysis (GSEA) [13, 14]. Indeed, we observed enrichment of RAS- and RAF-signaling signatures in the EZH2-KD situation (Fig. 2E). This included an extensive list of genes that activate the RAS-MAPK/ERK and other signal transduction cascades (Supplementary Table 4).

Fig. 2: EZH2 inactivation in RAS^mut myeloid cells amplifies MAPK/ERK-signaling and drives MEK inhibitor sensitivity.

The activation of the MAPK/ERK pathway was assessed by the phosphorylation of ERK (pERK) by Immunoblot in HL-60 cells (NRAS Q61L-mutated) after treatment with the EZH2 inhibitors GSK-126 (A) and DZNep (B). GSK-126 was added at a concentration of 3 µM for 7 days, DZNep at a concentration of 2 µM for 24 h. C These experiments were repeated after lentiviral shRNA-mediated EZH2 knockdown (EZH2-KD). The graphs denote the relative increase of pERK expression in the EZH2 inhibitor/KD conditions compared to controls and represent the mean ± standard deviation (SD) of at least three independent experiments. Comparisons against the control condition were performed using a one-sample t test against a reference value of 1. D HL-60 cells with and without EZH2-KD were treated with the MEK inhibitor U0126 (5 µM for 24 h). Subsequently, pERK was assessed by Immunoblot and apoptosis was measured by Annexin-V/7-AAD assay. The graphs denote the x-fold increase in apoptosis in U0126-treated cells compared to the respective vehicle-treated control situation in at least three independent experiments and represent the mean ± SD. Differences between cells with and without EZH2-KD were assessed by paired t test. E Gene set enrichment analysis (GSEA) demonstrating that signatures associated with RAS- and RAF-signaling are enriched within the EZH2-KD situation. All signatures displayed exhibited a false discovery rate of below 25%. NES, normalized enrichment score.

Finally, we re-analyzed a previously published ChIP-seq dataset of AML cells with EZH2 loss [9] via the NCBI Gene Expression Omnibus (GSE61785). By focusing on genes with a well-described activator function of RAS-signaling on the one hand, and a significant upregulation in our RNA-seq data of EZH-KD cells on the other hand, we were able to demonstrate decreased H3K27me3 signals in the condition with EZH2 loss (Supplementary Fig. 7). These data suggest that the upregulation of these genes in EZH2^inact cells could indeed be mediated through modification of H3K27me3 within their promoter and/or adjacent genomic regions.

In conclusion, we demonstrate that mutations within genes modifying RAS frequently co-occur with inactivation of the epigenetic modifier EZH2 in MN, and that this co-existence is linked to a dismal outcome in affected patients. We further demonstrate that inactivation of EZH2 amplifies the activation of RAS-MAPK/ERK-signaling in myeloid cells carrying RAS-modifying mutations. Most importantly, however, we present preclinical data showing that the co-existence of EZH2 inactivation and RAS-modifying mutations might confer increased sensitivity to MEK inhibitors, thereby providing a potential novel therapeutic rationale for these difficult to treat patients.