Distinct gene expression pattern of RUNX1 mutations coordinated by target repression and promoter hypermethylation in acute myeloid leukemia

Abstract

Runt-related transcription factor 1 (RUNX1) is an essential regulator of normal hematopoiesis. Its dysfunction, caused by either fusions or mutations, is frequently reported in acute myeloid leukemia (AML). However, RUNX1 mutations have been largely under-explored compared with RUNX1 fusions mainly due to their elusive genetic characteristics. Here, based on 1741 patients with AML, we report a unique expression pattern associated with RUNX1 mutations in AML. This expression pattern was coordinated by target repression and promoter hypermethylation. We first reanalyzed a joint AML cohort that consisted of three public cohorts and found that RUNX1 mutations were mainly distributed in the Runt domain and almost mutually exclusive with NPM1 mutations. Then, based on RNA-seq data from The Cancer Genome Atlas AML cohort, we developed a 300-gene signature that significantly distinguished the patients with RUNX1 mutations from those with other AML subtypes. Furthermore, we explored the mechanisms underlying this signature from the transcriptional and epigenetic levels. Using chromatin immunoprecipitation sequencing data, we found that RUNX1 target genes tended to be repressed in patients with RUNX1 mutations. Through the integration of DNA methylation array data, we illustrated that hypermethylation on the promoter regions of RUNX1-regulated genes also contributed to dysregulation in RUNX1-mutated AML. This study revealed the distinct gene expression pattern of RUNX1 mutations and the underlying mechanisms in AML development.

Keywords

RUNX1 / gene mutation / acute myeloid leukemia / transcriptional repression / DNA methylation

1 Introduction

Hematopoietic transcription factors (TFs) are essential regulators of hematopoiesis, and their genetic alterations, including mutations and chromosomal translocations, cause the differentiation blockage observed in many types of leukemia. These TFs exert their regulatory activities via downstream targets controlled in coordination by epigenetic regulations, including DNA methylation [1]. Given their essential roles in controlling hematopoiesis and their genetic abnormalities in the development of leukemia, elucidating the molecular classification of hematopoietic TF-associated leukemia with the integration of the ensemble of genome-wide targets and DNA methylation status is crucial.

Runt-related TF 1 (RUNX1, also named AML1, CBFA2, and PEBP2aB) encodes a TF that plays pivotal roles in normal hematopoiesis, including definitive hematopoietic stem cell formation and megakaryocyte maturation and granulocytic differentiation [2,3]. It can form a heterodimeric complex with the core-binding factor b subunit (CBFB) to exert its transactivation function on target genes [4]. RUNX1 abnormalities include fusions caused by chromosomal translocations and acquired mutations, among which point mutations in RUNX1 are frequently detected in hematologic malignancies, such as myelodysplastic syndromes, particularly acute myeloid leukemia (AML). The intensive investigations of the genome-wide binding maps of RUNX1-associated fusion proteins, such as RUNX1–RUNX1T1 generated by t(8;21), ETV6–RUNX1 by t(12;21), and CBFB–MYH11 by inv(16), have considerably improved our understanding of the molecular signatures and mechanisms of RUNX1 fusion-mediated leukemogenesis. In contrast to favorable prognosis-associated RUNX1 fusions, RUNX1 mutations usually portend poor outcomes and chemotherapy resistance in patients with AML [5,6], suggesting that the molecular mechanisms underlying RUNX1 mutations are distinct from those underlying fusion forms. Genetic and clinical analyses show that RUNX1 mutations rarely coexist with NPM1 and CEBPA mutations [7]. Given their independent prognosis, RUNX1 mutations have been adopted as a new but provisional entity in the latest (2016) revision of the World Health Organization (WHO) AML classification to stratify patients for risk-adapted treatments [8]. However, the molecular signatures associated with RUNX1 mutations are limited, and few common signatures have been identified to depict the mechanisms of RUNX1 mutations in AML.

In this study, we identified a 300-gene signature for RUNX1-mutated AML that was highly associated with genes directly repressed by RUNX1 and genes with the hypermethylated status. We first investigated the spectrum of RUNX1 mutations in a large combined AML cohort. Next, we performed differential gene expression analysis and developed a gene signature that could distinguish RUNX1-mutated AML from wild-type cases. Moreover, we integrated chromatin immunoprecipitation sequencing (ChIP-seq) data to evaluate the effects of RUNX1 mutations on RUNX1 target genes and analyzed the association of DNA hypermethylation with RUNX1 mutations. We collectively demonstrated that RUNX1-mutated AML possessed a distinct gene expression pattern coordinated by target repression and promoter hypermethylation. Our study provides insights into the pathogenic mechanisms caused by RUNX1 mutations and suggests promising targets for prognostic and therapeutic approaches.

2 Materials and methods

2.1 Data resources

The data of patients with AML with mutation information and clinical features were collected from three public cohorts containing 1741 qualified AML cases: TCGA (LAML project, 144 cases), EGA (European Genome-phenome Archive, EGAS00001000275, 1328 cases), and GEO (Gene Expression Omnibus, GSE23312, 269 cases). For the AML cohort from TCGA, RNA-seq, gene mutation, clinical, and DNA methylation data were downloaded from the GDC Data Portal website. For the AML cohort from EGA, gene mutation and clinical data were downloaded from the supplementary information of Papaemmanuil’s work [9] and website of Gerstung’s work [10]. The cohort from EGA contained de novo AML cases and secondary or therapy-related AML cases given that the original source (n = 1540) of the EGA cohort contained 8.3% (n = 128) secondary or therapy-related AML cases [9].

In the analysis of RUNX1 mutation distribution, we used all the 1741 cases from the three cohorts (TCGA, GSE23312, and EGAS00001000275). In the analysis of gene expression, we used 413 de novo AML cases with gene expression data from the TCGA AML cohort and GEO cohort. In the analysis of DNA methylation, we used 113 de novo AML cases with gene expression and DNA methylation data from the TCGA AML cohort. In the analysis of prognostic impact, we used 135 de novo non-APL cases with gene expression data and detailed survival information from TCGA, in which the treatment protocols had a low impact on prognosis (Tables S2–S4).

2.2 Data integration and differential gene expression analysis

For the TCGA AML data, the RNA-seq files of 151 cases (Level 3, HTSeq counts) were downloaded. The whole-exon sequencing files of 149 cases (Level 3, somatic mutations) were downloaded. Mutation calls were retained if detected by at least two of the following software packages: MuTect, MuSE, SomaticSniper, and VarScan. Our analyses were based on the gene expression and mutation data produced using the updated reference genome (GRCh38/hg38). For the cohort from GEO (GSE23312), the microarray data were filtered out to remove batch effects and failed samples (with>10% failed probes) before hierarchical clustering.

Differentially expressed genes (DEGs) were identified by using RNA-seq read count data and the R package DESeq2 (version 1.22.1) [11]. We identified DEGs with a false discovery rate of less than 0.05 and absolute LFC (log₂ fold change) of more than 0.5849 (FC>1.5 or<2/3).

2.3 Sample clustering

Hierarchical clustering based on the gene expression data of 413 patients (144 from TCGA and 269 from GSE23312) was conducted by using the pheatmap package (version 1.0.12). The variance-stabilizing transformation provided by DESeq2 was used to normalize RNA-seq count data for calculating sample distances. The complete linkage method and Euclidean distance were used to measure the distances. Principal components analysis (PCA) was conducted with the plotPCA function in DESeq2.

2.4 ChIP-seq analysis

The target genes of RUNX1 were retrieved from the published ChIP-seq data of three AML cell lines, U937 (GSM1632640 and GSM1632641), HL60 (GSM2871154 and GSM2871145), and THP-1 (GSM2108052), and reanalyzed as follows. Raw sequencing data were downloaded from the NCBI Sequence Read Archive. Short read data were aligned to the UCSC human reference genome hg38 by using Bowtie2 (version 2.3.4.3) [12]. PCR duplicates were removed by using SAMtools (version 1.9) [13]. RUNX1 binding peaks were called by utilizing MACS2 (version 2.1.2) [14] at a cut-off of q= 0.05 (data with input) or P = 10e−6 (data without input). High-confidence binding peaks were defined as overlapping regions in at least two cell lines. UCSC RefGene data were applied to extract gene promoter regions (-2 kb/+0 kb to transcription start sites (TSS) for genes shorter than 500 bp and -2 kb/+0.5 kb to TSS otherwise). BEDTools [15], which assigned binding peaks to RUNX1 target genes, were used to overlap high-confidence peak regions with gene promoter regions. The Cistrome Data Browser and the UCSC genome browser were used for visualization [16,17].

2.5 DNA methylation analysis

We reanalyzed the Illumina Infinium 450k DNA Methylation Array data of 113 patients from the TCGA AML cohort from scratch to update methylation profiles on the basis of the reference genome hg38. The raw IDAT files (Level 1) for each sample were downloaded from the online supplementary material of TCGA’s work [18]. The probe annotation data (released on September 9, 2018) of hg38 were from Zhou et al’s paper [19]. The R package minfi (version 1.28.3) [20] was used to adjust the background and normalize the red/green dye bias across samples to produce the beta values of each probe. Poorly performing probes with detection P<0.05 and probes on the sex chromosomes were filtered out first. On the basis of the probe annotation data, the probes were further removed if they had SNPs close to the 3′ end, multiple mappable locations, or partial overlaps with repeated elements in the bisulfite-converted genome. We finally obtained 407 071 high-quality probes on the array. On the basis of the beta values of the probes, we performed moderated t-test with the limma package (version 3.38.3) [21] to evaluate probe-wise differences. The probe coordinates in the genome were then used to generate CpG clusters and detect differentially methylated regions (DMRs) with the bumphunter package (version 1.24.5) [22]. We set the parameter maxGap to 500 bp and the parameter B to 4000 and defined family-wide error rate<0.1 to be statistically significant. The gene promoter regions were extracted as mentioned in the ChIP-seq analysis.

2.6 Bioinformatics analyses

Gene Ontology (GO) enrichment analysis was performed by using the clusterProfiler package [23]. We conducted gene set enrichment analysis (GSEA) in accordance with the user guide and ranked all the genes by gene expression fold change (RUNX1-mutated vs. RUNX1 wild-type). Survival analysis was conducted by using the packages survival and survminer. Prognostic impacts were evaluated with the Kaplan–Meier method and the log-rank test. Statistical analyses were performed with the environment R version 3.5.1.

3 Results

3.1 RUNX1 mutations were mainly in the Runt domain and exclusive with NPM1 mutations

To investigate the characteristics of RUNX1 mutations in AML, we carried out a meta-analysis that included 1741 patients with AML collected from three AML cohorts (detailed in the section of “Materials and methods”). First, we found that RUNX1 mutations were mutually exclusive with frequently occurring fusions in core-binding factor AML (CBF-AML), including RUNX1–RUNX1T1 generated by t(8;21) and CBFB–MYH11 generated by inv(16) (Fig. 1A). Given that RUNX1–RUNXT1 and CBFB–MYH11, which are both driver fusions that dominantly inhibit CBF functions, are considered as independent factors promoting the development of AML [24], the data indicated that RUNX1 mutations might independently promote leukemogenesis. Second, the detailed analysis of the distribution of RUNX1 mutations showed that 125 patients with RUNX1 mutations carried 135 RUNX1 mutations, including 57 frameshifts (42.2%) and 18 nonsense variants (13.3%). Most of the mutations (n = 102, 75.6%) were located in the highly conserved Runt domain (Fig. 1B) that binds to the Runt-binding DNA element [25]. Additionally, we calculated the frequency of mutated loci and found that the top recurrent mutations included p.R166Q, p.D198G, p.R201*, p.R204*, and p.R204Q (5 times), and p.R162k (4 times). Interestingly, RUNX1 mutations were far less frequently observed in the Runx1 inhibition domain than in the Runt domain (Fig. 1B). Third, we compared the concurrent mutations with mutant RUNX1 vs. wild-type RUNX1.We used samples with cytogenetically normal AML (CN-AML) to exclude the interference of chromosomal abnormalities. As shown in Fig. 1C, the following genes were more frequently mutated among patients with AML and RUNX1 mutations than those with the wild-type counterpart: SRSF2 (P<0.001, the Fisher’s exact test), ASXL1 (P<0.001), BCOR (P = 0.001), and SF3B1 (P = 0.038). By contrast, significantly fewer NPM1 (P<0.001), CEBPA (P = 0.006), and DNMT3A (P = 0.02) mutations were observed in RUNX1-mutated cases than in other cases.

3.2 Patients with AML and RUNX1 mutations show a distinct gene signature

Considering that RUNX1 mutations showed marked distribution characteristics in AML cohorts, we hypothesized that RUNX1-mutated AML might exhibit common gene expression features. Given that NPM1 mutations were frequently observed (49%) in patients with wild-type RUNX1 but rarely observed in patients with RUNX1 mutations (Fig. 1C), DEGs might result from either NPM1 mutations or RUNX1 mutations. Therefore, we removed NPM1-mutated samples to compare patients with RUNX1 mutations and wild-type RUNX1. DEG analysis (FDR<0.01, FC>1.5 or<2/3) showed that 476 genes were significantly upregulated in RUNX1-mutated AML, whereas 531 were significantly downregulated (Fig. 2A). On the basis of these DEGs, we further developed a gene signature for RUNX1 mutations to characterize the most noticeable differences between patients with RUNX1 mutations and wild-type RUNX1. The clustering performance of several gene signature settings (100, 200, 250, 300, and 400 of the most variable DEGs) was compared (Fig. S1). The 300-gene signature was chosen for the minimum number of genes that yielded the best classification in hierarchical clustering: RUNX1-mutated samples clustered as a distinct class, except only one RUNX1-mutated sample (Fig. 2B; Table S1). We performed PCA to test the capability of the 300-gene signature to identify RUNX1-mutated samples and found that the RUNX1-mutated samples separated from the other samples in the CN-AML cohort (Fig. 2C). Furthermore, we used this gene signature for hierarchical clustering in the whole TCGA AML cohort and the GEO AML cohort (GSE23312). As shown in Fig. 2D, the majority of RUNX1-mutated samples clustered as one group, confirming the performance of this signature. Interestingly, the t(8:21) AML samples also formed a separate cluster (Fig. 2D), demonstrating the extended utility of this signature. Thus, the gene signature consisting of the top 300 DEGs represented the distinct gene expression pattern of RUNX1-mutated AML.

We performed GO enrichment analysis to explore the biological function associated with this signature. The results obtained for 199 downregulated genes showed the enrichment of GO terms related to myeloid functions, including neutrophil degranulation, neutrophil activation, and neutrophil-mediated immunity (Fig. 2E). Specially, the expression of four vital genes (AZU1, CTSG, LRG1, and MPO) in these neutrophil-associated GO-terms were significantly repressed in RUNX1-mutated cases (Fig. 2F), suggesting that RUNX1 mutations might dysregulate neutrophil functions in myeloid cells to contribute to the development of AML. These findings were also supported by a recent work showing that RUNX1 mutations could lead to the blockage of granulocytic differentiation in in vitro models and primary AML samples [26]. For upregulated genes, no pathways were significantly enriched.

3.3 RUNX1 target genes were generally repressed in RUNX1-mutated AML

Given that RUNX1 is a crucial TF in hematopoiesis, RUNX1 mutations in AML may impact its target genes. We analyzed high-confidence RUNX1 binding peaks from the ChIP-seq data of three AML cell lines without RUNX1 aberrations to identify RUNX1 target genes (Fig. 3A). High-confidence binding peaks were defined as those occurring in at least two of the cell lines. Genes with high-confidence RUNX1 binding peaks in their promoters were considered as RUNX1 target genes. Accordingly, a total of 1371 RUNX1 target genes were identified.

Next, we performed GSEA to investigate the general effects of RUNX1 mutations on these RUNX1 target genes. RUNX1 targets were significantly enriched in the genes with the negative log₂ fold change in RUNX1-mutated AML (Fig. 3B), indicating that RUNX1 target genes were generally repressed in AML with RUNX1 mutations. The aforementioned 300-gene signature that represented the distinct expression pattern in RUNX1-mutated AML was chosen for in-depth investigation to determine the effects of mutated RUNX1 on the genes with the most differential expression. The intersection between the RUNX1 targets and the signature yielded 15 genes that were deregulated RUNX1 targets in RUNX1-mutated AML (Fig. 3C). Among these 15 RUNX1 targets, 14 were downregulated and one was upregulated, suggesting that the direct effect of RUNX1 mutations on transcription was predominantly repressive (Fig. 3C). For example, CTSG, which is associated with neutrophil-associated host defense and immune response, was a target gene of RUNX1 in all three AML cell lines (Fig. 3D) and was significantly downregulated in patients with RUNX1 mutations relative to in patients with wild-type RUNX1 (Fig. 3E). Our previously published study has demonstrated that the suppression of CTSG in t(8;21) AML is caused by RUNX1–RUNX1T1 direct targeting and plays a critical role in the pathogenesis of AML [27]. The results of the current work showed that the role of mutant RUNX1 in transcriptional repression might be similar to that of RUNX1–RUNX1T1 and might contribute to the progression of AML by inhibiting the expression of CTSG.

3.4 Association of RUNX1 mutations with DNA hypermethylation

RUNX1 mutations might also influence epigenetic modifications, such as DNA methylation. Therefore, the DNA methylation microarray data of the TCGA AML cohort were reanalyzed. We improved the analysis accuracy by utilizing the updated hg38 human reference genome and a newly developed probe-masking method for Infinium HumanMethylation450K BeadChip probes [19]. Accordingly, we identified 210 DMRs, among which 183 were hypermethylated, and 27 were hypomethylated in RUNX1-mutated AML. This result indicated that RUNX1 mutations might be associated with DNA hypermethylation.

We assigned these DMRs to gene promoter regions and detected 51 differentially methylated genes (DMGs). Next, we intersected DEGs and DMGs to investigate whether the methylation status change caused dysregulated gene expression. The presence of 10 DEGs among 51 DMGs might partially suggest the contribution of the altered DNA methylation levels to gene expression changes caused by RUNX1 mutations. The majority of these genes (8/10) were in the aforementioned gene signature of RUNX1 mutations: one was hypomethylated and upregulated, whereas seven were hypermethylated and downregulated due to RUNX1 mutations (Fig. 4A). For example, MS4A3 is involved in the regulation of innate immune pathways and the G1/S cell cycle in hematopoiesis [28]. This result suggested that RUNX1 mutations might lead to promoter hypermethylation and MS4A3 expression repression (Fig. 4B). These events might affect cell cycle regulation and contribute to the differentiation blockage of AML.

Next, the TCGA AML cohort was used to evaluate the prognostic influence of the expression of hypermethylated and downregulated genes (see the section of “Materials and methods”). As shown in Fig. 4C, the low expression of MS4A3 was associated with increased hazard ratio and a poor prognosis, indicating that MS4A3 downregulation might be crippling in RUNX1-mutated AML. Furthermore, the low expression of two additional hypermethylated and downregulated genes, namely, CD96 and LTK, was also correlated with the poor prognosis of patients with AML (Fig. S2). Additionally, the clinical outcome was less influenced by treatment choices than other factors (Table S2, Table S3, and Table S4). In summary, the results suggested that DNA hypermethylation was associated with RUNX1 mutations and contributed to gene repression caused by RUNX1 mutations.

4 Discussion

Somatic gene mutations constitute key events in AML pathogenesis, and RUNX1-mutated AML has been added as a provisional category in accordance with the updated (2016) WHO classification. In this study, we investigated the genetic characteristics of RUNX1 mutations in AML and showed that patients with RUNX1-mutated AML had a distinct gene expression pattern. We found that RUNX1 mutations generally repressed RUNX1 target genes and were associated with DNA hypermethylation. Our investigation helps understand the pathobiology of RUNX1-mutated AML.

Using a large number of samples helps us obtain an improved understanding of the characteristics of mutation distribution. First, we identified several previously unreported recurrent mutations in RUNX1, such as p.R166Q, p.D198G, p.R201fs, and p.R204fs. In addition to recurrent mutations, we found some interesting results regarding co-occurrent or mutually exclusive mutations with RUNX1 mutations. Cases with CN-AML were chosen to analyze the co-occurrent or mutually exclusive mutations with RUNX1 mutations because RUNX1 mutations mainly occur in patients with AML and normal karyotypes [6]. In addition to confirming the previous finding on the mutual exclusivity of RUNX1 mutations with NPM1 mutations [7], we obtained several novel findings: (1) mutations previously reported as concurrent, such as IDH1, IDH2, and DNMT3A [29], did not show a significantly higher frequency in the RUNX1-mut cohort than in the RUNX1-wt cohort and (2) DNMT3A mutations even had a statistically significantly lower frequency in the RUNX1-mut cohort than in other cohorts. These new findings were mainly attributed to the effective analysis based on CN-AML because the exclusion of AML cases with abnormal karyotypes removed biases and avoided underestimating the frequency of mutations, such as DNMT3A and IDH1/2, that are highly prevalent in CN-AML [30,31].

We developed a gene expression signature that represented the most distinguishable expression features of patients with RUNX1 mutations and could separate these patients from patients with wild-type RUNX1 in sample clustering. The performance of this signature had improved compared with that of a previously reported signature [6] because the effects of NPM1 mutations were excluded. High-frequency NPM1 mutations were almost exclusively enriched in RUNX1 wild-type AMLs. The removal of NPM1-mutated cases prior to gene signature development could clearly identify the differences caused by RUNX1 rather than the combinatory effects of RUNX1 and NPM1 mutations. Moreover, compared with the noncommercial microarray data used in previous studies [6,32], we used the updated TCGA RNA-seq data of hg38 (release on September 27, 2018), which covered more of the genome and had high levels of reproducibility. In addition, GO enrichment analysis based on our gene signature showed that pathways related to neutrophil functions were downregulated in RUNX1-mutated AML; this result was consistent with the finding of a recently published work showing that RUNX1 mutations lead to a blockage in granulocytic differentiation in human CD34⁺ progenitor cells [26].

Our results offer insights into the mechanisms underlying gene expression alterations in cases with RUNX1 mutations. First, we considered that RUNX1 mutations might lead to the dysfunction of RUNX1 in transcriptional regulation. When we took all the expression differences into consideration using GSEA, RUNX1 target genes tended to be downregulated in RUNX1-mutated AML. When we used relatively loose cut-offs to conduct differential gene expression analysis (FDR<0.05, FC>1.5 or<2/3), approximately 7.3% (n = 100) of RUNX1 targets were found to be differentially expressed with statistical significance and generally downregulated (n = 68). This result was suggestive of noticeable direct gene repression caused by RUNX1 mutations. Second, we investigated the effects of RUNX1 mutations on DNA methylation. We found that 10 out of 1007 DEGs were differentially methylated, indicating that DNA methylation partially accounted for the altered gene expression levels. Three DEGs, namely, MS4A3, PLPPR3, and C20orf197, were found to be RUNX1 targets that were hypermethylated and downregulated in RUNX1-mutated AML. In normal hematopoietic cells, RUNX1 regulates DNA demethylation by recruiting the DNA demethylation enzymes TET2, TET3, TDG, and GADD45 [33]. We may reasonably hypothesize that in RUNX1-mutated AML, RUNX1 mutations led to a mechanism that blocked the recruitment of demethylation modifiers. This effect increased the methylation of RUNX1 target genes and reduced their gene expression. However, only bioinformatics analyses were conducted in our study, and further wet experiments on regulatory targets and methylation status are required for in-depth investigation.

Our work not only indicates that RUNX1-mutated AML is an entity with a distinct gene expression pattern but also provides insights into the pathogenic mechanisms caused by RUNX1 mutations. Furthermore, the repression of RUNX1 and its targets has been reported to inhibit growth and induce the apoptosis of AML cells expressing mutant RUNX1 [34], suggesting that expression mimickers that deplete RUNX1 and its target gene expression levels may become novel therapeutic agents. Therefore, our findings on the gene signatures that are especially dysregulated in RUNX1-mutated AML may help identify promising targets with prognostic and therapeutic utilities.

References

[1] Zhu H, Wang G, Qian J. Transcription factors as readers and effectors of DNA methylation. Nat Rev Genet 2016; 17(9): 551–565

[2] Guo H, Ma O, Speck NA, Friedman AD. Runx1 deletion or dominant inhibition reduces Cebpa transcription via conserved promoter and distal enhancer sites to favor monopoiesis over granulopoiesis. Blood 2012; 119(19): 4408–4418

[3] Tober J, Yzaguirre AD, Piwarzyk E, Speck NA. Distinct temporal requirements for Runx1 in hematopoietic progenitors and stem cells. Development 2013; 140(18): 3765–3776

[4] Sood R, Kamikubo Y, Liu P. Role of RUNX1 in hematological malignancies. Blood 2017; 129(15): 2070–2082

[5] Tang JL, Hou HA, Chen CY, Liu CY, Chou WC, Tseng MH, Huang CF, Lee FY, Liu MC, Yao M, Huang SY, Ko BS, Hsu SC, Wu SJ, Tsay W, Chen YC, Lin LI, Tien HF. AML1/RUNX1 mutations in 470 adult patients with de novo acute myeloid leukemia: prognostic implication and interaction with other gene alterations. Blood 2009; 114(26): 5352–5361

[6] Gaidzik VI, Bullinger L, Schlenk RF, Zimmermann AS, Röck J, Paschka P, Corbacioglu A, Krauter J, Schlegelberger B, Ganser A, Späth D, Kündgen A, Schmidt-Wolf IG, Götze K, Nachbaur D, Pfreundschuh M, Horst HA, Döhner H, Döhner K. RUNX1 mutations in acute myeloid leukemia: results from a comprehensive genetic and clinical analysis from the AML study group. J Clin Oncol 2011; 29(10): 1364–1372

[7] Mendler JH, Maharry K, Radmacher MD, Mrózek K, Becker H, Metzeler KH, Schwind S, Whitman SP, Khalife J, Kohlschmidt J, Nicolet D, Powell BL, Carter TH, Wetzler M, Moore JO, Kolitz JE, Baer MR, Carroll AJ, Larson RA, Caligiuri MA, Marcucci G, Bloomfield CD. RUNX1 mutations are associated with poor outcome in younger and older patients with cytogenetically normal acute myeloid leukemia and with distinct gene and microRNA expression signatures. J Clin Oncol 2012; 30(25): 3109–3118

[8] Arber DA, Orazi A, Hasserjian R, Thiele J, Borowitz MJ, Le Beau MM, Bloomfield CD, Cazzola M, Vardiman JW. The 2016 revision to the World Health Organization classification of myeloid neoplasms and acute leukemia. Blood 2016; 127(20): 2391–2405

[9] Papaemmanuil E, Gerstung M, Bullinger L, Gaidzik VI, Paschka P, Roberts ND, Potter NE, Heuser M, Thol F, Bolli N, Gundem G, Van Loo P, Martincorena I, Ganly P, Mudie L, McLaren S, O’Meara S, Raine K, Jones DR, Teague JW, Butler AP, Greaves MF, Ganser A, Döhner K, Schlenk RF, Döhner H, Campbell PJ. Genomic classification and prognosis in acute myeloid leukemia. N Engl J Med 2016; 374(23): 2209–2221

[10] Gerstung M, Papaemmanuil E, Martincorena I, Bullinger L, Gaidzik VI, Paschka P, Heuser M, Thol F, Bolli N, Ganly P, Ganser A, McDermott U, Döhner K, Schlenk RF, Döhner H, Campbell PJ. Precision oncology for acute myeloid leukemia using a knowledge bank approach. Nat Genet 2017; 49(3): 332–340

[11] Love MI, Anders S, Kim V, Huber W. RNA-Seq workflow: gene-level exploratory analysis and differential expression. F1000 Res 2015; 4: 1070

[12] Langmead B, Wilks C, Antonescu V, Charles R. Scaling read aligners to hundreds of threads on general-purpose processors. Bioinformatics 2019; 35(3): 421–432

[13] Li H, Handsaker B, Wysoker A, Fennell T, Ruan J, Homer N, Marth G, Abecasis G, Durbin R; 1000 Genome Project Data Processing Subgroup. The Sequence Alignment/Map format and SAMtools. Bioinformatics 2009; 25(16): 2078–2079

[14] Zhang Y, Liu T, Meyer CA, Eeckhoute J, Johnson DS, Bernstein BE, Nusbaum C, Myers RM, Brown M, Li W, Liu XS. Model-based analysis of ChIP-Seq (MACS). Genome Biol 2008; 9(9): R137

[15] Quinlan AR, Hall IM. BEDTools: a flexible suite of utilities for comparing genomic features. Bioinformatics 2010; 26(6): 841–842

[16] Mei S, Qin Q, Wu Q, Sun H, Zheng R, Zang C, Zhu M, Wu J, Shi X, Taing L, Liu T, Brown M, Meyer CA, Liu XS. Cistrome Data Browser: a data portal for ChIP-Seq and chromatin accessibility data in human and mouse. Nucleic Acids Res 2017; 45(D1): D658–D662

[17] Haeussler M, Zweig AS, Tyner C, Speir ML, Rosenbloom KR, Raney BJ, Lee CM, Lee BT, Hinrichs AS, Gonzalez JN, Gibson D, Diekhans M, Clawson H, Casper J, Barber GP, Haussler D, Kuhn RM, Kent WJ. The UCSC Genome Browser database: 2019 update. Nucleic Acids Res 2019; 47(D1): D853–D858

[18] The Cancer Genome Atlas Research Network. Genomic and epigenomic landscapes of adult de novo acute myeloid leukemia. N Engl J Med 2013; 368(22): 2059–2074

[19] Zhou W, Laird PW, Shen H. Comprehensive characterization, annotation and innovative use of Infinium DNA methylation BeadChip probes. Nucleic Acids Res 2017; 45(4): e22

[20] Aryee MJ, Jaffe AE, Corrada-Bravo H, Ladd-Acosta C, Feinberg AP, Hansen KD, Irizarry RA. Minfi: a flexible and comprehensive Bioconductor package for the analysis of Infinium DNA methylation microarrays. Bioinformatics 2014; 30(10): 1363–1369

[21] Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, Smyth GK. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 2015; 43(7): e47

[22] Peters TJ, Buckley MJ, Statham AL, Pidsley R, Samaras K, V Lord R, Clark SJ, Molloy PL. De novo identification of differentially methylated regions in the human genome. Epigenetics Chromatin 2015; 8(1): 6

[23] Yu G, Wang LG, Han Y, He QY. clusterProfiler: an R package for comparing biological themes among gene clusters. OMICS 2012; 16(5): 284–287

[24] Duployez N, Marceau-Renaut A, Boissel N, Petit A, Bucci M, Geffroy S, Lapillonne H, Renneville A, Ragu C, Figeac M, Celli-Lebras K, Lacombe C, Micol JB, Abdel-Wahab O, Cornillet P, Ifrah N, Dombret H, Leverger G, Jourdan E, Preudhomme C. Comprehensive mutational profiling of core binding factor acute myeloid leukemia. Blood 2016; 127(20): 2451–2459

[25] Fukunaga J, Nomura Y, Tanaka Y, Amano R, Tanaka T, Nakamura Y, Kawai G, Sakamoto T, Kozu T. The Runt domain of AML1 (RUNX1) binds a sequence-conserved RNA motif that mimics a DNA element. RNA 2013; 19(7): 927–936

[26] Gerritsen M, Yi G, Tijchon E, Kuster J, Schuringa JJ, Martens JHA, Vellenga E. RUNX1 mutations enhance self-renewal and block granulocytic differentiation in human in vitro models and primary AMLs. Blood Adv 2019; 3(3): 320–332

[27] Jin W, Wu K, Li YZ, Yang WT, Zou B, Zhang F, Zhang J, Wang KK. AML1-ETO targets and suppresses cathepsin G, a serine protease, which is able to degrade AML1-ETO in t(8;21) acute myeloid leukemia. Oncogene 2013; 32(15): 1978–1987

[28] Kutok JL, Yang X, Folkerth R, Adra CN. Characterization of the expression of HTm4 (MS4A3), a cell cycle regulator, in human peripheral blood cells and normal and malignant tissues. J Cell Mol Med 2011; 15(1): 86–93

[29] Khan M, Cortes J, Kadia T, Naqvi K, Brandt M, Pierce S, Patel KP, Borthakur G, Ravandi F, Konopleva M, Kornblau S, Kantarjian H, Bhalla K, DiNardo CD. Clinical outcomes and co-occurring mutations in patients with RUNX1-mutated acute myeloid leukemia. Int J Mol Sci 2017; 18(8): E1618

[30] O’Brien EC, Brewin J, Chevassut T. DNMT3A: the DioNysian MonsTer of acute myeloid leukaemia. Ther Adv Hematol 2014; 5(6): 187–196

[31] Ok CY, Loghavi S, Sui D, Wei P, Kanagal-Shamanna R, Yin CC, Zuo Z, Routbort MJ, Tang G, Tang Z, Jorgensen JL, Luthra R, Ravandi F, Kantarjian HM, DiNardo CD, Medeiros LJ, Wang SA, Patel KP. Persistent IDH1/2 mutations in remission can predict relapse in patients with acute myeloid leukemia. Haematologica 2019; 104(2): 305–311

[32] Greif PA, Konstandin NP, Metzeler KH, Herold T, Pasalic Z, Ksienzyk B, Dufour A, Schneider F, Schneider S, Kakadia PM, Braess J, Sauerland MC, Berdel WE, Büchner T, Woermann BJ, Hiddemann W, Spiekermann K, Bohlander SK. RUNX1 mutations in cytogenetically normal acute myeloid leukemia are associated with a poor prognosis and up-regulation of lymphoid genes. Haematologica 2012; 97(12): 1909–1915

[33] Suzuki T, Shimizu Y, Furuhata E, Maeda S, Kishima M, Nishimura H, Enomoto S, Hayashizaki Y, Suzuki H. RUNX1 regulates site specificity of DNA demethylation by recruitment of DNA demethylation machineries in hematopoietic cells. Blood Adv 2017; 1(20): 1699–1711

[34] Mill CP, Fiskus W, DiNardo CD, Qian Y, Raina K, Rajapakshe K, Perera D, Coarfa C, Kadia TM, Khoury JD, Saenz DT, Saenz DN, Illendula A, Takahashi K, Kornblau SM, Green MR, Futreal AP, Bushweller JH, Crews CM, Bhalla KN. RUNX1-targeted therapy for AML expressing somatic or germline mutation in RUNX1. Blood 2019; 134(1): 59–73

RIGHTS & PERMISSIONS

Higher Education Press