Berger, Kurt

Introduction: Truncating mutations in Additional Sex Combs-Like 1 (ASXL1) are associated with a high-risk disease phenotype in myeloid malignancies. In chronic myelomonocytic leukemia (CMML), truncating ASXL1 mutations are known to increase transcriptional activity of leukemic driver genes and have been associated with gene body hypermethylation. We interrogated the transcriptome and methylome of patients with ASXL1-mutant (MT) and -wildtype (WT) CMML using a multi-omics approach to test the hypothesis that gene expression is mediated through gene body methylation. Methods: Bone marrow mononuclear cells from patients with ASXL1 WT (n=8) and MT (n=8) CMML were subjected to targeted NGS of DNA, whole transcriptome shotgun sequencing (RNA-seq), and immunoprecipitation of DNA methyl residues (DIP-seq). After quality control all samples were sequenced on an Illumina HiSeq 4000 before further processing and data analysis. Differential gene expression analysis was performed to identify genes up-regulated in MT CMML. The samples in the two groups were treated as biological replicates and subjected to a consensus peak calling strategy requiring an overlap of at least 30% between samples and an adjusted p-value < 5x10-5 for a methylation peak to be considered statistically significant. For validation purposes methylation analysis was performed on 3 ASXL1 MT and 3 WT CMML patients using Illumina Infinium MethylationEPIC microarrays and differentially methylated regions were identified using a bump hunting strategy. Gene body methylation was defined as methylation in gene bodies (outside the promoter region, i.e. transcription start site ±2kb). Gene body methylation was compared between WT and MT CMML for the up-regulated genes and correlated with expression of all genes in MT CMML. Results: Sixteen WHO-defined CMML patients were included, median age 69 years (48-77), 63% male, 50% had truncating ASXL1 frame shift mutations. Abnormal karyotypes were observed in the same number of patients and the burden of co-mutations was similar between the two groups (median number per group 3 vs. 3, p=0.508). This included several modulators of DNA methylation including TET2, DNMT3A, and IDH2 (median number per group 1 vs. 1, p=0.699). There was a predominant up-regulation of gene expression in MT CMML: 707 genes up- and 124 down-regulated (FDR<0.050) without evidence of differential methylation in promoter regions of the differentially expressed genes. There was no difference in the number of genes with consensus methylation peaks in the gene body or the extent of gene body methylation (area under the consensus peaks) between MT and WT CMML for the differentially expressed genes (Figure 1a). We further validated these results using methylation microarrays (Figure 1b). Among the MT CMML patients, there was a curvilinear relationship between gene body methylation and gene expression across all genes with intermediate gene body methylation being associated with the highest gene expression (Figure 1c). As an alternative unbiased approach, we identified 1595 differentially methylated regions (DMR, regions with FDR<0.05) in the validation data set. We mapped all hypermethylated regions to gene bodies requiring that there was no concurrent hypermethylation of the promoter region of the same gene. With this approach we identified one of the 707 up-regulated genes (0.14%) with evidence of isolated gene body hypermethylation (HBZ, log2-fold change in gene expression 4.25, FDR=0.0008, DMR area=1.61, DMR FDR=0.018). Representative examples of up-regulated mitotic kinases without evidence of differential gene body methylation are shown in Figure 1d (pie charts represent the mean β-values of microarray probes, p-values represent the comparisons of mean β-values between MT and WT CMML per gene). Conclusions: Gene body methylation was positively associated with gene expression in MT CMML. However, the lack of differential gene body methylation between WT and MT CMML for the up-regulated genes make it an unlikely explanation for the observed increase in transcriptional activity among patients with MT CMML. Figure 1 Disclosures Ordog: Millipore Sigma: Patents & Royalties.

Introduction: Additional Sex Combs-Like 1 (ASXL1) is a chromatin modifier frequently affected by truncating mutations in myeloid malignancies. In chronic myelomonocytic leukemia (CMML), truncating ASXL1 mutations are associated with the overexpression of leukemogenic driver genes, however the molecular mechanisms underlying this transcriptional activation remain controversial. We performed single-cell chromatin accessibility studies of ASXL1-mutant (MT) and -wildtype (WT) CMML to identify cis-regulatory elements (CREs) such as distal enhancer elements that may explain the observed transcriptional activity in MT CMML. Methods: Bone marrow mononuclear cells from a patient with WT and a patient with MT CMML were subjected to single-cell DNA transposase accessibility studies (scATAC-seq). The cryopreserved cells were thawed and resuspended and approximately 100000 viable mononuclear cells per sample were subjected to transposase assays before proceeding to single-cell partitioning into gel beads in emulsion, barcoding, library construction, and sequencing following an established 10X Genomics workflow. The target cell recovery was approximately 2000 cells per sample. Genomic libraries were sequenced on an Illumina HiSeq 4000 before demultiplexing, alignment to the reference genome, and post-alignment quality control. The 10X Genomics Cell Ranger ATAC software was used for demultiplexing, alignment of the reads to the GRCh38 reference genome, filtering and quality control, counting of barcodes and unique molecular identifiers, identification of transposase cut sites, detection of accessible chromatin peaks, count matrix generation for peaks and transcription factors, clustering, and differential accessibility analysis. The two-sample Kolmogorov-Smirnov test for equality of distribution functions was used to compare the distributions of single-cell entropies. Single-cell cis-co-accessibility scores were estimated using Cicero. Results: To mitigate potential effects on the analysis introduced by unequal cell recovery due to experimental conditions, we randomly sampled 1500 single cells from each genotype. Visualizing the first two principal components of the chromatin accessibility data revealed the differences in chromatin conformation between genotypes and also the increased heterogeneity within the MT sample (Figure 1a). To quantify this increase in intratumoral heterogeneity of chromatin accessibility in MT CMML, we calculated the single-cell entropies for both genotypes. There was an increase in the dispersion of single-cell entropy values (mean 0.793±0.026 versus 0.811±0.012, p=6.19x10-151) in MT compared to WT disease (Figure 1b). To leverage the power of the single-cell chromatin accessibility data, we estimated cis-co-accessibility scores (two-way chromatin interactions). There were 631 genes involved in two-way chromatin interactions that were either specific to MT CMML or were shared between MT and WT CMML and showed an increase in co-accessibility in MT CMML. Co-accessibility was higher in MT compared to WT CMML, regardless whether the chromatin interaction was MT-specific or shared between the genotypes (Figure 1c). Gene ontology analysis revealed increased chromatin accessibility in genes involved in the regulation of cell cycle, nuclear division, chromosome organization, cell cycle phase transition, DNA conformation change, and DNA replication (FDR<1.00x10-5). The construction of cis-co-accessibility networks demonstrated a remodeling of the chromatin landscape with MT-specific chromatin interactions involving key mitotic kinases such as MAP2K3 (Figure 1d). Conclusions: MT CMML is characterized by an increase in intratumoral heterogeneity and increased chromatin accessibility. The increased chromatin co-accessibility involved key mitotic kinases and may serve as a plausible explanation for the increased transcriptional activity observed in MT CMML. Figure 1 Disclosures Ordog: Millipore Sigma: Patents & Royalties.

Introduction: Chronic myelomonocytic leukemia (CMML) is a clonal hematopoietic neoplasm characterized by sustained peripheral blood (PB) monocytosis and an inherent risk for leukemic transformation. Clonal origins of the disease can be detected in hematopoietic progenitor cells (CD34+/CD38-), while the complete spectrum of mutational evolution can be seen in circulating monocytes (CD14+). Cell sorting strategies have been employed to select cells in CMML, and while there are adequate monocyte numbers in the PB, there are very few circulating progenitor cells. In addition, attrition related to the selection process significantly depletes primary cells available for biological experiments and multiomics studies such as RNA-seq, ChIP-seq, ATAC-seq, and DIP-seq. While single-cell methods may be able to overcome this challenge, bulk sequencing methods remain a robust and cost-effective approach. We hypothesized that, secondary to the stem cell origin of this disease and significant myeloproliferation, PB mononuclear cells (MNC) would provide comparable results with regards to transcriptomic analysis, in comparison to cell selection procedures. Methods: Peripheral blood obtained from 15 molecularly annotated patients with WHO-defined CMML was ACK-lysed and subjected to a Ficoll procedure for collection of MNC. MNC were left unsorted (n=5) or further selected for CD34+/CD38- (n=5) and CD14+ (n=5) using a fully automated RoboSep-S (StemCell Technologies) protocol. All samples were then subjected to bulk whole transcriptome shotgun sequencing (using Illumina TruSeq and an Illumina HiSeq 4000). After data quality control, counts of detectable transcripts were log2-normalized and Pearson's product-moment correlation coefficients were calculated to evaluate the correlation between the two cell-sorting strategies and unsorted cells in terms of detectable transcripts. To visualize sample differences log2-normalized transcripts counts were centered and scaled per gene for a select number of genes relevant to myeloid biology as well as a number of housekeeping genes. Results: Fifteen patients with WHO-defined CMML, median age 69 years (55-73 years), 66% male, were included. Next generation sequencing for somatic mutations was performed on PB MNC obtained at CMML diagnosis (Figure 1, top heatmap). Considering the small sample size, mutations were evenly distributed among groups with the exception of ASXL1 (higher frequency in CD14+ and CD34+/CD38- cells), ZRSR2 (higher frequency in unsorted cells), and TET2 (lower frequency in CD14+ cells). The three groups were also well matched with regards to other CMML-related variables such as WHO and FAB morphological subtypes, cytogenetic abnormalities, and risk stratification by the Mayo Molecular Model. Transcriptomic analysis revealed a strong positive correlation between the median number of log2-normalized detectable transcripts in unsorted cells and CD34+/CD38- cells (ρ = 0.96, p < 0.001, top scatterplot). Likewise, there was a strong positive correlation between the median number of log2-normalized detectable transcripts in unsorted cells and CD14+ cells (ρ = 0.91, p < 0.001, bottom scatterplot). The latter correlation was marginally lower, which was explained by increased global gene expression in 3 of the 5 CD14+ samples (bottom heatmap). Increased gene expression in these 3 samples involved key myeloid genes and housekeeping genes known to have stable expression across human tissues alike. In comparison to PB MNC, both cell sorting strategies resulted in significant depletion of primary cells required for other experiments, and for procedures such as ChIP-seq, DIP-seq and ATAC-seq (CD34+/CD38- had greater depletion than CD14+). Additional experiments to assess this strategy for the above mentioned epigenetic studies are currently being planned. Conclusions: Accounting for sample differences, different cell sorting strategies (unsorted, CD34+/CD38- selection, and CD14+ selection) yielded similar results when performing bulk transcriptomic assessments on PB MNC from patients with CMML. For the purpose of gene expression profiling there was no clear advantage with CD34+/CD38- or CD14+ selection. These results support the use of unsorted cells for bulk transcriptomic analysis in CMML. Figure 1 Disclosures Patnaik: Stem Line Pharmaceuticals.: Membership on an entity's Board of Directors or advisory committees.

Introduction: Truncating mutations in the Additional Sex Combs-Like 1 (ASXL1) gene are associated with a proliferative disease phenotype and poor survival outcomes across the spectrum of myeloid malignancies including chronic myelomonocytic leukemia (CMML). ASXL1 is thought to act as a chromatin modifier regulating transcriptional activity, however the exact mechanisms and resulting chromatin states remain controversial. We interrogated the epigenome of 16 patients with ASXL1-mutant and -wildtype CMML using a multiomics approach. Methods: Bone marrow mononuclear cells from patients with CMML (8 ASXL1-mutant, 8 -wildtype) were subjected to targeted NGS of DNA, whole transcriptome shotgun sequencing (RNA-seq), immunoprecipitation (IP) of DNA (hydroxy-)methyl residues (DIP-seq), IP of the histone modifications H3K4me1, H3K4me3, and H3K27me3 (ChIP-seq), and DNA transposase accessibility studies (ATAC-seq). After quality control all samples were sequenced on an Illumina HiSeq 4000 before further processing and data analysis. Global assessments of DNA (hydroxy-)methylation, DNA accessibility, and histone modifications between ASXL1-mutant and -wildtype CMML were performed. Differential gene expression was performed to define the up-regulated genes in ASXL1-mutant disease. The promoter regions of these up-regulated genes (defined as transcription start site ±3kb) were compared using the aforementioned multiomics approach. Epigenomic modification of the promoter region facilitating up-regulation of transcription was defined as the presence of a signal peak in ASXL1-mutant disease (in the absence of a signal peak in -wildtype disease) or 25% increase in a common signal peak (H3K4me1/3, 5hmC), the presence of a unique signal peak in ASXL1-mutant disease (ATAC), or the absence of a signal peak in ASXL1-mutant disease (in the presence of a signal peak in -wildtype disease), or 25% decrease in a common signal peak (H3K27me3, 5mC). Results: Sixteen patients with CMML, median age 69 years (48 - 77), 63% male, were included. Half of the patients had proliferative disease (pCMML) and half of them had truncating frameshift mutations in ASXL1 (heatmap). All ASXL1 variant allele frequencies were compatible with heterozygosity (31 - 48%). The burden of co-mutations was similar between ASXL1-wildtype and ASXL1-mutant disease (21 versus 23 per group; no difference in the median number of co-mutations, p = 0.684). The spectrum of co-mutations was typical for CMML, involving spliceosome components, epigenetic regulators, chromatin regulators, and cell signaling molecules (heatmap). There was a predominant up-regulation of gene expression in ASXL1-mutant patients: 707 genes up- and 124 down-regulated (volcano plot, FDR < 0.05 for all genes). Functional annotation of the up-regulated genes showed cell division, mitotic nuclear division, sister chromatid cohesion, DNA replication, and G1/S transition to be the 5 most enriched processes (accounting for 29% of all up-regulated genes, FDR < 1x10-10 for all terms). The up-regulated genes included several potential therapeutic targets and HOXA family members (including HOXA9). There were global increases in H3K4me1/3, 5mC, and 5hmC, decreases in H3K27me3, as well as a more relaxed chromatin conformation (bar graphs). Many of these epigenomic changes affected non-coding regions. When focusing on the promoter regions of the 707 up-regulated genes there was evidence of one or more of the interrogated epigenomic mechanisms facilitating transcription for 519 of the genes (73%). The most abundant mechanism was histone modification, followed by changes in DNA (hydroxy-)methylation, and increased chromatin accessibility, with considerable overlap (Venn diagram). For HOXA9, a known driver of leukemogenesis, the data supported a loss of H3K27me3 as the most prominent among the interrogated epigenomic regulatory mechanisms (average signal tracks). Conclusions: The transcriptome and chromatin conformation of ASXL1-mutant CMML are skewed towards proliferation and mirror the aggressive disease phenotype observed in practice. There is evidence of histone modification as well as changes in DNA methylation, and chromatin conformation facilitating transcriptional activity including known leukemogenic drivers. Additional regulatory mechanisms such as gene body methylation and enhancer elements require further exploration. Figure Disclosures Patnaik: Stem Line Pharmaceuticals.: Membership on an entity's Board of Directors or advisory committees.

Introduction: Chronic myelomonocytic leukemia (CMML), an aggressive myeloid malignancy, can be categorized into two subtypes, proliferative CMML (pCMML) and dysplastic (dCMML), based on a white blood cell (WBC) count of ≥ 13 x 109/L for the former (Arber et al. Blood 2016). While this WBC cut off is somewhat arbitrary, patients with pCMML have unique phenotypic features and a shorter survival. We carried out this study to assess the genomic, transcriptomic and epigenetic landscapes of these two CMML subtypes. Methods: Peripheral blood (PB) and bone marrow (BM) mononuclear cells (MNC) were obtained from WHO-defined CMML patients. Next generation sequencing (NGS) using a 36-gene panel was performed on 350 patients with Illumina HiSeq4000 platform with median read depth of 400X. RNA sequencing (RNA-seq) was performed on 25 patients by bulk whole transcriptome sequencing using Illumina TruSeq. DNA immunoprecipitation sequencing (DIP-seq) was performed on 18 patients using 5-methylcystocine (5mC), 5-hydroxymethylcytosine (5hmC) and bridging monoclonal antibodies with subsequent paired-end sequencing using HiSeq4000. Chromatin immunoprecipitation sequencing (ChIP-seq) was performed on 30 patients with Illumina HiSeq2500 to a depth of 25 million for histone 3 lysine 4 monomethylation (H3K4me1) and histone 3 lysine 4 trimethylation (H3K4me3) and 50 million reads for histone 3 lysine 27 trimethylation (H3K27me3) and Input per sample. Results: Five hundred and seventy-three patients with WHO defined CMML were included; median age 71 years (range 18-95 years), 67% males. 282 patients had pCMML (49%), while 291 (51%) had dCMML. As pre-defined, patients with pCMML were more likely to have higher absolute monocyte counts (p<0.0001), circulating immature myeloid cells (p<0.0001), PB blasts (p<0.0001), and higher lactate dehydrogenase levels (p=0.03). At last follow up 234 (41%) deaths and 70 (20%) leukemic transformations were documented. The median OS for pCMML vs dCMML in this cohort was 19 months vs 30 months (p<0.0001, Figure 1A) and validated in an independent Austrian cohort (p=0.02). Genomic profiling: NGS performed on 350 patients (BM MNC) revealed a higher frequency of NRAS (35 vs 17%, p=0.004), cumulative RAS pathway (NRAS, KRAS, CBL and PTPN11) (73 vs 47%, p=0.001), ASXL1 (p=0.003) and JAK2V617F (p=0.04) mutations in pCMML relative to dCMML (Figure 1B); while dCMML had a higher frequency of SF3B1 mutations (p=0.02). There were no differences in distribution of TET2 and SRSF2Transcriptomic analysis: RNA-seq was performed on PB MNC from RAS pathway mutant pCMML patients (n=12) and RAS pathway wildtype dCMML patients (n=13). Unsupervised clustering analysis resulted in appropriate segregation revealing distinct expression profiles between disease subtypes (Figure 1C). Compared to dCMML, 3729 genes were significantly differentially upregulated and 2658 genes were differentially downregulated in pCMML. Among genes most highly upregulated were mitotic checkpoint kinases including AURBK, PLK1, PLK2, PLK4 andEpigenetic profiling: ChIP-seq of PB and BM MNC from pCMML (n=18) and dCMML (n=12) patients and healthy, age-matched controls (n=10) revealed a global increase in H3K4me1, without significant differences in H3K4me3 or H3K27me3 occupancies (regardless of stratification by ASXL1 mutational status; 40% ASXL1mt in pCMML, 30% dCMML) in pCMML vs dCMML (Figure 1D). H3K4me1 occupancy was also increased in a sequence-specific manner at the transcription start sites of the aforementioned mitotic kinases (PLK1 and WEE1). DIP-seq was performed on PB MNC to assess global differences in 5-mC and 5-hmC levels, between pCMML (n=9) and dCMML (n=9), with no differences seen between the two subtypes (regardless of TET2 mutational status, 40% TET2mt in each subtype) (Figure 1E). Conclusions: Despite the somewhat arbitrary WBC distinction between pCMML and dCMML, clear phenotypic, genetic, transcriptomic, epigenetic and survival differences exist between the two subtypes, providing strong biological rationale for this distinction. pCMML patients have a higher frequency of oncogenic RAS pathway mutations, a unique transcriptomic profile enriched in mitotic check point kinases and a unique chromatin configuration with global and sequence specific enrichment in H3K4me1, with no significant global differences in 5mC, 5hmC, or H3K4me3 and H3K27me3 occupancy. Figure 1 Disclosures Geissler: AOP: Honoraria; Pfizer: Honoraria; AstraZeneca: Honoraria; Novartis: Honoraria; Celgene: Honoraria; Roche: Honoraria; Abbvie: Honoraria; Ratiopharm: Honoraria; Amgen: Honoraria. Al-Kali:Astex Pharmaceuticals, Inc.: Research Funding. Patnaik:Stem Line Pharmaceuticals.: Membership on an entity's Board of Directors or advisory committees.

Introduction: Additional Sex Combs-Like 1 (ASXL1) is a chromatin modifier frequently affected by truncating mutations in myeloid malignancies. These mutations are associated with poor survival outcomes and increased rates of acute leukemic transformation. In chronic myelomonocytic leukemia (CMML), ASXL1 mutations are thought to affect transcriptional activity mainly by modifying histone marks, however additional epigenomic mechanisms have not been fully explored. We interrogated the epigenome of patients with ASXL1-mutant (MT) and -wildtype (WT) CMML using a multiomics approach to define cis-regulatory elements (CREs) such as distal enhancers (DEs). Methods: Bone marrow mononuclear cells from patients with CMML were subjected to targeted NGS of DNA, whole transcriptome shotgun sequencing (RNA-seq), immunoprecipitation (IP) of DNA (hydroxy-)methyl residues (DIP-seq), IP of the histone modifications H3K4me1, H3K4me3, and H3K27me3 (ChIP-seq), and DNA transposase accessibility studies (ATAC-seq). After quality control all samples were sequenced on an Illumina HiSeq 4000 before further processing and data analysis. Global assessments of DNA (hydroxy-)methylation, DNA accessibility, and histone modifications between ASXL1 MT and WT CMML were performed. The samples in the two groups were treated as biological replicates and subjected to a consensus peak calling strategy requiring an overlap of at least 30% between samples and an adjusted p-value < 5x10-5 for a signal peak to be considered statistically significant. Differential gene expression was estimated to define the up-regulated genes in ASXL1 MT CMML. Potential CREs were defined as sites with statistically significant signal peaks overlapping in at least two of the three epigenomic marks: H3K4me1, 5hmC, and ATAC. Potential DEs were defined as CREs in non-coding regions outside promoter regions (defined as transcription start site ±3kb) that were annotated in GeneHancer. Annotated DEs only present in ASXL1 MT but not WT CMML (specific DEs) were intersected with the list of up-regulated genes and the ReMap atlas. Results: Sixteen WHO-defined CMML patients were included, median age 69 years (48 - 77), 63% male; of which 8 patients (50%, all truncating frame shift mutations) were ASXL1 MT and 8 (50%) WT. The burden and spectrum of co-mutations was similar between ASXL1 WT and MT CMML (21 versus 23 per group; p = 0.684; heatmap). There was a predominant up-regulation of gene expression in ASXL1 MT CMML: 707 genes up- and 124 down-regulated (volcano plot, FDR < 0.050 for all genes). There were 64336 potential CREs, the vast majority (97%) being present in both ASXL1 MT and WT CMML (left Venn diagram). These CREs were most commonly located in introns, promoter regions, and distal non-coding regions (bar graph and pie chart). There were 1303 CREs unique to ASXL1 MT CMML (specific DEs), 1161 (90%) of which were annotated in GeneHancer (left Euler diagram). Of these 1161 annotated specific DEs 859 (74%) were located outside promoter regions and 34 (4%) of them were known to be associated with genes up-regulated in ASXL1 MT CMML (Euler diagrams). These specific DEs were characterized by an increase in H3K4me1 occupancy and DNA accessibility (average signal tracks, purple bars indicating annotated DEs, thin bars below peaks indicating statistical significance). We previously observed epigenomic modification of promoter regions in 519 of the 707 up-regulated genes (73%) facilitating transcriptional activity in ASXL1 MT CMML. For 13 of the up-regulated genes (right Venn diagram, blue genes in volcano plot) the specific DEs were the sole identified mechanism, while for the other 21 genes there were additional mechanisms noted in the promoter region. The top five transcription factor candidates binding the 34 specific DEs included JMJD1C, MYC, KDM5B, RCOR1, and HDAC2 (-log10(E) > 40 for all candidates). Conclusions: Using a multiomics approach based on H3K4me1, 5hmC, and ATAC data we identified potential CREs in ASXL1 MT CMML and characterized potential DEs using publicly available annotation data. Specific DEs were associated with up-regulated genes serving as a possible explanation for the observed transcriptional activity, shedding further light on the adverse prognostic impact associated with ASXL1 mutations. Figure 1 Disclosures Patnaik: Stem Line Pharmaceuticals.: Membership on an entity's Board of Directors or advisory committees.