Critical role of Jumonji domain of JMJD1C in MLL-rearranged leukemia

Abstract

JMJD1C, a member of the lysine demethylase 3 family, is aberrantly expressed in mixed lineage leukemia (MLL) gene-rearranged (MLLr) leukemias. We have shown previously that JMJD1C is required for self-renewal of acute myeloid leukemia (AML) leukemia stem cells (LSCs) but not normal hematopoietic stem cells. However, the domains within JMJD1C that promote LSC self-renewal are unknown. Here, we used clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated protein-9 nuclease (Cas9) negative-selection screening and identified a requirement for the catalytic Jumonji (JmjC) domain and zinc finger domain for leukemia cell survival in vitro and in vivo. In addition, we found that histone H3 lysine 36 methylation (H3K36me) is a marker for JMJD1C activity at gene loci. Moreover, we performed single cell transcriptome analysis of mouse leukemia cells harboring a single guide RNA (sgRNA) against the JmjC domain and identified increased activation of RAS/MAPK and the JAK-STAT pathway in cells harboring the JmjC sgRNA. We discovered that upregulation of interleukin 3 (IL-3) receptor genes mediates increased activation of IL-3 signaling upon JMJD1C loss or mutation. Along these lines, we observed resistance to JMJD1C loss in MLLr AML bearing activating RAS mutations, suggesting that RAS pathway activation confers resistance to JMJD1C loss. Overall, we discovered the functional importance of the JMJD1C JmjC domain in AML leukemogenesis and a novel interplay between JMJD1C and the IL-3 signaling pathway as a potential resistance mechanism to targeting JMJD1C catalytic activity.

Graphical abstract

Figure 1.

Negative-selection screen using CRISPR/Cas9. (A) Schematics of the screen. Vector used to establish a clonal Cas9-expressing MLL-AF9 leukemia cell line and vector used for sgRNA transduction (upper). Experimental scheme (lower). (B) Fold changes in TdTomato positivity (day 2/day 14) during 14 days in culture. Each bar represents an sgRNA targeting Jmjd1c. The black bar represents sgRNA against Dot1l. Shaded areas are sgRNAs targeting ZFD (green) and Jumonji domain (blue). (C-F) Time course of flow cytometry analysis of TdTomato level after lentiviral transduction of sgRNAs. (B-F) Data are mean ± standard deviation of independently transduced triplicate samples.

Figure 2.

Effect of mutating JMJD1C ZFD and Jumonji domain on MLL-AF9 leukemogenesis. (A-B) CFC assay after transduction of MLL-AF9-Cas9 cells with sgRNAs against ZFD (e12.3, e12.4) and Jumonji domain (e20.1, e21.1). (A) Colonies in methylcellulose culture (upper panels). Individual colony morphology (original magnification ×80; lower panels). (B) Colony counts 1 week after plating. (C) Proliferation assay of sorted GFP⁺ (marks MLL-AF9) TdTomato⁺ (marks sgRNA) MLL-AF9-Cas9 leukemia cells after sgRNA transduction. Data are mean ± standard deviation (SD) of 3 independent experiments. The P value was determined by 2-way analysis of variance (ANOVA). For multiple comparisons, P < .05 for day 7 and P < .0001 for day 9 between Renilla and all other sgRNAs, with the exception of Renilla vs e12.3 on day 7 (not statistically significant). (D) Apoptosis assay (left panel). Data are mean ± SD from 3 independent experiments. Representative flow plot; numbers within the upper left quadrant indicate average percentage (right panel). (E) Representative flow cytometry analysis of Mac1 of MLL-AF9-Cas9 leukemia cells 7 days after transduction (left panel). Mean fluorescence intensity (MFI; geometric mean) of Mac1 expression (right panel). Data are mean ± SD from 3 independent experiments. (F) Survival curves of secondary recipient mice that received MLL-AF9-Cas9 cells after transduction with respective sgRNAs. P = .040 overall, log-rank test. Pairwise log-rank test P values between Renilla and JMJD1C sgRNAs: P = .025 for e12.3; P = .137 for e12.4; P = .096 for e20.1, and P = .048 for e21.1. (G) Peripheral blood GFP⁺TdTomato⁺ cell percentage 8 weeks after transplantation. P = .05, multiple comparison, 1-way ANOVA. Renilla vs e12.3, P = .06; Renilla vs e12.4, e20.1, and e21.1, P = .08 . *P < .05, **P < .01, ***P < .001, unpaired 2-tailed Student t test.

Figure 3.

Changes in histone modifications upon loss of JMJD1C. (A) Representative snapshots of RNA-seq and H3K36me3 ChIP-seq (triplicate and duplicate biological replicates, respectively; 1 representative replicate is shown) results in Jmjd1c sgRNA vs Renilla sgRNA cells: Irf8 and Gas7 are genes repressed by JMJD1C, and Actb is a control gene. (B) Heat map of differentially expressed genes (adjusted P < .05 and fold change > 1.5) between MLL-AF9 cells harboring Renilla vs JMJD1C sgRNAs. (C) Venn diagram of overlapping differentially regulated genes in the knockout and sgRNA RNA-seq data set. P < 8.5⁻⁹⁷, exact hypergeometric test. GSEA enrichment plots of differentially regulated genes between Jmjd1c^f/f and Jmjd1c^−/− cells (D), indicated gene sets (E, top and bottom) in MLL-AF9 cells harboring Renilla (Renilla) vs JMJD1C sgRNAs (REST) RNA-seq data. (F) GSEA enrichment plots of differentially regulated genes between MLL-AF9 cells harboring Renilla vs JMJD1C sgRNAs in H3K36me3 ChIP-seq data of Renilla vs JMJD1C sgRNAs (REST).

Figure 4.

Single-cell transcriptome analysis of JMJD1C Jumonji domain–mutated MLL-AF9 leukemia cells. (A-B) T-distributed stochastic neighbor embedding plot of single-cell gene-expression data of mouse MLL-AF9-Cas9 leukemia cells 7 days after transduction with sgRNA against Renilla or JMJD1C JmjC domain (Jumonji). (A) Cell phenotype. (B) Cell clusters identified within phenotypes. The numbers of cells are indicated in parenthesis. (C) Cyclone cell cycle status. (D) Expression level of Somerville LSC signature. (E) Monocle single-cell trajectory analysis. (F) SCENIC regulon analysis. Black bars indicate that a regulon is activated within a cell, across all clusters identified in panel B. (G-H) Decomposed Z score of MAST analysis on C6 oncogenic signature from the Molecular Signature Database (all Renilla cells vs all Jumonji cells). (G) Pink: enrichment in Jumonji; green: enrichment in Renilla sample, and hematopoietic fingerprint gene sets (Jumonji 2 cells vs Jumonji 3 cells). (H) Pink: enrichment in Jumonji 2; green: enrichment in Jumonji 3. GSVA analysis of top KRAS pathway (I; identified in panel G) and RAS/MAPK and JAK-STAT pathways (J). P < .01 for all pairwise comparisons in panel D, with the exception of Renilla 1 vs Renilla 2 (not significant), unpaired 2-tailed Student t test with Bonferroni correction. Gran, granulocytes; Mye, myeloid; Mono, monocytes; NK, natural killer cells.

Figure 5.

Increased IL-3 signaling in MLL-AF9 cells with mutated Jumonji domain and ZFD. (A) Western blotting of phospho- and total ERK1/2 and STAT5 in MLL-AF9-Cas9 leukemia cells 6 days after transduction with respective sgRNAs. Dose response curve (4-variable slope model) in the presence of RAS/ERK inhibitors (B) or JAK inhibitors (C). Two or 3 independent experiments, each done with technical triplicates. Shown are representative experiments (mean ± standard deviation [SD] of technical triplicates). Bar graphs on the far right show EC₅₀ (mean ± SD of 2-3 independent experiments) of inhibitors normalized to that of Renilla control. Numbers on top of the bar are average EC₅₀ (nM). *P < .05, **P < .01, unpaired 2-tailed Student t test. (D) Violin plot of relative gene expression data of Il3ra/Csf2rb/b2 derived from single-cell transcriptome. *****Adjusted P < .00001 by MAST. (E) Competitive cell-proliferation assay using sgRNAs against JMJD1C in the presence or absence of sgRNA against Csf2rb (marked by BFP). Data are mean ± SD of percentage of BFP⁺ and TdTomato⁺ double-positive cells from independently transduced triplicate samples. Curves were fitted with an asymmetric sigmoidal 5 parameter logistical model. (F) Model of transcription regulation of IL-3 receptor gene by JMJD1C.

Figure 6.

Presence of activating RAS mutations renders cells resistant to JMJD1C loss. Fold change in JMJD1C sgRNA abundance (A) and ATARIS score (B) in leukemia cell lines. Cell lines in blue bear RAS mutations according to the Broad Institute Cancer Cell Line Encyclopedia. (C) MOLM13, MOMO-MAC-1, NOMO1, THP1, U937, and OCI-AML3 cells were infected with the indicated sgRNA-Cas9-TdTomato viruses and subjected to flow cytometry analysis over a period of 28 days. Fold changes in TdTomato positivity compared with day 3 are plotted. Data are mean ± standard deviation of independently transduced triplicate samples. (D) Fold depletion of sgRNAs in MOLM13 cells overexpressing NRAS^G12V and in mouse MLL-AF9-Cas9 cells overexpressing NRAS^G12V and KRAS^G12D. Data are mean ± standard deviation of independently transduced triplicate samples. (E) Western blotting of NRAS/KRAS expression in cells in panel D.