Whole-genome CRISPR screening identifies N-glycosylation as a genetic and therapeutic vulnerability in CALR-mutant MPNs

Abstract

Calreticulin (CALR) mutations are frequent, disease-initiating events in myeloproliferative neoplasms (MPNs). Although the biological mechanism by which CALR mutations cause MPNs has been elucidated, there currently are no clonally selective therapies for CALR-mutant MPNs. To identify unique genetic dependencies in CALR-mutant MPNs, we performed a whole-genome clustered regularly interspaced short palindromic repeats (CRISPR) knockout depletion screen in mutant CALR-transformed hematopoietic cells. We found that genes in the N-glycosylation pathway (among others) were differentially depleted in mutant CALR-transformed cells as compared with control cells. Using a focused pharmacological in vitro screen targeting unique vulnerabilities uncovered in the CRISPR screen, we found that chemical inhibition of N-glycosylation impaired the growth of mutant CALR-transformed cells, through a reduction in MPL cell surface expression. We treated Calr-mutant knockin mice with the N-glycosylation inhibitor 2-deoxy-glucose (2-DG) and found a preferential sensitivity of Calr-mutant cells to 2-DG as compared with wild-type cells and normalization of key MPNs disease features. To validate our findings in primary human cells, we performed megakaryocyte colony-forming unit (CFU-MK) assays. We found that N-glycosylation inhibition significantly reduced CFU-MK formation in patient-derived CALR-mutant bone marrow as compared with bone marrow derived from healthy donors. In aggregate, our findings advance the development of clonally selective treatments for CALR-mutant MPNs.

Graphical abstract

Figure 1.

Whole-genome CRISPR knock-out depletion screen identifies genes differentially required for the growth of mutant CALR-transformed BA/F3-MPL cells. (A) Experimental setup of the whole-genome CRISPR depletion screen on BA/F3 cells. n = 2 independent biological replicates each. (B) Growth curve of BA/F3-MPL cells after splitting into 2 culturing conditions at day 7 (+IL3/−IL3). (C) Volcano plot depicting significance and fold change of depleted genes, separated by the conditions stated, highlighting the most significant depleted genes for each condition. Log₂ fold change threshold = ±1. FDR adjusted P < .05. (D) Genes involved in protein glycosylation among the 10 most significantly depleted genes comparing CALR^Δ52 −IL3 vs EV +IL3 are shown, ranked by log₂ fold change. FDR, false discovery rate; neg. LFC, negative log₂ fold change; LV, lentivirus.

Figure 2.

Whole-genome CRISPR depletion screen identifies pathways differentially required for the growth of mutant CALR-transformed BA/F3-MPL cells, validated in a secondary CRISPR pooled screen. GSEA on the whole-genome CRISPR screen showing that genes in the N-glycan biosynthesis pathway (A), the protein secretion pathway (B), as well as the UPR pathway (C) are differentially depleted in CALR^Δ52 −IL3 cells as compared with EV plus IL3 cells. GSEA ranking for the pathways indicated was performed with the genes rank-ordered based on fold change. Genes that are more depleted in EV +IL3 condition are represented on the left and genes that are more depleted in CALR^Δ52 −IL3 condition are represented on the right. The comparison for CALR^Δ52 −IL3 vs CALR^Δ52 +IL3 is shown on the right side of each panel. (D-F) Results of the CRISPR pooled screen for validation of the whole-genome CRISPR screen. (D) Gene ranking, comparing CALR^Δ52 −IL3 with empty vector +IL3. The 10 most differentially depleted genes for CALR^Δ52 (as compared with EV +IL3 cells) are shown, ranked by corrected P values. Genes involved in protein glycosylation are highlighted in dark blue. (E) Volcano plot depicting significance and fold change of depleted genes, separated by the conditions stated, highlighting Dpm2 and important control genes. Log₂ fold change threshold = ±1. FDR adjusted P threshold = .05. (F) Venn diagram depicting the significantly depleted genes, comparing overlapping hits in the comparisons stated. FDR, false discovery rate; neg. LFC, negative log₂ fold change; NES, normalized enrichment score.

Figure 3.

In-depth validation of the N-glycan biosynthesis pathway as an essential pathway for growth of mutant CALR-expressing hematopoietic cells. (A) Experimental setup: parental BA/F3 cells were infected with MPL-expressing virus, selected for 21 days with hygromycin, and infected with either EV- or CALR^Δ52-expressing virus carrying GFP. GFP⁺ cells were subsequently sorted and transduced with Cas9-carrying virus. Following 7 days of puromycin selection (1 mg/mL), cells were then infected with RFP-expressing viruses containing either 1 of 2 NT sgRNAs (NTG) or 1 of 2 sgRNAs directed against Dpm2. RFP⁺ cells were sorted and subsequently subjected to functional assays. N = 2 independent biological replicates for BA/F3-MPL-EV and -CALR^Δ52-Cas9. Two NTGs and 2 targeting sgRNAs per biological replicate with a total of n = 4 biological replicates per genotype. (B) Growth curves of independent biological replicates and the 2 different sgRNAs were combined in the analysis. Cells were assayed either in the presence or absence of IL3 (+IL3/−IL3) for up to 96 hours. The assay was performed n = 3 for all 4 biological replicates of both genotypes. Statistical significance was determined by 2-way analysis of variance (ANOVA). Mean plus and minus standard error of the mean (SEM). ****P < .00001. The most important statistical analysis is highlighted. (C) Cell surface expression for MPL of the cells used in (B), determined by flow cytometry. (D) phosphorylated STAT5 levels 24 hours upon withdrawal of IL3 of indicated cell lines. (C-D) Statistical significance was determined by 1-way ANOVA. Mean plus SEM. **P < .01; ****P < .0001. (E) NGS results of the CRISPR-targeted regions in the Dpm2 gene in the cell lines depicted. In the CALR^Δ52 Dpm2-targeted −IL3 condition (96 hours post IL3 withdrawal), there is a strong enrichment for Dpm2 WT subclones as compared with the CALR^Δ52 Dpm2-targeted +IL3 condition, indicating that Dpm2 is required for the growth of CALR^Δ52-transformed cells. Mean plus SEM. (F) N-glycan profile of immunoprecipitated MPL from BA/F3-MPL-EV^Δ52-Cas9-NTG and BA/F3-MPL-EV^Δ52-Cas9-ΔDpm2 cells. (G) Growth curve of BA/F3-MPL-CALR^Δ52-Cas9-ΔDpm2-WT and -Dpm2-sgRNA-resistant (sgR) cells grown for 96 hours upon IL3 withdrawal. N = 3 in duplicate. Statistical significance was determined by 1-way ANOVA. ****P < .0001. Mean plus or minus SEM. (H) Colony-forming assays on RFP-sorted CALR^Δ52 VaviCre Cas9 BM transduced with NTG or Dpm2-targeting (ΔDpm2) RFP virus. N = 4 to 5 in technical duplicates. Statistical significance was determined by 1-way ANOVA. Mean plus SEM. ***P < .001. a.u., arbitrary units; esc, escapee: these are cells that survived IL3 withdrawal for 96 hours.

Figure 4.

N-glycosylation inhibitors preferentially reduce growth of mutant CALR-transformed cells and reduce MPL surface expression. (A) Overview of N-glycosylation biosynthesis and endoplasmic reticulum (ER)-resident trimming proteins (labeled in black) differentially required for growth of CALR^Δ52-transformed cells (compared with EV or CALR^Δ52 cells grown in the presence of IL3) as determined by scoring as a “hit” in either of our 2 CRISPR screens (negative log₂ fold change > 1). After biosynthesis of the N-glycan tree (dolichol-P-P-GlcNAc₂Man₉Glc₃), it is transferred to a newly synthesized glycoprotein such as MPL (red line) by the OST complex. OST complex members that scored in our screens include OST4, OSTC, STT3A, RPN1, RPN2, and DAD1. CALR then binds and folds the glycoprotein. The graph also depicts the mode of action of N-glycosylation inhibitors (labeled in red) used in B-I. (B) Growth curves of BA/F3-MPL-EV or -CALR^Δ52–expressing cells in the presence or absence of IL3, respectively. Cells were treated with DMSO, 10 μM or 50 μM 2-deoxyglucose (2-DG). Data were normalized to DMSO control for each cell line. (C) Normalized geometric mean MPL (CD110) cell surface expression of BA/F3-MPL-EV, -CALR^Δ52, or –Jak2^V617F–expressing cells following 2-DG treatment for 24 hours. EV control cells were grown in the presence of IL3, CALR^Δ52, and Jak2^V617F cells in the absence of IL3. (D-I) Growth curves and MPL surface expression as described in panels B-C. DMSO, 1 μM tunicamycin (D-E), DMSO, 1 μM or 10 μM ML414, 5-[(Dimethylamino)sulfonyl]-N-(5-methyl-2-thiazolyl)-2-(1-pyrrolidinyl)-benzamide (NGI1) (F-G), and DMSO, 20, or 50 μM castanospermine (H-I). Statistical significance in panels B, D, F, and H was calculated using 2-sided Student t tests at the 72-hour time point. Growth curves were performed in duplicate 2 to 4 times. Statistical significance in panels C, E, G, and I was determined using 1-way analysis of variance. MPL measurements were performed in duplicate 2 to 3 times. The most important statistical analyses are highlighted. Mean plus or minus standard error of the mean. *P < .05; **P < .01; ***P < .001; ****P < .0001. Fruc, fructose; Glc, glucose; Man, Mannose; mRNA, messenger RNA; P, phosphate.

Figure 4.

N-glycosylation inhibitors preferentially reduce growth of mutant CALR-transformed cells and reduce MPL surface expression. (A) Overview of N-glycosylation biosynthesis and endoplasmic reticulum (ER)-resident trimming proteins (labeled in black) differentially required for growth of CALR^Δ52-transformed cells (compared with EV or CALR^Δ52 cells grown in the presence of IL3) as determined by scoring as a “hit” in either of our 2 CRISPR screens (negative log₂ fold change > 1). After biosynthesis of the N-glycan tree (dolichol-P-P-GlcNAc₂Man₉Glc₃), it is transferred to a newly synthesized glycoprotein such as MPL (red line) by the OST complex. OST complex members that scored in our screens include OST4, OSTC, STT3A, RPN1, RPN2, and DAD1. CALR then binds and folds the glycoprotein. The graph also depicts the mode of action of N-glycosylation inhibitors (labeled in red) used in B-I. (B) Growth curves of BA/F3-MPL-EV or -CALR^Δ52–expressing cells in the presence or absence of IL3, respectively. Cells were treated with DMSO, 10 μM or 50 μM 2-deoxyglucose (2-DG). Data were normalized to DMSO control for each cell line. (C) Normalized geometric mean MPL (CD110) cell surface expression of BA/F3-MPL-EV, -CALR^Δ52, or –Jak2^V617F–expressing cells following 2-DG treatment for 24 hours. EV control cells were grown in the presence of IL3, CALR^Δ52, and Jak2^V617F cells in the absence of IL3. (D-I) Growth curves and MPL surface expression as described in panels B-C. DMSO, 1 μM tunicamycin (D-E), DMSO, 1 μM or 10 μM ML414, 5-[(Dimethylamino)sulfonyl]-N-(5-methyl-2-thiazolyl)-2-(1-pyrrolidinyl)-benzamide (NGI1) (F-G), and DMSO, 20, or 50 μM castanospermine (H-I). Statistical significance in panels B, D, F, and H was calculated using 2-sided Student t tests at the 72-hour time point. Growth curves were performed in duplicate 2 to 4 times. Statistical significance in panels C, E, G, and I was determined using 1-way analysis of variance. MPL measurements were performed in duplicate 2 to 3 times. The most important statistical analyses are highlighted. Mean plus or minus standard error of the mean. *P < .05; **P < .01; ***P < .001; ****P < .0001. Fruc, fructose; Glc, glucose; Man, Mannose; mRNA, messenger RNA; P, phosphate.

Figure 5.

2-Deoxyglucose normalizes key MPN features in mice. (A) Schematic overview of 2-DG treatment in primary mice. (B-C) Spleen N-glycan analyses in vehicle- or 2-DG–treated Calr^Δ52/+ mice upon 18 days of treatment, showing the top 10 most abundant N-glycans (B) and the immature high-mannose N-glycans (C). Statistical analysis performed using 1-way analysis of variance (ANOVA). Mean plus standard error of the mean (SEM). *P < .05; **P < .01. (D) Peripheral blood platelet count of vehicle- or 2-DG–treated Calr^+/+ and Calr^Δ52/+ mice 18 days after treatment start. Statistical analysis performed using 1-way ANOVA. Mean plus SEM. ***P < .001; ****P < .0001. (E) Megakaryocyte-erythroid progenitor (MEP) frequency. N = 7 per genotype and condition. Statistical analysis performed using 1-way ANOVA. Mean plus SEM. *P < .05. (F) MPL surface expression (normalized geometric mean) on megakaryocyte progenitors of vehicle- or 2-DG–treated Calr^+/+ and Calr^Δ52/+ mice. N = 3. Statistical analysis performed using 1-way ANOVA. Mean plus SEM. *P < .05. (G-H) GSEA on RNA-seq data of MEPs isolated from Calr^+/+ MxCre and Calr^Δ52/+ MxCre mice treated with vehicle or 2-DG. (G) GSEA showing enrichment of the Hallmark apoptosis pathway in Calr^Δ52/+ MxCre MEPs from mice treated with 2-DG as compared with vehicle, whereas the opposite was found for Calr^+/+ MxCre mice treated with 2-DG as compared with vehicle, shown in (H). (I) Platelet (PLT) values of Jak2^V617F/+ MxCre and Jak2^+/+ MxCre mice treated with 2-DG for 14 days. N = 3. Statistical analysis performed using 1-way ANOVA. *P < .05. (J) MPL surface expression (normalized geometric mean) on megakaryocyte progenitors of vehicle- or 2-DG–treated Jak2^V617F/+ MxCre and Jak2^+/+ MxCre mice. N = 3. Mean plus SEM. Statistical analysis performed unpaired Student t test. *P < .05. CBC, complete blood cell count; FDR, false discovery rate; NES, normalized enrichment score.

Figure 6.

2-Deoxyglucose preferentially targets Calr mutant cells in a preclinical MPN mouse model. (A) Schematic overview of chimeric transplantation experiment. CD45.1⁺ competitor BM cells were mixed in a 1:1 ratio with CD45.2⁺ Calr^Δ52/+ MxCre UBC-GFP BM and transplanted into lethally irradiated CD45.1⁺ recipient animals. N = 5 to 7 mice per group. (B-D) Peripheral blood values of engrafted and pI:pC-induced vehicle or 2-DG–treated chimeric mice as described in (A). (B) Platelet (PLT) count over time. (C) WBC over time. (D) GFP chimerism in peripheral blood platelets over time. (E) Calr^Δ52/+ mutant chimerism (percentage of GFP⁺ cells) in long-term hematopoietic stem and progenitor cells in the bone marrow following 6 weeks of treatment. Statistical significance was determined using Skidak’s multiple comparisons tests (B-D) or 2-sided Student t tests (E). Mean plus or minus standard error of the mean. **P < .01. CBC, complete blood cell count; LT-HSC, long-term hematopoietic stem cell.

Figure 7.

N-glycosylation–related pathways are enriched in platelets of CALR-mutant human MPN as compared with human healthy control platelets. (A) Significantly enriched glucose and glycosylation-related pathways in platelets of CALR-mutant essential thrombocythemia and myelofibrosis patients (n = 13 samples) vs healthy controls (n = 21 samples) as determined by GSEA on platelet RNA-seq data. (B) GSEA of the Reactome N-glycan trimming in the ER and calnexin calreticulin cycle pathway in platelets of CALR-mutant vs control platelets. (C) GSEA of the KEGG fructose and mannose metabolism pathway in platelets of CALR-mutant vs control platelets. (D-E) CFU-MK of HC BM and BM from patients with a CALR mutation. N = 3 in duplicate. Mean plus or minus standard error of the mean. Statistical analysis performed using Student t tests. *P < .05; **P < .01. (D) Normalized colony count of BM grown in the presence of either DMSO or 2-DG (20 µM). (E) Normalized colony count of BM grown in the presence of either DMSO or NGI1 (1 µM). Mean plus standard error of the mean. CFU-MK, megakaryocyte colony-forming unit; FDR, false discovery rate; NES, normalized enrichment score.