A novel LGALS1-depended and immune-associated fatty acid metabolism risk model in acute myeloid leukemia stem cells

Abstract

Leukemia stem cells (LSCs) are recognized as the root cause of leukemia initiation, relapse, and drug resistance. Lipid species are highly abundant and essential component of human cells, which often changed in tumor microenvironment. LSCs remodel lipid metabolism to sustain the stemness. However, there is no useful lipid related biomarker has been approved for clinical practice in AML prediction and treatment. Here, we constructed and verified fatty acid metabolism-related risk score (LFMRS) model based on TCGA database via a series of bioinformatics analysis, univariate COX regression analysis, and multivariate COX regression analysis, and found that the LFMRS model could be an independent risk factor and predict the survival time of AML patients combined with age. Moreover, we revealed that Galectin-1 (LGALS1, the key gene of LFMRS) was highly expressed in LSCs and associated with poor prognosis of AML patients, and LGALS1 repression inhibited AML cell and LSC proliferation, enhanced cell apoptosis, and decreased lipid accumulation in vitro. LGALS1 repression curbed AML progression, lipid accumulation, and CD8⁺ T and NK cell counts in vivo. Our study sheds light on the roles of LFMRS (especially LGALS1) model in AML, and provides information that may help clinicians improve patient prognosis and develop personalized treatment regimens for AML.

Fig. 1. Identification of subtypes and construction of the LFMRs in AML.

The survival analyses of ssGSEA score of fatty acid metabolism-related gene sets from KEGG (A), Hallmark (B), Reactom (C) and Wp (D) in the TCGA-LAML cohort. E The differential expression genes between LSCs and HSCs in the three databases (GSE17054, GSE68172, GSE24395) were shown in a heatmap. Red represents the significantly upregulated genes in LSCs compared with HSCs. Blue represents the significantly downregulated genes in LSCs compared with HSCs. Function and pathway enrichment analysis of the significantly upregulated genes (F) and downregulated genes (G) in LSCs versus HSCs by Metascape. The image shows the histogram of the top 20 enriched pathway. H The Venn diagrams were used to screen the differential expression related with fatty acid metabolism in LSCs. I Univariate COX regression analysis of the 9 potential genes in TCGA-LAML. J Consensus matrix when k = 4. K Kaplan-Meier OS curves for AML patients among C1, C2, C3, and C4 in TCGA-LAML. This table under Kaplan-Meier OS curves shows that the remaining patients who do not have the end point event (death) under the indicated time in this subgroup. They are at risk of an endpoint event, which is called number at risk. L The expression of LFMGs among C1, C2, C3, and C4. M Heatmap of correlation between lipid metabolism-associated genes in LSCs with clinicopathological characteristics of AML patients in the TCGA-LAML cohort. N Lasso COX regression analysis of five OS-related genes. O The overall survival (OS) in the TCGA-LAML cohort database was analyzed by the univariate COX regression with the 5 potential genes and summarized in Forest plots. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns, not significant.

Fig. 2. Verification of prognostic model (LFMRS) in AML.

A Kaplan-Meier analysis of OS between high- and low-LFMRS groups in the TCGA cohort. B The 1-, 2- and 3-year ROC curves of the LFMRS in the TCGA cohort. Kaplan-Meier analyses of OS between high- and low-LFMRS groups in the BeatAML database (C), GSE71014 (D), GSE12417 (E) and GSE37642 (F) cohorts. G Overview of the correspondence between LFMRS and other clinical features of AML patients. LFMRS expression among distinct clusters (H), between alive and dead patients (I), between <60 and ≥60 patients (J), among distinct FAB subtypes (K), among different cytogenetic risks (L) of AML. M The correlation of LFMRS with WBC counts. N The correlation of LFMRS with IC50 values of first-line drugs from GDSC database quantified by pRRophetic. O Correlation of LFMRS with immune cells quantified by CIBERSORT, CIBERSORT-ABS, QUANTISEQ, MCPCOUNTER, XCELL and EPIC in AML. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns not significant.

Fig. 3. LGALS1 is highly expressed in LSCs and associated with poor prognosis of AML patients.

A The transcript levels of LGALS1 in AML samples compared with that in healthy individuals were identified from TNMplot database. B The mRNA levels of LGALS1 in primary AML cases (n = 13, AML#1-AML#13) and healthy control cases (n = 13). C The protein levels of LGALS1 in indicated primary AML cases and healthy control cases. D The relative mRNA expression of LAGLS1 from the single cell sequencing data of GSE116256. E The mRNA levels of LGALS1 in HSCs and LSCs from healthy control cases and AML cases, respectively (LSCs from AML#1-AML#16). F Kaplan-Meier plots of overall survival in TCGA cohorts for AML patients, stratified on the basis of LGALS1 expression above (LGALS1^high) or below (LGALS1^low) the median. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns, not significant.

Fig. 4. LGALS1 promotes cell proliferation and inhibits cell apoptosis of LSCs.

A–D LSCs were sorted from AML#17. A Efficiencies of LGALS1 silence in LSCs were determined by qRT-PCR. B Cell growth was determined by colony formation assay under a light microscope, and the percentage of colony formation units were shown. C Cell apoptosis was determined by flow cytometric analysis. D Cell cycle distribution was detected by flow cytometric analysis via Ki67 staining (upper) and EdU staining (lower), respectively, and the bar graph showed the percentage of G0/G1, S, and G2/M phase cells. E–H LSCs were sorted from AML#18. E Efficiencies of LGALS1 silence in LSCs were determined by qRT-PCR. F Cell growth was determined by colony formation assay under a light microscope, and the percentage of colony formation units were shown. G Cell apoptosis was determined by flow cytometric analysis. H Cell cycle distribution was detected by flow cytometric analysis via Ki67 staining (upper) and EdU staining (lower), respectively, and the bar graph showed the percentage of G0/G1, S, and G2/M phase cells. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns not significant.

Fig. 5. LGALS1 plays a key role in lipid metabolism reprogramming of LSCs.

A–D LSCs transfected with shRNA against GALS1, or treated with DMSO or OTX008 were cultured. A, B The mRNA levels of CD36 and PPAR-γ were detected by qRT-PCR. C The protein levels of CD36 were determined by FCM using anti-CD36-APC (1:100, BioLegend, America). D Representative images of Oil Red O staining. Differentiated 3T3-L1 cell was used as a positive control. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns not significant.

Fig. 6. LGALS1 enhances lipid metabolism reprogramming, an immunosuppressive microenvironment, and AML progression in vivo.

A Kaplan-Meier analysis of the survival curves of the mice in each group (n = 5). B The percentage of GFP⁺ leukemia cell in bone marrow were detected through flow cytometric analysis (n = 3). C Immature cells from the bone marrow were checked using Wright’s stain (Left), and spleen and liver infiltration were analyzed by H&E staining (Right). The representative pictures were shown. D The relative expression of CD36 and PPAR-γ in GFP⁺ leukemia cells were determined by qRT-PCR (n = 3). E Representative images of Oil Red O staining. F The schema chart of MLL-AF9-induced leukemia was shown (n = 3). G The percentage of GFP⁺ leukemia cell in peripheral blood were detected through flow cytometric analysis (n = 3). H Kaplan-Meier analysis of the survival curves of the mice in each group (n = 7). I Immature cells from the bone marrow were checked using Wright’s stain (Left), and spleen and liver infiltration were analyzed by H&E staining (Right). The representative pictures were shown. J The proportion of LSCs in bone marrow was detected by flow cytometric analysis (n = 3). K The relative expression of CD36 and PPAR-γ in GFP⁺ leukemia cells were determined by qRT-PCR (n = 3). L Representative images of Oil Red O staining. M The proportion of CD8⁺ T cells and NK cells in bone marrow were determined by flow cytometric analysis (n = 3). *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns, not significant.