Phosphatidic acid phosphatase LPIN1 in phospholipid metabolism and stemness in hematopoiesis and AML

Abstract

Targeting metabolism represents a promising approach to eradicate leukemic stem cells (LSCs) that are considered critical drivers of relapse in acute myeloid leukemia (AML). In this study, we demonstrate that the phosphatidic acid phosphatase LPIN1, which regulates the synthesis of diacylglycerol, the key substrate for triacylglycerol, and phospholipid production, is crucial for the function of healthy and leukemic hematopoietic stem and progenitor cells (HSPC and LSC). LPIN1 mRNA was highly expressed in the CD34+ compartment of primary human AML samples. LPIN1 suppression inhibited the proliferation of primary leukemic cells and normal HSPCs in vitro and in xenotransplantation assays. Lipidomics analyses revealed a reduction of phosphatidylcholine (PC) and phosphatidylethanolamine and an upregulation of sphingomyelin upon LPIN1 depletion. Distinct phospholipid composition was associated with genetic AML groups, and targeting PC production by choline kinase inhibitors showed strong anti-leukemic activity. In summary, our data establish a regulatory role of LPIN1 in HSPC and LSC function and provide novel insights into the role of glycerophospholipid homeostasis in stemness and differentiation.

Figure 1

LPIN1 expression correlates with leukemic stem cell frequency and prognosis in acute myeloid leukemia. (A) Dot plot of LPIN1 mRNA expression in NK AML samples (n = 56) with low (n = 25, round symbol, lilac bar), medium (n = 20, upside triangle symbol, blue bar), or high (n = 10, square symbol, green bar) LSC frequency. LSC frequencies were determined in immunocompromised mice (NSG) via limiting dilution assays (LSChigh: LSC frequencies > 1 in 32,000; LSClow: LSC frequencies < 1 in 3 × 10⁶; LSCmedium: in between the groups). The data are part of the Leucegene project. Symbols represent individual samples, horizontal bars show median LPIN1 expression, and asterisks show results from Mann–Whitney U test. (B) Left panel: Schematic sketch of the sorting strategy of NK AML samples (n = 10); right panel: from left to right LPIN1, LPIN2, and LPIN3 mRNA expression in CD34–GPR56– (lilac), CD34–GPR56+ (blue), and CD34+GPR56+ sorted fractions (green). Asterisks show results from a repeated measures one‐way ANOVA with Geisser–Greenhouse correction and Holm‐Sidak's multiple comparison test with individual variances computed for each comparison. (C) LPIN1 mRNA expression in 22 paired primary AML samples from the Leucegene project at initial diagnosis (lilac) and relapse (blue) shown as stacked bar plot (left) and column dot plot (right). Paired t‐test. (D) Adapted Kaplan–Meyer plot showing overall survival in dependence of LPIN1 mRNA expression from TCGA PanCancerAtlas. Survival data with >1 day follow‐up time were available for 157 patients. Patient groups were divided by median LPIN1 expression. (E) Left: Heatmap showing expression of genes of the Kennedy pathway in different maturation stages during myelopoiesis using RNA‐seq data from the Leucegene project. Yellow indicates higher expression values and lilac indicates lower expression values (log10 RPKM). CB: cord blood, BM: bone marrow. Right: Graphic illustrating triacylglycerol and phospholipid synthesis in HSCs (lilac) and differentiated granulocytes (green). Bold letters and arrows indicate higher expression. (F) Western blot showing LPIN1 protein expression in OCI‐AML3 upon lentiviral transduction with shRNAs against LPIN1 or Luciferase (shLuc). Actin was used as a loading control. (G) Knockdown (KD) efficiency of three different shRNAs against LPIN1 compared to shLuc in OCI‐AML3 measured by q‐RT‐PCR (shLuc: lilac, shLPIN1.1159: blue, shLPIN1.452: turquoise, shLPIN1.957: green). Three replicates were used per condition. Data are shown as mean + SD. Symbols represent individual replicates. Asterisks show results from ordinary one‐way ANOVA corrected for multiple comparisons with Dunnett's multiple comparison test with a single pooled variance. (H) Left: Cell proliferation for OCI‐AML2 transduced with shLPIN1.1159 (blue), shLPIN1.452 (turquoise), or shLPIN1.957 (green) versus shLuc (lilac). Shown is the fold‐increase in absolute cell counts per well until day 14, normalized to the fourth day of the culture. For each condition, 8 replicates were started. Cells were counted by HTS‐FACS. Right: Proliferation of OCI‐AML3 transduced with shLPIN1.1159 (blue), shLPIN1.452 (turquoise), or shLPIN1.957 (green) versus shLuc (lilac). Shown is the fold‐increase in absolute cell counts per well until day 18, normalized to the third day of the culture. For each condition, eight replicates were started. Cells were counted by HTS‐FACS. (I) Apoptosis assay for OCI‐AML3 transduced with shLPIN1.1159 (blue) or shLPIN1.957 (green) versus shLuc (lilac). Shown is the fraction of apoptotic cells. Three replicates were used per condition. Data are shown as mean + SD. Symbols represent individual replicates. Asterisks show results from ordinary one‐way ANOVA corrected for multiple comparisons with Dunnett's multiple comparison test with a single pooled variance. DG, diacylglycerol; FA‐CoA, fatty‐acyl‐CoA; Gly3‐P, glycerol‐3‐phosphate; LPA, lyso‐phosphatidic acid; PA, phosphatidic acid; PC, phosphatidylcholine; PE, phosphatidylethanolamine; TG, triacylglycerol.

Figure 2

LPIN1 is required for AML cell proliferation, in vivo engraftment, and maintenance of the stem cell‐enriched CD34+ compartment. (A) Top: Proliferation of patient‐derived xenograft (PDX) AML cells transduced with shLPIN1.1159 (blue), shLPIN1.452 (turquoise), or shLPIN1.957 (green) versus shLuc (lilac). Shown is the fold‐increase in absolute cell counts per well normalized to day 5. For each condition, five replicates were started. Cells were counted by HTS‐FACS. A two‐way ANOVA was performed, followed by the Holm‐Sidak test to correct for multiple comparisons. Bottom: Bar plots indicating the CD34+ bright fraction of AM+ cells at the respective time points. A two‐way ANOVA was performed, followed by the Holm‐Sidak test to correct for multiple comparisons. (B) Representative FACS plots showing CD34 expression and FSC‐A of AML‐602 cells on day 15 after lentiviral transduction with the indicated shRNAs. Values indicate CD34+ bright fractions in percent. (C) Graphic illustrating the experimental setup of in vivo experiments. See “Materials and Methods” section for details. For initial gene transfer, see Table S4. Created with BioRender.com. (D) Proliferation of primary AML‐491 cells transduced with shLPIN1.1159 (blue) versus shLuc (lilac). Shown is the fold‐increase in absolute cell counts per well normalized to the fourth day of culture. For each condition, five replicates were started. Cells were counted by HTS‐FACS. (E) Left: Representative FACS plots showing CD34 and GPR56 expression of primary AML‐491 cells transduced with shLPIN1.1159 (bottom) versus shLuc (top) on day 12. Values indicate gated fractions in percent. Right: Bar plot showing CD34 and GPR56 expression of primary AML‐491 cells in vitro transduced with shLPIN1.1159 (blue) versus shLuc (lilac). For each condition, three replicates were measured. Data are shown as mean + SD. Symbols represent individual replicates. (F) Left panel: Representative FACS plots showing FSC‐A and ametrine expression of primary AML‐491 cells transduced with shLPIN1.1159 (bottom) versus shLuc (top) 21 weeks after transplantation into immunocompromised NSGW41 mice. Values indicate gated fractions in percent. Right panel: Dot plots showing the overall human leukemic engraftment (left) and the fraction of Ametrine‐positive cells harboring shRNAs against LPIN1 or control (shLPIN1.1159 versus shLuc) within total bone marrow 21 weeks after injection (right). Human CD45 and CD11b were used to detect overall leukemic engraftment of AML‐491 including non‐transduced Ametrine negative cells, which was similar in both conditions, thus excluding technical issues during injections. Each group consisted of seven mice represented by symbols. Mann–Whitney U test. (G) Left: Representative FACS plots showing CD34 expression and SSC‐A of primary AML‐491 cells transduced with shLPIN1.1159 (bottom) versus shLuc (top) 21 weeks after transplantation into immunocompromised NSGW41 mice. Values indicate gated fractions in percent. Right: Dot plot showing CD34 surface expression of primary AML‐491 cells transduced with shLPIN1.1159 (blue) versus shLuc (lilac) 21 weeks after transplantation. Each group consisted of seven mice represented by symbols. Mann–Whitney U test. (H) Left: Representative FACS plots showing CD14 and CD11b expression of primary AML‐491 cells transduced with shLPIN1.1159 (bottom) versus shLuc (top). Values indicate gated fractions in percent. Center: Dot plot showing the frequency of CD14 and CD11b co‐positive fractions of primary AML‐491 cells transduced with shLPIN1.1159 (blue) versus shLuc (lilac). Each group consisted of seven mice, and each symbol represents one mouse. Mann–Whitney U test. Right: Dot plot showing the mean intensity of CD11b in primary AML‐491 cells transduced with shLPIN1.1159 (blue) versus shLuc (lilac). Unpaired t‐test after passing normal distribution test.

Figure 3

Normal CD34+ HSPCs are susceptible to LPIN1 loss. (A) Proliferation curves for HSPCs transduced with shLPIN1.1159 (blue) or shLPIN1.957 (green) versus shLuc (lilac). Shown is the fold‐increase in absolute cell counts per well on day 21 normalized to the fifth day of the culture. For each condition, five replicates were started. Cells were counted by HTS‐FACS. (B) Colony‐forming cell assay of pooled HSPCs, which were transduced with shLPIN1.1159 (blue) or shLPIN1.957 (green) versus shLuc (lilac) as negative control. Ametrine‐positive cells were directly sorted into methylcellulose 72 h post‐transduction. 200 cells were plated per well and colonies were counted 10 days post‐plating. (C) Left: Representative FACS plots showing CD34 and CD45RA expression of HSPCs 10 days after transduction with shLPIN1.1159 (middle) and shLPIN1.957 (right) versus shLuc (left). Values indicate gated fractions in percent. Right: Bar plot of the measured CD34 and CD45RA expression of HSPCs 10 days after transduction with shLPIN1.1159 (blue) or shLPIN1.957 (green) versus shLuc (lilac). For each condition, three replicates were measured. Data are shown as mean + SD. Symbols represent individual replicates. (D) Graphic illustrating the experimental setup of in vivo experiments. For initial gene transfer, see Table S4. See “Materials and Methods” section for details. Created with BioRender.com. (E) Left: Representative FACS plots showing ametrine and CD45 expression of HSPCs transduced with shLPIN1.1159 (right) versus shLuc (left) 14 weeks after transplantation into immunocompromised mice. Values indicate gated fractions in percent. Right: Dot plot showing engraftment of HSPCs transduced with shLPIN1.1159 (blue) versus shLuc (lilac) 14 weeks after transplantation into immunocompromised mice. Each group consisted of seven mice represented by symbols. Mann–Whitney U test. (F) Left: Representative FACS plots showing CD19 and CD33 expression in HSPCs transduced with shLPIN1.1159 versus shLuc 14 weeks after transplantation into immunocompromised mice. Values indicate gated fractions in percent. Right: Dot plot showing the fraction of CD19 and CD33 positive output of engrafted HSPCs transduced with shLPIN1.1159 (blue) versus shLuc (lilac) 14 weeks after transplantation into immunocompromised mice. Each group consisted of seven mice represented by symbols. Mann–Whitney U test. BFU‐E, burst‐forming unit‐erythroid; CFU, colony‐forming unit; G, granulocyte; GEMM, granulocyte, erythrocyte, megakaryocyte, macrophage; GM, granulocyte/macrophage; M, macrophage.

Figure 4

LPIN1 regulates the expression of genes for glycerophospholipid metabolism in OCI‐AML3 and HSPCs. (A) Volcano plot showing the log2‐fold changes (log2FC) of gene expression (x‐axis) and log10‐transformed adjusted p values (y‐axis) for OCI‐AML3 from RNA‐seq data, 48 h after lentiviral transduction with shLPIN1.1159, shLPIN1.452, or shLPIN1.957 versus shLuc. Data points in gray represent genes without significant regulation. Data points highlighted in blue represent significantly downregulated genes (p ≤ 0.05, log2FC shLPIN1/shLuc < 0), while data points highlighted by red dots represent significantly upregulated genes (p ≤ 0.05, log2FC shLPIN1/shLuc > 0). Not all gene symbols are displayed due to space constraints. See Table S5 for full gene lists. (B) Bubble plot showing the fold enrichment of significantly regulated gene sets in OCI‐AML3 (top) and HSPCs (bottom). Pathway analysis was performed using Shiny GO and the hallmark.MSigDB database. (C) Heatmap illustrating the regulation of genes belonging to the “FATTY ACID METABOLISM” (top) and “MTORC1 SIGNALING” gene set (bottom) in OCI‐AML3 and HSPCs. Bold gene names indicate significant regulation in OCI‐AML3, italic gene names indicate significant regulation in HSPCs, and bold italic gene names indicate significant regulation in both. Fold enrichment is given as log2FC shLPIN1/shLuc. Upregulation is indicated by blue, whereas downregulation is indicated by red color. (D) Proliferation of grouped OCI‐AML3 single clones with either equal and more (blue), or less (turquoise) than 30% remaining LPIN1 protein after knockout (KO) using CRISPR/Cas9 compared to sgGFP single clones (lilac) as control. For each sgLPIN1 clone, six replicates were started. For sgGFP, eight single clones at six replicates were started as a reference. Cells were counted by HTS‐FACS. Data are given as mean + SEM. (E) Western blot showing LPIN1 overexpression upon lentiviral transduction of OCI‐AML3 with LPIN1 full length (FL) (middle), an enzymatically impaired version of LPIN1 (ΔPAP) (right), or empty vector (ev) (left). Actin was used as a loading control. (F) Proliferation of OCI‐AML3 sgLPIN1 single clone 8 (sc #8) after lentiviral transduction with LPIN1 full length (FL), the enzymatically impaired version of LPIN1 (ΔPAP), or empty vector (ev) compared to a sgGFP single clone (lilac) upon lentiviral transduction with ev as control. Shown is the fold‐increase in absolute cell counts per well normalized to the second day of the culture. For each condition, eight replicates were started. Cells were counted by HTS‐FACS. Data are given as mean + SD. Asterisks show results from two‐way ANOVA with Dunn's multiple comparison test.

Figure 5

LPIN1 regulates phosphatidylcholine and ‐ethanolamine levels in AML. (A) Bar plot showing the distribution of lipid classes in OCI‐AML3 after transduction with shLPIN1.452 (turquoise) or shLPIN1.957 (green) versus shLuc (lilac) measured by mass spectrometry. For each condition, four replicates were measured. Data are shown as mean. Symbols represent individual replicates. (B) Bar plot showing changes in phospholipid composition regarding PA chain length in OCI‐AML3 after transduction with shLPIN1.452 (turquoise) or shLPIN1.957 (green) versus shLuc (lilac) measured by mass spectrometry. For each condition, four replicates were measured. Data are shown as mean. Symbols represent individual replicates. (C) Bar plot showing the lipid composition of primary AML samples measured by mass spectrometry. Each point represents a sample. Data are shown as mean + SD. (D) XY‐plot displaying the relative abundance of phosphatidylcholine and sphingomyelin in primary AML samples, measured by mass spectrometry, in mol%. Pearson coefficient was calculated and a simple linear regression model using a 95% confidence interval was applied. (E) Bar plot showing the ratio of PC/SM in OCI‐AML3 after transduction with shLPIN1.452 (turquoise) or shLPIN1.957 (green) versus shLuc (lilac). Data are shown as mean + SD. Symbols represent individual replicates. Ordinary one‐way ANOVA corrected for multiple comparisons with Dunnett's multiple comparison test with a single pooled variance. (F) Heatmap showing gene expression of sphingomyelin synthases and hydrolases as well as LPIN1 in different maturation stages during myelopoiesis using RNA‐seq data from the Leucegene project. Yellow indicates higher expression values and lilac indicates lower expression values (log10 RPKM). (G) Graphic illustrating the connection between phospho‐ and sphingolipid metabolism. Metabolic processes that are assumed to be upregulated in HSCs and LSCs based on RNA‐seq data are indicated in pink, and processes that are assumed to be upregulated in differentiated cells are indicated in lilac. (H) Principal component analysis (PCA) plot showing PC1 and PC2 from lipidomic analyses in primary AML. The figure includes loading vectors radiating from the origin, representing the contributions of individual lipids to the principal components. Points represent individual samples. The mutational status of the samples is indicated by color: nucleophosmin 1 (NPM1c) mutation and internal tandem‐repeat of fms‐like tyrosine‐kinase 3 (FLT3‐ITD) are highlighted in purple, NPM1c mutation and mutation of the tyrosine kinase domain (TKD) of FLT3 are shown in pink, no mutation of NPM1 (NPM1wt) and FLT3‐ITD are indicated in turquoise, and NPM1c mutation and no mutation in FLT3 (FLT3wt) are shown in green. The shaded area is characterized by similar lipid composition and comprises all four samples with NPM1c/FLT3‐ITD co‐mutations. BM, bone marrow; CB, cord blood; CE, cholesterol ester; Cer, ceramide; Chol, cholesterol; DG, diacylglycerol; HexCer, hexosyl ceramide; Hex2Cer, di‐hexosyl ceramide; LPC, lysophosphatidylcholine; PA, phosphatidic acid; PA O‐, ether‐linked phosphatidic acid; PC, phosphatidylcholine; PC O‐, ether‐linked phosphatidylcholine; PE, phosphatidylethanolamine; PE O‐, ether‐linked ethanolamine; PE P‐, ethanolamine plasmalogen; PG, phosphatidylglycerol; PG O‐, ether‐linked phosphatidylglycerol; SM, sphingomyelin; TG, triacylglycerol.

Figure 6

Choline pathway inhibition has strong anti‐leukemic activity in patient‐derived xenograft cells. (A) Graphic illustrating the CDP‐choline pathway and the inhibitors used to block key enzymes in human PDX samples. (B) Left: Representative FACS plots showing CD34 expression and FSC‐H of AML‐602 cells treated with 20 µM geranylgeraniol (center), 2 µM GW4869 (right), or DMSO as control (left). Values indicate gated fractions of CD34 bright (++) cells in percent. Right: Bar plot showing CD34 surface expression of AML‐602 and AML‐661 cells treated with 20 µM geranylgeraniol (blue), 2 µM GW4869 (green), or DMSO as control (lilac). Data are shown as mean + SD. Symbols represent individual replicates. Numbers indicate results from ordinary two‐way ANOVA and Holm‐Sidak's multiple comparison test with a single pooled variance computed for each comparison. (C) Dose–response curves of five PDX samples (AML‐602, AML‐661, AML‐663, AML‐346, and AML‐372) versus two independent batches of healthy HSPCs (CD34+ #1 comprising four donors, CD34 #2 comprising two donors) treated with RSM‐932A, EB‐3P,, ICL‐CCIC‐0019, and EB‐3D., , , Curves were fitted using a sigmoidal model. Numbers in brackets indicate the half‐maximal inhibitory concentrations (IC50). (D) Heatmap comparing the IC50 for the drugs RSM‐932A (RSM), EB‐3P,, ICL‐CCIC‐0019 (ICL), and EB‐3D, , , in AML versus HSPCs (CD34). Lilac indicates a lower IC50 in AML than in HSPCs, blue indicates a higher IC50 in AML than in HSPCs. Annotations indicate the shortened mutational status of the used samples. See Table S3 for further sample information. NK, normal karyotype; WT1mut, mutated wilms tumor protein 1; del7q, deletion on the long arm of chromosome 7; del6p, deletion on the short arm of chromosome 6; molAdv: adverse risk molecular mutations; NPM1: nucleophosmin; FLT3‐ITD: fms‐related tyrosine kinase 3 with internal tandem duplication; 5q‐, loss of the long arm of chromosome 5; 13q‐, loss of the long arm of chromosome 13. cKIT, tyrosine kinase KIT; cplx, complex karyotype; TP53mut, mutation of tumor protein 53. (E) Left: XY plot showing expression of choline kinase ɑ (CHKA) versus LPIN1 (log10 RPKM) in samples from the Leucegene cohort. Highlighted are HSPCs (CD34, lilac), normal karyotype samples with NPM1 and FLT3‐ITD co‐mutation (NK NF, green), and complex karyotype samples with TP53 mutation (CK TP53mut, blue). Right: Dot plot of CHKA mRNA expression in samples from the Leuegene cohort (log10 RPKM). Symbols represent individual samples, bars show median CHKA expression. Statistical significance was tested using ordinary one‐way ANOVA corrected for multiple comparisons with Holm‐Sidak test. CD34: healthy HSPCs; NK_FLT3ITD_NPM1c: normal karyotype samples with NPM1 and FLT3‐ITD mutation; complex_TP53_mol_mut: complex karyotype samples with TP53 mutation; complex_TP53_mol_wt: complex karyotype samples without TP53 mutation; NK_FLT3ITD_NPM1wt: normal karyotype samples with FLT3‐ITD and NPM1 wild type; NPM1c_FLT3wt: samples with NPM1 mutation without FLT3‐ITD, t(8;21): translocation of a part from chromosome 8 to chromosome 21; PML::RARA: reciprocal translocation of retinoic acid receptor ɑ (RARA) with promyelocytic leukemia protein (PML). (F) Graphic illustrating the proposed functions of LPIN1 in AML and HSPCs. ADP, adenosine diphosphate; ATP, adenosine triphosphate; CCTɑ/CCTβ, choline‐phosphate cytidylyltransferase‐ɑ/choline‐phosphate cytidylyltransferase‐β; CDP‐choline, cytidine‐diphosphate‐choline; Cer, ceramide; CHKɑ1/CHKß, choline kinase‐ɑ1/choline kinase‐β; CMP, cytidine monophosphate; CTP, cytidine triphosphate; nSMase, neutral sphingomyelinase; PC, phosphatidylcholine; P‐choline, phosphocholine; PPi, inorganic phosphate; SM, sphingomyelin; SMS1/SMS2, sphingomyelin synthase 1/sphingomyelin synthase 2.