Lipin-1 regulation of phospholipid synthesis maintains endoplasmic reticulum homeostasis and is critical for triple-negative breast cancer cell survival

Abstract

Cancer cells reprogram their metabolism to increase the synthesis of macromolecules for rapid proliferation. Compared to fatty acids, much less is known about the synthesis of phospholipids, which is essential for membrane biogenesis in cancer cells. We found that LPIN1, which encodes lipin-1, a phosphatidic acid phosphatase (PAP) controlling the rate-limiting step in the phospholipid synthesis pathway, is highly up-regulated in basal-like triple-negative breast cancer (TNBC). Moreover, high LPIN1 expression correlates with the poor prognosis of these patients. Knockdown of LPIN1 increases apoptosis in basal-like TNBC cell lines, whereas it has minimal or less effect on normal human mammary gland epithelial cells (HMECs) and estrogen receptor-positive breast cancer cell lines. Fatty acid incorporation and lipidomics analyses showed that LPIN1 knockdown blocks phospholipid synthesis and changes membrane lipid compositions that ultimately induce the activation of 1 of the 3 branches of unfolded protein responses, the inositol-requiring enzyme-1α pathway. We also show for the first time, to our knowledge, that lipin-1 knockdown significantly inhibits tumor growth in vivo using an orthotopic xenograft breast mouse model. Our results suggest that lipin-1 is a potential target for cancer therapy.—He, J., Zhang, F., Tay, L. W. R., Boroda, S., Nian, W., Levental, K. R., Levental, I., Harris, T. E., Chang, J. T., Du, G. Lipin-1 regulation of phospholipid synthesis maintains endoplasmic reticulum homeostasis and is critical for triple-negative breast cancer cell survival.

Increased membrane biogenesis is critical to rapid proliferation of cancer cells. Biologic membranes are composed of proteins and lipids. In addition to the structural role in cells, lipids can be used to store energy and function as signaling molecules (1). It has been known for some time that de novo fatty acid synthesis occurs at a very high rate in tumor tissues (2). Up-regulation of the rate-limiting enzyme, fatty acid synthase (FASN), correlates strongly with cancer progression (1, 2). However, some recent studies have revealed that fatty acid uptake from blood and stromal cells can also supply the lipids that support cancer cell growth in some settings (3, 4); thus, inhibition of fatty acid synthesis may have limited clinical success. Indeed, supplementing the culture medium with palmitic acid completely rescues cancer cells from apoptosis induced by the knockdown of either acetyl-CoA carboxylase (ACC) or FASN, 2 essential fatty acid synthesis enzymes (5). Inhibition of the master regulator of lipid synthesis, sterol regulatory element-binding protein-1, results in cell death only when exogenous lipid supplies are limited (6). It has also been reported that oncogenic Ras mutation increases the uptake of fatty acids of cancer cells from the extracellular spaces, therefore potentially limiting their dependence on de novo synthesis of these molecules (7, 8). Compared to fatty acids, very little is known about the metabolism of phospholipids in cancer cells. Some recent studies have revealed alterations in phospholipid metabolism and phospholipid metabolism genes in cancer (9–11). However, little is known about how phospholipid metabolizing enzymes, especially those directly involved in the biosynthesis of phospholipids, contribute to cancer initiation and progression.

Phospholipid and membrane proteins are mainly synthesized on the surface of the endoplasmic reticulum (ER) (12, 13). Physiologic and pathologic processes that disrupt the ER protein folding can lead to the accumulation of unfolded or misfolded proteins in the ER, a condition called ER stress (12). Some recent studies have shown that dysregulation of phospholipid metabolism can lead to ER stress response (13–15). Three highly specific signaling pathways, termed the unfolded protein response (UPR), have been evolved to protect the cell from ER stress: protein-kinase/endoribonuclease inositol-requiring enzyme (IRE)-1α, protein kinase R-like ER kinase/pancreatic eIF2 kinase (PERK), and activating transcription factor 6 (ATF)-6 (12, 13). Activation of the UPR maintains and restores ER homeostasis by increasing protein folding capacity through induction of ER chaperones that mediate protein folding and by proteasomal degradation of unfolded and aggregated proteins. If the UPR remains unresolved, ER stress triggers apoptosis through activation of CCAAT/enhancer-binding protein homologous protein (CHOP) or JNK (16). Thus, ER stress is essential for tumor proliferation and survival in diverse types of human cancer cells, and induction of persistent ER stress in cancer cells can be used for cancer therapy (16, 17).

In the present study, we showed that LPIN1, a gene encoding the phosphatidic acid phosphatase (PAP) lipin-1 that is critical for the de novo synthesis of phospholipids and triglycerides (18–20), is significantly up-regulated in basal-like triple-negative breast cancer (TNBC), and the overexpression of LPIN1 correlates highly with poor patient survival. Lipin-1 knockdown reduces the survival of TNBC cells through inhibition of phospholipid synthesis and the persistent activation of the IRE1α-JNK ER stress response pathway. Knockdown of LPIN1 significantly blocked tumor growth in an in vivo mouse xenograft tumor model. Our results suggest that the phospholipid synthesis pathway could be a good target for cancer therapy.

MATERIALS AND METHODS

Cell culture, virus production, and viability measurement

All cancer cell lines were obtained from American Type Culture Collection (Manassas, VA, USA). HCC1806 and BT474 breast cancer cells were cultured in Roswell Park Memorial Institute (RPMI)-1640 medium supplemented with heat-inactivated 10% fetal bovine serum (FBS; 10437-036, Sigma-Aldrich, St. Louis, MO, USA). MDA-MB-231, MCF-7, and HEK293-TLA (Thermo Fisher Scientific, Waltham, MA, USA), and HEK293-GP2 (Takara Bio, Mountain View, CA, USA) cells were cultured in DMEM supplemented with 10% FBS. Normal human mammary gland epithelial cells (HMECs) were cultured in mammary epithelial cell basal medium (MEBM), a growth medium (CC-3151) with growth factors and other supplements (CC-4136) from Lonza (Allendale, NJ, USA). Unless indicated, experiments were performed 3 d after viral infection, as previously described (21, 22). To inhibit kinase activity, cells were treated for 4 h with inhibitors for mTOR (250 nM torin 1; 10997), p38 MAPK (2 μM JX-401; 16898), AMPK (2 μM dorsomorphin; 11967), and Src (500 nM bosutinib; 12030), from Cayman Chemical (Ann Arbor, MI, USA); PI3K (10 μM LY-294002; 440202) from EMD Millipore (Billerica, MA, USA); MEK1/2 (10 μM U0126; 9903) from Cell Signaling Technology (Danvers, MA, USA); and JNK (10 μM SP600125; S5567) from Sigma-Aldrich. For viability measurement, cells were stained with trypan blue, and counted with a hemocytometer.

Plasmids

The small hairpin RNAs (shRNAs) for firefly luciferase (shLuc) and LPIN1 (shLPIN1-C and -D) were cloned in pLKO.1 lentiviral vector with the target sequences 5′–GATTTCGAGTCGTCTTAAT–3′, 5′–GTGGTTGACATAGAAATCA–3′, and 5′–GCAGAACTCTTCCTAATGA–3′, respectively. Two mouse Lpin1 splicing isoforms, Lpin1a and -1b and Lpin1b-D712E, were amplified from the pcDNA vector and cloned into a modified pQCXIP retroviral vector in which the CMV promoter was replaced by a phosphoglycerate kinase (PGK)-1 promoter and 3 tandem influenza hemagglutinin (HA) tags at the N terminus. Wobble mutations that change Lpin1b cDNA sequences, but not the encoded proteins, were introduced by PCR mutagenesis to make their mRNAs insensitive to the degradation of shLPIN1-C with the following primers: forward, 5′–GAGAGAAAGTAGTGGATATCGAGATAAATGGGGAGTCC–3′; reverse: 5′–TATCTCGATATCCACTACTTTCTCTCGGGAGCGGAGGAC–3′.

Lipin-1 antibody production

Rabbit polyclonal lipin-1 antibody was generated by injection of the purified glutathione S-transferase (GST)–fused human lipin-1 (1926) protein expressed from a pGEX-4T1 vector to 2 rabbits, as described in Grimsey et al. (23), and affinity purified by incubating rabbit sera from the final bleeds with the GST fusion proteins conjugated to beads. All the procedures were performed at Novoprotein (Summit, NJ, USA).

Western blot analysis and immunofluorescence staining

Western blot analysis with the Odyssey infrared imaging system (Li-Cor Biosciences, Lincoln, NE, USA) and immunofluorescence staining were performed as previously described (21, 22). For dephosphorylation, stably expressed HA-lipin-1 was immunoprecipitated from HCC1806 cells with HA-conjugated beads (11-815-016-001; Roche, Indianapolis, IN, USA), and treated with λ phosphatase (P0753S, New England Biolabs, Ipswich, MA, USA) as described in Eaton et al. (24). Antibody against lipin-1 was used at 1:100 dilution. Anti-lipin-2 (ABS524 Sigma-Aldrich) was used at 1:300 dilution. Eukaryotic initiation factor (eIF)-2α (sc-11365) and HA antibodies (sc-805) were from Santa Cruz Biotechnology (Dallas, TX, USA), and used at 1:200 and 1:300 dilution, respectively. Antibodies for phospho-eIF2α (Ser51; 9721), SAPK/JNK (9258), phospho-SAPK/JNK (Thr183/Tyr185; 9255), and CHOP (2895P) were from Cell Signaling Technology and were used at 1:1000 dilution. Anti-XBP1 (ab37152) from Abcam (Cambridge, MA, USA) was used at 1:1000 dilution. Goat anti-rabbit secondary antibody conjugated with Alex 594 (A-11037) was from Thermo Fisher Scientific. Goat anti-mouse or goat anti-rabbit secondary antibody conjugated with fluorescent IRDye 680LT (925-68020 or 925-68021) or IRDye 800CW (926-32210 or 926-32211) were from Li-Cor Biosciences and used at 1:5000 dilution. The intensity of protein bands was analyzed with ImageJ (National Institutes of Health, Bethesda, MD, USA).

Measurement of PAP activity

The basic protocol for assaying PA-specific activity by measuring the amount of [³²P]_i release was published by Han and Carman. (25). Phosphatidic acid was purified and labeled with [³²P] using γ-[³²P] (NEG035C; Perkin Elmer, Waltham, MA, USA) (24). PAP activity was measured in the presence of 0.5 mM MgCl₂, and the Mg²⁺-independent activity (PAP2) was measured in the absence of MgCl₂ and in the presence of 1 mM EGTA. The reactions were performed with 40 μg cell homogenate for 20 min at 30°C. Background activity (reaction containing everything but protein) was subtracted from PAP activity. The PAP2 activity was subtracted from that of PAP1 to reflect Mg²⁺-dependent PAP1 activity in the cell homogenates.

Flow cytometry analysis of apoptosis

Cells were harvested by trypsin digestion 3 d after infection with lentiviruses and stained with 5 μg/ml annexin V-FITC from Biolegend (640906; San Diego, CA, USA) and 5 μg/ml propidium iodide (PI) for 15 min. At least 10,000 cells per sample were collected by an LSRFortessa Cell Analyzer (BD Biosciences, San Jose, CA, USA) and analyzed with FlowJo software (Ashland, OR, USA). Cells without staining or stained with annexin V-FITC or PI only were used as compensation controls.

Measurement of phospholipid synthesis

HCC1806 cells were labeled with 2 μCi/ml palmitic acid [9, 10-³H(N)] (ART 0129) from American Radiolabeled Chemicals (St. Louis, MO, USA). After they were washed 3 times with PBS, the cells were harvested in ice-cold methanol, followed by addition of chloroform and water (methanol:chloroform:water = 6:5:4). The cell extract was vortexed and then spun at 13,300 g for 5 min. Organic phase lipid (100 μl) was mixed with 900 μl 1% formic acid in acetonitrile (Sigma-Aldrich), vortexed for 30 s, and then centrifuged at 5000 g for 3 min. The resulting supernatant was transferred to a HybridSPE-PL column (55261-U; Sigma-Aldrich), and centrifuge at a speed of 1500 g for 2 min. HybridSPE-PL column was then washed with 1% formic acid in acetonitrile once and discard the flow through. Phospholipids bound to the HybridSPE-PL column were then eluted with 1 ml 5% ammonium hydroxide in acetonitrile. The radioactivity of total lipids and phospholipids was measured by liquid scintillation counter from Perkin Elmer. The degree of phospholipid synthesis was normalized to total protein.

Lipidomics by electron spray ionization and tandem mass spectrometry

The lipids used were ceramide (Cer), cholesterol (CL), diacylglycerol (DAG), glucosyl/galactosyl ceramide (HexCer), phosphatidic acid (PA), phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylglycerol (PG), phosphatidylinositol (PI), and phosphatidylserine (PS). Their respective lysospecies were lysoPA, lysoPC, lysoPE, lysoPI, and lysoPS, and their ether derivatives: PC O-, PE O-, LPC O-, LPE O- sterol ester (SE), sphingomyelin (SM), and triacylglycerol (TAG). Lipid species were annotated according to their molecular composition as follows: (lipid class)-(sum of carbon atoms in the fatty acids):(sum of double bonds in the fatty acids) (e.g., SM-32:2). Where available, 2 fatty acid compositions of a lipid is separated with a hyphen (e.g., PC-18:1–24:2).

Lipids extraction and analysis using electron spray ionization and MS/MS were performed at Lipotype, GmbH (Dresden, Germany) (26, 27). Automated processing of acquired mass spectra, and identification and quantification of detected lipid species were performed by LipidXplorer software. The abundance of lipids was presented as picomoles of lipids per microgram of protein.

Quantitative RT-PCR

Total RNAs were prepared by using the RNeasy Plus Mini Kit (74134; Qiagen, Hilden, Germany). Total RNA (1 μg) was reverse transcribed to cDNA by using M-MLV reverse transcriptase (280250130; Thermo Fisher Scientific). The levels of lipin-1 transcriptional target genes (28) were measured by SYBR Green Real-time PCR method in C1000 Thermal Cycler a (Bio-Rad, Hercules, CA, USA). The mRNA level of each gene was normalized to that of GAPDH using 2^−ΔΔCt (29). The primers for PCR reaction were as the following: CPT1A, forward: 5′–CCTCCAGTTGGCTTATCGTG–3′, CPT1A, reverse: 5′–TTCTTCGTCTGGCTGGACAT–3′; ACADVLforward: 5′–TGGGTGAGGTTGGG-AGTG–3′, ACADVL, reverse: CAAACTGGGTACGATTA-GTGG–3′; ACADM, forward: GAATTAAACAT-GGGCCA–3′, ACADM, reverse: TACAGGTCTGGTTT-TATCAA–3′; GAPDH, forward: CCACTCCTCCACCT-TTGAC–3′, and GAPDH, reverse: ACCCTGTTGCTG-TAGCCA–3′.

ATP measurement

ATP level was measured by the ATPlite Luminescence ATP Detection Assay System (601694; Perkin Elmer). In brief, cells in the 96-well plate were lysed by adding 50 μl mammalian cell lysis solution to each well of cells and shaken for 5 min. Then, 50 μl substrate solution was added for 5 min. The ATP level was determined by measuring firefly luciferase luminescence and normalized to the amount of proteins.

Orthotopic breast cancer mouse model

Mouse experiments were performed in accordance with Association for Assessment and Accreditation of Laboratory Animal Care guidelines, and with University of Texas Health Science Center Institutional Animal Care and Use Committee approval. HCC1806 cells were infected with shLuc, shLPIN1-C, and shLPIN1-D lentivirus for 2 d and collected after trypsinization. Cells were washed with Dulbecco’s phosphate-buffered saline (DPBS) and centrifuged at 300 g for 5 min at room temperature to remove dead cells. After resuspension in DPBS, the cells were diluted to a final concentration of 2 × 10⁷/ml. The fourth mammary fat pad of 5-wk-old female BALB/c severe combined immunodeficiency mice from The Jackson Laboratory (Bar Harbor, ME, USA) were injected with 2 million cells under 2.5% isoflurane. The length and width of tumors were measured with a caliper every 2 to 3 d. The tumor volume was calculated by (width)² × length/2.

Bioinformatic analysis

We downloaded the RSEM-processed gene expression data and merged level 1 clinical annotations (including ER, HER2, and PR status; overall survival time) for the TCGA breast cancer samples (release 160128) from the Broad Institute Genome Data Analysis Center Firehose database (Cambridge, MA, USA; http://gdac.broadinstitute.org/runs/stddata__2016_01_28/). We acquired the PAM50 subtype calls (30). From the gene expression data, we selected the samples from primary tumors and logged the TPM values. From this matrix, we selected the genes previously shown to be involved in phospholipid synthesis, including GPAT1, GPAT2, GPAT3, GPAT4, AGPAT1, AGPAT2, AGPAT3, AGPAT4, AGPAT5, LPIN1, LPIN2, LPIN3, CHKA, CHKB, PCYT1A, PCYT1B, CEPT1, EPT1, ETNK1, ETNK2, PEMT, PTDSS1, and PTDSS2. To compare their gene expression values across subtypes, we used GraphPad Prism (ver. 6) for Mac OS X (GraphPad Software, La Jolla, CA, USA). We performed Kaplan-Meier analysis with the same software.

Statistical analysis

All data were obtained from at least 3 experiments. A Student’s t test was used to compare 2 groups. Kruskal-Wallis 1-way ANOVA followed by the Dunn or Holm-Sidak test was used to compare 3 or more groups. Two-way ANOVA followed by the Holm-Sidak test was used for growth curve analysis. Data are presented as means ± sd, unless otherwise specifically demonstrated. A value of P < 0.05 was considered statistically significant.

RESULTS

LPIN1 is overexpressed in human TNBCs, and its overexpression strongly correlates with poor prognosis

The synthesis of major phospholipids, such as PC, PE, and PS, requires the conversion of PA to DAG on the ER by a series of enzymes (1, 31, 32; Fig. 1A). Conceivably, inhibition of phospholipid synthesis would reduce the levels of phospholipids required for membrane biogenesis and signaling molecules in cancer cells, therefore inhibiting their proliferation or viability. Because of the lack of effective therapy for TNBC (33), we analyzed the mRNA expression levels of the genes encoding the major phospholipid synthesis enzymes in Fig. 1A in a patient with TNBC from The Cancer Genome Atlas TCGA database (30). The expression levels of several genes are significantly up-regulated in TNBCs (e.g., LPIN1, PTDSS1, AGPAT4, AGPAT9, and PCYT1B) (Fig. 1B), indicating that up-regulated expression of phospholipid synthesis genes is a common feature in TNBC. The most up-regulated gene, LPIN1, encodes lipin-1, which functions as a PAP to convert PA to DAG, a rate-limiting step in phospholipid and triglyceride synthesis (18–20). Further, analysis of LPIN1 expression across different types of breast cancers revealed that its expression is also significantly elevated in basal-like breast cancer, as compared to other types of breast cancer (i.e., Her2, luminal A, luminal B, and normal breast-like) (Fig. 1C), and TNBC showed a significant up-regulation of LPIN1 expression as compared to non-TNBC (Fig. 1D). Analysis of prognosis data in the TCGA database revealed that high LPIN1 expression correlates positively with poor survival of patients with basal-like TNBC (Fig. 1E).

Figure 1.

LPIN1 is up-regulated in basal-like TNBC, and its high expression level correlates with poor prognosis of patients with TNBC. A) The biochemical pathway of phospholipid synthesis. The enzymes and lipids are labeled with green and black fonts, respectively. B) Expression changes of phospholipid synthesis genes (points) in TNBC. The x axis shows the log₂ of the average fold change in expression between TNBC and other breast tumors. Positive numbers indicate higher expression in TNBC. The y axis is the negative log of the P value indicating the statistical significance of the difference in expression. Higher numbers indicate an increased significance. The most significant up- and down-regulated genes are labeled with red and blue fonts, respectively. C) The expression level of LPIN1 mRNA in different breast cancer types: basal-like, Her2, luminal A, luminal B, and normal breast-like. Post hoc tests between pairs of subtypes are performed with unpaired Student’s t tests. ****P < 0.0001 (ANOVA test), indicates the significance of the difference in expression levels. D) Comparison of the expression level of LPIN1 in TNBC and non-TNBC. ****P < 0.0001 (unpaired Student's t test). E) A high LPIN1 expression level correlated with shorter overall survival of patients with TNBC. P values determined by log–rank test.

Lipin-1 is necessary for the survival of basal-like TNBC cells

The studies on lipin-1 have mainly focused on metabolic diseases, such as lipodystrophy, obesity, and insulin resistance, rhabdomyolysis, and myopathy (18–20). The roles of lipin-1 in nonmetabolic tissues are less known. Recently, inhibition of lipin-1 is reported to potentiate the treatment of rapamycin in prostate cancer cells (34). It remains unknown whether lipin-1 also contributes to tumor initiation and maintenance in other cancers and how it regulates cancer progression in general. We knocked down LPIN1 by using 2 short hairpin (sh)RNAs, shLPIN1-C and -D, in basal-like TNBC HCC1806 cells (Fig. 2A). Lipin-1 knockdown did not change the level of lipin-2 protein, supporting the previous observation that the compensatory expression of lipin-2 in lipin-1 down-regulated cells is cell type dependent (34). Furthermore, lipin-1 knockdown eliminated most of the Mg²⁺-dependent PAP activity in an in vitro assay using total cell lysates, suggesting that lipin-1 accounts for most Mg²⁺-dependent PAP activity in these cells (Fig. 2B). Functionally, both LPIN1 shRNAs greatly reduced the number of viable cells (Fig. 2C). The reduced number of viable cells can be fully rescued by the re-expression of 2 major splicing variants of the mouse LPIN1 cDNAs, Lpin1a and Lpin1b (Fig. 2D), supporting the notion that the observed phenotype is specific to the loss of lipin-1 protein.

Figure 2.

Lipin1 is critical to the survival of basal-like TNBC cells. A) Western blot analysis showed that LPIN1 shRNAs, shLPIN1-C, and shLPIN1-D reduced the levels of lipin-1 without affecting lipin-2 in HCC1806 cells. B) Loss of most of the PAP activity in HCC1806 cells expressing LPIN1 shRNAs (n = 4–9). C) LPIN1 knockdown reduced the number of viable cells in HCC1806 cells. Viable HCC1806 cells were counted daily 3 days after infection with lentiviruses carrying the control or LPIN1 shRNAs (n = 3). D) Reexpression of mouse Lpin1a and -1b fully rescued the decreased number of viable cells in cells expressing shLPIN1-C (n = 3). ***P < 0.001 compared with shLuc; ^###P < 0.001 compared with shLPIN1-C. E) HCC1806 cells lacking LPIN1 expression undergo cell apoptosis. Left: flow cytometry analyses showed LPIN1 knockdown increased annexin V⁺ apoptotic cells without affecting necrosis (PI⁺). Right: statistical analyses of flow cytometry data (n = 3). *P < 0.05, ***P < 0.001. F) LPIN1 knockdown induced apoptosis in MDA-MB-231 cells. Left: confirmation of LPIN1 knockdown by shLPIN1-C using Western blot analysis. Middle: flow cytometry analysis showed LPIN1 knockdown increased annexin V⁺ apoptotic cells without affecting necrosis (PI⁺). Right: statistical analyses of flow cytometry data (n = 3). ***P < 0.001.

To determine the cause of the reduction of viability in LPIN1-knockdown cells, we performed flow cytometry analysis of HCC1806 cells staining with FITC-labeled annexin V and PI. The proportion of annexin V⁺/PI⁻ negative cells, which indicates early apoptotic cells, was significantly increased in cells expressing the 2 LPIN1 shRNAs (15.4 and 8.6% in shLPIN1-C and -D cells, respectively) as compared to the control (3.3%; Fig. 2E). In contrast, LPIN1 shRNAs did not cause a significant difference in the proportion of PI⁺ necrotic cells. Similarly, knockdown of LPIN1 in MDA-MB-231, another basal-like TNBC cell line, also increased the proportion of apoptotic cells (11.4% in shLPIN1-C cells vs. 1.0% in control shLuc cells; Fig. 2F). This result suggests that LPIN1 is important for cancer cell survival, and the reduction of a viable number of cells in LPIN1-knockdown cells is caused by apoptosis.

HMECs and estrogen receptor–positive breast cancer cells are less sensitive to LPIN1 knockdown

To determine whether the dependence of lipin-1 for cell survival is specific to TNBC, we knocked down LPIN1 in 2 estrogen receptor–positive cancer cell lines, MCF-7 and BT-474, and primary normal HMECs. In contrast to TNBCs, although LPIN1 knockdown eventually caused visual decreases in the number of viable cells in MCF-7 and BT-474 cells, no increase in cell death was observed in LPIN1-knockdown cells 3 d after lentiviral infection of shLPIN1-C shRNA (Fig. 3A, B), when significant cell death was observed in both HCC1806 and MDA-MB-231 cells (Fig. 2E, F). Similarly, knockdown of LPIN1 did not induce cell death in primary normal HMECs by flow cytometry analysis (Fig. 3C), although it slightly reduced the number of viable cells that became evident only on d 5 of the measurement (Fig. 3D), most likely caused by a reduction of proliferation rate rather than cell death. These results suggest that lipin-1 is critical for the survival of TNBC cells.

Figure 3.

HMECs and estrogen receptor–positive breast cancer cells are less sensitive to LPIN1 knockdown. Analyses of cell death in MCF-7 (A), BT-474 (B), and normal HMECs (C). Left: confirmation of LPIN1 knockdown by shLPIN1-C. Middle: flow cytometry analysis. Right: statistical results of flow cytometry data in the middle. D) LPIN1 knockdown only slightly reduced proliferation in normal HMECs.

Lipin-1 is necessary for membrane lipid remodeling but not for ATP production

Since lipin-1 functions as a PAP to convert PA to DAG, which is a substrate for phospholipid and TAG synthesis (18–20), we hypothesized that down-regulation of lipin-1 would inhibit the general phospholipid synthesis. Indeed, phospholipid synthesis was significantly reduced in lipin-1-knockdown HCC1806 cells, as measured by the incorporation of [³H]-palmitic acids into phospholipids (Fig. 4A). To gain insight into the effects of chronic depletion of lipin-1 on the composition of steady-state cellular lipids, we performed lipidomic analysis by using MS. The levels of CL, SM, and CE were significantly reduced in LPIN1-knockdown cells (Fig. 4B). The total levels of PA, DAG, TAG, and major phospholipids, such as PC, PE, PS, and PI, remained largely unchanged, similar to a previous report that liver-specific LPIN1 knockout did not alter the global levels of PA, DAG, and phospholipids (35). The levels of PA and DAG in both HCC1806 cells expressing shLuc and shLPIN1 were very low compared with other cell types (26, 35, 36), implying that they are actively converted to other lipids in these cells (31, 32). In contrast to unchanged total lipid levels, significant changes were observed in the compositions of most lipid classes in LPIN1-knockdown cells (Fig. 4C). Despite some variability for each lipid class, there were consistent reductions in short acyl chains and increases in long acyl chains in LPIN1-knockdown cells (Fig. 4C, D). Finally, LPIN1 knockdown also led to a decrease in the saturation of membrane lipids (Fig. 4D).

Figure 4.

Lipin-1 is necessary for phospholipid synthesis and membrane lipid remodeling in HCC1806 cells. A) Inhibition of palmitic acid incorporation into phospholipids by LPIN1 knockdown (n = 4). B) MS analysis of lipids in LPIN1-knockdown cells. The content of individual lipid classes was determined by summing up all identified species (n = 3). C) Changes in lipid species that may be directly regulated by lipin-1 activity. Except PA, where only 3 species were identified, only the top 15 most abundant species of each phospholipid class are presented. D) Changes of acyl chain length and saturation in LPIN1-knockdown cells. The length is reported as a combined value across the 2 acyl chains per lipid species. The unsaturation index is calculated as the sum of the mole percentage of each lipid species multiplied by the number of double bonds within that species. *P < 0.05, **P < 0.01, ***P < 0.001.

In addition to the PAP activity, lipin-1 can also function in the nucleus to regulate the expression of genes involved in mitochondrial fatty acid oxidation for energy production upon nutrient depletion (18, 20, 28). In contrast to PAP activity and phospholipid synthesis, LPIN1 knockdown had no effect on the expression of 3 genes known to be regulated by the transcriptional coactivator activity: CPT1A, ACADVL, and ACADM (Fig. 5A). Consequently, LPIN1 knockdown also did not alter ATP production (Fig. 5B). Furthermore, reexpression of a cDNA encoding a lipin-1b mutant, D712E, which loses the PAP activity but retains transcriptional coactivator activity (18, 19, 28), failed to rescue the cell death phenotype in LPIN1-knockdown cells (Fig. 5C). Taken together, these data suggest that the PAP activity, and not the transcriptional coactivator activity, regulates lipid remodeling and the survival of HCC1806 breast cancer cells.

Figure 5.

The nuclear function of lipin-1 does not regulate the viability of HCC1806 cells. A) LPIN1 knockdown did not change the expression of selected lipin-1 transcriptional target genes (n = 4). B) LPIN1 knockdown had no influence on cellular ATP level. C) D712E mutant failed to rescue the cell death phenotype induced by LPIN1 knockdown (n = 3). ***P < 0.001.

Lipin-1 deficiency activates the IRE1-XBP1 UPR pathway

We hypothesized that lipin-1-mediated phospholipid synthesis on the ER is necessary for membrane biogenesis in highly proliferative cancer cells, and its inhibition would lead to either unbalanced production of phospholipids and proteins or changes in ER structure, both eliciting the ER stress response. We first examined ER structure by ER tracker staining. The ER formed a network of tubules that extends from the cell center to the periphery in control shLuc cells (Fig. 6A). In contrast, the ER appeared not only fragmented but also clustered, mainly around nuclei in shLPIN1 cells. We next performed Western blot analysis to determine the activation status of the 3 UPR pathways by measuring the phosphorylation of eIF2α for the PERK pathway, expression of ATF6α and BIP/GRP78 for the ATF6 pathway, and spliced XBP1 for the IRE1α pathway. LPIN1 knockdown did not affect the activation status for the ATF6α and PERK pathways (Fig. 6B). In contrast, the level of the spliced form of XBP1 (sXBP1) was significantly up-regulated, indicating activation of the IRE1α-signaling pathway (Fig. 6B, C). Prolonged ER stress condition can severely impair ER function, thus leading to cell death through the activation of either the PERK-CHOP or IRE1α-JNK pathways (16, 37, 38). To delineate the specific mechanism of LPIN1-knockdown–induced cell death, we examined the activation of these 2 pathways by Western blot analysis. LPIN1 knockdown did not have an effect on the expression of CHOP (Fig. 6D), supporting the results based on eIF2α phosphorylation that the PERK pathway is not altered in LPIN1-knockdown cells (Fig. 6B). In contrast, the phosphorylation of JNK is significantly increased after LPIN1 knockdown (Fig. 6E, F). Taken together, these results indicate that activation of the IRE1α-JNK signaling pathway is responsible for cell death in LPIN1-knockdown cells.

Figure 6.

Knockdown of LPIN1 activates the IRE1α branch of UPRs. A) Disruption of the ER structure in HCC1806 cells expressing LPIN1 shRNAs. ER was labeled with ER tracker. Scale bars, 20 μM. B) LPIN1 knockdown led to a specific elevation of spliced XBP1 (sXBP1) in HCC1806 cells, whereas it had no effect on eIF2α phosphorylation or ATF6α and BIP/GRP78 levels. C) Quantification of relative levels of the sXBP1 in B. sXBP1 levels were normalized by α-tubulin (n = 4) **P < 0.01. D) The protein level of CHOP was increased in thapsigargin-treated cells, but it was undetectable in LPIN1-knockdown cells. E) JNK phosphorylation levels increased in HCC1806 cells expressing LPIN1 shRNAs. F) Quantification of the relative phosphorylation levels of phospho-JNK p46 and phospho-JNK p54 in E. The levels of phospho-JNK isoforms were normalized by their respective total JNKs (n = 4). *P < 0.05; **P < 0.01.

mTOR signaling regulates the phosphorylation and localization of lipin-1 in HCC1806 cells

Lipin-1 is subjected to phosphorylation at multiple sites that are important for its localization and activity (24, 39, 40). To determine how lipin-1 phosphorylation is regulated in HCC1806 cells, the cells were treated with inhibitors for 7 common signaling pathways and examined the mobility shift of lipin-1 on SDS-PAGE, which was used as an indicator of multiple phosphorylations (24, 39, 40). Among all inhibitors tested, only the mTOR inhibitor torin 1 and PI3K inhibitor LY-294002 made a portion of lipin-1 move visibly faster than the control (Fig. 7A). λ-Phosphatase treatment of immunoprecipitated HA-lipin-1 shifted its mobility from ∼140 kDa in control and multiple bands in torin 1-treated samples to a single band of ∼120 kDa (Fig. 7B), confirming that the change of lipin-1 mobility is indeed caused by phosphorylation. In addition, there is no difference in the levels of lipin-1 in the control and torin 1-treated samples, which was more apparent after λ-phosphatase treatment, suggesting that mTOR does not regulate the stability of lipin-1. Because PI3K acts upstream of mTOR and treatment of mTOR inhibitor showed stronger effect on lipin-1 mobility, it is likely that mTOR is directly responsible for the phosphorylation of lipin-1 in HCC1806 cells, as reported in other cell types (24, 40). The existence of multiple lipin-1 bands in cells treated with PI3K and mTOR inhibitors suggests that lipin-1 phosphorylation is also regulated by other unknown kinases in HCC1806 cells that remain to be identified. Lipin-1 phosphorylation by mTOR regulates its subcellular localization in nontumor cells (24, 39, 40), and we therefore also examined the localization of lipin-1 in torin 1-treated HCC1806 cells. Similar to other cell types, inhibition of mTOR activity promoted the translocation of lipin-1 from cytoplasm to nucleus (Fig. 7C).

Figure 7.

mTOR regulation of lipin-1 in HCC1806 cells. A) Mobility changes of lipin-1. Cells expressing HA-lipin-1 were treated with DMSO and indicated inhibitors for 4 h. B) Dephosphorylation of lipin-1 by λ-phosphatase increased the mobility of lipin-1 and abolished the difference between control and torin-1-treated cells. C) Inhibition of mTOR activity promoted the nuclear translocation of lipin-1. After treatment with or without torin 1 for 4 h, cells expressing HA-lipin-1 were fixed and stained with a rabbit HA antibody followed by Alexa 594–conjugated goat anti-rabbit secondary antibody. Nuclei were stained with DAPI. Scale bars, 20 μM.

Lipin-1 down-regulation impaired tumorigenesis in an orthotopic xenograft mouse breast cancer model

To determine the role of LPIN1 in tumorigenesis, mouse mammary fat pads were injected with HCC1806 cells stably expressing control shLuc, shLPIN1-C, or shLPIN1-D that showed successful knockdown of lipin-1 (Fig. 8A). All of the control mice started to develop palpable tumors by 11 d after injection, whereas only half of the mice bearing cells expressing shLPIN1-D developed tumors on 15 d after injection, and no mice bearing cells expressing shLPIN1-C developed tumors until 30 d after injection (Fig. 8B). The tumors in control mice also grew at a faster rate than those derived from cells expressing LPIN1 shRNAs. Analysis of lipin-1 expression in tumors collected at the endpoints of the experiments showed that the repression of lipin-1 expression was greatly lost in tumors expressing LPIN1 shRNAs (Fig. 8C). Because lipin-1 knockdown increased cell death (Fig. 2), the growth of tumors derived from cells expressing LPIN1 shRNAs is most likely related to the expansion of cancer cells in which LPIN1 knockdown was inefficient. This result also supports that lipin-1 is critical to tumor growth in vivo.

Figure 8.

Lipin-1 is critical for tumor growth in an orthotopic xenograft mouse breast cancer model. A) Confirmation of lipin-1 knockdown in HCC1806 cells. B) LPIN1 knockdown impaired tumor growth (n = 6), data are presented as means ± se. C) Loss of repression of lipin-1 expression in LPIN1 shRNA–expressing tumors collected at the endpoints of the experiment (as compared to that in HCC1806 cells initially injected in A. Lipin-1 expression in 3 independent tumors was examined for each shRNA. ***P < 0.001.

DISCUSSION

In the current study, we found that many phospholipid synthesis genes are up-regulated in TNBC, supporting the notion that the increased phospholipid synthesis is a general phenomenon in aggressive cancers. Among those, LPIN1 was the most up-regulated gene in patients with TNBC (Fig. 1B). High LPIN1 expression correlated with poor prognosis of basal-like TNBC (Fig. 1E). We demonstrated that lipin-1 regulates phospholipid synthesis and maintains ER structure (Figs. 4 and 6). LPIN1 knockdown in TNBC cells led to reduced viability (Fig. 2), due to activation of the IRE1α-JNK ER stress response pathway (Fig. 6). We show for the first time that lipin-1 knockdown significantly reduced tumor growth in vivo using an orthotopic xenograft breast cancer mouse model (Fig. 8). Primary normal HMECs and estrogen receptor–positive breast cancer cells were less sensitive to lipin-1 knockdown (Fig. 3). Our current study extends the widely studied role of lipin-1 in metabolic disease to cancer and suggests that lipin-1 inhibition is a therapeutic strategy for treating patients with TNBC.

Lipin-1 is a dual-function protein. By converting PA to DAG, lipin-1 controls the synthesis of major phospholipids. In addition, PA can also be used to synthesize some lipids through CDP-DAG (31, 32). Our data showed that lipin-1 is the major PAP in HCC1806 cells (Fig. 2B), and its chronic down-regulation led to the remodeling of lipids that eventually caused ER stress and cell death (Figs. 4 and 6). Even though lipin-1 accounts for the most of the PAP activity in HCC1806 cells, lipin-1 knockdown unexpectedly did not lead to the accumulation of PA and reduction of DAG (Fig. 4). One explanation is that these 2 lipids are actively metabolized in HCC1806 cells, which can be partially supported by our finding that both PA and DAG are in very low abundance in HCC1806 cells, as compared to those in other cells (26, 35, 36). In addition to its function as a PAP on the ER that regulates phospholipid synthesis, lipin-1 can function as a transcriptional coactivator to regulate the expression of genes for fatty acid oxidation (20, 28). In our experimental conditions, we found that lipin-1 transcriptional targets and mitochondrial ATP production are not altered by lipin-1 knockdown (Fig. 5A, B). Moreover, in contrast to wild-type, the expression of exogenous D712E that is inactive in PAP activity but otherwise normal in shLPIN1-C-expressing cells failed to rescue cell death (Figs. 2D and 4C). These results suggest that lipin-1-regulated cell viability is primarily contributed by its phospholipid synthesis function. Some studies on the transcriptional coactivator function of lipin-1 were studied under starved or nutrient-depleted conditions (20, 28). Therefore, our current study cannot fully rule out that lipin-1 also contributes to cancer progression under nutrient or growth factor-deprived conditions during rapid tumor growth or therapy.

One of the key characteristics of malignant cells is uncontrolled proliferation that necessitates the synthesis of macromolecules including membrane phospholipids (41). However, most of the previous studies focused on fatty acid synthesis and oxidation (1, 2). Very few studies have been conducted on phospholipid synthesis and its metabolizing enzymes in cancer, partially related to the complicated phospholipid metabolism routes, the nomenclature of enzymes, and technical challenges in handling phospholipids. Several FASN inhibitors show severe side effects in animal models, including significant weight loss (1). It is possible that phospholipid synthesis may comprise more effective therapeutic targets because the bulk of phospholipids are synthesized in the ER (31), and it is phospholipids, but not their building components, fatty acids, that make up cell membranes. Buildup of free fatty acids or other lipid intermediates causes lipotoxicity to cells (1, 2). In addition to membrane lipid remodeling and the persistent activation of the IRE1α-JNK pathway, it is also possible that fatty acid accumulation by inhibition of phospholipid synthesis is another reason of toxicity in lipin-1-knockdown cells, as lipin-1 knockdown inhibited fatty acid incorporation into phospholipids (Fig. 4A). Therefore, our current study supports the notion that the phospholipid synthesis pathway is a target for cancer therapy. A general concern about targeting membrane lipid synthesis that remains to be addressed is the potential specificity. However, like other targeted cancer therapy, the increases in the expression and activity of certain lipid synthesis enzymes and changes in lipid compositions in human cancers may provide a therapeutic window for selectively targeting cancer cells. Although the potential toxicity must be further evaluated in animal models, our current finding that normal HMECs are insensitive to lipin-1 knockdown supports the potential success of targeting the phospholipid synthesis pathway.

ER is the main site for the production and folding of membrane proteins and phospholipid synthesis (12, 13). Persistent ER stress triggers UPR and leads to cell death via several pathways. In our study, lipin-1 knockdown led to cell death by specifically activating the JNK pathway, likely though IRE1α, but not the other 2 UPR branches, PERK and ATF6. Overexpression of sXBP1, the active form of the XBP1 generated by UPR-mediated splicing of XBP1 mRNA, specifically elevates the expression and activity of PC/phospholipid-biosynthesized enzymes, increases levels of membrane phospholipids, and expands surface area and volume of rough ER (14). The mutual regulation of the IRE1 UPR and phospholipid synthesis pathways suggests the critical role of these 2 pathways in maintaining phospholipid levels during ER stress.

Glossary

ACC

acetyl-CoA carboxylase

CE

cholesteryl ester

Cer

ceramide

CHOP

CCAAT/enhancer-binding protein homologous protein

CL

cholesterol

DAG

diacylglycerol

DPBS

Dulbecco’s phosphate-buffered saline

ER

endoplasmic reticulum

FASN

fatty acid synthase

FBS

fetal bovine serum

GST

glutathione S-transferase

HA

influenza hemagglutinin

HexCer

glucosyl/galactosyl ceramide

HMEC

human mammary gland epithelial cell

IRE1α

inositol-requiring enzyme 1α

MEBM

mammary epithelial cell basal medium

mTOR

mechanistic target of rapamycin

MS/MS

tandem mass spectrometry

PA

phosphatidic acid

PAP

phosphatidic acid phosphatase

PC

phosphatidylcholine

PE

phosphatidylethanolamine

PG

phosphatidylglycerol

PI

propidium iodide

PS

phosphatidylserine

SE

sterol ester

shRNA

small hairpin RNA

SM

sphingomyelin

TAG

triacylglycerol

TCGA

The Cancer Genome Atlas

TNBC

triple-negative breast cancer

UPR

unfolded protein response

AUTHOR CONTRIBUTIONS

J. He, F. Zhang, and G. Du designed the research; J. He, F. Zhang, L. W. R. Tay, W. Nian, S. Boroda, T. E. Harris, and J. T. Chang performed the research; J. He, K. R. Levental, I. Levental, J. T. Chang, and G. Du analyzed the data; and J. He and G. Du wrote the paper.