sign in sign up
<<
Home , Ether lipid, Glycerophospholipid

Ether generating AGPS alters the balance of structural and signaling to fuel cancer pathogenicity

Daniel I. Benjamina, Alyssa Cozzoa, Xiaodan Jib, Lindsay S. Robertsa, Sharon M. Louiea, Melinda M. Mulvihilla, Kunxin Luob, and Daniel K. Nomuraa,1

aProgram in Metabolic , Department of Nutritional Sciences and Toxicology, and bDivision of Cell and Developmental Biology, Department of Molecular and Cell Biology, University of California, Berkeley, CA 94720

Edited by David W. Russell, University of Texas Southwestern Medical Center, Dallas, TX, and approved July 29, 2013 (received for review June 7, 2013)

Aberrant is an established hallmark of cancer metabolism to favor generation of oncogenic signaling lipids, such cells. In particular, levels have been shown to be as LPAe, lyosphosphatidic acid (LPA), and , which fuel elevated in tumors, but their specific function in cancer remains aggressive and tumorigenic features of cancer (12, 13, 16). Our elusive. We show here that the metabolic enzyme alkylglycer- studies thus reveal a heretofore unrecognized role of AGPS and onephosphate synthase (AGPS), a critical step in the synthesis of ether lipids in shifting the balance of fatty acid utilization from ether lipids, is up-regulated across multiple types of aggressive structural membrane lipids toward generation of oncogenic sig- human cancer cells and primary tumors. We demonstrate that naling lipids. ablation of AGPS in cancer cells results in reduced cell survival, Results and Discussion cancer aggressiveness, and tumor growth through altering the Cancer Cells Exhibit Heightened AGPS Expression and Ether Lipid balance of ether lipid, fatty acid, , and fatty acid– Metabolism. AGPS converts acyl-glycerone-3-phosphate into al- derived metabolism, resulting in an over- kyl-glycerone-3-phosphate, which is a requisite step in the gen- all reduction in the levels of several oncogenic signaling lipids. eration of all ether lipids. We hypothesized that cancer cells and Taken together, our results reveal that AGPS, in addition to main- tumors, which possess elevated levels of ether lipids, would have taining ether lipids, also controls cellular utilization of fatty acids, heightened AGPS expression. Consistent with this premise, we favoring the generation of signaling lipids necessary for promot- find that AGPS is highly expressed across aggressive breast ing the aggressive features of cancer. (231MFP), melanoma (C8161), and prostate cancer (PC3) cells compared with less aggressive cancer cells (MCF7, MUM2C, cancer metabolism | metabolomics | lipid signaling | and LNCaP, respectively) (Fig. 1A; Fig. S1A). We have reported eicosanoids that these aggressive cancer cells have heightened migratory, invasive, and tumorigenic properties compared with less ag- – ancer cells have fundamentally altered metabolism that gressive cells (17 19). We also show that AGPS expression is Cdrives their pathogenic features (1, 2). One hallmark of elevated two- to fourfold in Nottingham grade I (low-grade), II cancer cells is a heightened de novo lipogenic signature that (intermediate-grade), and III (high-grade) primary human breast tumors (Fig. 1B), as well as in estrogen receptor-positive/pro- serves as a critical foundation for generating lipids required for gesterone receptor-positive [ER(+)/PR(+)] and ER-negative/ cell proliferation (3, 4). For nearly half a century, it has also been PR-negative [ER(−)/PR(−)] breast tumors (Fig. 1C) compared known that cancer cells possess dramatically higher levels of – with normal breast tissue, indicating that AGPS expression may ether lipids compared with normal cells (5 8). Ether lipids have be heightened early in breast cancer development. Accordingly, we an or alkenyl chain on one or more carbons of the find that the expression of the commonly dysregulated transforming backbone bonded through an ether or vinyl linkage, rather than Harvey-Rat sarcoma oncogene (HRAS) (20) is heightened in the usual linkage. The physiological roles of ether lipids are not well understood, but they have been implicated in maintaining Significance physicochemical properties of cell membranes, such as membrane fluidity, membrane fusion events, and lipid raft microdomains fi (9, 10). Certain ether lipids, such as lysophosphatidic acid-ether Ether lipid levels are higher in tumors, but their speci c func- (LPAe) or platelet-activating factor-ether (PAFe), are signaling tion in cancer has remained unclear. We show here that the molecules that have been shown to possess bioactive and even metabolic enzyme alkylglyceronephosphate synthase (AGPS), oncogenic properties through binding specific receptors (11–15). a critical step in the synthesis of ether lipids, is up-regulated In the late 1960s, Snyder and Wood first reported that rodent across multiple types of aggressive human cancer cells and fi and human tumors possess significantly higher levels of ether lipids primary tumors. Inactivation of AGPS leads to signi cant relative to normal tissue. Over the ensuing decades, dramatic ele- impairments in cancer pathogenicity through not only lower- vations in ether lipid content have been confirmed for a wide range ing the levels of cellular ether lipids, but also by altering fatty of cancer cells and primary tumors from several tissues of origin and acid, eicosanoid, and glycerophospholipid metabolism, result- have been correlated with the proliferative capacity and tumorigenic ing in an overall reduction in the levels of several oncogenic potential of cancer cells (5–8). Nonetheless, whether elevated ether signaling lipids. lipids are causally linked, or merely associated with, cancer patho- Author contributions: D.I.B., L.S.R., S.M.L., K.L., and D.K.N. designed research; D.I.B., A.C., genicity has remained unclear. X.J., L.S.R., S.M.L., M.M.M., and D.K.N. performed research; D.I.B., K.L., and D.K.N. con- In this study, we show that inactivation of the critical enzyme tributed new reagents/analytic tools; D.I.B., A.C., X.J., L.S.R., S.M.L., M.M.M., and D.K.N. for ether lipid synthesis, alkylglycerone phosphate synthase (AGPS), analyzed data; and D.I.B., K.L., and D.K.N. wrote the paper. lowers ether lipid levels and impairs cancer pathogenicity, whereas The authors declare no conflict of interest. AGPS overexpression elevates ether lipid levels and increases This article is a PNAS Direct Submission. cancer cell motility, survival, and tumor growth. We also show that 1To whom correspondence should be addressed. E-mail: [email protected]. AGPS has a larger role beyond generating ether lipids to include This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. controlling fatty acid, eicosanoid, and acylglycerophospholipid 1073/pnas.1310894110/-/DCSupplemental.

14912–14917 | PNAS | September 10, 2013 | vol. 110 | no. 37 www.pnas.org/cgi/doi/10.1073/pnas.1310894110 Downloaded by guest on September 25, 2021 HRAS transformation of MCF10A mammary epithelial cells (21) A AGPS expression in B AGPS expression in C AGPS expression in human cancer cells primary human tumors primary human tumors induces up-regulation of AGPS expression (Fig. 1F). We thus show breast cancer melanoma that transformation of cells by HRAS is one regulatory route through 25 5 15 * 20 * * * 20 4 * 15 which cells may up-regulate AGPS expression. We also show that the ion 10 * 15 3 10 aggressive human breast 231MFP, melanoma C8161, and prostate

10 press 2 5 expression ex expression 5 5 1 expression PC3 cancer cells, as well as HRAS-transformed MCF10A cells, 0 0 0 0 possess significantly higher ether lipid levels compared with their

MCF7 C8161 less aggressive MCF7, MUM2C, and LNCaP counterparts, or grade 3 231MFP MUM2C grade 1/2 G B–F ER(-)/PR(-) ER(+)/PR(+) MCF10A control cells (Fig. 1 ; Fig. S1 ; Dataset S1). Height- normal tissue normal tissue breast tumorsbreast tumors breast tumorsbreast tumors ened lipid species include -ether (PAe), LPAe, D correlation between E F HRAS and AGPS expression HRAS expression in AGPS expression upon phosphatidyl -ether (PIe), -ether (PCe), in primary breast tumors human cancer cells RAS transformation lysophosphatidylcholine-ether (LPCe), -ether HRAS expression primary breast tumors breast cancer melanoma AGPS expression (PSe), -ether (PGe), lysophopshatidylglycerol- in primary breast tumors 4 4 5 * ether(LPGe)lipids,andplasmalogen ether lipids, such as 15 15 r = 0.59 * * * 4 G p<0.0001 3 3 phosphatidylethanolamine-plasmalogen (PEp) (Fig. 1 ; Fig. S1 10 * 10 3 2 2 B–F fi 2 ). We annotate ether lipids in this study by the speci c ether (e) pression

5 5 x expression expression e 1 1 expression 1 or (p) species and alkyl chain length and type (carbon

0 HRAS expression 0 0 0 0 number:degree of unsaturation—e.g., C16:0e is a palmityl-ether 0 5 10 15 G B–F AGPS expression MCF7 C8161 linkage with no unsaturation) in Fig. 1 and Fig. S1 . Our results MUM2C grade 3 231MFP MCF10A grade 1/2 HRAS-10A show that AGPS expression and ether lipid levels are heightened in normal tissue breast tumorsbreast tumors multiple types of aggressive human cancer cells and upon RAS ether lipid levels in human cancer cells G transformation and that AGPS expression is increased in primary breast cancer melanoma RAS transformation breast tumors.

AGPS Is a Critical Enzyme in Cancer Pathogenicity. We next sought to MCF7 231MFP MUM2C C8161 MCF10AHRAS-10A C16:0e/18:1 PAe C16:0e/18:1 PAe C16:0e/18:1 PAe determine the extent to which AGPS was necessary for maintaining C16:0e/20:4 PAe C16:0e/20:4 PAe C16:0e/20:4 PAe C16:0e LPAe C16:0e LPAe C16:0e LPAe the pathogenic features of cancer cells. We generated two inde- C16:0e/18:1 PIe C16:0e/18:1 PIe C16:0e/18:1 PCe C16:0e/20:4 PIe C16:0e/20:4 PIe C16:0e/20:4 PCe pendent stable short hairpin knockdown lines of AGPS (shAGPS-1 C16:0e/18:1 PCe C16:0e/18:1 PCe C16:0e LPCe and shAGPS-2) with greater than 90% knockdown in breast C16:0e/20:4 PCe C16:0e/20:4 PCe C16:0e/18:1 PEe C16:0e LPCe C16:0e LPCe C18:0e/C18:1 PEe 231MFP and melanoma C8161 cancer cells (Fig. 2A; Fig. S2A). C16:0e/18:1 PEe C16:0e/18:1 PEe C16:0e/20:4 PEe fi C16:0e/20:4 PEe C16:0e/20:4 PEe C16:0e LPEe AGPS inactivation results in a signi cant decrease in cellular C16:0e LPEe C16:0e LPEe C16:0e/18:1 PSe C16:0e/18:1 PSe C16:0e/18:1 PSe C16:0e/18:1 PGe motility, invasiveness, and anchorage-independent growth in soft C16:0e/18:1 PGe C16:0e/18:1 PGe C16:0e LPGe agar in both breast and melanoma cancer cells and reduces cell C16:0e/20:4 PGe C16:0e/20:4 PGe C16:0p/C20:4 PEp C16:0p/C20:4 PEp survival in breast cancer cells (Fig. 2 B–E; Fig. S2 B–E). AGPS inactivation also reduces breast 231MFP and melanoma C8161 tu- lower levels higher levels fi F F of metabolite of metabolite mor xenograft growth in immune-de cient mice (Fig. 2 ; Fig. S2 ). We also stably overexpressed AGPS in the less aggressive Fig. 1. AGPS is highly expressed in aggressive cancer cells, primary human breast (MCF7) and melanoma (MUM2C) cancer cells to de- tumors, and RAS-transformed cells. (A) AGPS gene expression is heightened termine whether AGPS was sufficient to confer pathogenic fea- across aggressive breast (231MFP) and melanoma (C8161) cancer cells (in tures (Fig. S2G). AGPS overexpression increased cell migration red) compared with less aggressive breast (MCF7) and melanoma (MUM2C) and serum-free cell survival in situ and tumor xenograft growth cancer cells (in black), as measured by qPCR. (B) AGPS gene expression is H I fi in vivo (Fig. S2 and ). Our results indicate that AGPS not only signi cantly higher in Nottingham grade 1 (low-grade) and grade 2 (in- is important to maintain aggressive and tumorigenic features, but termediate-grade), as well as grade 3 (high-grade) primary human breast fi tumors compared with normal breast tissue as measured by qPCR. (C)AGPS also is suf cient to confer these effects in less aggressive human gene expression is also significantly higher in ER(+)/PR(+)andER(−)/PR(−) cancer cells. human breast tumors compared with normal breast tissue. (D)HRASex- pression is higher in grade 1/2 and 3 primary human breast tumors com- Functional Metabolomics Reveals Widespread Alterations in Cellular pared with normal breast tissue. HRAS expression from matching normal Lipid Levels on AGPS Knockdown or Overexpression. We next wanted tissue and breast tumors is significantly correlated with AGPS expression to understand the mechanism through which AGPS drives cancer with a Pearson correlation coefficient of r = 0.59. (E)HRASexpressionis pathogenicity. We performed both targeted and untargeted met- higher in aggressive breast and melanoma cancer cells compared with less abolomic analyses to comprehensively identify alterations in can- aggressive cells. (F) HRAS overexpression in MCF10A nontransformed cer cell metabolites upon AGPS knockdown in 231MFP breast mammary epithelial cells increases AGPS expression. (G) Aggressive breast and C8161 melanoma cells (Fig. 3 A–C; Figs. S3 A–C and S4 A–C; and melanoma cancer cells and HRAS-transformed MCF10A cells possess Dataset S1). We used single reaction monitoring (SRM)-based significantly higher levels of multiple species of ether lipids compared with targeted metabolomic analysis to quantify the levels of ∼100 less aggressive or empty vector-infected MCF10A control cells, respectively. commonlipids,anduseduntargetedmetabolomicprofiling to Data in A–F are presented as mean ± SEM; n = 4 per group for A, E,andF, fi > = – – = – fi – comparatively pro le levels of an additional 3,000 ions. We used n 7 26 per group for B D,andn 4 5pergroupforG.Signi cance in A F the METLIN database to generate putative identifications for many is presented as *P < 0.05 compared with less-aggressive cancer cells (A and E), fi normal breast tissue (B–D), or empty vector-infected MCF10A cells (F). Sig- of the altered metabolites and quanti ed these metabolites by SRM fi < analysis (22). We then filtered for metabolites that were changed ni cance in G is presented as P 0.05 for lipid designations that are bolded fi in red comparing 231MFP to MCF7, C8161 to MUM2C, or HRAS-10A to signi cantly in both shAGPS lines (shAGPS-1 and shAGPS-2). MCF10A groups. Consistent with the function of AGPS as an early critical step in ether lipid synthesis, AGPS inactivation in 231MFP and C8161 cancer cells showed global reduction in the levels of multiple structural ether lipid and plasmalogen species, including PCe, low-, moderate-, and high-grade primary human breast tumors LPCe, PEe, LPEe, PSe, PGe, monoalkylglycerol ether (MAGe), fi D and correlates signi cantly with AGPS expression (Fig. 1 ). We PAe, PIe, lysophosphatidyl inositol-ether (LPIe), PEp, lysophos- also find that HRAS expression is higher in the aggressive 231MFP phatidyl ethanolamine-plasmalogen (LPEp), phosphatidylcholine- breast and C8161 melanoma cancer cells compared with the less- plasmalogen (PCp), and phosphatidyl serine-plasmalogen (PSp)

aggressive MCF7 breast and MUM2C melanoma cancer cells (Fig. (Fig. 3 A–C; Figs. S3 A–C and S4 A–C; Dataset S1). We also find BIOCHEMISTRY 1E). Consistent with this association between HRAS and AGPS, lower levels of the signaling ether lipids LPAe and LPA-plasmalogen

Benjamin et al. PNAS | September 10, 2013 | vol. 110 | no. 37 | 14913 Downloaded by guest on September 25, 2021 suggesting that additional enzymatic pathways may be required, such as cyclooxygenase (16), to generate these lipids. AB C Isotopic Fatty Acid Tracing Reveals Alterations in Arachidonate Utilization on Inactivation of AGPS. Our results showing that AGPS knockdown leads to lower and eicosanoid levels and an increase in certain arachidonoyl-containing acyl- implies a shift in fatty acid use toward specific diacylated structural lipids and away from lysophospho- lipids and eicosanoids. Consistent with this premise, isotopic fatty acid tracing with deuterated d8-arachidonic acid in 231MFP breast cancer cells reveals heightened d8-incorporation into DEFacylglycerophospholipids and reduced incorporation into ether lipids and eicosanoids on AGPS knockdown (Fig. 3D). In con- trast, isotopic tracing with d4-palmitic acid did not show in- creased d4-incorporation into acylglycerophospholipids, despite reduced incorporation into ether lipids upon AGPS knockdown (Fig. 3E). These results indicate preferential incorporation of arachidonic acid into acylglycerophospholipids, compared with saturated fatty acids. This shift in fatty acid utilization may be attributed to either increased lysophospholipid acyltransferase activity, reduced Fig. 2. AGPS ablation leads to impairments in breast cancer pathogenicity. hydrolysis, or lower cyclooxygenase activity. Upon (A) AGPS expression was stably knocked down in 231MFP breast cancer cells testing for these possibilities by quantitative PCR (qPCR), we using two independent shRNA oligonucleotides (shAGPS-1 and shAGPS-2), found that AGPS knockdown led to increased levels of LPC acyl resulting in >90% reduction in AGPS expression compared with shControl – transferase (LPCAT1), but no change in cytosolic cells determined by qPCR. (B E) AGPS inactivation in 231MFP cells decreases A2 (PLA2G4A) or cyclooxygenase 2 (PTGS2) expression (Fig. serum-free cell survival (B), cell migration (C), invasion (D), and anchorage- 4A). These results indicate that AGPS knockdown directs fatty independent growth in soft-agar (E). Serum-free cell survival was de- termined by measuring cell viability in serum-free media 24 h after seeding acid use to generating arachidonoyl acylglycerophospholipids 10,000 cells using a WST cell viability assay. Migration and invasion assays through increased acylation of fatty acids onto glycerol-phos- were performed by transferring cancer cells to serum-free media for 4 h be- phate and lysophospholipids, depleting the arachidonic acid pool fore seeding 50,000 cells into inserts with 8-μmporesizecontainingmem- that generates eicosanoids. branes coated with collagen (10 μg/mL) or BioCoat Matrigel, respectively. Migrated or invaded cells were fixed and stained after 8 h, and these cells AGPS Affects Cancer Pathogenicity Through Multiple Lipid Signaling were counted over four independent fields of view at 400× magnification and Pathways. Among the many lipidomic changes conferred by averaged for each biological replicate. For soft-agar assays, 4,000 cells were AGPS knockdown, we were particularly intrigued by the lower seeded in agar and colonies were counted by 3-(4,5-dimethylthiazol-2-yl)-2,5- levels of LPAe and PGE2 in 231MFP breast cancer cells and diphenyl tetrazolium bromide (MTT) staining after 4 wk. (F)AGPSknockdown LPAe and LPA in C8161 melanoma cancer cells. Both LPAe and impairs tumor growth in immune-deficient SCID mice compared with shControl its acyl-LPA counterpart act as oncogenic signaling lipids that bind 6 cells; 2 × 10 231MFP cells/100 μL were injected s.c. into the flank of female LPA receptors to drive multiple aspects of cancer (13). PGE2 is mice, and tumor growth was measured using calipers. Data are presented as ± = – fi < also a signaling lipid that acts through EP2 receptors to fuel mean SEM; n 3 7 per group. Signi cance is presented as *P 0.05 com- proliferative, malignant, and tumorigenic features of cancer (16). pared with shControl. Consistent with their function, C18:0e LPAe fully rescues and PGE2 partially rescues the migratory and invasive deficits con- (LPAp) that, similar to their acyl-LPA counterparts, stimulate ferred by AGPS knockdown in 231MFP breast cancer cells (Fig. 4 B and C). LPAe, but not PGE2, fully rescues the migratory and LPA receptors (12, 13) (Fig. 3 B and C; Fig. S3; Dataset S1). invasive impairments in shAGPS C8161 cells, consistent with the Intriguingly, AGPS knockdown also lowers levels of saturated lipidomic signature showing that eicosanoids are not detected in and unsaturated free fatty acids (FFAs) in both breast and mela- these cells. We find that C18:0 LPA also partially or fully rescues noma cancer cells, with a striking reduction in the levels of arachi- – the migratory impairments in shAGPS 231MFP and C8161 cells, donic acid derived signaling eicosanoids, E2/D2 respectively (Fig. S6 C and D), likely because either acyl-LPA or (PGE2/PGD2) (16), in 231MFP breast cancer cells (Fig. 3C; Fig. C LPAe can rescue overall reduced LPA receptor signaling caused S4 ). We also observe increases in the levels of several diacylated by lower LPAe or LPA levels. We also show that palmitate (C16:0 glycerophospholipids, with a particular enrichment in arachidonoyl- FFA) partially rescues the pathogenic deficits of AGPS knock- containing acyl-glycerophospholipids (e.g., C16:0/C20:4 PC, C16:0/ down, likely through the generation of acyl-LPA (Fig. 4 B and C; C20:4 PE, and C16:0/C20:4 PG), coupled with lower levels of Fig. S6 C and D). Indeed, we show that d4-C16:0 FFA labeling of B several acyl-lysophospholipids (e.g., LPA, LPE, and LPC) (Fig. 3 231MFP cancer cells results in the generation of d4-C16:0 LPA C B C and ; Fig. S4 and ; Dataset S1). Overall, this lipidomic sig- (Fig. S6E). We note that no other ether lipids such as PAF, C16:0e/ nature suggests that AGPS knockdown has broader effects beyond C20:4 PCe, or C16:0e LPCe rescue the oncogenic impairments, ether lipid pathways that include fatty acid, eicosanoid, and acyl- confirming the specificity of our proposed mechanism. glycerophospholipid metabolism. Interestingly, C18:0e LPAe, but not PGE2, significantly res- Consistent with the conjecture of Welsh et al. that lack of AGPS cues the up-regulation of LPCAT1 expression in AGPS knock- activity is responsible for the general deficiency of ether lipids in down breast cancer cells, suggesting that the broader alterations MCF7 breast cancer cells (23), we find that AGPS overexpression in fatty acid, acylglycerophospholipid, and PGE2 levels may be inMCF7cellsissufficient to increase the levels of several ether downstream of reduced LPAe signaling (Fig. 4D). In AGPS- lipids. We also show that oleic and arachidonic acid levels are also overexpressing cancer cells that show higher levels of LPAe but increased (Fig. S5A; Dataset S1). In AGPS overexpressing MUM2C undetectable levels of PGE2, we show that the increased mi- melanoma cancer cells, we see a similar increase in the levels of gration observed in AGPS-overexpressing MCF7 and MUM2C ether lipids, as well as reduced levels of arachidonoyl-containing cancer cells is significantly reversed by an LPA receptor antago- glycerophospholipids (Fig. S5B; Dataset S1). PGE2/PGD2 levels nist Ki16425, but not by the cyclooxygenase inhibitor ibuprofen were not detectable in either mock or AGPS-overexpressing cells, (Fig. S7).

14914 | www.pnas.org/cgi/doi/10.1073/pnas.1310894110 Benjamin et al. Downloaded by guest on September 25, 2021 Downloaded by guest on September 25, 2021 eta iis n asstelvl fsvrldayae lcrpopoiis ( glycerophospholipids. diacylated several of levels the raises and lipids, neutral signi are that levels metabolites which in The groups, shControl and shAGPS between ion each eaoois akrbu hdn nteha a orsod ohge eaielvl fmtblt,weeswieo ihe lesaigidctslower indicates shading blue lighter or white whereas metabolite, of levels relative higher to corresponds map ( heat levels. the on shading blue Darker metabolomics. signi are that metabolites ellrlpd.Dt in Data lipids. ( cellular lipids. cellular complex certain into incorporation The respectively. dots, red and blue as represented are knockdown AGPS on levels in increasing or decreasing are that Metabolites cells. spoie in provided is in # ( ( eicosanoids. and acids, fatty 3. Fig. ejmne al. et Benjamin Signi P P aae S1 Dataset < < .5 leain ntelvl fmn eaoie.Ec on ntevlaopo orsod oadsic o eetdi hGS1adsCnrl231MFP shControl and shAGPS-1 in detected ion distinct a to corresponds plot volcano the on point Each metabolites. many of levels the in alterations 0.05) y fi .5btend-1: n 8C04FAtetdsAP s acigsCnrlgop.Alrwdt o eaielvl n bouelvl r shown are levels absolute and levels relative for data raw All groups. shControl matching vs. shAGPS FFA-treated d8-C20:4 and d4-C16:0 between 0.05 ac in cance B D xsidctsterltv odcag ntelvl ftemtblt ewe hoto n hGSgop.( groups. shAGPS and shControl between metabolite the of levels the in fold-change relative the indicates axis C d8-C20:4 A C eaooi pro Metabolomic ersnaielpd from lipids Representative ) shAGPS-2 shAGPS-1 shControl

FFA relative levels 0.0 0.5 1.0 1.5 0.0 relative0.5 levels1.0 1.5 shControl shCont -+ PGE2/PGD2 ute uniaiedtisfrLA n Pplpd r hw in shown are lipids LPAp and LPAe for details quantitative Further . d7-PGE2/PGD2 shAGPS-1 i.S4 Fig. D

shAGPS-2 ** * and shAGPS Isotopic ArachidonicAcidTracingintoComplexLipids n h eaooi pro metabolomic the and , # * E relative levels 0.0 0.5 1.0 1.5

A C16:0e/18:1 PCe speetda * as presented is shControl – C18:0e LPAe fi fi E shAGPS-1 0.0 relative0.5 1.0 levels1.5 2.0 2.5 C18:0e/C18:1 PCe atyatrdi ees( levels in altered cantly igo GSkokoni ratcne el eel iepedatrtosi h eeso te iis acylglycerophospholipids, lipids, ether of levels the in alterations widespread reveals cells cancer breast in knockdown AGPS of ling C16:0/d8-C20:4 PC r rsne smean as presented are * -+

hotsAP hotsAP hotsAP hotshAGPS shCont shAGPS shCont shAGPS shCont shAGPS shCont shAGPS-2 A C16:0e/20:4 PCe * – C

* C18:0e/C20:4 PCe eaooi nlsso GSkokon21F el oprdwt hoto el.( cells. shControl with compared cells 231MFP knockdown AGPS of analyses Metabolomic ) relative levels 0.0 0.5 1.0 1.5 C16:0e LPCe shControl B # * sh C20:4 FFA C18:0e LPCe r hw sbrgah.AP ncdw o nylwr h eeso te iis u loftyais ioaod,and eicosanoids, acids, fatty also but lipids, ether of levels the lowers only not knockdown AGPS graphs. bar as shown are AGPS fold-change * P sh AGPS-2-1

C16:0e/18:1 PEe 100

relative levels Metabolic SignatureofAGPSKnockdownin231MFPBreastCancerCells < * 0 1 2 3 4 C16:0/d8-C20:4 PE 10 - - - - -+ -+ -+ Levels ofRepresentativeLipidsAlteredUponAGPSInactivation .5btend-1: F n 8C04FAgop oprdwt oiooi at cdtetetgop,and groups, treatment acid fatty nonisotopic with compared groups FFA d8-C20:4 and FFA d4-C16:0 between 0.05

C18:0e/C18:1 PEe 1 relative levels 10 0.0 0.5 1.0 1.5

* C16:0e/20:4 PEe shCont 0 fi C16:0 FFA

eo GSoeepeso nMF n U2 el spoie in provided is cells MUM2C and MCF7 in overexpression AGPS of le sh rol C18:0e/C20:4 PEe Significantly AlteredMetabolitesin231MFPBreastCancerCells E

P AGPS-1 10 agtdmtblmcaayi fiooi 4C60FA(amtcai)icroainit eti complex certain into incorporation acid) (palmitic FFA d4-C16:0 isotopic of analysis metabolomic Targeted ) * #

* sh ± < AGPS-2 C18:0e LPEe * SEM; .5 pnAP ncdw n21F el.Teemtbltswr quanti were metabolites These cells. 231MFP in knockdown AGPS upon 0.05) -1

C18:0p LPEp 786 587 relative levels 847 317 0.0 0.5 1.0 1.5 421 C16:0e/d8-C20:4 PCe 454 482

0.0 relative0.5 1.0 levels1.5 2.0

10 2 C16:0/C20:4 PC 771 332

C16:0e/18:1 PSe 81 509 599 shControl 735 747 732 n 255 sh C18:0e/C18:1 PSe -2 822 782 = AGPS * 279 ether 283 829 * shAGPS-2-1 857 768 4 reduced metabolites elevated metabolites 775 C18:0e/C20:4 PSe lipids 10 * – 754 of metabolite 392 768 423 746 e ru.Signi group. per 5 # lower levels * 752 724 C16:0e/18:1 PGe 778 718 -3 850 466 780 796 748 relative levels C16:0/C18:1 PC 468 0.0 0.5 1.0 1.5 798

C18:0e/C18:1 PGe 726 754

relative levels 401 740 10 738 0.0 0.5 1.0 1.5 shCont 794 C18:0p/d8-C20:4 PEp shAGPSrol C16:0e MAGe 908 746 659 728 -4 p-value 749 776 750 1000 sh 719 -1 906 868 AGPS-2 C16:0e/18:1 PAe 687 968 702 704 * 10 898 996 763 843 C18:0e/C18:1 PAe 972 i.S3 Fig. 876 748 D C16:0/C20:4 PE 874 921 -5 relative levels 774 894 0 1 2 3 agtdmtblmcaayi fiooi 8C04FA(rcioi acid) (arachidonic FFA d8-C20:4 isotopic of analysis metabolomic Targeted ) 848

# shControl

* C18:0e/C20:4 PAe 756 900 603 599 sh 846 910 AGPS-1 946 10 870

fi C18:0e LPAe 876 d4-C16:0 878 924 *

shAGPS 954 E 926 h eaooi pro metabolomic The . ac sepesdin expressed is cance 722 998 577 712 -6 * C18:0p LPAp 429 -2 709 of metabolite higher levels FFA fi relative levels 974 0.0 0.5 1.0 1.5 atyatrd( altered cantly

C16:0/d4-C18:1 PC C18:0e LPIe 10 C16:0/C18:1 PE hotsAP hotsAP hotsAP hotshAGPS shCont shAGPS shCont shAGPS shCont shAGPS shCont -+ 736

Isotopic PalmiticAcidTracingintoComplexLipids relative levels 0.0 0.5 1.0 1.5 2.0 940 891 shCon 16:0p/20:4 PEp 351 -7 970 976 888 * sh trol AGPS-1 18:0p/20:4 PEp

sh * AGPS 10 936 PNAS

* 18:0p/20:4 PCp * -2 -8 950 C16:0e/C18:1 PCe

16:0p/20:4 PSp 978 relative levels 0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5 relative levels 10 P shControl C16:0 FFA | 303 C16:0/d4-C18:1 PE -+ 902 912 < shAGPS-1 -9

etme 0 2013 10, September fatty

C C18:0 FFA .5 r ipae otergto h otdline. dotted the of right the to displayed are 0.05) fi shAGPS-2 ** eo GSiatvto nC11mlnm cells melanoma C8161 in inactivation AGPS of le 10 * s* as acids C18:1 FFA 980 956 C20:4 FFA -10 P C16:0e/C18:1 PEe * 0.0 relative0.5 levels1.0 1.5

< shControl PGD2/PGE2 prostaglandins i.S5 Fig.

.5cmae ihsCnrlgroups. shControl with compared 0.05 sh AGPS 0.0 0.5 1.0 1.5 B relative levels C16:0/C20:4 PC A * C18:0p/d4-C16:0 PCp sh h etmpsosalidenti all shows map heat The ) AGPS-1 GSkokonsossigni shows knockdown AGPS ) - - -+ -+ -2 * C16:0/C18:1 PE .

* C16:0/C20:4 PE 0.0 0.5 1.0 1.5 relative levels C16:0p/C20:4 PCp acyl | shControl C16:0 LPE glycero- * # fi s

o.110 vol. hAGPS-1

db R-ae targeted SRM-based by ed C16:0/C18:1 PG shAGPS-2 ** x C16:0/C20:4 PG 0.0 relative0.5 levels1.0 1.5 xsdenotes axis C18:0p/d4-C16:0 PEp C18:0/C18:1 PI

0.0 relative0.5 levels1.0 1.5 C18:0p/C20:4 PEp | shControl * C18:0 LPI

o 37 no. shAGPS C18:0/C18:0/C18:0 TAG shAGPS-1 * neutral lipids # *

-2 * C18:0/C18:1/C18:0 TAG P au of value | fi 14915 cant fi ed

BIOCHEMISTRY We profiled expression of LPCAT1 in primary human breast mediated changes to the lipidome may yield further insights into tumors to determine whether LPCAT1 expression was lower the roles of AGPS in cancer. Although we have established the compared with normal breast tissue. Instead, we find that LPCAT1 importance of AGPS in controlling the levels of oncogenic sig- expression is significantly higher in primary breast tumors com- naling lipids, such as LPAe, LPA, and eicosanoids, there may be pared with normal breast tissue (Fig. S8). Our finding that reduced other enzymes or regulatory pathways, even within the ether lipid LPAe signaling leads to an up-regulation in LPCAT1 expression metabolic pathway, that can be targeted to achieve similar may occur only upon knockdown of AGPS or depletion of cellular effects. For example, we have reported that monacylglycerol li- ether lipids, under conditions in which cancer cells have adapted to pase is up-regulated across aggressive cancer cells and primary a state of higher ether lipid levels. tumors where it controls release of FFAs from monoacyl- Overall, AGPS expression is heightened in aggressive human to fuel generation of FFA-derived oncogenic signal- cancer cells and primary human breast tumors, and AGPS in- ing lipids, such as LPA and PGE2 to drive malignant features activation significantly impairs aggressive features of cancer cells of cancer (18). Chiang et al. demonstrated that blocking of and attenuates tumor growth, not only through lowering the levels KIAA1363, a deacetylase of 2-acetyl MAGe, lowers levels of its of oncogenic ether lipids, but also through altering arachidonic neutral ether lipid product MAGe, which decreases the levels of acid utilization away from protumorigenic eicosanoids and lyso- its downstream metabolite LPAe, leading to impaired cellular phospholipids (Fig. 4E). migration and tumor growth (15). It will be of future interest to determine the effects of polypharmacologically blocking several Conclusion of these pathways simultaneously in cancer cells to achieve We show here that AGPS serves as a metabolic node in cancer maximal anticancer activity. Alternatively, controlling the supply cells that maintains levels of structural and signaling ether lipids of fatty alcohols in cancer cells may also modulate levels of ether and also modulates fatty acid utilization pathways to favor gen- lipids similarly to controlling AGPS expression (24). eration of key oncogenic lysophospholipids and eicosanoids, which In conclusion, we provide a potential mechanistic link between promote survival, migratory, invasive, and tumorigenic features. the historical associations between aggressive or hyperproliferative We used advanced metabolomic platforms to gain broad overview tumors and higher levels of ether lipids (7). Our studies indicate into changes in cancer cell lipid metabolism that are controlled by a previously unrecognized cross-talk between ether lipid, fatty acid, AGPS, but there are many metabolic alterations that we have eicosanoid, and acylglycerophospholipid metabolism in cancer not yet been able to identify. These yet uncharacterized AGPS- cells that leads to an optimal landscape of lipid signaling pathways

A LPCAT1 PLA2G4A PTGS2 Fig. 4. AGPS fuels cancer pathogenicity in breast 4 1.5 1.5 cancer cells through altering fatty acid utilization to * favor the generation of oncogenic signaling lipids. (A) 3 1.0 1.0 AGPS knockdown results in altered levels of not only 2 ether lipids, but also reduced levels of fatty acids and 0.5 0.5 1 eicosanoids and increased levels of acylglycerophos- fold over control fold over control fold over control pholipids. This may be due to heightened acyltrans- 0 0.0 0.0 ferase or reduced phospholipase or cyclooxygenase activity. AGPS ablation results in a significant increase shAGPS shAGPS shAGPS shControl shControl shControl in the expression of lysophosphatidyl acyl- transferase 1 (LPCAT1) without significantly altering B migration C invasion D LPCAT1 expression the expression of cytosolic phospholipase A2 (PLA2G4) 150 150 2.5 * 2.5 * * or cyclooxygenase 2 (PTGS2), as determined by qPCR. # # 2.0 2.0 (B and C) The migratory (B) and invasive (C) impair- 100 ##* 100 # * * # 1.5 1.5 ments in shAGPS 231MFP cells are significantly rescued * * * * * upon treatment of cells with low concentrations (100 ** 1.0 1.0 50 50 * ** % of control nM) of C18:0e LPA, PGE2, and C16:0 FFA, but not other % of control 0.5 0.5 fold over control fold over control ether lipids such as C16:0e LPCe, C16:0e/C20:4 PCe, and - 0 - 0 0.0 0.0 C16:0e PAFe. Treatment with the aforementioned lip- +PCe +PCe + FFA + FFA +PAFe +PGE2 +LPAe +PAFe +PGE2 +LPAe ids was initiated concurrently with the seeding of cells + LPCe + LPCe shAGPS shAGPS shAGPS shAGPS shControl shControl shControl shControl shControl shControl for assessment of cancer cell migration (8 h) and in- shAGPS shAGPS LPAe -+-+ PGE2 -+-+ vasion (24 h). (D) Treatment with C18:0e LPAe (100 nM), but not PGE2 (100 nM), significantly rescued the AGPS Regulates Ether Lipids, Arachidonoyl-acyl-phospholipids, and Eicosanoids E O heightened expression of LPCAT1 conferred by AGPS O O R1 knockdown, as assessed by qPCR. (E) Model depicting O OX the metabolic role of AGPS in exerting control over ether lipid metabolism, fatty acid metabolism, and glycerophospholipid metabolism. We show that AGPS 1-acyl-2-arachidonoyl knockdown in 231MFP breast cancer cells leads to re- sn glucose LPCAT1 - -glycerophospholipids duced levels of ether lipids, including the oncogenic

O LPAe signaling lipid (boxed in red), as well as an al- lysophosphatidic glycerol-3-phosphate PLA2G4A acid OH teration in arachidonic acid utilization toward mem- dihydroxyacetone brane phospholipids through LPAe-mediated LPCAT1 phosphate O O up-regulation, at the expense of generating oncogenic arachidonic acid O OH O O O (boxed in red). R1 denote acyl or alkyl O P OH O- chains and X refers to lipid head-group such as phos- acyl-glycerone-phosphate pyruvate phocholine, phosphoethanolamine, or phosphate. Data HO fatty alcohol OH are presented as mean ± SEM; n = 3 per group. Lipid AGPS PGE2 lactic acid rescue of migratory and invasive impairments on O AGPS knockdown in C8161 melanoma cells and C18:0 O O O O P OH O O OR1 LPA rescue experiments are provided in Fig. S6. Sig- O- HO alkyl-glycerone-phosphate O P OH R2 O nificance is represented as *P < 0.05 compared with OH OX # LPAe ether lipids shControl and P < 0.05 compared with DMSO-trea- ted shAGPS groups.

14916 | www.pnas.org/cgi/doi/10.1073/pnas.1310894110 Benjamin et al. Downloaded by guest on September 25, 2021 that subserve cancer pathogenesis. Manipulating this landscape Overexpression Studies in Human Cancer Cell Lines. Stable AGPS over- may yield novel strategies for thwarting the lipid signals that drive expression was achieved by subcloning the AGPS gene into the pMSCVpuro cancer pathogenicity and may provide unique approaches to vector (Clontech), generating retrovirus using the AmphoPack-293 Cell Line, cancer therapy. as described above with the RNA interference studies. The human AGPS was subcloned into the pMSCVpuro (Clontech) using Xho1 and AgeI restriction Materials and Methods sites using the following primers: 5′GTACGTACCTCGAGAGGCGGCGGCTG-3′ Cell Lines. The C8161 and MUM2C lines were provided by Benjamin Cravatt and 5′GTACGTACGAATTCGTTTCTGTTTCC-3′. (The Scripps Research Institute, La Jolla, CA). MCF10A and HRAS–transformed cells were obtained from Stefano Piccolo at the University of Padua (25). The Cancer Phenotypic Studies. Migration, invasion, cell proliferation, and survival 231MFP cells were generated from explanted xenograft tumors of MDA-MB- studies were performed as described previously (18). Details are providing in 231 cells, as described previously (26). PC3 and LNCaP cells were obtained SI Materials and Methods. from ATCC. Tumor Xenograft Studies. Human tumor xenografts were established by Conditions. HEK293T cells were cultured in DMEM media con- transplanting cancer cells ectopically into the flank of C.B17 SCID mice taining 10% (vol/vol) FBS and maintained at 37 °C with 5% (wt/vol) CO .C8161 2 (Taconic Farms) as described previously (18). Briefly, cells were washed two cells were cultured in RPMI1640 media containing 10% FBS and glutamine and times with PBS, trypsinized, and harvested in serum-containing medium. maintained at 37 °C at 5% (wt/vol) CO2. 231MFP cells were cultured in L15 media containing 10% FBS and glutamine and were maintained at 37 °C in Next, the harvested cells were washed two times with serum-free medium and resuspended at a concentration of 2.0 × 104 cells/mL, and 100 μLwas 0% CO2. MCF10A cells were cultured as previously described (25). injected. Growth of the tumors was measured every 3–6 d with calipers. qPCR. qPCR was performed using the manufacturer’s protocol for Fischer Maxima SYBRgreen, with 10 μM primer concentrations. Primer sequences Metabolomic Profiling of Cancer Cells. Metabolite measurements were con- were derived from Primer Bank (27). ducted using modified versions of previous procedures (18, 28). Details are provided in SI Materials and Methods. Breast Tumor Array qPCR. Breast cancer tumor array 1 was purchased from Origene,andqPCRwasperformedonthearrayusingtheprotocoldescribedabove. Analysis of Isotopic Incorporation of Palmitate and Arachidonate into Lipids. Isotopically labeled fatty acid incorporation into lipids was measured by Constructing AGPS Knockdown Cells. We used shRNA using two independent labeling cells with d4-palmitic acid or d8-arachidonic acid and measuring silencing oligonucleotides to knockdown the expression of AGPS using pre- isotopic incorporation into lipids. Cells were treated either with palmitate, viously described procedures (18). For construction of stable shRNA knock- d4-palmitate, arachidonate, or d8-arachidonate (10 μM). Cells were har- down lines, lentiviral plasmids (pLKO.1) containing shRNA (Open Biosystems) vested 4 h after labeling, and the nonpolar metabolome was extracted as against human AGPS were transfected into HEK293 cells using Fugene previously described and analyzed by SRM-based targeted LC-MS/MS for fi (Roche). Lentivirus was collected from ltered culture media and delivered to the presence of isotopically labeled lipids. the target cancer cell line with polybrene. These target cells were subsequently selected over 3 d with 1 μg/mL puromycin. The short-hairpin sequences used ACKNOWLEDGMENTS. We thank the members of the Nomura Research for constructing the two independent AGPS knockdowns were as follows: Group and the Luo laboratory for critical reading of the manuscript. This shRNA#1: AATTCGCTCAAACATTCCTTC; shRNA #2: AAGGATTTCTTCTCTAG- work was supported by National Institutes of Health Grants R21CA170317, CAGC. The control shRNA was targeted against GFP with the target sequence R01CA172667, and R00DA030908 (to D.I.B., A.C., L.S.R., S.M.L., M.M.M., and GCAAGCTGACCCTGAAGTTCAT. We confirmed knockdown by qPCR. D.K.N.) and Grant R01CA101891 (to K.L.) and the Searle Scholar Foundation (D.K.N.).

1. Vander Heiden MG, Cantley LC, Thompson CB (2009) Understanding the Warburg 15. Chiang KP, Niessen S, Saghatelian A, Cravatt BF (2006) An enzyme that regulates effect: The metabolic requirements of cell proliferation. Science 324(5930):1029–1033. ether lipid signaling pathways in cancer annotated by multidimensional profiling. 2. Benjamin DI, Cravatt BF, Nomura DK (2012) Global profiling strategies for mapping Chem Biol 13(10):1041–1050. dysregulated metabolic pathways in cancer. Cell Metab 16(5):565–577. 16. Wang D, Dubois RN (2010) Eicosanoids and cancer. Nat Rev Cancer 10(3):181–193. 3. Menendez JA, Lupu R (2007) Fatty acid synthase and the lipogenic phenotype in 17. Jessani N, Liu Y, Humphrey M, Cravatt BF (2002) Enzyme activity profiles of the se- cancer pathogenesis. Nat Rev Cancer 7(10):763–777. creted and membrane proteome that depict cancer cell invasiveness. Proc Natl Acad 4. Louie SM, Roberts LS, Nomura DK (2013) Mechanisms linking obesity and cancer Sci USA 99(16):10335–10340. [published online ahead of print March 5, 2013]. Biochim Biophys Acta, doi:10.1016/j. 18. Nomura DK, et al. (2010) Monoacylglycerol lipase regulates a fatty acid network that bbalip.2013.02.008. promotes cancer pathogenesis. Cell 140(1):49–61. 5. Snyder F, Wood R (1969) Alkyl and alk-1-enyl ethers of glycerol in lipids from normal 19. Nomura DK, et al. (2011) Monoacylglycerol lipase exerts dual control over endo- and neoplastic human tissues. Cancer Res 29(1):251–257. cannabinoid and fatty acid pathways to support prostate cancer. Chem Biol 18(7): 6. Roos DS, Choppin PW (1984) Tumorigenicity of cell lines with altered lipid composi- 846–856. tion. Proc Natl Acad Sci USA 81(23):7622–7626. 20. Pylayeva-Gupta Y, Grabocka E, Bar-Sagi D (2011) RAS oncogenes: Weaving a tumori- 7. Howard BV, Morris HP, Bailey JM (1972) Ether-lipids, -glycerol phosphate de- genic web. Nat Rev Cancer 11(11):761–774. hydrogenase, and growth rate in tumors and cultured cells. Cancer Res 32(7): 21. Dawson PJ, Wolman SR, Tait L, Heppner GH, Miller FR (1996) MCF10AT: A model for 1533–1538. the evolution of cancer from proliferative breast disease. Am J Pathol 148(1):313–319. 8. Albert DH, Anderson CE (1977) Ether-linked glycerolipids in human brain tumors. 22. Tautenhahn R, et al. (2012) An accelerated workflow for untargeted metabolomics Lipids 12(2):188–192. using the METLIN database. Nat Biotechnol 30(9):826–828. 9. Wallner S, Schmitz G (2011) the neglected regulatory and scavenging 23. Welsh CJ, et al. (1994) Accumulation of fatty alcohol in MCF-7 breast cancer cells. Arch lipid species. Chem Phys Lipids 164(6):573–589. Biochem Biophys 315(1):41–47. 10. Brites P, Waterham HR, Wanders RJ (2004) Functions and biosynthesis of plasmal- 24. Honsho M, Asaoku S, Fujiki Y (2010) Posttranslational regulation of fatty acyl-CoA ogens in health and disease. Biochim Biophys Acta 1636(2–3):219–231. reductase 1, Far1, controls ether glycerophospholipid synthesis. J Biol Chem 285(12): 11. Tsukahara T, Haniu H, Matsuda Y (2013) Effect of alkyl glycerophosphate on the 8537–8542. activation of proliferator-activated receptor gamma and glucose uptake 25. Cordenonsi M, et al. (2011) The Hippo transducer TAZ confers cancer stem cell-related in C2C12 cells. Biochem Biophys Res Commun 433(3):281–285. traits on breast cancer cells. Cell 147(4):759–772. 12. Xu Y, Tanaka M, Arai H, Aoki J, Prestwich GD (2004) Alkyl lysophosphatidic acid and 26. Jessani N, et al. (2004) Carcinoma and stromal enzyme activity profiles associated with fluoromethylene phosphonate analogs as metabolically-stabilized agonists for LPA breast tumor growth in vivo. Proc Natl Acad Sci USA 101(38):13756–13761. receptors. Bioorg Med Chem Lett 14(21):5323–5328. 27. Spandidos A, Wang X, Wang H, Seed B (2010) PrimerBank: A resource of human and 13. Mills GB, Moolenaar WH (2003) The emerging role of lysophosphatidic acid in cancer. mouse PCR primer pairs for gene expression detection and quantification. Nucleic Nat Rev Cancer 3(8):582–591. Acids Res 38(Database issue):D792–D799. 14. Melnikova V, Bar-Eli M (2007) Inflammation and melanoma growth and metastasis: 28. Kopp F, et al. (2010) The glycerophospho metabolome and its influence on amino acid The role of platelet-activating factor (PAF) and its receptor. Cancer Metastasis Rev homeostasis revealed by brain metabolomics of GDE1(-/-) mice. Chem Biol 17(8): 26(3-4):359–371. 831–840. BIOCHEMISTRY

Benjamin et al. PNAS | September 10, 2013 | vol. 110 | no. 37 | 14917 Downloaded by guest on September 25, 2021