Metabolic Profiling Reveals PAFAH1B3 As a Critical Driver Of

Chemistry & Biology Article

Metabolic Proﬁling Reveals PAFAH1B3 as a Critical Driver of Breast Cancer Pathogenicity

Melinda M. Mulvihill,1 Daniel I. Benjamin,1 Xiaodan Ji,2 Erwan Le Scolan,2 Sharon M. Louie,1 Alice Shieh,1 McKenna Green,1 Tara Narasimhalu,1 Patrick J. Morris,1 Kunxin Luo,2 and Daniel K. Nomura1,* 1Department of Nutritional Sciences and Toxicology, University of California, Berkeley, 127 Morgan Hall, Berkeley, CA 94720, USA 2Department of Molecular and Cell Biology, University of California, Berkeley, GL41 Koshland Hall, Berkeley, CA 94720, USA *Correspondence: [email protected] http://dx.doi.org/10.1016/j.chembiol.2014.05.008

SUMMARY Studies over the past decade have uncovered a subpopula- tion of cancer cells termed cancer stem/precursor cells (CSCs) Many studies have identified metabolic pathways that, like embryonic stem cells, can undergo a process of epithe- that underlie cellular transformation, but the meta- lial-to-mesenchymal transition (EMT), which is associated with bolic drivers of cancer progression remain less well self-renewing and tumor-initiating capabilities, poor prognosis, understood. The Hippo transducer pathway has and chemotherapy resistance within breast tumors (Polyak and been shown to confer malignant traits on breast Weinberg, 2009). Recent studies have identified TAZ, a trans- cancer cells. In this study, we used metabolic map- ducer of the Hippo pathway, as a critical factor that is elevated in poorly differentiated human breast tumors, is correlated with ping platforms to identify biochemical drivers of poor prognosis, promotes EMT, is stabilized by EMT, and is cellular transformation and malignant progression required to sustain self-renewal and tumor-initiation capacities driven through RAS and the Hippo pathway in breast of breast CSCs (Cordenonsi et al., 2011; Hong and Guan, cancer and identified platelet-activating factor ace- 2012). Identifying altered metabolic pathways that underlie tylhydrolase 1B3 (PAFAH1B3) as a key metabolic both cellular transformation and malignant progression through driver of breast cancer pathogenicity that is upregu- TAZ may thus yield therapeutic strategies for combatting malig- lated in primary human breast tumors and correlated nant breast cancers. with poor prognosis. Metabolomic profiling suggests Whereas targeting dysregulated metabolism is a promising that PAFAH1B3 inactivation attenuates cancer therapeutic strategy for cancer treatment, nearly all research in pathogenicity through enhancing tumor-suppress- cancer metabolism has focused on well-understood metabolic ing signaling lipids. Our studies provide a map of pathways in central carbon metabolism and has largely ignored the majority of cellular metabolic pathways, despite genetic or altered metabolism that underlies breast cancer metabolic evidence of their involvement in cancer. This is in large progression and put forth PAFAH1B3 as a critical part because of our incomplete understanding of metabolic net- metabolic node in breast cancer. works in normal physiology, let alone in dysregulated or rewired biochemical networks in human cancer cells (Benjamin et al., 2012). INTRODUCTION In this study, we used proteomic, functional proteomic, and metabolomic platforms to more broadly identify commonly Cancer cells possess fundamentally dysregulated metabolism altered metabolic pathways that underlie both mammary epithe- that underlies their pathogenic features (Benjamin et al., 2012; lial cell transformation by the oncogene HRAS as well as malig- Vander Heiden et al., 2009). Targeting altered metabolism in can- nant progression, EMT, and cancer-stem-cell-like features cer cells has historically served as a validated strategy for treat- through the constitutive activation of TAZ in the MCF10A non- ing cancer patients. These therapies include chemotherapeutics transformed mammary epithelial cell line background. Through such as 5-fluorouracil, methotrexate, and anastrozole that target these profiling efforts, we have identified platelet-activating nucleotide, folate, and estrogen metabolism, respectively factor acetylhydrolase 1B3 (PAFAH1B3) as a critical enzyme in (National Cancer Institute, 2013). More recent discoveries have maintaining breast cancer cell aggressiveness and tumor growth focused on targeting other aspects of dysregulated cancer through regulating tumor-suppressing lipids. metabolism, including glycolysis, tricarboxylic acid (TCA) cycle, and lipogenesis with pyruvate kinase M2 (PKM2) activators, iso- RESULTS AND DISCUSSION citrate dehydrogenase inhibitors, and fatty acid synthase (FASN) inhibitors, respectively (Benjamin et al., 2012; Menendez and Breast Cancer Progression Model Lupu, 2007; Vander Heiden et al., 2009). Whereas targeting dys- In this study, we used integrated metabolic mapping platforms to regulated metabolism is a promising strategy for cancer treat- identify enzymes and metabolites that are dysregulated in a ment, currently available chemotherapies, including those that MCF10A breast cancer progression model, a series of isogeni- target certain aspects of cancer metabolism, fail to eradicate cally derived human mammary cell lines consisting of nontrans- the most aggressive cancers (Chambers et al., 2002). formed MCF10A cells, HRAS-transformed MCF10A-T1k cells

relative levels 10A TAZS89A

10A TAZ S89A B MII TAZS89A 0 1 2 3 4 A 10A TAZS89A C hmsr Biology & Chemistry

10A MII TAZS89A MII TAZ S89A * spectral counts * MII 100 200 300 400 10A TAZ S89A * # 0

spectral counts 10A TAZS89A MII TAZS89A * 100 MIV # 10A 20 40 60 80 MIV 10A 10A TAZ S89A 0

MII MII TAZ S89A * ENO1 MIV 10A

10A TAZ S89A relative levels 10A MII MII * 0 1 2 3 1:/1: PS C16:0/C18:1 MII TAZ S89A * * DPP9 10A ## C16:0/C18:1 PC * C18:0/C18:1 PC MIV * MII TAZ S89A MII HK1 * C16:0/C20:4 PC C18:0/C20:4 PC GPI * # C16:0 LPC * MIV PFKP MII 10A C18:0 LPC * MIV MII ALDOA C18:1 LPC * # C20:4 LPC GAPDH C16:0/C18:1 PE spectral counts PGK1 *

MIV # C18:0/C18:1 PE 100 150 200 250 PGAM1 glycolysis C16:0/C20:4 PE 50

21 ENO1 C18:0/C20:4 PE 10A TAZ S89A

0 PKM2 C16:0 LPE phospholipids spectral counts C18:0 LPE DPP9 LDHA 100 200 300

3–4,Jl 7 2014 17, July 831–840, , 10A TAZ S89A relative levels C18:1 LPE 10A LDHB C20:4 LPE 10A TAZ S89 0 2 4 6

GAPDH PC

C16:0/C18:1 PS 0 APEH MII TAZ S89A * 10A C18:0/C18:1 PS PYGB C16:0/C20:4 PS 10A UGP2 1: SM C18:0 MII C18:0/C20:4 PS MII * PGM1 glycogen * C18:0 LPS MII MGLL PYGL * * TAZ C18:1 LPS APEH C20:4 LPS TAZ A PGD

MII * metabolism

* G6PD C16:0/C18:1 PG MII PREP MIV * S89 C18:0/C18:1 PG * S DCXR

C16:0/C20:4 PG * A 89A activity-based proteinprofilingofserinehydrolases PDHA1 C18:0/C20:4 PG * MI # C20:4 LPG FAP PDHB V MI * C16:0/C18:1 PA V CS C18:0/C18:1 PA ACLY C16:0/C20:4 PA PPME1 spectral counts ACO1 C18:0/C20:4 PA

10A TAZ relative levels 100 150 200 250 C16:0 LPA ACO2 0 1 2 3 4 C18:0 LPA 10A T 50 IDH1

PAFAH1B2 0 10A C18:1 LPA spectral counts IDH2 C16:0/C16:0 PI IDH3A S89A C16:0/C18:1 PI AZ S89A1 MII 100 200 300 400 0A IDH3B * C18:0/C18:1 PI NCEH1

TAZ S89A PEP C16:0/C20:4 PI 10A TA MI OGDH * PKM2 C18:0/C20:4 PI 0 MII * lipids I TAZ S89A OGDHL C16:0 LPI PREPL GLUD1 TCA cycle C18:0 LPI 10A MII * * Z S89A SUCLG1 C18:1 LPI

* SUCLG2

C20:4 LPI MII # * * MIV PREP # BAT5 C16:0e/C18:1 PCe lipids ether TAZ S89A SDHA C16:0e/C18:1 PEe M ** SDHB MII * I C16:0e/C18:1 PSe V FH C18:0e/C20:4 PGe ABHD6 10A TAZ S89A relative levels C16:0e/C18:1 PIe * MDH1

0 2 4 6 8 C18:0e/C18:1 PIe MDH2 ª

C16:0e/C20:4 PIe MIV * and neutrallipids PAFAH1B3 ME2 10A C12:0 AC DLD C16:0 AC spectral counts 04Esve t l ihsreserved rights All Ltd Elsevier 2014 C16:0 FFA DLSTP1

MI ACs, FFAs, *

C18:0 FFA PRCP 20 40 60 80 DLAT I AMP TAZ S89A C18:1 FFA 10A T 0 UGGT1 * M C22:6 FFA spectral counts NANS shotgun proteomics II C16:0 MAG 100 150 metabolomics SIAE AZ S * C18:0 MAG 10 GALNT2 glycosylation C18:1 MAG 10 50 A MOGS 0 MII 8 * C16:0/C18:1 DAG A MIV PGK1 PRKCSH # FASN 9A C18:0/C18:1 DAG T TAZ S89A ALDH7A1 C16:0/C20:4 DAG AZ 10A PAFAH1B2 MII * ALDH18A1 C18:0/C20:4 DAG S89A C16:0/C16:0/C16:0 TAG MI ASPH

LYPLA1 * * relative levels # 10A T C18:0/C18:0/C18:0 TAG I TAZ S GOT2

0 2 4 6 8 C16:0/C18:1/C16:0 TAG

* PYCR2

C18:0/C18:1/C18:0 TAG M M * IV GLS A 10A C16:0 NAE II DPP7 amino Z S89A * C18:0 NAE sphingo- 89 SHMT2 MII TAZ S89A C18:1 NAE A IVD * lipids C16:0 ceramide MIV * ABHD10 acid UMP ILVBL C18:0 ceramide PHGDH MII * C20:4 ceramide spectral counts C16:0 SM LIPE MCCC2 metabolism C18:0 SM *

# PFAS 20 40 60 C18:1 SM 10

0 ASS1 C20:4 SM A T * MIV # spectral counts glucose-1-phosphate PNPLA4 HIBADH glucose 6-phosphate AZ S89A10A NME2 fructose-6-phosphate 10 20 40 60 80 NNMT fructose-1,6-bisphosphate A TAZ 0 M *

FAM108A4 LDHA AK2 10A TAZ relative levels 10 1,3-bisphosphoglycerate II TAZ S89A nucleic

UDP-glucose glycolysis GMPS 0 2 4 6 8 3-phosphoglycerate 10A PAFAH1B3 G3P/DHAP S89A MII * GFPT1 10 PEP MII TAZ S89A ACOT1 acid * IMPDH2 * S A pyruvate # CTPS M 89A glycerol-3-phosphate * * DPYSL2 metabolism II TAZ S89A citrate M PLA2G15 MIV * isocitrate II # NAMPT * MII a-ketoglutarate TCA * FASN transaconitate GPD2

* PNPLA6 # succinate MI * OXCT1 malate cycle V oxaloacetate ACOT9 * M # IV ribulose-1,5-bisphosphate LYPLA2 spectral counts ECH1 ribose-5-phosphate ACAA2 UDP-glucuronic 10A TAZ S89A ribulose-5-phosphate ACAT1 PPP 20 40 60 relative levels erythrose-4-phosphate PNPLA8 1 0 2 4 6 8 phosphonogluconic acid 0A TAZ S89A0 ACSL4 10A glucuronic acid spectral counts ACSL3 glyoxylic acid 10A DCI N-acetylneuraminic acid hexoses 1 10 15 FAAH 0 CBR1 MII TAZ S89A acid inositol 0 5 ALDOA * A TAZ S89A M inositol-1,4-bisphosphate II TAZ * CPT1A inositol-4-phosphate inositols 10A GBA MII * ABHD12 lipid pantothenate M * ACADVL thiamine pyrophosphate vitamins * MII TAZ S89A II HSD17B2 # 3-ureidoproprionate * S89A * # SIAE SCPEP1 metabolism adenine ECHDC3

M * * # AMP MI ALDH3A2 I M * V ADP I IV # PTGES2 ATP * CES2 ACACA CDP nucleotide UMP M ACSL1 * ermncacid neuraminic

relative levels 10 10 UDP and IV ABHD11 ACAD9 A TAZ 0 2 4 6 8 UDP-GlcNAc spectral counts AGK UDP-glucose ECHS1 10A UDP-glucuronic acid metabolites ACADM N-acetyl dihydrouracil DAGLB 10A TAZ S89A 20 40 60 80 S uracil CPT2 M 8 0 *

II TAZ 9A adenine spectral counts DECR1 Cancer Breast in PAFAH1B3 cytosine 1000 1500 CPVL 1 AGPS * M thymidine 500 0 S89AII thymine 10A TAZ S89A A HADH NAD MII TAZ * HSD17B10 * FASN 0 # NADH nicotina- PAFAH2 CKMT1B NADP * MI * 10A MTHFD1 nicotinic acid M V II HSDL2 Biology & Chemistry N-methylnicotinamide ACOT2 S mides M 89A * valine * EPHX1 II TAZ FASN threonine MAOA

M * rela10 tive20 level30 s 40 phenylalanine 10 M * FAM108B1 IV SQRDL cofactor A TAZ 0 proline I serine S I FDXR 89A * 10A glycine SOD2 guanine FAM108C1 NQO1 metabolism

S8 M * MII TAZ proline histidine IV SOD1 9 * isoleucine amino A spectral counts NNT and aspartic acid LIPA 20 40 60 80 PRDX3 M * leucine 10A TAZ II lysine acids 0 OAT S arginine detoxi- 8 * PON2 # lgn nnx page) next on (legend 9A asparagine 10A methionine GSTK1

MI * S glutamic acid MII TAZ S89A MGST1 fication V * PYGB glutamine ALDH1B1 glutathione GSTP1 10A TAZ S89A MI * relative level10 s 15 I ALDH1A3 0 5 activity activity * glutathione # higher 89A 10A lower *

MIV expression expression M *

II TAZ higher levels levels higher lower lower

MII * S89A * #

MI * V # Chemistry & Biology PAFAH1B3 in Breast Cancer

(premalignant, also known as MII), and fully malignant MCF10A- analyses to identify commonly altered changes in protein CA1a (MIV) cells, derived in vivo from spontaneous progression expression of metabolic enzymes, activities of serine hydro- of MII cells (Santner et al., 2001; Figure S1 available online). With lases, and metabolite levels, respectively, that may underlie in vivo orthotopic models, MII cells generate low-grade tumors in cellular transformation and TAZ-mediated malignant progres- approximately 25% of xenografts, whereas the MIV lines form sion. Whereas shotgun proteomic profiling provides broad high-grade tumors, resembling grade III human breast tumors, coverage of alterations in protein expression, ABPP uses at a much higher frequency. This well-characterized progression active-site-directed chemical probes to identify dysregulated model displays many important features of breast cancer pro- activities of large numbers of enzymes (Nomura et al., 2010a). gression found in highly aggressive metaplastic and claudin- We chose to profile the serine hydrolase superfamily for this low breast tumor subtypes including EMT, expansion of CSC study because this enzyme class is one of the largest metabolic population, and the associated increase in expression of the enzyme classes in the human genome with a broad range of stem-cell-associated CD44+/CD24À/low antigenic profile, self- functions including esterase, lipase, hydrolase, deacetylase, renewal capabilities, and resistance to conventional therapies thioesterase, protease, and peptidase activities, and many (Chaffer and Weinberg, 2011; Gupta et al., 2009). In particular, serine hydrolases have been shown to be important in cancer Cordenonsi et al. (2011) recently reported that MIV cells show (Long and Cravatt, 2011). Through these profiling efforts, we significantly higher self-renewal ability, tumorigenic potential, identified several enzymes and lipids that were either specif- and an increased CSC population than MII cells, resembling ically upregulated by constitutive activation of TAZ or commonly the difference between grade III and grade I human breast upregulated in 10A TAZ S89A, MII, MII TAZ S89A, and MIV cells tumors. By analyzing a large human patient data set, they iden- (Figures 1A–1C and S1; Table S1). The dysregulated enzymes tified TAZ as a key signature that is overrepresented in poorly identified through shotgun proteomics include glycolytic en- differentiated high-grade tumors and correlates with increased zymes (enolase 1 [ENO1], glyceraldehyde-3-phosphate dehy- CSCs, metastasis, and reduced survival. TAZ, a transducer of drogenase [GAPDH], PKM2, phosphoglycerate kinase [PGK1], the Hippo-signaling pathway that mediates cell-cell contact lactate dehydrogenase A [LDHA], and aldolase A [ALDOA]), and polarity signals to control cell proliferation and organ size the de novo lipogenesis enzyme FASN, and the glycogen- (Chan et al., 2011), is also expressed at higher levels in MIV cells metabolizing enzyme glycogen phosphorylase B (PYGB) than MII cells and is required to sustain self-renewal and tumor- (Figures 1A and S1; Table S1). ABPP of serine hydrolases also initiation capacities in breast CSCs. Consistent with previous re- revealed FASN upregulation, in addition to peptidases (dipep- ports, we show that expression of a constitutively active TAZ, tidyl peptidase 9 [DPP9], acylpeptide hydrolase, and prolyl TAZ S89A, in MCF10A or MII cells results in increased EMT, col- endopeptidase), lipases (PAFAH1B2 and PAFAH1B3), and sialic ony formation in soft agar, and cellular migration (Cordenonsi acid acetylesterase (SIAE) (Figure 1B; Table S1). Metabolomic et al., 2011; Figure S1). analysis yielded several metabolites that were commonly heightened across the four cell lines, including lipids (phospha- Identifying Dysregulated Metabolic Pathways tidyl ethanolamine [PE], phosphatidyl serine [PS], and sphingo- Underlying Cellular Transformation and Malignant myelin [SM]), the glycolytic intermediate phosphoenolpyruvate Progression (PEP), nucleotides (AMP and uridine monophosphate [UMP]), Our goal was to employ multiple metabolic mapping platforms uridine-diphosphate-conjugated sugars (UDP-glucose and to broadly identify dysregulated metabolic pathways that under- UDP-glucuronic acid), the sialic acid N-acetylneuraminic acid, lie cellular transformation and malignant progression using the the amino acid proline, and the antioxidant glutathione (Fig- aforementioned breast cancer model. We performed shotgun ure 1C; Table S1). Certain enzymes such as monoacylglycerol proteomic analysis, activity-based protein profiling (ABPP) lipase (MGLL), the serine protease fibroblast activation protein using the serine hydrolase-directed activity-based probe, and (FAP), and protein methyl esterase 1 (PPME1) were upregulated targeted single reaction monitoring (SRM) liquid chro- specifically by constitutive activation of TAZ (Figure S1; matography/mass spectrometry (LC/MS)-based metabolomic Table S1). Antioncogenic enzymes such as succinate

Figure 1. Identifying Dysregulated Metabolic Enzymes in Breast Cancer Progression Model (A) Shotgun proteomic profiling of metabolic enzyme expression in the breast cancer progression model by MudPIT. Upper heatmap shows relative protein expression of each protein, normalized to the highest expression of each protein across the five cell lines. Dark blue corresponds to high expression, and white or light blue corresponds to lower expression. Lower bar graphs show the metabolic enzymes that were significantly upregulated across 10A TAZ S89A, MII, MII TAZ S89A, and MIV cells. (B) ABPP of serine hydrolase activities in the breast cancer progression model by ABPP-MudPIT using the fluorophosphonate-biotin serine hydrolase activity- based probe. Upper heatmap shows relative serine hydrolase activities of each protein, normalized to the highest expression of each protein across the five cell lines. Dark blue corresponds to high activity, and white or light blue corresponds to lower activity. Lower bar graphs show the serine hydrolase activities that were significantly upregulated across 10A TAZ S89A, MII, MII TAZ S89A, and MIV cells. (C) Metabolomic profiling of the breast cancer progression model using SRM-based targeted liquid chromatography-tandem mass spectrometry. Upper heatmap shows relative metabolite activities normalized to MCF10A control cells. Dark blue corresponds to high levels, and white corresponds to lower levels. Lower bar graphs show the metabolites that were significantly elevated in levels across 10A TAZ S89A, MII, MII TAZ S89A, and MIV cells. Data in bar graphs are presented as mean ± SEM; n = 4–6/group. Significance is presented as *p < 0.05 compared to MCF10A control cells, #p < 0.05 compared to MII cells. Detailed data are shown in Table S1. Characterization of breast cancer progression model, TAZ S89A-specific changes in metabolic enzymes, additional data showing regulation of these metabolic enzyme targets by the Hippo pathway, and additional data showing regulation of glycolytic metabolism by RAS and TAZ S89A are shown in Figure S1.

A expression B proliferation Figure 2. Screening for Nodal Metabolic siControl Enzymes in Breast Cancer 120 siRNA 120 100 100 * * (A) We transiently knocked down the expression of 80 80 * * * * representative metabolic enzymes in the pathways 60 * 60 40 * * * 40 that we identiﬁed as consistently dysregulated in

expression 20 * * * * * * proliferation 20 the breast cancer progression model with siRNA in 0 * * * * ** * 0 MII TAZ S89A cells.

ASN (B–D) Knockdowns were conﬁrmed by qPCR 48 hr SIAE PKM2LDHALDHBNANSUGP2PYGBPYGLF MGLL ENO1PGK1 SGMS ALDOA GAPDH AH1B2AH1B3 siSIAE after siRNA transfection. We screened for en- FAH1B3 siPKM2siLDHAsiLDHBsiNANSsiUGP2siPYGBsiPYGLsiFASNsiMGLL siENO1siPGK1 siSGMS siControl siALDOA siGAPDH AF PAFAH1B2PA P zymes that, when inactivated, impaired various si siPAF aspects of cancer pathogenicity, including cellular CDsurvival migration proliferation (B), serum-free cell survival (C), and 120 120 cell migration (D) in MII TAZ S89A cells. Cells were

n 100 100 * * * * * * * transfected with siRNA for 48 hr before seeding 80 * * * * 80 * * * *** * * 60 * * * * * * * * 60 * * * * into phenotypic experiments. Proliferation and cell

survival 40 40 survival were assessed 48 hr after seeding cells by migratio 20 20 the WST1 cell viability assay. Migration was as- 0 0 sessed by counting the number of cells migrated ASN ASN F through Transwell chambers over 24 hr. Data in F AH1B3 siSIAE AH1B3 siSIAE siPKM2siLDHAsiLDHBsiNANSsiUGP2siPYGBsiPYGLsi siMGLL siENO1siPGK1 siSGMS siPKM2siLDHAsiLDHBsiNANSsiUGP2siPYGBsiPYGLsi siMGLL siENO1siPGK1 siSGMS siControl siALDOA bar graphs are presented as mean ± SEM; n = 3–5/ siControl siALDOA siGAPDH siGAPDH AFAH1B2AF siPAFAH1B2siPAF group normalized to siControl. Significance is siP siP presented as *p < 0.05 compared to siControl MII TAZ S89A control cells. dehydrogenases (SDHAs) as well as several other citric acid cy- scriptional targets, we proceeded to knockdown both YAP cle enzymes were specifically downregulated upon induction and TAZ in MIV cells. Interestingly, YAP and TAZ dual knock- with TAZ S89A (Figure S1; Table S1). down led to more robust reductions in the expression Overall, our data indicate that cellular transformation and of most TAZ-regulated targets, including PAFAH1B2, ENO1, malignant progression heighten glycolysis and glycogen meta- ALDOA, PGK1, LDHA, PKM2, DPP9, MGLL, and NANS (Fig- bolism, de novo lipogenesis, nucleic acid metabolism, sialic ure S1). PAFAH1B3 was not altered, even upon TAZ and YAP acid metabolism, oxidative stress, and specific branches of lipid knockdown (Figure S1). We interpret this lack of change to metabolism. Many of these pathways, including glycolytic and perhaps indicate that other oncogenic pathways, such as glycogen metabolism, nucleic acid metabolism, sialic acid HRAS, may regulate this enzyme or that PAFAH1B3 may be metabolism, oxidative stress, and de novo lipogenesis, have regulated by posttranslational mechanisms. been previously shown to be critical pathways that underlie cancer metabolism and pathogenicity. Nonetheless, we show Screening for Metabolic Enzymes that Are Required for that both HRAS transformation as well as constitutive activation Breast Cancer Pathogenicity of TAZ induces glycolytic, glycogen, and sialic acid metabolism We next sought to identify metabolic enzymes that, when in- and that TAZ and HRAS act synergistically in many cases (e.g., activated, would impair cancer pathogenicity. We used small ENO1, GAPDH, PKM2, LDHA, ALDOA, PGK1, PYGB, UDP- interfering RNA (siRNA) to transiently knockdown the expression glucose, and N-acetylneuraminic acid) to enhance these of the dysregulated metabolic enzymes that we found through metabolic pathways (Figure S1). Additionally, we confirmed our profiling studies and identified those enzymes that impaired heightened glycolytic metabolism in these lines by measuring cellular proliferation, survival, or motility upon siRNA inactivation glucose consumption and lactic acid secretion and found that in MII TAZ S89A cells (Figures 2A–2D). To filter a large list both are heightened upon transduction with TAZ S89A or of dysregulated metabolic enzyme targets, we narrowed our HRAS and further heightened upon transduction with both TAZ focus on those enzymes found by shotgun proteomics and S89A and HRAS. The highly malignant MIV cells also show ABPP and enzymes that control metabolites found by meta- significantly heightened glycolytic metabolism beyond that of bolomics that were altered across 10A TAZ S89A, MII, MII MII cells (Figure S1). Interestingly, the enzymes specifically upre- TAZS89A, and MIV cells. Whereas inactivation of glycolytic-, gulated by TAZ, such as MGLL, FAP, and PPME1, or downregu- glycogen-, and sialic-acid-metabolizing enzymes modestly lated by TAZ, such as SDHA, have all been implicated as critical impaired proliferative, survival, or migratory capacity, we were drivers of cancer progression (Bachovchin et al., 2011; Kelly, surprised to find that knockdown of PAFAH1B2 and PAFAH1B3 2005; Nomura et al., 2011; Nomura et al., 2010b; Puustinen caused the most significant and dramatic impairments in the et al., 2009; Yang et al., 2013). proliferative, survival, and migratory phenotypes in MII TAZ Our data indicate that many of the aforementioned metabolic S89A cells (Figures 2A–2D). Thus, we decided to further investi- enzymes are upregulated upon induction with TAZ S89A. We gate the role of PAFAH1B2 and PAFAH1B3 in breast cancer wanted to determine whether these enzymes are regulated by pathogenicity. TAZ in aggressive MIV cells. Upon knocking down TAZ in MIV cells, we observed only modest reductions in the expres- Characterizing the Pathological Role of PAFAH1B2 and sion of the presumably TAZ-regulated targets, including PA- PAFAH1B3 in Breast Cancer FAH1B2, ENO1, ALDOA, DPP9, and NANS (Figure S1). Intrigued by our initial findings in our screen in the MII TAZ S89A Because both YAP and TAZ may regulate overlapping tran- cells, we next wanted to ascertain whether PAFAH1B2 and

PAFAH1B3 inactivation also impaired breast cancer pathoge- Metabolomic Characterization of PAFAH1B3-Regulated nicity in a more realistic patient-derived breast cancer cell line. Pathways in Breast Cancer We thus stably inactivated the expression of PAFAH1B2 and The putative function of PAFAH1B3 is to act as a deacetylase for PAFAH1B3 in the highly aggressive estrogen receptor/proges- the signaling lipid platelet-activating factor (PAF) (Hattori et al., terone receptor/HER2 receptor-negative (i.e., triple negative) 1995; Manya et al., 1999). Although recombinant PAFAH1B3 human breast cancer cells, 231MFP, a line derived from a breast can cleave PAF in vitro (Figure S3), we found that extracellular cancer patient adenocarcinoma MDA-MB-231 that has been and intracellular PAF levels are unchanged in shPAFAH1B2 in vivo passaged in mice to derive a malignant and more aggres- and shPAFAH1B3 231MFP cells and PAFAH hydrolytic activity sive variant (Jessani et al., 2004). Using two independent short was not reduced upon PAFAH1B2 and PAFAH1B3 knockdown, hairpin RNA oligonucleotides (shPAFAH1B2-1, shPAFAH1B2- indicating that PAFAH1B is not the dominant enzyme controlling 2, shPAFAH1B3-1, and shPAFAH1B3-2), PAFAH1B2 and PAF levels or PAF hydrolytic activity in 231MFP cells. Thus, PAFAH1B3 were knocked down by >80% as measured by PAFAH1B3 is likely not controlling cancer pathogenicity through both quantitative PCR (qPCR) and ABPP-MudPIT (Figure 3A; controlling PAF-signaling pathways (Figure S3). Figure S2). Knockdown of PAFAH1B2 or PAFAH1B3 significantly We next performed targeted and untargeted LC/MS-based impaired 231MFP cell proliferation, survival, migration, and inva- metabolomic profiling to identify metabolic changes in the lip- siveness in culture and slowed or eliminated in vivo tumor growth idome conferred by PAFAH1B3 knockdown in 231MFP breast in immune-deficient mice (Figures 3B–3F; Figure S2). Thus, our cancer cells (Figures 4A and 4B; Table S2). We subsequently results suggest that PAFAH1B enzymes are important in main- interpreted only those changes in lipid levels that were signifi- taining aggressive features of cancer. cant between the two sh-oligonucleotides and in more than Upon profiling the gene expression of PAFAH1B2 and one experiment to reduce false positives or artifacts. We found PAFAH1B3 in primary breast tumors, we found that PAFAH1B3, that PAFAH1B3 knockdown raises the levels of several lipid but not PAFAH1B2, was significantly heightened in human species including phosphatidylcholines (PCs), lysophosphatidyl breast tumors compared to normal mammary tissue (Figure 3G). choline (LPC), PEs, lysophosphatidyl ethanolamines (LPEs), Based on the online Kaplan-Meier plotter resource for breast phosphatidic acids (PAs), PSs, phosphatidyl glycerol (PGs), cancer recurrence-free survival, PAFAH1B3 expression is corre- SMs, ceramide, sphingosine, triacylglycerols, and monoalkyl- lated with low recurrence-free survival in breast cancer patients glycerols, as well as ether lipid counterparts (designated as overall, in lymph-node-positive tumors, and in grade 1–3 breast ‘‘e’’ after lipid nomenclature; Figure 4C; Table S2). We cancer patients (Figure 3H; Gyo¨ rffy et al., 2010). screened the lipids that contain potential hydrolyzable bonds From our phenotypic studies, we found that PAFAH1B3 against purified PAFAH1B3, but none of these lipids were inactivation produced more striking antitumorigenic effects direct substrates of PAFAH1B3 and were not turned over compared with that of PAFAH1B2. Furthermore, PAFAH1B3 (Figure S3). Thus, we interpret these changes to be indirect expression, but not PAFAH1B2 expression, was significantly alterations in lipid metabolism stemming from PAFAH1B3 higher in primary human breast tumors compared to normal inactivation. tissue. Thus, we focused our attention on further investigating Interestingly, several of the lipids that were elevated upon the pathophysiological and metabolic roles of PAFAH1B3 in PAFAH1B3 knockdown have been shown to possess antitu- driving breast cancer pathogenicity. Whereas we do not under- morigenic or proapoptotic effects. For example, ceramide stand the mechanistic basis for the differences in tumor growth levels are elevated upon PAFAH1B3 knockdown and ceramide deficits conferred by PAFAH1B2 or PAFAH1B3 knockdown, it is a well-established proapoptotic lipid (Morad and Cabot, will be of future interest to analyze the tumors that grow out in 2013). The metabolites involved in ceramide biosynthesis, shPAFAH1B2 xenograft tumors. including phosphatidylcholine (PC) and sphingomyelin, also We next overexpressed both the catalytically active and show heightened levels upon PAFAH1B3 knockdown. PS has the inactive serine to alanine mutant form of PAFAH1B3 been shown to be elevated in cancer cells undergoing (Ser47Ala or S47A) in MCF10A cells to determine whether this apoptosis after chemotherapy or radiation treatment and is enzyme was sufficient to confer oncogenic properties (Fig- involved in immune recognition and tumor clearance (Chao ure S2). Whereas overexpression of the wild-type form of et al., 2012). C16:0/C18:1 PC has been shown to act as a PAFAH1B3 did not alter proliferation or survival, wild-type PPARa agonist, and PPARa agonists have been shown to PAFAH1B3, but not the catalytically inactive mutant, significantly have antitumorigenic and antiangiogenic properties (Chakra- enhanced cell migration, indicating that PAFAH1B3 exerts at varthy et al., 2009; Huang et al., 2013; Liang et al., 2013; Pozzi least some of its pathogenic effects through its catalytic activity and Capdevila, 2008). Consistent with this premise, we found (Figure S2). Our studies also indicate that whereas PAFAH1B3 that the PPARa agonist fenofibrate impaired cancer cell sur- overexpression may be not sufficient to confer heightened pro- vival and proliferation in 231MFP cells and the PPARa antago- liferation or survival in the nontransformed MCF10A cells, its nist GW6471 rescued the proliferative and survival defects of activity is sufficient to heighten mammary epithelial cell migra- PAFAH1B3 knockdown (Figure S3). Interestingly, we also found tion. Because PAFAH1B3 has been shown to interact with that the elevated C16:0 LPC and C16:0/C18:1 PS also stimu- PAFAH1B2, PAFAH1B1, as well as ubiquitin ligase LNX1, it will lated PPARa activity (Figure S3). be of future interest to coexpress these other binding partners Whereas our metabolomic data still did not reveal the direct to determine whether this coupled expression is sufficient to substrate of PAFAH1B3, our results indicate that PAFAH1B3 drive other malignant features of cancer (Rual et al., 2005; Swee- inactivation leads to a metabolomic signature that is enriched ney et al., 2000). in tumor-suppressing signaling lipids that may be in part

effect of PAFAH1B2 or PAFAH1B3 knockdown in 231MFP breast cancer cells

ABexpression proliferation C survival PAFAH1B2 PAFAH1B3

100 100 100 100 n ion erat 50 50 f 50 * 50 * ** survival expressio

expression * proli * * * * * 0 0 ** 0 0

shControl AH1B2-2 shControl shControl AH1B2-2AH1B3-1 shControl AH1B2-2AH1B3-1 FAH1B3-1 FAH1B2-1 FAH1B3-2 A AFAH1B3-2 AF PAF PA shPAFAH1B2-1sh shP shP sh shP shPAF shPAFAH1B3-2 shPAFAH1B2-1shPAF shPAF shPA

DEmigration invasion F tumor growth shControl shControl 100 100 ) 1000 3

m shControl 800 * * * shPAFAH1B2 50 shPAFAH1B2 50 shPAFAH1B2 600 shPAFAH1B3 migration * invasion * ** * 400 * 0 0 200 shPAFAH1B3 *

shPAFAH1B3 tumor volume (m 0 0 1020304050 shControl shControl AH1B2-2 FAH1B3-1 F days AFAH1B2-1A P shPAFAH1B2-1shPAFAH1B2-2shPA shPAFAH1B3-2 sh shP shPAFAH1B3-1shPAFAH1B3-2

primary human breast tumors

GHexpression recurrence-free survival and PAFAH1B3 expression breast cancer patients (n=3455) lymph node-negative (n=1973) lymph node-positive (n=1178) 1.0 1.0 1.0 p-value=1.9e-5 p-value=0.096 p-value=0.0026 PAFAH1B2 PAFAH1B3 0.8 0.8 0.8 15 15 0.6 0.6 0.6 0.4 probability 0.4 0.4 probability * probability 0.2 low 0.2 low 0.2 low ion 10 10 high high high ssion 0 0 0 0 50 100 150 200 0 50 100 150 200 0 50 100 150 200 time (months) time (months) time (months)

xpress 5 5 expre e grade 1 patients (n=306) grade 2 patients (n=716) grade 3 patients (n=792) 1.0 p-value=0.00038 1.0 1.0 0 0 p-value=0.043 p-value=0.0072 0.8 0.8 0.8 0.6 0.6 0.6 normal normal 0.4 0.4

probability 0.4 probability probability 0.2 low 0.2 low 0.2 low breast cancer breast cancer high high high 0 0 0 0 50 100 150 200 0 50 100 150 200 0 50 100 150 200 time (months) time (months) time (months)

Figure 3. PAFAH1B2 and PAFAH1B3 Are Critical Metabolic Enzymes in Maintaining Breast Cancer Pathogenicity (A) We stably knocked down the expression of PAFAH1B2 and PAFAH1B3 in the triple-negative aggressive 231MFP breast cancer cells with two independent shRNA oligonucleotides by >80% as confirmed by qPCR. (B–E) PAFAH1B2 and PAFAH1B3 knockdown impairs proliferation (B), serum-free cell survival (C), migration (D), and invasion (E). (F) PAFAH1B2 and PAFAH1B3 knockdown also impairs tumor xenograft growth in immune-deficient SCID mice. (G) Expression of PAFAH1B2 and PAFAH1B3 in normal mammary tissue and primary human breast tumors assessed by qPCR. (H) Recurrence-free survival stratified by high versus low PAFAH1B3 expression in breast cancer patients overall and by lymph-node-negative/positive patients and grade 1–3 breast cancer patients. These data are derived from the online resource known as the Kaplan-Meier Plotter (http://www.kmplot.com; Gyo¨ rffy et al., 2010). Data are presented as mean ± SEM; n = 4–8/group for (A–F), 7–41 patients/group for (G), and 306–3,455 patients for (H) as indicated. Significance is presented as *p < 0.05 compared to normal tissue in (A) and compared to shControl 231MFP cells for (B)–(G). p values for (H) are indicated on the figure itself. Protein expression data of PAFAH1B2 and PAFAH1B3 knockdown cells, a recapitulation of tumor growth deficits in shPAFAH1B3 cells, and PAFAH1B3 overexpression studies are shown in Figure S2.

metabolomic changes upon PAFAH1B3 knockdown in 231MFP breast cancer cells targeted/untargeted ABmetabolomics lipidomic profiles 100 elevated metabolites reduced metabolites lipids phospho- ACs neutral lipids NAEs lipids FFAs eicosanoids ether and lipids sphingo- 10 A P fold-change C16:0/C18:1 PC C16:0/C20:4 PC C18:0/C20:4 PC C16:0 LPC C18:0 LPC C20:0 LPC C20:1 LPC C20:4 LPC C16:0/C20:4 PE C18:0/C18:1 PE C18:0/C20:4 PE C18:0 LPE C18:1 LPE C16:0/C18:1 PS C16:0/C20:4 PS C18:0/C18:1 PS C18:0 LPS C18:1 LPS C16:0/C18:1 PG C18:0/C18:1 PG C18:0/C20:4 PG C16:0/C20:4 PI C18:0/C18:1 PI C18:0/C20:4 PI C16:0 LPI C18:0 LPI C18:1 LPI C16:0/C18:1 C18:0p MAGp C16:0e/C2:0 MAGe C16:0/C20:4 PA C18:0/C18:1 PA C16:0 FFA C18:0 FFA C18:1 FFA C20:4 FFA C16:0 MAG C18:0 MAG C18:1 MAG C22:6 MAG C16:0/C18:1 DAG C16:0/C20:4 DAG C18:0/C20:4 DAG C18:0/C18:0/C18:0 TAG C12:0 AC C16:0 AC C16:0 NAE C18:0 NAE C18:1 NAE C20:4 NAE PGE2/PGD2 C18:0 SM C20:4 SM C16:0 ceramide C18:1 ceramide C20:4 ceramide sphinganine sphingosine C16:0e/C18:1 PCe C16:0e/C20:4 PCe C18:0e/C18:1 PCe C18:0e/C20:4 PCe C18:0p/C20:4 PCp C16:0e LPCe C18:0e LPCe C16:0p/C20:4 PEp C18:0e/C20:4 PEe C18:0p/C20:4 PEp C16:0e LPEe C18:0e LPEe C18:0p LPEp C16:0e/C18:1 PSe C18:0e/C18:1 PSe C18:0e/C20:4 PSe C18:0e LPSe C18:0p LPSp C16:0e/C20:4 PGe C18:0e LPGe C16:0e/C18:1 PIe C18:0e/C18:1 PIe C18:0e/C20:4 PIe C16:0e/C18:1 PAe C18:0e/C18:1 PAe C16:0e MAGe C18:0e MAGe C18:0/C18:1 PC C18:1 LPC C16:0/C18:1 PE C16:0 LPE C20:4 LPE C18:0/C20:4 PS C16:0 LPS C20:4 LPS C16:0/C20:4 PG C16:0/C16:0 PI C16:0/C18:1 PI C20:4 LPI C18:0e/C2:0 MAGe C18:0p/C2:0 MAGp C18:0/C20:4 PA C22:6 FFA C20:4 MAG C18:0/C18:1 DAG C18:0/C18:1/C18:0 TAG C16:0 SM C18:1 SM C2:0 ceramide C18:0 ceramide C18:0p LPCp C16:0e/C18:1 PEe C16:0e/C20:4 PEe C18:0e/C18:1 PEe C16:0e/C18:1 PGe C18:0e/C18:1 PGe C18:0e/C20:4 PAe shControl 1 0 -1 -2 -3 -4 -5 -6 -7 shPAFAH1B3-1 10 10 10 10 10 10 10 10 shPAFAH1B3-2 p-value lower levels higher levels

C significantly altered metabolites upon PAFAH1B3 knockdown

20 shControl 15 shPAFAH1B3-1* shPAFAH1B3-2* 10

5 relative levels 0 C e e e E E p A e M M M e C P P S S AG PC PCe PCp L 0 0SM 1S 4 1P L 0L 1 : : : : : 0p L :0e L: /C20:4 PS C16 C18 C18: C20: C16:0C18:0 LPCC18:1 LPCC20:0 LPCC20:1 LPC LPC C18 C18 /C18 :0 sphingosine C16:0eC18:0p LPEe LPE 6 C16:0eC18 C18 0e/C20:4 PA C16:0eC18:0e MAG MAGe :0e C16:0/C18:1C16:0/C20:4C18:0/C18:1 PC P PC C16:0/C20:4 PE C16:0/C20:4C1 P 8:0e/C18:1 PGe C18:1 ceramide

C18 C16:0e/C20:4C18:0e/C20:4 PEe PE C18: C1

C18:0/C18:1/C18:0 T

D tumor breast cancer PAFAH1B3 suppressing pathogenicity lipids

Figure 4. Metabolomic Profiling of PAFAH1B3 Knockdown Cells Reveals Elevations in Antitumorigenic Lipids (A) Volcano plot showing all significantly altered metabolites from untargeted and targeted lipidomic profiling in shPAFAH1B3 knockdown 231MFP cancer cells. The red and blue points correspond to elevated and reduced metabolites, respectively. The points to the left of the dotted line are ions detected, but not changing, upon PAFAH1B3 knockdown, whereas points to the right of the line are metabolites that are significantly changing (p < 0.05 compared to shControl cells). (B) Heatmap showing representative metabolites identified through targeted SRM-based lipidomic analysis. Dark blue corresponds to highest levels, and white corresponds to lowest levels. Metabolites highlighted in red are significantly changed (p < 0.05) in both shPAFAH1B3 lines and repeated in at least two independent experiments. (C) Bar graph showing relative levels of metabolites that are significantly altered in PAFAH1B3 knockdown cells normalized to shControl cells. (D) Our results show that PAFAH1B3 inactivation leads to a metabolomic signature enriched in heightened levels of tumor-suppressing lipids such as ceramides and PPARa ligands (e.g., PC, PS, and LPC). We postulate that this lipidomic signature contributes to reduced cancer pathogenicity upon PAFAH1B3 knockdown in breast cancer cells. Data are presented as mean ± SEM; n = 5–6/group. Significance is presented as *p < 0.05 compared to shControl 231MFP breast cancer cells. Detailed metabolomic data are shown in Table S2. PAF levels and PAFAH hydrolytic activity, lipid hydrolytic activities against PAFAH1B3, and studies testing the role of PPARa in PAFAH1B3 effects upon cancer are shown in Figure S3. responsible for the observed impairments in breast cancer path- incorporates cellular transformation through HRAS and EMT, ogenicity and tumorigenicity. CSC-like features, and malignant progression through the constitutive activation of TAZ. In addition to identifying several Conclusions canonical metabolic pathways, such as glycolysis, glycogen, Here, we have used multiple platforms, including protein expres- and lipogenic metabolism, which have already been shown to sion profiling, ABPP, and metabolomics, to broadly profile dysre- be metabolic drivers of cancer, we have discovered PAFAH1B3 gulated metabolism in a breast cancer progression model that as an important metabolic enzyme that is important in driving

aggressive and tumorigenic features of breast cancer. Metabo- and MGLL have been previously shown to be critical in cancer lomic profiling reveals that PAFAH1B3 controls cancer cell path- pathogenicity through controlling lipogenic and lipolytic pathogenicity through regulating an optimal landscape of signaling ways in cancer cells, respectively (Menendez and Lupu, 2007; lipids that may impair cancer aggressiveness. Nomura et al., 2010b, 2011). FAP, a serine protease selectively Whereas PAFAH1B3 has been implicated as a PAF acetylhy- expressed on tumor-associated fibroblasts and pericytes, has drolase, our data indicate that PAF is likely not the endogenous been shown to drive tumor stromagenesis and tumor growth substrate of PAFAH1B3 in cancer cells. While we still do not (Santos et al., 2009). PPME1, a demethylase and inhibitor of know the identification of this endogenous substrate, the protein phosphatase 2A, helps to promote protumorigenic answer may lie in the altered metabolites that still remain unan- phosphorylation signaling cascades (Bachovchin et al., 2011; notated from our metabolomic analyses of PAFAH1B3 inactiva- Puustinen et al., 2009). Interestingly, NCEH1 is upregulated tion. Identifying these unknown metabolites may further yield only upon induction of HRAS and TAZ S89A, indicating that mechanisms through which PAFAH1B3 exerts its pathogenic both oncogenic stimuli are required for elevating KIAA1363 effects in breast cancer cells. Alternatively, the endogenous activity. This enzyme has been shown to drive cancer cell substrate may also not be a lipid, or even a metabolite, and migration and tumorigenesis through controlling ether lipid- may operate on peptides, proteins, or protein posttranslational signaling pathways (Chang et al., 2011; Chiang et al., 2006). modifications. Thus, it would be of future interest to further Thus, we show that the Hippo transducer is a critical factor in investigate the direct substrates of PAFAH1B3 and how these controlling the metabolic programming that drives the patho- substrates lead to the lipidomic changes that we observe in genic properties of cancer. this study. Our studies underscore the utility of multidimensional meta- While we show that PAFAH1B2 and 1B3 knockdown lead to bolic profiling efforts in uncovering unique metabolic nodes in impairments in cellular proliferation, survival, migration, and cancer, and we put forth PAFAH1B3 as a potential therapeutic invasiveness, the motility and invasiveness deficits may in target for the treatment of aggressive breast cancers. part be due to the proliferative and survival deficits. Thus, it may be interesting to pursue those metabolic enzyme targets, SIGNIFICANCE such as SIAE or NANS, that do not show proliferative deficits but confer migratory impairments, because these targets may Whereas many recent studies have revealed key metabolic regulate invasiveness of the cancer cells. It will also be of future pathways that underlie cellular transformation, the meta- interest to investigate the roles of the other pathways and bolic drivers of cancer malignancy are less well understood. enzymes that are under the control of HRAS, TAZ, or HRAS Here, we have performed a massively parallel metabolic and TAZ. profiling study to identify metabolic drivers underlying a Overall, our results show many metabolic enzymes and breast cancer progression model that incorporates both biochemical pathways that are dysregulated during cellular cellular transformation by RAS and malignant progression transformation and also under malignant progression. Although by the Hippo transducer TAZ. We have discovered metabolic we focused on PAFAH1B3 in this study due to the striking phe- pathways including glycolytic, lipogenic, and lipolytic pro- notypes observed, many of the other targets may represent cesses that have been previously linked to cancer patho- unique metabolic nodes that influence various aspects of can- genicity that are coordinately regulated through both the cer pathogenicity. In this study, we have also identified multiple RAS and Hippo pathways. Most notably, we have uncovered metabolic genes, activities, metabolites, and pathways that are an important metabolic enzyme PAFAH1B3 as a critical regulated by the Hippo-signaling pathway. Previous studies driver of breast cancer pathogenicity through controlling have shown that the Hippo transducer pathway controls a tumor-suppressing signaling lipids. Overall, we show the diverse transcriptional network that drives EMT, cancer stem- utility of multidimensional metabolic profiling strategies in like properties, proliferation, survival, motility, invasiveness, assembling a metabolic map of breast cancer progression metastasis, and transformation (Cordenonsi et al., 2011; Lamar and put forth PAFAH1B3 as a potential enzyme target for et al., 2012). Here, we show that the Hippo transducers YAP breast cancer therapy. and TAZ collectively regulate the expression and activities of critical metabolic drivers of cancer pathogenicity. We show EXPERIMENTAL PROCEDURES that the Hippo transducer pathway strongly regulates glycolytic enzymes and glycolytic metabolism alone or in conjunction with Cell Lines HRAS, thus likely contributing to the glycolytic addiction or the The MCF10A, MII, MII TAZ S89A, and MIV lines were provided by Professor Stefano Piccolo at the University of Padua (Cordenonsi et al., 2011). The ‘‘Warburg effect’’ that subserves cancer pathogenicity. We also 231MFP cells were generated from explanted xenograft tumors of MDA- show here that the Hippo transducer pathway regulates many MB-231 cells, as described previously (Jessani et al., 2004). other metabolic enzymes that have been linked to controlling pathogenic features of cancer, including SDHA, FASN, MGLL, Cell Culture Conditions FAP, PPME1, and neutral cholesteryl ester hydrolase 1 Human embryonic kidney 293T cells were cultured in Dulbecco’s modified (NCEH1, also known as KIAA1363). SDHA loss-of-function Eagles’ medium (DMEM) media containing 10% fetal bovine serum (FBS) and maintained at 37C with 5% CO . 231MFP cells were cultured in L15 mutations, through accumulation of succinate and inhibition 2 media containing 10% FBS and glutamine and were maintained at 37Cin of alpha-ketoglutarate-dependent oxygenases, have been 0% CO2. MCF10A and MCF10A-derived lines were cultured in DMEM/F12K shown to enhance glycolysis, angiogenesis, and epigenetic media containing 5% (vol/vol) horse serum, 1% (vol/vol) penicillin-strepto- alterations that fuel cancer (Yang et al., 2013). Both FASN mycin-glutamine, 20 ng/ml epidermal growth factor, 100 ng/ml cholera toxin,

10 ng/ml insulin, and 500 ng/ml hydrocortisone and maintained at 37Cat5% Tumor Xenograft Studies

(wt/vol) CO2 as described previously (Cordenonsi et al., 2011). All experimental protocols used were approved by the Animal Care and Use Committee of the University of California, Berkeley. Human tumor xenografts were established by transplanting cancer cells ectopically into the flank of qPCR C.B17 severe combined immunodeficiency (SCID) mice (Taconic Farms) as qPCR was performed using the manufacturer’s protocol for Fischer Maxima described previously (Nomura et al., 2010b). Briefly, cells were washed two SYBRgreen, with 10 mM primer concentrations. Primer sequences were times with PBS, trypsinized, and harvested in serum-containing medium. derived from Primer Bank (Spandidos et al., 2010). Next, the harvested cells were washed two times with serum-free medium and resuspended at a concentration of 2.0 3 104 cells/ml and 100 ml was in- Cancer Phenotype Studies jected. Growth of the tumors was measured every 3–6 days with calipers. Migration, invasion, cell proliferation, and survival studies were performed as described previously (Nomura et al., 2010b). SUPPLEMENTAL INFORMATION

Shotgun Proteomic Profiling of Cancer Cells Supplemental Information includes Supplemental Experimental Procedures, m À Proteins (100 g) were TCA precipitated in 20% TCA at 80 C overnight and three figures, and two tables and can be found with this article online at 3 then centrifuged at 10,000 g at 4 C for 10 min to pellet protein. The protein http://dx.doi.org/10.1016/j.chembiol.2014.05.008. pellets were washed three times with 8 M urea in PBS. After solubilization, 30 ml of 0.2% Protease Max Surfactant (Promega) was added and the resulting AUTHOR CONTRIBUTIONS mixture was vortexed followed by the addition of 40 ml of 100 mM ammonium bicarbonate and 10 mM tris(2-carboxyethyl)phosphine. After 30 min, 12.5 mM M.M.M. designed research, performed research, analyzed data, and wrote the iodoacetamide was added and allowed to react for 30 min in the dark before paper. D.I.B. performed research and analyzed data. A.S., M.G., T.N., and adding 120 ml PBS and 1.2 ml 1% Protease Max Surfactant (Promega). The pro- P.J.M. performed research. P.J.M. contributed reagents. E.L.S. and X.J. tein solution was vortexed and 0.5 mg/ml Sequencing Grade Trypsin (Promega) performed research, contributed new reagents, and analyzed data. K.L. was added and allowed to react overnight at 37 C. The peptide solution was designed research, analyzed data, contributed new reagents, and wrote the subjected to a 30 min 10,000 3 g spin before the supernatant was transferred paper. D.K.N. designed research, performed research, analyzed data, and to a new tube and loaded onto a biphasic (strong cation exchange/reverse wrote the paper. phase) capillary column and analyzed by 2D liquid chromatography in combi- nation with tandem mass spectrometry on a nanospray Thermo LTQ-XL MS/ MS. The data were analyzed using the Integrated Proteomics Pipeline (IP2) ACKNOWLEDGMENTS as described previously (Nomura et al., 2010b). We thank the members of the D.K.N. research group and the K.L. laboratory for critical reading of the manuscript. This work was supported by grants from the Activity-Based Protein Profiling of Cancer Cells NIH (R21CA170317, R01CA172667, and R00DA030908 to M.M.M., D.I.B., ABPP-MudPIT analysis was performed using previously established proce- S.M.L., and D.K.N. and R01CA101891 to K.L.) and the Searle Scholar dures (Nomura et al., 2010b). Briefly, 1 mg protein was labeled with 5 mM fluo- Foundation (to D.K.N.). rophosphonate-biotin in 1 ml PBS for 1 hr at room temperature, solubilized in 1% Triton X-100 for 1 hr, denatured, and labeled enzymes were enriched using Received: February 11, 2014 avidin beads, reduced and alkylated, and trypsinized. Tryptic peptides were Revised: May 12, 2014 analyzed on a Thermo LTQ-XL MS/MS as described previously (Nomura Accepted: May 16, 2014 et al., 2010b). Published: June 19, 2014

Metabolomic Profiling of Cancer Cells REFERENCES Metabolite measurements were conducted using modified versions of previous procedures (Kopp et al., 2010; Nomura et al., 2010b). Details are Bachovchin, D.A., Mohr, J.T., Speers, A.E., Wang, C., Berlin, J.M., Spicer, described in Supplemental Experimental Procedures. T.P., Fernandez-Vega, V., Chase, P., Hodder, P.S., Schu¨ rer, S.C., et al. (2011). Academic cross-fertilization by public screening yields a remarkable RNA Interference of Enzymes class of protein phosphatase methylesterase-1 inhibitors. Proc. Natl. Acad. For siRNA knockdown, cells were treated with ON-TARGETplus SMARTpool Sci. USA 108, 6811–6816. siRNA (Thermo) per the manufacturer’s protocol. Two hundred thousand cells Benjamin, D.I., Cravatt, B.F., and Nomura, D.K. (2012). Global profiling strate- were treated with 50 nM siRNA using DharmaFECT 1 Transfection Reagent. gies for mapping dysregulated metabolic pathways in cancer. Cell Metab. 16, Assays were performed after 48 hr of treatment. We confirmed knockdown 565–577. by qPCR. Chaffer, C.L., and Weinberg, R.A. (2011). A perspective on cancer cell metas- We used short-hairpin RNA (shRNA) using two independent silencing oligo- tasis. Science 331, 1559–1564. nucleotides to knockdown the expression of PAFAH1B2 and PAFAH1B3 using previously described procedures (Nomura et al., 2010b). Details are described Chakravarthy, M.V., Lodhi, I.J., Yin, L., Malapaka, R.R., Xu, H.E., Turk, J., and in Supplemental Experimental Procedures. Semenkovich, C.F. (2009). Identification of a physiologically relevant endogenous ligand for PPARalpha in liver. Cell 138, 476–488.

Overexpression of PAFAH1B2 or PAFAH1B3 in MCF10A Cells Chambers, A.F., Groom, A.C., and MacDonald, I.C. (2002). Dissemination and growth of cancer cells in metastatic sites. Nat. Rev. Cancer 2, 563–572. Stable PAFAH1B3 overexpression was achieved by subcloning the PAFAH1B3 gene from fully sequenced human cDNA (Thermo Scientiﬁc) into Chan, S.W., Lim, C.J., Chen, L., Chong, Y.F., Huang, C., Song, H., and Hong, the pLenti CMV puro DEST vector (Addgene 17452) using the gateway cloning W. (2011). The Hippo pathway in biological control and cancer development. system (Life Technologies). The PAFAH1B3 Ser 47 Ala mutant was made using J. Cell. Physiol. 226, 928–939. Quick-Change Mutagenesis to mutate base 139 (Stratagene). Chang, J.W., Nomura, D.K., and Cravatt, B.F. (2011). A potent and selective inhibitor of KIAA1363/AADACL1 that impairs prostate cancer pathogenesis. Breast Tumor Array qPCR Chem. Biol. 18, 476–484. Breast cancer tumor array 1 was purchased from OriGene, and qPCR was per- Chao, M.P., Majeti, R., and Weissman, I.L. (2012). Programmed cell removal: a formed on the array using the protocol described above. new obstacle in the road to developing cancer. Nat. Rev. Cancer 12, 58–67.

Chiang, K.P., Niessen, S., Saghatelian, A., and Cravatt, B.F. (2006). An enzyme Morad, S.A., and Cabot, M.C. (2013). Ceramide-orchestrated signalling in that regulates ether lipid signaling pathways in cancer annotated by multidi- cancer cells. Nat. Rev. Cancer 13, 51–65. mensional profiling. Chem. Biol. 13, 1041–1050. Nomura, D.K., Dix, M.M., and Cravatt, B.F. (2010a). Activity-based protein Cordenonsi, M., Zanconato, F., Azzolin, L., Forcato, M., Rosato, A., Frasson, profiling for biochemical pathway discovery in cancer. Nat. Rev. Cancer 10, C., Inui, M., Montagner, M., Parenti, A.R., Poletti, A., et al. (2011). The Hippo 630–638. transducer TAZ confers cancer stem cell-related traits on breast cancer cells. Cell 147, 759–772. Nomura, D.K., Long, J.Z., Niessen, S., Hoover, H.S., Ng, S.W., and Cravatt, B.F. (2010b). Monoacylglycerol lipase regulates a fatty acid network that Gupta, P.B., Onder, T.T., Jiang, G., Tao, K., Kuperwasser, C., Weinberg, R.A., promotes cancer pathogenesis. Cell 140, 49–61. and Lander, E.S. (2009). Identification of selective inhibitors of cancer stem cells by high-throughput screening. Cell 138, 645–659. Nomura, D.K., Lombardi, D.P., Chang, J.W., Niessen, S., Ward, A.M., Long, Gyo¨ rffy, B., Lanczky, A., Eklund, A.C., Denkert, C., Budczies, J., Li, Q., and J.Z., Hoover, H.H., and Cravatt, B.F. (2011). Monoacylglycerol lipase exerts Szallasi, Z. (2010). An online survival analysis tool to rapidly assess the effect dual control over endocannabinoid and fatty acid pathways to support pros- of 22,277 genes on breast cancer prognosis using microarray data of 1,809 tate cancer. Chem. Biol. 18, 846–856. patients. Breast Cancer Res. Treat. 123, 725–731. National Cancer Institute. (2013). Drugs approved for breast cancer. http:// Hattori, K., Hattori, M., Adachi, H., Tsujimoto, M., Arai, H., and Inoue, K. (1995). www.cancer.gov/cancertopics/druginfo/breastcancer. Purification and characterization of platelet-activating factor acetylhydrolase II Polyak, K., and Weinberg, R.A. (2009). Transitions between epithelial and from bovine liver cytosol. J. Biol. Chem. 270, 22308–22313. mesenchymal states: acquisition of malignant and stem cell traits. Nat. Rev. Hong, W., and Guan, K.L. (2012). The YAP and TAZ transcription co-activators: Cancer 9, 265–273. key downstream effectors of the mammalian Hippo pathway. Semin. Cell Dev. Pozzi, A., and Capdevila, J.H. (2008). PPARalpha Ligands as Antitumorigenic Biol. 23, 785–793. and Antiangiogenic Agents. PPAR Res. 2008, 906542. Huang, J., Das, S.K., Jha, P., Al Zoughbi, W., Schauer, S., Claudel, T., Sexl, V., Vesely, P., Birner-Gruenberger, R., Kratky, D., et al. (2013). The PPARa agonist Puustinen, P., Junttila, M.R., Vanhatupa, S., Sablina, A.A., Hector, M.E., fenofibrate suppresses B-cell lymphoma in mice by modulating lipid meta- Teittinen, K., Raheem, O., Ketola, K., Lin, S., Kast, J., et al. (2009). PME-1 bolism. Biochim. Biophys. Acta 1831, 1555–1565. protects extracellular signal-regulated kinase pathway activity from protein phosphatase 2A-mediated inactivation in human malignant glioma. Cancer Jessani, N., Humphrey, M., McDonald, W.H., Niessen, S., Masuda, K., Res. 69, 2870–2877. Gangadharan, B., Yates, J.R., 3rd, Mueller, B.M., and Cravatt, B.F. (2004). Carcinoma and stromal enzyme activity profiles associated with breast tumor Rual, J.F., Venkatesan, K., Hao, T., Hirozane-Kishikawa, T., Dricot, A., Li, N., growth in vivo. Proc. Natl. Acad. Sci. USA 101, 13756–13761. Berriz, G.F., Gibbons, F.D., Dreze, M., Ayivi-Guedehoussou, N., et al. (2005). Kelly, T. (2005). Fibroblast activation protein-alpha and dipeptidyl peptidase IV Towards a proteome-scale map of the human protein-protein interaction (CD26): cell-surface proteases that activate cell signaling and are potential network. Nature 437, 1173–1178. targets for cancer therapy. Drug Resist. Updat. 8, 51–58. Santner, S.J., Dawson, P.J., Tait, L., Soule, H.D., Eliason, J., Mohamed, A.N., Kopp, F., Komatsu, T., Nomura, D.K., Trauger, S.A., Thomas, J.R., Siuzdak, Wolman, S.R., Heppner, G.H., and Miller, F.R. (2001). Malignant MCF10CA1 G., Simon, G.M., and Cravatt, B.F. (2010). The glycerophospho metabolome cell lines derived from premalignant human breast epithelial MCF10AT cells. and its influence on amino acid homeostasis revealed by brain metabolomics Breast Cancer Res. Treat. 65, 101–110. of GDE1(-/-) mice. Chem. Biol. 17, 831–840. Santos, A.M., Jung, J., Aziz, N., Kissil, J.L., and Pure´ , E. (2009). Targeting fibro- Lamar, J.M., Stern, P., Liu, H., Schindler, J.W., Jiang, Z.G., and Hynes, R.O. blast activation protein inhibits tumor stromagenesis and growth in mice. (2012). The Hippo pathway target, YAP, promotes metastasis through its J. Clin. Invest. 119, 3613–3625. TEAD-interaction domain. Proc. Natl. Acad. Sci. USA 109, E2441–E2450. Spandidos, A., Wang, X., Wang, H., and Seed, B. (2010). PrimerBank: a Liang, H., Kowalczyk, P., Junco, J.J., Klug-De Santiago, H.L., Malik, G., Wei, resource of human and mouse PCR primer pairs for gene expression detection S.J., and Slaga, T.J. (2013). Differential Effects on Lung Cancer Cell and quantification. Nucleic Acids Res. 38, D792–D799. Proliferation by Agonists of Glucocorticoid and PPARa Receptors. Mol. Carcinog. Published online April 26, 2013. http://dx.doi.org/10.1002/mc. Sweeney, K.J., Clark, G.D., Prokscha, A., Dobyns, W.B., and Eichele, G. 22029. (2000). Lissencephaly associated mutations suggest a requirement for the PAFAH1B heterotrimeric complex in brain development. Mech. Dev. 92, Long, J.Z., and Cravatt, B.F. (2011). The metabolic serine hydrolases and their 263–271. functions in mammalian physiology and disease. Chem. Rev. 111, 6022–6063. Manya, H., Aoki, J., Kato, H., Ishii, J., Hino, S., Arai, H., and Inoue, K. (1999). Vander Heiden, M.G., Cantley, L.C., and Thompson, C.B. (2009). Biochemical characterization of various catalytic complexes of the brain Understanding the Warburg effect: the metabolic requirements of cell prolifer- platelet-activating factor acetylhydrolase. J. Biol. Chem. 274, 31827–31832. ation. Science 324, 1029–1033. Menendez, J.A., and Lupu, R. (2007). Fatty acid synthase and the lipogenic Yang, M., Soga, T., and Pollard, P.J. (2013). Oncometabolites: linking altered phenotype in cancer pathogenesis. Nat. Rev. Cancer 7, 763–777. metabolism with cancer. J. Clin. Invest. 123, 3652–3658.