Metabolite Profiling Stratifies Pancreatic Ductal Adenocarcinomas

Home , Metabolic network

Metabolite profiling stratifies pancreatic ductal PNAS PLUS adenocarcinomas into subtypes with distinct sensitivities to metabolic inhibitors

Anneleen Daemena, David Petersonb,1, Nisebita Sahub,1, Ron McCordb,1, Xiangnan Duc, Bonnie Liuc, Katarzyna Kowanetzb, Rebecca Hongc, John Moffatd, Min Gaoc, Aaron Boudreaub, Rana Mroueb, Laura Corsonc, Thomas O’Brienc, Jing Qingc, Deepak Sampathc, Mark Merchantc, Robert Yauchb, Gerard Manninga, Jeffrey Settlemanb, Georgia Hatzivassiliouc, and Marie Evangelistab,2

aBioinformatics and Computational Biology, Genentech, South San Francisco, CA 94080; bDiscovery Oncology, Genentech, South San Francisco, CA 94080; cTranslational Oncology, Genentech, South San Francisco, CA 94080; and dBiochemical Pharmacology, Genentech, South San Francisco, CA 94080

Edited by Ronald A. DePinho, University of Texas M.D. Anderson Cancer Center, Houston, TX, and approved June 22, 2015 (received for review January 29, 2015) Although targeting cancer metabolism is a promising therapeutic therapeutic targets (9, 10). Just as tumors vary greatly in genomic strategy, clinical success will depend on an accurate diagnostic alterations that impact signaling and regulatory pathways, meta- identification of tumor subtypes with specific metabolic require- bolic transformation is also heterogeneous and dependent on tissue ments. Through broad metabolite profiling, we successfully identified type, proliferation rate, and isoenzyme use (9, 11). In addition, the three highly distinct metabolic subtypes in pancreatic ductal adeno- observed differences in the dependence on and utilization of the carcinoma (PDAC). One subtype was defined by reduced proliferative major nutrients—glutamine and glucose—are linked to oncogenic capacity, whereas the other two subtypes (glycolytic and lipogenic) signaling and the genomic features of a cancer cell (12). showed distinct metabolite levels associated with glycolysis, lipo- Large-scale pharmacogenomic screening is a powerful method genesis, and redox pathways, confirmed at the transcriptional level. for identifying biomarkers of drug response and can accelerate The glycolytic and lipogenic subtypes showed striking differences in the search for new cancer therapies (13, 14). In this study, we

glucose and glutamine utilization, as well as mitochondrial function, used broad baseline metabolite profiling in cell line models of CELL BIOLOGY and corresponded to differences in cell sensitivity to inhibitors of pancreatic ductal adenocarcinoma (PDAC), a disease context glycolysis, glutamine metabolism, lipid synthesis, and redox balance. previously associated with altered metabolism (15–18), to iden- In PDAC clinical samples, the lipogenic subtype associated with the tify metabolic subtypes within PDAC and predict their sensitivity epithelial (classical) subtype, whereas the glycolytic subtype to various metabolic inhibitors. strongly associated with the mesenchymal (QM-PDA) subtype, suggesting functional relevance in disease progression. Pharmaco- Results ∼ genomic screening of an additional 200 non-PDAC cell lines vali- Baseline Metabolite Profiling Identifies Three Metabolic Subtypes in dated the association between mesenchymal status and metabolic PDAC. We examined cell lines derived from naturally occurring drug response in other tumor indications. Our findings highlight the tumors because they recapitulate many aspects of the tissue type utility of broad metabolite profiling to predict sensitivity of tumors to a variety of metabolic inhibitors. Significance

metabolite profiling | metabolic subtypes in PDAC | glycolysis | lipid synthesis | biomarkers for metabolic inhibitors Targeting cancer metabolism requires personalized diagnostics for clinical success. Pancreatic ductal adenocarcinoma (PDAC) is characterized by metabolism addiction. To identify metabolic de- etabolic reprogramming during tumorigenesis is an essential pendencies within PDAC, we conducted broad metabolite profiling Mprocess in nearly all cancer cells. Tumors share a common and identified three subtypes that showed distinct metabolite phenotype of uncontrolled cell proliferation and must efficiently profiles associated with glycolysis, lipogenesis, and redox path- generate the energy and macromolecules required for cellular ways. Importantly, these profiles significantly correlated with growth. The first example of metabolic reprogramming was dis- enriched sensitivity to a variety of metabolic inhibitors including coveredmorethan80yagobyOttoWarburg:tumorcellscanshift inhibitors targeting glycolysis, glutaminolysis, lipogenesis, and re- from oxidative to fermentative metabolism in the course of onco- dox balance. In primary PDAC tumor samples, the lipid subtype genesis (1). More recently, there has been a resurgence of interest was strongly associated with an epithelial phenotype, whereas the in targeting cancer metabolism (2–4) because it may not only be ef- glycolytic subtype was strongly associated with a mesenchymal fective in inhibiting tumor growth, but may also provide a therapeutic phenotype, suggesting functional relevance in disease progression. window (5, 6). For example, inactivation of lactate dehydrogenase-A Our findings will provide valuable predictive utility for a number of (LDHA), an enzyme that catalyzes the final step of aerobic gly- metabolic inhibitors currently undergoing phase I testing. colysis, thereby reducing pyruvate to lactate, decreases tumorigenesis and induces regression of established tumors in mouse models of Author contributions: A.D., G.H., and M.E. designed research; A.D., D.P., K.K., R.Y., G.M., J.S., KRAS G.H., and M.E. performed research; A.D., D.P., N.S., R. McCord, X.D., B.L., K.K., R.H., J.M., lung cancer driven by oncogenic or epidermal growth factor M.G., A.B., R. Mroue, L.C., T.O., J.Q., and M.E. contributed new reagents/analytic tools; receptor (EGFR) while minimally affecting normal cell function A.D., D.P., N.S., R. McCord, X.D., B.L., K.K., R.H., J.M., M.G., A.B., R. Mroue, L.C., T.O., J.Q., (7). The finding that cancers have altered metabolism has prompted D.S., M.M., G.H., and M.E. analyzed data; and A.D. and M.E. wrote the paper. substantial investigation, both preclinically and in clinical trials, of The authors declare no conflict of interest. several metabolically targeted agents, including those that elevate This article is a PNAS Direct Submission. reactive oxygen species (ROS) or block glycolysis, lipid synthesis, Freely available online through the PNAS open access option. mitochondrial function, and glutamine synthesis pathways (8). 1D.P., R. McCord, and N.S. contributed equally to this work. The identification of distinct metabolic reprogramming events 2To whom correspondence should be addressed. Email: [email protected]. or metabolic subtypes in cancer may inform patient selection for This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. investigational metabolic inhibitors and in the selection of new 1073/pnas.1501605112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1501605112 PNAS Early Edition | 1of8 Downloaded by guest on October 2, 2021 and genomic context of cancer (13, 14, 19, 20). Levels of 256 with the lipogenic lines (P < 0.05; Fig. 1E and Fig. S1K). In con- metabolites were quantified in 38 pancreatic cancer cell lines (five trast, cell lines within the lipogenic subtype were enriched for ex- biological replicates per cell line) in logarithmic growth phase using pression of lipogenesis genes involved in cholesterol and de novo media with physiological glucose and glutamine concentrations lipid synthesis including 7-dehydrocholesterol reductase (DHCR7), (Datasets S1–S3). We applied nonnegative matrix factorization stearoyl-CoA desaturase (SCD), and fatty acid synthase (FASN) (NMF) (21), a recently established approach for consensus clus- (adjusted P < 0.1; Fig. 1 E and F, Fig. S1 H and L,andDataset tering (22–24), to 153 metabolites with reproducible variation, S6). Thus, PDAC-derived cell lines can be clustered by their allowing the capture of the strongest signal of metabolic de- metabolite profiles and these differences appear to be determined pendency (SI Materials and Methods). This analysis revealed three in part by differences in gene expression. stable and reproducible subtypes with adequate data coherence (Fig. S1 A and B). The metabolite profiles of the cell lines ordered Glycolytic and Lipogenic Subtypes Use Glucose and Glutamine in a by subtype are shown in Fig. 1A for metabolites with distinct in- Different Manner. The metabolic and transcriptional profiles sug- tensities in at least one subtype compared with the other two gested that these two subtypes may differ in their use of glucose subtypes (F test, P < 0.05). These three subtypes provided a useful and glutamine, the most abundant carbon sources available to and interpretable basis for further analysis. cancer cells. We predicted that the lipogenic subtype would preferentially use glucose for the tricarboxylic acid (TCA) cycle Metabolic Characterization Reveals a Slow Proliferating, Glycolytic, and lipid synthesis, whereas the glycolytic subtype would use glu- and Lipogenic Subtype. The metabolite intensities within each cose more for aerobic glycolysis, and consequently, use more glu- subtypewerethenmappedtoknown,previouslyestablished tamine for TCA anaplerosis. 13C metabolic mass isotopomer metabolic ontologies (Dataset S1 and SI Materials and Methods) distribution analysis (MIDA) using either [U-13C5]glutamine or (25). One subtype (34% of all lines) was especially low in amino [U-13C6]glucose revealed a significant increase in the contribution of acids and carbohydrates (Fig. 1A,leftsubtype,andFig. S1C). Cell [U-13C6]glucose to TCA metabolites in representative cell lines lines in this subtype had an average doubling time that was sig- from the lipogenic subtype relative to the glycolytic subtype (Fig. nificantly higher (Fig. S1D) and were named the slow proliferating 2A; P < 0.05). In contrast, representative glycolytic lines in- subtype. Doubling times for cell lines from the other two subtypes corporated [U-13C5]glutamine into TCA metabolites at signifi- were more similar (Fig. S1D); however, these two subtypes dis- cantly higher levels than lines from the lipogenic subtype (Fig. 2B; played strikingly distinct metabolic profiles, independent of pro- P < 0.05). Moreover, lipogenic cell lines incorporated 14C-glucose into liferation rate (SI Materials and Methods). Thus, these metabolic lipid metabolites at a significantly higher level than cell lines from subtypes have unique metabolic profiles that are independent of the glycolytic subtype (Fig. 2C; P < 0.01). Consistent with these growth rate. observations, lipogenic lines showed on average higher O2 con- We further explored the metabolic differences between the two sumption (Fig. 2D; P < 0.01) and a greater mitochondrial content subtypes with similar proliferation rates. One subtype (27% of all [Mitotracker and tetramethylrhodamine ethyl ester (TMRE) in- lines; Fig. 1A) exhibited, on average, elevated levels of various tensity] compared with glycolytic subtype lines (Fig. 2E; P < 0.01; components of the glycolytic and serine pathways, mainly phos- Dataset S7). Thus, cell lines from the glycolytic and lipogenic phoenolpyruvate (PEP), glyceraldehyde-3-phosphate, lactate, and subtypes appear to use glucose and glutamine in a different manner. serine (Fig. 1 B and C and Fig. S1E), and was named the glycolytic subtype. This subtype was also distinguished by lower levels of Glycolytic and Lipogenic Cell Lines Show Distinct Sensitivity to metabolites important for redox potential such as nicotinamide Various Metabolic Inhibitors in Vitro. Based on their distinct meta- adenine dinucleotide (NAD) reduced (NADH), NAD phos- bolic wiring, we predicted that glycolytic and lipogenic cell lines phate (NADP), NAD phosphate reduced (NADPH), glutathione would show differential sensitivity to inhibitors targeting aerobic disulfide (GSSG), glutathione (GSH), and flavine adenine di- glycolysis (oxamate and the LDHA inhibitor GNE-140) (26), nucleotide (Fig. 1 B and C, Fig. S1F,andDataset S4). In contrast, glutaminolysis [bis-2-(5-phenylacetimido-1,2,4,thiadiazol-2-yl)ethyl the other subtype (39% of all lines; Fig. 1A) was enriched for sulfide (BPTES)], and de novo lipid synthesis [FASN inhibitor various lipid metabolites such as palmitic acid (C16:0), oleic acid GSK1195010 (27), SCD inhibitor (28), cerulenin, and orlistat]. (C18:cis[9]1), palmitoleic acid (C16:cis[9]1), and myristic acid Indeed, as predicted, the glycolytic subtype was enriched for lines (C14:0) (Fig. 1 B and D and Dataset S4), as well as mitochondrial that were sensitive to the LDHA inhibitor, oxamate, and BPTES, [oxidative phosphorylation (OXPHOS)] metabolites important for whereas the lipogenic subtype was enriched for lines that were the electron transport chain such as coenzyme Q10 and coenzyme sensitive to inhibitors targeting lipid synthesis (Fig. 3A and Fig. Q9 and components of the aspartate-malate shuttle such as as- S2A; P < 0.05; Dataset S7 and SI Materials and Methods). More- partate and glutamate (Fig. S1G and Dataset S4), and was named over, glycolytic cell lines showed higher rates of fatty acid (FA) the lipogenic subtype. uptake (Fig. S2B) and increased sensitivity to media with reduced lipid content (Fig. S2C), suggesting these lines may be more re- Differences Between Glycolytic and Lipogenic Subtypes Are Confirmed liant on FA pathways for generating lipids. Transcriptionally. We next determined whether differences in me- Maintaining redox balance is another key requirement for can- tabolite levels observed between the glycolytic and lipogenic subtypes cer cells (29). The differences in redox-related metabolites between could be explained by differences in the expression of genes glycolytic and lipogenic cell lines suggested that they may also show known to be associated with the metabolic ontologies (Dataset S5 differential sensitivity to ROS-inducing agents or inhibitors of en- and SI Materials and Methods). Consistent with the differences in zymes or transporters important for maintaining glutathione syn- metabolite levels, expression of genes associated with glycolysis and thesis and NADP/NADPH balance in cells. Indeed, we found that the pentose phosphate pathway were found to be relatively ele- cell lines within the glycolytic subtype showed enhanced sensitivity vated in cell lines from the glycolytic subtype (Fig. 1 E and F, Fig. to a variety of such agents including inhibitors of gamma-gluta- S1 H and I, and Dataset S6). For example, most glycolytic lines mylcysteine synthetase [buthionine sulphoximine (BSO)], demonstrated higher expression of neuron-specific enolase [ENO2; and the cystine transporter xCT {(S)-4-carboxyphenylglycine adjusted P = 0.0016; Fig. 1 E and F], along with higher levels of its [(S)-4-CPG]} (Fig. 3B and Dataset S7). product PEP, whereas other enolase homologs were not differen- In addition to short-term (3 d) culture assays, we tested the tially expressed (Fig. S1J). We also noted that protein (and not efficacy profile of LDHA inhibitor, oxamate, and the SCD in- mRNA) abundance of the lactate transporter, monocarboxylate hibitor in long-term (12 d) culture assays and observed similar transporter 1 (MCT1) was elevated in the glycolytic lines compared results (Fig. 3 A and C).

2of8 | www.pnas.org/cgi/doi/10.1073/pnas.1501605112 Daemen et al. Downloaded by guest on October 2, 2021 PNAS PLUS

3.9 1.9 1.5 2.0 0 +2.1 +4.2 D A *** *** 1.0 1.0 Slow Proliferating Glycolytic Lipogenic 0.5 0.0 0.0

-0.5 -1.0 log2 Oleic acid (RU) (RU) acid Oleic log2 -1.0

log2 Palmitoleic acid (RU) -2.0 Carbo- hydrates

Glycolytic Lipogenic Glycolytic Lipogenic 2.0 3.0 *** *** Amino acids 2.0 1.0 1.0 REDOX 0.0 0.0 -1.0 Fatty acids -1.0 log2 Myristic acid (RU) (RU) acid Myristic log2

log2 Palmitoleic acid (RU) -2.0 -2.0 c

Glycolytic Lipogenic Glycolyti Lipogenic E Glucose G1P G6P Ribulose-5-P

2,6BP F6P

F1,6BP PPP PPP Glyceraldehyde 3-P Glycerol-3-P Glycolysis MVD PSPH 3-Phosoglycerate Serine HMGCS1 FDFT1 DGAT1 AsPCYAPCCFPKP TCCHs 766TCapanPancPanc 02.03Panc 04.03Panc 03.27KLM 05.04PancMIA 10.05 KPPacaPSN1HUPPK PL45PA KP4PK SW P1990HPACKCI P SUITPA H SU.86.86PK PK D HUPBxPCPK KP ENO2 A ANC P AN AF PEP DHCR7 SCD

TU TU TU CELL BIOLOGY AC 3 2 45H 45P MOH1 8 1 59 3 P 1 T3 G T4 L 1 AN2 1 2 II 3 MCT1 1 2 8988T 8902 8988S SCD5DL FASN 2 Lactate Pyruvate DHCR24 ACAT2 PDH PDK1 Citrate Lipids Acetyl-CoA B Citrate Isocitrate Sphingolipids Oxaloacetate

Malate Amino acids Succinate Fumarate a-KG

Fatty acids Glycolytic Lipogenic a-KG p < 0.05 Aspartate Glutamate OXPHOS / p < 0.1 Oxaloacetate mitochondria−related REDOX p < 0.2 Glutamine Other complex lipids F 6 8 Carbohydrates *** ** 5 7 4 Glycolysis / PPP 6 3 Cholines 5 2 log2 ENO2 (RPKM +1) 1 +1) (RPKM DHCR7 log2 4 −5 0 5 10 C JG score 4.0 Glycolytic Lipogenic Glycolytic Lipogenic 14.0 6.0 10 8 ** * ** 2.0 * * 12.0 4.0 8 7 0.0 10.0 -2.0 6 6 2.0

log2 PEP (RU) 8.0 log2 GAP (RU)

-4.0 (RU) GSSG log2 4 5 log2 SCD (RPKM +1) 6.0 -6.0 0.0 2 log2 FASN (RPKM +1) 4 nic

Glycolytic Lipogenic Glycolytic Lipogenic Glycolytic Lipoge Glycolytic Lipogenic Glycolytic Lipogenic

Fig. 1. Identification of distinct metabolic subtypes in PDAC through baseline metabolite profiling. (A) Hierarchical clustering of identifiable metabolites with significant intensity differences between any of the three subtypes (F test, P < 0.05; 99 metabolites). Cell lines were grouped by subtype, with the order per subtype defined by unsupervised clustering. Log2 intensity ratio data per metabolite are scaled across all cell lines to mean = 0 and SD = 1. Blue indicates low scaled intensity, and yellow indicates high for each metabolite. Highlighted in gray are functionally related metabolites. Slow proliferating lines are labeled in gray, glycolytic lines in purple, and lipogenic in cyan. (B) Relative enrichment of the eight metabolic ontology classes in the glycolytic and lipogenic subtypes, represented by JG score (47). Positive scores represent ontologies enriched for metabolites with high intensities in the glycolytic subtype. See Dataset S1 for a description and list of metabolites per ontology and Dataset S4 for the list of differentially expressed metabolites. (C) Normalized metabolite intensity levels for metabolites involved in glycolysis/pentose phosphate and redox pathways that were differentially expressed between glycolytic and lipogenic lines. RU stands for relative unit, with intensity levels normalized to a reference pool of samples for metabolites from the Broad Profiling platform (Dataset S2) and to a universal 13C-labeled internal standard for metabolites from the Energy platform (Dataset S3). (D) Normalized metabolite intensity levels for metabolites involved in lipid synthesis that were differentially expressed between glycolytic and lipogenic lines. (E) Detailed metabolite map with genes differentially expressed between cell lines from the glycolytic vs. lipogenic subtype indicated with various shades of color depending on P value corrected for multiple testing. For MCT1, P value is based on protein expression level. We refer to Dataset S6 for a list of differentially expressed genes. (F) Expression levels of ENO2, DHCR7, SCD, and FASN involved in glycolysis and lipid synthesis that were differentially expressed between glycolytic and lipogenic lines (Dataset S6). Asterisks denote a statistically significant difference by unpaired t test with Welch’s correction (*P < 0.05, **P < 0.01, ***P < 0.001).

Daemen et al. PNAS Early Edition | 3of8 Downloaded by guest on October 2, 2021 A B 40 40 40 25 * ** * * 20 30 30 30 15 20 20 20 10 10 10 13C-Glutamine 10 13C-Glutamine

5 % a-KG (M5) from % Malate (M4) from 0 0 0 0 % a-KG (M2) from 13C-Glucose % Citrate (M2) from 13C-Glucose from (M2) Citrate %

Glycolytic Lipogenic Glycolytic Lipogenic Glycolytic Lipogenic Glycolytic Lipogenic 40 15 10 ** * * 30 8 10 6 20 4 5 10 13C-Glutamine 2 % Aspartate (M2) from 13C-Glucose from % Glutamate (M5) from from (M5) Glutamate % 0 0 % Malate (M2) from 13C-Glucose

C Glycolytic Glycolytic Lipogenic Glycolytic Lipogenic Lipogenic

C 25000 E 1200 1200 ** ** ** 1000 1000 20000 800 800 15000 600 600 10000 400 400 5000 (CPM/mg protein) (CPM/mg 200 200 TMRE Intensity/cell

14C-Glucose into Lipids Lipids into 14C-Glucose 0 Mitotracker Intensity/cell Mitotracker 0 0 c

Glycolytic Lipogenic Glycolytic Lipogenic Glycolytic Lipogeni D 0.12 ** 0.10 Glycolyticy y Lipogenicp g 0.08 0.06

0.04 TMRE 0.02

OCAR (pmoles/min/cell) OCAR 0.00 Mitotracker Glycolytic Lipogenic SW 1990 HUP-T3 MIA Paca-2 PA-TU-8988S HPAC SUIT-2

Fig. 2. Functional characterization of glycolytic and lipogenic subtypes. (A) Comparison of relative contribution of glucose oxidation to the TCA metabolites, 13 determined by M2 labeling from [U- C6]glucose for citrate, αKG, malate, and aspartate between glycolytic and lipogenic cell lines. (B) Comparison of relative 13 contribution of reductive glutamine metabolism to TCA metabolites, determined by M5 labeling from [U- C5]glutamine for αKG and glutamate, and M4 labeling for malate between glycolytic and lipogenic cell lines. (C) Comparison of relative contribution of glucose metabolism to de novo lipid synthesis 14 14 between glycolytic and lipogenic cell lines. Cells were labeled with 1 μCi/mL D[U- C] glucose for 6 h, and lipids were extracted. The incorporation of Cinto lipids was determined by scintillation counting. (D) Comparison of oxygen consumption rates (OCRs) between glycolytic and lipogenic cell lines. (E) Com- parison of relative mitochondria number (Mitotracker intensity per cell) and potential/fitness (TMRE per cell) between glycolytic and lipogenic cell lines. For A–E, the mean and SD between cell lines belonging to the glycolytic subtype vs. lipogenic subtype is plotted where each cell line is shown as one dot, representing the mean of three replicates. Data are normalized to sample protein content (A–C) or cell number (D and E). Asterisks denote a statistically significant difference by unpaired t test with Welch’s correction (*P < 0.05, **P < 0.01, ***P < 0.001).

Functional Confirmation of the Glycolytic and Lipogenic Subtype in an SCD inhibitor showed no efficacy (Fig. 3E), although phar- Vivo. To translate these findings in vivo and generate proof-of- macodynamic inhibition of SCD was seen (Fig. 3F). In contrast, concept findings for our two metabolic subtypes, we evaluated HPAC xenograft tumors showed minimal sensitivity to LDHA xenografts of MIA Paca-2, a glycolytic cell line, and HPAC, a knockdown (9% TGI; Fig. S2 D and E) but showed significant lipogenic cell line, for their sensitivity to glycolysis vs. lipid syn- tumor growth inhibition to SCD inhibitor treatment (52% TGI) thesis inhibition. Because oxamate and LDHA inhibitors have (30). Thus, glycolytic and lipogenic subtypes are functionally dis- poor pharmacokinetics in mice (26), we inhibited glycolysis by tinct and show differential sensitivity to glycolytic and lipid engineering MIA Paca-2 and HPAC cells to express a doxycline biosynthesis inhibition. (DOX)-inducible shRNA against LDHA. MIA Paca-2 xenograft tumors treated with DOX showed undetectable levels of LDHA Glycolytic and Lipogenic Subtypes Are Associated with Known (Fig. 3D) and 68% tumor growth inhibition (TGI) compared with Subtypes of PDAC, Driven by Mesenchymal Status. We next set tumors expressing LDHA (Fig. 3E). In contrast, administration of out to determine how our defined metabolic subtypes associated

4of8 | www.pnas.org/cgi/doi/10.1073/pnas.1501605112 Daemen et al. Downloaded by guest on October 2, 2021 AC PNAS PLUS

B D CELL BIOLOGY F

Fig. 3. Glycolytic and lipogenic cell lines show distinct sensitivity to various metabolic inhibitors both in vitro and in vivo. (A)ComparisonofIC50 values to various metabolic inhibitors between representative glycolytic vs. lipogenic cell lines in short-term (3 d) viability assays. Saturated IC50 values correspond to cell lines where an IC50 was not reached at the maximum drug concentration. The mean and SD between cell lines belonging to the glycolytic vs. lipogenic subtype is plotted where each cell line is shown as one dot, representing the mean of three replicates. Asterisks denote a statistically significant difference by Mann–Whitney test

(*P < 0.05, **P < 0.01, ***P < 0.001). (B) Comparison of IC50 values to various ROS-inducing agents between representative glycolytic vs. lipogenic cell lines in short- term (3 d) viability assays, similar to A.(C) Comparison of sensitivity to oxamate, LDHA, or SCD inhibitors between representative glycolytic vs. lipogenic cell lines in longer-term (12 d), low seeding density growth assays. (D) Western blots showing 98% in vivo knockdown of LDHA levels in MIA Paca-2 xenografts administered with doxycycline (1 mg/mL) for 8 d vs. 5% sucrose. (E) In vivo knockdown of LDHA (n = 10 for each group) results in 68% TGI, 95% confidence interval [48, 83] in the MIA Paca-2 shLDHA model of a glycolytic subtype tumor, whereas treatment with an SCD inhibitor (75 mg/kg, orally, BID) resulted in no significant change in tumor volume. (F) Confirmed pharmacodynamic inhibition of lipid metabolism by SCD inhibitor. The SCD inhibitor reduces desaturation of palmitate and stearate in MIA Paca-2 shLDHA xenograft tumor tissues and in mouse liver and plasma (n = 5 per group). Data are presented as mean ± SD.

with primary PDAC tumor samples from patients. Three clinical were associated with the classical subtype (Fig. 4A; P = 0.0006; subtypes of PDAC were recently identified through molecular Dataset S7). Thus, our metabolite subtypes derived from pan- profiling of PDAC tumors: classical (characterized by high ex- creatic cell lines strongly correlate with known subtypes of pression of adhesion-associated and epithelial genes), quasi-mesen- PDAC tumors, with the glycolytic subtype strongly associating chymal (QM-PDA, characterized by mesenchyme-associated with mesenchymal features and the lipogenic subtype associ- genes), and exocrine-like (22). Because exocrine-like cell lines ating with epithelial features. have not been reported, we simplified the three-subtype PDAC signature to a 42-gene expression signature that distinguishes Metabolic and Mesenchymal Markers Predict Response to Glycolytic classical from QM-PDA (22), and applied it to our cell line panel. and Glutaminolytic Inhibitors in PDAC and Other Tumor Types. Car- We found that all cell lines within the glycolytic subtype associated cinomas with mesenchymal features (including PDAC) tend to be with the quasimesenchymal subtype, whereas most lipogenic lines more aggressive and typically have an overall poorer prognosis

Daemen et al. PNAS Early Edition | 5of8 Downloaded by guest on October 2, 2021 LDHAi BPTES A Glycolytic B PA TU 8988S Lipogenic HPAF II OXPHOS OXPHOS HUP T4 HPAC KCI MOH1 Amino acids KP 3L Amino acids PK 59 SUIT 2 PK 8 Lipids Lipids SU.86.86 KP 2 3 BxPC PPARA Glycolysis / PPP PA TU 8902 PANC 1 HUP T3 DAN G Glycolysis / PPP PPARA SW 1990 PL45 PK 45H 50 5 15 10 50 5 PK 45P JGJG scorescore JGJG score PA TU 8988T PSN1 Up in Up in Up in Up in KP4 resistant sensitive resistant sensitive MIA Paca 2

2 10 12D hymal Score EpithelialEpithelial / / MesenchymalMesenc Score 60 **** 30 **** **** Classical QMPDA (Epithelial) (Mesenchymal) 40 20 1000

20 10 BSO IC50 ( IC50 BSO

C BPTES IC50 ( M) 100 Oxamate IC50 (mM) IC50 Oxamate −2.4 +1.7 −4.4 0 +3.8 0 0 h w h w w

EMT Lo DGAT1 EMT Hig EMT Lo EMT Hig EMT Lo EMT High DHCR7 E 500 **** 30 **** **** FDFT1 400 1000 HMGCS1 20 300 MVD 200 10 Lipid Ave 100 BSO IC50 (nM)

BPTES IC50 ( IC50 BPTES M) 100 ENO2 Oxamate IC50 (mM) 0 0 ENO2 / Lipid Ave

Vim High Vim Low Vim High Vim Low Vim High Vim Low

F Glycolytic Subtype Lipogenic Subtype Glucose Glucose

Pentose Pathway

Serine

Lactate Pyruvate Pyruvate Lipid synthesis Mitochondria Mitochondria (TCA cycle) (TCA cycle)

Glutamine

Mesenchymal Tumors Epithelial Tumors Sensitive to: Sensitive to: -Glycolytic, Glutamine inhibitors, - Lipid inhibitors ROS-inducing agents

Fig. 4. Metabolic and mesenchymal markers predict response to glycolytic and glutaminolytic inhibitors in PDAC and other tumor types. (A)Epithelial/mesenchymal score for the glycolytic and lipogenic cell lines based on a 42-gene set characteristic of the classical and QM-PDA subtypes (22). The score is defined as the difference in average expression of QM-PDA vs. classical genes, with a positive score indicative of QM-PDA and a negative score of classical. Cell lines are colored by metabolic subtype, with glycolytic lines in purple and lipogenic lines in cyan. All glycolytic cell lines are of the QM-PDA subtype, whereas lipogenic cell lines are associated with the classical subtype (Fisher’s exact test, P = 0.0006). (B) Relative enrichment of the five curated metabolism gene sets in cell lines that are sensitive (positive JG score) or resistant (negative JG score) to LDHA inhibitor or BPTES in a pan-cancer panel of 204 and 167 nonpancreatic cell lines, respectively, after exclusion of cell lines with intermediate response. See Dataset S5 for a list of genes per gene set. (C) Metabolic dependency preference in the panel of ∼200 nonpancreatic cell lines is based on the ratio of ENO2 expression to the average expression of five lipid genes, and labeled on top of the heatmap as glycolytic in purple (ratio > third quartile), lipogenic in cyan (ratio < lower quartile), and undefined type in gray (ratio between lower and third quartile). Shown are expression (log2 RPKM + 1) of glycolytic gene ENO2, five lipid genes DGAT1, DHCR7, FDFT1, HMGCS1,andMVD, average expression of the five lipid genes (Lipid Ave), and the ratio of ENO2 to average lipid expression (ENO2/Lipid Ave). Data from Dataset S8.(D) High expression of a pan-cancer EMT signature (EMT) associates with sensitivity to oxamate, BPTES, and BSO across a variety of tumor types. EMT low is defined by RPKM values < lower quartile, EMT high = RPKM values > third quartile. Asterisks denote a statistically significant difference by Mann–Whitney test (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). (E) High expression of mesenchymal marker vimentin (Vim) associates with sensitivity to oxamate, BPTES, and BSO across a variety of tumor types. Vim low is defined by RPKM values < lower quartile, Vim high = RPKM values > third quartile. Asterisks as per D.(F) Model of preferential glucose and glutamine utilization in the glycolytic vs. lipogenic subtype.

6of8 | www.pnas.org/cgi/doi/10.1073/pnas.1501605112 Daemen et al. Downloaded by guest on October 2, 2021 (22, 31, 32). Given the strong association between quasi-mesenchymal ENO2, which converts 2-phosphoglycerate (2-PG) to PEP, was PNAS PLUS status and glycolytic dependency in the PDAC lines, we asked also one of the most differentially expressed genes between these whether this association might also exist in other tumor types. two subtypes, suggesting that inhibitors of ENO2 may be par- We screened ∼200 nonpancreatic cancer cell lines, representing ticularly effective against glycolytic tumors. Enolases act down- various tumor types, for sensitivity to inhibitors of aerobic gly- stream of phosphoglycerate mutase (PGAM1) and regulate colysis and glutaminolysis, as well as to ROS-inducing agents pyruvate kinase (PK) M2 isoform (PKM2), genes that are par- (Dataset S8). As in PDAC (Fig. 1E and Fig. S1H), we found that ticularly active in glycolytic tumors and have recently attracted cell lines most sensitive to the LDHA inhibitor, oxamate, and attention for their role in serine biosynthesis through regulation BPTES were associated with a glycolytic signature, whereas cell of 3-phosphoglycerate dehydrogenase (PHGDH) (34). ENO2 has lines that were most resistant to these inhibitors were associated also been proposed as a target in ENO1-deleted glioblastomas with an OXPHOS signature (Fig. 4B and Fig. S3A). We next (35). Our findings further substantiate the biological rationale assigned each cell line to a metabolic subtype (glycolytic vs. for targeting ENO2 in a subset of cancers. lipogenic) using the glycolytic and lipid genes that were most Finally, we demonstrated that the observed metabolic sub- differentially expressed in the PDAC metabolic subtypes (ad- types correlate with epithelial vs. (quasi)-mesenchymal cell states justed P < 0.05). Using the ratio of expression of glycolytic gene both in PDAC and other cancer types. We propose a model (Fig. ENO2 to the average expression of lipid genes diacylglycerol 4F) in which mesenchymal tumors are metabolically wired to O-acyltransferase 1 (DGAT1), DHCR7, farnesyl-diphosphate preferentially use glucose for glycolysis and lactate production farnesyltransferase 1 (FDFT1), 3-hydroxy-3-methylglutaryl-CoA and use glutamine for generating TCA metabolites, whereas synthase 1 (HMGCS1), and mevalonate (diphospho) decarbox- epithelial tumors preferentially use glucose for the TCA cycle ylase (MVD) clearly distinguished nonpancreatic cell lines by and de novo lipogenesis. Moreover, our analysis suggests that metabolic dependency preference (Fig. 4C). A glycolytic pref- mesenchymal tumors may be more vulnerable to ROS-inducing erence in nonpancreatic lines associated with sensitivity to LDHA agents, potentially through differences in NADPH balance and inhibitor, oxamate, and BPTES (Fig. S3B; P < 0.05; Dataset S8). In antioxidant responses (36). addition, consistent with our findings in the PDAC tumors, mes- Such differences in metabolic vulnerabilities between epithelial enchymal tumors [according to a pan-cancer epithelial-mesenchy- and mesenchymal states could arise from the activation of signaling mal transition (EMT) signature (33) or vimentin] were more pathways associated with these states. For example, epithelial sensitive to the LDHA inhibitor, oxamate, BPTES, and ROS- subtypes have previously been shown to be enriched for activating

S CELL BIOLOGY inducing agents [BSO and ( )-4-CPG] across a variety of tumor mutations in receptor tyrosine kinases (RTK) such as EGFR (37) D E C P < types (Fig. 4 and and Fig. S3 ; 0.001; Dataset S8). A and PI3K/AKT signaling pathways (23), leading to activation of the similar discrepant dependency was observed in the slow pro- mechanistic target of rapamycin (mTOR). mTOR increases both liferating PDAC cell lines, with six lines more glycolytic and/or protein synthesis and lipogenesis through mechanisms including mesenchymal and six lines more lipogenic and/or epithelial, despite enzyme phosphorylation and transcriptional activation of EIF1A D E their slower proliferation (Fig. S3 and ). Thus, mesenchymal (38) and SREBP1 (39–41). In contrast, mesenchymal states are tumors, regardless of indication, appear to share common metabolic associated with increased c-Myc expression and HIF1A, which have vulnerabilities, and agents that block glycolysis, glutamine metabo- been shown to drive a glycolytic profile (42, 43). Regardless of the lism, or redox balance may be particularly effective. These results nature or mechanism of action for the metabolic variation we ob- support a model in which metabolic plasticity with regard to bio- served, our data provide valuable predictive utility and thereby energetic pathways is limited, and, consequently, unique metabolic inform clinical evaluation of a variety of metabolic inhibitors such dependencies exist in tumors that can be exploited for cancer as MCT and glutaminase inhibitors currently undergoing phase I therapy based on tumor subtype. testing across a variety of tumor indications. Discussion Materials and Methods Using broad metabolite profiling, we successfully stratified Detailed materials and methods can be found in SI Materials and Methods. PDAC-derived cell lines into discrete metabolic subtypes. Pre- All cell lines listed in Dataset S9 were grown in RPMI (without glucose, vious metabolic profiling studies have been conducted in tumors without glutamine) media (US Biological #R9011) supplemented with 6 mM and in cell lines of the NCI-60 panel with different end points glucose, 2 mM glutamine, 5% FBS, 100 μg/mL penicillin, and 100 U/mL (9). However, this study is the first, to our knowledge, to suc- streptomycin. Metabolite profiling was performed as previously described cessfully identify metabolic subtypes through profiling of a large (44). For flux analysis, cells were cultured for ∼18 h in RPMI with 10% 13 number of samples within one tissue type and to demonstrate (vol/vol) dialyzed FBS supplemented with either 3 mM D[U- C]glucose or 13 that each subtype is enriched for drug sensitivity to unique 1mML[U- C]glutamine. Data analysis was carried out with the MultiQuant classes of metabolic inhibitors. software. For short-term viability assays, cells were plated using optimal Although metabolic clustering accounted for a substantial frac- seeding densities in 384-well plates. The following day, cells were treated tion of the drug response variation observed across cancer cell with LDHA inhibitor GNE-140 (26), oxamate (Sigma cat# O2751), SCD inhibitor (28), FASN inhibitor GSK1195010 (27), cerulenin, orlistat, BSO, lines, some heterogeneity in drug response within the lipogenic SI Text S-4-CPG, aminooxyacetic acid (AOA), and BPTES (45), using a 6-pt dose titra- subtype remained (see and Figs. S4 and S5 for a discussion tion scheme. After 72 h, cell viability was assessed using the CellTiter-Glo on heterogeneity). Some cell lines were clearly “hard-wired” for Luminescence Cell Viability assay. Absolute inhibitory concentration (IC) lipogenesis and showed sensitivity to all lipid inhibitors tested, values were calculated using four-parameter logistic curve fitting and are whereas the more refractory lines appeared to be capable of averages from a minimum of two independent experiments. For long-term switching to alternative pathways, perhaps those involving fatty acid growth assays, glycolytic cell lines (MIA Paca-2, SW 1990, PSN1, and HUP-T3) uptake. Further understanding of the nature and plasticity of met- and lipogenic cell lines (PA-TU-8902, PK-8, KP-3L, and SUIT-2) were seeded in abolic networks in these cancer cells will be required to more ac- a 6-well dish at 3,000 cells per well overnight (RPMI, 5% serum, 2 mM glu- curately predict their sensitivity to specific classes of metabolic tamine) and then treated in media with indicated concentrations of oxamate, inhibitors. In addition, although we successfully translated our in SCD inhibitor, or DMSO for 12 d at 37 °C and 5% CO2. Fatty uptake assays were performed using the Free Fatty Acid Uptake Assay Kit (ab176768) according vitro findings in vivo, additional factors within the tumor microen- to the manufacturer’s protocol. Reduced serum experiments were carried vironment (tumor-stroma signaling, angiogenesis, and hypoxia) will out using 3% delipidated serum (SeraCare 502099) and 1% FBS (SeraCare influence sensitivity and adaptation to metabolic inhibition in vivo. CC5010-500). Seahorse Bioscience assays were used for oxygen consump- Our study also identified PEP as one of the most differentially tion. All procedures involving animals were reviewed and approved by the expressed metabolites between glycolytic and lipogenic cell lines. Institutional Animal Care and Use Committee (IACUC) at Genentech and

Daemen et al. PNAS Early Edition | 7of8 Downloaded by guest on October 2, 2021 carried out in an AAALAC (Association for the Assessment and Accreditation of ACKNOWLEDGMENTS. We thank Richard Bourgon, Eva Lin, Billy Lam, Laboratory Animal Care) accredited facility. All statistical analyses were per- Yihong Yu, and Arjan Gower for help with cell-based drug screens and 13 formed in R 3.0.0 (46). The optimal number of metabolic subtypes was data analysis, Mandy Kwong for advice on C metabolic mass isotopomer distribution analysis (MIDA), Allison Bruce for assistance with obtained with nonnegative matrix factorization, using the NMF package. The the metabolic diagram, and Metanomics Health (Lisette Leonhardt, DESeq2 package was used for differential expression analysis. Metabolic on- Ulrike Rennefarhrt, Oliver Schmitz, and Hajo Schiewe) for technical sup- tology and gene set enrichment analyses were based on GSEAlm. port on metabolite profiling.

1. Warburg O, Posener K, Negelein E (1924) On the metabolism of carcinoma cells. 32. Peinado H, Olmeda D, Cano A (2007) Snail, Zeb and bHLH factors in tumour pro- Biochem Z 152:309–344. gression: An alliance against the epithelial phenotype? Nat Rev Cancer 7(6):415–428. 2. Dang L, et al. (2009) Cancer-associated IDH1 mutations produce 2-hydroxyglutarate. 33. Cheng WY, Ou Yang TH, Anastassiou D (2013) Development of a prognostic model Nature 462(7274):739–744. for breast cancer survival in an open challenge environment. Sci Transl Med 5(181): 3. Deberardinis RJ, Sayed N, Ditsworth D, Thompson CB (2008) Brick by brick: Metabo- 181ra50. lism and tumor cell growth. Curr Opin Genet Dev 18(1):54–61. 34. Hitosugi T, et al. (2012) Phosphoglycerate mutase 1 coordinates glycolysis and bio- 4. Hsu PP, Sabatini DM (2008) Cancer cell metabolism: Warburg and beyond. Cell 134(5): synthesis to promote tumor growth. Cancer Cell 22(5):585–600. 703–707. 35. Muller FL, et al. (2012) Passenger deletions generate therapeutic vulnerabilities in 5. Anastasiou D, et al. (2012) Pyruvate kinase M2 activators promote tetramer formation cancer. Nature 488(7411):337–342. and suppress tumorigenesis. Nat Chem Biol 8(10):839–847. 36. Gorrini C, Harris IS, Mak TW (2013) Modulation of oxidative stress as an anticancer 6. Le A, et al. (2012) Glucose-independent glutamine metabolism via TCA cycling for strategy. Nat Rev Drug Discov 12(12):931–947. proliferation and survival in B cells. Cell Metab 15(1):110–121. 37. Yauch RL, et al. (2005) Epithelial versus mesenchymal phenotype determines in vitro 7. Xie H, et al. (2014) Targeting lactate dehydrogenase–a inhibits tumorigenesis and sensitivity and predicts clinical activity of erlotinib in lung cancer patients. Clin Cancer tumor progression in mouse models of lung cancer and impacts tumor-initiating cells. Res 11(24 Pt 1):8686–8698. – Cell Metab 19(5):795 809. 38. Lomakin IB, Steitz TA (2013) The initiation of mammalian protein synthesis and mRNA 8. Available at www.clinicaltrials.gov. Accessed December, 2014. scanning mechanism. Nature 500(7462):307–311. 9. Jain M, et al. (2012) Metabolite profiling identifies a key role for glycine in rapid 39. Gingras AC, Raught B, Sonenberg N (2001) Regulation of translation initiation by – cancer cell proliferation. Science 336(6084):1040 1044. FRAP/mTOR. Genes Dev 15(7):807–826. 10. Sreekumar A, et al. (2009) Metabolomic profiles delineate potential role for sarcosine 40. Luyimbazi D, et al. (2010) Rapamycin regulates stearoyl CoA desaturase 1 expression – in prostate cancer progression. Nature 457(7231):910 914. in breast cancer. Mol Cancer Ther 9(10):2770–2784. 11. Hu J, et al. (2013) Heterogeneity of tumor-induced gene expression changes in the 41. Menendez JA, Lupu R (2007) Fatty acid synthase and the lipogenic phenotype in – human metabolic network. Nat Biotechnol 31(6):522 529. cancer pathogenesis. Nat Rev Cancer 7(10):763–777. 12. Dang CV (2010) Glutaminolysis: Supplying carbon or nitrogen or both for cancer cells? 42. Kim JW, Gao P, Liu YC, Semenza GL, Dang CV (2007) Hypoxia-inducible factor 1 and – Cell Cycle 9(19):3884 3886. dysregulated c-Myc cooperatively induce vascular endothelial growth factor and 13. Barretina J, et al. (2012) The Cancer Cell Line Encyclopedia enables predictive mod- metabolic switches hexokinase 2 and pyruvate dehydrogenase kinase 1. Mol Cell Biol elling of anticancer drug sensitivity. Nature 483(7391):603–607. 27(21):7381–7393. 14. Garnett MJ, et al. (2012) Systematic identification of genomic markers of drug sen- 43. Yoo YG, Christensen J, Huang LE (2011) HIF-1α confers aggressive malignant traits on sitivity in cancer cells. Nature 483(7391):570–575. human tumor cells independent of its canonical transcriptional function. Cancer Res 15. Ying H, et al. (2012) Oncogenic Kras maintains pancreatic tumors through regulation 71(4):1244–1252. of anabolic glucose metabolism. Cell 149(3):656–670. 44. van Ravenzwaay B, et al. (2007) The use of metabolomics for the discovery of new 16. Son J, et al. (2013) Glutamine supports pancreatic cancer growth through a KRAS- biomarkers of effect. Toxicol Lett 172(1-2):21–28. regulated metabolic pathway. Nature 496(7443):101–105. 45. Shukla K, et al. (2012) Design, synthesis, and pharmacological evaluation of bis-2-(5- 17. Zhang G, et al. (2013) Integration of metabolomics and transcriptomics revealed a phenylacetamido-1,2,4-thiadiazol-2-yl)ethyl sulfide 3 (BPTES) analogs as glutaminase fatty acid network exerting growth inhibitory effects in human pancreatic cancer. inhibitors. J Med Chem 55(23):10551–10563. Clin Cancer Res 19(18):4983–4993. 46. R Development Core Team (2009) R: A Language and Environment for Statistical 18. Commisso C, et al. (2013) Macropinocytosis of protein is an amino acid supply route in Computing (R Foundation for Statistical Computing, Vienna). Ras-transformed cells. Nature 497(7451):633–637. 47. Jiang Z, Gentleman R (2007) Extensions to gene set enrichment. Bioinformatics 23(3): 19. McDermott U, et al. (2007) Identification of genotype-correlated sensitivity to selec- 306–313. tive kinase inhibitors by using high-throughput tumor cell line profiling. Proc Natl 48. Gaujoux R, Seoighe C (2010) A flexible R package for nonnegative matrix factoriza- Acad Sci USA 104(50):19936–19941. tion. BMC Bioinformatics 11:367. 20. Heiser LM, et al. (2012) Subtype and pathway specific responses to anticancer com- 49. Kim PM, Tidor B (2003) Subsystem identification through dimensionality reduction of pounds in breast cancer. Proc Natl Acad Sci USA 109(8):2724–2729. – 21. Brunet JP, Tamayo P, Golub TR, Mesirov JP (2004) Metagenes and molecular pattern large-scale gene expression data. Genome Res 13(7):1706 1718. discovery using matrix factorization. Proc Natl Acad Sci USA 101(12):4164–4169. 50. Wu TD, Nacu S (2010) Fast and SNP-tolerant detection of complex variants and – 22. Collisson EA, et al. (2011) Subtypes of pancreatic ductal adenocarcinoma and their splicing in short reads. Bioinformatics 26(7):873 881. differing responses to therapy. Nat Med 17(4):500–503. 51. Klijn C, et al. (2015) A comprehensive transcriptional portrait of human cancer cell – 23. Cancer Genome Atlas Network (2012) Comprehensive molecular portraits of human lines. Nat Biotechnol 33(3):306 312. breast tumours. Nature 490(7418):61–70. 52. Possemato R, et al. (2011) Functional genomics reveal that the serine synthesis – 24. Cancer Genome Atlas Research Network (2014) Comprehensive molecular charac- pathway is essential in breast cancer. Nature 476(7360):346 350. terization of gastric adenocarcinoma. Nature 513(7517):202–209. 53. Mootha VK, et al. (2003) PGC-1alpha-responsive genes involved in oxidative phos- 25. Mutch DM, et al. (2009) Metabolite profiling identifies candidate markers reflecting phorylation are coordinately downregulated in human diabetes. Nat Genet 34(3): – the clinical adaptations associated with Roux-en-Y gastric bypass surgery. PLoS One 267 273. 4(11):e7905. 54. McCleland ML, et al. (2013) Lactate dehydrogenase B is required for the growth of 26. O’Brien T, et al. (2014) Abstract 964: Inhibiting glycolysis with an LDHA inhibitor: A KRAS-dependent lung adenocarcinomas. Clin Cancer Res 19(4):773–784. new solution to an old problem. Cancer Res 74(19, Suppl):964. 55. Subramanian A, et al. (2005) Gene set enrichment analysis: A knowledge-based 27. Chaudhari AM, Hallman J, Laudeman CP, Musso DL, Parrish CA (2011) Azabenzimi- approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci USA dazoles as fatty acid synthase inhibitors and their preparation and use for the 102(43):15545–15550. treatment of cancer. US Patent WO 2011066211 A1 (June 3, 2011). 56. Cheng WY, Ou Yang TH, Anastassiou D (2013) Biomolecular events in cancer revealed 28. Du X, et al. (2012) FGFR3 stimulates stearoyl CoA desaturase 1 activity to promote by attractor metagenes. PLOS Comput Biol 9(2):e1002920. bladder tumor growth. Cancer Res 72(22):5843–5855. 57. Gray DC, et al. (2007) pHUSH: A single vector system for conditional gene expression. 29. Acharya A, Das I, Chandhok D, Saha T (2010) Redox regulation in cancer: A double- BMC Biotechnol 7:61. edged sword with therapeutic potential. Oxid Med Cell Longev 3(1):23–34. 58. Pinheiro J, Bates D, DebRoy S, Sarkar D, R Development Core Team (2009) nlme: 30. Ashkenazi A, Du X, Qing J (2013) Methods of using scd1 antagonists. US Patent Linear and nonlinear mixed effects models. R Package Version 3.1-96. Available at WO2013056148 A2 (April 18, 2013). CRAN.R-project.org/package=nlme. Accessed July 8, 2015. 31. Thiery JP, Acloque H, Huang RY, Nieto MA (2009) Epithelial-mesenchymal transitions 59. Benjamini Y, Hochberg Y (1995) Controlling the false discovery rate: A practical and in development and disease. Cell 139(5):871–890. powerful approach to multiple testing. J R Stat Soc B 57(1):289–300.

8of8 | www.pnas.org/cgi/doi/10.1073/pnas.1501605112 Daemen et al. Downloaded by guest on October 2, 2021