PHGDH Expression Is Required for Mitochondrial Redox Homeostasis

Published OnlineFirst June 8, 2016; DOI: 10.1158/0008-5472.CAN-16-0530 Cancer Molecular and Cellular Pathobiology Research

PHGDH Expression Is Required for Mitochondrial Redox Homeostasis, Breast Cancer Stem Cell Maintenance, and Lung Metastasis Debangshu Samanta1,2, Youngrok Park1, Shaida A. Andrabi1,3, Laura M. Shelton4, Daniele M. Gilkes1,2,5, and Gregg L. Semenza1,2,5,6

Abstract

Intratumoral hypoxia stimulates enrichment of breast cancer drial redox homeostasis, and increased apoptosis, which abrogat- stem cells (BCSC), which are critical for metastasis and patient ed BCSC enrichment under hypoxic conditions. PHGDH-deficient mortality. Here we report a metabolic adaptation that is required cells exhibited increased oxidant levels and apoptosis, as well as for hypoxia-induced BCSC enrichment and metastasis. Hypoxia- loss of BCSC enrichment, in response to treatment with carbopla- inducible factors coordinately regulate expression of genes encod- tin or doxorubicin. PHGDH-deficient cells were relatively weakly ing phosphoglycerate dehydrogenase (PHGDH) and five down- tumorigenic and tumors that did form were deficient in BCSCs, stream enzymes in theserinesynthesis pathway andmitochondrial abolishing metastatic capacity. Our findings highlight a role for one-carbon (folate) cycle. RNAi-mediated silencing of PHGDH PHGDH in the formation of secondary (recurrent or metastatic) expression in both estrogen receptor–positive and negative breast tumors, with potential implications for therapeutic targeting of cancer cells led to decreased NADPH levels, disturbed mitochon- advanced cancers. Cancer Res; 76(15); 1–13. 2016 AACR.

Introduction increased mitochondrial reactive oxygen species (ROS) production that occurs due to decreased electron transport chain effi- Breast cancer mortality occurs in patients whose cancer cells ciency under hypoxic conditions (16–23). metastasize to distant sites, such as the lungs, bones, and brain. Oncogenic transformation also activates pathways that generate Only a small percentage of the breast cancer cells in a primary ROS and place cancer cells at risk for apoptosis (24). Redox tumor have self-renewal capacity, which is necessary to form a homeostasis is dependent on a balance between levels of oxidants metastatic tumor, and are designated as breast cancer stem cells and antioxidants. The latter are dependent upon the generation of (BCSC) or tumor-initiating cells (1, 2). Compared with bulk NADPH, which is used to maintain glutathione, the principal breast cancer cells, BCSCs exhibit increased survival when treated cellular antioxidant, in a reduced form. Two glycolytic shunt path- with cytotoxic chemotherapy (3, 4), which actively induces the ways utilize glucose metabolites for NADPH generation: the pen- BCSC phenotype (5–7). Intratumoral hypoxia is common in tose phosphate pathway (PPP) diverts glucose-6-phosphate, advanced breast cancers (8) and induces the metastatic (9) and whereas the serine synthesis pathway (SSP) converts 3-phospho- BCSC (10) phenotypes through transcriptional activation of glycerate into serine via three reactions that are catalyzed by target genes by hypoxia-inducible factor 1 (HIF-1) and HIF-2 phosphoglycerate dehydrogenase (PHGDH), phosphoserine ami- (11–15). Adaptation of mammalian cells to chronic hypoxia notransferase 1 (PSAT1), and phosphoserine phosphatase (PSPH). involves a HIF-1–dependent switch from oxidative to glycolytic Serine is utilized as a substrate for one-carbon (folate cycle) metabolism, which is an adaptive response to, and ameliorates, metabolism (1CM), either in the cytosol or mitochondria. In the mitochondria (mito1CM), serine hydroxymethyl transferase 2 (SHMT2) catalyzes the reaction of serine and tetrahydrofolate 1 Institute for Cell Engineering, Johns Hopkins University School of (THF) to glycine and 5,10-methylene-THF (MTHF). MTHF dehy- Medicine, Baltimore, Maryland. 2McKusick-Nathans Institute of Genet- ic Medicine, Johns Hopkins University School of Medicine, Baltimore, drogenase 2 (MTHFD2) catalyzes the reaction of MTHF and þ Maryland. 3Department of Neurology, Johns Hopkins University NADP to generate formyl-MTHF and NADPH. Finally, MTHFD1L 4 School of Medicine, Baltimore, Maryland. Human Metabolome Tech- splits formyl-THF into THF and formate (Fig. 1A). The cytosolic nologies America, Inc., Boston, Massachusetts. 5Sidney Kimmel Com- prehensive Cancer Center, Johns Hopkins University School of Med- (cyto1CM) reactions are catalyzed by SHMT1 and MTHFD1 (which icine, Baltimore, Maryland. 6Departments of Pediatrics, Medicine, performs reactions catalyzed by both MTHFD2 and MTHFD1L). Radiation Oncology, and Biological Chemistry, Johns Hopkins Univer- PHGDH catalyzes the reaction that diverts 3-phosphoglycerate sity School of Medicine, Baltimore, Maryland. from the Embden–Meyerhof pathway (EMP) to the SSP. A short Note: Supplementary data for this article are available at Cancer Research hairpin RNA (shRNA) screen revealed that transformed breast Online (http://cancerres.aacrjournals.org/). cells required PHGDH expression for tumor xenograft formation Corresponding Author: Gregg L. Semenza, Johns Hopkins University School of (25). PHGDH gene amplification was found in 6% of breast Medicine, 733 N. Broadway, Suite 671, Baltimore, MD 21205. Fax: 443-287-5618; cancers and shRNA-mediated knockdown of PHGDH expression E-mail: [email protected] inhibited proliferation of breast cancer cells with PHGDH ampli- doi: 10.1158/0008-5472.CAN-16-0530 fication (25, 26). PHGDH overexpression was observed in 70% of 2016 American Association for Cancer Research. estrogen receptor negative (ER ) breast cancers (25), indicating

www.aacrjournals.org OF1

Samanta et al.

A C Glucose 3-Phosphoglycerate Lactate NAD+ PHGDH NADH PHGDH MTHFD1L MTHFD2 PSPH 3-Phosphohydroxypyruvate SHMT2 MTHFD1 PSAT1 SHMT1 Glutamate PSAT1 α-Ketoglutarate MDA-231 Phosphoserine SUM-149 PSPH HCC-1954

Serine Glycine MCF-7 SHMT2 ZR-75.1 THF MTHF BT-474 GSH NADP+ MTHFD2 GSSG NADPH Hypoxic induction Formyl-THF No hypoxic induction Cytosol Mitochondrion MTHFD1L Figure 1. Hypoxia induces expression of mRNAs Formate encoding SSP and mito1CM enzymes. +THF A, enzymatic reactions. Glucose- derived 3-phosphoglycerate is metabolized to glycine and NADPþ is reduced to NADPH through the B P 20% O2 1% O2 No induction 1% O2 Induction ( < 0.001) activity of SSP (blue) and mito1CM (purple) enzymes. B, mRNA expression. Breast cancer cell lines 30 PHGDH 3.5 PSAT1 2 were exposed to 20% or 1% O for 300 2 3 24 hours and expression of mRNAs 250 1.5 2.5 encoding SSP and 1CM enzymes were 20 200 2 analyzed by RT-qPCR. The expression 1 150 1.5 of each mRNA was quantiﬁed relative 10 to 18S rRNA and then normalized to 100 1 0.5 the result obtained from MDA-MB-231 50 0.5 (MDA-231) cells at 20% O (mean 0 0 2 0 0 SEM; n ¼ 3). C, summary of mRNA expression data (columns) in breast cancer cell lines (rows). Red, signiﬁcantly increased expression 0.8 at 1% as compared to 20% O 12 2 3.5 7 (P < 0.001; Student t test); gray, no PSPH SHMT2 10 0.6 induction at 1% O2. 2.5 5 8 0.4 6 1.5 3 4 0.2 0.5 1 2 0 0 Relative mRNA expression mRNA Relative

60 2 1 12 MTHFD2 MTHFD1L 50 10 1.5 0.8 8 40 0.6 6 30 1 0.4 4 20 0.5 0.2 2 10 0 0 0 0

that a mechanism other than gene ampliﬁcation must underlie 1CM enzymes, is required to maintain redox homeostasis in PHGDH overexpression in most breast cancers. We hypothesized hypoxic breast cancer cells, especially in BCSCs, which are par- that increased expression of PHGDH, as well as other SSP and ticularly sensitive to ROS (27).

OF2 Cancer Res; 76(15) August 1, 2016 Cancer Research

PHGDH Is Required for Breast Cancer Progression

Materials and Methods Metabolite analysis Metabolites in culture media and cells were analyzed by cap- For details, see Supplementary Materials and Methods. illary electrophoresis and single or tandem mass spectrometry (MS/MS) relative to internal standards (Human Metabolome Cell culture Technologies) as described previously (29, 30). MCF-7, MDA-MB-231, HCC-1954, SUM-149, and SUM-159 cells were cultured as described previously (6). BT-474, ZR75.1, and T47D cells were cultured in RPMI1640 with 10% FBS. The cell Bioinformatics lines were obtained from Dr. Sara Sukumar (Johns Hopkins For the HIF signature, the The Cancer Genome Atlas Breast University, Baltimore, MD) in 2012. Cell authentication was Invasive Carcinoma Gene Expression Dataset of 1,215 patients performed by PCR analysis of short tandem repeats. was analyzed (31, 32). Tumor grade was analyzed using GOBO (33). Kaplan–Meier curves were generated using KM plotter (34). Lentivirus transduction Vectors encoding shRNA targeting HIF-1a and HIF-2a, and Results generation of MDA-MB-231 and MCF-7 subclones, were SSP and mito1CM enzyme expression is induced in hypoxic described previously (13, 28). pLKO.1-puro lentiviral vectors breast cancer cells þ encoding shRNA targeting PHGDH (Supplementary Table S1) Breast cancers are classified as ER , progesterone receptor þ were purchased from Sigma-Aldrich. Lentiviruses were packaged positive (PR ), human epidermal growth factor receptor 2 pos- þ and transduced cells were selected as described previously (28). itive (HER2 ), or triple negative (ER /PR /HER2 ). We exposed six representative human breast cancer cell lines to 20% or 1% O2 þ þ þ þ Reverse transcription and quantitative real-time PCR for 24 hours: BT-474 (ER /PR /HER2 ), HCC-1954 (HER2 ), þ þ Total RNA was extracted from cells and tumors using TRIzol MCF-7 (ER /PR ), MDA-MB-231 (ER /PR /HER2 ), SUM-149 þ (Invitrogen) and treated with DNase I (Ambion). cDNA synthesis (ER /PR /HER2 ), and ZR-75.1 (ER ; ref. 35). Total RNA was was performed using the iScript cDNA Synthesis system (Bio-Rad). isolated and analyzed for expression of SSP (PHGDH, PSAT1, qPCR was performed using human-specific primers (Supplemen- PSPH), mito1CM (SHMT2, MTHFD2, MTHFD1L), and cyto1CM tary Table S2) and iQ SYBR Green Supermix (Bio-Rad; ref. 28). (SHMT1, MTHFD1) mRNAs. In MDA-MB-231, HCC-1954, MCF- 7 and BT-474 cells, hypoxic exposure induced the expression of all Immunoblot assays three SSP and all three mito1CM mRNAs, and all six breast cancer Whole-cell lysates were prepared in modified RIPA buffer (14). cell lines exhibited induction of PHGDH and SHMT2 mRNA Blots were probed with antibodies against HIF-1a, PHGDH, (Fig. 1B and C), which encode enzymes catalyzing the first PSAT1, and PSPH (Novus Biologicals). HRP-conjugated anti- reaction of the SSP and mito1CM, respectively (Fig. 1A). In rabbit and anti-mouse secondary antibodies (Santa Cruz Biotech- contrast, expression of cyto1CM mRNAs was induced by hypoxia nology) were used. Blots were reprobed with anti-actin antibody in only one or two cell lines (Fig. 1C). (Santa Cruz Biotechnology). The SSP and PPP represent alternate mechanisms by which glucose metabolites are utilized to generate NADPH. The first BCSC assays enzyme of the PPP is glucose-6-phosphate dehydrogenase Aldefluor and mammosphere assays were performed as (G6PD). In contrast to the SSP and 1CM mRNAs, expression of described previously (13). G6PD mRNA was repressed by hypoxia in all breast cancer lines analyzed (Supplementary Fig. S1). Taken together, these data MitoSOX staining indicate that hypoxia selectively induces the expression of mRNAs Cells were incubated in 5 mmol/L MitoSOX Red (Molecular encoding SSP and mito1CM enzymes in cell lines derived from þ þ þ Probes) in PBS/5% FBS at 37C for 45 minutes and rinsed with ER ,PR , HER2 , and triple-negative breast cancers. PBS. Stained cells were filtered and subjected to flow cytometry. HIFs are required for hypoxic induction of SSP and mito1CM Apoptosis and viability assays enzymes Apoptotic cells were quantified by FITC-Annexin V and APC-7- MDA-MB-231 subclones that were stably transfected with a AAD staining followed by flow cytometry. Viable cells were vector encoding a nontargeting control shRNA (NTC) or shRNA quantified by MTT assay (Invitrogen). targeting HIF-1a (sh1a), HIF-2a (sh2a), or both HIF-1a and HIF- 2a [double knockdown (DKD)] have been used to investigate the Glutathione and NADPH assays role of HIFs in breast cancer progression (6, 13, 28). Hypoxic Cell lysates were analyzed for glutathione and NADPH using induction of PHGDH, PSAT1, PSPH, SHMT2, MTHFD2, and GSH/GSSG-Glo and NADP/NADPH-Glo assays (Promega). MTHFD1L mRNA expression, which was observed in the NTC subclone, was impaired when HIF-1a or HIF-2a or both were Glucose uptake assay silenced (Fig. 2A). Immunoblot assays demonstrated hypoxic Cells were incubated in 150 mmol/L 2-[N-(7-nitrobenz-2-oxa- induction of PHGDH, PSAT1, and PSPH protein expression in the 1,3-diazol-4-yl) amino]-2-deoxy-D-glucose (Molecular Probes) NTC subclone, which was impaired in the knockdown subclones and subjected to flow cytometry. (Fig. 2B). Similar results were obtained in MCF-7 subclones (Sup- plementary Fig. S2A). Hypoxia-induced PHGDH, PSAT1, and Seahorse assays PSPH expression in parental MCF-7 cells was abrogated, in a Oxygen consumption and extracellular acidification were mea- dose-dependent manner, by treatment with acriflavine (Supple- sured using the XF24-Analyzer (Seahorse Bioscience). mentary Fig. S2B), which is a drug that inhibits the

www.aacrjournals.org Cancer Res; 76(15) August 1, 2016 OF3

Samanta et al.

MDA-MB-231

20% O2 1% O2 A PHGDH 5 * 5 * SHMT2 4 4 3 # 3 # # # 2 # 2 # 1 1 * * 0 0 a a NTC sh1a sh2a DKD NTC sh1 sh2 DKD

8 PSAT1 4 * MTHFD2 * 3 6 Figure 2. SSP and mito1CM expression is HIF- 4 2 # dependent and increased in BCSCs. A, # # # analysis of mRNA expression in MDA- 2 # 1 * # * * MB-231 subclones, which expressed a 0 0 non-targeting control shRNA (NTC) or NTC sh1a sh2a DKD NTC sh1a sh2a DKD shRNA targeting HIF-1a (sh1a), HIF- Relative mRNA expression mRNA Relative Relative mRNA expression mRNA Relative 2a (sh2a), or both HIF-1a and HIF-2a * (DKD), and were exposed to 20% or 1% 4 2.5 * PSPH MTHFD1L O2 for 24 hours. Data were normalized 2 to NTC at 20% O (mean SEM; n ¼ 3 # 2 # P < 1.5 # 3). , 0.01 versus NTC at 20% O2; # P < 2 # , 0.001 versus NTC at 1% O2. 1 * * B, immunoblot assays of lysates # # 1 0.5 * prepared from MDA-MB-231 subclones, which were exposed to 0 0 ﬂ a a NTC sh1a sh2a DKD 20% or 1% O2 for 48 hours. C, Alde uor NTC sh1 sh2 DKD assay of MDA-MB-231 subclones

exposed to 20% or 1% O2 for 72 hours. The percentage of cells expressing B NTC sh1a1sh1a10 sh2a1 DKD C aldehyde dehydrogenase (ALDHþ) was determined (mean SEM; n ¼ 3). O2 (%) 20 1 20 1 20 1 20 1 20 1 20% O2 1% O2 , P < 0.05 versus NTC at 20% O2; 5 #, P < 0.01 versus NTC at 1% HIF-1α * 4 O2. D, analysis of gene expression PHGDH in adherent monolayers and Cells 3 mammospheres. MDA-MB-231 cells + PSAT1 2 were cultured on standard or ultra-low adherence plates for 7 days in # 1 # 20% O2 and adherent cells and PSPH % ALDH * * # mammospheres, respectively, were 0 Actin a a harvested for RT-qPCR analyses. NTC sh1 sh2 DKD Results were normalized to adherent cells (mean SEM; n ¼ 3). , P < 0.001 versus adherent cells D (Student t test). Adherent Mammospheres 5 12 *

4 * * 10 8 3 * * 6 2 * * 4 1 * 2 * * *

Relative mRNA expression mRNA Relative 0 0 a a

OF4 Cancer Res; 76(15) August 1, 2016 Cancer Research

PHGDH Is Required for Breast Cancer Progression

heterodimerization of HIF-a and HIF-1b subunits (36). Thus, esis, we exposed MDA-MB-231 and MCF-7 subclones to20% or 1% genetic and pharmacologic approaches indicate that HIFs coordi- O2 for 72 hours in adherent culture and stained the cells with nately regulate the expression of SSP and mito1CM enzymes when MitoSOX Red, which is selectively targeted to mitochondria and breast cancer cells are exposed to hypoxia. generates fluorescence when oxidized by superoxide radicals, there- BCSCs are characterized by high aldehyde dehydrogenase by serving as an indicator of mitochondrial ROS in live cells. The (ALDH) activity and can be identified by the Aldefluor assay, in NTC subclones showed no increase in ROS after hypoxic exposure, þ which BODIPY-aminoacetaldehyde is converted to the fluores- whereas the percentage of MitoSOX cells was significantly cent product BODIPY-aminoacetate (37). Exposure of NTC sub- increased in PHGDH knockdown subclones (Fig. 4A). Analysis of clones of MDA-MB-231 (Fig. 2C) and MCF-7 (Supplementary Fig. Annexin V and 7-amino-actinomycin D (7-AAD) staining revealed þ S2C) to hypoxia for 72 hours increased the percentage of ALDH no increase in apoptosis of hypoxic NTC cells, whereas the per- þ BCSCs, whereas this response was impaired in knockdown sub- centage of Annexin V /7-AAD cells was increased in the PHGDH clones. Treatment of MCF-7 cells with acriflavine also blocked knockdown subclones (Fig. 4B). Exposure of cells to hypoxia in the hypoxic induction of the BCSC phenotype as determined by the presence of manganese (III) tetrakis (1-methyl-4-pyridyl) porphy- mammosphere assay (Supplementary Fig. S2D), which is based rin (MnTMPyP), a cell-permeable superoxide scavenger (39), res- on the selective ability of BCSCs to generate multicellular spher- cued the apoptosis of PHGDH knockdown subclones under hyp- oids under nonadherent culture conditions (38). oxia (Fig. 4C), indicating that increased apoptosis was due to increased ROS levels. Expression of SSP and mito1CM mRNAs is increased in BCSCs We hypothesized that PHGDH was required under hypoxic The preceding results demonstrated a correlation between loss conditions for NADPH generation to maintain glutathione in a of hypoxia-induced SSP and mito1CM expression and loss of reduced state. Exposure of NTC subclones to hypoxia increased the hypoxia-induced BCSC enrichment. To determine whether ratio of reduced to oxidized glutathione (Fig. 4D), which was mRNAs encoding these enzymes were overexpressed in BCSCs associated with a modest decrease in NADPH levels (Fig. 4E). In relative to non-BCSCs, we cultured MDA-MB-231 and MCF-7 cells contrast, PHGDH knockdown was associated with an impaired as either adherent monolayers or mammospheres for 7 days. hypoxic induction of reduced glutathione (Fig. 4D) and a signif- HIF1a, HIF-2a, PHGDH, PSAT1, SHMT2, MTHFD2, and icant decrease in NADPH levels (Fig. 4E). Thus, PHGDH deficiency MTHFD1L mRNA expression was increased in BCSC-enriched impairs NADPH production, which becomes a liability specifically mammosphere cultures of MDA-MB-231 cells, whereas expres- under hypoxic conditions. sion of the cyto1CM enzymes SHMT1 and MTHFD2 was decreased in mammospheres relative to adherent cells, as was PHGDH plays a major role in determining the utilization of the PPP enzyme G6PD (Fig. 2D). Increased expression of HIF-1a, glucose metabolites HIF-2a, PHGDH, SHMT2, MTHFD2, and MTHFD1L mRNA in Glucose metabolism via the EMP leads to the production of BCSCs relative to non-BCSCs was also observed in MCF-7 cells acetyl CoA, which is utilized for ATP generation through oxidative (Supplementary Fig. S2E). Thus, HIF, SSP, and mito1CM mRNAs phosphorylation, and lactic acid, which is the terminal product of are preferentially expressed in BCSCs, suggesting that they play an glycolysis. PHGDH diverts glucose metabolites to the SSP, thereby important role in the BCSC phenotype. reducing production of both acetyl CoA and lactic acid. We analyzed the O2 consumption rate (OCR) and extracellular acid- PHGDH knockdown abrogates hypoxia-induced BCSC ification rate (ECAR) to monitor oxidative phosphorylation and enrichment glycolysis, respectively, in NTC and PHGDH knockdown sub- We chose to analyze the effect of PHGDH loss-of-function in clones. PHGDH deficiency increased the OCR (Fig. 5A) and ECAR breast cancer cells for three reasons: (i) PHGDH is required for the (Fig. 5B) in both MDA-MB-231 and MCF-7 cells. Glucose uptake diversion of glucose metabolites to the SSP and 1CM; (ii) PHGDH was not increased in PHGDH knockdown subclones (Fig. 5C). expression was hypoxia-inducible in all breast cancer lines ana- Taken together, these results indicate that the increased OCR and lyzed; and (iii) PHGDH was preferentially expressed in BCSCs. ECAR in PHGDH knockdown subclones are due to decreased MDA-MB-231 and MCF-7 cells were stably transfected with an shunting of glucose metabolites from the EMP to the SSP. expression vector encoding either of two independent shRNAs The effect of PHGDH knockdown on metabolite levels in MDA- targeting PHGDH (designated sh2 and sh4). Knockdown efficiency MB-231 cells was analyzed by MS. PHGDH knockdown was was validated at the mRNA (Fig. 3A) and protein (Fig. 3B) levels in associated with increased extracellular (Fig. 5D) and intracellular both cell lines. PHGDH knockdown did not impair proliferation of (Fig. 5F) lactic acid levels under both nonhypoxic and hypoxic either MDA-MB-231 (Supplementary Fig. S3A) or MCF-7 (Supple- conditions, which was consistent with the increased ECAR mentary Fig. S3B) cells cultured for 72 hours at either 20% or (Fig. 5B). Hypoxia increased extracellular serine levels in the NTC 1% O2. In contrast, PHGDH knockdown markedly impaired the subclone, whereas PHGDH deficiency reduced serine levels under hypoxia-induced enrichment of BCSCs as determined by Aldefluor both nonhypoxic and hypoxic conditions (Fig. 5E). Thus, serine assays (Fig. 3C) or primary and secondary mammosphere assays synthesis increases in NTC cells under hypoxic conditions as a result (Fig. 3D). These results indicate that PHGDH expression is specif- of increased PHGDH expression, leading to decreased import and/ ically required for hypoxic induction of the BCSC phenotype. or increased export of serine. Hypoxia or PHGDH knockdown increased intracellular levels of all of the metabolites in the EMP at PHGDH is required to maintain redox homeostasis and or downstream of the SSP branch point: 3-phosphoglyceric acid survival of hypoxic breast cancer cells (Fig. 5G), 2-phosphoglyceric acid (Fig. 5H), phosphoenolpyruvic Acute hypoxia leads to increased mitochondrial ROS generation acid (Fig. 5I), and pyruvic acid (Fig. 5J). Hypoxia decreased intra- (23). We hypothesized that PHGDH deficiency would lead to cellular levels of 6-phosphogluconic acid (Fig. 5K), the G6PD increased ROS levels and increased apoptosis. To test this hypoth- reaction product, which is consistent with decreased G6PD

www.aacrjournals.org Cancer Res; 76(15) August 1, 2016 OF5

Samanta et al.

MDA-MB-231 MCF-7

A 1.2 1.2 1 1 0.8 0.8 0.6 0.6 Figure 3. 0.4 0.4 Decreased PHGDH expression and 0.2 hypoxia-induced BCSC enrichment in

0.2 PHGDH Relative Relative PHGDH Relative mRNA expression mRNA mRNA expression mRNA * * * * knockdown subclones. A and B, 0 0 analysis of PHGDH expression. NTC sh2 sh4 NTC sh2 sh4 Subclones of MDA-MB-231 (left) and MCF-7 (right) expressing NTC shRNA or either of two different shRNAs B NTC sh2 sh4 NTC sh2 sh4 targeting PHGDH (sh2 and sh4) were exposed to 20% or 1% O for 24 (A)or O (%) 20 1 20 1 20 1 2 O2 (%) 20 1 20 1 20 1 2 48 (B) hours and analyzed for PHGDH PHGDH expression of PHGDH mRNA by RT- qPCR assay (A) and PHGDH protein by Actin Actin immunoblot assay (B). RNA data were normalized to NTC (mean SEM; n ¼ 3). , P < 0.001 versus NTC. C, Aldeﬂuor assay. Subclones were 7 9 C * * exposed to 20% or 1% O2 for 72 hours þ 6 8 and the percentage of ALDH cells 7 was determined by ﬂow cytometry 5 n ¼ P < Cells 6 (mean SEM; 3). , 0.01; Cells + 4 P < + 5 , 0.001 versus NTC at 20% O2; # P < 3 4 , 0.001 versus NTC at 1% O2. 3 D, mammosphere assays. Subclones 2 # 2 # were exposed to 20% or 1% O2 for % ALDH

% ALDH # 1 # 72 hours, transferred to ultra-low ** 1 ** ** 0 0 attachment plates, and 7 days later the NTC sh2 sh4 NTC sh2 sh4 number of primary mammospheres per ﬁeld was counted (mean SEM; 20% O2 n ¼ 3). Primary mammospheres were 1% O2 collected, dissociated, transferred to ultra-low attachment plates, and secondary mammospheres were D Primary Secondary Primary Secondary counted 7 days later (mean SEM; n ¼ P < P < 80 30 ** 60 25 3). , 0.01; , 0.001 versus ** # P < ** NTC at 20% O2; , 0.001 versus ** 25 50 20 60 NTC at 1% O2. 20 40 15 40 15 30 # # # 10 10 20 # # 20 # # # 5

Mammospheres 5 10 * * * 0 0 0 0 NTC sh2 sh4 NTC sh2 sh4 NTC sh2 sh4 NTC sh2 sh4

expression in hypoxic cells (Supplementary Fig. S1). In contrast, (5–7). Carboplatin is a chemotherapy agent that is used to treat PHGDH knockdown did not affect 6-phosphogluconic acid levels, breast cancer (41, 42). Platinum compounds increase mitochon- as expected, because the PPP shunt is upstream of the SSP shunt in drial ROS by forming adducts on mitochondrial DNA, thereby the EMP (Fig. 5L). Taken together, the data presented in Fig. 5 impairing the transcription of electron transport chain compo- indicate that PHGDH shunts a significant proportion of glucose- nents (43). MDA-MB-231 and MCF-7 subclones were exposed to derived 3-phosphoglycerate from the EMP to the SSP in breast increasing concentrations of carboplatin for 72 hours and cell cancer cells. viability was determined by MTT assay. PHGDH deficiency sensitized breast cancer cells to carboplatin (Supplementary Fig. S4). PHGDH knockdown increases the sensitivity of breast cancer Compared with NTC subclones, treatment of PHGDH-deficient cells to chemotherapy cells with carboplatin at IC50 led to increased mitochondrial ROS As many cytotoxic cancer chemotherapies increase ROS levels (Fig. 6A) and apoptosis (Fig. 6B). Carboplatin treatment of NTC þ (40), we hypothesized that PHGDH deficiency would also impair subclones induced enrichment of ALDH BCSCs, which was the enrichment of BCSCs that occurs in response to chemotherapy abrogated in the knockdown subclones (Fig. 6C).

OF6 Cancer Res; 76(15) August 1, 2016 Cancer Research

Downloaded from cancerres.aacrjournals.org on September 29, 2021. © 2016 American Association for Cancer Research. www.aacrjournals.org T t1 O 1% at NTC T t1 O 1% at NTC rasneo 50 of absence or h ucoe eeepsdt 0 or O 20% 1% to exposed were subclones The a eemnd(mean determined was P O 20% at NTC to normalized and measured ( ratio essNCa 0 O 20% at NTC versus a eemndby determined was Red MitoSOX cells for of positive percentage the and hours 72 for O 1% or 20% to exposed were subclones (right) MCF-7 and (left) MDA-MB-231 production. ROS mitochondrial of survival. cell redox and homeostasis on knockdown PHGDH of Effect 4. Figure ## # xoe o2%o %O 1% or 20% to exposed were subclones The scavenger. ROS .0 essNCa %O 1% at NTC versus 0.001 eemnd(mean determined nei V Annexin 2husadtepretg fAnnexin V of percentage the and hours 72 r1 O 1% or D (mean , þ 2 and , < , P /7-AAD P (mean P .0 essNCa 0 O 20% at NTC versus 0.001 < 2 < < o 2husadtepretg of percentage the and hours 72 for E, D .0 essNCa %O 1% at NTC versus 0.001 2 .0 essNCa 0 O 20% at NTC versus 0.001 .0 essn MnTMPyP. no versus 0.001 n AP ees( levels NADPH and ) ucoe eeepsdt 20% to exposed were subclones o 2husadteGSH/GSSG the and hours 72 for SEM; þ Downloaded from n 7-AAD and 2 2 ppoi el was cells apoptotic SEM; . . n B, ¼ nlsso apoptosis. of analysis m 3). n o/ nMy for MnTMPyP mol/L ﬂ 2 ¼ ; wcytometry ow # SEM; # , 3). , 2 P P ntepresence the in ppoi cells apoptotic SEM; 2 < < . ﬂ .0 versus 0.001 .0 versus 0.001 , C, n uorescence P Published OnlineFirstJune8,2016;DOI:10.1158/0008-5472.CAN-16-0530 A, E ¼ eceby rescue n < were ) analysis 2 3). ¼ 2 0.01 ; ; # 3). cancerres.aacrjournals.org 2 P ; < 2 , MnTMPyP E D C B A

NADPH GSH/GSSG 100 200 300 400 500 0.5 1.5 % Apoptotic cells % Apoptotic cells % MitoSox+ cells 0 1 0 10 15 0 2 4 6 8 0 1 2 3 4 5 6 7 8 9 0 5 - T h sh4 sh2 NTC T sh2 NTC T h sh4 sh2 NTC T h sh4 sh2 NTC T h sh4 sh2 NTC MDA-MB-231 Cancer Research. * * - on September 29, 2021. © 2016American Association for - ** * * 50 # # # # # - * * # ## 50 # # # HD sRqie o ratCne Progression Cancer Breast for Required Is PHGDH 20% O MnTMPyP 1% O NADPH GSH/GSSG

100 + 0.5 1.5

% Apoptotic cells % Apoptotic cells 2 20 40 60 80 % MitoSox cells 2 0 0 1 0 1 2 3 4 5 6 7 8 9 0 2 4 6 8 0 2 4 6 8 acrRs 61)Ags ,2016 1, August 76(15) Res; Cancer - T h sh4 sh2 NTC T h sh4 sh2 NTC T sh4 NTC T h sh4 sh2 NTC T h sh4 sh2 NTC * - MCF-7 ** - * ** # # 50 # # # - ** * # # ## # 50 # OF7 OF8 acrRs 61)Ags ,2016 1, August 76(15) Res; Cancer aat tal. et Samanta D ABC K HI FG OCR Concentration Concentration (pmol/min/mg) Concentration Concentration (pmol/106 cells) 6 (pmol/10 cells) 0.05 0.15 120,000 150,000 180,000 0.1 0.2

6 10,000 20,000 6 30,000 60,000 90,000 (pmol/10 cells) 6-Phosphogluconic acid (pmol/10 cells) 0 Downloaded from 100 200 300 400 2-Phosphoglyceric acid 20 40 60 0 0 0 0 * 0 1% 20% 0 1% 20% 20% 0 1% 20% Lactic acid * * * * Lactic acid (2PG) (Lac) * * * ** Published OnlineFirstJune8,2016;DOI:10.1158/0008-5472.CAN-16-0530 (6PG) * # 1% # # * cancerres.aacrjournals.org # # #

* ECAR (mpH/min/mg) 0 1 2 3 4 5 6 7 Glc L E Phosphoenolpyruvic acid 100 150 200 250 * 50 3-Phosphoglyceric acid 200 400 0 100 200 300 400 500 600 * 0 0 0 1% 20% G6P 3PG 3PG G6P 6PG 6PG PPP * ** 0 1% 20% 20% (PEP) * * (3PG) * Serine * * Cancer Research. 5 Glucose uptake * 1,000 1,200 1,400 on September 29, 2021. © 2016American Association for # SSP Ser Ser 200 400 600 800 1% ## # 0 # 3 # 2PG 2PG 1,000 2,000 3,000 4,000 5,000 J 0 Pyruvic acid Extracellular 0 1% 20% PEP Intracellular * MDA-231 MDA-231 (Pyr) * Media sh4 sh2 NTC sh4 sh2 NTC MDA231 sh2 MDA231 NTC MCF7 sh4 MCF7 sh2 MCF7 NTC AcCoA # Lac Lac Pyr # o2%o %O 1% or 20% to exposed were subclones MDA-MB-231 ( ( extracellular of concentrations absolute eemndby determined mean the ucoe utrdudr2%O 20% under cultured Subclones C, acidi extracellular and (OCR) rate consumption oxygen knockdown. PHGDH of consequences Metabolic 5. Figure -l amino]-2-deoxy- 4-yl) 2-[N-(7-nitrobenz-2-oxa-1,3-diazol- 150 with stained were O 0 O 20% P n Ser. and 3PG between reactions three the and 3PG, and (G6P) glucose-6-phosphate omits the that form abbreviated in shown are (SSP) pathway synthesis serine only (PPP; pathway phosphate pentose pathway, Embden pathways. SEM; (mean standards reference using MS ( ( OCR nuae t2%O 20% at incubated subclones MDA-MB-231 and MCF-7 in F B , 2 , – P SEM; mean ; . nlsso lcs uptake. glucose of analysis P K L, fi < eaoie eedtrie by determined were metabolites ) < eezmtcratosbetween reactions enzymatic ve n A .0 essMF7NTC; MCF-7 versus 0.001 lcs Gc metabolic (Glc) glucose 2 .0 essMAM-3 NTC. MDA-MB-231 versus 0.001 ¼ mean ; n ; # ¼ 3). , fl fi P SEM; 3). oecneitniywas intensity uorescence s ecini hw) and shown), is reaction rst A < D , D P and .0 essNCa 1% at NTC versus 0.001 2 fl and – < SEM; n wctmty(mean cytometry ow o 2husadthe and hours 72 for K, fi ¼ .0 essNCat NTC versus 0.001 acrResearch Cancer ainrt (ECAR). rate cation B, – eaooi data. metabolomic E 2 eehf(main) Meyerhof )wr measured were 3) n intracellular and ) o 2hours. 72 for esrmn of measurement n D m ¼ guoeand -glucose mol/L )adECAR and 3) 2 Published OnlineFirst June 8, 2016; DOI: 10.1158/0008-5472.CAN-16-0530

PHGDH Is Required for Breast Cancer Progression

Vehicle MDA-MB-231 MCF-7 Carboplatin # # A 12 In vitro 10 10 8 Cells 8 Cells + * + 6 * 6 4 4 MitoSox 2 MitoSox 2 % % 0 0 NTC sh2 NTC sh4

B 10 # 10 #

8 8 * 6 6

4 4

2 2 % Apoptotic cells Apoptotic % 0 cells Apoptotic % 0 Figure 6. NTC sh2 NTC sh4 Effect of PHGDH knockdown on the response to chemotherapy. A–C, response to chemotherapy in vitro. MDA-MB-231 (left) and MCF-7 (right) subclones * were exposed to vehicle or carboplatin for 72 hours at 3 15 * m m C IC50 (75 mol/L for MDA-MB-231 and 200 mol/L for 2.5 MCF-7) and the percentage of MitoSOXþ (A), apoptotic (B), þ Cells and ALDH (C) cells was determined (mean SEM; 2 Cells 10 +

# + n ¼ 3). , P < 0.01 versus NTC at 20% O2; , P < 0.001 vs. NTC 1.5 – in vivo at 1% O2. D G, response to chemotherapy . MDA-MB- # 5 231 subclones were implanted into the mammary fat pad of 1 # female NSG mice. When tumor volume reached 200 mm3 0.5 % ALDH (day 0), the mice were randomly assigned to receive % ALDH 0 0 intraperitoneal injections of saline (Vehicle) or carboplatin NTC sh2 NTC sh4 (15 mg/kg) on days 0, 5, and 10. Tumors were harvested on day 13, and samples were analyzed for the percentage of MitoSOXþ (D), apoptotic (E), and ALDHþ (F) cells and the number of mammosphere-forming cells (G). MDA-231 Data are presented as mean SEM (n ¼ 3). , P < 0.01; , P < 0.05, versus vehicle-treated NTC; #, P < 0.01 versus Vehicle carboplatin-treated NTC. D E # Carboplatin 15 20 # In vivo

Cells 15

+ 10 * 10 5 5 * ** MitoSox %

0 cells Apoptotic % 0 NTC sh2 NTC sh2

F * G 25 120 *

20 100 80 Cells 15 + 60 10 40 5 # 20 #

** Mammospheres % ALDH 0 0 NTC sh2 NTC sh2

www.aacrjournals.org Cancer Res; 76(15) August 1, 2016 OF9

Samanta et al.

We also investigated the effect of PHGDH knockdown on the formed tumors in only 3 out of 7 mice (Fig. 7G). PHGDH mRNA response to doxorubicin, which is an anthracycline that is levels in knockdown tumors were significantly less than in NTC commonly used in the United States to treat breast cancer and tumors (Fig. 7H). was the first chemotherapeutic shown to induce HIF activity (44). Compared with NTC subclones, treatment of PHGDH- Analysis of SSP and 1CM expression in primary human breast deficient cells with doxorubicin at IC increased mitochondrial 50 cancers ROS (Supplementary Fig. S5A) and apoptosis (Supplementary To investigate the clinical relevance of our experimental find- Fig. S5B). Doxorubicin treatment induced enrichment of þ ings, we mined gene expression databases. We first investigated ALDH BCSCs, which was abrogated in the PHGDH-knock- whether expression of mRNA encoding SSP and mito1CM down subclones (Supplementary Fig. S5C). Thus, both plati- enzymes, either individually or in aggregate (designated S1C), num and anthracycline chemotherapy induce PHGDH-depen- was correlated with the HIF signature, which comprised expres- dent BCSC enrichment. sion of HIF-1a mRNA and 13 HIF target-gene mRNAs (PLOD1, To investigate whether carboplatin-induced BCSC enrichment VEGFA, LOX, P4HA2, NDRG1, SLC2A1, ERO1L, ADM, LDHA, in vivo also requires PHGDH expression, MDA-MB-231 subclones PGK1, ANGPTL4, SLC2A3, and CA9), in 1,215 breast cancer (2 106 cells) were implanted in the mammary fat pad (MFP) of specimens (31, 32) using Pearson correlation test. For each of female NSG mice. When tumors reached a volume of 200 mm3, the 6 mRNAs encoding an SSP or mito1CM enzyme, expression mice were treated with 15 mg/kg of carboplatin by intraperitoneal was significantly correlated with the HIF signature (P < 0.0001 in injection every 5 days for 3 doses. Tumors were harvested 3 days each case; Fig. 7I). These results are consistent with the data after the last dose for analysis. Compared with NTC subclones, obtained from breast cancer cell lines demonstrating that hypox- PHGDH knockdown subclones exhibited a greater increase in the þ ia-induced expression of these genes is HIF-dependent (Fig. 1 percentage of MitoSOX (Fig. 6D) and apoptotic (Fig. 6E) cells. þ and 2). Analysis of the GOBO database (33) revealed that S1C Carboplatin treatment increased the percentage of ALDH cells gene expression increased significantly (P < 0.00001) with increas- (Fig. 6F) and number of mammosphere-forming cells (Fig. 6G) in ing tumor grade (Supplementary Fig. S6A). Analysis of PHGDH or the NTC subclone, and these effects were abrogated by PHGDH S1C mRNA expression in > 3,500 human breast cancer specimens knockdown (Fig. 6F and G). These results indicate that PHGDH using KM Plotter (34) revealed that levels greater than the median deficiency sensitizes breast cancer cells to chemotherapy and were associated with decreased relapse-free survival (HR ¼ 1.34 abrogates chemotherapy-induced BCSC enrichment. and P < 10 7 for PHGDH; HR ¼ 1.74 and P < 10 16 for S1C; Fig. 7J). Greater-than-median expression of SHMT2 or MTHFD2, but PHGDH expression promotes tumor initiation and is required not PSAT1 or PSPH, was also associated with a significant decrease for breast cancer metastasis in relapse-free survival (Supplementary Fig. S6B). Thus, expres- To investigate whether PHGDH regulates other aspects of sion of SSP and mito1CM mRNAs in primary breast cancers is breast cancer progression, MDA-MB-231 NTC and PHGDH HIF-regulated and predictive of patient mortality. knockdown subclones were implanted in the MFP of NSG mice and the resulting tumors were harvested on day 35. PHGDH knockdown was associated with a significant increase in tumor Discussion mass after orthotopic transplantation of sh2 and sh4 cells Recent studies have attempted to determine the mechanisms (Fig. 7A), which was consistent with cell culture data demonstrat- and consequences of PHGDH enzyme expression in breast cancer, ing increased growth of the sh2 and sh4 (as well as sh3 and sh5) but these studies have focused on cell proliferation in vitro or PHGDH-knockdown subclones in vitro (Supplementary Fig. S3A). primary tumor growth in vivo (25, 26). Here, we demonstrate that The tumors analyzed in Fig. 7A were derived from the injection of hypoxia induces expression of PHGDH and other SSP and 2 106 cells; under these conditions, BCSCs are not limiting for mito1CM enzymes that is mediated by HIF-1 and HIF-2. This primary tumor formation. PHGDH knockdown was associated coordinate regulation of multiple genes provides a mechanism to þ with an increased percentage of MitoSOX cells (Fig. 7B) and increase flux through the pathway. The expression of at least four decreased BCSCs as determined by Aldefluor (Fig. 7C) and out of six genes encoding SSP/mito1CM enzymes was induced by mammosphere (Fig. 7D) assays. IHC revealed PHGDH expres- hypoxia in each of the cell lines analyzed, which included repre- þ þ sion in perinecrotic (hypoxic) regions of NTC tumors but not in sentative lines derived from ER , HER2 , and triple-negative knockdown tumors (Fig. 7E). breast cancers. PHGDH and SHMT2 expression was hypoxia- To determine whether PHGDH is required for spontaneous induced in all six breast cancer lines and was increased in metastasis from breast to lungs, MDA-MB-231 subclones (2 106 BCSC-enriched cell populations. In addition, increased expres- cells) were again injected into the MFP of female NSG mice. To sion in primary breast cancers of mRNA encoding PHGDH, take into account differences in primary tumor growth, tumors SHMT2, MTHFD2, or all six SSP/mito1CM enzymes was associ- and lungs were isolated when primary tumors reached a volume ated with HIF target-gene expression and an increased risk of of 1000 mm3. Despite growing faster, PHGDH knockdown patient mortality. Regulation of PHGDH, PSAT1, PSPH, SHMT1, tumors did not generate lung metastases, whereas in mice bearing and SHMT2 expression by the transcription factors NRF2 and NTC tumors, metastatic cells occupied large areas of the lung ATF4 in lung cancer and MYC in liver cancer was recently reported parenchyma (Fig. 7F). (45, 46). Further studies are required to determine whether ATF4 To investigate whether PHGDH promotes the tumor-initiating or MYC regulates basal expression of these genes in breast cancer. potential of breast cancer cells, we injected only 1 103 cells into We studied the consequences of PHGDH deficiency because of the MFP of female SCID mice, so that BCSCs would be limiting for its role as the first enzyme in the SSP that is responsible for tumor formation. NTC cells formed tumors by day 71 after diverting glucose metabolites from the EMP and because the injection in 7 of 7 injected mice, whereas sh2 and sh4 cells each finding of PHGDH gene amplification in 6% of primary breast

OF10 Cancer Res; 76(15) August 1, 2016 Cancer Research

PHGDH Is Required for Breast Cancer Progression

ABCD 4 4 50 1.2 * * 1 * 40 3 3 Cells 0.8 Cells + 30 + 0.6 2 2 * * 20 0.4 * 1 1 10 *

0.2 % ALDH Mammospheres Tumor weight weight (g) Tumor %MitoSox Figure 7. 0 0 0 0 Effect of PHGDH knockdown on orthotopic tumor initiation and metastasis, and clinical correlates. E NTC sh2 A–F, analysis of primary tumors and metastases. MDA-MB-231 subclones (2 106 cells) were implanted in the MFP of female NSG mice. After 35 days, the tumors and lungs were harvested. N N Tumor mass was determined (A)and tumor tissue was dissociated for þ þ analysis of MitoSOX (B), ALDH (C), M1 M2 M3 and mammosphere-forming (D) cells F (mean SEM, n ¼ 3). , P < 0.01 versus NTC. Tumor tissue was sectioned for PHGDH IHC (E; scale bar, 200 mm; N, necrosis). Lung sections (2 2 mm) NTC were stained with hematoxylin and eosin to identify metastasis in three mice (M1, M2, M3), each bearing 1,000- mm3 tumors derived from NTC or PHGDH knockdown cells (F). G and H, analysis of tumor-initiating capacity.

MDA-MB-231 subclones were implanted sh2 in the MFP of female SCID mice (1 103 cells per mouse, 7 mice per subclone). Mice were scored for palpable tumors after 71 days. , P < 0.05 vs. NTC, c2 test (G). RNA was extracted from those

tumors that formed, PHGDH mRNA sh4 levels were determined, and results were normalized to NTC (mean SEM; n ¼ 3); , P < 0.01 versus NTC (H). I, analysis of gene expression data from human primary breast cancers. Pearson G 100% H 5 correlation test was performed to Tumor compare expression of the HIF + 3 signature with expression of the six 50% - SSP/mito1CM mRNAs in aggregate ** (S1C) or individually using data from 1 * *

1,215 breast cancer samples. , P < Mice with tumor 0%

PHGDH mRNA NTC sh2 sh4 0.0001, two-tailed t test. J, correlation NTC sh2 sh4 –1 of gene expression with patient mortality. Kaplan–Meier analyses of 0.6 * * * relapse-free survival over ten years I * were performed using data for 3,554 * * breast cancer patients, who were 0.3 *

ﬁ HIF with strati ed by PHGDH (left) or S1C (right) Pearson signature correlation mRNA expression levels in the primary 0 tumor, which were greater (red) or less (black) than the median level. The hazard ratio (HR) and P value (log-rank J 1.0 test) for each comparison are shown. 1.0 0.8 0.8 0.6 0.6 HR = 1.34 HR = 1.74 P -07 - = 7.4×10 P < 1×10 16 probability PHGDH expression S1C expression 0.2 0.4 High = 1,774 0.2 0.4 Relapse-free survival survival Relapse-free High = 1,777 Low = 1,780 Low = 1,777 0.0 0.0 0 20 40 60 80 100 120 0 20 40 60 80 100 120 Time (months) Time (months)

www.aacrjournals.org Cancer Res; 76(15) August 1, 2016 OF11

Samanta et al.

cancers suggested a critical role for this enzyme in breast cancer included impaired capacity for tumor initiation and lung metas- progression. Our data suggest that a major determinant of tasis. We interpret the previously reported failure of PHGDH increased PHGDH expression in breast cancers without gene knockdown cells to form tumor xenografts (25) as due to a amplification is intratumoral hypoxia, which is a common find- deficiency of BCSCs rather than an effect on cell proliferation, ing in advanced breast cancer (8). The correlation between which we did not observe in vitro or in vivo. Similarly, the loss of expression of PHGDH and HIF target genes in primary breast metastatic capacity we demonstrated is likely due, at least in part, cancers strongly supports this conclusion. Similar to PHGDH, to a deficiency of BCSCs (49, 50). þ SHMT2 expression was induced by hypoxia in all breast cancer PHGDH knockdown sensitized both ER and ER breast lines studied and SHMT2 overexpression in breast cancers was cancer lines to chemotherapy, with increased mitochondrial ROS, associated with patient mortality, suggesting that SHMT2 is also increased apoptosis, and loss of chemotherapy-induced BCSC important for breast cancer progression. The expression of SHMT2 enrichment. These findings suggest that combining chemothera- is induced by hypoxia in neuroblastoma cells and SHMT2 knock- py with an inhibitor of PHGDH may improve the survival of down was shown to increase ROS levels and cell death under women with advanced breast cancer by blocking counter-thera- hypoxic conditions, but effects on cancer stem cells and metastasis peutic induction of BCSCs. HIF inhibitors may be useful to target were not studied (47). These findings point to the importance of BCSCs because, in addition to blocking induction of SSP and coordinate, HIF-mediated regulation of SSP and mito1CM mito1CM enzymes, they block multiple other pathways, which enzyme expression, and suggest that this metabolic requirement are induced by hypoxia (13, 14) or chemotherapy (6, 7) and is not limited to breast cancer. which promote BCSC specification or maintenance. Our findings þ Metabolic analyses revealed that both ER and ER breast underscore the importance of adaptive responses to the hypoxic cancer cells divert a considerable proportion of glucose-derived tumor microenvironment and to chemotherapy, whereby cancer 3-phosphoglyceric acid to the SSP, as PHGDH knockdown led to cells maintain metabolic and redox homeostasis, which is significantly increased oxidative and glycolytic metabolism under required for execution of the BCSC and metastatic programs that nonhypoxic conditions. Consistent with these findings, metabo- underlie the lethal cancer phenotype. lomic analyses revealed that PHGDH knockdown led to increased concentrations of EMP intermediates downstream of and includ- Disclosure of Potential Conflicts of Interest ing 3-phosphoglyceric acid, the substrate for PHGDH. Gene The authors disclose no potential conflicts of interest. expression and clinical outcome data from thousands of breast cancers suggest that this metabolic reprogramming is not limited Authors' Contributions to tissue culture cells. The orthotopic transplantation studies Conception and design: D. Samanta, G.L. Semenza provide evidence that PHGDH expression is a major determinant Development of methodology: D. Samanta, S.A. Andrabi, G.L. Semenza of BCSC abundance and metastasis. Acquisition of data (provided animals, acquired and managed patients, Our data indicate that PHGDH expression is critical under provided facilities, etc.): D. Samanta, Y. Park, S.A. Andrabi, D.M. Gilkes hypoxic conditions to increase the availability of reduced gluta- Analysis and interpretation of data (e.g., statistical analysis, biostatistics, thione for maintenance of redox homeostasis. In PHGDH knock- computational analysis): D. Samanta, S.A. Andrabi, G.L. Semenza Writing, review, and/or revision of the manuscript: D. Samanta, G.L. Semenza down subclones, NADPH levels were decreased and mitochon- Administrative, technical, or material support (i.e., reporting or organizing drial ROS levels increased under hypoxic conditions, leading to data, constructing databases): L.M. Shelton increased apoptosis compared with NTC subclones. Gene expres- Study supervision: G.L. Semenza sion and metabolomic studies revealed increased expression of the first enzyme of the SSP and mito1CM (PHGDH and SHMT2) and Acknowledgments repression of the first enzyme of the PPP (G6PD) in hypoxic breast We thank Haiquan Lu for helpful comments and Karen Padgett (Novus cancer cells, indicating a switch from cytosolic to mitochondrial Biologicals) for providing IgG and antibodies against PHGDH, PSAT1, PSPH, production of NADPH to combat increased ROS generated by the and Annexin V. electron transport chain (23). Thus, under hypoxic conditions, HIFs mediate a metabolic switch from oxidative to glycolytic Grant Support metabolism, which reduces mitochondrial oxidant generation This work was supported by Breast Cancer Research Program Impact Award (18), and a switch from cytosolic to mitochondrial NADPH W81XWH-12-1-0464 from the Department of Defense and a grant from the generation, which augments antioxidant defenses. The increase Cindy Rosencrans Fund for Triple Negative Breast Cancer (G.L. Semenza). The costs of publication of this article were defrayed in part by the payment of in NADPH complements increased glutathione synthesis, which is page charges. This article must therefore be hereby marked advertisement in also induced in hypoxic breast cancer cells by HIF-1 (7, 48). accordance with 18 U.S.C. Section 1734 solely to indicate this fact. PHGDH knockdown reduced the number of BCSCs under both nonhypoxic and hypoxic conditions in vitro and within orthotopic Received February 18, 2016; revised May 23, 2016; accepted May 31, 2016; tumors. The functional consequences of this loss of BCSCs published OnlineFirst June 8, 2016.

References 1. Al-Hajj M, Wicha M, Benito-Hernandez A, Morrison S, Clarke M. Prospec- metastatic capacity and a distinct molecular signature. Cancer Res tive identiﬁcation of tumorigenic breast cancer cells. Proc Natl Acad Sci 2009;69:1302–13. U S A 2003;100:3983–8. 3. Li X, Lewis MT, Huang J, Gutierrez C, Osborne CK, Wu MF, et al. Intrinsic 2. Charafe-Jauffret E, Ginestier C, Iovino F, Wicinski J, Cervera N, Finetti P, resistance of tumorigenic breast cancer cells to chemotherapy. J Natl Cancer et al. Breast cancer cell lines contain functional cancer stem cells with Inst 2008;100:672–9.

OF12 Cancer Res; 76(15) August 1, 2016 Cancer Research

PHGDH Is Required for Breast Cancer Progression

4. Creighton CJ, Li X, Landis M, Dixon JM, Neumeister VM, Sjolund A, et al. 28. Zhang H, Wong CC, Wei H, Gilkes DM, Korangath P, Chaturvedi P, et al. Residual breast cancers after conventional therapy display mesenchymal as HIF-1-dependent expression of angiopoietin-like 4 and L1CAM mediates well as tumor-initiating features. Proc Natl Acad Sci U S A 2009;106:13820–5. vascular metastasis of hypoxic breast cancer cells to the lungs. Oncogene 5. Bhola NE, Balko JM, Dugger TC, Kuba MG, Sanchez V, Sanders M, et al. 2012:31:1757–70. TGF-b inhibition enhances chemotherapy action against triple-negative 29. Soga T, Ohashi Y, Ueno Y, Naraoka H, Tomita M, Nishioka T. Quantitative breast cancer. J Clin Invest 2013;123:1348–58. metabolome analysis using capillary electrophoresis mass spectrometry. J 6. Samanta D, Gilkes DM, Chaturvedi P, Xiang L, Semenza GL. Hypoxia- Proteome Res 2003;2:488–94. inducible factors are required for chemotherapy resistance of breast cancer 30. Sugimoto M, Wong DT, Hirayama A, Soga T, Tomita M. Capillary elec- stem cells. Proc Natl Acad Sci U S A 2014;111:E5429–38. trophoresis mass spectrometry-based saliva metabolomics identified oral, 7. Lu H, Samanta D, Xiang L, Zhang H, Hu H, Chen I, et al. Chemotherapy breast and pancreatic cancer–specificprofiles. Metabolomics 2009;6: triggers HIF-1- dependent glutathione synthesis and copper chelation that 78–95. induces the breast cancer stem cell phenotype. Proc Natl Acad Sci U S A 31. Goldman M, Craft B, Swatloski T, Ellrott K, Cline M, Diekhans M, et al. The 2015;112:E4600–9. UCSC Cancer Genomics Browser: update 2013. Nucleic Acids Res 2013;41: 8. Harris AL. Hypoxia – a key regulatory factor in tumor growth. Nat Rev D949–54. Cancer 2002;2:38–74. 32. The Cancer Genome Atlas Network. Comprehensive molecular portraits of 9. Sullivan R, Graham CH. Hypoxia-driven selection of the metastatic phe- human breast tumors. Nature 2012;490:61–70. notype. Cancer Metastasis Rev 2007;26:319–31. 33. Ringner M, Fredlund E, H€akkinen J, Borg Å, Staaf J. GOBO: gene expression 10. Conley SJ, Gheordunescu E, Kakarala P, Newman B, Korkaya H, Heath AN, based outcome for breast cancer online. PLoS One 2011;6:e17911. et al. Antiangiogenic agents increase breast cancer stem cells via the 34. Gyorffy€ B, Lanczky A, Eklund AC, Denkert C, Budczies J, Li Q, et al. An generation of tumor hypoxia. Proc Natl Acad Sci U S A 2012;109:2784–9. online survival analysis tool to rapidly assess the effect of 22,277 genes on 11. Schwab LP, Peacock DL, Majumdar D, Ingels JF, Jensen LC, Smith KD, et al. breast cancer prognosis using microarray data of 1,809 patients. Breast Hypoxia- inducible factor 1a promotes primary tumor growth and tumor- Cancer Res Treat 2010;123:725–31. initiating cell activity in breast cancer. Breast Cancer Res 2012;14:R6. 35. Neve RM, Chin K, Fridlyand J, Yeh J, Baehner FL, Fevr T, et al. A collection of 12. Regan Anderson TM, Peacock DL, Daniel AR, Hubbard GK, Lofgren KA, breast cancer cell lines for the study of functionally distinct cancer subtypes. Girard BJ, et al. Breast tumor kinase (Brk/PTK6) is a mediator of hypoxia- Cancer Cell 2006;10:515–27. associated breast cancer progression. Cancer Res 2013;73:5810–20. 36. Lee K, Zhang H, Qian DZ, Rey S, Liu JO, Semenza GL. Acriflavine inhibits 13. Xiang L, Gilkes DM, Hu H, Takano N, Luo W, Lu H, et al. Hypoxia-inducible HIF-1 dimerization, tumor growth, and vascularization. Proc Natl Acad Sci factor 1 mediates TAZ expression and nuclear localization to induce the U S A 2009;106:17910–5. breast cancer stem cell phenotype. Oncotarget 2014;5:12509–27. 37. Ginestier C, Hur MH, Charafe-Jauffret E, Monville F, Dutcher J, Brown M, 14. Zhang C, Samanta D, Lu H, Bullen JW, Zhang H, Chen I, et al. Hypoxia et al. ALDH1 is a marker of normal and malignant human mammary stem induces the breast cancer stem cell phenotype by HIF-dependent and cells and a predictor of poor clinical outcome. Cell Stem Cell 2007;1: ALKBH5-mediated m6A-demethylation of NANOG mRNA. Proc Natl Acad 555–67. Sci U S A 2016;113:E2047–56. 38. Ponti D, Costa A, Zaffaroni N, Pratesi G, Petrangolini G, Coradini D, et al. 15. Semenza GL. The hypoxic tumor microenvironment: a driving force for Isolation and invitro propagation of tumorigenic breast cancer cells with breast cancer progression. Biochim Biophys Acta 2016;1863:379–81. stem/progenitor cell properties. Cancer Res 2005;65:5506–11. 16. Iyer NV, Kotch LE, Agani F, Leung SW, Laughner E, Wenger RH, et al. 39. Gardner PR, Nguyen DD, White CW. Superoxide scavenging by Mn(II/III)

Cellular and developmental control of O2 homeostasis by hypoxia-induc- tetrakis (1-methyl-4- pyridyl) porphyrin in mammalian cells. Arch Bio- ible factor 1a. Genes Dev 1998;12:149–62. chem Biophys 1996;325:20–8. 17. Seagroves TN, Ryan HE, Lu H, Wouters BG, Knapp BG, Thibault P, et al. 40. Doroshow JH. Redox modulation of chemotherapy-induced tumor cell Transcription factor HIF-1 is a necessary mediator of the Pasteur effect. Mol killing and normal tissue toxicity. J Natl Cancer Inst 2006;98:223–5. Cell Biol 2001;21:3436–44. 41. Perez EA. Carboplatin in combination therapy for metastatic breast cancer. 18. Kim JW, Tchernyshyov I, Semenza GL, Dang CV. HIF-1-mediated expres- Oncologist 2004;9:518–27. sion of pyruvate dehydrogenase kinase: a metabolic switch required for 42. Wang H, Li M, Rinehart JJ, Zhang R. Pretreatment with dexamethasone cellular adaptation to hypoxia. Cell Metab 2006;3:177–85. increases antitumor activity of carboplatin and gemcitabine in mice 19. Papandreou I, Cairns RA, Fontana L, Lim AL, Denko NC. HIF-1 mediates bearing human cancer xenografts: invivo activity, pharmacokinetics, and adaptation to hypoxia by actively downregulating mitochondrial oxygen clinical implications for cancer chemotherapy. Clin Cancer Res 2004;10: consumption. Cell Metab 2006;3:187–97. 1633–44. 20. Huang D, Li T, Li X, Zhang L, Sun L, He X, et al. HIF-1-mediated suppression 43. Marullo R, Werner E, Degtyareva N, Moore B, Altavilla G, Ramalingam SS, of acyl-CoA dehydrogenases and fatty acid oxidation is critical for cancer et al. Cisplatin induces a mitochondrial-ROS response that contributes to progression. Cell Rep 2014;8:1930–42. cytotoxicity depending on mitochondrial redox status and bioenergetic 21. Zhang H, Bosch-Marce M, Shimoda LA, Tan YS, Baek JH, Wesley JB, et al. functions. PLoS ONE 2013;8:e81162. Mitochondrial autophagy is an HIF-1-dependent adaptive metabolic 44. Cao Y, Eble JM, Moon E, Yuan H, Weitzel DH, Landon CD, et al. et al.Tumor response to hypoxia. J Biol Chem 2008;283:10892–903. cells upregulate normoxic HIF-1a in response to doxorubicin. Cancer Res 22. Bellot G, Garcia-Medina R, Gounon P, Chiche J, Roux D, Pouyssegur J, et al. 2013;73:6230–42. Hypoxia- induced autophagy is mediated through hypoxia-inducible 45. DeNicola GM, Chen PH, Mullarky E, Sudderth JA, Hu Z, Wu D, et al. NRF2 factor induction of BNIP3 and BNIP3L via their BH3 domains. Mol Cell regulates serine biosynthesis in non-small cell lung cancer. Nat Genet Biol 2009;29:2570–81. 2015;47:1475–81. 23. Chandel NS, McClintock DS, Feliciano CE, Wood TM, Melendez JA, 46. Sun L, Song L, Wan Q, Wu G, Li X, Wang Y, et al. cMyc-mediated activation Rodriguez AM, et al. Reactive oxygen species generated at mitochondrial of serine biosynthesis pathway is critical for cancer progression under complex III stabilize hypoxia-inducible factor-1a during hypoxia: a mech- nutrient deprivation conditions. Cell Res 2015;25:429–44.

anism of O2 sensing. J Biol Chem 2000;275:25130–8. 47. Ye J, Fan J, Venneti S, Wan YW, Pawel BR, Zhang J, et al. Serine catabolism 24. Schumacker PT. Reactive oxygen species in cancer: a dance with the devil. regulates mitochondrial redox control during hypoxia. Cancer Discov Cancer Cell 2015;27:156–7. 2014;4:1406–17. 25. Possemato R, Marks KM, Shaul YD, Pacold ME, Kim D, Birsoy K, et al. 48. Harris IS, Treloar AE, Inoue S, Sasaki M, Gorrini C, Lee KC, et al. Gluta- Functional genomics reveal that the serine synthesis pathway is essential in thione and thioredoxin antioxidant pathways synergize to drive cancer breast cancer. Nature 2011;476:346–50. initiation and progression. Cancer Cell 2002;27:211–22. 26. Locasale JW, Grassian AR, Melman T, Lyssiotis CA, Mattaini KR, Bass AJ, 49. Brooks MD, Burness ML, Wicha MS. Therapeutic implications of cellular et al. Phosphoglycerate dehydrogenase diverts glycolytic ﬂux and contri- heterogeneity and plasticity in breast cancer. Cell Stem Cell 2015;17: butes to oncogenesis. Nat Genet 2011;43:869–74. 260–71. 27. Suda T, Takubo S, Semenza GL. Metabolic regulation of hematopoietic 50. Oskarsson T, Batlle E, Massague J. Metastatic stem cells: sources, niches, and stem cells in the hypoxic niche. Cell Stem Cell 2011:9:298–310. vital pathways. Cell Stem Cell 2014;14:306–21.

www.aacrjournals.org Cancer Res; 76(15) August 1, 2016 OF13

PHGDH Expression Is Required for Mitochondrial Redox Homeostasis, Breast Cancer Stem Cell Maintenance, and Lung Metastasis

Debangshu Samanta, Youngrok Park, Shaida A. Andrabi, et al.

Cancer Res Published OnlineFirst June 8, 2016.

Updated version Access the most recent version of this article at: doi:10.1158/0008-5472.CAN-16-0530

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2016/06/08/0008-5472.CAN-16-0530.DC1

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/early/2016/07/17/0008-5472.CAN-16-0530. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.