Cancer Immunology, Immunotherapy (2019) 68:1107–1120 https://doi.org/10.1007/s00262-019-02347-3

ORIGINAL ARTICLE

Metabolic remodeling contributes towards an immune‑suppressive phenotype in glioblastoma

Pravin Kesarwani1 · Antony Prabhu1 · Shiva Kant1 · Prakash Chinnaiyan1,2

Received: 31 October 2018 / Accepted: 17 May 2019 / Published online: 22 May 2019 © Springer-Verlag GmbH Germany, part of Springer Nature 2019

Abstract Glioblastoma (GBM) is one of the most aggressive tumors. Numerous studies in the feld of immunotherapy have focused their eforts on identifying various pathways linked with tumor-induced immunosuppression. Recent research has dem- onstrated that metabolic reprogramming in a tumor can contribute towards immune tolerance. To begin to understand the interface between metabolic remodeling and the immune-suppressive state in GBM, we performed a focused, integrative analysis coupling metabolomics with -expression profling in patient-derived GBM (n = 80) and compared them to low- grade astrocytoma (LGA; n = 28). Metabolic intermediates of , , , and adenosine emerged as immuno-metabolic nodes in GBM specifc to the mesenchymal and classical molecular subtypes of GBM. Integrative analyses emphasized the importance of downstream of several of these metabolic pathways in GBM. Using CIBERSORT to analyze immune components from the transcriptional profles of individual tumors, we demonstrated that tryptophan and adenosine metabolism resulted in an accumulation of Tregs and M2 , respectively, and was recapitulated in mouse models. Furthermore, we extended these fndings to preclinical models to determine their potential utility in defning the biologic and/or immunologic consequences of the identifed metabolic programs. Collectively, through integrative analysis, we uncovered multifaceted ways by which metabolic reprogramming may contribute towards immune tolerance in GBM, providing the framework for further investigations designed to determine the specifc immunologic con- sequence of these metabolic programs and their therapeutic potential.

Keywords Glioblastoma · Immune metabolism · Tryptophan · Arginase · Prostaglandin · Adenosine

Abbreviations CBR Carbonyl reductase 1-L-MT 1-Methyl-l-tryptophan CIBERSORT Cell-type identifcation by estimating rela- AD-H Adenosine pathway metabolites-high tive subsets of RNA transcripts AD-L Adenosine pathway metabolites-low CKMT1 , mitochondrial 1A AHR Aryl hydrocarbon COX ARG-H Arginine pathway metabolites-high GBM Glioblastoma ARG-L Arginine pathway metabolites-low GSCs Glioma stem cells ARG2 Arginase 2 IDH1 1 ASL Argininosuccinate IL2-Rα Interleukin 2 receptor-subunit alpha KEGG Kyoto encyclopedia of and genomes KMO 3-Mono- Electronic supplementary material The online version of this KYAT aminotransferase article (https​://doi.org/10.1007/s0026​2-019-02347​-3) contains KYNU supplementary material, which is available to authorized users. LGA Low-grade astrocytoma * Prakash Chinnaiyan MGMT O6-methylguanine–DNA [email protected] methyltransferase MSP Methylation-specifc PCR 1 Department of Radiation Oncology, Beaumont Health, 3811 NAD Nicotinamide adenine dinucleotide West Thirteen Mile Road, Royal Oak, MI 48073, USA NOS1 Nitric oxide synthase 1 2 Oakland University William Beaumont School of Medicine, PG-H Prostaglandin pathway metabolites-high Royal Oak, MI, USA

Vol.:(0123456789)1 3 1108 Cancer Immunology, Immunotherapy (2019) 68:1107–1120

PG-L Prostaglandin pathway metabolites-low gliomagenesis, and we performed integrative, cross-platform PGE2 Prostaglandin E2 analyses consisting of global metabolomics coupled with PGF2A Prostaglandin F2 alpha gene-expression profling in patient-derived tumors. In addi- PGG2 Prostaglandin G2 tion to demonstrating clear metabolic reprogramming asso- PGH2 ciated with these immune modulatory pathways in GBM, PLS Partial least squares these comprehensive studies uncovered transcriptional pro- PTGES Prostaglandin E synthase grams designed to drive the observed metabolic phenotype, PTGIS Prostaglandin I synthase putative metabolic targets, and a framework for future stud- PTGS Prostaglandin-endoperoxide synthase ies designed to determine how these specifc metabolic pro- QPRT Quinolinate phosphoribosyltransferase grams may actively infuence the immune state. TCGA The cancer genome atlas TDO Tryptophan 2,3- TRP-H Tryptophan pathway metaboliteshigh Materials and methods TRP-L Tryptophan pathway metaboliteslow VIP Variable importance in projection Human tumor samples and metabolomic profling

Introduction Metabolomic profling was performed in glioma using a combination of liquid chromatography/tandem mass spec- trometry (LC/MS) and gas chromatography/mass spectrom- Glioblastoma (GBM) is the most common adult primary etry (GC/MS) using a robust library of standards enriched brain tumor [1]. Despite continued advances in surgery, radi- with intermediates of tryptophan, arginine, prostaglandin, ation, and the identifcation of novel molecularly targeted and adenosine metabolism, comparing patient-derived GBM agents, outcomes remain poor. Recent clinical advance- (n = 80) to low-grade astrocytoma (LGA; n = 28). All tumors ments using immune checkpoint inhibitors designed to were newly diagnosed, fresh-frozen, and their integrity and target tumor-mediated immune tolerance have revolution- histology confrmed by a staf pathologist [14]. Metabolomic ized our approach to cancer therapy. Cytotoxic T-lympho- studies were conducted at Metabolon (Morrisville, NC) and cyte-associated protein 4 and programmed death-1, which analyzed using methods previously described [14, 15]. negatively regulate T-cell activation, represent two specifc immune checkpoints that have received recent attention, Microarray and database analysis with inhibitors targeting these immune pathways demon- strating unprecedented clinical activity in multiple tumors [2–4]. Unfortunately, as with many novel therapies, these The cancer genome atlas (TCGA) data for glioma were approaches have yet to translate to meaningful beneft in downloaded from http://xena.ucsc.edu. Genes associated GBM [5]. Therefore, continued investigations designed to with tryptophan, arginine, prostaglandin, and adenosine both understand the immune landscape of GBM and identify metabolism were identifed using the kyoto encyclopedia novel strategies to revert its immune-suppressive microenvi- of genes and genomes (KEGG) pathway database. mRNA ronment are warranted. expression data for these gene sets were compared between Aberrant cellular metabolism is emerging as a novel ther- low-grade glioma (LGG) and GBM to generate log2 apeutic target, and the interplay between metabolic remod- fold change (LGG at the baseline). Furthermore, Benja- p eling and immune regulation in cancer represents an active mini–Hochberg corrected value < 0.05 (−log10) between area of investigation [6, 7]. For example, enhanced glycoly- these two groups was used to generate volcano plots. Gene- sis in tumors that appears to be driving its aggressive pheno- expression profling and molecular subtyping for O6-methyl- type leads to a microenvironment depleted of glucose, which guanine–DNA methyltransferase (MGMT) promoter methyl- is a critical to help support the rapid and dynamic ation and isocitrate dehydrogenase 1 (IDH1) mutation status transitions between naïve and activated states in a variety were performed as previously described [16]. of immune cells [2]. In addition to passive consequences of metabolic reprogramming, tumors have co-opted vari- Cell culture and animal handling ous metabolic strategies to actively modulate the immune landscape. Tryptophan [8, 9], arginine [10], prostaglandin Human GBM cell lines U251 and T98G were obtained and [11, 12], and adenosine [13] metabolism represent some of grown in conditions described previously [17]. The geneti- the most studied pathways actively contributing towards cally engineered murine GBM TRP line and human GBM immune suppression. To begin to understand the relevance, mesenchymal (MES83 and MES326) and proneural (PN19 these immuno-metabolic pathways may play a role in and PN84) glioma stem cells (GSCs) were generated and

1 3 Cancer Immunology, Immunotherapy (2019) 68:1107–1120 1109

a class LGA Kynurenate GBM Kynurenine Serotonin Tryptophan Kynurenine/Kynurenate Indoleacetate Indolelactate 3-Indoxyl sulfate 5-Hy droxyindoleacetate Tryptophan 1-Methylnicotinamide ADP- ribose pathway N1-Methyl-2-py-5-carb Nicotinamide NA D+ NA DH Nicotinamide riboside Quinolinate Arginine Urea Ornithine Pr oline Citrulline Arginine Creatine Creatine-P pathway 4-Hy droxyproline N-Acetylputrescine Putrescine Spermidine Spermine N-Acetylspermine Pr ostaglandin E2 Pr ostaglandin F2alpha Prostaglandin Pr ostaglandin J2 6-Keto PG F1alpha pathway 5-HETE Adenine Adenosine 2'-AMP cAMP Adenosine 3'-AMP ADP pathway ADP ribose AMP Adenosine 2',3'-cMP

5-0 5

Pathways Tryptophan b Arginine Prostaglandin Adenosine e lu Va p 10 Lo g −

Log2 Fold Change (GBM / LGG)

Fig. 1 Immuno-metabolic pathways are altered in GBM when com- a heat map after normalization. b mRNA expression data for GBM pared to LGA. a Global metabolomic profling was performed on and low-grade glioma (LGG) from the TCGA were used to analyze patient-derived glioblastoma (GBM; n = 80) and LGA (n = 28) tumor genes specifc to tryptophan, arginine, prostaglandin, and adenosine samples using LG/GC–MS. Biochemicals involved in metabolic metabolism. Diferentially expressed genes (p < 0.05) are presented as pathways with known immune consequence, including tryptophan, a volcano plot arginine prostaglandin, and adenosine metabolism, are presented as

1 3 1110 Cancer Immunology, Immunotherapy (2019) 68:1107–1120 grown as previously described [18, 19]. Flank and orthotopic Fig. 2 Tryptophan metabolism in GBM. a Schematic of tryptophan ◂ tumor implants were performed as previously described [8]. (TRP) metabolism. Red indicates metabolites upregulated in GBM when compared to LGA. Green indicates metabolites downregulated in GBM; brown indicates that the metabolite was analyzed, but was Human PBMCs not signifcantly diferent. Metabolites in black were not detected or analyzed. Numbers in bracket demonstrate fold diference between Leukapheresis packs were obtained from healthy donors by GBM and LGA. b Hierarchical clustering using metabolites spe- cifc to tryptophan metabolism was performed on molecularly Research Blood Components (Boston, MA) and PBMCs subtyped, patient-derived GBM (n = 56), resulting in two distinct were isolated using Ficoll–Paque PLUS (GE Bio-Sciences, clusters defned as TRP-High (TRP-H) and TRP-Low (TRP-L). c Pittsburgh, PA). PBMCs were activated for 3 days using Gene-expression profles of GBM defned as TRP-H and TRP-L were plate-bound anti-CD3 (clone-OKT3) and anti-CD28 (clone- analyzed using CIBERSORT. *p < 0.05. d Described cell lines were cultured with ± 100 ng/ml human IFN-γ or 100 ng/ml human IFN-γ CD28.2) antibodies (Biolegend; San Diego, CA). and 100 µM of 1-methyl-l-tryptophan (1-L-MT) for 3 days and ana- lyzed for IDO1 and TDO2 expression. Supernatant from this cell cul- Kynurenine estimation ture was used for kynurenine estimation. e Described cell lines were cultured with ± 100 ng/ml human IFN-γ for 3 days and analyzed for genes involved in tryptophan metabolism. Comparisons were made to Mouse cells were cultured in the presence or absence of naïve and activated PBMCs murine recombinant IFN-γ (100 ng/ml) and human cells were cultured in the presence of human recombinant IFN-γ (50 ng/ml) from Peprotech (Rocky Hill, NJ) for 3 days in ­CD45+CD68+CD11b+F4/80+ (Biolegend). CD8­ + T cells the presence or absence of 1-methyl-l-tryptophan (Sigma- were isolated from C57BL/6 mouse splenocytes and purifed Aldrich, St. Louis, MO). Cell-culture supernatants were using an untouched mouse CD8 cell-isolation kit (Invitrogen/ collected for kynurenine estimation using Ehrlich’s reagent Thermo Fisher). ­CD8+ T cells were stained with CellTrace using methods previously described [20, 21]. (CFSE; GIBCO/Thermo) and activated with plate-bound anti-CD3 (1 µg/ml; clone-145-2C11) and anti-CD28 (5 µg/ Western blot ml; clone-37.51) (Biolegend). Polarized M2 cells were added at indicated ratios ± adenosine to evaluate their capacity to Western blot was performed using methods previously suppress ­CD8+ T-cell proliferation using CFSE dilution. described [8]. Human indoleamine 2,3-dioxygenase (IDO1), tryptophan 2,3-dioxygenase-2 (TDO), prostaglandin I syn- CIBERSORT thase (PTGIS), cyclooxygenase-1/2 (PTGS1/PTGS2), and β-actin antibodies were obtained from Cell-Signaling Normalized gene-expression data from GBM were used for Technology (Boston, MA). Tryptophan 2,3-dioxygenase-2 performing cell-type identifcation by estimating relative (TDO2) antibody was obtained from Santa Cruz Biotechnol- subsets of RNA transcripts (CIBERSORT) allowing the ogy (Dallas, TX). identifcation of 22 immune subtypes within a tumor through a deconvolution algorithm that uses a set of signature gene- T‑regulatory cell (Treg) polarization expression values specifc for these immune subsets [22].

CD4+ T cells were isolated from C57BL/6 mouse spleno- cytes and purifed using an untouched mouse CD4 cell-iso- Results lation kit (Invitrogen/Thermo Fisher). Cells were activated using plate-bound anti-CD3 (1 µg/ml; clone-145-2C11) and To begin to understand how metabolic reprogramming may anti-CD28 (5 µg/ml; clone-37.51) antibodies (Biolegend) contribute towards the immune-suppressive microenviron- along with TGF-β1 (5 ng/ml) for 3 days, in ± kynurenine, ment observed in GBM, metabolomic profling was per- and analyzed for Tregs ­(CD45+CD4+FoxP3+CD25+) using formed on patient-derived gliomas, comparing low-grade fuorochrome-conjugated antibodies (Biolegend). astrocytoma (LGA; n = 28) with GBM (n = 80), specifcally focusing on metabolites/metabolic pathways implicated in M2‑ polarization and cell‑suppression immune tolerance, which included tryptophan, arginine, assay prostaglandin, and adenosine metabolism [7]. A total num- ber of 55 biochemicals involved in these described metabolic Bone-marrow cells were isolated from the femur of C57BL/6 pathways were identifed in human tumors and 52% demon- mice, polarized to an M2 macrophage phenotype in the pres- strated diferential accumulation between LGA and GBM. ence of murine GM-CSF (20 ng/ml) for 6 days followed by Hierarchical clustering performed on this focused panel murine IL-4 and IL-13 (20 ng/ml) (Peprotech) and analyzed of metabolites resulted in a clear separation between LGA for polarization with fuorochrome-conjugated antibodies, and GBM (Fig. 1a). Consistent with metabolomic fndings,

1 3 Cancer Immunology, Immunotherapy (2019) 68:1107–1120 1111

a Deamination pathway 3-Indoxyl sulfate (5.7) Indole-acetate (2) Indole-lactate (2.2) D Serotonin biosynthesis KYNU DD C TPH1 5-Hydroxyl 5-Hydroxy IDO1 N-Formyl AFMI Kynurenine Anthralic Serotonin Tryptophan (1.4) Indole acetate (0.12) L-Tryptophan TDO Kynurenine (6.1) acid

Kynurenine pathway D O KYAT I

U KMO KYAT II 3HAA

ACMS 3-Hydroxy Acetyl Glutaryl KYN 3-Hydroxy Kynurenate (0.23) AMS ACMS anthralic CoA CoA kynurenine acid

Quinolinic NA NA Nicotinamide NAD+ (8.5)

acid (64) mononucleotide 3 dinucleotide (NAM) , T1

T2 NAMPT QPRT

NMNA NMNAT1 NADYSN NAM

NMNA NMNAT2, 3 mononucleotide b De novo NAD biosynthesis NAD salvage IDH MGMT PN MES Subtype d 3 TRP-H TRP-L Nicotinamide 26 S3 19 ribose 84

Kynurenate ES8 M ME PN Serotonin PN 6 5-Hydroxyindoleacetate ~48 IDO1 Kynurenine 4 Tryptopha ~50 TDO2 n 1-Methylnicotinamide 2 ~55 Tubulin ADP- 0 IFN γ − + − + − − Niribosecotinamide + + Kynurenine Indoleacetate ) -2 50

Indolelactate µM 40 3-Indoxyl sulfate -4 n( io 30 Kynurenine/Kynurenate at -6 tr 20 Quinolinate en 10 NADH nc 0 Co t t t N1-Methyl-2-PY t N N N N N N N N en en en NAD en IF IF IF IF IF IF IF IF tm tm tm tm T+ T+ T+ c T+ ea ea ea ea LM LM LM LM tr tr tr tr ) 1- 1- 1- 1- No No No 45 No (% TRP-H TRP-L

n 40 MES326MES83PN19PN84

o e 35

30 )

Fracti 25

12 FU (R

Cell 8 * 20 ues 4 **l * * 15 va … … … … … … … … 0 r … … … … … … … … y s s e a 1 0 g g e Relative 8 d ls l ry t ls ry s s s s s i e e s 8 a s s c c l e ed n ls l ls ls ls ls ls ng i l ls i ed ed iv ll ll iv D i in or l ll l t il i ll M2

o 10 te t p e ha h ha el t hi ti t it l tory iv te yt el el el el iz at el e ated el h ri c C ting s s d p sM sM mo e a c na na CD ce ce ce ce va la va c ph c p ce dr i es e iv o e em cy em h al n a roph t re ls o es t nd g Tc Tc Tc Tc T T K o ls Bc ls no ct iv r in ct cr op cr op cr op st st me l

m 5 s t en no e u sr sr el ma n D4 ar ac l o a acti rm ll NK N a e l s ll l ma el sr c sm si 4m regu l D D os l m u a s c e sa l M eu phages l ls Ne Ma Ma M Ma Ma l ls D4 T E a c l el ce ce a Mo sC i No l P Bc N l lls T Eo c el el ro tc CD ll ce 0 K B P e c ce c crophage cropha s el c sg sC fo l B l ls a c N l ti cc l i st s e t Tc el M Ma Ma Ma T ll NK ri ce e d Ma Tc T Tc n Tc Dendri De PN PN+IFN MES MES+IFN Naïve PBMCs Activated PBMCs clear separation between LGA and GBM was observed when Supplementary Table 1). Using integrative analyses, we evaluating the expression of genes involved in these meta- went on to metabolically, molecularly, and functionally to bolic pathways using the TCGA, supporting the potential defne these metabolic pathways in further detail. role of these metabolic pathways in gliomagenesis (Fig. 1b;

1 3 1112 Cancer Immunology, Immunotherapy (2019) 68:1107–1120

Tryptophan metabolism. Tryptophan can be metabolized available. MGMT promoter methylation status and IDH1 to kynurenine, which is driven by the rate-limiting mutation represent two of the strongest prognostic factors IDO/TDO. Kynurenine can then be exported to the microen- in GBM [28, 29]. We, therefore, went on to determine if vironment by tumors, contributing towards immune suppres- these molecular subtypes diferentially co-opted tryptophan sion at many levels, with its most notable role in contributing metabolism to modulate the immune response. Although towards materno-fetal immune tolerance [23–25]. Although IDH1 and MGMT methylation status did not appear to cor- several recent studies have implicated these enzymes and/ relate with the observed metabolic phenotype when tran- or kynurenine in gliomagenesis [8, 9, 26], its intermediary scriptional profles were molecularly subtyped [30], the metabolism has not been studied in detail. When compared immuno-metabolic phenotype of tryptophan metabolism to LGA, in addition to an expected accumulation of tryp- was unique to mesenchymal and classical subtypes of GBM tophan, kynurenine, and its biosynthetic enzymes IDO1/ (Fig. 2b). TDO [7], an accumulation of several indoles was observed To determine if tryptophan metabolism infuenced the in GBM (Fig. 2a). Similar to kynurenine, indoles appear to immune landscape in GBM, immune phenotypes were have the capacity of activating the aryl hydrocarbon recep- defned using transcriptional profles generated from indi- tor (AHR) [27] and, therefore, may have immune modula- vidual tumors and analyzed using CIBERSORT [22], allow- tory roles [26]. However, the -linking indoles with ing for in silico cell sorting of specifc immune components. tryptophan metabolism are not typically present in humans Integrative analyses coupling these metabolomic signatures [27]; therefore, the specifc metabolic pathways contributing with specifc cellular immune subsets suggested that trypto- to the presence of these metabolites in the tumor microen- phan metabolism contributes towards an immunosuppressive vironment and their biologic consequence remains unclear. phenotype in GBM, with signifcantly higher levels of Tregs A particularly striking fnding these integrative analyses and M0 macrophages and lower levels of memory T cells. In uncovered was a clear shift towards the downstream metab- addition, a trend in diminished CD8 cells was observed in olism of kynurenine in GBM. Most notable was an acti- tryptophan ‘high’ tumors (Fig. 2c). To functionally extend vation of the de novo NAD+ biosynthetic pathway, includ- fndings linking kynurenine metabolism with the accumula- ing an accumulation of the metabolites quinolinic acid and tion of Tregs in GBM, we sought to determine if kynurenine NAD+ that were coupled to several of the biosynthetic contributed towards Treg polarization. As demonstrated in enzymes involved in this pathway. Supplementary Fig. 2, ­CD4+ T cells isolated from murine Despite clear clustering when compared to LGA, con- splenocytes demonstrated a 44% increase in Treg polariza- siderable metabolic heterogeneity was still observed within tion when cultured in the presence of kynurenine. GBM. Therefore, we sought to both defne this metabolic Finally, we sought to determine if this immune-met- heterogeneity and understand its molecular context. Of the abolic phenotype was recapitulated in preclinical mod- 80 GBM specimens that were metabolomically profled, 56 els. Although the activation of the IDO1/TDO pathway had additional tissue available to allow for cross-platform has been previously described in GBM [8, 26], a unique genomic/transcriptional analyses. Hierarchical clustering fnding our integrative analysis uncovered was that this performed using metabolites specifc to tryptophan metab- phenotype was specific to the mesenchymal/classical olism in these tumors identifed two subtypes defned as subtypes of GBM. Using subtype-specific GBM lines tryptophan ‘high’ and ‘low’ (Fig. 2b). Next, we evaluated [18, 31], consistent with our fndings involving patient- gene-expression profles of these two metabolic subtypes to derived tumors, we were able to demonstrate that upon provide molecular context to the observed metabolic hetero- IFN-γ induction, mesenchymal lines demonstrated geneity. Consistent with the integrative analysis comparing increased expression of IDO1 and kynurenine accumula- GBM with LGA, IDO1 emerged as the top gene separating tion (Fig. 2d). Another novel fnding from our studies was tryptophan ‘high’ and ‘low’ GBM on VIP analysis (Supple- the observation of continued metabolism downstream of mentary Fig. 1a). Interestingly, these studies also identifed kynurenine in GBM. However, it remains unclear whether quinolinate phosphoribosyltransferase (QPRT) expression this occurs in tumor cells or a component of the inter- to be a central mediator driving this metabolic phenotype, mediary metabolism of immune cells within the tumor further supporting the relevance of the downstream metabo- microenvironment [32, 33]. We, therefore, extended our lism of tryptophan in GBM. studies to evaluate its downstream metabolism. As an We next sought to determine if the observed metabolic initial investigation, we determined the expression level heterogeneity of tryptophan metabolism in GBM could be of kynurenine 3-mono-oxygenase (KMO), kynureninase a direct consequence of established molecular subtypes in (KYNU), and QPRT in our preclinical models, which this malignancy. To accomplish this, we performed cross- represent aberrantly expressed enzymes involved in the platform analyses using RNA and DNA isolated from the downstream metabolism of kynurenine (Fig. 2a). Consist- 56 samples, where a matched aliquot of tumor tissue was ent with clinical data, mesenchymal lines demonstrated

1 3 Cancer Immunology, Immunotherapy (2019) 68:1107–1120 1113

Urea cycle L-Arginosuccinate(2.2) a NOS1, NO ASS1 ASL NOS2, NOS3 Creatinine biosynthesis CKM NOS1, NOS2, NOS3 Creatine (0.6) Citruline(1.5) Arginine CKMT1 CKMT1 ARG1, ARG2 Creatine-P OTC CKB Creatinine Ornithine Urea (1.5) (0.4) (5.8) Putrescine & Spermine pathway ODC1 OAT biosynthesis

βAlanine SMS Spermidine SRM Spermine Putrescine Glutamate 5- biosynthesis (0.9) semialdehyde SAT1 N-Acetyl- N-Acetlyspermine putrescine LAP3 (19.5) Proline (3.1) Peptides MAO B MAO A P4H4 1-Proline •ADC pathway N4-Acetyl- DAO 4-Hydroxyproline 4hydroxyl 2 •Butenoate pathway aminobutanal • aspartate glutamate (1.7) carboxylate b IDH WT MT MGMT UNMM Subtype MCLN PN d ARG-H ARG-L Putrescine 20 Creatine Spermidine on 15 Spermine ssi ) N-acetylspermine FU pre 10 Ornithine (R ex Proline ed

Citrulline iz 5 Creatine-P al rm Urea 0 No Creatinine ARG2 ASL 4-Hydroxyproline Arginine PN N-Acetylputrescine MES 42 0-2-4 Naïve PBMCs c Activated PBMCs

) e 40 (% ARG-H ARG-L LGA MESXeno tion 30

ac Creatine-P

Fr 4-Hydroxyproline ll 20 Citrulline

Ce Ornithine Proline 10 Creatine * * Spermine * Urea

0 r Relative y s s d s e a y 0 y 1 2 g d s g e 8 g

l Arginine il ll i lt ng n e ed ed iv iv i i in te te in or or t or pe t t tory Spermidine a at at ce CD st de s sM sM sM m el a cy na na va iv e iv

iv Creatinine i e e em em o t r t a roph re s rest ul ls re noph g ge t ls rh n ct ct c N-acetylspermine g m ma s el m u D4 ac l a ls el la l lls l c hage e re 4m

4m Putrescine ls m u as sa sa s Mo p l Eosi N T a ll ll ce ce ce el

sC N-Acetylputrescine ic ls o opha opha Pl Bc l e el r r ll l CD CD el ce st ic cr c

e 42 0 -2 -4 o sg t f i Bc a ac NK ls ls ll cc r Kc i st s t Tc d el Ma el Ma M M l Tc N ce ri n el d Ma T Tc Tc De en Tc D

Fig. 3 Arginine metabolism in GBM. a Schematic of arginine metab- ters were defned as ARG-High (ARG-H) and ARG-Low (ARG-L). c olism. Red indicates metabolites upregulated in GBM when com- CIBERSORT analysis for 22 immune subsets was performed compar- pared to LGA. Green indicates metabolites downregulated in GBM ing the immune phenotype of ARG-H and ARG-L tumors. *p < 0.05. when compared to LGA; brown indicates that the metabolite was ana- d Mesenchymal (MES83 and MES326) and proneural (PN19 and lyzed, but was not signifcantly diferent. Metabolites in black were PN84) GBM tumor lines were analyzed for genes involved in arginine not detected or analyzed. Numbers in bracket demonstrate fold difer- metabolism and compared with naïve and activated PBMCs. e MES ence between GBM and LGA. b GBM (n = 56) with known molec- orthotopic tumor xenografts in nude (nu/nu) mice (green; n = 6) were ular subtype, MGMT methylation status, and IDH1 mutation status analyzed for arginine pathway metabolites using global metabolomic was clustered using arginine pathway metabolites. Two distinct clus- profling and compared with LGA (red; n = 28) 1 3

1114 Cancer Immunology, Immunotherapy (2019) 68:1107–1120 ◂ increased expression of all three enzymes when compared Fig. 4 Prostaglandin metabolism in GBM. a Schematic of prosta- to proneural, and importantly, their relative expression was glandin metabolism. Red indicates metabolites upregulated in GBM when compared to LGA. Green indicates metabolites downregulated consistent with the rate-limiting enzyme IDO1 (Fig. 2e). in GBM; brown indicates that the metabolite was analyzed, but was To further explore the potential role immune cells may not signifcantly diferent between GBM and LGA. Metabolites in play a role in the downstream metabolism of kynurenine, black were not detected or analyzed. Numbers in bracket demonstrate we extended these investigations to both naïve and active fold diference between GBM and LGA. b GBM (n = 56) with known tumor subtype, MGMT methylation, and IDH1 status were clustered human PBMCs. Increased expression of interleukin 2 using prostaglandin pathway metabolites. Two distinct clusters were receptor-subunit alpha (IL2R-α) and TNF-α in activated labeled as PG-High (PG-H) and PG-Low (PG-L). c 56 GBM tumors PBMCs served as a positive control. Although QPRT divided into PG-H and PG-L and used for CIBERSORT analysis for expression was diminished in PBMCs, the remainder of 22 immune subsets. *p < 0.05. d Described cell lines were analyzed for enzymes central to prostaglandin metabolism by western blot. e the enzymes displayed equivalent expression when com- MES orthotopic tumors xenografts in nude (nu/nu) mice (n = 6) were pared to mesenchymal lines. Based on these fndings, both analyzed for using global metabolomic profling and tumor cells and immune cells appear to have a similar compared with LGA tumor metabolite data ability for downstream metabolism of kynurenine. Further studies designed to delineate the intermediary metabolism of kynurenine in individual cell types and the resulting As was noted in studies involving tryptophan metabo- immune and/or non-immune consequences are warranted. lism, it is unclear if metabolomic profles generated from Arginine metabolism. Arginine represents an important patient-derived tumors are refective of the tumor itself or substrate utilized for ornithine and urea production, which that of intra-tumoral immune cells. We, therefore, extended contribute towards M2 macrophage-mediated immune sup- our investigations to preclinical models to better defne these pression [7]. In addition to an accumulation of both ornithine pathways and their relevance to gliomagenesis. As an ini- and urea in GBM, integrative analyses identifed additional tial investigation, we evaluated the expression of enzymes aspects of arginine metabolism to potentially be relevant in involved in arginine metabolism in our preclinical models to gliomagenesis (Fig. 3a). In addition to an accumulation of provide insight into its intermediary metabolism, focusing metabolites central to arginine synthesis, similar to trypto- specifcally on the arginine synthetic and catabolic enzymes phan metabolism, a particularly striking fnding was an accu- ASL and ARG2, respectively. Unlike fndings involving mulation of biochemicals downstream of arginine, including tryptophan metabolism, differential expression was not spermine and proline. Accordingly, aberrant expression of observed between mesenchymal and proneural GBM cell several genes specifc to the metabolism of these biochemi- lines or in PBMCs (Fig. 3d). In addition, other than an cals was observed in GBM, including increased expression accumulation of spermidine and proline in mesenchymal of enzymes associated with arginine synthesis and metabo- cells, pathway activation did not appear to be recapitulated lism, argininosuccinate lyase (ASL) and arginase-2 (ARG2), in preclinical models when metabolomically evaluated respectively, and its further along the proline and in vitro Supplementary Fig. 3. Conversely, accumulation of spermine pathways. metabolites was observed in vivo when grown in an immune Similar to our approach evaluating the kynurenine path- defcient model and compared to LGA (Fig. 3e), further sup- way, we went on to evaluate for heterogeneity of arginine porting the relevance of this pathway in GBM and suggest- metabolism in GBM, define genomic programs driving ing that (1) tumor cells have the capacity to independently this metabolic phenotype, and determine its immune con- metabolize arginine to its downstream intermediaries and (2) sequence. Hierarchical clustering of metabolites involved pathway activation may be stimulated by factors associated in arginine metabolism identifed both arginine ‘high’ and with the unique microenvironment in these tumors. ‘low’ subtypes (Fig. 3b). Interestingly, similar to what was Prostaglandin metabolism. Prostaglandins are derived observed with tryptophan metabolism, mesenchymal and from the metabolism of , which is driven classical molecular subtypes, rather than MGMT methyla- by cyclooxygenase (COX or prostaglandin-endoperoxidase tion or IDH1 mutation status, appeared to display aberrant synthase [PTGS]). Prostaglandin-E2 (PGE2) represents one arginine metabolism in GBM. In addition to upregulation of the most well-studied downstream metabolic intermediar- of the arginine synthetic enzyme ASL, decreased expres- ies of this pathway and a key mediator in immunopathology, sion of enzymes involved in alternate metabolic pathways regulating infammation at many levels [11, 12]. PGE2 has of arginine, including nitric oxide synthase 1 (NOS1) and a paradoxical role of both promoting active infammation mitochondrial 1A (CKMT1), ranked as the but also shifting from an anti-tumor to an immunosuppres- most important genes defning this metabolic pathway (Sup- sive response within the tumor microenvironment. Metabo- plementary Fig. 1b). The immune phenotype of arginine lomic profling uncovered several aspects of prostaglandin metabolism consisted of signifcant increases in M0 mac- metabolism unique to GBM. Although the overexpres- rophages and a trend in Treg populations (Fig. 3c). sion of PTGS2 (COX2) has been described in numerous

1 3 Cancer Immunology, Immunotherapy (2019) 68:1107–1120 1115

a

Prostaglandins DHET GPX1 Thromboxane A GPX3 EPHX2 5-HPETE GPX4 5-HETE 6-Keto-PGF1α GPX7 GPX2 (1.8) TBXA1 (21.2) EET ALOX5 PGF2α (2.8) 8(S)-HPETE 8(S)-HETE CBR1 PTGIS CYP2 CBR3 PTGS1 PTGS1 ALOX12B PGE2 (6.2) PGH2 PGG2 Arachidonate 12(R)-HPETE 12(R)-HETE PTGS2 PTGS2 PTGDS PLA2G6 ALOX12 PTGES PLA2G10 PGD 12(S)-HPETE 12(S)-HETE PGA2 2 PLA2G12 PLA2G1 ALOX15 AKR1C3 GPX1 PLA2G2 GPX3 15(S)-HPETE 15(S)-HETE PLA2G4 GPX4 PGJ2 (7.6) GPX7 11-epi-PGF2α Phospholipids GPX2 b

IDH WT MT MGMT UNMM 3 Subtype MCLN PN 2 d PG-H PG-L 5-HETE 1

6-Keto Prostaglandin 0 ProstaglandinF2al -1

ProstaglandinE2 -2 -3 ProstaglandinJ2

c

40 PG-H PG-L 30 20 * 10 * * r r s s y s g g 2 0 8 4 0 4 g e d g e d d a ls ls ls 4 ry e ll il ll ils n ed n lt e e e la D iv iv t M1 e el el ti lp ytes at at c at cel CD CD ce ph s stin stin sM s sM cu de ator mo CD na na v v v t esti e cc cc sC he T lli tiva oc ro ti ti re ti ls ls s ul re a noph ll ti ls ls o c c n me ma yr ut el el e sr riti m ac l ls a a ac el r eg el l d hage Ma r sf m as y el Mo lls n ndri ll p ls Ne Tc Tc Tc Eosi l s ce c e Tc l mo e ophage o ophage Pl Bc or ga t r De De el ce cr s c cr c Tc me Bc a NK em T Ma M Ma Ma m NK

e LGA MES Xeno

Prostaglandin F2al

6-Keto Prostaglandin

Prostaglandin E2

Prostaglandin J2

3210-1 -2 -3

malignancies, including GBM, an accumulation of its result- GBM, including the accumulation of PGE2, PGF2a, and ant metabolites, PGG2 and PGH2 were not detected in our 6-keto-PGF1a, along with increased expression of their studies (Fig. 4a). However, interestingly, increased activity biosynthetic enzymes (PTGES, CBR1, CBR3, and PTGIS). downstream of this metabolic node was clearly present in Beyond the generation of prostaglandins, additional aspects

1 3

1116 Cancer Immunology, Immunotherapy (2019) 68:1107–1120 ◂ of arachidonic acid metabolism were also uncovered through Fig. 5 Adenosine metabolism infuences the induction of M2-like these integrative analyses, including metabolic programs macrophages in GBM. a Schematic of adenosine metabolism. Red indicates metabolites upregulated in GBM when compared to LGA. designed to generate this biochemical through phospholipid Green indicates metabolites downregulated in GBM; brown indicates metabolism and alternate modes of its subsequent metabo- that the metabolite was analyzed but was not signifcantly diferent lism through the eicosanoid pathway. between GBM and LGA. Metabolites in black were not detected. Unlike studies involving both tryptophan and arginine Numbers in bracket demonstrate fold diference between GBM/LGA. b GBM (n = 56) with known tumor subtype, MGMT methylation, phenotypes, metabolites associated with prostaglandin and IDH1 status were clustered using the top 5 expressed adenosine metabolism were unevenly distributed, with only a small pathway metabolites. Two distinct clusters were defned as AD-High proportion of tumors (14/56) defned as ‘high’ and integra- (AD-H) and AD-Low (AD-L). c GBM (n = 56) tumors were defned tive analyses did not identify specifc molecular subtypes or as AD-H and AD-L and their expression profles were analyzed by CIBERSORT to defne immune subsets. d Murine bone-marrow- genetic programs (Fig. 4b, Supplementary Fig. 1C) that may derived macrophages were polarized to the M2 phenotype and ana- be driving this phenotype or clear changes in the immune lyzed for CD11b and F4/80. This experiment has been replicated four landscape relative to its metabolism (Fig. 4c). We went on to times with similar results. *p < 0.05 determine if these fndings were recapitulated in preclinical models. In addition to validating increased expression of PTGIS in mesenchymal lines and established human GBM Discussion cell lines, we also uncovered an unexpected correlation with the expression levels of PTGS1 and PTGS2. Specifcally, This study represents one of the frst of its kind to perform increased expression of PTGIS appeared to be consistent integrative analyses coupling metabolomics and expres- with increased PTGS1 expression, yet these were inversely sion profling to comprehensively defne the interface related to PTGS2 expression (Fig. 4d). Therefore, the inter- between metabolic reprogramming and immune response play of PTGS1 and PTGS2 and their role in the potential in GBM. Through these studies, both tryptophan and diferential metabolism of prostaglandins in GBM is worthy arginine metabolism emerged as metabolic pathways of further investigation. In addition, fndings were supported particularly relevant in gliomagenesis. Numerous studies by metabolomic profling of mesenchymal GBM cells grown have identifed adenosine as a potent anti-infammatory in vivo, demonstrating accumulation of prostaglandin E2, J2, molecule involved in the restoration of tissue homeosta- and 6-ketoprostaglandin F1-α (Fig. 4e). sis through the modulation of the innate and adaptive Adenosine metabolism. Although we did not demonstrate immune response [34, 35]. Extracellular adenosine levels an accumulation of adenosine in GBM when compared to are typically very low; however, necrotic tumors generate LGA (Fig. 5a), considerable heterogeneity was observed. high levels of extracellular adenosine, which is metabo- We, therefore, further evaluated differential adenosine lized through the dynamic and sequential actions of cell- metabolism in GBM and its possible immune consequence. surface enzymes (ectoenzymes), thereby contributing to Approximately 2/3 of the tumors were designated as adeno- an immune-suppressive state [36–38]. Surprisingly, these sine “high” and this metabolic phenotype did not appear studies did not identify the accumulation of the potent to be specifc to established molecular subtypes (Fig. 5b). immune-suppressing metabolite adenosine in GBM when However, adenosine “high” tumors did appear to confer an compared to LGA. Moreover, this also appears to contra- immune-suppressive phenotype, consisting of an accumula- dict a recent report, suggesting that this pathway is acti- tion of M2 macrophages when analyzed by CIBERSORT vated, including the expression of the key ectoenzymes (Fig. 5c). To functionally extend fndings linking adenosine associated with its formation, and targetable in GBM [39]. metabolism with the accumulation of M2 macrophages in As our integrated analyses only provide a static metabolic GBM, we sought to determine if this metabolic pathway picture, one possibility is that this metabolite is rapidly contributed towards M2 polarization and functional sup- metabolized following activation. Therefore, further work pression. Murine bone-marrow-derived macrophages dem- designed to carefully defne the intermediary metabolism onstrated both an increase in M2 polarization (Fig. 5d) and of this pathway is required to better determine its potential functional suppression of CD8 cells (Supplementary Fig. 4) role in gliomagenesis. when cultured in the presence of adenosine. However, Gene-expression profiles identified numerous genes similar to what was observed in tumors, an accumulation that were signifcantly either up- or downregulated within of adenosine was not observed in our preclinical models each metabolic pathway (Fig. 1b, Supplementary Table 1). in vitro or in vivo Supplementary Fig. 5A/B. Therefore, Although the expression of many of these genes still requires further investigation is required to better determine if this validation when evaluated in the context of their metabolic metabolic pathway plays a role in gliomagenesis. profles, these fndings provide a window into potential mechanisms contributing towards individual phenotypes. For example, decreased expression of a specifc enzyme

1 3 Cancer Immunology, Immunotherapy (2019) 68:1107–1120 1117

a Ribosylation ADP-ribose (0.68) APRT AK9 PNP NT5C1B AK1, 4-9 NME1-6 ADCY1 Adenine (0.85) Adenosine (0.83) AMP ADP ATP cAMP (0.32) -RDH14 AK3 PKLR, PKM PDE7B PDE10A 2’3’ -cyclic AMP 3’-AMP (1.72) ATP metabolism

b IDH WT MT MGMT UNMM Subtype MCLN PN AD-L AD-H ADP 2’-AMP

Adenosine ADP-ribose

AMP 420-2-4

c ) 45 AD-H (% AD-L * tion

ac 30 Fr ll Ce 15 ve ti la Re

0 r y s s s s s g g 2 0 8 e a e d d l l a l il es v ed lt ng e ng e l D4 t D4 i iv in or t i M1 t per el h t a ated a c at ce ll CD C ce s p sM sM m de cu CD4 na na i v iv iv l es t i esti iv e es e sC s hel t rophil ic t ic re restin ls l ll t t g g nocy ct ls ol i me ma f sr el el sr ri ag ma sino o ac ac t ac ells ry el ll ce c c s d dr ha regulatory l ls sa as M y n p ph l ls Neut T Eo T T am ce ll ce el Tc ls en el o o Bc Pl or el el g r D De el c st c cr cropha Bc tc ls Tc memo a a NK l em e Tc M Ma Ma M as NK m M Tc d M2 M2+Adenosine Isotypes b CD 11

F4/80

may drive metabolism towards a more ‘oncogenic’ upstream of an individual enzyme were identifed to be diferentially or parallel pathway. Furthermore, particularly relevant to expressed. These fndings may stimulate further investi- prostaglandin metabolism in this study, several isoforms gations designed to determine the afnity of individual

1 3 1118 Cancer Immunology, Immunotherapy (2019) 68:1107–1120 isoforms to a given substrate and how this may lead to meta- immune-cell population. M2 macrophages are anti-infam- bolic reprogramming. matory, cytokine producing macrophages and aid in tumor One notable fnding these integrative analyses ofered and progression through immune suppression. was that although many of the observed immuno-metabolic This suggests that concerted eforts designed to target these phenotypes appeared to be independent of MGMT promoter immune-suppressive cells may revert the potent immune methylation and IDH1 mutation status, two of the strongest tolerance observed in this malignancy. Surprisingly, ARG- prognostic factors in GBM, they were enriched in the mesen- high tumors also demonstrated very similar results, with a chymal and classical transcriptional subtypes. Interestingly, reduction in both plasma cells and NK cells. Tumors with rather than representing inter-tumoral heterogeneity, we have increased tryptophan and/or arginine metabolism appeared recently demonstrated that these molecular subtypes refect to have higher levels of resting M0 macrophages; however, intra-tumoral heterogeneity in GBM, with mesenchymal and the immune consequence of this fnding remains unclear. classical subtypes enriched in perinecrotic regions within an Although an accumulation of adenosine did not appear to individual tumor [40]. We, therefore, hypothesize that the diferentiate GBM from LGA, the subtype of increased observed immuno-metabolic subtypes in GBM are a direct adenosine metabolism within GBM was the only identi- consequence of the diverse tumor microenvironment in this fed metabolic program associated with increased M2 mac- heterogeneous malignancy. rophages. These fndings support a potential immune conse- A common theme that emerged from these comprehen- quence of this metabolic pathway and, again, reinforce more sive metabolomic studies was an enrichment of numerous focused investigations to better understand its intermediary biochemicals’ downstream of the established metabolic metabolism and immune consequence. pathways in GBM. For example, although tryptophan In summary, these comprehensive, integrative analyses metabolism and kynurenine accumulation have been previ- provide insight into how metabolic remodeling contributes ously described in GBM, this study represents one of the towards an immune-suppressive phenotype in GBM. As our frst to delineate the potential relevance of its intermediary understanding of how tumor-specifc metabolic programs metabolism, including an accumulation downstream metab- contribute to immune suppression grows, this database can olites involved in de novo NAD biosynthesis and salvage. be utilized to determine their role in gliomagenesis. Fur- Downstream of ornithine and urea, proline, and spermine thermore, by extending these investigations into preclinical metabolism was enriched in the arginine pathway. In addi- models, this work provides the framework for future stud- tion, eicosanoid metabolism, along with several other bio- ies designed to defne the intermediary metabolism of these chemicals’ downstream of prostaglandins, accumulated in pathways in further detail, evaluate their role in immune GBM that appeared to be recapitulated in preclinical models. tolerance, and determine their therapeutic implications in This suggests that targeting this metabolic pathway may be this aggressive malignancy. extended beyond COX inhibitors, including studies designed to test the novel targets PTGIS and PTGS1. Therefore, these fndings provide the framework for a series of investigations Author contributions Study design: PK and PC; experiments: PK, AP, and SK; data analysis: PK and PC; reagents: PC; and manuscript designed to better defne the intermediary metabolism of preparation: PK and PC. these individual pathways and their potential to be targeted. Another novel aspect of these studies is that it allows cou- Funding This work was supported by the National Institute of Health pling of metabolomic signatures with CIBERSORT analy- (NIH)/National Institute of Neurological Disorders and Stroke ses of gene-expression profles, which defnes immune-cell (NINDS) (R21NS090087), American Cancer Society (RSG-11-029- 01), Bankhead-Coley Cancer Research Program and Cancer Research populations within an individual tumor, providing a previ- Seed Grant Awards from Beaumont Health to Prakash Chinnaiyan. ously undescribed window into the immune consequences of a given metabolic phenotype. Of the metabolic pathways Compliance with ethical standards evaluated, tryptophan metabolism appeared to play the most dominant role in immune tolerance in GBM, with over a Conflict of interest The authors declare that they have no potential twofold increase in Tregs and a strong trend in decreased confict of interest. ­CD8+ T cells. These tumors also demonstrated a signifcant Ethical approval GBM/glioma tissue samples were obtained from the reduction in natural killer cells (NK cells), which are impor- Moftt Cancer Center Tissue Core Facility. Institutional Review Board/ tant for innate immune response. These results suggest that Human Subjects approval (MCC16197) was obtained for this retro- an accumulation of tryptophan-related metabolites inhibits spective study from the ethics committee of the Moftt Cancer Center. both innate and acquired response resulting in mitigating All animal studies were carried out under protocols approved by the IACUC (AL-16-09 and AL-18-10) at William Beaumont Research anti-tumor immunity. Collectively, however, the immune Institute. landscape of GBM appears to be primarily influenced by M2 macrophages, which represent nearly 40% of the

1 3 Cancer Immunology, Immunotherapy (2019) 68:1107–1120 1119

Informed consent Patients gave written informed consent for the use of 12. Wang D, DuBois RN (2016) The role of prostaglandin E(2) in their specimens and clinical data for research and publication prior to tumor-associated immunosuppression. Trends Mol Med 22:1–3. surgical resection or following diagnosis under the Total Cancer Care https​://doi.org/10.1016/j.molme​d.2015.11.003 Tissue Repository program at the Moftt Cancer Center. 13. Deaglio S, Dwyer KM, Gao W et al (2007) Adenosine genera- tion catalyzed by CD39 and CD73 expressed on regulatory T Animal source C57BL/6 (H-2b, CD45.2) and athymic nu/nu (NU- cells mediates immune suppression. J Exp Med 204:1257–1265. nu Foxn1 ) mice were purchased from Charles River Laboratories (Wilm- https​://doi.org/10.1084/jem.20062​512 ington, MA). 14. Chinnaiyan P, Kensicki E, Bloom G et al (2012) The metab- olomic signature of malignant glioma reflects accelerated Cell line authentication Human GBM cell lines U87, U251, and T98G anabolic metabolism. Cancer Res 72:5878–5888. https​://doi. were purchased from ATCC and authenticated by STR analysis at org/10.1158/0008-5472.CAN-12-1572-T The University of Arizona Genetics Core. MES83, MES326, PN19, 15. Evans AM, DeHaven CD, Barrett T, Mitchell M, Milgram E and PN84 were generated, obtained, and authenticated by Dr. Ichiro (2009) Integrated, nontargeted ultrahigh performance liquid Nakano’s lab at Ohio State University. Murine TRP cell lines were chromatography/electrospray ionization tandem mass spectrom- generated, obtained, and authenticated by Dr. C. Ryan Miller’s lab at etry platform for the identifcation and relative quantifcation The University of North Carolina School of Medicine. of the small-molecule complement of biological systems. Anal Chem 81:6656–6667. https​://doi.org/10.1021/ac901​536h 16. Prabhu AH, Kant S, Kesarwani P, Ahmed K, Forsyth P, Nakano I, Chinnaiyan P (2018) Integrative cross-platform analyses iden- References tify enhanced heterotrophy as a metabolic hallmark in glioblas- toma. Neuro Oncol. https​://doi.org/10.1093/neuon​c/noy18​5 17. Prabhu A, Sarcar B, Kahali S, Yuan Z, Johnson JJ, Adam 1. Ostrom QT, Gittleman H, Liao P, Vecchione-Koval T, Wolinsky KP, Kensicki E, Chinnaiyan P (2014) catabolism: a Y, Kruchko C, Barnholtz-Sloan JS (2017) CBTRUS statistical novel metabolic pathway contributing to glioblastoma growth. report: primary brain and other central nervous system tumors Cancer Res 74:787–796. https​://doi.org/10.1158/0008-5472. diagnosed in the United States in 2010-2014. Neuro Oncol 19:v1– CAN-13-1423 v88. https​://doi.org/10.1093/neuon​c/nox15​8 18. Mao P, Joshi K, Li J et al (2013) Mesenchymal glioma stem cells 2. Biswas SK (2015) Metabolic reprogramming of immune cells are maintained by activated glycolytic metabolism involving alde- in cancer progression. Immunity 43:435–449. https​://doi. hyde dehydrogenase 1A3. Proc Natl Acad Sci USA 110:8644– org/10.1016/j.immun​i.2015.09.001 8649. https​://doi.org/10.1073/pnas.12214​78110​ 3. Larkin J, Chiarion-Sileni V, Gonzalez R et al (2015) Combined 19. Vitucci M, Karpinich NO, Bash RE et al (2013) nivolumab and ipilimumab or monotherapy in untreated mela- between MAPK and PI3 K signaling activation is required for noma. N Engl J Med 373:23–34. https​://doi.org/10.1056/NEJMo​ glioblastoma pathogenesis. Neuro Oncol 15:1317–1329. https​:// a1504​030 doi.org/10.1093/neuon​c/not08​4 4. Pardoll DM (2012) The blockade of immune checkpoints in can- 20. Braun D, Longman RS, Albert ML (2005) A two-step induction of cer immunotherapy. Nat Rev Cancer 12:252–264. https​://doi. indoleamine 2,3 dioxygenase (IDO) activity during dendritic-cell org/10.1038/nrc32​39 maturation. Blood 106:2375–2381. https://doi.org/10.1182/blood​ ​ 5. Reardon DA, Omuro A, Brandes AA et al (2017) OS10.3 ran- -2005-03-0979 domized phase 3 study evaluating the efficacy and safety of 21. Grant RS, Naif H, Thuruthyil SJ, Nasr N, Littlejohn T, Takikawa nivolumab vs bevacizumab in patients with recurrent glioblas- O, Kapoor V (2000) Induction of indolamine 2,3-dioxygenase in toma: CheckMate 143. Neuro-Oncology 19:iii21. https​://doi. primary human macrophages by human immunodefciency virus org/10.1093/neuon​c/nox03​6.071 type 1 is strain dependent. J Virol 74:4110–4115 6. Buck MD, Sowell RT, Kaech SM, Pearce EL (2017) Meta- 22. Newman AM, Liu CL, Green MR, Gentles AJ, Feng W, Xu Y, bolic instruction of immunity. Cell 169:570–586. https​://doi. Hoang CD, Diehn M, Alizadeh AA (2015) Robust enumeration of org/10.1016/j.cell.2017.04.004 cell subsets from tissue expression profles. Nat Methods 12:453– 7. Kesarwani P, Kant S, Prabhu A, Chinnaiyan P (2017) The 457. https​://doi.org/10.1038/nmeth​.3337 interplay between metabolic remodeling and immune regula- 23. Munn DH, Sharma MD, Hou D et al (2004) Expression of tion in glioblastoma. Neuro Oncol 19:1308–1315. https​://doi. indoleamine 2,3-dioxygenase by plasmacytoid dendritic cells in org/10.1093/neuon​c/nox07​9 tumor-draining lymph nodes. J Clin Invest 114:280–290. https​:// 8. Kesarwani P, Prabhu A, Kant S, Kumar P, Graham SF, Buelow doi.org/10.1172/JCI21​583 KL, Wilson GD, Miller CR, Chinnaiyan P (2018) Tryptophan 24. Munn DH, Zhou M, Attwood JT, Bondarev I, Conway SJ, Mar- metabolism contributes to radiation-induced immune checkpoint shall B, Brown C, Mellor AL (1998) Prevention of allogeneic fetal reactivation in glioblastoma. Clin Cancer Res 24:3632–3643. https​ rejection by tryptophan catabolism. Science 281:1191–1193 ://doi.org/10.1158/1078-0432.CCR-18-0041 25. van Baren N, Van den Eynde BJ (2015) Tryptophan-degrading 9. Wainwright DA, Balyasnikova IV, Chang AL, Ahmed AU, Moon enzymes in tumoral immune resistance. Front Immunol 6:34. https​ KS, Aufnger B, Tobias AL, Han Y, Lesniak MS (2012) IDO ://doi.org/10.3389/fmmu​.2015.00034​ expression in brain tumors increases the recruitment of regula- 26. Opitz CA, Litzenburger UM, Sahm F et al (2011) An endogenous tory T cells and negatively impacts survival. Clin Cancer Res tumour-promoting of the human aryl hydrocarbon receptor. 18:6110–6121. https://doi.org/10.1158/1078-0432.CCR-12-2130​ Nature 478:197–203. https​://doi.org/10.1038/natur​e1049​1 10. Rath M, Muller I, Kropf P, Closs EI, Munder M (2014) Metabo- 27. Hubbard TD, Murray IA, Bisson WH, Lahoti TS, Gowda K, Amin lism via arginase or nitric oxide synthase: two competing argi- SG, Patterson AD, Perdew GH (2015) Adaptation of the human nine pathways in macrophages. Front Immunol 5:532. https​:// aryl hydrocarbon receptor to sense microbiota-derived indoles. doi.org/10.3389/fmmu​.2014.00532​ Sci Rep 5:12689. https​://doi.org/10.1038/srep1​2689 11. Kalinski P (2012) Regulation of immune responses by pros- 28. Dang L, White DW, Gross S et al (2009) Cancer-associated IDH1 taglandin E2. J Immunol 188:21–28. https​://doi.org/10.4049/ mutations produce 2-hydroxyglutarate. Nature 462:739–744. https​ jimmu​nol.11010​29 ://doi.org/10.1038/natur​e0861​7

1 3 1120 Cancer Immunology, Immunotherapy (2019) 68:1107–1120

29. Hegi ME, Diserens AC, Gorlia T et al (2005) MGMT gene silenc- 36. de Lourdes Mora-Garcia M, Garcia-Rocha R, Morales-Ramirez ing and beneft from temozolomide in glioblastoma. N Engl J Med O et al (2016) Mesenchymal stromal cells derived from cervi- 352:997–1003. https​://doi.org/10.1056/NEJMo​a0433​31 cal cancer produce high amounts of adenosine to suppress cyto- 30. Verhaak RG, Hoadley KA, Purdom E et al (2010) Integrated toxic T lymphocyte functions. J Transl Med 14:302. https​://doi. genomic analysis identifies clinically relevant subtypes of org/10.1186/s1296​7-016-1057-8 glioblastoma characterized by abnormalities in PDGFRA, 37. Leone RD, Lo YC, Powell JD (2015) A2aR antagonists: Next IDH1, EGFR, and NF1. Cancer Cell 17:98–110. https​://doi. generation checkpoint blockade for cancer immunotherapy. Com- org/10.1016/j.ccr.2009.12.020 put Struct Biotechnol J 13:265–272. https​://doi.org/10.1016/j. 31. Prabhu A, Sarcar B, Miller CR, Kim SH, Nakano I, Forsyth P, csbj.2015.03.008 Chinnaiyan P (2015) Ras-mediated modulation of pyruvate dehy- 38. Saldanha-Araujo F, Ferreira FI, Palma PV, Araujo AG, Queiroz drogenase activity regulates mitochondrial reserve capacity and RH, Covas DT, Zago MA, Panepucci RA (2011) Mesenchymal contributes to glioblastoma tumorigenesis. Neuro Oncol 17:1220– stromal cells up-regulate CD39 and increase adenosine produc- 1230. https​://doi.org/10.1093/neuon​c/nou36​9 tion to suppress activated T-lymphocytes. Stem Cell Res 7:66–74. 32. Heyes MP, Achim CL, Wiley CA, Major EO, Saito K, Markey SP https​://doi.org/10.1016/j.scr.2011.04.001 (1996) Human microglia convert l-tryptophan into the neurotoxin 39. Xu S, Shao QQ, Sun JT et al (2013) Synergy between the ecto- quinolinic acid. Biochem J 320(Pt 2):595–597 enzymes CD39 and CD73 contributes to adenosinergic immuno- 33. Sahm F, Oezen I, Opitz CA et al (2013) The endogenous tryp- suppression in human malignant gliomas. Neuro Oncol 15:1160– tophan metabolite and NAD + precursor quinolinic acid confers 1172. https​://doi.org/10.1093/neuon​c/not06​7 resistance of gliomas to . Cancer Res 73:3225– 40. Prabhu A, Kesarwani P, Kant S, Graham SF, Chinnaiyan P (2017) 3234. https​://doi.org/10.1158/0008-5472.CAN-12-3831 Histologically defned intratumoral sequencing uncovers evolu- 34. Ohta A, Sitkovsky M (2001) Role of G-protein-coupled aden- tionary cues into conserved molecular events driving gliomagen- osine receptors in downregulation of and pro- esis. Neuro Oncol 19:1599–1606. https​://doi.org/10.1093/neuon​ tection from tissue damage. Nature 414:916–920. https​://doi. c/nox10​0 org/10.1038/41491​6a 35. Thiel M, Caldwell CC, Sitkovsky MV (2003) The critical role of Publisher’s Note Springer Nature remains neutral with regard to adenosine A2A receptors in downregulation of infammation and jurisdictional claims in published maps and institutional afliations. immunity in the pathogenesis of infectious diseases. Microbes Infect 5:515–526

1 3