Better understanding of pancreatic diseases, including pancreatic ductal adenocarcinoma (PDAC), is an urgent medical need, with little advances in preoperative differential diagnosis, preventing rational selection of therapeutic strategies. The clinical management of pancreatic cancer patients would benefit from the identification of variables distinctively associated with the multiplicity of pancreatic disorders. We investigated, by 1H nuclear magnetic resonance, the metabolomic fingerprint of pancreatic juice (the biofluid that collects pancreatic products) in 40 patients with different pancreatic diseases. Metabolic variables discriminated PDAC from other less aggressive pancreatic diseases and identified metabolic clusters of patients with distinct clinical behaviors. PDAC specimens were overtly glycolytic, with significant accumulation of lactate, which was probed as a disease-specific variable in pancreatic juice from a larger cohort of 106 patients. In human PDAC sections, high expression of the glucose transporter GLUT-1 correlated with tumor grade and a higher density of PD-1+ T cells, suggesting their accumulation in glycolytic tumors. In a preclinical model, PD-1+ CD8 tumor–infiltrating lymphocytes differentially infiltrated PDAC tumors obtained from cell lines with different metabolic consumption, and tumors metabolically rewired by knocking down the phosphofructokinase (Pfkm) gene displayed a decrease in PD-1+ cell infiltration. Collectively, we introduced pancreatic juice as a valuable source of metabolic variables that could contribute to differential diagnosis. The correlation of metabolic markers with immune infiltration suggests that upfront evaluation of the metabolic profile of PDAC patients could foster the introduction of immunotherapeutic approaches for pancreatic cancer.

Pancreatic ductal adenocarcinoma (PDAC) is the fourth leading cause of death and projected to be the second within a decade (1–3). Although surgical resection is the only curative option, PDAC is affected by a high frequency of early relapse events and postsurgical complications. Preoperative diagnostic tools that improve patient profiling could limit unnecessary surgery in rapidly evolving diseases and could promote the use of neoadjuvant therapies. Current diagnostic strategies are not always adequately efficient in discriminating between PDAC and other benign pancreatobiliary diseases with overlapping symptoms for which surgical resection may not be necessary. The complex and challenging process of surgical decision-making would benefit from the identification of variables distinctively associated with the multiplicity of pancreatic disorders. So far, considerable advances toward a better definition of pancreatic cancer profiles have been primarily achieved through genomic analyses (4–6). However, pancreatic cancer cells are renowned for their extensively rewired metabolism, largely driven by intrinsic oncogenic events and by interactions with noncancerous stromal cells (7). Metabolic reprogramming is considered a hallmark of cancer progression (8–10), and its analysis could represent a tool to identify novel disease-specific biomarkers. Previously, global analysis of metabolites by metabolomic platforms performed on body fluids has emerged as an information-rich tool that can ultimately reveal indicators of perturbed biological systems and disease-specific profiles. Blood and urine metabolic variables can discriminate between chronic pancreatitis and PDAC (11–13), suggesting an important connection between tumor metabolic alterations and diagnosis.

The pathways that regulate tumor metabolism and immune cell function are tightly linked (14–16). In homeostatic conditions, correct functionality of T cells requires a finely regulated metabolic network that provides energy and biosynthetic precursors (17). As a result, the metabolic dysfunction frequently observed in pathologic states can affect the efficacy of the immune response. Nutrient deprivation is commonly present in tumor settings, suggesting that metabolic microenvironmental cues could impact the mounting and implementation of an effective antitumor immune response (9, 18). In pancreatic cancer, PDAC cells escape immune recognition and orchestrate an immunosuppressive microenvironment by various mechanisms, including limited immunogenicity, inadequate immune cell infiltration, and myeloid immune regulatory networks (19–21). Surprisingly, despite the well-known metabolic derangements that characterize pancreatic cancer (22), the influence of metabolic cues on immune cell function has not been addressed. Notwithstanding the compromised nature of immune cells in PDAC, this type of cancer has not emerged at the forefront of immunotherapeutic strategies. In particular, although checkpoint inhibitors represent an important breakthrough in the field of immunotherapy, they still have not been proven effective in PDAC patients (19, 23), suggesting that a better comprehension of the immune-metabolic microenvironment could improve therapeutic strategies.

The immune microenvironment of human pancreatic cancer has been the object of intense investigations (24–28). Along this line, we have previously reported the clinical relevance of immune cells as potential novel biomarkers of prognosis (24) and response to therapy (25) in pancreatic cancer patients. In an attempt to characterize environmental cues affecting immune cell function, in this study we have investigated the metabolic profile and how it relates to immune infiltration in patients with pancreatic pathologies. In order to unearth early metabolic alterations, we analyzed pancreatic juice, a relatively unexplored body fluid collected during surgical procedures and enriched with pancreatic tumor–derived metabolites. Our results identified discriminative metabolic variables that allowed us to single out PDAC from other pancreatic pathologies, which were associated with distinct clinical behaviors. Also, we revealed a link between glucose metabolism and accumulation of PD-1+ cells in pancreatic tumors and provide preclinical evidence that metabolic rewiring of cancer cells results in modification of the immune contexture. Collectively, in this work we uncovered the diagnostic and prognostic relevance of measurable metabolic parameters in pancreatic juice.

Human samples and study design

The study population included two cohorts of patients (referred to as cohorts 1 and 2), who underwent surgery at the Humanitas Clinical and Research Center from 2015 to 2018. All patients were aged more than 18 and nonmetastatic, except for 1 patient for whom a hepatic metastasis was found intraoperatively. For all patients enrolled, peripheral blood, tumor tissue, and adjacent normal tissue were collected at the time of surgery according to protocols approved by the Ethical Committee of the Institution. Cohort 1 comprised 40 patients, from which pancreatic juice was withdrawn during pancreasectomy and used for metabolomics analysis (Supplementary Table S1). Patients were diagnosed postoperatively with pancreatic adenocarcinoma (n = 31), pancreatitis (n = 2), papillary-ampulla tumors (n = 4), neuroendocrine tumors (n = 2), IPMN (n = 1; Supplementary Table S1). Cohort 2 comprised 106 patients (inclusive of cohort 1), including 79 PDAC patients and 27 non-PDAC patients, from which pancreatic juice was withdrawn during pancreasectomy and lactate measured by colorimetric assay (Supplementary Table S2). Clinicians prospectively assembled a clinical retrospective database by collecting patient demographics and clinical and histopathologic data, as detailed in Supplementary Tables S1 and S2. T (tumor) and G (grade) were categorized according to the 7th edition of the TNM staging system as follows: T1–T2 as low-T tumors, T3 as high-T tumors; G1–G2 as low-grade tumors and G3–G4 as high-grade tumors.

CD8+ T cells were obtained from buffy coats of healthy donors (n = 3) obtained from the blood bank of the institution. All the patients were enrolled in the study after signed informed consent including collection of biological specimens and clinical data. The study was approved by the Ethical Committee of the Institution (protocol number ICH-595, approval issued on May 2009).

Plasma and pancreatic juice collection and preparation

Blood was collected at the time of surgery in EDTA-containing BD Vacutainer vials (BD Biosciences, 366473). Plasma was then obtained via centrifugation and stored at −80°C until further use. Pancreatic juice was collected intraoperatively by Wirsung puncture immediately before pancreas transection and stored in EDTA-containing BD Vacutainer vials (BD Biosciences, 366473) at 4°C until transportation to the laboratory of the Biobank Unit within the Hospital for processing. Once in the laboratory, the sample was centrifuged within 5 hours of collection, aliquoted, and stored at −80°C. Patients fasted overnight prior to collection.

Nuclear magnetic resonance spectroscopy

Nuclear magnetic resonance (NMR) spectra of pancreatic juice and plasma samples were acquired using a Bruker IVDr 600 MHz spectrometer (Bruker BioSpin), operating at 600.13-MHz proton Larmor frequency, equipped with a 5-mm PATXI H/C/N with 2H-decoupling probe, including a z-axis gradient coil, an automatic tuning-matching, and an automatic refrigerated sample changer (SampleJet). Pancreatic juice (180 μL) was mixed with 20 μL of D2O buffer [1.5 M KH2PO4 dissolved in 99.9% D2O, pH 7.4; 2 mmol/L sodium azide; and 0.1% 3-(trimethyl-silyl) propionic acid-d4 (TSP)]. A total of 200 μL of this mixture was transferred into a 3-mm NMR tube (Bruker BioSpin) for the analysis. Plasma samples were prepared as previously described (29). Temperature was regulated to 300 ± 0.1 K for pancreatic juice and 310 ± 0.1 K for plasma, with a BTO 2000 thermocouple. Both one-dimensional (1D) Nuclear Overhauser Effect Spectroscopy (NOESY) and 1D Carr-Purcell-Meiboom-Gill (CPMG) spectra (30) were collected for plasma and pancreatic juice samples, whereas 1D-diffusion experiment was additionally acquired for plasma samples. Both pancreatic juice and plasma spectra were bucketed [0.02 parts per million (ppm) buckets width] by AMIX software (Bruker Biospin). After the OPLS-DA (orthogonal projection latent structure discriminant analysis), the produced latent variables plot (LV) components and the estimated variable importance in the projection (VIP) score of each group (variables with VIP > 1.0 were initially considered significant for each model) were used for the detection of the weighted variables (NMR signals–buckets) responsible for the group classification. These signals corresponded to the metabolites reported in the Results section. The statistical significance (P) of each metabolite was calculated by univariate one-way ANOVA analysis refined by an n-fold cross-validation implementation by splitting the data into n sub-data sets while performing the ANOVA test (e.g., cross-validated ANOVA). Metabolite signals presenting a strong overlapping with EDTA were removed before multivariate statistical analysis.

Multivariate statistical analysis applied to metabolite quantification by NMR

All multivariate and univariate analyses of the NMR-based metabolomics study were performed by PLS_Toolbox 8.2.1 and homemade scripts in Matlab 2016a. Bucketed NMR data were normalized by total area normalization for both plasma and pancreatic juice spectra. Total area normalization is a “mild” normalization method and proved suitable for our data because no extreme outliers [identified as values with high distance to the model (DModX) or values outside the 95% confidence interval (CI) of the multivariate analysis] were observed in any of our multivariate analyses (31). Deconvolution and integration of metabolite signals were accomplished by an in-house developed software in MATLAB programming suited by Dr. P.G. Takis. The algorithm is based upon the unconstrained nonlinear minimization (fitting) of the metabolite NMR signals, using a combination of Lorentzian–Gaussian functions. By this approach, each NMR region of interest is decomposed and deconvoluted into its component parts, and then integrated to obtain the metabolite concentrations in arbitrary units. For the n × fold concentration changes for each metabolite, the following equation was used:

formula

Lactate and glucose measurement

Concentration of lactate in pancreatic juice was assessed by a colorimetric assay (Sigma-Aldrich, MAK064), following the manufacturer's instruction, without prior pretreatment of samples. Samples were diluted 10-fold or 25-fold and concentration was calculated on standard curve. Concentration of glucose in interstitial fluid was measured by a colorimetric assay (Sigma-Aldrich, MAK263), following the manufacturer's instructions. Samples were diluted 10-fold or 30-fold and concentration was calculated on standard curve.

IHC

Human FFPE-PDAC tissues from 18 patients included in cohort 2 were provided as tissue blocks, by the Pathology Department of the Humanitas Clinical and Research Center. Consecutive tissue sections (2 μm thick) were obtained from each specimen and stained as previously described (24). In brief, sections (one section for each marker) were stained with primary antibodies raised against CD8 (DAKO, clone C8/144B, diluted 1:100), PD-1 (Abcam, clone NAT105, diluted 1:50), GLUT-1 (Abcam, clone EPR3015, diluted 1:2,000), CD68 (DAKO, KP-1 clone, diluted 1:1,000), and CD20 (DAKO, L26, diluted 1:200). Investigators who performed the assessment of immune variables were blinded to the clinical data.

Quantification of the immune-reactive area

After staining procedure, tissue sections were digitalized using the slide scanner VS120 dotSlide (Olympus). To evaluate the percentage of immune-reactive area (IRA%) of immune cells, at least two independent operators blinded to any patient clinical data selected three nonoverlapping and noncontiguous areas comprising of approximately 50% of tumor and 50% of stromal tissue. Evaluation of GLUT-1 was performed by selecting three pictures inclusive of pancreatic ducts and corresponding to the hotspot regions. Both sampled microscopic area and light density were maintained throughout the analysis. Selected areas were quantified by computer-assisted image analysis, with ad hoc software (Image Pro Premiere), to obtain the IRA% of the digitalized tissue surface. The mean value, obtained from the three different regions, was calculated for each marker and subsequently used for analyses.

Cell lines

Human BxPC3 were purchased from ATCC in 2007, and PT45 cell lines were kindly provided by Prof. Lorenzo Piemonti (San Raffaele Diabetes Research Institute, Milan, Italy) in 2001. Neither was authenticated. Cells were cultured in RPMI-1640 medium (Lonza) supplemented with 10% FBS and 2 mmol/L L-glutamine in a humidified incubator with 5% CO2. Cells were routinely tested for Mycoplasma contamination. For supernatant collection, 5 × 106 cells were seeded with fresh medium, and medium was collected after 24 hours, centrifuged to remove debris, and stored at −80°C until further use. The murine Panc02 cell line (32) was obtained from Prof. Lorenzo Piemonti (San Raffaele Diabetes Institute, Milan, Italy) as previously described (24) and DT6606 from Prof. Francesco Novelli (Center for Experimental Research and Medical Studies, Torino, Italy) as described (33). Briefly, DT6606 is a cell line obtained from the LSL-KrasG12D-Pdx1-Cre mouse, a genetically modified mouse developing pancreatic cancer due to conditional expression of the RasG12D mutation under the Pdx pancreatic-specific promoter. Cells were cultured in RPMI-1640 medium (Lonza) supplemented with 10% FBS and 2 mmol/L L-glutamine in a humidified incubator with 5% CO2 for one passage before in vivo injection. Cells were routinely tested for Mycoplasma contamination.

Knockdown of Pfkm in Panc02 cells

Stable genetic inhibition of glycolysis in Panc02 cells was achieved by silencing the phoshofructokinase gene (Pfkm) with a lentiviral vector coding for a short hairpin RNA (shPFK; Sigma-Aldrich, SHCLNV-NM_021514). A scramble shRNA (shSCR) was used as control (Sigma-Aldrich, SHC016V). For transduction, Panc02 cells were cultured to 60% to 80% confluence in Dulbecco Modified Eagle's Medium (Lonza) with 10% FBS and 2 mmol/L L-glutamine in a humidified incubator with 5% CO2. Transduced cells were selected by addition of puromycin (1 μg/mL) to the medium 72 hours after infection. Gene silencing was verified by real-time qPCR.

Glycolytic rate assay

Glycolytic rates of PT45 and BxPC3 cells were assessed using a commercially available kit (Agilent Seahorse XF Glycolytic Rate Assay Kit, Agilent Technologies, 103344-100) following the manufacturer's protocol. Briefly, 104 cells/well were seeded in RPMI-1640 medium (Lonza) supplemented with 10% FBS and 2 mmol/L L-glutamine in a humidified incubator with 5% CO2. The following day, oxygen consumption rates (OCR) and extracellular acidification rates (ECAR) of PT45 and BxPC3 cells were measured in XF medium (nonbuffered RPMI medium containing 10 mmol/L glucose, 1 mmol/L pyruvate, and 2 mmol/L glutamine) using an XF24 analyzer (Seahorse Bioscience). OCR and ECAR were measured under basal conditions and in response to 500 nmol/L rotenone/antimycin A and 50 mmol/L 2-deoxy-D-glucose (2-DG; all Agilent Technologies), mitochondrial and glycolytic inhibitors, respectively.

In vitro T-cell cultures

Peripheral blood mononuclear cells (PBMC) were isolated from buffy coats of healthy donors by gradient centrifugation using Ficoll. CD8+ T cells were then separated using immunomagnetic positive selection (CD8 MicroBeads, Miltenyi Biotec) and seeded at 2 × 105 cells/well in glucose-free RPMI-1640 medium (Lonza) containing 10% FBS. D-glucose (Sigma) was added at concentrations varying from 90 mg/dL (referred to as “normal”) to 10 mg/dL (referred to as “low”). Cancer cell line supernatants from PT45 or BxPC3 cells (33% concentration), lactic acid (5 mmol/L), or sodium lactate (5 mmol/L) were added from day 0 for 3 days before measuring PD-1 expression by flow cytometry. Cells were stained with CFSE (10 μmol/L for 10 minutes at 37°C in PBS), stimulated with anti-CD3/anti-CD28-coupled beads (Dynabeads, Thermo Fisher Scientific) at a 1:100 bead/T-cell ratio, and cultured for 7 days before measuring PD-1, CD25, CD127 expression and CFSE dilution by flow cytometry. The following antibodies were used: PD-1 (clone EH12.1, BD Biosciences), CD25 (clone M-A251, BD Biosciences), and CD127 (clone HIL-7R-M21, BD Biosciences). Sample acquisition was performed on a BD LSRFortessa (BD Biosciences), and data were analyzed with FlowJo software.

PDAC murine models

All mice used for the subcutaneous implantation of murine cell lines were 8-week-old C57BL/6J females purchased from Charles River and housed in a specific pathogen–free animal facility of the Humanitas Clinical and Research Center in individually ventilated cages. Procedures involving mice and their care were conformed to EU and Institutional Guidelines (protocol ID 121/2016-PR). Pancreatic cancer–prone LSL-KrasG12D-Pdx1-Cre mice (KC mice) were generated by crossing single-mutated KrasG12D with C57BL/6 mice expressing Cre recombinase. Mice were screened by PCR using tail DNA amplified by specific primers to the Lox-P cassette flanking Kras and wild-type Kras genes and Cre recombinase genes. Mice were bred, maintained, and treated at the saprophytic and pathogen-free animal facility of the Molecular Biotechnologies Center (Torino, Italy), as previously described (33).

For subcutaneous implantation of all murine cell lines (Panc02, DT6606, shPFK-Panc02, and shSCR-Panc02), cells were cultured in RPMI-1640 medium (Lonza) supplemented with 10% FBS and 2 mmol/L L-glutamine in a humidified incubator with 5% CO2 for one passage before in vivo injection. Cells (106) were injected subcutaneously in the right flank, and tumor growth was monitored using a caliper twice a week. In all experiments, mice were sacrificed 21 days after injection of tumor cells. Once excised, tumor weight and size (using caliper) were recorded.

Isolation of tumor interstitial fluid

Tumor interstitial fluid (TIF) was collected from subcutaneously implanted Panc02 and DT6606 tumors via a centrifugation method as previously described (34). Briefly, tumors were excised, rinsed in PBS, blotted on filter paper, placed on 40-μm nylon filters on top of 50-mL conical tubes and centrifuged for 10 minutes at 4°C at 400 × g. Flow through from this step was collected as TIF and analyzed for the indicated metabolite concentrations.

Optical imaging

In vivo evaluation of glucose uptake was assessed on subcutaneously implanted Panc02 and DT6606 tumors using the fluorescent glucose analogue XenoLight RediJect 2-DG-750 probe (PerkinElmer), following the manufacturer's instructions. The probe was injected i.v. 24 hours prior to sacrifice of the mice. Imaging was performed ex vivo on excised tumors using the IVIS Lumina III InVivo Imaging System (PerkinElmer). Acquisition was performed with four filter pairs to allow spectral unmixing (excitation 680, 700, 720, and 740 nm; emission 790 nm). Binning: 8; f/stop: 2. Images were analyzed with Living Image 4.3.1 software (PerkinElmer).

Flow cytometry of tumors

Single-cell suspensions were obtained by incubating fragmented tumors with collagenase type IV (0.5 mg/mL; Sigma-Aldrich) for 30 minutes at 37°C. Tumor-infiltrating leukocytes were isolated by immunomagnetic positive selection (TIL CD45 MicroBeads, Miltenyi Biotec). The following fluorophore-conjugated primary antibodies were used: anti-CD45 (BD Pharmingen, clone 30-F11), anti-CD3 (eBioscience, clone 145-2C11), anti-CD8 (BioLegend, clone 53-6.7), and anti–PD-1 (BioLegend, clone 29F.1A12). Sample acquisition was performed on a BD LSRFortessa (BD Biosciences), and data were analyzed with FlowJo software. Zombie Aqua (BioLegend) amine-reactive fluorescent dye was used to perform dead cell exclusion. 2-NDBG (2-Deoxy-2-[(7-nitro-2,1,3-benzoxadiazol-4-yl)amino]-D-glucose) uptake was measured by incubating digested tumors in a 200 μmol/L 2-NBDG (Sigma-Aldrich) solution for 30 minutes at 37°C.

Gene expression

RNA was extracted with the RNeasy Plus Mini Kit (QIAGEN) following the manufacturer's instructions and reverse transcribed with the High Capacity cDNA Reverse Transcription Kit (Applied Biosystem) following the standard protocol. cDNA was then used for real-time qRT-PCR, performed using SYBR Green PCR Master Mix (Applied Biosystems) and SYBR Green RTPCR Reagents Kit (Applied Biosystems) following kit protocol. Real-time qPCR reactions were performed on ViiA7 Real-Time PCR (Life Technologies) instrument with standard protocol. The results were normalized to Hprt expression. Differences in gene expression were analyzed by the comparative threshold cycle (Ct) method (2–ΔΔCt). The following primers were used: Pfkm (forward: TCCATGAGGGTTACCAAGGC; reverse: GCACTTCCAATCACTGTGCC), Cd274 (forward: TCTCCTCGCCTGCAGATAGT; reverse: AGCCGTGATAGTAAACGCCC), Hprt (forward: CTCAGACCGCTTTTTGCCG; reverse: CGCTAATCACGACGCTGGG).

Statistical analysis

Correlation between lactate concentration measured by different assays and between IRA% values of GLUT-1 and PD-1 was estimated by a nonparametric Spearman rank correlation coefficient test and linear regression analysis. Differences between experimental groups were estimated by a nonparametric Mann–Whitney U test. Differences between distribution of patients according to clinical variables were estimated by a χ2 test. Survival analysis was performed by Kaplan–Meier curves. Differences in overall survival time were calculated by the Mantel–Cox log-rank test using the R software. For each test, only two-sided P values lower than 0.05 were considered statistically significant. All the analyses were done using GraphPad Prism software (Version 4.1) or R software.

Alteration in glucose metabolism in pancreatic juice from PDAC patients

Cancer cells have metabolic alterations that can result in biochemical changes in biological fluids and, therefore, represent a source of relevant metabolic variables. In order to capture a comprehensive overview of the metabolic profile of pancreatic specimens, we collected pancreatic juice from patients who underwent pancreatectomy at our institute. Pancreatic juice is the liquid that collects pancreatic products and is readily accessible during endoscopic procedures (35) or pancreatectomy. Due to its proximity to the pancreas, it could be more informative compared with plasma or urine. We performed a metabolomics analysis by 1H-NMR on pancreatic juices from 40 patients (cohort 1), including 31 PDAC and 9 other non-PDAC clinically relevant controls [2 chronic pancreatitis, 4 papillary-ampulla tumors, 2 neuroendocrine tumors, 1 intraductal papillary mucinous neoplasia (IPMN); Supplementary Table S1]. Availability of limited information of this biofluid led us to use 1H-NMR spectroscopy, which despite low sensitivity presents with high reproducibility and accuracy in detection (31). 1d CPMG spectra obtained from 1H-NMR analysis of pancreatic juices allowed us to appreciate in detail small molecular weight metabolites by filtering the broad NMR signals of macromolecules (e.g., lipoproteins, lipids etc.; Fig. 1A). Supervised OPLS-DA analysis indicated that the metabolic profile of pancreatic juice could discriminate between PDAC patients and non-PDAC controls, which differentially segregated with an accuracy of 82.4% obtained by cross-validation (Fig. 1B; Supplementary Fig. S1A). In the NMR profile of the pancreatic juice samples, 18 metabolites (L-phenylalanine, citrate, L-glutamine, L-tyrosine, acetate, L-tryptophan, L-valine, succinate, L-leucine, lactate, L-isoleucine, 3-hydroxy-butyrate, glycine, glucose, formate, L-alanine, acetone, acetoacetate) were responsible for the group classification (Fig. 1C). By analyzing the contribution of each metabolite, lactate, the end product of glycolysis, was the only one significantly increased in PDAC patients (1.57-fold) compared with non-PDAC patients (P = 0.011), whereas the concentration of most of the other metabolites detected was decreased (Fig. 1C). This result suggests a selective alteration of glucose metabolism in PDAC cells. The concentration of lactate was not significantly different according to the occurrence of PDAC-associated diabetes (Supplementary Fig. S1B) or according to tumor size (Supplementary Fig. S1C), and metabolomics analysis on 29 plasma samples from the same patients did not achieve equal discriminative power (Supplementary Fig. S2).

PDAC accumulates lactate and upregulates the glycolytic marker GLUT-1

We then asked whether the concentration of a single metabolite could discriminate the two groups of patients. In a preliminary analysis on 27 juices, we confirmed that the concentration of lactate, measured by 1H-NMR, correlated with the concentration of lactate measured by a colorimetric assay (r = 0.954, P < 0.0001; Fig. 2A). We then analyzed pancreatic juice samples from 106 patients (cohort 2; Supplementary Table S2) by the colorimetric assay and found that the concentration of lactate was significantly enriched in pancreatic juice from PDAC patients (n = 79) compared with juice from non-PDAC patients (n = 27; P = 0.012; 95% CI, 5.93-61.86; Fig. 2B). No difference was seen in lactate concentration in plasma from PDAC and non-PDAC patients (Supplementary Fig. S2). We next tested pancreatic tissues to confirm our findings obtained in pancreatic juice. Because lactate is the end product of glycolysis and accelerated glucose metabolism in tumor cells (the so-called “Warburg effect”) relies on modulation of key glycolytic enzymes (10), including glucose transporter 1 (GLUT-1), we assessed the rate of glycolysis in human PDAC tissues by analyzing the expression of GLUT-1 (22, 36) in 18 PDAC specimens by IHC. The expression of GLUT-1 on pancreatic ducts was significantly higher compared with the healthy region of the same patient, and increased expression was seen in regions where tumor cells invaded normal pancreas (Fig. 2C and D). Despite a certain degree of heterogeneity of GLUT-1 expression within the same specimen, GLUT-1 expression was higher in specimens from high-grade (G3–G4) compared with low-grade (G1–G2) tumors (Fig. 2E and F), confirming an association between a more undifferentiated phenotype and a higher glycolytic activity of pancreatic tumor cells (22).

Metabolic profile of pancreatic juice identifies patients with distinct clinical outcomes

We, thus, investigated whether any association of the metabolic profile with the clinical behavior of PDAC patients exists (cohort 1; n = 31). Unsupervised K-means cluster analysis of the metabolomics data, without imposing any threshold, identified 4 metabolic clusters (Fig. 3A). By pairing the groups with clinical data, we found a distinct correlation of specific metabolic clusters (clusters 1–3) with clinical parameters [i.e., tumor size (T) and grade (G)]. In particular, although metabolic cluster 1 comprised primarily patients with T1–T2 tumors, patients belonging to metabolic clusters 2 and 3 had, for the most part, T3 tumors (Fig. 3B, top; P = 0.02). Similarly, metabolic cluster 1 comprised primarily patients with G1–G2 tumors, whereas patients belonging to metabolic clusters 2 and 3 had, for the most part, G3–G4 tumors (Fig. 3B, bottom; P = 0.009). Kaplan–Meier survival curves confirmed that the metabolic profile detected in pancreatic juice associated with different clinical outcomes, with cluster 1 having the best and cluster 2 having the worst survival (Fig. 3C; P = 0.00079). In addition to lactate, L-valine, succinate, L-phenylalanine, L-leucine, and acetoacetate were significantly increased in clusters 2 and 3. Nonetheless, the clustering was ascribable to the total metabolic fingerprint rather than to any one metabolite in particular (37), possibly due to the small number of samples, thus resulting in a composite biomarker.

Tissue metabolic alterations correlate with distinct immune contextures in human PDAC

We next analyzed whether the increased glucose metabolism correlated with the immune contexture of PDAC. To this aim, we investigated potential associations of GLUT-1 expression with the presence of tumor-infiltrating leukocyte populations. By IHC, we quantified the density [immunoreactive area (IRA%)] of GLUT-1 (Fig. 2C) and of immune cells, including CD8+ tumor-infiltrating lymphocytes (CD8-TIL), CD20+ B lymphocytes (CD20-TILs), CD68+ tumor–associated macrophages (CD68-TAM; Supplementary Fig. S3A), and PD-1+ cells (Fig. 4A). The density of PD-1+ TILs in 18 PDAC specimens correlated with the expression of GLUT-1 (r = 0.52, P = 0.023; Fig. 4B). Other tumor-infiltrating immune cells did not show a correlation with the expression of GLUT-1 (Supplementary Fig. S3B). To identify the single metabolic determinants involved in PD-1 upregulation, we cultured sorted CD8+ T cells in control medium (containing 90 mg/dL of D-glucose) or in low-glucose medium (containing 10 mg/dL of D-glucose), and added supernatants from two pancreatic cancer cell lines with different glycolytic profiles (PT45 and BXPC3; Supplementary Fig. S4), lactate, or lactic acid (Fig. 4C). Low-glucose concentration was the main driver for PD-1 upregulation (P < 0.0001; Fig. 4C). Based on these preliminary results, we additionally performed dose–response experiments on the effect of glucose on CD8+ T-cell proliferation and expression of PD-1, CD25, and CD127 (Fig. 4D). Although proliferation progressively declined with decreasing glucose concentrations (Fig. 4D), a glucose concentration of 30 mg/dL induced a significant upregulation of PD-1 (P = 0.002 vs. control), and only marginally affected the expression of CD127 and CD25 (Fig. 4D).

We confirmed the correlation between GLUT-1 expression by pancreatic ducts and PD-1 infiltration in the genetic preclinical model of PDAC Pdx1-Cre;KrasLSLG12D. Although very low or absent expression of GLUT-1 was detected in the pancreas of wild-type (WT) mice (Fig. 4E, left), weak to very intense immunoreactivity was present in the tumors of Pdx1-Cre;KrasLSLG12D mice (Fig. 4E, right). An analogous expression pattern was shown by PD-1+ TILs (Fig. 4F) and, consistent with our observations in human PDAC, a positive correlation between GLUT-1 expression and PD-1+ TIL density was present (r = 0.79, P = 0.0007; Fig. 4G).

Targeting glycolysis impacts the accumulation of PD-1+ TILs in PDAC preclinical models

To further investigate the association between the tumor metabolic state and PD-1+ cell infiltration, we modeled the immune–metabolic axis observed in human specimens in preclinical models of PDAC. We analyzed implanted tumors obtained from two different pancreatic cell lines, DT6606 and Panc02, derived, respectively, from early- and late-stage murine PDAC. Despite the fact that tumors obtained from the two cell lines were not significantly different in size (Supplementary Fig. S5A), they displayed a different glycolytic activity. In particular, Panc02 tumors displayed a higher glycolytic activity measured by both in vitro uptake of a fluorescent glucose analogue (2-NBDG; Fig. 5A) and ex vivo by optical imaging on the excised tumor mass, after injection of the fluorescent probe deoxy-D-glucose (2-DG-750; Fig. 5B and C). The increased metabolism was reflected in an increased glucose consumption, which led to decreased glucose concentration in the interstitial fluid obtained from Panc02 tumors compared with DT6606 tumors (Fig. 5D). Tumor cells accounted for most of the glycolytic activity (Supplementary Fig. S5B), as assessed by glucose uptake ex vivo after tumor mass digestion. Consistent with the correlation between GLUT-1 expression and PD-1+ cell infiltration observed in both human and murine PDAC, Panc02 tumors contained a higher percentage of PD-1+ cells within CD8+ lymphocytes compared with DT6606 tumors (Fig. 5EG), despite a similar density of total CD3+ and CD8+ cells (Supplementary Fig. S5C). This suggests a selective increase of PD-1 expression in highly metabolic tumors.

To probe whether the increased metabolism of Panc02 cells was sufficient to determine PD-1+ TIL accumulation, we blunted the glycolytic metabolism of tumor cells. We selectively silenced glycolysis in Panc02 cells by genetically targeting the rate-limiting glycolytic enzyme phosphofructokinase-m (PFK-m), which, despite its relatively low expression (Supplementary Fig. S6A), is exclusively involved in the glycolytic reaction. We generated Panc02 cells stably knocked-down with an shRNA targeting Pfkm (shPFK) and the corresponding control cell line stably expressing a scramble sequence (shSCR; Fig. 5H). The two cell lines were implanted into mice, to originate tumors with opposite glycolytic rates (Supplementary Fig. S6B). Knockdown of Pfkm did not affect tumor growth (Supplementary Fig. S6C) nor the frequency of CD3+ and CD8+ lymphocytes (Supplementary Fig. S6D). Genetic targeting of glucose metabolism significantly reduced the frequency of PD-1+ TILs within CD8+ cells infiltrating Panc02 tumors (Fig. 5I) and decreased Cd274 (PD-L1) expression in tumor cells (Fig. 5J), ultimately confirming the link between glucose metabolism and PD-1+ cell infiltration. This result also suggests that targeting tumor metabolism could be used as a strategy to alter immune infiltration.

Metabolomic analyses of cancer biospecimens grant the possibility to appreciate the variety of molecules downstream of cancer metabolic adjustments. In pancreatic cancer, well known for its metabolic alterations, metabolomic analyses of blood (12, 38, 39), urine (11), salivary fluids (40), and exhaled breath (13) have been performed, primarily documenting differences between PDAC and control subjects or patients with pancreatitis. In this study, we tested the value of pancreatic juice—a far less explored body fluid—as a source of metabolic markers of disease. 1H-NMR spectral data indicated significant group discrimination based on a disease-specific metabolic profile that singled out PDAC from clinically relevant controls, including benign pancreatitis and other pancreatobiliary diseases with overlapping symptoms (tumors of the ampullary region, neuroendocrine, and cystic tumors). This result suggests that upfront evaluation of the metabolic profile could implement diagnostic accuracy and help in detecting early-stage tumors. The metabolic fingerprint of pancreatic juice associated with distinct clinical outcomes, confirming that it could represent a composite biomarker (31, 37) capable of prognostic classification, in contrast to plasma samples. This could be important in identifying rapidly evolving diseases, avoiding, when unnecessary, considerable morbidity still associated with surgical resection. Despite pancreatic juice was collected intraoperatively in this study, in the perspective of identifying early informative clinical markers, the collection of pancreatic juice could also be performed by minimally invasive procedures, such as endoscopic ultrasound, retrograde cholangio-pancreatography, or by endoscopic collection of duodenal juice secretion (35), although such techniques could alter the composition of the fluid collected and results should, therefore, be carefully evaluated. Also, in order to detect important biochemical details and potential biomarkers, it may be beneficial to perform a more comprehensive metabolic assessment by other metabolomics assays, such as mass spectrometry.

Among the 18 metabolites identified in pancreatic juice, only lactate exhibited a higher concentration in samples from PDAC patients, whereas a number of other metabolites were decreased. The distinct metabolic profiles could suggest a rewiring of PDAC tumors toward aerobic glycolysis (lactate accumulation) and increased lipid and nucleotide biosynthesis (decreased concentration of essential amino acids, glycine, and acetoacetate). This comprehensive analysis does not allow to investigate in detail the biochemical pathways ongoing in tumor cells, as it is the result of all the different cell types that contribute to the tumor metabolism, including stromal and immune cells. Nevertheless, the increased expression of the glucose transporter GLUT-1 in PDAC paraffin-embedded specimens compared with healthy regions of the same patient reflects an increased glucose uptake (36). GLUT-1 expression in PDAC cells has been implicated in resistance to chemotherapy (41) and metastasis (22), confirming the clinical relevance of alterations in glucose metabolism that can be detected in the biological fluid immediately adjacent to the tumor. The concentration of lactate in the pancreatic juice collected from a larger cohort of patients also increased. This finding suggests that, with further development, the assessment of this metabolite could be feasible in clinical routines, where the performance of other biomarkers (such as carbohydrate antigen 19-9) is disappointing in terms of differential diagnostic power. It is important to underline that in our cohort of patients, we were not able to detect in plasma the specific alterations seen in pancreatic juice.

Our data suggest that the glucose metabolism of human PDAC affects its immune contexture. Metabolic reprogramming, a hallmark of cancer progression, is known to severely impact on the efficacy of the antitumor immune response (8–10). Our results entail clinical consequences because presurgical identification of markers correlating with immune variables could represent a considerable advancement in the introduction and choice of immunotherapeutic approaches for pancreatic cancer. When we attempted to define the mechanism behind the metabolic modulation of PD-1, we found that low-glucose concentration was the main driver of PD-1 upregulation in T cells. The mechanisms of resistance and variable response to checkpoint inhibitors remain poorly understood and suggest that factors other than the immunologic ones could be important dictators of response to PD-1–targeted therapies, including metabolic pathways (42). Along this line, baseline blood concentration of lactate dehydrogenase (LDH) has been identified as a predictive biomarker of response to anti–CTLA-4 treatment (43–45) and intratumor expression of metabolic genes was associated with resistance to anti–PD-1 therapy (46). Because anti–PD-1 works predominantly within tumors, looking for circulating biomarkers of response could be unrevealing, and pancreatic juice could represent a valuable option as a body fluid, collecting metabolites released in the microenvironment of the tumor mass. However, differently from the other tumor types mentioned, issues related to checkpoint inhibitors are still out of reach in the case of PDAC, for which these therapeutic approaches have been introduced to a minor extent, and patients still have not achieved tumor regression (19, 23).

Genetic inhibition of glycolysis in a preclinical model resulted in a decreased accumulation of PD-1+ cells, without impairing CD8+ T-cell recruitment, and in decreased expression of Cd274 (PD-L1) in tumor cells. This provides a rationale to metabolic targeting of tumors as an effective option to reboot the antitumor response. Other studies have shown that neutralization of tumor acidity improves antitumor responses to immunotherapy with checkpoint inhibitors (47), although further studies are needed to dissect the function of PD-1+ T cells in tumors with different metabolic activity.

Collectively, our analysis promotes metabolomics as a valuable diagnostic tool, while introducing pancreatic juice as a source of metabolic biomarkers to advance surgical management and clinical decision-making of pancreatic diseases. Although promising, our metabolic analysis cannot be considered definitive and further development is needed to design a clinically feasible method based on our results. The correlation of metabolic markers with immune infiltration suggests that implementation of preoperative patient metabolic profiling could foster the introduction of immunotherapeutic approaches for pancreatic cancer.

A. Mantovani is a consultant/advisory board member for Novartis, Roche, Ventana, Pierre Fabre, Verily, AbbVie, Compugen, Macrophage Therapeutics, AstraZeneca, Biovelocita, Verseau, Third Rock, and BG Fund; reports receiving commercial research grants from Novartis; and is an inventor of patents related to PTX3 and other innate immunity molecules and receives royalties for reagents related to innate immunity. No potential conflicts of interest were disclosed by the other authors.

Conception and Design: F. Marchesi, G. Capretti, A. Zerbi

Development of methodology: P.G. Takis, G.F. Castino

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N. Cortese, G. Capretti, M. Barbagallo, A. Rigamonti, G.F. Castino, D. Vignali, G. Nappo, G. Donisi, M. Erreni, R. Avigni, P. Spaggiari, P. Cappello, F. Novelli, A. Zerbi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): G. Capretti, P.G. Takis, D. Vignali, G. Maggi, M. Erreni, F. Novelli, P. Monti, F. Marchesi

Writing, review, and/or revision of the manuscript: N. Cortese, G. Capretti, P.G. Takis, G. Nappo, P. Cappello, P. Monti, P. Allavena, A. Mantovani, F. Marchesi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): G. Capretti, F. Gavazzi, C. Ridolfi, D. Rahal

Study supervision: A. Mantovani, F. Marchesi

Other (NMR data acquisition): P.G. Takis

Other (study of the cases): M. Roncalli

We thank Roberta Migliore for technical assistance, Daniela Pistillo and the Biobank within the Humanitas Clinical and Research Center, Dr. Antonio Inforzato for helpful discussion on biochemical pathways, and Dr. Floriana Farina for assistance in glycolytic rate assessment. The research leading to these results has received funding from Associazione Italiana per la Ricerca sul Cancro (AIRC) under IG2016-ID.18443 project: principal investigator F. Marchesi, AIRC fellowship 18011 to N. Cortese, AIRC 5 × 1000 IG-12182 to P. Allavena and F. Novelli and IG-15257 to F. Novelli, and AIRC 5 × 1000 21147 to A. Mantovani. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Costello
E
,
Greenhalf
W
,
Neoptolemos
JP
. 
New biomarkers and targets in pancreatic cancer and their application to treatment
.
Nat Rev Gastroenterol Hepatol
2012
;
9
:
435
44
.
2.
Siegel
RL
,
Miller
KD
,
Jemal
A
. 
Cancer statistics, 2018
.
CA Cancer J Clin
2018
;
68
:
7
30
.
3.
Neoptolemos
JP
,
Kleeff
J
,
Michl
P
,
Costello
E
,
Greenhalf
W
,
Palmer
DH
. 
Therapeutic developments in pancreatic cancer: current and future perspectives
.
Nat Rev Gastroenterol Hepatol
2018
;
15
:
333
48
.
4.
Bailey
P
,
Chang
DK
,
Nones
K
,
Johns
AL
,
Patch
AM
,
Gingras
MC
, et al
Genomic analyses identify molecular subtypes of pancreatic cancer
.
Nature
2016
;
531
:
47
52
.
5.
Singhi
AD
,
George
B
,
Greenbowe
JR
,
Chung
J
,
Suh
J
,
Maitra
A
, et al
Real-time Targeted genome profile analysis of pancreatic ductal adenocarcinomas identifies genetic alterations that might be targeted with existing drugs or used as biomarkers
.
Gastroenterology
2019
;
156
:
2242
53
.
6.
Karasinska
JM
,
Topham
JT
,
Kalloger
SE
,
Jang
GH
,
Denroche
RE
,
Culibrk
L
, et al
Altered gene expression along the glycolysis-cholesterol synthesis axis is associated with outcome in pancreatic cancer
.
Clin Cancer Res
2020
;
26
:
135
46
.
7.
Halbrook
CJ
,
Lyssiotis
CA
. 
Employing metabolism to improve the diagnosis and treatment of pancreatic cancer
.
Cancer Cell
2017
;
31
:
5
19
.
8.
Kroemer
G
,
Pouyssegur
J
. 
Tumor cell metabolism: cancer's Achilles' heel
.
Cancer Cell
2008
;
13
:
472
82
.
9.
Hanahan
D
,
Weinberg
RA
. 
Hallmarks of cancer: the next generation
.
Cell
2011
;
144
:
646
74
.
10.
Ward
PS
,
Thompson
CB
. 
Metabolic reprogramming: a cancer hallmark even warburg did not anticipate
.
Cancer Cell
2012
;
21
:
297
308
.
11.
Napoli
C
,
Sperandio
N
,
Lawlor
RT
,
Scarpa
A
,
Molinari
H
,
Assfalg
M
. 
Urine metabolic signature of pancreatic ductal adenocarcinoma by (1)h nuclear magnetic resonance: identification, mapping, and evolution
.
J Proteome Res
2012
;
11
:
1274
83
.
12.
Mayerle
J
,
Kalthoff
H
,
Reszka
R
,
Kamlage
B
,
Peter
E
,
Schniewind
B
, et al
Metabolic biomarker signature to differentiate pancreatic ductal adenocarcinoma from chronic pancreatitis
.
Gut
2018
;
67
:
128
37
.
13.
Markar
SR
,
Brodie
B
,
Chin
ST
,
Romano
A
,
Spalding
D
,
Hanna
GB
. 
Profile of exhaled-breath volatile organic compounds to diagnose pancreatic cancer
.
Br J Surg
2018
;
105
:
1493
500
.
14.
MacIver
NJ
,
Michalek
RD
,
Rathmell
JC
. 
Metabolic regulation of T lymphocytes
.
Annu Rev Immunol
2013
;
31
:
259
83
.
15.
Pearce
EL
,
Pearce
EJ
. 
Metabolic pathways in immune cell activation and quiescence
.
Immunity
2013
;
38
:
633
43
.
16.
Pollizzi
KN
,
Powell
JD
. 
Integrating canonical and metabolic signalling programmes in the regulation of T cell responses
.
Nat Rev Immunol
2014
;
14
:
435
46
.
17.
Fox
CJ
,
Hammerman
PS
,
Thompson
CB
. 
Fuel feeds function: energy metabolism and the T-cell response
.
Nat Rev Immunol
2005
;
5
:
844
52
.
18.
Kim
DH
,
Sarbassov
DD
,
Ali
SM
,
King
JE
,
Latek
RR
,
Erdjument-Bromage
H
, et al
mTOR interacts with raptor to form a nutrient-sensitive complex that signals to the cell growth machinery
.
Cell
2002
;
110
:
163
75
.
19.
Feig
C
,
Jones
JO
,
Kraman
M
,
Wells
RJ
,
Deonarine
A
,
Chan
DS
, et al
Targeting CXCL12 from FAP-expressing carcinoma-associated fibroblasts synergizes with anti-PD-L1 immunotherapy in pancreatic cancer
.
Proc Natl Acad Sci U S A
2013
;
110
:
20212
7
.
20.
Beatty
GL
,
Winograd
R
,
Evans
RA
,
Long
KB
,
Luque
SL
,
Lee
JW
, et al
Exclusion of T cells from pancreatic carcinomas in mice is regulated by Ly6C(low) F4/80(+) extratumoral macrophages
.
Gastroenterology
2015
;
149
:
201
10
.
21.
Jiang
H
,
Hegde
S
,
Knolhoff
BL
,
Zhu
Y
,
Herndon
JM
,
Meyer
MA
, et al
Targeting focal adhesion kinase renders pancreatic cancers responsive to checkpoint immunotherapy
.
Nat Med
2016
;
22
:
851
60
.
22.
Chaika
NV
,
Yu
F
,
Purohit
V
,
Mehla
K
,
Lazenby
AJ
,
DiMaio
D
, et al
Differential expression of metabolic genes in tumor and stromal components of primary and metastatic loci in pancreatic adenocarcinoma
.
PLoS One
2012
;
7
:
e32996
.
23.
Brahmer
JR
,
Tykodi
SS
,
Chow
LQ
,
Hwu
WJ
,
Topalian
SL
,
Hwu
P
, et al
Safety and activity of anti-PD-L1 antibody in patients with advanced cancer
.
N Engl J Med
2012
;
366
:
2455
65
.
24.
Castino
GF
,
Cortese
N
,
Capretti
G
,
Serio
S
,
Di Caro
G
,
Mineri
R
, et al
Spatial distribution of B cells predicts prognosis in human pancreatic adenocarcinoma
.
Oncoimmunology
2015
;
5
:
e1085147
.
25.
Di Caro
G
,
Cortese
N
,
Castino
GF
,
Grizzi
F
,
Gavazzi
F
,
Ridolfi
C
, et al
Dual prognostic significance of tumour-associated macrophages in human pancreatic adenocarcinoma treated or untreated with chemotherapy
.
Gut
2016
;
65
:
1710
20
.
26.
Carstens
JL
,
Correa de Sampaio
P
,
Yang
D
,
Barua
S
,
Wang
H
,
Rao
A
, et al
Spatial computation of intratumoral T cells correlates with survival of patients with pancreatic cancer
.
Nat Commun
2017
;
8
:
15095
.
27.
Mahajan
UM
,
Langhoff
E
,
Goni
E
,
Costello
E
,
Greenhalf
W
,
Halloran
C
, et al
Immune cell and stromal signature associated with progression-free survival of patients with resected pancreatic ductal adenocarcinoma
.
Gastroenterology
2018
;
155
:
1625
39
.
28.
Wartenberg
M
,
Cibin
S
,
Zlobec
I
,
Vassella
E
,
Eppenberger-Castori
S
,
Terracciano
L
, et al
Integrated genomic and immunophenotypic classification of pancreatic cancer reveals three distinct subtypes with prognostic/predictive significance
.
Clin Cancer Res
2018
;
24
:
4444
54
.
29.
Takis
PG
,
Taddei
A
,
Pini
R
,
Grifoni
S
,
Tarantini
F
,
Bechi
P
, et al
Fingerprinting acute digestive diseases by untargeted NMR based metabolomics
.
Int J Mol Sci
2018
;
19
:
E3288
.
30.
Hart
CD
,
Vignoli
A
,
Tenori
L
,
Uy
GL
,
Van To
T
,
Adebamowo
C
, et al
Serum metabolomic profiles identify ER-positive early breast cancer patients at increased risk of disease recurrence in a multicenter population
.
Clin Cancer Res
2017
;
23
:
1422
31
.
31.
Vignoli
A
,
Ghini
V
,
Meoni
G
,
Licari
C
,
Takis
PG
,
Tenori
L
, et al
High-throughput metabolomics by 1D NMR
.
Angew Chem Int Ed Engl
2019
;
58
:
968
94
.
32.
Corbett
TH
,
Roberts
BJ
,
Leopold
WR
,
Peckham
JC
,
Wilkoff
LJ
,
Griswold
DP
, et al
Induction and chemotherapeutic response of two transplantable ductal adenocarcinomas of the pancreas in C57BL/6 mice
.
Cancer Res
1984
;
44
:
717
26
.
33.
Cappello
P
,
Rolla
S
,
Chiarle
R
,
Principe
M
,
Cavallo
F
,
Perconti
G
, et al
Vaccination with ENO1 DNA prolongs survival of genetically engineered mice with pancreatic cancer
.
Gastroenterology
2013
;
144
:
1098
106
.
34.
Sullivan
MR
,
Danai
LV
,
Lewis
CA
,
Chan
SH
,
Gui
DY
,
Kunchok
T
, et al
Quantification of microenvironmental metabolites in murine cancers reveals determinants of tumor nutrient availability
.
Elife
2019
;
8
:
E44235
.
35.
Yu
J
,
Sadakari
Y
,
Shindo
K
,
Suenaga
M
,
Brant
A
,
Almario
JAN
, et al
Digital next-generation sequencing identifies low-abundance mutations in pancreatic juice samples collected from the duodenum of patients with pancreatic cancer and intraductal papillary mucinous neoplasms
.
Gut
2017
;
66
:
1677
1687
.
36.
van Baardwijk
A
,
Dooms
C
,
van Suylen
RJ
,
Verbeken
E
,
Hochstenbag
M
,
Dehing-Oberije
C
, et al
The maximum uptake of (18)F-deoxyglucose on positron emission tomography scan correlates with survival, hypoxia inducible factor-1alpha and GLUT-1 in non-small cell lung cancer
.
Eur J Cancer
2007
;
43
:
1392
8
.
37.
Turano
P
. 
Colorectal cancer: the potential of metabolic fingerprinting
.
[editorial]
.
Expert Rev Gastroenterol Hepatol
2014
;
8
:
847
9
.
38.
Bathe
OF
,
Shaykhutdinov
R
,
Kopciuk
K
,
Weljie
AM
,
McKay
A
,
Sutherland
FR
, et al
Feasibility of identifying pancreatic cancer based on serum metabolomics
.
Cancer Epidemiol Biomarkers Prev
2011
;
20
:
140
7
.
39.
Mayers
JR
,
Wu
C
,
Clish
CB
,
Kraft
P
,
Torrence
ME
,
Fiske
BP
, et al
Elevation of circulating branched-chain amino acids is an early event in human pancreatic adenocarcinoma development
.
Nat Med
2014
;
20
:
1193
8
.
40.
Sugimoto
M
,
Wong
DT
,
Hirayama
A
,
Soga
T
,
Tomita
M
. 
Capillary electrophoresis mass spectrometry-based saliva metabolomics identified oral, breast and pancreatic cancer-specific profiles
.
Metabolomics
2010
;
6
:
78
95
.
41.
Shukla
SK
,
Purohit
V
,
Mehla
K
,
Gunda
V
,
Chaika
NV
,
Vernucci
E
, et al
MUC1 and HIF-1alpha signaling crosstalk induces anabolic glucose metabolism to impart gemcitabine resistance to pancreatic cancer
.
Cancer Cell
2017
;
32
:
71
87
.
42.
Brand
A
,
Singer
K
,
Koehl
GE
,
Kolitzus
M
,
Schoenhammer
G
,
Thiel
A
, et al
LDHA-associated lactic acid production blunts tumor immunosurveillance by T and NK cells
.
Cell Metab
2016
;
24
:
657
71
.
43.
Weide
B
,
Martens
A
,
Hassel
JC
,
Berking
C
,
Postow
MA
,
Bisschop
K
, et al
Baseline biomarkers for outcome of melanoma patients treated with pembrolizumab
.
Clin Cancer Res
2016
;
22
:
5487
96
.
44.
Kelderman
S
,
Heemskerk
B
,
van Tinteren
H
,
van den Brom
RR
,
Hospers
GA
,
van den Eertwegh
AJ
, et al
Lactate dehydrogenase as a selection criterion for ipilimumab treatment in metastatic melanoma
.
Cancer Immunol Immunother
2014
;
63
:
449
58
.
45.
Martens
A
,
Wistuba-Hamprecht
K
,
Geukes Foppen
M
,
Yuan
J
,
Postow
MA
,
Wong
P
, et al
Baseline peripheral blood biomarkers associated with clinical outcome of advanced melanoma patients treated with ipilimumab
.
Clin Cancer Res
2016
;
22
:
2908
18
.
46.
Ascierto
ML
,
McMiller
TL
,
Berger
AE
,
Danilova
L
,
Anders
RA
,
Netto
GJ
, et al
The intratumoral balance between metabolic and immunologic gene expression is associated with anti-PD-1 response in patients with renal cell carcinoma
.
Cancer Immunol Res
2016
;
4
:
726
33
.
47.
Pilon-Thomas
S
,
Kodumudi
KN
,
El-Kenawi
AE
,
Russell
S
,
Weber
AM
,
Luddy
K
, et al
Neutralization of tumor acidity improves antitumor responses to immunotherapy
.
Cancer Res
2016
;
76
:
1381
90
.