Deprogramming metabolism in pancreatic cancer with a bi-functional GPR55 inhibitor and biased β2 adrenergic agonist

Metabolic reprogramming contributes to oncogenesis, tumor growth, and treatment resistance in pancreatic ductal adenocarcinoma (PDAC). Here we report the effects of (R,S′)-4′-methoxy-1-naphthylfenoterol (MNF), a GPR55 antagonist and biased β2-adrenergic receptor (β2-AR) agonist on cellular signaling implicated in proliferation and metabolism in PDAC cells. The relative contribution of GPR55 and β2-AR in (R,S′)-MNF signaling was explored further in PANC-1 cells. Moreover, the effect of (R,S′)-MNF on tumor growth was determined in a PANC-1 mouse xenograft model. PANC-1 cells treated with (R,S′)-MNF showed marked attenuation in GPR55 signal transduction and function combined with increased β2-AR/Gαs/adenylyl cyclase/PKA signaling, both of which contributing to lower MEK/ERK, PI3K/AKT and YAP/TAZ signaling. (R,S′)-MNF administration significantly reduced PANC-1 tumor growth and circulating l-lactate concentrations. Global metabolic profiling of (R,S′)-MNF-treated tumor tissues revealed decreased glycolytic metabolism, with a shift towards normoxic processes, attenuated glutamate metabolism, and increased levels of ophthalmic acid and its precursor, 2-aminobutyric acid, indicative of elevated oxidative stress. Transcriptomics and immunoblot analyses indicated the downregulation of gene and protein expression of HIF-1α and c-Myc, key initiators of metabolic reprogramming in PDAC. (R,S′)-MNF treatment decreased HIF-1α and c-Myc expression, attenuated glycolysis, shifted fatty acid metabolism towards β-oxidation, and suppressed de novo pyrimidine biosynthesis in PANC-1 tumors. The results indicate a potential benefit of combined GPR55 antagonism and biased β2-AR agonism in PDAC therapy associated with the deprogramming of altered cellular metabolism.


Online Materials and Methods
Sample preparation and multiplatform metabolomics study I) Samples preparation. The method for sample preparation was adapted from González-Peña et al. 2017. Homogenization All samples were stored at -80°C until analysis. In the day of analysis, the samples were immersed in LN2 and a representative cross-section of tissue was cut and weighted. A mixture of cold (-20°C MeOH/H2O (1:1) in ratio 1:10) was added to each sample. The exact volume of the extraction solution was adjusted according to weight of the sample. Tissue lysis and metabolite extraction was carried out with QIAGEN TissueLyser LT bead-mill homogenizer. The process was performed by using 5 mm (mean diameter) stainless steel beads, vibrating at 50 Hz for 5 min. Three repeated cycles with 1 min break were applied with samples and TissueLyser adapter cooled on ice. Additionally, ultrasounds (UP 200S Ultrasonic lab homogenizer, Hielscher GmbH) were introduced. Parameters of sonication were set as follow: 50 amplitudes, 0.8 cycles, sonication time 30 sec. The process was performed on ice.

Metabolites extraction for LC-MS analysis
Homogenate (100 µL) was reconstituted in 320 µL of MeOH and vortex-mixed for 15 min. Subsequently 80 µL methyl tertbutyl ether (MTBE) was added and vortex-mixed 1 h at room temperature. Samples were centrifuged for 20 min at 4000 × g at 20°C. Hundred microliters (100 µL) of supernatant was used for LC-MS analysis.

Metabolites extraction for GC-MS analysis
Homogenate (150 μL) protein was precipitated with cold methanol (300 μL) and separated by centrifugation (16000 × g, 15 min, 4°C). Resulting supernatant was transferred to GC vial with insert and then evaporated to dryness (Speedvac Concentrator (Thermo Fisher Scientific). Twenty microliters (20 μL) of O-methoxyamine hydrochloride in pyridine (15 mg/mL) was added to each GC vial, and mixture was vigorously vortex-mixed and ultrasonicated. Methoxymation was carried out in darkness at room temperature for 16 h. N, O-Bistrifluoroacetamide (BSTFA) with 1% chlorotrimethylsilane (TMCS) (20 μL) was then added as catalyst. For silylation process samples were heated in an oven for 1 h at 70°C. Finally, 100 μL of heptane containing 10 ppm of C18:0 methyl ester as internal standard (IS) was added to each GC vial and vortex-mixed before GC analysis.

Metabolites extraction for CE-MS analysis
Two hundred microliters (200 µL) of homogenate were vortex-mixed with 200 μL of 0.2 M formic acid, centrifuged (16000 × g 10 min, 4°C) and transferred to a centrifree ultracentrifugation device (Millipore Ireland Ltd., Ireland) with 30 kDa protein cutoff for deprotenization through centrifugation (2000 × g, 70 min, 4°C). The filtrate was then transferred to the chromacol vial, dried using a SpeedVac, and resuspended in 100 μL of 0.1 M formic acid with 0.2 mM methionine sulfone (IS) before CE-MS analysis.

QC sample preparation
Quality control (QC) samples were prepared by pooling equal volumes of pancreatic cancer tissue homogenate from each of the 20 samples. Five QC samples were independently prepared for each technique following the same procedure as applied for the experimental samples. QC samples were analyzed throughout the run to provide a measurement of the system's stability, performance, and the reproducibility of the sample treatment procedure.

Randomization
The samples were randomized before homogenization, metabolite extraction and for analytical run.

LC-QTOF/MS
A HPLC system (1200 series, Agilent Technologies, Waldbronn, Germany), equipped with a degasser, two binary pumps, and a thermostated autosampler coupled with Q-TOF LC/MS (6520) system (Agilent), was used in the ESI+ and ESI− mode to increase the number of detected metabolite ions.
Briefly, 5 μL of extracted pancreatic tissue samples was injected into a thermostated (60°C) Agilent Poroshell 120 EC-C8 column (150 mm × 2.1 mm, 2.7 μm) with a guard column Ascentis® Express C8 (5 mm × 2.1 mm, 2.7 μm). The flow rate was 0.5 mL/min with solvent A (5 mM ammonium formate in Milli-Q water), and solvent B (5 mM ammonium formate in methanol) for analysis in positive ionization mode and solvent A (water with 0.1% formic acid), and solvent B (methanol with 0.1% formic acid and 15% isopropanol) for analysis in negative ionization mode. Initial conditions at time 0 were 82% B, increasing to 96% B in 30 min. This was then held until 38 min. The gradient then increased to 100% B by 38.5 min and held until 40.5 min. The starting condition was returned to by 42 min, followed by an 8-min re-equilibration time, taking the total run time to 50 min. Capillary voltage was set to 3.5 kV for positive and 4.5 kV for negative ionization mode; the drying gas flow rate was 12 L/min at 250°C and gas nebulizer at 52 psi; fragmentor voltage was 175 V for positive and 125 V for negative ionization mode; skimmer and octopole radio frequency voltage (OCT RF Vpp) were set to 65 V and 750 V, respectively. Data were collected in the centroid mode at a scan rate of 1.2 spectrum per second. Mass spectrometry detection was performed in both positive and negative ESI mode in full scan from 100 to 1200 m/z. The reference mass ions used were 121.050873, 922.009798 (positive ion mode) and 119.036320, 966.000725 (negative ion mode). These masses were continuously infused into the system to allow constant mass correction. Samples were analyzed in separate runs (positive and negative ionization modes), in a randomized order.

GC-Q-MS
A GC system (Agilent Technologies 7890A), equipped with an autosampler (Agilent 7693) and interfaced to an inert mass spectrometer with triple-Axis detector (5975C, Agilent), was used for pancreatic tissue cancer fingerprinting. Briefly, 2 μL of the derivatized sample was injected in a GC column DB5-MS (30 m length, 0.25 mm, 0.25 μm film 95% dimethyl/ 5% diphenylpolysiloxane) coupled to a pre-column (10 m Agilent J&W Capillary GC column integrated with Agilent 122-5532G). The injector port was held at 250°C, and the helium carrier gas flow rate was set at 1.0 mL/min. The split ratio was 1:10. The temperature gradient was programmed as follows: the initial oven temperature was set to 60°C (held for 1 min), increased to 325°C at a rate of 10°C/min; the system was allowed to cool down for 10 min before the next injection. The detector transfer line, the filament source and the quadrupole temperature were set to 280°C, 230°C and 150°C, respectively. MS detection was performed in electron impact (EI) mode at −70 eV. The mass spectrometer was operated in scan mode over a mass range of 50-600 m/z at a rate of 2.7 scan/s.

CE-TOF-MS
An Agilent 7100 (CE) system, coupled to a TOF Mass Spectrometer (6224 Agilent), was used for sample analysis. In brief, a fused-silica capillary (Agilent Technologies; total length, 96 cm; i.d., 50 μm) was pre-conditioned with 1 M NaOH for 30 min, followed by MilliQ® water for 30 min and background electrolyte (BGE; 0.8 M formic acid in 10% methanol) for 30 min. Before each analysis, the capillary was flushed for 5 min (950 mbar pressure) with BGE. The MS was operated in positive polarity, with a full scan from 80 to 1000 m/z at a rate of 1.4 scan/s. Drying gas was set to 10 L/min, nebulizer to 10 psi, voltage to 3.5 kV, fragmentor to 125 V, gas temperature to 200°C and skimmer to 65 V. The sheath liquid composition was methanol/water (1/1, v/v), containing 1.0 mM formic acid with two reference masses (121.050873 -purine (C5H4N4) and 922.009798 -HP-921 (C18H18O6N3P3F24)), which allows for correction and provides more accurate mass determination. Flow rate was 0.6 mL/min and split was set to 1/100. Samples were injected at 50 mbar for 50 s. After each injection, along with the samples, BGE was co-injected for 10 s at 100 mbar pressure to improve repeatability. Separations were performed at a pressure of 25 mbar and a voltage of +30 kV; current under these conditions was 100 μA.

Data Acquisition
Quality of the raw data was first inspected by the analysis of the chromatograms/electrochromatograms acquired for experimental samples, QC samples, blanks and internal standard if used. Raw data acquired were processed to provide structured data in an appropriate format for data analysis. The data collected by LC-MS and CE-MS were cleaned of background noises and unrelated ions in recursive analysis in Mass Hunter Profinder (B.06.00, Agilent Technologies) software. Feature extraction is the reduction of acquired data size and complexity through the removal of redundant and nonspecific information by identifying the important variables (features) associated with the data. Molecular feature extraction (MFE) performs chromatographic deconvolution to find the features in the analyzed samples. The features are aligned across all of the selected sample files using mass and retention/migration time. Recursive Feature Extraction first performs MFE and then uses the MFE results, feature mass and retention time, to perform a targeted feature extraction referred to as Find by Ion (FbI). Find by Ion uses the median mass, median retention time, and composite spectrum calculated from the aligned features to improve the reliability in finding the features in the data. GC-MS raw data files were translated to appropriate format with MassHunter Workstation GC/MS Translator (B.04.01, Agilent Technologies) and then processed with the MassHunter Quantitative (B.08.00, Agilent Technologies) software. Deconvolution process and metabolite identification was conducted applying Agilent MassHunter Unknowns Analysis Tool 7.0.

Data normalization and filtration
Quality assurance procedure was applied to check overall data quality (Dudzik et al. 2017). Non-supervised PCA-X projection method was used to evaluate instrumental signal drift, sensitivity loss and variation of the measurement in QC samples. The control Shewhart´s charts were used to plot the sum of acquired signals of detected metabolic features for every analyzed sample against the acquisition order, that enables for fast detection of the measurement abnormalities. When required, the data were normalized. Normalization according to internal standard, methionine sulfone was applied to CE-MS data and normalization by fold change to correct GC-MS data. Variation of the compound concentrations in QC samples expressed as coefficient of variation (%CV) was also calculated. For data filtration, a threshold of 20% for LC-MS and CE-MS and 30% for GC-MS was set for the CV values of metabolites in the QC samples.

Statistical analysis
Data normality was verified by evaluation of the Kolmogorov−Smirnov−Lillefors and Shapiro−Wilk tests and variance ratio by the Levene's test. Differences between experimental groups were evaluated by unpaired t test (equal or unequal variance) or nonparametric (Mann−Whitney test) with post hoc Benjamini Hochberg (FDR) correction. The levels of statistical significance were set at 95% level (P < 0.05). Statistical analyses were performed using Matlab R2015 (Mathworks) software. MetaboAnalyst v5.0 data annotation tool (metaboanalyst.ca) was used for testing the relationships between variables (Xia et al., 2016). Multivariate (unsupervised and supervised) analysis as well as other multivariate calculation and plots was performed by using SIMCA-P+ 14.0 (Sartorius AG/Umetrics, Umea, Sweden; sartorius.com). Combination of VIP-p(corr) (correlation coefficient combined with VIP, Variable Influence on the Projection) based on selected OPLS-DA, with the cutoff set as VIP ≥ 1.0 and p(corr) ≥ 0.4 model was applied for specified interpretations.

IV) Compounds Identification
Accurate mass of statistically significant metabolic features was searched for possible ID against the online available databases as Kegg, Metlin, LipidMaps and HMDB using online available advanced CEU Mass Mediator (CMM) tool (Gil de la Fuente et al., 2018). Isotopic distribution, accurate m/z, retention/migration time for each metabolite feature (LC-MS and CE-MS) as well as fragmentation spectra obtained from LC-MS/MS analysis and analysis of commercially chemical standards if available have been studied for final metabolite identification. Compound identification by GC-MS was performed with the target metabolite Fiehn GC/MS Metabolomics RTL library (G1676AA, Agilent), the CEMBIO-library and the NIST 14 Mass Spectral Library (Babushok et al., 2007).

V) Experiment Validation
OPLS/O2PLS-DA models that were obtained according to multivariate calculations were validated by cross-validation tool. Validation was performed by using the leaving-1/3-out approach. A randomized data set was divided into three parts, and 1/3 of samples were excluded to build a model with the remaining 2/3 of samples. Then, the excluded samples were predicted by the new model, and the process was repeated until all samples have been predicted at least once. Each time the percentage of correctly classified samples was calculated. In the models obtained with data from LC-MS, 63% (ESI+) and 90% (ESI−) of all excluded samples were classified correctly; in data from GC-MS and CE-MS, 80% and 50%, respectively, of the excluded samples were classified correctly. To estimate the predictive power of statistically significant metabolites the multivariate model was created and validated resulting in 92% correct classification of the samples.

Microarray Analysis
For the calculation of pairwise distances between samples, each microarray was considered as a point in a high-dimensional space since we treated each probe as a variable. For parametric analysis of gene set enrichment (PAGE), our expression data was tested using the PAGE method as previously described (Kim and Volsky, 2005). Briefly, for each pathway under each pair 5 of conditions, an aggregated Z score and P value were computed (JMP 6.0 software) to the total Z-ratio in comparison by Ztest. Ingenuity Pathways Analysis© was performed by using the tools supplied by Ingenuity Inc. (Ingenuity Systems; Redwood City, CA).  The experimental groups are vehicle control for (R,S′)-MNF (Control), 20 mgkg -1 (R,S′)-MNF (Arm 1), and 40 mgkg -1 (R,S′)-MNF (Arm 2). The presented data was collected before initiation of treatment (Day 8), at the end of the third cycle of drug administration on Day 29 and at the end of the study on Day 33. Compounds significantly altered in response to (R,S′)-MNF, as per univariate analyses and VIP-p(corr) scores (gray background), were depicted on Figure 4A in the main text. Remaining metabolites were identified based only on the VIP-p(corr) values, thus need to be interpreted with caution as applied OPLS-DA models are prone to over-fitting. Mass, monoisotopic molecular weight in Da; RT, Retention Time; MT, Migration Time; Error, mass error in PPM; p(corr), multivariate correlation coefficient; VIP, Variable Influence on the Projection; % of change expressed in (R,S′)-MNF group; FC, Fold Change; CV, coefficient of variation calculated for QC samples; STD , compounds confirmed by standards; #, FDR P-value; n/s, not significant; n.a., not available. The values represent the average ± SD. Control group, n = 10; (R,S′)-MNF group, n = 9. *, **, ***, ****, P-value < 0.05, 0.01, 0.001, 0.0001; n/s, not significant. Significance is defined as Z-score > 1.5 in either direction, false discovery rate (fdr) < 0.3 and p < 0.05. N = 10 mice per experimental group. The number of genes in a given GO Term is provided. * The z-ratios of significantly up (+)-and down (-)-regulated genes in the (R,S′)-MNF:Vehicle pairwise comparison are shown. Significance is defined as z-ratio > 1.5 in either direction, false discovery rate (fdr) < 0.3 and p < 0.05. N = 10 mice per experimental group.