Metabolic profiling reveals distinct metabolic alterations in different subtypes of pituitary adenomas and confers therapeutic targets

All tumor samples were obtained following transsphenoidal surgery performed at Beijing Tiantan Hospital. Fresh tumor samples were stored in liquid nitrogen. 63 tissue samples include 56 pituitary adenomas and seven normal pituitary glands. 56 pituitary adenomas were classified according to the 2017 WHO classification. In the present study, 56 cases include 10 cases of corticotroph adenomas (five silent ACTH and five functional ACTH), ten cases of gonadotroph adenomas, eight cases of somatotroph adenomas, eight cases of mammosomatotroph adenoma, five cases of lactotroph adenomas, ten cases of oncocytomas, and five cases of null cell adenomas.

This study was approved by the Ethics Committee of Beijing Tiantan Hospital (KY2013-015-02). Written informed consent was obtained from all of the enrolled subjects, and the study was performed in compliance with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

For the gas chromatography–mass spectrum (GC–MS) analysis, tissue samples were mixed with 600 µl methanol/water (v/v 4:1) solution containing internal standards followed by thorough homogenization in a mixed grinding apparatus (MM400, Retsch, Germany) at 25 Hz for 2 min. Supernatants were lyophilized for subsequent oximation and silylation reactions. A QP 2010 GC–MS system (Shimadzu, Japan) with a DB-5 MS fused-silica capillary column (30 m × 0.25 mm × 0.25 µm, Agilent Technologies, Santa Clara, CA, USA) was used for the metabolic profiling. A pseudotargeted GC–MS metabolomics method was established elsewhere [11‐13]. A total of 280 ion features assigned to 23 groups were defined for the data collection and quantification. The system parameter settings were previously described [13]. Metabolite identities were determined based on commercial libraries (Mainlib, NIST, Wiley, and Fiehn) and an internal metabolite library.

Total RNA was extracted and purified using the mirVana™ miRNA Isolation Kit (Cat# AM1561, Ambion, Austin, TX, US) following the manufacturer’s instructions and checked for an RIN number to inspect the RNA integration using an Agilent Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA, US). Total RNA was amplified and labeled with the Low Input Quick Amp Labeling Kit, One-Color (Cat# 5190-2305, Agilent Technologies, Santa Clara, CA, US), following the manufacturer’s instructions. Labeled cRNA was purified using an RNeasy mini kit (Cat# 74106, Qiagen, GmBH, Germany). Each slide was hybridized with 1.65 µg Cy3-labeled cRNA using the Gene Expression Hybridization Kit (Cat# 5188-5242, Agilent Technologies, Santa Clara, CA, US) in a hybridization oven (Cat# G2545A, Agilent Technologies, Santa Clara, CA, US), according to the manufacturer’s instructions. After 17 h of hybridization, the slides were washed in staining dishes (Cat# 121, Thermo Shandon, Waltham, MA, US) with the Gene Expression Wash Buffer Kit (Cat# 5188-5327, Agilent Technologies, Santa Clara, CA, US), following the manufacturer’s instructions. Slides were scanned by an Agilent Microarray Scanner (Cat# G2565CA, Agilent Technologies, Santa Clara, CA, US) with the default settings: dye channel = green, scan resolution = 5 µm, PMT 100%, 10%, 16 bit. Data were extracted with Feature Extraction software 10.7 (Agilent Technologies, Santa Clara, CA, US). The raw data were normalized using the quantile algorithm and limma packages in R.

Proteins from pituitary adenoma tissues and normal pituitary glands were extracted using a total protein extraction kit (Millipore, Billerica, MA, USA). Protein concentrations were determined using the bicinchoninic acid protein assay (Pierce, Rockford, IL, USA). Proteins from each group of tissues were equally combined into a single pool as previously described [14]. A mass of 100 μg of each pooled sample was then denatured, reduced and alkylated as described in the iTRAQ protocol (Thermo Fisher Scientific, Waltham, MA, USA) and digested overnight with 0.1 µg/µl trypsin solution at 37 °C. The digested pooled samples were subsequently labeled with 113, 115, 117, 118, 119 and 121 iTRAQ tags, respectively, according to the manufacturer’s protocol (AB Sciex, Framingham, MA, USA). The tagged peptides were dried via vacuum centrifugation and combined in one tube. Strong cation-exchange (SCX) chromatography was then performed as previously described [15, 16]. In brief, pooled samples were separated on an apoly-LC SCX column (4.6 × 250 mm, 5 μm, 100 Å) using an LC 100 instrument (Eksigent Technologies, Dublin, CA, USA), and then the labeled peptides were detected by ultraviolet radiation using SPD-20 (Shimadzu, Kyoto, Japan). A total of 48 fractions were collected, dried by speed vacuum centrifugation, and combined into 10 fractions according to the SCX chromatogram. Each fraction was injected onto a desalting column (350 µm × 0.5 mm, 3 µm C18, 120 Å) and separated on an analytical column (75 µm × 150 mm, 3 µm C18, 120 Å) using an Eksigent NanoLC instrument (Eksigent, Dublin, CA, USA). The samples were separated by capillary high-performance liquid chromatography and were subsequently analyzed using a Triple TOF 5600+ system (AB Sciex).

Protein identification and proteome annotation were performed using the ProteinPilot software package, version 4.5 (Applied Biosystems), and cross-referenced against the SwissProt database (March 2013) using the Mascot 2.2 search engine (Matrix Science, London, UK). The search parameters utilized to analyze the MS/MS data included: trypsin as the digestion enzyme with a maximum of two missed cleavages allowed; fixed modifications of carbamidomethyl (C) and iTRAQ Plex (K and N-terminus); variable modifications of oxidation (M); peptide mass tolerance of ± 20 ppm; fragment mass tolerance of ± 0.1 Da; and a peptide false discovery rate (FDR) of less than or equal to 0.01.

The GH3 rat pituitary cell line was purchased from the China Institute of Cell Line Resources and cultured in phenol red-free Dulbecco’s modified Eagle’s medium (DMEM, Invitrogen, China) supplemented with 10% fetal bovine serum (Gibco, Auckland, USA) in a humidified incubator at 37 °C with 5% CO₂. IDH2-targeted shRNA (TF702283A-D) was obtained from OriGene Clone (OriGene Technologies, MD, USA). Transfection was performed using Lipofectamine 3000 (Invitrogen, Carlsbad, CA, USA) following the manufacturer’s instructions. A total of 1 µg of plasmid was used to transfect 1 × 10⁶ GH3 cells.

GH3 cells were transfected with shRNA or empty control vectors (noneffective 29-mer scrambled shRNA cassette in pGFP-C-shLenti vector). Cultures were adjusted to a density of 1 × 10⁵ cells/ml, and 100 µl of cell suspension was plated into each well of a 96-well plate and cultured for 0 h, 24 h or 48 h before adding 20 µl of MTS tetrazolium solution to each well with incubation for an additional 4 h. Absorbance at 490 nm was measured using an ELISA plate reader (Thermo, USA).

Cell migration was assayed on fibronectin- and Matrigel-coated polycarbonate filters in modified Transwell chambers (Corning, USA). GH3 cells (5 × 10⁴ cells) were introduced into the upper chambers. The time of incubation in the chambers was 24 h. Cells adhering to the lower membrane surface were fixed in 4% paraformaldehyde and stained with hematoxylin (Zhongshan Company, Beijing, China). The average number of migrated cells in five randomly chosen high-power fields was determined under light fields with fluorescent microscope (ZEISS, Jena, Germany). The assays were performed in triplicate.

Total RNA was extracted from GH3 cells with shRNA or empty control vectors using TRIzol reagent (Qiagen), and qRT-PCR was performed as previously described [17] with an Applied Biosystems 7500 Fast System (Life Technologies, Carlsbad, USA). The fold-change in differential expression for each gene was calculated using the comparative CT method (2−∆∆CT method) as previously described [17]. GAPDH was used as the housekeeping gene. The primers are listed in Additional file 1: Table S1).

We collected ACTH-PA, silent ACTH-PA, GT-PA, GH-PA, PRL-PA, GH-PRL-PA, oncocytomas, NC-PA and normal pituitary glands from patients or donors and subjected these samples to metabolomic analysis. The heatmap of the metabolite profiles showed that all subtypes of pituitary adenomas were clustered together and generally different from the normal pituitary glands (Additional file 2: Figure S1). As expected, GH-PA, GH-PRL-PA, and PRL-PA formed a subclass, showing a close relationship among these three subtypes of adenomas. Intriguingly, the metabolomes of oncocytoma, GT-PA, and NC-PA seemed very similar. Silent ACTH-PA displayed widespread downregulation of metabolites compared to the other subtypes of adenomas, which is consistent with its silent clinical and biochemical manifestations [18].

When further looked into the metabolites involved in different biological processes, we generally found decreased levels of tricarboxylic acid cycle (TCA) flux, glycolysis (Fig. 1a), and essential amino acids (Fig. 1b). However, the metabolites involved in nucleotide metabolism and short-chain fatty acids (Fig. 1c) were upregulated compared to that in the normal pituitary tissues.

In terms of energy metabolism, all eight subtypes of adenomas studied showed reduced levels of glucose, glucose-6-P, and fructose-6-P (Fig. 1a). Phosphoenolpyruvate (PEP) was also decreased in all samples except GT-PA and PRL-PA, while no significant reductions in pyruvate and lactate levels were observed. These findings suggested that glucose uptake is inhibited in pituitary adenomas, seemingly contradictory to Warburg’s effect. In addition, we noticed that TCA cycle-related metabolites were generally recovered downstream of isocitrate, implying that anaplerosis of glutamate might be enhanced to supplement the TCA cycle flux (Additional file 3: Figure S2). In agreement, the level of glutamate was significantly higher in GH-PA and GH-PRL-PA (p < 0.05), which also showed higher levels of alpha-ketoglutarate (α-KG) than those in normal pituitary glands (p < 0.05) (Additional file 4: Table S2). Therefore, decreased energy metabolism was a common alteration in pituitary adenomas.

The levels of essential amino acids were predominantly decreased. In particular, methionine, valine, and leucine were downregulated in all subtypes of pituitary adenomas. The abundance of lysine, tryptophan, and phenylalanine was also reduced except in ACTH-PA (Fig. 1b). Furthermore, the levels of pyrophosphate, zymosterol, and adenosine-5-monophosphate were increased, while those of uric acid was upregulated or unchanged in pituitary adenomas (Fig. 1c), suggesting enhanced nucleotide metabolism. The levels of short-chain fatty acids such as nonanoic acid, heptanoic acid, octanoic acid, capric acid, and hexanoic acid were mostly increased, especially in ACTH-PA and GH-PA (Fig. 1c). Although the detailed relationship is unclear, changes in amino acids, nucleotides, and fatty acids are suggested to be connected with the endocrinal function of these cells.

To further understand the genetic regulation of metabolic alterations in pituitary adenomas, we performed RNA microarray in the eight subtypes of adenomas, with normal pituitary glands as the controls. The heatmap showed similar clustering of these adenomas compared to the metabolome (Additional file 5: Figure S3). Notably, the two clusters (Cluster 1: PRL-PA/GH-PA/GH-PRL-PA, and Cluster 2: GT-PA/NC-PA/oncocytoma) were again automatically formed. Intriguingly, the transcriptomes of ACTH-PA and silent ACTH-PA were quite similar, though their metabolomes were distinct, suggesting that synthesis of ACTH was associated with gene expression while secretion of ACTH was more likely to be associated with cellular metabolism.

Since subtype-specific metabolic reprogramming may provide novel targets for intervention in the corresponding tumors, we performed a proteomic analysis in the eight subtypes of adenomas, with normal pituitary glands as the controls. The integrated results of the metabolomics and proteomics analyses in the eight subtypes of adenomas showed several changes in metabolic pathways based on the enrichment and topology analyses (Additional file 6: Table S3). Among these, glucose metabolism and arginine/proline metabolism pathways were found to be GH-PA-specific when compared to the other subtypes of pituitary adenomas (Fig. 2a and Additional file 6: Table S3). The integrated results of the metabolome and proteome analyses in GH-PA implied that three parts of the glucose metabolism pathway were significantly altered (Fig. 2b). We inferred that the overexpression of IDH2 was biologically significant, with downregulated levels of isocitrate and upregulated levels of α-KG. Therefore, we hypothesized that IDH2 inhibition might be effective for limiting the growth of GH-PA cells.

To test the therapeutic effects of IDH2 inhibition in GH-PA, we used four IDH2 shRNA plasmids (SH-A/B/C/D) to block IDH2 expression in GH3 cells. The IDH2 mRNA levels (SH-A/B/C/D) were 103 ± 21.5%, 77.4 ± 15.2%, 21.3 ± 11.7% and 16.2 ± 8.3%, respectively, of that in the control group 72 h after transfection, as shown in Fig. 3a. The IDH2 protein levels as determined by Western blotting showed a similar trend (Fig. 3b, c). Therefore, SH-C and SH-D were selected for the further experiments. As expected, cell viability was inhibited 3 days after IDH2 knockdown (Fig. 3d). The average number of migrated cells was reduced after IDH2 SH-C and SH-D blockage in GH3 cell lines in Fig. 3e.

The exocrine function of GH3 cells was downregulated, and decreased levels of GH and insulin-like growth factor 1 (IGF-1) (Fig. 4a) were observed 3 days after IDH2 knockdown. We further detected the mRNA levels of some key GH-related and migrated-related genes, and revealed that all these genes were dramatically inhibited, including MMP2, N-cadherin, IL-6, STAT3 and STAT5 (Fig. 4b, c). These findings suggest that targeting IDH2 could reduce both the viability and functional activity of GH-PA cells.