Introduction

The anterior lobe of the pituitary gland is a structurally and functionally complex endocrine organ comprising five types of hormone-producing cells and several non-endocrine cells. Accumulating studies show that in addition to regulation via the hypothalamic-pituitary-target organ axis, these cells interact with each other within the gland to maintain pituitary functions (Denef 2008; Mollard et al. 2012). The cellular interactions also form microenvironmental niches among non-endocrine stem/progenitor cells, which are considered important for supplying endocrine/non-endocrine cells to the gland (Yoshida et al. 2016, 2017). Until now, several tissue stem/progenitor cells have been identified including some populations of folliculostellate cells expressing S100β protein (Inoue et al. 2002; Vankelecom 2007; Devnath and Inoue 2008), Sox2-positive cells (Fauquier et al. 2008; Chen et al. 2009; Andoniadou et al. 2013; Rizzoti et al. 2013) and more recently, a cell type originating from the neural crest (Ueharu et al. 2017). These cells have the potential to differentiate into certain pituitary cell types in response to stimuli from adjacent cells to maintain cellular homeostasis within the gland. However, the molecular mechanisms underlying the differentiation processes of pituitary cells and their induction stimuli are not clear. Establishment of a cell model for such pituitary stem/progenitor cells is expected to enrich our understanding of differentiation mechanisms in pituitary cells.

TtT/GF is a mouse pituitary cell line derived from a radiothyroidectomy-induced thyrotropic pituitary tumor (Inoue et al. 1992). These cells do not produce the classic adenohypophyseal hormones but react with antiserum against glial fibrillary acidic protein (GFAP), which is expressed in folliculostellate cells in the anterior lobe of the pituitary gland. Similar to folliculostellate cells, these GFAP-positive tumor cells cover the neighboring glandular cells with their long cytoplasmic processes and display phagocytic activity and follicular formation with S100β positivity (Inoue et al. 1992). From these profiles, TtT/GF cells have been recognized for a long time as a model of folliculostellate cells to investigate their molecular and cellular functions (Renner et al. 1997; Tierney et al. 2003; Stilling et al. 2005; Vitale and Barry 2015). On the other hand, TtT/GF cells express Sox2 and several stem/progenitor cell marker genes such as drug-efflux ATP-binding cassette (ABC) transporters and stem cell antigen-1 (Sca1) (Chapman et al. 2002; Mitsuishi et al. 2013). Our recent microarray analysis further demonstrated that TtT/GF cells display some characteristics of endothelial cells and pericytes (capillary mural cells), as they express hematopoietic stem/progenitor cell markers (Cd34) as well as pericyte markers like chondroitin sulfate proteoglycan 4 (Cspg4/Ng2) and nestin (Nes) (Mitsuishi et al. 2013; Yoshida et al. 2014). From these observations, we hypothesized that TtT/GF cells are not terminally differentiated cells and have the potential to differentiate into folliculostellate cells, endothelial cells, or pericytes, i.e., they do not merely represent a model for folliculostellate cells.

In order to examine the plasticity of TtT/GF cells, we explored factors that could push the cells toward a more differentiated state and found that transforming growth factor beta (TGFβ) reinforces pericyte properties. Briefly, TGFβ treatment increases the transcripts of pericyte markers (Cspg4 and Nes) and reduces several stem/progenitor cell markers (Sox2, Sca1 and Cd34) (Tsukada et al. 2018). However, the protein level changes in TtT/GF cells upon TGFβ treatment remain unclear. Therefore, the present study utilizes mass spectrometry analysis combined with the stable isotope labeling of amino acids in a cell culture (SILAC) system (Berger et al. 2002; Ong et al. 2002; Zhu et al. 2002) to extensively characterize TGFβ-treated TtT/GF cells.

Materials and methods

Cell culture and labeling of TtT/GF cells

The TtT/GF cell line was kindly provided by Dr. Kinji Inoue (Saitama University) and was used for experiments at passage 20–25. For light medium, DMEM-F/12 containing 10% dialyzed fetal bovine serum was supplemented with 100 μg/ml L-lysine, 100 μg/ml L-arginine, 2.5 mM L-glutamine, 3.2 g/l D-glucose, 8 mg/l phenol red (supplemented with SILAC DMEM-Flex Media; Life Technologies, Carlsbad, CA, USA) and antibiotics (0.5 U/ml penicillin and 0.5 μg/ml streptomycin; Life Technologies). For heavy medium, L-lysine and L-arginine were replaced with 13C6-labeled L-lysine and L-arginine, respectively. The cell culture protocol including treatment with recombinant TGFβ (human TGFβ2; PeproTech, Rocky Hill, NJ, USA) and a selective TGFβ receptor I inhibitor (SB431542; Merck Millipore, Billerica, MA, USA) (Inman et al. 2002) is shown in Fig. 1. Briefly, cells were seeded in 60-mm dishes at a density of 1.0 × 105 cells/3 ml/dish and were maintained at 37 °C in an incubator with 5% CO2 for 3 days (> three doublings). The cells were then replated in 100-mm dishes at a density of 5.0 × 105 cells/10 ml/dish with poly-L-lysine-coated coverslips (Iwaki, Tokyo, Japan) and incubated with 10 ng/ml TGFβ or 10 μM SB431542 for 3 additional days (> three doublings) to ensure complete incorporation of 13C6-labeled L-lysine and L-arginine. Dimethyl sulfoxide and 0.1% bovine serum albumin were used as a vehicle and added to the light medium. An IX71 inverted microscope (Olympus, Tokyo, Japan) was used for observation of live cells.

Fig. 1
figure 1

Experimental workflow using SILAC. TtT/GF cells were cultured for 3 days (> three doublings) in either light medium (containing normal Arg/Lys) or in heavy medium (containing 13C6-labeled Arg/Lys) for incorporation. The cells were then replated onto new cell culture plates with fresh medium containing vehicle (light), TGFβ (heavy), or the selective TGFβ receptor I inhibitor (SB431542, heavy) at the indicated concentrations. After 3 additional days in culture (> three doublings), cells were lysed for protein extraction. The collected proteins were mixed at a 1:1 ratio (vehicle vs. TGFβ and vehicle vs. SB431542) and analyzed by mass spectrometry. Relative TGFβ vs. SB431542 comparisons were calculated and normalized based on the data obtained from vehicle vs. TGFβ and vehicle vs. SB431542 comparisons

Full size image

Protein extraction

Proteins were extracted from TtT/GF cells using ProteoExtract Complete Mammalian Proteome Extraction Kit (Merck, Darmstadt, Germany). In addition to the manufacturer’s protocol, genomic DNA was broken down by five cycles of 28 kHz sonication (30 s per cycle) in ice-cold water using an ultrasonic machine (VS-100III; AS ONE, Osaka, Japan) and then needle sheared before centrifugation to obtain extracted proteins in the supernatant. The protein concentration was measured by Bradford assay (Bio-Rad, Hercules, CA, USA) using bovine serum albumin as a standard protein.

Tryptic digestion and peptide fractionation

Detailed procedures of digestion by trypsin to produce peptides are described in our previous report (Kito et al. 2016). Briefly, proteins from vehicle-treated cells (cultured in light medium) were mixed with an equal amount of proteins from the TGFβ- or SB431542-treated cells (cultured in heavy medium). After precipitation of 240 μg proteins by methanol/chloroform, proteins were re-dissolved in 0.1 M Tris-HCl buffer (pH 8.5) containing 8 M urea, followed by cysteine reduction with DTT (final concentration, 5 mM) and alkylation by iodoacetamide (final concentration, 10 mM). Urea concentration was then reduced to 2 M, followed by trypsin addition (substrate to enzyme ratio, 20:1) and overnight incubation at 37 °C. Tryptic peptides were separated into 24 fractions by gel-free isoelectric focusing using a model 3100 OFFGEL fractionator (Agilent Technologies, Santa Clara, CA, USA) according to the manufacturer’s protocol and a previous report (Kito et al. 2016).

Mass spectrometric analysis

Identification and quantification of fractionated peptides were carried out by liquid chromatography-tandem mass spectrometry (LC-MS/MS) on a DiNa nano LC system (KYA technologies, Tokyo, Japan) and an LTQ-Orbitrap XL mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA), according to the experimental procedures described in a previous publication (Okada et al. 2017). The MS/MS spectra acquired in a data-dependent mode were searched against a database containing the amino acid sequences of Mus musculus, which was originally downloaded from the UniProt database (http://www.uniprot.org; Proteome ID, UP000000589; released on 17-Apr-2016). To reduce the redundancy of protein sequences, 12,494 protein sequences that were completely covered by others with more than 99% identity were removed. Mass shift of the isotopically labeled L-lysine and L-arginine (+ 6.020) was considered in the database search and the list of peptides identified with a false discovery rate less than 1% was subjected to calculation of ratios of peak intensities between light (non-labeled) and heavy (labeled) peptide ions. Each protein ratio was determined as a median value of the calculated ratios of peptide ion intensities. The list of proteins whose heavy-to-light ratios were successfully calculated from two or more ratios of peptide ions was used for further analysis. Proteins that showed logarithmic ratios greater than 0.6 or less than − 0.6 with corrected p values (Benjamini and Hochberg method) less than 0.05 were determined to be altered in expressional abundance between the two different conditions.

Gene ontology analysis

Gene ontology (GO) analysis was performed using the Database for Annotation, Visualization and Integrated Discovery v6.8 (DAVID, https://david.ncifcrf.gov) to demonstrate the biological processes enriched in the altered proteins between the different conditions. Among the protein names sharing identical quantified peptides, which seemed to be members of protein isoforms or families, only one protein name was subjected to GO analysis in order to avoid overestimation of biological meaning of the enriched GO terms. The resulting GO terms with p values under 0.05 were considered enriched biological processes.

STRING analysis

To determine the presence of interactions/partnerships among TGFβ-induced proteins, protein-protein interaction networks of upregulated and downregulated proteins were extracted from the STRING database (https://string-db.org) and drawn by STRING v10 (Szklarczyk et al. 2015).

Phalloidin staining and cell count

To examine the actin cytoskeleton architecture, we stained TGFβ- and/or SB431542-treated cells with phalloidin. In this experiment, the cells were seeded in 35-mm dishes at a density of 6.25 × 104 cells/3 ml/dish with poly-L-lysine-coated coverslips and incubated with 10 ng/ml TGFβ and/or 10 μM SB431542 for 3 days. The cells attached to coverslips were briefly washed twice in Hank’s balanced salt solution (Life Technologies) and fixed with 4% paraformaldehyde in 25 mM phosphate buffer (PB; pH 7.4) for 20 min. The cells were then permeabilized in phosphate-buffered saline (PBS) containing 0.2% Triton X-100 (Sigma-Aldrich, St. Louis, MO, USA) for 20 min and incubated with phalloidin-conjugated Alexa Flour 488 (1:100 dilution; Life Technologies) and 4′,6-diamidino-2-phenylindole (DAPI; 500 ng/mL, Dojindo Laboratories, Kumamoto, Japan) for 1 h. After washing with PBS, the coverslips were mounted on slides using Aqua Poly/Mount (Polysciences, Warren, PA, USA) and subsequently analyzed on an FV1000 confocal laser microscope (Olympus). Images were processed for presentation using Photoshop CS and Illustrator CS (Adobe Systems, San Jose, CA, USA).

For cell counting, stained cells were picked randomly and imaged with a 20× objective lens. DAPI-positive cells/image/1.527 mm2 were counted using ImageJ software v1.43 (National Institutes of Health, Bethesda, MD, USA). The results are based on five independent experiments (n = 5).

Transmission electron microscopy

Vehicle- and TGFβ-treated cells were processed for transmission electron microscopy according to a protocol described in our previous report (Horiguchi et al. 2011). Briefly, cells were fixed with 2.5% glutaraldehyde in 0.1 M PB (pH 7.4) and then were postfixed with 2% OsO4 in 0.1 M PB. After dehydration in a series of graded alcohols followed by embedding in epoxy resin (Quetol 812; Nissin EM Co., Tokyo, Japan), the cells were sectioned into ultrathin slices with a Reichert-Nissei Ultracut S (Leica Microsystems, Wetzlar, Germany). The sections were stained with uranyl acetate and lead citrate and then observed under a HT7700 electron microscope (Hitachi, Tokyo, Japan).

Statistical analysis

Cell count results are presented as the mean ± SEM. One-way analysis of variance (ANOVA) followed by Tukey’s test for multiple comparisons was performed using Prism v6 (GraphPad Software, San Diego, CA, USA). p value < 0.05 was used to indicate statistical significance.

Results

Effect of TGFβ and SB431542 on cell morphology

According to the experimental workflow shown in Fig. 1, TtT/GF cells were treated with TGFβ for 3 days in heavy medium containing 13C6-Lys/Arg. As TtT/GF cells express TGFβ, we used SB431542 (a selective TGFβ receptor I inhibitor) to exclude the endogenous TGFβ activity. The efficiencies of TGFβ and SB431542 have been determined previously (Tsukada et al. 2018). The present study confirmed that the dose of TGFβ in SILAC media was sufficient to alter the intracellular localization of Smad2 and the pericyte and stem/progenitor cell marker gene expression and that SB431542 nulled the actions of TGFβ (Electronic Supplementary Material, Figs. 1 and 2). Before protein extraction, we checked the effect of TGFβ and SB431542 on cell morphology (Fig. 2). Compared with vehicle-treated cells, TGFβ-treated cells displayed an elongated shape and appeared less dense. The shape of SB431542-treated cells was similar to that of vehicle-treated cells.

Fig. 2
figure 2

TGFβ induces cellular elongation in TtT/GF cells. TtT/GF cells were treated with TGFβ (10 ng/mL) or TGFβ receptor I inhibitor (SB431542; 10 μM) for 3 days and cell morphology was observed. Magnified views of a, b and c are shown in a′, b′ and c′, respectively. TGFβ-treated TtT/GF cells showed an elongated shape (b, b′). Bars = 100 μm (both)

Full size image

Extraction of TGFβ-induced proteins from mass spectrometry data

All mass spectrometry analyses were performed in duplicate and in total, three individual experiments (E1, E2 and E3) were performed to verify the actions of TGFβ and SB431542. Approximately 3000–4000 proteins were identified in each mass spectrometry analysis and approximately 1200–2000 of these were successfully quantified (Electronic Supplementary Material, Table S1). In order to compare the expression profile obtained by mass spectrometry analysis in each treatment, we first generated a heat-map for proteins that showed differential expression between different treatments (Fig. 3). The result showed an inverse relationship between TGFβ and SB431542 treatments. From the volcano plot in Fig. 4(a), we extracted proteins that were altered > 1.5-fold (|log2| > 0.6) with statistical significance (p < 0.05) upon TGFβ and SB431542 treatments and further narrowed them down by omitting proteins that carried significance in only one of three individual experiments (Fig. 4b and Electronic Supplementary Material, Table S2). Through the filtering process, a total of 105, 37 and 187 proteins shown in red in Fig. 4(b) were extracted when comparing vehicle vs. TGFβ, vehicle vs. SB431542 and TGFβ vs. SB431542, respectively. For subsequent data analysis, we utilized the TGFβ vs. SB431542 comparison, as vehicle-treated cells might be affected by endogenous TGFβ. Finally, proteins that are almost identical (e.g., merely different tissue localization as the protein annotation) were merged into a single protein, resulting in the identification of 163 proteins induced by TGFβ (51 upregulated and 112 downregulated; Electronic Supplementary Material, Tables S3 and S4).

Fig. 3
figure 3

Inverse profiles between TGFβ-treated and SB431542-treated TtT/GF cells. After three individual mass spectrometry analyses (E1, E2, and E3), protein expression profiles were depicted as a heat-map (on a logarithmic scale with base 2). Red and green indicate increase and decrease in protein abundance, respectively. The data show an inverse profile between vehicle vs. TGFβ and vehicle vs. SB431542 comparisons, suggesting that the identified proteins are influenced by TGFβ signaling. (Hierarchical clustering of protein expression profiles and heat-map generation were performed with R statistical software.)

Full size image
Fig. 4
figure 4

Identification of proteins altered by TGFβ and SB431542 treatments. a, a’, a” Volcano plot showing fold-change (as log2) of each comparison (representative plots). Red dots indicate proteins whose logarithmic values (log2) of ratios differed from the median value by more than 0.6 (approximately 1.5-fold-change) with corrected p values (Benjamini and Hochberg method) less than 0.5. b, b’, b” Venn diagrams showing the overlap of proteins with significant changes in three individual experiments (E1, E2 and E3). Significantly altered protein expression in more than two experiments was considered a significant change (red numbers). For individual data, see Electronic Supplementary Material, Tables S1 and S2

Full size image

GO and STRING analyses

TGFβ-induced proteins (163 proteins) were then subjected to GO analysis. GO terms related to actin cytoskeleton, adhesion, extracellular matrix (ECM) and DNA replication were highlighted. The GO analysis also detected proteins related to differentiation, endocytosis, metabolism and oxidation-reduction process (Table 1; for all data, see Electronic Supplementary Material, Table S5). To examine protein-protein interactions among the TGFβ-induced proteins, we performed STRING analysis (Fig. 5). Consistent with GO analysis, the STRING analysis depicted robust protein-protein networks related to actin cytoskeleton, cell adhesion, ECM, DNA replication, metabolic pathways and oxidation-reduction process.

Table 1 Gene ontology analysis (selected and sorted)
Full size table
Fig. 5
figure 5

TGFβ pathway mainly regulates actin cytoskeleton, cell adhesion, extracellular matrix, metabolic system and DNA replication. All upregulated and downregulated proteins induced by TGFβ (see Electronic Supplementary Material, Tables S3 and S4) were subjected to STRING analysis for identification of protein-protein interaction networks. Robust networks for actin cytoskeleton (blue, 11 proteins, GO:0015629), cell adhesion (red, 17 proteins, GO:0007155), extracellular matrix (green, 16 proteins, GO:0031012), metabolic pathways (magenta, 27 proteins, GO:01100) and DNA replication (yellow, five proteins, GO:0006270) were extracted. STRING analysis also extracted networks for cell differentiation (37 proteins, GO:0030154) cell projection (21 proteins, GO:0042995), and cell motility (12 proteins, GO:0048870). These networks were all connected to each other

Full size image

Effect of TGFβ and SB431542 on actin filament architecture and cell growth

To confirm the results of GO and STRING analyses in vitro, we examined whether TGFβ actually regulates the actin cytoskeleton and cell growth. We first stained actin filaments with phalloidin in TGFβ/SB431542-treated cells (Fig. 6a–d’). The result showed that the TGFβ-treated cells displayed directionally aligned actin filaments, whereas the vehicle- and SB431542-treated cells showed randomly aligned actin filaments. TGFβ-mediated actin filament reorganization was blocked in the presence of SB431542. We also checked actin architecture by transmission electron microscopy (Fig. 6e, f). TtT/GF were agranular cells as described (Inoue et al. 1992) and TGFβ-treated cells displayed more actin bundles compared with vehicle-treated cells. The number of cells was significantly reduced in TGFβ-treated cells compared with that of vehicle- and SB431542-treated cells (Fig. 6g).

Fig. 6
figure 6

TGFβ alters actin filament organization and inhibits cell growth. a–d Bright field images superimposed on images of filamentous actin stained with phalloidin conjugated to Alexa Fluor 488 (green). DAPI was used to stain the nucleus (blue). a’–d’ High-magnification images of each cell. Actin filaments were multi-directionally stretched in vehicle- and SB431542-treated TtT/GF cells (a, a’, c, c’); however, TGFβ-treated TtT/GF cells showed directionally aligned actin filaments (b, b’). TGFβ-induced actin filament reorganization was not observed in the presence of the TGFβ receptor I inhibitor (SB431542, d, d’). Bars = 50 μm (a–d), 10 μm (a’–d’). e, f) Electron micrographs of vehicle-treated cells and TGFβ-treated cells. Actin bundles (arrowheads) were often observed in TGFβ-treated cells (e). Bar = 1 μm (e, f). g Cell count in each treatment. TGFβ treatment significantly inhibited the cell growth of TtT/GF cells, whereas addition of SB431542 nulled the effect of TGFβ. *, **, p < 0.05, 0.01, respectively (Tukey’s test)

Full size image

Characterization of TGFβ-treated TtT/GF cells

From the list of upregulated and downregulated proteins in Electronic Supplementary Material, Tables S3 and S4, we searched for marker proteins that can characterize TGFβ-treated TtT/GF cells and summarized these in Table 2. Remarkably, several pericyte markers were found to be upregulated (CSPG4, NES, ACTA and TAGLN). TGFβ also upregulated smooth muscle cell markers, though some markers overlap with those for pericytes. Interestingly, proteins known to be expressed in pituitary adenomas (CTSB, EZR and LGALS3) were downregulated by TGFβ. In addition, LAMA5, THY1, and ANXA1, which are markers for stem cells, thyrotropes and folliculostellate cells, respectively, were downregulated by TGFβ.

Table 2 Characterization of TGFβ-treated TtT/GF cells (selected and sorted)
Full size table

Discussion

TGFβ is a multi-functional cytokine that promotes epithelial-mesenchymal transition (EMT), tissue development, cell differentiation and fibrosis (for reviews see Beyer et al. 2013; Weiss and Attisano 2013; Gonzalez and Medici 2014; Meng et al. 2016). Our previous study found that TtT/GF cells express TGFβ receptor I and II, which form a heterodimer to act as the Smad-dependent canonical TGFβ receptor and that TGFβ reinforces pericyte properties in these cells (Tsukada et al. 2018). The present study extensively explored the effect of TGFβ on TtT/GF cells by utilizing SILAC-assisted mass spectrometry and successfully identified 51 upregulated and 112 downregulated proteins induced by TGFβ.

Reorganization of actin cytoskeleton

One major action of TGFβ was reorganization of actin cytoskeletal architecture. GO and STRING analyses revealed many proteins associated with actin cytoskeleton components (e.g., smooth muscle actin, α-actinin and tropomodulin-3) and regulators (e.g., RRAS2, ROCK2 and spectrin). During EMT, TGFβ-induced actin cytoskeleton reorganization is thought to be regulated by LIM-domain proteins (Järvinen and Laiho 2012). Indeed, we found several LIM-related proteins including PDLIM1, 2 and 7, FHL1 and LIMA1 from our mass spectrometry data. STRING analysis also revealed that actin cytoskeleton-related proteins were intimately linked with cell adhesion molecules (e.g., integrins, NCAM-1 and Thy-1) and ECM components (e.g., collagen, laminin and fibronectin). Altered interaction among actin-binding proteins, cell adhesion molecules and ECM components may result in elongation with well-aligned actin filaments in TGFβ-treated TtT/GF cells.

Unlike previous observations (Tsukada et al. 2018), TGFβ did not induce colony formation in TtT/GF cells. The present study utilized dialyzed fetal bovine serum for the SILAC experiments instead of normal fetal bovine and horse sera; hence, small molecules including steroids and lysophospholipids in the serum were absent. Thus, TGFβ-induced colony formation might require another unknown factor present in normal serum.

Regulation of extracellular TGFβ signaling components

The present study showed upregulation of extracellular regulators of TGFβ such as fibrillin-1, fibronectin-1, latent TGFβ-binding protein-1 (LTBP1) and integrin α5. These proteins bind TGFβ extracellularly to form a large latent TGFβ complex and TGFβ is activated by liberation from the complex (Gressner et al. 2007). Our previous study showed that TtT/GF cells express TGFβ as well as the TGFβ receptor (Tsukada et al. 2018), suggesting that endogenous TGFβ acts in a paracrine/autocrine fashion. It is likely that endogenous TGFβ acts on TtT/GF cells and suppresses its own action by increasing the extracellular regulators of TGFβ through negative feedback. As knockdown and aberrant expression of these regulators result in Marfan and Ehlers-Danlos syndromes in humans and developmental abnormalities in model animals (ten Dijke and Arthur 2007), TtT/GF cells can be used to study the pathological processes involved in these syndromes and developmental abnormalities.

Tumor suppression

TtT/GF cells are a mouse cancer cell line that originated from a thyrotropic pituitary tumor (Inoue et al. 1992). The present study revealed that TGFβ significantly downregulated cathepsin B, ezrin and galectin-3 expression. These proteins are known to be expressed in several pituitary adenomas (Righi et al. 2013; Tanase et al. 2014; Chen et al. 2017) and reduction of ezrin and galectin-3 exerts an inhibitory action on cancer progression, invasion and proliferation (Huang et al. 2014; Chen et al. 2017). Consistently, we observed an inhibitory action of TGFβ on cell proliferation in TtT/GF cells, accompanied with downregulation of pCNA and DNA replication licensing factors (MCM2, 4, 5, 6 and 7) as indicated by GO and STRING analyses. Conversely, we observed upregulation of metalloproteinase inhibitor 3 (TIMP3) and integrin α5, which have been shown to suppress the progression of pancreatic cancer (Hezel et al. 2012) and malignant behavior in colorectal cancer cells (Lin et al. 2012), respectively, suggesting that TGFβ acts as a tumor suppressor in TtT/GF cells. The TGFβ-induced proteins found in this study can be used as novel biomarkers for pituitary tumor progression and invasion and as therapeutic targets for pituitary adenomas.

Effect of endogenous TGFβ on TtT/GF cells

Although TGFβ is expressed in TtT/GF cells (Renner et al. 1997; Tsukada et al. 2018), the present study did not show significant differences between vehicle- and SB431542-treated cells regarding cell morphology, actin architecture and cell growth. This is likely because endogenous TGFβ is secreted as a latent form (Gressner et al. 2007). Indeed, our previous and present studies showed that Smad2 is localized mostly in the cytoplasm of vehicle-treated cells (Electronic Supplementary Material, Fig. 1) (Tsukada et al. 2018). The results suggested that the effect of endogenous TGFβ on TtT/GF cells is subtle. Although we found that Sca-1 and Prrx2 were upregulated by SB431542 (Electronic Supplementary Material, Fig. 2) (Tsukada et al. 2018), SB431542 did not influence other genes of interest. This implied that the regulatory system of these transcription factors is more sensitive to endogenous TGFβ compared with those of other genes.

Reinforcement of pericyte and smooth muscle cell properties

Our previous study demonstrated that TGFβ significantly increased the mRNA level of pericyte markers Cspg4 (Ng2), nestin and type I collagen (Tsukada et al. 2018). Although the increase in pericyte markers implied that TGFβ induces pericyte differentiation, further characterization is required to verify the properties of TGFβ-treated TtT/GF cells, because pericytes are not a single-cell entity and there is no single and/or reliable pericyte marker. SILAC-assisted mass spectrometry analysis allows broad identification and quantification of TGFβ-induced proteins. Using this method, we found that smooth muscle actin and transgelin as well as Cspg4, nestin and type I collagen were upregulated by TGFβ, all of which have been used as pericyte markers in previous studies (Lu and Shenoy 2017). Furthermore, TGFβ also induced the expression of TIMP3, thrombospondin-1 and filamin-A. These proteins are expressed in pericytes (Schor et al. 1995; Shojaee et al. 1998; Saunders et al. 2006; Azuma et al. 2015) but are not used as reliable markers. We also found upregulation of smooth muscle cell-related proteins, though some of them overlap with pericyte markers. Observation of GO terms related to muscle contraction suggested that TGFβ-treated TtT/GF cells might be differentiated into cells with contractile activity like pericytes and smooth muscle cells.

More recently, Lu and Shenoy demonstrated that some EMT events upregulate multiple pericyte markers in cancer cells (Lu and Shenoy 2017). They called this event epithelial-to-pericyte transition (EPT) and suggested that EPT in cancer cells suppresses tumor metastasis due to stabilization of vasculature by increasing pericyte coverage (Lu and Shenoy 2017). We found suppression of DNA replication and increased pericyte markers, suggesting that TGFβ might induce EPT in TtT/GF cells resulting in acquisition of pericyte properties.

In summary, the present study supports our previous observation that TGFβ reinforces pericyte properties and revealed many other TGFβ-induced proteins related to actin cytoskeleton organization, ECM including extracellular regulators of TGFβ and DNA replication. Of course, due to the size, quantity, ionization and database limitations of mass spectrometry analysis, it should be noted that there could be more TGFβ-induced proteins and molecules in TtT/GF cells other than the lists shown in Electronic Supplementary Material, Tables S3 and S4. Even though there are such technical limitations, SILAC-assisted mass spectrometry analysis is a powerful tool to characterize cellular properties and the use of TtT/GF cells will contribute to understanding the mechanisms related to actin cytoskeleton and ECM regulation/interaction, pituitary tumor progression and invasion, pathological processes and pericyte/smooth muscle cell differentiation including EPT in pituitary cells. Lastly, in order to effectively share the current mass spectrometry proteomics data, including raw data, the data file was deposited in the ProteomeXchange Consortium (Vizcaíno et al. 2014) via the Japan Proteome Standard Repository (jPOST) (Deutsch et al. 2017) with the identifier PXD009245.