Introduction

The anterior lobe of the pituitary gland is a master endocrine organ consisting of five different hormone-producing cell types, which originate in the cranial placode, preceding invagination of the oral ectoderm. Accumulating studies have identified many factors involved in gene regulation and morphogenesis that determine the fate of each hormone-producing cell (for reviews, see Kelberman et al. 2009; de Moraes et al. 2012; Rizzoti 2015). In addition, the induction of hormone-producing cells in vitro from murine and human embryonic stem cells has been reported (Suga et al. 2011; Ozone et al. 2016). Moreover, in adult tissue, circumstantial evidence suggests that SOX2-positive cells serve as multipotent stem/progenitor cells and are a cell resource in the anterior lobe (Fauquier et al. 2008; Chen et al. 2009; Andoniadou et al. 2013; Rizzoti et al. 2013). In addition to SOX2-positive cells, non-endocrine cells expressing S100β (S100β-positive cells) consist of heterogeneous populations and include folliculostellate (FS) cells; as such, a portion of these cells is believed to be candidate pituitary stem/progenitor cells (Inoue et al. 2002; Vankelecom 2007; Devnath and Inoue 2008). Both SOX2-positive and S100β-positive cells form their own niches in the marginal cell layer and parenchyma of the anterior lobe (Gremeaux et al. 2012; Yoshida et al. 2016). The third potential source of stem/progenitor cells is capillary mural cells, namely, pericytes. In addition to their functions in angiogenesis and vascular homeostasis (Armulik et al. 2005; Birbrair et al. 2014), sub-populations of pericytes possess stemness and can differentiate into fibroblasts, osteoblasts, chondrocytes, adipocytes, skeletal muscle and neural cells (Collett and Canfield 2005; Karow et al. 2012; Birbrair et al. 2013). Pericytes are located throughout the anterior lobe, as the pituitary is a highly vascularized organ and capillary architecture is important for vital functions. Pericytes might therefore contribute endocrine and/or non-endocrine cells to the anterior lobe. However, there have been few studies on pericytes in the anterior lobe. Understanding the cellular and molecular mechanisms of pericytes in this tissue will provide important insights into pituitary stem cell research and, accordingly, the development of model cells to study pituitary pericytes is crucial.

TtT/GF is a mouse anterior pituitary cell line derived from a radiothyroidectomy-induced thyrotropic pituitary tumor (Inoue et al. 1992), which does not have the characteristics of typical endocrine cells. As these cells display long cytoplasmic protrusions, follicle formation, phagocytic activity and S100β positivity, they are used as a model of FS cells. Many investigations into the molecular and cellular properties of FS cells have been performed using TtT/GF cells (Renner et al. 1997; Tierney et al. 2003; Stilling et al. 2005; Vitale and Barry 2015). In addition, stem/progenitor characteristics of these cells have been revealed by the presence of Sox2, drug-efflux ATP-binding cassette transporters and stem cell antigen-1 (Sca1) (Chapman et al. 2002; Mitsuishi et al. 2013). Recently, our molecular profiling through microarray analysis suggested the intriguing hypothesis that TtT/GF cells are not terminally differentiated and might have some properties of stem/progenitor cells, vascular endothelial cells, and pericytes rather than representing a model solely for FS cells (Yoshida et al. 2014).

The aim of this study is to examine whether TtT/GF cells have the potential to differentiate into vascular endothelial cells and/or pericytes. To this end, we focus on transforming growth factor beta (TGFβ), known to be essential for epithelial-mesenchymal transition (EMT) (Gonzalez and Medici 2014) and vascular development and function (Jakobsson and van Meeteren 2013), as a potential factor that can be used to elicit differentiation in TtT/GF cells. The present study examines the effects of TGFβ on TtT/GF cells; in addition, novel properties, possible origins and potential uses for TtT/GF cells are discussed.

Materials and methods

Cell culture

The TtT/GF cell line was kindly provided by Dr. Kinji Inoue (Saitama University) and used for experiments at passages 20–25. DMEM/F-12 medium (Life Technologies, Carlsbad, CA, USA) was supplemented with 2.5% fetal bovine serum, 10% horse serum (Biowest, Riverside, MO, USA), 0.5 U/mL penicillin and 0.5 μg/mL streptomycin (Life Technologies). Cells were seeded in 60-mm dishes with poly-l-lysine-coated coverslips (Iwaki, Tokyo, Japan) at a density of 1.8 × 105 cells/3 mL/dish and were maintained at 37 °C in an incubator with 5% CO2. Cells were harvested 3 days after plating and then processed for each analysis.

Treatment with TGFβs and TGFβ receptor I inhibitor (SB431542)

Recombinant human TGFβ1, 2 and 3 (PeproTech, Rocky Hill, NJ, USA) and a selective TGFβ receptor I inhibitor (SB431542; Merck Millipore, Billerica, MA, USA) (Inman et al. 2002) were diluted in Hank’s balanced salt solution (HBSS; Life Technologies) containing 0.1% bovine serum albumin and dimethyl sulfoxide, respectively and stored at −20 °C until use. On the day of experimentation, TGFβ ligands and SB431542 were diluted to the indicated concentrations with fresh medium and added to dishes containing TtT/GF cells. An IX71 inverted microscope (Olympus, Tokyo, Japan) was used for observation of live cells.

Sample preparation for immunofluorescence

For SMAD2 immunocytochemistry (30-min treatment with TGFβ/SB431542), cells attached to coverslips were briefly washed in HBSS twice and fixed with 4% paraformaldehyde (PFA) in 20 mM HEPES (pH 7.5) for 20 min. For immunocytochemistry probing for neuron-glial antigen 2 (NG2), NESTIN and type-I collagen, cells treated for 3 days with TGFβ/SB431542 were dispersed with 0.025% trypsin and re-suspended in fresh medium. The cells were then re-plated onto poly-l-lysine-coated coverslips and incubated for an additional 30 min. Attached cells were washed, fixed and permeabilised as described previously. For immunohistochemistry, male ICR mice (8 weeks old; Japan SLC, Shizuoka, Japan) were anesthetized with intraperitoneally injected pentobarbital sodium (Kyoritsu Seiyaku, Tokyo, Japan) and then perfused through the left ventricle with ice-cold 4% PFA in 50 mM phosphate buffer (PB; pH 7.4) for 5 min. Pituitary glands were then excised and immersed in the same fixative for 24 h at 4 °C, after which tissues were immersed in 50 mM PB (pH 7.2) containing 30% sucrose for 2 days at 4 °C. The tissues were then embedded in frozen section compound (Leica Microsystems, Wetzlar, Germany) and cryosections (8 μm) were obtained using a cryostat (CM1950; Leica Microsystems). All animal experiments were performed in a humane manner after receiving approval from Toho University and protocols were based on the NIH Guidelines for the Care and Use of Laboratory Animals.

Immunofluorescence microscopy

All staining procedures were performed at room temperature and 20 mM HEPES (pH 7.5) containing 100 mM NaCl was used for antibody dilution and washing. Fixed TtT/GF cells and mouse pituitary cryosections were incubated in blocking solution (2% normal goat serum) for 20 min and then incubated with primary antibodies for 90 min. The primary antibodies included rabbit monoclonal anti-SMAD2, rabbit polyclonal anti-type I collagen, anti-NG2 and chicken monoclonal anti-NESTIN (see Table 1 for information regarding primary antibodies). The cells/sections were then incubated for 30 min with secondary antibodies and with 4′,6-diamidino-2-phenylindole (DAPI, 500 ng/mL) for nuclear staining. Secondary antibodies included Alexa Fluor 488-conjugated goat anti-rabbit IgG and Alexa Fluor 568-conjugated goat anti-chicken IgG (both 1:200; Life Technologies). Coverslips were mounted using Aqua Poly/Mount (Polysciences, Warren, PA, USA). Stained cells were subsequently analyzed with an FV1000 confocal laser microscope (Olympus). Images were processed for presentation using Photoshop CS5 (Adobe Systems, San Jose, CA, USA).

Table 1 Information on primary antibodies used
Full size table

Reverse transcription PCR

After 3 days in culture, total RNA was extracted from TtT/GF cells, which were attached to coverslips, using an RNeasy mini-kit and an RNase-free DNase set according to the manufacturer’s instructions (Qiagen, Hilden, Germany). cDNA was synthesized using the PrimeScript RT reagent kit (Takara Bio, Otsu, Japan) with oligo-(dT)20 primers (Life Technologies). cDNA was mixed with gene-specific primers (see Table 2), Blend Taq DNA polymerase, buffer and dNTPs according to the manufacturer’s instructions (Toyobo, Osaka, Japan) and then subjected to PCR as follows: 2 min at 94 °C; 32–34 cycles of 30 s at 94 °C, 30 s at 58 °C and 30 s at 72 °C; and an additional 7 min at 72 °C using a GeneAmp PCR System 9700 (Applied Biosystems, Foster City, CA, USA). The products were run on 1.5% agarose gels and visualized with ethidium bromide.

Table 2 Primers for reverse transcription PCR and real-time PCR
Full size table

Real-time PCR-based quantification of mRNA levels

Total RNA was extracted after a 3-day treatment with TGFβ/SB431542 and cDNA was synthesized as described previously. Quantitative real-time PCR (Applied Biosystems StepOnePlus Real-Time PCR; Applied Biosystems) was performed using SYBR Green Real-time PCR Master Mix Plus (Toyobo) and specific primer sets at 0.6 μM for each target gene (Table 2). Each sample was measured in duplicate and results are based on four independent experiments; data were analyzed by the comparative CT method (ddCt method) to estimate gene copy number relative to that of the TATA box binding protein (Tbp), used as an internal standard. The DNA sequence of the PCR product of each sample was confirmed by nucleotide sequencing (data not shown).

Statistical analysis

All results are presented as the mean ± SEM (n = 4). A one-way analysis of variance (ANOVA) followed by Tukey’s test for multiple comparisons was performed using Prism v.6 (GraphPad Software, San Diego, CA, USA). P < 0.05 was used to indicate statistical significance.

Results

Identification of TGFβ signaling pathway in TtT/GF cells

Reverse transcription PCR was performed to determine whether TtT/GF cells possess TGFβ ligands and receptors (Fig. 1). Transcripts for all three TGFβ isoforms (Tgfb1–3) were detected in TtT/GF cells; however, the expression of TGFβ1 was very low. TGFβ receptor I and II (encoded by Tgfbr1 and Tgfbr2, respectively), which form a heterodimer and act as canonical TGFβ receptors, were expressed. TGFβ receptor III (Tgfbr3), a non-canonical TGFβ receptor known as betaglycan, was also expressed. Because canonical TGFβ receptors were expressed in TtT/GF cells, we next examined if these receptors are functional. Three days after initiation of culture, TtT/GF cells were treated with recombinant human TGFβs for 30 min and immunocytostaining for SMAD2, an intracellular TGFβ signaling mediator, was performed. Strong nuclear staining of SMAD2 was observed when cells were treated with 10 ng/mL TGFβ (Fig. 2d, d′ for TGFβ2, Supplementary Fig. 1 for TGFβ1 and 3), whereas SMAD2 staining was diffuse in the cytoplasm in the absence of or with lower concentrations of TGFβ2 (0–1 ng/mL) (Fig. 2a–c′).

Fig. 1
figure 1

All TGFβ ligand isoforms and receptors are expressed in TtT/GF cells. The expression of TGFβ1–3 (Tgfb1–3, 32 cycles) and TGFβ receptor I–III (Tgfbr1–3, 34 cycles) in TtT/GF cells was determined by reverse transcription PCR. All PCR products were of the predicted sizes, as indicated in Table 2

Full size image
Fig. 2
figure 2

TGFβ2 induces Smad2 nuclear translocation. TtT/GF cells were cultured for 3 days and treated with (Left half) TGFβ2 (0–10 ng/mL) or (Right half) TGFβ2 (10 ng/mL) with a TGFβ receptor inhibitor (SB431542, 0–10 μM) for 30 min. Treated cells were immunostained for SMAD2 (black and white panels). The color panels show the merged images (SMAD2: green, DAPI: blue). Diffuse cytoplasmic staining for SMAD2 was observed with lower concentrations of TGFβ2 treatments (a–c′); however, intense nuclear staining was observed with 10 ng/mL TGFβ2 (d, d′). The TGFβ2-induced Smad2 nuclear translocation was completely blocked by 10 μM SB431542 (h, h′). The effects of TGFβ1 and TGFβ3 on SMAD2 nuclear translocation are shown in Supplementary Fig. 1. Bar 100 μm

Full size image

As SMAD2 nuclear translocation was induced by TGFβ ligands, we further tested the function of the receptors using the selective TGFβ receptor I inhibitor, SB431542. When SB431542 was co-administered with TGFβ, TGFβ-induced SMAD2 nuclear translocation was completely blocked at a dose of 10 μM (Fig. 2h, h′ for TGFβ2, Supplementary Fig. 1 for TGFβ1–3). For the following experiments, we selected TGFβ2, as it is a major TGFβ ligand that is expressed in TtT/GF cells and because there was no difference between TGFβ2 and TGFβ3 in terms of the effect on SMAD2 nuclear translocation (Supplementary Fig. 1).

Effect of TGFβ2 on TtT/GF cell morphology

TtT/GF cells were treated with TGFβ2 for 3 days and cell morphology was analyzed (Fig. 3). It is known that TtT/GF cells typically spread out over the entire surface during culture (Fig. 3a, a′). However, TGFβ2-treated cells did not spread but rather formed colonies (Fig. 3b, b′). Interestingly, TGFβ2-induced colony formation was completely blocked when SB431542 was added (Fig. 3d, d′), whereas SB431542 treatment alone had no effect on cell morphology (Fig. 3c, c′).

Fig. 3
figure 3

TGFβ2 induces cell colony formation. TtT/GF cells were treated with TGFβ2 (10 ng/mL) and/or a selective TGFβ receptor I inhibitor (SB431542; 10 μM) for 3 days and cell morphology was observed. Magnified views of a, b, c, and d are shown in a′, b′, c′, and d′, respectively. After TGFβ2 treatment, TtT/GF cells formed cell colonies (b, b′), and TGFβ2-induced colony formation was completely blocked by SB431542 (d, d′). Bars  (left) 1 mm, (right) 100 μm

Full size image

Effect of TGFβ2 on TtT/GF cell differentiation

Quantitative real-time PCR was performed to determine whether the TGFβ signaling pathway is associated with cell differentiation in TtT/GF cells. As previous studies have suggested that TtT/GF cells have properties of stem/progenitor cells, mesenchymal cells, pericytes, endothelial cells and FS cells (Inoue et al. 1992; Mitsuishi et al. 2013; Yoshida et al. 2014), the expression of marker genes for each cell type was examined after TGFβ/SB431542 treatment (Fig. 4). The transcript levels of stem cell markers (Sox2 and Cd34), mesenchymal cell marker (Vimentin) and pituitary endothelial progenitor cell marker (Sca1) were significantly reduced with TGFβ2 and levels were recovered to those observed after vehicle treatment when TtT/GF cells were cultured in the presence of both TGFβ2 and SB431542. Conversely, SB431542 significantly increased Sca1 and Prrx2 expression. Nestin and Ng2, pointers for the pericyte characteristics, were significantly increased with TGFβ2 treatment, whereas SB431542 completely blocked this effect. Previously, the expression of type-I collagen was found to be exclusively expressed in pituitary pericytes (Fujiwara et al. 2010; Tofrizal et al. 2016). In the present study, we observed that TGFβ2 treatment enhanced expression of type-I collagen in TtT/GF cells. TGFβ2 treatment had no significant effect on the transcription of an endothelial cell marker (VE-cadherin) and an FS cell marker (S100b). We next examined protein levels by immunocytochemistry to support the elevated gene expressions (Fig. 5). Consistent with the quantitative real-time PCR results, TGFβ2 increased the immunofluorescence signals for NG2, NESTIN (Fig. 5a–b′′) and type-I collagen (Fig. 5c–d′′), suggesting enhanced protein production.

Fig. 4
figure 4

TGFβ2 alters the gene expression profiles of TtT/GF cells. TtT/GF cells were treated with TGFβ2 (10 ng/mL) and/or a selective TGFβ receptor I inhibitor (SB431542; 10 μM) for 3 days and gene expression was evaluated by quantitative real-time PCR (n = 4, mean ± SEM). Genes included stem cell markers (a Sox2, b Cd34, c Prrx2), pituitary endothelial progenitor cell marker (d Sca1), mesenchymal cell marker (e Vimentin), pericyte markers (f Nestin, g Ng2, h type I collagen), a folliculostellate cell marker (i S100b) and an endothelial cell marker (j VE-cadherin). mRNA copy numbers were normalized to those of TATA-binding protein (Tbp). TGFβ2 increased pericyte marker gene expression and decreased stem cell marker gene expression. *, **, ***, ****p < 0.05, 0.01, 0.001, 0.0001, respectively (Tukey’s test)

Full size image
Fig. 5
figure 5

TGFβ2 enhances the production of NG2, NESTIN and type-I collagen in TtT/GF cells. Three days after TGFβ2 treatment (10 ng/mL), TtT/GF cells were re-plated and stained for a–b′′) NG2 (green) and NESTIN (red) and c–d′′) type I collagen (green). DAPI (blue) was used for nuclear staining. Left panels show bright-field images. Middle and right panels show lower- and higher-magnification images of immunofluorescence, respectively. TGFβ2 treatment enhanced immunoreactivity for NG2, NESTIN and type I collagen (b–b′′, d–d′′). Bars (left and middle panels) 100 μm, (right panels) 10 μm

Full size image

Localization of NESTIN/NG2-double positive cells in mouse anterior pituitary gland

NESTIN expression in pericytes of the anterior lobe was first demonstrated in vivo by Krylyshkina et al. (2005). The present study further determined immunohistochemically whether NESTIN-positive pericytes express NG2. The results showed that NESTIN-positive cells had an elongated shape and were located in the perivascular space of the anterior lobe and that NG2 was expressed in NESTIN-positive cells (Fig. 6; arrowheads).

Fig. 6
figure 6

NG2/NG2 double-positive cells are present in the mouse anterior lobe. Double immunohistochemistry of NG2 and NESTIN was performed using mouse pituitary tissue (a NG2, b NESTIN). A merged image (NG2 green, NESTIN red, DAPI blue) and differential interference contrast images superimposed on fluorescence images are shown in (c and d), respectively. Higher-magnification images of a, b, c and d are shown in a’, b’, c’ and d’, respectively. NG2/NG2 double-positive cells (arrowheads) were localized around the capillary (*). Bars (top) 50 μm, (bottom) 10 μm

Full size image

Discussion

TGFβ signaling pathway in TtT/GF cells

The present study showed that TtT/GF cells express canonical TGFβ receptors that are associated with Smad pathways. TtT/GF cells were also shown to express all three TGFβ isoforms (Tgfb1–3) and Tgfb2 and Tgfb3 were predominant. This suggests that TGFβ ligands secreted from TtT/GF cells act in a paracrine and/or autocrine manner. In the rat anterior lobe, canonical TGFβ receptors have been shown to be present in FS cells, pericytes and endothelial cells (Syaidah et al. 2016; Tsukada et al. 2016). In addition to FS cell properties, our previous microarray analysis indicated that TtT/GF cells have vascular endothelial cell/pericyte characteristics (Yoshida et al. 2014). Although there are differences between TtT/GF cells derived from mouse pituitary adenoma and the ‘non-endocrine’ cells of the rat anterior pituitary, based solely on the TGFβ receptor profile, TtT/GF cells exhibit similarity to all three cell types. In contrast, regarding ligands, our recent study revealed that FS cells isolated from S100β-GFP transgenic rats (Itakura et al. 2007) highly express Tgfb2 and Tgfb3 but marginally express Tgfb1 (Tsukada et al. 2016). Based on the TGFβ ligand expression profile, similarity exists only between TtT/GF cells and FS cells. Interestingly, we detected transcripts of the Smad-independent non-canonical TGFβ receptor III (Tgfbr3) in TtT/GF cells, which supports canonical TGFβ signaling by delivering TGFβ to TGFβ receptor II (López-Casillas et al. 1993); however, our previous study did not detect Tgfbr3 expression in the rat anterior lobe (Tsukada et al. 2016). Recent studies indicate that the TGFβ receptor III acts as a cancer promoter or cancer suppressor in several cancer cell types (Dong et al. 2007; Jovanović et al. 2014). As TtT/GF cells are a cancer cell line, it is possible that TtT/GF cells acquire Tgfbr3 expression during tumorigenesis.

TGFβ signaling is associated with TtT/GF cell colony formation

The present study demonstrated a pronounced TGFβ-mediated effect on TtT/GF cell colony formation. Co-administration of the TGFβ receptor I inhibitor (SB431542) completely blocked the action of this axis, suggesting that cell colony formation is linked to canonical TGFβ signaling. Cell colony formation is mainly caused by reorganization of cell adhesion and motility. One study showed that TGFβ1-induced E-cadherin expression results in cell cluster formation in CD34+ hematopoietic progenitor cells (Riedl et al. 2000), which is important for the differentiation of epithelial Langerhans cells; however, TGFβ is well known to have opposing effects, reducing E-cadherin-mediated adhesion during EMT. In addition, TGFβ promotes Lin11, Isl-1 and Mec-3 (LIM domain proteins) expression to regulate actin cytoskeleton organization (Järvinen and Laiho 2012). Similar to CD34+ hematopoietic progenitor cells, TGFβ-induced cell colony formation might be associated with cell differentiation and/or transformation by modulating such adhesion and cytoskeletal molecules in TtT/GF cells. However, further study is required to understand the biological relevance of TGFβ-induced cell colony formation.

TGFβ signaling is associated with differentiation of TtT/GF cells

During EMT, TGFβ upregulates Sox2, Podocalyxin (a member of the Cd34 family) and Vimentin in glioblastoma cells and the lung adenocarcinoma cell line A549 (Ikushima et al. 2009; Meng et al. 2011) to induce mesenchymal properties. However, TGFβ2 administration in the present study resulted in downregulation of these stem and mesenchymal cell marker genes in TtT/GF cells. Moreover, inhibition of endogenous TGFβ by SB431542 significantly increased the expression of the other stem/progenitor cell marker Sca1 (Gussoni et al. 1999) and Prrx2 (Susa et al. 2012). These results suggest that TGFβ2 attenuates the stemness of TtT/GF cells via these transcription factors rather than induction of dedifferentiation or EMT. Indeed, TGFβ2 was shown to upregulate both mRNA and protein of type I collagen in TtT/GF cells. Our previous studies revealed that type I collagen is produced almost exclusively by pericytes in the rat and human anterior pituitary (Fujiwara et al. 2010; Tofrizal et al. 2016) and that TGFβ2 upregulates collagen synthesis in pituitary pericytes (Tsukada et al. 2016). Concomitant with type I collagen expression, the expression of Nestin and Ng2, major pericyte markers, was significantly increased by TGFβ2 treatment and this increase was completely blocked by SB431542, suggesting that TGFβ2 attenuates the stemness of TtT/GF cells and endows pericyte properties via the Smad pathway. Renner et al. (2002) showed that TGFβ1 and TGFβ3 dramatically increase vascular endothelial growth factor (Vegf) expression in TtT/GF cells; we saw a similar increase in the presence of TGFβ2 (Supplementary Fig. 2). Although VEGF was first isolated from FS cells in the pituitary gland (Ferrara and Henzel 1989; Leung et al. 1989), pericytes also express Vegf to induce endothelial cells to undergo neovascularization and VEGF expression is upregulated by TGFβ (Thanabalasundaram et al. 2011). Similar to pericytes, TGFβ-treated TtT/GF cells might have angiogenic capacity. However, we note that our finding of TGFβ-induced TtT/GF cell differentiation should be interpreted cautiously, as it is possible that genetic drift might have occurred in the TtT/GF cell line with repeated passaging such that the TtT/GF cells were no longer a single biological entity.

NESTIN/NG2-double positive pericytes are located in the mouse anterior lobe

In the present study, we demonstrated that NESTIN/NG2-double positive cells are present in the perivascular area where pericytes are located. Accumulating evidence suggests that pericytes are not a single cell population. Indeed, Birbrair et al. (2013) isolated two bona fide pericyte subtypes from mouse skeletal muscle: one expresses Ng2 but not Nestin (NESTIN-negative/NG2/NG2-positive: type-1), while the other expresses both Nestin and Ng2 (NESTIN/NG2-double positive: type-2). Type-1 pericytes express collagens in the presence of TGFβ, suggesting fibrogenic potency, whereas type-2 pericytes have muscular regenerative capacity that restores injured muscles. According to the expression profile of Nestin/Ng2, TGFβ2-induced TtT/GF cells share some molecular characteristics with type-2 pericytes. However, the TGFβ-induced cells in the present study likely lost stemness and thus possess fibrogenic potential similar to that of type-1 pericytes in skeletal muscle. Although further study is required to associate TGFβ2-induced TtT/GF cells with NESTIN/NG2-double positive cells in the mouse anterior lobe, it is possible that more than two different pericyte populations are present in vivo.

Possible origin and potential use of TtT/GF cells

As suggested by our previous microarray analysis (Yoshida et al. 2014), the present study indicated that TtT/GF cells have the capacity to differentiate into cells with some of the characteristics of pericytes in the presence of TGFβ. Accordingly, studies on FS cells that have utilized TtT/GF cells should be accepted with caution. TtT/GF cells were derived from a radiothyroidectomy-induced thyrotropic pituitary tumor (Inoue et al. 1992) and have been used as a model of FS cells. In the thyrotropic pituitary tumor, glial fibrillary acidic protein (GFAP)-immunoreactive FS cells encapsulate lobules of parenchymal tissue, suggesting that these cells represent a possible origin of TtT/GF cells. Intriguingly, our recent study demonstrated the presence of four types of collagen-producing cells in human pituitary adenomas, including pericytes, fibroblasts, myofibroblasts and myoepithelial-like cells (Tofrizal et al. 2016). Among collagen-producing cells, the location of myoepithelial-like cells is similar to that of GFAP-immunoreactive FS cells in the thyrotropic pituitary tumor. As myoepithelial cells are not isolated from the healthy anterior pituitary lobe, this suggests that EMT is a possible source of myoepithelial-like cells in pituitary adenomas. The origin of TtT/GF cells might also be related to the origin of myoepithelial-like cells. Alternatively, TtT/GF cells could be of neural crest origin, as we and other groups have found that neural crest cells invade during the early phase of adenohypophyseal development and differentiation into pericytes (Davis et al. 2016; Ueharu et al. 2016). Taking all possibilities into account, experiments on TtT/GF cell lines may provide useful pointers for investigating the functions of pituitary stem/progenitor cells and pericytes as well as the mechanisms of organogenesis and tumorigenesis.