Isolation and characterisation of CD9-positive pituitary adult stem/progenitor cells in rats

S100β protein and SOX2-double positive (S100β/SOX2-positive) cells have been suggested to be adult pituitary stem/progenitor cells exhibiting plasticity and multipotency. The aim of the present study was to isolate S100β/SOX2-positive cells from the adult anterior lobes of rats using a specific antibody against a novel membrane marker and to study their characteristics in vitro. We found that cluster of differentiation (CD) 9 is expressed in the majority of adult rat S100β/SOX2-positive cells, and we succeeded in isolating CD9-positive cells using an anti-CD9 antibody with a pluriBead-cascade cell isolation system. Cultivation of these cells showed their capacity to differentiate into endothelial cells via bone morphogenetic protein signalling. By using the anterior lobes of prolactinoma model rats, the localisation of CD9-positive cells was confirmed in the tumour-induced neovascularisation region. Thus, the present study provides novel insights into adult pituitary stem/progenitor cells involved in the vascularisation of the anterior lobe.


Results
Microarray analysis using S100β/GFP-TG male rats (P5 and P60). Haematoxylin Fig. S1A). Although GFP-positive cells were also present in the MCL of the intermediate lobe, we focused on the MCL and parenchyma in the anterior lobe in the present study. As shown in Fig. 1A, most S100β/GFP-positive cells in the parenchyma at P5 were immunonegative for SOX2; however, a large number were positive for SOX2 at P60. Dispersed cells from the anterior lobes of S100β/GFP-TG male rats were separated into GFP-positive and -negative cell fractions by a cell sorter (Fig. 1B). We performed a comparative microarray analysis of GFP-positive cells at P5 and P60 to identify CD antigens specific for GFP-positive cells at P60. Ten novel CD antigen genes that were up-regulated (fold change > 2.0) in the GFP-positive fraction at P60 compared with levels at P5 were identified (Fig. 1C). In addition, three CD antigen genes that were down-regulated at P60 (fold change <−2.0) were identified (Cd90: NM_012673, Cd200: NM_031518, and Cd231: NM_00108815). We subsequently performed quantitative polymerase chain reaction (qPCR) to determine the mRNA levels of these 10 genes relative to that of β-actin (Actb) in the GFP-positive cell fraction at P60. Cd9, Cd24, and Cd81 were clearly expressed in the S100β/GFP-positive cell fraction (Fig. 1D).
Identification and isolation of CD9-positive cells in the anterior lobe. We further examined whether anterior lobe cells expressed these candidate 10 genes by in situ hybridisation. Cd9-expressing cells and Cd24-expressing cells were clearly detectable in the MCL of the anterior and intermediate lobes and the parenchyma of the anterior lobe; however, they were undetectable in the posterior lobe ( Supplementary Fig. S1B). In contrast, only a few signals were detected for the other eight candidate genes, suggesting that the sensitivity of in situ hybridisation was too low to detect these mRNAs. Next, immunohistochemistry was performed using anti-CD9 and anti-CD24 antibodies. The specificity of the anti-rat CD9 antibody was determined by western blot analysis using lysates prepared from the rat adult anterior lobe. An immunoreactive protein band was detected at approximately 25 kDa, which was consistent with the expected size of rat CD9 ( Fig. 2A). In contrast, only non-specific immunoreactivity was observed for the anti-CD24 antibody following western blotting and immunohistochemistry (Supplementary Fig. S1C and D). CD9-immunopositive (CD9-positive) cells were localised in the MCL of the anterior and intermediate lobes and the parenchyma of the anterior lobe, but not in the posterior lobe by immunohistochemistry using the anti-CD9 antibody and in situ hybridization ( Fig. 2B and Supplementary Fig. S1B). Despite the nonspecific immunoreactivity observed in western blotting, immunohistochemistry for CD24 suggested that this protein was localised in the anterior and intermediate lobe. These locations coincided with the results of RNA in situ hybridisation, suggesting that this immunoreactivity did not represent nonspecific staining. In the anterior lobe, some CD24-positive cells were immunonegative for S100β ( Supplementary Fig. S1D). Based on these results, we selected the anti-CD9 antibody for further study. We performed double-staining using in situ hybridisation for Cd24 and immunohistochemistry for CD9. CD9-positive cells expressed Cd24 (Fig. 2C), indicating that Cd24-expressing cells were the same as CD9-positive cells. CD9 is a member of the tetraspanin superfamily and is a small protein with four transmembrane domains 20,21 . Typically, CD9 forms a complex with CD81, which is also a member of the tetraspanin superfamily and was also detected in the present microarray data (Fig. 1C). Finally, we examined whether CD9-positive cells expressed GFP in S100β-GFP male rats at P5 and P60 by immunohistochemistry. As shown in Fig. S1E, GFP-expressing cells in S100β-GFP male rats were also clearly stained with the anti-CD9 antibody in both the MCL and parenchyma in the anterior lobe at P5 and P60 (Fig. S1E). Most CD9-positive cells expressed GFP and the number of GFP-expressing CD9-positive cells was higher at P60 ( Supplementary Fig. S1E). In conclusion, we selected CD9 as a tool to perform the desired function in the present study.
Next, double immunostaining for CD9 with S100β or SOX2 was performed (Fig. 2D). CD9 and S100β-double positive (CD9/S100β-positive) cells and CD9 and SOX2-double positive (CD9/SOX2-positive) cells were present in the rat adult anterior lobe. The proportions of S100β-positive and SOX2-positive cells among CD9-positive cells were 94.9% and 86.1%, respectively (Fig. 2D). In contrast, the proportions of CD9/S100β-positive cells Immunofluorescence staining of SOX2 in the anterior lobe of S100β/GFP-TG rats at P5 and P60. Open white arrowheads indicate S100β-positive cells negative for SOX2. White arrowheads indicate S100β-positive cells positive for SOX2. Right images are high magnifications of boxed areas in left images at P5 and P60. AL, anterior lobe; IL, intermediate lobe; PL, posterior lobe. (B) GFP intensity of anterior pituitary cells from S100β/GFP-TG rats at P5 and P60, separated by a cell sorter. Numbers indicate gated cell frequencies (n = 3). (C) Comparative expression levels of CD antigens of interest from microarray data of S100β-positive cells at P5 and P60. (D) Expression levels of 10 CD genes and S100β mRNA in GFP-positive cells from S100β/GFP-TG rats at P60 as determined by qPCR and normalised with an internal control (β-actin gene, Actb). among S100β-positive cells and CD9/SOX2-positive cells among SOX2-positive cells were 81.2% and 79.1%, respectively (Fig. 2D). We finally performed triple-staining using in situ hybridisation for Cd9 and double immunostaining for S100β and SOX2. We observed that triple positive (CD9/S100β/SOX2-positive) cells were present in the rat adult anterior lobe, and the proportion of CD9/S100β/SOX2-positive cells among CD9-positive cells was 84.7% (Fig. 2E).

Purification of CD9-positive cells localised in the MCL and parenchyma of the adult anterior lobe.
First, we analysed the proportion of CD9-positive cells among rat anterior lobe cells using FACS. We next attempted to purify CD9-positive cells the monoclonal using anti-rat CD9 antibody combined with the pluriBead-cascade cell isolation system (Fig. 3A). The results showed that the proportion of CD9-positive cells among anterior lobe cells was 3.9% ± 0.2% (Fig. 3B). One drop of suspension of the CD9-positive fraction was used for smear preparation and immunocytochemistry. We observed that most of the cells were immunopositive for CD9 (Fig. 3C). Cd9 mRNA levels were 14.0-fold higher in the CD9-positive cell fraction than in the CD9-negative fraction (Fig. 3D). In addition to the S100β mRNA levels, those of Cd13, Cd24, Cd81, Cd133, and Cd184 among the 10 CD antigen genes were also significantly higher in the CD9-positive cell fraction than in the CD9-negative cell fraction (Fig. 3D) ) were significantly lower in the CD9-positive cell fraction than in the CD9-negative fraction (Fig. 3D). Major stem/progenitor cell markers (Sox2, Sox9, Prop1, Cadh1, Efnb2, and Cxcr4) in the rat anterior lobe were expressed at significantly higher levels in the CD9-positive cell fraction than in the CD9-negative fraction (Fig. 3D). The mRNA levels of these genes relative to that of Actb in the CD9-positive cell fraction are shown in Supplementary Fig. S2.
We cultured isolated CD9-positive cells on a non-coated surface with 10% foetal bovine serum (FBS) at a density of 1.0 × 10 5 cells/cm 2 for 72 h, followed by immunostaining for CD9 (Fig. 3E) or in situ hybridisation for Cd9 and double immunostaining for S100β and SOX2 (Fig. 3F). After 72 h of cultivation, the proportion of CD9-positive cells among total cells was 86.3% (Fig. 3E), and that of CD9/S100β/SOX2-positive cells among CD9-positive cells was 88.9% (Fig. 3F). CD9 has the ability to associate with various integrins 20,21 . In previous studies, we found that S100β-positive cells exhibited marked proliferation activity under the influence of laminin, an ECM component of the basement membrane, through the ECM receptors integrin-α3 (ITGA3) and β1 (ITGB1) 22,23 . To examine the functional role of CD9, we knocked down Cd9 gene expression using small interfering RNAs (siRNAs) in CD9-positive cells primary cultured on laminin-coated surfaces at 1.0 × 10 5 cells/cm 2 . Cd9 expression levels were successfully down-regulated by siRNA treatment (Fig. 4A). Conversely, the expression levels of Itga3 and Itgb1 were up-regulated (Fig. 4A). The number of 5-bromo-2ʹ-deoxyuridine (BrdU)-positive signals in cells treated with Cd9 siRNA was clearly lower than that in control cells (Fig. 4B). In addition, the percentage of BrdU-positive cells among Cd9 siRNA-treated cells (3.6% ± 0.7%) was significantly lower (P < 0.01) than that among control cells (10.6% ± 1.3%) (Fig. 4C).

Differentiation of CD9-positive cells into endothelial-like cells. CD9-positive cells were cultured
for 120 h on laminin-coated surfaces with 0.1% bovine serum albumin (BSA) or 10% FBS in Medium 199 at 2.0 × 10 4 cells/cm 2 . CD9-positive cells cultured with 0.1% BSA were solitary or paired, and some were flattened and showed extension of their cytoplasmic processes a short distance (Fig. 5A). Almost all CD9-positive cells cultured with 0.1% BSA were immunopositive for S100β protein ( Supplementary Fig. S3). In contrast, when cultured with 10% FBS, the cytoplasmic processes of CD9-positive cells extended over 100 μm and were interconnected between cells, followed by the formation of a capillary-like network (Fig. 5A). The cells forming the capillary-like network were immunopositive for the endothelial cell marker VE-cadherin and expressed Kdr, one of a vascular endothelial growth factor receptor (Fig. 5B). Cells positive for isolectin B4 (an endothelial cell marker) were weakly immunopositive for S100β and SOX2 (Fig. 5C, white arrowhead) but were negative for CD9 (Fig. 5C). This demonstrated that a subset of S100β/SOX2-positive cells differentiated into endothelial cells. However, most S100β-positive or SOX2-positive cells were negative for isolectin B4 (Fig. 5C, white open arrowhead). After primary cultivation of the CD9-positive fraction with 10% FBS for 120 h, an average of 8.3% (ranging from 6.6% to 9.4%) of cultured cells was transformed into isolectin B4-positive cells and formed a capillary-like network. The remaining isolectin B4-negative cells, which were still positive for CD9, S100β, and SOX2, did not form an interconnected architecture. Cultivation with 10% FBS resulted in clear decreases in the expression levels of S100β and Sox2 with moderate reductions in Cd9, Prop1, Cadh1, and Cxcr4 in comparison with levels in cells cultivated with 0.1% BSA (Fig. 5D). In contrast, a significant increase in expression was observed for several genes, including Id2 and the endothelial cells markers Sox18, Nrp1, Kdr, Pecam1, and Edn (Fig. 5D).
Among these genes, we focused on the expression of inhibitor of differentiation (ID) 2, a transcription factor that contains a helix-loop-helix domain 24 and is a regulator for differentiation that controls cell fate. Indeed, in situ hybridisation and immunohistochemistry showed that CD9-positive cells expressed Id2 ( Supplementary  Fig. S4). Notably, since the rapid induction of Id2 expression by serum stimulation has been reported 24 , the shows the proportion of S100β and SOX2-double positive (S100β+ and SOX2+) cells among CD9-positive (CD9+) cells. Numbers of CD9, S100β, and SOX2-triple positive cells were counted in random areas of the anterior lobe, and the ratio of triple-positive cells to CD9+ cells was calculated for in situ hybridisation and immunohistochemistry.
SCientifiC REPORts | (2018) 8:5533 | DOI:10.1038/s41598-018-23923-0 up-regulation of Id2 expression under the influence of FBS is likely to be a trigger for the differentiation of CD9-positive cells. Hence, we attempted to knock down Id2 expression using siRNAs in CD9-positive cells cultured on laminin-coated surfaces with 10% FBS. Id2 siRNA treatment decreased the expression of the target gene and inhibited the formation of the capillary-like network (Fig. 6A). qPCR analysis revealed that Cd9 and Sox2 expression levels in Id2 siRNA-treated cells were higher than those in control CD9-positive cells. However, Kdr and Pecam1 mRNA levels decreased significantly (P < 0.01; Fig. 6B). Although S100β mRNA levels were not significantly altered, Id2 siRNA-treated cells were strongly immunoreactive for S100β (Fig. 6A,B). These findings suggested that the up-regulation of Id2 expression did not directly trigger endothelial cell differentiation in CD9-positive cells.
Next, we analysed the signalling pathway responsible for endothelial cell differentiation in CD9-positive cells. Bone morphogenic protein (BMP) signalling, which exhibits pleiotropic activity in cell differentiation and tissue morphogenesis, is known to be responsible for serum-induced Id2 expression 24 . CD9-positive cells on laminin-coated surfaces were incubated with vehicle or dorsomorphin, which specifically inhibits the BMP signalling pathway by targeting the type I receptors ALK2, ALK3, and ALK6. As shown in Fig. 6C, dorsomorphin treatment led to a failure in capillary-like formation in CD9-positive cells. Dorsomorphin-treated cells were strongly immunoreactive for S100β (Fig. 6C). Furthermore, the S100β and Sox2 expression levels in the dorsomorphin-treated cells were higher than those in vehicle-treated cells; however, dorsomorphin significantly repressed the expression of Id2, Kdr, and Pecam1 (Fig. 6D).
Vascularisation mediated by CD9/S100β-positive cells in the anterior lobes of prolactinoma model rats. To characterise CD9-positive cells under pathological conditions, rats were treated with diethylstilbestrol (DES), establishing a model of prolactinoma, in which an excess amount of prolactin is secreted. Prominent prolactinoma is accompanied by increased pituitary gland weight, high serum prolactin levels 25 , and neovascularisation 26 . After treatment with DES, as shown in Fig. 7A, the tumour mass gradually increased in size in DES-treated rats but not in control rats. In the tumour tissue, neovascularisation was notable, and aggregation of extravascular erythrocytes was often observed among the pituitary cells. We observed many blood vessels in the anterior lobes of DES-treated rats (4-12 weeks of treatment) by HE staining of the tumour tissue: the capillaries were tortuous, and lumens were larger than those in control rats.
The mRNA levels of Cd9, S100β, and Sox2 were assessed by qPCR in the anterior lobes of control and DES-treated rats. The results showed a marked decrease in the mRNA levels in DES-treated rats (Fig. 7B). Concomitantly, the levels of Id2, Kdr, and Pecam1 mRNAs were significantly increased (P < 0.01) in the DES-treated anterior lobes in comparison with those in the control lobes (Fig. 7B). We also observed via in situ hybridisation and immunohistochemistry that levels of CD9 transcripts and S100β proteins were reduced in rats after 1 week of DES treatment (Fig. 7C), followed by further time-dependent decreases ( Supplementary Fig. S5). Interestingly, S100β-positive cells in DES-treated rats were localised around blood vessels (Fig. 7C). We examined the localisation pattern of Id2-expressing cells and PECAM1-positive cells under the neovascularisation of the anterior lobe in DES-treated rats. These were localised around the blood vessels, and their positive signals were stronger in DES-treated rats (Fig. 7C). In addition, some Id2-expressing cells were weakly immunopositive for CD9 and VE-cadherin (Fig. 7D). To determine the direct effect of DES, we cultured CD9-positive cells in the presence or absence of DES with medium containing charcoal/dextran-treated FBS. Regardless of the presence or    Fig. S6). These data indicated that DES has no direct effect on the differentiation of CD9-positive cells to endothelial cells.

Discussion
The present study succeeded in the isolation of SOX2/S100β-positive stem/progenitor cells from the rat anterior lobe using a monoclonal anti-CD9 antibody and demonstrated that BMP signalling was associated with endothelial cell differentiation in a subset of CD9/S100β/SOX2-positive cells. In addition, the hyperplasia of blood vessels in the anterior lobes of DES-treated prolactinoma model rats was driven in part by the differentiation of CD9/ S100β/SOX2-positive cells. Thus, CD9 is a novel indicator that can be used to determine the roles of stem/progenitor cells in the anterior lobe. S100β-positive cells are known to be composed of heterogeneous subpopulations and to play several biological roles in the pituitary gland. Yoshida et al. 12 showed that the proportion of SOX2-positive cells among S100β-positive cells in the adult rat anterior lobe was approximately 85%. To understand the multiple functions of S100β-positive cells, it is important to be able to isolate S100β/SOX2-positive cells from the anterior lobe in order to utilise several in vitro assays. In this study, we showed that SOX2/S100β-positive cells accounted for 84.7% of CD9-positive cells and that the proportion of CD9-positive cells in the anterior lobe was 3.9% according to FACS analysis. These results indicate that the proportion of CD9/SOX2/S100β-positive cells in the anterior lobe was 3.3%. In the CD9-positive fraction, CD9-positive cells accounted for 86%, whereas the proportion of S100β/SOX2-positive cells among CD9-positive cells was about 89%. Therefore, the proportion of CD9/S100β/ SOX2-positive cells in the CD9-positive fraction was 77%. These data reflected an enrichment of 23-fold from anterior lobe tissue, indicating that this method provided a powerful tool for investigating the functions of SOX2/ S100β-positive cells by utilising CD9 expression.
CD9 was first described as a motility-related factor in 1991, when it was reported that specific anti-CD9 antibody inhibited the migration of multiple cancerous cell lines 27,28 . Thereafter, CD9 has been shown to be associated with various integrins, including α3 and β1, and the migratory functions of CD9 have been attributed to its modulatory activity towards integrin complexes 29 . In previous studies, we found that S100β-positive cells exhibited marked proliferation activity under the influence of ECMs through integrin β1 signalling, which activated the mitogen-activated protein kinase pathway 22 . Downregulation of the Cd9 gene upregulates the expression of integrins α3 and β1 and inhibits BrdU incorporation in primary culture. These results suggest that the function of CD9 contributes to sustaining proliferation activity through integrin signalling. The downregulation of Cd9 may trigger the differentiation, rather than the proliferation, of CD9-positive cells under the influence of the ECM.
In the present study, we revealed that approximately 8% of the CD9-positive fraction differentiated into endothelial cells in vitro through the involvement of Id2 expression. These data suggested that S100β/ SOX2-positive cells were multipotent progenitor cells and that their differentiation capacity was s influenced by cell culture conditions. Yoshida et al. isolated the dense cell clusters originating from the parenchymal niche, termed PS clusters, which are composed of S100β/SOX2-positive cells, from the adult rat anterior lobe. By taking advantage of its tight structure and resistance to protease treatment 11,16 , they showed that S100β/SOX2-positive cells exhibited the capacity for differentiation into endocrine cells in a three-dimensional cultivation system 16 . In contrast, the S100β/SOX2-positive cells in the present study were individually isolated from the whole anterior lobe including the MCL and parenchyma niches by protease reaction and a pluriBead kit with an anti-CD9 antibody. It may be that S100β/SOX2-positive cells outside of PS clusters differentiate into endothelial cells or that the removal of a subset of S100β/SOX2-positive cells from PS clusters using the protease reaction and pluriBead kit triggers their differentiation into endothelial cells. In either case, our findings suggested that S100β/SOX2-positive cells were comprised of heterogeneous subpopulations in the anterior lobe.
To maintain the physiological functions of the pituitary, a blood capillary network composed of endothelial cells and pericytes is essential. We have observed in the developing pituitary gland that S100β/isolectin B4-double positive cells, which may be endothelial-like cells, appear to enter into the embryonic anterior lobe 30 . A portion of S100β-positive cells extend their cytoplasmic processes into the basement membrane around blood capillaries 31 . Some of the CD9/S100β/SOX2-positive cells in the parenchyma may be tissue-resident vascular precursor cells and may participate in vascularisation as suppliers of endothelial cells in the adult anterior lobe. However, we do not currently have definitive data regarding the characteristics of CD9-positive cells. Further studies are necessary to determine whether CD9-positive cells have sphere-forming capacity and differentiate into all types of endocrine cells or whether CD9/S100β/SOX2-positive cells differentiate into endothelial cells in vivo using lineage tracing assays.
ID2 functions as a regulator of basic helix-loop-helix transcription factors, and its expression is rapidly induced by serum stimulation of BMP signaling 24 . Lasorella et al. 32 reported that ID2 mediates tumour initiation, proliferation, and angiogenesis in the mouse anterior pituitary. In fact, the present study demonstrated that CD9-positive cells express Id2 in the rat anterior lobe and that the inhibition of BMP signalling down-regulates Id2 expression. These observations suggest that the expression of Id2 contributes to sustaining the plasticity of CD9-positive cells in the niche and that its down-regulation may lead to neovascularisation or tumourigenesis. in the early stage of DES-treatment, as capillaries were extended and tortuous and haemorrhages were often present in tumour tissues at 12 weeks. Moreover, our results suggested that CD9-positive cell-derived endothelial cells maintained Id2 expression in tumourigenesis.
In conclusion, we found that Cd9 was expressed in the majority of S100β/SOX2-positive adult stem/progenitor cells in the rat anterior lobe of the pituitary gland. A subset of CD9/S100β/SOX2-positive cells were shown to differentiate into endothelial cells via stimulation of the BMP/ID2 cascade, indicating that a subset of CD9/S100β/ SOX2-positive cells played a role in vascularisation as tissue-resident vascular precursor cells. These CD9-positive cells were observed around the neo-vascular vessels in the pituitary glands of DES-induced prolactinoma model rats. These findings should provide a better understanding of the adult tissue stem cells of the anterior lobe and further preclinical and clinical studies on tumour vascularisation.

Methods
Animals. Wistar-crlj S100β/GFP-TG rats that express GFP under the control of the S100β promoter were provided by Professor K. Inoue of Saitama University and bred in our laboratory. Male Wistar F344 rats were purchased from Japan SLC, Inc. The day of birth was designated as P0. Eight-to 10-week-old rats weighing 200-250 g were given ad libitum access to food and water and housed under a 12-h light/dark cycle. Rats were killed by exsanguination from the right atrium under deep nembutal anaesthesia and were then perfused with Hanks' balanced salt solution (Thermo Fisher Scientific, Waltham, MA, USA) for culture and FACS or with 4% paraformaldehyde in 0. 05  Immunohistochemistry and immunocytochemistry. Tissue preparation and immunohistochemistry were performed as described previously 33 . Frozen frontal sections of rat pituitary (8 μm thickness) were obtained using a cryostat (Tissue-tek Polar DM; Sakura Finetek, Tokyo, Japan; CM1860, Leica, Wetzlar, Germany). The standard ABC method was performed using a Vectastain ABC kit (Vector Laboratories, Burlingame, CA, USA) with 3,3ʹ-diaminobenzidine (Dojindo Laboratories, Kumamoto, Japan) as the substrate 33 . Primary and secondary antibodies are listed in Supplementary Table S1. The absence of an observable nonspecific reaction was confirmed using normal mouse, rabbit, or goat serum. Sections were scanned using an epifluorescence microscope (BX61, Olympus, Tokyo, Japan) with the cellSens Dimension system (Olympus).
Cultured cells fixed with 4% paraformaldehyde in 0.025 M PB for 20 min at room temperature (21-23 °C) were first immersed in phosphate-buffered saline (PBS) containing 2% normal goat or donkey serum for 20 min at 30 °C, then incubated overnight with biotinylated isolectin B4 (1:25; Vector Laboratories) or primary antibodies as listed in Supplementary Table S1 at room temperature. After washing with PBS, cells were incubated with secondary antibodies. Biotinylated isolectin B4 was detected by Alexa Fluor 568-conjugated streptavidin (1:400; Thermo Fisher Scientific). The absence of an observable nonspecific reaction was confirmed using normal rabbit or donkey serum. Cells were scanned using a fluorescence microscope (cellSens Dimension system; Olympus). cDNA microarray. Dispersed cells from the anterior lobes of S100β/GFP-TG male rats at P5 and P60 were separated into GFP-positive cell fractions by a cell sorter (MoFlo XDP, Beckman Coulter, Fullerton, CA, USA). Total RNA as prepared with TRIzol reagent (Thermo Fisher Scientific) from the GFP-positive cell fraction at P5 and P60 and incubated with RNase-free DNase I (1 U/tube; Promega, Madison, WI, USA). Microarrays were performed on total RNA samples by a custom analysis service (TORAY, Tokyo, Japan) using 3D-Gene (Rat Oligo Chip 20 k). Microarray data were normalised by median levels.
qPCR. qPCR was performed as described previously 33 . Briefly, qPCR assays were conducted on a Thermal Cycler Dice Real Time System II (Takara, Shiga, Japan) using gene-specific primers and SYBR Premix Ex Taq II (Takara) containing SYBR Green I. Gene-specific primer sequences are listed in Supplementary Tables S2 and  S3. For normalisation, levels of β-actin (Actb), TATA binding protein (Tbp), and glyceraldehyde-3-phosphate dehydrogenase (Gapdh) were quantified 34 . The relative gene expression was calculated by comparing cycle times for each target PCR. Cycle threshold values were converted to relative gene expression levels using the 2-(ΔCt sample −ΔCt control) method.
In situ hybridisation. In situ hybridisation was performed with digoxigenin labelled cRNA probes, as described in our previous report 33 . DNA fragments were amplified from rat pituitary cDNA using PCR with the primer sets listed in Supplementary Table S2. After in situ hybridisation, sections or cells were subsequently stained using immunohistochemistry, as described above. A control experiment using the sense cRNA probe was performed, and no specific signal was detected. Cells were scanned using a microscope (BX61, Olympus).
Western blotting. Immunoblot analysis was performed as described previously 33 . We applied 20 μg of protein from rat anterior lobes to Ssodium dodecyl sulfate-polyacrylamide gel electrophoresis followed by electrophoretic transfer to an Immobilon-P transfer membrane (Merck Millipore). The membrane was blocked with