Hedgehog signaling activation induces stem cell proliferation and hormone release in the adult pituitary gland

Hedgehog (HH) signaling is known to be essential during the embryonal development of the pituitary gland but the knowledge about its role in the adult pituitary and in associated tumors is sparse. In this report we investigated the effect of excess Hh signaling activation in murine pituitary explants and analyzed the HH signaling status of human adenopituitary lobes and a large cohort of pituitary adenomas. Our data show that excess Hh signaling led to increased proliferation of Sox2+ and Sox9+ adult pituitary stem cells and to elevated expression levels of adrenocorticotropic hormone (Acth), growth hormone (Gh) and prolactin (Prl) in the adult gland. Inhibition of the pathway by cyclopamine reversed these effects indicating that active Hh signaling positively regulates proliferative processes of adult pituitary stem cells and hormone production in the anterior pituitary. Since hormone producing cells of the adenohypophysis as well as ACTH-, GH- and PRL-immunopositive adenomas express SHH and its target GLI1, we furthermore propose that excess HH signaling is involved in the development/maintenance of hormone-producing pituitary adenomas. These findings advance the understanding of physiological hormone regulation and may open new treatment options for pituitary tumors.

state. Consequently, HH pathway activity was proposed to prevent the development of PA 21 . However, the proof of this hypothesis is missing.
We here assessed the effect of activation and inactivation of Hh signaling on morphology, hormone expression/release and proliferation of pituitary explants and single cells isolated from Ptch flox/flox CreERT2 mice 23 . In addition, we analyzed the activation status of HH pathway in the human adenopituitary (qRT-PCR: n = 12, specific immunohistological stainings: n = 15) and in a large cohort of human pituitary tumors (qRT-PCR: n = 48, specific immunohistological stainings: n = 96). In contrast to former reports, our results demonstrate that active Hh signaling induces proliferation of Sox2 + and Sox9 + adult pituitary stem cells and hormone release in the adult pituitary gland. Finally our data suggest that activation of the HH pathway may be involved in the formation and/ or maintenance of pituitary tumors. Therefore, inhibition of Hh signaling could be a promising new target for the treatment of aggressive PA.

Hh signaling activation induces hormone secretion and proliferation of Sox2 + and Sox9 + adult pituitary stem cells in murine pituitaries. To investigate if activation of Hh signaling triggers expression
of pituitary hormones, we analyzed pituitaries from Ptch flox/flox CreERT2 +/− mice 17 d after tamoxifen-mediated induction of a biallelic Ptch deletion (tPtch −/− mice) 23 . The floxed Ptch flox loci of the posterior and the anterior tPtch −/− pituitaries were efficiently deleted (Fig. 1a), which resulted in significantly elevated Gli1 transcription (Fig. 1b, p = 0.023) and thus in activation of Hh signaling in the respective glands. Additionally, the Pomc expression levels of the pituitaries and the Acth serum levels of tPtch −/− mice were increased compared to the controls (Fig. 1c,d).
Since the animals are in a very poor general condition 17 d after induction of the Ptch deletion 23 , the elevated Acth levels in tPtch −/− mice could be stress-induced or correlated with activation of the Hh signaling pathway. To prevent potential hormonal feed-back loops that may have distorted the results concerning the direct influence of Hh signaling on hormone production, we isolated Ptch flox/flox CreERT2 +/− pituitaries and induced the Ptch mutation ex vivo by administrating tamoxifen (for verification of successful culture see Fig. 2 showing immunohistological stainings of cultured explants). This resulted in recombination of the Ptch flox loci (Ptch del ) (Fig. 1e), a significant increase of mutant Ptch transcripts (Ptch del ) ( Fig. 1f) (p = 0.0013) and a loss of wt Ptch expression in Ptch flox/flox CreERT2 +/− compared to control pituitaries (Fig. 2). As expected, the mutation furthermore resulted in elevated Gli1 expression levels and thus activation of Hh signaling in Ptch flox/flox CreERT2 +/− glands (Fig. 1g). Similar results were obtained after incubation of normal murine pituitaries with rShh-N, which likewise resulted in an upregulation of Gli1 and wt Ptch transcription, however to a much lesser extent (Suppl. Fig.  S1a,b). When the glands were treated with cyclopamine, the Gli1 expression levels were considerably reduced in tamoxifen-treated Ptch flox/flox CreERT2 +/− , in rShh-N-treated and solvent-treated organs ( Fig. 1g; Suppl. Fig. S1b). Most interestingly, Ptch deletion also resulted in a tendency towards increased Pomc and Glycoprotein hormones alpha chain (Cga) transcript levels (Fig. 1j,k), in a significantly elevated Acth release (Fig. 1m) (p = 0.023) and upregulation of Gh and Prl expression (Fig. 1h,i). These effects were abrogated by cyclopamine treatment (Fig. 1h-k), indicating a direct involvement of Hh signaling in the transcriptional activation of these genes. In contrast, the expression levels of Oxt, Tshβ , Lhβ and Fshβ were not altered by Ptch depletion, cyclopamine or rShh-N treatment (Suppl. Fig. S2).
Immunohistological examinations of short-term cultured explants of tamoxifen-treated Ptch flox/flox CreERT2 +/− and control pituitaries revealed no obviously changed expression patterns of Gh, Prl and Acth, the adult pituitary stem cell marker Sox2 and Sox9 24 , Gli1, and Ptch (Fig. 2). Remarkably, the expression of Sox2, Sox9, Gli1 and Ptch overlapped in cells of the intermediate zone, the marginal zone and the anterior lobe although Gli1 or Ptch positivity was not exclusively restricted to Sox2 + or Sox9 + cells (Fig. 2). Beyond that double immunofluorescent stainings against Sox2 and Gli1 revealed that all Sox2 + cells of the intermediate lobe, the marginal zone (data not shown) and the anterior lobe also are Gli1 + (Fig. 3a,b). However Gli1 expression was not exclusively restricted to Sox2 + cells (Fig. 3a,b).
Gli-dependent activation of the murine and human Pomc/POMC promoter. Gli1 and Sox2 expression patterns largely overlap in the intermediate lobe, the marginal zone and the anterior lobe, but Gli1 positivity was not restricted to Sox2 + cells (see Fig. 3a,b). This suggests that Hh signaling might be also implicated in the function of endocrine cells. Sequence analyses of the mouse, rat and human Pomc/POMC promoter revealed two Gli binding sites upstream of the first ATG (Suppl. Fig. S4a). To validate our finding from the ex vivo explant experiments we tested whether activation of Hh signaling directly induces Acth-secretion in the murine PA cell line AtT-20 that has been widely used as a model system for Acth-expressing pituitary cells 20,21,[26][27][28] . However, AtT-20 cells express only low levels of Ptch and no Gli1 or Gli2, and an induction of these genes upon Shh stimulation was not possible (data not shown). Nevertheless Gli1 or Gli2 overexpression resulted in transcriptional activation of the Ptch gene (Suppl. Fig. S4b), higher Pomc transcription and activation of the human POMC promoter (Suppl. Fig. S4c,e). Moreover Gli1 or Gli2 overexpression -contrary to Shh treatment -did not alter the proliferative capacity of AtT-20 cells (Suppl. Fig. S4d) which is in line with the observation that Ptch depletion does not induce proliferation of Acth-expressing cells in pituitary explants (see Fig. 3o,p). Besides, these results indicate that Shh induces proliferation in AtT-20 cells in a Gli1/Gli2-independent manner.

Expression of SHH and GLI1 in corticotropic, somatotropic and lactotropic cells of the human adenohypophysis and related adenoma subtypes.
To transfer our results to the human situation we analyzed the HH signaling activity in 15 normal adenopituitaries by GLI1 in situ hybridization ( Fig. 4a,b) and/or anti-SHH immunostaining using an antibody generated against the N-terminal part of SHH protein ( Fig. 4c-f). Cells throughout the adenohypophysis homogenously expressed GLI1 (Fig. 4a,b) and stained positive for SHH to, however, variable extent. We found cell populations displaying an intense cytoplasmic and predominantly granular SHH staining pattern as well as cells with restricted immunoreactivity to particularly perinuclear structures (Fig. 4c). Double immunofluorescent stainings using antibodies against SHH and several hormones produced in the adenohypophysis (ACTH, GH and PRL) support former studies 21 that ACTH-expressing (corticotropic) cells show a distinct, homogenous SHH immunoreactivity (Fig. 4d). However, besides co-localization of SHH and corticotrophs circumscribed, perinuclear, dot-like SHH positivity in obviously lactrotrophs and somatotrophs was observed (Fig. 4e,f). Since binding of SHH to PTCH results in internalization of both proteins to perinuclear lysosomes or endosomes 29-31 the distinct SHH signals in these endocrine active cells (Fig. 4e,f) might reflect its localization to the respective vesicles and therefore might stain SHH-responding cells. In contrast, SHH might co-localize to ACTH storing vesicles in corticotropic cells (Fig. 4d).
We next quantified the SHH and GLI1 transcript levels of 12 normal adenopituitaries and 48 PA by qRT-PCR (Table 1, Fig. 5a,b). GLI1 expression levels were significantly elevated in ACTH-, GH-and PRL-expressing adenomas (p = 0.026; p = 0.035; p = 0.0059, respectively) whereas null cell adenomas and FSH-expressing tumors showed similar GLI1 expression levels compared to normal adenopituitary lobes (Fig. 5a). Although SHH expression levels of the tumors did not reach significance a generally higher SHH expression of all PA in comparison to the controls was detected (Fig. 5b). Correlation analysis revealed that the SHH and GLI1 expression levels significantly correlate on transcript level throughout the tumor collection ( Fig. 5c; Spearman r-coefficient 0.8571; p = 0.0238).
Subsequently we analyzed the SHH and GLI1 expression levels in a large cohort of 92 PA and 4 pituitary carcinomas by GLI1 in situ hybridization and SHH immunohistology ( Table 2, Fig. 5d-g, Suppl. Figs S5 and S6). Throughout the tumor collection a strong and significant correlation of GLI1 and SHH expression was observed ( Fig. 5h; Spearman r-coefficient 0.8062; p = 0.0065). Very intense signals were detected in ACTH-and GH-producing tumors ( Fig. 5d-g, Suppl. Figs S5 and S6). Prolactinomas as well as mixed GH/PRL-producing adenomas showed high GLI1 expression levels and moderate immunoreactivity for SHH ( Fig. 5d-g, Suppl. Figs S5 and S6). TSH-and FSH-expressing tumors and null cell adenomas were moderately positive for both ( Fig. 5d-g, Suppl. Figs S5 and S6). LH-and FSH/LH-adenomas expressed low to moderate GLI1 and only low SHH levels ( Fig. 5d-g, Suppl. Figs S5 and S6). In pituitary carcinomas GLI1 and SHH expression could only be hardly detected ( Fig. 5d-g, Suppl. Figs S5 and S6). There was no association between HH signaling activation and proliferation rates within the tumor specimens (Fig. 5i,j).
Taken together, these data show that, in contrast to a former study 21 , not only corticotrophs but at least lactotrophs and somatotrophs of the normal human adenopituitary express SHH. Moreover, our results suggest that HH signaling activity may play a role in formation and/or maintenance of PA subtypes, especially in ACTH-, GH-or PRL-immunopositive tumors that show the highest GLI1 and SHH expression levels. However, this is pure speculation and remains to be investigated in the future.

Discussion
Within this paper, we demonstrate that excess Hh signaling in the pituitary gland elevates the hormone expression and induces proliferative processes of Gli1 + /Sox2 + and Sox9 + adult stem cells. However, in contrast to neural stem cells in which Hh signaling activation results in stem cell accumulation 32 Ptch-depletion in the pituitary does not impact the absolute cell number but enhances the percentage of proliferating stem cells. Due to the fact that early Sox2 + progenitors can give rise to Sox2 + /Sox9 + transit-amplifying cells 25  hormone producing cell subtypes 24 , our data thus indicate that Hh signaling activation induces proliferation and differentiation processes of Sox2 + cells, most likely by asymmetric divisions. This might result in maintenance of the stem cell pool and simultaneously in generation of more differentiated hormone producing daughter cells. Moreover our findings that Gli1 expression is not solely restricted to Sox2 + cells furthermore indicate that the Hh pathway -similarly to neural fate specification and cell cycle progression in the retina 33 -might be involved in the   Table 2. ACTH, adreno corticotrophic hormone; GH, growth hormone; PRL, prolactin; FSH, follicle stimulating hormone; LH, luteinizing hormone; TSH, thyroid stimulating hormone; CARC, carcinoma. Mean+/−SEM, *p < 0.05, **p < 0.01. Scale bar: 20 μ m. function of different cellular pituitary subtypes. In fact excess Hh signaling did not induce proliferation of differentiated endocrine cells (see Fig. 3) and Acth-producing AtT-20 cells but enhanced the hormone expression in pituitary explants and Pomc/POMC promoter activity in AtT-20 cells. This implies that activation of Hh signaling in endocrine pituitary cells induces hormone production rather than proliferative processes. This is in contrast to Sox2 + /Sox9 + pituitary stem cells, in which Hh signaling activates proliferation. Contrary to excess Hh signaling activation due to Ptch-depletion a moderate pathway activity induced by Shh did not alter the proliferative capacity of primary cultured pituitary cells. In contrast, Shh induced proliferative processes in AtT-20 cells. Since this was not accompanied by elevated Gli1 or Gli2 expression levels we conclude that it is impossible to induce and study the effects of canonical Hh signaling in AtT-20 cells upon Shh treatment. The fact that Gli1 or Gli2 overexpression in AtT-20 cells also did not affect the proliferation rate of these cells supports our suggestion that activation of Hh signaling has no impact on the proliferation rate of hormone-producing pituitary cells. Beyond that these data indicate that exceeding of a certain Hh signaling threshold is necessary to induce proliferation of pituitary stem cells.
In accordance with our murine explant experiments human adenopituitaries also express GLI1 and SHH. Moreover SHH expression pattern overlaps with hormonally active cells (e.g. ACTH-, PRL-and GH-producing cells). Similar to pituitary gland development in zebrafish 34 these data suggest that HH signaling may also play a role in these cellular subpopulations in the adult human pituitary. Remarkably, also ACTH-, GH-or PRL-immunopositive human PA express high SHH/SHH and GLI1 levels whereas TSH-, LH-, or FSH-producing tumors showed only moderate or low HH signaling activity. Previously it has been proposed that active HH signaling is restricted to ACTH-producing endocrine cells and that SHH maintains pituitary cells in a non-proliferative state. Moreover it was suggested that a down-regulation of HH signaling may be involved in the pathogenesis of PA 21 . In contrast to these studies we here quantified the GLI1 transcript level which is the best indicator of HH signaling activity. Furthermore, we used a SHH antibody that detects the active N-terminal SHH fragment and thus visualizes SHH-expressing and SHH-responding cells. Finally, we revealed a significant and strong correlation of GLI1 and SHH/SHH expression levels in the human PA collection (p = 0.0065 and p = 0.0238). Consequently, our results suggest that HH signaling is active in ACTH-, GH-and PRL-expressing cells of the adenohypophysis and in the respective PA subtypes and might indicate a role of HH signaling in the development or maintenance of PA (e.g. ACTH-, PRL-, GH-immunopositive PA). It is furthermore tempting to speculate that HH signaling activation (e.g. by PTCH mutations) itself drives tumor formation (e.g. in a paracrine, ligand dependent manner) and can drive hormone secretion in pituitary adenomas. However the fact that GLI1 or SHH/SHH expression did not correlate with the proliferation status of PA may be explained by intratumor heterogeneity. Furthermore, Sox2 and Sox9 expression have been associated with tumor growth 35,36 , self-renewal of oncogene target cells, tumor initiation and invasion 37 . Moreover Sox2 + pituitary cells have tumor-inducing potential 36 indicating that Sox2 + / Sox9 + cells indeed play a role in PA formation. Beyond that it is interesting that pituitary carcinomas exhibit the lowest levels of HH signaling activity. These tumors are most commonly ACTH-or PRL-secreting, invasive macroadenomas that spread distant metastases. Although the analyzed case number was low, it is tempting to speculate that the conversion of a rather benign pituitary tumor (i.e. adenoma) into a metastatic tumor necessitates and correlates with downregulation of HH signaling.
To our knowledge mutations in components of the HH signaling pathway have not been reported in tumors of the pituitary. PA are the third most common intracranial tumors and the estimated prevalence in the general population is approximately 17% 38 . Surgical resection of this tumor is the primary therapy for most patients. However in a portion of patients, surgery does not result in cure 39 . Our findings suggest that the HH pathway plays a role in the pathogenesis of PA and hormone production, which however have to be verified by future experiments. If this comes true, it would be interesting to test HH pathway inhibitors (e.g. the FDA-approved HH-inhibitor vismodegib) alone or in combination with other drugs to target pituitary tumor cells (reviewed in 40 ).
In summary our data demonstrate that activation of Hh signaling (e.g. by enhanced Gli1 expression), similarly to neural fate specification and cell cycle progression in the retina 33 , induces cell type-specific cellular processes in the pituitary (e.g. hormone-release of endocrine cells and proliferation of adult pituitary stem cells). Finally, we present data which might be indicative for a role of HH signaling in the development or maintenance of PA.

Methods
Ethical approval and informed consent. All  Mice. Experiments using animals were performed in compliance with all German legal and ethical requirements. The Rosa26CreERT2 (CreERT2) knock-in mouse strain (kindly provided by Dr. Anton Berns, The Netherlands Cancer Institute, Amsterdam, The Netherlands) expresses a fusion gene encoding Cre recombinase and a modified ligand-binding domain for the estrogen receptor under control of the endogenous Rosa26 promoter 41 . Generation, genotyping and CreERT2-activation of Ptch flox/flox CreERT2 +/− mice are described in Uhmann et al. 23 . Briefly, CreERT2 mice were bred to Ptch flox/flox mice to obtain Ptch flox/flox CreERT2 +/− mice. Eight-week-old Ptch flox/flox CreERT2 +/− mice were injected intraperitoneally with 1 mg tamoxifen dissolved in a 1:10 ethanol-sunflower oil emulsion 42  Isolation of genomic DNA and analysis of genomic recombination at the Ptch flox locus. Isolation of genomic deoxyribonucleic acid (DNA) from freshly isolated and cultured pituitary glands and quantification of the recombination efficiency at the Ptch flox locus were performed as previously described 23 . For polymerase chain reaction (PCR)-based detection of the CreERT2-mediated Ptch flox recombination the primer pair 5′ -gcatgtgacctgcctactaattc-3′ /5′ -cctacttatctgatggtctgcatc-3′ was used.
Plasmids. The Gli-binding site luciferase reporter construct (p9xGliBS) has been described previously 43
Hormone determination. Acth blood serum levels were quantified using an enzyme-linked immunosorbent assay (ELISA) (Uscn Life Science Inc., BIOZOL Diagnostica). Acth concentrations of supernatants of cultured pituitary were quantified using the IMMULITE ® 2000 Immunoassay System (Siemens AG).
In situ hybridization, immunohistological and immunofluorescent stainings of cultured murine pituitary gland. Cultured murine pituitary glands were fixed in 4% PFA, embedded in paraffin and sectioned. The sections were stained with hematoxylin and eosin (H&E) for histopathological analyses or were analyzed by in situ hybridization, immunohistological or immunofluorescent stainings. The in situ hybridization procedure and the probes for detecting wt Ptch and whole Ptch transcripts have been described previously 48 . Immunohistological and immunofluorescent stainings were conducted by boiling the sections in 10 mM citric acid, pH 6.0 (which also allows for specific staining with anti-BrdU antibodies) 49  Negative controls were carried out by incubation in the absence of the primary antibody and always yielded negative results.
Human pituitary tissue samples. Surgical specimens from 142 different patients with sellar lesions (n = 145) were retrieved from the Department of Neuropathology of the University Hospital of Erlangen. Each tumor specimen was classified and graded according to the currently valid version of the World Health Organization classification system of tumors of endocrine organs 22 . Twelve normal pituitary tissues and five different tumor subtypes (n = 48) were analyzed by qRT-PCR for relative GLI1 and SHH expression level (data for each case are summarized in Table 1). Fifteen normal pituitary tissues and ten different tumor subtypes (n = 96; data for each case are summarized in Table 2) were analyzed by specific histological stainings (GLI1 in situ hybridization, SHH immunohistological stainings). Tumors were grouped according to clinical symptoms and hormone release as following: 1) ACTH producing adenomas with and without clinical signs of Cushing's disease (qRT-PCR: 7 male, 9 female, median age 49.7 years; specific histological stainings: 23 tumors from 22 different patients; 8 male,14 female, median age 42.8 years); 2) adenomas from patients with clinical signs of acromegaly and immunohistochemical expression of GH (qRT-PCR: 6 male, 6 female, median age 39.4 years; specific histological stainings: 6 male, 8 female, median age 45.2 years); 3) mixed adenomas from patients with clinical signs of acromegaly and combined immunohistochemical expression of GH and PRL (specific histological stainings: 4 male, 6 female, median age 31.8 years); 4) clinically nonfunctioning tumors with expression of FSH (qRT-PCR: 4 male, 2 female, median age 54.1 years; specific histological stainings: 5 male, 2 female, median age 51.7 years); 5) clinically nonfunctioning adenomas with expression of LH (specific histological stainings: 4 male, 3 female, median age 61.4 years); 6) clinically nonfunctioning tumors with combined expression of FSH and LH (specific histological stainings: 6 male, 1 female, median age 58.3 years); 7) nonfunctioning adenomas without detectable hormone expression (qRT-PCR: 2 male, 1 female, median age 49.0 years; null cell adenomas; specific histological stainings: 5 male, 3 female, median age 46.8 years); 8) TSH producing adenomas (specific histological stainings: 2 male, 5 female, median age 38 years); 9) PRL producing adenomas (qRT-PCR: 3 male, 8 female, median age 46.7 years; specific histological stainings: 6 male, 3 female, median age 39.4 years); 10) pituitary carcinomas (specific histological stainings: 4 tumors and/or metastases from 3 different patients; 2 male, 1 female, median age 39 years). Data for each case included in the study are shown in Table 1 (qRT-PCR) and Table 2 (specific histological stainings).
Normal pituitary tissue samples (n = 27) were acquired from patients with sellar exploration in cases of magnetic resonance imaging negative microadenomas and Rathke's cleft cysts. The latter specimens showed a regular lobulated reticulin fiber network typically observed in the adenohypophysis and IHC confirmed the regular spectrum of hormone production.
The study concept was approved by the Ethical Committee of the University of Erlangen-Nürnberg.
IHC procedure and evaluation on human samples. IHC was performed as described using a semiautomated benchmark apparatus (Nexes; Ventana, Illkirch, France) and the Ventana DAB staining system 50 . Positive and negative controls were used to validate the staining of the primary antibodies listed in Supplemental Table S3.
The specificity of the antibodies was tested by staining human adenopituitaries showing selective populations of cells only. The specificity of the anti-SHH antibody was verified by staining human pancreas carcinoma and intestine. Negative controls were carried out by incubation in the absence of the primary antibody and always yielded negative results. The intensity of SHH staining was assessed semi-quantitatively blinded to any diagnosis concerning the hormonal status of patients (D.S. and R.B.) and grouped into five different categories (0 = no SHH expression; 1 = very low intensity, 2 = moderate intensity, 3 = strong intensity, 4 = highest intensity of all). In cases without detectable SHH expression, the IHC was repeated to verify the result. The proliferation rate of each tumor was assessed quantitatively (D.S.) by counting a minimum of 1,000 tumor cell nuclei in a region with the highest amount of distinct positive Ki76 + cell nuclei.
In situ hybridization and evaluation on human samples. In situ hybridization was conducted as described 48,51 using antisense and sense (negative control) riboprobes reversely transcribed from a pBS-hGLI1 plasmids containing the hGLI1 cDNA (GenBank: X07384.1) (provided by Mark Wijgerde).
The intensity of GLI1 in situ hybridization was assessed semi-quantitatively blinded to any diagnosis concerning the hormonal status of patients (A.U., D.S. and R.B.) and grouped into 7 different categories (0 = no expression; 0.5 = very low intensity, 1 = low intensity, 2 = intermediate intensity, 2.5 = intermediate to strong intensity, 3 = strong intensity, 3.5 = highest intensity of all).
Statistical analyses. Statistical analyses were conducted using the software GraphPadPrism 6 (GraphPad Software Inc.). Correlation or significance of GLI1 and SHH/SHH expression in human PA was tested by a nonparametric Spearman correlation test or unpaired nonparametric Mann-Whitney test, respectively. Statistical significance of the recombination efficiency at the Ptch flox locus, the gene expression in pituitary glands and human PA, the Acth blood serum levels and Acth concentration of media were tested using Holm-Sidak's multiple comparison test (unpaired t test). Statistical significance of gene expression levels of ex vivo cultured pituitaries was tested after outliers correction using ROUT method (Q = 1%) and testing the Gaussian distribution by D' Agustino and Pearson omnibus normality test followed by an one-way-ANOVA (analysis of variance; Holm-Sidak's multiple comparison test). Statistical significance of proliferative Sox2 + or Sox9 + cells, of BrdU incorporation assays of in vitro cultured pituitary cells or AtT-20 cells, of dual-luciferase assays and expression analyses of AtT-20 cells were tested using nonparametric Mann-Whitney test, Dunn's multiple comparison test or Kruskal-Wallis test, respectively.