Author contributions: J.C.: conception and design, collection and/or assembly of data, data analysis and interpretation, final approval of manuscript; L.G.: conception and design, collection and/or assembly of data, data analysis and interpretation, final approval of manuscript; Q.F.: collection and/or assembly of data, data analysis and interpretation, final approval of manuscript; D.L.: collection and/or assembly of data, data analysis and interpretation, final approval of manuscript; S.V.L.: collection and/or assembly of data, data analysis and interpretation, final approval of manuscript; H.V.: conception and design, collection and/or assembly of data, data analysis and interpretation, manuscript writing; final approval of manuscript. J.C. and L.G. contributed equally to this work.
First published online in Stem CellsExpress February 26, 2009
The pituitary gland represents the endocrine core, governing the body's hormonal landscape by adapting its cellular composition to changing demands. It is assumed that stem/progenitor cells are involved in this remodeling. Recently, we uncovered a candidate stem/progenitor cell population in the anterior pituitary. Here, we scrutinized this “side population” (SP) and show that, unexpectedly, not the subset expressing high levels of “stem cell antigen-1” (Sca1high) but the remainder non-Sca1high fraction clusters the pituitary progenitor cells. Transcriptomal interrogation revealed in the non-Sca1high SP upregulated expression of the pituitary stem/progenitor cell markers Sox2 and Sox9, and of multiple factors critically involved in pituitary embryogenesis. The non-Sca1high SP encloses the cells that generate spheres and display multipotent hormone differentiation capacity. In culture conditions selecting for the non-Sca1high subset within the SP, stem cell growth factors that induce SP expansion, affect transcription of embryonic factors, suggesting impact on a developmental program that unfolds within this SP compartment. Non-Sca1high SP cells, revealed by Sox2 expression, are observed in the postulated periluminal stem/progenitor cell niche, but also in small groups scattered over the gland, thereby advocating the existence of multiple niches. In early postnatal mice undergoing a pituitary growth wave, Sox2+ cells are more abundant than in adults, concordant with a larger SP and higher non-Sca1high proportion. Together, we tracked down pituitary progenitor cells by SP phenotype, and thus provide a straightforward method to isolate and scrutinize these cells from the plastic pituitary ex vivo, as well as a culture system for in-depth exploration of their regulatory network. Stem Cells 2009;27:1182–1195
The pituitary is known as the “master gland”, functioning as central endocrine regulator of growth, puberty, reproduction, metabolism, stress, and immune defense. To exert these core activities, the pituitary anterior lobe is populated by distinct endocrine cell types that produce prolactin (PRL; lactotropes), growth hormone (GH; somatotropes), adrenocorticotropic hormone (ACTH; corticotropes), thyroid-stimulating hormone (TSH; thyrotropes), or luteinizing hormone/follicle-stimulating hormone (LH/FSH; gonadotropes). TSH, LH, and FSH are heterodimers with a common α-glycoprotein subunit (αGSU) and a specific β-subunit. The different hormonal cell lineages arise during embryogenesis from common progenitors in the pituitary primordium, Rathke's pouch, under the influence of a well-orchestrated interplay between morphogens, growth, and transcription factors (reviewed in [1–3]). After birth, pituitary size and cell composition mature further. Once adult, the gland displays a slow but continual replacement of cells. In addition, hormonal cell populations dynamically adapt to fluctuating endocrine demands of the organism. It has repeatedly been suggested that the pituitary harbors tissue-specific stem/progenitor cells that participate in the generation of new hormonal cells during postnatal turnover and plasticity (reviewed in [4, 5]).
Recently, our group identified a “side population” (SP) in the anterior pituitary of mouse, rat, and chicken [6, 7], characterized by the efficient efflux of Hoechst dye through multidrug-transporters, in particular ABCG2. Today, it has been widely documented that a SP is present in many tissues and frequently enriched in (presumptive) stem/progenitor cells [4, 5, 8–22]. We found that also the pituitary SP enriches for cells with stem/progenitor-like features, including the high expression of “stem cell antigen-1” (Sca1), and the clonal generation of spheres [6, 7]. The Sca1high SP subset was subjected to a first molecular characterization which revealed expression of additional stem/progenitor cell-associated markers (Oct4, Nanog, Bmi1, CD133, nestin)  as well as of Notch signaling components . In the present study, we aimed at a more profound characterization of the Sca1high SP by comparing it with the remainder of the SP not expressing Sca1 at high levels and referred to as the non-Sca1high SP. Surprisingly, we discovered that not the Sca1high subset of the SP, but its non-Sca1high fraction clusters the pituitary-specific progenitor cells, as based on gene expression profile, sphere-forming ability, multipotency, topographical localization, and regulatory characteristics.
MATERIALS AND METHODS
SP Analysis and Sorting by Sca1 Level
Pituitary anterior lobes were isolated from female FVB mice, dissociated into single cells using trypsin, incubated with Hoechst 33342 (Sigma-Aldrich, Bornem, Belgium, http://www.sigmaaldrich.com), stained for Sca1 expression (for antibodies, supporting information Table 1), and examined or sorted using a FACSVantage (BD Biosciences, San Jose, CA, http://www.bdbiosciences.com), all as previously described [6, 7]. Verapamil control of the SP phenotype and isotype staining controls were performed as before [6, 7] (data not shown). Animal manipulations were approved by the University of Leuven Ethical Committee.
FACS Analysis of Endothelial Phenotype
Anterior pituitary cells were incubated with DiI-labeled acetylated low-density lipoprotein (Ac-LDL; 10 μg/ml; Biomedical Technologies, Stoughton, MA, http://www.btiinc.com/), added together with Hoechst. Labeling was measured by fluorescence-activated cell sorting (FACS) with standard PE-filters .
To analyze CD31 (Pecam1) immunophenotype by FACS , anterior pituitaries were dispersed using a collagenase/dispase mixture (Liberase Blendzyme 1; Roche Diagnostics, Vilvoorde, Belgium, http://www.roche-diagnostics.be) as trypsin destroyed the CD31 epitope. Although cell viability was lower (70% vs. 90%), collagenase dispersion also allowed to discern a SP.
Growth and Differentiation of Pituispheres
To grow spheres, FACS-sorted cells were handled as described in detail earlier  with the exception of using a coculture system here. Total anterior pituitary cells (270,000) were deposited at the bottom of the well to serve as feeder layer, and 20,000 sorted cells of each population were seeded in the well insert having a 1-μm pore-sized PET membrane (BD Biosciences). Cells were cultured in serum-free defined medium (SFDM; Invitrogen, Grand Island, NY, http://www.invitrogen.com) containing B27 (1:50; Invitrogen) and bFGF (20 ng/ml; R&D Systems Inc., Minneapolis, MN, http://www.rndsystems.com).
To explore differentiation capacity (method modified from ), pituispheres were manually picked at day 6 of culture, carefully rinsed in SFDM, transferred to coverslips coated with growth factor-reduced Matrigel (1:10; BD Biosciences), and cultured in SFDM without supplements for 1–2 weeks.
FACS-sorted cytospun cells, pituitary vibratome sections, and pituispheres were fixed using paraformaldehyde (4%), permeabilized with saponin (0.5%), or Triton-X100 (0.4%), incubated with primary antibodies overnight and finally stained with Alexa Fluor-conjugated antisera (supporting information Table 1), all as described earlier [6, 24]. As negative control, primary antibodies were omitted. Nuclei were counterstained with DAPI (Sigma-Aldrich) or ToPro3 (Invitrogen) for epifluorescent examination or confocal scanning (Zeiss LSM 510; Zeiss, Zaventem, Belgium, http://www.zeiss.com), respectively.
Serum-Free Aggregate Culture
Anterior pituitary cells from adult female mice were cultured as reaggregates in SFDM, in previous work extensively shown to mimic in vivo pituitary biology . Cultures were treated from their initiation with bFGF (20 ng/ml), EGF (20 ng/ml; R&D Systems Inc.), or LIF (10 ng/ml; Millipore, Brussels, Belgium, http://www.millipore.com) for 10–14 days. Aggregates were then dispersed into single cells, analyzed for SP and endothelial phenotype, and SP cells sorted for (semi-)qRT-PCR analysis.
Whole-Genome Expression Profiling
Sca1high SP, non-Sca1high SP, and “main population” (MP) cells (representing the bulk of the anterior pituitary) were sorted by FACS into cold lysis solution of the RNeasy Micro kit (Qiagen, Venlo, The Netherlands, http://www.qiagen.com). To obtain a Sca1high subset as pure as possible, gates (Fig. 1A, 1B) were positioned to include only the cells with highest signals (approximately 60% of the SP) which is more stringent than before [6, 7]. RNA was immediately extracted according to the manufacturer's protocol, starting from 25,000–50,000 cells after a typical sort. RNA quality and concentration were determined using Picochips on a BioAnalyzer 2100 (Agilent Technologies, Santa Clara, CA, http://www.agilent.com). Only when the RNA Integrity Number was ≥8.0, samples were used for microarray examination.
RNA (20 ng) was subjected to two successive rounds of linear amplification by in vitro transcription while incorporating Cy3-dCTP . cRNA probes were hybridized onto Agilent whole mouse genome oligonucleotide arrays in a 1 × 44k (G4122A; n = 1) and 4 × 44k (G4122F; n = 2) print format. Scanned images were processed using Agilent's Feature Extraction Software. Reliability of the amplification method was verified for several genes using (semi-)qRT-PCR; expression differences between the distinct cell populations were comparable in the original RNA extracts and the corresponding amplified samples (not shown).
Three independent cell sortings were done. Statistical analysis was carried out using the data sets from the two independent 4 × 44k arrays. The first sorting, hybridized on the 1 × 44k array format, was found to confirm the conclusions (not shown). Raw signals were background-corrected, subjected to quantile normalization, and log2-transformed. First, Sca1high SP was compared with non-Sca1high SP by principal component analysis (PCA), hierarchical clustering, and volcano plotting with BioConductor in R (www.bioconductor.org), all performed on the set of genes with highest expression levels; i.e., above the 25th percentile of all signal intensities (18,181 probe sets). Second, the complete set of genes with signal intensities above the arrays' background were examined by Student's t test. Genes altered between the Sca1high and non-Sca1high SP by more than or equal to twofold with p < .05 were considered differentially expressed. Gene Set Enrichment Analysis in R was performed on this list of differentially regulated genes. Functional annotation is not discussed in detail here as it does not directly serve the goal of the present study. Some limited observations on selected biological processes and pathways are reported (supporting information Fig. 3). A comprehensive analysis and discussion will be presented elsewhere, and array data files will be deposited on that occasion. Third, gene expression was also compared between the non-Sca1high SP and the MP using the same significance criteria of more than or equal to twofold difference with p < .05.
Semi-qRT-PCR was used to validate and extend the findings obtained from the microarray interrogation and was performed as described earlier [6, 7]. Briefly, RNA was serially diluted starting from the maximum amount available (typically 10–100 ng), reverse-transcribed and subjected to PCR using the primers and conditions listed in supporting information Table 2. The ribosomal protein L19 was used as internal control and normalization standard. Amplicons were sequenced by Lark Technologies (Essex, U.K., http://www.larktechnologies.com) to confirm their identity.
qRT-PCR was also done as outlined earlier [6, 7] using Taqman Gene expression assays (Applied Biosystems, Foster City, CA, http://www.appliedbiosystems.com). 18S ribosomal RNA was used as internal control and results analyzed with the ΔΔCT method (User Bulletin #2 from Applied Biosystems).
Whole-Genome Expression Profiling Reveals a Striking Dichotomy Between the Sca1high and the Non-Sca1high SP Subsets
Recently, we identified a SP in the mouse anterior pituitary enriched in cells expressing high levels of Sca1. A first molecular characterization of the Sca1high SP subset exposed the expression of additional stem/progenitor cell-associated markers and signaling factors [6, 7]. Here, we intended to more thoroughly characterize this Sca1high SP fraction by comparing it with the remainder of the SP, referred to as the non-Sca1high subset (Fig. 1A, 1B).
First, from the dual-wavelength FACS plot it is obvious that Sca1high cells (approximately 60% of the SP) are found in a different location of the SP streak than the non-Sca1high cells, the former comprising the right-lower portion and the latter the left-upper rim with some clustering in the proximal part (supporting information Fig. 1). Both SP subsets were subjected to microarray gene expression profiling. PCA of the genes with highest signal intensities (18,181 probe sets) clearly separates the Sca1high from the non-Sca1high SP, a finding confirmed by pairwise comparisons (supporting information Fig. 2). A total of 2,738 genes are significantly upregulated (more than or equal to twofold, p < .05) in the Sca1high SP, whereas 2,968 are expressed at higher levels in the non-Sca1high SP (depicted as volcano plot in Fig. 1C). Unsupervised hierarchical clustering further corroborates the striking gene-expression dichotomy between the Sca1high and the non-Sca1high SP (Fig. 1D).
A comprehensive gene ontology analysis is beyond the goal of the current study. Gene set enrichment of some selected biological processes and of signaling pathways are summarized in supporting information Figure 3. Genes functioning in the Notch, transforming-growth factor β (TGFβ), and Jak/Stat pathways, and genes associated with apoptosis and angiogenesis are upregulated in the Sca1high SP, whereas genes involved in mitogen-activated protein kinase (MAPK), Wnt, and Notch signaling, and those implicated in negative regulation of apoptosis, are enriched in the non-Sca1high SP.
The Sca1high SP Is Largely Enriched in Cells with Endothelial Phenotype
As microarray analysis revealed in the Sca1high SP prominent expression of angiogenesis-related genes (for a limited overview, Table 1), we first wanted to know the frequency of endothelial-type cells in this fraction. By examining the uptake of Ac-LDL as a hallmark of endothelial cells (EC) , we found that the majority of the Sca1high SP cells (79.0 ± 2.8%; mean ± SEM; n = 11) display an endothelial phenotype, whereas the non-Sca1high SP is much more depleted from Ac-LDL-labeled cells (21.9% ± 2.7%; Fig. 1E, 1F). In accord with these findings, Ac-LDL+ cells are mainly located in the lower-right portion of the SP streak whereas Ac-LDL− cells reside in the left-upper border and proximal part (supporting information Fig. 4). In addition, the endothelial marker CD31 (Pecam1)  was detected by FACS in a similar large proportion of the Sca1high SP cells (78.3 ± 4.3%; n = 4), whereas the non-Sca1high SP is again much more devoid of CD31-expressing cells (20.7 ± 1.8%; n = 4). Together, these data indicate that the pituitary Sca1high SP predominantly clusters cells with endothelial phenotype.
Table 1. Comparison of non-Sca1high SP, Sca1high SP, and MP for expression of genes implicated in stem/progenitor cells, early pituitary embryogenesis, Notch signaling, growth-factor activity, and angiogenesis, as extracted from the microarray gene expression profiles
Not the Sca1high but the Non-Sca1high SP Shows an Expression Profile Indicative of Pituitary Stem/Progenitor Cells
To find out whether the Sca1high SP, besides its major endothelial nature, still displays molecular features of (pituitary) stem/progenitor cells, we screened its transcriptome for stem/progenitor cell-associated genes vis-à-vis that of the non-Sca1high SP (Table 1). Remarkably, not the Sca1high fraction but the remainder SP subset of non-Sca1high cells shows upregulated expression of Sox2 and Sox9, transcription factors detected in stem/progenitor cells of a variety of tissues [26–28] and just recently advanced as markers of pituitary cells with stem/progenitor cell properties [23, 29]. Immunofluorescence evaluation confirmed these transcriptional data; Sox2+ and Sox9+ cells segregate to the non-Sca1high SP (representing 52.8 ± 3.3% and 36.7 ± 3.7% of those SP cells, respectively; n = 3), and are virtually absent from the Sca1high SP (Fig. 2A). These observations suggest the clustering of putative pituitary stem/progenitor cells in the non-Sca1high SP rather than in the Sca1high SP. Therefore, we further compared transcriptomes not only between the two SP subsets but also between the non-Sca1high SP and the MP, the latter encompassing the bulk of anterior pituitary (differentiated hormonal) cells. Also when compared with the MP, expression of Sox2 and Sox9 is upregulated in the non-Sca1high SP (Table 1; Fig. 2A–2C). In addition, also CD44  and Lgr5, a recently discovered stem cell marker , are specifically elevated in the non-Sca1high fraction (Table 1). The other stem/progenitor cell-associated factors CD133  and Cxcr4  show a tendency (p > .05) of higher expression in the non-Sca1high SP, whereas Oct4, Bmi1, and nestin [6, 24, 27] display transcript levels not different between the non-Sca1high and the Sca1high SP, but elevated when compared with the MP (Table 1; Fig. 2B, 2C). Expression of the pluripotency regulator Nanog  is not detectable in the microarrays, but is exposed by the more sensitive reverse-transcription polymerase chain reaction (RT-PCR) technique solely in the non-Sca1high SP (Fig. 2B, 2C). In contrast, transcripts encoding the ABCG2 transporter are less abundant in the non-Sca1high than the Sca1high SP (Table 1) which is in accordance with the clustering of non-Sca1high SP cells more proximally in the SP streak indicating lower efficiency of Hoechst efflux. Higher expression of Abcg2 in both SP subsets when compared with the MP is concordant with their SP phenotype. Finally, highest expression of Sca1 (Ly6a) in the Sca1high SP (Table 1) validates our sorting procedure.
We then zoomed in on the expression of transcription factors with critical role in pituitary early-embryonic development (reviewed in [1–3]), given the general idea that adult stem/progenitor cells may recapitulate the embryonic program when generating new cells [6, 7, 35, 36]. Pitx1, Pitx2, Isl1, Lhx3, Lhx4, Pax6, Six1, Six6, Msx1, and Otx2  are all more abundant in the non-Sca1high than the Sca1high SP of the adult pituitary (Table 1; Fig. 2B, 2C). Moreover, Hesx1 and Prop1, not measurable above the microarrays' background, were detected using RT-PCR in the non-Sca1high but not in the Sca1high SP (Fig. 2B, 2C). Of all these transcription factors, those with expression demonstrated to persist after embryogenesis and with established hormone-regulatory role in the postnatal gland (Pitx1, Pitx2, Lhx3, Isl1) [1–3] are equally or higher expressed in the MP. Interestingly, however, factors shown to largely extinguish after embryonic development and to control the embryonic progenitor cells (Hesx1, Prop1, Lhx4, and Pax6) [1–3] are higher expressed in the non-Sca1high SP than in the MP.
Notch signaling also plays a key role in pituitary organogenesis by regulating the embryonic progenitor cell pool [38–42]. As examined here in the adult gland, the receptors Notch1 and Notch4, the target genes Hes1 and Hes5, and the Notch inhibitor Numb display elevated expression levels in the Sca1high when compared with the non-Sca1high SP, and similar or higher levels in the non-Sca1high SP than the MP (Table 1; Fig. 2B, 2C). Interestingly, Notch2 and Notch3 are most abundant in the non-Sca1high SP. The latter Notch receptors are expressed in the periluminal progenitor zone of the embryonic pituitary, whereas Notch1 and Notch4 are not detectable in the developing gland. Also the Notch target gene, Hey1, is expressed in the embryonic progenitor area  and displays the highest mRNA levels in the non-Sca1high SP (Table 1; Fig. 2B, 2C). Of the Notch ligands Dll1-3, Jag1-2, and Dlk1 (Table 1; Fig. 2C), only Dlk1 displays upregulated expression in the non-Sca1high SP when compared with the Sca1high fraction. However, levels are even more elevated in the MP which is consistent with its expression and role in somatotropes .
All together, these data indicate that, unexpectedly, not the SP subset with highest Sca1 levels but the non-Sca1high fraction clusters the pituitary candidate stem/progenitor cells given their expression profile embracing the just recently proposed pituitary stem/progenitor cell markers Sox2 and Sox9, and multiple factors critically implicated in early pituitary embryogenesis.
The Non-Sca1high SP Isolates the Cells that Form Pituispheres and Display Multipotent Differentiation Capacity
Sphere formation is a property of stem/progenitor cells from a variety of tissues [44–46]. We were the first to demonstrate that also the anterior pituitary contains cells that clonally expand into spheres, and we found that these stem/progenitor-like cells segregate to the SP . To further pin down the SP fraction that encloses the sphere-initiating cells, non-Sca1high SP and Sca1high SP cells (as well as MP cells) were seeded in sphere-growing conditions as described earlier . As shown in Figure 3A–3C, pituispheres were only obtained from non-Sca1high SP cells (approximately 10–50 spheres from the sorted 20,000 cells). These spheres were similar in growth and morphology to the spheres obtained from the total anterior pituitary cell population that we described in detail earlier . In 6-day-old spheres, the majority of the cells contain Sox2 protein, and also nestin expression was found in most of the spheres (Fig. 3D, 3E). Clonal derivation of the pituispheres and ability to give rise to second-generation spheres, both previously described in detail , were confirmed here (data not shown). Further-generation spheres could so far not be obtained ( and Discussion).
To investigate whether the sphere-initiating non-Sca1high SP cells are multipotent, 6-day-old first-generation spheres were transferred to differentiation conditions (Matrigel). Although freshly sorted non-Sca1high SP contains a small portion of PRL+ cells (9.1 ± 1.1%; n = 3) and some sporadic αGSU+ cells (supporting information Fig. 5), no hormone-immunopositive cells could be detected in the spheres from the day of appearance (days 2–3; not shown) till the day of transfer to differentiation conditions (day 6; Fig. 3F). However, after 1–2 weeks in differentiation medium, all endocrine cell types of the anterior pituitary including GH+, PRL+, ACTH+, and αGSU+ cells (the latter including TSHβ+ thyrotropes and LHβ+/FSHβ+ gonadotropes; Fig. 3G–3J and not shown) were detected in the spheres, whereas Sox2 protein was not observed anymore (Fig. 3K). Hormone-immunoreactive cells were present at the border as well as on the inside of the spheres as shown by serial sections (supporting information Fig. 6) and 3D reconstruction (supporting information videos 1 and 2). Most spheres had adopted a more irregular shape after differentiation.
In summary, the non-Sca1high SP encloses the pituisphere-initiating cells that are capable of differentiating into all endocrine lineages, thereby supporting the identification of pituitary stem/progenitor cells by the non-Sca1high SP phenotype. As extended serial passaging of the pituispheres could not yet be obtained, it is at present more appropriate to term these cells “progenitor cells” instead of “stem cells” [23, 36].
Non-Sca1high SP Cells, Revealed by Sox2, Reside in the Postulated Periluminal Stem/Progenitor Cell Niche, but Additionally in Clusters Scattered Over the Anterior Pituitary
To obtain information on the localization of non-Sca1high SP cells within the adult pituitary, we examined Sox2 expression by immunofluorescence. Sox2 signals were detected in the periluminal zone (i.e., around the cleft; Fig. 4A), a Rathke's pouch remnant in the past repeatedly postulated to embody the pituitary stem/progenitor cell niche (reviewed in [4, 5]). This observation is in line with just recently reported findings [23, 29]. However, a careful look revealed some additional remarkable aspects. Sox2 protein is predominantly localized in the nucleus in the marginal cell layer that lines the cleft, whereas immunoreactive signal is mainly cytoplasmic in the submarginal zone, in cells located either directly apposed to the marginal layer (supporting information Fig. 7) or at some further distance (Fig. 4A). Moreover, nuclear Sox2 protein is also present in clusters of cells, scattered over the anterior lobe and often surrounding a small lumen (Fig. 4B, 4C). Cells with cytoplasmic Sox2 are again found in the vicinity of these clusters. The lumen-lining Sox2+ cells do not represent EC because they are not bordering sinusoidal capillaries, visualized using fibronectin  (Fig. 4C), or using autofluorescent red blood cells (data not shown). Finally, cytoplasmic Sox2+ cells without noticeable affiliation with nuclear Sox2+ cells were also detected, although less frequently, disseminated over the anterior pituitary and often in small groups of 2–4 cells (Fig. 4E, 4F).
In previous work, we detected nestin immunoreactivity in the periluminal zone including in some cleft-bordering cells, as well as in a pericyte location close to sinusoidal EC, as analyzed in rat pituitary . Accordingly, we here observed nestin in some of the marginal cells of the mouse pituitary (Fig. 4D) , as well as in the submarginal zone and in the body of the gland in which nestin+ cells mainly display a mesenchymal-like morphology. All nestin+ cells of the marginal cell layer express Sox2 (Fig. 4D), but not vice versa. No colocalization could be found in the rest of the anterior lobe.
Nuclear Sox2 protein was never observed in hormone-immunoreactive cells. Although cytoplasmic Sox2 signals were also not found together neither with ACTH nor GH immunoreactivity (Fig. 4E, 4F), colocalization was detected with αGSU or PRL in some cells (Fig. 4G, 4H), the incidence of which is region-dependent and estimated to be 1%–20% of the particular hormone-expressing cells. Interestingly, some of these double-positive cells were found in the neighborhood of the nuclear Sox2+ cells of both the marginal layer (Fig. 4H) and of the small clusters within the lobe (data not shown).
We further observed that Sox2+ cells are more abundant in the pituitary of early-postnatal mice at the age of the first pituitary growth wave (1-week-old) than in adult animals. In particular, the scattered nuclear Sox2+ cell clusters, as well as the cytoplasmic Sox2+ cells in close vicinity of cells with nuclear Sox2, are more prevalent (Fig. 4I). In agreement, the proportion of the SP in the neonatally developing pituitary is threefold to fourfold higher, and its non-Sca1high subset 1.5-fold larger, than in the adult pituitary (see  and supporting information Fig. 8).
Together, these data suggest the existence of multiple stem/progenitor cell niches in the anterior pituitary as revealed by Sox2 expression, embracing the periluminal cell layer but also small-cell clusters disseminated over the lobe.
Serum-Free Culture Selects for the Non-Sca1high Cells Within the SP, Which Are Responsive to Stem Cell Growth Factors
In our previous study, we observed persistence of SP cells when anterior pituitary cells were cultured as reaggregates in serum-free conditions . However, the SP proportion was smaller than ex vivo (approximately 0.4% vs. approximately 1.5%), and the number of Sca1high SP cells much lower (supporting information Fig. 9A, 9B). We here reasoned that the Sca1high SP, mainly encompassing EC, is disadvantaged in serum-free conditions given that EC typically require high amounts of serum to survive and grow in culture . Indeed, the SP after aggregate culture only contains a very low proportion of Ac-LDL+ cells (Fig. 5A), and CD31 mRNA levels are 500-fold lower than ex vivo, as measured by qRT-PCR (data not shown).
We previously also demonstrated that the stem cell growth factors LIF, bFGF, and EGF [44–46] induce proportional expansion of the SP in aggregate culture . These actions are in line with the present microarray data revealing higher expression of the cognate receptors in the non-Sca1high SP when compared with the MP (Table 1). In addition, we here observed that, although the SP expands 2.0-, 2.7-, and 4.6-fold when aggregates are cultured in the presence of LIF, bFGF, or EGF, respectively, the Ac-LDL+ cell proportion within the SP does not change (Fig. 5A, 5B), neither does the CD31 transcript level (data not shown) and the Sca1high proportion (supporting information Fig. 9). Hence, expansion of the SP under growth-factor treatment is not due to a prosurvival or mitogenic action on its EC compartment. Finally, these growth-signaling molecules do not only affect SP size, but also expression of embryogenesis-implicated factors within the SP. LIF stimulates the transcription of Isl1 and Hesx1 while reducing the level of Prop1 mRNA (Fig. 5C). Transcription of Lhx4 is stimulated by all three growth factors (Fig. 5D).
In summary, culturing anterior pituitary cells as reaggregates under serum-free conditions selects for the non-Sca1high cells within the SP. This culture method is expected to be valuable for exploring the regulatory network of pituitary progenitor cells. As a start, we found that LIF, bFGF, and EGF influence the expression of pituitary early-embryonic factors within the aggregates' SP.
In previous work, we discovered a SP in the postnatal anterior pituitary enriched for candidate stem/progenitor cells expressing high levels of Sca1 [6, 7]. Here, we further scrutinized the SP after Sca1-based dissection and found that, surprisingly, not the subset with high Sca1 expression (Sca1high SP), but the remainder of the SP referred to as the non-Sca1high SP, clusters the pituitary progenitor cells, as indicated by the following arguments. (a) The non-Sca1high SP displays upregulated expression of Sox2, Sox9, CD44, Lgr5, and Nanog, all markers of multiple stem/progenitor cells [26–28, 30, 31, 34]. (b) The non-Sca1high subset isolates the cells that expand into spheres, a property displayed by various tissue stem/progenitor cells [44–46]. (c) The sphere-initiating non-Sca1high SP cells are multipotent because they can generate all hormonal lineages of the anterior pituitary. (d) The non-Sca1high subset concentrates the cells that express transcription and signaling factors with a critical as well as primary role in early embryogenesis of the gland (see later). Increasingly, it is recognized that stem/progenitor cells of adult tissues recapitulate the embryonic developmental program en route to differentiated tissue cells [6, 7, 35, 36]. (e) The non-Sca1high SP cells, persisting in serum-free culture, are regulated by growth factors that affect stem/progenitor cells in other tissues [44–46]. These growth factors are also produced within the pituitary [4, 5]. In addition, bFGF plays a role in maintaining the embryonic progenitor pool . (f) The higher SP proportion with enlarged non-Sca1high fraction, and the correspondingly increased incidence of Sox2-expressing cells during the first early-postnatal wave of pituitary growth, further supports a stem/progenitor cell role. Recently, also two other groups provided evidence for stem/progenitor cells in the pituitary gland [23, 29]. In line with our findings, Sox2 and Sox9 were advanced as markers, and cells in the marginal layer bordering the cleft were endowed with stem/progenitor cell properties.
In clear contrast to the non-Sca1high subset, the SP fraction expressing high levels of Sca1 largely enriches for cells with endothelial phenotype that agrees with findings in other organs in which this was examined [9, 10]. In many tissues, it has been shown that the SP contains endothelial-type cells, either concluded from transcriptional profiling (like in mammary gland; ) or from analysis by FACS (like in bone marrow, lung, skeletal muscle, and heart [9, 10, 12, 14, 15]). At least for skeletal muscle and heart [9, 10], it has been demonstrated that the nonendothelial SP fraction clusters the tissue-specific stem/progenitor cells, which in essence corresponds to our findings. In view of the high expression of the endothelial progenitor marker CD14  (Table 1) and of general stem cell-associated markers Oct4, Bmi1, and nestin [6, 24, 27] in the Sca1high SP, we further hypothesize that this SP fraction does not only collect mature EC but also (or mainly?) endothelial progenitors, likely involved in capillary network remodeling during pituitary plasticity. A similar hypothesis has been formulated for the SP of mammary gland , and the idea is further supported by recent detection of Sca1 on endothelial progenitor cells [50–52]. As also suggested by others, the SP as a whole appears to enclose multiple stem/progenitor cell types (endothelial, mesenchymal, tissue-specific) that are present in the tissue [4, 5, 22].
Expression of Sca1 is observed in the SP of many, if not all, tissues [9–17, 19–21]. Surprisingly, Sca1-based dissection of the SP, followed by functional stem/progenitor cell assays of the fractions obtained, has so far only been reported for bone marrow, in which stem cells were found to be most enriched in the Sca1+ fraction of the SP (isolated in combination with other markers) [13, 14]. However, whether Sca1 is a marker of the tissue-specific stem/progenitor cells appears to differ between organs. Spermatogonial stem cells are enriched in a small testis cell fraction that is Sca1negative, apparently not coinciding with the SP . Unfortunately, the SP holding a vast proportion (90%) of Sca1high cells was not subfractionated, and stem cell properties of the non-Sca1high SP were not explored. Along the same line, muscle progenitor (satellite) cells do not express Sca1 . The muscle SP, of which a small fraction is capable of generating satellite cells and myocytes after intramuscular injection, contains a large proportion (90%) of Sca1+ cells but was also not dissected for Sca1 level . In contrast, the Sca1+ population (20%) from mouse mammary gland primary cultures enriches for the mammary gland progenitor cells and is suggested to partially overlap with the SP (2%–3%) that contains a large proportion (75%) of Sca1+ cells as well as a fraction (20%) of CD31+ cells [18–20]. However, how these markers are related in the mammary gland SP and whether the SP subpopulation that gives rise to mammary gland outgrowths belongs to the epithelial and/or endothelial lineage (as also raised in ) has, to our knowledge, not been clarified. In hepatic stem/progenitor cell cultures, SP (2%–4%) and Sca1+ (15%–40%) phenotype appear to diverge, and SP cells have more pronounced stem cell characteristics than did Sca1+ cells . In conclusion, our study in essence demonstrates that subfractionation of the SP can provide additional information as to the phenotype of the tissue's stem/progenitor cells and more accurately designate these cells. In addition, it might clarify some apparent inconsistencies between studies, in particular, regarding the relationship between stem/progenitor cell, SP, Sca1, and EC phenotypes. It is obvious that still more refined dissection of the pituitary non-Sca1high SP (e.g., using the cell membrane markers CD44, CXCR4, CD133) may further narrow down the pituitary stem/progenitor cell population.
The pituitary primordium, Rathke's pouch, initially consists of an epithelial layer of uniformly looking cells. Under the coordinated action of growth factors and morphogens, cells proliferate to enlarge the pouch and start to adopt different fates in a defined spatio-temporal order [1–3]. Cells around the lumen of Rathke's pouch (the cleft) continue to proliferate as undifferentiated progenitor cells during embryogenesis. Their progeny exits the cell cycle and starts the differentiation program toward hormone-producing cells as they move to, or become integrated in, the developing anterior lobe. Pituitary specification, as well as maintenance, expansion and fate determination of the embryonic progenitor cells are dictated by transcriptional regulators among which Hesx1, Pitx1, Pitx2, Lhx3, Lhx4, Isl1, Prop1, Pax6, Six1, Six6, Msx1, and Otx2 play critical roles [1–3, 37]. Expression of all these factors in the non-Sca1high SP suggests that an embryonic-like developmental program is executed within this cell compartment of the adult gland. The detection of multiple factors likely reflects their inductive, repressive, and/or cooperative interactions as unfolded during pituitary embryogenesis [1–3, 37], and the presence of progenitor cells at various stages of early development toward a differentiated hormonal cell. Postnatal functions for Hesx1, Prop1, Lhx4, and Pax6 which are all downregulated after early embryogenesis, as well as for Six6 and Otx2 [1–3, 37], constitute a provocative idea that awaits further clarification. Our data suggest a task in postnatal stem/progenitor cell regulation and differentiation. In support, we have found that the non-Sca1high fraction within the SP is enlarged when Prop1 is transgenically expressed in the pituitary in a persistent manner  (H. Vankelecom, unpublished observations). Unfortunately, these promising results could not be further elaborated because of breeding deficiencies of the mice.
Also the Notch signaling pathway plays an essential role during early-embryonic development of the pituitary [38–42]. Of the Notch receptors, only Notch2 and Notch3 are detected in the embryonic gland. Remarkably, these particular receptors show highest expression in the non-Sca1high SP. During embryogenesis, Notch2 is found in the periluminal progenitor zone at the juncture between the highly proliferating and the differentiating cells. Transgenic studies have shown that Notch2 expression is essential to keep the embryonic progenitor cells in an undifferentiated state after cell cycle exit and that downregulation is a prerequisite to initialize transition to differentiation [38, 39]. Notch2 may play a similar role in the progenitor (non-Sca1high SP) compartment of the adult gland. Of the Notch downstream genes, the Hey family (Hey1, Hey2, and Heyl) predominates in the non-Sca1high SP (Table 1). Hey1 transcripts are observed in a pattern similar to Notch2 during the earliest phase of pituitary development . Both Hey1 and Hey2 have been shown to maintain stem/progenitor cells in other tissues , but to date no role has been reported in the embryonic pituitary. Also Hes1 is expressed in the developing gland, but most studies so far did not confer an essential role, which may, at least partly, be due to redundant activity of other Notch effector genes [41, 42]. We found upregulated Hes1 expression in the non-Sca1high SP when compared with the MP, but still more elevated levels in the Sca1high subset, which may be in line with the absence of an embryonic role in the forefront, and may rather suggest a more prominent function in the regulation of the endothelial compartment (including endothelial progenitors) in the pituitary. The latter idea may also apply to Notch1 and Notch4, prominently expressed in the pituitary Sca1high SP and particularly implicated in angiogenesis as reported for other systems . In plain contrast to the progressively elucidated role of Notch signaling in pituitary organogenesis [38–42], almost nothing is known regarding its function in the postnatal pituitary. Our data highlight at least one possible task, i.e., the regulation of the stem/progenitor cell pool, already hinted by our previous findings that Notch signaling affects SP size [6, 7]. Moreover, as found in the embryonic gland [41, 42], Notch signaling may affect expression of the critical transcriptional regulators described earlier. In addition, Sox2 and Sox9 have been shown to maintain certain stem/progenitor cells in an undifferentiated state by inducing the expression of Notch effectors . Whether such interactions also occur in the stem/progenitor cells of the postnatal pituitary remains to be established.
The topographical map of Sox2 expression in the anterior pituitary promotes some interesting ideas. Cells with nuclear Sox2 protein inhabit the cell layer lining the cleft (this study and [23, 29]), a frequently postulated pituitary stem cell niche (reviewed in [4, 5]), but in addition constitute small groups scattered throughout the parenchyma of the anterior lobe and often surrounding a small lumen. These cell clusters may represent Rathke's pouch remnants (cysts) or invaginations of the marginal cell layer that are “pinched off,” or even still connected. Their configuration is clearly different from follicles lined by folliculo-stellate cells as described earlier . Remarkably, Sox2 protein is also found in a cytoplasmic pattern in some cells, often residing in the vicinity of the nuclear Sox2+ cells. Together, our data propose the model that, on top of the extended marginal stem/progenitor cell niche, additional small niches are disseminated over the gland to allow a swift reaction to hormonal requests of the body. As stem/progenitor cells move into differentiation, Sox2 protein becomes restricted from entry into the nucleus to extinguish its activity as gatekeeper of multipotency [26, 27]. Of note, although cytoplasmic Sox2 was not observed together with GH and ACTH, it colocalized with PRL and αGSU in some cells, and a small number of PRL+ and αGSU+ cells were also found in freshly sorted non-Sca1high SP. It is not unthinkable that in female mice as used here, lactotropes and gonadotropes are the cell types that are predominantly renewed, a suggestion also made by others . Moreover, higher expression of the cell proliferation marker Ki67 in the non-Sca1high SP than the MP (2.9-fold, p < .05) points to a more active cell division in the progenitor than the differentiated cell compartment, at least under the basal (low-turnover) conditions examined. These findings raise the hypothesis that homeostatic renewal of pituitary endocrine cells is not, or at least not only, by self-duplication of mature hormonal cells (reviewed in [4, 5]), but also through division and differentiation of progenitor cells. However, direct demonstration requires further extensive studies.
Finally, we here present a serum-free culture method to enrich for the non-Sca1high cells within the SP, expected to be valuable in deciphering the regulatory network and genetic programs of the pituitary stem/progenitor cells. As an initial impetus, we show that LIF, bFGF, and EGF affect the expression of early-embryonic factors, each presumably with particular accents. Interestingly, Hesx1 goes up and Prop1 down under the influence of LIF, thereby mirroring the repressive interplay of both factors during pituitary embryogenesis [1–3]. Further in-depth exploration will start from the transcriptomal profiles and, in particular, from the expression data on the signaling pathways of Notch, Wnt, MAPK, TGFβ, bone morphogenetic protein, and Jak/Stat, enriched in many mammalian stem cell populations [6, 7, 35, 36]. In previous work, we already demonstrated activation of Notch signaling under the influence of LIF, bFGF, and EGF . This culture method provides a valuable alternative to pituispheres that do not display cell–cell contacts with other nonstem (including hormonal and mesenchymal) cells as likely present in the pituitary niches in situ. In addition, as of today, serial pituisphere passaging is limited to the second/third generation (this study and ), possibly—or at least partly—due to technical obstacles of sphere picking and dispersion yielding only restricted numbers of viable cells from primary spheres. Consequently, demonstration of self-renewal capacity of the pituisphere-initiating cells awaits further investigation and improvement, if possible, of the pituisphere assay.
We here demonstrate that the non-Sca1high SP phenotype identifies the progenitor cells of the postnatal pituitary. Moreover, our study is the first to provide a method to isolate (living) progenitor cells from the pituitary gland, valuable for their phenotypical, transcriptional, and functional characterization in various normal and pathological conditions. In-depth exploration of the stem/progenitor cells is expected to considerably increase our knowledge on pituitary (postnatal) development and plasticity, and on the etiology of pituitary dysfunctions, many aspects of which remain unknown.
This work was supported by grants from the Fund for Scientific Research (F.W.O.)-Flanders (Belgium), the Research Fund of the K.U.Leuven, the Flemish Ministry of Science Policy (Belgium), and the National Natural Science Foundation of China (Project 30500248). J.C. obtained an IRO Fellowship and Q.F. a SBA Scholarship from the K.U.Leuven, L.G. a Ph.D. Fellowship from the F.W.O. The authors are very grateful to Vik Van Duppen (Hematology, K.U.Leuven) for his invaluable help with fluorescence-activated cell sorting, and to the Microarray Core Facility (VIB, Leuven). The authors also thank Yvonne Van Goethem and Kristine Rillaerts for their technical help.
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest.