Pituitary tumors contain a side population with tumor stem cell-associated characteristics

Pituitary adenomas

Human pituitary adenoma samples were obtained immediately after transsphenoidal resection at the University Hospitals Leuven (Neurosurgery Division). Informed consent was obtained from the patients. Hormonal phenotype was determined by the Department of Imaging and Pathology (University Hospitals Leuven). An overview of the samples with clinical data is provided in Table 1. Sexes were represented approximately equally with an average age of 53 years (range 19–84 years) at the time of surgery. The majority of the tumors were NF-A (n=34) and GH-A (n=21). The other types of resected pituitary tumors were typically microadenomas, which only yielded sufficient study material in a limited number of cases (ACTH-A (n=3)) or were not surgically removed but were instead treated pharmacologically (resected PRL-A (n=2)).

Tumor samples were either immediately processed for flow cytometric analysis after resection, or they were cryopreserved as small pieces in Leibovitz-L-15 medium (Life Technologies) supplemented with 10% fetal bovine serum (FBS; Lonza, Rockland, ME, USA) and 10% DMSO (Sigma–Aldrich). They were then stored at −196 °C for later examination. The freezing procedure did not change the SP proportion (Supplementary Fig. S1, see section on supplementary data given at the end of this article) or the expression characteristics, as was evident from the results of the microarray analyses.

Dissociation of human pituitary adenomas into single cells

Tumor samples were rinsed in DMEM (Life Technologies) supplemented with 0.3% BSA (Serva, Heidelberg, Germany) and incubated with collagenase type IV (1 mg/ml in Medium 199; Life Technologies) for 2–2.5 h at 37 °C. Cell debris was removed by centrifugation through a 3% BSA layer. The resulting cell pellet was resuspended in pituitary-optimized, serum-free chemically defined medium (SFDM, Life Technologies; supplemented with 0.5% BSA; Chen et al. 2005, 2009) for further analyses.

Immunohistology of human pituitary adenoma sections

Formalin-fixed paraffin-embedded pituitary adenoma blocks were obtained from the Biobank (University Hospitals Leuven). Antigen retrieval was performed on 5 μm sections with EnVision FLEX Target Retrieval Solution (Dako, Glostrup, Denmark). Endogenous peroxidase activity was blocked with EnVision Peroxidase-Blocking Reagent, and sections were incubated with primary antibodies directed against human E-cadherin (ready-to-use; Dako), vimentin (VIM; 1:500; Dako), NOTCH2 (1:200; Abcam, Cambridge, UK), and interleukin 6 (IL6; 1:200; Abcam). Subsequently, samples were processed using EnVision Dual Link (Dako). The complex formed was visualized with 3,3′-diaminobenzidine (Dako), and sections were counterstained with hematoxylin (Dako).

Isolation and dissociation of mouse pituitaries

C57/Bl6 mice were purchased from Elevage Janvier (BioServices, Uden, The Netherlands), and dopamine receptor D2 knockout (Drd2^−/−) mice were purchased from The Jackson Laboratory (Bar Harbor, ME, USA). Mice were kept or bred in the KU Leuven Animal Facility under conditions of constant temperature, humidity, and day–night cycle in a sterile environment. The mice were allowed to consume food and water ad libitum. Animal experiments were approved by the KU Leuven Ethical Committee.

The pituitary was isolated from killed mice as previously described (Chen et al. 2005, 2009). The anterior lobes (anterior pituitary (AP)) of the Drd2^−/− mice that grow prolactinomas (from 6 to 14 months of age; Cristina et al. 2006) and those of the respective control mice (Drd2^+/− and Drd2^+/+) were dispersed into single cells using collagenase type IV as explained in the previous sections for human pituitary adenomas.

Culture and treatment of AtT20

The mouse corticotrope AtT20 cell line was obtained from the American Type Culture Collection (ATCC, CCL-89; Manassas, VA, USA). The cell line was authenticated by hormone expression analysis with quantitative RT-PCR (qRT-PCR; see the subsequent section on qRT-PCR). The cells were cultured in DMEM/F12 (Life Technologies) supplemented with 10% FBS. To investigate the involvement of the transforming growth factor beta (TGFβ) or Cxcr4 pathways, AtT20 cells were cultured in serum-free medium for 18–24 h and were then treated with TGFβ1 (5 ng/ml; R&D Systems, Minneapolis, MN, USA) or AMD3100 (octahydrochloride, 100 ng/ml; Sigma–Aldrich) respectively. Medium was refreshed every 24 h. To analyze gene expression and SP proportion after treatment (see the subsequent sections on these methods), cells were detached and dissociated using trypsin (0.05%) with EDTA (0.02%; Life Technologies).

Flow cytometry

Dissociated cells (from human adenoma or mouse AP, AtT20 cell line, or xenograft tumors – see subsequent sections) were processed for SP analysis as described in detail previously (Chen et al. 2005, 2009). In short, cells were incubated with the vital dye Hoechst 33342 (Sigma–Aldrich) at a final concentration of 2.5 μg/ml (for mouse AP and AtT20) or 5 μg/ml (for adenoma and xenograft samples). Subsequently, cells were either immediately analyzed by FACS (Vantage or Aria III; BD Biosciences, Erembodegem, Belgium) or they were first stained with phycoerythrin (PE)-conjugated mouse anti-human CD45, FITC-conjugated mouse anti-human CD31, PE-conjugated mouse anti-mouse CD45, and/or FITC-conjugated rat anti-mouse CD31 antibodies using dilutions recommended by the manufacturer (BD Biosciences). Samples were examined for SP on the basis of dual-wavelength emission using appropriate filters (Vantage: 424/44 and 630/22; Aria III: 450/40 and 630/20) after (near-)u.v. excitation (360 and 375 nm on Vantage and Aria III respectively). The SP phenotype was verified by adding verapamil (100 μM; Sigma–Aldrich), which blocks the Hoechst efflux. The bulk cell population that did not expel Hoechst was designated as the ‘main population’ (MP). CD31 and CD45 staining was controlled using isotype antibodies (BD Biosciences; data not shown).

Xenograft tumors and in vivo tumorigenic activity

In an attempt to grow human pituitary adenoma in immunodeficient mice, tumor pieces or dissociated cells (1–1.5×10⁶) were transplanted s.c. (GH-A and NF-A (n=8)) or under the kidney capsule (GH-A and ACTH-A (n=3)). Male mice (6–10 weeks old) with increasing grades of immunodeficiency (from SCID to NOD–SCID to NOD–SCID^{IL2Rγ−/−}) were purchased from Elevage Janvier and were anesthetized with Avertin (tribromoethanol; Sigma–Aldrich). Before injection, cells were resuspended in DMEM mixed with Matrigel (1:1, BD Biosciences). Three to six months later, the mice were killed and examined for tumor growth. Tissue from the implantation sites was dissected, and hematoxylin and eosin staining was performed.

To grow xenograft tumors from AtT20 cells, SCID mice (males; 6–10 weeks old) received s.c. injections of 5×10⁵ cells. To assess the tumorigenic (TSC) activity of the AtT20 SP versus the MP, cell populations were sorted into DMEM/F12/10% FBS, mixed with Matrigel (1:1) and s.c. injected at different cell numbers into SCID mice (SP cells into one flank and a similar number of MP cells into the other flank of the same mouse). Tumor size was measured using a caliper, and volume was calculated with the following formula: ((large side)×(small side)²)×0.52.

To explore the involvement of the Cxcr4 pathway, AtT20 xenograft tumors that developed in SCID mice were treated with AMD3100 by intratumoral injection (5 μg, two times per week) starting from day 21 after the s.c. AtT20 cell implantation.

Tumorsphere- and colony-forming assay

Dissociated human pituitary adenoma cells were seeded at a density of 100 000 cells/ml into 35 mm nontreated culture dishes (Iwaki, Scitech, Chiba, Japan) and cultured in SFDM supplemented with B27 (1:50; Life Technologies) and recombinant basic fibroblast growth factor (bFGF, 20 ng/ml; R&D Systems). Fresh medium was added every 3 days. To assess their self-renewal capacity, the spheres were harvested and dissociated into single cells using trypsin (Chen et al. 2009), and the cells were reseeded under similar conditions.

To evaluate their colony-forming capacity, cells were cultured at low density (1000–4000 cells/60 mm dish) in Ultraculture medium (Life Technologies) to which 5% FBS, 20 ng/ml bFGF, and 50 ng/ml cholera toxin (Sigma–Aldrich) were added. Medium was changed every 3 days. Cultures were fixed with methanol and stained with 1,9-dimethyl-methylene blue (Sigma–Aldrich), and the number of colonies was determined. Live pictures of the formed spheres and colonies were taken using a Nikon Eclipse TS100 microscope (Nikon Instruments, Melville, NY, USA).

Cell motility assay

Cell motility was evaluated using the ‘scratch assay’. After reaching 90% confluence, AtT20 cells were serum-starved for 24 h and then treated with mitomycin C (10 μg/ml; Kyowa, Tokyo, Japan) to inhibit cell proliferation. A straight scratch was created, and cells were further kept in DMEM/F12, together with TGFβ1 (5 ng/ml), AMD3100 (100 ng/ml), or vehicle. Medium was changed every 24 h. The migration of cells into the scratch was evaluated by light microscopy, and live pictures were taken with the Nikon Eclipse TS100 microscope at different time points. The open area was calculated using analysis-algorithm-based TScratch Software (CSElab–ETH, Zurich, Switzerland).

Small-animal magnetic resonance imaging

Magnetic resonance imaging (MRI) was performed on Drd2^−/− and Drd2^+/− mice using a horizontal 9.4-Tesla nuclear magnetic resonance spectrometer and a 40 mm-diameter birdcage coil as the radio frequency transmitter and receiver. Mice were anesthetized using 1% isoflurane in a mixture of 75% air and 25% oxygen. T1-weighted images were acquired using a two-dimensional multiple-slice spin echo sequence, with an echo time of 11 ms, a repetition time of 500 ms, an in-plane resolution of 0.1×0.1 mm, 24 coronal slices with 0.3 mm slice thickness, and ten signal averages. Image reconstruction was performed using the spectrometer console. Ten minutes before the scanning procedure, mice received an i.p. injection of the contrast agent gadolinium (0.50 mmol/kg body weight).

Immunofluorescence analysis

For the immunofluorescence examination of mouse pituitaries, whole glands were isolated and fixed with paraformaldehyde (4%; Riedel-de Haën, Seelze, Germany). After agarose embedding, pituitaries were coronally sectioned into 45 μm slices using a vibratome (Microm HM 650 V; Prosan, Merelbeke, Belgium). Sections were permeabilized with 0.4% Triton X-100 (Sigma–Aldrich) and incubated overnight with the following primary antibodies: goat anti-human SOX2 (1:750; Immune Systems, Devon, UK) and rabbit anti-mouse PRL (1:10 000; from Dr A F Parlow, NHPP, Harbor–UCLA Medical Center, Torrance, CA, USA). Finally, sections were incubated with AlexaFluor 488-conjugated donkey anti-rabbit and/or AlexaFluor 555-conjugated donkey anti-goat antibodies (1:1000; Life Technologies), and nuclei were labeled with ToPro3 (Life Technologies). Immunofluorescence was examined with a Zeiss LSM 510 confocal laser-scanning microscope (Zeiss, Zaventem, Belgium) accessible through the Cell Imaging Core (CIC; KU Leuven). Z-stacks were collected and analyzed with the Zeiss LSM Image Browser and ImageJ (http://imagej.nih.gov/ij/), and pictures were prepared using the same programs and Microsoft PowerPoint (2007).

Immunofluorescence staining of paraffin-embedded mouse pituitary sections (5 μm) was performed after antigen retrieval with citrate buffer (pH 6) and permeabilization with Triton X-100 (0.1% in PBS). Following incubation with primary and secondary antibodies (as described earlier), sections were covered with Vectashield (containing 4′,6-diamidino-2-phenylindole (DAPI)) and analyzed with a Leica DM5500 epifluorescence microscope (Leica Microsystems, Diegem, Belgium) accessible through InfraMouse (KU Leuven–VIB). Recorded images were converted to pictures with ImageJ and Microsoft PowerPoint (2007).

Cytospin samples of dissociated mouse AP cells were fixed with paraformaldehyde (4%), permeabilized with saponin (0.5%; Sigma–Aldrich), and further immunostained for Sox2 (with goat anti-human SOX2 at 1:250; Immune Systems) as described in the previous sections. In some of the analyses, cells were simultaneously immunostained for the proliferation marker Ki67 (with rabbit anti-Ki67 at 1:50; Thermo Scientific, Fremont, CA, USA). Nuclei were counterstained with DAPI (0.5 μg/ml). To enumerate Sox2-immunopositive cells, pictures were captured using a Leica DM5500 microscope, and cells were counted with ImageJ in five to ten random fields per group in each experiment (about 1500 cells/group in each experiment). The proportions of the Sox2⁺ cells (of total cells counted) were determined. Because of the increase in total AP cells in the Drd2^−/− adenomatous pituitary, comparison of the proportions is not appropriate. Thus, the absolute numbers of Sox2⁺ cells were calculated (using their proportion and the total number of AP cells obtained per experimental group of mice) for comparisons between Drd2^−/−, Drd2^+/−, and Drd2^+/+ mice.

For the immunofluorescence examination of tumorspheres developed from human pituitary adenomas, cells were fixed with paraformaldehyde (4%), permeabilized with Triton X-100 (0.4% in PBS), and incubated with goat anti-human SOX2 (1:250; Immune Systems), rat anti-mouse nestin (1:500; Abcam), and/or rabbit anti-human GH (for spheres originating from GH-A; 1:2000; Dako). Secondary antibodies included AlexaFluor 555-conjugated donkey anti-goat, AlexaFluor 488-conjugated donkey anti-rat, and AlexaFluor 488-conjugated donkey anti-rabbit (all at 1:1000; Life Technologies). Pictures were obtained using a Leica DM5500 epifluorescence microscope as described in the previous sections.

AtT20 cell death, apoptosis, and proliferation

The number of dying or dead AtT20 cells after AMD3100 treatment (100 ng/ml) was determined after staining with Trypan blue (0.4%; Sigma–Aldrich). Apoptosis was scored by TUNEL analysis in cytospin samples of AtT20 cells from culture or xenograft tumors using the FragEL Kit (Calbiochem, Darmstadt, Germany). Cell proliferation was quantified by the immunofluorescence staining of cytospin samples for Ki67 as described earlier in the present study. Pictures were taken and cells enumerated (see previous sections).

Whole-genome expression profiling

Total SP and MP cells (NF-A (n=5) and GH-A (n=4)) or CD31⁻/CD45⁻ SP (purified SP (pSP)) and pMP cells (NF-A (n=5)) were sorted by FACS into cold lysis solution of the RNeasy Micro Kit (Qiagen). Total RNA was extracted immediately after cell collection in accordance with the manufacturer's guidelines. RNA samples were stored at −80 °C until further processing for whole-genome expression profiling (in collaboration with the VIB Nucleomics Core, KU Leuven). RNA quality and concentration were determined with Agilent Picochips on an Agilent BioAnalyzer 2100 (Agilent Technologies, Santa Clara, CA, USA). Two nanograms of high-quality RNA (with RNA integrity number ≥8.0) was amplified using the NuGen Pico WTA Kit (NuGen Technologies, San Carlos, CA, USA). Cy3 label was incorporated into the cRNA probes, which were then hybridized onto Agilent whole-human genome 44K oligonucleotide arrays. Slides were scanned and data were processed using Agilent's Feature Extraction Software for background correction, quantile normalization, principal component analysis (PCA), hierarchical clustering, and MA and volcano plots. Microarray data are available from the NCBI's Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/projects/geo/) through the series accession number GSE62960.

To determine differentially expressed genes (probes), a stringent analysis was initially applied to the normalized data sets. Taking into account that the data are paired (SP versus MP for each sample), analysis was based on the log₂ ratios computed for each pair. Log₂ ratios were compared with the Linear Models for Microarray Data (Limma) package of Bioconductor in R, and a moderated t-statistic (implemented in Limma) was performed to test the significance of the obtained differences. For the first microarray data set (total SP versus MP), the resulting P values were corrected for multiple testing with Benjamini–Hochberg to control the false discovery rate. Differentially expressed genes were withheld when the corrected P value was lower than 0.05. For the second microarray data set (CD31⁻/CD45⁻ SP versus CD31⁻/CD45⁻ MP), Benjamin–Hochberg could not be applied because of the low number of samples. Therefore, a cutoff based on the uncorrected P values (<0.001) was utilized. For both data sets, the cutoff for the P values was combined with a cutoff for the fold change (≥2) to finally identify the differentially expressed genes.

Given the rather small number of adenomas analyzed thus far, which causes a very high cutoff for differential expression and hence the discarding of potentially interesting differences, we additionally applied a less stringent analysis by performing paired Student's t-tests on the individual probe sets. Genes were judged to be differentially expressed when P<0.05 and fold change ≥1.5. We further considered genes that showed a significant expression difference in five to eight out of the nine adenomas analyzed for total SP and MP, and in two to four out of the five adenomas examined for CD31⁻/CD45⁻ fractions.

Functional clustering analysis of differentially expressed genes (gene ontology (GO)) was performed using the Database for Annotation, Visualization and Integrated Discovery version 6.7 (DAVID, www.david.abcc.ncifcrf.gov). Gene clusters with an enrichment score of >1.5 (−log of the median P value for the functional group) were retained. Finally, we narrowed the gene sets by focusing on factors that have been shown to be important in CSC/‘tumor stemness’, including EMT, NOTCH, WNT/β-CATENIN, and TGFβ/BMP.

Gene expression analysis by qRT-PCR

Amplified cDNA of the CD31⁻/CD45⁻ SP and MP fractions from the five microarrayed NF-A samples, as well as that from two additional samples (NF-A (n=1) and GH-A (n=1); see Table 1), was examined by qRT-PCR to validate the expression of a selection of genes with Fast SYBR Green Master Mix and an ABI 7900HT Fast RT-PCR system according to the manufacturer's instructions (Life Technologies).

Gene expression analysis of AtT20 cells and dispersed xenograft tumors was performed using a similar approach (but without the RNA amplification step). cDNA was subjected to qPCR in a LightCycler 480 with the LightCycler FastStart DNA Master Plus SYBR Green I (Roche).

Forward and reverse primers were designed using PerLPrimer (http://perlprimer.sourceforge.net) or were obtained from a public database (http://medgen.ugent.be/rtprimerdb), and they are described in Supplementary Table S1, see section on supplementary data given at the end of this article. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH), β2-microglobulin (B2M), and/or hypoxanthine phosphoribosyltransferase 1 (HPRT1) were used to normalize gene expression. Their expression was found to be stable (data not shown). Relative gene expression levels were calculated as change in the threshold cycle (ΔCt) values (‘Ct target−Ct reference’) and were compared between the SP and MP as fold change (2^−ΔCt SP/2^−ΔCt MP). Statistical analysis was performed using GraphPad Prism version 5.2 (GraphPad Software, La Jolla, CA, USA) by applying Mann–Whitney U tests. P values of <0.05 were considered statistically significant.

For authentication of the AtT20 cell line used, hormone gene expression was examined by qRT-PCR. The corticotrope phenotype of the cell line was confirmed by the detection of proopiomelanocortin (POMC; a precursor of ACTH) expression, whereas GH and TSHβ gene transcription was not found (see Supplementary Table S2, see section on supplementary data given at the end of this article).

Human pituitary tumors contain a SP

Human pituitary adenomas (n=60; see Table 1) were examined for the presence of a SP. In all of the tumors analyzed, a Hoechst^low cell population was detected that was no longer observed when the efflux-blocker verapamil was added together with Hoechst (Fig. 1A), thereby demonstrating the SP phenotype (mean SP±s.e.m. 1.9±0.3% of the total viable tumor cells; range 0.4–17.2%; Table 1).

View Full Size

Identification and characterization of the side population (SP) in human pituitary adenomas. (A) Dot plots of dual-wavelength FACS analysis of human pituitary adenoma cells after incubation with Hoechst 33342 alone (left) or together with verapamil (right). The SP (black) and MP (white) are gated, and SP proportions are indicated. A representative example is shown. (B) Boxplot of SP proportions obtained from NF-A (n=33) and GH-A (n=20) as well as from all samples taken together (n=60). Minimum and maximum values (whiskers) and outliers (dots) are indicated. (C) Dot plots of FACS analysis of human pituitary adenoma cells after incubation with Hoechst 33342 and immunostaining for CD31 and CD45. The CD31⁺, CD45⁺, and CD31⁻/CD45⁻ cells within the SP are also shown (inset). (D) Percentage of CD31⁺ (upper panel) and CD45⁺ cells (lower panel) within the SP versus the SP proportion within the pituitary adenomas analyzed (n=37 and n=31 respectively). (E) Expression analysis by quantitative RT-PCR of a selection of markers identified as differentially expressed in the pituitary adenoma CD31⁻/CD45⁻ SP (pSP) versus the pMP by microarray analysis. Bars indicate means of fold expression pSP/pMP±s.e.m. (n=7 pituitary adenomas). *P<0.05.

Citation: Endocrine-Related Cancer 22, 4; 10.1530/ERC-14-0546

• Download Figure
• Download figure as PowerPoint slide

SP proportions were not different between the samples analyzed immediately after surgical resection (n=26) or those analyzed after cryopreservation (n=34) (Table 1 and Supplementary Fig. S1A, see section on supplementary data given at the end of this article) or between the different hormonal phenotypes examined (Table 1 and Fig. 1B). The adenoma SP predominantly clusters cells of small to medium size (low ‘forward scatter’) and low granularity (low ‘side scatter’) (Supplementary Fig. S1B).

Whole-genome expression profiling of the adenoma SP points to EMT and ‘tumor stemness’

SP cells and the tumor bulk MP cells from somatotropinomas (GH-A (n=4)) and NF-A (n=5) (see Table 1) were sorted by FACS and subjected to gene expression profiling using whole-genome microarrays. PCA indicated a similar segregation of the SPs from the corresponding MPs for the majority of the samples analyzed (Supplementary Fig. S1C). Hierarchical clustering of differentially expressed genes also largely separated the SPs from the MPs (Supplementary Fig. S1D).

The set of genes that was most differentially expressed (P<0.001; 2760 probe sets) was submitted to DAVID for GO analysis, which yielded a total of 1688 corresponding human ENSEMBL IDs. Gene set enrichment exposed in the SP the upregulation of biological functions that are involved in cell proliferation, vascular development, cell migration and motility, and lymphocyte activation (Supplementary Table S3, see section on supplementary data given at the end of this article).

A focused analysis of a selection of genes using the expression data of the nine adenomas (Table 2 and Supplementary Table S4, see section on supplementary data given at the end of this article) revealed upregulated expression (i.e., >1.5-fold in at least two-thirds, or in five, of the nine tumors, P<0.05; see the ‘Materials and methods’ section) in the SP of the multidrug transporters ABCB1 and ABCG2, which typically underlie the SP phenotype. This finding therefore validates the SP identification and sorting procedure. Remarkably, the transcriptome displayed an expression pattern that supports the occurrence of EMT in the SP compartment. The epithelial markers E-cadherin (CDH1) and claudin-1 (CLDN1) were downregulated in the pituitary adenoma SP, whereas the mesenchymal markers VIM and fibronectin 1 (FN1) were upregulated (Table 2 and Supplementary Table S4). Moreover, recognized markers of EMT, such as SNAI1, ZEB1, and ZEB2 (all of which are transcriptional repressors of the CDH1 gene), were more highly expressed in the SP vs the MP, together with other inductors and regulators of EMT (KLF8, LEF1, TCF4, FOXC1, HOXB9, FOSL2, MMP1/2, and CXCR4). Some of the pathways that were identified as being upregulated are also well known for their regulatory input in EMT (WNT, TGFβ, epiregulin, and amphiregulin). In the present analysis, we also observed that factors often associated with ‘tumor stemness’ or CSCs, including CD44, CXCR4, KIT, KLF4, and NESTIN were up-regulated in the SP. In addition, the signaling pathways of NOTCH, WNT, and TGFβ–BMP have also been related to ‘tumor stemness’. Overall, no obvious differences were observed between the SP transcriptomes of the NF-A and the GH-A samples.

Notably, the microarray analysis also revealed the upregulated expression of genes associated with an endothelial phenotype (such as PECAM1/CD31, VWF, CDH5, VCAM1, FLK1/KDR, TIE1, and TEK/TIE2) or an immune character (such as CD45/PTPRC) in the adenoma SP (see Table 2 and Supplementary Table S4). These expression characteristics are consistent with the enriched biological processes described earlier (viz., vascular development and lymphocyte activation; see Supplementary Table S3), and they indicate the presence of endothelial and immune/hematopoietic cells in the SP. Previously, we and others discovered that the SP, either from healthy tissue or tumors, represents a still heterogeneous population that segregates not only cells with (tumor) stem cell properties but also some other cells of endothelial and immune/hematopoietic phenotype (Pfister et al. 2005, Uezumi et al. 2006, Chen et al. 2009).

Taken together, our microarray expression data indicate that, as compared with the MP, the pituitary adenoma SP is enriched in cells with EMT- and ‘tumor stemness’-associated molecular traits, but it is still heterogeneous in cell phenotypes.

EMT- and ‘tumor stemness’-related expression characteristics remain present in the adenoma SP depleted from endothelial and immune cells

The pituitary adenoma SP was then analyzed for the presence of endothelial (CD31⁺) and immune/hematopoietic (CD45⁺) cells. Flow cytometry revealed a CD31⁺ cell population of varying abundance in the SP (mean±s.e.m. 50.9±4.8% of the SP; range 5.8–95.2%; n=37; Fig. 1C and D; Table 1). Also, CD45⁺ cells were found, but in general they constituted a lower fraction of the adenoma SP (mean±s.e.m. 16.2±3.0% of the SP; range 0.2–66.6%; n=31; Fig. 1C and D; Table 1). The non-endothelial, non-immune (CD31⁻/CD45⁻) SP fraction finally comprised on average one-third of the total SP (mean±s.e.m. 32.3±2.2%; range 7.2–64.1%), 0.5% (±0.1) of all the CD31⁻/CD45⁻ adenoma cells and 0.1% (±0.01) of the total tumor cell population. In general, no correlation was found between the size of the total SP and the proportion of CD31⁺ or CD45⁺ cells (Table 1 and Fig. 1D), although the two samples with the highest SP proportion analyzed for these antigens (PA 26, and PA 43) contained the largest fractions of CD31⁺ cells in the SP.

To determine the whole-genome expression profiles of the non-endothelial, non-immune cell populations, SP and MP were depleted from CD31⁺/CD45⁺ cells during flow cytometric sorting (see Fig. 1C for gating; NF-A (n=5); Table 1). Stringent analysis revealed that 101 genes (probe sets), which correspond to 74 human ENSEMBL IDs in DAVID, were differentially expressed between the CD31⁻/CD45⁻ SP (pSP) and the CD31⁻/CD45⁻ MP (pMP). Eighty of these were upregulated in the pSP, and 21 were downregulated (Supplementary Fig. S2A, B and C, see section on supplementary data given at the end of this article). The relatively small number of samples analyzed thus far caused a high cutoff for differential expression (as is clear in Supplementary Fig. S2A). Therefore, we performed an additional, less stringent examination (see the ‘Materials and methods’ section) by analyzing genes that were altered (≥1.5-fold, P<0.05) not only in all of the adenomas but also in four out of the five tumors profiled. A total of 2374 genes (probe sets) were found to be differentially expressed in the pSP as compared with the pMP in at least four of the adenoma samples (corresponding to 1576 human ENSEMBL IDs), with 1662 and 712 genes being upregulated and downregulated respectively (Supplementary Table S5, see section on supplementary data given at the end of this article). The gene sets obtained in both the stringent and less stringent analyses (101 and 2374 genes respectively) were submitted to GO examination using DAVID. Overviews of the most significantly enriched biological processes are provided in Supplementary Tables S6 and S7, and they include, among others, cell motility and migration, vascular development, cell proliferation, and organ development.

Subsequent analysis of the pSP was focused on a selection of genes as described earlier for the total SP, including genes that are associated with EMT and ‘tumor stemness’ (Table 3 and Supplementary Table S8, see section on supplementary data given at the end of this article). On the whole, the expression picture of the total SP appeared to be retrieved in the CD31⁻/CD45⁻ SP, viz. The upregulated expression of ABCG2 and of other CSC-associated markers, including CD44, CD34, LIFR, and CXCR4. KIT expression was upregulated in four of the five samples, and NESTIN was upregulated in three of the five (Table 3). In particular, the gene expression picture that underlies EMT was also present in the pSP, showing elevated expression of the mesenchymal markers VIM and FN1 and of several transcriptional EMT regulators (including SNAI1, SNAI2, ZEB1, and ZEB2; Table 3). Because GO enrichment indicated gene clusters involved in cell motility and migration, and because EMT is known to drive the motile and invasive behavior of cells, additional genes involved in these processes were examined. A considerable number were found to be upregulated or downregulated in the pSP versus the pMP, including EPHB1 and TM4SF1 (Supplementary Table S8). Interestingly, epidermal growth factor receptor (EGFR) was also highly (ninefold) upregulated in the pSP (Supplementary Table S8); EGFR has been found to be involved in tumor cell migration in craniopharyngiomas (Hölsken et al. 2011). In addition, several components that function in NOTCH, WNT, and TGFβ signaling were elevated (Table 3 and Supplementary Table S8). These pathways play a role in EMT and ‘tumor stemness’ but also in pituitary development (Zhu et al. 2007, Vankelecom 2010). In general, gene clusters associated with organ development and differentiation processes were found to be enriched in the pSP (see Supplementary Tables S4 and S5). Finally, genes that are implicated in angiogenesis were also upregulated in the pSP versus the pMP (Table 3 and Supplementary Tables S6, S7 and S8).

Differential expression as determined from the pSP microarray interrogations was verified by qRT-PCR for a number of the most interesting genes (Fig. 1E). For these analyses, the five microarrayed NF-A samples were complemented with two additional tumors (one NF-A and one GH-A; see Table 1). The results of qRT-PCR essentially confirmed the microarray data, although statistical significance was not always reached in the relatively limited number of samples analyzed thus far. All 17 genes examined showed upregulated expression in the pSP as compared with the pMP (nine with statistical significance, eight with an upward trend; Fig. 1E). Most remarkable were the elevated expression levels of CD44, IL6, NOTCH2, PDGFRB, SNAI2, VIM, and ZEB2 in the pSP over the pMP (>55-fold). A higher expression level (>55-fold) was also found for CXCR4, FN1, HES1, PDGFC, and TGFBR2, in at least five out of the seven samples. The expression of CDH1 was also checked but was not significantly different in the pSP versus the pMP, although a trend of downregulation was measured in the pSP of four out of the seven adenomas analyzed. The expression of E-cadherin, VIM, IL6, and NOTCH2 was further analyzed by immunohistochemistry in paraffin-embedded sections of pituitary adenomas that were used for the gene expression analysis described earlier in this paper (n=6; see Table 1 and Supplementary Fig. S2D). Tumors with high E-cadherin immunoreactivity contained some zones and cell clusters wherein E-cadherin was downregulated, and VIM was expressed in some regions or in small nests of tumor cells (in addition to the pericapillar cells), which together suggest that some of the tumor cells might have undergone EMT. IL6 immunoreactivity was observed in a couple of clustered cells (two to four per section) in three out of the six adenomas analyzed, and NOTCH2 immunoreactivity occurred in one to two cells per section in four out of the six adenomas (Table 1 and Supplementary Fig. S2D). IL6 and NOTCH2 immunoreactivity were found in the same adenomas, except for PA 34.

Human pituitary tumors contain sphere-forming cells that segregate to the pSP

TSCs generate spheres (‘tumorspheres’) when they are brought under appropriate suspension-culture conditions (Vankelecom & Gremeaux 2010, Clevers 2011). When pituitary adenoma cells were cultured in the presence of B27 and bFGF, free-floating spheres developed progressively. After 7 days of culture, bright and mostly smooth-edged spheres were obtained that clearly differed from the irregular clumps of aggregated cells that also formed (NF-A and GH-A (n=13); Fig. 2A and B). Sphere cells were found to express the stemness markers SOX2 and NESTIN (Chen et al. 2009), which thereby supported their TSC character (n=3; NF-A (n=1) and GH-A (n=2); Fig. 2C and Supplementary Fig. S3A and B, see section on supplementary data given at the end of this article). At the end of the culture period (days 7–10), some of the spheres contained cells that expressed the hormone produced by the original tumor, which suggests some degree of spontaneous differentiation and further supports their TSC phenotype (Supplementary Fig. S3C). To examine the self-renewal capacity of the sphere-initiating cells (as a further property of TSC), primary spheres were dissociated, and cells were seeded again. Secondary spheres with similar morphology developed, although they were fewer in number (n=9; Fig. 2D and E), probably because of a considerable loss of cells after the dispersion of the primary spheres. Third-generation spheres were also obtained, but these were still scarcer (Fig. 2F).

View Full Size

Human pituitary tumors contain sphere-forming cells segregating to the pSP. (A) Smooth-edged spheres (arrows) develop in cultures of total pituitary adenoma cells, which are clearly distinct from clumps of cells sticking together (arrowheads). (B) Examples of spheres at higher magnification. (C) Spheres (arrow) show Sox2 immunofluorescent signal, whereas clumps do not (arrowheads). Nuclei are stained with ToPro3. A representative example is shown (left: light-microscopic image; right: overlay). (D) Secondary spheres (arrows) form after the dissociated primary sphere cells are reseeded. (E) Example of a secondary sphere at higher magnification. (F) Third-generation spheres (arrow) develop after the dissociated second-generation sphere cells are reseeded. (G) Spheres develop from pSP cells of pituitary adenoma, although their number is low. In pMP cell cultures, only clumps of cells are observed (inset). All of the pictures were taken at 7 days of culture. Scale bars: 50 μm. A full colour version of this figure is available at http://dx.doi.org/10.1530/ERC-14-0546.

Citation: Endocrine-Related Cancer 22, 4; 10.1530/ERC-14-0546

• Download Figure
• Download figure as PowerPoint slide

To investigate whether the sphere-forming cells belong to the pituitary tumor's SP, CD31⁻/CD45⁻ SP cells were sorted from adenomas and seeded under sphere-forming conditions (n=3). Spheres developed in the pSP cultures (Fig. 2G) and were absent from pMP cultures, which contained only irregular clumps of cells (Fig. 2G, inset). However, only very few spheres could be obtained (two to four spheres per 20 000 pSP cells originally seeded), which made it impossible to obtain enough viable cells for further passaging after dispersion.

Together, the present findings provide supportive evidence that human pituitary adenomas contain self-renewing sphere-forming cells that give rise to spheres that express stemness markers, which are considered to be properties of TSCs. The findings further show that the sphere-initiating cells segregate to the pSP.

In addition to the in vitro sphere-forming assay, CSC/TSC activity is typically tested in vivo by (xeno-) transplantation in immunodeficient mice (Vankelecom & Gremeaux 2010, Clevers 2011, Chen et al. 2012, Schepers et al. 2012, Boumahdi et al. 2014). However, we observed that human pituitary adenomas, irrespective of their hormonal phenotype, do not grow (i.e., expand) in immunodeficient SCID mice. Although small tumor pieces survived for 3–4 months after s.c. implantation, no enlargement was detected (data not shown). Xenograft adenoma growth failed not only in SCID mice but also in the more immunodeficient NOD–SCID and NOD/SCID^{IL2Rγ−/−} models after the s.c. implantation of the tumor parts as well as the implantation of dissociated cells (data not shown). Similarly, pituitary adenoma fragments engrafted under the kidney capsule survived but did not visibly grow in size (as analyzed after 3–6 months; Supplementary Fig. S4, see section on supplementary data given at the end of this article). Therefore, to be able to further functionally characterize the pituitary tumor SP and to explore its TSC-associated phenotype, we turned to a well-established pituitary tumor cell line, the mouse corticotrope AtT20, which has previously been shown to grow tumors after s.c. implantation in immunodeficient mice (Taguchi et al. 2006).

Characterization of AtT20 SP and tumorigenic activity

First, we examined whether the AtT20 cell line contains a SP. A verapamil-blockable Hoechst^low cell population was detected that amounted to 1.0±0.3% of the total cells (n=12; Fig. 3A). This AtT20 SP showed upregulated expression of the analyzed ‘tumor stemness’ genes Cd44 and Cxcr4 (Fig. 3B).

View Full Size

Identification and characterization of the SP in the AtT20 pituitary tumor cell line. (A) Dot plots of dual-wavelength FACS analysis of AtT20 cells after incubation with Hoechst 33342 alone (left) or together with verapamil (right). The SP (black) and MP (white) are gated, and SP proportions are indicated. A representative example is shown. (B) Cd44 and Cxcr4 gene expression in the SP and MP of AtT20 cells in culture, as determined by qRT-PCR. Bars show means of fold SP/MP±s.e.m. (n=3). *P<0.05. (C) Volume of xenograft tumors grown in SCID mice as measured at days 40 and 47 after s.c. injection of the indicated number of AtT20 SP cells (black bars) and MP cells (gray bars). Bars show means±s.e.m. (five mice for 10 000 cells and four mice each for 5000 and 1000 cells; injected in two or three independent experiments; see Supplementary Fig. S5B). *P<0.05.

Citation: Endocrine-Related Cancer 22, 4; 10.1530/ERC-14-0546

• Download Figure
• Download figure as PowerPoint slide

To study in vivo tumorigenic activity, we first verified the tumor-growing capacity of AtT20 cells in SCID mice. The s.c. injection of 5×10⁵ cells resulted in a gradually developing tumor (Supplementary Fig. S5A, see section on supplementary data given at the end of this article), which was palpable after 15–20 days and further expanded in volume until the mouse became moribund, most probably as a result of corticoid overproduction because of increased adrenal stimulation by the AtT20-produced ACTH (Taguchi et al. 2006). A SP remained present in the AtT20-derived xenograft tumors (as analyzed after 50–55 days), although the proportion was lower than that in the cell line in culture (0.5% versus 1%; n=3).

SP and MP cells were then sorted by FACS from cultured AtT20 cells, and different numbers (10 000, 5000, and 1000 cells) were s.c. injected into SCID mice. Although all of the cell populations induced tumor growth, SP-derived tumors expanded faster (Supplementary Fig. S5B) and/or became larger (Fig. 3C and Supplementary Fig. S5B) as compared with MP-derived tumors. The SP cells developed a tumor with a reconstitution of the SP/MP profile (0.5% SP; n=2), whereas the MP cells generated tumors with a lower proportion of SP cells (0.2%; n=2). Together, the present results provide indications that the AtT20 SP has tumor-proliferative dominance, and that TSC (and/or TSC-like cells) are enriched, although not exclusively segregated, in the SP.

Effects of EMT regulatory pathways in the AtT20 cell line

As described earlier in the present study, the transcriptomic data of the human pituitary adenomas supported the presence of an EMT process in the (p)SP. Therefore, we tested the effects of two important EMT regulatory pathways (De Craene & Berx 2013), that is, the TGFβ and Cxcr4 pathways, in the AtT20 cell line. Moreover, Cxcr4 has been identified as a marker of CSC(-like cells) in various types of cancer (Hermann et al. 2007, Vankelecom & Gremeaux 2010, Clevers 2011). As mentioned in the previous sections, its expression was upregulated in the human pituitary adenoma pSP and the AtT20 SP (Table 3; Figs 1E and 3B). It is worth noting that these studies could not be performed with primary human pituitary adenoma cells, because these cells cannot efficiently be kept in culture and do not grow (well) in vitro (Hofland & Lamberts 2002).

The treatment of AtT20 cells with TGFβ1 for 72 h significantly stimulated the expression of the EMT factors Snai2 and Vim, and there was a clear upward trend for Snai1 and Cxcr4 (Fig. 4A). However, TGFβ1 treatment did not significantly enhance the motility of AtT20 cells, which is an expected functional consequence of EMT (Fig. 4B). Treatment of AtT20 cells in culture with TGFβ1 did not significantly increase the SP proportion (1.2±0.1% versus 1.0±0.3% in control culture; n=3; P>0.05). Along the same lines, the pretreatment of AtT20 cells with TGFβ1 did not lead to a higher tumor-growing capacity when they were s.c. transplanted into SCID mice, either in terms of onset or in terms of progression (volume expansion) of the tumor (Supplementary Fig. S5C).

View Full Size

Involvement of the TGFβ and Cxcr4 pathways in EMT-/TSC-associated properties of the AtT20 pituitary tumor cell line. (A) Gene expression in TGFβ1-treated (72 h) and untreated AtT20 cells as determined by qRT-PCR. Bars show means of fold TGFβ1/control±s.e.m. (n=3). *P<0.05. (B) Effect of TGFβ1 and AMD3100 on the motility of AtT20 cells. Light-microscopic pictures of the AtT20 cell culture after applying a scratch ‘wound’ treated or not (control) with TGFβ1 or AMD3100 for the indicated time period. Representative examples are shown. Scale bar: 100 μm. The percentage open area was determined with TScratch Software (CSElab–ETH, Zurich, Switzerland) in the cultures without compound (black bars) or with TGFβ1 (gray bars) or AMD3100 (dark gray bars). Bars show means±s.e.m. (n=4). *P<0.05. (C) Effect of the Cxcr4-antagonist AMD3100 on AtT20 tumor growth. Immunodeficient SCID mice with growing AtT20-derived subcutaneous tumors were treated with intratumoral injections of AMD3100 (gray) or vehicle (PBS; black) starting from day 21 (arrow) after s.c. AtT20 cell implantation. Volumes of individual tumors are shown, with horizontal lines indicating the mean for each group at each time point (n=3, with an initial five mice per independent experiment, a number that declined toward days 35–42 because of the gradual death of the mice). *P<0.05. (D) Gene expression in AMD3100-treated (20 days) and untreated AtT20 xenograft tumors as determined by qRT-PCR. Bars show means of fold AMD3100/control±s.e.m. (n=3, except for Snai1 and Snai2, where n=2). *P<0.05. (E) Effect of AMD3100 on AtT20 cell viability. Percentage of dying/dead AtT20 cells (as determined by Trypan blue exclusion) after treatment with AMD3100 (gray bars) or vehicle (black bars). Bars show means±s.e.m. (n=3). *P<0.05. (F) Effect of AMD3100 on AtT20 cell apoptosis. Percentage of TUNEL⁺ cells after treatment with AMD3100 (gray bars) or vehicle (black bars). Bars show means±s.e.m. (n=3). *P<0.05. (G) Effect of AMD3100 on AtT20 cell proliferation. Percentage of Ki67⁺ AtT20 cells after treatment with AMD3100 (gray bars) or vehicle (black bars). Bars show means±s.e.m. (n=3). *P<0.05.

Citation: Endocrine-Related Cancer 22, 4; 10.1530/ERC-14-0546

• Download Figure
• Download figure as PowerPoint slide

To analyze the involvement of the Cxcr4 pathway, we treated developing AtT20 xenograft tumors in SCID mice with the Cxcr4 antagonist AMD3100. Interestingly, tumors treated with injections of AMD3100 grew at a slower rate and remained smaller than tumors that received injections of vehicle (Fig. 4C).

We further observed that AMD3100 significantly reduced the migratory activity of AtT20 cells as analyzed in vitro (Fig. 4B), which thus indicated an effect on EMT. However, AMD3100 treatment did not significantly affect the expression levels of the EMT-associated genes Zeb2, Snai1, Snai2 (although a downward trend was observed), Vim, and Cdh1 or of the CSC-associated marker Cd44, as analyzed in the xenograft tumors after 20 days of treatment (Fig. 4D). In vitro, 72 h AMD3100 treatment also did not affect the expression levels of these genes, and it had no significant effect (although there was a tendency towards a decrease) on the proportion of the SP (0.5±0.3% vs 0.9±0.2% in controls; n=3; P>0.05). Notably, there was a small increase in Cxcr4 expression in the AMD3100-treated xenograft tumors (Fig. 4D); the upregulation of Cxcr4 after AMD3100 treatment has also been reported in other tumors, supposedly as a result of the adaptation of the tumor to sustained Cxcr4 signal inhibition (Domanska et al. 2012). Together, EMT genes – at least the ones examined in the present analysis – do not seem to be involved in the AMD3100-induced inhibition of cell motility and of xenograft tumor growth (at least as examined after 20 days of treatment). To search for possible other mechanisms of tumor growth reduction, cell viability and proliferation were evaluated. Treatment of cultured AtT20 cells with AMD3100 resulted in a twofold to threefold increase in dying/dead cells (from 2–4% in controls to 8–11% after AMD3100 treatment; Fig. 4E). In analogy, AMD3100 augmented the number of AtT20 cells that went into apoptosis (about twofold; Fig. 4F). However, no difference could be detected in the number of apoptotic cells after the 20-day AMD3100 treatment in vivo (data not shown). The treatment of cultured AtT20 cells with AMD3100 also reduced the proportion of proliferating cells (from about 90 to 60% after 72 h; Fig. 4G). From these in vitro data, it can be assumed that the tumor-growth-inhibitory effect of AMD3100 at least partially involves a reduction in cell viability and proliferation. From our results so far, there are no indications as of yet that EMT is also implicated.

Pituitary tumorigenesis and pituitary stem cells

In addition to the finding of SOX2 expression in tumorsphere cells, the results of the present microarray analysis also revealed upregulated gene expression of SOX2 in the adenoma pSP (more than 1.5-fold in three of the five human pituitary tumors analyzed; P<0.05). Interestingly, the AtT20 SP also showed higher expression of Sox2 (as compared with the MP; Supplementary Fig. S6A, see section on supplementary data given at the end of this article). A similar finding was observed in an additional pituitary tumor cell line, the rat lactosomatotrope GH3, which also contains a SP (0.5±0.05%; n=18; Supplementary Fig. S6A and B). Sox2 is a stemness factor that also marks the recently identified stem cells of the pituitary gland (Chen et al. 2009). In general, a link may exist between CSCs/TSCs of tumors in organs and the resident stem cells, as has been recently supported by results from studies of several cancers (Barker et al. 2009, Llaguno et al. 2009, Zhu et al. 2009, Mulholland et al. 2010, Vankelecom & Gremeaux 2010, Clevers 2011, Lapouge et al. 2011, Chen et al. 2012, Schepers et al. 2012, Boumahdi et al. 2014, Vankelecom & Chen 2014, Vanner et al. 2014).

To explore the pituitary stem cells in connection with the gland's tumors, we turned to the Drd2^−/− mouse model (Cristina et al. 2006). In the pituitaries of Drd2^−/− female mice, prolactinomas gradually develop starting at 6–8 months of age, which can be detected by MRI and after necropsy (Supplementary Fig. S7A, see section on supplementary data given at the end of this article). The CD31⁻/CD45⁻ SP (pSP) proportion in the tumor-bearing AP from Drd2^−/− mice (1.7±0.3%; n=5; Supplementary Fig. S7B) was higher than that in heterozygous Drd2^+/− (1.4±0.6%; n=2) and WT Drd2^+/+ control littermates (1.3±0.2%; n=2). Calculated using absolute cell numbers (see the ‘Materials and methods’ section), the Drd2^−/− AP contains about tenfold more pSP cells than the control mice (Fig. 5A). Similarly, the absolute number of colony-forming cells (which is also often used to investigate TSC activity in vitro) was higher in Drd2^−/− than in Drd2^+/− pituitaries (3.9±0.7-fold; n=3). In addition, colonies that developed from the Drd2^−/− pituitary were generally larger in size and more smoothly shaped than were the colonies that developed from the heterozygous mice (Supplementary Fig. S7C).

View Full Size

Pituitary stem cells in the Drd2^−/− pituitary tumor model. (A) Fold difference in the absolute number of pSP cells in the anterior pituitary (AP) of homozygous Drd2^−/− (n=5) and heterozygous Drd2^+/− mice (n=2), relative to WT Drd2^+/+ mice (set as 1). *P<0.05 versus Drd2^+/+ mice. (B) Absolute number of Sox2⁺ cells in the AP of homozygous Drd2^−/−, heterozygous Drd2^+/− mice, and WT Drd2^+/+ mice. Bars represent means±s.e.m. (n=3), with an indication of the nuclear-Sox2⁺ cells (black segments) and the cytoplasmic-Sox2⁺ cells (gray segments). *P<0.05 versus Drd2^+/+ mice (for total Sox2⁺ cell number). (C) Immunofluorescent staining of dissociated AP cells (cytospins) from WT Drd2^+/+ mice (upper panel) and homozygous Drd2^−/− mice (middle and lower panels) for Sox2 (red) and Ki67 (green). Double Sox2⁺/Ki67⁺ cells are indicated (arrows). Nuclei are stained with DAPI. Representative examples are shown (female mice, 10–12 months of age). Scale bars: 50 μm. (D) Confocal images of pituitary vibratome sections of Drd2^−/−, Drd2^+/−, and WT Drd2^+/+ mice immunofluorescently stained for Sox2 (red) and prolactin (PRL, green). Representative examples are shown (female mice, 10–13 months of age). In all of the mouse phenotypes, Sox2⁺ cells are found in the marginal cell layer along the cleft. In Drd2^−/− pituitary, Sox2⁺ cells are clustered in small groups and are often observed within the PRL⁺ tumor cell mass (arrows), but clusters of Sox2⁺ cells are also found scattered in the AP parenchyma of control Drd2^+/− and Drd2^+/+ mice. Nuclei are stained with ToPro3. Scale bars: 50 μm. Insets show a lower-magnification overview. IL, intermediate lobe. A full colour version of this figure is available at http://dx.doi.org/10.1530/ERC-14-0546.

Citation: Endocrine-Related Cancer 22, 4; 10.1530/ERC-14-0546

• Download Figure
• Download figure as PowerPoint slide

We next examined the prolactinoma-bearing pituitaries of Drd2^−/− mice for Sox2⁺ cells. According to countings in dissociated APs (see the ‘Materials and methods’ section), the absolute number of Sox2⁺ cells was higher in the Drd2^−/− AP as compared with the WT gland (2.4±0.6-fold; n=3; Fig. 5B). At least part of the increase in Sox2⁺ cells is the result of higher proliferative activity. Whereas proliferating (Ki67⁺) Sox2⁺ cells were virtually absent in WT Drd2^+/+ AP at the ages tested (6–12 months), double Sox2⁺/Ki67⁺ cells were more readily observed in Drd2^−/− AP (Fig. 5C). Furthermore, as has been shown previously (Chen et al. 2009, Fu et al. 2012, Gremeaux et al. 2012), Sox2 immunoreactive signal was not only found in the expected location in the nucleus (nuclear-Sox2⁺ (nSox2⁺) cells); some cells showed Sox2⁺ signals in the cytoplasm (cytoplasmic-Sox2⁺ (cSox2⁺) cells). The Drd2^−/− cSox2⁺ cells showed a higher increase than the nSox2⁺ cells as compared with Drd2^+/+ WT pituitaries (3.1±0.3- and 2.3±0.7-fold respectively; Fig. 5B and C). In situ, Sox2⁺ cells in the Drd2^−/− pituitary were found in the marginal zone bordering the cleft, which is not different from the WT pituitary (Fig. 5D and Supplementary Fig. S7D). On top of this location, frequent clusters of Sox2⁺ cells were observed within the PRL⁺ tumor cell mass in the Drd2^−/− pituitary (Fig. 5D and Supplementary Fig. S7D). In the WT gland, Sox2⁺ cell clusters are also found spread within the endocrine AP parenchyma (Fig. 5D and Supplementary Fig. S7D). Double Sox2⁺/PRL⁺ cells were not detected in either of the pituitary phenotypes. Taken together, the number of Sox2⁺ cells is clearly increased in the Drd2^−/− versus the WT pituitary, but their overall appearance in the putative stem cell niches seems to be not very different.