MAPK pathway control of stem cell proliferation and differentiation in the embryonic pituitary provides insights into the pathogenesis of papillary craniopharyngioma

In vertebrates, the pituitary gland is considered to be master regulator of homeostasis because it regulates a wide range of essential physiological functions, such as metabolism, growth, fertility and the stress response. The pituitary gland comprises anterior, intermediate and posterior lobes (AL, IL and PL, respectively). The AL and IL derive from Rathke's pouch, a dorsal evagination of the oral ectoderm at the boundary with the pharyngeal endoderm. The PL develops from a diverticulum at the ventral midline of the diencephalon: the infundibulum. At 18.5 dpc, the AL contains five hormone-expressing cell types – somatotrophs, lactotrophs, thyrotrophs, gonadotrophs and corticotrophs – that secrete growth hormone (GH), prolactin (PRL), thyroid-stimulating hormone (TSH), gonadotropins (FSH and LH) and adrenocorticotropin hormone (ACTH), respectively (Kelberman et al., 2009). In addition, the AP contains Sox2⁺/Sox9⁺ cells, a proportion of which represent stem cells in the postnatal pituitary (Castinetti et al., 2011; Andoniadou et al., 2013; Rizzoti et al., 2013). These cells are concentrated around the pituitary cleft in the IL and the ‘marginal zone’, a region of the dorsal AL mostly devoid of hormone-producing cells (Fauquier et al., 2008; Garcia-Lavendeira et al., 2009; Goldsmith et al., 2016; Pérez Millán et al., 2016). Dispersed Sox2⁺/Sox9⁺ cells are also detected throughout the parenchyma of the AL, intermingled with hormone-expressing cells (Mollard et al., 2012). The IL contains melanotrophs, which express melanocyte-stimulating hormone (MSH). The PL is devoid of endocrine cell types and contains the axonal projections from hypothalamic neurons.

Several signalling pathways and transcription factors regulate the normal development of the pituitary gland. Around 9.5 days post coitum (dpc), secreted signals from the prospective hypothalamus, such as FGF8, FGF10, BMP4 and SHH are required for RP induction and for the expression of transcription factors, including Hesx1, Sox2, Lhx3 and Lhx4 in RP progenitors (Ericson et al., 1998; Treier et al., 1998, 2001). RP progenitors divide actively from 9.5 to 14.5 dpc (Davis et al., 2011) to generate post-mitotic precursors that initiate cell-lineage commitment by expressing: (1) Sf1 (Nr5a1), in the gonadotroph cell lineage (Schimmer and White, 2010); (2) the T-box transcription factor Tpit, which is responsible for the activation of the pro-opiomelanocortin (Pomc1) gene in corticotrophs (and melanotrophs of the intermediate lobe) (Lamolet et al., 2001); and (3) POU Class 1 homeobox 1 (Pou1f1 or Pit1), which control the differentiation of thyrotrophs, lactotrophs and somatotrophs (Dollé et al., 1990). Mouse mutants and humans carrying inactivating mutations in most of these genes develop hypopituitarism (Fang et al., 2016).

The mitogen-activated protein kinase (MAPK) pathway encompasses different signalling cascades, of which the Ras-Raf-Mek-ERK1/2 (hereafter MAPK/ERK) is one of the most dysregulated in human cancer and, additionally, plays important roles during normal physiology (Dhillon et al., 2007; Zhang and Liu, 2002). Extracellular growth factors (FGFs, EGF and PDGF, among others) bind to and activate receptor tyrosine kinases, causing a downstream phosphorylation cascade (RAS-RAF-MEK-ERK), which eventually leads to transcription of target genes controlling cellular proliferation, differentiation, apoptosis and senescence. Gain-of-function mutations in components of the pathway, such as BRAF p.V600E and KRAS p.G12D, have been identified in numerous tumours and cancers (Dhillon et al., 2007). These mutations lead to the overactivation of the MAPK/ERK pathway and increase cell proliferation and survival, resulting in cell transformation and tumourigenesis. In pituitary tumours, biochemical evidence indicates that this pathway is activated in many types of pituitary adenoma in humans (Dworakowska et al., 2009). However, mutations in components of the pathway have been identified only in papillary craniopharyngioma (PCP); specifically, the BRAF p.V600E mutation is present in the vast majority of the tumours analysed (Brastianos et al., 2014).

In this manuscript, we have addressed the role of MAPK/ERK pathway during normal pituitary development and in tumourigenesis by conditionally activating this pathway in RP progenitors during embryonic development. Our results demonstrate that persistent activation of the pathway leads to a drastic increase in the proliferative capacity of Sox2⁺ cells and impairment of their differentiation properties, resulting in enlargement of the Sox2⁺ stem cell compartment by the end of gestation. Additionally, expression analysis of human tumour samples strongly suggests that similar mechanisms underlie the pathogenesis of PCP.

We have previously shown that the Hesx1-Cre mouse line drives robust Cre-mediated activity in the developing pituitary gland by 9.0 dpc (Andoniadou et al., 2007; Gaston-Massuet et al., 2011; Jayakody et al., 2012). To assess the role of the MAPK/ERK pathway during development, we crossed the Hesx1^Cre/+ mice with either Braf^V600E/+ or Kras^G12D/+ animals (Mercer et al., 2005; Tuveson et al., 2004). Genotypic analysis of 10.5-18.5 dpc embryos showed no statistically significant variation from the expected Mendelian ratios (Table S1). In contrast, genotyping of postnatal mice from birth to 3 weeks failed to identify any viable Hesx1^Cre/+;Braf^V600E/+ or Hesx1^Cre/+;Kras^G12D/+ mice (Table S1). Histological examination revealed the presence of expanded airway structures in both mouse models at 18.5 dpc, suggesting that abnormal lung development could be the cause of the perinatal death observed (Fig. S1) (Tang et al., 2011).

Haematoxylin and Eosin staining of Hesx1^Cre/+;Braf^V600E/+ and Hesx1^Cre/+;Kras^G12D/+ mutants at 10.5 dpc revealed no gross morphological defects in the developing RP of these mutants compared with control littermates (Fig. 1A-C). The first clear evidence of a morphological defect, typically anterior pituitary hyperplasia, was observed at 12.5 dpc and was further pronounced by 14.5 dpc (Fig. 1D-I). At 18.5 dpc, a fully penetrant phenotype of severe anterior pituitary hyperplasia with branched cleft was observed in all embryos analysed (Fig. 1J-L). Cell counts of dissociated pituitaries at 18.5 dpc revealed a total of 96,000±2.7% in the Hesx1^Cre/+;Braf^V600E/+mutant (P≤0.01), 101,666±4.3% in the Hesx1^Cre/+;Kras^G12D/+ mutant (P≤0.01) and 67,200±3.5% in control wild-type mice (Fig. 1M). The posterior lobe was present and apparently normal in Hesx1^Cre/+;Braf^V600E/+ and Hesx1^Cre/+;Kras^G12D/+ mutant pituitaries (Fig. 1J-L). These data suggest that RP induction occurs normally in the Hesx1^Cre/+;Braf^V600E/+ and Hesx1^Cre/+;Kras^G12D/+ mutants, followed by an increase in proliferation, leading to hyperplasia of the anterior pituitary by the end of gestation.

The expression patterns of Braf and Kras mRNA and pERK1/2 protein expression, as a readout of activated MAPK/ERK pathway, were respectively analysed by in situ hybridisation and immunostaining on histological sections of wild-type, Hesx1^Cre/+;Braf^V600E/+ and Hesx1^Cre/+;Kras^G12D/+ embryos. At 10.5 dpc wild-type embryos, Braf and Kras mRNA were clearly detected in the prospective hypothalamus, including the posterior diencephalon and pre-optic area, in addition to the developing RP (Fig. 2A,D). At 12.5 and 14.5 dpc, transcripts were detected throughout the developing pituitary, including the infundibulum, periluminal epithelium and anterior lobe (Fig. 2B,E; data not shown). At 18.5 dpc, the expression domain of Kras and Braf was restricted to the periluminal area, including the intermediate lobe and marginal zone, with scattered cells dispersed throughout the anterior lobe (Fig. 2C,F). The expression patterns of Braf in Hesx1^Cre/+;Braf^V600E/+ (Fig. 3A-C) and Kras in Hesx1^Cre/+;Kras^G12D/+ (data not shown) mutants were comparable with those in wild-type embryos.

These expression patterns of mRNA expression did not correlate with pERK1/2 detection at all embryonic stages. At 10.5 dpc, pERK1/2 expression was comparable between genotypes in both the developing hypothalamus and RP (Fig. 4A,E,I), corresponding with Braf and Kras mRNA expression (Fig. 2). In contrast, at 12.5 and 14.5 dpc, very few pERK1/2⁺ cells were detected, despite the broad expression of Braf and Kras mRNA in mutant and wild-type embryos (Fig. 4B,C,F,G,J,K) as well as the extensive expression of BRAF-V600E mutant protein throughout the Hesx1^Cre/+;Braf^V600E/+ pituitary (Fig. 3D-F). At 18.5 dpc, the expression of pERK1/2 was noticeably increased in both Hesx1^Cre/+;Braf^V600E/+ and Hesx1^Cre/+; Kras^G12D/+ mutants compared with the control pituitaries (Fig. 4D,H,L), and correlated with the in situ data (Figs 2 and 3; data not shown). Together, these expression studies demonstrate that the MAPK/ERK pathway is temporally regulated with the highest levels observed at 10.5 and 18.5 dpc. In addition, we show that most of the cells upregulating the pathway are located in the epithelium lining the cleft, an area enriched for Sox2⁺ undifferentiated embryonic precursors and stem cells (Andoniadou et al., 2013).

To assess the effects of the overactivation of the MAPK/ERK pathway in Rathke's pouch (RP) induction, cell lineage commitment and terminal differentiation, we performed detailed expression analyses at 10.5, 14.5 and 18.5 dpc in both mutants in comparison with controls. The expression of Fgf10, Bmp4 and Shh within the hypothalamus is required for Lhx3 activation in RP progenitors (Ericson et al., 1998; Treier et al., 1998, 2001). At 10.5 dpc, the expression domains of these markers were indistinguishable between mutants and controls, supporting the notion that early RP induction occurs normally in the Hesx1^Cre/+;Braf^V600E/+ and Hesx1^Cre/+;Kras^G12D/+ mutant embryos (Fig. S2). Immunostaining against POU1F1 (PIT1), NR5A1 (SF1) and TPIT, markers of specific cell-lineage precursors of the anterior pituitary, revealed evident defects in the Hesx1^Cre/+;Braf^V600E/+ mutants relative to control embryos at 14.5 dpc (Fig. 5A-I). Specifically, the proportion of PIT1⁺ cells was significantly reduced relative to total cells (control, 29%; Braf mutant, 4%; P=0.001), SF1⁺ cells were almost absent (control, 4%; Braf mutant, 0.5%; P≤0.05) and TPIT⁺ cells were increased (control, 10%; Braf mutant, 26%; P=0.001) (Fig. 5J). In the Hesx1^Cre/+;Kras^G12D/+ mutants, PIT1⁺ and SF1⁺ cell lineage precursors were also significantly reduced (PIT1⁺: control, 29%; Kras mutant, 13%; P≤0.01; SF1⁺: control, 4.5%; Kras mutant, 0.6%; P≤0.05), albeit to a lesser extent than in Hesx1^Cre/+;Braf^V600E/+ mutants. Numbers of TPIT⁺ cells were similar (control, 10%; Kras mutant, 12%; P=0.6156) (Fig. 5J).

Total numbers of hormone-producing cells were calculated by adjusting the proportion of each of these cell types to the size of the pituitary gland of the Hesx1^Cre/+;Braf^V600E/+, Hesx1^Cre/+;Kras^G12D/+ and control embryos at 18.5 dpc (see Materials and Methods). We reasoned that this approach could avoid any bias in the quantitative analysis due to the hyperplasia observed in the mutant pituitaries; for example, numbers of a particular cell type could be the same in mutant and controls but distributed in more histological sections in the mutant pituitary, hence seeming proportionally reduced. Overall, specific immunostaining revealed abnormal terminal differentiation of hormone-producing cells in both Hesx1^Cre/+; Braf^V600E/+ and Hesx1^Cre/+;Kras^G12D/+ mutants, although the effects were more accentuated in the former (Fig. 6A-R). Specifically, numbers of somatotrophs (GH⁺) (wild type, 23560; Braf mutant, 3185), lactotrophs (PRL⁺) (wild type, 17604; Braf mutant, 3233) and thyrotrophs (TSH⁺) (wild type, 4043; Braf mutant, 664) (i.e. the Pit1 cell lineage) were significantly reduced, and gonadotrophs (FSH⁺) (wild type, 6134; Braf mutant, 1686) and LH⁺ (wild type, 1564; Braf mutant, 23) almost completely absent in Hesx1^Cre/+;Braf^V600E/+ mutant pituitaries relative to controls (Fig. 6S). Immunostaining against αGSU (wild type, 6861; Braf mutant, 2977) confirmed the reduction of thyrotrophs and gonadotrophs in the mutant pituitaries (Fig. S3). In contrast, total numbers of corticotrophs (ACTH⁺) (wild type, 15958; Braf mutant, 28778) appeared significantly elevated in Hesx1^Cre/+;Braf^V600E/+ mutants (Fig. 6S). Abnormal differentiation was also observed in the Hesx1^Cre/+;Kras^G12D/+ mutant pituitaries, but to a lesser extent, with a reduction in numbers of thyrotrophs (wild type, 4043; Kras mutant, 1371), gonadotrophs (FSH⁺) (wild type, 6134; Kras mutant, 4044) and lactotrophs (wild type, 17604; Kras mutant, 6710). Similar to Hesx1^Cre/+;Braf^V600E/+ mutants, elevated numbers of corticotrophs were also observed in Hesx1^Cre/+;Kras^G12D/+ mutant pituitaries (wild type, 15958; Kras mutant, 21370) (Fig. 6S). Furthermore, absolute quantification of mRNA transcripts using quantitative RT-PCR from 18.5 dpc control, Hesx1^Cre/+;Kras^G12D/+ and Hesx1^Cre/+;Braf^V600E/+ pituitaries revealed a significant decrease in Gh and a trend towards increased Acth mRNA in both mutant genotypes (Fig. 6T), in agreement with the cell counts. Together, these analyses demonstrate that cell-lineage commitment and terminal differentiation are severely disrupted upon activation of the MAPK pathway in RP embryonic precursors, with an overall reduction of hormone-producing cells at the end of gestation.

Analysis of cell proliferation of RP progenitors was carried out at 12.5, 14.5 and 18.5 dpc in Hesx1^Cre/+;Braf^V600E/+, Hesx1^Cre/+;Kras^G12D/+ and control pituitaries. Quantitative analysis of the proliferation index using Ki67 staining, a marker of cycling cells, revealed no difference at 12.5 dpc, with around 70% of total cells being Ki67⁺ (Fig. 7A-C,J). However, a significant increase in Ki67⁺ cells at 14.5 and 18.5 dpc was observed in both mutant genotypes, relative to stage-matched controls (Fig. 7D-J). Of note, while the Ki67 proliferation index decreased gradually in control pituitaries from 72% at 12.5 dpc to 22% at 18.5 dpc, this progressive quiescence was not observed to the same degree in the Hesx1^Cre/+; Kras^G12D/+ or Hesx1^Cre/+;Braf^V600E/+ mutants. At 18.5 dpc, the proportion of proliferating cells fell to only 42% in Hesx1^Cre/+;Kras^G12D/+ and 55% in Hesx1^Cre/+;Braf^V600E/+ mutant pituitaries (Fig. 7J). Increased proliferation was also confirmed by immunostaining against Cyclin D2, which controls G1/S cell cycle progression and is mainly expressed in RP progenitors in the periluminal epithelium (Bilodeau et al., 2009). Quantification of Cyclin D2-expressing cells out of total DAPI-stained nuclei within the periluminal epithelium revealed an increased trend in both mutant genotypes at 12.5 dpc compared with stage-matched controls (Fig. S4).

The restricted expression of pERK1/2 and increased proliferation of periluminal cells, prompted us to assess whether the Sox2⁺ cells could be the main population responding to the overactivation of the MAPK/ERK pathway in our mouse models. Histological sections of Hesx1^Cre/+;Braf^V600E/+, Hesx1^Cre/+;Kras^G12D/+ and control pituitaries were analysed by immunostaining at different developmental stages. At 12.5 dpc, the proportion of Sox2⁺ cells was initially similar between genotypes at around 80% of total cells (Fig. 8A-C,J). However, in control embryos, the number of Sox2⁺ cells fell to 21% by 18.5 dpc (P=0.0004) (Fig. 8G,J). Strikingly, in Hesx1^Cre/+;Kras^G12D/+ and Hesx1^Cre/+;Braf^V600E/+ mutant pituitaries, the high proportion of Sox2⁺ cells found at earlier developmental time points was maintained, falling to only 53% (P=0.0519) and 72% (P=0.068) of total cells at 18.5 dpc, respectively. This suggests a trend towards a decrease in Sox2⁺ cell numbers although not statistically significant (Fig. 8H-J). Further to this, levels of Sox2 mRNA were observed to be significantly increased in Hesx1^Cre/+;Braf^V600E/+ mutants by quantitative RT-PCR (Fig. 8K). Corroborating the histological findings, the clonogenic potential in Hesx1^Cre/+;Kras^G12D/+ and Hesx1^Cre/+;Braf^V600E/+ pituitaries at 18.5 dpc was nearly three and five times higher, respectively, than in controls when analysed in stem cell-promoting medium (Fig. 8L,M).

The described defects suggest that the MAPK pathway may control the balance between self-renewal and differentiation of Sox2⁺ progenitors during development. To further characterise this defect, we performed a pulse-chase experiment with the nucleotide analogue EdU, which is incorporated during the S phase of the cell cycle. Specifically, we injected EdU at 14.5 dpc, when most of the replicating cells are Sox2⁺ progenitors, and analysed the pituitaries 48 h later at 16.5 dpc, when progenitors have exited cell cycle to commit to specific cell lineages (Fauquier et al., 2008; Bilodeau et al., 2009). This was performed in both control and Hesx1^Cre/+;Braf^V600E/+ embryos, owing to their more pronounced phenotype.

In control embryos, ∼25% of the initially labelled EdU cells expressed the commitment marker PIT1 and around 21% expressed SOX2 (Fig. 9A,C,D,F). In contrast, in Hesx1^Cre/+;Braf^V600E/+ mutants, the percentage of EdU⁺/PIT1⁺ cells dropped to 3% and the proportion of EdU⁺/Sox2⁺ cells was elevated to 40% (Fig. 9B,C,E,F). Despite the significant increase in ACTH⁺ cells in the mutant pituitary at 18.5 dpc, the proportion of EdU⁺/TPIT⁺ positive in this tracing experiment remained similar between genotypes (8% in controls and 9% in mutants) (Fig. 9H,I). This tracking experiment also revealed a higher proportion of EdU⁺/Ki67⁺ cells in the Hesx1^Cre/+;Braf^V600E/+ mutants (25%) relative to the controls (18%) (Fig. 9J-L). To assess which cells were labelled at 14.5 dpc, we performed a short pulse experiment by injecting EdU in pregnant females at 14.5 dpc and analysed the embryos 2 h later. This analysis confirmed that the majority of the EdU+ cells expressed SOX2, specifically 74% of labelled EdU⁺ cells expressed SOX2 in the control and 87% in the Hesx1^Cre/+;Braf^V600E/+ mutants pituitaries, suggesting a trend towards an increase, although not significant (P=0.1599) (Fig. S5A-C). In the control pituitary, 6% of the EDU⁺ cells expressed PIT1 and 1.4% expressed TPIT, while in the mutant pituitary these proportions were 0.1% for PIT1 and 3.7% for TPIT, both significantly different (P=0.0246, PIT1 and P=0.0490, TPIT) (Fig. S5D-I). Three conclusions can be drawn from these EdU studies: (1) the activation of the MAPK/ERK pathway drives self-renewal versus differentiation of Sox2⁺ cells; (2) there is a blockade in the differentiation of Sox2⁺ cells into PIT1 progenitors; and (3) there is no bias in the differentiation of Sox2⁺ cells into the TPIT cell lineage; rather, TPIT progenitor respond to the pathway activation by increasing proliferation. Together, these data demonstrate that the activation of the MAPK pathway in embryonic precursors results in significant expansion of the Sox2⁺ cell population and increased clonogenic potential. In addition, we show that an abnormal balance between self-renewal and differentiation contribute to these defects.

Higher numbers of organ-specific stem cells with reduced differentiation potential can lead to tumourigenesis in many organs, including the pituitary gland (Gaston-Massuet et al., 2011). As BRAF p.V600E is the sole genetic event associated with human papillary craniopharyngioma (PCP), we hypothesised that overactivated MAPK in these tumours may lead to similar cellular alterations to those observed in the Hesx1^Cre/+;Braf^V600E/+ mutant pituitary, i.e. expansion of the Sox2⁺/Sox9⁺ cells with decreased differentiation capacity.

PCPs are histologically characterised by the presence of cores of fibrovascular stroma lined by basal cells and suprabasal layers of well-differentiated non-keratinised squamous epithelium (Fig. 10A). PCPs are non-functional tumours, which do not express pituitary hormones (Louis et al., 2016). In a series of archival formalin-fixed paraffin-embedded (FFPE) PCP specimens (n=5), we confirmed the presence of p.BRAF(V600E) mutations (Table S2) and expression of BRAFV600E in all tumour cells (Fig. 10B). pERK1/2 staining was more restricted and focused in areas around the fibrovascular cores (Fig. 10E). Most of the cells in the tumours were positive against a pan-cytokeratin antibody recognising cytokeratins, but cells from the basal layer were negative (Fig. 10C). This was further evidenced by immunostaining against cytokeratin 19 (Fig. 10D). Double immunostaining revealed that the squamous epithelial tumour components robustly expressed SOX2 and SOX9 in a graded manner, with the highest expression observed in basal cells immediately in contact with the fibrovascular cores and a progressive loss of signal in cells away from the SOX2⁺/SOX9⁺ areas (Fig. 10F-H). Quantification analyses demonstrated that ∼16% of these Sox2⁺ cells expressed Ki67 (Fig. 10I-L). Moreover, up to 91% of all cycling cells within the tumour (excluding host-derived stroma) co-expressed SOX2, suggesting that proliferating SOX2⁺ cells may be driving PCP growth (Fig. 10L).

For comparison purposes, we assessed the proliferative potential of Sox2⁺ cells during normal human embryonic pituitary development and adulthood. At CS20 (Carnegie stage 20; gestational day 49), an embryonic stage that is analogous to 14.5 dpc in mouse, human foetal pituitaries were predominantly SOX2⁺;SOX9⁺, and almost the majority of these cells expressed Ki67 (Fig. S6A-C,J-L). In contrast, adult human pituitary samples contained very few SOX2⁺ cells in the anterior lobe (Fig. S6D-I), which very rarely expressed Ki67 with an almost mutual exclusivity between these two markers (Fig. S6M-R). These observations suggest that, as in the mouse model, the activation of the MAPK pathway in human PCP tumours result in increased proliferative capacity and reduced terminal differentiation of the SOX2⁺ cells.

In this article, we show that the overactivation of the MAPK pathway in RP embryonic precursors results in increased sustained proliferation and impaired terminal differentiation of Sox2⁺ cells during development. Consequentially, by the end of gestation, the pituitary gland contains vast numbers of Sox2⁺ cells, leading to an enlargement of the pituitary stem cell compartment. Because the Hesx1-Cre mouse line drives Cre-mediated recombination in the anterior hypothalamus, we cannot completely rule out a hypothalamic contribution to the observed phenotype. However, the normal hypothalamic patterning at 10.5 dpc supports a more relevant role for the dysregulation of the MAPK/ERK pathway in the developing RP. Of translational relevance, we reveal the presence of a population of Sox2⁺ cells in human PCP tumours endowed with proliferating capacity but unable to differentiate into hormone-producing pituitary cells.

Our data suggest that the activation of the MAPK/ERK pathway by FGF ligands is crucial during pituitary development. The FGF family of secreted factors binds to tyrosine kinase FGF receptors (FGFRs) to signal through the MAPK/ERK and other pathways (Ornitz and Itoh, 2015). Several FGFs, including Fgf8, Fgf10 and Fgf18 are expressed in the developing ventral diencephalon around 9.5-10.5 dpc (Treier et al., 1998, 2001), and loss-of-function mutants of FGF and FGFR2 demonstrate that FGF signalling is required for normal proliferation of RP progenitors and pituitary morphogenesis (De Moerlooze et al., 2000; Ohuchi et al., 2000). Ex vivo culture of RP tissue exposed to FGFs and FGF inhibitors has revealed a crucial function for the pathway in controlling both RP proliferation, and its downregulation is required for normal PIT1-cell lineage specification (Ericson et al., 1998; Norlin et al., 2000). Of note, transgenic ectopic expression of Fgf8 in the developing pituitary using a Cga (a-GSU) promoter results in pituitary hyperplasia and expansion of ACTH-expressing cells (corticotrophs and melanotrophs) with a concomitant severe loss of gonadotrophs, somatotrophs, thyrotrophs and gonadotrophs (Treier et al., 1998), a phenotype resembling our mouse models that activate the MAPK-ERK pathway. The data presented here raise the possibility that other ligands such as EGF, PDGF and TGFα, which, like FGFs, can activate the MAPK/ERK pathway, might be important during pituitary development. Alternatively, it is also possible that FGF is the only secreted signal activating the MAPK/ERK pathway in RP. Further analyses are required to clarify the contributions of other signals besides FGFs.

We reveal a crucial function of the MAPK/ERK pathway in the control of the stem cell compartment in the pituitary by the end of gestation. The proportion of Sox2⁺ cells is drastically increased and comprises almost three-quarters of the entire cell population in the Hesx1^Cre/+;Braf^V600E/+ at 18.5 dpc. Our analyses show that this substantial expansion is due to a failure of Sox2⁺ embryonic progenitors to exit the cell cycle and differentiate at earlier developmental stages, resulting in increased clonogenic potential by the end of gestation. In addition, pERK1/2 staining is strongly detected in the marginal zone, a region enriched in Sox2⁺ stem cells at 18.5 dpc, suggesting that the MAPK pathway has a role in controlling stem cell specification during development and perhaps during stem cell homeostasis postnatally.

BRAF and KRAS mutations are uncommon in pituitary adenomas (Newey et al., 2013; Reincke et al., 2015; Ronchi et al., 2016; Välimäki et al., 2015), but the overactivation of the MAPK/ERK pathway has been reported in these neoplasias. A mutation was identified in codon 12 of the HRAS gene (Gly to Val) in a recurrent prolactinoma that was highly invasive and ultimately proved to be fatal (Karga et al., 1992). Overexpression of both BRAF mRNA and protein has also been reported in non-functioning pituitary adenomas (Ewing et al., 2007). Persistent MAPK/ERK pathway activation, by expressing oncogenic KRAS-G12V, promotes differentiation of the bi-hormonal somatolactotroph GH4 precursor cell line into a prolactin-secreting cell type but is not sufficient to drive tumourigenesis (Booth et al., 2014). In our mouse models, lactotroph differentiation is impaired rather than promoted, suggesting that the cell context where the MAPK/ERK pathway is activated (i.e. GH4 versus embryonic precursors) and/or approach used (i.e. in vitro versus in vivo) may have a profound effect on the phenotypic outcomes.

PCPs are benign tumours of the sellar region, mostly affecting the elderly, and can be associated with significant morbidity when behaving aggressively. The majority of the PCP tumours carry BRAF p.V600E mutations, but little is known about the consequences of the expression of this oncogenic protein (Brastianos et al., 2014). Of interest, preliminary data indicate that the use of inhibitors against BRAF-V600E and MEK may be of clinical relevance (Aylwin et al., 2016; Brastianos et al., 2016). Our mouse models have provided important insights into the molecular pathogenesis of PCP. We reveal the presence of a population of cycling Sox2⁺ cells in the basal layer surrounding the fibrovascular cores of PCP tumours, suggesting that these cells may be driving tumour growth. Indeed, our analysis shows that most of the cycling cells in the tumour are contained in the Sox2⁺ cell compartment. It is interesting that despite the broad expression of the BRAF-V600E mutant protein in the tumours, pERK1/2 staining is mostly restricted to cells surrounding the fibrovascular cores, a region enriched in Sox2⁺ cells. This is very similar to our observations in the mouse models, where the activation of the pathway, as evidenced by pERK1/2 staining, mostly occurs in the Sox2⁺ cells, even if all the pituitary cells express BRAF-V600E. In the human tumours, the squamous epithelium, which is thought to derive from the basal cell layer, does not express any pituitary hormone or cell-lineage commitment marker (PIT1, TPIT and SF1, data not shown), suggesting that, as in the mouse, the activation of the MAPK/ERK pathway in human PCP leads to sustained proliferation and impaired differentiation of the Sox2⁺ cells. Therefore, although there are differences between human PCP and the mouse models presented here, both have in common that Sox2⁺ cells represent the major proliferative cell population and that these show impaired differentiation potential. The perinatal lethality in Hesx1^Cre/+;Braf^V600E/+ mice precluded further analysis and, therefore, we cannot conclude whether increased numbers of mutated Sox2⁺ cells in the pituitary may eventually drive the formation of PCP-like tumours in these mice at an older age. Further refinement may be required to model PCP more accurately [e.g. using the Sox2-CreERT2 mouse line (Andoniadou et al., 2013)]. Although hampered by the rarity of these tumours, further research aiming to test whether these cycling Sox2⁺ cells may represent a tumour-initiating cell population within human PCP is also warranted.

The Hesx1^Cre/+, Braf^V600E/+or Kras^G12D/+ mice have previously been described (Andoniadou et al., 2007; Mercer et al., 2005; Tuveson et al., 2004). In both the Braf^V600E/+ and the Kras^G12D/+ mouse lines, the expression of the oncogenic protein is activated upon Cre-mediated recombination. Genotyping of mice and embryos was carried out by PCR on ear punch biopsies or pieces of tissue from embryos digested in DNAreleasy (Anachem) as per manufacturer's instructions. The data presented in this work are representative of examples of at least three individual embryos per genotype. All the experiments performed in mice were carried out according to UK Home Office guidance and approved by a local ethical committee.

Haematoxylin and Eosin staining was performed as previously described (Gaston-Massuet et al., 2011). Histological processing of embryos and in situ hybridisation on paraffin sections was performed as previously described (Gaston-Massuet et al., 2008; Sajedi et al., 2008). The antisense riboprobes used in this study (Fgf10, Shh, Bmp4 and Lhx3) have been described (Gaston-Massuet et al., 2008; Sajedi et al., 2008; Jayakody et al., 2012; Trowe et al., 2013). Full-length Braf and Kras antisense riboprobes were obtained from Source Bioscience (PX00999A07 and IRAVp968D072D, respectively).

For EdU labelling of embryonic pituitaries, pregnant females at 14.5 dpc were subjected to a single intra-peritoneal injection of EdU at a dose of 100 mg/kg and embryos dissected either 2 h later for short-term tracing experiments or 48 h later at 16.5 dpc for longer term tracing. Double immunostaining against specific markers (SOX2, PIT1, TPIT and KI67) and EdU was conducted using the Click-It EdU imaging kit (Invitrogen) according to manufacturer's instructions.

Embryos were fixed in 4% PFA and processed for immunodetection as previously described (Jayakody et al., 2012; Andoniadou et al., 2013). Detection of hormones was carried out using antibodies for α-GH (NHPP AFP-5641801), α-ACTH (10C-CR1096M1), α-TSH (NHPP AFP-1274789), α-PRL (NHPP AFP-425-10-91), α-LH (NHPP AFP-C697071P) and α-FSH (AFP-7798-1289) (Developmental Studies Hybridoma Bank) at a 1:1000 dilution. Detection of pituitary lineage commitment was performed using antibodies for α-PIT1 (a gift from S. Rhodes, Indiana University School of Medicine, Indianapolis) and α-TPIT (a gift from J. Drouin, Montreal Clinical Research Institute, Montreal) at a 1:1000 dilution. α-SF1 (434200; Thermo) was used at a dilution of 1:200. Antibodies for markers of pituitary stem cells α-SOX2 (GT15098, Immune Systems) and α-SOX9 (AB5535, Millipore) were used at a dilution of 1:250 and 1:500, respectively. Proliferation marker antibodies α-KI67 (ab16667, Abcam) and α-Cyclin D2 (SC-593, Santa Cruz) were used at a dilution of 1:100 and 1:200, respectively. Detection of activated MAPK signalling cells was performed using α-pERK1/2 (9101, Cell Signaling) at a dilution of 1:250. Cytokeratins were detected using α-pan-cytokeratin recognising cytokeratins 5, 6, 8, 17 and 19 (DMNF116; Dako M0821) at a dilution of 1:300, and an antibody specifically recognising cytokeratin 19 (DAKO clone RCK108; M088). Detection of oncogenic BRAF was performed using a α-BRAF V600E antibody (Spring Bioscience; E19290) at a dilution of 1:50.

All cell counting analysis were performed in three to five non-consecutive histological sections immunostained using the relevant antibodies. PIT1, SF1, TPIT, KI67, SOX2 and Cyclin D2 expressing populations were determined by taking the proportion of positive cells relative to total DAPI-positive nuclei. Around 4000 to 7000 DAPI-positive cells and 300 to 1000 marker-positive (PIT1+, SF1+, TPIT+, KI67+, SOX2+ and Cyclin D2+) cells were counted for each genotype. For analysis of EdU-tracing experiments, the proportions of SOX2-, PIT1- or KI67-positive cells were determined relative to the EdU-positive cell population. Around 500-1500 EdU-positive cells and 200-600 marker-positive cells were counted for each genotype. Analysis of hormone-producing cell populations (GH⁺, PRL⁺, TSH⁺, FSH⁺, LH⁺, ACTH⁺ and αGSU⁺) was assessed through determining the relative proportion of hormone-positive population to total DAPI-positive nuclei in a section. These proportions were then applied to average cell counts of dissociated pituitaries to determine an approximation of the absolute numbers of these populations in the pituitary. Around 5500-8500 DAPI-positive cells and 200-2500 marker-positive cells were counted for each genotype.

Pituitaries from Hesx1^Cre/+;Braf^V600E/+, Hesx1^Cre/+;Kras^G12D/+ and control littermates were dissected at 18.5 dpc and the posterior lobe was removed. Anterior pituitaries were dissociated mechanically into a single cell suspension following incubation in an enzyme mix (0.5% w/v collagenase type 2, Lorne Laboratories), 0.1× trypsin (Gibco) and 50 μg/ml DNaseI (Worthington) with 2.5 μg/ml Fungizone (Gibco) in Hank's balanced salt solution (HBSS, Gibco) for 4 h at 37°C. Cells were then washed in HBSS, suspended in pituitary stem cell medium and seeded at densities of 2000, 4000 and 8000 cells per well of a six-well plate for adherent culture. Cells were cultured for 3 days at which point they were fixed with 4% PFA and colonies were stained for Haematoxylin, as previously described (Gaston-Massuet et al., 2011). The proportion of colonies observed after 3 days of culture relative to seeded cells was used to estimate total clonogenic cells in 18.5 dpc pituitaries by multiplying this value by the total number of cells quantified following dissociation of the pituitary.

Pituitaries from Hesx1^Cre/+;Braf^V600E/+ or Hesx1^Cre/+;Kras^G12D/+ and control littermates were dissected at 18.5 dpc. The posterior lobe of the pituitary was discarded, and the intermediate and anterior lobes were processed for total RNA extraction using the RNeasy Micro kit (Qiagen). Approximately 1 μg of total RNA was reverse transcribed to cDNA using the iScript Reverse Transcription Kit and random hexamers (Bio-Rad). Quantitative real-time PCR reactions were run in triplicate using the iTaq SYBR Green (Bio-Rad) and repeated for a minimum of three independent samples for each genotype. Primer sequences are available on request. The absolute number of mRNA copies was determined for each sample by comparison of obtained Cq values with that of a standard of known copy number for each transcript investigated. Variations in input cDNA used for each reaction were accounted for by normalising differences in Gapdh expression between samples and genotypes, and applying this normalisation factor to absolute copy numbers obtained for each mRNA species of interest. Primer sequences can be found in Table S3.

Anonymised archival frozen and FFPE specimens of PCP were accessed through Brain UK (University of Southampton, UK). Human foetal pituitary tissue was accessed through the Human Developmental Biology Resource (HDBR, Newcastle and London, UK). Anonymised post-mortem human adult pituitary was accessed from the Medical Research Council Edinburgh Brain and Tissue bank and Royal Victoria Infirmary (Newcastle, UK) (Ethical approval 14/LO/2265).

DNA was diluted fivefold with nuclease-free water (Ambion). Each digital PCR reaction was made up to 20 µl with 2 µl diluted DNA, 10 µl ddPCR Supermix for Probes (No dUTP; Bio-Rad), forward and reverse primers, dual-labelled probes for wild-type and mutant templates, and nuclease-free water. Sequences for the BRAF-V600E assay oligonucleotides are from Hindson et al. (2011), and were used at the concentrations stated. Reactions were partitioned into ∼23,000 droplets with Droplet Generation Oil for Probes (Bio-Rad) using the QX200 Droplet Generator (Bio-Rad). All samples were tested in duplicate, with no-template controls included in all columns. Standard cycling conditions, as recommended by the manufacturer were used with a T_a of 60°C. Results were analysed with QuantaSoft software, v1.7. A mean of 19,080 droplets were successfully read in each well, including controls. Two droplets positive for wild-type template were seen in the negative controls, indicating a negligible degree of contamination.

Mendelian ratios were evaluated using the chi-squared test. Clonogenic potential of control, Hesx1^Cre/+;Kras^G12D/+ and Hesx1^Cre/+;Braf^V600E/+ pituitaries was evaluated using a paired t-test. Total cell counts of control and mutant pituitaries and quantitative real-time PCR data were analysed by unpaired t-test.

We thank the Developmental Studies Hybridoma Bank (University of Iowa) and the National Hormone and Peptide Program (Harbor–University of California, Los Angeles Medical Center) for providing some of the antibodies used in this study. The human embryonic and foetal material was provided by the Joint Medical Research Council (MRC)/Wellcome Trust Human Developmental Biology Resource (www.hdbr.org) (099175/Z/12/Z). PCP tissue samples were also obtained from Plymouth Hospitals NHS Trust as part of the UK Brain Archive Information Network (BRAIN UK), which is funded by the Medical Research Council and The Brain Trust Charity. Adult pituitary tissue was obtained from the Medical Research Council Edinburgh Brain and Tissue Bank and from Dr Abhijit Joshi (Royal Victoria Infirmary, Newcastle, UK).