Insm1 controls development of pituitary endocrine cells and requires a SNAG domain for function and for recruitment of histone-modifying factors

The pituitary gland is a central neuroendocrine organ that relays information from the central nervous system to peripheral targets by secreting hormones and other factors. The anterior pituitary gland contains six different endocrine cell types that control physiological processes, such as growth, stress response, metabolic status, reproduction and lactation (Cushman and Camper, 2001). During embryogenesis, transcription factors and signaling events control the development of these hormone-secreting cell types (reviewed by Zhu et al., 2005; Kelberman et al., 2009). The discovery of these regulatory cascades facilitated genetic analyses of patients with hormone deficiencies. All hormone-secreting cell types of the pituitary gland derive from Rathke’s pouch. Rathke’s pouch is an invagination of the oral ectoderm. Inductive interactions between surrounding tissues and Rathke’s pouch control the expression of transcription factors, such as Lhx3, Lhx4, Hesx1, Prop1 and Pitx1, that specify fates of pituitary cells (reviewed by Ericson et al., 1998; Rosenfeld et al., 2000). Differentiation of the six endocrine cell types (thyrotropes, corticotropes, melanotropes, somatotropes, lactotropes, gonadotropes) depends on transcription factors, such as Pit1 (Pou1f1), Math3 (NeuroD4), Nr5a1 (Sf1), Tbx19 and NeuroD1 (Camper et al., 1990; Shinoda et al., 1995; Pulichino et al., 2003; Lamolet et al., 2004; Zhu et al., 2006). Sox2 is expressed in Rathke’s pouch and in endocrine progenitor cells during development and in the adult, but is downregulated upon differentiation and hormone expression (Fauquier et al., 2008; Jayakody et al., 2012).

The Insm1 gene encodes a zinc finger transcription factor that is expressed in various endocrine cell types as well as in tumors that derive thereof (Goto et al., 1992; Breslin et al., 2002; Gierl et al., 2006; Mellitzer et al., 2006; Wildner et al., 2008). Insm1 contains five zinc finger domains of the C2H2 type, a motif frequently associated with transcription factors (Razin et al., 2012). At the very N-terminus, Insm1 contains a SNAG domain (Grimes et al., 1996), which was first identified in Snail/Slug and Gfi1 and Gfi1b zinc finger transcription factors and recruits Rcor1, the histone deacetylases Hdac1 and Hdac2 and the histone demethylase Kdm1a (Saleque et al., 2007; Lin et al., 2010). Rcor1 acts as a co-repressor and was first identified together with the histone deacetylases Hdac1 and Hdac2 in the REST silencing complex (Andrés et al., 1999). Together, these proteins modulate gene expression by chromatin modification and provide key epigenetic information. Zinc finger factors related to Insm1 exist in Drosophila (Nerfin-1) and Caenorhabditis elegans (EGL-46), where they function in neuronal differentiation, but the invertebrate homologs do not contain SNAG domains (Wu et al., 2001; Kuzin et al., 2005). In accordance with a potentially repressive function, Insm1 is known to bind Hdac1 and Hdac3 (Liu et al., 2006), but whether the Insm1 SNAG domain is responsible for this interaction has not been assessed.

Genetic analysis in mice demonstrated that Insm1 encodes a key factor in development of endocrine cells of pancreas, intestine and adrenal gland. Insm1 is required for differentiation but not for the specification of these endocrine cell types (Gierl et al., 2006; Wildner et al., 2008). Insm1 is expressed also in endocrine cells of the pituitary (our unpublished observations) and pituitary tumor cells (Goto et al., 1992), but a potential function in the pituitary has not been assessed. Here, we show that Insm1 expression in the pituitary is initiated when endocrine cells differentiate, and persists during further development and in the adult. Mutation of Insm1 leads to the formation of an anterior pituitary gland of normal overall morphology, but this pituitary is devoid of the entire repertoire of differentiated endocrine cell types expressing the appropriate hormones. In the mutant endocrine cells, skeletal muscle-specific genes are ectopically expressed, indicating that Insm1 suppresses an inappropriate differentiation program. We also define the role of the SNAG domain in the differentiation of pituitary endocrine cells. We used biochemistry to show that Insm1, via its SNAG domain, interacts with a number of histone-modifying factors (i.e. Kdm1a, Hdac1/2 and Rcor1-3) and with factors implicated in transcriptional regulation (Hmg20a/b and Gse1). Furthermore, deletion of the SNAG domain in vivo disrupts the differentiation of endocrine cells in the pituitary. Our work demonstrates that Insm1 acts in the epigenetic and transcriptional network that controls differentiation of endocrine cells in the anterior pituitary gland, and requires the SNAG domain to exert its role in differentiation and in suppression of inappropriate gene expression programs.

We analyzed expression of Insm1 during organogenesis of the pituitary gland by immunohistological analysis. The first Insm1⁺ cells appear at embryonic day (E) 11.5 and locate to the most ventral part of the pituitary anlage (Fig. 1A). This domain is known to contain proliferating progenitors that initiate differentiation into the early thyrotropes of the pituitary (Pope et al., 2006; Bilodeau et al., 2009). Counterstaining with DAPI demonstrated nuclear localization of the Insm1 protein (Fig. 1A, inset). At E13.5, Insm1 expression has expanded (Fig. 1B), and is detected below the progenitor zone known to contain proliferating progenitors (c.f. Bilodeau et al., 2009). By E17.5, Insm1 is present in the entire anterior pituitary (Fig. 1C,D). This indicates that Insm1 expression accompanies the spreading of endocrine cell differentiation. Sox2 and Sox9 are expressed in endocrine progenitors of the pituitary (Fauquier et al., 2008; Jayakody et al., 2012), and we detected few cells that co-express Insm1 and Sox2, or Insm1 and Sox9 at E17.5 (Fig. 1C,C′). Instead, the majority of Insm1⁺ cells co-expressed hormones, as assessed by immunohistology using a mix of antibodies directed against pituitary hormones (Fig. 1D,D′). Further analysis indicated that Insm1 was co-expressed with growth hormone (GH), prolactin (Prl), thyroid-stimulating hormone (TSH), follicle-stimulating hormone (FSH), adrenocorticotrope hormone (ACTH), as well as melanocyte-stimulating hormone (MSH) (Fig. 1E-J). Thus, Insm1 is present in all differentiated endocrine cell types, and its expression persists into adulthood (supplementary material Fig. S1A). In accordance with this, we observed only very rarely Insm1⁺ cells that expressed Ki67 at E17.5, indicating that most Insm1⁺ cells have exited the cell cycle (Fig. 1K,K′). Insm1 was not co-expressed with markers for fibroblasts (ER/TR7), folliculostellate cells (S100) or endothelia (PCAM), and is thus restricted to the endocrine lineage of the pituitary (Fig. 1L-N). This pattern indicates that Insm1 expression initiates in pituitary progenitors when differentiation of endocrine cells begins. Insm1 continues to be expressed in differentiated endocrine cells during development and in adulthood. Furthermore, Insm1 is expressed in all endocrine cell types of the pituitary (see Fig. 1O for a summary of the different endocrine cell types and their expression of Insm1).

Next, we used mouse genetics to define the role of Insm1 in the development of endocrine cells, using a previously generated null allele in which the Insm1 coding sequence is replaced by lacZ (Gierl et al., 2006). As assessed by immunohistochemistry, GH was produced at E17.5 in heterozygous Insm1^lacZ animals, which served as controls, but was not detectable in homozygous Insm1^lacZ mice (Fig. 2A,B; we refer subsequently to heterozygous and homozygous Insm1^lacZ mice also as control and Insm1^lacZ mutants). In situ hybridization demonstrated a strong reduction in Gh transcripts in Insm1^lacZ mutants (supplementary material Fig. S1B,C). Thus, somatotropes require Insm1 for their differentiation. Similarly, TSH, FSH and LH proteins were present in the pituitary of control animals, but were not detectable in Insm1^lacZ mutants (Fig. 2C-H), and the corresponding mRNAs, as well as the mRNA encoding the common α-subunit of the three hormones, Cga, were strongly downregulated (supplementary material Fig. S1D-K). Thus, thyrotropes and gonadotropes also require Insm1 for differentiation. Lactotropes express Prl, which was markedly downregulated in Insm1^lacZ mutants compared with control animals (Fig. 2I,J). Finally, melanotropes and corticotropes produce MSH and ACTH, respectively, which are generated from a common precursor protein, pro-opiomelanocortin (POMC). Antibodies that detect mature MSH and ACTH demonstrated a strong reduction of these hormones in pituitaries of homozygous Insm1^lacZ mutants (Fig. 2K-N), whereas Pomc mRNA was mildly downregulated (supplementary material Fig. S1L,M). A further hallmark of endocrine cell differentiation is the production of granin proteins, which are associated with secretory vesicles. Chromogranin A (Chga) transcripts were present at reduced levels in the pituitary of homozygous Insm1^lacZ mutants compared with control mice (Fig. 2O,P). We conclude that none of the endocrine cell types of the pituitary differentiates correctly in Insm1^lacZ mutants, which is reflected by a lack or marked downregulation of hormones and proteins associated with secretory vesicles. The deficit in differentiation of endocrine cells was accompanied by a marked increase in the number of cells expressing Sox2 and Sox9 (Fig. 3A-D′). The supernumerary Sox2⁺ and Sox9⁺ cells co-expressed β-gal in Insm1^lacZ mutants, and we noted that β-gal⁺ nuclei in the anterior lobe of the pituitary were packed more densely in mutant than in control mice (Fig. 3A-D′). Thus, the disrupted differentiation of endocrine cells is accompanied by an increase in the number of endocrine progenitor cells. We did not detect obvious changes in the distribution or morphology of other cells types of the pituitary (fibroblasts, folliculostellate cells, endothelia) in Insm1^lacZ mutant mice (Fig. 3B,B′,F,F′).

We next investigated whether the disrupted differentiation affected proliferation and survival of endocrine cells in the pituitary. Cell proliferation was assessed by BrdU incorporation, which revealed no significant changes when control and Insm1^lacZ mutants were compared (supplementary material Fig. S2A-C). Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining did not detect apoptosis in the pituitary at E13.5 or E15.5 (supplementary material Fig. S2D-G). The overall shape of pituitary glands of control and homozygous mutant mice were similar at E17.5 (Fig. 3G,H). Nissl-stained consecutive sections of the pituitary were used to estimate the size of the pituitary at E17.5, which detected a small decrease (∼10%) in mutant mice (Fig. 3I-K). Thus, the pituitary size is reduced mildly in Insm1^lacZ mutants, apparently owing to a denser packing of cells. We conclude that the mutation of Insm1 interferes in a striking manner with endocrine differentiation but not cell cycle exit in the pituitary.

To assess global changes in gene expression, we performed gene profiling of control and Insm1^lacZ mutant pituitary glands using Illumina microarrays. This confirmed the marked downregulation of genes encoding pituitary hormones, and demonstrated that many other genes that function in hormone production and secretion were also downregulated, e.g. proprotein convertases and various granins, such as secretogranin II/III (Table 1). Interestingly, comparison with previous experiments that analyzed genes deregulated in the adrenal gland and pancreas of Insm1^lacZ mutants (Gierl et al., 2006; Wildner et al., 2008) detected a set of genes that are deregulated in all three endocrine organs (Table 2; supplementary material Fig. S5). Thus, Insm1 functions in many endocrine cell types and its mutation affects the expression of an overlapping gene set in different endocrine organs.

Remarkably, further analysis of the microarray data showed that a number of genes of the Notch pathway were upregulated in the pituitary of Insm1^lacZ mutant mice (Table 1). Among these were Dll1 and the Notch target genes Hey1, Hes1 and Hes5. We confirmed the enhanced expression of Dll1 and Hey1 using in situ hybridization and qPCR (Fig. 4A-F,K). Notch2 is expressed during pituitary development (Raetzman et al., 2004), and in situ hybridization demonstrated a pronounced upregulation of Notch2 expression in the Insm1^lacZ mutant pituitary at E15.5; at E17.5, altered Notch2 expression was detected by qPCR but not by in situ or microarray experiments.

In addition, the array analysis identified a marked upregulation of a number of genes typically expressed in skeletal muscle, e.g. mRNAs encoding myosin light polypeptide 1 (Myl1), alpha cardiac muscle 1 actin (Actc1) expressed in skeletal and cardiac muscle, and musculin (Msc), a myogenic transcription factor (Table 1). Upregulated expression of Myl1 and Actc1 mRNAs in a subpopulation of cells was confirmed by in situ hybridization and qPCR (Fig. 4G-K). We conclude from these data that Insm1 is required for repression of components of the Notch signaling pathway and of skeletal muscle-specific genes in endocrine cells of the pituitary.

The development of different endocrine cell types depends on a number of transcription factors. In particular, somatotrope, lactotrope and thyrotrope lineages depend on Pit1 for differentiation (Camper et al., 1990), and Pit1 expression was indeed not detected at E14.5 and was markedly downregulated at E17.5 in Insm1^lacZ mutants (Fig. 5A-D). Math3 controls differentiation of somatotropes and lactotropes, and Nr5a1 the differentiation of gonadotropes (Shinoda et al., 1995; Zhu et al., 2006). Math3 and Nr5a1 transcripts were markedly downregulated in Insm1^lacZ mutant mice (Fig. 5E-H). Tbx19 and NeuroD1 mark differentiating corticotropes (Tbx19, NeuroD1) and melanotropes (Tbx19) and control their differentiation as well as POMC expression (Lamolet et al., 2001; Pulichino et al., 2003; Lamolet et al., 2004); expression of Tbx19 and NeuroD1 was downregulated in mutant mice (Fig. 5I-L). In conclusion, transcription factors that direct the differentiation of endocrine cell types of the pituitary are not correctly expressed in Insm1^lacZ mutant mice. We also assessed the early specification of endocrine cells by analyzing transcription factors, such as Hesx1, Pitx1, Pitx2, Lhx3 and Prop1, which are known to be essential for pituitary development (Sheng et al., 1996; Dattani et al., 1998; Wu et al., 1998; Szeto et al., 1999; Suh et al., 2002). No obvious change in the expression of the corresponding transcripts was detected at E12.5 (supplementary material Fig. S3A-P). Prop1 expression was compared at various stages, and was unchanged at E12.5 and mildly upregulated at E14.5 and E15.5 (supplementary material Fig. S3I-P). We conclude that in the absence of Insm1, endocrine progenitor cells of the pituitary are correctly specified, but their differentiation is disrupted.

To discover Insm1-interacting proteins, we combined stable isotope labeling by amino acids in cell culture (SILAC) and affinity purification, an approach that can identify protein-protein interactions with very high confidence (Selbach and Mann, 2006; Paul et al., 2011). This technology relies on the quantification of proteins that co-immunoprecipitate with Insm1. The proteome of AtT-20 cells, which is a mouse corticotrope tumor cell line, was labeled with amino acids carrying either heavy or light stable isotopes (‘heavy’ and ‘light’ cells). ‘Heavy’ and ‘light’ cells were transduced with retroviruses encoding non-tagged Insm1 (Insm1^wt) and FLAG-tagged Insm1 (Insm1^FLAG), respectively. Immunoprecipitates obtained from the ‘heavy’ and ‘light’ cells using anti-FLAG antibodies were combined and subjected to mass spectrometry-based quantification (see Fig. 6A for an outline of the experiment). As expected, mass spectrometry found Insm1 among the proteins most abundant for the ‘light’-label state when anti-FLAG immunoprecipitates from ‘heavy’-labeled Insm1^wt and ‘light’-labeled Insm1^FLAG cells were compared. In addition, Kdm1a, Rcor1-3, Hdac1/2, Gse1 and Hmg20a/b were identified and quantified with similar protein ratios (Table 3). Swapping isotope labels resulted in reciprocal abundance ratios for Insm1 and detected co-precipitated proteins, supporting specificity of the interaction. Western blot analysis was used to confirm these interactions, using protein extracts from AtT-20 cells and from a second pituitary cell line, the rat somatotrope line GH3. In these experiments, Kdm1a, Rcor1 and Hdac1/2 were co-precipitated with endogenous Insm1 (Fig. 6B).

We next tested biochemically whether the SNAG domain of Insm1 is essential for the interaction with Kdm1a, Rcor1 and Hdac1/2. For this, cDNAs encoding FLAG-tagged Insm1 with intact SNAG domain (Insm1^FLAG) and a truncated FLAG-tagged variant that lacks seven highly conserved amino acids at the N-terminus were constructed (Insm1ΔSNAG^FLAG); the deletion of these seven amino acids disrupts the SNAG domain of Insm1 (Fig. 6C). Insm1^FLAG and Insm1ΔSNAG^FLAG cDNAs were expressed in AtT-20 cells and immunoprecipitated with anti-FLAG and anti-Insm1 antibodies (Fig. 6D). Kdm1a, Rcor1 and Hdac1/2 were no longer co-immunoprecipitated with Insm1ΔSNAG^FLAG, but were present in Insm1^FLAG precipitates (Fig. 6D). Similarly, Kdm1a, Rcor1 and Hdac1/2 antibodies co-immunoprecipitated full-length Insm1^FLAG, but not Insm1ΔSNAG^FLAG (Fig. 6E). Thus, Kdm1a, Rcor1 and Hdac1/2 are recruited to Insm1 via its N-terminal SNAG domain.

To assess the function of the Insm1 SNAG domain in the developing pituitary, we generated a mouse strain in which wild-type Insm1 was replaced by a sequence encoding Insm1ΔSNAG by the use of homologous recombination in embryonic stem cells and blastocyst injections (supplementary material Fig. S4). We compared Insm1 protein in pituitary extracts from mice with the genotypes Insm1^ΔSNAG/+ (also called control) and Insm1^ΔSNAG/lacZ (also called Insm1^ΔSNAG mutants), using anti-Insm1 and anti-FLAG antibodies. This demonstrated that the Insm1ΔSNAG protein was present and stable, and it was detected even at mildly increased amounts in the pituitaries of Insm1^ΔSNAG mutants compared with control mice (Fig. 7A). To test whether this is caused by transcriptional autoregulation, we quantified lacZ transcripts by qPCR and found upregulated lacZ expression in Insm1^ΔSNAG/lacZ compared with Insm1^lacZ/+ pituitaries mice (Fig. 7B). The heterozygous Insm1^ΔSNAG animals that were used as controls were viable and fertile and displayed no overt phenotype.

We next compared pituitary differentiation in control and Insm1^ΔSNAG mutant mice. None of the endocrine cell types of the pituitary differentiated correctly in Insm1^ΔSNAG mutants. In particular, we did not detect GH, TSH and FSH proteins in the pituitary (Fig. 7C-H). Furthermore, LH, Prl, MSH and ACTH were absent or present at reduced levels (Fig. 7I-P). The expression of pan-endocrine genes, such as proprotein convertases and granins, were compared using qPCR and were downregulated to similar extents in the pituitary of Insm1^lacZ and Insm1^ΔSNAG mutants (Fig. 7U). By contrast, muscle-specific genes, such as Myl1, Actc1 and Msc were upregulated to a similar extent, as assessed by in situ hybridization and/or qPCR (Fig. 7Q-T,V). qPCR analysis also demonstrated that Dll1, Notch2 and the Notch target genes Hey1, Hes1 and Hes5 were upregulated to a similar extent in the pituitary of Insm1^ΔSNAG and Insm1^lacZ mutants (Fig. 7W). Overall, our data show that the SNAG domain of Insm1 is essential for its function during the differentiation of pituitary endocrine cell types.

The fact that mutant endocrine cells continue to express the truncated Insm1ΔSNAG protein allowed us to use anti-Insm1 antibodies to analyze and quantify cells co-expressing (wild-type or truncated) Insm1, Sox2 or Sox9 protein in control and Insm1^ΔSNAG mutants. First, we quantified overall numbers of Sox2⁺, Sox9⁺ and Insm1⁺ cells in the anterior lobe of the pituitary (supplementary material Fig. S6). This demonstrated a significant increase in the number of Sox2⁺ and Sox9⁺, but not Insm1⁺ cells. Furthermore, the proportion of Sox2⁺ or Sox9⁺ cells co-expressing Insm1 was increased in Insm1^ΔSNAG mutants compared with control mice (Fig. 8A-F), providing further evidence that the mutant cells remain in a progenitor state for a prolonged period. Interestingly, a combination of in situ hybridization and immunohistology demonstrated that a subpopulation of cells expressing the truncated Insm1 protein co-express Myl1 (Fig. 8G,H). Thus, Insm1 acts cell-autonomously when repressing the expression of muscle-specific genes in the pituitary.

We show here that all endocrine cell types of the pituitary express Insm1 and depend on Insm1 for differentiation. Early specification of endocrine cells proceeded correctly, whereas transcription factors that control differentiation of endocrine cells were not correctly expressed in Insm1 mutants. Furthermore, the endocrine differentiation program was disrupted in Insm1 mutants, and hormones, proprotein convertases and granins were not expressed or expressed at markedly reduced levels. In addition, we noted prolonged expression of progenitor markers (Sox2/9), as well as ectopic expression of Notch signaling components and of genes typical for the muscle lineage in the Insm1 mutant pituitary (summarized in Fig. 8I). Our previous work has demonstrated a severe disruption of the differentiation of endocrine cells in pancreas, intestine and adrenal medulla (Gierl et al., 2006; Wildner et al., 2008). A comparison of the deregulated genes in the pancreas, adrenal and pituitary glands detects a small set of overlapping genes (Table 2). Insm1 is thus emerging as a pan-endocrine differentiation factor.

The SNAG domain consists of fewer than ten amino acids located at the immediate N-terminus of the Insm1 protein. We use here mouse genetics to show that this domain is essential for Insm1 function. The SNAG domain is found in other transcription factors, such as Gfi1 and Snail1 (Snai1), two well characterized zinc finger factors that act as transcriptional regulators (Hock and Orkin, 2006; Moreno-Bueno et al., 2008; Möröy and Khandanpour, 2011). We show in biochemical experiments that the SNAG domain of Insm1 recruits a battery of chromatin modifying enzymes, such as Kdm1a, Rcor1-3 and Hdac1/2, as well as additional factors, such as Gse1 and Hmg20a/b.

Snail1 is a repressor of E-cadherin (cadherin 1) gene expression and controls epithelial-mesenchymal transition of cells, for instance in tumorigenesis (Cano et al., 2000). The SNAG domain of Snail1 also recruits Kdm1a, Rcor1/2 and Hdac1/2; it resembles the tail of histone H3, and the active site of Kdm1a binds the histone H3 tail and the SNAG domain (Lin et al., 2010). To the Snail1-Kdm1a complex, Rcor is recruited, and the formation of the ternary complex stabilizes Snail1 (Lin et al., 2010). By contrast, the steady state levels of Insm1ΔSNAG are not markedly lowered, indicating that Insm1 stability does not depend on the formation of the ternary Insm1-Rcor-Kdm1a complex. The transcription factor Gfi1 is essential for hematopoietic differentiation, and a single point mutation in the SNAG domain of Gfi1 interferes with Gfi1 function in hematopoietic differentiation as well as in recruitment of Kdm1a, Rcor1/2 and Hdac1/2 (Saleque et al., 2007). Thus, the function of the SNAG domains of Insm1, Snail1 and Gfi1 are conserved, and they act as hub to recruit chromatin-modifying factors. The recruitment of these modifying factors is essential for the function of these zinc finger transcription factors.

Our analysis of pituitary development in Insm1^lacZ mutant mice indicates that the differentiation of the different endocrine cell types is disrupted and that neither the hormones nor pan-endocrine-specific genes, such as proprotein convertases or granins, are expressed correctly. However, morphogenesis of the pituitary remains intact and, for instance, endocrine progenitors leave the proliferative zone in the Insm1 mutant pituitary. This requires an epithelial-mesenchymal transition, which thus occurs in the absence of Insm1. The first changes that we detect during development of endocrine cells in the pituitary of Insm1^lacZ mutants are associated with the failure to initiate expression of transcription factors that control the lineage-specific differentiation (e.g. Pit1, Sp1, Math3). Insm1 might thus participate in the activation of such genes or indirectly preclude the expression of such factors. We also observed upregulated expression of the genes encoding Notch2, the Notch ligand Dll1 and of target genes of the Notch signaling pathway, such as Hey1, Hes1 and Hes5. In normal development, the Notch2 and Notch3 receptors and Hes1 are expressed in progenitor cells, but are excluded from differentiating endocrine cells (Raetzman et al., 2004). A tight control of Notch signaling is required for pituitary development, and sustained Notch activation in progenitor cells by expression of Notch1/2 intracellular domains disrupts and dramatically delays endocrine differentiation (Raetzman et al., 2006; Zhu et al., 2006). Furthermore, the maintenance of progenitors and formation of melanotropes depend on the Notch target genes Hes1 and Hes5 (Kita et al., 2007; Raetzman et al., 2007). Upregulated expression of Notch signaling components is associated with a prolonged expression of Sox2/9, indicating that the Insm1 mutant endocrine cells remain in a progenitor stage. The Insm1^lacZ mutation arrests differentiation of pituitary endocrine cells in fetal development, but the lethal phenotype of the mutation precludes analysis during postnatal stages. Thus, it is possible that by upregulating Notch signaling, the Insm1^lacZ mutation merely delays differentiation, as described for the Notch2 gain-of-function mutation (Raetzman et al., 2006). Upregulation of the Notch signaling pathway, particularly of Dll1 and Hes1, was also noted in our previous analysis on the development of the sympathetic nervous system in Insm1^lacZ mutant mice (Wildner et al., 2008). The fact that we observe upregulated Dll1 and Hes1 expression in several tissues might indicate that activation of Notch is a direct and primary event in the deregulation of gene expression caused by the Insm1^lacZ mutation.

Deletion of sequences encoding the SNAG domain of Insm1 (Insm1^ΔSNAG) and the null mutation of Insm1 (Insm1^lacZ) have indistinguishable effects on the development of endocrine cells of the pituitary, indicating that the recruitment of proteins by the SNAG domain is essential for Insm1 function. Remarkably, conditional mutation of Kdm1a (also known as Lsd1) in the entire anlage of the pituitary results in stunningly similar deficits and, like the Insm1 mutation, blocks differentiation of all endocrine cell types of the pituitary (Wang et al., 2007). Furthermore, the expression of Prop1 but not Pit1 is correctly initiated, and Hey1 is upregulated in Kdm1a and Insm1 mutant pituitaries (this work) (Wang et al., 2007). Kdm1a acts as a histone demethylase that demethylates both Lys-4 (H3K4me) and Lys-9 (H3K9me) of histone H3 and acts thus in a context-dependent manner as co-activator or co-repressor (Lee et al., 2005; Metzger et al., 2005). It is possible that Insm1 also takes over such dual roles in the pituitary. In support for a function as a repressor, we observed many genes that were upregulated, among them Dll1 and Notch target genes. Interestingly, among the top 20 upregulated genes, nine are strongly expressed in muscle tissue in normal development. Muscle and pituitary derive from mesoderm and ectoderm, respectively. Activation of particular gene expression programs that are characteristic of different germ layers at such late developmental stages is unusual. It should be noted that this ectopic expression in the pituitary did not reach levels comparable to those observed in mature skeletal muscle, and the changed expression does therefore not represent a complete fate shift. The data indicate that Insm1 participates in the repression of inappropriate differentiation programs. We conclude that Insm1 recruits chromatin-modifying factors and thus directs differentiation of all endocrine cell types of the pituitary.

Generation and genotyping of Insm1^lacZ mutant mice were carried out as described (Gierl et al., 2006). The Insm1ΔSNAG targeting vector was generated by homologous recombination in bacteria (Lee et al., 2001); in particular, the Insm1 coding sequence was replaced by an Insm1ΔSNAG sequence that lacks the seven N-terminal amino acids (Fig. 6C; supplementary material Fig. S4) and contains a C-terminal 3xFLAG-tag (Sigma-Aldrich) sequence. The targeting vector contained a self-excision neo cassette (Bunting et al., 1999). R1/E embryonic stem cells were electroporated and colonies that had incorporated the targeting vector into their genome were selected by G418 and analyzed for homologous recombination by Southern blot analysis. Chimeric mice obtained after blastocyst injection were mated to C57BL/6 mice for transmission of the Insm1ΔSNAG allele. Routine genotyping was performed by PCR using the following primers: Insm1ΔSNAG-fw CCCAACCAGTGCGTCGCCCTT and Insm1ΔSNAG-rv AGCAGCGCCCGGTCACTGTCC.

For in situ hybridization, tissue was embedded into OCT compound directly after dissection. Digoxygenin-labeled riboprobes, which were used for hybridization, were generated by in vitro transcription. β-Galactosidase activity was assessed by X-Gal staining as described previously (Lobe et al., 1999). For immunohistochemistry, tissue was fixed with 4% paraformaldehyde in 0.1 M sodium phosphate buffer (pH 7.4) and cryoprotected in 25% sucrose in PBS prior to embedding, and cryosections were cut at 14 μm thickness. The following primary antibodies were used: rabbit/guinea pig anti-Insm1 (1:10,000; antibodies were raised against an antigen produced in Escherichia coli and the plasmid used for production of the antigen was kindly provided by G. Gradwohl, IGBMC Illkirch, France), rabbit anti-ACTH (1:2000; AFP15610278912, NHPP, Torrance, USA), chicken anti-β-galactosidase (1:5000; Ab9361, Abcam), rat anti-BrdU antibodies (1:200; AbD Serotec), rat-anti ER/TR7 (1:500; Ab51829, Abcam), rabbit anti-FSH (1:2000; AFP7798-1289, NHPP), rabbit anti-GH (1:1000; Ab940, Merck-Millipore), rat anti-Ki67 (1:500; M724901-8, Dako Cytomation), rabbit anti-LH (1:2000; AFPC697071P, NHPP), rabbit anti-MSH (1:1000; Peninsula Laboratories LLC), goat anti-NeuroD1 (1:500; sc-1084, Santa Cruz), rat anti-PECAM (1:300; 553930, BD Pharmingen), rabbit anti-Prl (1:2000; AFP131078rb, NHPP), rabbit anti-S100 (1:1000; Z0311, Dako), goat anti-Sox2 (1:500; GT15098, Immune Systems), rabbit anti-Sox9 (1:2500; Ab5535, Merck-Millipore) and rabbit anti-TSH (1:1000; Ab976, Merck-Millipore). Secondary antibodies conjugated to Cy2, Cy3 or Cy5 were used at a dilution of 1:500 (Jackson ImmunoResearch). All quantifications from histological sections were performed on at least three sections of three or more animals/genotype. Unpaired Student’s two-tailed t-tests were performed using Graphpad Prism 5 to determine significance. Average and s.e.m. are displayed. For visualization of Myl1 and Insm1ΔSNAG co-expression, in situ hybridization was combined with immunohistochemistry. TUNEL was performed by the use of a kit (57110, Merck-Millipore) according to the manufacturer’s instructions. DAPI was used as a nuclear counterstain. Fluorescence was imaged on a Zeiss LSM 700 confocal microscope and images were processed using Adobe Photoshop software.

For the estimation of the volume and area of the pituitary gland, 14-μm-thick sagittal sections were collected and Nissl stained. A total of 75 sections of the pituitary gland were obtained per animal (three animals/genotype). The area of every third section was measured using imaging analysis software (ImageJ; NIH, version 1.34n). Estimation of the total volume of the pituitary gland was obtained by application of Cavalieri’s method (West, 2012).

E17.5 pituitary glands were dissected and total RNA was isolated using Trizol reagent (Invitrogen). qRT-PCR analysis after first-strand cDNA synthesis was performed using a CFX96 RT-PCR system (Bio-Rad), ABsolute QPCR SYBR Green Mix (Thermo Fisher Scientific) and the primers are listed in supplementary material Table S1. Total RNA from pituitary glands was labeled and hybridized to MouseRef-8 v2.0 Expression BeadChips (Illumina, San Diego, CA, USA) according to the manufacturer’s protocol.

All microarray datasets were analyzed using R Bioconductor package (Gentleman et al., 2004). To compare samples, we applied log2 transformation and normalized using the ‘quantile’ method, and adjusted P-values were obtained using the Benjamini-Hochberg FDR method (cutoff P<0.05 for all cell types). Expression data of control and Insm1 mutant pituitary glands have been deposited in NCBI’s Gene Expression Omnibus (Edgar et al., 2002) and are accessible through accession number GSE46139.

AtT-20 and GH3 cells were grown in high-glucose DMEM (Gibco/Invitrogen) containing 10% fetal calf serum (FCS; Sigma-Aldrich). For SILAC experiments, AtT-20 cells were grown in SILAC DMEM (PAA Laboratories, Velizy-Villacoublay, France), 10% dialyzed FCS (SAFC), 4 mM glutamine and amino acids labeled either with ‘light’ [¹²C;¹⁴N]arginine/lysine (Sigma-Aldrich) or ‘heavy’ [¹³C;¹⁵N]arginine/lysine (Sigma-Aldrich) isotopes for four passages.

For immunoprecipitation, cells were washed with PBS, harvested in lysis buffer containing 50 mM Tris pH 7.4, 140 mM NaCl, 1% Triton X-100, complete protease inhibitors (Roche) and phosphatase inhibitor cocktails 2 and 3 (Sigma-Aldrich). After pre-clearing, supernatants were incubated with antibodies conjugated and covalently coupled to Dynabeads ProteinA/G for 3 hours at 4°C. Precipitates of the corresponding ‘heavy’ and ‘light’ pull-downs were combined and washed three times with lysis buffer. The last wash was performed with 5 mM Tris pH 7.4, 140 mM NaCl. Bound proteins were eluted with 3×100 μl 100 mM glycine pH 2.5 and subsequently precipitated by adding 1 μl GlycoBlue (Ambion/Life Technologies), 80 μl 2.5 M sodium acetate pH 5.0 and 1500 μl ethanol. Sedimentation was carried out at 20,000 g for 30 minutes. Protein pellets were dissolved in 6 M urea/2 M thiourea/10 mM HEPES (pH 8.0), reduced with 1 mM dithiothreitol, alkylated with 5.5 mM iodoacetamid and subsequently digested by lysyl-endopeptidase (Lys-C, Wako Chemicals) and sequencing grade modified trypsin (Promega) in 50 mM ammonium bicarbonate following standard procedures (Shevchenko et al., 2007). Peptides were desalted by stop and go extraction (STAGE) tips containing C18 Empore disks (3M) (Rappsilber et al., 2003). Peptide mixtures were analyzed by C18 reversed phase nanoflow HPLC on an Eksigent nanoLC-1D Plus system (Eksigent Technologies) coupled to an LTQ-Orbitrap XL mass spectrometer (Thermo Fisher Scientific). Mass spectrometers were operated in the data-dependent mode with a TOP5 method (Olsen et al., 2005). Raw data files were processed using the MaxQuant software platform with standard settings (Cox and Mann, 2008; Cox et al., 2009). Peak lists were searched against an in-house curated mouse International Protein Index (IPI) human protein database using MASCOT (Matrix Science, Boston, USA) and a database containing common contaminants. Proteins with one unique peptide were considered identified, and peptide and protein false discovery rates were restricted to 0.01.

For western blot analysis, immunoprecipitated proteins were eluted from beads by boiling in 1 × Laemmli buffer for 5 minutes and then subjected to SDS-PAGE. Proteins were transferred to PVDF membranes. The following antibodies were used for western blot analysis: rabbit anti-Kdm1a (1:1000, #2139, Cell Signaling Technology), goat anti-CoREST (1:200, sc-23448, Santa Cruz), rabbit anti-Hdac1 (1:1000, #2062, Cell Signaling Technology), rabbit anti-Hdac2 (1:2000, ab32117, Merck-Millipore), rabbit/guinea pig anti-Insm1 (1:6000), mouse anti-tubulin (1:10,000, T9026, Sigma-Aldrich), mouse anti-Flag (1:1000, F1804, Sigma-Aldrich). Secondary antibodies coupled to horseradish peroxidase (Dianova, Hamburg, Germany) were used, and blots were developed on a Chemi-smart 3000 (Vilber, Eberhardzell, Germany). Wild-type and modified Insm1 cDNAs [Insm1^FLAG (Insm1 containing 3xFLAG-tag, Sigma-Aldrich), Insm1ΔSNAG^FLAG (Insm1ΔSNAG containing 3xFLAG-tag, Sigma-Aldrich)] were expressed in AtT-20 cells using a MMULV-based system developed by the Nolan lab (http://www.stanford.edu/group/nolan/retroviral_systems/phx.html).

We particularly thank Ivonne Schiffner, Bettina Brandt and Karolin Fuchs for technical assistance; Claudia Päseler and Petra Stallerow for help with the animal husbandry; Walter Birchmeier, Mathias Treier, Thomas Müller and Dominique Bröhl for critical discussions and for reading the manuscript; and Hagen Wende (all MDC, Berlin) for help with the analysis of microarray data. We are very grateful to Ronald Naumann and his team (MPI-CBG Dresden, Germany) for the gift of R1/E ES cells and blastocyst injections.