The selector gene Pax7 dictates alternate pituitary cell fates through its pioneer action on chromatin remodeling

Pax7 is specifically expressed in pituitary melanotropes

Expression profiling of transcription factors during pituitary ontogenesis led us to identify Pax7 for its preferential expression in the IL. Analysis of mouse pituitary RNA by RT-qPCR confirmed the highly restricted expression of Pax7 in the adult IL compared with ALs (Fig. 1A). Furthermore, Pax7 mRNA was detectable in E16 pituitaries but not at E14, in agreement with IL melanotrope differentiation within that developmental interval. By immunohistochemistry (IHC), Pax7 was found only in IL melanotropes and no other pituitary cells (Fig. 1B). In order to correlate Pax7 expression with cell differentiation in the developing pituitary, its expression was assessed by immunohistofluorescence together with Tpit. At E14.5 of mouse embryonic development, Tpit is already expressed in AL corticotropes, but neither Tpit nor Pax7 is expressed in ILs (Fig. 1C). By the next day (E15.5), most IL nuclei express Pax7, and a subset of those express Tpit, indicating that Pax7 expression precedes Tpit (Fig. 1E). At E16.5, most IL cells coexpress Pax7 and Tpit (Fig. 1G). The expression of these transcription factors is correlated with expression of POMC in both the AL and IL (Fig. 1D,F,H). In addition, IL expression of PC2 starts at E15.5, following expression of Tpit (Fig. 1F). PC2-positive cells of the AL at E14.5–E16.5 are presumptive gonadotropes. The onset of Pax7 expression in IL melanotropes thus precedes the terminal differentiation revealed by Tpit, POMC, and PC2 expression.

Pax7 in human/dog pituitary and corticotrope adenomas (Cushing's disease)

The human adult pituitary does not have an IL because this tissue regresses after the 15th week of gestation; it is nonetheless interesting to find out whether Pax7-expressing pituitary cells are present in humans, since their POMC cells expand to form adenomas that cause Cushing's disease, and this results in excessive and uncontrolled ACTH and glucocorticoid (Gc) production (Drouin et al. 2007). Typically, these adenomas are resistant to Gc feedback, and hence they may be of IL or AL origin if corticotropes have developed Gc resistance (Bilodeau et al. 2006). We identified Pax7 and Tpit-positive cells in IL remnants of a 25-wk human embryo (Fig. 1I,J), indicating that the human fetal IL normally expresses Pax7. Cushing's disease is more frequent in dogs than in humans, and hence we showed that the dog pituitary has a distribution of Pax7-expressing IL cells similar to the mouse (Fig. 1K). Using a panel of human and dog Cushing adenomas that we previously investigated for mechanisms of Gc resistance (Bilodeau et al. 2006; Roussel-Gervais et al. 2010), we show that ∼10% of human and 30% of dog Cushing adenomas express Pax7 with varying penetrance (Fig. 1L; Supplemental Fig. S1). It will be interesting to investigate the clinical significance of Pax7 expression in this subset of adenomas and, in particular, its putative usefulness for the development of therapeutic strategies that rely on melanotrope-specific functions, such as expression of dopamine receptors.

Pax7-dependent expression of melanotrope-specific markers

The highly restricted expression profile suggests that Pax7 might be an important regulator of the melanotrope cell fate. We addressed this question using Pax7 knockout mice. These mice die during the perinatal period if bred in a C57bl/6 background. On a 129/sv background, however, they survive up to 30 d after birth but die shortly after for as-yet-to-be-determined reasons (Kuang et al. 2006). At that age, homozygous (−/−) mice are much smaller than their wild-type or heterozygous (+/−) littermates and, accordingly, display reduced body weight. This general size reduction affected the pituitary roughly within the same proportion (Supplemental Fig. S2A).

Lineage markers were used in order to assess the differentiation status of IL cells in Pax7^−/− pituitaries. Expression of the POMC cell-restricted factor Tpit is unaffected by the Pax7 mutation, suggesting that IL cells are still directed toward a POMC lineage (Fig. 2, cf. D and A). However, expression of POMC as assessed by IHC is markedly reduced in the IL but not in AL corticotropes (Fig. 2B,E). Similarly, expression of PC2 is down to background level in IL Pax7^−/− pituitaries, while its expression is unaffected in AL gonadotropes (Fig. 2C,F). We used postnatal day 24 (P24) dissected neurointermediate lobes (NILs) from normal and Pax7^−/− pituitaries in order to quantify by RT-qPCR the effect of Pax7 deficiency on melanotrope markers (Fig. 2G). These analyses confirm that Tpit expression is not altered by the Pax7 mutation. In contrast, POMC mRNA levels are decreased 8.6-fold in Pax7^−/− NILs compared with either wild-type or Pax7^+/− pituitaries. An even greater loss of PC2 expression (56-fold) was observed in Pax7^−/− NILs with a slight decrease in PC2 mRNA levels in Pax7^+/− NILs. And similarly, Drd2 expression is 13-fold lower in Pax7^−/− NILs compared with wild type. Collectively, these analyses indicate a loss of melanotrope-specific gene expression (POMC, PC2, and Drd2) but without affecting expression of the lineage marker Tpit.

Melanotropes switch to a corticotrope fate in the absence of Pax7

The Pax7^−/− IL cells still express POMC but have lost important melanotrope features. They could be blocked in a partially differentiated state or may have switched to another cell fate. In order to address this, we first ruled out expression of other pituitary hormones in the Pax7^−/− IL (Supplemental Fig. S2B). Alternatively, the POMC-expressing cells of the Pax7^−/− IL may have adopted a corticotrope-like fate. AL corticotropes are marked by expression of the Gc receptor (GR) in contrast to the GR-negative melanotropes; we used IHC to reveal the GR in pituitary sections. As expected, the GR is widely expressed in the AL but not in the IL of normal pituitaries. In contrast, the GR is readily detectable in the Pax7^−/− IL (Fig. 2H). We used RT-qPCR analysis of mRNA from wild-type and Pax7^−/− NILs in order to quantify expression of corticotrope markers. These analyses indicated a fourfold up-regulation of the GR mRNA in the Pax7^−/− NILs compared with wild type or heterozygotes (Fig. 2I). The basic helix–loop–helix (bHLH) transcription factor NeuroD1 is restricted to pituitary AL cells—in particular, corticotropes—during development (Drouin 2010). Interestingly, NeuroD1 expression was increased 14-fold in the Pax7^−/− NIL, consistent with the idea that the Pax7^−/− melanotropes might have acquired corticotrope features. This idea was further supported by assessment of CRH receptor 1 (CRH-R1) and AVP receptor 1b (AVP-R1b) expression in the mutant NIL (Fig. 2I). Expression of these two membrane receptors is normally restricted to AL corticotropes, and their significant up-regulation in the Pax7 mutant NIL supports the interpretation that Pax7^−/− IL cells have switched to a corticotrope fate.

In order to better appreciate the extent of this fate change, we compared these results with expression data obtained from purified corticotropes and melanotropes. Expression profiles derived using FACS-sorted cells from POMC-EGFP transgenic mice (Lavoie et al. 2008) were used to select a panel of genes that are enriched in either corticotropes or melanotropes (Fig. 2J). Including the known markers mentioned above, we considered 26 expressed genes. Among these, 13 genes showed preferential expression in corticotropes (cortico-enriched) (blue in Fig. 2J). Concordantly, 11 of these showed higher expression in Pax7^−/− than in wild-type ILs (Fig. 2J). Of the 12 genes that are preferentially expressed in melanotropes (melano-enriched) (yellow in Fig. 2J), eight displayed decreased expression in Pax7^−/− ILs. Expression of the remaining four melanotrope-enriched genes (Gpr50, Gldc, Scn7a, and Fndc3cl) was not affected by loss of Pax7, suggesting that there may be other determinants of melanotrope identity in addition to Pax7. Other transcription factors or transcriptional coregulators may thus contribute together with Pax7 to the expression of melanotrope-specific genes; this may also account for the quantitative rather than absolute changes in expression profiles observed in the Pax7^−/− IL. Notwithstanding, the switch of expression signature observed in the Pax7^−/− IL supports the conclusion that Pax7 is an important determinant of melanotrope versus corticotrope cell identity.

Relative roles of Pax7 and Tpit in differentiation

In contrast to the role of Pax7 as a positive factor for the melanotrope program and a negative regulator of the corticotrope fate, Tpit is essential for differentiation into both cell fates, since Tpit^−/− pituitaries fail to express POMC, and cells destined for these lineages mostly remain undifferentiated (Pulichino et al. 2003b). An upstream action of Tpit would be consistent with its maintained expression in Pax7^−/− pituitaries (Fig. 2A,D), and accordingly, expression of Pax7 is decreased, but still present, in developing ILs of Tpit^−/− E16.5 pituitaries (Fig. 3A,B). However, Pax7 is expressed before Tpit in the developing IL (Fig. 1E). In postnatal (P2) Tpit^−/− ILs, Pax7 expression is restricted to cells on the ventral/luminal side, with more dorsal cells appearing completely negative for Pax7 expression (Fig. 3, cf. D and C). This pattern of Pax7 expression is thus more consistent with independent activation of Pax7 and Tpit expression and with Tpit exerting a positive feedback on Pax7 expression.

Prior work had shown that the bulk of adult Tpit^−/− IL cells is blocked in a precursor state and that 10%–15% of the IL cells switch fate to become gonadotropes, as marked by expression of SF1 (Pulichino et al. 2003b; Bilodeau et al. 2009). At P2, cells that have undergone this fate switch are essentially observed on the dorsal side of the Tpit^−/− IL, and SF1 expression appeared to be mutually exclusive with Pax7 (Fig. 3E,F). Thus, most dorsal cells of the Tpit^−/− IL do not express Pax7 and have differentiated into gonadotropes. It is likely that these cells expressed Pax7 earlier, as most IL cells express Pax7 in early development, albeit at a lower level in Tpit^−/− pituitaries (Fig. 3B).

Pax7 and progenitor cell cycle exit

The restriction of Pax7-positive cells on the luminal side of the Tpit^−/− IL is intriguing, since a population of Sox2-expressing pituitary progenitors is localized around the lumen of the adult gland (Fauquier et al. 2008). We used coimmunofluorescence for Sox2 and Pax7 in order to define the relationship between pituitary stem cells and IL Pax7-positive cells. In P2 pituitaries, Sox2-positive cells are present around the lumen of the gland, as described previously, and Pax7-positive cells occupy the remainder of the IL (Fig. 3G). A few scattered cells are positive for both markers, suggesting that the transition from progenitor to putative precursor is accompanied by overlapping Sox2 and Pax7 expression. In the Tpit^−/− IL, this overlap is widespread, with most Pax7-positive cells retaining Sox2 expression (Fig. 3H). Interestingly, the Pax7-negative cells of the Tpit^−/− IL do not express Sox2. In order to assess whether Pax7/Sox2 double-positive cells are still in cell cycle, we performed colabeling for Pax7 and Ki67; this revealed few Ki67-positive cells and little overlap in either normal or Tpit^−/− ILs (Fig. 3I,J), indicating that Pax7 cells have mostly exited the cell cycle in both genotypes.

In view of the apparent failure of IL cells to fully exit their progenitor state in Tpit^−/− pituitaries (i.e., they maintain Sox2 in the presence of Pax7) and of the role of p57^Kip2 as the cell cycle inhibitor driving pituitary progenitors out of cell cycle (Bilodeau et al. 2009), we assessed expression of Pax7 in relation to p57^Kip2 and p27^Kip1 in Tpit^−/− ILs. p57^Kip2 expression was mostly observed in the dorsal Pax7-negative cells of the Tpit^−/− ILs (Fig. 3K,L) that have undergone a fate switch to SF1-positive gonadotropes (Fig. 3E,F). The Sox2/Pax7 double-positive cells do not appear to express either p57^Kip2 or p27^Kip1 (Fig. 3K–N). Interestingly, the distribution of p57^Kip2, p27^Kip1, and Sox2 expression was unaffected in the Pax7^−/− ILs (Supplemental Fig. S3A), suggesting that Pax7, in contrast to Tpit, is not involved in cell cycle progression of pituitary cells. Furthermore, Pax7^−/− IL cells do not exhibit changes in proliferation or apoptosis, as assessed by Ki67 or cleaved caspase3 labeling (Fig. 3I,J; data not shown). In normal development, Tpit is expressed later than Pax7 (Fig. 1E), and p57^Kip2 expression starts earlier than either in the IL (Bilodeau et al. 2009). In support of this sequence, we observed some cells double-positive for p57^Kip2 and Pax7 in the E16.5 IL (Supplemental Fig. S3B–D) but no p57^Kip2/Tpit double-positive cells (Supplemental Fig. S3E–G). Thus, Tpit and Pax7 play different roles in the progression of pituitary progenitors toward differentiation and in the repression of the progenitor state.

Pax7 and pituitary differentiation

In order to explore further the role of Pax7 in engaging differentiation of pituitary progenitors, we designed a transgenic gain-of-function experiment where Pax7 expression is driven by the Pitx1 promoter (Fig. 4A). Pitx1 is expressed broadly within the oral ectoderm and pituitary primordium—much earlier than any pituitary differentiation event (Lanctôt et al. 1997). The Pitx1 promoter fragment used in these experiments exhibits appropriate expression, albeit with variable penetrance both spatially and temporally (Bilodeau et al. 2009). Of a total of 13 transgenic embryos, five showed transgene expression and phenotypes of varying severity. The most severely affected pituitary had very few cells remaining at E17.5 (Fig. 4B–D). Pax7 was expressed throughout this tissue (Fig. 4B,C), and some cells coexpressed Sox2 (Fig. 4B). Other cells coexpressed Tpit with Pax7 (Fig. 4C), and that is associated with expression of POMC and the melanotrope marker PC2 (Fig. 4D). Other transgenic pituitaries analyzed at either E14.5 or E17.5 revealed a mosaic pattern of Pax7 expression, indicating less penetrant transgene expression. While some transgenic pituitaries exhibited reduced pituitary tissue, one pituitary presented with expanded tissue extending all the way to the oral ectoderm (Fig. 4E,H) and extensive but not complete Pax7 expression (Fig. 4I). Both Pax7-positive and Pax7-negative cells exhibited cortico/melanotrope differentiation, as revealed by Tpit (Fig. 4I) and POMC (Fig. 4J) expression. Interestingly, transgenic expression of Pax7 in this E14.5 anterior pituitary did not cause premature expression of the melanotrope-specific PC2 (Fig. 4J). Some Pax7-positive and Pax7-negative cells have undergone somatotrope differentiation, as revealed by Pit1 (Fig. 4K) and GH expression (data not shown). Gonadotrope differentiation, marked by expression of SF1 (Fig. 4L) and LH (data not shown), is also observed, but SF1-positive cells never coexpress Pax7, suggesting that Pax7 expression is incompatible with gonadotrope differentiation; this is consistent with the mutually exclusive differentiation between POMC and gonadotrope lineages (Pulichino et al. 2003b).

Collectively, the ectopic expression of Pax7 during pituitary development appears to engage progenitors into differentiation pathways. Although Pax7 expression itself did not lead to complete differentiation, it appears permissive for cortico/melanotrope and Pit1-dependent, but not gonadotrope, differentiation. Transgenic pituitaries with expanded pituitary tissue exhibited more Pax7-negative cells than those with impaired pituitary expansion, suggesting that an early timing of Pax7 expression may prevent expansion of progenitors, whereas its later or less-penetrant expression may contribute to their expansion. Consistent with this, we found Pax7-negative and Pax7-positive cells that are also positive for either p57^Kip2 (Fig. 4M) or Ki67 (Fig. 4N). These transgenic pituitaries also contain a mixture of Sox2-positive cells that are either Pax7-negative or Pax7-positive (Fig. 4O), but no apoptotic cells (Fig. 4P). By engaging progenitors into differentiation pathways, Pax7 expression thus appears to exhaust the progenitor pool in the most severe transgenic pituitaries (Fig. 4B–D). These gain-of-function experiments further support the model of Pax7 acting as an upstream selector gene for modulation of Tpit-driven differentiation.

Pax7 exerts its selector function through chromatin reprogramming

In order to determine how Pax7 may exert its selector function on cell differentiation, we expressed Pax7 in AtT-20 cells, an established tissue culture model of pituitary corticotrope cells. This was achieved using a retrovirus-mediated Pax7 vector and compared with cells infected with a similar empty vector conferring neomycin resistance. For each, pools of at least 10³ colonies were expanded and characterized (Fig. 5A). These cells express Pax7, as ascertained by Western blot, and this expression does not affect Tpit levels but increased the melanotrope-specific PC2 and slightly decreased GR levels (Fig. 5B). In order to more broadly assess changes in gene expression elicited by Pax7, we used RT-qPCR to quantify expression of melanotrope- and corticotrope-specific genes in AtT-20Pax7 compared with control AtT-20neo cells (Fig. 5C). These analyses showed increased expression of melanotrope-specific genes such as PC2 and DrD2 and slight repression of corticotrope-specific markers such as CRH-R1, AVP-R1b, and GR. Pax7 expression thus exerts some reprogramming of AtT-20 cells with a major effect on activation of melanotrope genes and less repression of corticotrope genes when compared with their relative levels in the IL of Pax7^−/− mice (Fig. 5C).

In order to investigate how Pax7 may reprogram AtT-20 cells, we conducted chromatin immunoprecipitation (ChIP) combined with deep sequencing (ChIP-seq) for Pax7 in AtT-20Pax7 cells. Genome-wide, this led to the identification of 43,741 peaks of Pax7 recruitment with a P-value ≥10⁻⁶; their characterization is described below. Since Pax7 modulates POMC cell differentiation that is primarily dependent on Tpit, we assessed whether Pax7 altered Tpit action by performing Tpit ChIP-seq in both AtT-20neo and AtT-20Pax7 cells (Supplemental Fig. S4A). While the total number of Tpit sites (∼11,000) is similar in these cells, only 43% of all Tpit sites are the same (conserved sites) in both (Fig. 5D; Supplemental Fig. S4B). In contrast, ∼35% of the Tpit sites present in AtT-20neo cells were lost in AtT-20Pax7 cells: In general, these tend to be weaker sites of lesser P-value, although a subset of strong Tpit sites was completely lost. Finally, a last subset of new Tpit sites (22%) appeared in AtT-20Pax7 cells. Overall, 73% of Tpit sites are within 1 kb of a Pax7 site. In view of this close relationship, we investigated the effect of Pax7 on not only Tpit binding, but also chromatin surrounding the binding sites. We thus conducted FAIRE-seq (formaldehyde-assisted isolation of regulatory elements [FAIRE] and deep sequencing) (Giresi and Lieb 2009; Song et al. 2011) and ChIP-seq for histone H3 Lys 4 methyl-1 (H3K4me1) that constitutes chromatin marks of active regulatory sequences, particularly enhancers (Heintzman et al. 2007). While FAIRE reveals regions of accessible DNA, likely nucleosome-free, active enhancers are marked by a bimodal H3K4me1 signature. We assessed these chromatin marks at the different subgroups of Tpit peaks (Fig. 5E). Whereas conserved Tpit peaks present in both AtT-20neo and Pax7 cells exhibit the expected FAIRE-seq signal flanked by H3K4me1 ChIP-seq signals, similar but weaker signatures are also observed at Tpit sites that are lost in AtT-20Pax7 cells, consistent with the idea that these sites identify active regulatory sequences. The recruitment of Pax7 to these sites likely reflects a transcriptional modulatory role of Pax7. A similar chromatin signature (FAIRE peak with bimodal H3K4me1) appears at new Tpit sites induced by Pax7 (Fig. 5E). Strikingly, this signature is not present at these positions in AtT-20neo cells; in fact, they exhibit a single peak of H3K4me1 and no FAIRE signal (Fig. 5E). The appearance of active chromatin marks in Pax7-expressing cells is consistent with a pioneer transcription factor role of Pax7 that would open chromatin to allow Tpit binding. In contrast, these active chromatin marks are not erased at Tpit-binding sites that are lost in AtT-20Pax7, but their intensity is diminished, consistent with the greater ability of Pax7 to activate melanotrope genes than repress corticotrope genes (Fig. 5C). In order to assess genome-wide whether Pax7 leads to chromatin remodeling and whether chromatin opening revealed by FAIRE is generally correlated with the switch in H3K4me1 signature from single to double peaks, we investigated these marks at all FAIRE-seq peaks that are constant (i.e., ratio of 0.8–1.2 comparing AtT-20Pax7 over neo cells) and at positions where a FAIRE peak appears (ratio >3) in AtT-20Pax7 compared with neo cells. Thus, conserved FAIRE peaks are associated with constant H3K4me1 bimodal signatures, whereas the appearance of new FAIRE peaks is associated with a shift from single to bimodal H3K4me peaks (Fig. 5F). This switch in H3K4me1 profile is thus clearly associated with chromatin remodeling and, in this instance, Pax7-dependent pioneering action.

Pax7, a pioneer transcription factor

The action of Pax7 that allows new Tpit binding has the hallmark of so-called active pioneer factors (Zaret and Carroll 2011), since its activity is associated with remodeling of chromatin (Fig. 5E,F). In order to better define this activity, we analyzed our ChIP-seq data to identify Pax7 and Tpit DNA-binding sites. The search for conserved motifs present under the Tpit-binding peaks identified a predominant palindromic motif corresponding to the TpitREpal (Fig. 6A; Supplemental Fig. S4C) that we previously identified in the POMC gene −7-kb enhancer (Langlais et al. 2011). The TpitREpal contains two half-sites related to TCACACC separated by 2 base pairs (bp); a similar motif is present under 70% of Tpit-binding sites revealed by ChIP-seq (Supplemental Fig. S4D). The TpitREpal half-site corresponds to the previously described Brachyury-binding site (Lamolet et al. 2001).

The search for conserved motifs at Pax7 ChIP-seq binding sites revealed three motifs that are related to each other and to previously identified binding sites for other Pax factors (Soleimani et al. 2012). Indeed, a Pax7 motif (Fig. 6B; Supplemental Fig. S5A) comprised of juxtaposed paired and HD motifs is most prevalent at Pax7-only peaks (i.e., >3 kb away from Tpit sites) and at Pax7 sites associated with new Tpit sites (Supplemental Fig. S5B). As for the latter (Fig. 5E), the Pax7-only sites that have a Pax7 motif are globally associated with chromatin remodeling (and pioneer activity), as revealed by FAIRE and H3K4me1 profiles (Fig. 6B). A large number of Pax7 peaks do not have a complete Pax7 motif but rather either of its half-sites, as they contain only paired or HD motifs. Whereas Pax7 peaks that only have the HD motif are clearly associated with pre-existing active FAIRE and H3K4me1 marks (Fig. 6C), the peaks associated with paired-only sequences have weaker marks and less clearly bimodal H3K4me1 in AtT-20neo cells (Fig. 6D): This may reflect the weaker conservation of the HD half of the Pax motif and the associated difficulty for search algorithms to identify it. Hence, these data suggest that Pax7 interaction with the HD motif may not be sufficient for pioneer activity, but the conclusion is less clear for the paired motif. In contrast, the clear association of the full-length Pax7 motif with pioneer activity indicates that Pax7 interactions with DNA that involve both paired and HD domains represent the privileged mode of pioneering action.

The PC2 locus provides an interesting example of Pax7 action that allows Tpit recruitment and expression. The PC2 gene lies within a 4-Mb region that is devoid of Tpit-binding sites in AtT-20 cells and contains no other gene that has altered expression in the Pax7^−/− IL (Fig. 6E). Pax7 binds several positions within this locus, but one Pax7-binding site located at −146 kb relative to the PC2 transcription start site (TSS) is very close to an induced Tpit-binding site (Fig. 6F). In fact, the well-conserved Pax7 site present at the Pax7 peak is 337 bp away from the TpitREpal present under the Tpit peak; there is also a smaller Tpit peak that colocalizes with the Pax7 peak. Consistent with a pioneering action of Pax7, binding of Pax7 results in the appearance of a FAIRE peak flanked by H3K4me1 peaks, indicative of chromatin remodeling (Fig. 6F). Since Tpit does not occupy its −146-kb target sequence in the PC2 locus in AtT-20neo cells, this suggests that Tpit does not have pioneer activity, in contrast to Pax7.

We assessed the putative enhancer properties of the 1.4-kb fragment that is delimited by the H3K4me1 profile in AtT-20Pax7 cells and that corresponds to conserved DNA sequences. When inserted upstream of the POMC promoter, this fragment was inactive, but Pax7 enhanced expression of the luciferase reporter, indicating that its activity requires Pax7 (Fig. 6G). These data suggest that Pax7 has at least two roles for transcription of the PC2 gene: first, a pioneer activity that opens chromatin structure at the −146-kb enhancer, and second, a canonical transcription factor activity when Pax7 is bound to the enhancer.

Pax7, a selector gene that modulates Tpit-driven differentiation

The role of Pax7 in the differentiation of the two pituitary POMC lineages is unique in comparison with many other transcription factors that have been implicated in cell fate decisions. The role of Pax7 in this context may be the first example of the selector gene concept developed to define developmental fields >35 years ago by Garcia-Bellido (1975). In order to delineate this role, it is interesting to first discuss what Pax7 does not do in comparison with the large number of transcription factors that have been implicated in differentiation and cell fate decisions. First and foremost, Pax7 is not a pituitary differentiation gene, since its presence in the Tpit^−/− IL is not sufficient to drive differentiation. Also, Pax7 does not play a maintenance role in the pituitary, as it does in muscle satellite cells (Oustanina et al. 2004; Lepper et al. 2009). Thus, the IL Pax7-positive cells are engaged in the differentiation pathway but need Tpit to complete the process. In contrast, Tpit is truly a differentiation gene, since Tpit^−/− IL cells remain either nondifferentiated (Sox2/Pax7 double-positive) or differentiate into the alternate gonadotrope cell fate.

In normal IL development, progenitors would exit the cell cycle under the action of p57^Kip2 and switch on Pax7 expression followed by Tpit (Fig. 7). Tpit exerts a positive feed-forward effect on Pax7 expression, and together, these factors set up a melanotrope-specific gene expression program. In the absence of Pax7, this program shifts to a corticotrope-like profile, in agreement with the importance of Tpit for AL corticotrope differentiation (Pulichino et al. 2003b). Pax7 thus has a dual role of activating the melanotrope and repressing the corticotrope programs (Fig. 7). In the absence of the terminal differentiation function provided by Tpit, the IL accumulates Pax7/Sox2 double-positive cells that either remain in an intermediate precursor state or switch fate to become SF1-positive gonadotropes (Pulichino et al. 2003b).

Pax7 thus fits well within the original description of selector genes as cell-autonomously acting genes that provide positional identity and define developmental fields. The highly restricted expression of Pax7 in the pituitary IL thus defines the identity of this tissue relative to the other pituitary lobes: the anterior and the posterior pituitaries. However, the contribution of Pax7 to the establishment of a unique cell-specific genetic program is critically dependent on and only realized after Tpit-driven terminal differentiation. In promoting melanotrope-specific genes and repressing corticotrope-specific genes, Pax7 provides tissue identity to the cell differentiation process, as postulated for early-acting developmental selector genes. The action of Pax7 is thus very different from that of terminal differentiation genes, such as Tpit. The role of Pax7 in IL pituitary cells represents the first well-defined selector role for a transcription factor that specifies compartment identity and modulates the differentiation process.

Selector function is exerted through pioneer action

A major effect of Pax7 as a selector gene is the activation of melanotrope-specific genes such as PC2 and DrD2. Interestingly, the mechanism underlying these actions is consistent with the concept of a pioneer transcription factor (Zaret and Carroll 2011). Indeed, Pax7 expression in AtT-20 cells results in many new Tpit-binding sites; for example, at the PC2 and DrD2 loci (Fig. 6E,F). These new Tpit-binding sites contain TpitREpal sequences that are indistinguishable from other Tpit-binding sites active in AtT-20 cells. The pioneer action of Pax7 at these putative regulatory sequences creates a chromatin environment that is permissive for Tpit binding. This environment includes a FAIRE signal (similar to DNase-1-hypersensitive regions) that has been associated with nucleosome-free regions (Song et al. 2011). Another hallmark of these regions is the presence of flanking peaks of H3K4me1 (Heintzman et al. 2007). This signature appeared in Pax7-expressing cells at putative enhancers that were previously silent but marked by a single shallow peak of H3K4me1 (Fig. 5E,F). This single peak may be a previously unrecognized mark of silent enhancers, such as those revealed by the presence of p300 (Heintzman et al. 2007). The shift in the profile of H3K4me1 observed when putative enhancers switch from silent to active may reflect an important role of the flanking H3K4me1 peaks in defining an “open” region of chromatin. The switch of the H3K4me profile was correlated with chromatin opening, as revealed by FAIRE independently of Pax7 or Tpit peak positions (Fig. 5F), highlighting their genome-wide correlation.

Pax factors have the ability to bind DNA through either or both of their DNA-binding domains: the paired and HD domains. It was recently reported that Pax7 may have a higher affinity for HD motifs than the related Pax3 and that this may account for the greater number of Pax7 targets (Soleimani et al. 2012). We found both paired and HD motifs associated with Pax7 peaks (in agreement with this study in muscle cells) but a high frequency of juxtaposed sites that define the Pax7 motif (Fig. 6B). It is noteworthy that many peaks have multiple copies of these motifs but with no specific spacing (Supplemental Fig. S5C). However, when we focused our analyses on peaks that contain only one of these motifs (based on search algorithms that are not fail-safe), we found the highest frequency of Pax7 motifs associated with peaks that exhibit chromatin opening (Fig. 6B) compared with Pax7 peaks that contain only HD motifs (Fig. 6C) and that have FAIRE and H3K4me1 profiles of accessible chromatin.

These data thus suggest that the Pax7 motif has a greater pioneering potential than HD and possibly paired domains on their own. Binding of Pax7 to these latter motifs may thus be associated with canonical transcription factor action, consistent with such a role for Pax7 action on the PC2 gene −146-kb enhancer (Fig. 6G).

While pioneer activity has been ascribed to Fox transcription factors (Zaret and Carroll 2011) and associated with epigenetic changes (Serandour et al. 2011), others have argued that combinatorial action of lineage-determining factors may suffice for cell fate determination (Heinz et al. 2010). Given that transcription factors may have both pioneer and canonical transcription activator functions, the association of preferred DNA motif-binding preferences and open chromatin profiles may be required to demarcate these two properties and their gene targets.

Mice, histology, and IHC

Tpit mutant and POMC-EGFP transgenic mice were described previously (Pulichino et al. 2003b; Lavoie et al. 2008). Pax7 mutant mice (Mansouri and Gruss 1998) were provided by Mona Nemer. IHC was performed on formalin-fixed paraffin-embedded tissue sections, as described in Lamolet et al. (2001). Antibody sources and dilutions are indicated in Supplemental Table 1. For IHC, biotinylated anti-mouse or anti-rabbit secondary antibodies (Vector Laboratories) were revealed with HRP-coupled streptavidin (Perkin Elmer). Immunofluorescence was performed under the same conditions using fluorescent (Alexa 488 or Alexa 546) anti-mouse, anti-rabbit, or anti-guinea pig antibodies obtained from Molecular Probe.

Animal experimentation was approved by the IRCM Animal Ethics Review Committee. Studies using human samples were approved by the IRCM Human Ethics Review Committee.

RT-qPCR and ChIP

For analysis of Pax7 knockout NIL mRNAs, mice were sacrificed at P24, and RNA was prepared from dissected ALs and NILs. RNA was extracted using RNeasy columns according to the manufacturer's recommendations (Qiagen), and cDNA was synthesized using SuperScript III reverse transcriptase (Invitrogen) and purified using QIAquick columns. qPCR was performed using PerfeCta reagent (Quanta) on a MX-3005 device (Stratagene), and results were analyzed using the accompanying software. All quantifications were relative to GAPDH or TBP mRNA as stated in the figure legends. The primers used are provided in Supplemental Table 2.

ChIP and ChIP-seq were performed and analyzed as described (Langlais et al. 2012).

ChIP-seq peak-finding analysis

Sequence reads mapping to the mouse mm9 genome assembly were performed with Bowtie using the following settings: bowtie −t−trim5 1−best mm9 −S−chunkmbs 1024 (Langmead et al. 2009). About 95% of the sequence reads were mapped to the reference genome. To identify significant binding events, we processed the mapped sequence reads with MACS version 1.4.1 using a matching number of IgG control reads. We used the following setting in MACS: −to-small−bw 250-g mm−wig (Zhang et al. 2008). The MACS option –mfold was determined with the prior diag option. The complete number of sequence reads and peaks are listed in Supplemental Figure S4A. We kept peaks with a P-value >10⁻⁵ and a length <5 kb.

To compare Tpit genomic recruitment in AtT-20 cells versus AtT-20Pax7 cells, we used a similar number of reads (∼67 million reads after filtering) for each sample. To analyze the best subgroups of new, conserved, and lost Tpit sites, we ranked the peaks in each sample based on P-values and then compared their ranking.

We generated the density profiles with MACS using these attributes: −wig,−space 50.

De novo motif analyses

We extracted 100 bp of DNA sequence surrounding the binding peaks from the University of California at Santa Cruz Web site (Kent et al. 2002). The sequences of each cluster were processed using MEME (Bailey et al. 2006) and HOMER (Heinz et al. 2010). The following settings were used in MEME: −dna −nmotifs 5 −mod zoops −minsites 25. Graphical representation of the position weight matrices obtained from these analyses were generated with WebLogo (http://demo.tinyray.com/weblogo) (Crooks et al. 2004).

Transcription factor-binding motif searches

We searched for sequence motifs in 60-bp fragments surrounding ChIP-seq binding peaks using HOMER. The resulting occurrence for each motif was compared with the number of predicted binding sites in a set of 7311 randomly chosen genomic sequences.