Characterization of Intermediate Lobe Tumors in p27^−/− Pituitaries.

We first examined P27 by immunofluorescence and observed widespread nuclear expression throughout the anterior pituitary with varying levels of expression (Fig. 1A), in agreement with fluctuations of P27 according to the cell cycle. Anterior endocrine and stem cells proliferate rarely (20, 22) and P27 levels are relatively high in these. However, even though IL melanotrophs have been shown to be quiescent (6), P27 immunostaining is much weaker compared with the AL and SCs. We then examined SOX2, which is expressed both in SCs (18), and, at lower levels, in melanotrophs (19) (Fig. 1A). Coexpression of SOX2 and its repressor P27 probably reflects complex regulatory mechanisms controlling Sox2 expression and/or availability of corepressors recruited by P27 mechanisms.

Fig. 1.

The expression of stem cell and differentiation markers is affected in p27^−/− IL. (A) P27 and SOX2 immunofluorescence in adult pituitaries. In p27^−/− IL, SOX2 staining is more intense. (B) Quantification of SOX2 expression levels in p27^−/− versus wild type (WT) after immunofluorescence. The ratio of staining intensity between SOX2⁺melanotroph/SOX2⁺ SC in AL parenchyma is represented. In WT, SOX2 staining in melanotrophs is less intense than in AL SCs (0.63 ± 0.20) while in p27^−/− it becomes similar (0.91 ± 0.16); (*P = 0.019, n = 4–6 in each group). (C and D) Bulk RNAseq selected pathway analysis in 2-mo-old IL (p27^−/− n = 2 vs. WT n = 3, adjusted P value <0.05) (C), and in 7-mo-old hyperplastic IL (p27^−/− n = 4 vs. WT n = 2) (D). Blue tones show down-regulated pathways while red-yellow represent up-regulation. (E) Heatmap of representative genes expression from pathways shown in C and D. (F) POMC immunofluorescence in adult pituitaries. (G) Quantification of POMC intensity in melanotrophs in relation to corticotrophs at E18.5 in WT (1.01 ± 0.17, n = 5) and p27^−/− (1.10 ± 0.19, n = 5) and in 2- to 4-mo-old animals. Levels are significantly reduced in mutants (WT: 2.42 ± 0.76, n = 3; p27^−/−: 1.1 ± 0.07, n = 4, *P = 0.0286). (H) Radioimmunoassays for MSH and ACTH. p27^−/− pituitaries contain less MSH (*P = 0.0306, n = 4 in each group) (Left graph); ACTH levels are not affected (Right graph). (I) RT-qPCR in 2- to 3-mo-old p27^−/− and WT. Pomc (WT: 1.03 ± 0.27, n = 3; p27^−/−: 0.32 ± 0.09, n = 3,**P = 0.0013) and Tbx19 levels (WT: 0.94 ± 0.08, n = 3; p27^−/−: 0.30 ± 0.10, n = 3, **P = 0.0071) were reduced in mutants. (J) Average cell size of melanotrophs. p27^−/− melanotrophs (99 ± 5.6, n = 5) are smaller compared to WT (132 ± 10, n = 3, ***P = 0.0007). (K–M) Analysis of cell proliferation. EdU incorporation was quantified in melanotrophs (EdU;SOX2;PAX7 triple positive/SOX2;PAX7 double positive in L or EdU;SOX2 double positive; SOX9 negative/SOX2 positive;SOX9 negative positive cells in M and N) and SCs (EdU;SOX2 double positive; PAX7 negative/SOX2 positive;PAX7 negative cells in L or EdU;SOX2;SOX9 triple positive/SOX2;SOX9 double positive in M and N). SCs were distinguished as ventral (flanking AL) or dorsal (flanking IL). In 1-mo-old mice (M), cell proliferation increases in p27^−/− melanotrophs and dorsal SCs (n = 3–4 mice/genotype, melanotrophs *P = 0.0235, dorsal cleft *P = 0.0419). In 7-mo-old mice, (N) cell proliferation increases in melanotrophs and both dorsal and ventral SCs (n = 3 mice/genotype, melanotrophs *P = 0.0167, dorsal cleft **P = 0.0085, ventral cleft *P = 0.0229). (N) SOX2, SOX9, and EdU triple staining in 1-mo-old pituitaries. (O) Immunofluorescence for SOX2 on IL pituispheres from 2- to 3-mo-old animals. The proportion of pituispheres is increased in mutant IL (***P < 0.0001, n = 6–8 in each group). (P) Immunofluorescence for CD31 in 1-mo-old pituitaries. p27^−/− ILs show ectopic blood vessels (arrows). ns, nonsignificant.

In p27^−/− pituitaries, the SC layer appears thicker in mutants (15). In agreement with a repressive action of P27 on Sox2 expression (15), SOX2/Sox2 levels are significantly up-regulated both in SCs and melanotrophs in p27^−/− pituitaries (Fig. 1 A and B and SI Appendix, Fig. S1 A and B).

To further characterize initiation and development of the tumors in p27^−/− mice, we performed transcriptomic analyses from dissected ILs (Fig. 1 C–E and SI Appendix, Fig. S1C). Before tumorigenesis, at 2 mo of age, we detected 250 differentially expressed genes (DEGs) between p27^−/− and control samples. In 7-mo-old animals, when hyperplasia is very clear, there were 3,736 DEGs (Dataset S1). We then performed a pathway analysis (Dataset S1). At both stages, up-regulation of cell cycle markers was observed, in agreement with the CDKI function of P27. We also observed down-regulation of genes associated with the cytoskeleton and cell adhesion which may reflect disruption of the direct interaction between P27 and RhoA (23). Furthermore, down-regulation of POMC and genes associated with its processing suggests an alteration of melanotroph function and/or identity in mutants.

We further investigated whether melanotroph function and/or identity have been altered, something that has not been reported previously (910–11). We observed a significant down-regulation of POMC processing in 2-mo-old mutants, while expression of Pax7 and of the POMC-lineage marker Tnxb (24) are reduced in 7-mo-old mutants (Dataset S1). We then quantified POMC/Pomc expression levels, and detected a down-regulation exclusively in p27^−/− melanotrophs, postnatally, but prior to tumorigenesis (Fig. 1 F–H), while corticotrophs appear unaffected. In agreement with these data, radioimmunoassays show that levels of MSH are reduced while ACTH is unaffected in p27^−/− pituitaries (Fig. 1H). This shows that the anterior pituitary is affected by loss of p27. TBX19 is a direct upstream regulator required for Pomc expression (25). In agreement with a reduction in Pomc levels, Tbx19 levels are significantly reduced in the IL, while Pax7 levels (5) also appear reduced, although not to a statistically significant degree in 2- to 3-mo-old animals (Fig. 1I). Moreover, p27^−/− melanotrophs are smaller than those in wild-type (WT) pituitaries (Fig. 1J). Altogether, these results suggest that melanotroph identity and/or postnatal maturation are altered in p27^−/− mutants, and this occurs before tumors develop.

We then examined cell proliferation by quantifying incorporation of 5-ethynyl-2′-deoxyuridine (EdU) in the IL, dorsal (flanking IL), and ventral SC layer (flanking the AL) (Fig. 1 K–N). In E18.5 embryos, proliferation tends to increase in both IL and its flanking SC layer in the mutants; however, this is not statistically significant (Fig. 1K). In contrast, in 1-mo-old mutant animals, we observed a significant increase in proliferation of both melanotrophs and adjacent SCs (Fig. 1 L–N). In 7-mo-old animals, increased proliferation was observed in all compartments of the mutants compared to controls (Fig. 1M), in agreement with the transcriptomic analysis results (Fig. 1 E and F). To further test the proliferative capacity of SCs, we performed pituisphere assays (20) separately from anterior and intermediate lobe cells and observed an increased sphere-forming ability from p27^−/− IL cells (Fig. 1O). Altogether these results suggest that, while both SC layers behave abnormally in p27^−/− pituitaries, the SCs in direct contact with melanotrophs are more affected, suggesting that there may be some noncell autonomous consequences to the loss of p27.

HIF-1 activation is associated with tumor angiogenesis and its targets are up-regulated in 7-mo-old p27^−/− IL transcriptomes (Fig. 1E). Moreover, p27^−/− IL tumors are hemorrhagic (9). We thus examined CD31 expression and observed development of ectopic blood vessels in juvenile p27^−/− ILs, coincident with the abnormal increase in rates of cell proliferation (Fig. 1P).

Furthermore, p27^−/− mice are affected by gigantism, which has been attributed to general loss of cell cycle repression (910–11); however, an additional explanation could be the significant increase in pituitary growth hormone (GH) content we observed (SI Appendix, Fig. S1D). While anterior lobe tumors are not observed in mice (12), in both rats and humans carrying mutations in P27 GH adenomas are observed (26). It would be of interest to examine whether in mice somatotroph numbers are affected.

In conclusion, p27 deletion results in increased cell proliferation and ectopic vascularization in the IL. While P27 starts to be expressed in differentiating embryonic endocrine cells (13), defects consecutive to its loss appear to only become significant postnatally, both in melanotrophs and SCs. In addition, melanotroph identity and/or maturation appears affected in mutants.

Characterization of the SOX2-P27 Interaction in p27^−/− Mutants.

Deletion of one copy of Sox2 is sufficient to dramatically reduce occurrence of IL tumors (Fig. 2A and SI Appendix, Fig. S2B) (15); however, hyperplasia is still observed in p27^−/−; Sox2^+/− IL (Fig. 2A). We further characterized the SOX2-P27 interaction by demonstrating that survival of p27^−/− mutants (12.5 mo, n = 28) was significantly improved by Sox2 haploinsufficiency (17.1 mo, n = 10) (SI Appendix, Fig. S2A). However, in agreement with cell-type specificity for the SOX2-P27 interaction, while pituitary and duodenum tumors were sensitive to Sox2 dosage, lung tumor development was unaffected (SI Appendix, Fig. S2B).

Fig. 2.

Deletion of one copy of Sox2 in p27^−/− animals improves survival and results in reduction of cell proliferation and impairment of tumorigenesis. (A) Brightfield pictures taken at the same magnification of WT, Sox2^+/−, p27^−/− , and p27^−/−; Sox2^+/− pituitaries (Left). Histological analysis of 10- to 12-mo-old WT, p27^−/−, and p27^−/−; Sox2^+/− pituitaries (Right). In p27^−/−; Sox2^+/− IL tumorigenesis is impaired, but hyperplasia is observed. (B) Comparison modalities for RNAseq analysis of wild-type, p27^−/−, and p27^−/−; Sox2^+/− 2- to 3-mo-old IL samples (WT n = 4, p27^−/− n = 3, p27^−/−; Sox2^+/− n = 3, adjusted P value <0.05) (Left). Selected enriched pathways are represented (Center) and heatmap of representative gene expression from affected pathways (Right). (C) Analysis of cell proliferation. EdU incorporation was quantified in melanotrophs (EdU; SOX2 double positive; SOX9 negative/SOX2 positive;SOX9 negative cells) and SCs (EdU; SOX2; SOX9 triple positive/SOX2; SOX9 double positive cells). SCs were further distinguished as ventral or dorsal, respectively, as flanking the anterior or intermediate lobe. Deletion of one copy of Sox2 results in a reduction of cell proliferation compared to p27^−/− samples in dorsal cleft SCs (n = 3 to 4 mice/genotype, *P = 0.0208), melanotrophs. (D) SOX2 and EdU double staining in 2- to 3-mo-old WT, p27^−/−, and p27^−/−; Sox2^+/− animals. There is a clear reduction in SOX2 levels in p27^−/−; Sox2^+/− ILs. Cleft and IL are outlined. (E) Proportion of pituispheres obtained from AL and IL WT, Sox2^+/−, p27^−/−, and p27^−/−; Sox2^+/− 2- to 3-mo-old animals. There is a reduction in the proportion of pituispheres formed from p27^−/−; Sox2^+/− ILs compared to p27^−/− samples. WT (1 ± 0.11, n = 5), Sox2^+/− (0.2 ± 0.08, n = 2), p27^−/− (2.09 ± 0.31, n = 4), and p27^−/−; Sox2^+/− (1 ± 0.52, n = 3), ****P < 0.0001 and ***P = 0.0002; the reduction in Sox2^+/− samples in agreement with the slightly hypomorphic pituitaries observed in Sox2 heterozygous mice (25). (F) Quantitative analysis of melanotroph marker expression levels by RT-qPCR in 2- to 3-mo old WT, Sox2^+/−, p27^−/−, and p27^−/− ; Sox2^+/− IL. In p27^−/−; Sox2^+/− animals, Tbx19 expression levels are restored to WT levels. Pomc; WT (n = 3) vs. p27^−/−; Sox2^+/− (n = 3), *P = 0.03 and WT vs. p27^−/− (n = 3), **P = 0.0022, Tbx19; WT vs. p27^−/−; Sox2^+/− (n = 3, P = nonsignificant [ns]) and WT vs. p27^−/− (n = 3, *P = 0.0169).

We then performed comparisons of the transcriptomes of wild type, p27^−/−, and p27^−/−; Sox2^+/− IL (Fig. 2B and SI Appendix, Fig. S2C). To pinpoint genes underlying IL tumor impairment in compound mutants, and conversely those associated with higher levels of SOX2 and tumorigenesis, we focused our analyses on those that were exclusively differentially expressed in p27^−/−, and therefore restored to control levels in p27^−/−; Sox2^+/− samples. This analysis was thus conducted differently than the one presented in Fig. 1 C and D, where we simply compared p27^−/− vs. wild type. When compared, the two lists of DEGs comprise both similarities and differences (Figs. 1E and 2B and Datasets S1 and S2). However, when we performed a pathway analysis, we obtained a list of pathways comparable in our two analyses (Figs. 1 C and D and 2B and Datasets S1 and S2). This implies that in the IL, the different processes affected by P27 loss are restored to normal levels upon deletion of one copy of Sox2. This thus suggests that SOX2 is not required for a specific aspect of tumorigenesis in p27^−/− mice, but rather that it has a broad and therefore probably early role in tumor formation. Reduced Sox2 dosage is associated with a decrease in proliferation and HIF1-target gene expression, in agreement with less angiogenesis linked with reduced overproliferation in p27^−/−; Sox2^+/− mutants. We also observed an effect on genes involved in the cytoskeleton, suggesting that reduction of Sox2 allows restoration of a cellular phenotype resembling that of wild type. Alternatively, alteration of cellular shape and contacts may be linked to variation in rates of cell proliferation.

We then analyzed SOX2 expression, cell proliferation, and pituisphere formation in p27^−/−; Sox2^+/− pituitaries (Fig. 2 C–E). There was a reduction in SOX2/Sox2 levels in p27^−/−; Sox2^+/− IL, both in melanotrophs and stem cells (Fig. 2D and SI Appendix, Figs. S2D, S3E, and S4J). Moreover EdU incorporation in compound mutants, both in melanotrophs and SCs was not significantly different from controls, in agreement with impaired tumorigenesis (Fig. 2C). A similar effect was observed in pituisphere assays (Fig. 2E). We moreover observed a reduction in pituisphere formation efficiency in Sox2^+/− samples, suggesting a reduced proliferative capacity (Fig. 2E). This is in agreement with the slightly hypomorphic pituitaries observed in Sox2 heterozygous mice (27) and with SOX2 being required for proliferation of pituitary progenitors in the embryo (19, 28).

Alteration of p27^−/− melanotroph identity (Fig. 1I) also appears to depend on SOX2 dosage, because levels of Tbx19 which are reduced in p27^−/− IL, are not significantly affected in p27^−/−; Sox2^+/− compared to control IL (Fig. 2F and see Fig. 6J below). However, levels of Pomc were still significantly reduced in p27^−/−; Sox2^+/− IL compared to control.

Sox2 Regulatory Region 2 (Srr2) Mediates P27 Effect on Sox2 Expression In Vivo.

Transcriptomic studies and proliferation and marker analyses all indicate that reduction of SOX2 dosage compromises most aspects of tumorigenesis in p27^−/− animals, implying an essential role for SOX2 in p27^−/− melanotrophs and adjacent SCs for tumor development. To test whether the SOX2-P27 interaction relies solely on Srr2 (15), we generated mice deleted for this region (Fig. 3 A and B). Srr2^del/del mutants were viable (SI Appendix, Fig. S3A) and pituitary morphology was normal (Fig. 3C). However, on a p27^−/− background, the Srr2 deletion prevents tumorigenesis, while hyperplasia of IL is still evident, as observed in p27^−/−; Sox2^+/− mutants (Fig. 3C and SI Appendix, Fig. S3B). In addition, the morphology of the retina, which is affected in absence of p27 (11), is also improved in the p27^−/−; Srr2^del/del mutants. In particular, and again similarly to p27^−/−; Sox2^+/− mutants, the inner and outer layers have a reduced thickness and better organization compared to p27^−/− retina (SI Appendix, Fig. S3C) (15). In the pituitary, levels of SOX2 are reduced in both melanotrophs and SCs in p27^−/−; Srr2^del/del mutants compared to p27^−/− samples (Fig. 3 E and F and SI Appendix, Fig. S3D). Intriguingly, EdU incorporation analyses revealed a specific reduction of proliferation in SCs in compound mutants compared to p27^−/− samples (Fig. 3D). In agreement with this result, while cultures from p27^−/− IL give rise to significantly more pituispheres than controls, the efficiency of pituisphere formation from p27^−/−; Srr2^del/del samples is reduced (Fig. 3G). These analyses strongly suggest that Srr2 has an exclusive role in mediating the SOX2-P27 interaction (see model SI Appendix, Fig. S3E). Moreover, the consequences of its deletion are more obvious in SCs, suggesting that Srr2 plays a more important role in this compartment. Furthermore, this implies that SCs may be critical for IL tumorigenesis in p27^−/− mutants.

Fig. 3.

Deletion of Srr2 in p27^−/− animals results in reduction of cell proliferation in SCs and impairment of tumorigenesis. (A) Schematic representation of the Sox2 murine locus and Srr2 deletions. Two pairs of sgRNA (pairs 1 and 2, protospacer adjacent motif sequence in orange) were designed. The primers used to genotype Srr2-deleted animals are marked by red arrows. Three founders deleted for Srr2 were used to generate stable lines (founders 1 through 3); the size of the deletion is indicated for each line. (B) Gel electrophoresis showing mosaic loss of Srr2 in founder mice. The number above each lane corresponds to a individual founder. Boxed numbers below indicate strain derived from that particular founder. (C) Brightfield pictures of WT, Srr2^del/del, p27^−/−, and p27^−/−; Srr2^del/del pituitaries. Deletion of two copies of Sox2 Srr2 on a p27^−/− background is sufficient to delay tumorigenesis. (D) Analysis of cell proliferation. EdU incorporation was quantified in melanotrophs (EdU; SOX2 double positive; SOX9 negative/SOX2 positive; SOX9 negative cells) and SCs (EdU; SOX2; SOX9 triple positive/SOX2; SOX9 double positive cells) in 7-mo-old WT, p27^−/−, and p27^−/−; Srr2^del/del animals. In p27^−/−; Srr2^del/del animals, there is a reduction of proliferation in stem cells compared to p27^−/− (p27^−/− (12.33 ± 5.37, n = 4) vs. p27^−/−; Srr2^del/del (2.11 ± 1.88, n = 4), **P = 0.0061), but not in melanotrophs (p27^−/− [8.75 ± 2.36] vs. p27^−/−; Srr2^del/del [6.45 ± 4.01], P = ns). (E) SOX2 and EdU double staining in 7-mo-old p27^−/− and p27^−/−; Srr2^del/del animals. SOX2 expression appears decreased in p27^−/−; Srr2^del/del IL. Cleft is outlined. (F) Quantification of SOX2 expression levels in WT, Srr2del/del, p27^−/−, p27^−/−;Sox2^+/−, p27^−/−;Srr2del/del. Levels of expression were quantified after immunofluorescence in SCs and melanotrophs. In p27^−/−;Srr2del/del, SOX2 staining is less intense than in p27^−/− (*P = 0.017 SCs and **P = 0.0078 in melanotrophs, n = 11 to 4 in each group). In contrast SOX2 staining is more intense in p27^−/− SC vs. WT (***P = 0.0007) and p27^−/−;Sox2^+/− (***P = 0.0009). (G) Proportion of pituispheres obtained from IL in WT, p27^−/−, and p27^−/−; Srr2^del/del. The percentage of spheres formed from p27^−/−; Srr2^del/del (0.02 ± 0.007, n = 3, P = ns) is not significantly higher than WT (0.006 ± 0.005, n = 3), in contrast with p27^−/− samples (0.04 ± 0.02, n = 3, *P = 0.0344), demonstrating that loss of Srr2 has an impact on sphere-forming efficiency in mutants. ns, nonsignificant.

Requirement for SOX2 in p27^−/− Melanotrophs.

Conditional deletion of p27 in melanotrophs results in formation of tumors; therefore, transformation is cell autonomous in this context (29). To better characterize the role of SOX2 for tumorigenesis, we deleted one copy of the gene exclusively in melanotrophs in p27^−/− mutants. We initially used Pomc-CreERT2 which displays a mosaic pattern of recombination in the IL (SI Appendix, Fig. S4A). To look at heterozygosity for Sox2, we induced Cre activity in 4-wk-old Pomc-CreERT2; Sox2^fl/+; p27^−/− animals and controls, which were then examined 6 to 8 mo later. Tumor development was unaffected by the mosaic deletion of one copy of Sox2 in p27^−/− melanotrophs (Fig. 4A). In agreement with this observation, proliferation was unaffected by Sox2 reduction, as we observed a similar increase in EdU incorporation in Pomc-CreERT2; Sox2^fl/+; p27^−/− and Pomc-CreERT2; p27^−/− IL compared to controls (Fig. 4 A and B). To analyze more directly the effect of the loss of one copy of Sox2 on melanotrophs’ proliferative ability, we analyzed EdU incorporation in Pomc-CreERT2; Sox2^fl/+; p27^−/−; Rosa26^ReYFP samples. We compared the percentages of eYFP; PAX7; EdU triple positive and eYFP negative, PAX7; EdU double positive melanotrophs, assuming that Rosa26^ReYFP recombination also reflected Sox2 heterozygous deletion. We observed a significant reduction of EdU incorporation in eYFP-positive vs. -negative cells (Fig. 4 C and D). Therefore, reduction of SOX2 dosage in p27^−/− melanotrophs diminishes their proliferative capacity.

Fig. 4.

Deletion of one copy of Sox2 in p27^−/− melanotrophs prevents IL tumorigenesis. (A) SOX2 and EdU double staining in 6- to 8-mo-old Pomc-CreERT2, Pomc-CreERT2; p27^−/−, and Pomc-CreERT2; Sox2^fl/+; p27^−/− pituitaries. (B) Analysis of cell proliferation. EdU incorporation was quantified in melanotrophs (EdU; SOX2 low; POMC triple positive/DAPI nuclei in IL). EdU levels are similarly elevated in both p27^−/− samples. Pomc-CreERT2; Sox2^fl/+; p27^−/− (7.53 ± 2.92, n = 4, **P = 0.0023) and Pomc-CreERT2; p27^−/− (7.68 ± 1.20, n = 8, ***P = 0.0007) vs. Pomc-CreERT2 (controls), (1.60 ± 0.55, n = 3). (C and D) EdU incorporation was quantified in Sox2^+/+ melanotrophs (PAX7^+ve; eYFP^−ve) (2.40 ± 1.31, n = 6) and in Sox2^+/− melanotrophs (PAX7; eYFP double positive) (0.78 ± 0.86, n = 6) in 2.5- to 6-mo-old Pomc-CreERT2; Sox2^fl/+; p27^−/−; Rosa26^ReYFP mice, assuming that Rosa26^ReYFP recombination reflects Sox2 heterozygous deletion. In Sox2^+/− melanotrophs EdU incorporation is significantly reduced compared to Sox2^+/+ ones (*P = 0.0210), demonstrating that loss of one copy of Sox2 results in reduced proliferation in p27^−/− melanotrophs. (D) EdU, eYFP, and PAX7 triple staining. Most EdU; PAX7 positive cells (arrows) do not express eYFP in Pomc-CreERT2; Sox2^fl/+; p27^−/−; Rosa26^ReYFP IL. (E) Brightfield pictures of Pomc-Cre, Pomc-Cre; p27^−/−, and Pomc-Cre; Sox2^fl/+; p27^−/− pituitaries. Deletion of one copy of Sox2 using Pomc-Cre prevents IL tumorigenesis. (F) Double immunofluorescence for POMC and SOX9 in 6- to 10-mo-old animals, illustrating reduced IL size in Pomc-Cre; Sox2^fl/+; p27^−/−. (G) Measurement of IL area in Pomc-Cre; p27^−/− (1,472,054 ± 808,382, n = 3) and Pomc-Cre; Sox2^fl/+; p27^−/− (495,612 ± 185,861, n = 3). IL area in Pomc-Cre; Sox2^fl/+; p27^−/− pituitaries represents a third of IL area in p27^−/− mutants (**P = 0.0041). (H) Quantification of POMC intensity in melanotrophs in relation to corticotrophs in 6- to 10-mo-old Pomc-Cre; p27^−/− (1.13 ± 0.11, n = 3) vs. Pomc-Cre; Sox2^fl/+; p27^−/− (1.89 ± 0.29, n = 3) animals, where levels are significantly increased, *P = 0.0127. (I) In situ hybridization for Sox2 in 6.25- to 10-mo-old wild-type, p27^−/−, p27^−/−; Sox2^+/−, and Pomc-Cre; Sox2^fl/+; p27^-/− animals. (J) Quantification of Sox2 levels following in situ hybridization. Sox2 expression levels are reduced in both p27^−/−; Sox2^+/− and Pomc-Cre; Sox2^fl/+; p27^-/− melanotrophs and SCs compared to p27^−/−. Quantification of Sox2 levels in melanotrophs: p27^−/− (12.17 ± 2.15, n = 4), WT (3.5 ± 2.5, **P < 0.003), p27^−/−; Sox2^+/− (5.4 ± 1, *P = 0.0276), Pomc-Cre; Sox2^fl/+; p27^−/− (2.56 ± 0.1, **P = 0.0038). Quantification of Sox2 levels in SCs: p27^−/− (28 ± 6, n = 4), WT (11.4 ± 8, n = 5, ****P < 0.0001), p27^−/−; Sox2^+/− (11.3 ± 1.4, n = 3, ****P < 0.0001), Pomc-Cre; Sox2^fl/+; p27^−/− (19 ± 3, n = 3, **P = 0.0039). IL is outlined in A. IL and stem cell cleft are outlined in I. ns, nonsignificant.

We then hypothesized that a more efficient Cre-driver may block tumorigenesis; we therefore generated Pomc-Cre; Sox2^fl/+; p27^−/−animals. We showed previously that SOX2 is required for acquisition of melanotroph identity; however, homozygous deletion of the gene once melanotrophs have up-regulated PAX7 does not have an apparent effect on cell fate acquisition (19). There were no IL tumors or signs of hyperplasia in any 4- to 7-mo-old Pomc-Cre; Sox2^fl/+; p27^−/− animals (Fig. 4 E–G and SI Appendix, Fig. S4B). Moreover POMC staining intensity was increased in Pomc-Cre; Sox2^fl/+; p27^−/− melanotrophs compared to p27^−/− samples (Fig. 4H), indicating restoration of a normal melanotroph phenotype. This demonstrates that SOX2 is required in melanotrophs for p27 deletion to result in tumorigenesis. Strikingly, the morphology of IL appears completely normal in Pomc-Cre; Sox2^fl/+; p27^−/− (Fig. 4 E and F) compared to p27^−/−; Sox2^+/− IL where hyperplasia is clear (Fig. 2A). This difference in phenotype is unexpected, because in both mutants the genotype of melanotrophs is the same: only one allele of Sox2 is active while p27 is deleted. However, different regulatory mechanisms may affect the levels of expression of the remaining copy of Sox2 in Pomc-Cre; Sox2^fl/+; p27^−/− vs. p27^−/−; Sox2^+/− melanotrophs. We thus quantified Sox2 levels and observed an up-regulation in p27^−/− , and to a lesser extent in p27^−/−; Sox2^+/− samples (Fig. 4 I and J). In agreement with the absence of hyperplasia, we observed a further reduction in Sox2 expression in Pomc-Cre; Sox2^fl/+; p27^−/− vs. p27^−/−; Sox2^+/− samples, suggesting that regulation of Sox2 is indeed modulated according to the spatiotemporal pattern of its deletion. Furthermore we also observed, albeit to a lesser extent, reduction of Sox2 expression in Pomc-Cre; Sox2^fl/+; p27^−/− SCs.

In conclusion, while cell transformation in p27^−/− melanotrophs is known to be cell autonomous (29), we demonstrate here that SOX2 is required in these cells for tumor development in p27^−/− mutants. Moreover, reduction of Sox2 expression in SCs in Pomc-Cre; Sox2^fl/+; p27^−/− compared to p27^−/− suggests a noncell autonomous role and implies an interaction between melanotrophs and SCs.

SOX2 Is Required Independently in SCs for Induction of Melanotroph Tumors in p27^−/− Animals.

Because deletion of p27 induces up-regulation of SOX2 in SCs, we asked whether the SCs flanking the IL directly give rise to tumor cells or induce tumorigenesis. Given that, in the context of the pituitary, SOX9 is expressed uniquely in the SCs (20), we first performed lineage tracing experiments using Sox9^iresCreERT2 (30). Cre activity was induced in 4-wk-old Sox9^{iresCreERT2/+}; Rosa26^ReYFP/+ animals and pituitaries were harvested 6 to 11 mo later. In controls, the vast majority of eYFP-positive cells are SOX9-positive SCs (21) (Fig. 5A). In agreement with a low turnover of melanotrophs (6), very few eYFP-positive cells are present in the IL (Fig. 5A); these are negative for SOX9 and PAX7, where the latter indicates they are not melanotrophs. Their identity is unclear; they may originate from the pituitary SC layer or the posterior lobe where SOX9 is also expressed. In p27^−/− animals we also observed that most eYFP-positive cells are SOX9 positive SCs. However, in contrast with Sox9^{iresCreERT2/+}; Rosa26^ReYFP/+ controls, we observed some rare eYFP; PAX7-positive melanotrophs. These results clearly demonstrate that SCs are not at the origin of the melanotroph tumors in p27^−/− animals. However, tumorigenesis and/or lack of p27 in SCs, drives commitment of rare SCs toward the melanotroph lineage.

Fig. 5.

SOX2 is required in SCs for tumorigenesis in p27^−/− IL. (A) eYFP, SOX9 (Middle), and eYFP, PAX7 (Right) double stainings in 7- to 12-mo-old Sox9^{iresCreERT2/+}; Rosa26^ReYFP/+ and Sox9^{iresCreERT2/+}; p27^−/−; Rosa26^ReYFP/+ animals. eYFP; SOX9 double positive cells are localized in the epithelium lining the cleft in both genotypes (arrows). Very few eYFP⁺, SOX9⁻, and PAX7⁻ cells are present in Sox9^{iresCreERT2/+}; Rosa26^ReYFP/+ animals (arrows). In Sox9^{iresCreERT2/+}; p27^−/− Rosa26^ReYFP/+ pituitaries some rare eYFP; PAX7 double-positive cells are present, demonstrating low levels of differentiation into melanotroph (Left). (B) Brightfield pictures (Upper) and SOX2 immunohistochemistry (Bottom) on sections of Sox9^{iresCreERT2/+}; Rosa26^ReYFP/+, Sox9^{iresCreERT2/+}; p27^−/−, and Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/− pituitaries. In Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/− pituitaries tumorigenesis is impaired (1-y-old 14% tumor incidence, n = 7) compared to Sox9^{iresCreERT2/+}; p27^−/− (1-y-old, n = 5, 100% tumor incidence). The thickness of the SOX2 positive SC epithelium resembles that of control, demonstrating that reduction of Sox2 dosage in SCs impairs p27^−/− IL tumorigenesis. (C, D, and F) Analysis of cell proliferation. EdU incorporation was quantified in SCs and melanotrophs in 7- to 12-mo-old Sox9^{iresCreERT2/+}, Sox9^{iresCreERT2/+}; p27^−/−, and Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/− IL. (C) In SCs (EdU; SOX2; SOX9 triple positive/SOX2; SOX9 double-positive cells), there is a significant reduction of proliferation in Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/− samples (0.20 ± 0.35, n = 3, **P = 0.0011) compared to Sox9^{iresCreERT2/+}; p27^−/− ones (3 ± 0.68, n = 3). In fact, proliferation in Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/− SC is similar to Sox9^{iresCreERT2/+} controls (0.28 ± 0.32, n = 3, P = ns). (D) In melanotrophs (POMC positive cells in IL/DAPI positive nuclei), there is a reduction of cell proliferation in Sox9^{iresCreERT2/+};Sox2^fl/+;p27^−/− samples (1.88 ± 0.72, n = 6, **P = 0.0033) compared to Sox9^{iresCreERT2/+};p27^−/− (3.7 ± 0.76, n = 4); again proliferation in Sox9^{iresCreERT2/+};Sox2^fl/+;p27^−/− is similar to Sox9^{iresCreERT2/+} controls (0.60 ± 0.4, n = 4, *P = 0.03) while proliferation is increased in Sox9^{iresCreERT2/+};p27^−/− vs Sox9^{iresCreERT2/+} controls (****P = 0.0001). (E) Quantification of the vascular structures ectopic development in 7 to 12 mo old. Sox9^{iresCreERT2/+}, Sox9^{iresCreERT2/+}; p27^−/−, and Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/− animals. Deletion of one copy of Sox2 in SC leads to a reduction in the development of ectopic blood vessels in IL. IL vascular structures area: Sox9^{iresCreERT2/+} (0.32 ± 0.28, n = 3), Sox9^{iresCreERT2/+}; p27^−/− (2.56 ± 1.26, n = 3, (*P < 0.03), and Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/− (1.15 ± 0.43, n = 3). (F) EdU, SOX2 double staining in 7- to 12-mo-old Sox9^{iresCreERT2/+}, Sox9^{iresCreERT2/+}; p27^−/−, and Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/− pituitary glands. Right, immunofluorescence for CD31 showing ectopic blood vessel formation reduction in Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/− sample. SC layer is outlined in A, IL in F. ns, nonsignificant.

To investigate the role of SOX2 in SCs during tumorigenesis, one allele of the gene was removed exclusively in SCs in Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/− animals. Cre activity was induced in 4-wk-old animals and pituitaries harvested when the animals were 6 to 12 mo old. Strinkingly, only one among seven 12-mo-old Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/− animals developed an IL tumor, while all p27^−/−controls were affected (Fig. 5 B, Upper and SI Appendix, Fig. S5A). The IL is however consistently hyperplastic in Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/− animals, similarly to p27^−/−; Sox2^+/− mice. These results unequivocally demonstrate that SOX2 is required independently in SCs for tumors to develop in p27^−/− IL. The thickness of the SOX2-positive SC layer appeared reduced in Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/− compared to Sox9^{iresCreERT2/+}; p27^−/− (Fig. 5 B, Lower), and there were fewer SCs upon Sox2 reduction (SI Appendix, Fig. S5B). SOX2 staining intensity in SCs was also reduced (Fig. 5F and SI Appendix, Fig. S5C). Furthermore, we observed a significant reduction of proliferation in both SCs and melanotrophs in 6-mo-old Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/− IL compared to Sox9^{iresCreERT2/+}; p27^−/− samples (Fig. 5 C, D, and F). In addition, formation of ectopic vascular structures is significantly decreased (Fig. 5 E and F).

Altogether these results indicate that, while SCs are not themselves giving rise to melanotroph tumors in p27^−/− IL, SOX2 activity in the SCs and consequently the SCs themselves are required for tumorigenesis in p27^−/− IL.

Characterization of the Transcriptional Consequences of Altering Sox2 Levels in p27^−/− IL SCs.

To analyze the role of SOX2 in SCs, we performed single cell RNA sequencing analysis (scRNAseq) on ILs from 3-mo-old Sox9^{iresCreERT2/+}, Sox9^{iresCreERT2/+}; p27^−/−, and Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/− animals, where Cre activity had been induced at birth. Dataset clustering was performed by generating uniform manifold approximation and projection (UMAP) plots (Fig. 6A and SI Appendix, Fig. S6A). The three datasets were integrated and clusters identified according to the expression of known DEGs, and by comparison with published studies (24, 31). The major cell cluster corresponds to melanotrophs as expected (Tbx19, Pomc, and Pax7 positive). The SC cluster (Sox2 and Sox9 positive) is proportionally larger in our IL dataset in comparison with studies performed on the whole pituitary (24, 31). We independently reanalyzed the SC (Fig. 6 B and C) and melanotroph clusters (Fig. 6 F and G). Within each cell type we observed that cells were mostly, but not perfectly, clustered according to their genotype. We thus aimed at examining DEGs between the clusters to characterize the direct effect of Sox2 dosage on SCs, and the consequences on melanotrophs. More precisely, we analyzed DEGs in Sox9^{iresCreERT2/+}; p27^−/− compared to Sox9^{iresCreERT2/+} and Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/−.

Fig. 6.

Characterization of the transcriptomic effects of Sox2 dosage reduction in SCs in p27^−/− IL by single cell analysis. (A) Identification of clusters in integrated UMAP. (B) Subclustering of the stem cell fraction (0.9 resolution) shows that segregation only partially correlates with genotype (Right). (C) Representation of the proportion of cells of the indicated genotype in each cluster, Sox9^{iresCreERT2/+} (n = 137 cells), Sox9^{iresCreERT2/+}; p27^−/− (n = 104 cells), and Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/− (n = 82 cells). (D) Heatmap showing expression of 14 genes selected out of 36 differentially expressed genes (adjusted P value <0.05) in Sox9^{iresCreERT2/+}; p27^−/− compared with Sox9^{iresCreERT2/+} or Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/−. (E) Immunohistochemistry for p-ERK on Sox9^{iresCreERT2/+}, Sox9^{iresCreERT2/+}; p27^−/−, and Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/− pituitaries. While the MAPK/ERK pathway seems overactive in p27^−/− SCs, p-ERK levels return to normal levels upon deletion of one copy of Sox2 in SCs. (F) Subclustering of the melanotroph fraction (0.2 resolution) shows that segregation correlates with genotype (Right). (G) Representation of the proportion of cells of the indicated genotype in the most abundant cluster, Sox9^{iresCreERT2/+} (n = 1,272 cells), Sox9^{iresCreERT2/+}; p27^−/− (n = 3,561 cells), and Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/− (n = 3,006 cells). (H) Strategy followed for scRNAseq analysis of Sox9^{iresCreERT2/+}; p27^−/− vs. Sox9^{iresCreERT2/+} or/and Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/− melanotrophs. (I, Upper) Pathway analysis on genes differentially expressed (adjusted P value <0.05) in melanotrophs following the strategy delineated in I. Pathways associated with secretory function and cytoskeleton are down-regulated in Sox9^{iresCreERT2/+}; p27^−/− cells. (Bottom) Process network analysis performed on genes differentially expressed between Sox9^{iresCreERT2/+}; p27^−/− and Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/− melanotrophs (adjusted P value <0.05). Processes associated with translation and transcription are down-regulated in Sox9^{iresCreERT2/+}; p27^−/− melanotrophs, while genes promoting cell cycle progression are up-regulated. (J) RT-qPCR analysis of Tbx19 levels in 5- to 12-mo-old Sox9^{iresCreERT2/+}, Sox9^{iresCreERT2/+}; p27^−/−, Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/−, and p27^−/−; Sox2^+/− IL. Tbx19 levels are increased in Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/− (0.99 ± 0.20, n = 3, *P = 0.02) and p27^−/−; Sox2^+/− (1.07 ± 0.15, n = 3, **P = 0.0044) vs. Sox9^{iresCreERT2/+}; p27^−/− (0.58 ± 0.28, n = 4).

In SCs, (Fig. 6 B and C) 36 DEGs were identified by comparing subcluster 1 (which has the highest proportion of p27^−/− cells) to the two other subclusters (Dataset S3). Thirty of the DEGs were up-regulated in subcluster 1, and among them, 14 have been associated with tumorigenesis (Fig. 6D). Among these, 5 are immediate early response genes (Ier2, Fos, Junb, Jun, and Egr1) that can be activated by the MAPK pathway. This pathway was also activated in our p27^−/− bulk RNAseq dataset (Fig. 1D). In embryonic pituitary progenitors, activation of the MAPK pathway is associated with expansion of this compartment (32). We thus performed immunohistochemistry for phosphorylated ERK and observed a clear and specific up-regulation of the signal in p27^−/− SCs, in contrast with both control and Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/− samples (Fig. 6E). This suggests that SOX2 positively regulates activity of the pathway in SCs, either directly or indirectly, and this correlates with tumorigenesis.

Subclustering of the melanotrophs, which were more numerous than the SCs, was better associated with the different genotypes (Fig. 6 F and G). In agreement with the absence of tumorigenesis, there was a reduction in melanotroph numbers in Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/− compared to Sox9^{iresCreERT2/+}; p27^−/−, but these were still more numerous than in wild type, in agreement with the hyperplasia observed in compound mutants (Fig. 6F). Furthermore, while melanotrophs are genetically equivalent in Sox9^{iresCreERT2/+}; p27^−/− and Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/− samples, their transcriptome is clearly different, further confirming the SOX2-dependent effect of SCs on melanotrophs. There were 597 DEGS in cluster 0 (Sox9^{iresCreERT2/+}; p27^−/−) melanotrophs compared to clusters 2 (Sox9^{iresCreERT2/+}) and 1 (Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/−) (Fig. 6H). Pathway analysis of the 573 down-regulated genes in Sox9^{iresCreERT2/+}; p27^−/− melanotrophs revealed that vesicular and endosomal traffic, and cytoskeleton remodeling, as we observed in our bulk analysis (Figs. 1 C and D and 2B), were predominantly affected, suggesting that the differentiated, secretory phenotype of melanotrophs was restored upon reduction of SOX2 in SCs (Fig. 6 I, Upper). In agreement with this hypothesis, Tbx19 was among the down-regulated genes in Sox9^{iresCreERT2/+}; p27^−/− melanotrophs. Recovery of normal expression levels of Tbx19 upon Sox2 down-regulation in SCs was further validated by quantitative PCR (Fig. 6J). Finally, pairwise comparison between clusters 0 and 1, comprising Sox9^{iresCreERT2/+}; p27^−/− and Sox9^{iresCreERT2/+}; Sox2^fl/+; p27^−/− cells, respectively, revealed a significant enrichment in process networks associated with translation and transcription (Fig. 6 I, Bottom), suggesting reduced levels of both transcription and translation in p27^−/− cells, restored in melanotrophs upon reduction of Sox2 levels in SCs. SCs and CSCs are known to have reduced levels of translation. Moreover, SOX2 was recently shown to be directly involved in alterations in the rate of protein synthesis in a mouse model of squamous tumor initiating cells (33). This suggests that SOX2 may play a similar role in p27^−/− melanotrophs, acting as a repressor of translation and transcription of genes related to differentiation.

Altogether these data show that SOX2 up-regulation in p27^−/− SCs results in the induction of a dedifferentiated, protumoral phenotype in melanotrophs. In SCs, SOX2 overexpression leads to increased MAPK pathway activation, which correlates with and may mediate the role of SOX2 in tumorigenesis, at least in p27^−/− IL.