GFAP-directed Inactivation of Men1 Exploits Glial Cell Plasticity in Favor of Neuroendocrine Reprogramming

Abstract

Background & aims: Efforts to characterize the signaling mechanisms that underlie gastroenteropancreatic neoplasms (GEP-NENs) are precluded by a lack of comprehensive models that recapitulate pathogenesis. Investigation into a potential cell-of-origin for gastrin-secreting NENs revealed a non-cell autonomous role for loss of menin in neuroendocrine cell specification, resulting in an induction of gastrin in enteric glia. Here, we investigated the hypothesis that cell autonomous Men1 inactivation in glial fibrillary acidic protein (GFAP)-expressing cells induced neuroendocrine differentiation and tumorigenesis.

Methods: Transgenic GFAP^ΔMen1 mice were generated by conditional GFAP-directed Men1 deletion in GFAP-expressing cells. Cre specificity was confirmed using a tdTomato reporter. GFAP^ΔMen1 mice were evaluated for GEP-NEN development and neuroendocrine cell hyperplasia. Small interfering RNA-mediated Men1 silencing in a rat enteric glial cell line was performed in parallel.

Results: GFAP^ΔMen1 mice developed pancreatic NENs, in addition to pituitary prolactinomas that phenocopied the human MEN1 syndrome. GFAP^ΔMen1 mice exhibited gastric neuroendocrine hyperplasia that coincided with a significant loss of GFAP expression. Men1 deletion induced loss of glial-restricted progenitor lineage markers and an increase in neuroendocrine genes, suggesting a reprogramming of GFAP⁺ cells. Deleting Kif3a, a mediator of Hedgehog signaling, in GFAP-expressing cells attenuated neuroendocrine hyperplasia by restricting the neuroendocrine cell fate. Similar results in the pancreas were observed when Sox10 was used to delete Men1.

Conclusions: GFAP-directed Men1 inactivation exploits glial cell plasticity in favor of neuroendocrine differentiation.

Keywords: Enteric Glia; Gastrinomas; Hedgehog Signaling; KIF3A; Primary Cilia; SOX10.

Graphical abstract

Figure 1

GFAP-directed inactivation of Men1 promotes pancreatic islet hyperplasia and the development of pancreatic neuroendocrine neoplasms. (A) Hematoxylin and eosin (H&E) stains of well differentiated and poorly differentiated tumors in GFAP^ΔMen1 mice compared with WT and heterozygous groups. (B–C) H&E stains of well differentiated GFAP^ΔMen1 PNETs compared with a poorly differentiated PNEC (D). (E) Immunofluorescent staining of a PNET for CHGA and menin. Insets: representative images of a PNET showing absent or cytoplasmic menin in the tumor (white). Immunofluorescent staining of a PNET (F) and PNEC (G) for SYP and the alpha cell-specification factor IRX-2. Insets: IRX-2 is expressed in the nucleus of tumor cells (white). (H–I) Ki-67 and smooth muscle actin (SMA) staining of PNETs compared with a PNEC (J). Proliferative index (PI) is indicated by the percentage of Ki67-positive tumor cells in each HPF. K) H&E stain of a GFAP^ΔMen1 PNET stained for GCG (L) and INS1 (M). (N) Quantitation of Chga and Syp mRNA in WT pancreas and GFAP^ΔMen1 PNETs. N = 3–5 mice per group;∗∗P < .01. Data are represented as mean ± standard deviation. (O) Number of mice presenting with pancreatic neuroendocrine neoplasms as stratified by sex. (P) Number of PNETs per mouse in males vs females. Levels of GCG (Q) INS1 (R) and glucose (S) in sera of WT and GFAP^ΔMen1 mice, with symbols indicating male or female mice with and without tumors. N = 11–12 mice per group; ∗P < .05; ∗∗P < .01. Data are represented as mean ± standard deviation.

Figure 2

GFAP^ΔMen1PNETs exhibit loss of GFAP expression and maintain stemness in 3D culture. (A) Immunofluorescent staining of a WT mouse pancreatic islet for the glial-restricted progenitor lineage marker GFAP, the nerve fiber marker NF-H, SYP, the beta-cell hormone INS1, the alpha-cell markers GCG and IRX-2, and menin. (B) Representative images of a normal islet and GFAP^ΔMen1 PNET stained for GFAP in green and the nerve fiber marker NF-H in red, indicating the absence of GFAP expression in and surrounding the PNET. (C) Quantitation of Gfap, S100b and Men1 mRNA in WT pancreas and GFAP^ΔMen1 PNETs. N = 4–5 mice per group; ∗P < .05; ∗∗P < .01. Data are represented as mean ± standard deviation. (D) Schematic of GFAP^ΔMen1-tdTomato construct for endogenous labeling of GFAP. (E) Immunofluorescent images of frozen pancreas sections from WT GFAP-tdTomato mice co-stained for the glial-restricted progenitor lineage markers GFAP, S100B, and P75^NTR, in addition to NF-H as a negative marker for the glial lineage. Immunofluorescent images were merged on a transmitted light micrograph to distinguish islets from acinar cells. (F) Top to bottom: Macroscopic image of a GFAP^ΔMen1 pancreas, with an arrow indicating the presence of a PNET and phase contrast image of a corresponding PNET organoid. PNET organoids were stained for CHGA, SYP, GCG, INS1, and Ki-67 and imaged by confocal microscopy. Images are representative of PNET organoids derived from 4 tumor-bearing mice (n = 4).

Figure 3

GFAP^ΔMen1mice develop pitNETs that phenocopy human MEN1 prolactinomas. (A) Hematoxylin and eosin (H&E) image of a sagittal section of a GFAP^ΔMen1 pitNET labeled with adjacent brain structures (C.b., Cerebellum; Hyp., hypothalamus; M.b., midbrain). Inset shows a representative macroscopic image of a pitNET with an arrow indicating to the tumor. (B) Higher magnification image of a pitNET showing classical “salt and pepper” nuclei characteristic of well-differentiated NETs. Immunofluorescent staining of a GFAP^ΔMen1 pitNET for SYP and CHGA (C), prolactin (D), and (E) Ki-67 and the pituitary specification factor PIT-1. Green arrow indicates a single Ki-67⁺ cell in the tumor. (F) Representative immunofluorescent image of a GFAP^ΔMen1 pitNET stained for the anterior pituitary hormones ACTH and GH, showing increased expression of the latter. As pitNETs were highly vascularized, immunofluorescent images were merged on a transmitted light micrograph to distinguish tumor cells from red blood cells (Cap., Capillary). (G) Number of mice presenting with pitNETs as stratified by sex. (H) Serum prolactin levels in male and female mice with or without pitNETs. N = 12–15 mice per group; ∗P < .05 by the unpaired Student t test. Data are represented as mean ± standard deviation. (I) Quantitation of neuroendocrine-related transcripts in WT pituitary and GFAP^ΔMen1 pitNETs. N = 4–5 mice per group, with the exception of the WT pituitary group, which represents 4 samples of 3 pooled pituitaries, for a total of 12 tissues in this group; ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001. (J) PitNETs show absent expression of menin and GFAP compared with the adjacent hypothalamus (Hyp). Left panel: a capillary (Cap.), characterized by auto fluorescent red blood cells, separates the 2 compartments. Middle panel: Robust GFAP staining (red) in a GFAP^ΔMen1 hypothalamus with few cells co-expressing menin (green). Right panel: Tumor is negative for menin and GFAP. (K) Quantitation of Gfap and Men1 mRNA in WT pituitary and GFAP^ΔMen1 pitNETs. Data are represented as mean ± standard deviation. (L) Representative images of tumor neurospheres generated from 3 tumor-bearing mice (n = 3). Phase contrast images show growth at days (d) 2, 10, and 21. Tumor neurospheres were stained for prolactin (PRL) and the neural stem cell markers nestin and SOX2. (M) Specificity of select antibodies was tested by preadsorption. Immunofluorescent images of a pitNET stained with non-adsorbed and pre-adsorbed GH and SOX2 antibodies.

Figure 4

GFAP-directed deletion of Men1 stimulates hyperplastic reprogramming of the gastro-intestinal epithelium and promotes neuroendocrine cell hyperplasia. (A) Number of mice presenting with gastric hyperplasia and antral tumors, as stratified by sex. (B) Representative hematoxylin and eosin (H&E) and immunofluorescent images of WT and GFAP^ΔMen1 corpus stained for CHGA, SYP, and gastrin. Bottom H&E indicates parietal cell atrophy consistent with metaplasia. (C) Representative H&E and immunofluorescent images of WT and GFAP^ΔMen1 gastric antrum showing increased numbers of CHGA- (green), SYP- (red), and gastrin- (white) positive cells. Colored arrows indicate the degree of co-localization amongst the 3 markers: yellow indicates cells where all 3 markers are expressed, whereas magenta indicates cells expressing gastrin and SYN only. (D) Expression of Chga, Syp, and Gast mRNA in WT and GFAP^ΔMen1 corpus (CP), gastric antrum (AT), and duodenal mucosa (DUO). N = 6 mice per group; ∗P < .05; ∗∗P < .01. (E) Average number of CHGA-, SYP-, and gastrin-positive cells per crypt counted across three 400× fields in the mid-antrum. N = 7 mice per group. ∗P < .05 by 2-way analysis of variance. (F) GFAP^ΔMen1 mice exhibit elevated serum gastrin levels compared with WT littermates. N = 12–15 mice per group; ∗P < .05 by unpaired t test. (G) Evaluation of gastric acidity in GFAP^ΔMen1 mice and controls. N = 8–9 mice per group. ∗P < .05 by unpaired t- test. The expression of non-enteroendocrine cell transcripts was evaluated in the corpus (H) and gastric antrum (I) of WT and GFAP^ΔMen1 mice. N = 6–8 mice per group; ∗P < .05; ∗∗P < .01 by 2-way analysis of variance. All data are represented as mean ± standard deviation.

Figure 5

Conditional deletion of menin in GFAP⁺cells stimulates reprogramming from a glial-restricted progenitor lineage. (A) Whole tissue mounts of proximal duodenum from Cre-negative and tdTomato-expressing mice showing tdTomato fluorescence localized to the myenteric plexus (MP). (B) Immunofluorescent images of frozen stomach sections from WT GFAP-tdTomato mice co-stained for the glial-restricted progenitor lineage markers GFAP, S100B, and p75^NTR, showing strong localization to the same cell types. Negative localization with the nerve fiber marker NF-H serves as a control. Immunofluorescent images were merged on a transmitted light micrograph to distinguish submucosal layers. Widefield images (C) and quantitation (D) of tdTomato signal in the stomach and duodenum of WT and GFAP^ΔMen1 mice expressing tdTomato reporter. N = 4–5 mice per group; ∗∗∗P < .001; ∗∗∗∗P < .0001. (E) Representative images of cryosections of corpus (CP), gastric antrum (AT), and proximal duodenum (DUO) from WT and GFAP^ΔMen1 mice expressing tdTomato. (F) Co-immunoprecipitation (Co-IP) of menin from rat EGC lysate followed by Western blot for GFAP. Input is 5% of lysate used for IP. (G) Quantitation of band density in 3 Co-IP experiments comparing expression of GFAP with IP with IgG isotype control in cell extracts. (H) Expression of Men1 and the glial transcripts Gfap and S100b following siRNA-mediated Men1 silencing in cultured rat EGCs. N = 3 independent experiments; ∗P < .05; ∗∗∗P < .001; ∗∗∗∗P < .0001. (I) Western blot of menin and GFAP in whole cell lysates following 72-hour Men1 silencing in rat EGCs with (J) quantitation of band density normalized to loading control. N = 3 experiments; ∗∗P < .01; ∗∗∗P < .001 by unpaired t test. All data are represented as mean ± standard deviation.

Figure 6

Transcriptome-wide analysis of gastric neuroendocrine hyperplasia and NENs identifies a role for glial-to-neural reprogramming in epithelial differentiation. (A) Volcano plot of significant DEGs in gastric antra of wild type and GFAP^ΔMen1 mice. (B) Volcano plot of significant DEGs in pooled WT pituitaries and pitNETs of GFAP^ΔMen1 mice. Heat maps of significant DEGs mapped to neuroendocrine differentiation in GFAP^ΔMen1 gastric antra (C) and pitNETs (D). Heatmap of significant DEGs mapped to the neuroglial lineage in GFAP^ΔMen1 gastric antra (E) and pitNETs (F) compared with littermate WT controls. Arrow indicates GFAP as among the genes most significantly downregulated in tumors. (G) Immunofluorescent staining for GRP and SYP in WT and GFAP^ΔMen1 gastric antra, shown as green and red respectively. Insets: Co-localization of GRP and SYP in the gastric mucosa (yellow). (H) Immunofluorescent staining for GRP and the nerve fiber marker NF-H, indicating localization of NF-H to the myenteric plexus, with no overlap among GRP⁺ cells of the mucosa. (I) Staining of a WT mouse gastric antrum with non-adsorbed and pre-adsorbed GRP antibody to confirm antibody specificity. (J) Quantitation of Syp and Grp mRNA following 48-hour siRNA-mediated silencing of Men1 in rat EGCs. N = 3 independent experiments; ∗P < .05 and ∗∗P < .01 by unpaired Student t test. Data are represented as mean ± standard deviation.

Figure 7

Transcriptome analysis identifies potential pathways in neuroendocrine differentiation leading to gastric neuroendocrine hyperplasia and NET development in GFAP^ΔMen1mice. Heat maps of significant DEGs mapped to cell cycle regulation (A) and development (B) in GFAP^ΔMen1 pitNETs compared with WT pituitaries. (C) Validation of select DEGs in pitNETs vs WT pituitary by qPCR. N = 4–5 mice per group. WT pituitary group represents 4 samples of 3 pooled pituitaries, for a total of 12 tissues in this group. ∗P < .05; ∗∗∗P < .001 by 2-way analysis of variance. (D) Cross-validation of select pitNET-enriched transcripts in GFAP^ΔMen1 PNETs and WT pancreas. Data are represented as mean ± standard deviatin. (E) Kyoto Encyclopedia of Genes and Genomes (KEGG) Ontology Pathway analysis of wild type pituitary and GFAP^ΔMen1 pitNETs showing the number of genes mapped to enriched pathways and their level of statistical significance. (F) KEGG Ontology Pathway analysis of gastric antra from WT and GFAP^ΔMen1 mice showing the number of genes mapped to enriched pathways and their level of statistical significance. (G) Heat map showing enrichment of kinesin motor protein transcripts in GFAP^ΔMen1 pitNETs. Heat maps depicting downregulation of cytokeratins (H) and immune-related transcripts (I) in GFAP^ΔMen1 antral extracts.

Figure 8

Blockade of SHH signaling in GFAP⁺cells attenuates gastric neuroendocrine differentiation in GFAP^ΔMen1mice. (A) Representative hematoxylin and eosin (H&E) and immunofluorescent images of corpus and gastric antrum of GFAP^ΔMen1;Sst^-/-, GFAP^{ΔMen1;ΔKif3a};Sst^-/- and littermate control mice showing reduced numbers of CHGA- and gastrin-positive cells with additive Kif3a deletion. CHGA (B) and gastrin (C) mRNA expression is reduced in the gastric antra following Kif3a deletion. (D) Expression of antral gastrin peptide among the different genotypes as evaluated by enzyme immunoassay. N = 7–8 mice per group; ∗∗P < .01; ∗∗∗P < .001. All data are represented as mean ± standard deviation.

Figure 9

Loss of the ciliary motor protein KIF3A in GFAP-expressing cells impairs SHH signaling and GFAP expression. (A) Immunofluorescent staining for acetylated-tubulin (Ac-TUB) and GFAP in the gastric antra of WT and GFAP^{ΔMen1;ΔKif3a};Sst^-/- mice. Arrows indicate primary cilia marked by acetylated tubulin. (B) Quantitation of primary cilia length in GFAP-expressing cells of the gastric antrum. N = 34 cells counted across 10 random 1000× magnification images per group. ∗∗∗∗P < .0001 by the unpaired Student t test. (C) Western blot of GLI2 protein in gastric antral lysates of Sst^-/- (Con), GFAP^ΔMen1;Sst^-/-, and GFAP^{ΔMen1;ΔKif3a};Sst^-/- mice. N = 3 mice. (D) Gli2 mRNA in gastric antra and duodenal mucosa of respective genotypes. N = 4–7 mice per group. (E) Representative images of cryosections of corpus (CP), gastric antrum (AT), and proximal duodenum (DUO) from WT, GFAP^ΔMen1;Sst^-/- and GFAP^{ΔMen1;ΔKif3a};Sst^-/- mice expressing tdTomato. (F) Quantitation of relative tdTomato fluorescence intensity following whole tissue ex vivo imaging. ∗P < .05; ∗∗P < .01 by 2-way analysis of variance. (G) Quantitation of GFAP mRNA in the gastric antrum and duodenum of different groups. N = 4–7 mice per group. All data are represented as mean ± standard deviation.

Figure 10

SOX10-driven deletion of Men1 recapitulates hyperplastic reprogramming of the gastric epithelium and leads to accelerated development of PNETs. (A) Survival curve for Sox10^ΔMen1 mice prior to delaying the age of weaning from 3 weeks of age to 4 weeks. Body weight (B) and gastric pH (C) of 10-week-old Sox10^ΔMen1 mice prior to adjusting the age of weaning from 3 weeks to 4 weeks of age. Blue triangles = male mice; Red circles = female mice. (D) Number of Sox10^ΔMen1 mice presenting with PNETs by 11 months of age, as stratified by sex. Levels of serum GCG (E) glucose (F) and prolactin (G) in Sox10^ΔMen1 mice compared with littermate controls, with symbols indicating male and female mice with and without the tissue-involved tumors. N = 6–7 mice per group; ∗∗P < .01 by the unpaired Student t test. (H) Representative hematoxylin and eosin and immunofluorescent images of a well-differentiated Sox10^ΔMen1 PNET stained for neuroendocrine and hormone markers. Hormone-expressing tumor cells are negative for expression of the nerve fiber marker NF-H. (I) Macroscopic images of tissues from a 21-month-old Sox10^ΔMen1 mouse presenting with corpus hyperplasia, multiple antral adenocarcinomas, and a PNET. (J) Hematoxylin and eosin and immunofluorescent staining of the previous stomach tissues indicating the presence of a gastric NET and neuroendocrine cell hyperplasia. Colored arrows indicate the degree of co-localization of CHGA (green), SYP (red), and gastrin (white), with several gastrin⁺ cells expressing SYP but not CHGA. (K) Serum gastrin peptide levels in 6–11-month-old Sox10^ΔMen1 mice as evaluated by enzyme immunoassay. N = 9–12 mice per group; ∗P < .01 by the unpaired Student t test. All data are represented as mean ± standard deviation.