Mice lacking neurofibromin develop gastric hyperplasia

Abstract

Gastrointestinal (GI) neoplasms are among many manifestations of the genetic disease neurofibromatosis type 1 (NF1). However, the physiological and pathological functions of the Nf1 gene in the GI system have not been fully studied, possibly because of a lack of mouse models. In this study, we generated conditional knockout mice with Nf1 deficiency in the GI tract. These mice develop gastric epithelial hyperplasia and inflammation together with increased cell proliferation and apoptosis. The gastric phenotypes observed in these mutant mice seem to be the consequence of loss of Nf1 in gastric fibroblasts, resulting in paracrine hyperactivation of the ERK pathway in the gastric epithelium. These mice provide a useful model to study the pathogenesis of GI lesions in a subset of patients with NF1 and to investigate the role of the Nf1 gene in the development of GI neoplasms.

neurofibromatosis type i (NF1) is an autosomal dominantly inherited disorder that appears in about 1/3,500 births worldwide (36). Major clinical features of this disorder are café-au-lait spots and benign tumors of peripheral and optic nerves. Less penetrant manifestations include lesions involving the skin, bones, endocrine organs, and blood vessels (25). The gastrointestinal (GI) tract can also be affected in these patients; gastric tumors are observed in about 2–25% of patients with NF1 with variable tumor locations and tumor types, including those of epithelial origin (adenoma and adenocarcinoma), stromal origin (GI stromal tumor), and others (34, 37, 40, 48). In rare cases, GI tumors can cause pyloric obstruction (2). Nonneoplastic lesions in the GI tract such as inflammatory fibroid polyps and hyperplastic polyps have also been reported in patients with NF1 (12, 32); however, the pathophysiological features and etiological mechanisms of these lesions have not been investigated.

NF1 is caused by mutation in the Nf1 gene (51). Neurofibromin, the protein encoded by Nf1, functions as a negative regulator of Ras-mediated signaling pathways by stimulating conversion of the active GTP-bound form to the inactive GDP-bound form (5). Two major signaling pathways downstream of Ras are the mitogen-activated protein kinase kinase (MEK)/extracellular-regulated kinase (ERK) pathway, which is involved in proliferation, apoptosis, and differentiation, and the phosphatidyl-inositol-3-kinase/AKT pathway, which is important for cell survival (59). Both pathways can thus be regulated or modified by neurofibromin. Nf1 has cell-autonomous effects in different cell types. For example, homozygous Nf1 mutant (Nf1−/−) embryonic sensory neurons from mice can survive in the absence of neurotrophic support, and adult sensory neurons with Nf1 deficiency can grow longer neurites both in vitro and in vivo after injury (38, 50). Heterozygous Nf1 mutant (Nf1+/−) astrocytes exhibit increased proliferation compared with controls (1). In addition, Nf1 deficiency can also lead to noncell-autonomous paracrine effects. For example, Nf1−/− Schwann cells secrete Kit ligand (KitL), which stimulates mast cell migration, and Nf1+/− mast cells can secrete increased amount of transforming growth factor-β (TGF-β) (55, 56); the interaction of these two cell types is believed to be important for the development of neurofibroma (57). Also, Nf1+/− microglia can secrete insulin-like growth factor-1 (IGF-1) and hepatocyte growth factor (HGF) to stimulate growth of Nf1−/− astrocytes and glioma (9).

Gastric cancer is the second most prevalent cancer in the world and the second most common cause of global cancer death (14, 49). Existing rodent models to study the pathogenic mechanisms of gastric tumors include mice infected with Helicobacter pylori, mice with disturbed levels of gastrin, a hormone that can stimulate acid secretion in stomach, and mice with genetic modification (11, 52, 58). The targeted genes in mice that develop gastric hyperplasia include ion transporters (H/K-ATPase b subunit, Na/K exchanger 2, Kvlqt1 channel), signal transducers (TGF-α, sonic hedgehog), transcription factors (Fox1, NF-κB), and cell adhesion molecules (13, 17, 21, 26, 31, 35, 41–44). Moreover, gastric cancers are observed in mice deficient for Trefoil factor 1, Smad4, or Runx3 and mice with point mutations in IL-6 cytokine coreceptor gp130 (28, 30, 47, 54).

Nf1 traditional knockout mice are embryonic lethal attributable to defects in multiple organs (18), precluding investigation into the physiological and pathological role of the Nf1 gene in the GI system. Using mice carrying two Nf1 flox alleles and a Cre transgene specifically expressed in the GI tract, we demonstrate that region-specific deletion of Nf1 in the GI tract leads to gastric epithelial hyperplasia and inflammation together with increased cell proliferation and apoptosis. These phenotypes appear to be mediated by noncell-autonomous contribution from Nf1-deficient gastric fibroblasts. Hyperactivation of the ERK pathway is detected in the mutant gastric epithelium. These mice can thus be useful animal models to study the pathogenesis of GI lesions in a subpopulation of patients with NF1, as well as to investigate the role of the Nf1 gene in gastric tumor initiation and/or progression.

MATERIALS AND METHODS

Breeding and genotyping of Nf1^flox/flox and Islet1-Cre mice.

Nf1^flox/flox mice were crossed with mice carrying Islet1-Cre (provided by Dr. Thomas Jessell at Columbia University) to generate Nf1^flox/flox;Islet1-Cre conditional knockout mice (Nf1-Isl1-CKO). Rosa26^lacZ;Islet1-Cre mice (R26-Isl1) were generated by crossing Islet1-Cre mice with Rosa26-Stop-lacZ reporter mice (Jackson Laboratories, Bar Harbor, ME). All mice were kept on a mixed genetic background. Age-matched littermates with the genotypes of Nf1^flox/flox, Nf1^flox/wt, and Nf1^flox/wt;Islet1-Cre were phenotypically indistinguishable and were thus all pooled as controls. Genotyping for the Nf1^flox allele and Cre transgene was done by PCR as previously described (60). All mouse protocols were approved by the Institutional Animal Care and Research Advisory Committee and adhered to guidelines of University of Texas Southwestern Medical Center.

X-gal staining.

β-gal activity was examined in R26-Isl1 mice at different ages. Embryos were collected in PBS and then fixed in 2% paraformaldehyde (PFA) for 1 h at 4°C. Postnatal mice were transcardially perfused with PBS and 2% PFA, and organs were then carefully dissected out, washed with PBS, and postfixed in 2% PFA for 1 h at 4°C. For sections, after fixation, embryos or organs were protected in 30% sucrose until sunk and cryostat sections were cut at 14 μm. X-gal staining for whole-mount tissues or sections was performed as previously described (38).

Detection of recombined Nf1 allele.

Genomic DNA was extracted from tissues or cells. Recombination of the Nf1 allele was examined as previously described (24). The PCR primers were 5′-AATGTGAAATTGGTGTCGAGTAAGGTAACCAC-3′, 5′-TTAAGAGCATCTGCTGCTCTTAGAGGGAA-3′, and 5′-TCAGACTGATTGTTGTACCTGATGG TTGTACC-3′. The sizes of the PCR products were Nf1wt allele = 493 bp, Nf1flox allele = ∼600 bp, and recombined Nf1 allele = ∼300 bp.

Histology.

Mice were transcardially perfused with PBS and 4% PFA. Stomachs were removed and cut along the greater curvature, pinned out, and photographed. After paraffin processing, tissues were sectioned at 5 μm and stained with hematoxylin and eosin, periodic acid-Schiff staining (PAS; Sigma, St. Louis, MO), or Alcian-blue staining (pH 2.5, Sigma) according to manufacturer's instructions.

Immunohistochemistry.

Paraffin sections were deparaffinized and rehydrated. After microwave antigen retrieval, sections were incubated with one of the following primary antibodies: H⁺-K⁺ ATPase (mouse, 1:1,000; Sigma), pepsinogen II (sheep, 1:200; Abcam, Cambridge, MA), gastrin (rabbit, 1:400; Dako, Glostrup, Denmark), PCNA (mouse, 1:10,000; Santa Cruz Biotechnology, Santa Cruz, CA), smooth muscle actin (SMA, mouse, 1:500; Dako), vimentin (mouse, 1:500; Invitrogen, Carlsbad, CA), phospho-p44/42 MAP Kinase (pERK, rabbit, 1:200; Cell Signaling Technology, Danvers, MA), phospho-AKT (rabbit, 1:50; Cell Signaling Technology), and phospho-signal transducer and activator of transcription 3 (STAT3) (rabbit, 1:200; Cell Signaling Technology). The visualization of primary antibodies was performed with a horseradish peroxidase system (Vectastain ABC kit; Vector Laboratories, Burlingame, CA). Digital images were captured with a CCD camera using bright-field illumination (Olympus).

Immunofluorescent staining on cryostat sections was carried out by incubating with primary antibodies in 5% normal donkey serum overnight at 4°C. Primary antibodies included lacZ (rabbit, 1:2,000; ICN, Costa Mesa, CA), Tuj1 (mouse, 1:500; Covance, Berkeley, CA), SMA (mouse, 1:50; Dako), and vimentin (mouse, 1:50; Invitrogen). Immunofluorescence was visualized after 1-h incubation with appropriate Cy2- or Cy3-conjugated secondary antibodies (1:200, Jackson Laboratories) at room temperature. DAPI (1 μg/ml; Fluka, Buchs, Switzerland) was added together with the secondary antibodies to stain nuclei. The immunostaining was visualized using a Nikon Eclipse TE2000-U fluorescence microscope. Digital images were analyzed using the Metamorph software (Universal Imaging, Downingtown, PA).

Morphometric and statistical analysis.

At least three mice each for control and Nf1-Isl1-CKO were included at each time point of every measurement. Image J software was used to analyze the data. For gland length measurement, the average value of at least 10 longitudinally sectioned glands in the pyloric region was shown for each animal. Gastric cell types were examined as previously described (22). The number of parietal cells (H⁺-K⁺ ATPase positive), chief cells (pepsinogen II positive), and total cells was quantified in 10 randomly selected glands in the corpus of the stomach, whereas the number of gastrin-positive cells was counted in the antrum. Results are shown as the average number of cells per gland.

Cell proliferation.

Mice were injected intraperitoneally with 100 μg 2-bromo-5′-deoxyuridine (BrdU, Sigma)/g of body weight 2 h before euthanasia. Anti-BrdU staining was performed as described above using an antibody against BrdU (mouse, 1:100; DAKO). The average number of BrdU-positive cells in at least 10 longitudinally sectioned glands was reported. All cell counts were performed in a genotype-blind manner.

Apoptosis.

Detection of apoptotic cells was performed as described above using an antibody against cleaved Caspase-3 (rabbit, 1:400; Cell Signaling Technology). Epithelial cells positively stained for cleaved Caspase-3 were counted in at least 10 randomly selected fields per section. All cell counts were performed in a genotype-blind manner.

Gastric fibroblast culture.

Primary gastric fibroblast cultures were established as previously described with modification (33). Briefly, stomachs were removed from postnatal day 7 (P7) animals, and the mucosa layer was separated from the muscular layer by forceps. The mucosa was then minced and digested in collagenase/dispase (Roche, Indianapolis, IN) for 30 min followed by trypsin digestion for 15 min at 37°C. The tissues were then triturated to single cell solution and plated onto six-well plates. Cells were maintained in growth medium containing DMEM, 10% fetal bovine serum, and 1% penicillin/streptomycin in 5% CO₂ at 37°C. Medium was changed 3 days after plating and every other day thereafter. The identity of fibroblasts was confirmed by immunostaining for SMA (mouse, 1:50; Dako) and vimentin (mouse, 1:50; Invitrogen). Cells at passage 2 were used for subsequent experiments.

Statistical analysis.

Statistical analysis was done using Student's t-test with P values ≤ 0.05 as being statistically significant. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001. Error bar represents means ± SE.

RESULTS

Islet1-Cre is expressed in the GI tract.

Islet1 is a transcription factor widely expressed in neural crest derivatives such as sensory ganglia, cardiovascular progenitors, and foregut endoderm (46). We examined the expression and activity of an Islet1 regulatory element-controlled Cre recombinase transgene (Islet1-Cre) using Rosa26-stop-lacZ reporter mice (R26-Isl1) and found it to be also prominently expressed in the GI tract.

At embryonic day 14.5 (E14.5), β-gal activity was detected throughout the stomach and proximal duodenum of R26-Isl1 embryos as revealed by whole-mount X-gal staining (Supplemental Fig. S1A; supplemental material for this article is available online at the American Journal of Physiology Gastrointestinal and Liver Physiology website). Histological analysis showed that β-gal activity was mainly restricted to the muscular layer and lamina propria of the GI tract at E14.5 (Supplemental Fig. S1B) as well as at P1 (Supplemental Fig. S1C). Double immunofluorescence staining was then performed to confirm the identity of β-gal-positive, Cre-expressing cells. We observed that lacZ staining colocalized with SMA, a mesenchymal marker (Supplemental Fig. S1D), but not with Tuj1, a neuronal marker (Supplemental Fig. S1E), suggesting that Islet1-Cre is not expressed in enteric neurons, which are also derivatives of neural crest cells (15). At P14, β-gal activity was observed in lamina propria and muscular layers of the stomach from the R26-Isl1 mice (Fig. 1, A and B). Within the pyloric region, β-gal signal was detected mainly in the muscular layer and periglandular fibroblasts (Fig. 1, B and C), the identity of which was confirmed by double immunofluorescent staining of lacZ and SMA (Fig. 1, D–F) or vimentin (data not shown), markers for smooth muscle cells and fibroblasts (27). In regions distant from the pylorus, β-gal signal was also observed in some epithelial cells (data not shown). Consistent with endogenous Islet1 expression and with a previous report, β-gal activity was also found outside the GI tract, namely in the trigeminal ganglia, dorsal root ganglia, hind limbs, and spinal cord (data not shown) (45).

Fig. 1.

Islet1-Cre is expressed in the gastrointestinal tract. At postnatal day 14 (P14), β-gal activity was detected in the stomach of Rosa26^lacZ;Islet1-Cre (R26-Isl1) mice (B, left brackets indicate the mucosal layer and right brackets indicate the muscular layer. Sto, stomach; Duo, duodenum), but not in controls (A). C: higher magnification of the square in B showed β-gal activity mainly in gastric fibroblasts (arrows) and muscular layer within the pylorus. D–F: double immunofluorescent staining showed colocalization of lacZ (red, arrows) and smooth muscle actin (SMA) (green, arrows), suggesting that Islet1-Cre is expressed in gastric fibroblasts and smooth muscle cells. G: PCR showed the presence of recombined neurofibromatosis type 1 (Nf1) alleles in pyloric tissues from P30 Nf1-Isl1-CKO (Nf1^f/f;Cre+) mice, where CKO is conditional knockout. PCR on gastric fibroblasts from Nf1-Isl1-CKO mice (CKO-GF) amplified only the recombined Nf1 allele, whereas only the Nf1^flox allele was amplified in control mice (Con-GF). Arrows indicate three different bands, namely wild-type Nf1 allele (Nf1wt), Nf1 allele with loxP sites (Nf1flox), and recombined Nf1 allele (Nf1 delta). H: primary gastric fibroblast cultures were established from P7 control and Nf1-Isl1-CKO mice, and the identity of these cells was confirmed by immunofluorescent staining with SMA antibodies. Scale bars = 500 μm in A and B, 50 μm in C, 100 μm in D–F and H.

Thus, in addition to the anticipated neural crest expression, Islet1-Cre is expressed in gastric fibroblasts and gastric smooth muscle within the pyloric region, as well as in some epithelial cells residing in the nonpyloric area of the stomach.

Nf1-Islet1-CKO mice exhibit progressive body weight loss and morbidity.

To study the role of Nf1 in the GI system, we crossed Nf1^flox/flox mice to Islet1-Cre mice to generate Nf1 conditional knockout animals (Nf1-Isl1-CKO). We first examined the efficiency of Islet1-Cre-mediated recombination in the GI tract. DNA was extracted from whole pyloric tissue of P30 control and Nf1-Isl1-CKO mice and subjected to PCR analysis. The recombined Nf1 allele was detected in Nf1-Isl1-CKO tissues (Fig. 1G), indicating that Islet1-Cre-mediated recombination was effective. Primary gastric fibroblast cultures were then established from control and Nf1-Isl1-CKO mice. The identity of these cultured cells was confirmed by positive staining for SMA (Fig. 1H) and vimentin (data now shown). DNA amplified from the mutant gastric fibroblast cultures showed complete recombination of Nf1 alleles (Fig. 1G), suggesting that most gastric fibroblasts in the mutant mice were deficient for Nf1.

Nf1-Isl1-CKO mice were born in Mendelian ratios with no obvious abnormalities at birth. However, starting as early as 1 wk after birth, they began to exhibit significant loss of body weight compared with their littermate controls. The loss increased with age (Fig. 2A). The majority of Nf1-Isl1-CKO mice died between 2 and 3 mo of age (mean survival time = 70 days), possibly attributable to malnutrition; the mutant mice had no fat deposits under their skin, and few survived longer than 4 mo (data not shown and Fig. 2B). The premature death and age-dependent weight loss in these mice cannot be attributed to the Islet1-Cre transgene because no significant difference was observed in survival rate or body weight of Nf1^flox/wt; Islet1-Cre, Nf1^wt/wt;Islet1-Cre, and non-Cre animals (Fig. 2, A and B). These phenotypes were also not sex specific because both female and male Nf1-Isl1-CKO mice exhibited similar survival rate and progressive body weight loss (Supplemental Fig. S2, A and B).

Fig. 2.

Nf1-Isl1-CKO mice exhibit decreased survival rate, age-dependent body weight loss, and stomach enlargement. A: Nf1-Isl1-CKO mice (CKO) exhibited significant decrease in body weight as early as 1 wk after birth compared with their littermate controls (HET, Nf1^flox/wt;Islet1-Cre mice; WT, Nf1^wt/wt;Islet1-Cre and non-Cre mice). This weight loss became more severe as the mice aged. B: Nf1-Isl1-CKO mice usually died between 2 and 3 mo of age, with a mean survival time of 70 days (N = 25 for each genotype). C: mice were euthanized at specified ages, and stomachs were removed for macroscopic examination. Stomach from P0, P7, and P90 control and Nf1-Isl1-CKO mice are shown. Enlargement of stomach was observed in Nf1-Isl1-CKO mice as early as 1 wk after birth and became more obvious as the mice aged. Data represent means ± SE. **P < 0.01, ***P < 0.001, Student's t-test. Scale bars = 1 mm.

Nf1-Isl1-CKO mice develop gastric hyperplasia.

To understand the cause of premature mortality in Nf1-Isl1-CKO mice, we performed detailed examination of the mice at different ages. We observed age-dependent enlargement of the stomach in mutant mice compared with littermate controls as early as 1 wk after birth (Fig. 2C), which coincided with the beginning of weight loss in these mice (Fig. 2A). As the animals aged, the stomach enlargement became more severe (Fig. 2C). In mice older than 3 mo, a huge abdominal mass could be observed. Macroscopic examination revealed the presence of antropyloric lesions in Nf1-Isl1-CKO mice with 100% penetrance. Although not noticeable at 2 wk of age (Fig. 3D), these lesions were macroscopically evident by 1 mo of age (Fig. 3E) and, in mice older than 3 mo, often resulted in obstruction of the gastric outlet (Fig. 3F), the probable cause for the weight loss and premature death in these mice. These lesions were never observed in littermate controls (Fig. 3, A–C).

Fig. 3.

Nf1-Isl1-CKO mice develop gastric hyperplasia. Stomachs from control (A–C) and Nf1-Isl1-CKO mice (D–F) at different ages (A and D, P14; B and E, P30; and C and F, P90) were subjected to macroscopic examination. Antropyloric lesions were visible in Nf1-Isl1-CKO mice by 1 mo of age (E, arrowhead) and became more obvious and severe at 3 mo of age (F, arrowhead). G–J: hematoxylin and eosin (H & E) staining of sections from pyloric region of 2-wk-old mice (G and I) and 3 mo-old mice (H and J); control (G and H) and Nf1-Isl1-CKO mice (I and J) are shown. Right brackets indicate the epithelium. K: quantification of average gland length at different time points indicated that in aged Nf1-Isl1-CKO mice the gland length was significantly increased compared with controls. Each dot represents the average length of 10 pyloric glands in 1 mouse; N = 5 for each genotype at each time point. Data represent means ± SE. **P < 0.001, Student's t-test. Scale bars = 1 mm in A–F, 100 μm in G and I, and 200 μm in H and J.

Hematoxylin and eosin staining revealed hyperplasia of gastric mucosa in the pyloric region of Nf1-Isl1-CKO mice compared with controls, starting as early as 2 wk after birth (Fig. 3, I and G, respectively) and becoming more pronounced in older animals (Fig. 3, J and H, respectively). Signs of inflammation were already present in the submucosal layer of gastric mucosa in P14 Nf1-Isl1-CKO mice (Fig. 3I), when no obvious difference in the length of pyloric glands was observed between the mutant and controls (Fig. 3K). In older mutant mice, the gastric glands were enlarged and the pits were elongated (Fig. 4B) compared with controls (Fig. 4A). Infiltrating neutrophils and lymphocytes as well as neovascularization were observed (Fig. 4, C and D). In a few surviving older mutant animals, atypical epithelial cells (Fig. 4, E and F) could be identified.

Fig. 4.

Histological analysis of Nf1-Isl1-CKO gastric epithelium. H & E staining of P90 (B–F) pyloric tissues from Nf1-Isl1-CKO mice revealed hyperplastic and inflammatory changes within the gastric mucosa compared with controls (A). B: hyperplastic changes consisted of elongated and branching glands. Inflammation, as evidenced by infiltrating neutrophils and lymphocytes (arrowhead in C) and the proliferation of small blood vessels (asterisks in D), was also observed. Atypical cells (arrowheads in E and F) were found in aged mutant mice. Sections of pyloric region from P90 Nf1-Isl1-CKO (H and J) and control (G and I) mice were stained with periodic acid-Schiff stain (PAS) (G and H) and Alcian blue (I and J). PAS and Alcian-blue-positive cells were detected in Nf1-Isl1-CKO pyloric glands. Scale bars = 100 μm in A–C and G–J, 50 μm in D–F.

To further study the properties of these hyperplastic lesions, we performed PAS staining and Alcian-blue staining to detect mucin-secreting cells. PAS staining showed that the majority of hyperplastic cells were PAS positive and therefore mucin-secreting pit cells (Fig. 4, G and H). More cells positive for Alcian-blue staining were detected within the pyloric glands of Nf1-Isl1-CKO mice (Fig. 4J) compared with controls (Fig. 4I), suggesting that mucous-secreting cells were expanded in the CKO epithelium.

We also compared different types of gastric epithelial cells in the control and Nf1-Isl1-CKO mice at the age of 1 mo. The number of total cells, parietal cells and chief cells per gland was counted in the corpus of the stomach from control and CKO mice, and no significant difference was observed. The number of gastrin-positive cells in the antrum was also comparable between the two genotypes (Supplemental Table S1).

Taken together, these data suggest that Nf1-Isl1-CKO mice develop progressive gastric hyperplasia of the mucin-secreting epithelial compartment accompanied by inflammation and, in some cases, atypical cellular changes.

Increased cell proliferation and cell death in Nf1-Isl1-CKO gastric epithelium.

The gastric hyperplasia observed in Nf1-Isl1-CKO mice could either be due to increased cell proliferation, decreased cell apoptosis, or some combination thereof in the gastric epithelium. To examine the status of cell proliferation in the gastric epithelium, control and Nf1-Isl1-CKO mice at different ages were intraperitoneally injected with BrdU, a thymidine analog that is incorporated into the DNA of dividing cells during S phase of the cell cycle and thus used as an indicator of cell proliferation. At P14 and P30, the majority of BrdU-positive cells were found localized to the isthmus region of pyloric glands in both control and Nf1-Isl1-CKO mice (Fig. 5, A and B, respectively, and data not shown), a reported location for gastric progenitor cells (11). No significant difference in the number of BrdU-positive cells per pyloric gland was detected between mutant and controls at these time points (Fig. 5E and data not shown). However, in gastric epithelium of P60 Nf1-Isl1-CKO mice (Fig. 5D), the number of BrdU-positive cells significantly increased compared with controls (Fig. 5, C and E), and BrdU-positive cells were found throughout the pyloric glands in the mutant mice. Similar results were found upon immunostaining for PCNA, another marker of proliferating cells (data not shown). BrdU-positive lymphocytes were also found outside the pyloric glands in P60 Nf1-Isl1-CKO mice (Fig. 5D, inset), whereas few nonepithelial cells were positive for BrdU in the controls.

Fig. 5.

Increased proliferation and apoptosis in Nf1-Isl1-CKO gastric epithelium. 2-Bromo-5′-deoxyuridine (BrdU) staining of control (A and C) and Nf1-Isl1-CKO (B and D) gastric epithelium at different stages (A and B, P14; C and D, P60) showed increased proliferation in Nf1-Isl1-CKO mice at older ages (E). Higher magnification showed the presence of BrdU-positive lymphocytes (arrowheads in inset of D) in the CKO mucosa. Apoptotic cells were visualized by staining with antibody against cleaved Caspase-3. Pictures from 2-wk-old mice (F and G) and 3-mo-old mice (H and I) and control (F and G) and Nf1-Isl1-CKO (H and I) mice were shown. Arrows in I indicate the apoptotic cells. J: quantification of cleaved Caspase-3-positive cells showed increased apoptosis in gastric epithelium of 3-mo-old Nf1-Isl1-CKO mice. Data represent means ± SE. *P < 0.05; ***P < 0.001, Student's t-test. Scale bars = 100 μm and 30 μm in inset.

To examine the status of cell death, we performed immunostaining using an antibody against cleaved Caspase-3, which specifically labels apoptotic cells. At P14, no difference in the number of cleaved Caspase-3-positive cells was detected in the gastric epithelium of control and Nf1-Isl1-CKO mice (Fig. 5, F and G, respectively). However, the number of apoptotic cells increased dramatically in the gastric epithelium of mutant mice older than 3 mo compared with controls (Fig. 5, H–J). Because most cleaved Caspase-3-positive cells were located at the surface of the gastric epithelium or in the lumen of the glands (Fig. 5I), this increase of apoptotic cells in mutant mice might be a secondary effect of enhanced proliferation at this time point (Fig. 5C).

Taken together, these results suggest that both cell proliferation and cell death are affected in the gastric epithelium of Nf1-Isl1-CKO mice.

Noncell-autonomous role of Nf1 in gastric epithelium hyperplasia.

Because expression of Islet1-Cre was found mainly in the smooth muscle cells and fibroblasts within the pyloric region (Fig. 1), it is likely that the hyperproliferation of gastric epithelial cells in Nf1-Isl1-CKO mice was brought about by a paracrine effect of Nf1 loss in other cell types rather than by a cell-autonomous role in the epithelial cells themselves. To explore this possibility, we took advantage of the R26-stop-lacZ alleles present in a subset of Nf1-Isl1-CKO mice and performed X-gal staining. Cells with β-gal activity were thus deficient for Nf1. At different time points (P7, P14, and P60), β-gal activity was mainly detected in the muscular layer and the periglandular fibroblasts, while the majority of epithelial cells in the pyloric region lacked β-gal reactivity (Fig. 6, A–D). Immunofluorescence double staining for lacZ and SMA (Fig. 6, E–G) or lacZ and vimentin (data not shown) showed that lacZ-positive cells also expressed SMA and vimentin and thus were fibroblasts and smooth muscle cells.

Fig. 6.

LacZ staining of pyloric lesions in Nf1-Isl1-CKO mice. Sections from P7 (A), P14 (B), and P60 (C and D) Nf1^flox/flox;Rosa2^lacZ;Islet1-Cre mice were subjected to X-gal staining. β-gal activity was detected mainly in muscular layer (right brackets in A–C) and fibroblasts (arrowheads in D) within the pyloric region. Little β-gal activity was observed in the epithelium at different time points. E–G: double immunofluorescent staining revealed colocalization (G, arrows) of lacZ with SMA, a marker for smooth muscles (E: lacZ staining, red; F: SMA staining, green). Scale bars = 100 μm in A–D, 50 μm in E–G.

These results indicate that the loss of Nf1 in gastric fibroblasts or smooth muscle cells may play a role in the proliferative response of gastric epithelial cells in our mouse model. A study in which Nf1 was exclusively deleted in the smooth muscle compartment did not reveal gastric hyperplasia, thus pointing to fibroblasts as the source of paracrine stimulants (29, 53) (see discussion).

Hyperactivation of ERK pathway in Nf1-Isl1-CKO gastric epithelium.

To study the possible molecular mechanisms underlying the observed phenotypes in the Nf1-Isl1-CKO mice, we examined the status of two major signaling cascades involved in gastric hyperplasia and tumor development, the ERK pathway and the STAT3 pathway (10, 14). We found that phospho-ERK was dramatically increased in gastric epithelial cells of P90 Nf1-Isl1-CKO mice compared with controls (Fig. 7, A and B). Despite the limited number of Nf1-deficient epithelial cells in the pyloric region as revealed by X-gal staining (Fig. 7D and Fig. 6D), hyperactivation of the ERK pathway was observed throughout the pyloric glands, suggesting that the activation may be the result of cytokine or growth factor stimulation of the gastric epithelium, rather than the loss of Nf1 in these cells. Preliminary results indicated that phospho-STAT3 signal was also increased in gastric epithelial cells of Nf1-Isl1-CKO mice (data not shown), whereas the level of phospho-AKT, another downstream effector of neurofibromin, did not change in the mutant gastric epithelium compared with controls (data not shown). It is interesting that, in Nf1-Isl1-CKO mice older than 2 mo, BrdU-positive cells could be observed throughout the pyloric glands as well (Fig. 7C and Fig. 5F), suggesting that the increased proliferation in the epithelial cells may also be attributable to external stimulation.

Fig. 7.

The ERK pathway is activated in gastric epithelium of Nf1-Isl1-CKO mice. Immunohistochemical staining of sections from P90 mice revealed increased levels of phospho-ERK (p-ERK) in the gastric epithelium of Nf1-Isl1-CKO mice (B) compared with controls (A). For comparison, BrdU immunohistochemical staining (C) and X-gal staining (D) of P60 Nf1-Isl1-CKO pyloric tissues are shown. Arrowheads in B–D indicate epithelial cells. Scale bars = 100 μm.

DISCUSSION

Gastric epithelial tumors are present in a subset of NF1 patients; however, the specific features and etiological mechanisms underlying these lesions have not been fully studied, possibly because of the lack of animal models. In this study, we generated conditional knockout mice with Nf1 deleted in the GI tract. These mice exhibited progressive loss of body weight and decreased survival. They developed age-dependent stomach enlargement that, in most cases, lead to obstruction of the gastric outlet. Histological examination showed gastric hyperplasia, inflammation, increased cell proliferation, and increased cell death in the mutant gastric epithelium. Atypical epithelial cells were also observed in some aged mutant mice. In addition, the ERK pathway was found to be hyperactivated in the mutant gastric epithelium.

Mouse models of gastric hyperplasia and gastric tumor.

Several rodent models have been developed to study the physiology and pathology of the GI system. The role of the Nf1 gene in GI tract has not been clearly studied possibly because the traditional knockout mice are embryonic lethal (6, 18). Using Islet1-Cre, we deleted Nf1 in gastric smooth muscle cells, gastric fibroblasts, and few gastric epithelial cells. These mice developed gastric hyperplasia and inflammation, with atypical cells in some cases, which are considered hallmarks of precancerous lesions for gastric tumor and are closely associated with tumor formation (11). Nf1-Isl1-CKO mice usually died before 3 mo of age possibly owing to obstruction of the gastric outlet and malnutrition. Thus a reason for the absence of tumor formation in the gastric epithelium could simply be the precocious death caused by the obstruction. This is not surprising because, in other mouse models of gastric cancer, tumor development was not observed until the mice were several months old, although gastric hyperplasia could be observed much earlier (20, 47).

More importantly, Cre-mediated deletion of Nf1 in the GI tract resulted in 100% penetrance of the gastric hyperplasia phenotype in Nf1-Isl1-CKO mice. This strongly suggests a possible role for the Nf1 gene in gastric epithelium and gastric tumor development. This mouse model can thus be used to study the features of gastric epithelial lesions present in a subpopulation of NF1 patients, as well as to study the role of the Nf1 gene in gastric tumor development.

Noncell-autonomous role of Nf1 in gastric hyperplasia development.

The origin of gastric epithelial tumors has been controversial. Gastric epithelial cells themselves obviously play a prominent role in tumor initiation and progression. It has also been reported that genetic modification in lymphocytes is sufficient for gastric tumor development in some mouse models (Smad4, Runx3) (7, 23). Bone marrow stem cells are able to incorporate into the epithelial tumors (16). Additionally, of note, stromal cells play a significant role in the initiation and progression of epithelial tumors. These cells can secrete various factors that affect the proliferation, apoptosis, and transformation of epithelial cells (4). Mice with TGF-β type II receptor deleted specifically in fibroblasts developed squamous cell carcinoma of the forestomach, possibly through paracrine activation of the HGF pathway (3).

In our mouse model, Islet1-Cre expression was mainly restricted to the muscular layer and gastric fibroblasts in the pyloric region, where the lesions developed. The hyperplasia of epithelial cells, in which Cre was not active, was thus not due to an autonomous effect of Nf1 in these cells. It is possible that other types of cells, after losing Nf1, secrete factors to stimulate the growth of epithelial cells. The source of these paracrine factors remains to be further investigated. It is unlikely that the loss of Nf1 in epithelial cells distant from the pyloric region affects the growth of pyloric glands. The loss of Nf1 in gastric smooth muscles cannot result in the gastric hyperplasia that we observed in Nf1-Isl1-CKO mice because conditional Nf1 mutant mice carrying a widely expressed smooth muscle-specific Cre (SM22α-Cre) develop normally (29, 53), demonstrating that Nf1 deletion in the smooth muscle cells only is not sufficient for the mice to develop gastric tumors. The most likely candidates are thus the gastric fibroblasts, which have already been shown to be able to affect growth of the epithelium (4). The loss of Nf1 immortalizes mouse embryonic fibroblasts, which show growth advantage in vitro (8, 39). It would be interesting to investigate the in vivo functions of Nf1−/− fibroblasts and their possible role in gastric hyperplasia development using a fibroblast-specific Cre line (FSP-Cre) (3). We also cannot exclude the possibility that infiltrating lymphocytes and their secreted factors initiates or promotes gastric epithelial hyperplasia because local aggregation of lymphocytes is observed well before the change in the gastric mucosa (Figs. 3 and 4).

The microenvironment is an important contributor to the development of Nf1-related tumors and, indeed, to many forms of cancer. In NF1-related neurofibroma formation, it is the mast cells that are critical for neurofibroma initiation (57). Interestingly, our data indicate a similar situation in which loss of Nf1 in gastric fibroblasts apparently initiates hyperplasia of gastric epithelial cells. This chronic hyperplasia could in turn lead to development of cancer phenotype. The loss of Nf1 in different types of cells can result in secretion of factors such as TGF-α, KitL, IGF, and HGF (9, 55, 56). We are also interested in identifying and analyzing the possible secreted factors (trophic factors and cytokines) involved in the development of gastric hyperplasia in Nf1-Isl1-CKO mice.

Molecular changes in Nf1-Isl1-CKO gastric epithelium.

Cytokines and their related signaling pathways are important in gastric tumor development. Studies of point mutations in the IL-6 receptor-gp130 revealed two reciprocally regulated downstream pathways associated with different pathogenesis. Through inactivation of the SH2 domain-containing protein tyrosine phosphatase SHP2 binding site on gp130, the STAT3 pathway can be activated while the ERK pathway is inhibited. Mice with this mutation (gp130^757F/757F) develop gastric tumors, whereas mice with activated SHP2/Ras/ERK pathway do not (47). STAT3 is also necessary for the tumor genesis because further crossing of heterozygous STAT3 mutant mice (STAT3+/−) to gp130^757F/757F mice partially reverses the gastric phenotype (19). However, studies looking at Helicobacter pylori infection and gastric tumor suggested that the Ras/ERK pathway is also important for tumor development. Upon binding to the epithelial cells, Helicobacter bacteria inject cytotoxin-associated gene A (CagA) proteins into the cells. CagA then interacts with SHP2, leading to ERK activation, which may be the reason for the pathogenesis of Helicobacter infection (14).

In Nf1-Isl1-CKO gastric epithelium, we found hyperactivation of both the ERK (Fig. 7B) and STAT3 (preliminary data) pathways throughout the pyloric glands, even though few epithelial cells lack Nf1 (Figs. 1 and 6). Therefore, activation of these pathways may not be the direct effect of Nf1 loss in these epithelial cells but rather attributable to stimulative effects of secreted factors present in the gastric epithelium. This is consistent with the hypothesis that it is mainly Nf1-deficient nonepithelial cells that contribute to the gastric hyperplasia phenotypes we observed. It would be interesting to test the direct effects of Nf1 deletion in gastric epithelial cells and examine whether Nf1 has cell-autonomous effects on these cells as well.

GRANTS

This work was supported by grants from the ACS, the NINDS (P50NS052606-01), and the DOD (W81XWH-05-1-0265). L. Parada is an American Cancer Society Research Professor.