EHS
EHS

BMP feed-forward loop promotes terminal differentiation in gastric glands and is interrupted by H. pylori-driven inflammation


Myh11+ myofibroblasts respond to H. pylori infection

To investigate what impact the stroma along the entire gastric gland has on gland homeostasis, we generated Myh11CreErt2/Rosa26-mTmG mice to induce GFP expression in the Myh11-myofibroblast lineage. Confocal microscopy identified GFP+ cells beneath and between the glands in direct proximity to epithelial cells, and staining with α-smooth muscle actin (SMA) antibody showed co-localization with the GFP signal (Fig. 1a), indicating that it is specific to mesenchymal cells. Signal quantification confirmed a strong overlap between Myh11 and α-SMA signal (Fig. 1b), and gene expression analysis from sorted gastric GFP+ cells revealed that Myh11+ cells indeed express high levels of mesenchymal markers, whereas markers of epithelial and immune cells were absent (Fig. 1c).

Fig. 1: H. pylori infection reduces BMP signaling in Myh11+ myofibroblasts.
figure 1

a Confocal microscopy images of stomach antrum from Myh11CreErt2/Rosa26-mTmG mice. b Quantification of the signal overlap between α-SMA+ and Myh11-eGFP signal from Myh11CreErt2/Rosa26-mTmG mice (n = 3). c Heat map for selected genes expressed in sorted Myh11-eGFP cells from Myh11CreErt2/Rosa26-mTmG mice (n = 3). The gene expression level is presented as the read count. d Confocal microscopy images of gland hyperplasia (left) and Myh11+ myofibroblasts (right) in uninfected and 2-month H. pylori-infected Myh11CreErt2/Rosa26-mTmG mice. e Quantification of actin and Myh11 area per picture in uninfected (n = 3) and 2-month H. pylori-infected (n = 3) Myh11CreErt2/Rosa26-mTmG mice. f Heat map of RNAseq data with genes differentially expressed in uninfected (n = 3) and 2-month H. pylori-infected (n = 3) Myh11+ myofibroblasts. g Highlight of RNAseq data for differentially expressed genes involved in BMP signaling in uninfected (n = 3) and 2-month H. pylori-infected (n = 3) Myh11+ myofibroblasts. h qPCR validation of RNAseq data for BMP genes (uninfected: n = 3; infected: n = 3). Images are representative of at least three biological replicates. Scale bar: 100 µm. Data are presented as mean ± SEM. Statistical analyses were performed using Student’s t-test (two-tailed) for (e) and (h). RNAseq data are from three biological replicates per group. Source data are provided as a Source Data file.

Next, Myh11CreErt2/Rosa26-mTmG mice were infected with H. pylori for 2 months. Infection induced gland hyperplasia, which was accompanied by an expansion of surrounding Myh11+ cells (Fig. 1d, e). Next, we asked how the infection affects gene expression in myofibroblasts. Cells isolated from the antrum and transitional zone of infected animals were sorted for Myh11+/ Epcam− cells (Supplementary Fig. 1a), which were further processed for RNA-seq. Myofibroblasts from uninfected mice expressed a variety of genes involved in the regulation of epithelial stem cells, including Wnt, Rspo, and Bmp, as well as activators and inhibitors of these pathways. Upon infection with H. pylori, 104 genes were significantly upregulated and 93 downregulated (Fig. 1f), indicating that myofibroblasts actively respond to infection. We noticed that several of these genes were involved in BMP signaling. Further analysis revealed that the upregulated genes included several physiological BMP inhibitors, including Grem1, Bambi, Crim1, and Chrdl1. In contrast, expression of BMP ligands, including Bmp1–7, as well as the BMP pathway target genes Id1 and Id2 was decreased (Fig. 1g). We were able to confirm a significant loss of Bmp2, Bmp6, and Bmp7 expression, as well as a significant increase of Chrdl1 via qPCR (Fig. 1h). Accordingly, H. pylori infection induces a shift towards inhibition of BMP signaling in Myh11+ myofibroblasts.

BMP signaling is spatially organized along the gland axis

To visualize the spatial distribution of different BMP signaling molecules in the uninfected gastric epithelium, we performed single-molecule in situ hybridization (sm-ISH) for a number of BMP ligands. We noticed that Bmp2, the most highly expressed gene in both epithelial and stromal cells, was expressed predominantly at the gland surface and gradually decreased towards the base, where it was not present at all (Fig. 2a). Other BMP ligands were expressed at lower levels, e.g., Bmp4, which was distributed equally along the entire gland axis but was mostly restricted to the stroma and only occasionally co-localized with epithelial cells (Fig. 2b). As the expression of Bmp6 and Bmp7 was very low (Fig. 2c), we focused on the most abundant ligands, Bmp2 and Bmp4. To quantify the distribution along the gland axis, we divided the tissue into top, middle, and base parts and quantified the signal for Bmp2 and Bmp4 in each. In the top of the gland, the average Bmp2 expression was ten times higher than in the base (Fig. 2d). Quantification of the Bmp4 signal confirmed no significant difference in expression between the top, middle, and base of the gland (Fig. 2e).

Fig. 2: BMP signaling shows highly organized distribution in gastric glands, with strong expression in the pit cells and absence or inhibition in the gland base.
figure 2

ac Single-molecule in situ hybridization (sm-ISH) for Bmp2, Bmp4, Bmp6, and Bmp7 in antrum tissue (n = 3). d, e High magnification and quantification for Bmp2 and Bmp4 signal area per picture divided into top, middle, and base of the gland (n = 3). f Sm-ISH for Grem1, Grem2, and Chrdl1 in antrum tissue with high magnification of the gland base (n = 3). g Sm-ISH for Noggin in antrum tissue. h Quantification for Grem1, Grem2, and Chrdl1 signal area per picture divided for top, middle, and base (n = 3). Images are representative of at least three biological replicates. Scale bar: 100 µm. Data are presented as mean ± SEM. Statistical analyses were performed using one-way ANOVA, followed by Tukey’s multiple comparisons test for (d, e, h). Source data are provided as a Source Data file.

We then analyzed the spatial distribution of BMP inhibitors. ISH showed high expression of Grem1, Grem2, and Chrdl1 in the stroma beneath the base of the glands (Fig. 2f). Noggin, which acts as a BMP inhibitor in the intestine9, was not detected in the stomach (Fig. 2g). Quantification of ISH showed that Grem1, Grem2, and especially Chrdl1 expression was limited to the stromal cells beneath the gland base in the lamina muscularis mucosae (Fig. 2h). Together, these data depict a differential distribution of BMP signaling molecules along the gastric gland axis—with strong expression of Bmp2 in the pit region and absence of Bmp2 along with expression of BMP inhibitors in the base. To further dissect the stromal compartment in the stomach, we performed single-cell RNA sequencing (scRNA-seq) of non-epithelial cells from the gastric antrum. Unbiased clustering revealed the presence of 6 populations, which are representative for smooth muscle cells, immune cells, remaining epithelial cells, endothelial cells, and two populations of mesenchymal stromal cells that expressed the contractile gene Acta2 (Supplementary Fig. 2a, d). To validate these results, we analyzed data from a published scRNA-seq experiment on gastric stromal cells10, which confirmed the presence of two mesenchymal cell populations (stromal 1 and 2) expressing the pan-fibroblast marker vimentin, Acta2, as well as extracellular matrix-related genes, such as Col4a5 and Col6a2 (Supplementary Fig. 2e, f).

Mapping of Bmp-associated genes revealed differential expression in the two stromal clusters: While Chrdl1 and Grem1 were almost exclusively expressed in population 2, Bmp2 was highly expressed in population 1 (Supplementary Fig. 2b, c). Interestingly, expression of Id1 was almost absent in the cell cluster that expressed BMP inhibitors, while the Bmp2-expressing cluster contained numerous cells expressing high levels of Id1, indicating that BMP signaling in the stroma is inhibited in cells surrounding the gland base. Population 1 also showed high expression of Foxl1, which was absent in population 2 (Supplementary Fig. 2d). Differential expression of BMP ligands and inhibitors was also consistently found in the validation data set (Supplementary Fig. 2f) and recent scRNA-seq data from the intestine11.

H. pylori regulates BMP ligand and inhibitor expression

To further map gene expression changes upon H. pylori infection in situ, we examined the distribution of BMP ligand and inhibitor transcripts in the infected mouse antrum. ISH from gastric tissue of 2-month infected mice revealed a strong decrease of Bmp2 expression at the gland surface, in both stroma and epithelial cells, whereas Bmp4 expression remained largely unchanged (Fig. 3a). Quantification of the signal showed an almost sevenfold lower expression of Bmp2 in the infected samples compared to those from uninfected mice (Fig. 3b). A significant decrease of Bmp2 expression was also confirmed in an additional set of mice infected with H. pylori for only 2 weeks (Supplementary Fig. 3a).

Fig. 3: Infection with H. pylori reduces BMP ligand at the surface and enhances BMP inhibitors in the base of the gland.
figure 3

a Sm-ISH for Bmp2 and Bmp4 in antrum tissue of uninfected (n = 3) and 2-month H. pylori-infected (n = 3) C57BL/6 mice. b Quantification for Bmp2 and Bmp4 signal area per picture in antrum tissue of uninfected (n = 3) and 2-month H. pylori-infected (n = 3) C57BL/6 mice. c Sm-ISH for Grem1, Grem2, and Chrdl1 in antrum tissue of uninfected (n = 3) and 2-month H. pylori-infected (n = 3) C57BL/6 mice. d Quantification for Grem1, Grem2, and Chrdl1 signal area per picture in antrum tissue of uninfected (n = 3) and 2-month H. pylori-infected (n = 3) C57BL/6 mice. e, f Sm-ISH and quantification for Grem1 and Grem2 signal area per picture in the transitional zone between antrum and corpus of uninfected (n = 3) and 2-month H. pylori-infected (n = 3) C57BL/6 mice. g, h Sm-ISH for Id1 and quantification of the Id1-free compartment in the gland base of uninfected (n = 3) and 2-month H. pylori-infected (n = 3) C57BL/6 mice. Images are representative of at least three biological replicates. Scale bar: 100 µm. Data are presented as mean ± SEM. Statistical analyses were performed using Student’s t-test (two-tailed) for (b, df, h). Source data are provided as a Source Data file.

We next visualized the BMP inhibitors in the antrum and confirmed an increased expression of Chrdl1 as well as Grem2 in infected samples (Fig. 3c, d). Grem1 did not show a significant difference upon infection when analyzing the whole antrum, but we noticed a striking increase of Grem1 and Grem2 in the proximal antrum and the transitional zone between the antrum and corpus (Fig. 3e, f). This area is particularly highly colonized by gland-associated H. pylori12 and exhibits particularly prominent gland hyperplasia and antral metaplasia. Notably, Grem1 and Grem2 expression was not only increased but was no longer restricted to the lamina muscularis mucosae, instead of extending upwards to the area surrounding the basal antral cells. Next, we performed ISH for Id1 in uninfected and infected mice. We found expression of Id1 at the surface but not at the base in uninfected mice, which is consistent with the presence of BMP inhibitors specifically at the base (Fig. 3g). Infected mice showed a significant expansion specifically of the Id1-free compartment at the gland base (Fig. 3g, h). Together, H. pylori infection downregulates Bmp2 expression at the top of the gland and upregulates BMP inhibitors at the gland base, resulting in an expansion of Id1-negative epithelial cells in the bottom of gastric glands.

BMP promotes stem cell differentiation into surface cells

The antral epithelium is characterized by a strict cellular organization: proliferative Axin2 cells in the lower isthmus either self-renew and give rise to Lgr5+ secretory cells in the base, which express markers such as Muc6 and Gif, or they lose their self-renewal capacity and differentiate into Muc5ac+ mucous pit cells in the surface region (Fig. 4a). To investigate how BMP signaling affects gastric epithelial proliferation and differentiation, we established 3D organoids from mouse antrum. The addition of recombinant BMP proteins to the medium rapidly inhibited organoid growth (Figs. 4b and S3b). However, the addition of BMP inhibitors such as CHRDL1, GREM1, and GREM2 to the culture with BMP ligands was sufficient to block the effect of BMPs, resulting in organoids that were phenotypically similar to those cultured in normal medium, as confirmed by quantification of their size (Supplementary Fig. 3b).

Fig. 4: BMP promotes mucous pit cell differentiation and inhibits basal marker expression.
figure 4

a Schematic representation of gastric antrum markers for characteristic gland areas. b Images of organoids and quantification of the diameter of organoids from antral epithelium either untreated or treated with BMP2 (n = 3). c qPCR for the BMP target gene Id1 and markers characteristic for specific gland regions from organoids untreated or treated with BMP2 (n = 3). d qPCR for antimicrobial genes from organoids untreated or treated with BMP2 (n = 3). e Representative pictures and diameter quantification of organoids from antral epithelium cultured in a full medium, medium without Rspo, medium without Rspo, and reduced WNT or with the addition of BMP2 protein (n = 3). f qPCR for BMP target gene Id1 from organoids cultured in indicated conditions (n = 3). g qPCR for markers characteristic for specific gland regions from organoids cultured in indicated conditions (n = 3). h Immunofluorescence images of organoids grown in indicated conditions and quantification of relative abundance of Ki67+ cells, Muc5ac+ pit cells, and GSII+ gland base secretory cells (n = 3). Scale bar: 100 µm. i, j Representative organoid images and quantification of antral organoids per well treated with indicated conditions, passaged and cultured in normal medium (n = 3). Images are representative of at least three biological replicates. Scale bar: 250 µm. Data are presented as mean ± SEM. Statistical analyses were performed using Student’s t-test (two-tailed) for (b, c, d, j); using one-way ANOVA, followed by Tukey’s multiple comparisons test for (e, f, g, h). Source data are provided as a Source Data file.

We then performed qPCR to examine the expression of cell type-specific markers in organoids. We first confirmed that the BMP pathway was efficiently induced by validating that expression of the BMP target gene Id1 was strongly increased upon treatment with BMP proteins (Fig. 4c). We selected marker genes characteristic of either antrum gland base or gland surface and noticed a strong downregulation of gland base markers Lgr5 as well as Muc6 upon BMP treatment (Figs. 4c and S3c). In contrast, the surface pit cell marker Muc5ac was significantly upregulated (Figs. 4c and S3c). Expression of Axin2 itself, as well as enteroendocrine cell markers such as gastrin (Gast) and chromogranin A (Chga), remained unchanged (Figs. 4c and S3c). BMP2 and BMP4 had a similar effect on gene expression, although Muc5ac expression was much higher upon BMP2 compared to BMP4 treatment (Figs. 4c and S3c). To test whether BMP antagonists are able to block the effects of BMP on differentiation, we repeated the experiment, adding CHRDL1, GREM1, or GREM2 to the medium in addition to BMP ligands, and found that all of them blocked the BMP-driven differentiation towards Muc5ac-expressing pit cells (Supplementary Fig. 3d).

We recently showed that secretory Lgr5+ cells produce antimicrobial factors capable of counteracting H. pylori infection8. Since expression of Bmp ligands, which inhibit Lgr5 expression, is blocked upon infection, we wondered if the expression of antimicrobial genes is also inhibited by BMP signaling. Therefore, we examined transcription of the gland base antimicrobial factors Itln1, Reg3b, and Reg3g in organoids treated with exogenous BMP2 and BMP4. qPCR showed a strong decrease for all three genes upon BMP treatment (Figs. 4d and S3e). Next, we examined the influence of the three BMP inhibitors and found that all of them restored expression of Itln1, Reg3b, and Reg3g, compared to samples treated with BMP4 alone (Supplementary Fig. 3f). Together, these data indicate that activation of BMP signaling induces increased expression of surface mucous cell markers and that inhibition of BMP signaling promotes expression of antimicrobial genes found in the gland base upon infection.

BMP signaling enforces rapid terminal differentiation

Our previous work showed that Rspo3-driven stabilization of WNT signaling promotes stem cell differentiation into gland base secretory cells4. Expression of Rspo3 is strictly restricted to the base, which correlates with the expression pattern of WNT target genes in the gland. Using organoid cultures, it has been shown that surface enterocyte markers are enriched upon removal of WNT or Rspo from the culture medium, indicating that cells that exit the gland base compartment automatically differentiate into surface cells13. Therefore, we compared the effects of Rspo/WNT-depletion or BMP activation on epithelial stemness, proliferation, and differentiation.

Treatment with BMP2 led to a more marked decrease in organoid size compared to a reduction of WNT ligand concentration or removal of Rspo from the medium (Fig. 4e). qPCR confirmed that the BMP target gene Id1 was significantly upregulated after BMP2 treatment (Fig. 4f). As expected, the lack of WNT signaling, as well as the addition of BMP2, reduced the expression of gland base markers, but the effect of BMP2 was much stronger, leading to an almost complete loss of these markers (Fig. 4g). In contrast, expression of the surface cell marker Muc5ac was increased – here again, BMP2 had a much stronger effect (Fig. 4g). We stained organoids for Muc5ac, as well as Ki67 and GSII to visualize gland base secretory cells. While removal of WNT/Rspo induced a slight increase of Muc5ac+ cells with some Ki67+ cells still present, BMP2-treated organoids almost completely lost both the GSII+ and the Ki67+ cells, with almost 100% of cells now positive for Muc5ac (Fig. 4h). Thus, we conclude that BMP2 induces rapid terminal differentiation into surface mucous cells.

To further dissect whether this is driven by lineage commitment or death of the other cell types, we performed lineage tracing in organoids from Lgr5CreErt2/Rosa26-tdTomato mice (Supplementary Fig. 4a) and Axin2CreErt2/Rosa26-tdTomato mice (Supplementary Fig. 4b). We tracked organoids on a daily basis and confirmed that BMP2 inhibits organoid growth (Supplementary Fig. 4c). Lgr5+ or Axin2+ cell lineage tracing was induced at the same time as treatment with BMP2. We found that although organoids grew much more slowly in the presence of BMP2, there was an expansion of initially labeled clones. Quantifying the number of organoids with fluorescent signal showed that BMP2 treatment did not significantly affect the proportion that contained tdTomato+ cells, indicating that most labeled cells did not undergo cell death (Supplementary Fig. 4a, b). We conclude that induction of BMP signaling promotes rapid differentiation of stem cells into Muc5ac+ mucous cells.

Next, we passaged the organoids grown in a medium that was Rspo/WNT depleted, BMP2 supplemented, or both and transferred them back to a normal stem cell medium to assess their colony-forming efficiency. Cells kept in medium without Rspo and with reduced WNT were able to generate organoids when returned to normal medium, but cultures treated with BMP2 were not, regardless of whether they grew in normal or Rspo/WNT depleted medium before passaging (Fig. 4i, j), demonstrating that BMP2 treatment completely blocks organoid forming capacity.

Since we observed that Bmp2 has a profound effect on epithelial differentiation and is also expressed in surface epithelial cells themselves, we investigated how its expression is controlled in the epithelium. Organoids expressed low levels of Bmp2 when grown in the control medium as well as in a medium with reduced Rspo/WNT. In contrast, the addition of BMP2 itself, as well as its homolog BMP4, strongly upregulated Bmp2 expression (Fig. 5a), indicative of a positive auto- and paracrine BMP feedback loop that, once activated, can be sustained and enhanced autonomously by the Bmp2-expressing cell compartment. By exposing organoids to different concentrations of recombinant BMP2 (rBMP2), we observed that low concentrations of rBMP2 induce a slight increase of Muc5ac and a slight decrease of Lgr5 expression, resulting in reduced organoid forming capacity (Fig. 5b, c). However, once endogenous Bmp2 is induced by rBMP2, a hyper-additive effect on all measured parameters is observed, resulting in a rapid and complete loss of Lgr5 and Muc6 expression as well as organoid forming capacity (Fig. 5b, c). As the normal organoid culture medium contains the BMP inhibitor noggin, we repeated this experiment in organoids grown in a noggin-free medium. Treatment with rBMP2 still had the same effects, but the concentrations required were lower (Fig. 5d). To investigate the impact of endogenous BMP2, organoids in the noggin-free medium were treated with 5 ng/ml rBMP2 for four days. Following treatment, they were washed and grown either in full medium with noggin or medium without noggin (Fig. 5e). We noticed that 5 ng/ml rBMP2 induced inhibition of organoid forming efficiency. While organoid forming efficiency could be fully recovered in cells grown with noggin at passage 2 (Fig. 5f, dark blue line), cells grown without noggin showed a progressive loss of organoid forming capacity (Fig. 5f, light blue line). This was accompanied by suppression of endogenous Bmp2 expression in organoids grown with noggin at passage 2, while organoids previously exposed to BMP2 and grown without noggin showed an increase in Bmp2 expression (Fig. 5g). This demonstrates that once its expression is induced, endogenous BMP2 feed-forward signaling is sufficient to enforce differentiation and loss of proliferative capacity.

Fig. 5: BMP signaling has a strong impact on gland turnover through the feed-forward loop and regulates R-spondin3 expression in myofibroblasts.
figure 5

a qPCR for Bmp2 from organoids cultured in indicated conditions (n = 3). b qPCR for indicated genes and forming capacity from organoids exposed to different concentrations of rBMP2 (n = 3). c Representative images of organoids that re-grew in the full medium after exposure to the indicated rBMP2 concentrations. d qPCR for indicated genes and forming capacity from organoids grown without noggin, exposed to different concentrations of rBMP2 (n = 3). e Schematic representation: organoids were grown without noggin and treated with 5 ng/ml rBMP2 (blue) for 4 days at passage 0. Organoids were washed and passaged and either kept in full medium with noggin (dark blue) or without noggin (blue). f Organoid forming capacity at different passages from organoids treated (blue) or untreated (yellow) with rBMP. Organoids exposed to full medium with noggin (dark blue) recovered their forming capacity, while organoids without noggin progressively lost forming capacity (n = 2). g qPCR from the organoids from e at passage 2: Bmp2 expression triggered by rBMP is not observed in organoids that were re-grown in full medium, while in organoids grown without noggin, endogenous Bmp2 expression is strongly upregulated (n = 4). h Immunofluorescence image of myofibroblasts from Myh11CreErt2/Rosa26-tdTomato mice cultured in 2D. Scale bar: 100 µm. i qPCR for Id1 and Rspo3 from myofibroblasts cultured with noggin and BMP2 (n = 3). j qPCR for Id1 and Rspo3 from myofibroblasts cultured with different BMP2 concentrations (n = 2). k Immunofluorescence images of EdU proliferation assay performed with myofibroblasts untreated and treated with BMP2. Scale bar: 100 µm. l Quantification of the proportion of EdU positive cells (n = 3). Images are representative of at least three biological replicates. Scale bar: 250 µm, except where indicated. Data are presented as mean ± SEM. Statistical analyses were performed using Student’s t-test (two-tailed) for (g, i, l); using one-way ANOVA, followed by Tukey’s multiple comparisons test for (a). Source data are provided as a Source Data file.

We recently reported that WNT signaling in the stomach is controlled by Rspo3 expressed exclusively in myofibroblasts beneath the gland base but not in the stroma surrounding the surface of the gland4. Since Bmp2 is only expressed at the gland surface, we asked whether BMP2 affects Rspo3 expression. We therefore isolated and cultured primary Myh11+ myofibroblasts from Myh11CreErt2/Rosa26-tdTomato mice and confirmed that they were mesenchymal cells by staining for vimentin expression (Fig. 5h). Treatment with BMP2 led to a decrease of Rspo3 expression, while inhibition of BMP signaling by noggin increased Rspo3 expression in myofibroblasts (Fig. 5i, j). To exclude that this is due to a selective effect on the proliferation of Rspo3+ myofibroblasts, we analyzed proliferation via the EdU proliferation assay. No significant changes in proliferation upon BMP2 treatment were observed, indicating that BMP2 signaling has a direct inhibitory effect on Rspo3 expression in myofibroblasts (Fig. 5k, l). Thus, our data demonstrate the existence of feedback loops controlling both strength and spatial distribution of WNT and BMP signals in gastric glands.

The deficiency of BMP signaling promotes gland hyperplasia

Smad4 is one of the central mediators through which transforming growth factor-β/BMP signaling regulates gene expression. To study if inhibition of BMP signaling in Axin2+ cells alters epithelial cell composition in the stomach, we generated Axin2CreErt2/Smad4fl/fl mice. In the antrum of these mice, we did indeed find Smad4-depleted glands (Supplementary Fig. 4d, dashed red lines) that had almost completely lost expression of Id1 (Supplementary Fig. 4e, dashed black lines). The KO efficiency of this model was relatively low. However, these KO glands specifically displayed an expansion of the gland base module and hyperproliferation (Supplementary Fig. 4f).

Due to the low efficiency of Smad4 KO, we established another model to study the effect of BMP signaling on stem cell fate. We generated Axin2-specific BMP type I receptor Bmpr1a KO mice by breeding Bmpr1afl/fl mice with Axin2CreErt2 mice and found that the KO in these mice was efficient. Expression of Id1 was reduced by 90.7% (Fig. 6a), which was associated with a significant decrease of Bmp2 expression in the glands (Fig. 6b), confirming that BMP signaling induces a positive feed-forward loop via Bmp2. Histological analysis revealed that KO mice had severe hyperplasia (Fig. 6c), which was characterized by an expansion of the gland base module (gland base secretory cells and proliferative cells), mimicking our observations in mice infected with H. pylori. We also infected KO mice with H. pylori. Infection caused an inflammatory response but did not augment antrum hyperplasia (Supplementary Fig. 4g), suggesting that BMP inhibition upon infection is a dominant event in the context of H. pylori-induced hyperplasia. Overall, these functional genetic experiments confirm that depletion of Bmpr1a in Axin2+ cells causes increased proliferation and accumulation of gland base secretory cells.

Fig. 6: Deficiency of BMP signaling blocks the BMP feedback loop and promotes gland hyperplasia.
figure 6

a Sm-ISH for Id1 and quantification of Id1 signal area per image in antrum tissue from WT (n = 3) and Axin2CreErt2/Bmpr1afl/fl (n = 3) mice. b Sm-ISH for Bmp2 and quantification of Bmp2 signal area per image in antrum tissue from WT (n = 3) and Axin2CreErt2/Bmpr1afl/fl (n = 3) mice. c Immunofluorescence images of antrum tissue from WT (n = 3) and Axin2CreErt2/Bmpr1afl/fl (n = 3) mice. Gland height of antrum tissue and relative abundance of the Ki67+ and GSII+ compartments were quantified. Images are representative of at least three biological replicates. Scale bar: 100 µm, data are presented as mean ± SEM. Statistical analyses were performed using Student’s t-test (two-tailed) for (ac). Source data are provided as a Source Data file.

H. pylori induces BMP2 loss and gland hyperplasia via T4SS

Since we had observed that BMP signaling is inhibited upon H. pylori infection, we asked whether infection triggers similar changes in the glands as those observed upon BMP inhibition in organoids. Indeed, staining for cell type-specific markers revealed an expansion of GSII+ mucous gland base cells upon infection, as well as an expansion of the proliferative compartment, overall leading to gland hyperplasia, whereas Muc5ac+ cells were relatively underrepresented (Fig. 7a). These observations are consistent with our previous findings showing that infection leads to an expansion of stem cells as well as gland base secretory cells4,8.

Fig. 7: H. pylori induce an IFN-γ response.
figure 7

a Immunofluorescence images of antrum tissue from uninfected (n = 3) and 2-month PMSS1 WT H. pylori-infected (n = 3) mice. Relative abundance of the Muc5ac+, Ki67+, and GSII+ compartments were quantified. b Representative images of H&E staining and gland height quantification of antrum tissue from C57BL/6 mice infected for two months with PMSS1 WT (n = 3) or ΔCagE H. pylori (n = 3). c Immunofluorescence images labeled for Ki67 and quantification of Ki67+ cells per gland in antrum tissue from C57BL/6 mice infected for two months with PMSS1 WT (n = 3) or ΔCagE H. pylori (n = 3). d Sm-ISH for Bmp2 and quantification of Bmp2 signal area per image in antrum tissue from C57BL/6 mice infected for 2 months with PMSS1 WT (n = 3) or ΔCagE H. pylori (n = 3). e GSEA with the hallmark of IFN-γ response genes for PMSS1 WT H. pylori-infected (n = 2) compared to uninfected C57BL/6 mice (n = 2). f qPCR for Ifnγ in the gastric antrum from C57BL/6 mice infected for 2 months with PMSS1 WT (n = 3) or ΔCagE H. pylori (n = 3). g Immunofluorescence images of antrum tissue from 6-week H. pylori-infected C57BL/6 mice (n = 4) and IfnγR KO mice (n = 4). Gland height of antrum tissue and relative abundance of the Muc5ac+, Ki67+, and GSII+ compartments were quantified. h CFUs from 6-week H. pylori-infected C57BL/6 mice (n = 4) and IfnγR KO mice (n = 4). i, j Sm-ISH images and quantification for Bmp2 and Id1 in antrum tissue of 6-week H. pylori-infected C57BL/6 mice (n = 4) and IfnγR KO mice (n = 4). Images are representative of at least three biological replicates. Scale bar: 100 µm, data are presented as mean ± SEM. Statistical analyses were performed using Student’s t-test (two-tailed) for (ad, fh, j). Source data are provided as a Source Data file.

The H. pylori T4SS is required to translocate CagA, its most prominent virulence factor, which we have previously shown to be required for the expansion of the Lgr5+ cell compartment upon infection5. We wondered whether the inhibition of Bmp2 expression upon infection also depends on a functional T4SS. To address this, we infected mice with an isogenic mutant lacking the CagE gene, which has a dysfunctional T4SS and is not able to translocate CagA into host cells14. Mice infected with ΔCagE PMSS1 showed less gastric hyperplasia (Fig. 7b) and fewer Ki67+ cells compared to mice infected with wild-type PMSS1 (Fig. 7c). By performing ISH for Bmp2, we observed that the signal was still present at high levels in mice infected with ΔCagE PMSS1, whereas wild-type PMSS1 infected mice showed an almost complete loss of Bmp2 expression (Fig. 7d). Thus, the downregulation of Bmp2 in response to infection depends on a functional T4SS.

We, therefore, investigated the mechanism underlying the downregulation of Bmp2 expression. Microarray data comparing gene expression in the antrum of uninfected versus 2-month-infected mice followed by Gene Set Enrichment Analysis (GSEA) showed the highest level of enrichment for IFN-γ response genes (Fig. 7e). Using qPCR to compare the expression of Ifnγ in the gastric antrum from mice infected with wild-type vs. ΔCagE PMSS1, we found that expression is significantly higher in wild-type-infected animals, indicating that the strong induction of Ifn-γ expression upon infection requires the T4SS (Fig. 7f). Thus, we conclude that translocation of CagA by H. pylori leads to induction of IFN-γ signaling as well as repression of Bmp2 expression.

To investigate the role of IFN-γ for BMP-driven gastric pathology, we infected IFN-γ receptor KO (IFNγR KO) mice with H. pylori. We observed that KO mice showed less hyperplasia (Fig. 7g). Moreover, the absence of the IFNγR rescued infection-driven BMP alterations: we found no loss of Bmp2 or Id1 expression (Fig. 7i, j), but reduced Grem2 expression in infected IFNγR KO mice compared to WT littermates (Supplementary Fig. 5a). While infected WT mice showed an expansion of gland base cells, this was not seen in IFNγR KO mice (Fig. 7g). Moreover, colony-forming unit (CFU) analysis showed higher colonization in KO mice (Fig. 7h). Overall, these data demonstrate that IFN-γ is required for the changes in BMP signaling observed after H. pylori infection.

H. pylori is known to trigger an immune response with infiltration of IFN-γ producing T cells15. Since we noticed that H. pylori already induced Bmp2 loss after two weeks of infection (Supplementary Fig. 3a), we investigated to what extent T cells are found in the mucosa. Indeed, at two weeks, there was a strong infiltration of T cells (Supplementary Fig. 5b), indicating that the inflammatory responses and subsequent changes in epithelial cell fate determination appear at an early time point of H. pylori infection.

IFN-γ inhibits BMP signaling in epithelial and stromal cells

Next, we asked whether IFN-γ can directly influence the expression of BMP ligands or inhibitors in epithelial and stromal cells. We treated gastric organoids with recombinant IFN-γ and confirmed a strong upregulation of the IFN-γ target gene Irf1 by qPCR (Figs. 8a and S5e). Interestingly, Bmp2, the BMP2 target gene Id1, as well as other target genes of BMP signaling, were downregulated upon IFN-γ treatment, while expression of Lgr5 was increased by 85.4% (Figs. 8a, S5c, and S5e). These effects required the IFN-γ receptor, as no expression changes were observed in organoids from IFN-γ receptor KO mice (Supplementary Fig. 5d). To investigate this further, we performed microarray analysis from organoids treated with IFN-γ followed by GSEA. IFN-γ led to a significant enrichment of genes that were previously identified as being downregulated by BMP2 (Fig. 8b), whereas genes upregulated by BMP2 showed significant negative enrichment (Fig. 8c). β-catenin target genes were not significantly regulated, indicating that the IFN-γ-driven upregulation of Lgr5 expression was driven by inhibition of BMP2 signaling and not by interfering with the WNT pathway (Supplementary Fig. 5f). As we had observed inhibition of Bmp2 expression in stromal cells upon infection, we now asked whether this, too, was induced by IFN-γ. Indeed, we found a significant decrease of Bmp2 expression in isolated stromal cells that were treated with IFN-γ (Fig. 8d), indicating that both epithelial and stromal BMP signaling is inhibited by IFN-γ. Since Bmp2 signaling inhibits stromal Rspo3, we asked how its expression is affected by IFN-γ and found that Rspo3 was indeed upregulated in stromal cells exposed to IFN-γ (Fig. 8d). Moreover, expression of Grem1 and Grem2 was increased, consistent with the data from stromal cells of mice infected with H. pylori (see Fig. 3c–f).

Fig. 8: IFN-γ blocks BMP signaling.
figure 8

a qPCR for indicated genes in untreated or IFN-γ treated organoids (n = 3). b, c GSEA with the hallmark of targets downregulated or upregulated by BMP2 (Lee) for IFN-γ treated organoids compared to untreated ones (n = 2). d qPCR for indicated genes from primary gastric antrum stromal cells untreated or treated with IFN-γ (n = 4). e Immunofluorescence images of organoids either untreated or treated with IFN-γ; relative abundance of different cell types was quantified (n = 3). f Quantification of forming capacity of organoid either untreated or pretreated with IFN-γ (n = 3). g Images and quantification of tdTomato positive organoids from Lgr5CreErt2/Rosa26-tdTomato mice untreated or treated with IFN-γ, passaged and cultured in normal medium (tamoxifen was given 24 h before passaging; n = 3). Scale bar: 250 µm. h Images and quantification of tdTomato positive organoids from Axin2CreErt2/Rosa26-tdTomato mice untreated or treated with IFN-γ, passaged and cultured in normal medium (tamoxifen was given 24 h before passaging; n = 3). Scale bar: 250 µm. i Representative images and quantification of organoids pretreated with BMP2 or with BMP2 and IFN-γ together before passaging and then cultured in normal medium (n = 3). Scale bar: 250 µm. j qPCR for indicated genes with untreated, BMP2, or BMP2 and IFN-γ treated organoids (n = 4). k Immunofluorescence images from antrum of untreated (n = 3) or IFN-γ treated mice (n = 3). l, m Quantification of gland height and relative abundance of different cell compartments in mice upon IFN-γ treatment (ctrl: n = 3; +IFN-γ: n = 3). n Sm-ISH images and quantification for Bmp2 and Id1 in mice either untreated (n = 3) or treated with IFN-γ (n = 3). Images are representative of at least three biological replicates. Scale bar: 100 µm, except where indicated. Data are presented as mean ± SEM. Statistical analyses were performed using Student’s t-test (two-tailed) for (a, dh, I, l, m, o); using one-way ANOVA, followed by Tukey’s multiple comparisons test for (j). Source data are provided as a Source Data file.

To further investigate the direct effects of IFN-γ on gastric epithelial cells, we stained organoids treated with IFN-γ and observed enrichment of GSII+ cells, while Muc5ac+ cells were lost (Fig. 8e). Although in our organoids, IFN-γ induced higher levels of Lgr5 expression and blocked BMP2 signaling, organoid growth was not increased upon treatment. Instead, IFN-γ treated organoids appeared slightly smaller, and fewer organoids formed after passaging (Fig. 8f).

As Lgr5 is expressed in gland base secretory cells as well as in a subset of proliferative stem cells, we asked how IFN-γ affects these subpopulations. Lineage tracing was induced with tamoxifen in IFN-γ-treated and non-treated organoids from Lgr5CreErt2/Rosa26-tdTomato mice. After 24 h, organoids were dissociated into single cells and passaged to determine the proportion of tdTomato positive organoids formed, i.e., cells that derive from Lgr5+ cells or their early progeny (Fig. 8g). Upon treatment with IFN-γ, the proportion of organoids deriving from tdTomato+ cells was decreased, indicating a selective advantage for organoid-forming capacity of cells that were Lgr5-negative cells compared to Lgr5-positive cells or their immediate progeny (Fig. 8g). Thus, we conclude that IFN-γ prevents the proliferation of Lgr5+ cells, which are likely to be programmed into gland base secretory cells. In contrast, when we performed the same experiment using Axin2CreErt2/Rosa26-tdTomato mice, in which both Lg5+ gland base cells and proliferative isthmus stem cells are labeled, we found that the proportion of Axin2+ cell-derived organoids was slightly higher upon IFN-γ-treatment (Fig. 8h), suggesting that stem cell capacity, in general, is maintained in Axin2-positive cells. We next treated organoids with BMP2 and IFN-γ to address whether IFN-γ affects stem cell differentiation under these conditions. We quantified the organoid forming efficiency when organoids were passaged, and a defined number of cells was re-plated in the regular organoid medium. BMP2 alone led to an almost complete loss of organoid forming capacity, which could be partially rescued by IFN-γ, although the absolute numbers were still much lower compared to untreated controls (Fig. 8i). Moreover, BMP2-driven overexpression of Muc5ac and Bmp2, but not loss of Lgr5, were inhibited by IFN-γ (Fig. 8j). Together, these data demonstrate that IFN-γ inhibits BMP2 signaling, shifting cell fate towards phenotypes present in the gland base, including mucous gland base cells and proliferative stem cells.

To address whether these effects of IFN-γ are also observed in vivo, we obtained gastric tissue from mice treated with IFN-γ for 14 days. Consistent with previous results, IFN-γ induced gland hyperplasia (Fig. 8k, l). Immunofluorescence revealed an expansion of Ki67+ cells in the base/isthmus, as well as increased numbers of Ki67+/Muc5ac+ cells (Fig. 8k, m). In situ hybridization for Bmp2 and Id1 revealed an expansion of cells at the base that express neither Id1 nor Bmp2 (Fig. 8n, o). Moreover, mice that were treated with IFN-γ expressed significantly higher levels of Grem2 in the stroma (Supplementary Fig. 5g), similar to the mice infected with H. pylori (Fig. 3c, d). Together, we conclude that IFN-γ inhibits BMP signaling in the stomach in both epithelial and stromal cells. This expands the gland base module by promoting stem cell self-renewal and accumulation of gland base secretory cells.



Source link

EHS
Back to top button