Gremlin 1 depletion in vivo causes severe enteropathy and bone marrow failure

Abstract The intestinal epithelium is perpetually renewed from a stem cell niche in the base of crypts to maintain a healthy bowel mucosa. Exit from this niche and maturation of epithelial cells requires tightly controlled gradients in BMP signalling, progressing from low BMP signalling at the crypt base to high signalling at the luminal surface. The BMP antagonist gremlin 1 (Grem1) is highly expressed by subepithelial myofibroblasts adjacent to the intestinal crypts but its role in regulating the stem cell niche and epithelial renewal in vivo has not been explored. To explore the effects of Grem1 loss in adulthood following normal growth and development, we bred mice (ROSA26CreER‐Grem1 flx/flx) in which Grem1 could be deleted by tamoxifen administration. While Grem1 remained intact, these mice were healthy, grew normally, and reproduced successfully. Following Grem1 depletion, the mice became unwell and were euthanised (at 7–13 days). Post‐mortem examination revealed extensive mucosal abnormalities throughout the small and large intestines with failure of epithelial cell replication and maturation, villous atrophy, and features of malabsorption. Bone marrow hypoplasia was also observed with associated early haematopoietic failure. These results demonstrate an essential homeostatic role for gremlin 1 in maintaining normal bowel epithelial function in adulthood, suggesting that abnormalities in gremlin 1 expression can contribute to enteropathies. We also identified a previously unsuspected requirement for gremlin 1 in normal haematopoiesis. © 2020 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of Pathological Society of Great Britain and Ireland.


Introduction
The intestinal mucosa is covered by a perpetually selfrenewing layer of epithelium that is sustained by intestinal epithelial stem cells in each crypt's base. This stem cell niche is tightly controlled by key regulatory signals that are involved in intestinal stem cell renewal and differentiation, including Wnt, bone morphogenetic protein (BMP), and notch pathways [1,2]. Strict gradients of interacting morphogens control the exit of stem cell progeny from the niche driven by high Wnt and low BMP signalling in the lower half of the crypt, before progressively differentiating into post-mitotic specialised cells, controlled by low Wnt and high BMP signalling at the luminal surface [3].
A number of proteins that antagonise BMP functions including gremlin 1 (GREM1), gremlin 2 (GREM2), and noggin (NOG) are expressed locally by subepithelial myofibroblasts adjacent to the intestinal crypts. In vitro BMP antagonists are essential for culture of intestinal organoids or crypts, suggesting that the normal Wnt-BMP gradient requires the restricted paracrine secretion of these antagonists [4,5]. Of these BMP antagonists, gremlin 1, encoded by GREM1, which blocks BMP-2, BMP-4, and BMP-7 signalling, has been most directly linked to disease pathogenesis [6]. Human mixed polyposis syndrome is caused by a 40 kbp genetic duplication that results in excessive epithelial expression of GREM1 mRNA and transgenic overexpression of Grem1 initiates intestinal tumourigenesis in animal models [4,7]. Furthermore, GREM1 mRNA is highly expressed in the stroma of common colon cancers [7,8]. Taken together, those studies suggest that reducing GREM1 mRNA (and thus protein) expression or inhibiting GREM1 function may be a novel therapeutic strategy in intestinal disorders characterised by dysregulated Wnt-BMP signalling or aberrant gremlin 1 expression.
In contrast to the well-described role of increased GREM1 mRNA in disease pathogenesis, no studies to date have reported pathogenetic effects of reduced GREM1 mRNA in the bowel. The importance of the individual BMP antagonists in intestinal homeostasis in vivo has not previously been investigated. To assess the role of Grem1 in intestinal epithelial homeostasis, we induced widespread Grem1 deletion in adult mice, at a stage when normal growth and development were complete, by using tamoxifen-activated cre recombinase (ROSA26CreER-Grem1 fl/fl ), thus circumventing the perinatal lethality caused by complete loss of Grem1 in utero [9]. Our results demonstrate that Grem1 plays an indispensable role in maintaining the normal bowel epithelium. Unexpectedly, we also discovered an essential role for Grem1 in haematopoiesis.

Materials and methods
Adult male mice (sexually mature, 3-6 months old) expressing tamoxifen-activated cre recombinase driven by the ubiquitously expressed ROSA26 promoter and in which the coding sequence of both alleles of Grem1 had been flanked by loxP sites (ROSA26CreER-Grem1 fl/fl ) were given dietary tamoxifen to induce Grem1 depletion [10,11]. At the end of the experimental protocols, mice were deeply anaesthetised and then euthanised by cervical dislocation. Systematic postmortem examinations were undertaken and tissues fixed for histological examination. Marrow was isolated from the femur and tibia to count total cell numbers. In further mice, organs were removed post-mortem and immediately flash frozen for later analysis of mRNA. All protocols were approved by the UCD Animal Research Ethics Committee and licensed by the Department of Health, Ireland. Detailed methods may be found in supplementary material, and methods.

Results
In the absence of tamoxifen, adult ROSA26CreER-Grem1 fl/fl mice remained well for periods up to 18 months, successfully mated, and produced healthy offspring with equal numbers of male and female pups. When adult mice were fed tamoxifen to induce gremlin 1 depletion (Grem1 depl ), they initially appeared healthy. However, after 12-13 days of administration the first four mice given tamoxifen died unexpectedly. Postmortem examination of these revealed markedly dilated intestines, with a translucent intestinal wall, that contained watery yellow or gelatinous fluid mixed with small amounts of white digesta ( Figure 1A).
Following this, an intensive monitoring regimen was instituted, and any mice given tamoxifen that lost more than 25% of body weight or appeared unwell were euthanised after periods ranging from 7 to 12 days. All of the matched ROSA26Cre-Grem1 fl/fl mice in the control groups fed a normal diet (Grem1 intact ) remained well ( Figure 1B). Grem1 mRNA expression was markedly reduced ( Figure 1C) in the intestines of Grem1 depl mice (n = 8) when compared with the Grem1 intact mice (n = 8). Expression of Grem2 and Noggin (Nog), the two other BMP antagonists that are expressed in intestinal crypts and also block BMP-2, BMP-4, and BMP-7, was unchanged [12,13]. Wild-type mice and mice expressing the tamoxifen-inducible ROSA26 Cre

118
SC Rowan, H Jahns et al recombinase alone (i.e. in the absence of floxed Grem1 alleles) remained well during tamoxifen administration (data not shown), in keeping with previous reports [14]. A group (n = 8) of Grem1 depl mice was examined post-mortem and compared with Grem1 intact mice (n = 6). Dilated translucent intestines filled with clear or yellow gelatinous material were observed in four of the eight Grem1 depl mice, similar to the four mice initially exposed to tamoxifen that had died unexpectedly. Histopathological examination of the stomach was normal ( Figure 2). However, examination of the duodenum, jejunum, and ileum revealed marked villous stunting and loss ( Figure 2). The epithelial lining was mainly intact but consisted predominantly of large polygonal and cuboidal cells of varying sizes with abundant cytoplasm and large, centrally located nuclei and evidence of arrested cell proliferation (supplementary material, Figure S1). Crypts were collapsed or lost at multiple sites, while those that remained were lined by abnormal epithelial cells and frequently had no lumen ( Figure 2). Mitotic figures were rarely observed in the crypt-lining cells of Grem1 depl mice and there was marked reduction of Ki67 expression, indicating reduced cell replication (supplementary material, Figure S1C,D).
The caecum and colon of Grem1 depl mice showed similar widespread changes of the epithelial cells, with varying mucosal thickness and more superficial, abnormal crypts (Figure 3 and supplementary material, Figure S1). Mitotic figures were infrequent in the crypt epithelium. Staining for proteoglycans (Alcian blue) in the large intestine illustrated a loss of regularly arranged goblet cells (supplementary material, Figure S1E-H). Grem1 depletion in vivo causes enteropathy and bone marrow failure 119

SC Rowan, H Jahns et al
Grem1 intact mice showed normal mucosal structures in the caecum and colon. Serum urea, creatinine, and cholesterol were significantly reduced in Grem1 depl mice (supplementary material, Table S1), compatible with malabsorption and loss of muscle mass. In Grem1 intact mice, Grem1 mRNA expression was seen in pericryptal fibroblasts and in the muscularis mucosa along the entire gastrointestinal tract but most prominently in the ileum (Figures 2 and 3). Similar, but less intense Grem1 ISH staining was observed in stromal cells of the submucosa and in the tunica muscularis (supplementary material, Figure S2), in keeping with previous findings in the normal gastrointestinal tract [4,5,7,8,15,16]. In the Grem1 depl mice, Grem1 expression was markedly and extensively reduced throughout the gastrointestinal tract (Figures 2 and 3), although Grem1 expression was seen occasionally in small clusters of cells (supplementary material, Figure S3B,D).
The second major abnormality noted in Grem1 depl mice was that the four mice affected by the intestinal changes showed severe depletion of myeloid, lymphoid, and erythroid lineages in the bone marrow with replacement of cells by large vascular sinuses, an appearance similar to aplastic anaemia ( Figure 4A). Only scattered small groups of progenitor cells and megakaryocytes remained in the bone marrow. The number of bone marrow cells isolated from Grem1 depl mice was significantly less than that from Grem1 intact mice ( Figure 4B). Bone marrow cells obtained from the Grem1 depl mice had reduced Grem1 mRNA expression when compared with Grem1 intact mice ( Figure 4C). The peripheral blood reticulocyte count ( Figure 4D) was reduced in Grem1 depl mice compared with Grem1 intact mice (supplementary material, Table S2), compatible with the early stages of failure of erythropoiesis.
Histological examination of the liver, kidneys, adrenal glands, spleen, testes, lungs, and heart was normal in all Grem1 depl mice (data not shown).

Discussion
Our findings show for the first time that gremlin 1 plays an essential role in the maintenance of mucosal function in the adult bowel in vivo. Taken together with previous evidence that BMP signalling is required for maintenance of normal bowel structure and function [17], our results demonstrate that carefully regulated interactions of both the ligands and the antagonist are needed for a healthy bowel in vivo. Disruption of this balance may contribute to tumour development, as previously demonstrated, but may also play a role in enteropathies characterised by mucosal atrophy and failure.
The severe disruption of bowel structure and function that we observed after Grem1 depletion contrasts sharply with the absence of bowel disturbance reported by Davis et al [4] following Grem1 depletion induced using a different, tamoxifen-inducible cre-recombinase (CAGG-CreErT2), which was at least as effective as the ROSA26-CreERT2 that we used. Differences in the genetic backgrounds of the different mouse colonies may explain the very different phenotypes. Another potential explanation for the differences is that the bowel abnormalities were secondary to immunosuppression caused by the bone marrow failure that we found. This failure was unexpected since even though Grem1 is expressed in the normal marrow [15,18], a requirement for Grem1 in normal haematopoiesis has not previously been reported. However, the peripheral blood white cell counts were unchanged at the time our mice became unwell and histopathological examination did not show the epithelial ulceration, bacterial invasion, and necrosis typical of neutropenic enterocolitis [19]. Interestingly, in a paper published while our manuscript was under review, McCarthy et al [20] reported that ablation of Grem1-expressing cells in the small intestinal wall (achieved using a tamoxifen-inducible diphtheria toxin receptor regulated by the Grem1 promoter) produced a small intestinal phenotype closely similar to that which we report. Taken together with our data, this suggests that the bowel abnormalities were a result of the direct effects of Grem1 loss in the bowel. It is also worth noting that although Davis et al [4] did not find any bowel abnormality in the absence of other genetic mutations, they observed a reduced polyp burden following Grem1 depletion in mice carrying a tumour-promoting mutation of the adenomatous polyposis coli locus, a finding compatible with the role for Grem1 in the regulation of epithelial proliferation that we found.
The abnormalities caused by Grem1 depletion were seen only in tissues with very rapid turnover, i.e. bowel epithelium and bone marrow cells. Thus, even though we found no abnormalities in any other organs, Grem1 may have important functions in those organs that were not manifested because of the rapidly fatal effects of Grem1 loss in the bowel and marrow.
In summary, our results demonstrate for the first time an essential homeostatic role for gremlin 1 in the maintenance of normal bowel function in adulthood in vivo and demonstrate a previously unsuspected, essential requirement for normal gremlin 1 expression in bone marrow function. Moreover, our findings show that gremlin 1 is non-redundant among the BMP antagonists. Taken together, our findings suggest that abnormally reduced or increased expression of gremlin 1 may play a significant role in disease development in the gastrointestinal tract and bone marrow.  Figure S3. Grem1 mRNA expression was markedly and extensively reduced in the gastrointestinal tract of Grem1 depl mice, although some staining was occasionally evident. Representative images of the intestine of Grem1 intact (first column) and Grem1 depl mice (second column) Table S1. Results of the serum biochemical analyses