Dendritic cell–derived hepcidin sequesters iron from the microbiota to promote mucosal healing

See allHide authors and affiliations

Science  10 Apr 2020:
Vol. 368, Issue 6487, pp. 186-189
DOI: 10.1126/science.aau6481

Ironing out the details of mucosal healing

Anemia is a frequent complication of disorders such as inflammatory bowel disease, occurring in part as a result of increased bleeding into the intestine. Bessman et al. show that the peptide hormone hepcidin, which regulates systemic iron homeostasis, is required for intestinal repair in a mouse model of inflammatory bowel disease (see the Perspective by Rescigno). This effect was independent of hepatocyte-produced hepcidin and systemic iron levels. Instead, production of hepcidin by conventional dendritic cells was necessary and sufficient to promote local iron sequestration by macrophages, which in turn modulated the makeup of the gut microbiota to one with a more beneficial distribution of species.

Science, this issue p. 186; see also p. 129


Bleeding and altered iron distribution occur in multiple gastrointestinal diseases, but the importance and regulation of these changes remain unclear. We found that hepcidin, the master regulator of systemic iron homeostasis, is required for tissue repair in the mouse intestine after experimental damage. This effect was independent of hepatocyte-derived hepcidin or systemic iron levels. Rather, we identified conventional dendritic cells (cDCs) as a source of hepcidin that is induced by microbial stimulation in mice, prominent in the inflamed intestine of humans, and essential for tissue repair. cDC-derived hepcidin acted on ferroportin-expressing phagocytes to promote local iron sequestration, which regulated the microbiota and consequently facilitated intestinal repair. Collectively, these results identify a pathway whereby cDC-derived hepcidin promotes mucosal healing in the intestine through means of nutritional immunity.

Inflammatory bowel disease (IBD), colorectal cancer, and gastrointestinal infections cause tissue inflammation that drives bleeding, malabsorption, and diarrhea (13). As a result, patients frequently exhibit anemia that is difficult to treat, and bleeding introduces a new source of iron to the intestine (4, 5). Hepcidin, the master regulator of systemic iron homeostasis, is produced as a peptide hormone from the liver and promotes degradation of the cellular iron efflux transporter ferroportin (4, 610). Ferroportin is expressed on red pulp macrophages and the basolateral surface of duodenal enterocytes, where it facilitates iron recycling from senescent red blood cells and import of dietary iron, respectively (4, 610). Despite these advances, it remains unclear whether hepcidin has a role in gastrointestinal health or disease.

To address this, we exposed wild-type (Hamp+/+) and hepcidin-deficient (Hamp–/–) mice to a model of intestinal tissue damage, inflammation, and repair by administering dextran sodium sulfate (DSS) in their drinking water. During DSS administration, Hamp+/+ and Hamp–/– mice exhibited similar weight loss (Fig. 1A), indicative of comparable inflammation and tissue damage. However, upon removal of DSS, Hamp–/– mice exhibited persistent weight loss, continued disruption of epithelial crypt architecture, and significantly reduced colon lengths relative to controls (Fig. 1, A to C). Surprisingly, we observed significant DSS-dependent reductions in liver hepcidin expression, as well as reduced systemic hepcidin protein levels, relative to naïve controls (fig. S1, A and B). These data are consistent with negative feedback on hepcidin production due to anemia and erythropoiesis (11, 12). To test the role of hepatocyte-derived hepcidin, we bred mice with a floxed gene (HampF/F) to mice expressing Cre recombinase under the control of the albumin promoter, thereby generating HampΔliver mice that develop systemic iron overload comparable to hepcidin-deficient mice (13). After exposure to DSS, HampΔliver mice and controls exhibited comparable recovery of body weight, tissue architecture, and colon length (Fig. 1, D to F). Thus, hepcidin is essential for mucosal healing, but this occurs independent of hepatocyte expression and systemic iron regulation.

Fig. 1 Extrahepatic hepcidin promotes mucosal healing.

(A to C) Mice were given DSS for 7 days, and disease and recovery were monitored by weight loss (A), hematoxylin and eosin (H&E) staining of the distal colon (B), and colon shortening (C) at day 12. (D to F) Mice were given DSS for 9 days, and recovery was monitored by weight loss (D), H&E staining of the distal colon (E), and colon shortening (F). Data in (A), (B), (C), and (E) are representative of n = 3 to 5 mice per group replicated in two or more independent experiments; data in (D) and (F) are pooled from two independent experiments with n = 3 or 4. Data are means ± SEM. **P < 0.01 (unpaired two-tailed Student t test); ns, not significant. In (A) and (D), weights at killing, normalized to starting weight, were analyzed by unpaired two-tailed Student t test. Scale bars, 200 μm.

To interrogate the source(s) of hepcidin that promote mucosal healing, we analyzed tissues of naïve mice. We observed expression within the mesenteric lymph node (mLN) and lamina propria of the colon (cLP), which was maintained upon administration of DSS (Fig. 2A and fig. S1, C and D). Previous in vitro studies indicated that macrophages produce hepcidin (14). Surprisingly, we observed that type 2 conventional dendritic cells (cDC2s), and not macrophages or type 1 cDCs, were the dominant myeloid source of hepcidin in the colon after DSS administration (Fig. 2, B and C). Bacteria and bacteria-derived molecules were potent inducers of hepcidin expression in both bone marrow–derived DCs and sort-purified cDC2s (Fig. 2D and fig. S2). Intestinal biopsies from Crohn’s disease and ulcerative colitis patients revealed a significant increase in hepcidin expression relative to healthy controls, as well as significant correlations with DC-associated genes (Fig. 2E and fig. S3). Further, a recently described antibody to hepcidin (15) revealed that cDCs were major producers of hepcidin in the inflamed intestine of IBD patients (Fig. 2, F and G). Thus, cDCs are a previously unappreciated source of hepcidin in the intestine of mice and humans that is induced by microbes.

Fig. 2 Conventional dendritic cells are a source of hepcidin in the inflamed intestine.

(A) Hepcidin expression was analyzed by quantitative polymerase chain reaction (qPCR) in naïve mouse tissue. (B and C) Mice were given DSS for 7 days; colon lamina propria myeloid cells were sorted (B) and hepcidin expression was analyzed by qPCR (C). (D) cDC2s sorted from spleen were stimulated and hepcidin expression was analyzed. (E) Hepcidin expression was quantified from intestinal biopsies of humans. In (A), (C), and (D), representative data with n = 3 to 5 per group are shown, and data were replicated in at least two independent trials. In (E), n = 5 for the healthy group and n = 21 for the Crohn’s disease (CD) and ulcerative colitis (UC) groups. Data were analyzed by unpaired two-tailed Student t test [(A), (D), and (E)] or by Mann-Whitney U test (C). (F and G) Lamina propria cells from the inflamed ileum of pediatric CD patients were analyzed for hepcidin protein. In (F), representative histograms are shown. In (G), four different patients were tested and all data were pooled. Data are means ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001 [one-way analysis of variance (ANOVA) with Tukey multiple-comparisons test].

We deleted hepcidin in cDCs by crossing CD11cCre mice with HampF/F mice. HampΔCD11c mice exhibited a selective loss of hepcidin expression in cDCs (Fig. 3A and fig. S4A). DC development and systemic iron in HampΔCD11c mice were comparable to controls (fig. S4, B to D). Furthermore, lymphocyte, myeloid, and granulocyte responses were similar in HampΔCD11c mice and controls during naïve conditions and after administration of DSS (figs. S4, E and F, and S5, A to C). Global transcriptional profiling also revealed minimal changes in cDC subsets from HampΔCD11c mice relative to controls (fig. S6). Thus, cDC-derived hepcidin does not affect immune responses in these contexts. By contrast, HampΔCD11c mice exhibited significantly reduced body weight after removal of DSS, abnormal colon tissue architecture, and shortened colons relative to controls (Fig. 3, B to D). Zbtb46Cre-mediated deletion of hepcidin in cDCs resulted in a similar impairment of tissue repair relative to controls (fig. S7). Thus, cDC-derived hepcidin is essential for mucosal healing.

Fig. 3 Dendritic cell–derived hepcidin acts on ferroportin-expressing phagocytes to facilitate mucosal healing.

(A) Hepcidin expression was determined by qPCR in mice exposed to DSS for 7 days. (B to D) Mice were given DSS for 8 days, and recovery was monitored by weight change (B), H&E staining of the distal colon (C), and colon shortening (D). (E) Sort-purified cells from the naïve mouse colon were analyzed for Slc40a1 expression by qPCR. (F to H) Mice were given DSS for 7 days, and recovery was monitored by weight change (F), H&E staining of the distal colon (G), and colon shortening (H). Data in (D) and (H) were analyzed by unpaired two-tailed Student t test. In (B) and (F), weights at killing, normalized to starting weight, were analyzed by unpaired two-tailed Student t test. Data in (A) to (D) are representative of at least two independent experiments with n = 3 to 5 per group; data in (F) and (H) are pooled from, and data in (G) are representative of, three independent experiments with n = 1 to 3 per group. Data are means ± SEM. **P < 0.01. Scale bars, 200 μm.

We next profiled the colonic expression of ferroportin (Slc40a1) and observed high expression in epithelium, neutrophils, and macrophages (Fig. 3E). To determine whether these are the targets of hepcidin that facilitate mucosal healing, we used mice in which a hepcidin-resistant ferroportin variant, Slc40a1C326Y, is expressed from the endogenous locus after Cre-mediated recombination (fig. S8A) (16). The expression of Slc40a1C326Y in DCs or intestinal epithelial cells had no impact on mucosal healing (fig. S8, B to E). By contrast, Slc40a1C326Y expression in macrophages and neutrophils via LysMCre resulted in significantly reduced body weight, abnormal colonic tissue architecture, and shortened colons relative to controls and after removal of DSS (Fig. 3, F to H). Consistent with posttranslational regulation of ferroportin, DC-derived hepcidin did not have an impact on Slc40a1 or Hmox1 mRNA levels in macrophages (fig. S8F). Thus, ferroportin-expressing macrophages and/or neutrophils are a critical target for hepcidin-mediated mucosal healing.

To test whether this intestinal hepcidin-ferroportin axis regulates local iron distribution in the gut, we used quantitative mass spectrometry imaging. Strikingly, iron levels within the cecal tissue of DSS-treated HampΔCD11c mice were decreased relative to controls (Fig. 4A and fig. S9, A and B). Consistent with this, non-heme iron levels were increased in the luminal content of HampΔCD11c mice relative to controls after DSS-induced damage, but not in naïve mice (Fig. 4B and fig. S9, C and D). This likely involves conversion of heme-bound iron from erythrocytes into non–heme-bound iron via heme oxygenase 1 in phagocytes (17), which would then efflux to extracellular space through ferroportin unless regulated by hepcidin. Iron sequestration is a key component of nutritional immunity (4, 18, 19), so we examined whether cDC-derived hepcidin alters the microbiota. HampΔCD11c mice exhibited a significant shift in microbiota composition relative to littermate controls (Fig. 4C). Fecal microbiota transplantation (FMT) from HampΔCD11c mice to wild-type germ-free recipients was sufficient to transfer impaired mucosal healing relative to controls (fig. S10). Catenibacterium and Bifidobacterium were significantly different genera in HampΔCD11c mice relative to controls (fig. S11A). Notably, Bifidobacterium species support epithelial barrier function, and dietary iron supplementation can suppress Bifidobacterium species and exacerbate inflammation (20). We found that Bifidobacterium species expanded with restricted dietary iron, and that oral administration of Bifidobacterium species increased expression of intestinal tight junctions in wild-type mice and enhanced mucosal healing in HampΔCD11c mice (fig. S11, B to E). Bifidobacterium only partially restored normal mucosal healing, and the pathways by which DC-derived hepcidin promotes colonization with this microbe remain unclear. In addition, HampΔCD11c mice also exhibited significantly increased levels of tissue-infiltrating bacteria relative to controls after DSS exposure, and antibiotic treatment eliminated DSS-induced phenotypes (Fig. 4D and fig. S12). To determine whether excess extracellular iron impairs healing in HampΔCD11c mice, we administered deferoxamine (DFO), which sequesters extracellular iron from bacteria by chelation (21). DFO treatment in DSS-exposed HampΔCD11c mice was sufficient to completely restore mucosal healing (Fig. 4, E and F).

Fig. 4 Dendritic cell–derived hepcidin sequesters iron to shape the intestinal microbiota.

(A and B) Mice were exposed to DSS for 7 days. Whole cecal tissues were analyzed for iron levels by quantitative mass spectrometry imaging (A), and iron levels were quantified in colon lumen contents (B). (C) Fecal microbiota were analyzed by 16S rRNA gene sequencing and principal coordinates analysis. (D) Mice were exposed to DSS for 7 days, and bacterial colony-forming units (CFU) were quantified from colon tissue homogenates. (E and F) Mice were given DSS in drinking water for 7 days and treated daily with either phosphate-buffered saline (PBS) vehicle or DFO from day 0 through day 11. DSS-induced disease and recovery were monitored by weight loss (E) and H&E staining of the distal colon (F). In (A), two independent experiments with n = 1 to 5 per group were performed; representative data are shown. Data in (B) and (E) are pooled from two independent experiments, each with n = 3 to 5 per group. Data in (C) and (D) are representative of two independent experiments with n = 5 per group. In (B) and (D), groups were compared by unpaired two-tailed Student t test. In (C), P value was determined using a permutational multivariate ANOVA test. In (E), weights at killing, normalized to starting weight, were analyzed by one-way ANOVA using Tukey multiple comparisons. In (F), representative data are shown from two independent experiments with n = 3 to 5 per group. Data are means ± SEM. Scale bars, 200 μm. *P < 0.05, ** P < 0.01.

Our results outline a model in which cDCs produce hepcidin in response to microbiota-derived signals, and subsequently limit iron release from intestinal phagocytes to prevent tissue infiltration by the microbiota and thus promote mucosal healing (fig. S13). This contrasts with liver-derived hepcidin, which acts as an endocrine hormone, is induced by inflammatory cytokines, and has the potential to protect against systemic infection (7, 22, 23). It will be important to interrogate whether DC-derived hepcidin has the potential for a direct impact on the immune response (although this was not observed in our models) or on systemic iron homeostasis in other contexts. Furthermore, our results indicate that hepcidin mimetics could be a beneficial therapeutic strategy in the context of FMT or gastrointestinal diseases where mucosal healing is an emerging therapeutic goal.

Supplementary Materials

Materials and Methods

Figs. S1 to S13

References (2430)

References and Notes

Acknowledgments: We thank members of the Sonnenberg laboratory for discussions and critical reading of the manuscript, the Epigenomics Core of Weill Cornell Medicine, and S. Mozumder and K. Kim for technical assistance. Funding: Research in the Sonnenberg laboratory is supported by NIH fellowship F32AI124517 (N.J.B.); Crohn’s and Colitis Foundation fellowship 608975 (L.Z.); and NIH grants R01AI143842, R01AI123368, R01AI145989, R21CA249274, and U01AI095608, the NIAID Mucosal Immunology Studies Team (MIST), the Searle Scholars Program, an American Asthma Foundation Scholar Award, an Investigators in the Pathogenesis of Infectious Disease Award from the Burroughs Wellcome Fund, a Wade F. B. Thompson/Cancer Research Institute (CRI) CLIP Investigator grant, the Meyer Cancer Center Collaborative Research Initiative, Linda and Glenn Greenberg, and JRI (G.F.S.). G.F.S. is a CRI Lloyd J. Old STAR. Funding support also included the European Research Council (FP7/2011-2015 #261296); the “Fondation pour la Recherche Médicale” (DEq. 20160334903); the Laboratory of Excellence GR-Ex (ANR-11-LABX-0051); a labex GR-Ex fellowship (J.R.R.M. and S.L.); the French National Research Agency (ANR-11-IDEX-0005-02); the “Fondation ARC pour la recherche sur le cancer” (S.Z.); NIH grants PP30ES023515 and 1U2CES030859 (C.A. and M.A.); NIH grant R00HL125899 (S.M.C.); NICHD grant R00HD087523 (C.A.); and Science Foundation Ireland grant FRL4862 (S.M.C.). The JRI IBD Live Cell Bank is supported by the JRI, Jill Roberts Center for IBD, Cure for IBD, the Rosanne H. Silbermann Foundation, and Weill Cornell Medicine Division of Pediatric Gastroenterology and Nutrition. Author contributions: N.J.B. and G.F.S. conceived the project; N.J.B., L.Z., T.C.F., K.C.F., J.B.M., J.R.R.M., C.R., S.L., and S.Z. performed experiments and analyzed data; S.V. and S.L.-L. provided mouse models and expertise; H.S. performed pathological analyses; T.A. provided tools and expertise; C.P. provided mouse models and designed and supervised experiments; N.J.A. and G.G.P. analyzed sequencing data; C.A. and M.A. performed and analyzed iron imaging; R.E.S. provided essential advice and guidance; S.M.C. provided guidance and iron measurements; and N.J.B. and G.F.S. wrote the manuscript with input from all the authors. Competing interests: G.F.S. holds stock and is a member of an advisory board for Celsius Therapeutics Inc. T.A. is an employee of Amgen Inc. H.S. is a co-founder of Exeliom Biosciences and has received unrestricted study grants from Danone, Biocodex, and Enterome; board membership, consultancy, or lecture fees from Carenity, Abbvie, Astellas, Danone, Ferring, Mayoly Spindler, MSD, Novartis, Roche, Tillots, Enterome, Maat, BiomX, Biose, Novartis, and Takeda. The other authors declare no competing interests. Data and materials availability: All data necessary to understand and evaluate the conclusions of this paper are provided in the manuscript and supplementary materials. Microarray and 16S rRNA sequencing data are available from the GEO database with accession numbers GSE143869 and GSE139371. Floxed mice are available with a material transfer agreement with INSERM.

Stay Connected to Science

Navigate This Article