Bmp2 controls iron homeostasis in mice independent of Bmp6

Abstract

Hepcidin is a key iron regulatory hormone that controls expression of the iron exporter ferroportin to increase the iron supply when needed to support erythropoiesis and other essential functions, but to prevent the toxicity of iron excess. The bone morphogenetic protein (BMP)-SMAD signaling pathway, through the ligand BMP6 and the co-receptor hemojuvelin, is a central regulator of hepcidin transcription in the liver in response to iron. Here, we show that dietary iron loading has a residual ability to induce Smad signaling and hepcidin expression in Bmp6−/− mice, effects that are blocked by a neutralizing BMP2/4 antibody. Moreover, BMP2/4 antibody inhibits hepcidin expression and induces iron loading in wildtype mice, whereas a BMP4 antibody has no effect. Bmp2 mRNA is predominantly expressed in endothelial cells of the liver, where its baseline expression is higher, but its induction by iron is less robust than Bmp6. Mice with a conditional ablation of Bmp2 in endothelial cells exhibit hepcidin deficiency, serum iron overload, and tissue iron loading in liver, pancreas and heart, with reduced spleen iron. Together, these data demonstrate that in addition to BMP6, endothelial cell BMP2 has a non-redundant role in hepcidin regulation by iron.

INTRODUCTION

Iron is an essential element for the basic functions of life, including mitochondrial respiration, DNA synthesis, and oxygen transport.¹ Iron deficiency reduces cell survival and is a global problem in human health.^1,2 However, excess iron is also toxic by catalyzing the production of reactive oxygen species that can damage cells and tissues.³ A clinical scenario including liver cirrhosis, hepatocellular cancer, cardiomyopathy and diabetes mellitus characterizes diseases that lead to the accumulation of iron in humans, including hereditary hemochromatosis and iron loading anemias.^4–5 For these reasons, iron concentration at the cellular and organismal level must be kept within a narrow range.

The peptide hormone hepcidin is a key regulator of systemic iron balance by mediating the degradation of the iron exporter ferroportin on the surface of duodenal enterocytes, macrophages, and hepatocytes to inhibit dietary iron absorption and iron recycling from body storage sites.⁶ Hepcidin expression in the liver is controlled by stimulatory signals from iron and inflammation and inhibitory signals from erythropoietic drive.⁶ One of the most important regulators of hepcidin expression is BMP6^7–8, a member of the TGF-β superfamily. BMP6 is produced in liver endothelial cells in response to iron loading^9–10 and acts in a paracrine fashion on hepatocytes to bind the co-receptor hemojuvelin (HJV, also known as HFE2) and BMP type I and type II receptors to induce phosphorylation of SMAD1/5/8 proteins, which complex with SMAD4 and translocate to the nucleus to induce hepcidin transcription.^7,11 The importance of BMP6 and the BMP-SMAD signaling pathway in hepcidin and iron homeostasis regulation is demonstrated by the finding that mutation or ablation of Bmp6^7–8, HJV^12–14, type I receptors Acvr1 or Bmpr1A¹⁵, type II receptors Acvr2a and Acvr2b¹⁶, Smad1/5¹⁷, or Smad4¹⁸ result in hepcidin deficiency and hemochromatosis.

Previously, it was reported that iron is still able to stimulate hepcidin expression in global Bmp6 knockout (Bmp6−/−) mice, albeit to a lesser extent than in wildtype mice, suggesting the existence of a residual pathway for iron mediated hepcidin expression in the absence of BMP6.¹⁹ Moreover, HJV, the critical BMP co-receptor for hepcidin regulation^11–12, binds to several other BMP ligands with similar or higher affinity as BMP6, including BMP2.^20–21 Here, we used Bmp6 knockout mice, neutralizing BMP antibodies, and conditional Bmp2 knockout mice to investigate whether BMP2 contributes to hepcidin regulation and iron homeostasis.

MATERIALS AND METHODS

Animals

All animal protocols were approved by the Institutional Animal Care and Use Committee at the Massachusetts General Hospital. To generate conditional endothelial Bmp2 knockout mice, we purchased mice homozygous for a Bmp2 floxed allele, containing loxP sites flanking exon 3 of the targeted gene (B6;129S4-Bmp2^tm1Jfm/J, Jackson).²² Bmp2^fl/fl mice were crossed with mice expressing Cre recombinase under the control of a receptor tyrosine kinase Tek (Tie2) promoter/enhancer (B6.Cg-Tg(Tek-cre)1Ywa/J, hereafter Tek-Cre, Jackson). We previously reported efficacy and specificity of the Tek-Cre for recombination in liver endothelial cells compared with other liver cell populations using Cre reporter mice as well as genomic DNA and mRNA analysis of isolated hepatocytes and fluorescence-activated cell sorted endothelial cells and Kupffer cells.⁹ Since previous studies demonstrated that offspring from female Tek-Cre mice exhibit germline recombination^9,23, we used only male Tek-Cre mice for breeding. All mice were maintained on a mixed C57BL/6,129SvEv background with ad libitum access to water and a Prolab 5P75 Isopro RMH 3000 diet containing 380 ppm iron. Mice were genotyped by PCR of genomic DNA extracted from the tail and primers in Supplemental Table S1. Gender-matched littermate Bmp2^fl/fl;Tek-Cre− mice were used as controls throughout the study.

For BMP4 and BMP2/4 antibody injection experiments in wildtype mice, 8-week-old male 129S6/SvEvTac mice received an intraperitoneal (IP) injection of 10 mg/kg or 20 mg/kg monoclonal anti-human BMP4 (MAB757, R&D Systems) or 10 mg/kg anti-human BMP2/4 antibody (MAB3552, R&D Systems) daily for three days. Control mice received an IP injection of an equal volume of phosphate buffered saline (PBS) using the same dosing regimen. Six hours after the last injection, mice were sacrificed and blood, livers, and spleen were harvested for measurement of iron parameters and hepcidin expression.

For Bmp6−/− mouse experiments, male and female Bmp6−/− mice were generated as previously described⁹ and fed a low iron diet (2–6 ppm, Harlan Teklad) for 3 weeks upon weaning. Mice were then either kept on the low iron diet or switched to the standard rodent diet, which is high in iron (380 ppm), for 1 week without or with 4 IP injections of neutralizing BMP2/4 antibody (10 mg/kg body weight) or an equal volume of PBS alone, which were administered every other day. Animals were harvested 6 hours after the last injection.

For the iron challenge experiments, four-week-old C57BL/6 male mice were placed on a standard diet (380 ppm iron), a high iron diet containing 2% carbonyl iron (Harlan Teklad), or a low iron diet (2–6 ppm, Harlan Teklad) for 3 weeks. Alternatively, 129S6/SvEvTac male mice were placed on an iron sufficient diet (40 ppm, Research Diets) or high iron diet containing 1% carbonyl iron (~10 000 ppm, Research Diets) for 4 weeks upon weaning.

RNA Isolation, Reverse Transcription, and qRT-PCR

Total liver RNA was isolated using QIAshredder and RNeasy purification kits (Qiagen), and cDNA was synthesized from 1000 ng RNA using the iScript cDNA synthesis kit (Bio-Rad). qRT-PCR was performed as described previously²⁴ using the Power SYBR Green PCR Master Mix (Life Technologies) and primers in Supplemental Table S2. Transcript copy numbers of Bmp2 and Bmp6 were determined using Taqman Universal Master Mix, TaqMan primers and probes selected to target all known mRNA variants (Applied Biosystems), and standard curves generated from the plasmids in Supplemental Table S3.

Luciferase assay

Hepcidin promoter luciferase reporter assays in Hep3B cells were carried out using the Dual-Luciferase Reporter Assay System (Promega) as previously described.^7,11 In brief, Hep3B cells were transfected with 250 ng Hepcidin promoter luciferase reporter¹¹ and 25 ng pRL-TK Renilla luciferase control reporter. Forty-eight hours after transfection, cells were serum starved in 1% FBS for 6 h and treated with 50 ng/ml BMP2 (R&D Systems 355-BM), BMP4 (R&D Systems 314-BP), BMP5 (R&D Systems 615-BMC-020), BMP7 (R&D Systems 354-BP-010), BMP2/6 (R&D Systems 7145-BP-010), or 5ng/ml BMP9 (R&D Systems 3209-BP-010) for 16 h in combination with 0 to 30 µg/ml neutralizing BMP2/4 or BMP4 antibodies.

Iron Analysis

Serum iron and transferrin saturation were determined by colorimetric assay (Pointe Scientific) and quantitative tissue nonheme iron concentrations were determined by the method of Torrence and Bothwell as described previously.²⁴ Tissue iron is reported as µg iron/g wet weight tissue.

Prussian Blue Staining

Tissues were fixed in neutral buffered formalin for 24 h, embedded in paraffin, sectioned at 7 µm, deparaffinized and rehydrated, and subjected to Perls’ Prussian blue staining using an iron stain kit (American MasterTech). For tissues with less iron loading, a modified diaminobenzidine (DAB)-enhanced Perls’ stain containing cobalt chloride was used.²⁵

Hepatocyte, Kupffer cell, and endothelial cell isolation

Livers were perfused via the inferior vena cava, followed by collagenase digestion, centrifugation, and Percoll gradient centrifugation to obtain viable hepatocytes and nonparenchymal cell fractions as previously described.⁹ Liver endothelial cells and Kupffer cells (>99% purity) were isolated from the nonparenchymal cell fraction by fluorescence-activated cell sorting (FACS) as previously described.⁹

Immunoblot

Liver lysates were prepared, electrophoretically separated on a 10% polyacrylamide gel, and transferred to nitrocellulose as previously described.²⁶ Blots were incubated with blocking buffer (5% nonfat dry milk in TBST) containing rabbit anti-Smad5 (1:1000; Abcam ab40771), rabbit anti-P-Smad5 (1:500; Abcam ab92698, hereafter called P-Smad1/5 antibody due to crossreactivity with P-Smad1) overnight at 4°C. Antibodies were verified previously.²⁶ Membranes were washed, incubated with the appropriate HRP-conjugated secondary antibodies, and immunoreactivity was visualized by enhanced chemiluminescence (SuperSignal West Pico; Pierce) and x-ray film. Chemiluminescence quantitation of scanned films was performed using ImageJ 1.46v.

Statistics

Statistical significance was determined by 2-tailed Student’s t-test or one-way analysis of variance (ANOVA) with Dunnett’s or Tukey’s post hoc test for pairwise multiple comparisons using Prism 7 (GraphPad). P < 0.05 was considered significant.

RESULTS

Iron increases liver hepcidin expression and Smad1/5 phosphorylation in Bmp6−/− mice

A previous publication reported that dietary iron loading still induces hepcidin expression in iron depleted Bmp6−/− mice, although to a lesser extent than wildtype mice.¹⁹ However, the mechanism for this residual hepcidin induction by iron in Bmp6−/− mice is not known. To confirm this finding in our Bmp6−/− mice generated by a different strategy⁹, we tested the effects of maintaining Bmp6−/− mice on an iron deficient diet (2–6 ppm iron) for 4 weeks after weaning, compared with switching mice to a standard rodent diet (which is high in iron, 380 ppm) for the last 1 week prior to sacrifice. Bmp6−/− mice switched to the standard diet had significantly increased serum iron and transferrin saturation compared to low iron diet mice (Figure 1A–B, left 2 bars), demonstrating iron loading by the standard diet. Liver hepcidin (Hamp) mRNA was also significantly induced by dietary iron loading in Bmp6−/− mice (Figure 1C, left 2 bars), similar to the prior report.¹⁹ Notably, phosphorylation of liver Smad1/5 proteins (P-Smad1/5) was also induced by dietary iron loading in parallel with hepcidin in Bmp6−/− mice (Figure 1D). Together, these data confirm that there is another pathway for iron to induce hepcidin in the absence of BMP6, and suggest that this pathway may involve activation of SMAD signaling by another BMP ligand.

Figure 1. Dietary iron loading induces liver hepcidin expression and Smad phosphorylation in Bmp6^−/− mice, effects that are blocked by a neutralizing BMP2/4 antibody.

Bmp6^−/− mice were fed a low iron diet for 3 weeks upon weaning, and then kept on the low iron diet or switched to a standard diet for 1 week, either alone or in combination with 10mg/kg neutralizing BMP2/4 antibody (αBmp2/4) IP dosed every other day or an equal volume of vehicle alone (PBS). Mice were analyzed for serum iron (A), transferrin saturation (B), liver hepcidin (Hamp) relative to Rpl19 mRNA expression by qRT-PCR (C), liver phosphorylated Smad1/5 (P-Smad1/5) relative to total Smad5 by immunoblot and chemiluminescence quantitation (D, F), and liver iron content (G). n=5–11 males (blue circles) and 3–6 females (red triangles) per group in panels A, B, C, G. A subset of 5 mice per group was analyzed in panels D, F. Bars represent mean ± SEM. For qRT-PCR experiments, the average of low iron diet mice was set to 1. * P<.05, ** P<.01, ***P<.001 relative to the low iron diet or PBS group by Student’s t test. E) Hep3B cells were transfected with a hepcidin promoter firefly luciferase reporter and pRL-TK Renilla luciferase control reporter, followed by treatment without or with 50 ng/ml BMP2, BMP4, BMP2/6, BMP6, BMP5, BMP7, or 5ng/ml BMP9 for 16 hours in combination with 0 to 30 µg/ml αBMP2/4. Relative luciferase activity of ligand-treated cells in the absence of αBMP2/4 was set to 100%. Values represent mean ± SEM of 5 experiments for BMP2, BMP4, and BMP6, and 3 experiments for other BMPs, each performed in triplicate.

BMP2/4 neutralizing antibody inhibits hepcidin induction by iron and increases liver iron loading in Bmp6−/− mice

We previously demonstrated by surface plasmon resonance that HJV, the BMP co-receptor required for optimal hepcidin expression, has a high binding affinity not only for BMP6 (8.1 nM), but also for BMP2 and BMP4 (9.4 and 4.5 nM respectively).²⁰ HJV also binds to BMP5 and BMP7, although with lower affinity.²⁰ To investigate whether other BMP ligands, in particular BMP2 and/or BMP4, mediate hepcidin induction by iron in Bmp6−/− mice, we tested the ability of a neutralizing BMP2/4 antibody to inhibit hepcidin induction by dietary iron loading in Bmp6−/− mice. The efficacy and specificity of the BMP2/4 antibody was first examined in hepatoma-derived Hep3B cells using a hepcidin promoter luciferase reporter assay.⁷ The BMP2/4 antibody potently inhibited hepcidin promoter luciferase activity induced by either BMP2 or BMP4 (Figure 1E). However, at higher concentrations, this antibody also inhibited hepcidin promoter luciferase activity induced by other BMP ligands, including BMP5, BMP7, a BMP2/6 heterodimer, and, to a lesser extent, BMP6 (Figure 1E). BMP2/4 antibody did not inhibit BMP9 at the highest concentrations tested (Figure 1E) or other TGF-β super family members tested, including Activin B (data not shown).

Next, we placed Bmp6−/− mice at weaning on a low iron diet for 3 weeks followed by a standard diet in combination with the neutralizing BMP2/4 antibody or vehicle alone (PBS) for 1 week. The BMP2/4 antibody completely inhibited the induction of both P-Smad1/5 (Figure 1F) and Hamp by dietary iron loading (Figure 1C, compare right 2 bars). Blockade of this residual hepcidin induction by iron was physiologically relevant because it resulted in increased liver iron loading in BMP2/4 antibody treated mice compared with vehicle treated mice (Figure 1G), although serum iron, which was already high in Bmp6−/− mice, was not further increased (Figure 1A–B). Potential off-target effects of the BMP2/4 antibody to inhibit BMP6 were not relevant in this model where Bmp6 is ablated. These data support the hypothesis that hepcidin induction by iron in Bmp6−/− mice is dependent on other BMP ligands.

BMP2/4 antibody, but not BMP4 antibody, inhibits hepcidin expression and induces iron loading in wildtype mice

To determine whether BMP2/4 antibody inhibits hepcidin expression under control conditions when BMP6 is present and to begin to determine the identity of the ligand involved, we treated 8-week-old wildtype mice with BMP2/4 antibody (10mg/kg) or a specific neutralizing BMP4 antibody (10mg/kg) daily for 3 days. Efficacy and specificity of the BMP4 antibody was confirmed by hepcidin promoter luciferase reporter assay (Figure 2A). We were unable to validate a commercial antibody that specifically neutralized BMP2 activity without impacting BMP4 activity (data not shown). We found that BMP2/4 antibody, but not BMP4 antibody, significantly reduced liver Hamp mRNA in wildtype mice (Figure 2B). The reduction in Hamp in BMP2/4 antibody treated mice was physiologically relevant because it was associated with a reduction in spleen iron content, an increase in serum iron, and an increase in liver iron content (Figure 2C–E), consistent with other models of hepcidin deficiency.^{7–9, 13–14} Even higher doses of the BMP4 antibody (20mg/kg) failed to suppress Hamp mRNA, and the small increase in serum iron seen with the lower dose of BMP4 antibody was not reproducible in the higher dose study (Figure 2F–G). Together, these results suggest that BMP4 does not contribute to hepcidin and iron homeostasis regulation in vivo, but do suggest a possible role for BMP2. However, firm conclusions are limited by the off-target effects of the neutralizing BMP2/4 antibody on other BMP ligands at higher concentrations.

Figure 2. Bmp2 is predominantly expressed in liver endothelial cells, and neutralizing BMP2/4 antibody, but not BMP4 antibody, reduces Hamp mRNA and induces iron loading in wildtype mice.

A) Hep3B cells were transfected with a hepcidin promoter firefly luciferase reporter and pRL-TK Renilla luciferase control reporter, followed by treatment without or with 50 ng/ml BMP2, BMP4, BMP2/6, BMP6, BMP5, BMP7 or 5ng/ml BMP9 for 16 hours in combination with 0 to 30 µg/ml neutralizing BMP4 antibody (αBMP4). Relative luciferase activity was analyzed as in Figure 1E. Values represent mean ± SEM of 2 experiments, each performed in triplicate. B–E) Eight-week-old male wildtype mice received an IP injection of 10 mg/kg αBmp2/4, 10 mg/kg αBmp4, or PBS alone daily for 3 days. Animals were analyzed for liver Hamp relative to Rpl19 mRNA expression by qRT-PCR (B), spleen iron content (C), serum iron (D), and liver iron content (E). F–G) Eight-week-old male wildtype mice were injected with 20 mg/kg αBmp4 or PBS alone daily for 3 days. Animals were analyzed for liver Hamp relative to Rpl19 mRNA expression by qRT-PCR (F) and serum iron (G). A–G) n=4–6 mice per group. Bars represent mean ± SEM. For qRT-PCR results, the average of PBS treated mice was set to 1. ** P<.01, ***P<.001 relative to the PBS group by one-way ANOVA with Dunnett’s post hoc test (B–E) or Student’s t test (F–G). H–I) Hepatocytes, endothelial cells (EC), and Kupffer cells (KC) were isolated from wildtype male mouse livers (n=3 mice) and analyzed for Bmp2 (H) and Bmp6 mRNA (I) by qRT-PCR. Transcripts were normalized to Rpl19 and the average expression from KCs was set to 1. Bars represent mean ± SEM. ***P<.001 for ECs relative to hepatocytes and KCs by one-way ANOVA with Tukey post hoc test.

BMP2 is predominantly expressed in endothelial cells in the liver

To definitively establish whether BMP2 plays a role in hepcidin and iron homeostasis, we sought to develop a genetic mouse model. However, in contrast to Bmp6, Bmp2 global knockout mice are embryonic lethal.²⁷ We therefore determined the cellular source of Bmp2 in the liver using our previously described method to isolate highly purified hepatocytes, endothelial cells, and Kupffer cells (liver macrophages) from mouse livers using a FACS based approach⁹, followed by measurement of Bmp2 mRNA expression by qRT-PCR. We found that similar to Bmp6, endothelial cells are the predominant source of Bmp2 mRNA in the liver (Figure 2 H–I).

Generation of mice with a conditional ablation of Bmp2 in endothelial cells

Next we mated Bmp2^fl/fl mice²² with mice expressing Cre recombinase under the control of receptor tyrosine kinase Tek (Tie2) promoter/enhancer (B6.Cg-Tg(Tek-cre)1Ywa/J) to generate mice with an ablation of Bmp2 in endothelial cells (Bmp2^fl/fl;Tek-Cre+) and littermate Cre- controls (Bmp2^fl/fl;Tek-Cre−). We previously used a similar approach to generate endothelial Bmp6 conditional knockout mice (Bmp6^fl/fl;Tek-Cre+), which established a key role for endothelial cells in producing Bmp6 for hepcidin and iron homeostasis regulation.⁹ Efficacy and specificity of this Cre model for recombination in endothelial cells and other liver cell populations was evaluated in our prior study, which demonstrated >98% recombination efficiency in liver endothelial cells.⁹ Recombination in Kupffer cells was also seen with this Cre model.⁹ Bmp2^fl/fl;Tek-Cre+ mice exhibited >95% reduction in total liver Bmp2 mRNA expression (Figure 3A), confirming a functional loss of liver Bmp2 in this model. Low levels of residual Bmp2 mRNA in Cre- mice may be due to <100% recombination efficiency in endothelial cells or a small contribution of another liver cell type to Bmp2 production.

Figure 3. Mice with a conditional ablation of Bmp2 in endothelial cells (Bmp2^fl/fl;Tek-Cre+) exhibit reduced liver Bmp2 and Hamp mRNA expression, serum iron overload, tissue iron overload in liver, heart, and pancreas, and reduced spleen iron.

Eight-week-old littermate male (blue circles) and female (red triangles) Bmp2^fl/fl;Tek-Cre+ mice compared with littermate controls (Bmp2^fl/fl;Tek-Cre−) were analyzed for total liver Bmp2 (A), Hamp (B), and Id1 (C) relative to Rpl19 mRNA by qRT-PCR; serum iron (D); transferrin saturation (E); and tissue iron in liver (F,J), spleen (G,J), pancreas (H,J), and heart (I,J) by biochemical analysis (F–I) or Perls’ Prussian blue (J, liver and spleen) or DAB-enahanced Perls’ stain (J, pancreas and heart). n=4–13 mice per gender per group. Bars represent mean ± SEM. For qRT-PCR results, the average of Cre- mice for each gender was set to 1. Tissues from one representative mouse per group are shown in J (original magnification ×20; scale bar represents 100µm). *P<.05, **P<.01, ***P<.001 relative to the respective Cre- controls by Student’s t test.

Bmp2^fl/fl;Tek-Cre+ mice exhibit hepcidin deficiency and iron overload

Bmp2^fl/fl;Tek-Cre+ mice exhibited significantly reduced liver Hamp mRNA (Figure 3B) and the Bmp-Smad target transcript Id1 (Figure 3C). Consistent with a functional deficiency of hepcidin, Bmp2^fl/fl;Tek-Cre+ mice exhibited significantly increased serum iron and transferrin saturation (Figures 3D,–E), massive liver iron overload (Figure 3F, J), reduced spleen iron (Figure 3G, J), and extra-hepatic iron loading in pancreas and heart (Figure 3H–J) compared with littermate Bmp2^fl/fl;Tek-Cre− mice at 8 weeks of age. Both male and female mice exhibited hemochromatosis. However, although our study was not powered to detect gender differences in addition to genotype differences, female mice seemed to have a less severe phenotype in regard to hepcidin deficiency (Figure 3B) and extra-hepatic iron loading (Figure 3H–J) compared with male mice, similar to prior reports in Bmp6 global and endothelial knockout mice.^9,28

Iron treatment alters the ratio of Bmp2 to Bmp6 mRNA expression in the liver

Liver Bmp6 mRNA expression was not only preserved, but was upregulated in Bmp2^fl/fl;Tek-Cre+ mice (Figure 4A), confirming that the iron overload phenotype of the Bmp2^fl/fl;Tek-Cre+ mice is not due to Bmp6 deficiency. The increased liver Bmp6 mRNA levels in the Bmp2^fl/fl;Tek-Cre+ mice is consistent with previous literature showing that liver Bmp6 mRNA is positively regulated by dietary iron and correlates with liver iron content.¹⁰ To test whether Bmp2 mRNA is also regulated by iron, we treated wildtype mice with diets of different iron content. Total liver Bmp2 mRNA was negatively regulated by a low iron diet and positively regulated by a high iron diet, although to a lesser extent than Bmp6 (Figure 4B–C). A quantitative assessment of liver endothelial cell Bmp2 versus Bmp6 copy number showed that Bmp2 mRNA is 11.6-fold higher compared with Bmp6 mRNA on an iron sufficient diet (40 ppm), but only 2.9-fold higher on a 1% carbonyl iron diet (10,000 ppm, Figure 4D). Thus, although both Bmp2 and Bmp6 mRNA levels are regulated by iron, the higher baseline expression of Bmp2 and higher induction of Bmp6 by iron leads to a different ratio of Bmp2 to Bmp6 under conditions of different iron status.

Figure 4. Iron regulates liver Bmp2 mRNA expression to a lesser extent than Bmp6 and alters the ratio of Bmp2 to Bmp6.

A) Eight-week-old littermate male (blue circles) and female (red triangles) Bmp2^fl/fl;Tek-Cre+ mice compared with littermate controls (Bmp2^fl/fl;Tek-Cre−) were analyzed for total liver Bmp6 relative to Rpl19 mRNA by qRT-PCR. n=4–13 mice per gender per group. Bars represent mean ± SEM. The average of Cre- mice for each gender was set to 1. ***P<.001 relative to the respective Cre- controls by Student’s t test. B–C) C57BL/6 male mice were treated with a standard diet (380 ppm, Ctrl) compared to a low iron diet (2–6 ppm, B) or high iron diet (2% carbonyl iron, ~20,000 ppm, C) for 3 weeks followed by analysis of total liver Bmp2 and Bmp6 mRNA relative to Rpl19 expression by qRT-PCR. n=6–12 mice per group. Bars represent mean ± SEM. The average of standard diet mice was set to 1. **P< .01, ***P<.001 relative to standard diet mice by Student’s t test. D) Liver endothelial cells were isolated by FACS from 129S6/SvEvTac male mice treated with an iron sufficient diet (40 ppm) or a 1% carbonyl iron supplemented diet (~10,000 ppm) for 4 weeks, and Bmp2 and Bmp6 copy numbers were quantitated by qRT-PCR. Fold difference in Bmp2 relative to Bmp6 copy number for each diet is shown above bars. Fold change for each ligand on 1% carbonyl iron diet compared to iron sufficient diet is shown inside bars. n=4–6 mice per group. **P< .01, ***P<.001 for Bmp2 compared with Bmp6 expression for each diet.

DISCUSSION

Inappropriate hepcidin regulation leads to several disorders of iron homeostasis including hemochromatosis, iron loading anemias, iron refractory iron deficiency anemia, and anemia of inflammation.^1,6 The BMP-SMAD pathway is a central regulator of hepcidin expression in the liver.²⁹ Mutations in BMP-SMAD pathway components and modifiers of this pathway have a role in the pathogenesis of many of these iron disorders^{7–9, 12–18, 29}, and pharmacologic BMP-SMAD pathway modulators have shown efficacy in treating some of these disorders in animal models.^30–32 Identifying the specific components of the BMP-SMAD pathway crucial for hepcidin regulation may help to elucidate the pathophysiology of iron disorders and yield new treatment targets.

We and others previously demonstrated that BMP6 is a key endogenous regulator of hepcidin expression and systemic iron homeostasis since Bmp6 knockout mice have a hemochromatosis phenotype manifest by hepcidin deficiency, serum iron overload, and massive tissue iron loading in the liver, heart, and pancreas with reduced spleen iron content.^7–8 Here, we identify BMP2 as another BMP ligand with a nonredundant functional role in hepcidin and iron homeostasis regulation in vivo. Key evidence supporting an important functional role for BMP2 comes from demonstrating that mice with a conditional knockout of Bmp2 in endothelial cells (Bmp2^fl/fl;Tek-Cre+ mice) exhibit a hemochromatosis phenotype similar to Bmp6 knockout mice.

During the preparation of this manuscript, another group reported that ablation of Bmp2 in liver endothelial cells using another Cre driver (Stabilin-2 or Stab2) causes a similar iron overload phenotype.³³ These findings provide additional confirmation of an important role for BMP2 in hepcidin and iron homeostasis regulation. In contrast to the Tek-Cre used in the current study, which also causes recombination in Kupffer cells⁹, the Stab2 Cre is reported to be specific for liver endothelial cells as demonstrated by Cre reporter mice.³³ However, Tek-Cre-mediated recombination in Kupffer cells was only demonstrated in our prior study by higher exposure fluorescence imaging of Cre reporter mice, since expression from the Rosa26 locus in Kupffer cells is much lower compared to liver endothelial cells.⁹ Tek-Cre-mediated recombination in Kupffer cells in our prior study was also demonstrated by genomic DNA and RNA analysis of isolated liver cells⁹, which was not reported in the Stab2-Cre mice.³³ Notably, stabilin-2 expression has been reported in human monocyte derived macrophages, human alveolar macrophages, and several macrophage cell lines, where it has been proposed to play a role in clearance of aged red blood cells and other apoptotic cells.³⁴ It therefore remains possible that the Stab2 Cre is also not fully specific for endothelial cells in the liver. Nevertheless, the predominant expression of Bmp2 mRNA in liver endothelial cells (75- to 240-fold higher) compared with other liver cell populations does support the notion that endothelial cells are most likely the predominant source of Bmp2 for hepcidin and iron homeostasis regulation, similar to Bmp6.

A notable difference between the Bmp2^fl/fl;Tek-Cre+ mice in our study and the Bmp2 conditional knockout mice in the prior study³³ is that the Bmp2^fl/fl;Tek-Cre+ mice appear to have a stronger phenotype with increased iron loading in liver, heart, and pancreas. Moreover, the Bmp2^fl/fl;Tek-Cre+ mice exhibit reduced spleen iron, a characteristic feature of hemochromatosis due to increased iron export out of iron recycling macrophages as a consequence of hepcidin deficiency. In contrast, no changes in spleen iron were reported in the prior study.³³ Comparisons between the studies, however, are limited by the fact that the ages, genetic background, and details of dietary iron content, were not reported in the prior study³³, all of which have major impacts on systemic iron homeostasis. The use of different Cre drivers may also play a role. The hemochromatosis phenotype of the Bmp2^fl/fl;Tek-Cre+ mice is similar in severity to what we previously reported for endothelial Bmp6 conditional knockout mice generated using a similar strategy (Bmp6^fl/fl;Tek-Cre+)⁹, although the mixed background and differences in background strain also preclude a true quantitative comparison. Future studies comparing littermate endothelial Bmp2 conditional knockout mice and Bmp6 knockout mice on the same genetic background will be useful to determine the relative roles of BMP2 and BMP6 in vivo.

BMPs are dimeric proteins and can exist either as homodimers or as heterodimers. It has previously been reported that heterodimeric ligands may have a more potent effect compared with homodimeric ligands, and that in some biologic contexts BMPs must function as heterodimeric ligands, whereas homodimeric ligands are not able to compensate.^35–36 One reason for increased potency of heterodimeric ligands such as BMP2/6 is that BMP2 binds with higher affinity to type I receptors, whereas BMP6 binds with higher affinity to type II receptors. Heterodimeric BMP2/6 ligand may therefore bind with higher affinity to the type I/type II receptor complex compared with homodimeric BMP2 or BMP6 ligands.³⁶ It is also possible that heterodimeric BMP ligands facilitate the formation of heteromeric receptor complexes since BMP ligands also have different affinities for different type I and type II receptors.³⁵ Since BMP2 and BMP6 both play a role in hepcidin regulation and both appear to be made in endothelial cells, it is tempting to speculate that BMP2/6 heterodimeric ligands may have an important functional role in hepcidin and iron homeostasis regulation. Interestingly, Bmp2 is the dominant Bmp ligand expressed in liver endothelial cells in mice under basal conditions, whereas Bmp6 expression approaches that of Bmp2 after treatment with a high iron diet due to its greater induction by iron, at least at the mRNA level. Although secreted protein may not necessarily correlate with mRNA levels, this suggests that BMP2 homodimers may be the predominant species under control/low iron conditions, whereas BMP2/BMP6 heterodimer formation and/or BMP6 homodimer formation may increase under high iron conditions. Notably, the residual ability of iron to stimulate hepcidin in Bmp6−/− mice, and the ability of neutralizing BMP2/4 antibody to block this induction, suggests that BMP2 can function independent of BMP6 to stimulate hepcidin in vivo. Nevertheless, the iron overload phenotype of both Bmp6−/− mice and Bmp2^fl/fl;Tek-Cre+ mice demonstrates that neither BMP2 nor BMP6 can fully compensate for the loss of the other ligand, despite an appropriate induction of Bmp6 mRNA by iron in the Bmp2^fl/fl;Tek-Cre+ mice. Whether this inability to compensate is due to lesser amounts of total BMP ligand, or whether the ligand species (BMP2 homodimer, BMP6 homodimer, or BMP2/6 heterodimer) plays a role is an interesting area for future study.

The ability of the neutralizing BMP2/4 antibody to completely block hepcidin induction by iron in Bmp6−/− mice suggests that hepcidin induction by iron is fully dependent on the BMP signaling pathway. An important limitation of the BMP2/4 neutralizing antibody is that it is not specific for inhibiting BMP2 and BMP4 ligands, but can also inhibit most other ligands including BMP5, BMP7, BMP2/6, and, to a lesser extent, BMP6. A functional role for BMP2 is clearly established by the hemochromatosis phenotype of the Bmp2 conditional knockout mice. However, it is still possible that other BMP ligands contribute to hepcidin regulation in addition to BMP2 and BMP6. Indeed, we have previously demonstrated that many other BMP ligands can stimulate hepcidin expression when added exogenously, including BMP4, BMP5, BMP7, and BMP9.³⁷ Moreover, BMP4, BMP5, and BMP7 also bind to the co-receptor HJV that plays a critical role in hepcidin regulation in vivo.²⁰

A key role for BMP-SMAD signaling in hepcidin regulation in humans is supported by the fact that mutations in the BMP co-receptor HJV are a major cause of juvenile-onset hereditary hemochromatosis.¹² Although, evidence supporting a role for specific BMP ligands in humans is less robust than mice, heterozygous mutations in the BMP6 prodomain were recently linked to inappropriately low hepcidin levels and iron overload in humans.^38–39 Additionally, a significant association between serum ferritin and a common SNP in the BMP2 gene has been reported in humans.⁴⁰ Finally, BMP2 was reported to contribute to hepcidin excess in multiple myeloma patients⁴¹, although this findings relied on the BMP2/4 neutralizing antibody that we have demonstrated is not specific for inhibiting BMP2 or BMP4. Our data raise possibility that mutations in BMP2 may function as another cause of hereditary hemochromatosis of a modifier of disease penetrance.

In summary, our findings reveal a crucial, non-redundant role for BMP2 in hepcidin and iron homeostasis regulation. These results may provide new insights into the pathophysiology of iron disorders and may help in the development of new more targeted therapies.