Erythroferrone lowers hepcidin by sequestering BMP2/6 heterodimer from binding to the BMP type I receptor ALK3

TO THE EDITOR:

Hepcidin is a key iron regulatory hormone that downregulates the iron exporter ferroportin to limit iron entry into the circulation from dietary sources and body stores.¹  Hepcidin is suppressed by the bone marrow–derived hormone erythroferrone (ERFE) in response to erythropoietin (EPO).²  This action has a physiological role to ensure adequate iron for erythropoiesis in the context of anemia but contributes to iron overload in diseases of ineffective erythropoiesis.³  We recently discovered that ERFE suppresses hepcidin by interfering with bone morphogenetic protein (BMP)-SMAD signaling, hepcidin’s central transcriptional regulatory pathway in the liver.⁴  This signaling pathway is activated by endothelial cell–derived BMP6 and BMP2 ligands^5-7  binding to hepatocyte BMP type I and type II receptors^8-10  and the coreceptor hemojuvelin¹¹  to phosphorylate SMAD1/5/8 transcription factors,¹²  which activate the hepcidin promoter. We showed that ERFE and EPO inhibited SMAD1/5/8 phosphorylation and hepcidin transcription in primary hepatocytes and mice but failed to suppress hepcidin when Smad1 and Smad5 were ablated.⁴  ERFE was subsequently shown to inhibit hepcidin induction by BMP5, BMP6, and BMP7 (but not BMP2, BMP4, or BMP9) in vitro. It was postulated that ERFE suppresses hepcidin by acting as a ligand trap for BMP5/6/7 subfamily ligands.¹³  However, a direct interaction between ERFE and BMPs was not shown, and it remains possible that ERFE interacts with BMP receptors that preferentially use BMP5/6/7. Importantly, the detailed mechanism by which ERFE suppresses hepcidin in vivo remains unknown. Here, we examined whether ERFE-mediated hepcidin suppression depends on specific BMP receptors in vitro and in mice, and we determined the binding interactions between ERFE, BMP ligands, and BMP receptors.

Primary hepatocytes isolated from C57BL/6x129S5 background mice were cultured as previously described.⁴  Cells were then transfected with 80 nM of small interfering RNA (siRNA) targeting BMP receptors or control siRNA (Invitrogen) for 30 hours, and treated with conditioned medium containing 50% (vol/vol) supernatant from control HEK293T cells (CTRL-CM) or HEK293T cells overexpressing mouse ERFE (ERFE-CM)²  for 15 hours. Hep3B cells were transfected with 200 ng of mouse Erfe complementary DNA or empty vector (Origene) for 48 hours and stimulated with 5 ng/mL of BMP ligand for 6 hours. Animal protocols were approved by the Institutional Ethics Committee of the North Rhine-Westphalian Agency for Nature, Environment, and Consumer Protection at Muenster University, Muenster, Germany (permit number Az. 84-02.04.2014.A015, Az.84-02.04.2013.A281). Hepatocyte-specific Alk2 (Alk2^fl/fl;Alb-Cre) or Alk3 (Alk3^fl/fl;Alb-Cre)–deficient mice and Cre- controls on a C57BL/6 background⁸  were fed a chow diet (198 parts per million iron). Eight-week-old female mice were given an intraperitoneal injection of EPO (200 U; Janssen) or phosphate-buffered saline (PBS) and were euthanized after 15 hours. RNA isolation and quantitative reverse transcription–polymerase chain reactions were performed and analyzed as described^4,10,14  by using the primers shown in supplemental Table 1 (available on the Blood Web site). Transcript levels were normalized to Rpl19 or 18S as internal controls. For immunoprecipitation and pull-down assays, 0.5 to 2 μg of ALK3-Fc, ALK2-Fc, BMP2, BMP6, or BMP2/6 (R&D Systems) were mixed with 0.3 to 3 μL of ERFE-FLAG at 52.9 ng/μL concentrated from serum-free ERFE-CM using a 30 KDa Filter Unit (Ambion) in NETN buffer. The ERFE-FLAG concentration was determined by using an enzyme-linked immunosorbent assay (Intrinsic LifeSciences). Samples were incubated with anti-FLAG affinity gel (MilliporeSigma) or protein A agarose (Thermo Scientific) at 4°C overnight, and eluted with 3× FLAG peptide (MilliporeSigma) or 2× Laemmli buffer before immunoblot using antibodies (supplemental Table 2). Data are presented as box plots with minimum to maximum whiskers or mean ± SEM. Means were compared by using the Student paired t test, one-way analysis of variance with Tukey’s post hoc test, or (for nonparametric data sets) the Mann-Whitney U test using Prism 7 (GraphPad Software). P < .05 was considered significant.

To explore the role of BMP receptors in hepcidin suppression by ERFE, we first performed siRNA knockdown experiments in primary hepatocytes. siRNA knockdown efficiency and specificity were verified (supplemental Figure 1). ERFE suppressed hepcidin messenger RNA (mRNA) in cells transfected with siRNAs targeting each of the three type II receptors and the type I receptor Alk2 but not Alk3 (Figure 1A-B). To determine whether ALK3 is also important for hepcidin suppression by ERFE in vivo, we injected EPO or PBS into mice with or without hepatocyte Alk2 or Alk3 deficiency. EPO similarly induced spleen Erfe mRNA in all mouse groups (supplemental Figure 2) and decreased liver hepcidin mRNA in both Cre- controls (Figure 1C-D). EPO also decreased liver hepcidin in Alk2^fl/fl;Alb-Cre mice, albeit to a lesser extent (Figure 1C). However, EPO failed to decrease hepcidin in Alk3^fl/fl;Alb-Cre mice (Figure 1D). Although basal hepcidin expression was low in Alk3 siRNA transfected cells and Alk3^fl/fl;Alb-Cre mice, hepcidin was further suppressible by serum-free medium (in cells) and a low-iron diet (in mice), showing that hepcidin was not yet at its nadir (supplemental Figure 3). Thus, ALK3 is functionally important for hepcidin suppression by EPO and ERFE, whereas ALK2 plays a minor role.

We next used immunoprecipitation/pull-down assays to explore whether ERFE directly binds ALK3 or ALK2. In addition, BMP receptor knockdown experiments were used to examine whether the preferential utilization of different receptor subsets by BMP ligands could account for ERFE’s differential inhibition of BMP5/6/7 but not BMP2/4/9. Finally, we tested whether ERFE directly binds BMP ligands as recently proposed,¹³  focusing on BMP6 and BMP2, the confirmed physiological regulators of hepcidin in vivo.^5-7  ERFE selectively bound BMP6 but required higher concentrations to bind BMP2 (Figure 2A). In contrast, ERFE did not bind ALK3 or ALK2 (Figure 2B; supplemental Figure 4). As positive controls, both BMP2 and BMP6 bound ALK3, and BMP6 bound ALK2 (supplemental Figure 5). However, whereas BMP2 preferentially used ALK3, BMP6 favored ALK2 for hepcidin induction in Hep3B cells (Figure 2C). These data provide the first demonstration of direct binding between ERFE and BMP ligands but not BMP receptors. They also confirm the selectivity of ERFE for binding BMP6 compared with BMP2. However, these data do not explain why EPO failed to suppress hepcidin in Alk3^fl/fl;Alb-Cre mice, in which BMP6-ALK2 signaling should remain intact.

To understand the in vivo observations, we took advantage of recent data that hemochromatosis severity was not exacerbated in double endothelial Bmp2 and Bmp6 knockout mice compared with single knockout mice,¹⁵  suggesting that BMP2 and BMP6 work collaboratively in vivo, potentially as a heterodimer. We thus examined whether heterodimeric BMP2/6 binds ERFE and its type I receptor utilization. We found that unlike BMP6, BMP2/6 preferentially used ALK3 for hepcidin induction (Figure 2D). Moreover, BMP2/6-mediated hepcidin induction was suppressed by ERFE (Figure 2E). As previously reported, ¹³  ERFE also inhibited hepcidin induction by BMP6. In contrast, although hepcidin levels were lower in cells treated with BMP2 and ERFE compared with BMP2 treatment alone, the fold-induction was not reduced after adjusting for lower basal hepcidin levels in ERFE-treated control mice (Figure 2E). This outcome suggests that ERFE does not functionally inhibit BMP2 at concentrations sufficient to inhibit BMP6 and BMP2/6. Finally, ERFE bound directly to BMP2/6 heterodimers and prevented BMP2/6 binding to ALK3 (Figure 2F-G; supplemental Figure 6). These data suggest that ERFE inhibits hepcidin by acting as a ligand trap for the physiological form of BMP ligand in vivo, BMP2/6, which preferentially utilizes ALK3 for signaling (Figure 2H). These results are consistent with the major role of ALK3 compared with ALK2 in hepcidin suppression by EPO/ERFE, and they help explain the more robust iron overload phenotype in Bmp6^−/− and Alk3^fl/fl;Alb-Cre mice compared with Alk2^fl/fl;Alb-Cre mice.^7,8,16