Iron-dependent BMP6 Regulation in Liver Sinusoidal Endothelial Cells Is Instructed by Hepatocyte-derived Secretory Signals.

The hepatocytic peptide hormone hepcidin regulates systemic iron homeostasis. Increases in iron levels stimulate hepcidin synthesis to reduce dietary iron uptake and iron release from iron recycling macrophages by blocking iron export into the circulation via ferroportin (Fpn).[1] The liver monitors both plasma and liver iron levels and controls hepcidin transcription by bone morphogenetic protein (BMP)/Son Of Mothers Against Decapentaplegic (SMAD) signaling. Critically, bone morphogenetic protein 6 (BMP6) is released in response to elevated hepatic iron levels,[2] but the underlying mechanisms to control Bmp6 expression are unresolved.

The growth factor BMP6 was initially identified as a regulator of bone development. However, lack of BMP6 does not affect bone homeostasis, but causes systemic iron overload in mice.[3] Moreover, BMP6 mutations are associated with iron overload in patients,[4-8] supporting its crucial regulatory function in iron metabolism. In the liver, BMP6 is mainly produced by liver sinusoidal endothelial cells (LSECs). Indeed, specific Bmp6 deletion in endothelial cells, but not in macrophages or hepatocytes (HCs), causes iron overload in the mouse due to low hepcidin levels.[9] Previous data suggest that the iron-dependent Bmp6 induction in LSECs is cell autonomous[10-12] and mediated by the activation of the antioxidant nuclear factor Nuclear factor erythroid 2-related factor 2 (NRF2) via intracellular iron accumulation.[12] This model is not supported by the analysis of LSECs isolated from mice with hemochromatosis (Hjv knockout [KO]), hallmarked by elevated Bmp6 levels but iron-deficient molecular signature in LSEC.[13] Here, we systematically dissect the mechanism of BMP6 control in response to liver iron overload and reveal that a factor produced by HCs is required for Bmp6 induction in LSECs.

We first explored the requirement of iron accumulation in LSECs for increased Bmp6 messenger RNA (mRNA) expression in vivo by comparing wild-type (wt) mice maintained on a “high iron” diet’ (Fehigh diet)[14] and Fpn(C326S) mice exerting iron overload due to a point mutation in Fpn that disrupts hepcidin binding[15] (Suppl. Figure S1A–H). From these mice, we isolated HCs and LSECs and analyzed ferritin L protein and transferrin receptor (Tfr) 1 mRNA levels, as markers of intracellular iron content. HCs and LSECs isolated from Fehigh diet mice show a molecular signature consistent with intracellular iron accumulation, with elevated ferritin protein and reduced Tfr1 mRNA levels (Suppl. Figure S1I–L; Figure 1B). Similarly, Fpn(C326S) mice show iron-loaded HCs. However, LSECs express similar levels of ferritin and increased Tfr1 mRNA in comparison with wt controls (Suppl. Figure S1I, J; Suppl. Figure S1M, N; Figure 1C); indicating that they are not iron-loaded but rather iron-deficient. We hypothesize that, in LSECs isolated from Fpn(C326S) mice, iron import via TfR1 (Figure 1A) and the non-transferrin bound iron importers is not sufficient to compensate for the increased Fpn-mediated iron export. Despite the absence of iron overload in LSECs isolated from Fpn(C326S) mice, Bmp6 mRNA expression is elevated in total liver (Suppl. Figure S1B, F) and purified HCs and LSECs (Figure 1E, F) from both mouse models. These findings extend and support data reported in Hjv KO mice by Rausa et al.[13] This strongly suggests that iron accumulation in LSECs is not the driving force for Bmp6 induction.

Figure 1.

Iron accumulation in LSECs is not a prerequisite for an iron-dependent increase of Bmp6 mRNA expression. (A–F), Total RNA was extracted from HCs and LSECs isolated from male wt mice maintained either on an iron-balanced or iron-loaded (Fehigh) diet for 4 wk or from male Fpn(C326S) mice and Fpn(wt) controls. All groups were sacrificed at 11 wk of age. Primary mouse LSECs were serum starved 2 h and then incubated with (G–I) 50 μM of FeNTA for 1, 3, 6, and 15 h and with (J) 20 μM of ML334 for 8 h or left untreated. Freshly prepared LSECs were serum starved for 2 h and then left untreated or incubated with (K and L) 30 μM of Apo-TF or Holo-TF for 18 h and (M and N) increasing concentration of ferritin (0.5 to 100 μg/mL). mRNA expression of (A–C, G, J, K, M) Tfr1 (Tfrc), (D–F, I, J, L, N) Bmp6, (H, J) Ho1, and (J) Nqo1 was analyzed by qRT-PCR and normalized to the housekeeping gene Rpl19. All treatments were performed in serum-free medium. A minimum of 3 biological replicates per condition is shown. A minimum of 2 independent experiments with at least 3 biological replicates was performed. In (G–I, M and N) results of 2 independent experiments are shown and they are divided by a dashed line. Data are reported as mean ± SEM and represented as (A, D) relative expression to HC (B and C, E and F) to control mice or (G–N) to vehicle treated or untreated cells (as indicated with the dashed line). Two-tailed Student t test and 1-way ANOVA were calculated: *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Apo-TF = Apo-transferrin; BMP6 = bone morphogenetic protein 6; Fehigh = high iron; FeNTA = iron nitrilotriacetate; Fpn = ferroportin; HC = hepatocyte; Ho1 = hemoxygenase1; Holo-TF = Holo-transferrin; LSECs = liver sinusoidal endothelial cells; mRNA = messenger RNA; ns = not significant; qRT-PCR = quantitative real time PCR; RQ = relative quantification; Tfr = transferrin receptor; wt = wild-type.

Among the different hepatic cell types, LSECs express the highest levels of Bmp6 (Figure 1D) and the biological relevance of this was demonstrated by endothelial cell-specific Bmp6 deletion.[9] Available LSEC cell lines are unable to recapitulate all features of primary LSECs.[16] We therefore aimed to validate data obtained in mice in a primary culture of mouse LSECs. As previously shown, this cell preparation is highly pure and maintains fenestrae, denoting lack of cell trans-differentiation.[14] Exposure of LSECs to iron nitrilotriacetate (FeNTA) decreases Tfr1 mRNA expression and activates hemoxygenase1 (Ho1) transcription (Figure 1G, H), indicating that intracellular iron accumulation induces oxidative stress (Figure 1G). By contrast, Bmp6 transcription was not induced (Figure 1I). Similar results were obtained in primary mouse LSECs incubated with ferric ammonium citrate (FAC) (Suppl. Figure S2). This finding supports our data obtained in Fpn(C326S) mice and suggests that iron accumulation in LSECs may not be a direct signal activating Bmp6 in an oxidative stress-dependent manner, as previously proposed.[12] To analyze whether the antioxidant transcription factor NRF2 controls Bmp6 expression, we applied ML334, a drug promoting NRF2 activation. We observed induction of NRF2 target genes (Ho1; Nqo1); however, mRNA expression of Tfr1 and Bmp6 remained unchanged (Figure 1J). These data demonstrate that NRF2 activation by iron and ML334 does not activate Bmp6 transcription in freshly prepared primary mouse LSECs.

We next explored whether different iron sources can elicit a BMP6 response. Transferrin saturation increases to approximately 100% in Fpn(C326S) mice.[15] However, holo-transferrin (holo-Tf) treatment of LSECs did not activate Bmp6 expression (Figure 1K, L), suggesting that a very high transferrin saturation does not contribute to Bmp6 upregulation. In addition, delivery of iron in the form of ferritin did not increase Bmp6 mRNA levels, despite eliciting Tfr1 reduction in a dose-dependent manner (Figure 1M, N). This indicates that the previously reported treatment of mice with ferritin[17] may have induced Bmp6 expression levels in LSECs indirectly. Taken together, these findings suggest that the iron-dependent Bmp6 regulation in LSECs is not cell autonomous.

In contrast to LSECs, HCs accumulate iron in Fehigh and Fpn(C326S) mice (Suppl. Figure S1I–N; Figure 1B, C). Given the spatial proximity of LSECs and HCs and reports of their reciprocal cross-talk, we hypothesized that HCs may be required for correct iron sensing and subsequent Bmp6 induction in LSECs. Therefore, we cultivated primary mouse LSECs and HCs, either alone or in co-culture, and treated with iron (Figure 2A, B). Remarkably, Bmp6 is only induced in the co-culture setting (Figure 2B). We next explored if the secretome of iron-loaded HCs is sufficient to induce Bmp6. Therefore, we incubated LSECs with the supernatant of a monoculture of HCs that was treated or not with iron (med(HC)). As controls, LSECs were also cultivated with unconditioned medium supplemented or not with iron (med). Importantly, we observed Bmp6 induction when LSECs were incubated with conditioned medium derived from iron-loaded HCs (Figure 2C), clearly demonstrating that molecule(s) released by iron-loaded HCs are sufficient to induce Bmp6 mRNA expression in LSECs. Of note, mice lacking the non-transferrin-bound iron importer Zrt- and Irt-like Protein 14 (ZIP14) and that are maintained on a Fehigh diet do not accumulate iron in HCs. Despite this, hepatic Bmp6 levels are increased,[18] suggesting that increased iron content in HCs may not be essential to observe Bmp6 upregulation in vivo. Therefore, we investigated whether the HC molecule(s) required for the iron-dependent Bmp6 induction in LSECs are secreted by HCs under steady state conditions, that is, in the absence of iron. LSECs were incubated with conditioned medium derived from untreated HCs, whereby iron (FeNTA) was added retrospectively to the supernatant. Iron import is not affected by adding HC secretome to LSECs, as Tfr1 mRNA expression is reduced (Figure 2D). Notably, Bmp6 is induced 6 hours after iron treatment (Figure 2E). Similar data were observed by replacing primary mouse HCs with the murine HC cell line Hepa1-6 (Suppl. Figure S3A–F). In contrast, the combined treatment of HC secretome and holo-Tf did not alter Bmp6 mRNA expression in primary LSECs (Suppl. Figure S3G); this suggests that only an unbound form of iron synergizes with HC-released molecules for the regulation of Bmp6. To explore if the factor released by HCs is a protein, we subjected the conditioned medium obtained from an HC monoculture to heat-inactivation, a process causing heat-dependent protein denaturation and precipitation, or to proteinase K digestion. Both treatments strongly reduced the amount of total protein within the HC-conditioned medium (Figure 2F, H). Importantly, when LSECs were exposed to the protein-depleted medium, Bmp6 induction in the presence of iron was prevented (Figure 2G, I).

Figure 2.

Crosstalk between primary LSECs and HCs is critical to induce Bmp6 mRNA levels in response to iron. (A and B), Primary mouse LSECs and HCs were maintained together (LSEC + HC) or separately and incubated with 50 μM of FeNTA for 6 and 15 h. The 2 time points represent 2 independent experiments. (C), Freshly prepared primary mouse LSECs were treated with unconditioned medium (med) supplemented or not with 50 μM of FeNTA for 6 h, or with medium from HC primary culture treated or not with 50 μM of FeNTA for 18 h (med(HC + FeNTA)). (D and E), Primary mouse LSECs were cultivated in HC derived medium, whereby HC remained untreated (med(HC)). Fifty μM of FeNTA was then added for 1, 3, and 6. (F and G), Hepatocyte conditioned medium (med(HC)) was heat-inactivated at 95°C, 30 min, and insoluble proteins were precipitated by high-speed centrifugation. (F), Coomassie blue staining of med(HC) boiled or unboiled. (G), Primary mouse LSECs were incubated with untreated or heat-inactivated med(HC) supplemented or not with 50 μM FeNTA. (H and I), Proteinase K was immobilized on agarose resin and used to digest proteins secreted in the med(HC). (H), Proteinase K treated and untreated medium was loaded on a SDS-PAGE and stained with Coomassie blue or (I) used to cultivate primary LSECs in presence or absence of 50 μM FeNTA. Total RNA was extracted and mRNA expression of (A, D) Tfr1 (Tfrc) and (B, C, E, G, I) Bmp6 was analyzed by qRT-PCR. Gene expression was normalized to the housekeeping gene Rpl19. All treatments were performed in serum-free medium. A minimum of 3 biological replicates per condition is shown. A minimum of 2 independent experiments with at least 3 biological replicates was performed. (J), Schematic representation of BMP6 regulation in response to iron. LSEC-secreted BMP6 induces hepcidin by activating SMAD1/5/8 signaling pathway in hepatocytes. The combination of a ubiquitously secreted hepatocyte molecule and iron induces BMP6 in LSECs that, subsequently, contributes to hepcidin upregulation. Data are represented as relative quantification to vehicle treated cells and reported as mean ± SEM. Two-tailed Student t test and 2-way ANOVA were calculated: *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. BMP6 = bone morphogenetic protein 6; BMPR = bone morphogenetic protein receptor; BMP-RE = bone morphogenetic protein responsive element; FeNTA = iron nitrilotriacetate; HC = hepatocyte; LSECs = liver sinusoidal endothelial cells; med = medium; mRNA = messenger RNA; MW = molecular weight; ns = not significant; ProtK = proteinase K; qRT-PCR = quantitative real time PCR; RQ = relative quantification; SDS-PAGE = sodium dodecyl sulphate - polyacrylamide gel electrophoresis; SMAD = Son Of Mothers Against Decapentaplegic; Tfr = transferrin receptor.

In conclusion, our findings demonstrate that in contrast to previous reports,[10-12] the iron-dependent Bmp6 regulation in LSECs is not cell autonomous. None of the tested iron sources, such as FeNTA, FAC, holo-Tf, or ferritin induced Bmp6 mRNA expression in primary culture of LSECs. We expect freshly prepared primary mouse LSECs used in this study to be superior in reflecting the iron response observed in vivo compared with endothelial cell lines or commercially available primary LSECs. By establishing a co-culture system of primary mouse LSECs and HCs, we show that cell-to-cell communication between HCs and LSECs is essential for iron sensing and subsequent Bmp6 regulation. Specifically, we demonstrate that proteins secreted by HCs induce Bmp6 in LSECs when combined with iron treatment (Figure 2J). This study uncovers a novel regulatory step of iron metabolism and establishes a basis for the future identification of regulators of the BMP6-hepcidin axis. Our results unravel a novel model of iron-dependent BMP6 regulation where HCs secrete protein(s) that function as BMP6 activator in the presence of iron. This model explains data from genetically modified mice, demonstrating that intracellular iron accumulation in both, HCs and LSECs, is not essential for increased BMP6 expression in response to iron overload. Finally, we hypothesize that component(s) of the HC secretome interact with iron, possibly in the noncellular space between HCs and LSECs (space of Disse), functioning as BMP6 activator(s).

ACKNOWLEDGMENTS

We thank the personnel of the animal facility at Heidelberg University and European Molecular Biology Laboratory (EMBL) for mouse housing and care. Also, we thank SMART Servier Medical Art (https://smart.servier.com/) for the graphic tools we used.

AUTHOR CONTRIBUTIONS

SC and MUM involved in conceptualization. SC, SA, OM, KM, and ARA involved in investigation. SC involved in visualization. SC and MUM involved in writing—original draft. SA, OM, and MWH involved in writing—review & editing. MWH and MUM involved in supervision. MUM involved in funding acquisition.

DISCLOSURES

MUM is a HemaSphere Associate Editor. All the other authors have no conflicts of interest to disclose.

SOURCES OF FUNDING

SFB1036, SFB1118, and DFG (FerrOs—FOR5146) provided research funding. OM is supported by a Junior Research Grant from the European Hematology Association and an Olympia-Morata-Programme fellowship provided by the Medical Faculty of the University of Heidelberg.