Hepcidin upregulation by inflammation is independent of Smad1/5/8 signaling by activin B

To the editor:

Activin B, which is strongly induced by inflammatory stimuli in the mouse liver, has recently appeared as a potent inductor of hepcidin in vitro, via the crossactivation of noncanonical SMAD1/5/8 signaling.^1,2  To confirm the cause and effect relationships among activin B, Smad1/5/8 phosphorylation, and hepcidin in vivo, we challenged Inhbb^−/− mice³  (deficient in activin B) with lipopolysaccharide (LPS) or infected them with an Escherichia coli septicemic strain, as indicated in supplemental Methods (available on the Blood Web site).

To examine whether, as observed in human hepatoma cell lines and in mouse primary hepatocytes, activin B also stimulates canonical SMAD2/3 and noncanonical SMAD1/5/8 signaling in vivo, we compared by western blot analysis Smad2 and Smad5 phosphorylation levels in the liver of wild-type and Inhbb^−/− mice 4 hours after LPS stimulation (Figure 1A) or E coli infection (Figure 1B). Whereas these inflammatory stimuli strongly induce Smad2 and Smad5 phosphorylation in wild-type mice, none of these Smad effectors are activated in mice deficient in activin B. These data demonstrate that activin B, whose messenger RNA (mRNA) expression is strongly induced after LPS administration (Figure 1C) or E coli infection (Figure 1D), is required for activation of Smad1/5/8 signaling in these mice.

In parallel to activating SMAD1/5/8 phosphorylation in vitro, activin B was also shown to induce hepcidin expression.^1,2  Remarkably, pretreatment of hepatocytes with the BMP type I receptor inhibitor LDN-193189 prevented both the induction of SMAD/5/8 phosphorylation and the upregulation of the hepcidin (HAMP) gene expression by activin B,¹  suggesting that, at least in vitro, the effect of activin B on hepcidin expression was entirely attributable to the activation of noncanonical SMAD signaling. Since activin B also activates noncanonical Smad1/5/8 signaling in vivo, we expected that induction of hepcidin expression in response to LPS administration or E coli infection would be impaired in Inhbb^−/− mice. However, the magnitude of hepcidin mRNA induction and its evolution over time was unexpectedly similar in wild-type and in Inhbb^−/− mice challenged with LPS (Figure 1E) or infected with E coli (Figure 1F), pointing out the limitations of in vitro studies. To confirm the data at the protein level, serum hepcidin was quantified by competitive enzyme-linked immunosorbent assay (ELISA) in wild-type and in Inhbb^−/− mice before and 4 hours after a LPS challenge. As suggested by the quantitative polymerase chain reaction (PCR) data, and similarly to wild-type mice, Inhbb^−/− mice produce on average 3 times more hepcidin after endotoxin administration (Figure 1G). These results show that neither activation of Smad1/5/8 signaling nor activin B induction is necessary for upregulation of hepcidin production by inflammatory stimuli such as LPS administration or E coli infection. Hepcidin induction during inflammation was previously shown to be due at least in part to direct transcriptional regulation by the interleukin-6/STAT3 pathway.^4-6  Four hours after challenge with LPS or infection with E coli, Stat3 phosphorylation was similarly induced in wild-type and Inhbb^−/− mice (Figure 1A-B), suggesting a preponderant role of Stat3 activation in hepcidin induction by these inflammatory stimuli.

Mice lacking the iron-inducible Smad1/5/8-activating ligand Bmp6 have very low basal hepcidin levels.⁷  However, they respond to LPS by inducing liver expression of Inhbb as much as do wild-type mice¹  and, as shown in Figure 2A, this leads to a similar induction of Smad1/5/8 phosphorylation in the 2 categories of mice. Importantly, despite a marginally significant increase in the amount of circulating hepcidin 4 hours after a LPS challenge in Bmp6^−/− mice, the level reached after stimulation remains similar to that in unchallenged wild-type animals and ∼3 times lower than that in LPS-challenged wild-type mice (Figure 2B). These data highlight a lack of proportionality between Smad1/5/8 signaling and hepcidin production in the inflammatory context.

Circulating iron and tissue iron both activate the Smad1/5/8 signaling cascade in the hepatocyte, which leads to the induction of not only hepcidin mRNA but also the mRNA of other targets, such as Id1 and Smad7.^8,9  We therefore examined whether activation of Smad1/5/8 phosphorylation by inflammation was also accompanied by an induction of Id1 and Smad7 gene expression. As shown in Figure 2C-F, neither LPS nor E coli infection led to the induction of Id1 or Smad7 mRNA in the liver of either wild-type or Inhbb^−/− mice. Id1 mRNA expression was strongly repressed at the earliest time point (2 hours) but returned to baseline at 4 hours (Figure 2C-D). Smad7 mRNA expression was also strongly repressed at 2 hours post-LPS challenge or post–E coli infection, and it remained below baseline for the whole time period (Figure 2E-F). Notably, the kinetics of expression of these 2 genes were similar in wild-type and in Inhbb^−/− mice. These data point out major differences between Smad1/5/8 activation by iron and by inflammation. Indeed, whereas there is a good correlation between Smad1/5/8 activation by iron and induction of Hamp, Id1, and Smad7 gene expression,^8,9  such a concordance is totally lacking in the inflammatory context. It would be interesting to know whether activation of Smad1/5/8 signaling by inflammatory stimuli takes place in liver cells other than the hepcidin-producing hepatocytes, as this would explain the discrepancies observed between iron and inflammation.

To determine whether activin B gene expression is induced in any type of inflammation, we infected mice with different pathogens (bacteria and parasite) and injected them with turpentine to cause sterile tissue abscess. As shown in supplemental Figure 1, liver Inhbb mRNA expression was induced not only in a mouse model of infection with the gram-negative extracellular pathogenic bacteria E coli but also in mice infected with the gram-negative intracellular pathogenic bacteria Salmonella enterica serovar Typhimurium or the gram-positive Staphylococcus aureus. However, despite notable induction of hepcidin (Hamp) mRNA, liver Inhbb mRNA expression was not significantly increased in the malaria model induced by Plasmodium berghei K173–infected red blood cells or in the sterile inflammation induced by turpentine. These data suggest that production of activin B by the liver is a biomarker of bacterial infection rather than a key player in anemia of inflammation.

Bmp6^−/− mice have strong impairment of Bmp signaling, likely similar to the one achieved when treating wild-type animals with the BMP type I receptor inhibitor LDN-193189 or the BMP ligand antagonist ALK3-Fc. When challenged with LPS, they hardly increase their hepcidin production to the level seen in unchallenged wild-type mice. Much higher circulating hepcidin levels would be expected if the inflammation and iron (BMP) signals had additive transcriptional effects on hepcidin promoter. Rather, our in vivo observations suggest transcriptional synergy between these 2 signals and explain why lowering one of them using LDN-193189 or ALK3-Fc is sufficient to attenuate the induction of hepcidin gene expression by various inflammatory stimuli.^10-12  A similar attenuation is observed when BMP signaling is genetically impaired, as here in Bmp6^−/− mice or in mice with liver-specific deletion of Alk3.¹³  The present data are compatible with the previously proposed synergy between interleukin-6/STAT3 and BMP/SMAD signaling in regulating hepcidin.¹⁴  They show that full induction of hepcidin expression by inflammatory stimuli requires a functional BMP6-activated signaling pathway in the hepatocyte but is independent of activin B and its activation of Smad1/5/8 signaling that likely occurs in other cells of the liver.