|Home | About | Journals | Submit | Contact Us | Français|
The hepatic peptide hormone hepcidin controls the duodenal absorption of iron, its storage and its systemic distribution. Hepcidin production is often insufficient in chronic hepatitis C and alcoholic liver disease, leading to hyperabsorption of iron and its accumulation in the liver. Hepatocyte growth factor (HGF) and epidermal growth factor (EGF) mediate the hepatic regeneration after liver injury. We examined the effect of the growth factors on hepcidin synthesis by hepatocytes. Results: HGF and EGF treatment of primary mouse hepatocytes, as well as EGF administration in mice, suppressed hepcidin mRNA synthesis. The suppression of hepcidin by these growth factors was transcriptional, and was mediated by a direct effect of HGF and EGF on the BMP pathway regulating hepcidin synthesis. We further showed that growth factors interfered with nuclear localization of activated Smads and increased the nuclear pool of the BMP transcriptional co-repressor TG-interacting factor (TGIF). In a kinase screen with small-molecule kinase inhibitors, inhibitors in the PI3 kinase pathway and in the MEK/ERK pathway prevented HGF suppression of hepcidin in primary mouse hepatocytes. Conclusion: HGF and EGF suppress hepatic hepcidin synthesis, in part through PI3 kinase MEK/ERK kinase pathways which may be modulating the nuclear localization of BMP pathway transcriptional regulators including activated Smads1/5/8 and the co-repressor TGIF. EGF, HGF and possibly other growth factors that activate similar pathways may contribute to hepcidin suppression in chronic liver diseases, promote iron accumulation in the liver and exacerbate the destructive disease processes.
The hepatic hormone hepcidin controls the flows of iron from dietary absorption, storage and recycling into blood plasma, and thereby regulates plasma iron concentrations and stores (1). In turn, plasma iron concentrations and hepatic iron stores transcriptionally modulate hepcidin synthesis (2, 3), completing a homeostatic feedback loop. The BMP pathway is essential for iron and hepcidin regulation (4). The BMP receptor complex and a range of BMP ligands, including BMP6, induce hepcidin expression by activating Smad4 and Smad1/5/8 (5–7). Human mutations that cause hereditary hemochromatosis either ablate the hepcidin gene (rare) or affect iron-specific hepcidin regulatory proteins that are thought to interact with the BMP pathway (1).
Hepcidin insufficiency and hepatic iron loading are seen in chronic hepatitis of multiple etiologies, including alcoholic hepatitis and viral hepatitis (8–10) and the resulting chronic iron loading in the liver worsens disease prognosis (11). The mechanism of hepcidin suppression in chronic hepatitis is not known.
Chronic hepatitis is characterized by repeated liver injury and repair. Growth factors mitogenic for hepatocytes are important mediators of the liver repair and regeneration. Hepatocyte growth factor (HGF) and epidermal growth factor (EGF) are well-characterized mediators of hepatic regeneration following experimental injury (12–14). We explored the modulation of hepcidin synthesis by these growth factors.
* Detailed methods are provided in the Supplement
Murine EGF, HGF, IGF-1 and IGF-2, rat PDGF-BB, and human BMP6 were from R&D Systems, Minneapolis, MN. Recombinant mouse IL-6 and recombinant human EGF were from Millipore, Billerica, MA. Kinase inhibitors EHT1864, PHA665752, NSC23766, 10-DEBC hydrochloride were from Tocris Bioscience, St. Louis, MO, and U73112, LY294002, Calphostin, JNK Inhibitor II, STAT3 inhibitor VII, U0126, ERK inhibitor peptide II FR180204, Akt inhibitor II and Akt inhibitor X from Millipore.
Hepatocytes were isolated from six to eight-week old wild-type C57BL/6 mice by a two-step portal vein collagenase perfusion method, and used within hours or after 18h incubation with serum-free William’s E Medium (serum-starved). Transfection of hepatocytes and HepG2 cells was done with Nucleofector (Lonza Group Ltd, Basel, Switzerland) according to the manufacturer’s instructions. The hepcidin-luciferase reporter included human hepcidin promoter spanning −1 to −2997 (15), and the BRE-luciferase reporter was obtained from H.Y. Lin (16). Luciferase activity was measured by a Veritas Microplate Luminometer (Turner Biosystems, now Promega, Sunnyvale CA).
Quantitative real-time RT-PCR data are presented as either fold-change relative to control or using the ddCt method which naturally yields a logarithmic scale. Fold-change was calculated by the method of Pfaffl (17), where the target gene (Hepc1 or ID1) was referenced to a housekeeping gene (β-actin) and the data presented as a ratio to the control treatment within each experiment. The average relative expression value of the triplicate control treatments was assigned as 1 in each experiment. Relative quantification was performed using the comparative Ct, or ddCt method (18). The target gene (Hepc1 or ID1) was first normalized to a reference housekeeping gene (β-actin) and then presented as the difference from the control treatment within each experiment. The average ddCt value of the triplicate control treatments was zero in each experiment.
Cells for whole-cell lysates were plated in 60 mm collagen-coated dishes (BD Biocoat, Franklin Lakes, NJ) and cells for fractionated lysates (nuclear and cytosolic) were plated in 10-cm collagen-coated dishes. After treatment, whole-cell lysates were collected in ice-cold NETT buffer (150 mM NaCl, 10 mM EDTA, 10 mM Tris, 1% Triton X-100) containing HALT protease/phosphatase inhibitor cocktail (ThermoScientific/Pierce, Rockford, IL). Fractioned lysates were separated into nuclear and cytosolic fractions using the NE-PER kit (Pierce) according to the manufacturer’s instructions. Lysates (30 μg of fractionated lysates, 50 μg of whole-cell lysates) were electrophoresed on 4–20% LongLife iGels (NuSep Inc, Lawrenceville, GA) and transferred to Immobilon-P PVDF membrane (EMD Millipore) for Western blotting, visualization by chemiluminescence, and quantification with ChemiDoc XRS+ System with Image Lab software (Biorad, Hercules, CA).
Twenty-four hours after isolation, serum-starved hepatocytes were pre-treated with kinase inhibitors in a 0.5 μl volume of DMSO. Each kinase inhibitor was added to duplicate or triplicate wells and after 1h, 20 ng/ml HGF was added, then 1h later 10 ng/ml BMP6 was added. Depending on the kinase inhibitor, cultures were incubated for a minimum of 8 hours or overnight prior to sample collection.
WT 6-week-old C57BL/6 mice were mildly iron depleted by being placed on a diet with less than 4 ppm iron (Harlan Teklad) for seven days prior to the experiment. On the day of injection, mice received a series of three intraperitoneal (IP) injections (spaced six hours apart) of 50 μg human EGF (Millipore Corporation, Billerica, MA) or saline. Some mice were injected with 5 mg holotransferrin (Sanquin, Amsterdam, The Netherlands) with the third dose of EGF or saline. Liver samples were collected 24 hours after the first dose of EGF was administered.
HGF dose-dependently suppressed hepcidin mRNA in hepatocytes (Figure 1, A). When hepcidin was induced by its physiological stimuli, holotransferrin or BMP, HGF significantly lowered both baseline hepcidin expression and the maximal induction of hepcidin by holotransferrin (Figure 1, B) or by a range of BMP6 concentrations (Figure 2, A). At each concentration of BMP6, HGF addition caused 10- to 20-fold suppression of hepcidin mRNA. In experiments where IL-6 was used as the inducing cytokine, HGF suppression of hepcidin mRNA was overcome by increasing concentrations of IL-6, even though it is a less potent hepcidin inducer than BMP6. (Supplementary Figure S1). Among other growth factors tested, only EGF suppressed hepcidin mRNA similarly to HGF (Figure 2, B). PDGF (Figure 2, C) and IGFs −1 and −2 (Supplementary Figure S2) had no effect on BMP induction of hepcidin mRNA.
To test whether the suppressive effect of growth factors could be relevant in vivo, we injected mice with EGF, holotransferrin, or their combination (Figure 3), using recombinant human EGF because of much lower cost. Increased expression of the known EGF target transcript osteopontin confirmed that EGF had a detectable effect in the liver (Figure 3, B). EGF significantly suppressed hepcidin responses to holotransferrin (Figure 3, A), with hepcidin mRNA approximately 20-fold lower than in mice that received holotransferrin alone.
In primary mouse hepatocytes transfected with hepcidin promoter-luciferase reporter, HGF strongly suppressed the induction of the hepcidin promoter by BMP2 (Figure 4, A). We tested a broader range of BMPs in HepG2 cells transfected with hepcidin-luciferase reporter and found that HGF suppressed the induction of the hepcidin reporter by BMP-2, 4 6 and 9) (Figure 4, C, Supplementary Figure S3A). Thus HGF is a broadly active transcriptional suppressor of the BMP response of the hepcidin promoter.
We also tested the effect of HGF on another BMP-sensitive luciferase reporter containing the BMP-responsive element (BRE) from the promoter of the gene for ID1 (inhibitor of DNA binding 1), a known direct target gene for BMP (16). In transfected mouse hepatocytes and HepG2s, HGF suppressed the induction of the BRE-luciferase reporter by BMP-2, −4, −6 and −9 (Figure 4, B and D, and Supplementary Figure S3B). Further, in primary mouse hepatocytes HGF and EGF similarly modulated the BMP-dependent induction of ID1 mRNA (Supplementary Figure S4). Taken together, these data indicate that HGF and EGF inhibit transcription of BMP-sensitive genes including hepcidin, likely by modulating BMP pathway signaling or BMP-dependent assembly of transcriptional machinery.
When BMPs bind to their receptor (BMP-R), the receptor phosphorylates and activates cytosolic signaling proteins R-Smads 1, 5 and/or 8, which form complexes with the common mediator Smad4. These complexes translocate into the nucleus where they transactivate BMP-dependent transcription (19). The induction of hepcidin mRNA by BMP6 occurred within 4h, and co-stimulation with HGF or EGF suppressed the maximal induction of hepcidin mRNA within the same time-frame (Supplementary Figure S5). The short timeframe favored a mechanism based on rapid, covalent modifications of signaling mediators rather than the synthesis of new transcriptional regulators. We hypothesized that HGF and EGF were initiating kinase signaling that resulted in decreased activation of Smad1/5/8, or in inhibitory modification of Smad1/5/8, through prevention or removal of the activating C-terminal phosphorylation (20) or by targeting Smads 1/5/8 for degradation.
BMP-dependent activating phosphorylation of Smad 1/5/8 was equal in the growth factor-treated and BMP-only series (Figure 5, A, B), indicating that despite the presence of HGF or EGF the BMP signal was still fully transduced to the R-Smads. This makes it unlikely that BMP ligand-trap proteins, modification or degradation of BMP-Rs, or R-Smad de-activation play an important role in the modulation of the BMP effect by HGF or EGF. Further, total nuclear Smad1/5/8 is not decreased by HGF treatment (Figure 5A)..
Additional modulation at the ligand-receptor level is provided by BMP pathway inhibitors including BAMBI (21), Smad 6 (22, 23) and Smad7, the latter already known to play a role in hepcidin regulation (7, 24). The concentrations of BAMBI and inhibitory Smads determine their effect on signaling (23). We used whole-cell lysates of primary mouse hepatocyte cultures treated with BMP and HGF or EGF to examine the protein levels of the three inhibitors. After overnight incubation, neither Smads 6 and 7 (Supplementary Figure S6A, B) nor BAMBI (data not shown) were induced by growth factor treatments.
Growth factors have been reported to decrease the total Smad1 pool by proteasomal degradation (25). Treatment with HGF had no effect on total Smad1 or Smad5 (Supplementary Figure S6C, D) in the 2h following HGF treatment or overnight (data not shown). Treatment with EGF also did not cause a change in total Smad 1/5/8 in whole cell lysates (Figure 5, B).
The common mediator Smad4 is also a target for regulatory input and its ubiquitination leads to its degradation in the proteasome (26). Decreased Smad4 in the context of hepcidin reporter suppression by hypoxia was recently described (27). From hepatocytes treated with BMP6 with and without HGF, we blotted nuclear lysates for Smad4. Smad4 levels in the nucleus were unaffected by HGF, indicating that the BMP signal had adequate access to co-Smad for formation of transcription complexes (Supplementary Figure S6E). Thus, the mechanism for growth factor suppression of hepcidin does not include overall degradation of the receptor-activated Smad pool or Smad4.
The linker region between the two globular domains of Smad1 can be phosphorylated by several kinases, including the mitogen-activated protein kinase (MAPK) ERK2, cyclin-dependent kinases (CDK) and glycogen synthase kinase-3β (GSK3β) (25) and the modification inhibits nuclear translocation of Smads. Growth factors including HGF and EGF induce linker phosphorylation (28) acting to limit BMP signaling during development.
After growth factor treatment of BMP6-stimulated hepatocytes, immunoblots for phospho-Smad1/5/8 showed moderately decreased nuclear localization of phospho-Smad/1/5/8 (Figure 6). The difference between the growth-factor treated nuclear lysates and the control lysates was statistically significant for both growth factors by pairwise t-test when four repeated experiments were analyzed together for each growth factor.
We next considered modes of BMP pathway suppression that target the Smad transcriptional complex. The Smad transcriptional complex is nucleated by R-Smad/Smad4, but the binding affinity is regulated by DNA-binding co-activators or co-repressors such as TGIF (29). Figure 7 demonstrates increased protein levels of TGIF in nuclear lysates of cells treated with HGF, suggesting a mechanism for HGF interference with Smad transcriptional complexes.
Taken together, these data indicate that the mechanism for HGF suppression is downstream of the multiple levels of Smad regulation and may involve a combination of decreased nuclear localization of activated Smad1/5/8 as well as induction of transcriptional co-repressors such as TGIF.
HGF activates signaling pathways through its receptor, tyrosine kinase Met. Met signaling is complex, branching into multiple distinct but interacting signaling modules, so that HGF suppression of BMP signaling to hepcidin may be the product of more than one downstream signal from HGF/Met (Supplementary Figure S7). Using primary hepatocytes treated with BMP6, we performed a limited screen with small-molecule kinase inhibitors against individual kinase pathways known to be activated by HGF.
The proof of principle experiment tested for inhibition of HGF signaling by a kinase inhibitor for the Met receptor itself (PHA665752). Inhibition of the Met receptor abrogated HGF suppression of both hepcidin mRNA and ID1 mRNA (Figure 8, A, B). Interestingly, the dose required to inhibit HGF (1 μM) was 20-times the IC50 (25–50 nM) for inhibition of receptor activation in epithelial cell lines (kidney, lung and mammary cells). The requirement for high doses of inhibitor may be due to the hepatocyte cell membrane resistance to permeation of small molecule kinase inhibitors, akin to difficulties with the transfection of primary hepatocyte using liposomal methods. Alternatively, the known catabolic activity of hepatocytes towards small organic molecules may cause rapid degradation of many of our inhibitors. Bearing this in mind, we examined a higher range of inhibitor concentrations.
Two MAPK pathways are known to be activated by HGF: Rac1/JNK and Ras/MEK/ERK. Two Rac1 inhibitors (5 μM EHT1864, 94 μM NSC23766) did not inhibit HGF suppression of hepcidin mRNA, nor did JNK inhibition (1 μM, JNK Inhibitor II). With MEK1/2 inhibitor U0126, we observed partial reversal of HGF suppression of hepcidin mRNA (Figure 8, C) as well as ID1 mRNA (Figure 8, D). The ERK inhibitor peptide II (5 μM) recapitulated these data (data not shown). The high dose of U0126 (25 μM) reversed HGF suppression of hepcidin and ID1, but it also affected the baseline hepcidin and ID1 mRNA, indicating that the activity of the inhibitor at 25 μM may have effects not specific to HGF. These data indicate at most a partial role for HGF/MEK signaling to hepcidin; alternatively, inhibition of MEK1/2 may result in mild hepcidin increase by mechanisms independent of HGF.
Broadening our focus, we sought to rule out other major pathways downstream of the Met receptor (Supplementary Figure S7). Small-molecule inhibitors of protein kinase C (1.25 μM Calphostin), phospholipase C (5 μM U73112) or STAT3 (1 μM inhibitor VII) neither affected BMP induction of hepcidin, nor did they reverse suppression of hepcidin by HGF. We conclude that none of these major pathways plays a role in the regulation of hepcidin by HGF.
Treatment of primary mouse hepatocytes with PI3K inhibitor LY294002 at a moderate concentration (8 μM, 5X IC50) significantly reversed HGF suppression of hepcidin (p = 0.04, t-test compared to controls) (Figure 8, E) without affecting baseline hepcidin mRNA in the controls or maximal hepcidin induction by BMP6. ID1 suppression was similarly reversed (Figure 8, F). Increased phosphorylation of AKT confirmed activation of PI3K by HGF, and loss of AKT phosphorylation confirmed the effectiveness of the PI3K inhibitor (Supplementary Figure 8). Pre-treatment with the Met inhibitor also prevented AKT activation (Supplementary Figure S9A). In agreement with hepcidin mRNA suppression in primary hepatocytes, only HGF and EGF, but not PDGF, IGF-1 and IGF-2, caused activation of AKT (Supplementary Figure S9B).
We report the growth factors HGF and EGF as a new category of hepcidin suppressors that robustly block hepcidin transcriptional regulation by known physiologic inducers, iron and BMPs. The ability of EGF to suppress iron-induced hepcidin mRNA was also confirmed in mice. Our data also indicate that HGF and EGF regulate hepcidin by suppressing BMP signaling upstream of the hepcidin promoter, a suppressive effect that extends to the unrelated BMP-sensitive promoter and mRNA transcript of ID1 The rapid onset of suppression suggests direct molecular crosstalk between BMP and growth factor signaling mediators. The crosstalk does not extend to the IL-6 pathway as HGF does not significantly suppress hepcidin mRNA at higher IL-6 concentrations.
Growth factor regulation of BMP signaling through MAPK-mediated nuclear exclusion of R-Smads has been extensively reported in culture systems using transfected, highly overexpressed tagged Smad constructs. The data from such studies suggest that Smad linker phosphorylation by growth factor-activated MAPK/ERK nearly entirely abrogates BMP-dependent nuclear localization of activated Smads (25). Our findings indicate that in hepatocytes the endogenous R-Smad pool is less strictly regulated. The trend we observed for nuclear exclusion of activated Smads and the increased regulatory phosphorylation at MAPK motifs on the Smad linker is modest at best and seems unlikely to account for the dramatic inhibition of hepcidin induction by HGF and EGF. Furthermore, the activation of the R-Smads was not suppressed by the growth factors, nor was the cellular pool of Smads1 and 5 and co-Smad4 degraded. We also detected no evidence of transcriptional induction of BMP negative regulators such as inhibitory Smads.
R-Smad by itself interacts weakly with its cognate promoter element and its association with other transcription factors is thought to be required for optimal activity. The increased protein levels of a transcriptional co-repressor, TGIF, in the nuclei of hepatocytes treated with HGF suggests a likely mechanism for HGF crosstalk with BMP signaling. Phosphorylation of the co-repressor TGIF by EGF-activated Ras/MEK signaling has been reported; TGIF phosphorylation resulted in stabilization of the repressor and formation of R-Smad/TGIF transcriptionally suppressive complexes (30). We surmise that HGF may suppress hepcidin induction by BMP through MAPK stabilization of TGIF.
HGF is a pleiotropic growth factor that activates a multitude of downstream signaling pathways; many of the mitogenic, morphogenic and motogenic effects of Met are regulated by more than one of these downstream signals. Our kinase inhibitor screen in primary hepatocytes identified at least two signaling pathways (MEK and PI3K) that appear to regulate hepcidin. The activity of the MEK1/2 inhibitor U0126 in our studies suggested a role for MEK in HGF suppression. It was previously reported that Ras/MEK activation by EGF results in phosphorylation and stabilization of the Smad transcriptional co-repressor TGIF (30). HGF may cause a similar stabilization of TGIF via MEK activation. A more detailed exploration of the similarities and differences between HGF and EGF pathways will be undertaken in a future study.
In view of the role of growth factors HGF, EGF, and TGF-α, which also binds to the EGF receptor, as mediators of the hepatic regenerative response (14), the suppression of hepcidin by growth factors may be relevant to hepcidin deficiency and hepatic iron loading in chronic liver diseases. Elevated liver tissue concentrations of growth factors in chronic viral and alcoholic hepatitis could be repressing maximal hepcidin response to iron thereby increasing dietary iron absorption and worsening the liver injury. As in hereditary hemochromatosis, the relative lack of hepcidin induction by iron in chronic hepatitis results in chronic hyperabsorption of dietary iron. Excess iron accumulates particularly in the liver due to the avid uptake of non-transferrin-bound iron (NTBI) by hepatocytes, as well as the first-pass effect of portal circulation from the gut. The iron deposition is often parenchymal and compounds preexisting liver injury from hepatitis, worsening disease prognosis. In CHC, iron correlates with development of cirrhosis and hepatocellular carcinoma (HCC) (11). The role of iron in disease progression has been supported by studies in which phlebotomy improved disease indices in nonalcoholic steatohepatitis and chronic hepatitis C (31, 32). However, the effects of iron on hepatitis C may be complex; excess iron promotes tissue damage but it also suppresses viral replication, perhaps accounting for the divergent outcomes of phlebotomy interventions (33).
Regulation of hepcidin by growth factors may be important for normal iron homeostasis as well. Hepcidin must be physiologically suppressed during early years of life, when continuing growth and development require greater iron absorption than in the mature adult (34). Though few studies exist of hepcidin levels in children and adolescents, a recent study in children and adult patients undergoing hemodialysis found that the pediatric control group had serum hepcidin concentrations that were only a third as high as the adult control group (35). Interestingly, hepatic hepcidin mRNA is not detectable by Northern blot in mice from E15.5 to postnatal day 56 apart from a transient induction at birth extending to postnatal day 2 (36). Hepcidin expression, therefore, only reaches a high level in the adult mouse liver, concordant with the human studies suggesting that hepcidin is repressed during early growth and maturation.
Finally, better understanding of pathways that mediate hepcidin suppression may help identify useful targets for new treatments for iron restrictive disorders (anemia of inflammation, anemia of chronic kidney disease) in which hepcidin excess contributes to the pathogenesis of anemia and to erythropoietin resistance.
We are grateful for the excellent technical assistance of Victoria Gabayan with all the mouse studies in this manuscript.
This work was supported by Roche Foundation for Anemia Research (RoFAR) to T.G., National Institutes of Health grants R01 DK065029 (to T.G.), R01 DK082717 (to E.N.), the Will Rogers Fund (to T.G.), Ruth L. Kirschstein National Research Service Award (NRSA) F30 DK082151 (to J.G.) and Ruth L. Kirschstein NRSA GM007185 (to E.R.).