Gluconeogenic signals regulate iron homeostasis via hepcidin in mice.

BACKGROUND & AIMS: Hepatic gluconeogenesis provides fuel during starvation, and is abnormally induced in obese individuals or those with diabetes. Common metabolic disorders associated with active gluconeogenesis and insulin resistance (obesity, metabolic syndrome, diabetes, and nonalcoholic fatty liver disease) have been associated with alterations in iron homeostasis that disrupt insulin sensitivity and promote disease progression. We investigated whether gluconeogenic signals directly control Hepcidin, an important regulator of iron homeostasis, in starving mice (a model of persistently activated gluconeogenesis and insulin resistance).
METHODS: We investigated hepatic regulation of Hepcidin expression in C57BL/6Crl, 129S2/SvPas, BALB/c, and Creb3l3-/- null mice. Mice were fed a standard, iron-balanced chow diet or an iron-deficient diet for 9 days before death, or for 7 days before a 24- to 48-hour starvation period; liver and spleen tissues then were collected and analyzed by quantitative reverse-transcription polymerase chain reaction and immunoblot analyses. Serum levels of iron, hemoglobin, Hepcidin, and glucose also were measured. We analyzed human hepatoma (HepG2) cells and mouse primary hepatocytes to study transcriptional control of Hamp (the gene that encodes Hepcidin) in response to gluconeogenic stimuli using small interfering RNA, luciferase promoter, and chromatin immunoprecipitation analyses.
RESULTS: Starvation led to increased transcription of the gene that encodes phosphoenolpyruvate carboxykinase 1 (a protein involved in gluconeogenesis) in livers of mice, increased levels of Hepcidin, and degradation of Ferroportin, compared with nonstarved mice. These changes resulted in hypoferremia and iron retention in liver tissue. Livers of starved mice also had increased levels of Ppargc1a mRNA and Creb3l3 mRNA, which encode a transcriptional co-activator involved in energy metabolism and a liver-specific transcription factor, respectively. Glucagon and a cyclic adenosine monophosphate analog increased promoter activity and transcription of Hamp in cultured liver cells; levels of Hamp were reduced after administration of small interfering RNAs against Ppargc1a and Creb3l3. PPARGC1A and CREB3L3 bound the Hamp promoter to activate its transcription in response to a cyclic adenosine monophosphate analog. Creb3l3-/- mice did not up-regulate Hamp or become hypoferremic during starvation.
CONCLUSIONS: We identified a link between glucose and iron homeostasis, showing that Hepcidin is a gluconeogenic sensor in mice during starvation. This response is involved in hepatic metabolic adaptation to increased energy demands; it preserves tissue iron for vital activities during food withdrawal, but can cause excessive iron retention and hypoferremia in disorders with persistently activated gluconeogenesis and insulin resistance.

Adaptation to different states, such as exercise, rest, and starvation or overnutrition, is essential for life. In turn, dysfunction and perturbation of these networks can lead to metabolic imbalances, which if uncorrected induce diseases such as obesity or diabetes. Metabolic adaptation is largely controlled by transcriptional co-regulators and transcription factors responsible, respectively, for sensing metabolic disturbances and fine-tuning the transcriptional response. During starvation, this adaptive response is essential for species survival, and the liver plays a central role in this process as a main site for gluconeogenesis and energy production. At early stages, the liver mobilizes glucose from its glycogen stores; as fasting progresses, it oxidizes fat to provide both energy for gluconeogenesis and substrate for ketogenesis. Generation of sugar from nonsugar carbon substrates (gluconeogenesis) involves several enzyme-catalyzed reactions that take place in both cytosol and mitochondria.

Iron is essential for vital redox activities in the cell, in particular it is required for respiration and energy production in mitochondria (which are also the unique site for heme synthesis and the major site for Fe-S cluster biosynthesis), and likewise is important for mitochondria biogenesis. A number of iron abnormalities, ranging from low serum iron/iron-restricted anemia to hepatic/systemic iron overload, have been reported in human disorders with activated gluconeogenic signaling pathways, including obesity, metabolic syndrome,5, 6, 7 and diabetes.8, 9 Interestingly, iron excess has been associated with worsened insulin sensitivity and disease progression, whereas iron removal has been found to be beneficial.6, 8, 10 Based on these premises, we asked whether iron status could be regulated directly by gluconeogenic signals.

Systemic and local iron status are under the control of hepcidin, a defensin-like circulating peptide that degrades the iron exporter ferroportin (FPN1), thereby dictating the extent of iron release or retention in the cell. Hepcidin expression is transcriptionally controlled by a number of factors that deliver the relevant stimulatory or inhibitory signals to the nuclear machinery and turn on or off the hepcidin (HAMP) gene. The main stimulatory transcription factors include small mother against decapentaplegic (SMAD) proteins, which bind the bone morphogenetic protein responsive element and deliver the “iron signal,”12, 13 STAT-3, mainly involved in the inflammatory signal,14, 15, 16 and cyclic adenosine monophosphate (cAMP) response element binding protein 3–like 3, CREB3L3 (also known as CREBH), more recently found to mediate hepcidin induction by endoplasmic reticulum (ER) stress triggered by a variety of physiological and pathophysiological states.18, 19, 20

Therefore, we focused on investigating the regulation of hepcidin expression in the liver in response to gluconeogenic stimuli. To this end, we studied mice undergoing prolonged starvation, a classic model of persistently activated gluconeogenesis and insulin resistance.

Materials and Methods

Animal Studies

The starvation experiment was as follows: 8- to 10-week-old male C57BL/6Crl, 129S2/SvPas, BALB/c wild-type mice, and Creb3l3-/- null mice (The Jackson Laboratory, Bar Harbor, ME) were allowed free access to water and fed a standard, iron-balanced chow diet in pellets (2018S Teklad Global 18% Protein Rodent Diet; Harlan Laboratories, (San Pietro Al Natisone, UD, Italy); iron content, 225 mg/kg) or starved up to 48 hours starting at the beginning of the light cycle.

Iron-deficient diet experiments were as follows: 8-week-old male C57BL/6Crl wild-type mice were fed an iron-deficient diet (ssniff EF R/M Iron Deficient; Charles River, Calco, LC, Italy; iron content, <10 mg/kg) for 9 days before death, or for 6 days before the 24- to 48-hour starvation period.

All animals received humane care according to the criteria outlined by the Federation of European Laboratory Animal Science Associations. The study was approved by the Ethics Committee for Animal Studies at the University of Modena and Reggio Emilia.

Blood Measurements and Tissue Iron Content

Serum iron, serum ferritin (Tina-quant Ferritin kit; Roche Diagnostics, Milan, Italy), hemoglobin, and glucose were determined using an automated COBAS C501 counter (Roche, Milan, Italy) at the clinical-chemical laboratory of the University Hospital of Modena. Serum hepcidin was determined using an enzyme-linked immunosorbent assay kit (USCN Life Science, Hubei, China) according to the manufacturer's instructions, as previously reported. Serum ketone bodies were analyzed using a β-Hydroxybutyrate Assay Kit (Sigma-Aldrich, Milan, Italy) following the manufacturer's instructions.

Liver and spleen tissue specimens were analyzed for non-heme iron content as previously reported.

Real-Time Quantitative Reverse-Transcription Polymerase Chain Reaction and Semiquantitative Reverse-Transcription Polymerase Chain Reaction

Total cellular RNA was obtained by incubating cells in iScript quantitative reverse-transcription polymerase chain reaction (qRT-PCR) Sample Preparation Reagent (Bio-Rad, Milan, Italy) according to the manufacturer's instructions. Total hepatic RNA was extracted as described. Complementary DNA was generated by reverse transcription of 2 μL of iScript buffer (for cultured cells) or 1 μg (for liver) with 200 U ImProm-II Reverse Transcriptase (Promega, Milan, Italy) following the manufacturer's instructions. Expression of mRNA was analyzed using SsoFast EvaGreen Supermix (Bio-Rad). Primer sequences are listed in Supplementary Table 1. Cycling conditions were as follows: 30 seconds at 98°C, followed by 40 cycles of 2 seconds at 98°C and 10 seconds at 60°C. After 40 amplification cycles, threshold cycle values were calculated automatically using the default settings of the CFX Manager software (version 2.0; Bio-Rad), and femtograms of starting complementary DNA were calculated from a standard curve covering a range of 5 orders of magnitude. At the end of the PCR run, melting curves of the amplified products were obtained and used to determine the specificity of the amplification reaction. In each experiment, the change of specific mRNA expression was reported as the fold increase as compared with that of control cells or mice. Normalization of qRT-PCR data was based on RPL19 housekeeping mRNA expression after validation using the target stability value obtained from the CFX Manager software (version 2.0; Bio-Rad). X-box binding protein 1 (Xbp1) splicing was analyzed as described by Vecchi et al. Primer sequences are listed in Supplementary Table 1. The Hamp oligos detects total Hamp mRNA (Hamp1 and Hamp2 mRNA).

Western Blot Analyses

For the FPN1 assay, mouse liver specimens were homogenized in lysis buffer (150 mmol/L NaCl, 10 mmol/L Tris, pH = 8, 1 mmol/L EDTA, 0.5% Triton X-100) containing 1:100 protease inhibitor cocktail (Sigma-Aldrich). After centrifugation at 13,000 × g at 4°C for 15 minutes, the supernatant was collected and the protein concentration was assayed by the Bradford method. A total of 60 μg of liver extracts were loaded without boiling on 10% acrylamide gels with Laemmli sample buffer, and run in sodium dodecyl sulfate–polyacrylamide gel electrophoresis buffer.

Membranes were probed with specific antibodies: rabbit anti-FPN1 (1:1000; Alpha Diagnostic, Inc, San Antonio, TX), as previously reported, and mouse anti-tubulin (1:3000; Sigma-Aldrich), followed by appropriate horseradish-peroxidase–conjugated secondary antibodies. Western blot analysis was performed by Western Lightning Ultra substrate (PerkinElmer, Waltham, MA) according to the manufacturer's instructions. Chemiluminescence was detected and quantified using the Molecular Imager ChemiDoc XRS+ with Image Lab Software (Bio-Rad).

Cell Cultures and Primary Hepatocyte Isolation

Human hepatoma HepG2 cells were cultured in Modified Eagle's Medium (MEM) (containing 1 g/L glucose), supplemented with 1 mmol/L glutamine, 100 U/mL penicillin, 100 μg/mL streptomycin, and 10% heat-inactivated fetal bovine serum, in a 5% CO2 atmosphere at 37°C.

Mouse primary hepatocytes from 8- to 10-week-old male C57BL/6Crl mice were isolated as previously described.

HepG2 cells and mouse primary hepatocytes were incubated for 8 hours in the presence of 1 mmol/L of 8Br cAMP (Sigma-Aldrich) or for 6 hours in the presence of 100 nmol/L of glucagon (Sigma-Aldrich), both in 2% fetal bovine serum culture medium.

Plasmids, Small Interfering RNAs, Transfection, and Luciferase Assay

Hepcidin promoter construct, plasmid encoding Flag-tagged CREB3L3-N (the active form of the factor), CREB3L3 small interfering RNA (siRNA) transfection, and luciferase analysis have been reported elsewhere. Plasmid encoding peroxisome proliferator-activated receptor gamma coactivator 1-α (PPARGC1A) was kindly provided by Dr Chang Liu (Nanjing, China). PPARGC1A siRNA were obtained from Invitrogen (Life Technologies Italia, Monza, Italy) (PPARGC1AHSS116799).

Chromatin Immunoprecipitation

Chromatin immunoprecipitation (ChIP) was described elsewhere with the following modifications. Briefly, HepG2 cells were transfected using X-tremeGENE transfection reagent (Roche Applied Science, Milan, Italy) with plasmid encoding Flag-tagged CREB3L3-N. Forty-eight hours after transfection, cells were treated with 1 mmol/L 8Br cAMP for 8 hours and fixed for formaldehyde cross-linking and ChIP. Protein–DNA complexes were immunoprecipitated overnight using the following antibodies: anti-Flag (Sigma-Aldrich), anti-PPARGC1A (anti-PGC1A; Santa Cruz Biotechnology, Dallas, TX), or anti-green fluorescent protein (GFP) (Abcam, Cambridge, UK) as negative control.

Statistical Analyses

All data were controlled for normal distribution (Kolmogorov–Smirnov and Shapiro–Wilk tests). When comparing a variable in 2 groups, a paired t test or the Wilcoxon–Mann–Whitney test was used, depending on the presence or absence of normal data distribution and/or small sample size. When making multiple statistical comparisons on a single data set, for normally distributed data a 1-way analysis of variance with the Tukey or Dunnett post hoc tests, depending on the presence or absence of homoscedasticity, was used. For skewed data, the Kruskal–Wallis test was used. In all statistical analyses, a P value less than .05 was considered significant. Data presented in Figures are mean ± SEM. All analyses were conducted using Prism 5 for mac OS X version 5.0a software (GraphPad Software, Inc, La Jolla, CA).

Results

In starving mice, phosphoenolpyruvate carboxykinase 1 (Pck1) mRNA, known to be readily responsive to gluconeogenic stimuli, rapidly increased at 2 hours (Figure 1A), whereas Hamp mRNA increased at 5 hours, in concomitance with a marked serum glucose decrease, and remained increased for up to 48 hours (Figure 1B). In addition, serum hepcidin showed a sharp increase at 5 hours, although slightly decreased at later time points (Figure 1C). Hamp induction led to a decrease of serum iron, and a progressive increase of serum ferritin and iron content in the spleen and the liver (Table 1). In agreement with the hepcidin model of iron regulation, which implies a post-translational down-regulation of ferroportin protein by hepcidin, hepatic Fpn1 mRNA was unchanged (Figure 1D) whereas FPN1 protein was degraded in a time-dependent manner starting at 5 hours' starvation (Figure 1E and F). A visible ferroportin down-regulation in the spleen was not detected (data not shown). As discussed earlier, the main stimuli for hepcidin transcription in vivo are increased serum and hepatic iron, and cytokines produced during inflammation and infection, particularly interleukin 6 (IL6), IL22, tumor necrosis factor-α, and ER stress. In mice undergoing prolonged starvation, we were unable to detect up-regulation of cytokines such as IL6 and tumor necrosis factor-α, whereas IL22 actually was depressed by food withdrawal (Supplementary Figure 1A–C). IL1β was induced by short-term fasting but returned to normal at 48 hours (Supplementary Figure 1D), when hepcidin mRNA expression was still increased markedly. Similar negative results were found when analyzing inflammation marker C-reactive protein (Crp) mRNA (Supplementary Figure 1E) and ER stress markers (namely, Xbp1 mRNA splicing; Supplementary Figure 1F). To address whether hypoferremia in starving mice was caused by lower iron intake associated with food deprivation, we studied mice premaintained on an iron-deprived diet for 1 week. After the iron-deficient diet, this group of mice showed normal serum iron levels (Figure 2A), but almost halved spleen iron stores compared with fed mice maintained on an iron-balanced diet (Figure 2B), suggesting a marked iron redistribution from the storage site toward the bloodstream to sustain red cell production and maintain normal hemoglobin levels (Figure 2C). However, even under this circumstance, starvation led to a progressive decrease of serum iron (Figure 2A). Moreover, hepcidin mRNA expression, although depressed in control mice (iron-deficient group) likely because of the latent iron-deficiency state and active marrow activity, still dramatically was induced by starvation (Figure 2D). Activation of hepcidin and perturbation of iron homeostasis during starvation-induced gluconeogenesis also was found in other tested mouse strains, such as BALB/c (Supplementary Figure 2A–C) or 129S2 (Supplemental Figure 2D–F). Overall, these data suggested that, in starving mice, stimuli that are independent of inflammation and/or stress may be responsible for hepcidin induction.

Figure 1

In vivo time course of hepcidin and ferroportin expression in mice during starvation. (A) Real-time qRT-PCR analysis of Pck1 mRNA and (B) Hamp mRNA expression relative to housekeeping Rpl19 mRNA in C57BL/6 mice fed a standard diet (white bar) or starved for the indicated time periods (gray bars). (C) Enzyme-linked immunosorbent assay quantification of serum hepcidin levels. (D) Fpn1 mRNA expression relative to housekeeping Rpl19 mRNA. (E) Western blot analysis of FPN1 protein expression in the liver, with tubulin as loading control. The arrow indicates the specific FPN1 band, whereas the nonspecific upper band is owing to the secondary antibody. (F) Densitometric quantification of FPN1 protein expression relative to tubulin. Results are mean ± SEM of 6–8 mice per group. In Western blot analysis, 3 representative mice per group are shown. For mRNA expression analysis, mean control values for the fed mice group are set to 1. P values are reported for comparisons between fed mice and mice fasted at each time point. *P < .05, ***P < .001.

Table 1

In Vivo Effects of Starvation on Biochemical Parameters in Mice

	Fed	2-h Fast	5-h Fast	16-h Fast	24-h Fast	48-h Fast
Serum glucose, mg/dL	249.7 ± 36.83	276.4 ± 29.31	188.0 ± 9.09a	193.2 ± 13.61a	146.7 ± 39.11b	180.7 ± 38.53b
Serum iron, μg/dL	136.3 ± 3.33	143.6 ± 2.85	138.0 ± 8.12	84.00 ± 2.96b	86.62 ± 3.06b	87.08 ± 4.41b
Serum ferritin, ng/mL	204.6 ± 8.78	190.0 ± 25.61	193.5 ± 10.78	286.5 ± 32.53	240.0 ± 12.58	277.0 ± 13.58a
Spleen iron, μg/gdry weight	776.1 ± 29.31	837.5 ± 64.33	706.7 ± 27.62	905.2 ± 13.66	852.0 ± 56.25	1418 ± 58.28b
Liver iron, μg/gdry weight	224.3 ± 7.39	226.4 ± 9.42	214.9 ± 11.12	231.3 ± 19.73	262.8 ± 16.56	367.1 ± 26.99b

NOTE. Serum glucose, serum iron, serum ferritin, and tissue iron levels were analyzed in spleen and liver tissue of wild-type starving and nonstarving mice. Results are mean ± SEM of 6–8 mice per group. P values are reported for comparisons between fed and each time-point fasted mice.

P < .01.

P < .001.

Figure 2

Fasting induces hepcidin gene expression also in mice premaintained on an iron-deficient diet. Eight- to 10-week-old male C57BL/6Crl mice were fed an iron-balanced diet or an iron-deficient diet for 9 days before death (IB and ID, respectively), or for 6 days before the 24- to 48-hour starvation period (ID fast 24-hr and 48 hr). (A) Serum iron quantification, (B) spleen iron content, (C) hemoglobin (Hb) levels, and (D) Hamp mRNA expression relative to housekeeping Rpl19 mRNA expression. Results are expressed as the mean ± SEM of 5–6 mice per group. P values are reported for comparisons between the indicated groups. *P < .05, **P < .01, ***P < .001.

To identify the molecular basis for this novel hepcidin regulatory mechanism, we used an in vitro approach.

The hepatic expression of genes encoding gluconeogenic enzymes, such as PCK1, is regulated by a network of transcription factors and cofactors, including CREB proteins28, 29 and PPARGC1A. We recently found that a member of the CREB family, CREBH, is engaged constitutively on the hepcidin promoter and readily transactivates it during ER stress. Both Ppargc1a and Creb3l3 mRNA are induced by hepatic gluconeogenesis in vivo during starvation (Figure 3A and B). We hypothesized that CREBH is a target for PPARGC1A coactivation during hepcidin induction by active gluconeogenesis. In line with this hypothesis, PPARGC1A silencing in HepG2 cells led to a 60% decrease of hepcidin mRNA expression, similar to the effect obtained by CREB3L3 knockdown (Figure 3C).

Figure 3

Ppargc1a and Creb3l3 are induced by starvation and are involved in hepcidin expression. (A) Real-time qRT-PCR analysis of Ppargc1a mRNA and (B) Creb3l3 mRNA expression in liver of C57BL/6 mice fed an iron-standard diet (white bar) and starved for the indicated time points (gray bars). (C) Basal expression of HAMP mRNA in HepG2 cells transfected with siRNAs against PPARGC1A, CREB3L3, or both. Results are mean ± SEM of 6–8 mice per group or 3–4 independent experiments each repeated in triplicate. Mean control values for the fed mice group for in vivo experiments or unspecific (US) RNA interference (RNAi) for in vitro experiments are set to 1 and are normalized relative to housekeeping Rpl19 mRNA. P values are reported for comparisons between control and treated cells, or between indicated groups. *P < .05, **P < .01, ***P < .001.

Gluconeogenesis induced by food deprivation involves cAMP as the main intracellular second messenger in response to hormonal stimuli.31, 32 HepG2 cells exposed to 8Br cAMP, a cAMP analog, showed a significant increase of both PCK1 and HAMP mRNA in a time-dependent manner (Figure 4A). A similar trend of hepcidin activation also was found in primary hepatocytes exposed to either glucagon or 8Br cAMP. Both treatments induced Pck1 and Hamp mRNA expression in cultured hepatocytes, although Hamp response was significantly but appreciably lower than in HepG2 cells (Figure 4B). Hepcidin stimulation by 8Br cAMP in HepG2 cells transfected with siRNA for either PPARGC1A or CREB3L3 was appreciably lower as compared with 8Br cAMP-treated control cells (Figure 4C). A similar effect was documented when we tested the response of Hamp promoter to 8Br cAMP in the presence of PPARGC1A or CREB3L3 siRNAs (Figure 4D). To prove that PPARGC1A cooperates with CREBH to turn on hepcidin in response to gluconeogenesis, we assessed if the coactivator PPARGC1A/CREBH transduces and binds the hepcidin promoter in response to gluconeogenic stimuli. Overexpression of PPARGC1A in HepG2 cells led to a significant transactivation of the Hamp promoter, indicating that the transcription factor is involved in hepcidin promoter regulation (Figure 4E). In a previous study we showed that CREBH constitutively occupies the HAMP promoter and transactivates it in response to ER stress. Here, the ChIP assay showed that, in addition to the known constitutive hepcidin promoter occupancy by CREBH (Figure 4F, αFlag, control cells), PPARGC1A also constitutively binds to the same region (Figure 4F, αPGC1A, control cells). In agreement with the studies reported earlier, after exposure of HepG2 cells to 8Br cAMP, more CREBH was stabilized on the HAMP promoter in the presence of stable PPARGC1A binding (Figure 4F, 8Br cAMP-treated cells).

Figure 4

Hepcidin is induced by gluconeogenic signals through PPARGC1A/CREBH. (A) HepG2 cells were cultured in the presence of a cAMP analog (8Br cAMP) and analyzed at different time points for PCK1 and HAMP mRNA expression by real-time qRT-PCR. (B) Pck1 and Hamp mRNA expression in primary mouse hepatocytes isolated from C57BL/6 mice and exposed to glucagon or 8Br cAMP. (C) HAMP mRNA expression and (D) Hamp-promoter luciferase activity in HepG2 cells after silencing of PPARGC1A and CREB3L3. (E) Hamp-promoter luciferase activity in HepG2 cells transfected with control plasmid (empty) or construct encoding PPARGC1A protein. (F) ChIP assay of HepG2 cells transfected with Flag-tagged CREB3L3-N vector and exposed to 8Br cAMP. CREBH (αFlag) and PPARGC1A (αPGC1A) occupancy of CREBH site on hepcidin endogenous promoter was evaluated by real-time qRT-PCR and expressed as a percentage of the input signal. αGFP (green fluorescent protein) antibody is used as control, unrelated antibody. Results are mean ± SEM of 3–4 independent experiments, each repeated in triplicate. For mRNA and luciferase analysis, mean control values are set to 1. ChIP data are mean ± SEM representative of 2 separate experiments. P values are reported for comparisons (A and B) between control and treated cells, (C–E) between indicated groups, or between αGFP and specific antibodies. *P < .05, **P < .01, ***P < .001.

In Creb3l3 null mice, in agreement with the in vitro studies, starvation correctly induced Pck1 mRNA (Figure 5A), but was unable to activate hepcidin mRNA (Figure 5B), modify serum hepcidin levels (Figure 5C), or cause hypoferremia (Figure 5D). Of note, Ppargc1a mRNA was still induced by starvation (Figure 5E), but it apparently was unable to stimulate hepcidin expression in the absence of CREBH. These data support a role for CREBH in hepcidin activation by gluconeogenic stimuli in the liver. Interestingly, serum glucose levels were significantly lower in starving Creb3l3 null mice as compared with starving wild-type mice (Table 2). Seemingly, the increase of serum ketone bodies during starvation was more pronounced in the Creb3l3 null mice (4.7- to 5.6-fold) as compared with control mice (3.1- to 3.5-fold) (Table 2).

Figure 5

Starvation fails to induce hepcidin gene expression in Creb3l3-/- mice. Eight- to 10-week-old wild type (WT) or Creb3l3-/- male mice were starved for 24 or 48 hours before death. (A) Pck1 mRNA and (B) Hamp mRNA expression were assessed by real-time qRT-PCR. (C) Enzyme-linked immunosorbent assay quantification of serum hepcidin levels and (D) serum iron levels. (E) Ppargc1a mRNA expression in starved mice. Results are mean ± SEM of 6–8 mice per group. (A, B, and E) Mean control values for the fed mice group are set to 1 and are normalized relative to housekeeping Rpl19 mRNA. P values are reported for comparisons between fed and 24- or 48-hour fasted mice, within each genotype. *P < .05, **P < .01, ***P < .001.

Table 2

In Vivo Effects of Starvation on Glucose and Ketone Body Status in Creb3l3 Null Mice

	WT			Creb3l3 -/-
	Fed	24-h Fast	48-h Fast	Fed	24-h Fast	48-h Fast
Serum glucose, mg/dL	243.7 ± 8.32	146.2 ± 26.91a	102.2 ± 12.78a	237.6 ± 8.26	95.2 ± 9.24a	64.8 ± 9.41a
Serum ketone bodies, mmol/L	0.54 ± 0.13	1.68 ± 0.10a	1.88 ± 0.24a	0.39 ± 0.02	1.85 ± 0.17a	2.20 ± 0.17a

NOTE. Serum glucose and ketone bodies levels were analyzed in WT and Creb3l3 null mice starved for 24 or 48 hours before sacrifice. Results are mean ± SEM of 6–8 mice per group. P value are reported for comparisons between fed and 24- or 48-hour fasted mice, within each genotype.

P < .001.

Discussion

Hepcidin is constitutively produced by the liver to maintain plasma iron levels within a narrow physiologic range. To do so it senses a variety of physiologic and pathophysiologic stimuli that tend to alter blood iron levels, and responds by inhibiting ferroportin, the main iron-exporter in mammals. In this study we showed that hepcidin is regulated transcriptionally also by gluconeogenic signals through PPARGC1A/CREBH. Induction of this regulatory pathway in a classic model of insulin resistance/activated gluconeogenesis, ie, starvation, leads to tissue iron retention and circulatory iron deficiency. Hypoferremia is clearly secondary to increased tissue iron retention after hepcidin induction and not to reduced food iron intake because it still is preserved in mice premaintained on an iron-deprived diet (Figure 2). Activation of hepcidin and perturbation of iron homeostasis during starvation-induced gluconeogenesis seem to represent a general defensive response in rodents because it was found in other tested mouse strains. However, differences in terms of the time course of hepcidin induction and the extent of iron status modifications were detected clearly among various starving mice strains. This could be explained by the fact that both the gluconeogenic response/gluconeogenic gene expression and iron status/iron gene expression may vary appreciably among mouse strains, as also documented by the significantly higher expression of the Pck1 gene in C57BL/6 mice (an optimal mouse model for studying gluconeogenesis/insulin resistance34, 35 and the model that most closely parallels the gluconeogenic response to starvation seen in human beings) as compared with 129S2, BALB/c, and Creb3l3 null mice (which actually display a mixed genetic background of 129S1, 129X1, C57BL/6, FVB/N). A close look at the time course induction of Pck1/Hamp (Figure 1A and B) and Ppargc1a/Creb3l3 RNAs (Figure 3A and B) suggests that the initial 5-hour burst of Hamp transcription largely depends on increased Creb3l3 expression. Later, the increase in Ppargc1A expression likely sustains hepcidin transcription by enhancing and further stabilizing CREBH binding on the Hamp promoter (Figure 4F, ChIP study). We were able to reproduce the effect of starvation in vitro, in a hepatoma cell line and cultured primary hepatocytes, using different gluconeogenic stimuli (Figure 4). However, the Hamp gene response to gluconeogenic signals in primary hepatocytes was lower than in hepatoma cells. This may depend on the fact that in primary hepatocytes the gluconeogenic signals may be attenuated, as indicated by the lower Pck1 induction in primary hepatocytes exposed to 8Br cAMP as compared with HepG2 cells (Figure 4B vs A), and/or that additional factors essential for the hepcidin transcriptional machinery are lost in primary hepatocytes after the disruption of liver architecture/microenvironment.

The newly identified regulatory pathway links glucose and iron metabolism in the liver and identifies hepcidin, the iron hormone, as a gluconeogenic sensor.

PPARGC1A is a transcriptional coactivator that regulates the genes involved in energy metabolism. During starvation, PPARGC1A readily is activated to turn on the gluconeogenic machinery, but also to stimulate mitochondrial biogenesis and respiration, which are essential to support the increased energy demands. Interestingly, in osteoclasts, mitochondrial biogenesis involves CREB/PPARGC1A proteins, but requires iron uptake and supply to mitochondrial respiratory proteins. Here, we found that PPARGC1A constitutively occupies the hepcidin promoter and, in response to gluconeogenic stimuli, stabilizes CREBH binding and transactivates HAMP promoter. CREBH is an ER stress–associated liver-specific transcription factor originally involved in the induction of acute-phase response genes (such as serum amyloid protein and C-reactive protein), and subsequently has been found to activate the transcription of HAMP. Based on recent publications and this report, CREBH now emerges as a key metabolic regulator in the liver: it is activated by fatty acids and PPARα,39, 40 and regulates the expression of genes involved in hepatic lipogenesis, fatty acid oxidation, and lipolysis under metabolic stress. Interestingly, CREBH also has been found to transcriptionally regulate Pck1 and glucose-6-phosphatase, the critical genes in hepatic gluconeogenic response. Here, we report that CREBH is engaged constitutively on the hepcidin promoter to sense metabolic gluconeogenic stress and modify, accordingly, iron traffic. Of note is that starving Creb3l3 null mice show reduced glucose and increased ketone body output.

Adaptation to starvation is essential for species survival. Seemingly, defense against pathogens represents a priority in species evolution. The liver, as the main source for hepcidin, seems to play a central role in both processes. During infection, hepcidin limits vital iron that is needed by invading microorganisms, thus contributing to host defense. During prolonged starvation, hepcidin likely preserves tissue iron and helps to maintain energy balance and support gluconeogenesis in the liver (this report). Most likely, this response originally evolved to protect human beings during food withdrawal.

Paradoxically, in human disorders associated with food excess and storage, such as type 2 diabetes, obesity, and the metabolic syndrome, persistently activated gluconeogenesis may result in overstimulation of hepcidin, iron accumulation, and potential damage. Cases of unexplained hepatic iron excess, characterized by high serum ferritin levels with normal or subnormal transferrin saturation, and associated with metabolic abnormalities, originally were reported by Moirand et al, who also introduced the term iron overload-associated insulin resistance (recently renamed dysmetabolic iron overload syndrome). Hepatocellular and/or mesenchymal iron deposition, usually slight or mild, has been reported since then in nonalcoholic fatty liver disease (NAFLD) and nonalcoholic steatohepatitis. The clinical relevance of iron excess in these disorders, in terms of fibrosis development and cancer risk, is actively debated, but increasing data indicate that iron may sustain disease activity and/or contribute to its progression.46, 47, 48, 49 Interestingly, NAFLD patients with mixed or mesenchymal iron overload (a pattern of iron deposition consistent with a “hepcidin-excess model”) seem more likely to develop fibrosis than those with pure parenchymal iron deposits (a pattern of iron deposition consistent with a “hepcidin-deficient model”).47, 49 The mechanism of iron deposition in NAFLD/dysmetabolic iron overload syndrome likely is multifactorial: sex, diet, disease activity, genetic background (HFE hemochromatosis gene mutations), ethnicity, and (micro)inflammation all may account for the variability of both iron excess and its pattern of distribution. We hypothesize that a fraction of dysmetabolic/NAFLD patients with normal-low transferrin saturation and mixed/mesenchymal hepatic iron deposits may represent a subgroup of patients with prominent insulin resistance and hepcidin induction via the gluconeogenic PPARGC1A/CREBH-driven pathway described here. In these patients, hepcidin, depending on the degree and duration of its induction, may modify iron traffic locally or systemically and lead, respectively, to simple hepatic iron retention with marginal systemic reflections (ie, mesenchymal/mixed hepatic iron accumulation with normal or subnormal transferrin saturation), or substantial tissue iron retention, hypoferremia, and iron-restricted anemia. Further studies are needed to prove that the gluconeogenic signal-driven induction of hepcidin in starving mice also takes place in other instances of activated gluconeogenesis and insulin resistance, such as diabetes, obesity, or NAFLD. If so, because of the increasingly recognized negative effect of iron excess on the progression of these disorders, the novel regulatory pathway reported here may offer potential new therapeutic targets to prevent or correct iron disturbances in common metabolic disorders.

Supplementary Table 1

List of Primers Used for Real-Time qRT-PCR

Accession number	Gene name	Forward	Reverse
Murine oligonucleotides
NM_032541.1	Hamp	5′-GCCTGTCTCCTGCTTCTCCT-3′	5′-GCTCTGTAGTCTGTCTCATCTGTT-3′
NM_011044.2	Pck1	5′-AACTGTTGGCTGGCTCTC-3′	5′-GAACCTGGCGTTGAATGC-3′
NM_145365.3	Creb3l3	5′-GATACCCTGTACCCGGAGGAG-3′	5′-CGGACAGCAGCAGTTCCTTC-3′
NM_008904.2	Ppargc1a	5′-CCGTAAATCTGCGGGATGATG-3′	5′-CAGTTTCGTTCGACCTGCGTAA-3′
NM_031168.1	Il6	5′-ATGGATGCTACCAAACTGGAT-3′	5′-TGAAGGACTCTGGCTTTGTCT-3′
NM_008361.3	Il1b	5′-CCTTTTCGTGAATGAGCAGACAG-3′	5′-TCTCTTTGAACAGAATGTGCCATG-3′
NM_016971.2	Il22	5′-ATGAGTTTTTCCCTTATGGGGAC-3′	5′-GCTGGAAGTTGGACACCTCAA-3′
NM_013693.2	Tnf	5′-ATGGCCTCCCTCTCATCAGTT-3′	5′-GGCTACAGGCTTGTCACTCG-3′
NM_007768.4	Crp	5′-CTGCACAAGGGCTACACTGT-3′	5′-TCTCCCACCAAAGACTGCTTT-3′
NM_009078.2	Rpl19	5′-ATGAGTATGCTCAGGCTACAGA-3′	5′-GCATTGGCGATTTCATTGGTC-3′
NM_016917.2	Fpn1	5′-AAGACTCCAACATCCGTGAACTT-3′	5′-GACCCATCCATCTCGGAAAGTG-3′
Human oligonucleotides
NM_021175.2	HAMP	5′-TGTTTTCCCACAACAGACGGG-3′	5′-CGCAGCAGAAAATGCAGATGG-3′
NM_000981.3	RPL19	5-′GGGCATAGGTAAGCGGAAGG-3′	5′-TCAGGTACAGGCTGTGATACA-3′
NM_032607.1	CREB3L3	5′-CCTCTGTGACCATAGACCTGG-3′	5′-ACGGTGAGATTGCATCGTGG-3′
NM_013261.3	PPARGC1A	5′-GCTTTCTGGGTGGACTCAAGT-3′	5′-TCTAGTGTCTCTGTGAGGACTG-3′
NM_002591.3	PCK1	5′-GCGGCTGAAGAAGTATGA-3′	5′-GGAACCTGGCATTGAACG-3′