Description
Please read this paper concerning how to read a science article -https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3687192/Please read the attached article and answer the questions listed in Table 4 (Questionnaire for original research articles) in the linked article.
Unformatted Attachment Preview
Adipocyte iron regulates adiponectin and
insulin sensitivity
J. Scott Gabrielsen, … , William T. Cefalu, Donald A.
McClain
J Clin Invest. 2012;122(10):3529-3540. https://doi.org/10.1172/JCI44421.
Research Article
Metabolism
Iron overload is associated with increased diabetes risk. We therefore investigated the effect
of iron on adiponectin, an insulin-sensitizing adipokine that is decreased in diabetic
patients. In humans, normal-range serum ferritin levels were inversely associated with
adiponectin, independent of inflammation. Ferritin was increased and adiponectin was
decreased in type 2 diabetic and in obese diabetic subjects compared with those in equally
obese individuals without metabolic syndrome. Mice fed a high-iron diet and cultured
adipocytes treated with iron exhibited decreased adiponectin mRNA and protein. We found
that iron negatively regulated adiponectin transcription via FOXO1-mediated repression.
Further, loss of the adipocyte iron export channel, ferroportin, in mice resulted in adipocyte
iron loading, decreased adiponectin, and insulin resistance. Conversely, organismal iron
overload and increased adipocyte ferroportin expression because of hemochromatosis are
associated with decreased adipocyte iron, increased adiponectin, improved glucose
tolerance, and increased insulin sensitivity. Phlebotomy of humans with impaired glucose
tolerance and ferritin values in the highest quartile of normal increased adiponectin and
improved glucose tolerance. These findings demonstrate a causal role for iron as a risk
factor for metabolic syndrome and a role for adipocytes in modulating metabolism through
adiponectin in response to iron stores.
Find the latest version:
http://jci.me/44421-pdf
Research article
Adipocyte iron regulates adiponectin
and insulin sensitivity
J. Scott Gabrielsen,1 Yan Gao,1 Judith A. Simcox,1 Jingyu Huang,1 David Thorup,1 Deborah Jones,1
Robert C. Cooksey,1,2 David Gabrielsen,1 Ted D. Adams,3 Steven C. Hunt,3 Paul N. Hopkins,3
William T. Cefalu,4 and Donald A. McClain1,2
1Departments of Medicine and Biochemistry, University of Utah School of Medicine, Salt Lake City, Utah, USA. 2VA Medical Center, Research Service,
Salt Lake City, Utah, USA. 3Division of Cardiovascular Genetics, Department of Medicine, University of Utah School of Medicine, Salt Lake City, Utah, USA.
4Pennington Biomedical Research Center, Louisiana State University System, Baton Rouge, Louisiana, USA.
Iron overload is associated with increased diabetes risk. We therefore investigated the effect of iron on adiponectin, an insulin-sensitizing adipokine that is decreased in diabetic patients. In humans, normal-range serum
ferritin levels were inversely associated with adiponectin, independent of inflammation. Ferritin was increased
and adiponectin was decreased in type 2 diabetic and in obese diabetic subjects compared with those in equally
obese individuals without metabolic syndrome. Mice fed a high-iron diet and cultured adipocytes treated with
iron exhibited decreased adiponectin mRNA and protein. We found that iron negatively regulated adiponectin
transcription via FOXO1-mediated repression. Further, loss of the adipocyte iron export channel, ferroportin,
in mice resulted in adipocyte iron loading, decreased adiponectin, and insulin resistance. Conversely, organismal iron overload and increased adipocyte ferroportin expression because of hemochromatosis are associated with decreased adipocyte iron, increased adiponectin, improved glucose tolerance, and increased insulin
sensitivity. Phlebotomy of humans with impaired glucose tolerance and ferritin values in the highest quartile
of normal increased adiponectin and improved glucose tolerance. These findings demonstrate a causal role
for iron as a risk factor for metabolic syndrome and a role for adipocytes in modulating metabolism through
adiponectin in response to iron stores.
Introduction
Increased iron stores are associated with increased risk of type 2
diabetes (1–4), gestational diabetes (5), prediabetes (6), metabolic
syndrome (MetS) (7), central adiposity (8), and cardiovascular disease (9, 10). The mechanisms underlying these associations are
poorly understood. The commonly used marker for total body
iron stores, serum ferritin, is also responsive to inflammatory
stress (11, 12), so increased ferritin in diabetes could simply reflect
the inflammatory component of that disease (13). On the other
hand, phlebotomy improves glycemia and MetS traits (14–17),
arguing that iron may play a causal role in diabetes.
The possible mediators of the association between iron and diabetes risk are not known. Decreases in both insulin secretion and sensitivity have been linked to iron. Excess iron impairs pancreatic β cell
function and causes β cell apoptosis (18–21). Recent studies have
also found a negative correlation between serum ferritin and the
insulin-sensitizing adipokine, adiponectin (3, 22–24). The hypothesis that adiponectin links iron and insulin resistance is appealing, as
decreased adiponectin levels are associated with obesity and type 2
diabetes (25) and are causally linked with insulin resistance (26).
We therefore investigated the mechanisms underlying the relationships among serum ferritin, adiponectin, and MetS in mice
and humans. We demonstrate in humans that the association
between serum ferritin and adiponectin is independent of inflammation and that serum ferritin, even within its normal ranges, is
among the best predictors of serum adiponectin. Studies in cell
culture, mouse models, and humans demonstrate that iron plays
a direct and causal role in determining adiponectin levels and diabetes risk. The adipocyte expresses specialized proteins related to
Conflict of interest: The authors have declared that no conflict of interest exists.
Citation for this article: J Clin Invest. 2012;122(10):3529–3540. doi:10.1172/JCI44421.
iron metabolism that make it well suited to perform as an iron
sensor, allowing it to integrate iron availability into its broader
nutrient-sensing function.
Results
Human ferritin levels are inversely associated with serum adiponectin independently of inflammation. We studied 110 individuals with (n = 49)
and without (n = 61) diabetes recruited for an independent study
of metabolic flexibility (27). Serum ferritin was negatively associated with serum adiponectin (r = –0.294, P = 0.0017). To mitigate
the effects of inflammation and/or extreme iron overload and
anemia, we next restricted the analysis to individuals with normal
serum ferritin (men, >30 ng/ml and 100
3530
mg/dl or diabetes (32). Those without MetS had no more than
1 factor in addition to obesity, and none had diabetes. Serum
ferritin was measured in all subjects (n = 125), and adiponectin
was measured in a randomly selected subset (n = 38). Overall,
the MetS group had significantly higher ferritin (260 ± 23 ng/
ml vs. 185 ± 21 ng/ml, P < 0.01) and lower adiponectin (11.5 ± 1.0
μg/ml vs. 18.9 ± 1.9 μg/ml, P < 0.005) than the non-MetS group.
Higher ferritin values were seen in the MetS subset with diabetes
compared with those in either the non-MetS group or the MetS
subgroup without diabetes (Figure 1D, P < 0.03). Adiponectin was
higher in the non-MetS group compared with that in either MetS
subgroup (Figure 1E, P < 0.03). Consistent with the data in Figure 1A, ferritin was inversely correlated with adiponectin in this
cohort, although in this smaller group the relationship did not
quite reach the level of statistical significance (r = 0.304, n = 38,
P = 0.06, data not shown). Insulin resistance estimated from the
homeostasis model (HOMA-IR) was correlated positively with
ferritin (r = 0.264, n = 125, P < 0.01) and negatively with adiponectin (r = 0.48, n = 38, P < 0.005).
Adipocyte iron increases and adiponectin mRNA and serum protein
levels decrease in dietary iron overload. Several studies have identified effects of iron on adipocyte metabolism (e.g., refs. 33, 34). To
explore the regulation of adipocyte iron levels, we first demonstrated that adipocyte iron levels respond to dietary iron content
in WT C57BL6/J mice by measuring mRNA levels of the transferrin receptor (Tfrc). Tfrc mRNA contains iron response elements in
its 3′ untranslated region that result in decreased Tfrc mRNA levels
as cellular iron levels increase (35). We observed a 40% decrease in
Tfrc mRNA in adipocytes from mice fed a high-iron diet (20 g/kg
iron) compared with that in mice fed normal chow (330 mg/kg
iron) (Figure 2A, P < 0.05). Technical difficulties precluded the
direct assay of cytosolic iron in isolated adipocytes. To investigate
a possible direct and causal role of iron in the regulation of adiponectin, we studied the effects of dietary iron overload on serum
adiponectin levels in mice. We fed 129SvEvTac male mice high(20,000 mg/kg carbonyl iron), normal (330 mg/kg), or low- (7 mg/
kg) iron diets for 2 months. Body weights were significantly lower
in mice fed high-iron diets and significantly higher in those fed
low-iron diets compared with those of mice fed normal chow (low
iron, 34.6 ± 1.1 g; normal chow, 29.0 ± 0.5 g; high iron, 26.7 ± 0.6 g,
P < 0.001 between groups by ANOVA, n = 8–12/group). Despite
decreased body weight, serum adiponectin levels were 29% lower in
iron-overloaded mice (Figure 2B, P = 0.0002). Conversely, dietary
iron restriction increased serum adiponectin levels by 31% despite
increased body weight (Figure 2B, P = 0.0036). A similar 29%
decrease in serum adiponectin was also seen in a different strain,
The Journal of Clinical Investigation http://www.jci.org Volume 122 Number 10 October 2012
research article
Table 1
Multiple regression models for the relationship of adiponectin to ferritin
Statistical model
Variables
Multiple regression,
men and women (n = 83)
Multiple regression,
men only (n = 47)
Ferritin correlation
coefficient
Ferritin P value
t ratio
P value
0.006228
0.006102
0.006124
0.005834
0.002041
0.003887
0.003567
0.003423
0.002770
0.0004
0.0004
0.0003
0.0006
0.23
0.021
0.033
0.039
0.10
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
1.681
1.485
1.012
0.441
0.10
0.15
0.32
0.66
Ferritin
Ferritin/log(CRP)
Ferritin/log(CRP)/BMI
Ferritin/log(CRP)/BMI/diabetes
Ferritin/log(CRP)/BMI/diabetes/gender
Ferritin
Ferritin/BMI
Ferritin/BMI/diabetes
Ferritin/BMI/diabetes/log(CRP)
Relative contribution of variables,
men only
Ferritin
Log(CRP)
Diabetes
BMI
Multivariate analysis of the relationship between ferritin and adiponectin corrected for the variables of C-reactive protein (log CRP), body mass index (BMI),
diabetes status, and gender. Results are presented for the entire cohort and for men only, with the relative contribution of variables also indicated for the
men-only cohort.
namely male C57BL6/J mice fed normal chow and high-iron diets
(6.46 ± 0.15 μg/ml vs. 4.55 ± 0.16 μg/ml, respectively, P < 0.0001).
Adiponectin mRNA levels in isolated epididymal adipocytes were
30% lower in iron-overloaded mice (Figure 2C, P = 0.07), mirroring
the changes in serum adiponectin levels. With a smaller cohort
of high-iron diet– and normal chow–fed mice (n = 4/group), we
determined body composition by magnetic resonance imaging,
and the high-iron diet caused not only a decrease in weight but an
even larger relative effect on fat mass because of a parallel increase
in lean body mass (Figure 2D). Both food intake and oxygen consumption rates were higher in the lower-weight, high-iron group
(Figure 2D). The differences in weight seen with the high-iron
diet were not observed in mice with genetic deletion of adiponectin (Figure 2E). The expected inverse linear relationship between
weight and adiponectin was observed in mice fed normal chow
(r = 0.48, n = 22, P = 0.02, not shown), but this relationship was
lost in mice fed the high-iron diet, in fact even trending toward a
positive relationship (r = 0.17, n = 17, P = 0.51).
Finally, to demonstrate that the change in adiponectin was
accompanied by changes in glucose metabolism, we performed
hyperinsulinemic clamp studies on WT mice on normal chow and
those that had been on the high-iron diet for 8 weeks. There was
a trend toward lower glucose disposal, normalized to total body
weight, in the mice fed high-iron diet (Figure 2F, P = 0.22). However, because most glucose uptake at hyperinsulinemia is into skeletal muscle, we also normalized to lean mass based on the magnetic resonance imaging findings, and the mice on high-iron diet
had a significant decrease in their maximal glucose disposal rate
per gram of lean tissue (P < 0.001).
Iron decreases adiponectin transcription. To demonstrate further that
the decreased adiponectin mRNA levels are due to decreased transcription and that iron regulates adiponectin directly, we examined the effects of iron on adiponectin in a cell culture model.
Treatment of 3T3-L1 adipocytes with iron sulfate decreased media
adiponectin protein levels in a dose-dependent manner (Figure 3A,
P < 0.0001). Adiponectin mRNA levels also decreased 30% with
iron treatment (Figure 3B, P = 0.02). We measured luciferase activ-
ity driven by the proximal 1,460 bp of the murine adiponectin promoter, which contains most of the previously identified sites that
regulate adiponectin transcription (36). Iron decreased promoter
activity by 28% (Figure 3C, P = 0.0025). Iron did not decrease the
half-life of the endogenous mRNA or the reporter construct measured after actinomycin C or cycloheximide D treatment of cells
(data not shown). Most physiologic regulation of adiponectin gene
transcription is attributable to the factors FOXO1 and PPARγ (36,
37). To explore the mechanism of regulation of adiponectin by
iron, we first examined posttranslational modification of FOXO1.
Iron caused decreased acetylation of FOXO1 without changing its
level of phosphorylation or total protein (Figure 3, D and E). In
agreement with the lack of change of FOXO1 phosphorylation, we
also detected no differences in basal or insulin-stimulated phosphorylation of AKT in iron-treated cells (Figure 3F).
Contrary to the observed effects of iron on adiponectin transcription, deacetylation of FOXO1 is generally associated with
increased adiponectin transcription (36, 38). We therefore measured FOXO1 occupancy at its 2 known sites of transcriptional
activation using ChIP. As predicted by FOXO1 acetylation status,
cells treated with iron exhibited a 3.1-fold increase in occupancy
by FOXO1 at the sites reported to stimulate adiponectin transcription (P < 0.01, Figure 3G). FOXO1, however, has also been reported
to transrepress adiponectin transcription when associated with
the PPARγ response element (PPRE) of the adiponectin promoter
(39). Iron treatment resulted in a 3.5-fold enhancement of FOXO1
binding to the PPRE (P = 0.01, Figure 3G). There was no significant increase in PPARγ binding to the PPRE (1.6 fold, P = 0.07,
Figure 3G). Because the interaction of C/EBPα with FOXO1 has
also been implicated in regulation of adiponectin transcription
(36), we also measured the association of C/EBPα with FOXO1 by
coimmunoprecipitation but saw no effect of iron on this association (P = 0.93, Figure 3H).
Adipocytes express the specialized iron channel ferroportin. Adipocytes
express genes that are generally restricted to iron-sensing tissues
(40, 41). We therefore sought to obtain evidence that adipocytes
might have a specialized iron-sensing capacity by assessing adi-
The Journal of Clinical Investigation http://www.jci.org Volume 122 Number 10 October 2012
3531
research article
Figure 2
Serum adiponectin and adipocyte mRNA levels decrease with dietary iron overload and with iron treatment in 3T3-L1 cells. (A) Tfrc mRNA
quantified by RT-PCR and normalized to cyclophilin in collagenased adipocytes from epididymal fat pads of mice fed normal chow (NC) or a
high-iron diet (HI) for 8 weeks. *P < 0.05. (B) Serum adiponectin levels were measured in 7-month-old 129/SvEvTac background mice following
2 months of being fed low-iron (7 mg/kg carbonyl iron), normal chow (330 mg/kg), or high-iron (20 g/kg) diets. *P = 0.004, low iron vs. normal chow;
‡P = 0.0002, high iron vs. normal chow. (C) Adiponectin mRNA levels in isolated epididymal adipocytes from mice fed high-iron diet or normal
chow. P = 0.07. (D) Body weights were determined in C57BL6/J mice after 8 weeks on normal chow or high-iron diets, and body composition
was determined by magnetic resonance imaging (n = 6/group or 8/group, *P < 0.05, ‡P < 0.01). (E) Body weights were determined in mice with
knockout of the adiponectin gene compared with those of controls after 8 weeks on normal chow or high-iron diets (n = 5–12/group, ‡P < 0.001).
(F) Euglycemic hyperinsulinemic clamps were performed on WT mice on normal or high-iron diets. Glucose infusion rates (GIRs) trended lower
when normalized to total body weight (P = 0.22) but differed significantly when normalized to lean body mass (‡P < 0.0005).
pocyte expression of the iron export channel ferroportin, whose
significant tissue expression has been reported to be limited to
gut enterocytes, placenta, and reticuloendothelial cells, including macrophages (42, 43). Ferroportin mRNA and protein were
detectable by quantitative RT-PCR and Western blot in differentiated 3T3-L1 adipocytes. Treatment with iron sulfate increased
ferroportin mRNA in a dose-dependent manner (Figure 4A,
P < 0.0001). Protein levels were responsive both to iron and to
treatment by hepcidin, a ferroportin ligand that results in ferroportin downregulation (Figure 4B and ref. 44).
Deletion of adipocyte ferroportin results in increased adipocyte iron
levels, decreased serum adiponectin, and increased insulin resistance. To
demonstrate that ferroportin serves as a functional iron channel and exporter in adipocytes, we generated mice lacking the
ferroportin gene in adipocytes (Fpn1–/– mice). An Fpn1fl/fl mouse
(45), provided by Nancy C. Andrews (Duke University, Durham,
North Carolina, USA), was crossed to a mouse expressing Cre
recombinase under control of the 5.4-kB AP2 promoter. The Ap2Cre:Fpn1fl/fl mice were subsequently backcrossed onto the 129
strain for at least 5 generations. Ferroportin mRNA was undetectable in adipocytes purified by collagenase digestion from Fpn1–/–
mice (Figure 4C). Because macrophages can also express the Ap2
gene, we examined ferroportin expression in splenocytes, wherein
the only cell expressing significant ferroportin is the macrophage.
There was no decrease in ferroportin mRNA in splenocytes from
the Ap2-Cre:Fpn1fl/fl mice (Figure 4C).
To demonstrate a role for ferroportin in modulating adipocyte
iron, we measured Tfrc mRNA in isolated adipocytes. Compared
with WT Fpn1fl/fl adipocytes, Ap2-Cre:Fpn1fl/fl adipocytes exhib3532
ited a 44% decrease in Tfrc (Figure 4D, P < 0.001), consistent with
increased cytosolic iron, and functionality of the ferroportin
channel is in adipocytes.
The increased levels of adipocyte iron in the Fpn1–/– mice resulted
in a 58% decrease in levels of adiponectin mRNA in adipocytes (Figure 4E, P < 0.01), and this was reflected in decreased
serum adiponectin (Figure 4F). Because of the heterogeneity in
the weights of the mice and the effect of weight on adiponectin, serum adiponectin was determined in a cohort of mice all
weighing less than 30 g. In control (Fpn1fl/fl) mice, the high-iron
diet resulted in a 12% decrease in serum adiponectin (Figure 4F,
P < 0.05). Serum adiponectin was also lower (13%, P < 0.05) in
the Ap2-Cre:Fpn1fl/fl mice on normal chow compared with controls
and did not decrease further in mice on the high-iron diet. No
changes were noted in the distribution of adiponectin molecular
weight isoforms, as analyzed by an adiponectin assay that detects
both total and high molecular weight isoforms (Figure 4G) and as
analyzed by native SDS-PAGE (data not shown).
To determine whether the change in adiponectin was physiologically significant, we performed glucose tolerance testing in
WT and Fpn1–/– mice. Fpn1–/– mice had significantly higher glucose excursions at 30 and 60 minutes after challenge (Figure 4H,
P < 0.05), and the areas under the glucose curve differed significantly between the groups (16,372 mg-min/dl in WT mice and
20,272 mg-min/dl in Fpn1–/– mice, 24% increase, P < 0.001, data not
shown). Fasting glucose and insulin levels were also determined
in WT and Fpn1–/– mice on different levels of dietary iron, and
insulin resistance, as determined by homeostasis model assessment (HOMA-IR), was increased in the Fpn1–/– mice (P < 0.01,
The Journal of Clinical Investigation http://www.jci.org Volume 122 Number 10 October 2012
research article
Figure 3
Transcriptional regulation of adiponectin by iron. (A) Media adiponectin levels in 3T3-L1 cells 12 hours following 12-hour pretreatment. P < 0.0001.
(B) RT-PCR quantification of adiponectin mRNA levels in 3T3-L1 adipocytes treated with no iron or 100 μM FeSO4 for 24 hours, normalized to
cyclophilin A. *P = 0.02. (C) Adiponectin promoter-driven luciferase activity in the presence or absence of 100 μM FeSO4. ‡P = 0.0025. (D) Western blot for acetylated FOXO1 (Ac-FOXO1), phosphorylated FOXO1 (P-FOXO1), total FOXO1 (Tot-FOXO1), and β-actin in 3T3-L1 adipocytes
treated with no iron or 100 μM FeSO4 for 8 hours. (E) Quantitation of Western blots (total n = 6 independent determinations) normalized to β-actin.
*P < 0.05. (F) Quantitation of Western blots for phosphorylated AKT in 3T3-L1 adipocytes treated with no iron or 100 μM FeSO4 for 8 hours and
insulin (10 nM) for 1 hour. (G) ChIP showing FOXO1 occupancy of adiponectin promoter FOXO1 sites and PPRE and PPARγ occupancy of PPRE
in 3T3-L1 adipocytes (n = 3 experiments each assayed in duplicate, *P < 0.05). (H) Immunoprecipitation of 3T3-L1 adipocyte extracts, treated
overnight in the presence or absence of 100 μM FeSO4, by antibodies to FOXO1, followed by immunoblotting for FOXO1 (t-FOXO1) and C/EBPα
(0.58 ± 0.15 density units for control, 0.61 ± 0. 26 density units for iron-treated extracts, P = 0.93).
data not shown). The effects of high-iron diet on body weight and
body composition were also lost in the Fpn1–/– mice. Body weights
of the Fpn1 –/– mice on normal chow compared with those of
mice on high-iron diets did not differ (normal chow, 32.1 ± 1.5 g,
high iron 31.5 ± 1.3 g, P = 0.68), and the changes in body composition induced by high-iron diet (Figure 2D) and replicated
in a cohort of control Fpn1fl/fl mice (Figure 4I, P < 0.05 for both
percentage of lean and fat mass) were also not seen in the Fpn1–/–
mice (Figure 4I, P = 0.37 and P = 0.56 for percentage of lean and
fat mass, respectively).
Lower adipocyte iron, higher serum adiponectin, and increased insulin
sensitivity in hereditary hemochromatosis. The effects of iron on adiponectin levels present a paradox. Namely, adiponectin decreases
with dietary iron overload, and yet serum adiponectin levels are
increased in a mouse model of genetic iron overload, wherein the
gene most commonly mutated in human hereditary hemochro-
matosis (HH) has been deleted (Hfe–/– mice) (46). It has been shown
that the relative lack of hepcidin in HH results in failure to downregulate ferroportin, so that cells that express significant amounts
of the iron channel are paradoxically less loaded with iron in HH
(47, 48). We therefore sought to determine whether the same
were true of adipocytes. Tfrc mRNA levels, which inversely reflect
cytosolic iron, were increased by 67% in adipocytes from Hfe–/–
mice compared with those from WT mice (Figure 5A, P = 0.05).
Thus, the previously reported increased adiponectin in a mouse
model of HH (46) is consistent with lower adipocyte iron levels.
Because humans with HH also trend toward increased insulin
sensitivity prior to the onset of clinical diabetes (20), we hypothesized that serum adiponectin levels would likewise be increased
in human HH. Serum adiponectin levels were increased by 89%
in male subjects with HH, compared with non-HH, male sibling
controls (Figure 5B, P = 0.04). In women, serum adiponectin lev-
The Journal of Clinical Investigation http://www.jci.org Volume 122 Number 10 October 2012
3533
research article
Figure 4
Functional expression of ferroportin in adipocytes. (A) Fpn1 mRNA levels quantified by RT-PCR in 3T3-L1 adipocytes exposed to different concentrations of iron (FeSO4) in the culture medium. *P < 0.05 compared with 0.1 mM, ‡P < 0.01, P < 0.0001 for overall trend. (B) FPN1 protein
levels in 3T3-L1 adipocytes, as detected by immunoblotting in adipocytes treated with no iron, 100 μM FeSO4, or 1 μg/ml hepcidin for 8 hours. (C)
Fpn1 mRNA in adipose tissue and spleen from WT (Fpn1fl/fl) and Fpn1–/– (AP2 Cre ferroportin knockout) mice. ‡P < 0.001. (D) Tfrc mRNA quantified by RT-PCR and normalized to cyclophilin in collagenased adipocytes from epididymal fat pads of WT and Fpn1–/– mice (n = 10–14/group,
‡P < 0.001). (E) Adiponectin mRNA in the adipocytes used in E. ‡P < 0.01. (F) Serum adiponectin in WT and Fpn1–/– mice (n = 9–12/group,
*P < 0.01). (G) High molecular weight (HMW) adiponectin determined as a percentage of total in the same group depicted in F. (H) Glucose
tolerance testing of WT and Fpn1–/– mice (n = 5–6/group, *P < 0.05 for individual glucose values). (I) Body composition by magnetic resonance
imaging in WT and Fpn1–/– mice on normal chow or high-iron diets (n = 11–20/group, *P < 0.05).
els also trended higher (136%) in patients with HH (Figure 5B,
3.3 μg/ml vs. 1.4 μg/ml, P = 0.06). In non-HH sibling controls of
the subjects with HH, serum adiponectin and BMI were closely
and inversely associated (Figure 5C, r = 0.77, P = 0.03), consistent
with previous reports (31, 49). However, the association between
BMI and adiponectin was lost in patients with HH (Figure 5C,
r = 0.09, P = 0.87). Serum ferritin, which largely reflects hepatic
iron stores, was not associated with serum adiponectin levels in
patients with HH (data not shown).
To determine whether the increased adiponectin levels in HH
mice were functionally significant, we generated mice with deletion of the adiponectin gene (APN–/–, provided by Phillip Scherer,
ref. 50) on the C57BL6/J-HH (Hfe–/–) background. Because the
effects of adiponectin deletion on glucose tolerance are more manifest in obese mice or mice exposed to a high-fat diet (26, 50), the
mice were fed a high-fat diet for 8 weeks. The Hfe–/– mice exhibited
a 19% decrease in fasting glucose compared with APN–/– mice, an
improvement that was completely lost in the APN–/–:Hfe–/– dou3534
ble-knockout mice (Figure 5D, P = 0.006 by ANOVA). A similar
trend was seen in animals on normal chow but was not significant
because of the relatively smaller effects of both the Hfe–/– and APN–/–
genotypes on glucose in mice on normal chow (fasting glucose levels, 126 ± 7 mg/dl in WT, 115 ± 4 mg/dl in Hfe–/–, and 129 ± 11 mg/dl
in APN–/–:Hfe–/–, P = 0.16, data not shown).
Phlebotomy increases adiponectin and improves glucose tolerance in
humans with high-normal serum ferritin. We next sought to determine whether iron plays a causal role in determining adiponectin
levels and the risk of MetS in humans. We studied humans with
impaired glucose tolerance (IGT) whose serum ferritin levels were
in the highest quartile of normal (221 ± 42 ng/ml, see Supplemental Table 2). Individuals with chronic inflammatory states, such as
hepatitis, arthritides, or infections, were excluded, as were individuals with HH. Subjects received oral and frequently sampled intravenous glucose tolerance tests (OGTTs and FSIVGTTs) before and
approximately 6 months after phlebotomy, which was sufficient
to result in a fall in serum ferritin to the lowest quartile of normal
The Journal of Clinical Investigation http://www.jci.org Volume 122 Number 10 October 2012
research article
Figure 5
Adiponectin in mouse and human hemochromatosis. (A) Tfrc mRNA levels in isolated
adipocytes from WT and Hfe–/– mice on normal chow, normalized to cyclophilin A (n = 5/
group, *P = 0.05). (B) Serum adiponectin levels in male (*P = 0.04) and female (P = 0.06)
subjects with HH compared with non-HH sibling controls. (C) Serum adiponectin levels
plotted as a function of BMI for subjects with
HH (black circles) and non-HH sibling controls (white circles). Sexes were combined for
linear regression analysis of HH (P = 0.87)
and control subjects (P = 0.03). (D) Fasting
glucose levels in WT, Hfe–/–, and Hfe–/–:APN–/–
double-knockout mice (n = 5–7/group,
*P = 0.006 by ANOVA).
(33 ± 12 ng/ml, P < 0.02 compared with before phlebotomy). The
average blood donation was 3.7 units.
After phlebotomy, all subjects improved in the area under the
glucose curve during OGTT, and the difference in the groups
before and after phlebotomy was significant (Figure 6A, P = 0.03).
All subjects had both IGT and impaired fasting glucose (IFG)
prior to phlebotomy. After phlebotomy, one exhibited correction
of both parameters, one exhibited correction of IGT only, and one
exhibited correction of IFG only. There was a nonsignificant trend
toward improvement in the FSIVGTT parameters of insulin secretory capacity (acute insulin response to glucose [AIRg], ~2.5 fold,
Figure 6B) and insulin sensitivity (Si, ~3 fold, Figure 6C). Similar
trends toward improvement were seen in the homeostasis model
indices of β cell function and insulin resistance (data not shown).
The disposition index, the product of Si and AIRg and the better predictor of overall diabetes risk (51), improved significantly
(Figure 6D, P < 0.05). After phlebotomy, all subjects showed an
increase in serum adiponectin (range of increase 9%–55%, see Supplemental Table 2, P < 0.02 by paired t test). Similar to the results
seen in mice on a low-iron diet (Figure 2A), adiponectin increased
despite an average weight gain of 2.6 kg, with weight gain occurring in two-thirds of the subjects.
Discussion
Increased serum ferritin is associated with insulin resistance and
increased risk for diabetes (1–3). Recent studies (3, 22, 23) have
Figure 6
Phlebotomy improves glucose tolerance. Human
men with impaired glucose tolerance and serum
ferritin levels in the highest quartile of normal underwent phlebotomy to decrease their ferritin values
to the lowest quartile of normal. (A) Results of oral
glucose tolerance testing before (open symbols) and
approximately 6 months after initiation of phlebotomy
(closed symbols). The integrated area under the glucose curve for the 120-minute test is shown. Squares
represent individual values, and circles represent the
means. (*P < 0.03 by paired t test). (B) Insulin secretory capacity (AIRg) and (C) Si were determined from
FSIVGTT performed before and after phlebotomy. (D)
The disposition index was calculated from the data in
B and C. *P < 0.05 by paired t test.
The Journal of Clinical Investigation http://www.jci.org Volume 122 Number 10 October 2012
3535
research article
also noted an association between serum ferritin and adiponectin, the adipocyte-specific, insulin-sensitizing hormone. We have
verified that serum ferritin levels, reflecting tissue iron stores, are
more tightly associated with adiponectin than its more common
predictor, obesity. More importantly, the relationship is causal,
reflecting regulation of adiponectin transcription by iron. We have
demonstrated this in cultured cells, by manipulation of iron stores
and adipocyte iron levels in rodents, and in humans. Adiponectin
is causally linked to insulin sensitivity (26), and, consistent with
this, the changes in adiponectin in response to iron are accompanied by changes in glucose tolerance and insulin sensitivity.
The fact that adipocytes use iron levels to regulate adiponectin suggests a role for adipocytes in coordinating organism-wide
metabolic responses to iron availability, as they do for responses
to overall macronutrient status. There is other evidence for crosstalk between iron and adipocyte metabolism. Insulin treatment,
for example, increases iron uptake by increasing cell surface
expression of transferrin receptor 1 in 3T3-L1 and rat adipocytes
(52, 53). Iron induces lipolysis in cultured adipocytes and modulates the lipolytic response to norepinephrine (34, 54). Close
coregulation of iron levels and metabolic parameters, such as fuel
preference, is conserved from yeast (55, 56) to mammals (46). The
need for this coregulation is consistent with the necessity of iron
for electron transport and other redox reactions combined with
its dangers as a potent oxidant.
Adipocytes are well suited for their iron-sensing role. They
express not only common regulators of iron homeostasis, such as
ferritin and iron regulatory proteins (57), but also iron-related proteins with restricted tissue expression, incl
