Department of Vascular Medicine, University Medical Center Utrecht, Utrecht, the Netherlands
Correspondence to: Frank L. J. Visseren, MD, PhD, Department of Vascular Medicine, University Medical Center Utrecht, P. O. Box 85500, Utrecht 3508 GA, the Netherlands. Tel.: +31 88 7555555; fax: +31 88 7555488; e-mail: F.L.J.Visseren@umcutrecht.nl
We investigated whether plasma ferritin levels through the pro-inflammatory effects of free iron are associated with adipose tissue dysfunction in a relevant population of patients with manifest vascular disease who would potentially benefit the most from further aetiological insights.
Materials and methods
In a cohort of 355 patients with vascular diseases, the association between plasma ferritin and adiponectin levels was quantified using linear regression analysis. Interleukin-6 and adiponectin levels were measured in medium from pre-adipocytes and adipocytes after incubation with increasing concentrations of Fe(III)-citrate and after co-incubation with iron chelators or radical scavengers.
Increasing ferritin plasma concentrations were not related to plasma adiponectin levels in patients without (β −0·13; 95% CI −0·30 to 0·04) or with the metabolic syndrome (β −0·04; 95% CI −0·17 to 0·10). Similar results were found in patients who developed a new cardiovascular event in the follow-up period. In vitro, incubation with increasing concentrations of Fe(III)-citrate-induced inflammation in pre-adipocyte cultures as witnessed by increased IL-6 secretion at 30 μm Fe(III)-citrate vs. control (500 ± 98 pg/mL vs. 194 ± 31 pg/mL, P = 0·03). Co-incubation of pre-adipocytes with iron chelators or radical scavengers prevented this inflammatory response. Incubation of adipocytes with 30 μm Fe(III)-citrate did not influence adiponectin secretion compared with control.
In patients with vascular disease, there is no association between plasma ferritin and adiponectin levels. In vitro, free iron induces an inflammatory response in pre-adipocytes, but not in adipocytes. This response was blocked by co-incubation with iron chelators or radical scavengers. Adiponectin secretion by adipocytes was not influenced by free iron.
Abdominal obesity is strongly associated with inflammation, insulin resistance, vascular complications and type 2 diabetes mellitus [1-3]. Adipose tissue dysfunction has been indicated as an intermediate between abdominal obesity and subsequent inflammation and insulin resistance through an increase in release of pro-inflammatory adipocytokines and a decrease in release of anti-inflammatory adipocytokines such as adiponectin . Adiponectin is a well-characterized adipokine with various beneficial effects such as insulin-sensitizing, anti-atherosclerotic and anti-inflammatory properties . Obesity has been shown to be associated with decreased plasma adiponectin levels, possibly through an increase in inflammation [5-7].
Body iron stores and catalytic iron are associated with an increased risk of the development of diabetes [8, 9], while iron depletion by bloodletting has been shown to improve insulin sensitivity in patients with type 2 diabetes mellitus and nonalcohol fatty liver disease [10, 11]. Indeed, higher plasma ferritin levels have been associated with inflammation and insulin resistance [12-14]. As plasma ferritin levels can be seen as an indicator of body iron stores, higher plasma ferritin levels may be associated with higher levels of free plasma iron . While the precise mechanisms behind the insulin-sensitizing effect of bloodletting are unclear, several options have been suggested: First, the insulin-sensitizing effect may occur through reduced reactive oxygen species (ROS) production. Activation of the IKKβ/nuclear factor-κB pathway by tumour necrosis factor-α (TNF-α), signalling via Toll-like receptors (TLRs) by, for example, free fatty acids (FFAs) and by ROS, all eventually lead to phosphorylation of IRS-1 at serine sites that inhibit normal signalling through the insulin receptor/IRS-1 axis in adipocytes . Unbound Fe(II) catalyses the Haber–Weiss reaction producing ROS, which lead to increased inflammation and insulin resistance [16-18]. Secondly, high body iron stores are also linked to vitamin A metabolism via higher plasma levels of the insulin resistance inducing adipokine retinol-binding protein-4 (RBP-4) [19, 20]. Indeed, depleting body iron stores with bloodletting decreases plasma RBP-4 levels . Higher plasma ferritin levels may also be the result of inflammation by interleukin-6 (IL-6)-induced hepcidin synthesis . In this view, abdominal obesity and adipose tissue dysfunction increase synthesis of hepcidin, which causes internalization and degradation of the transmembrane Fe(II) transporter ferroportin upon binding, leading to decreased intestinal uptake of iron and sequestering of iron in ferritin . This notion is supported by findings that lowering body weight is associated with lower plasma hepcidin and ferritin levels . It is thus unclear whether high plasma ferritin levels are causative in the development of obesity-related inflammation and insulin resistance via adipose tissue dysfunction or just a result of IL-6-dependent increased synthesis of hepcidin. To investigate the possibility that free iron may induce adipose tissue dysfunction, we set out to investigate the relation between plasma levels of ferritin and adiponectin in a cohort of patients with manifest vascular diseases and by investigating the causal relation between free iron and IL-6 and adiponectin production in pre-adipocytes and adipocytes in vitro.
Study design and patients
Patients originated from the Second Manifestations of ARTerial disease study (SMART), an ongoing prospective cohort study at the University Medical Centre Utrecht designed to establish the prevalence of concomitant arterial diseases and risk factors for atherosclerosis in a high-risk population . In this study, participants are screened noninvasively for other manifestations of atherosclerotic diseases and risk factors than the inclusion diagnosis. This study complies with the Declaration of Helsinki. The local Ethics Committee approved the research protocol, and all patients gave their written informed consent. For this study, we used data from an earlier published case-cohort study, which included patients with manifest vascular disease who developed a new event during follow-up (cases) and those who did not (controls) [,]. Of the 2398 patients enrolled in the SMART study between 1996 and 2003, 220 patients emerged, who suffered from a new cardiovascular event during follow-up. Second, a random sample of 10% (240 patients) was drawn as reference from the 2398 patients having a history of cardiovascular disease in the SMART population. Of these 240 patients, 16 had already been selected because of an outcome event in the initial 220 selected cases. A further 23 patients were excluded due to missing blood samples. To limit the effect of outliers in plasma ferritin, the data were truncated at the lowest and highest 1%. Patients with high-sensitive C-reactive protein (hs-CRP) plasma concentrations > 15 mg/L were excluded, because this may indicate the presence of an active inflammatory condition. Finally, 15 premenopausal patients were excluded to limit the influence of the menses on plasma ferritin levels, leaving 355 patients with a mean follow-up of 2·3 ± 1·8 years for final analysis (Fig. 1) (162 patients who suffered a new cardiovascular event and 193 patients who did not). The metabolic syndrome was defined by the revised National Cholesterol Education Program (NCEP-R) criteria . The metabolic syndrome was diagnosed when patients fulfilled three or more of the following five criteria: waist circumference ≥ 102 cm (men) or ≥ 88 cm (women), fasting triglycerides ≥ 1·7 mm or use of triglyceride-lowering medication, fasting HDL cholesterol < 1·03 mm (men) or < 1·30 mm (women), blood pressure ≥ 130 mmHg systolic or ≥ 85 mmHg diastolic or use of blood pressure-lowering medication and fasting glucose ≥ 5·6 mm or use of glucose-lowering medication.
Culturing pre-adipocytes and adipocytes with SGBS cell culture in vitro
Simpson–Golabi–Behmel Syndrome (SGBS) cells were generously provided by Dr. Martin Wabitsch from the University of Ulm, Germany, and differentiated to adipocytes using a procedure modified from previous publications [28, 29]. In short, SGBS cells were grown to confluence in Dulbecco's Modified Eagle's Medium/Nutrient Mixture F-12 Ham's supplemented with 10% foetal bovine serum, 33 μm biotin, 17 μm pantothenate, 100 μg/mL streptomycin and 62·5 μg/mL penicillin.
For tests in pre-adipocytes, SGBS cells were grown to confluence. Two days after confluence, the medium was replaced with serum-free medium for 24 h after which the experiments were performed. For differentiated adipocytes, SGBS cells were stimulated to differentiate starting 2 days postconfluence with serum-free growth medium supplemented with 20 nm insulin, 200 pm triiodothyronine, 1 μm cortisol, 500 μm IBMX, 25 nm dexamethasone, 0·01 mg/mL human transferrin and 2 μm rosiglitazone. After 4 days, the medium was replaced with the differentiation medium without IBMX, dexamethasone and rosiglitazone. Ten days after initiation of differentiation, the medium was replaced with serum-free medium for 24 h after which experiments were conducted.
All cell cultures were exposed to the experimental conditions for the duration of 24 h after which the medium was collected for further testing.
Iron, iron chelators and radical scavengers
A 2·5 mm Fe(III)(citrate) stock solution with a 1 : 10 iron-citrate molar ratio was freshly prepared before each experiment by mixing equal parts of a 5·0 mm FeCl3 and a 50 mm NaCitrate solution (both Sigma-Aldrich, St Louis, MO, USA). Iron chelator and radical scavenger stock solutions were prepared in PBS as follows: 18 mm deferiprone (Sigma-Aldrich), 6 mm deferoxamine (Sigma-Aldrich), 20 mm tempol (Sigma-Aldrich), 2 m thiourea (Sigma-Aldrich) and 2 m dimethyl-thiourea (DMTU) (Sigma-Aldrich).
Cell viability was assessed using a Cytotoxicity Detection Kit (Roche, Basel, Switzerland) in accordance with instructions by measuring lactate dehydrogenase (LDH) in medium of both pre-adipocyte and adipocyte cell cultures after exposure to the experimental conditions. Maximum cytotoxicity was deemed low at 6%.
In patient, serum adiponectin and hs-CRP were analysed with a quantitative enzyme immunoassay technique (R&D Systems, Minneapolis, MO, USA). Ferritin was measured by two-site sandwich immunoassay (ADVIA Centaur Ferritin assay; Bayer, Leverkusen, Germany). Interleukin-6 (IL-6) and adiponectin were assessed in cell culture medium using ELISA (Pelikine; Sanquin Reagents, Amsterdam, the Netherlands).
Baseline characteristics of the study population are presented as means ± SD or median with interquartile range in the case of a skewed distribution. The baseline characteristics of the cases and controls were compared using a two-tailed independent t-test for normally distributed variables, a two-tailed Mann–Whitney-U test in case of skewed distributed variables and a chi-squared test for categorical variables. The relation between ferritin and adiponectin was determined with linear regression. Plasma ferritin, adiponectin and hsCRP levels were log-transformed as they had a skewed distribution. Results are expressed as beta (β) coefficients with 95% confidence intervals (95% CI) denoting change in log-transformed adiponectin per point of log-transformed ferritin. Three models were used to assess the relation between ferritin as independent factor and adiponectin as dependent variable. In our first model, we show the crude model, while in the second model, additional adjustment was carried out for age and gender. In our third model, we added hsCRP to investigate the role of inflammation in the relation between ferritin and adiponectin. As HDL cholesterol also has anti-inflammatory properties and is linked to adipose tissue dysfunction via insulin resistance, we performed similar analyses using plasma ferritin levels as the main determinant and plasma HDL cholesterol as outcome measure to support our main findings concerning plasma adiponectin levels.
Interaction analyses showed significant interaction for whether or not patients developed a new vascular event during follow-up (P = 0·0001). We found further interaction for the presence of the metabolic syndrome as defined by the revised National Cholesterol Education Program (NCEP-R) criteria  in both the patients with manifest vascular disease who developed a new vascular events (cases) (P = 0·012) and in patients with manifest vascular disease who did not develop an event (controls) (P < 0·001). We therefore investigated the relation between plasma ferritin and adiponectin levels in strata of patients with or without the metabolic syndrome and in patients with or without a new cardiovascular event in the follow-up period. Statistical significance was established when P < 0·05. The results of the in vitro experiments are presented as means ± SEM. Differences in quantitative measures were tested for significance using the unpaired two-tailed Student's t-test. Significance was established when P < 0·05.
Baseline characteristics of patient cohort
Baseline characteristics of the study population are presented in Table 1. All patients had manifest vascular disease. Patients who developed a new cardiovascular event during follow-up were 79% male with a mean age of 64 ± 11 years. Patients with manifest vascular disease drawn from the complete cohort were 78% male with a mean age of 61 ± 10.
Table 1. Baseline characteristics of the study population
Cases N = 162
Controls N = 193
MDRD, Modification of Diet in Renal Disease; hsCRP, high sensitivity C-reactive protein.
All data:% (n), mean ± SD or median (interquartile range).
The relation between ferritin and adiponectin in patients with manifest vascular disease
Plasma ferritin levels were negatively related to plasma adiponectin levels in patients with manifest vascular disease without a new event during follow-up (controls) without the metabolic syndrome (β −0·19; 95% CI −0·37 to 0·02) (Table 2). However, this relation was not statistically significant after adjustment for age, gender and hsCRP (β −0·13; 95% CI −0·30 to 0·04). In patients with metabolic syndrome, the crude model showed a weak relation between plasma levels of ferritin and adiponectin (β −0·08; 95%CI −0·21 to 0·06), which was further weakened after adjustment for age, gender and hsCRP (β −0·04; 95%CI −0·17 to 0·10). A similar nonsignificant negative relation was found between plasma ferritin and HDL cholesterol levels (data not shown).
Table 2. Relation between plasma ferritin and adiponectin levels in patients with manifest vascular disease
N = 78 No metabolic syndrome
N = 115 Metabolic syndrome
HsCRP, High sensitivity C-reactive protein.
All data: Beta regression coefficients (b) with 95% confidence interval (CI).
Model I: [log]ferritine
−0·19 (−0·37 to −0·02)
−0·08 (−0·21 to 0·06)
Model II: model I + gender + age
−0·12 (−0·28 to 0·05)
−0·04 (−0·17 to 0·10)
Model III: model II + [log]hsCRP
−0·13 (−0·30 to 0·04)
−0·04 (−0·17 to 0·10)
Relation between ferritin and adiponectin plasma levels in patients with manifest vascular disease who developed a new cardiovascular event
In patients with manifest vascular disease without the metabolic syndrome who developed a new cardiovascular event during the follow-up period (cases), plasma ferritin levels were negatively related to plasma adiponectin levels (β −0·08; 95%CI −0·24 to 0·08) (Table 3). Adjustment for age, gender and hsCRP did not change this relation (β −0·08; 95%CI −0·25 to 0·08). Ferritin plasma levels were negatively related to plasma adiponectin levels (β −0·06; 95%CI −0·22 to 0·10), but this was statistically not significant. After adjustment for age, gender and hsCRP, this relation was unchanged (β −0·03; 95%CI −0·19 to 0·13) (Table 2). A similar nonsignificant negative relation was found between plasma ferritin and HDL cholesterol levels (data not shown).
IL-6 production of cultured pre-adipocytes and adipocytes after incubation with Fe(III) in vitro
Incubation of pre-adipocytes with increasing concentrations of Fe(III)-citrate resulted in increased IL-6 secretion by pre-adipocytes from 194 ± 31 pg/mL in control experiments without Fe(III)-citrate to 500 ± 98 pg/mL after exposure to 30 μm Fe(III)-citrate (P = 0·03) for 24 h (Fig. 2). Co-incubation with the iron chelators deferoxamine or deferiprone to cell medium with 30 μm Fe(III)-citrate completely blocked the increase in IL-6 secretion by pre-adipocytes (deferoxamine: 180 ± 28 pg/mL, P = 0·02 and deferiprone: 40 ± 31 pg/mL, P < 0·01).
Separate co-incubation with either tempol, a scavenger, or thiouracil, a hydroxylradical scavenger, also blocked the Fe(III)-citrate-induced increase in IL-6 secretion by pre-adipocytes (tempol: 180 ± 15 pg/mL, P = 0·02 and thiouracil: 219 ± 37 pg/mL, P = 0·04). Co-incubation with DMTU, a H2O2 and hydroxylradical scavenger, was unable to prevent the Fe(III)-citrate-induced increase in IL-6 secretion by pre-adipocytes (259 ± 18 pg/mL). Similar experiments were conducted in adipocyte cultures with increasing concentrations of Fe(III)-citrate and co-incubation of 30 μm Fe(III)-citrate with iron chelators and radical scavengers. There was no detectable concentration of IL-6 level in adipocyte medium with or without incubation with Fe(III)-citrate.
The effect of Fe(III) on adiponectin secretion by cultured adipocytes in vitro
Results on adiponectin secretion by adipocytes are shown in Fig. 3. There was no change in adiponectin secretion from adipocytes by incubation with increasing concentrations of Fe(III)-citrate (0 μm: 64·3 ± 5·1 ng/mL vs. 30 μm: 66·5 ± 4·6 ng/mL, P = 0·76). Co-incubation of 30 μm Fe(III)(citrate) with either iron chelators or radical scavengers did not influence adiponectin secretion by adipocytes.
In this study, there was no relation between plasma ferritin and adiponectin plasma levels in patients with manifest vascular disease with or without the metabolic syndrome. Adjustment for inflammation did not change these relations. In vitro, Fe(III) induced production of IL-6 by pre-adipocytes but not by adipocytes. The Fe(III)-induced IL-6 production by pre-adipocytes could be blocked by intracellular and extracellular iron chelation and by ROS scavenging pointing towards a role for NFκB upregulation by ROS as a result of the Haber–Weiss reaction (Table 3).
Table 3. Relation between plasma ferritin and adiponectin levels in patients with manifest vascular disease who developed a new cardiovascular event
N = 66 No metabolic syndrome
N = 96 Metabolic syndrome
HsCRP, High sensitivity C-reactive protein.
All data: Beta regression coefficients (b) with 95% confidence interval (CI).
Model I: [log]ferritine
−0·08 (−0·24 to 0·08)
−0·06 (−0·22 to 0·10)
Model II: model I + gender + age
−0·09 (−0·24 to 0·07)
−0·03 (−0·19 to 0·13)
Model III: model II + [log]hsCRP
−0·08 (−0·25 to 0·08)
−0·03 (−0·19 to 0·13)
Other studies have investigated the relation between plasma ferritin and (high molecular weight) adiponectin levels in the general population and in patients with diabetes without providing a potential explanation [30-32]. Two of these studies showed an inverse relation between plasma ferritin and adiponectin levels without adjustment for potential confounders and are therefore in agreement with our findings in patients with manifest vascular disease without the presence of the metabolic syndrome [30, 31]. Plasma ferritin and adiponectin levels are inversely related in healthy subjects after adjustment for age, body mass index, fasting blood glucose and hsCRP . As the results of that study are based on a population study, this population differs greatly from the study populations in the present study with manifest vascular disease.
Recently, Gabrielsen et al. [.] investigated the relation between iron metabolism and adipose tissue function and found a negative relation between plasma ferritin and adiponectin levels. In accordance with our findings, however, this relation was not statistically significant when adjusted for gender and hsCRP. Furthermore, while we detected no effect of free iron incubation on adiponectin expression, these authors found a lower adiponectin release after stimulation using a free iron concentration similar to our studies. Possible explanations for this discrepancy include the cellular model system used (murine 3T3-L1 adipocytes vs. human SGBS adipocytes) and the duration of incubation with iron (12 vs. 24 h). Earlier studies have investigated the effect of transferrin and iron on murine adipocytes isolated from epididymal fat pads, which is a murine model for human visceral adipose tissue [34, 35]. In these studies, iron and transferrin induced a small reduction in insulin sensitivity in murine adipocytes . Both transferrin and free iron were able to induce increased lipolysis in murine adipocytes, an effect which could be blocked by co-incubation with N-acetyl-L-cysteine, a known free radical scavenger. In the present study, we showed induction of inflammation in pre-adipocyte cultures which could be blocked with iron chelators and radical scavengers. In contrast, iron did not induce an inflammatory response in adipocyte cultures. Our study results in adipocyte cultures argue against the findings from both murine and human studies in which iron store depletion reduced oxidative stress and inflammation [10, 36]. However, these studies have been conducted in in vivo models with complete adipose tissue, consisting of adipocytes and the stromal vascular cell fraction (SVC), which contains both pre-adipocytes and immune cells . It is therefore conceivable that free iron does not directly induce inflammation and lower adiponectin secretion from adipocytes but may induce adipose tissue dysfunction, as measured by lower adiponectin secretion, indirectly through the pro-inflammatory effect on both pre-adipocytes and immune cells . Future studies employing co-cultures of adipocytes and pre-adipocytes or ex-vivo experiments with adipose tissue explants may clarify this.
Our findings add to the body of evidence showing that the effect of reducing body iron stores on inflammation is most likely due to a reduction in the Fe(II)-catalysed Haber–Weiss reaction producing ROS and subsequent inflammation in pre-adipocytes . Hydroxylradicals are a product of the Haber–Weiss reaction and induce NFκB translocation facilitating protein synthesis, such as IL-6. In the light of these findings, the absence of a relation between plasma ferritin and adiponectin levels in patients with manifest vascular disease is most likely due to an absence of a clear negative relation between plasma ferritin and iron levels in this specific population. Although we adjusted for hsCRP in the analyses, the most likely explanation is that in patients with manifest vascular disease, which is characterized by inflammation , plasma ferritin levels are highly dependent on circulating plasma IL-6 levels . The response of the body to sequester iron during inflammation in the form of ferritin would limit the amount of free iron significantly and would make a relation between ferritin as a marker of body iron stores and free iron-induced adipose tissue dysfunction less likely. This notion is supported by our findings that the most negative relation between plasma ferritin and adiponectin levels was found in patients with manifest vascular disease in the absence of the metabolic syndrome, a condition characterized by inflammation [27, 40]. It is therefore perceivable that the metabolic changes in patients with the metabolic syndrome or those patients at very high risk of new cardiovascular events disturb the relation between plasma ferritin and adiponectin levels.
One strength of this study is the cross-sectional design in a relevant patient category with a high risk of diabetes in which further understanding of this risk is necessary . A further strength is the combination of a clinical study in combination with in vitro studies, investigating a potential causal relation. Study limitations should also be considered. Although plasma ferritin levels are a good representation of body iron stores in healthy individuals , in patients with inflammation, as seen in vascular diseases, other markers such as soluble transferrin receptor levels have been proposed as a better representation of body iron stores [42, 43]. Furthermore, we estimated the concentration of free iron in adipose tissue for the concentrations of Fe(III)-citrate used in the cell culture experiments. Due to the need for stratification, as a consequence of significant interaction with the future development of new cardiovascular events and the metabolic syndrome, the size of the respective strata was small, and we were unable to perform subanalyses in patients with different types of vascular disease. As we performed stratified analyses, the statistical power was limited. Although plasma adiponectin levels are widely used to quantify adipose tissue function, other adipocytokines such as leptin and resistin are readily measureable. Adiponectin levels solely may not reflect adipose tissue function completely. Plasma insulin levels were not available for further interaction analyses into the role of insulin resistance. Finally, we could not adjust for the possible confounding effects of (over-the-counter) use of omega-3 supplements or vitamins as these data were not available. Future studies should be performed in larger study populations to better assess the relation between body iron stores and adiponectin in the different patient populations.
In conclusion, in patients with manifest vascular disease, there is no relation between plasma ferritin and adiponectin levels. However, in vitro, free iron induces an inflammatory response in pre-adipocytes, but not in adipocytes, which was blocked by co-incubation with iron chelators or radical scavengers. Adiponectin secretion by adipocytes was not influenced by free iron. These findings indicate that free iron may indeed induce adipose tissue dysfunction indirectly through pro-inflammatory effects on pre-adipocytes. Whether this is also relevant in patients with vascular diseases in which the existing pro-inflammatory state has secured body iron stores in the form of ferritin is uncertain.
The authors have nothing to disclose.
Department of Vascular Medicine, University Medical Center Utrecht, Heidelberglaan 100, 3584 CX Utrecht, the Netherlands (J. Westerink, D. R. Faber, H. Monajemi, F. L. J. Visseren); Department of Vascular Medicine, Tweesteden Hospital, Kasteellaan 2, 5141 BM Waalwijk, the Netherlands (J. K. Olijhoek); Department of Metabolic Diseases, University Medical Center Utrecht, Heidelberglaan 100, 3584 CX Utrecht, the Netherlands (A. Koppen, E. Kalkhoven); Department of Internal Medicine, University Medical Center Utrecht, Heidelberglaan 100, 3584 CX Utrecht, the Netherlands (B. S. van Asbeck); Julius Center for Health sciences and Primary care, University Medical Center Utrecht, Universiteitsweg 100, 3584 CG Utrecht, the Netherlands (Y. van der Graaf).