Leptin Receptor-Deficient db/db Mice Show Significant Heterogeneity in Response to High Non-heme Iron Diet

Recent studies have shown an association between iron homeostasis, obesity and diabetes. In this work, we investigated the differences in the metabolic status and inflammation in liver, pancreas and visceral adipose tissue of leptin receptor-deficient db/db mice dependent on high iron concentration diet. 3-month-old male BKS-Leprdb/db/JOrlRj (db/db) mice were divided into two groups, which were fed with different diets containing high iron (29 g/kg, n = 57) or standard iron (0.178 g/kg; n = 42) concentrations for 4 months. As anticipated, standard iron-fed db/db mice developed obesity and diabetes. However, high iron-fed mice exhibited a wide heterogeneity. By dividing into two subgroups at the diabetes level, non-diabetic subgroup 1 (<13.5 mmol/l, n = 30) significantly differed from diabetic subgroup two (>13.5 mmol/l, n = 27). Blood glucose concentration, HbA1c value, inflammation markers interleukin six and tumor necrosis factor α and heme oxygenase one in visceral adipose tissue were reduced in subgroup one compared to subgroup two. In contrast, body weight, C-peptide, serum insulin and serum iron concentrations, pancreatic islet and signal ratio as well as cholesterol, LDL and HDL levels were enhanced in subgroup one. While these significant differences require further studies and explanation, our results might also explain the often-contradictory results of the metabolic studies with db/db mice.


INTRODUCTION
Obesity is one of the major metabolic disorders today and an important risk factor for the development of type 2 diabetes (T2D) (1,2). In 2016, more than 1.9 billion adults were overweight and among those 650 million had obesity. Noteworthy, in 2017, ∼ 462 million individuals were affected by T2D corresponding to 6.28% of the world's population and over one million deaths per year can be attributed to diabetes alone (WHO) (1,3). Despite recent advances in our understanding how obesity develops and contributes to metabolic disorders, the precise pathomechanisms are not entirely clear (1,2). In particular, it remains open why only a subgroup of people with obesity and T2D develop complications. Rodent models may help to identify the root causes and pathogenic factors underlying the large human variation obesity related diseases and T2D complications. The leptin-receptor-deficient C57BLKS (BKS) inbred mouse (db/db) strain represents one of the first animal models of T2D (4). This strain is frequently used to study T2D mechanisms because of the many similarities to the development of human obesity-related T2D. Development of diabetes in db/db mice seems to be mediated by the interaction between the leptin receptor deficiency and additional genetic factors (5)(6)(7)(8)(9). T2D in db/db mice is associated with obesity and is characterized by insulin resistance as an initial alteration which subsequently leads to pancreatic ß-cell failure (10).
Recently, an association between iron homeostasis, obesity and T2D has been shown, but the link between dietary iron and metabolic dysfunction is poorly defined (11)(12)(13). Iron plays an important role in many biological functions such as oxygen binding and transport, oxygen metabolism and cell development (14). Iron is also a co-factor for heme and non-heme-dependent enzymes and it is involved in mitochondrial energy generation (14,15). Free iron increases formation of reactive oxygen species (ROS) via Fenton reaction (16). In turn, increased ROS levels could induce degradation of intracellular proteins, lipids and DNA (17). Interestingly, iron overload in pancreatic ß-cells causes cell damage and apoptosis, leading to impaired insulin synthesis and secretion (18). Apoptosis of pancreatic ß-cells, liver dysfunction as well as peripheral insulin resistance promote the development of T2D (16)(17)(18). Notably, the uptake of hemebound iron, but not of non-heme iron, is positively associated with an increased risk of T2D (11,12,19). Moreover, elevated free iron concentrations, accumulation of iron-related markers such as ferritin and transferrin and the expression of heme oxygenase 1 (Hmox1) in adipose tissue of mice and humans correspond with higher inflammatory process and lower adipogenesis (20)(21)(22)(23)(24)(25)(26). High iron burden in adipose tissue induces insulin resistance, but the mechanisms underlying adipose iron accumulation remain also unknown (13).
In our recent work, we investigated the role of iron in the development of peripheral diabetic neuropathy, one of the most common complications of T2D. We found that dietary iron deficiency, but not non-heme iron overload, leads to peripheral nerve dysfunction in obese diabetic db/db mice (27). In this study, the positive effect of high non-heme iron diet on nerve conduction velocities and reduced pro-inflammatory milieu have been shown in sciatic nerves of these animals (27). Generally, db/db mice fed with high iron diet had a lower blood glucose level and HbA1c content as well as higher serum insulin Abbreviations: ROS, reactive oxygen species; T2D, type 2 diabetes; Hmox1, hem oxygenase 1; TNFα, tumor necrosis factor α; Il, interleukine; ATMs, adipose tissue macrophages; Fth, ferritin heavy chain; Ftl, ferritin light chain; Fpn, ferroportin; Ccl2, chemokine ligand 2; Ccl20, chemokine ligand 20; Cdh 6, cadherin-6; Clmp, CXADR like membrane protein; DLK1, Delta like non-canonical Notch ligand 1; ErbB4, Erb-b2 receptor tyrosine kinase 4; Fslt3, Follistatin like 3; Sez6l2, Seizure related 6 homolog like 2; vAT, visceral adipose tissue; LDL, high-densitylipoprotein-cholesterol; HDL, low-density-lipoprotein-cholesterol; Aif1, allograft inflammatory factor 1; Dmt1, divalent metal transporter 1; Tfr1, transferrin receptor 1; ACAM, adipocyte adhesion molecule. concentrations (27). Surprisingly, a detailed analysis of glucose, lipid and iron metabolic parameters revealed two subgroups of db/db mice fed with high non-heme iron diet. Recently, our research group has investigated the influence of dietary non-heme iron on the development of peripheral neurological complications in the metabolic syndrome [leptin-deficient ob/ob mice, (28)], streptozotocin (STZ)-induced type 1 [STZ-rats, (29)] and type 2 [db/db mice, (27)] diabetes mellitus. In these studies, the most important criterion was blood glucose level as a determinant of the diabetic status of the experimental animals. It is generally known that hyperglycemia (blood glucose concentration >13.5 mmol/l) is one of the most important factors in the development of peripheral neuropathy. In all of our studies to date, the same high iron diet was used and, with the exception of the last work with db/db mice, no significant differences were found in response to high dietary iron levels within one animal group (27)(28)(29). Therefore, it was surprising to observe the dual response of db/db diabetic mice to high iron diet (27). Notably, reports on the effect of iron on the patient diabetic status and development of the diabetic complications are contradictory (11,13). In the present work, we aim to investigate differences in the metabolic status, as well as inflammation markers in liver and adipose tissue between two high iron diet fed db/db mouse subgroups. The detailed characterization of these two subpopulations of db/db mice reacting differently to dietary iron might help to explain the different effects of this nutrient on individual patients and to choose the appropriate adjuvant therapy against iron homeostasis disorders.

Animals and Phenotypical Analyses
3-month-old male BKS-Leprdb/db/JOrlRj (db/db) mice, obtained from the Janvier Labs (Saint Berthevin Cedex, France), were used in this study. The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Landesdirektion Sachsen, Leipzig, the local authority for animal care (reg. no.: TVV65/15). Animals were divided into two groups of db/db mice, which were fed with different chows (Altromin, Lage, Germany) containing high iron (29 g/kg -standard diet modified with 3% carbonyl iron) or standard iron (0.178 g/kg) concentrations for 4 months. In our previous neurological work, we used db/db mice fed with different amounts of dietary nonheme iron for the first time (27). Iron is an essential element, but it is known that some forms of dietary iron could be toxic (30)(31)(32). To exclude the toxic effect of dietary carbonyl iron on the mice used in this study, we also tested db/+ mice as an additional control. All db/+ mice fed with high iron diet were normoglycemic and they could not be divided into subgroups of high and low glucose. Despite the high level of serum iron, they did not show any signs of metabolic and hematopoietic disorders and all investigated metabolic parameters in this control group were similar to that observed in the standard and low iron diet fed db/+ control groups (27). In the present study, we fed mice with the same high iron diet. Since the high iron fed db/+ mice have not shown any differences in blood glucose level, they have been excluded as an additional control from the present experiment. The experiment was repeated six times. Body weight, blood glucose concentration and HbA1c were measured as described (27). Lipid status and serum iron levels were performed by the Institute of Laboratory Medicine, Clinical Chemistry and Molecular Diagnostics, University of Leipzig. Serum insulin (Mouse Ultrasensitive Insulin ELISA, Alpco Diagnostics, Salem, NH), C-peptide (Mouse Ultrasensitive Cpeptide ELISA, Alpco Diagnostics, Salem, NH) and adiponectin [Adiponectin (mouse) ELISA Kit, AdipoGen R LIFE SCIENCES, Liestal, Switzerland] concentrations were detected by ELISA according to the manufacturer's protocol. Serum concentrations of Chemokine ligand 2 (CCL2), Chemokine ligand 20 (CCL20), Cadherin-6 (Cdh6), Coxsackie virus and adenovirus receptorlike membrane protein (Clmp), Delta like non-canonical notch ligand 1 (DLK1), Erb-b2 receptor tyrosine kinase 4 (ErbB4), Follistatin like 3 (Fstl3) and Seizure related six homolog-like two (Sez16) were determined by mouse OLINK proteomics (Uppsala, Sweden). The first four experiments were performed as a part of previous neurological studies and parts of the values have already been used, but without dividing of the high iron group into two subgroups (body weight, blood glucose concentration, HbA1c, serum insulin, serum iron levels, see Figures 1A-G) (27).

Immunostaining and Quantification of Pancreatic Islet ß-Cells
Pancreatic tissue was fixed with 4 % paraformaldehyde, embedded within paraffin and sliced into 7 µm sections. Immunostaining was performed as previously described (27). Three slides each containing three slices of five animals per group were analyzed. The mice used for this analysis did not differ from the other animals in the given subgroup in terms of body weight, blood glucose and iron levels. Slices were incubated with insulin antibody (1:2,000, Abcam) overnight at 4 • C and secondary antibody Alexa 488 goat anti-rabbit (1:500) for 1 h at room temperature. Nuclei staining was performed with DAPI (1:10,000, Serva).
Immunofluorescence stained microscope slides were fully digitized at 20x magnification using a digital slide scanner (Pannoramic Scan II, 3D HISTECH Ltd., Budapest, Hungary) equipped with a quad band (DAPI/FITC/TRITC/Cy5) filter set. DAPI channel (blue) represents nuclei staining, FITC channel (green) represents insulin staining and TRITC channel (red) represents tissue autofluorescence. Images of the stained tissue slices were exported from data sets using CaseViewer (Version 2.3, 3D HISTECH Ltd., Budapest, Hungary) with pixel dimensions of 0.325 µm.
Image analysis was performed with Mathematica (Version 12.0, Wolfram Research, Inc., Champaign, IL, USA). Fluorescence images were imported and adjusted for brightness and contrast. Tissue area was segmented using Otsu's (cluster variance maximization) thresholding method (33), small disconnected tissue segments were removed and tissue artifacts like small cuts or fissures were corrected using a five pixel wide morphological closing filter, an operation which dilates the segmented tissue area by five pixels (enlarges the tissue area and closes gaps) and subsequently erodes the tissue by five pixels (reduces the tissue area but keeps gaps closed). Imported images were split into separate color channels. While tissue autofluorescence is negligible in the blue channel it can appear as bright as stained islets in the green channel. However, the autofluorescence is not limited to the green channel only as it also appears in the red channel with comparable intensity. Consequently, autofluorescence correction was performed by subtracting the red from the green channel. Supplemental figure shows the difference in segmented islets between original green channel and channel subtraction (green minus red). While detection results from the green channel (Supplementary Figures 1E,F) show many components or component appendages originating from unspecific staining (i.e. autofluorescence), only specifically stained and properly delimited islets remain in the detection results from channel subtraction (Supplementary Figures 1G,H). This comparison underlines the need for an autofluorescence correction method of the green channel and in this study we used fluorescence information already present in the red channel. The resulting images show minimal residue of autofluorescence along with brightly stained islets that can be easily detected. Pancreatic islets in the corrected green channel were also segmented using Otsu's thresholding method and subsequently enhanced by morphological closing since uncorrected islets may appear slightly porous due to the staining of individual islet cells. Segmented islets were visually examined and image masks for the removal (staining artifacts, tissue overlaps, blurry regions) or adding (poorly stained islets or islet regions) of structures were created by hand using GIMP (Version 2.10.2, The GIMP team, http://www.gimp.org) when necessary. These masks were subsequently imported in Mathematica and the respective islet segmentations were corrected. Based on the final segmentations, tissue and islet areas were calculated. Fluorescence intensity values of masked islets were accumulated both for high-intensity as well as low-intensity portions of all segmented islets. Finally, all values were normalized based on the tissue area of the respective images.

RNA Extraction and cDNA Synthesis
Liver and visceral adipose tissues (vAT) were stored at −80 • C. The mice used for this analysis did not differ from the other animals in the given subgroup in terms of body weight, blood glucose and iron levels. Tissues were transferred into 1,000 µl Trizol (Invitrogen), homogenized with Ultra Turax (IKA), centrifuged and the supernatant was collected. 200 µl Chloroform (Roth) was added, centrifuged and rinsed with ice cold isopropanol and ethanol. Pellet was solved in 30 µl RNasefree water. DNase digestion was carried out with TURBO DNAfree Kit (Invitrogen). RNA concentration was measured with a Qubit 3.0 Fluorometer (Invitrogen) and RNA BR Assay kit (Invitrogen). cDNA-Synthesis was performed with ProtoScript First Standard cDNA Synthesis Kit (BioLabs).
Quantitative RT-PCR with SYBR Green qPCR Master Mix (Thermo scientific) was performed using the CFX96 Real-Time System (Bio-Rad). mRNA expression of target genes was normalized to the expression of IPO8, ATCB and B2M mRNA. Values of the high iron fed subgroup 1 were compared with the

Hematoxylin-Eosin and Prussian Blue Staining
Liver and pancreatic paraffin slides were incubated with hematoxylin for 10 min for HE staining or with 2% potassium ferrocyanide and 2% hydrochloric acid for 30 min at room temperature for Prussian blue staining. Eosin or nuclear fast red were used for counterstaining. Stained microscope slides were fully digitized at 20x magnification using a digital slide scanner (Pannoramic Scan II, 3D HISTECH Ltd., Budapest, Hungary).

Statistical Analysis
Data are presented as mean ± SD. Differences between the groups were validated by a two-tailed unpaired Student's t-test or one-way-ANOVA and the Tukey test for the adoption of a normal distribution or Kruskal-Wallis and Mann-Whitney for non-parametric test using GraphPad Prism Eight Values of p < 0.05 were considered as statistically significant.

Phenotypic Characterization of High Iron Group and Development of Diabetic Alterations
Based on blood glucose concentrations after 4 months high iron diet, we divided this group of db/db mice into two subgroups: Subgroup 1 (Sub1) with blood glucose level lower than 13.5 mmol/l and subgroup 2 (Sub2) with blood glucose concentration equal to or higher than 13.5 mmol/l (Supplementary Figure 2). Notably, we did not observe such differences in the standard group (Supplementary Figure 2). Only two out of 42 animals showed a blood glucose level lower than 13.5 mmol/l (Supplementary Figure 2). At the start of the experiment there were no significant differences in blood glucose levels between the groups (Supplementary Figure 3). In Sub1, a significant decrease of blood glucose concentrations was observed during the experiment while in Sub2, blood glucose levels increased similarly to those observed in the standard group (Figures 1A-C). Surprisingly, body weight of Sub1 animals increased from 50 g (±3.5 g) to 59.4 (±5 g) while in Sub2 and in mice fed with standard iron diet, it decreased from 49.7 g (±3.5 g) to 47.2 (±8.5 g) and from 50 g (±4.3 g) to 46.6 g (±8.9 g), respectively (Figures 1B,D). In addition to blood glucose levels, the HbA1c content in Sub1 was significantly lower (6,3 % ± 2 %) as compared to Sub2 (11.97 % ± 2.6 %) and the db/db mouse standard group (9.8 % ± 2.8 %) (Figure 1E). Notably, serum insulin concentrations correlated with these findings and were increased in Sub1 (7.6 ng/ml ± 3.4 ng/ml) compared to Sub2 (1.4 ng/ml ± 1.2 ng/ml) as well as the standard group (2.9 ng/l ± 3.6 ng/l) (Figure 1F). Serum iron concentration of Sub1 was significantly higher (97.8 mmol/l ± 19.7 mmol/l) than in Sub2 (64.1 mmol/l ± 18.0 mmol/l). As expected, Sub2 showed a higher serum iron concentration compared to the db/db mice on standard diet (36.6 mmol/l ± 8.7 mmol/l) (Figure 1G).
In addition to the increased serum insulin concentrations, serum C-peptide levels in Sub1 were also higher than in Sub2 and standard group ( Figure 1H). Furthermore, serum levels of adiponectin were comparable in subgroup 1, subgroup 2 and standard group (Figure 1I). Next, we analyzed the serum lipid status of high iron Sub1 and Sub2 and standard group. Parallel to increased body weight and serum insulin concentration, the serum total cholesterol level and the concentrations of its fractions (high-density-lipoprotein-cholesterol (LDL) and lowdensity-lipoprotein-cholesterol (HDL) were significantly higher in Sub1 compared to the Sub2 and standard group (Figure 1J). Notably, there were no differences in triglyceride and free fatty acid concentrations between both animal subgroups (data not shown).
Due to the lack of division into two subgroups of standard iron diet fed animals, which do not differ in blood glucose levels, we have only analyzed the differences between Sub1 and Sub2 of the high iron fed db/db mice in the further part of the present study.
The mouse OLINK proteomic panel analysis was carried out to examine the levels of serum circulating proteins, which play an important role in glucose and lipid metabolism as well as in inflammation. Among 91 tested proteins, only Eight were significantly different between both high iron subgroups ( Table 1). We found higher levels of Ccl2, Cdh 6, Clmp, DLK1, Fstl3, and Sez6l2 in Sub1 compared to Sub2. Serum concentrations of Ccl20 and ErbB4 were significantly lower in Sub1 compared to Sub2 ( Table 1).

Quantification of Pancreatic Islet Area of ß-Cells
Previously, it has been shown that iron increases the apoptosis of pancreatic ß cells by inducing the production of ROS, which in turn promotes the development of diabetes mellitus (16)(17)(18). For this reason, we examined the pancreas morphology of db/db mice fed with high iron diet. An automatic quantification of pancreatic ß-cell islets was used to analyze differences in the signal ratio of insulin and average islet size (Figure 2). The analysis revealed a significant higher islet and signal ratio of insulin in pancreas of Sub1 compared to Sub2 (Figures 2A,B). Also, the average islet size was increased in Sub1 compared to Sub2 (Figure 2C). Values represent means ± SD * p < 0.05, ** p < 0.01, *** p < 0.001 (t-test). The mice used for this analysis did not differ from the other animals in the given subgroup in terms of body weight, blood glucose and iron levels.

Markers of Inflammation and Macrophages in Liver and Visceral Adipose Tissue
The association between obesity, T2D and chronic tissue inflammation with infiltration of immune cells and increased expression of inflammatory markers has been shown in several previous studies (34,35). Notably, macrophages, liver and visceral adipose tissue (vAT) play a significant role in systemic iron homeostasis (36). Therefore, we analyzed different inflammation and macrophage markers in liver and vAT after 4 months of high iron diet using RT-PCR. Relative gene expression analyses of pro-inflammatory markers Il1ß, Il6, and Tnfα showed no differences in liver between both subgroups. In vAT, a significant lower expression of Il6 and Tnfα was observed in Sub1 compared to Sub2. Whereas the expression of an antiinflammatory marker Il10 in both tissues was similar in the subgroups (Figure 3A). Relative gene expression of macrophage marker allograft inflammatory factor 1 (Aif1) in liver as well as in vAT was comparable in both high iron subgroups. In both tissues, pro-inflammatory M1 macrophage markers (CD11c and CD11b) and anti-inflammatory M2 macrophage marker CD206 were equally expressed in both subgroups ( Figure 3B). Relative gene expression analyses of adiponectin and leptin in the vAT revealed no differences in the two subgroups ( Figure 3D).

Markers of Iron Metabolism in Liver and vAT
Liver is one of the main iron storage tissue, but iron can also be stored in vAT. Therefore, we investigated the main iron storage protein ferritin in both tissues using RT-PCR. Analysis of Fth and Ftl (ferritin heavy chain and ferritin light chain) gene expressions showed no differences in both subgroups. Next, we examined other iron metabolism markers such as divalent metal transporter 1 (DMT1), ferroportin (Fpn) and transferrin receptor 1 (Tfr1) in the liver as well as vAT. No differences in mRNA expression of the tested proteins were observed between two high iron subgroups. Only the heme oxygenase1 (Hmox1) level was significant reduced in the vAT of the Sub1 compared to the Sub2 ( Figure 3C). Moreover, in the gut of both subgroups, we did not find any differences in mRNA expression of iron metabolism markers (data not shown). Then, we performed Prussian blue staining for visualization of iron in liver and pancreas. In the pancreas there were no detectable deposits of iron in all experimental groups (data not shown). Clusters of iron were visible in the liver tissue in Sub1 and Sub2. Iron deposits were found in sinusoids and hepatocytes with central alignment to the central vein. Analysis of HE-stained slices of all groups revealed an accumulation of fat in the liver (Figure 4).

Correlation Between Glucose Metabolic Parameters, Pancreas Morphology and the Serum Iron Status
The principal finding of our study is that db/db mice fed for 4 months with high iron diet can be divided into two metabolic subgroups ( Table 2). Our results of biochemical and morphological analyses clearly show a relationship between the improvement of glucose metabolic parameters and an increased volume of pancreatic islets with a simultaneous significant increase in insulin and C-peptide synthesis in Sub1. In context of these observations, the remarkably interesting finding of this study is increased circulating Sez6l2 protein levels in Sub1 mice. Sez6l2 occurs as a pancreatic islet-enriched cell surface protein and is expressed in endocrine progenitor islet cells during pancreatic development, although its role in islet cell division is still unknown (37). The increased level of this protein is associated with the increasing volume of the pancreatic islets in Sub1. The natural consequence of the above-mentioned changes was a reduction by half of the glucose level in this experimental group compared to the other mice. Notably, the highest level of serum iron has been also observed in Sub1, which indicates a possible direct or indirect protective effect of this essential but also potentially dangerous micronutrient on the ß-cells. Contrary to the present observations, Cooksey et al. have shown that dietary iron restriction or iron chelation protect from loss of β-cell function and diabetes in obese, leptin-deficient (ob/ob) mice (18). In their study, mice had hyperglycemia and were fed with high carbohydrate or highfat diet with different iron content. Unfortunately, it could not be excluded here that the high carbohydrate or the high-fat diets were additional factors that might influence insulin and glucose metabolism and β-cell function of these leptin-deficient animals (18).

Improved T2D Parameters and Body Weight Gain
An unexpected effect of the improvement of glucose metabolic parameters was a significant increase in body weight in Sub1 animals. It should be noted, however, that similar changes were observed in patients with too high doses of insulin (38). This may explain the increase in body weight in Sub1 mice, in whom insulin and C-peptide levels were significantly higher compared to the Sub2 animals. An interesting finding in the context of the above-mentioned results is the increased level of serum Clmp concentration observed in Sub1 mice. Clmp, also known as adipocyte adhesion molecule (ACAM), is a part of the epithelial tight junctions playing a role in cellcell adhesion in white adipose tissue. Clmp is upregulated in adipocyte maturation in humans and rodents with obesity (39). Interestingly, the study of Murakami et al. with transgenic ACAM mice has shown a reduction in adipocyte mass with smaller adipocytes and a protection against obesity and diabetes (40). In contrast to these observations, the increased level of serum ACAM in the Sub1 mice correlated with a weight gain. The explanation of this phenomenon requires further investigations. Additionally, we found higher serum Fstl3 and DLK1 levels in Sub1 animals. Fstl3 shows glucose regulatory effects in rodents and humans (41,42). Fstl3 is expressed and secreted in skeletal muscle. An overexpression of Fstl3 in mice fed with high-fat diet leads to fat gain and an improvement in insulin signaling (43)(44)(45). DLK1 (also called preadipocyte factor one) also plays a role in adipose tissue homeostasis  Values represent means ± SD *p < 0.05, **p < 0.01 (t-test, Mann-Whitney test).  Differences in biochemical and physiological parameters in serum, pancreas and vAT between the two subgroups. A single arrow -moderate differences; a double arrow -greater differences in subgroup 1 compared to subgroup 2.
and it is an inhibitor of adipogenesis by preventing adipocyte differentiation (46) as well as regulation of the whole-body glucose metabolism (47,48). Surprisingly, we found lower serum concentration of ErbB4 in Sub1 combined with an improvement in several biochemical parameters of diabetes. ErbB4 is an important factor for cell proliferation, differentiation, migration and apoptosis (49). It is induced by different ligands such as neuroregulin. These ligands play an important role in glucose transport and lipogenesis (50,51). ErbB4 deletion in mice fed with a moderate-fat diet leads to the development of metabolic syndrome with obesity, hyperglycemia, insulin resistance and hepatic steatosis (52). Further studies of adipocyte mass, size and physiology in accordance with the iron status are necessary to explain the exact role of the proteins mentioned here in the improvement of the diabetic status, the weight gain as well as the iron homeostasis in Sub1 and Sub2 animals. The increased body weight of mice from Sub1 was accompanied by an increased level of cholesterol and its fractions, which is consistent with other studies published so far (53).
Handa at el. examined male B6.BKS (D)-Leprdb/J diabetic mice fed with 2% carbonyl iron. They found hepatocellular ballooning in liver of diabetic iron-fed mice compared to control mice, which is a characteristic of human non-alcoholic steatohepatitis. Hepatic ballooning was combined with increased oxidative stress and inflammation as well as immune cell activation in liver of iron-fed mice. In accordance with our results, they also found higher serum cholesterol, LDL and HDL concentrations in serum of the high iron diet-fed db/db mice (54).
Interestingly, in another study, the opposite results regarding body weight in db/db mice fed with high iron diet have been demonstrated (32). However, in this experiment, the ferrous sulfate was used to enrich the diet with iron (32). Unlike this work, we used carbonyl iron. It is generally known that the ferrous sulfate is more toxic than carbonyl iron (30,31). For this reason, it is impossible to exclude a toxic effect of ferrous sulfate in the applied dose, which would explain the loss of body weight in these high iron db/db animals (32).

Body Weight Gain, Inflammatory Markers and Serum Iron Status
It is widely accepted that a major increase in adipose mass contributes to adipose tissue dysfunction. This dysfunction promotes metabolic disorders via a mild, chronic inflammation which is characterized by an increased expression of proinflammatory factors such as leptin, TNFα, IL6 or monocyte chemoattractant protein-1 (MCP-1 -also known as CCL2) (55). Interestingly, in the present study Tnfα and Il6 expression levels were the lowest in the vAT of Sub1 animals with higher body weight. Simultaneously, the serum iron concentrations were the highest in mice of this experimental group. Previously, we have shown that elevated serum iron levels are associated with lower Tnfα expression as well as decreased infiltration and activation of inflammatory cells in peripheral nerves, and consequently alleviation of the neuropathic symptoms in obese ob/ob and in obese and diabetic db/db mice (27,28). Notably, the panel protein analysis revealed significant lower CCL20 serum concentrations in Sub1 mice. CCL20 stimulates the migration of dendritic cells as well as T and B cells (56) and it is upregulated by pathogenic bacteria and inflammatory cytokines such as Il1ß and TNFα (57)(58)(59). Moreover, it has been shown that an iron chelator deferoxamine (DFO) could induce the production of CCL20 in human intestinal epithelial cells (60). The observed decrease of CCL20 level in Sub1 mice correlates with lower Tnfα and Il6 expressions as well as higher serum iron concentration and it is consistent with previously published results. The obtained results clearly suggest a relationship between serum iron levels and vAT inflammation, glucose metabolism and body weight in both mouse subgroups fed with high iron diet.
Visceral fat plays a significant role in the regulation of iron homeostasis. Moreover, it is known as the main determinant of insulin resistance, and it is involved in the regulation of inflammation state through proinflammatory molecule production (36). For these reasons, we focused on changes in the visceral fat of db/db mice. Interestingly, Dongiovanni et al. have shown a decrease in visceral fat mass and no change in subcutaneous fat in mice fed a similar diet as in our study. In opposition to our results, the blood glucose levels were significantly higher in mice fed with high-iron diet (20). It should be noted here, however, that in the study by Dongiovanni et al., healthy, non-obese, wild-type mice (C57Bl/6) were used, which were also 6 weeks younger than our obese db/db animals at the beginning of the experiment. These observations confirm an important link between fat mass as well as iron and glucose metabolism, but a thorough explanation of the mechanisms of this interaction remains an open question that requires further studies including also subcutaneous fat.

Iron Homeostasis Markers
All mice of similar genetic background fed with high iron diet kept in separate cages with free access to food have similar leptin levels but show differences in serum iron level concentrations between the two subgroups. To elucidate that in more detail, we analyzed markers of iron metabolism such as Dmt1, Fpn, Tfr, Fth and Ftl in liver, vAT as well as gut. We found no differences between the investigated animal groups. Based on these mRNA expression data we performed Prussian blue staining to detect iron in liver and pancreas. In pancreas there were no visible iron deposits in both subgroups. Accumulation of iron was found in livers of Sub1 and Sub2, but the amount was comparable. Taken together, observed differences in circulating iron levels in both subgroups cannot be explained by parameters of iron metabolism or iron storage capacity in the liver, vAT as well as in the gut.
Another important enzyme playing a crucial role in iron homeostasis is Hmox1. Handa et al. have previously shown a significant higher Hmox1 expression in the liver of db/db mice fed with high iron diet (54). Our analyses of Hmox1 expression show a significant reduction in vAT of Sub1 compared to Sub2 but no differences were observed in the liver. Hmox1 is induced by oxidative stressors such as ROS or pro-inflammatory cytokines (61,62). The reports of Hmox1 in relation to disorders of glucose and lipid metabolism in connection with inflammation and iron homeostasis are unfortunately not conclusive. On the one hand, Hmox1 induction in parallel with an increase in serum adiponectin levels decreased pro-inflammatory markers such as TNFα, Il6 and Il1ß in obese mice (63). Moreover, Nicolai et al. (2009) showed that Hmox1 induction improved insulin sensitivity and decreased adiposity in Zucker diabetic rats (64). On the other hand, it has been shown that an increase in Hmox1 activity by heme results in decreased glucose uptake and increased markers of inflammation, oxidative stress and iron accumulation in human adipocytes (11,24,65). Here we found a decreased Hmox1 expression in vAT in combination with reduced expression of pro-inflammatory markers such as Tnfα and Il6 and CCL20 in Sub1. In contrast, a higher expression of Hmox1 in vAT was observed in Sub2 in combination with higher pro-inflammatory marker levels. Further metabolic and molecular studies are required to explain these differences in serum iron levels between both subgroups fed with high iron diet.

CONCLUSION
In this study, we demonstrated the presence of two subpopulations of db/db mice within the high iron-fed group. These subpopulations differ in serum iron concentrations, pancreatic morphology, glucose, insulin, C-peptide levels, expressions of inflammatory markers, and body weight. These differences indicate a significant heterogeneity within this mouse strain especially in respect to glucose, lipid and iron metabolism and might be associated with impaired function of the pancreatic ß-cells. Moreover, we cannot exclude an impact of factors such as differences in food intake or absorption, metabolic rate, spontaneous activity, or individual insulin sensitivity. Further studies are required to explain these observations. These findings may also explain the contradictory results of studies with db/db mice and show how important it is to define the metabolic profile of each tested animal. Based on observations presented in this study and the genetic diversity of patients with T2D, there may also be differences in response to the non-heme