MECHANISMS OF NON-TRANSFERRIN-BOUND IRON UPTAKE BY HUMAN β CELLS AND THE ROLE OF IRON IN DIABETIC PATHOGENESIS

Transcription

1 MECHANISMS OF NON-TRANSFERRIN-BOUND IRON UPTAKE BY HUMAN β CELLS AND THE ROLE OF IRON IN DIABETIC PATHOGENESIS By RICHARD COFFEY A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2016

2 2016 Richard Coffey

3 To my parents

4 ACKNOWLEDGMENTS I would first and foremost like to thank my major advisor Dr. Mitchell Knutson who took a chance and gave me the opportunity to do research all those years ago. I would not be where I am right now if I did not end up in your undergraduate lab techniques class. Also, I would like to thank my past and current labmates Chia-Yu Wang, Supak Jenkitkasemwong, Wei Zhang, Hyeyoung Nam, Lin Zhang, Katie Sullivan, Lizzie Paulus, Laura Diez-Ricote, Nike Akinyode, and Emily Mejia who have either taught me most of what I know about lab work today or have helped me in innumerable other ways. I wish you the best wherever you find yourself in life and know that you will be successful. I would also like to thank my committee members Dr. Clayton Mathews, who has been instrumental in the design and execution of several experiments discussed in this dissertation, Dr. James Collins, and Dr. Michelle Gumz who have all taken time out of their busy lives to help me during this process. Finally I would like to thank my parents who have always given me the opportunity to pursue my interests and have been unconditionally supportive of my decisions. 4

5 TABLE OF CONTENTS 5 page ACKNOWLEDGMENTS... 4 LIST OF TABLES... 7 LIST OF FIGURES... 8 LIST OF ABBREVIATIONS ABSTRACT CHAPTER 1 INTRODUCTION LITERATURE REVIEW Basics of Iron Metabolism General Information Dietary Iron Absorption Disorders of Iron Metabolism Non-Transferrin-Bound Iron Import Proteins The Iron and Diabetes Connection Elevated Iron Stores and Diabetic Pathology Pancreatic Iron Accumulation Mechanistic Evidence from Human Studies Mechanistic Evidence from Animal Studies Iron and Autoimmune Diabetes MATERIALS AND METHODS Animals and Diets Iron Status Parameters and Blood Glucose Concentrations Pancreatic Mineral Concentrations Histological Analysis RNA Isolation and Assessment of RNA Integrity Microarray Analysis Relative mrna Quantification Western Blotting Cell Culture and Treatments In Vitro Glucose Stimulated Insulin Secretion Mouse Islet Isolation Determination of DMT1, ZIP8, and ZIP14 mrna Copy Numbers sirna Knockdown of DMT1, ZIP8, and ZIP

6 Overexpression of DMT1, ZIP8, and ZIP Immunofluorescencse Cellular NTBI Uptake Generation of Transgenic MIP-Zip14-HA Mice Statistical Analysis TRANSCRIPTIONAL PROFILING OF PANCREATIC GENE EXPRESSION IN RESPONSE TO DIETARY IRON LOADING OR DEFICIENCY Results Body Weight, Iron Status, and Blood Glucose Concentrations Pancreatic Mineral Concentrations Identification and Classification of Differentially Expressed Genes by Microarray Analysis Confirmation of Up-Regulation of Alox15 Expression by QRT-PCR and Western Blotting Confirmation of Reg Family mrna Levels by QRT-PCR Discrepancies Between Microarray and QRT-PCR Analysis Results Discussion MECHANISMS OF NTBI UPTAKE BY HUMAN β CELLS Results Overexpression of NTBI Transporters in Human β Cells sirna Knockdown of NTBI Transporters in Human β Cells Cellular Localization of NTBI Transporters in Human Islets Modulation of ZIP14 Expression by Iron in Human β Cells Modulation of ZIP14 Expression By IL-1β in Human β Cells Discussion THE INFLUENCE OF IRON STATUS ON DIABETIC PATHOLOGY AND β- CELL FUNCTION Results Effect of Iron Status on Spontaneous Autoimmune Diabetes in NOD Mice Effect of Dietary Iron on Rate of Growth and Systemic Iron Status Pancreatic Mineral Concentrations Testing of β cell function During the Prediabetic Period Effect of Iron Status on Human Islet GSIS Generation of Mice Selectively Overexpressing Zip14 in β Cells Discussion CONCLUSIONS, LIMITATIONS, AND FUTURE DIRECTIONS LIST OF REFERENCES BIOGRAPHICAL SKETCH

7 LIST OF TABLES Table page 4-1 Body weight, iron status, and blood glucose concentration of rats Pancreatic mineral concentrations Top 10 most up-regulated and down-regulated genes in FeD pancreata Top 10 most up-regulated and down-regulated genes in FeO pancreata Primers for qrt-pcr Functional categories of pancreatic genes differentially expressed in response to iron deficiency Functional categories of pancreatic genes differentially expressed in response to iron overload Iron parameters of type 1 diabetes-prone NOD mice Pancreatic mineral concentrations in NOD mice

8 LIST OF FIGURES Figure page 4-1 Functional classification of pancreatic genes up- or down-regulated in response to iron deficiency and iron overload Effect of iron deficiency and overload on rat pancreatic Alox15 expression Effect of iron deficiency and overload on the expression of pancreatic Reg family genes ZIP14 and ZIP8, but not DMT1, overexpression increases iron uptake by βlox5 cells When overexpressed in βlox5 cells, ZIP14 localizes to the plasma membrane whereas DMT1 mainly localizes intracellularly Endogenous iron uptake by βlox5 cells is decreased by sirna knockdown of ZIP14, but not ZIP sirna knockdown of ZIP14 decreases NTBI uptake by primary human islets ZIP14 is detected in human pancreatic β cells by immunofluorescent analysis Cellular iron levels and treatment with IL-1β increase ZIP14 levels in βlox5 cells but not primary human islets mrna copy numbers of NTBI transporters in primary human islets. qrt- PCR measurement of DMT1, ZIP14, and ZIP8 mrna copy numbers in total RNA isolated from nondiabetic human islets DMT1, but not ZIP8, is detected in human β cells by immunoflourescence staining Dietary iron deficiency, but not iron overload, results in a trend towards an increased incidence of spontaneous diabetes in female NOD mice NOD mice fed an iron-loaded diet initially experience diminished growth Iron stores of mice fed an iron-deficient diet increase with age Glucose tolerance and insulin secretory capacity is not affected by iron status in prediabetic NOD mice Iron-deficient prediabetic NOD mice show no differences in β cell iron status or insulitis compared with iron-adequate mice

9 6-6 Iron status does not affect glucose-stimulated insulin secretion by human islets in vitro Generation of mice selectively overexpressing Zip14 in β cells

10 LIST OF ABBREVIATIONS Alox15 AMPK Arachidonate 15-lipoxygenase 5 adenosine monophosphate-activated protein kinase Apoa1 Apolipoprotein A-1 BMI CCS CRP CS DAVID Body mass index Copper chaperone for superoxide dismutase C-reactive protein Cell surface Database for annotations, visualization, and integrated discovery DCT1 Divalent cation transporter 1 DcytB DFO Duodenal cytochrome B Deferoximine mesylate DMT1 Divalent metal-ion transporter 1 ELISA EV Enzyme-linked immunosorbent assay Empty vector Fabp1 Fatty acid binding protein 1 Fabp2 Fatty acid binding protein 2 FAC FeA FeD Fe-NTA FeO GFP GSIS HA Ferric ammonium citrate Iron adequate Iron deficient Ferric nitrilotriacetate Iron loaded Green fluorescent protein Glucose-stimulated insulin secretion Hemagglutinin antigen tag 10

11 HAMP HETE hgh HJV ICP-MS IgG IL-1β IP-GTT IRE KRB LPS LTCC MIP Mn-SOD Na + /K + ATPase NOD Hepcidin antimicrobial peptide Hydroxyeicosatetraenoic acid Human growth hormone intron region Hemojuvelin Inductively coupled plasma mass spectrometry Immunoglobulin G Interleukin 1β Intraperitoneal glucose tolerance testing Iron response element Kreb s-ringer Buffer Lipopolysaccharide L-type calcium channel Mouse insulin 1 promoter Manganese superoxide dismutase Sodium-potassium adenosine triphosphatase Non-obese diabetic NRAMP2 Natural-resistance associated macrophage protein 2 NTBI PPM qrt-pcr Reg ROS sirna Non-transferrin bound iron Parts per million Quantitative reverse transcriptase polymerase chain reaction Rengenerating islet-derived gene Reactive oxygen speices Small interfering ribonucleic acid SLC11a2 Solute carrier family 11 member 2 SLC39a14 Solute transporter family 39 member 14 11

12 SLC39a8 Solute transporter family 39 member 8 STEAP3 Six-Transmembrane epithelial antigen of prostate family member 3 stfr TBI TCL TF TF sat TFR Soluble transferrin receptor Transferrin bound iron Total cell lysate Transferrin Transferrin saturation Transferrin receptor TFR1 Transferrin receptor 1 TFR2 Transferrin receptor 2 Tg TIBC TTCC UTR Wt Transgenic Total iron binding capacity T-type calcium channel Untranslated region Wild type ZIP14 ZRT/IRT-Like protein 14 ZIP8 ZRT/IRT-Like protein 8 12

13 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy MECHANISMS OF NON-TRANSFERRIN-BOUND IRON UPTAKE BY HUMAN β CELLS AND THE ROLE OF IRON IN DIABETIC PATHOGENESIS Chair: Mitchell D. Knutson Major: Nutritional Sciences By Richard Coffey May 2016 The relationship between iron and diabetes has long been recognized as individuals with iron overload display an increased prevalence of diabetes and iron depletion is thought to protect against the development of diabetes. This link is attributed to the accumulation of iron in beta cells, which may impair cellular function. However, the mechanisms by which beta cells take up iron, as well as the specifics of how iron status affects diabetic pathogenesis, are undetermined. During iron overload plasma iron levels exceed the carrying capacity of transferrin resulting in non-transferrin bound iron (NTBI), which is rapidly taken up by tissues. Currently 3 mammalian proteins which transport NTBI have been identified: Divalent metal-ion transporter 1 (DMT1), ZRT/IRT-Like transporter 14 (ZIP14), and ZRT/IRT-Like transporter 8 (ZIP8). The aims of this project were to determine the contribution of DMT1, ZIP14, and ZIP8 to iron uptake by human beta cells and to investigate the influence of iron status on various aspects of diabetic pathology. I found that overexpression of ZIP14 and ZIP8, but not DMT1, resulted in increased iron uptake by Betalox5 cells, a human beta-cell line. sirna-mediated knockdown of ZIP14, but not ZIP8, resulted in 50% lower iron uptake in Betalox5 cells. In primary human islets, knockdown of ZIP14 also reduced iron 13

14 uptake by 50%. Immunofluorescence analysis of human pancreatic sections localized ZIP14 and DMT1, but not ZIP8, to beta cells. To determine how iron status may affect diabetic pathology I examined pancreatic gene expression in iron-deficient, ironadequate, and iron-loaded rats. Iron overload and deficiency were associated with increased pancreatic expression of genes associated with pancreatic stress and linked to the development of autoimmune diabetes. Additionally, non-obese diabetic mice fed an iron-deficient, but not iron-loaded, diet trended towards an increased incidence of diabetes compared with iron-adequate mice. However, this trend was not associated with a reduction in beta-cell function during the prediabetic period. Iron depletion or loading of isolated human islets also had no effect on beta-cell function. Overall, results from these studies indicate that ZIP14 contributes to beta-cell NTBI uptake and suggest that iron deficiency may not protect against the development of diabetes. 14

15 CHAPTER 1 INTRODUCTION Iron is an essential trace mineral that is necessary for many biological functions including, but not limited to, mitochondrial respiration (1), hemoglobin production (2), drug metabolism (3), and DNA synthesis (4). Disruptions of normal iron homeostasis, due to either iron deficiency or iron overload, are some of the most common nutritionrelated issues worldwide. One complication of iron overload is an increased prevalence of diabetes, as evidenced by humans with hemochromatosis or β-thalassemia major, which result in systemic iron accumulation (5, 6). Despite the well-documented association between excess iron and an increased incidence of diabetes several questions remain unanswered. For example, little is known regarding how pancreatic β cells take up iron, which is found in the plasma bound to transferrin, under normal conditions, and as non-transferrin bound iron (NTBI) during disorders of iron overload, when the carrying capacity of transferrin becomes exceeded. NTBI is known to be a major contributor to iron loading of hepatocytes and pancreatic acinar cells (7), but its contribution to iron loading of pancreatic β cells is not well understood. Additionally, the details of how iron deficiency and overload may affect the pathogenesis of diabetes have yet to be fully elucidated. The experiments described in this dissertation sought to answer questions concerning the link between iron metabolism and the pathogenesis of diabetes. Specifically, by determining the mechanisms by which NTBI is taken up by pancreatic β cells and by evaluating the impact of iron status on pancreatic gene expression and on aspects of diabetic pathogenesis. 15

16 To determine the molecular mechanisms of NTBI uptake by β cells I performed in vitro experiments testing how altering the expression of established mammalian NTBI transport proteins affects iron uptake by human β cells. Additionally I determined the cellular localization of these transporters within the human pancreas to identify the cell populations in which they may function. To evaluate the role of iron in the pathogenesis of diabetes I investigated potential mechanisms by which iron status could affect the development of diabetes by using microarray analysis to identify the altered expression of genes previously linked to various aspects of diabetic pathogenesis or β cell function, as described by other reports, in the pancreata of rats fed iron-deficient or -loaded diets. To test the proposed relationship between iron status and the risk of developing type 1 diabetes I also determined the effect of systemic iron status on the development of autoimmune diabetes in type 1-diabetes-prone NOD mice and measured the effect of cellular iron status on insulin secretion by human islets in vitro. Lastly, I document the generation of a novel transgenic mouse that overexpresses the iron transporter ZIP14, which may provide a model to study β cell iron loading in vivo. 16

17 CHAPTER 2 LITERATURE REVIEW Basics of Iron Metabolism General Information Iron in biological systems has two main oxidation states, ferrous (Fe 2+ ) or ferric (Fe 3+ ). These two oxidation states allow iron to readily exchange electrons and to participate in oxidation-reduction reactions within biological systems. Iron is generally categorized as either heme iron, in which iron is incorporated into a protoporphyrin ring as found in hemoglobin and myoglobin, or non-heme iron, a broad term used to describe all iron not incorporated into heme. The majority of dietary iron consumed is non-heme iron (8), which is found in both plant and animal foods whereas heme iron is only present in animal sources. Despite the greater bioavailablity of heme iron, the majority of iron absorbed by the body is obtained from non-heme sources (8, 9), due to the abundance of dietary non-heme iron. On average the human body contains 2.5 to 4 g of iron, for females and males respectively. The majority of body iron exists as hemoglobin or myoglobin, which function in oxygen transport, or as stored iron within ferritin, the iron-storage protein. The daily requirement for iron to support biological functions, such as erythropoiesis and the production of other iron-containing proteins, is approximately 24 mg/day. However, the majority of iron used daily is obtained from the catabolism of senescent erythrocytes by splenic and hepatic macrophages, which can then recycle the iron contained within erythrocytes for the production of new red blood cells and iron-containing proteins by the bone marrow and systemic tissues. While this process is efficient and provides approximately 90% of the daily iron requirement, minor quantities of iron are lost through various routes including sweating and the sloughing of 17

18 skin cells. Therefore, 1-2 mg of dietary iron is absorbed from the diet to offset daily losses and maintain systemic iron homeostasis. Dietary Iron Absorption Iron metabolism in humans differs from that of other minerals, such as copper or zinc, in that there is no major route of excretion by which substantial amounts of iron can be eliminated from the body. Rodents are able to remove some excess iron through biliary excretion, although this is insufficient to prevent iron overload (10). Therefore, the absorption of iron from food sources is tightly regulated. The majority of dietary iron is non-heme iron, which is solubilized at the acidic ph of gastric secretions and absorbed in the proximal small intestine via Divalent Metal-ion Transporter 1 (DMT1), a transmembrane protein that couples the transport of ferrous iron to a proton gradient (11). DMT1 is indispensable for intestinal NTBI uptake as mice lacking DMT1 in the intestine develop severe iron deficiency and display ablated iron absorption (12, 13). To ensure that dietary iron exists as ferrous iron, the form which can be transported by DMT1, ferric iron is thought to be reduced at the enterocyte brush border by the reductase duodenal cytochrome B (Dcytb). However mice lacking Dcytb show no difference in the uptake of radiolabeled ferric iron compared wild-type mice suggesting that Dcytb is dispensable for the reduction of ferric iron in vivo (14). It is possible that the absence of Dcytb can be compensated by the presence of unidentified brush-border reductases or reducing agents, such as ascorbate either consumed in the diet or secreted within digestive juices (15), to reduce iron within the gut lumen. After the uptake of ferrous iron by the enterocyte iron can either be stored within the enterocyte within ferritin, and eventually lost when the enterocyte is sloughed off into the gut lumen, or exported into the portal circulation. If systemic iron stores are adequate or elevated, 18

19 much of the iron taken up by enterocytes will be incorporated into the iron storage protein ferritin, a multisubunit spherical protein with ferroxidase activity containing a core composed of ferric iron. If systemic iron stores are low, or in response to stimuli including anemia (16, 17) or hypoxia (18), enterocyte iron is exported into the portal circulation by ferroportin, the only identified mammalian iron export protein. Ferroportin is located on the enterocyte basolateral membrane and is essential for iron export as mice selectively lacking intestinal ferroportin accumulate iron within enterocytes and develop severe anemia (19). After export from the enterocyte, iron in the plasma binds to the plasma iron transport protein transferrin (TF). However, ferrous iron exported from enterocytes must be oxidized to the ferric state before binding to transferrin. The oxidation of iron exported via ferroportin may be accomplished by the action of the ferroxidase hephaestin, located at the basolateral membrane of the enterocyte. In mice the loss of hephaestin function, due to genetic mutation (20) or genetic deletion (21), results in decreased dietary iron absorption and iron accumulation in enterocytes despite elevated levels of ferroportin (21, 22). Impaired iron transport by ferroportin in response to the loss of functional hephaestin may be due to a necessary interaction between ferroportin and hephaestin during iron export from the enterocyte. Hephaestin and ferroportin have been reported to physically interact in rat enterocytes (23) supporting this hypothesis. However, ferroportin is able to transport iron when overexpressed in xenopus oocytes, independent of hephaestin overexpression, suggesting that the interaction of ferroportin with hephaestin is not necessary for iron export activity (24). Ferric iron circulating as TBI in the plasma is taken up by cells through the binding of TF to transferrin receptor (TFR) on the cell surface, forming a 19

20 complex which is then endocytosed. Iron is released from TF within the endosome by endosomal acidification after which ferric iron is reduced to ferrous iron by STEAP3 (25), allowing for the export into the cytosol of ferrous iron via DMT1 and, potentially, ZRT/IRT-like Protein 14 (ZIP14) and ZRT/IRT-like Protein 8 (ZIP8) (26-28). In the bone marrow, the major site of transferrin-tfr iron acquisition, the export of endosomal iron is accomplished by the action of DMT1, as evidenced by loss-of-function mutations in DMT1 leading to impaired reticulocyte iron delivery by the endocytosed TF-TFR complex (29) and microcytic anemia (26, 30). In humans the majority of iron released into the circulation and delivered to tissues by TF is provided by the release of iron by the macrophages of the reticuloendothelial system and hepatocytes, with newly absorbed dietary iron being a minor contributor to plasma iron under normal conditions (31). To ensure that appropriate amounts of iron are provided for bodily functions, such as erythropoiesis, the release of stored iron from macrophages and hepatocytes is tightly regulated. The export of tissue iron stores into circulation is mediated by ferroportin (19, 32), similar to the release of iron from enterocytes, allowing iron mobilization from body stores and enterocytes to be regulated by a similar mechanism. The control of systemic iron homeostasis centers on the regulation of ferroportin levels through the action of hepcidin, a peptide hormone primarily produced by hepatocytes (33). Hepcidin binds to ferroportin leading to endocytosis and subsequent degradation (34), thus preventing the release of iron from cells. The expression of hepcidin is normally linked to systemic iron status and decreases during iron deficiency (35) to increase dietary iron absorption and the release of stored iron from macrophages and hepatocytes, while increasing during 20

21 iron loading (36), to prevent excess dietary iron uptake and the release of stored iron. The expression of hepcidin can also be regulated by factors including anemia, hypoxia, and inflammatory stimuli (37). Disorders of Iron Metabolism While under normal conditions iron homeostasis is tightly regulated by hepcidin to ensure adequate iron for cellular functions while preventing excess iron accumulation, conditions can arise during which iron homeostasis becomes perturbed. Dietary iron deficiency is the most common nutrient deficiency worldwide, with over 2 billion people estimated to be iron deficient (38). The greatest estimated incidence of dietary iron deficiency is seen in infants, children (39), and pregnant women in developing countries (38), but women of child-bearing age still suffer from iron deficiency in industrialized countries. In the United States 10-15% of women of childbearing age are iron deficient (40). Iron deficiency in men, as well as post-menopausal women, is less common in the United States (40). The negative effects of iron deficiency include fatigue (41), impaired cognitive function (42), and pica (43). On the opposite end of the spectrum from iron deficiency are disorders of iron overload, characterized by iron accumulation in various organs resulting in ironmediated tissue damage. Iron overload disorders can most often be linked to genetic mutations. One such disorder is hemochromatosis, a disease characterized by excessive iron loading in the liver, pancreas, and heart, resulting in hepatic fibrosis/cirrhosis (44), diabetes (45), and cardiomyopathy (46). The majority of hemochromatosis cases result from a single point mutation in the hemochromatosis gene, HFE (47) (48), and are referred to as type 1 hemochromatosis. HFE is involved in plasma-iron sensing by hepatocytes and interacts with transferrin receptor 2 (TFR2) to 21

22 regulate the hepatic production of hepcidin (48, 49). Homozygosity for this mutation occurs at a frequency of 1/200 in individuals of northern European descent but has incomplete penetrance (50). More severe and rapid iron accumulation is observed in humans with juvenile hemochromatosis, or type 2 hemochromatosis, resulting from mutations in the hemochromatosis type 2 gene (HFE2) (51), encoding Hemojuvelin (HJV) involved in hepatocyte iron sensing (type 2A) (52), or in the hepcidin antimicrobial peptide gene (HAMP) (53), encoding hepcidin itself (type 2B). Symptoms of iron overload in juvenile hemochromatosis occur rapidly and are often detected in patients with juvenile hemochromatosis before 30 years of age (54), earlier than type 1 hemochromatosis, which is usually diagnosed in middle-aged patients (55). The majority of juvenile hemochromatosis cases are due to mutations in HFE2, with HAMP mutations being less commonly documented (56). Type-3 hemochromatosis due to mutations in the gene transferrin receptor 2 (TFR2) (57), another protein involved in hepatocyte iron sensing (48), has also been documented to produce iron overload of an intermediate degree, with iron symptoms of iron overload appearing before those seen in with type 1 but later than type 2 hemochromatosis (58, 59).While these mutations occur in various genes, including HFE, HFE2, TFR2, and HAMP, the end result is a deficiency of hepcidin production by hepatocytes, resulting in the impaired downregulation of ferroportin and increased dietary iron absorption. A unique form of iron overload linked to mutations in ferroportin, rather than a dysregulation of hepcidin, is referred to as ferroportin disease or type-4 hemochromatosis. Mutations that inhibit the ability of hepcidin to bind ferroportin result in a phenotype similar to other forms of hemochromatosis (60, 61), in which ferroportin 22

23 expression is elevated, characterized by elevated dietary iron absorption, macrophage iron export, transferrin saturation, and liver iron accumulation. While gain-of-function mutations in ferroportin lead to iron overload, paradoxically, loss-of-function mutations in ferroportin have also been reported to result in iron accumulation. Mutations that impair the targeting of ferroportin to the plasma membrane (62) have been reported in humans with ferroportin disease, characterized by iron accumulation in macrophages, elevated serum ferritin prior to elevated transferrin saturation, and liver iron deposition (63). Currently it remains to be clarified how a loss-of-function mutation in ferroportin can simultaneously result in the loss of ferroportin activity in iron export by macrophages while allowing for dietary iron uptake and the establishment of elevated liver iron levels. Studies of intestinal iron absorption in humans or animals with loss-of-function ferroportin mutations are needed. Another genetic disorder that results in iron overload is β-thalassemia major, caused by mutations in the β globin gene (64) leading to impaired hemoglobin production. Treatment for β-thalassemia major involves regular blood transfusions that produce transfusional iron overload. Additionally, the failure to produce adequate hemoglobin results in persistent anemia that can suppress hepcidin production in response to iron overload, (65, 66), resulting in elevated intestinal iron absorption exacerbating transfusional iron loading (67). While patients with β-thalassemia major are usually treated with iron chelators, iron overload still occurs characterized by severe iron accumulation in peripheral tissues, often associated with the development of endocrine complications and heart failure (68). 23

24 Iron overload resulting from high dietary iron is far less common than hemochromatosis or β-thalassemia major and is best documented in individuals living in rural Sub-Saharan Africa. In this region, the practice of brewing beer in metal containers results in iron leaching into the beverage, which is often heavily consumed (69). Iron loading in response to this highly bioavailable iron affects some individuals to a greater degree than others indicating that there may be a genetic predisposition to iron accumulation (70). Mutations in ferroportin, similar to that observed in ferroportin disease, have been associated with African iron overload suggesting that this may explain the susceptibility of some individuals to dietary iron loading (71). However, other reports indicate that the presence of this mutation is not associated with markers of iron overload in African families with previously identified iron overload (72). The interplay between dietary iron consumption and genetic susceptibility to iron overload has yet to be elucidated regarding African iron overload. While many conditions lead to iron overload, a common phenotype of excess iron deposition in peripheral tissues, to varying degrees, is observed in response to systemic iron accumulation. During severe iron overload the amount of plasma iron exceeds the binding capacity of TF leading to the appearance of non-transferrin bound iron (NTBI). NTBI is cleared rapidly from the circulation by tissues, leading to iron deposition in organs such as the liver, pancreas, and heart (73, 74), potentially leading to tissue damage and dysfunction (60). While the deposition of iron in these tissues during iron overload is well established, the mechanisms by which NTBI is initially taken up remain to be elucidated for many cell types. 24

25 Non-Transferrin-Bound Iron Import Proteins The study of cellular iron transport in mammals has in large part centered on the action of transmembrane proteins that demonstrate the ability to transport free iron. The first discovered mammalian iron transport protein identified was DMT1, formerly referred to as divalent cation transporter 1 (DCT1) or natural resistance-associated macrophage protein 2 (NRAMP2), a transmembrane protein encoded by the solute carrier family 11, member 2 gene (SLC11A2) (11). DMT1 was originally identified by screening a cdna library from iron-deficient rat duodenum for iron transport activity (11). DMT1 expression was also found to be strongly induced in the duodenum of rats fed a low-iron diet, a treatment which greatly induces intestinal iron uptake. Intestinal expression of DMT1 is most abundant in the proximal small intestine, which has an acidic microenvironment near the brush border due to the presence of gastric secretions. In line with this localization, DMT1 functions optimally at an acidic ph with iron transport ability substantially decreasing, displaying only residual activity, at physiologic ph (75). The ph-dependent nature of DMT1-mediated NTBI transport is due to the coupling of iron transport with protons, necessitating a low ph for efficient activity. Since the initial discovery and characterization of DMT1, further studies have reported multiple isoforms of human DMT1 that differ at both the N and C-terminal regions, allowing for 4 isoforms. At the N terminus DMT1 isoforms can be identified as either 1A or 1B, differentiated by an additional amino acid sequence present on the 1A isoform proceeding the shared sequence by both 1A and 1B isoforms (76). 1A/1B isoforms of DMT1 display differential expression patterns, with 1B isoforms being expressed to some degree in all tissues examined whereas 1A-DMT1 expression is restricted to the duodenum and kidney (77). Isoforms of DMT1 differing in the C-terminal region can be classified as those translated 25

26 from mrna sequences with or without an iron response element (IRE), identified as +IRE or IRE isoforms. 3 iron response elements allow for posttranscriptional regulation of mrna levels, increasing mrna stability during iron deficiency (78). While the IREs are located in the 3 UTR, DMT1 translated from IRE transcripts differs from that translated from +IRE due to the substitution of the terminal 18 amino acids with a different 25 amino acid sequence (79). Tissue characterization of DMT1 expression between +IRE and IRE isoforms indicate that most tissues examined express both isoforms, with the exception of the liver, testis, and duodenum in which +IRE isoforms were more abundant (77). Differences in the C-terminal region between DMT1 isoforms have been documented to affect the intracellular targeting of DMT1 in a cell-type specific manner (76, 80). However, intracellular targeting attributed to differences between isoforms at the N-terminal region have yet to be reported. The isoform of DMT1 that functions in the duodenum to upregulate iron uptake in response to iron deficiency is likely DMT1+IRE, as the induction of DMT1+IRE, but not IRE, has been reported in the iron deficient mouse intestine (81), and 1A but not 1B isoforms, of DMT1 are regulated by iron status in Caco2 cells, an intestinal epithelial cell line (77). While the role of DMT1 in intestinal iron uptake and the export of iron from endosomes into the cytosol within developing erythrocytes is well established and previously discussed, the contribution of DMT1 to NTBI uptake by other cell types during iron overload is unclear. The initial characterization of DMT1 in rats reported low level mrna expression of Dmt1 in the liver, pancreas, and heart, relative to the level of Dmt1 expressed in the kidney or intestine (11), suggesting that NTBI uptake via DMT1 may be limited in these tissues. DMT1 has been detected in rat liver at the protein level; 26

27 however, hepatic DMT1 protein levels are strongly reduced in response to liver iron loading (82), suggesting that this pathway of NTBI uptake is unlikely to promote hepatic iron accumulation. A study of mice selectively lacking DMT1 in hepatocytes also argues against a role for DMT1 in NTBI uptake by the liver, as DMT1 was found to be dispensable for hepatic NTBI uptake and hepatocyte iron accumulation in a mouse model of hemochromatosis (83). Cardiac NTBI uptake through the action of DMT1 also appears unlikely as Dmt1 expression in the heart is observed to decrease in response to cardiac iron loading, similar to the trend observed in the liver (82). However, mechanistic studies regarding the role that DMT1 plays in cardiac NTBI uptake have yet to be carried out. Unlike DMT1 expression in the liver or heart, DMT1 expression is reportedly unchanged in response to pancreatic iron accumulation in rats (82). Acinar cells comprise the majority of the pancreas and therefore the unaltered DMT1 expression during iron loading likely reflects the state of this cell population. However, pancreatic iron loading is often viewed in the context of the pathogenesis of diabetes, requiring the study of pancreatic islets which constitute 1-2% of the pancreas. Due to the small contribution of islets to overall pancreatic mass, techniques that measure whole-tissue expression will be unable to draw accurate conclusions about islet gene expression as mrna or protein from islets will be diluted by that of acinar cells. Techniques that are able to discern cell-type-specific changes in gene expression have reported that the expression of Dmt1 within pancreatic islets in mice injected with iron decreases (84). Additionally, the pattern of DMT1 expression within the cell types of the pancreas is disputed in the literature. In humans pancreatic DMT1 is reported to be primarily 27

28 restricted to islets (85), whereas in mice contradictory reports exist demonstrating Dmt1 expression restricted to islets (60, 84) or to a similar level between islets and acinar cells (86). Mice selectively lacking Dmt1 in the β cell have been generated but no measure of NTBI uptake has been made in islets or β cells from this model (86). The second identified mammalian NTBI transporter is ZIP14, encoded by the gene SLC39A14. ZIP14 stands for ZRT/IRT-like Protein 14, named after the similarity between this protein and both zinc-regulated transporters (ZRT) and iron-regulated transporters (IRT). Members of the ZRT gene family transport zinc in Saccharomyces cerevisiae (87, 88), and IRT1 was identified as a route of iron transport in the roots of iron-deficient Arabidopsis thaliana (89). ZIP14 was originally identified as a zinc transporter and the iron transport capability of ZIP14 was not assessed until later, when ZIP14 expression was reported to affect NTBI uptake and iron accumulation in mammalian cells (90). Unlike DMT1, iron transport by ZIP14 is electrically neutral (75) indicating iron uptake is accompanied by either the co-transport of anionic or the export of a cationic species but the specifics of this have not yet been elucidated. ZIP14 has been demonstrated to transport iron optimally at a neutral ph, with a loss of iron transport as ph decreases (27, 75). The ability of Zip14 to transport iron at a physiologic ph, similar to that of plasma, is consistent with a role for ZIP14 in the clearance of NTBI at the plasma membrane. ZIP14 in mouse hepatocytes has been localized to the plasma membrane (82, 91), as well as intracellular locations (27). The expression of ZIP14 at the cell surface and intracellularly has also been reported in cell lines overexpressing ZIP14 (90-92) and those expressing ZIP14 at endogenous levels (27). Despite decreased iron transport ability at an acidic ph, ZIP14 has been demonstrated 28

29 to colocalize with transferrin within endosomes and promote the assimilation of iron from transferrin (27). More than 50% of TBI has been reported to dissociate from TF at ph 6.5 (93), at which ZIP14 still demonstrates iron transport activity (75), indicating that ZIP14 may contribute to the export of iron from the endosome into the cytosol. Four mrna transcript variants predicted to encode three unique protein isoforms of human ZIP14 have been recorded within the UniGene database. Transcript variants 1, 2, and 3 encode isoforms of human ZIP14 comprised of 492 amino acids. However, the predicted protein product of mrna variant 2 differs mid-sequence from the protein products of mrna variants 1 and 3, which are identical in amino-acid sequence. Transcript variant 4 encodes a 481-amino-acid peptide with a sequence identical to that of variants 1 and 3 with the exception of the C-terminal region. Currently, characterization of the differences in iron transport capabilities or intracellular targeting between isoforms of human ZIP14 has not been performed but investigation into the iron transport kinetics of mouse ZIP14 isoforms has been carried out in Xenopus oocytes. Three transcript variants of mouse Zip14 encoding 2 protein isoforms, both containing 489 amino acids but differing mid-sequence, have been reported and identified as isoforms A and B. Isoform B is reported to demonstrate a greater affinity for iron as well as a greater maximal rate of transport compared with isoform A (94). However, this study did not control for differences in the expression of individual ZIP14 isoforms complicating the interpretation of these results as similar sequences in the same expression vector may demonstrate differences in expression (76). Early characterization of ZIP14 identified this protein as a potential contributor to tissue NTBI uptake. Human ZIP14 mrna expression was reportedly highest in the liver, 29

30 pancreas, and heart, tissues known to accumulate iron during iron overload (92). The cellular/subcellular localization of ZIP14 in human tissues has not yet been performed. However, in rats Zip14 is detected at the hepatocyte sinusoidal membrane and throughout pancreatic acinar cells, at the plasma membrane as well as intracellular locations, but not β cells (82). Recent determination of the role ZIP14 plays in tissue NTBI uptake using mice lacking Zip14 has indicated that Zip14 is required for iron loading in hepatocytes and pancreatic acinar cells in response to both genetic and dietary iron overload (7). In the absence of Zip14, iron deposits are only observed in the non-parenchymal cells of the liver and within pancreatic connective tissue indicating that ZIP14 is likely the sole route of NTBI uptake within hepatocytes and acinar cells. ZIP14 is unlikely to contribute to cardiac NTBI accumulation as hearts from ZIP14 knockout mice display no difference in NTBI uptake compared with mice with intact ZIP14. In addition to providing a route of NTBI uptake within the liver and pancreas, ZIP14 expression has also been reported to increase in response to iron loading within these tissues (82). Mechanistic studies of the regulation of ZIP14 by cellular iron status have shown that iron regulates ZIP14 posttranscriptionally, by preventing the proteosomal degradation of ZIP14 (95). The upregulation of hepatic and pancreatic ZIP14 in response to iron loading suggests that iron accumulation in the liver and pancreas may increase the future rate of iron uptake in these tissues. In agreement with this hypothesis, iron loaded HepG2 cells (96) and rodent hepatocytes (97, 98) demonstrate increased NTBI uptake. However, while iron loading increases total-cell ZIP14 expression the abundance of ZIP14 on the plasma membrane is not increased, relative to non-iron treated HepG2 cells (95), arguing against increased plasma NTBI clearance 30

31 by ZIP14 during iron overload (although NTBI uptake in response to iron loading was not measured in this study). Discrepancies in the subcellular distribution of ZIP14 between studies may be attributed to differences in the degree of iron loading achieved (95, 96). Future studies will be required to determine if the increased rate of NTBI uptake associated with cellular iron loading is attributed to ZIP14 upregulation. The most recently described mammalian NTBI transporter is ZIP8, ZRT/IRT-Like Protein 8, another member of the ZIP protein family encoded by the gene SLC39A8. Within the ZIP protein family ZIP8 is the most similar to ZIP14 in amino-acid sequence, with mouse ZIP8 and ZIP14 having approximately 50% shared identity (99). ZIP8 was originally referred to as BIGM103 and identified as a protein induced in monocytes in response to LPS or TNFα (100). Similar to ZIP14, ZIP8 was originally found to transport zinc, as overexpression of ZIP8 increased zinc accumulation by CHO cells (100). In light of the ability of ZIP14 to transport iron and the similarity between ZIP14 and ZIP8, the iron transport activity of ZIP8 was measured revealing that ZIP8 overexpression and suppression increase and decrease NTBI uptake, respectively, in mammalian cells (28). ZIP8 is reported to transport iron optimally at ph 7.5, with a loss of transport activity with decreasing ph, and is localized to the plasma membrane supporting the role of this protein in iron uptake from the plasma (28). ZIP8 is also detected in endosomes suggesting that ZIP8 may be able to contribute to iron export from endosomes into the cytosol as ZIP8 demonstrates iron transport within mammalian cells at ph 6.5 (28, 101), at which iron will dissociate from transferrin within endosomes (93). However, iron transport in Xenopus oocytes is abrogated at ph 6.5, complicating the interpretation of the role ZIP8 plays in endosomal iron transport (28). ZIP8 reportedly functions as an 31

32 electrically neutral symporter (102), coupling the transport of cations with the transport of bicarbonate, as cells overexpressing ZIP8 demonstrate increased ion uptake in the presence of added bicarbonate and decreased metal uptake when bicarbonate transport is inhibited (101). However, the possibility that the effect of bicarbonate on metal transport was mediated through the action of a transporter other than ZIP8 was not accounted for in this experiment. ZIP8 mrna is detected in many tissues but concentrated in the pancreas, lung, placenta, liver, and thymus (100). Detection of ZIP8 at the protein level within these tissues, from either human or animal sources, has been carried out to a very limited degree, both in regards to tissue protein expression and the cellular/subcellular localization of ZIP8 in vivo. The lack of data regarding the expression profile of ZIP8 currently limits the ability to make conclusions as to the role of ZIP8 in tissue NTBI uptake during iron overload. ZIP8 is reportedly expressed at the plasma membrane of rat β cells (103) but experiments to determine the contribution of ZIP8 to β cell iron uptake have not been performed. Study of the role ZIP8 plays in tissue NTBI uptake has also proved difficult due to the embryonic lethality of Slc39a8 disruption in mice. Mice with Slc39a8 alleles disrupted by the neomycin-resistance cassette display hypomorphic ZIP8 expression and fail to survive past post-natal day 3 (104). The embryonic lethality observed in response to the disruption of ZIP8 is thought to be due to impaired hematopoiesis during embryonic development (104). Mice with selective deletion of Zip8 in tissues that accumulate iron during iron-overload disorders have yet to be investigated. However, it is unlikely that Zip8 plays a role in NTBI uptake by hepatocytes or pancreatic acinar cells as these cell populations display no iron loading 32

33 in the absence of ZIP14 (7). ZIP8 may play a role in NTBI uptake by other hepatic or pancreatic cell types as well as in other organs (e.g. heart) that are unaffected by the loss of ZIP14. Similar to ZIP14, ZIP8 expression is upregulated by cellular iron loading. However, unlike ZIP14 that demonstrates an increase in intracellular rather than cell surface expression in iron-treated HepG2 cells (95), ZIP8 levels were observed to increase at the cell surface in response to iron loading in H4IIE cells, a rat hepatoma cell line (28). However, an increase in the protein level of ZIP8 in response to iron loading in vivo has yet to be confirmed. More research is necessary to determine the pattern of ZIP8 expression in human tissues and the contribution of ZIP8 to iron transport. Some reports indicate that NTBI can be taken up into cells through both L-type Ca 2+ channels (LTCC) and T-type Ca 2+ channels (TTCC). Currently, the study of NTBI transport by calcium channels has been restricted to cardiomyoctes, exploring mechanisms by which iron may accumulate in cardiac tissue during iron overload. LTCC agonists increase iron uptake by the mouse heart and LTCC blockers decrease iron accumulation in mouse heart tissue perfused with ferrous sulfate (105). Additionally, treatment with LTCC or TTCC blockers in vivo reduces cardiac iron accumulation in mice injected with iron dextran and in a mouse model of β thalassemia (106, 107). While Ca 2+ channels show promising evidence for a role in cardiac NTBI uptake, the contribution of Ca 2+ channels to NTBI uptake by other cell populations has not been investigated. However, other cell types, such as pancreatic β cells, express both LTCC (108) and TTCC (109) and therefore may take up via calcium channels. The 33

34 contribution of both LTCC and TTCC in various cell populations should be a topic of future study. The Iron and Diabetes Connection Elevated Iron Stores and Diabetic Pathology A link between iron overload and the development of diabetes dates back to the initial case report of the disorder which would later become known as hemochromatosis. Autopsy of an individual who died due to diabetic complications in 1865 noted the bronze coloration of skin and organs referring to the condition as bronze diabetes. Many years later it was established that the bronze discoloration observed was due to excess iron deposition in tissues. The initial link between excess iron accumulation and diabetes has been strengthened since the initial discovery by studying the prevalence of diabetes in patients with pathological iron overload disorders. In patients with hemochromatosis the incidence of diabetes varies considerably among reports but is substantially greater than the incidence reported in the general population of middle-aged Americans and Northern Europeans, in which hemochromatosis is prevalent (110, 111). A general trend is that the reported incidence of diabetes in hemochromatotic patients declines with the passage of time, with the highest prevalence being reported in earlier manuscripts and a lower prevalence in more recent reports. Initial reports indicate that diabetes is observed in approximately 80% of patients with hemochromatosis (112), although the diagnostic methods used were not discussed. A later report based on data collected by physicians from an unspecified time until 1972 reported that 63% of patients with hemochromatosis, as defined by serum iron indices and liver biopsy, were diagnosed with diabetes (113). A study of hospital records from 1977 to 1997 found that 40% of patients admitted to the 34

35 hospital with hemochromatosis were diabetic (114). However, a substantially lower incidence of diabetes was reported in a study of French and Canadian hemochromatosis patients between 1970 and 1997 which only identified 15.9% of men compared with 7.4% of women to have diabetes (115). However this study may have under reported the prevalence of diabetes as patients were not tested for undiagnosed diabetes in this study. Glucose tolerance testing in hemochromatosis patients between 2000 and 2006 determined that 23% of patients were diabetic and 30% had impaired glucose tolerance or elevated fasting glucose (45). The same study corroborated these findings by examination of the medical records from the study center from 1975 to 2006 which reported that 26% of hemochromatosis patients were diabetic, as defined by a measurement of elevated fasting blood glucose. The general decrease in the reported prevalence of diabetes in patients with hemochromatosis over time may reflect improvements in diagnosis, such as genetic testing, and improved implementation of treatments. In hemochromatosis patients diagnosed after the advent of genetic testing, post 1995, the prevalence of diabetes was 17.7% compared with 35.6% in patients diagnosed with hemochromatosis by elevated iron indices between , prior to genetic testing (116). However, it should be noted that there was no difference in the age of diabetic diagnosis between the groups in this study allowing the older patients, diagnosed with hemochromatosis by iron parameters rather than by genetic testing, a longer time period to develop diabetes. In line with this concept the prevalence of impaired glucose tolerance is greater in patients diagnosed with hemochromatosis in recent years, 13%, compared with those diagnosed pre-1995 (6.7%) (116). It is difficult to interpret from this study whether impaired glucose 35

36 tolerance will progress to diabetes, resulting in a similar prevalence of diabetes as was observed in patients diagnosed with hemochromatosis by traditional metrics of iron overload, rather than by genetic testing, if given the same amount of time. The study carried out by Bussychard et al. (114) is in agreement with the concept that diabetes incidence likely is inversely related to the management of hemochromatosis and lifestyle choices as diabetes was prevalent in 53% of patients with cirrhosis but only 25% of those without cirrhosis. Ferritin levels were significantly higher in patients with cirrhosis suggesting poor disease management relative to those with no cirrhosis and a lower incidence of diabetes. The addition of other factors associated with diabetes risk, such as obesity, may also affect the development of diabetes in hemochromatosis patients. One report has documented that all patients with hemochromatosis who were identified as diabetic by glucose tolerance testing were overweight or obese (45), suggesting that the influence of iron overload on diabetic pathology may increase when coupled with additional risk factors for diabetes, such as obesity. However, another study reported no difference in BMI between patients with hemochromatosis alone, 24.2 kg/m 2, or with hemochromatosis accompanied by diabetes, 25 kg/m 2, (117). The development of diabetes in non-obese hemochromatosis patients indicates that obesity is not required for the development of diabetes in patients with iron overload. Evidence linking pathological iron overload to the development of diabetes is also provided by studies measuring the prevalence of diabetes in patients with thalassemia major. Thalassemic patients frequently develop diabetes, often at a young age when diabetes is rare in the general population. The prevalence of diabetes varies among studies likely reflecting differences in demographic factors including frequency of 36

37 transfusion, age of patients, and improvements in disease management or treatment options. One study reported an initial diabetic prevalence of 26%, increasing to 30% after a 2-year follow up (118). Another study documented a 14% prevalence of diabetes out of all cases seen over the course of a 30-year study (119). However, diabetes was only documented in patients over the age of 23, yielding a prevalence of 30% in this demographic within the study. The observed increase in the prevalence of diabetes with age supports the concept of longer-duration iron accumulation increasing diabetes risk in patients with thalassemia as well as novel treatment options potentially improving disease prognosis, as evidenced by improved overall survival with increased chelator treatment (119). A lower prevalence of diabetes, 6.5%, has been reported in patients from various centers in Italy which followed patients diagnosed with thalassemia between 1970 and The lower prevalence in this study may be partially by more stringent diagnostic criteria for diabetes, fasting glucose higher than 140 mg/dl for several consecutive days or overt symptoms of diabetes, such as glycosuria. It is possible that asymptomatic or mild diabetes may have been overlooked in this study (120). While an increased incidence of diabetes in pathological disorders resulting in iron overload is well established, elevated iron stores remaining within the normal range may also contribute to the pathogenesis of diabetes. Many prospective epidemiological studies have investigated the link between markers of iron status, most often determined by serum ferritin, and the risk of developing diabetes in the future ( ). Nearly all studies report increased baseline serum ferritin levels in subjects that develop diabetes at a later time point in the study supporting the link between elevated normal 37

38 iron stores and glucose dyshomeostasis. However, the use of serum ferritin as an indicator of iron stores is problematic due to the nature of ferritin as an acute-phase reactant induced by inflammatory signaling pathways (126). Chronic systemic inflammation is hypothesized to contribute to diabetic pathology (127) and therefore it is difficult to determine whether ferritin is elevated in subjects who develop diabetes due to increased systemic iron or simply in response to low-grade inflammation preceding the development of diabetes. In line with the concern that serum ferritin is a marker of inflammation rather than a barometer of iron stores, markers of inflammation, such as C-reactive protein (CRP), are often higher in the diabetic cohort at baseline relative to the control patients which did not develop diabetes during follow up (121, 123, 125). However, when relative risk is adjusted to account for differences in CRP, serum ferritin levels remain an independent risk factor for developing diabetes (123). Fewer studies make use of soluble transferrin receptor (stfr) levels in serum to measure iron status, either alone or in combination with serum ferritin ( ). stfr is reported to inversely correlate with iron stores independently of inflammatory stimuli (128, 129). Unlike serum ferritin, stfr as an indicator of iron status does not suggest elevated baseline iron stores in patients who developed diabetes arguing against the modulation of diabetes risk by iron stores within normal levels (123). Some studies do not directly report the value for stfr, instead referencing the ratio of stfr to serum ferritin (130). However, the data for ferritin listed individually in these studies suggest that stfr is largely similar at baseline between normal patients and future diabetics. Elevated stfr, suggesting lower iron stores, has even been reported in baseline measurements from patients who developed diabetes compared with controls (125). In 38

39 light of the confounding influence of inflammation on ferritin levels and the lack of a clear trend observed with stfr levels, the link between variation in non-pathological iron stores and the development of diabetes appears unsubstantiated, despite being often mentioned in the literature. Pathological iron accumulation leads to transferrin saturation and the appearance of NTBI in circulation. Additionally, plasma NTBI has been reported in type 2 diabetics despite normal levels of transferrin saturation and without pathological iron accumulation (131). Plasma NTBI present in diabetics or in patients with severe iron loading can be taken up by tissues such as the pancreas. Pancreatic Iron Accumulation Due to the role of the pancreas, specifically the pancreatic β cells, in glucose homeostasis and the pathogenesis of diabetes, pancreatic iron accumulation is thought to potentially account for the increased prevalence of diabetes in humans with pathological iron overload. Reports of iron loading in the human pancreas from patients with either hemochromatosis or thalassemia indicate that iron deposits heavily within the acinar cells of the exocrine pancreas and to a somewhat lesser degree within islets of the endocrine pancreas as determined by various iron-staining techniques ( ). Within the islet, iron staining has been determined to be primarily restricted to β cells, with α cells remaining relatively free of iron deposits (132, 134). No studies have reported iron accumulation in the other cell populations of the pancreatic islet (e.g. δ cells, ε cells). Due to the invasive nature of pancreatic biopsy, all data published on human pancreatic iron loading at this point are derived from autopsy cases. Therefore, the current reports indicate the pattern of pancreatic iron loading at end of life usually 39

40 after long periods of iron overload. Magnetic resonance imaging has been used to detect pancreatic iron accumulation and could be used to measure the progression of pancreatic iron overload over long periods in patients with iron overload (135). However, magnetic resonance imaging is not currently able to distinguish between cell populations within the pancreas precluding the use of this method to discern between iron loading in the exocrine versus endocrine pancreas. New non-invasive methods of measuring pancreatic iron in vivo will need to be developed before the nuanced characterization of pancreatic iron loading over time in humans is feasible. While the cell type-specific examination of pancreatic iron loading is not possible in humans, extensive studies in rodents have been carried out with both genetic and dietary models of iron overload. However, rodent models fail to appropriately model human pancreatic iron loading as both mouse and rat islets remain largely free of iron deposition even during severe iron overload (60, 134, 136, 137). Studies using Hjv (138) and Hamp (136) knockout mice, resulting in severe pancreatic iron accumulation, clearly show that β cells/islets fail to accumulate iron despite iron building up in the surrounding exocrine pancreas. In accord with the lack of islet iron deposition in these models no diabetic phenotype, characterized by impaired glucose tolerance resulting from insufficient insulin secretion, is detected even up to a year of age in HAMP knockout mice (136). The resistance of mice islets to iron loading and subsequent diabetic pathology during severe iron overload is best evidenced by a novel mouse model with a mutation in ferroportin resulting in impaired regulation by hepcidin leading to severe tissue iron loading (60). Massive iron accumulation in this mouse model results in death due to exocrine pancreas failure, a phenotype not reported in other 40

41 mouse models of iron overload, suggesting that this model potentially represents the most severe model of pancreatic iron loading. Yet, even in this mouse model islets are spared from iron deposition and glucose homeostasis is unchanged compared with mice that have the wild-type ferroportin allele. Rats also fail to recapitulate the human condition with regard to substantial islet iron loading and the development of a diabetic phenotype. Several feeding studies with high-iron diets at various concentrations and durations have been performed using rats (137, ). However, none report a diabetic phenotype even when pancreatic iron loading was observed. Studies that have achieved pancreatic iron loading in rats through prolonged feeding with high-iron diets report that, similar to mice, iron deposition occurs within the exocrine pancreas, in both acinar cells and interstitial areas, but islets were spared from iron loading (140). High dietary iron in rats has also been reported to lead to pancreatic atrophy, of both the exocrine and endocrine pancreas, during which pancreatic tissue is replaced by adipose tissue (141, 142). However, iron staining or the measurement of pancreatic iron was not reported in these studies to demonstrate that iron deposits were localized to atrophic cells. Currently it remains unclear whether islet atrophy in response to high dietary iron is due to islet iron accumulation or simply the loss of normal surrounding pancreatic morphology, as islets are reported to resist atrophy longer than acinar cells (142). The most extreme incidence of pancreatic iron loading through diet combined the surgical bypass of the liver by the portal circulation in conjunction with feeding a high-iron diet for an extended period of time resulted in severe pancreatic iron overload (137). However, within the islet very little iron deposition was observed relative to the surrounding acinar cells. No 41

42 determination of glucose homeostasis was carried out in this study, but the authors reported no overt diabetic phenotype despite an extreme treatment to induce pancreatic iron loading. Additionally pancreatic atrophy was only reported in one rat that received this treatment. It is possible that genetic differences between rat strains account for differences in susceptibility to pancreatic atrophy in response to iron loading. Iron loading of the pancreas has also been reported in rats injected with high doses of iron. Rats administered iron dextran demonstrate detectable iron staining within the exocrine pancreas, with the most prominent staining reportedly detected within macrophages, while islets remained free of iron deposits. Injection of ferric nitrilotriacetate (Fe-NTA) daily for 120 days in rats resulted in heavy iron deposits visible by Perl s staining in acinar cells, macrophages, and pancreatic connective tissue but iron loading was sparse within islets (143). Similar experiments repeated by the same group produced similar results reporting iron deposits in the exocrine pancreas, while the islet remained free of detectable iron accumulation until 6 months after the start of injections, when faint staining was detected within islets (144). Experiments that use Fe- NTA injections to produce iron overload are unique in that they report a diabetic phenotype, characterized by hyperglycemia and reduced insulin secretion (143, 144). However, these studies report the development of diabetes within 1-2 (143, 144) months of iron injections while iron deposits are not detectable in islets until 6 months after the start of injections (144) making it difficult to conclude that it is β cell iron loading that directly leads to impaired insulin production in this model. Additionally Fe-NTA administered via injection represents both a non-physiologic source and method of iron loading. 42

43 Mechanistic Evidence from Human Studies Currently there is a lack of an established physiologically relevant rodent model that accurately models the pancreas of humans with iron overload, characterized by significant iron deposition in β cells in response to genetic, transfusional, or dietary iron overload. Due to the limitation of in vivo models, mechanistic investigation into the pathogenesis of iron-induced diabetes has been difficult; however, studies using humans with iron-overload disorders provide some insight into the pathology of ironinduced diabetes. Diabetes in humans with iron overload can result from either impaired insulin production by β cells, decreased peripheral insulin sensitivity, or a combination of these factors (145). The study of humans with hemochromatosis suggests that impaired glucose homeostasis is due, at least in part, to impaired insulin secretory capacity. Hemochromatosis is associated with a trend toward decreased first and second phase insulin secretion during glucose tolerance testing and individuals with hemochromatosis and diabetes show drastically reduced first phase insulin secretion (117). This finding has been corroborated in the literature as middle-aged, non-obese patients with hemochromatosis demonstrate lower glucose-stimulated insulin secretion compared with relatives without hemochromatosis (45). Somewhat conflicting evidence exists for the role of insulin sensitivity in glucose homeostasis during hemochromatosis. One report indicates that insulin sensitivity is not different between control subjects and patients with hemochromatosis without cirrhosis or diabetes (117). However, in patients with hemochromatosis and either cirrhosis or diabetes, insulin sensitivity is significantly impaired, potentially attributable to greater iron loading, as indicated by higher serum ferritin levels, compared with non-symptomatic hemochromatosis. Another study 43

44 reported that hemochromatosis patients with normal glucose tolerance maintain greater insulin sensitivity, despite similar age and body mass indices, compared with normal patients, although again this was lost in patients with diabetic complications (45). Studies from animal models of hemochromatosis support the position of unchanged insulin sensitivity in human hemochromatosis as insulin sensitivity in both Hamp -/- and Hfe -/- mice is similar to that in wild-type mice (136, 146). Conflicting evidence also exists regarding the effect of iron depletion by phlebotomy on β cell function and insulin sensitivity in hemochromatosis. The removal of systemic iron stores, as indicated by diminished plasma ferritin levels, by phlebotomy is reported to improve insulin secretory capacity to varying degrees (147, 148), although this is interestingly associated with unchanged glucose tolerance attributed to either decreased insulin sensitivity or increased hepatic insulin clearance. The extent of improvement in insulin secretion, post phlebotomy, may be due to the amount of iron removed and the severity of initial iron accumulation, as a more prominent improvement in insulin secretory capacity was seen in individuals with greater initial iron loading (147) compared with more mild iron accumulation (148). However, normalization of iron stores by phlebotomy has also been reported to have no effect on either first-phase insulin secretion, although longer time points were not examined and insulin sensitivity was not measured (117). In diabetic patients without hemochromatosis but with elevated ferritin levels, phlebotomy improved both insulin secretory capacity and insulin sensitivity, supporting the idea that elevated iron levels negatively affect glucose homeostasis by modulating both β cell function and peripheral tissue insulin sensitivity (149). 44

45 In patients with β thalassemia major glucose dyshomeostasis may result from either decreased insulin sensitivity, impaired β cell function, or some degree of both of these factors. In thalassemics over the age of 18, insulin secretion in response to glucose was elevated but glucose tolerance was unchanged relative to control subjects, indicating diminished insulin sensitivity but increased β cell function (150). Decreased insulin sensitivity compensated by increased insulin production in patients with thalassemia has been reported elsewhere and it was documented that the phenotype became more pronounced with the extent of transfusional iron overload (151). With the onset of adulthood, glucose tolerance worsened and a loss of first-phase insulin secretion was observed, indicating impaired β cell function after initial insulin resistance (151). The finding of eventual β cell failure corroborates other reports of decreased insulin sensitivity and a lack of compensatory insulin secretion leading to impaired glucose tolerance, attributed to β cell failure (152). Impaired glucose tolerance in patients with thalassemia has also been attributed to impaired insulin secretion by other studies, however insulin sensitivity was not directly measured in these reports (153, 154). In patients with thalassemia improved treatment with iron chelators is reported to marginally improve glucose tolerance, attributed to improved insulin sensitivity rather than increased insulin secretion (155). Currently, the specifics of the relative contribution between systemic insulin resistance and declining β cell function to the development of diabetes in patients with thalassemia are still undefined. Variation in these reports likely reflects differences between the subjects studied possibly including lifestyle factors such as effectiveness of chelation therapy, extent of iron overload, or age of participants. 45

46 While the clinical characteristics of hemochromatosis and thalassemia differ with regard to insulin sensitivity, both disorders are linked to a diminished insulin secretory capacity suggesting impaired β cell function. Several reports indicate that islets from humans, as well as rodents, express low levels of or display low activity of antioxidant proteins, relative to other tissues such as the liver, suggesting that islets may be easily damaged by oxidative stress ( ). During iron overload it has been hypothesized that iron accumulation catalyzes the formation of reactive oxygen species (ROS), which damage islets and leads to a loss of insulin secretory capacity. However, the effect of iron loading on β cell function, evidenced by markers of oxidative stress and indicators of β cell function such as glucose-stimulated insulin secretion, has not been examined in human islets. In addition to iron accumulating directly in β cells leading to impaired function it has also been suggested that iron loading of the exocrine pancreas can result in pancreatic atrophy, compromising β cell function due to the loss of normal pancreatic morphology. Diminished pancreatic mass and an increased proportion of fatty tissue in the pancreas has been reported in patients with hemochromatosis that have developed diabetes (132). However, it is possible that in patients with both iron overload and diabetes diminished pancreatic mass is a result, rather than the cause, of decreased β cell function as insulin is thought to promote acinar cell growth (159) and a lack of insulin is associated with decreased pancreatic mass (160). Compelling evidence against the degeneration-dysfunction hypothesis is provided by a mouse model of iron overload that demonstrates severe degeneration of the exocrine pancreas and the failure to produce digestive enzymes resulting in death. However, glucose tolerance is 46

47 intact in this model despite pancreatic degradation (60) suggesting that iron-mediated damage of the exocrine pancreas has little effect on β cell function. Mechanistic Evidence from Animal Studies Studies carried out using mice are complicated by the lack of substantial β cell iron accumulation and the development of a diabetic phenotype. However, several studies have still been performed in mouse models of iron overload which have reported differences in β cell function and/or mass. Islets from Hfe knockout (Hfe -/- ) mice on a 129/SvEvTac genetic background are reported to have marginally elevated iron levels compared with Hfe +/+ mice and this is associated with diminished islet mass, pancreatic insulin content, and in vitro GSIS by isolated islets (146) attributed to increased β cell oxidative damage and apoptosis. However, in vivo glucose tolerance was improved in Hfe -/- mice compared with Hfe +/+ controls (146, 161), despite slightly reduced insulin secretory capacity during the first 30 min following intraperitoneal glucose injection (146). In aged Hfe -/- mice, on a C57BL6 background, glucose tolerance was reportedly reduced attributed to a failure to increase insulin secretion compared with Hfe +/+ mice, suggesting the burden of added iron results in increased β cell dysfunction with time. Follow up experiments with the Hfe -/- mice have suggested that the phenotype observed is due to a cellular manganese deficiency in response to elevated iron levels leading to insufficient Mn-SOD activity and impaired mitochondrial function (162). Supplementation with manganese improved insulin secretion and glucose tolerance in Hfe -/- mice. The Hfe -/- mouse represents a relatively mild model of iron overload, accumulating little additional iron in the pancreas (7), and it would be expected that the non-significant trend towards reduced insulin secretion in this model, on both 129/SvEvTac and C57BL6, backgrounds would manifest as an overt phenotype in a more drastic model of 47

48 iron overload, such as Hamp knockout mice. However, aged mice lacking hepcidin, resulting in massive iron overload, report no difference in either glucose tolerance or GSIS (136), suggesting the slight differences observed in Hfe -/- mice may be attributable to a factor other than iron overload. Additionally the phenotype of improved glucose tolerance, attributed to increased glucose disposal as a result of increased AMPK activity in Hfe -/- mice, was not reported in Hamp -/- mice, again suggesting other signaling pathways besides iron sensing may be altered in Hfe -/- mice. Leptin-deficient, Ob/Ob mice on a C57BL/6J background have also been used as a model to reveal the possible contribution of iron status to dysglycemia during conditions when β cell function is stressed. It has been reported that feeding of a diet containing 500 ppm iron vs 35 ppm results in increased insulin resistance and a loss of insulin secretory capacity in response to obesity (163). Similar findings were also reported in response to iron chelation in the same study. However, in this study no iron measurements were performed for any tissue making it difficult to conclude whether the observed differences were due to the impact of iron on the β cell or rather systemic differences. Dietary iron deficiency and phlebotomy have also been reported to increase pancreatic insulin content in obese rats compared with those fed a control diet, in which pancreatic insulin levels were unaffected by iron deficiency, (164) attributed to lower levels of islet fibrosis. However, it unlikely that the fibrosis was due to islet iron accumulation as iron staining did not reveal any iron deposits in the islets, with iron only accumulating in macrophages, and islet iron was not otherwise quantified in any manner. Rats in this study were not fed an iron-loaded diet but plasma ferritin levels 48

49 were elevated suggesting systemic inflammation, potentially accounting for the observed islet fibrosis and loss of function. Rodents are poor models of β cell iron loading, as previously discussed, complicating the interpretation that changes in β cell function in response to iron loading in vivo are due to β cell iron accumulation rather than other systemic factors as a result of systemic iron accumulation. Currently, investigation into the influence of islet iron status on function in vitro, without the confounding factors potentially introduced by changes in systemic iron status in vivo, has not been carried out. However, treatment of rat islets with iron has been demonstrated to modestly decrease islet viability, attributed to β cell death, and increase markers of oxidative stress (165). Future studies will be needed to determine if increased oxidative stress due to iron loading results in impaired islet function. Iron and Autoimmune Diabetes The vast majority of research examining the relationship between iron status and diabetic pathology has focused on the impact of iron on exacerbating β cell stress or worsening insulin resistance. The failure to produce adequate insulin to regulate glucose levels due to either β cell exhaustion, insulin resistance, or a combination of these factors is associated with type 2 diabetes; however, less is known about the impact of iron status on the development of autoimmune type 1 diabetes. Epidemiological evidence is less prevalent and provides a less compelling link between iron and autoimmune diabetes compared with type 2 diabetes. The consumption of greater dietary iron within the first 4 months of life has been associated with an increased risk of developing type 1 diabetes before the age of 6 (166). However, the collection of data for this study relied on 1 retrospective survey, administered 6 to 10 49

50 years after the feeding period of interest, in which the parents were asked about the child s food consumption. Therefore, the data obtained in this study, and the conclusions drawn, may be questionable. Elevated transferrin saturation, indicating higher iron status, was also associated with an increased risk of type1 diabetes in adults (167). However, it is also a possibility that in this study elevated transferrin saturation was a result of diabetes, rather than a causative factor, as hepatic insulin signaling has been demonstrated to regulate hepcidin production and iron absorption (168). One caveat in these epidemiological studies is that type 1 diabetes is defined as insulindependent diabetes rather than directly indicated as type 1A diabetes, indicating autoimmune involvement, or type 1B diabetes indicating an idiopathic origin. No indicators of the mechanistic origin of type 1 diabetes are measured in the current studies and therefore it is unclear whether insulin deficiency is attributable to the autoimmune destruction of β cells. Additionally it has been suggested that some cases of late-onset type 1 diabetes may be the result of undiagnosed hemochromatosis leading to the loss of β cell function. The prevalence of HFE C282Y homozygosity was greater in type1 diabetics who were diagnosed after the age of 30, allowing time for systemic iron loading, compared with the general population and that follow up investigation of these patients indicated iron overload (169). However this study did not determine the mechanism of diabetic development in patients with type 1 diabetes and iron overload and therefore it cannot be established if hemochromatosis was directly responsible for the loss of β cell function or if iron had an effect on autoimmunity in these patients. Currently, the epidemiological evidence linking iron status with autoimmune diabetes is tentative and very limited; a more compelling case can be 50

51 made from studies that treated rodent models of autoimmune diabetes with factors capable of modulating aspects of iron metabolism. Several reports provide indirect evidence that iron restriction may limit critical aspects of β cell death during autoimmune diabetes. Injection of the iron chelator desferrioxamine prolongs the survival of islets grafts in NOD mice, suggesting that iron restriction may limit T cell mediated β cell destruction (170). However, no measure of either systemic or cellular iron status or further mechanistic investigation was carried out in this study, thus complicating the interpretation of these results. Treatment with anti- TFR antibodies has also been demonstrated to impair T cell proliferation and cytotoxicity in vitro (171). The mechanism by which TFR interference affects T cell activation is unclear but could potentially be linked to impaired iron acquisition limiting activity or proliferation. However, the antibody used to bind TFR is reported to not interfere with TF-TFR interaction, arguing against the blockade of iron uptake as an explanation for impaired the impaired T cell response. It should be noted that in this study the actual measure of cellular iron uptake was not carried out and it was only assumed that if binding between TFR and TF was able to occur then successful iron acquisition was also unhindered (172). In line with TFR signaling playing a role in immune activation, apotransferrin has demonstrated a protective effect towards islets during diabetic pathogenesis. Treatment of cultured mouse islets with apotransferrin reduces the loss of viability observed in response to incubation with proinflammatory cytokines and administration of apotransferrin to rodent models of spontaneous autoimmune diabetes decreases the incidence and delays the onset of diabetes (173). The protective benefits of apotransferrin in these models have been attributed to 51

52 decreased production of proinflammatory cytokines and reduced insulitis, characterized by the infiltration of pancreatic islets by immune cells. The mechanism through which apotransferrin elicits this shift in the immune response has not been elucidated but could potentially affect cellular iron acquisition by interfering with cellular iron uptake via TF. However the binding affinity for apotransferrin is considerably lower compared with holotransferrin arguing against this theoretical mechanism (174, 175). While current findings indirectly suggest that iron depletion may be protective against the development of autoimmune diabetes no investigation into the direct influence of systemic or cellular iron status on diabetic pathology has been performed at this time. Future studies will need to be carried out to determine the impact of iron status on the development of autoimmune diabetes. 52

53 CHAPTER 3 MATERIALS AND METHODS Animals and Diets Weanling male Sprague-Dawley rats (Charles River Laboratories), used in experiments detailed within chapter 4, were randomized (n=6/group) to receive either iron-deficient (FeD), iron-adequate (FeA), or iron-overloaded (FeO) diets. Purified diets were prepared according to the AIN-93G formulation, but with no added iron (FeD), 35 mg/kg ferric citrate (FeA), or 2% carbonyl iron (Sigma-Aldrich) (FeO). Iron contents of the diets, as determined by inductively coupled plasma mass spectroscopy (ICP-MS), were 5 ppm (FeD), 36 ppm (FeA), and 20,275 ppm (FeO). Diets were also modified to contain Avicel microcrystalline cellulose instead of cellulose and 20% sucrose instead of 10% sucrose (while reducing the amount of cornstarch accordingly) (176). After 3 weeks of feeding, overnight-fasted rats were sacrificed by exsanguination from the descending aorta. Blood was collected into heparinized syringes and then centrifuged to obtain plasma. Pancreata were quickly harvested, immediately frozen in liquid nitrogen, and maintained at -80 C for subsequent analyses. Weanling female NOD/ShiLtJ mice (The Jackson Laboratory),used in experiments detailed within chapter 6, were randomized to receive either iron-deficient (FeD), iron-adequate (FeA), or iron-overloaded (FeO) diets. Purified diets were prepared according to the AIN-76A formulation, but with no added iron (FeD), 270 mg/kg ferric citrate, or mg/kg carbonyl iron. Diets were modified to contain 5% wheat, to increase the diabetic potential of the diet, and avicel microcrystalline cellulose, replacing cellulose, to reduce contaminate iron. Additionally diets were heated in a convection oven at 125 C for 30 m, as this treatment has previously been reported to increase the incidence of diabetes 53

54 in mice fed a purified diet, and subsequently irradiated. Iron contents of the finalized diets, as determined inductively coupled plasma mass spectroscopy (ICP-MS) analysis were 14 mg/kg (FeD), 359 mg/kg (FeA), and 6629 (FeO) mg/kg dry matter. To account for potential heat degradation of vitamins during diet preparation an additional 10 g/kg vitamin mix was added to all diets. Animals were fed their respective diets until removal from the study, due to either assignment to prediabetic analysis at 10 wk of age, study termination at 30 wk of age, or detection of diabetes during the study. Testing for diabetes was carried out biweekly starting at 8 wk of age by using a handheld glucometer (Accu-Chek). Animals were considered to be diabetic if glycosuria was detected on 2 consecutive days followed by 2 consecutive blood glucose readings of >250 mg/dl. Upon removal from the study, animals were fasted overnight and sacrificed by isofluorane inhalation. Iron Status Parameters and Blood Glucose Concentrations Hemoglobin levels in heparinized blood were measured by using a HemoCue Hb 201+ Hemoglobin Analyzer (HemoCue). Liver non-heme iron levels were determined by colorimetric analysis of acid-digested tissues, as described previously (177). Briefly, tissues were weighed and digested in a solution of HCl, trichloroacetic acid. After incubation at 65ᵒ C for 20 h, the iron content of dissolved samples was measured by the addition of the dissolved sample to a chromogen reagent containing bathophenanthroline disulphonate, thioglycolic acid, and saturated sodium acetate. Color change, indicating non-heme iron concentration, was measured by using a spectrophotometer and compared with a dilution series made from an iron-reference solution (Fisher). Plasma iron and total iron-binding capacity (TIBC) were measured colorimetrically by the methods described by Cook (1985). Briefly, for plasma iron 54

55 determination a solution of trichloroacetic acid, HCl, and thioglycolic acid was added to plasma followed by centrifugation to precipitate proteins and reduce plasma iron. Plasma supernatant was then added to a chromogen solution composed of sodium acetate and ferrozine and iron concentration of unknown samples and a reference iron dilution series (Fisher) measured by using a spectrophotometer. TIBC was calculated by saturating plasma transferrin with iron followed by addition of magnesium carbonate to remove excess iron prior to the measurement of plasma iron. Blood glucose concentrations were determined in freshly collected heparinized blood by using a handheld glucometer (Accu-Chek). Pancreatic Mineral Concentrations Concentrations of iron, zinc, manganese, copper, and cobalt in pancreas samples were determined by using ICP-MS. Analyses were performed by the Michigan State University Diagnostic Center for Population and Animal Health. Histological Analysis 10% buffered formalin phosphate-fixed and paraffin embedded pancreases from 10 wk NOD were stained with hematoxylin and eosin and scored for insulitis. Insulitis was scored based on the scale 0, no insulitis; 1, peripheral-insulitis, 2, mild-insulitis (<50% of islet infiltrated); 3, severe insulitis ( 50% islet infiltrated). An average of 35 islets from at least 3 unique pancreatic sections per animal were scored and the average insulitis score reported. RNA Isolation and Assessment of RNA Integrity Frozen pancreas samples were submerged in liquid nitrogen and finely ground by using a mortar and pestle. After grinding, total RNA was isolated by using the RNeasy Mini Kit (Qiagen) following the manufacturer s protocol. Integrity of isolated 55

56 RNA was confirmed by denaturing agarose gel electrophoresis followed by visualization of 18S and 28S ribosomal RNA bands. Prior to microarray analysis, RNA integrity was additionally assessed by using the Agilent 2100 Bioanalyzer and RNA 6000 Nano Kit (Agilent). Microarray Analysis RNA samples (n=6) from each dietary group were pooled and analyzed by using a Rat GE 4x44K v3 Microarray (Agilent). Both FeD and FeO pooled cdna samples were individually compared with FeA in duplicate measurements. The assignment of fluorescent Cy3 or Cy5 dye to each comparison group was alternated between the duplicates. To identify differential gene expression, values of signal intensity were log2 transformed and normalized before the Student s t-test was performed for probespecific comparisons. Genes showing a statistically significant (P<0.05) log2- transformed fold change of at least ± 2 were analyzed to identify functional biological categories by using the Database for Annotation, Visualization and Integrated Discovery (DAVID) (178). Microarray analysis was conducted at the Interdisciplinary Center for Biotechnology at the University of Florida. The microarray data discussed herein have been deposited in NCBI's Gene Expression Omnibus (179) and are accessible through GEO Series accession number GSE44699 Relative mrna Quantification cdna was synthesized from total RNA by using the High Capacity cdna Reverse Transcription Kit (Applied Biosystems). Specific primers for genes of interest were generated (Table 4-5) and confirmed for specificity by using the NCBI Basic Local Alignment and Search Tool (BLAST) (180). Quantitative reverse transcriptase polymerase chain reaction (qrt-pcr) was performed by using Power SYBR Green 56

57 PCR Mastermix (Applied Biosystems) and an Applied Biosystems 7300 Real-Time PCR System. Dissociation curve analysis of PCR products revealed single products and all PCR amplification efficiencies were 100 ± 10%. Quantitation of mrna was determined by comparison to standard curves generated by four 10-fold serial dilutions of standard cdna. Transcript levels were normalized to those of cyclophilin B (peptidylprolyl isomerase B, PPIB). Western Blotting Pancreas samples were homogenized in ice-cold buffer containing 0.05 M Tris- HCl (ph 7.4), 0.05 M NaCl, M EDTA, 0.25% Tween-20, and Complete, Mini Protease Inhibitor Cocktail (Roche). Tissue homogenates were clarified by centrifugation at 10,000 x g for 10 minutes at 4 ºC, followed by sonication of the supernatant. Protein concentrations were determined by using the RC DC Protein Assay Kit (Bio-Rad). Proteins were mixed with Laemmli buffer, incubated at 70 C for 10 minutes, and then electrophoretically separated by sodium dodecyl sulfatepolyacrylamide gel electrophoresis (SDS-PAGE) on a 7.5% gel. Separated proteins were transferred to a polyvinyl difluoride (PVDF) membrane (Bio-Rad), and incubated in blocking buffer (5% nonfat dry milk in Tris-buffered saline (TBS)-Tween 20 (TBS-T)) for 1 hour. The blot was then incubated with rabbit anti-rat Alox15 antibody (kindly provided by Dr. James F. Collins, University of Florida), 1:8,000 dilution for 2 hour. After washing with TBS-T, the blot was incubated with horseradish peroxidase (HRP)-conjugated donkey anti-rabbit IgG secondary antibody (Amersham Biosciences), 1:10,000 dilution for 45 minutes. After washing with TBS-T and TBS, antibody binding was observed by using enhanced chemiluminescence (SuperSignal West Pico, Pierce) and the Fluorchem E imaging system (ProteinSimple). To indicate lane loading, the blot was 57

58 stripped and reprobed with a mouse anti-α-tubulin antibody (Sigma) at a 1:5,000 dilution, followed by an HRP-conjugated goat anti-mouse IgG secondary antibody (Santa Cruz) at a 1:10,000 dilution. Densitometry was performed by using AlphaView software (ProteinSimple). Cell samples were lysed and sonicated in RIPA buffer containing 150 mm sodium chloride, 1% IGEPAL,.5% sodium deoxycholate,.1% sodium dodecyl sulfate, and 50 mm tris-base, and Complete, Mini Protease Inhibitor Cocktail (Roche). The RC DC Protein Assay Kit (Bio-Rad) was used to determine lysate protein concentrations and samples were mixed with Laemmli buffer and incubated at 37 C for 20 minutes prior to western blotting analysis of ZIP14, ZIP8, and DMT1 or incubated at 95 C for 10 min for other proteins. The procedure of western blotting was performed as previously discussed with the exception of nitrocellulose replacing PVDF membranes (139). Primary antibodies used were rabbit anti-mdmt1 antibody at a concentration of 1:1,000 (generously contributed by Dr. Francois Cannone-Hergaux), rabbit anti-hzip8 at 1:5,000 concentration (Prestige Antibodies,Sigma Aldrich), rabbit anti-hzip14 at 1:5,000 concentration (Prestige Antibodies, Sigma Aldrich), rabbit anti- CCS at 1:200 (Santa Cruz), mouse anti-na + /K + -ATPase at 1:200 (Santa Cruz), Goat anti-ferritin light chain 1:4,000 (Novus Biologicals), mouse anti-α tubulin 1:5,000 (Sigma Aldrich). Cell Culture and Treatments βlox5 cells were cultured in low glucose (1g/L) Dulbecco s Modified Eagle s Medium (Cellgro) containing 100 units/ml penicillin, 100 ug/ml streptomycin, 10% fetal bovine serum (FBS) (Atlanta Biologicals), 1% Minimum Eagle s Medium Non-Essential Amino Acids (Corning), and 15 mm HEPES (Cellgro). βlox5 cells were treated with 100 µm ferric ammonium citrate (MP Biomedicals) and 1 mm ascorbate (EMD Millipore) for 58

59 24 h to promote iron loading. Recombinant interleukin 1β (IL-1β) (Peprotech) was added to media at a concentration of 100 U/ml for specified times. Primary human islets of at least 90% purity and viability were obtained from the Integrated Islet Distribution Program (IIDP) and were cultured in Ultra-Low attachment plates (Corning) with CMRL1066 media containing 10% FBS, 100 units/ml penicillin, and 100 ug/ml streptomycin. Depletion of cellular iron or iron loading was performed by treating islets with 50 µm Deferoxamine (Hospira), or 100 µm ferric ammonium citrate (MP Biomedicals) and 1 mm ascorbate (EMD Millipore), respectively, for 48 h. Prior to transfection islets were dissociated by incubating in Accutase (Life Technologies) for 15 min at 37 C followed by pipetting to ensure a single cell suspension. Dissociated islets were transfected for 48 h. All cells were maintained in 5% CO2 at 37 C. In Vitro Glucose Stimulated Insulin Secretion Triplicate groups of approximately 10 human islets were selected for glucose stimulated insulin secretion (GSIS) testing and incubated in low glucose 2.8 mm Kreb sringer buffer (KRB) containing 25 mm HEPES, 115 mm NaCl, 24 mm NaHCO3, 5 mm KCl, 1 mm MgCl26H20,.1% BSA, 2.5 mm CaCl2, adjusted to ph 7.4 for 30 m. After incubation islets were transferred to fresh 2.8 mm glucose KRB and incubated for 1 hr after which KRB was sampled for determination of basal insulin secretion and replaced with high-glucose KRB containing 22 mm glucose. Islets were again incubated for 1 hr before high-glucose KRB was sampled and removed. Insulin concentrations in sampled KRB were measured by using the ALPCO Ultra-sensitive ELISA KIT (Alpco) and used to determine total insulin secretion during each time period. Results were normalized to levels of islet DNA measured by using the Quant-IT PicoGreen dsdna assay kit (Life Technologies). 59

60 Mouse Islet Isolation Islets from 10 wk mice used in glucose-stimulated insulin secretion testing were isolated by using a collagenase infiltration method with Liberase TL (Roche) as previously detailed (181). In brief, pancreases were incubated at 37 C for 15 m and islets released by manually agitating the tissues with ice-cold HBSS followed by centrifugation at 300 g at 4 C to pellet the tissue. Manual agitation and centrifugation was repeated 5 times. After dissolution of the pancreas, islets were handpicked and transferred to RPMI1640 medium for culture prior to analysis. Islets from 2-3 mice from each dietary group were pooled and lysed in RIPA buffer (50 mm Tris-HCl, 1% NP-40, 0.25% Na-deoxycholate, 150 mm NaCl, and 1 mm EDTA, adjusted to ph 7.4) for western blot analysis. Islet iron status was determined by western blotting for transferrin receptor by using mouse anti-transferrin receptor (Life Technologies) at a concentration of 1:4000 ug/ml. Determination of DMT1, ZIP8, and ZIP14 mrna Copy Numbers Total RNA was isolated from primary human islets by using RNAzol (Molecular Research center) following the manufacturer s protocol. cdna synthesis from isolated RNA was carried out by using the High Capacity cdna Archive Kit (Life Technologies). Quantitative RT-PCR was performed by using SYBR Select Master Mix (Life Technologies) and a CFX96 Real Time PCR Detection System (Bio-Rad). Copy numbers of DMT1, ZIP8, and ZIP14 were calculated by comparing the Ct values from human islet cdna samples to standard curves constructed from known quantities of the plasmids pbluescriptr-hdmt1 (BC100014; Addgene), pcmv-sport6-hzip8 (BC012125; Open Biosystems), and pcmv-xl4-hzip14 (BC015770; Open Biosystems). The primers used to detect DMT1 (F, 5 - TGCATCTTGCTGAAGTATGTCACC-3 and R, 5-60

61 CTCCACCATCAGCCACAGGAT-3 ), ZIP14 (F, 5 -CAAGTCTGCAGTGGTGTTTG-3 and R 5 -GTGTCCATGATGATGCTCATTT-3 ), and ZIP8 (F, 5 - CAGTGTGGTATCTCTACAGGATGGA-3 and R, 5 -CAGTTTGGGCCCCTTCAAA-3 ) were designed to detect all known mrna transcripts. sirna Knockdown of DMT1, ZIP8, and ZIP14 SMARTpool sirna targeting either human DMT1 or ZIP14 (Thermo Scientific) and Flexitube sirna targeting ZIP8 (Qiagen) were used to suppress mrna levels. Transfection was performed by using Lipofectamine sirnaimax (Life Technologies) and Opti-MEM Media (Life Technologies) for sirna and reagent suspension following the manufacturer s protocol to yield a final concentration of 12 nm sirna after addition to the complex to plated cells. In brief Opti-MEM media was added to separate vials of either sirna or Lipofectamine sirnaimax after which the contents of each vial were combined and allowed an incubation period of 15 min. After incubation 500 µl of the transfection mixture was then added to each well of a 6 well plate containing 2 ml of cell culture media and cultured for 48 h prior to collection. Successful knockdown was confirmed by immunoblotting. Overexpression of DMT1, ZIP8, and ZIP14 Cultured βlox5 cells were transiently transfected with either pcdna3.1hdmt1- (1A/+IRE) flag, generously contributed by Dr. Natascha Wolff, pcmv-sport6-hzip14 (BC015770), pcmv-sport6-hzip8 (BC012125), or pcmv-sport6-empty vector, by using Effectene Transfection Reagent (Qiagen) according to the manufacture s protocol. After 24 h cells were harvested for confirmation of overexpression or used in iron uptake experiments. 61

62 Immunofluorescencse Paraffin-embedded tail sections of human pancreas were obtained through the Network for Pancreatic Organ Donors with Diabetes (npod). Paraffin was cleared with xylene and tissues rehydrated in stages. After hydration slides underwent heat-induced epitope retrieval in sodium citrate buffer containing 10 mm sodium citrate,.05% tween 20, and adjusted to ph 6.0 with HCl. Slides were then briefly cooled in distilled H20 and washed with TBS to remove residual sodium citrate buffer. Washed slides were then incubated in blocking buffer containing 2% goat serum for 30 min to prevent nonspecific binding of secondary antibody. During the primary antibody incubation human sections were triple stained for insulin, glucagon, and either DMT1, ZIP8, or ZIP14 by using guinea pig anti-insulin (1:200, Abcam), mouse anti-glucagon (1:1,000, Abcam), and either rabbit anti-dmt1 (1:1,000, Prestige Antibodies Sigma Aldrich), rabbit anti-zip8 (1:250 Peprotech), and rabbit anti-zip14 (1:1,000, Prestige antibodies Sigma Aldrich) antibodies. During staining for ZIP8 slides were also permeabilized with 5% triton x-100 for 15 min after antigen retrieval. To control for nonspecific primary antibody binding serial sections were also triple stained with nonimmune rabbit IgG replacing the primary antibody for DMT1, ZIP8, or ZIP14 at the same concentration. The incubation with primary antibodies was carried out at 4 degrees C overnight in a humidified chamber. After the primary incubation slides were washed with TBS and incubated for 2 h at room temperature with the secondary antibodies goat anti-guinea pig Alexa Flour 594 (1:250, Life Technologies) for insulin, goat anti-mouse Pacific Blue (1:250, Life Technologies) for glucagon, and goat anti-rabbit Alexa Flour 488 (1:250, Life Technologies) for either DMT1, ZIP8, or ZIP14 to obtain a fluorescent signal. After this incubation period slides were washed with TBS and cover slips carefully mounted. Confocal microscopy was 62

63 performed and images obtained by using an Olympus IX2-DSU spinning disk confocal fluorescent microscope equipped with a Hamamatsu ORCA-AG camera and 3i SlideBook v4.2 software. Cellular NTBI Uptake Cultured βlox5 cells or isolated human islets were washed twice with serum free media (SFM) and incubated for 1 h in SFM containing 2% bovine serum albumin to bind residual transferrin and prevent iron uptake via transferrin-bound iron endocytosis. Following this incubation period cells were again washed with SFM before the addition of media containing 2 μm ferric ammonium citrate (MP Biomedicals) radiolabeled with 59 Fe and 1 mm ascorbate. Cells were incubated with radiolabeled media for 2 h during sirna experiments and for 1 h during overexpression experiments. After the incubation period media was aspirated and cells were washed 3 times with an iron chelator solution containing 1 mm diethylenetriaminepentaaetic acid and 1mM bathophenanthroline disulfonate to remove residual radiolabeled iron. Cells were then lysed in SDS lysis buffer and the lysate collected measured for radioactivity by using a WIZARD2 gamma counter (PerkinElmer). Counts for each sample were then normalized to cellular protein concentrations. To account for variation in 59 Fe uptake between independent experiments, NTBI uptake relative to cellular protein was measured in separate groups of βlox5 cells and dissociated human islets to establish a reference level of 59 Fe uptake in each independent experiment. Reference levels from separate independent experiments were compared and used to generate adjustment ratios by which experimental groups would be multiplied by to produce a control groups with a relative 59 Fe uptake value of 1 while preserving variance in the experimental control group for statistical analysis. 63

64 Generation of Transgenic MIP-Zip14-HA Mice The transgenic construct previously used to generate mice selectively overexpressing green fluorescent protein (GFP) in β cells under control of the mouse insulin 1 promoter, MIP-GFP was generously contributed by Drs. Graeme Bell and Manami Hara (University of Chicago, Chicago, IL) and has been detailed previously (182). In brief the construct consisted of a fragment of the mouse insulin 1 promoter, the GFP coding region, and an intron region of the human growth hormone gene to enhance expression. The region of this vector encoding GFP was excised by restriction digestion with Xho1 and hemagglutinin antigen (HA) tagged Zip14 was subcloned into the MIP vector, resulting in the generation of a MIP-ZIP14-HA construct. The MIP- Zip14-HA-hGH fragment was isolated from the vector backbone by restriction digestion with HindIII and SfiI followed by agarose gel separation, extraction, and purification. DNA was introduced into fertilized embryos from C57BL/6J mice by pronuclear microinjection carried out at the Mouse Models Core at the University of Florida. Founders were screened based on the presence of HA tag using the primers 5 TACCCTTACGACGTGCCT 3 and 5 AGGAGAGAGGCCAGGTTAAT 3 to differentiate between endogenous Zip14 and successful transgene insertion. Transgene copy numbers were determined by the comparison of genomic DNA to a standard curve constructed from known amounts of MIP-Zip14-HA plasmid DNA. The primer set used was 5 CCTTACGACGTGCCTGATTA 3 and 5 TTCAGCTGTGTCAGGGTAAG 3, targeting HA tag. Founders were bred with wild-type C57BL/6J mice to establish transgenic lines. Overexpression of Zip14-HA in β cells was confirmed by immunofluorescence detection of HA tag in mouse pancreatic sections (Mouse anti-ha 64

65 1:100, Roche) following the protocol previously detailed for IF in human pancreatic sections. Statistical Analysis Data were analyzed for statistical significance by using one-way ANOVA and Tukey s multiple comparison post-hoc test or Student s T-test (GraphPad Prism) where indicated. Unequal variance between groups was accounted for by log transformation, where applicable, to normalize variance before statistical analysis. Survival curves were analyzed for significance by using the Gehan-Breslow-Wilcoxon Test (Graphpad Prism). 65

66 CHAPTER 4 TRANSCRIPTIONAL PROFILING OF PANCREATIC GENE EXPRESSION IN RESPONSE TO DIETARY IRON LOADING OR DEFICIENCY 1 The association between excess iron and pancreatic dysfunction has long been observed in the iron overload disorder hereditary hemochromatosis (183). Patients with hemochromatosis have a higher prevalence of diabetes, decreased insulin secretory capacity, and impaired glucose tolerance relative to the normal population (45). The Hfe knockout mouse, the animal model of hemochromatosis, also displays alterations in pancreatic function, including decreased insulin secretory capacity (146). In humans, insulin secretory capacity and glucose tolerance improves after iron stores are normalized by phlebotomy, suggesting that tissue iron levels are an important determinant of insulin action (184). Consistent with this idea are animal studies showing that a decrease in iron stores (in response to phlebotomy or a low-iron diet) can increase insulin secretion and pancreatic insulin levels (163, 164). However, iron depletion to the point of iron deficiency and anemia has been shown to negatively affect glucose homeostasis by increasing blood glucose concentrations (185). The effects of iron overload and deficiency on glucose homeostasis are likely mediated, at least in part, by iron-related changes in the expression of genes involved in glucose metabolism. For example, iron deficiency has been reported to be associated with higher levels of rate-limiting gluconeogenic enzymes in rat liver (186) and ironloaded Hfe knockout mice display increased glucose uptake by isolated soleus muscle 1 Reprinted with permission from Coffey R, Nam H, Knutson MD. Microarray analysis of rat pancreas reveals altered expression of Alox15 and regenerating islet-derived genes in response to iron deficiency and overload. PLoS One. 2014;9:e

67 and decreased glucose oxidation by isolated hepatic mitochondria (161, 187). Little information, however, exists regarding iron-related gene expression in the pancreas. Given that the pancreas hormonally controls whole-body glucose homeostasis, the aim of the present study was to examine global changes in pancreatic gene expression in response to iron deficiency and overload. Identification of pancreatic genes that are regulated by iron status may offer insight not only into how iron status perturbs glucose homeostasis, but also how iron overload may contribute to β cell destruction and diabetes. Results Body Weight, Iron Status, and Blood Glucose Concentrations After 3 weeks of feeding the experimental diets, body weights were significantly lower in the FeD and FeO groups relative to FeA controls, but did not differ between FeD and FeO animals (Table 4-1). Liver non-heme iron concentrations, an indicator of body iron stores, confirmed that rats fed the FeD diet became iron deficient whereas rats fed the FeO diet became iron overloaded. In FeO animals, liver non-heme iron concentrations were nearly 40 times higher than controls. FeD rats also became anemic with hemoglobin levels that were 41% lower than normal (Table 4-1). Blood glucose concentrations were elevated in FeD rats compared with FeA controls, whereas those in FeO animals did not differ from controls (Table 4-1). Pancreatic Mineral Concentrations In FeO rats, pancreatic iron concentrations were 155% higher than those in FeA animals, whereas in FeD rats, iron concentrations were 40% lower than controls (Table 4-2). Given that iron deficiency and overload can affect tissue concentrations of other trace minerals (176), I measured pancreatic concentrations of zinc, manganese, copper, 67

68 and cobalt (Table 4-2). Pancreatic zinc concentrations were found to be 26% higher in FeD rats, and copper concentrations were 74% lower in FeO rats, when compared with FeA controls. By contrast, pancreatic manganese and cobalt concentrations did not differ among groups. It should be noted that the concentrations of zinc, manganese, copper, and cobalt did not differ among the experimental diets (data not shown). Identification and Classification of Differentially Expressed Genes by Microarray Analysis Microarray analysis was used to identify candidate genes that are differentially expressed in FeD and FeO pancreas, especially those that may influence the risk for diabetes. Using a log2 fold change of ± 2 and P < 0.05 as a cutoff, I identified a total of 230 genes as differentially expressed in FeD and FeO pancreas relative to FeA pancreas. In FeD pancreas 66 genes were differentially expressed (56 down-regulated and 10 up-regulated) (Figure 4-1A). In FeO pancreas 164 genes were differentially expressed (82 down-regulated and 82 up-regulated) (Figure 4-1B). The differentially expressed genes were analyzed by using DAVID bioinformatics resources to identify gene ontology categories. In FeD pancreas, the category with the highest number of genes was lipid transport (7 genes), followed by antimicrobial (4 genes), neuropeptide (4 genes), and pancreatitis-associated protein (3 genes) (Figure 4-1A). All but two of the genes in these categories were down-regulated in FeD pancreas. By contrast, in FeO pancreas, most gene ontology categories were enriched with upregulated genes (Figure 4-1B). For example, 6 of 8 genes were up-regulated in the pattern binding category in FeO pancreas. Of note, the gene ontology category pancreatitis-associated protein was identified in both FeD and FeO pancreas. Lists of the genes in each category along with fold change are provided in Table 4-6 and Table 68

69 4-7. The top 10 most up-regulated and down-regulated genes in FeD and FeO pancreas, ordered by mean magnitude change (P< 0.05), are shown in Table 4-3 and Table 4-4. The genes listed in Table 4-6, Table 4-7, Table 4-3, and Table 4-4 were surveyed in the literature to identify those with reported associations with diabetes and/or glucose homeostasis, and several of these were subsequently selected for validation by qrt-pcr of individual rat samples (n=6 group). Confirmation of Up-Regulation of Alox15 Expression by QRT-PCR and Western Blotting According to the microarray analysis, the most up-regulated gene in FeD pancreas was Alox15 (arachidonate 15-lipoxygenase) (Table 4-3). Alox15 catalyzes the oxidation of polyunsaturated fatty acids, such as arachidonic acid, during the formation of inflammatory mediators and has been linked to the development of type 1 diabetes (188, 189). QRT-PCR analysis confirmed the up-regulation of Alox15 mrna levels in FeD pancreas (Figure 4-2A), and western blot analysis revealed higher Alox15 protein levels in FeD pancreas (Figure 4-2B). Alox15 protein levels were also found to be higher in FeO pancreas compared with FeA controls despite no increase in Alox15 mrna levels. As western blotting controls for rat Alox15, jejunum samples from ironadequate (JA) and iron-deficient (JD) rats were analyzed in parallel with the rat pancreas samples. Consistent with a previous study by Collins et al. (190), Alox15 was detected at approximately 70 kda and was markedly up-regulated in iron-deficient jejunum (JD) (Figure 4-2B). Densitometric analysis of the western blots indicated that Alox15 protein levels were approximately 8- and 9-fold higher (P<0.001) in FeD and FeO pancreas, respectively, compared with FeA controls (n=6/group; data not shown). 69

70 Confirmation of Reg Family mrna Levels by QRT-PCR Of the genes showing positive regulation during FeO, the most elevated belonged to the Reg family of regenerating islet-derived genes (Table 4-4). The genes of the Reg family, most notably Reg1a, have been linked to pancreatic regeneration as well as cellular growth and survival during oxidative stress ( ). Consistent with the microarray data, qrt-pcr analysis revealed that mrna levels of these genes were up-regulated in FeO pancreas. Mean mrna levels of Reg1a, Reg3a, and Reg3b were found to be 21, 37, and 18 times higher, respectively, in FeO pancreas than FeA controls (Figure 4-3). Also consistent with the microarray, qrt-pcr analysis found that Reg1a mrna levels were up-regulated in FeD pancreas. Reg mrna levels varied considerably among rats, particularly in the FeO and FeD groups in which two or three values were notably higher than the others. Repeated analyses confirmed that the high values do not represent analytical artifacts. In the FeD, FeA, and FeO groups, the high values for Reg1a and Reg3a (but not Reg3b) mrna are from the same animals, suggesting that Reg1a and Reg3a are up-regulated in parallel. Discrepancies Between Microarray and QRT-PCR Analysis Results According to the microarray, both FeD and FeO pancreas showed large downregulations in the expression of Fabp1, Fabp2, and Apoa1 (Tables 3 and 4), which clustered in the gene ontology category of lipid transport (Table S3). As all of these genes have been associated with diabetes in the surveyed literature ( ) they were selected for follow-up study. However, the expression levels of these genes in pancreas were found to be below the detection limit of qrt-pcr, similar to previous studies that have failed to detect the expression of these genes in pancreas (198, 199). 70

71 Discussion Because iron status can affect glucose homeostasis, I sought to identify glucose metabolism-related genes in rat pancreas whose expression might be affected by iron deficiency or overload. Unexpectedly, our microarray data, and subsequent functional enrichment analysis by DAVID, did not identify any changes in the expression of genes known to be involved in glucose metabolism. A limitation to our study is that many of the glucose-responsive genes are found in islet cells (200), which constitute only 1-2% of the mass of the pancreas, and therefore changes in islet-cell gene expression may not be readily detectable in whole pancreas tissue. The most notable finding from our microarray analyses was the identification of differentially expressed genes that are associated with diabetes and/or pancreatic stress. More specifically, Alox15 was identified as the most up-regulated mrna in iron deficiency, and Reg family transcripts Reg1a, Reg3a, and Reg3b, were found to be markedly up-regulated in iron overload. Alox15 encodes arachidonate 15-lipoxygenase, a non-heme iron-containing enzyme that catalyzes the oxygenation of polyunsaturated fatty acids to form inflammatory mediators (201). Despite the name suggesting 15-lipoxygenase activity, Alox15 in rodents has been demonstrated to function primarily as a 12-lipoxygenase with secondary 15-lipoxygenase function (202). Therefore the term leukocyte 12- lipoxgenase, as well as the hybrid term 12/15-lipoxygenase, is often used in reference towards Alox15. A link between iron deficiency and Alox15 was first reported in a microarray study by Collins et al. (190), who identified Alox15 as the most strongly induced gene in the intestine of iron-deficient rats. Alox15 has also been identified as the most highly induced gene in microarray studies of iron-deficient rat liver (186) and brain (203). Similar to Collins et al. (190), I found that elevated Alox15 mrna levels 71

72 were associated with higher Alox15 protein levels. Protein levels of Alox15 were also found to be elevated in iron-loaded rat pancreas, despite no up-regulation of Alox15 mrna levels, suggesting post-transcriptional regulation under iron-overload conditions. In the pancreas, Alox15 is present in β cells (204) where it appears to play a role in the pathogenesis of diabetes. Genetic deletion of a locus containing Alox15 has been shown to protect nonobese diabetic (NOD) mice from developing autoimmune diabetes, with knockout mice exhibiting superior islet mass and glucose tolerance (188). Recent experiments using sirna against Alox15 provide evidence that diminished Alox15 levels are responsible for the protective phenotype (189). Resistance to the development of a diabetic phenotype induced via streptozotocin was also observed in mice lacking Alox15 (205). It has been proposed that Alox15 contributes to the development of diabetes via its ability to catalyze the formation of inflammatory mediators such as 12-HETE (hydroxyeicosatetraenoic acid), which causes β cell dysfunction and death (204, 206, 207), recently linked to excessive production of reactive oxygen species (208). Our observation that iron deficiency causes a marked elevation in Alox15 mrna and protein levels in the pancreas raises the possibility that iron deficiency in addition to iron overload may increase the risk of developing diabetes through up-regulation of Alox15. Such a possibility appears opposite to recent studies showing a protective effect of iron restriction on diabetes risk. For example, Cooksey et al. (163) observed that an iron-restricted diet enhanced β cell function and insulin sensitivity in the ob/ob mouse model of type 2 diabetes. Similarly, Minamiyama et al. (164) found that feeding an iron-restricted diet to type 2 diabetic rats normalized 72

73 plasma insulin levels. It should be noted, however, that in the study by Cooksey et al. (163), iron-restriction did not result in iron deficiency or anemia in contrast to our study. Although it is well known that individuals with iron overload are susceptible to developing diabetes (183), the molecular mechanisms involved remain poorly understood. Our observation that iron-overloaded rats have highly elevated Alox15 protein levels in the pancreas suggests that Alox15 may contribute to β cell loss and β cell dysfunction in iron overload. Indeed, the pancreases of iron-loaded rats appear to be under stress as indicated by the elevated expression of the regenerating isletderived gene family members Reg1a, Reg3a, and Reg3b. As indicated by their name, Reg genes were first identified by their strong induction in regenerating pancreatic islets in response to stress/damage (209). Reg1a is a 165-a.a secreted protein that has been shown to play an important role in β cell function in vivo (191). Disruption of murine Reg1 (the ortholog of rat Reg1a) resulted in decreased proliferative capacity of pancreatic β cells (210), whereas administration of recombinant rat Reg1a resulted in β cell regeneration and reversal of diabetes in rats after surgical resection of 90% of the pancreas (191). Similar to Reg1, Reg3a and Reg3b have been associated with islet regeneration and protection against diabetes (193, 211). Reg3 proteins are also known as pancreatitis-associated proteins (PAP) that become highly expressed in acinar cells in response to injury (212). Our observation of elevated Reg3 expression in iron-loaded rat pancreas is consistent with a previous report of hypotransferrinemic mice, which displayed pancreatic iron loading and markedly elevated expression of Reg3 mrna (213). However, in that study, a time course analysis of pancreatic iron loading indicated that Reg3/PAP mrna became detectable only when pancreatic non-heme 73

74 iron concentrations had reached levels that were ~50 times normal. In our study of ironloaded rats, I found that even modest elevations in pancreatic iron concentrations (2.5 times normal) are associated with enhanced expression of Reg3 mrna, suggesting that Reg mrna levels could serve as an early biomarker of iron-related pancreatic stress/damage in rats. The apparent discrepancy in pancreatic iron load required to elicit increased Reg3 expression between mice and rats is likely attributable to interspecies variability. Mice are largely resistant to the degenerative effects of pancreatic iron loading whereas rats exhibit acinar cell degradation, indicative of pancreatic damage, following dietary iron overload (136, 142). One caveat is that the elevated pancreatic Reg expression in iron-loaded rats could be confounded by the abnormally low (i.e., ~25% of normal) copper concentrations in these animals. Copper deficiency in rats has been shown to result in pronounced atrophy of the exocrine pancreas (214). Pancreatic atrophy is observed during pancreatitis, a state which promotes extensive expression of Reg family genes (215). Also, during copper deficiency islet hyperplasia and β cell neogenesis have been documented (216) in line with the islet-regenerating properties of Reg proteins. More research is needed to determine if low copper levels induce the expression of these genes. In conclusion, microarray analysis of rat pancreas has revealed that iron deficiency and overload increase the expression of one or more genes strongly associated with diabetes and pancreatic stress, thus highlighting the importance of iron status in the pancreas. 74

75 Table 4-1. Body weight, iron status, and blood glucose concentration of rats Group Body weight Liver non-heme iron Hemoglobin Glucose (g) (μg/g) (g/dl) (mg/dl) FeD ± 20.2 a 3.5 ± 3.6 a 7.5 ± 2.2 a ± 39.0 b FeA ± 11.1 b 25.4 ±17.7 b 12.8 ± 0.4 b 99.0 ± 16.0 a FeO ± 23.6 a ± c 13.6 ± 0.6 b ± 18.5 a FeD, iron deficient; FeA, iron adequate; FeO, iron overloaded. Values represent means ± SD, n = 6. Means without a common superscript are significantly different P<

76 Table 4-2. Pancreatic mineral concentrations Group Iron Zinc Manganese Copper Cobalt FeD 38.2 ± 5.7 a ± 18.0 b 8.7 ± ± 1.0 b 0.05 ± 0.02 FeA 63.7 ± 14.3 b 85.0 ± 16.0 a 6.2 ± ± 0.7 b 0.03 ± 0.01 FeO ± 59.5 c 75.2 ± 8.5 a 7.0 ± ± 0.0 a 0.04 ± 0.01 Mineral concentrations (μg/g dry weight) were measured by using ICP-MS. Values represent means ± SD, n = 6 Means without a common superscript are significantly different P<

77 Table 4-3. Top 10 most up-regulated and down-regulated genes in FeD pancreata Gene Name Symbol Accession Fold change a arachidonate 15-lipoxygenase Alox15 NM_ L-threonine dehydrogenase Tdh NM_ RT1 class I, locus CE5 RT1-CE5 NM_ S100 calcium binding protein A9* S100a9 NM_ transient receptor potential cation Trpc3 NM_ channel, subfamily C, member 3 vascular endothelial growth factor B Vegfb NM_ regenerating islet-derived 1 alpha* Reg1a NM_ secretoglobin, family 2A, member 1 Scgb2a1 NM_ potassium intermediate/small conductance Ca-activated channel, subfamily N, member 1 Kcnn1 NM_ alanine-glyoxylate aminotransferase 2 Agxt2 NM_ fatty acid binding protein 1, liver* Fabp1 NM_ fatty acid binding protein 2, intestinal* Fabp2 NM_ LOC protein LOC NM_ proline-rich acidic protein 1 Prap1 NM_ s100 calcium binding protein G S100g NM_ monoacylglycerol O-acyltransferase 2* Mogat2 NM_ apolipoprotein A-I* Apoa1 NM_ camp responsive element binding protein 3-like 3 Creb3l3 NM_ similar to carboxylesterase 5 LOC XM_ carboxylesterase 5-like LOC XR_ Fold change log2 relative to iron-adequate rat pancreas.* In gene ontology category in Figure 2-1 and supplemental Table S2. 77

78 Table 4-4. Top 10 most up-regulated and down-regulated genes in FeO pancreata Gene name Symbol Accession Fold change a regenerating islet-derived 3 alpha* Reg3a NM_ regenerating islet-derived 3 beta* Reg3b NM_ extracellular proteinase inhibitor Expi NM_ regenerating islet-derived 1 alpha* Reg1a NM_ prepronociceptin Pnoc NM_ beta-galactosidase-like protein Bin2a NM_ calmodulin-like 3 Calml3 NM_ vascular endothelial growth factor B* Vegfb NM_ phospholipase A2, group IIA* Pla2g2a NM_ upper zone of growth plate and cartilage matrix associated Ucma NM_ similar to Robo-1 LOC NM_ fatty acid binding protein 2, intestinal * Fabp2 NM_ fatty acid binding protein 1, liver* Fabp1 NM_ proline-rich acidic protein 1 Prap1 NM_ lectin, galactoside-binding, soluble, 4 Lgals4 NM_ apolipoprotein A-I * Apoa1 NM_ hydroxysteroid (17-beta) dehydrogenase 2 Hsd17b2 NM_ S100 calcium binding protein G S100g NM_ LOC protein* LOC NM_ hypothetical protein LOC LOC NM_ Fold change log2 relative to iron-adequate rat pancreas. *In gene ontology category in Figure 1 and supplemental Table S3. 78

79 Table 4-5. Primers for qrt-pcr Symbol Alox15 Reg1a Reg3a Reg3b Fabp1 Fabp2 Apoa1 Name arachidonate 15- lipoxygenase regenerating islet-derived 1 alpha regenerating islet-derived 3 alpha regenerating islet-derived 3 beta fatty acid binding protein 1, liver fatty acid binding protein 2, intestinal apolipoprotein A-I GenBank Accession No. NM_ NM_ NM_ NM_ NM_ NM_ NM_ Forward primer (5'-3') CCCTGTCGG GACTCGGAA GC TTGTCTCAGC CTGCAGAGA TTG CCGTGGTAA CTGTGGCAG TCT AAAGATGATG AGAGTTAAGA TGTTGCA AGGTCAAGG CAGTGGTTAA GAT TCACTGGGA CCTGGACCA TG TCCACTTTGG GCAAACAGC TGAAC Reverse primer (5'-3') CCAGTGCCC TCAGGGAGG CT CATGATGAG CAGCAGACT GTCTT GTGATGGTC TCCCCACTTC AG AGCAGCATC CAGGACATG ACT TGTCATGGTA TTGGTGATTG TGT CATATGTGTA GGTCTGGAT TAGT TCCTGTAGG CGACCAACA GTTGAA 79

80 Table 4-6. Functional categories of pancreatic genes differentially expressed in response to iron deficiency Functional category Gene symbol Description GenBank number Fold change Lipid transport Npc1l1 NPC1 (Niemann-Pick NM_ disease, type C1, gene)-like 1 Mttp microsomal NM_ triglyceride transfer protein Apoc3 apolipoprotein C-III NM_ Apoa1 apolipoprotein A-I NM_ Mogat2 monoacylglycerol O- NM_ acyltransferase 2 Fabp2 fatty acid binding NM_ protein 2, intestinal Fabp1 fatty acid binding NM_ protein 1, liver Antimicrobial S100a9 S100 calcium binding NM_ protein A9 Defa6 defensin alpha 6 NM_ Defa-rs1 defensin alpharelated NM_ sequence 1 Defa8 defensin alpha 8 NM_ Neuropeptide Calca calcitonin-related NM_ polypeptide alpha Cartpt CART prepropeptide NM_ Gal galanin NM_ prepropeptide Vip vasoactive intestinal NM_ peptide Pancreatitisassociated Reg1a regenerating islet- NM_ protein derived 1 alpha Reg3g regenerating isletderived NM_ gamma Reg4 regenerating isletderived family, member 4 NM_ Fold change log2 relative to iron-adequate rat pancreas 80

81 Table 4-7. Functional categories of pancreatic genes differentially expressed in response to iron overload Functional category Gene symbol Description Genbank number Fold change Pattern binding Vegfb vascular endothelial growth NM_ factor B pla2g5 phospholipase A2, group V NM_ Glutathione and drug metabolism Abp1 amiloride binding protein 1 NM_ (amine oxidase, coppercontaining) Colq collagen-like tail subunit (single strand of homotrimer) of asymmetric acetylcholinesterase NM_ Cyp2e1 cytochrome P450, family 2, NM_ subfamily e, polypeptide 1 Aox1 aldehyde oxidase 1 NM_ Pancreatitisassociated protein Itgam integrin, alpha M NM_ Ccl7 chemokine (C-C motif) ligand NM_ Prg4 proteoglycan 4, NM_ (megakaryocyte stimulating factor, articular superficial zone protein, camptodactyly, arthropathy, coxa vara, pericarditis syndrome) Tpsb2 tryptase beta 2 NM_ Fmo1 flavin containing NM_ monooxygenase 1 Gpx2 glutathione peroxidase 2 NM_ Gsta5 glutathione S-transferase Yc2 NM_ subunit Gsta2 glutathione S-transferase A2 NM_ Loc LOC protein NM_ Reg3a regenerating islet-derived 3 alpha Reg3b regenerating islet-derived 3 beta Reg1a regenerating islet-derived 1 alpha Reg3g regenerating islet-derived 3 gamma Reg4 regenerating islet-derived family, member 4 NM_ NM_ NM_ NM_ NM_

82 Table 4-7. Continued Functional category Gene symbol Description Genbank number Fold change Digestive system process Defensin propeptide Regulation of lipid transport Tff1 trefoil factor 1 NM_ Mogat2 monoacylglycerol O- NM_ acyltransferase 2 Fabp1 fatty acid binding NM_ protein 1, liver Fabp2 fatty acid binding protein 2, intestinal NM_ Defa24 defensin, alpha, 24 NM_ Defa-rs1 defensin alpha-related NM_ sequence 1 Defa8 defensin alpha 8 NM_ Adipoq adiponectin, C1Q and NM_ collagen domain containing Apoc3 apolipoprotein C-III NM_ Apoa1 apolipoprotein A-I NM_ Protease activity Cma1 chymase 1, mast cell NM_ Mcpt1l3 mast cell protease 1- ENSRNOT like 4 Tmprss8 transmembrane NM_ protease, serine 8 (intestinal) Spink4 serine peptidase NM_ inhibitor, Kazal type 4 Mep1b meprin 1 beta NM_ Mmp7 matrix NM_ metallopeptidase 7 Complement Cfd complement factor D NM_ activation (adipsin) C4bpa complement NM_ component 4 binding protein, alpha C6 complement component 6 NM_ Fold change Log2 relative to iron-adequate rat pancreas 82

83 Figure 4-1. Functional classification of pancreatic genes up- or down-regulated in response to iron deficiency and iron overload. Microarray analysis identified a total of 66 differentially expressed genes in response to iron deficiency (Panel A) and 164 genes in response to iron overload (Panel B). Genes were then subjected to DAVID analysis to identify functional categories. A) Functional gene categories identified in iron-deficient pancreas and the number of genes in each category. B) Functional gene categories identified in iron-overloaded pancreas and the number of genes in each category. 83

84 Figure 4-2. Effect of iron deficiency and overload on rat pancreatic Alox15 expression. A) Total RNA was isolated from rat pancreas and the relative transcript abundance of Alox15 was measured by using qrt-pcr. Transcript abundances were normalized to the housekeeping transcript cyclophilin B and are expressed relative to the FeA group mean (set to 1). B) Immunoblot analysis of Alox15 from a representative sample of FeD, FeA, and FeO rats. Jejunum from iron-adequate (JA) and iron-deficient (JD) rats were analyzed in parallel to serve as negative and positive controls respectively for immunodetection of Alox15. The blot was stripped and reprobed for tubulin to indicate protein loading among lanes. Values are expressed as the mean ± SEM, n=6. Asterisks indicate a significant difference relative to FeA controls, **P<

85 Figure 4-3. Effect of iron deficiency and overload on the expression of pancreatic Reg family genes.total RNA was isolated from rat pancreas and the relative transcript abundances of Reg family genes were determined by qrt-pcr. Transcript abundances were normalized to levels of cyclophilin B and are expressed relative to the FeA group mean (set to 1). Statistical significance was determined by one-way ANOVA. Asterisks indicate a significant difference relative to FeA controls *P<0.05, **P<0.01, ***P<

86 CHAPTER 5 MECHANISMS OF NTBI UPTAKE BY HUMAN β CELLS Iron is an essential trace mineral necessary for numerous biological functions, including oxidation-reduction reactions, due, in part, to the ability of iron to exist in multiple oxidation states. While these reactions are required for normal physiologic processes, iron can also catalyze the generation of hydroxyl radicals, which can damage lipids, protein, and DNA (217). Due to the duality of iron redox chemistry, iron transport and homeostasis are tightly regulated in vivo to prevent the production of reactive oxygen species by free iron. However in genetic disorders such as hemochromatosis, in which excessive amounts of dietary iron are absorbed, or β- thalassemia major, which requires blood transfusions, excess iron overwhelms the normal mechanisms of iron transport. One such consequence is the appearance of plasma non-transferrin-bound iron (NTBI), a form of iron that appears when the carrying capacity of transferrin, the circulating iron transport protein, becomes exceeded. The exact chemical nature of NTBI in the plasma is not known, but is thought to consist mainly of ferric citrate and other low-molecular-weight iron species (218, 219). Although it is generally believed that NTBI is a pathologic species that appears only when transferrin saturation exceeds 75% (220), plasma NTBI has been reported to be commonly present in diabetics with transferrin saturations below 60% (131). Studies in mice have shown that plasma NTBI is rapidly cleared mostly by the liver, and to a lesser extent, the pancreas, kidney, and heart (73, 74, 83). Accordingly, NTBI is a major contributor to iron loading of the liver and other tissues in iron overload disorders. In the liver and pancreas, NTBI is taken up mainly by hepatocytes and acinar cells via 86

87 ZRT/IRT-Like Protein 14, ZIP14 (SLC39A14) (7). How NTBI is taken up by the kidney, heart, and other organs/cell types remains to be established. Studies of iron-loaded human pancreas have revealed that iron not only accumulates in acinar cells, but also in β cells of the islets ( ). Iron loading of the β cell has been proposed to contribute to the well-known β cell dysfunction and diabetes in individuals with clinical iron overload (85, 133). Given the known role of NTBI uptake to iron loading of various organs and cells, we hypothesize that human β cells are able to take up NTBI. The aim of the present study was to examine the potential roles of the transmembrane transporters DMT1 (divalent metal-ion transporter 1), ZIP14, and ZIP8 in NTBI uptake by human β cells. We focused on these three transporters because of their well-documented roles in NTBI uptake/iron metabolism (7, 11, 13, 28, 90), and in the case of DMT1 and ZIP8, also because DMT1 has been reported to be expressed in human islets and ZIP8 has been reported in rat β cells (85, 103). Results Overexpression of NTBI Transporters in Human β Cells To determine whether the expression of established NTBI transporters could promote iron uptake in β cells, ZIP14, ZIP8, and DMT1 were overexpressed in βlox5 cells, a human β cell line (221), and NTBI uptake was measured. NTBI uptake was assessed at ph 7.4, the ph of blood plasma. I found that overexpression of ZIP14 or ZIP8, but not DMT1, increased the ability of βlox5 cells to take up NTBI when compared with cells transfected with empty vector control (Figure 5-1). To explore the possibility that the lack of DMT1-mediated NTBI transport in βlox5 cells results from poor DMT1 expression at the cell surface, I isolated cell-surface proteins from cells overexpressing DMT1. Western blotting analysis of total-cell lysate and isolated cell-surface proteins 87

88 revealed that the majority of DMT1 was indeed intracellular with little expression at the cell surface, thus potentially accounting for the lack of additional NTBI uptake during DMT1 overexpression (Figure 5-2A). In contrast to DMT1, overexpressed ZIP14 was enriched at the cell surface (Figure 5-2B). The proteins copper chaperone for superoxide dismutase (CCS) and Na + /K + ATPase were measured to indicate intracellular and cell-surface protein fractions, respectively. sirna Knockdown of NTBI Transporters in Human β Cells To define the contribution of endogenous NTBI transporter expression to iron uptake by human β cells, sirna was used to suppress the expression of ZIP14, ZIP8, and DMT1 in βlox5 cells. sirna-mediated suppression of ZIP14 expression decreased cellular iron uptake by approximately 50% (Figure 5-3A). By contrast, sirna knockdown of ZIP8 did not affect iron uptake (Figure 5-3B), suggesting that endogenous ZIP8- mediated NTBI uptake is negligible in βlox5 cells. I was unable to achieve successful knockdown of DMT1 in this cell line because the cells died shortly after transfection. Interestingly, cell death could be prevented by supplementing the cell culture medium with 50 µm ferric ammonium citrate (FAC), suggesting that decreased cellular viability was related to cellular iron deficiency (data not shown). Similar to βlox5 cells (Figure 5-3A), knockdown of ZIP14 in primary human islets decreased NTBI uptake by approximately 50% (Figure 5-4), suggesting that ZIP14 is a major route of NTBI uptake in human β cells. Analysis of mrna copy numbers in human primary islets (Figure 5-7) indicates that the number of mrna transcripts encoding ZIP14 is approximately 2 and 4 times the number of ZIP8 and DMT1 transcripts, respectively (Figure 5-7). 88

89 Cellular Localization of NTBI Transporters in Human Islets As pancreatic islets represent a non-homogenous population of cells, consisting primarily of β and α cells, my methods using whole islets are unable to discern the contribution of individual cell types to mrna expression and iron uptake. Therefore, immunofluorescence analysis was used to determine if ZIP14, ZIP8, and DMT1 are expressed at the protein level in β cells or in other cells comprising pancreatic islets. In the case of ZIP14 I found that protein expression is largely restricted to β cells with negligible expression in α cells (Figure 5-5). ZIP14 staining in β cells displayed a diffuse speckled pattern throughout the cytosol (Figure 5-5B). Staining for DMT1 in the human pancreas indicated that its expression was restricted to β cells with no signal detected from α cells (Figure 5-8A). DMT1 displayed a punctate, granular staining pattern suggesting an intracellular localization, consistent with the known role of DMT1 in endosomal iron transport in some cell types (26). Staining for ZIP8 in the human pancreas revealed only low-level diffuse staining in pancreatic acinar cells. No islet staining was observed beyond non-specific levels detected with non-immuned IgG substituted for anti-zip8 primary antibody (Figure 5-8B). Modulation of ZIP14 Expression by Iron in Human β Cells Previous reports have indicated that ZIP14 protein levels are modulated by cellular iron status. For example, in human hepatoma HepG2 cells, ZIP14 protein levels are induced by iron loading with ferric ammonium citrate (FAC) (82, 95). ZIP14 protein levels are also elevated in iron-loaded rat liver and pancreas (82). To determine if ZIP14 levels are induced by iron loading in human β cells, I treated βlox5 cells and primary human islets with FAC and measured ZIP14 protein expression. I found that cellular iron 89

90 loading, confirmed by elevated ferritin protein levels, increased ZIP14 protein expression in βlox5 cells (Figure 5-6A) but not primary islets (Figure 5-6B). Depletion of cellular iron levels by the iron chelator desferrioxamine (DFO) has been documented to decrease ZIP14 protein levels in HepG2 cells (95). However in primary human islets treated with DFO, I detected no difference in ZIP14 protein levels after DFO-induced iron depletion, as confirmed by elevated TFR1 protein levels (Figure 5-6B). I was unable to test the effect of iron deficiency on ZIP14 expression in βlox5 cells as DFO treatment did not successfully alter TFR1 levels in this cell line (data not shown). Modulation of ZIP14 Expression By IL-1β in Human β Cells IL-1β levels are elevated in primary islets isolated from type 2 diabetics. Additionally, ZIP14 mrna levels have been observed to increase in response to IL-1β in isolated mouse hepatocytes (222). To determine if IL-1β induces ZIP14 expression in human β cells, βlox5 cells were treated with IL-1β for either 8 or 24 h. Both of these treatment times increased ZIP14 protein levels to a similar degree (Figure 5-6C). However, treatment of human islets with IL-1β (for 24 h) resulted in no induction of ZIP14 protein (Figure 5-6D). Discussion Disorders of iron overload in humans are associated with β cell iron accumulation ( ), which is currently thought to impair β cell function (45). While β cell iron loading is has been documented during these disorders, little is known regarding the mechanisms by which β cells take up iron. In the present study I examined the contribution of the established NTBI transport proteins DMT1, ZIP14, and ZIP8 to β cell NTBI uptake. The observation that suppression of ZIP14 expression decreased NTBI uptake by approximately 50% in the human pancreatic β cell line βlox5 suggests that 90

91 ZIP14 is a major route of NTBI uptake by human β cells. A similar reduction in NTBI uptake was observed after suppression of ZIP14 expression in isolated primary human islets, which I found express ZIP14 in β but not cells. Iron loading in human islets is reported to be restricted to β cells, in line with the pattern of ZIP14 expression in human islets, suggesting that the lack of iron accumulation in α cells may be due to a lack of ZIP14 expression (132, 134). Although ZIP14 in the human pancreas is detected in β cells, more robust ZIP14 staining was observed in surrounding acinar cells, similar to our previous studies of ZIP14 expression in rat pancreas (82). Indeed, the more robust expression of ZIP14 in acinar cells likely explains why iron loads in the exocrine pancreas during iron overload (7). However, in contrast to the pattern of ZIP14 expression in human pancreas, ZIP14 in rat pancreas was not detectable in β cells (82). Based on these observations, I speculate that the lack of β cell ZIP14 in rodents accounts for the fact that rodent β cells do not load iron, even in the context of massive iron overload (60, 136, 138, 223, 224). I am aware of only 2 studies that have demonstrated iron loading by Perls staining in rodent β cells, both of which have utilized non-physiologic models of iron loading (e.g, portacaval shunting and iron dextran injections) (137, 144). Recently it has been reported that plasma NTBI levels are elevated in type 2 diabetics, even in the absence of systemic iron overload in which plasma iron levels exceed the binding capacity of transferrin (131). Due to the ability of cellular iron loading to increase ZIP14 expression in other cell populations (82, 95), plasma NTBI could initially be taken up by β cells, leading to an upregulation of ZIP14 and therefore an increased capacity for subsequent β cell NTBI uptake. Excess β cell iron is proposed to 91

92 decrease insulin secretory capacity (45). Thus, the mechanism by which iron uptake and accumulation increases subsequent iron loading may be relevant in the context of diabetic pathology. While I found that iron loading increased ZIP14 levels in βlox5 cells this trend was not observed in primary human islets arguing against a cyclic mechanism of iron uptake and ZIP14 upregulation. The upregulation of ZIP14 observed in βlox5 cells but not in primary human islets may be due to the iron status of these cell populations under normal culture conditions. Isolated islets are reported to be quiescent in vitro (225) whereas βlox5 cells rapidly proliferate resulting in a basal state of iron deficiency, evidenced by a lack of induction in TFR1 levels after treatment with DFO (data not shown). Therefore, it is possible that the upregulation of ZIP14 in βlox5 cells after treatment with iron is not due to cellular iron loading but rather the restoration of adequate iron status, a change which does not occur in cultured primary islets as iron status may be adequate to support cellular function, even after temporary iron chelation with DFO, due to a lack of proliferation. β cells from individuals with type 2 diabetes display increased levels of the cytokine IL-1β, attributed to exposure to elevated levels of glucose (226). Given that IL- 1β was previously demonstrated to increase Zip14 expression in isolated mouse hepatocytes (222), I hypothesized that IL-1β may increase β cell ZIP14 levels which, in diabetics with plasma NTBI, could increase β cell NTBI uptake. While ZIP14 levels increased in βlox5 cells following IL-1β treatment no effect was observed in primary human islets suggesting that IL-1β is unlikely to upregulate ZIP14 in islets of patients with diabetes. Differences in gene expression between βlox5 cells and primary human β 92

93 cells have previously been reported (221), potentially accounting for the differential effect of IL-1β treatment observed in primary islets and βlox5 cells. The expression of DMT1 in human islets has been reported previously and it has been hypothesized that DMT1 may be responsible for β cell iron loading (85). While I observed that DMT1 is expressed in β cells, and is also absent from α cells reflecting the pattern of islet iron loading in humans, other results from the present study argue against a role of DMT1 in the process of NTBI uptake by β cells. For example, I found that overexpression of DMT1 fails to increase NTBI uptake in βlox5 cells, likely due to the intracellular localization of DMT1 precluding iron uptake at the cell surface. Immunofluorescense analysis of DMT1 in human islets also suggests that DMT1 is localized intracellularly, due to the punctate, granular staining pattern observed. In addition to the intracellular localization of DMT1, the functional properties of DMT1, specifically the coupling of efficient iron transport to a proton gradient, argue against DMT1 contributing substantially to the uptake of plasma NTBI by β cells. DMT1 functions optimally at ph 5.5 (11), in line with the established function of DMT1 in intestinal (13) and endosomal NTBI transport (26), and transports iron relatively poorly at the physiologic ph of 7.4 for plasma. ZIP8 is reported to be expressed at the plasma membrane of rat β cells (103) and ZIP8 mrna is abundantly expressed in the human pancreas, relative to other tissues (100), suggesting that ZIP8 may contribute to β cell iron uptake. In the present study I found that the overexpression of ZIP8 in βlox5 cells increased NTBI uptake but that suppression of endogenous ZIP8 expression had no effect on NTBI uptake, suggesting that ZIP8 levels are not abundant in human β cells. Additionally, I detected 93

94 modest amounts of ZIP8 protein in acinar cells but not in β cells strengthening the finding that ZIP8 protein expression is negligible, and therefore is unlikely to contribute to NTBI uptake in human β cells. In conclusion I have identified ZIP14 as a major contributor to NTBI uptake by human β cells. The identification of ZIP14 as a route of β cell NTBI uptake provides a target for inhibitors that could be used to prevent β cell iron accumulation during iron overload. Future study of the role ZIP14 plays in in-vivo NTBI uptake by β cells, and β cell function, will need to be carried out using rodent models which successfully reflect the human phenotype characterized by β cell ZIP14 expression and β cell iron accumulation. 94

95 Figure 5-1. ZIP14 and ZIP8, but not DMT1, overexpression increases iron uptake by βlox5 cells. A) Western blot analysis of cell lysates from blox5 cells transfected with pcmv-sport6-empty vector (EV), DMT1, ZIP14, or ZIP8. Tubulin is shown to indicate lane loading. B) Effect of ZIP14, ZIP8, or DMT1 overexpression on the uptake of iron by βlox5 cells. To measure iron uptake, cells were incubated for 1 h in serum-free medium containing 2 μm [ 59 Fe] ferric citrate and 1 mm ascorbate and the cellular uptake of 59 Fe was measured by gamma counting. Data represent the mean ± S.E. of 3 independent experiments performed in triplicate. Group means were compared by unpaired Student s t-test. Asterisks indicate differences relative to cells transfected with EV (*P < 0.05). 95

96 Figure 5-2. When overexpressed in βlox5 cells, ZIP14 localizes to the plasma membrane whereas DMT1 mainly localizes intracellularly. Western blot analysis of ZIP14, DMT1, Na + /K + -ATPase, and copper chaperone for superoxide dismutase (CCS) in total-cell lysate (TCL) or cell-surface (CS) proteins isolated from βlox5 cells transfected with either empty vector (EV), A) ZIP14, or B) DMT1. Plasma membrane proteins were labeled with Sulfo- NHS-SS-Biotin and affinity purified by using streptavidin-agarose columns prior to western blotting. Na + /K + -ATPase and CCS serve as markers for plasma membrane and cytosolic proteins, respectively. 96

97 Figure 5-3. Endogenous iron uptake by βlox5 cells is decreased by sirna knockdown of ZIP14, but not ZIP8. A) Western blot analysis of lysates from βlox5 cells transfected with negative control sirna (sinc) or sirna targeting either ZIP14 (sizip14, left panel) or ZIP8 (sizip8, right panel). B) To measure NTBI uptake, cells were incubated for 2 h in serum-free medium containing 2 μm [ 59 Fe] ferric citrate and 1 mm ascorbate and the cellular uptake of 59 Fe was measured by gamma counting. Data represent the mean ± S.E. of 3 independent experiments performed in triplicate. Group means were compared by unpaired Student s t-test. Asterisks indicate differences relative to cells transfected with sinc (*P < 0.05). 97

98 Figure 5-4. sirna knockdown of ZIP14 decreases NTBI uptake by primary human islets. A) Western blot analysis of cell lysates from isolated human islets transfected with either negative control sirna (sinc) or sirna targeting ZIP14 (sizip14). B) To measure iron uptake, cells were incubated for 2 h in serum-free medium containing 2 μm [ 59 Fe] ferric citrate and 1 mm ascorbate and the cellular uptake of 59 Fe was measured by gamma counting. Data represent the mean ± S.E. of 3 independent experiments performed in triplicate. Group means were compared by unpaired Student s t-test. Asterisks indicate differences relative to cells transfected with sinc (**P < 0.01). 98

99 Figure 5-5. ZIP14 is detected in human pancreatic β cells by immunofluorescent analysis. Immunofluorescent images taken at either A) 20x or B) 60x of human pancreatic tail sections co-stained for ZIP14 (green), insulin (β cell marker, red), and glucagon (α cell marker, blue). Panels show the same tissue region as stained for ZIP14 only (I), ZIP14 with insulin (II), or ZIP14 with insulin and glucagon (III). Serial sections co-stained in parallel, but with non-immune IgG replacing the ZIP14 primary antibody are shown to indicate background staining (IV). 99

100 Figure 5-6. Cellular iron levels and treatment with IL-1β increase ZIP14 levels in βlox5 cells but not primary human islets. A) Western blot analysis of βlox5 cell lysates for ZIP14 and ferritin after 24 h incubation in control (CON) medium or medium supplemented with 100 μm ferric ammonium citrate +1 mm ascorbate (FAC). B) Western blot analysis of ZIP14, TFR1, and ferritin in human-islet lysates 48 h after treatment with CON medium or medium containing 50 µm deferoxamine (DFO) or 100 µm ferric ammonium citrate + 1mM ascorbate (FAC). Lysates from βlox5 cells transfected with either sinc or sizip14 sirna are shown to confirm the band size of ZIP14 protein. C) Western blot analysis of ZIP14 in βlox5 lysates after incubation in CON medium or medium supplemented with 100 U/ml recombinant human IL-1β for either 8 or 24 h. D) Western blot analysis for ZIP14 in human-islet lysates after incubation in CON medium or medium containing 100 U/ml recombinant human IL-1β for 24 h. Lysates from βlox5 cells transfected with either sinc or sizip14 sirna are shown to confirm the band size. Tubulin is shown to indicate lane loading. 100

101 Figure 5-7. mrna copy numbers of NTBI transporters in primary human islets. qrt- PCR measurement of DMT1, ZIP14, and ZIP8 mrna copy numbers in total RNA isolated from nondiabetic human islets. Copy numbers were calculated based on standard curves constructed from known concentrations of plasmid DNA encoding either DMT1, ZIP14, or ZIP8. Data represent the mean mrna copy numbers ± S.E. obtained from 4 independent donors. 101

102 Figure 5-8. DMT1, but not ZIP8, is detected in human β cells by immunoflourescence staining. A) Immunoflourescence images (60x) of human pancreatic tail sections co-stained for DMT1 (green), insulin (red), and glucagon (blue). Panels show the same tissue region as DMT1 signal only (I),DMT1 with insulin signal (II), or DMT1 with insulin and glucagon signal (III). Serial sections co-stained in parallel, but with non-immune IgG replacing the DMT1 primary antibody are shown to indicate background DMT1 signal (IV). B) Immunoflourescence images (20x) of human pancreatic tail sections costained for ZIP8 (green), insulin (red), and glucagon (blue). Panels show the same tissue region as ZIP8 signal only (I), ZIP8 with insulin signal (II), or ZIP8 with insulin and glucagon signal (III). Serial sections co-stained in parallel, but with non-immune IgG replacing the ZIP8 primary antibody are shown to indicate background ZIP8 signal (IV). 102

103 CHAPTER 6 THE INFLUENCE OF IRON STATUS ON DIABETIC PATHOLOGY AND β-cell FUNCTION The link between iron status and diabetes has been extensively documented in patients with iron overload disorders, in which the prevalence of diabetes is elevated compared with the general population (45, 113, 114, 118). Patients with the iron overload disorder hemochromatosis are reported to display diminished insulin secretion in response to glucose (45), suggesting that excess iron accumulation impairs the ability of β cells to regulate glucose homeostasis. Diabetes has historically been categorized as either type 2 diabetes, resulting from systemic insulin resistance, or type 1 diabetes, resulting from a loss of insulin secretory capacity. Type 1 diabetes most often results from the autoimmune-mediated destruction of pancreatic β cells. While the link between systemic iron status and diabetes has been extensively documented, little is currently known about the influence of iron in the pathogenesis of autoimmune diabetes. The evidence linking iron status to autoimmune diabetes in humans is currently limited to retrospective epidemiology demonstrating a potential link between iron and diabetes risk. Increased iron consumption during infancy is reported to be associated with a greater risk for developing diabetes during childhood (166) and elevated transferrin saturation, an indicator of iron status, is associated with an increased prevalence of type 1 diabetes in adults (167). However, direct mechanistic evidence for a causative role of iron in autoimmune diabetes is currently lacking. Studies carried out in animal models have indirectly suggested that iron depletion may be protective against autoimmune diabetes. Treatment with the iron chelator desferroxiamine protects islet grafts from autoimmune destruction in NOD mice (170), a mouse model of autoimmune diabetes. Also, the administration of apotransferrin to NOD mice reduces the incidence 103

104 of spontaneous diabetes (173), an effect hypothesized to be attributed to the binding of plasma NTBI, which may be elevated in diabetics (131). T-cell proliferation and cytotoxicity are also reduced by treatment with anti-transferrin receptor antibodies (171), although the ability of T cells to acquire iron in response to antibody treatment was not determined. Currently no controlled trials have determined the influence of systemic iron status on the development of autoimmune diabetes. Evidence for β cell iron status directly affecting β cell function is also lacking as studies evaluating glucose homeostasis during iron deficiency or overload have done so in the context of systemic changes in iron status (45, 146, 147, 163). Changes in wholebody iron levels may affect β cell function through the modulation of known or unknown systemic factors, such as erythropoietin, which promotes β cell proliferation and prevents apoptosis (227). Current in vitro studies investigating the direct impact of iron status on islets in isolation have been limited to rat islets (165) and have not determined the effect of iron status on islet function, as evidenced by insulin secretion, directly. To date no studies have determined the effect of iron depletion or loading on the insulin secretory capacity of human islets. Additionally, the study of β cell iron loading in vivo has been complicated by the lack of a mouse model which accumulates substantial iron within β cells, similar to the pattern observed in humans with iron overload (132, 133). β cells in mouse models of iron overload demonstrate a remarkable resistance to iron loading (60, 136, 138) making it difficult to determine the effect of progressive β cell iron accumulation on β cell function. In the current study I aimed to examine the role of iron in diabetic pathology by determining the influence of systemic iron status on the development of autoimmune 104

105 diabetes in NOD mice and the effect of iron status on glucose-stimulated insulin secretion (GSIS) by isolated human islets. I hypothesized that iron deficiency would be protective against the development of diabetes and increase GSIS by human islets while iron loading would result in an increased incidence of autoimmune diabetes in NOD mice and impair GSIS by human islets. Additionally I produced a transgenic mouse expressing the mammalian iron transporter Zip14 under control of the mouse insulin 1 promoter with the aim of generating a mouse model of β cell iron loading. Results Effect of Iron Status on Spontaneous Autoimmune Diabetes in NOD Mice To determine whether differences in systemic iron status could affect the development of autoimmune diabetes female NOD mice were fed either iron-deficient (FeD), iron-adequate (FeA), or iron-loaded (FeO) diets from weaning until 30 wk of age. FeD mice demonstrated a trend towards a greater incidence of diabetes relative to FeA mice, 80% diabetic at 30 wk of age compared with 60%, although this difference did not reach statistical significance (P=0.06) (Figure 6-1). The initial onset of diabetes was also earlier in FeD mice, first detected at 11 wk of age, compared with FeA mice in which diabetes was first detected at 14 wk of age. No differences were observed between the incidences of diabetes in FeO compared with FeA mice and the development of diabetes was initially detected at similar ages in these groups. Effect of Dietary Iron on Rate of Growth and Systemic Iron Status Iron deficient and loaded diets have previously been demonstrated to affect growth in rodents (176, 228, 229) and higher body weights are associated with increased susceptibility to autoimmune diabetes in BioBreeding rats (230). However, no differences were observed in the rate of growth between FeD and FeA mice at any point 105

106 in the study. Mice in the FeO group gained weight at a reduced rate initially but the difference was corrected during the prediabetic period, before 11 wk of age, and body weights between groups did not differ after this point (Figure 6-2). To confirm that systemic iron status was successfully altered by dietary treatments prior to the development of diabetes I measured indices of iron status in 10-wk-old mice that had consumed FeD, FeA, or FeO diets since weaning (Table 6-1). FeD mice had lower 12% lower hemoglobin levels compared with FeA mice but plasma iron levels were not diminished in response to dietary iron deficiency. FeO mice had elevated transferrin saturation, attributed primarily to diminished total iron binding capacity (TIBC), whereas no difference was detected between FeD and FeA mice, in line with the similar plasma iron concentrations between these groups. Iron stores, indicated by liver non-heme iron concentrations, were depleted in FeD mice and 10 times greater in FeO animals compared with FeA mice, indicating that iron status was successfully altered by dietary intervention during the period preceding the development of diabetes. Plasma iron was found to be greater in 30 wk-old-feo mice compared with other groups, although transferrin saturation was similar between age groups. Iron status in FeD mice normalized with age, as evidenced by the recovery of hemoglobin values in 30-wk-old FeD mice (Table 6-1). Additionally liver iron stores increased in 30-wk-old FeD mice relative to those measured in FeD mice at 10 wk of age. However, iron stores were found to still be significantly greater in FeA mice compared with FeD at 30 wk of age. Unlike at 10 wk of age, plasma iron was significantly greater in 30-wk-old FeO mice compared with the other groups. Measurement of liver non-heme iron levels from 106

107 FeD mice at various ages between 10 and 30 wk of age indicate that iron stores gradually increase in a linear fashion after 10 wk of age (Figure 6-3). Pancreatic Mineral Concentrations Our lab has demonstrated that alterations in dietary iron can alter pancreatic mineral concentrations, which may contribute to pancreatic dysfunction (139). To explore this possibility I measured trace minerals in the pancreases from prediabetic FeD, FeA, and FeO mice by using ICP-MS analysis. In 10-wk-old mice pancreatic iron levels in FeD and FeO mice were 52% and 172% of FeA levels, respectively (Table 6-2). Modest differences were also detected in pancreatic zinc and copper, which were elevated in FeD mice compared with the FeA and FeO animals, and in selenium which was lower in FeO mice. Pancreatic Iron levels in 30-wk-old mice were not different between FeD and FeA animals but were 3.5 times greater in FeO mice. Differences in pancreatic zinc levels were not detected between groups at 30 wk of age but FeD mice had slightly elevated copper levels. Testing of β cell function During the Prediabetic Period To investigate whether systemic iron status has an effect on β cell function in NOD mice I performed glucose tolerance testing using 10 wk prediabetic mice. No differences in glucose tolerance or fasting glucose levels between groups were detected (Figure 6-4A). However, all groups reported poor glucose tolerance, maintaining blood glucose levels >300 even 2 h post injection. GSIS capacity was also measured in conjunction with glucose tolerance and no significant differences in plasma insulin levels were detected between groups at any time point (Figure 6-4B). As β cell iron status has been hypothesized to affect β cell function, islets from mice used in glucose tolerance testing were isolated and iron status was determined by 107

108 measuring transferrin receptor 1 (TFR1) expression, which is well documented to inversely reflect cellular iron levels (78). TFR1 levels in FeO mice were lower compared with FeD and FeA mice, indicating increased levels of islet iron (Figure 6-5A). However, no difference was observed between FeD and FeA mice indicating that islet iron status was not different between FeD and FeA mice at 10 wk. Also, histological analysis of pancreas sections from 10 wk mice revealed no difference in the degree of insulitis between groups, with all groups reporting only mild insulitis (Figure 6-5B). Effect of Iron Status on Human Islet GSIS The feeding of NOD mice with an iron-deficient did not result in diminished β cell iron status in vivo. To determine the effect of iron status on β cell function I altered the iron status of isolated primary human islets in vitro prior to GSIS testing to indicate islet function. Islets were treated for 48 h with either control medium (CON), 50 μm deferoxamine mesylate (DFO), an iron chelator to deplete islets of iron, or 100 μm ferric ammonium citrate and 1 mm ascorbate (FAC), to load islets with iron. The successful alteration of islet iron status was confirmed by western blot analysis for both TFR1 and ferritin, an indicator of cellular iron stores. TFR1 expression was elevated in islets treated with DFO relative to CON islets, indicating a reduction in cellular iron levels, and decreased in islets treated with FAC, indicating cellular iron loading (Figure 6-6A). However, no differences in insulin secretion between groups, during either basal or glucose stimulated conditions, were measured during GSIS testing indicating that altered iron status does not affect GSIS (Figure 6-6B). Generation of Mice Selectively Overexpressing Zip14 in β Cells Previous studies using mouse models of severe iron overload have demonstrated that mouse islets are resistant to substantial iron loading in vivo (60, 136, 108

109 138). Recent findings by our lab have indicated that ZIP14 is required for iron loading of the exocrine pancreas (7) and that ZIP14 contributes to β cell iron uptake by human islets (unpublished results). In light of these findings I aimed to generate transgenic mice overexpressing ZIP14 in pancreatic β cells to create a novel mouse model predisposed to β cell iron loading. I generated a vector construct containing mzip14 tagged with human influenza hemagglutinin antigen (HA) under control of the mouse insulin 1 promoter (MIP) and containing a downstream intronic region of human growth hormone (Figure 6-7A). Four founder animals were obtained and bred to establish 4 individual transgenic mouse lines. Analysis of the number of transgene copies per genome by using qrt-pcr revealed that 2 founders possessed approximately 5 copies, referred to as low-copy lines, while other founders had approximately either 13 or 500 copies, referred to as moderate- or high-copy lines. Comparison between transgene copy numbers in founder animals and subsequent generations indicate that the transgenes are completely inherited in the lines carrying low- and moderate-copy lines. However, incomplete inheritance of the transgene was detected within the highcopy line with some offspring inheriting a low number of copies. Successful expression of the transgene in β cells was confirmed by immunostaining for HA in pancreatic sections (Figure 6-7B). Animals from 3 out of 4 transgenic lines demonstrated transgene expression to various degrees in line with the number of transgene copies detected. Discussion Excess iron accumulation is associated with an increased prevalence of diabetes and is believed to influence aspects of diabetic pathology (45, 113, 114, 118). Previous experiments have suggested that systemic iron levels may be a risk factor for the 109

110 development of autoimmune diabetes (167, 169) and that β cell iron accumulation results in diminished insulin secretory capacity (45, 146). However, direct evidence for these claims is currently lacking. The present study using NOD mice produced the unexpected results that iron overload did not increase the incidence of diabetes and that dietary iron restriction was not protective, potentially even promoting to the development of autoimmune diabetes. Diabetes was detected earlier in FeD mice and the overall incidence of diabetes trended strongly towards being increased by dietary iron restriction, suggesting that iron deficiency may increase susceptibility to the development of autoimmune diabetes. Furthermore, dietary iron overload failed to increase the incidence of diabetes compared with mice fed FeA diets, arguing against the hypothesis that elevated iron stores increase the risk of developing type 1 diabetes. While iron deficiency may negatively impact the development of diabetes in NOD mice, experiments in prediabetic mice failed to explain the trend toward an increased incidence of diabetes in in iron-restricted mice. Glucose tolerance testing in prediabetic NOD mice did not reveal any differences between groups regarding either glucose tolerance or GSIS, indicating that systemic iron status did not affect glucose homeostasis or β cell function during the prediabetic period. All groups investigated reported poor glucose tolerance, potentially attributed to the high sucrose diet used in this study (231). Other studies have reported that feeding an iron-deficient or loaded diet can affect insulin sensitivity in rodents ( ). In this study I did not detect any differences in insulin sensitivity as both glucose tolerance and insulin secretion were similar for all groups measured. Discrepancies between the current study and previous studies regarding the influence of iron loading or depletion on insulin sensitivity in 110

111 rodents may result from differences in either dietary iron content or feeding duration. In general the iron content of the diets and the duration of feeding used in the present study were less extreme compared with diets and timelines used in previous studies reporting differences in insulin sensitivity in response to dietary iron deficiency or overload ( ). The lack of difference in GSIS between groups during glucose tolerance testing indicates that β cell function was not affected by altering systemic iron stores during the prediabetic period. While iron stores were lower in 10-wk-old FeD mice relative to FeA mice, as evidenced by diminished liver non-heme iron and mild anemia in FeD animals, islet TFR1 levels were not different between these groups suggesting that the trend toward increased diabetic development was not attributable to β cell iron deficiency. Serum iron was also similar between FeD and FeA mice suggesting that circulating iron levels were adequate to supply islets with iron, thus preventing islets from becoming iron deficient. Additionally, dietary iron overload increased islet iron status but did not result in altered insulin secretory capacity or glucose tolerance, suggesting that islet iron loading, at least to the degree observed in this study, does not have an effect on β cell function. The finding that islet iron status does not affect GSIS is supported by the testing of iron-loaded and iron-depleted human islets in vitro, also carried out in this study. I determined that there was no effect of iron status on GSIS, during either basal or high glucose conditions, arguing against the hypothesis that β cell iron status impacts insulin secretory capacity. To my knowledge this is the first report of the effect of iron depletion or iron loading on human β cell function in an isolated cell-culture system, eliminating 111

112 the potential influence of systemic factors, on β cell function. The lack of effect on GSIS by both iron loading and iron depletion was unexpected based on previous reports demonstrating a role for ROS in the process of glucose stimulated insulin secretion (235) and the hypothesized role of cellular iron in the generation of intracellular ROS (86). It is possible that in the current study the degree of iron loading of human islets did not exceed the ability of the β cell to sequester iron within ferritin, preventing the buildup of free intracellular iron which would be capable of catalyzing ROS formation. While iron status was clearly different between control, iron-depleted, and iron-loaded islets in the current study, as indicated by TFR1 and ferritin levels, future investigation of the influence β cell iron status has on insulin secretory capacity may benefit from longer term or higher-dose iron loading. Additionally it is possible that the alteration of β cell iron status affects first-phase insulin secretion, an early indicator of impaired β cell function (236, 237), which would not be detected by the methods used in this study. Future studies may benefit from the use of more nuanced methods of GSIS testing, such as islet perifusion, which is capable of discerning differences in GSIS at individual time points. The study of long-term, progressive β cell iron accumulation would benefit from the availability of a mouse model that accumulates iron in β cells, similar to what has been observed in humans with iron overload disorders (132, 133). In this study I also detailed the generation of transgenic mice that overexpress mzip14 in pancreatic β cells under the control of the mouse insulin 1 promoter. Previous reports by our lab have indicated that ZIP14 is required for iron loading of pancreatic acinar cells (7) and that ZIP14 contributes to non-transferrin bound iron uptake by human islets (publication 112

113 under review), suggesting that overexpressing ZIP14 in mouse β cells may result in increased iron uptake and accumulation. In this study I reported the successful overexpression of mzip14 in β cells, localized to the plasma membrane and intracellular locations within β cells of transgenic mice. Future testing is needed to confirm that this novel model demonstrates β cell iron loading during systemic iron overload. A limitation of the current study is that iron deficiency resolved in the FeD mice with age, potentially attributable to the plateau of growth after mice reached approximately 10 wk of age. Due to the recovery of iron status it is possible that early trends observed regarding the incidence of diabetes in FeD mice were somewhat ablated in older mice and could potentially be more dramatic under conditions of more severe iron deficiency. In the current study the iron-deficient diet contained 14 ppm iron, a greater concentration than traditional iron-deficient diets (163, 238), due to the inclusion of 5% wheat in the diet. Wheat was added to the diets as additional wheat has been reported to increase the diabetic potential of purified diets (239), which usually result in a low incidence of diabetes in NOD mice (240, 241). Future studies investigating the role of iron deficiency in autoimmune diabetes may benefit from the use of BioBreeding rats, which also develop spontaneous autoimmune diabetes and in which iron deficiency may be easier to induce due to the greater increase in rat body weight relative to that of mice. In conclusion I report that iron deficiency, but not iron overload, may increase the development of autoimmune diabetes in NOD mice. Additionally β cell iron status did not affect β cell function, calling into question the long-hypothesized mechanism thought to account for the link between iron loading and diabetes. Future studies will be required 113

114 to determine the influence of both systemic and β cell iron content in the pathogenesis of diabetes. 114

115 Table 6-1. Iron parameters of type 1 diabetes-prone NOD mice Hemoglobin Liver non-heme iron Plasma Iron Age Group (g/dl) (μg Fe/g) (μg/dl) TIBC (μg/dl) TF Sat (%) 10 wk FeD 12.7 ± 1.3 a 16.7 ± 3.8 a ± 61.7 a ± 43.7 a 37.4 ± 15.5 a 10 wk FeA 14.5 ± 1.1 b ± 41.8 b ± 46.7 a ± 22.8 a 49.6 ± 12.1 ab 10 wk FeO 14.5 ± 0.6 b ± c ± 51.2 a ± 38.9 b 74.4 ± 18.3 b 30 wk FeD 14.1 ± 0.6 a ± 27.5 a# ± 16.7 a ± 22.5 a 33.9 ± 2.4 a 30 wk FeA 14.1 ± 0.9 a ± 23.3 b# ± 32.9 a ± 24.1 a 37.1 ± 8.4 a 30 wk FeO 15.1 ± 0.4 a ± c ± 97.0 b ± 74.2 a 71.6 ± 13.1 b Liver non-heme iron levels are reported as μg Fe/g wet tissue. Liver non-heme iron, plasma iron, and total iron-binding capacity (TIBC) were determined colorimetrically. Transferrin saturation (TF Sat) was calculated as plasma iron as a percentage of TIBC. Values are presented as means ± SD, n=3-8. Statistical significance was determined by one-way ANOVA. Means without a common superscript are significantly different compared with other groups at the same age. Values at 30 wk of age that are significantly different from those of the same group at 10 wk of age are indicated by # (P<0.05). 115

116 Table 6-2. Pancreatic mineral concentrations in NOD mice Age Group Iron Zinc Manganese Copper Cobalt Selenium Molybdenum 10 wk FeD 77.7 ± 13.4 a ± 57.6 a 7.6 ± 1.1 a 7.6 ± 1.3 a 0.16 ± 0.01 a 1.4 ± 0.20 a 0.5 ± 0.04 a 10 wk FeA ± 26.6 b ± 24.8 b 7.4 ± 1.3 a 6.3 ± 0.5 b 0.16 ± 0.02 a 1.3 ± 0.08 a 0.5 ± 0.08 a 10 wk FeO ± 67.3 c ± 25.0 b 7.9 ± 1.1 a 5.5 ± 0.5 b 0.14 ± 0.01 a 1.0 ± 0.20 b 0.5 ± 0.05 a 30 wk FeD ± 58.0 a# ± 40.6 a# 10.4 ± 0.5 a 6.9 ± 0.4 a 0.18 ± 0.03 a 1.3 ± 0.1 a 0.5 ± 0.03 a 30 wk FeA ± 25.5 a ± 40.6 a 8.5 ± 0.7 a 5.4 ± 0.5 b 0.14 ± 0.01 a 1.3 ± 0.2 a 0.4 ± 0.04 a 30 wk FeO ± b# ± 17.4 a 17.7 ± 11.8 a# 5.1 ± 0.5 b 0.13 ± 0.03 a 1.3 ± 0.3 a 0.5 ± 0.07 a Mineral concentrations (mg/g dry weight) were measured by using ICP-MS. Values represent means ± SD, n = 4-6. Means without a common superscript are significantly different P<0.05. Statistical significance was determined by one-way ANOVA. Means without a common superscript are significantly different compared with other groups at the same age. Values at 30 wk of age that are significantly different from those of the same group at 10 wk of age are indicated by # (P<0.05). 116

117 % Without Diabetes P=0.06 FeD (n=20) FeA (n=20) FeO (n=20) Age (weeks) Figure 6-1. Dietary iron deficiency, but not iron overload, results in a trend towards an increased incidence of spontaneous diabetes in female NOD mice. Cumulative diabetes incidence in female NOD mice fed either iron deficient (FeD), iron adequate (FeA), or iron loaded (FeO) diets starting at 3 wk of age. Spontaneous development of diabetes was monitored by glycosuria starting at 8 wk of age. Survival curves were compared to FeA by using the Gehan- Breslow-Wilcoxon test. 117

118 Body Weight (g) ** *** ** * * FeD (n=20) FeA (n=20) FeO (n=20) Age (Weeks) Figure 6-2. NOD mice fed an iron-loaded diet initially experience diminished growth. Body weights of female NOD mice fed either iron-deficient (FeD), ironadequate (FeA) or Iron-loaded (FeO) diets from 3 to 30 wk of age. Animals were weighed every 3 days until 12 wk of age, at which point body weights were recorded weekly. Body weights at individual time points were compared by one-way ANOVA. Asterisks indicate significant differences in bodyweights of FeO animals compared with FeA animals (*P<0.05, **P<0.01, ***P<0.001) 118

119 Liver Nonheme Iron ( g/g tissue) r= Age (Weeks) Figure 6-3. Iron stores of mice fed an iron-deficient diet increase with age. Liver nonheme iron concentrations of FeD mice were measured colorimetrcally in 10- wk-old prediabetic mice, diabetic mice of various ages, and 30 wk nondiabetic mice fed an iron-deficient diet. The correlation coefficient (r) was calculated by using a linear model. 119

120 Figure 6-4. Glucose tolerance and insulin secretory capacity is not affected by iron status in prediabetic NOD mice. A) Results of intraperitoneal glucose tolerance testing in fasted 10-wk-old prediabetic female NOD mice. Values reported as mean ± SEM, n=5 per group. B) Plasma insulin levels in mice used in glucose tolerance testing. Values reported as mean ± SEM, n=3-5 per group. Blood glucose and plasma insulin levels between groups at individual time points were compared by one-way ANOVA. 120

121 Figure 6-5. Iron-deficient prediabetic NOD mice show no differences in β cell iron status or insulitis compared with iron-adequate mice. A) Western blot analysis of mouse islet total-cell lysate from 10-wk-old prediabetic iron-deficient (FeD), iron-adequate (FeA), and iron-loaded (FeO) NOD mice for transferrin receptor (TFR1). Islets were pooled from 2-3 mice per group. Tubulin is shown to indicate lane loading. B) Average insulitis score from FeD, FeA, and FeO 10- wk-old prediabetic female NOD mice. Values are expressed as the mean ± SEM, n=6. Statistical significance was determined by one-way ANOVA. 121

122 Figure 6-6. Iron status does not affect glucose-stimulated insulin secretion by human islets in vitro. A) Western blotting of human islet total cell lysate for transferrin receptor (TFR1) and ferritin. Islets were treated for 48 h with either control medium (CON), 50 μm deferoxamine mesylate (DFO) or 100 μm ferric ammounium citrate and 1 mm ascorbate (FAC) prior to analysis. Tubulin is shown to indicate lane loading. B) Ability of islets to secrete insulin after DFO or FAC treatment. Total insulin secreted by islets during a 1-h incubation in media containing 2.8 mm D-Glucose followed by a 1-h incubation in media containing 22 mm D-glucose was measured and normalized to islet DNA. Data represent the mean ± SEM of 4 independent experiments performed in triplicate. Treatment group means were compared by one-way ANOVA. 122

123 Figure 6-7. Generation of mice selectively overexpressing Zip14 in β cells. A) Vector map of the construct used to generate MIP-Zip14-HA transgenic mice consisting of a region of the mouse insulin 1 promoter (MIP), HA-tagged ZIP14 (ZIP14-HA), and an intronic region of human growth hormone (hgh Intron). DNA was digested with HindIII and SfiI, to remove the vector backbone, purified, and introduced to fertilized embryos by pronuclear injection. B) Confirmation of transgene expression by Immunofluorescense analysis. Mouse pancreas sections were co-stained for HA (green), insulin (red), and DAPI (blue). Panels show the same tissue section as HA with insulin and DAPI (panel I), HA with DAPI (panel II), insulin with DAPI (panel III). A pancreatic section from a wild-type mouse co-stained in parallel for HA, insulin, and DAPI is shown to indicate non-specific background HA signal (panel IV). 123