University of Southern California Dissertations and Theses

Non-invasive live-cell imaging for monitoring and evaluating pancreatic islet and beta cell metabolism

(USC Thesis Other)

Non-invasive live-cell imaging for monitoring and evaluating pancreatic islet and beta cell metabolism

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Copyright 2020 Zhongying Wang Non-invasive live-cell imaging for monitoring and evaluating pancreatic islet and beta cell metabolism by Zhongying Wang A Dissertation Presented to the FACULTY OF THE USC DORNSIFE COLLEGE OF LETTERS, ARTS AND SCIENCES UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (Molecular Biology) May 2020 i Acknowledgments Foremost I would like to express my gratitude to my advisor Prof. Ray Stevens for his continuous support of my Ph.D. study and research for his patience and enthusiasm. Besides my advisor, I want to thank my thesis committee: Prof. Scott Fraser, Prof. Vadim Cherezov and Prof. Vsevolod Katritch, for their insightful comments on the thesis. I also like to thank everyone in Kate-Ray’s lab and all the collaborators, for their encouragement, help and support all the time. Meanwhile, I would like to acknowledge Elsevier Inc, Becker & Hicki GmbH, ISS and Seahorse for allowing me to use their figures in the thesis to illustrate basic principles of their studies. Finally, I would like to thank everyone, including my family, friends, and pets, who love me and I love. Many things happened over the past five years. I felt frustrated, confused, stressful and lost sometimes. It’s your love and companion that supported me went through the most difficult time. Thank you. ii Table of Contents Acknowledgments............................................................................................................................ i List of Figures ................................................................................................................................. v List of Tables ................................................................................................................................ vii Abstract ........................................................................................................................................ viii Chapter 1: Diabetes, pancreatic islets, and beta cells. ................................................................ 1 1.1 Diabetes Mellitus ................................................................................................................. 1 1.1.1 Type 1 Diabetes ................................................................................................................ 2 1.1.2 Type 2 Diabetes ................................................................................................................ 3 1.2 Pancreatic islets: functional unit of the pancreas ................................................................. 5 1.2.1 Cell Types ........................................................................................................................ 5 1.2.2 Microenvironment of islets .............................................................................................. 6 1.2.3 Beta cell forms a signaling hub ........................................................................................ 7 1.3 Pancreatic beta cells: insulin factory.................................................................................... 8 1.3.1 Glucose stimulated insulin secretion ................................................................................ 8 1.3.2 Glycolysis and oxidative phosphorylation in beta cells ................................................. 10 1.3.3 Metabolism reprogramming in diabetic beta cells ......................................................... 12 1.4 How to understand beta cell metabolism in diabetes development ................................... 13 Chapter 2: Investigating beta cell metabolism by live-cell imaging and biochemical assays .. 15 2.1 Background: Fluorescent lifetime imaging (FLIM) - characterizing metabolism without labeling .......................................................................................................................................... 15 2.1.1 Fluorescent lifetime ........................................................................................................ 15 2.1.2 Principles of FLIM imaging ........................................................................................... 17 2.1.2 Principles of phasor - FLIM analysis ............................................................................. 18 2.1.3 Photobleaching and phototoxicity .................................................................................. 20 2.1.4 Live cell imaging in deep tissues ................................................................................... 20 iii 2.1.2 NADH lifetime in live cells ........................................................................................... 21 2.2 Results ................................................................................................................................ 22 2.2.1 FLIM shows subcellular metabolism in the monolayer cell lines .................................. 22 2.2.2 Glucose stimulation enhances beta cell oxidative phosphorylation ............................... 25 2.2.3 Glucose affects basal respiration and spare respiration capacity of beta cells ............... 26 2.2.4 Exendin-4 can decrease insulin secretion by decreasing oxidative phosphorylation. .... 31 2.2.5 Dispersed rodent alpha and beta cells possess different FLIM character ...................... 33 2.2.6 Alpha and beta cells have the opposite response to glucose stimulation ....................... 35 2.3 Discussion .......................................................................................................................... 38 2.4 Method and Material .......................................................................................................... 41 2.4.1 Cell culture ..................................................................................................................... 41 2.4.2 Rodent islet isolation and dispersion .............................................................................. 42 2.4.3 FLIM imaging ................................................................................................................ 43 2.4.4 FLIM data analysis ......................................................................................................... 44 2.4.5 Seahorse ......................................................................................................................... 44 2.4.6 Glucose stimulation assay .............................................................................................. 44 2.4.7 Immunofluorescence ...................................................................................................... 45 Chapter 3: Glucose-stimulated metabolism in healthy vs. diabetic rodent islets ..................... 46 3.1 Background: metabolic reprogramming in diabetic islets ................................................. 46 3.1.1 Human IAPP vs. mouse IAPP ........................................................................................ 46 3.1.2 Cell stress leads to metabolic reprogramming ............................................................... 47 3.2 Results ................................................................................................................................ 48 3.2.1 Glucose enhanced beta cell oxidative phosphorylation in islets. ................................... 48 3.2.2 Alpha and beta cells also have the opposite response to glucose stimulation in islets .. 49 3.2.3 Human IAPP stress can suppress beta cell OX PHOS ................................................... 50 3.2.4 Beta cells form the “cell hub” to synchronize response of all beta cells ....................... 54 3.2.5 Identification of different cell types and islet structures by FLIM and SHG ................. 56 3.3 Discussion .......................................................................................................................... 57 3.4 Method and Material .......................................................................................................... 61 3.4.1 Rodent islet culture and immobilization ........................................................................ 61 3.4.2 FLIM imaging and SHG imaging .................................................................................. 61 iv 3.4.3 Data analysis .................................................................................................................. 62 Chapter 4: Understanding human islet metabolism in vitro ..................................................... 63 4.1 Background: Comparing human with rodent islets ........................................................... 63 4.1.1 Islet structure .................................................................................................................. 63 4.1.2 Aging and waste accumulation ...................................................................................... 64 4.1.3 Autofluorescence in live tissue ...................................................................................... 65 4.2 Results ................................................................................................................................ 66 4.2.1 Glucose enhances human islet oxidative phosphorylation ............................................. 66 4.2.2 Diabetic islets oxidative phosphorylation are suppressed .............................................. 69 4.2.3 K + channel opener restored insulin release and increased oxidative phosphorylation .. 70 4.2.4 Autofluorescence species in human islets ...................................................................... 71 4.3 Discussion .......................................................................................................................... 76 4.4 Method and Material .......................................................................................................... 80 4.4.1 Human islet culture and immobilization ........................................................................ 80 4.4.2 FLIM data analysis ......................................................................................................... 80 4.4.3 Hyper spectrum imaging ................................................................................................ 80 Chapter 5: Live cell imaging and multi-source data integration for in vivo study ................... 81 5.1 Integrating FLIM data with other data types ..................................................................... 81 5.1.1 Cell hub prediction ......................................................................................................... 86 5.1.2 Metabolic pathway modeling ......................................................................................... 88 5.1.3 Mitochondria network .................................................................................................... 89 5.2 In vivo study of islets is essential for understanding the microenvironment effect in diabetes ......................................................................................................................................... 91 5.3 Islet FLIM signature in high throughput drug screening platform .................................... 95 5.4 Experiment Summary ........................................................................................................ 96 5.5 Conclusions ........................................................................................................................ 98 References ..................................................................................................................................... 99 Appendix I .................................................................................................................................. 110 v List of Figures Figure 1. Blood glucose level of non-diabetic (blue) and diabetic (red) patient after OGTT ....................... 1 Figure 2. Biphasic insulin secretion of beta cell after glucose stimulation ................................................... 8 Figure 3. 11 Major metabolic pathways related to insulin secretion in pancreatic beta cells ....................... 9 Figure 4. Metabolism scheme of non-diabetic (ND) beta cells .................................................................. 11 Figure 5. Metabolism scheme of non-diabetic (ND) beta cells (left) and diabetic beta cells (right) .......... 13 Figure 6. Jablonski diagram illustrating the energy change of fluorophore after exciting. ........................ 16 Figure 7. Diagram of TCSPC microscope principles.................................................................................. 18 Figure 8. The linear combination of multiple species on phasor plot ......................................................... 19 Figure 9. FLIM lifetime map flow chart ..................................................................................................... 20 Figure 10. NAD(P)H molecular structure and its fluorescent structure ...................................................... 21 Figure 11. The NAD(P)H average lifetime increased and shifted to OX PHOS ........................................ 24 Figure 12. Quantitative analysis of the phasor plot G axis in different stimulation conditions .................. 26 Figure 13. Seahorse assay of INS-1E mitochondrial stress ........................................................................ 29 Figure 14. High glucose enhances mitochondria function and ATP production in INS-1E cells. .............. 30 Figure 15. Glucose stimulation enhanced OX PHOS by shifting phasor plot to lower G value. ............... 32 Figure 16. Alpha and beta cells at distinct metabolic states ....................................................................... 34 Figure 19. Protein sequences of IAPP between 20 -29 amino acids in different species. .......................... 46 Figure 20. Beta cells upregulated OX PHOS level in response to glucose stimulation .............................. 48 Figure 21. Alpha and beta cells phasor plot of G value in non-diabetic (ND) mouse islets ....................... 50 Figure 22. FLIM signature shifts to glycolysis under glucose stimulation in hIAPP mouse islets............. 52 Figure 23. Phasor plot of G values for alpha and beta cells in hIAPP transgenic mouse islets. ................. 53 Figure 24. Relative OX PHOS level change of alpha and beta cells in ND and hIAPP transgenic mouse islets ............................................................................................................................................................ 54 Figure 25. Hub cells synchronized all beta cells OX PHOS response in healthy islets but not in hIAPP islets ............................................................................................................................................................ 56 Figure 26. The autofluorescence signal of hIAPP islets. ............................................................................ 56 Figure 27. 3D projection of overlapped FLIM image and SHG signal ...................................................... 57 Figure 28. Excitation and emission spectrum of common auto-fluorescent substances in live tissue ........ 66 Figure 29. The auto-fluorescence signal of mouse islets in two channels .................................................. 67 Figure 30. The auto-fluorescence signal of human islets in two channels .................................................. 68 Figure 31. Human islets 3D view of FLIM signal ...................................................................................... 69 Figure 32. OX PHOS change in human nondiabetic islets and T2D islets from two patients. ................... 70 Figure 33. Slice view of human islets under 16 mM Glucose and 30 µM NN414. .................................... 71 Figure 34. FLIM imaging of a single beta cell stained by TMRM ............................................................. 73 Figure 35. Single beta cell lifetime color-coding map ................................................................................ 74 Figure 36. HySP phasor plot of auto-fluorescence in the human islet. ....................................................... 76 Figure 37. NADH and NADPH involved metabolic pathways .................................................................. 83 Figure 38. PBC Consortium whole cell model ........................................................................................... 85 Figure 39. Scheme of predicting and aligning hub cell location ................................................................. 87 Figure 40. Holotomography imaging .......................................................................................................... 90 vi Figure 41. Longitudinal hyperglycemia timeline scheme. .......................................................................... 94 vii List of Tables Table 1. Cell lines resources and cultures. .................................................................................................. 42 viii Abstract Both type 1 and 2 diabetes (T1D and T2D) are characterized by a progressive loss of beta cell function. In both T1D and T2D, impaired glucose-stimulated insulin secretion precedes diabetes onset, which declines further after diabetes and is attributed to glucose toxicity superimposed on beta cell stress. In healthy individuals, the rate of insulin secretion is tightly regulated by blood glucose concentration, although there may be heterogeneity between beta cells for the threshold glucose concentration at which they are recruited to secrete insulin. Glucose readily enters beta cells through membrane resident glucose transporter proteins and the rate of glycolysis is regulated by glucokinase with a Km in the mid physiological range. All pyruvate generated by glycolysis enters the TCA cycle generating ATP that prompts insulin exocytosis. In diabetes, metabolism pathways are reprogrammed, and glycolytic lactate production is upregulated. Glycolysis generates free nicotinamide adenine dinucleotide (NADH) while the TCA cycle generates NADH bound to complexes of the electron transport chain. By applying the phasor approach of fluorescent lifetime microscopy (FLIM), this thesis investigates and establishes strategies to quantify the relative abundance of free and bound NADH in living pancreatic islets and thus the relative rate of glycolysis and oxidative phosphorylation (OX PHOS) in response to glucose in individual cells within islets. For the first time, it has been possible to characterize alpha and beta cell identity and function non-invasively in dispersed and isolated islets enabling the observation of both the suppressed OX PHOS activity in the diabetic mouse model and beta cell hub signaling in healthy islets. The same approach was further applied to human islets in vitro. Finally, islet metabolism was successfully monitored and evaluated by live imaging of FLIM. The findings presented in this thesis will make it possible to non-invasively detect diabetes pathological changes in pharmaceutical and clinical studies in the near future. 1 Chapter 1: Diabetes, pancreatic islets, and beta cells. 1.1 Diabetes Mellitus Diabetes mellitus, known as Diabetes, is a group of metabolic disorders characterized by prolonged hyperglycemia. In normal people, plasma glucose levels are typically between 70 to 100 mg/dL after fasting overnight, but will quickly increase within the first 30 minutes after an oral glucose load, slipping below 140 mg/dL after 2 hours (Figure 1). In contrast, with diabetic patients, plasma glucose is higher than normal; over 126 mg/dL for a fasting test and over 200 mg/dL after a 2-hour oral glucose tolerance test (OGTT) [1, 2]. In addition, diabetic patients can also experience the production of an insufficient amount of insulin and/or the ineffective use of insulin. Figure 1. Blood glucose level of non-diabetic (blue) and diabetic (red) patient after OGTT. Adapted from diabetesbook.org. Based on differences in pathogenesis, diabetes is divided into two main Types – Type 1 and Type 2 (T1D and T2D, respectively). Both Types of diabetes can lead to similar long-term complications, such as cardiovascular disease, kidney damage and retinopathy due to continuous exposure to 2 hyperglycemia [3]. However, with ongoing cellular and molecular level research on diabetes, more subtypes are being discovered and defined [4]. In this chapter, the mechanisms of T1D and T2D are reviewed, focusing on T2D pathogenesis within the pancreas. Then the different T2D drugs on the market and their mechanisms are discussed in general terms. 1.1.1 Type 1 Diabetes T1D, also known as insulin-dependent diabetes, results from autoimmune destruction of insulin- producing pancreatic beta cells that, mistaken as foreign, are destroyed by CD8-positive T cells [5]. Autoantibodies targeting beta cells, such as insulin (IAA) and protein tyrosine phosphatase (IA2 and IA-2beta), are usually detected and can be used for T1D prediction in patients [6]. However, there are other pieces of evidence and cases in different regions with different ethnic groups show that autoantibodies are not the gold standard for T1D. For example, in the Osaka area of Japan, a subgroup of patients was characterized by rapid-onset ketoacidosis and the absence of autoantibodies mentioned above [7, 8]. Because of the clinical heterogeneity of T1D, researchers have classified autoimmune-driven diabetes as Type 1A and non-autoimmune-related diabetes with beta cell destruction and notable insulin deficiency are categorized as Type 1B [9]. Both genetics and environmental factors are considered as triggers of T1D. The overexpression of major histocompatibility complex (MHC) antigens accounts for the vulnerability of beta cells [10] and MHC human leukocyte antigen (HLA) alleles are highly linked to the risk of T1D. Meanwhile, several environmental factors are also associated with diabetes onset, including virus, exposure to seasonal antigens and nutritional imbalance [11]. 3 Insulin is the central solution to T1D for now. Since it is still not possible to prevent beta cell loss, insulin is prescribed as an exogenous supplement to maintain blood glucose levels [5]. Interstitial glucose monitors and various kinds of insulin injection pens and pumps make it easier for patients to maintain healthy blood glucose levels with minimum side effects. Other solutions, including transplanting islets (groups of beta cells) or the whole pancreas itself, offer the potential to relieve patients from daily insulin injections [12]. Despite that, transplantation is still not available to the majority of patients. Therefore, the main challenge in overcoming beta cell loss in T1D is beta cell regeneration. While, beta cells, as somatic cells, are fully differentiated and irreproducible, recent evidence of cell identity turnover and dedifferentiation in pancreatic islets suggests potential treatments stop and reverse beta cell loss in patients [13, 14]. 1.1.2 Type 2 Diabetes T2D is usually characterized by insulin resistance, reduced insulin secretion, and beta cell dysfunction without apparent beta cell loss [2]. There is a huge heterogeneity in pathogenesis and disease development within T2D patients. Even though there is evidence of genetic, diet, circadian rhythm and environmental influences, there is no single cause of T2D. The American Diabetic Association (ADA) defines the prediabetic state as exhibiting higher fasting glucose at 100-125 mg/dL or OGTT at 140-199 mg/dL. Prediabetes is the reversible early stage of diabetes development [15, 16]. Even without drug treatment, lifestyle modification and weight loss can reduce the relative risk of developing diabetes. This evidence provides encouragement to investigate the pathogenesis of diabetes in order to systematically understand diabetes 4 development and take advantage of the reversibility of the prediabetic stage, in order to prevent, control and reverse diabetes progression. T2D is a complex, progressive and chronic disease. One hypothesis is that T2D is a kind of protein misfolding disorders (PMDs) similar to Huntington’s and Alzheimer’s diseases [17, 18]. In PMDs, dysfunctional proteins aggregate in non-native conformations. These aggregated proteins may precipitate on the cell-matrix, cell membrane or subcellular organelles and interfere with cell function thereby increasing cell stress. Previous evidence suggests that human islets amyloid polypeptide (hIAPP) can form toxic aggregates, which are found in 90% of T2D patients [18, 19]. However, in diabetes and other PMDs, aggregates can be observed in aged individuals with no disease symptoms. Another hypothesis is that T2D can be a result of inflammation [20-22]. Inflammation induces oxidative stress and overexpression of reactive oxygen species (ROS). It has been shown that chronic pancreatitis increases the risk of diabetes development [23]. Cytokines can inhibit beta cell insulin section and induce beta cell apoptosis [24]. Proinflammatory cytokine tumor necrosis factor-α (TNF-α) can inhibit insulin receptor expression and cause insulin resistance in our tissue [25]. High levels of cytokines IL-1B and IL-6 can induce endoplasmic reticulum (ER) stress in beta cells and lead to diabetes in prediabetic mice [26]. Glucolipotoxicity is another potential pathogenesis of T2D [27-30]. Long-term hyperglycemia and free fatty acid levels induce impaired insulin section and beta cell apoptosis. This hypothesis also 5 suggests that obesity and cholesterol are risk factors of diabetes pathogenesis [31-34]. Research has shown that a high fat diet (HFD) can cause beta cell function loss by decreasing glucokinase activity and increasing ER stress in beta cells [35-37] and can also trigger macrophage infiltration and elevate inflammatory response in islets [38, 39]. 1.2 Pancreatic islets: functional unit of the pancreas 1.2.1 Cell Types There are three main types of cells in the pancreatic islets: alpha, beta and delta cells. Alpha cells secrete glucagon to upregulate glucose production and fatty acid levels in the liver [40, 41]. They help to maintain euglycemia and glucose homeostasis under starvation conditions. Beta cells are the key regulators of insulin secretion and glucose levels after glucose intake [42]. Beta cell loss is the most common underlying factor in all types of diabetes. Delta cells produce somatostatin to regulate insulin and glucagon secretion in the pancreas, and are also found in the stomach and intestines [43]. Another cell-type of low abundance is the Pancreatic polypeptide cells (PP-cells) that secret pancreatic polypeptide for regulating islet homeostasis and stimulating gastric juice secretion. Beta cell heterogeneity is an evolving concept since, while they are the most abundant cell in the islets, it now recognized that not all beta cells have the same transcriptional and translational features [44-47]. At the metabolic level, not all beta cells are insulin-responsive nor responsive at a similar level [48, 49]. Unresponsive beta cells need a higher concentration of glucose to trigger insulin secretion [50]. At the level of embryonic development, not all beta cells grow into mature 6 cells that secret insulin [51]. During islet development, beta cells showed heterogenous glucose- sensing during maturation in zebrafish [52]. The Ucn3 gene is a maturation marker in embryonic development since it is expressed in all beta cells in adult mice. As a sign of under-differentiation, Ucn3 negative beta cells express insulin but are functionally and transcriptionally immature. Huising et al. characterized Ucn3 negative beta cells as “virgin beta cells”, which are transdifferentiated alpha cells and form a “neogenic niche” [51]. These subtypes of the beta cell might be responsible for beta cell generation in adults. 1.2.2 Microenvironment of islets Of all the cells in mouse islets, beta cells make up more than 80% [53]. Even though the size of the islets varies, it usually forms a core of hundreds of beta cells. Most of the other types of cells are located at the periphery of the islets. Blood vessels and nerve fibers cross the central part of the islets to regulate homeostasis and provide oxygen. Human islets are quite different from mouse islets, as the islets are compartmentalized into multiple sections based on blood vessel distribution. Alpha cells are also located in the core area next to the blood vessel structure. In human islets, only 65% of the cells are beta cells with a higher proportion of alpha cells. But in both mouse and human islets, blood vessels are crucial in maintaining oxygen supply and hormone regulation. They bring in glucose to islets and take insulin, glucagon and other hormones to other organs. 7 Connective tissues and inflammatory-related macrophages have also been observed in pancreatic islets [54]. Beta cell response is mainly the result of glucose stimulation, it is also regulated by glucagon and somatostatin secreted by alpha and delta cells. At the same time, alpha and delta cells are also regulated by insulin secreted by beta cells directly [55]. The microenvironment of the islet is highly ordered and they self-regulate for maintaining functional unit homeostasis. In order to understand diabetes pathogenies, we need to investigate not only beta cell function but also its communication with other cell types and subtypes. 1.2.3 Beta cell forms a signaling hub To understand beta cell functional heterogeneity, researchers have started to map the functional architecture in islets. Studies have shown that some beta cells respond earlier than others like a pacemaker [56]. After glucose stimulation, Ca 2+ influx wave response in these cells tend to precede and last longer than that in the others. Johnston et al. [57] identified these cell “hubs” as a pacemaker with preceding activity in Ca 2+ signaling. The other “follower” cells have lagging and short Ca 2+ influx wave, synchronously responding after “hub” cells. If the cell-cell junction is interrupted, hub cells will fail to lead the follower cells synchronization and islet insulin secretion will be down-regulated. These hub cells are metabolic and transcriptionally immature with lower glucokinase expression than follower cells. This evidence shows that both the subtype beta cell identity and microenvironment are critical for functional insulin release. 8 1.3 Pancreatic beta cells: insulin factory 1.3.1 Glucose stimulated insulin secretion Pancreatic beta cell metabolism balance is associated with diabetes development and treatment. Unlike many other cell types and cancer cells in particular, the pancreatic beta cell tightly regulates its glucose intake via glucokinase [58]. The Km value of glucokinase is 10mM, which is close to the physiological glucose concentration after a meal [59]. Thus, beta cells sense higher levels of glucose and start to metabolize glucose to produce insulin. The amount of insulin secreted by beta cells is tightly regulated and glucose dose-dependent [60, 61]. The insulin secretion rate of beta cells appears in two-phases during the first hour after glucose stimulation (Figure 2). In the first phase, beta cells release the pre-stored granules from the granule pool [62]. This fast release happens within the first 5-10 minutes. After the first phase, beta cells start to make new granules and transport them to the cell membrane for continuous release. Figure 2. Biphasic insulin secretion of beta cell after glucose stimulation. Adapted from ‘Modeling phasic insulin release: immediate and time-dependent effects of glucose’ [63]. 9 In the first phase, glucose metabolism enhances ATP production in beta cells. ATP activates the ATP-sensitive K + channel on the cell membrane to depolarize it. Depolarization activates the Ca 2+ voltage channel and induces Ca 2+ ion influx. Ca 2+ signaling induces granule trafficking and exocytosis (Figure 3). It has been reported that beta cells have two different pools of granules that are regulated by different sources of Ca 2+ , which might be responsible for biphasic secretion [64, 65]. However, it’s still unclear how beta cells regulate their pools and the exchange of granules in between. Figure 3. 11 Major metabolic pathways related to insulin secretion in pancreatic beta cells[66]. Reprinted with the permission of ‘Mechanisms of the amplifying pathway of insulin secretion in the β cell’ by Kalwat M, Pharmacology & Therapeutics, 2017. 179: p. 17-30. Copyright 2017 Elsevier Inc. 10 Many other factors also regulate insulin section, such as glucose-like peptide 1 (GLP-1) and fatty acids (Figure 3). The beta cell appears as a simple system for glucose input and insulin output. However, inside the system is a big black box with hundreds of metabolic pathways forming a complex metabolic network. Therefore, in order to better understand beta cell function and its failure in diabetes, data must be integrated from multiple interconnected experiments that probe impacts between these pathways. 1.3.2 Glycolysis and oxidative phosphorylation in beta cells One of the key factors for maintaining beta cell function is the accurate regulation of glycolysis and oxidative phosphorylation (OX PHOS) required for providing energy for insulin exocytosis. Glycolysis is the first step for breaking down glucose into pyruvate in the metabolic pathway. As the glucose sensor, glucokinase triggers the metabolic flux and increases pyruvate production [67, 68]. In healthy human beta cells, to maximize energy production for insulin secretion, lactate dehydrogenase (LDH) activity is low and unfavorable for metabolism [69-71]. Metabolizing glucose to lactate, fermentation glycolysis through LDH only produces 2 ATP, while the TCA cycle and OXPHOS can produce 36. Pyruvate is transported into mitochondria and enters the tricarboxylic acid (TCA) cycle, fueling OX PHOS. Increased OX PHOS leads to higher levels of ATP production, which depolarizes the cell membrane through KATP channel closure and further induces Ca 2+ channel opening [72]. As a consequence, insulin granules are released in response to elevated glucose levels [67]. 11 In fasting conditions, both glycolysis and OX PHOS activity are low due to the high Km of glucokinase and low glucose influx. When beta cells are fed with a high concentration of glucose, glucokinase starts to process the glucose, leading to increased rates of glycolysis. As more pyruvate enters the TCA cycle, the OX PHOS level is also enhanced (Figure 4). Figure 4. Metabolism scheme of non-diabetic (ND) beta cells under glucose fasting (left) and feeding (right) conditions. The font size indicates the relative abundance of substrates. The weight of arrows reflects the relative abundance of energy flow. A critical metabolic intermediate, nicotinamide adenine dinucleotide (NADH), is produced and recycled in glycolysis and OX PHOS. As an electron carrier, NADH interacts with respiratory Complex I, the first enzyme complex on the electron transport chain (ETC), to produce a proton gradient for ATP production in the OX PHOS pathway. Since the NADH cannot cross the mitochondria membrane, the total pool of NADH and NAD + is the same. In mitochondria, pyruvate goes into the TCA cycle and produces NADH to be further oxidized into NAD + on Complex I to produce a proton gradient. Glucose consumption enhances glycolysis and the increases the supply of pyruvate into the TCA cycle. As a result, glucose consumption also 12 increases the activity of ETC to maintain the NAD + pool in the mitochondria. In other words, with the increased pyruvate supply, the turnover rate of NAD + to NADH and back to NAD + must increase in order to maintain the energy flow. This turnover rate can be represented by the ratio of NADH binding enzyme activity or enzyme-binding NADH to free-NADH. As illustrated in Figure 4, glucose stimulation can enhance the OX PHOS reaction in mitochondria and increase the bound- NADH/free-NADH ratio in mitochondria. 1.3.3 Metabolism reprogramming in diabetic beta cells In diabetic beta cells of both T1D and T2D, lactate production is not prevented by to metabolic reprogramming [58, 70, 71, 73, 74]. Increased HIF1α is commonly observed in diabetic beta cells, which can inhibit pyruvate dehydrogenase (PDH) and enhance LDH function. As a result, pyruvate metabolism is diverted from the TCA cycle to lactate production (Figure 5). The glucose- stimulated OX PHOS level is suppressed in diabetic beta cells, as is the bound-NADH/free-NADH ratio. 13 Figure 5. Metabolism scheme of non-diabetic (ND) beta cells (left) and diabetic beta cells (right) under glucose feeding conditions. The font size indicates the relative abundance of substrates. The weight of arrows reflects the relative abundance of energy flow. 1.4 How to understand beta cell metabolism in diabetes development Since beta cell healthiness can be determined by its metabolism state, understanding metabolism reprograming during diabetes development benefit the drug discovery of diabetes treatment and prevention. The increased activity of LDH diverts the metabolic flux into glycolysis as the Warburg-like effect in beta cells [74, 75]. Thus, the shift to glycolysis leads to a drop in OX PHOS, and ATP production responding to glucose stimulation. Even though the trigger of diabetes onset is unclear, it is true that either amyloid-induced ER stress or inflammation can lead to metabolic reprogramming [75, 76]. The enhanced lactate production observed is quite similar to cancer and stem cells, which are underdifferentiated and less mature. As beta cells are somatic cells that are fully differentiated, metabolic reprogramming might cause cell stress and even cell failure. 14 In this thesis, the bound-NADH/free-NADH ratio is exploited to investigate metabolic regulation in beta cells and to develop a less-invasive way to study beta cell function in islets. 15 Chapter 2: Investigating beta cell metabolism by live-cell imaging and biochemical assays 2.1 Background: Fluorescent lifetime imaging (FLIM) - characterizing metabolism without labeling 2.1.1 Fluorescent lifetime A substance that can re-emit light after light excitation is characterized as a fluorophore. The fluorescent lifetime refers to the average amount of time that a fluorophore stays in an excited state. As illustrated in Figure 6, a fluorophore can absorb energy from a short-wavelength photon (absorption spectrum), and, after a certain amount of time, will emit the energy in light form, i.e. fluorescence (emission spectrum). 16 Figure 6. Jablonski diagram illustrating the energy change of fluorophore after exciting [77]. Reprinted by permission of Elsevier from ‘Fluorescence lifetime imaging ophthalmoscopy’ by Dysli, C. Progress in Retinal and Eye Research.2017. 60: p. 120-143. Copyright 2017 The Authors. Published by Elsevier Ltd. This procedure follows first-order kinetics. Thus, the intensity of a fluorophore after excitation can be expressed as a function of time (t) in a single exponential decay: I ( t ) = I 0 e − t / τ [78] where I0 is the initial intensity. The fluorescent lifetime, τ, is the time it takes for the intensity to reach 1/e of I0. The lifetime of a fluorophore is independent of excitation light intensity or fluorophore concentration (I0). The molecular properties of an intrinsic fluorophore, such as spectrum and lifetime, can be affected by pH, temperature and molecular-protein interaction. 17 2.1.2 Principles of FLIM imaging In practice, usually more than one fluorophore is excited in samples, and a single fluorophore may have more than one lifetime. In that case, the intensity of the experimental sample is a sum of intensities of each decay, proportional to the fractional contribution of every single exponential decay contributed from each species: I ( t ) = I 0 ∑ α i i ∙ e − t / τ i where αi is the weighted amplitude as determined by molecule abundance. Thus, the average lifetime of multiple fluorophores is given by: τ α = ∑ α i n i = 1 τ i In traditional time-correlated single-photon counting (TCSPC) FLIM imaging, the photon distribution that we measured at each pixel should represent the sum exponential decay of all excited fluorophores in that pixel [79] (Figure 7). Due to instrument limitations, the photon distribution needs to be fitted into either a single exponential decay in order to calculate τα or into multiple exponential decays to get the corresponding lifetime, τi, and amplitude, αi. The accuracy of the fitting process is affected by both data quality and data complexity. Thus, in this study, the Leica SP8 FALCON was used for better data quality and phasor analysis of FLIM imaging for a more intuitive and fitting-free approach. 18 Figure 7. Diagram of TCSPC microscope principles. Photon times is collected after short laser pulses excitation. The photon histogram is formed by repetitively cycle of excitation-measurement [79]. Reprinted with the permission of ‘The bh TCSPC Handbook, Eighth Edition’ by Becker & Hicki GmbH. 2.1.2 Principles of phasor - FLIM analysis The phasor FLIM analysis uses Fourier transformation to transform the decay trace of each pixel into cosine and sine expressions (G and S) and plot them in the phasor plot [80] (Figure 8 and Figure 9B). The relationship between the phasor values and the lifetime τ is as follows: G = ∑ f i 1 + ω 2 τ i 2 N i = 1 and S = ∑ f i ω τ i 1 + ω 2 τ i 2 N i = 1 where ω is 2π times 80 MHz (as determined by instrument setting), N is the number of fluorescent species and fi is the fractional contribution of species i. Meanwhile, the decomposition of the phasor plot is fit free. The phasor point (G, S) of multiple species is a linear combination of each fluorescent species. Even in a complex biological system, it is still possible to study the relative contribution of two multi-component species. As in Figure 8, a single component species will have a phasor point that falls on the semicircle. Together, 19 species A and B are multi-component species. Their average lifetime depends on the fractional contribution of two species and will lie on the line between their individual phasor point. Figure 8. The linear combination of multiple species on phasor plot [80]. Reprinted with permission of ISS. Each point on a phasor plot (Figure 9B) represents the lifetime data of each pixel on the raw fluorescent image (Figure 9A). This one point to one pixel relationship gives us a more intuitive readout of lifetime distribution in space. A lifetime mask can be applied to the raw intensity fluorescent image, based on the phasor plot distribution, to study which pixel has a shorter lifetime. For example, in Figure 9C, the shorter lifetime is masked in blue. It is obvious that most of the blue pixels are located in the nuclear area of cells, indicating a shorter lifetime at the nucleus. 20 Figure 9. FLIM lifetime map flow chart. A&B. Original NADH fluorescent intensity of INS-1E cells and corresponding phasor plot. C. Rainbow lifetime map combined the original intensity figure and lifetime mask based on phasor distribution. 2.1.3 Photobleaching and phototoxicity Live-cell imaging is always challenging due to the inevitable photobleaching and phototoxicity induced by the excitation laser, especially for time-lapse imaging [81]. However, the photon count is also crucial for acquiring effective decay traces in FLIM imaging. Thus, a balance must be sought between sufficient photon counting and minimized damage by testing imaging parameters, such as appropriate excitation laser power, scanning repeats number and scanning speed. 2.1.4 Live cell imaging in deep tissues A 2-photon excitation laser was used to replace the traditional 1-photon laser in confocal imaging. Instead of using the energy of one photon to excite a fluorophore as in Figure 6, two less energetic but longer wavelength photons were used for the same purpose. The 2-photon excitation can limit the excitation area to a smaller area size around the focal point, so that it reduces the photobleaching effect, and it has the ability to penetrate deeper into tissue with less scattering [82, 83]. 21 2.1.2 NADH lifetime in live cells The target fluorophore, reduced NADH, as an endogenous autofluorescence biomarker is tightly linked with glycolysis and OX PHOS in cells [84]. It can be reduced into NAD(P) + , which has no fluorescence (Figure 10). The free form of NAD(P)H in the physiological buffer has a short lifetime of 0.4 ns. In a redox reaction, the binding of NAD(P)H to an enzyme will slow down the release of photons and increase the lifetime to ~ 2-4 ns. Figure 10. NAD(P)H molecular structure and its fluorescent structure [84]. Reprinted by permission of Elsevier from ‘Investigating mitochondrial redox state using NADH and NADPH autofluorescence’ by Blacker, T.S. Free Radic Biol Med, 2016. 100: p. 53- 65. Published by Elsevier Ltd. The exact lifetime of bound-NAD(P)H is determined by the binding enzyme type [85]. Thus, the metabolic pathway composition and activation relationship can influence and determine the average NAD(P)H lifetime in different cell types [86, 87]. The bound-NADH lifetime of certain kinds of enzymes has been quantified in vitro by the purified form of the protein [85, 88]. However, due to the stability and availability of different enzymes, not all key enzymes can be purified and 22 their NADH-bound lifetime quantified, such as Complex I. Thus, it is still challenging to determine the exact enzyme type that is affecting the FLIM signature in different cell types. Here, we hypothesis that two enzymes, LDH and PDH, are the key factors of the FLIM signature in pancreatic beta cells. NADH-LDH has a shorter lifetime (1.6 ns) than NADH-PDH (2.5 ns) [85]. As mentioned in Chapter 1, lactate generation is energetically unfavorable and prohibited in beta cell metabolism. Increased levels of glycolysis and a decreased average NADH lifetime can be monitored as a result of upregulated LDH activity and NADH-LDH abundance in diabetic beta cells. Thus, the NADH lifetime can reflect the contribution of energy generated through OX PHOS (PDH activity) and glycolysis (LDH activity). The term “glycolysis”, we are using in this metabolic trajectory, is referring in particular to the anerobic component of glycolysis from pyruvate to lactate. 2.2 Results 2.2.1 FLIM shows subcellular metabolism in the monolayer cell lines FLIM imaging was first applied on the pancreatic beta cell line INS-1E, grown as monolayer on glass coverslip, to test imaging parameters (Figure 11). The NAD(P)H signal was collected in a 440-500 nm emission range with 740 nm 2-photon excitation by Leica SP8 FALCON FLIM- Confocal microscopy. Imaging was done at more than 50% cell confluency. The INS-1E cell line forms colonies, growing into piles first instead of spreading out as other cancer cell lines. Therefore, we began imaging before the cells started to form multilayers (Figure 11A). 23 In Figure 11B, the NAD(P)H intensity map indicates that the cytoplasm and mitochondria have the highest NAD(P)H concentrations, while the nuclei are less abundant in NAD(P)H. From the fluorescence intensity decay at each pixel, the lifetime of each pixel was transformed into a point on the semicircle phasor plot. The lifetime distribution in the FLIM lifetime map corresponds to the rainbow color coding based on the phasor plot, which represents the relative free and bound ratio of NAD(P)H in each pixel. Under low glucose conditions (1.1 mM), the cell’s cytoplasm contained more bound-NAD(P)H, colored as green and red (Figure 11C). In cell nuclei, more free- NAD(P)H was observed, which corresponds to higher nuclear glycolysis activity through the pentose phosphate pathway (PPP). After being stimulated with 16.7 mM of glucose for 30 minutes, increased bound-NAD(P)H occurs in the cytoplasmic region, with more red pixels present as shown in Figure 11D. 24 Figure 11. The NAD(P)H average lifetime increased and shifted to OX PHOS under high glucose in INS-1E cells. A. The bright field view of INS-1E showed the cell boundary and morphology. B. NAD(P)H intensity map of INS-1E illustrated NAD(P)H distribution in the cells. C. NAD(P)H lifetime map of INS-1E under 1.1 mM glucose condition showed higher ratio of free- NAD(P)H. D. 16.7 mM glucose condition enhanced OX PHOS level with higher ratio of bound-NAD(P)H. (Scale bar, 10 µm) To better understand this color shift, we quantified the phasor distribution of each figure by calculating the median G value of all the phasor points. This included a few pre-processing steps to increase the signal-to-noise ratio. This analysis strategy is discussed in detail in the Method and 25 Material section of this chapter. Since the S axis range (0 - 0.5) was much smaller than the G axis range (0 – 1), the G value change was used to represent the lifetime change in the phasor plot. The higher the average lifetime, the lower the G value will be. The decrease of the G value represents the increased level of OX PHOS on the metabolic trajectory [89]. In addition, we defined ∆G as the metabolism indicator. ∆G = G (before treatment) – G (after treatment) In all glucose stimulation assays, ∆G = G (low glucose) – G (high glucose). The positive ∆G indicates enhanced OX PHOS activity after treatment while a negative value represents enhanced glycolysis. 2.2.2 Glucose stimulation enhances beta cell oxidative phosphorylation The G values of a single group of beta cells in low and high glucose conditions are compared in Figure 12A. Each data point represents the phasor of cells in one field of view. There was a significant drop in the G value for all data points, indicating increased OX PHOS activity after glucose stimulation. In Figure 12B, comparing to the control group, in which no high glucose treatment was performed, the ratio of OX PHOS/Glycolysis was much higher. This verified our hypothesis that in beta cells, glucose can enhance the OX PHOS level. 26 Figure 12. Quantitative analysis of the phasor plot G axis in different stimulation conditions. A. Median G value of phasor plots decreased after high glucose stimulation in four representative views. B. Relative OX PHOS level change under no treatment (control), 16.7 mM glucose stimulation, glucose with 100 µM koningic acid (KA) and glucose with 100 µM aspartate (ASN). Scale bar, 10 µm. n>3 for each condition. *P<0.5, ****P<0.0001, unpaired two-tail t-test was used of all comparisons. Data were plotted by Box-and-Whisker Min-Max to show the distribution. 2.2.3 Glucose affects basal respiration and spare respiration capacity of beta cells The mitochondrial response and OX PHOS enhancement observed in FLIM studies can also be verified by measuring the oxygen consumption rate in INS-1E cells (Figure 13A). Seahorse assays, which evaluate cell mitochondria stress, on INS-1E cells were performed under control of low (1.1 mM) and high (16.7mM) glucose concentrations, and the oxygen consumption was monitored by disrupting mitochondrial function as in the flow path shown in Figure 13B. Basal respiration, ATP production, proton leakage, and spare capacity of mitochondria were calculated as illustrated in the figure. 27 After glucose stimulation, the oxygen consumption rate (OCR) increased under the presence of high concentration glucose. The elevated OCR showed the energetic oxygen demand of beta cells after glucose in-flux. Due to the increased carbon supply, ATP production was also increased. Then ATP synthase was inhibited by oligomycin, to terminate ATP synthase linked oxygen consumption and measure ATP production rate in samples. This means all OX PHOS linked ATP was eliminated. Thus, the higher ATP production level of high glucose stimulation resulted in enhanced OX PHOS levels (Figure 14B). Carbonyl cyanide-4 (trifluoromethoxy) phenylhydrazone (FCCP) was used to uncouple OX PHOS from ATP synthase by disrupting the proton gradient across the mitochondria inner membrane. As a result, oxygen consumption through Complex IV was not inhibited by high proton concentration in the mitochondrial intermembrane spaces. The OCR reached its highest level, showing the maximum mitochondrial OX PHOS capacity. The spare respiratory capacity is the difference value of maximal and basal respiration, which reflects the cell mitochondria function and its ability to reach its full potential in responding to energy demands. With low glucose, INS-1E cells had quite a low level of spare respiratory capacity indicting poor mitochondria fitness and flexibility (Figure 14A). All OX PHOS was terminated through rotenone and antimycin A treatment, which inhibited Complexes I and III, respectively. The proton leak could be calculated by subtracting non- mitochondria OCR from oligomycin inhibited OCR as in Figure 14B. The proton leak level was quite similar in both low and high glucose. Therefore, it was speculated that glucose level does not 28 affect mitochondrial health but does influence its capacity and flexibility. Regulation of mitochondria capacity and flexibility may be fulfilled by the regulation of the mitochondria network and fragmentation, which is discussed later in this chapter. 29 Figure 13. Seahorse assay of INS-1E mitochondrial stress measures oxygen consumption rate (OCR) under multiple modulators of respiration. A. OCR was measured three times with 10 min intervals after modulators treatment in three conditions. Control, 1.1 mM glucose. 25MM GLUC, 25 mM glucose. 25MM GLUCOSE + EXENDIN4, 25mM glucose and 10nM exendin-4. n=12 for each group. Data were plotted as mean ± standard error of the mean (SEM). B. The cell mitochondria stress test profile show the key parameters of mitochondria function. Reprinted with the permission of ‘XF analyzer user guide’ by Agilent Technologies, Inc. 30 Figure 14. High glucose stimulation enhances mitochondria function and ATP production in INS-1E cells which can be weakened by administration of exendin-4. A-B. Basal respiration, spare reparation capacity, proton leak and ATP production evaluations. n=12 in each group, Data were plotted as mean ± standard error of the mean (SEM). 31 2.2.4 Exendin-4 can decrease insulin secretion by decreasing oxidative phosphorylation. It was also interesting to investigate how the commercial diabetic drug exenatide, Exendin-4 (Ex- 4), boosted insulin secretion through the metabolic pathway. Thus, mitochondria health and function with Ex-4 treatment was evaluated by Seahorse assay (Figures 13&14). Both basal respiration levels and ATP production levels were kept at a similar level with the Ex-4 additive. This indicated that Ex-4 was not affecting insulin secretion by producing more respiratory ATP. This finding aligned with our knowledge of its activation through the GLP-1 receptor and the cAMP pathway. However, it was noticeable that Ex-4 diminished the OCR that was initially induced by glucose and decreased the spare respiratory capacity of mitochondria. This indicated Ex-4 may redirect energy production from OX PHOS through mitochondria to glycolysis through fermentation respiration. This shift may be achieved by directly altering mitochondrial function or may in turn influence mitochondrial flexibility. 32 Figure 15. Glucose stimulation enhanced OX PHOS by shifting phasor plot to lower G value. Quantitative analysis of the phasor plot G axis in INS-1E cells under different stimulation conditions in time-lapse FLIM imaging. 0G, 0 mM glucose. 2.8G, 2.8 mM glucose. 25G, 25 mM glucose. 2.8G+GLP-1, 2.8 mM glucose and 10 nM GLP-1. 25G+GLP-1, 25 mM glucose and 10 nM GLP- 1. Data were plotted as G median value of a single experiment. From the FLIM imaging, increases in G value were detected in both high (25 mM) and low (2.8 mM) glucose conditions in the first 10 minutes of stimulation, meaning a decrease of OX PHOS levels (Figure 15). This aligned with the decreased OCR level observed in the seahorse assay in Figure 13. 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0 10 20 30 40 50 60 70 G TIME (MIN) INS 0G 2.8G 25G 2.8G+GLP-1 25G+GLP-1 33 2.2.5 Dispersed rodent alpha and beta cells possess different FLIM character After verifying that the FLIM phasor plot can monitor changes in cellular metabolism and evaluate the relative energy contributions of OX PHOS and glycolysis in cell lines, we applied this technique to investigate mouse pancreatic beta cells. Before directly looking into deep tissue, mouse islets were isolated and dispersed into single cells growing in a monolayer. The mouse islets were dispersed and seeded on grid-slides to allow for tracing the cell types (alpha, delta, and beta cells). To decrease stress caused by excessive manipulations, we did not sort the cells using antibodies against corresponding membrane receptors. The cell type was verified by fixing the cells after FLIM imaging and staining with cell type specific markers. Small morphology changes were detected after time-lapse imaging and fixing. Also, some cells were lost due to washing during stimulation and staining (Figure 16, white arrow), but for many cells we were able to trace back to the same cell by spatial distribution alignment. Beta cells were detected by the insulin antibody. They are usually bigger and more spread than alpha cells. (Figure 16B, green arrow) The diameter of beta cells can vary from 5-10 µm. In contrast, alpha cells tend to be smaller and more compact under the bright field with a diameter of 4-6 µm. Even though alpha cells are smaller and only make up around 15% of the total population, it is still difficult to distinguish between beta cells and alpha cells only under the bright field of view. We first investigated the metabolic signature of mouse alpha (red arrow) and beta cells (green arrow) under low (4 mM) glucose. It was clear from the FLIM color that alpha cells are more 34 reddish and have a higher OX PHOS ratio than beta cells under low glucose (Figure 16A). Under low glucose, beta cells did not respond to produce energy. However, alpha cells needed to produce glucagon to maintain glucose homeostasis. Thus, alpha cells have a much higher OX PHOS level than beta cells under low glucose conditions. This difference in FLIM signature enables us to distinguish alpha and beta cells before fixation and antibody staining. We further quantified the median G values of alpha and beta cells. Beta cells had significantly higher G values than alpha cells, representing lower OX PHOS levels in beta cells under low glucose (Figure 16C). Figure 16. Alpha and beta cells at distinct metabolic states under low glucose. A. FLIM imaging of dispersed islets under 4 mM glucose. B. Immunofluorescent staining of the dispersed islets. Green arrows indicate beta cells. Red arrow indicate alpha cells. White arrows point to where cells detached during processing. C. Phasor plot of median G value of alpha and beta cells. Scale bar, 10 µm. n>8 for each condition. ***P<0.005, the unpaired two-tail t-test was used. Data were plotted by Box-and-Whisker Min- Max to show the distribution. 35 2.2.6 Alpha and beta cells have the opposite response to glucose stimulation The differences between alpha and beta cells at the basal level encouraged further investigation into the metabolism shift of both cell types after glucose treatment. One of the challenges in studying dispersed cells is the low photon count due to the low density of cells in a larger view. As a result, it is necessary to zoom into single cells to more effectively count photons, of course at the expense of increased phototoxicity. In Figure 15, it was shown that continuous imaging and scanning of cells with the laser at 5-10 minute intervals can bleach the autofluorescence signal and slightly increase the OX PHOS level (0G condition). In order to image cells at a higher magnitude, the time points to be investigated must be decreased. Thus, only low glucose and high glucose conditions after 30 minutes of stimulation were investigated. Rainbow color-coding was used to represent the three-dimensional NAD(P)H lifetime distribution for a collection of eight alpha cells and eight beta cells. Under high glucose stimulation, the OX PHOS level increased in beta cells. We can learn from the FLIM color shifting from blueish green to reddish-green (Figure 17, bottom), which indicates a higher ratio of bound-NAD(P)H in beta cells under high glucose stimulation. Quantified phasor points distribution also shows a drop of G value in four representative cells (Figure 18). 36 Figure 17. Lifetime changes of alpha and beta cells under low (4 mM) and high (16 mM) glucose. Scale bar, 5 µm. White dashed and dotted lines indicate the cell and nuclear boundaries, respectively. 37 Figure 18. Relative OX PHOS level change is positive in beta cells and negative in alpha cells. n>8 for each cell Type. *P<0.5, **P<0.05, ****P<0.0001, the unpaired two-tail t-test was used of all comparisons. Data were plotted by Box-and-Whisker Min- Max to show the distribution. Alpha cells show the opposite character as they shift from red, a more OX PHOS state, to green, showing a suppression in OX PHOS activity under high glucose (Figure 17, top). In Figure 18A, round dots, the G value of alpha cells increased as glucose concentration was raised. From the ∆G analysis of the relative activity of OX PHOS/Glycolysis in Figure 18B, alpha cell metabolism was found to change in the opposite direction to beta cells (i.e. ∆G was negative). Beta cells showed significantly higher OX PHOS activity than an alpha cell, after glucose stimulation. 38 2.3 Discussion In this chapter, we investigated methods to apply NADH FLIM imaging to monitor metabolism changes in beta cell lines and dispersed mouse islets. The NADH metabolic trajectory in the phasor plot has been used previously to study metabolism changes in cancer and stem cells [89-92]. However, NADH is involved in more than 10 metabolic reactions [84]. While many studies quantified the bound NAD(P)H lifetime of many different enzymes, including LDH and PDH, many other key metabolic pathway enzymes, such as the NADH-Complex, are missing [84, 85, 88, 93, 94]. Therefore, it remained unclear exactly how the NADH free/bound ratio relates to the glycolysis and OX PHOS reaction. While there is no direct evidence of free NADH as a product of glycolysis, free NADH exists in all cellular compartments including the nucleus and cytosol. Meanwhile, the bound-NADH lifetime varies from ~2-4 ns, depending on the enzyme type. As a result, the FLIM signature of a cell is determined by: 1) free NADH production rate as determined by the kinetic ratio of enzymes that are producing or using NADH; 2) the bound NADH lifetime as determined by the character of enzyme type; and 3) the relative abundance of these enzymes. Thus, in this chapter, the hypothesis that LDH and PDH are two key regulators in beta cells for determining the free/bound NADH ratio was put forward. Firstly, in metabolism from glucose to pyruvate, there are mainly two NAD(P)H enzymes. The first is G6PDH in the PPP, which is mainly located in the nucleus and is responsible for DNA synthesis[88]. Its influence was eliminated by masking out the nuclear area. The other is GAPDH (glyceraldehyde 3-phosphate dehydrogenase), which produces 2 free NADH for every glucose molecule. Free NADH produced here is translocated into the mitochondria for further utilization by the electron transport chain (ETC). 39 Thus, a higher intensity of NADH in the mitochondria was observed, as was a lower average NADH lifetime. When beta cells were stimulated with glucose, the energy flux also activates ETC, consuming oxygen to produce ATP. Thus, the free NADH could not stay free indefinitely and would bind to Complex I, generating a proton gradient. As a result, an overall increase of the bound NADH ratio was observed and taken to be an indicator of OX PHOS activity on ETC. The data also showed that by inhibiting GAPDH activity, no free NADH was produced. Therefore, there was an increased bound-NADH ratio since all NADH was loaded to ETC for maintaining the energy supply (Figure 4). Then the question was how the FLIM signature could uncouple glycolysis from OX PHOS since GAPDH in glycolysis is continuously fueling mitochondria OX PHOS. The research on FLIM signatures in stem and cancer cells indicated that undifferentiated cells have shorter average NADH lifetime and stronger glycolysis rate [95-101]. One of the key features of undifferentiated cells is lactate-favored energy production [102]. After pyruvate production, undifferentiated cells prefer to produce energy through LDH instead of going into the mitochondria for OX PHOS. NADH-LDH has a quite short lifetime compared to PDH and some other enzymes [85]. Thus, the NADH average lifetime in stem cells, cancer cells, and even in beta cells is not only indicative of mitochondrial OX PHOS levels, but also reflects LDH participation level in energy supply. Thus, in NADH-FLIM phasor plot, the metabolism trajectory of glycolysis refers in particular to the fermentation component of glycolysis from pyruvate to lactate. 40 Therefore, in healthy beta cells, due to the prohibited expression of LDH [71, 103], NADH-LDH does not contribute to the bound-NADH lifetime. Then an increase in the bound-NADH ratio, contributed by NADH binding to PDH and Complex I from OX PHOS, was observed after glucose stimulation. In the alpha cells, it was much more complicated, where there is a certain level of LDH expression and hormone-induced suppression of energy production under high glucose.[41, 104] As a result, the decrease in average NADH lifetime in alpha cells resulted from the double effect of increased fermentation and decreased OX PHOS activity. Moreover, this significant difference of the FLIM signature between alpha and beta cells provides an opportunity to directly distinguish beta cells within islets without using external dye or genetic labeling for live islet imaging. It also provides a new marker for evaluating cell function and metabolism changes in a non-invasive way. This study of Ex-4 showed that it is possible to use the FLIM signature for monitoring the metabolic regulation of drugs. Ex-4 has been reported to promote insulin secretion and beta cell metabolism through the HIF1α pathway [105-107]. HIF1α has been reported to play an important role in metabolism remodeling in cancer and other diseases, [108, 109] and it can suppress PDK activity by inducing PDH synthesis and enhance the energetically unfavorable LDH production of lactate. This finding is aligned with the results presented above that Ex-4-treated INS1E cells have a shorter NADH lifetime, indicating the upregulation of glycolysis by LDH. 41 Even though NADH autofluorescence is quite low in live tissue, it was still possible to see the mitochondria network in single cells. It was challenging to monitor mitochondria by NADH in time-lapse, since phototoxicity and photobleaching are currently inevitable. Therefore, better strategies are still being sought for tracking mitochondrial morphology and function at the same time. The evidence of increased spare respiratory capacity after glucose stimulation and decreased capacity with Ex-4 treatment may be related to the mitochondria fusion and fission process [110, 111]. Higher spare respiratory capacity may be the result of mitochondria network formation. The fusion of mitochondria could further benefit the proton synchronization across a large area of mitochondria and induce more efficient OX PHOS. This study on subcellular organelle FLIM signatures remains insufficient. To better understand exactly how different NADH-enzymes are contributing to the metabolism pathway and FLIM signature, it is still necessary to incorporate may other techniques, such as metabolomics data and modeling in multiscale. These future directions are discussed further in Chapter 5. 2.4 Method and Material 2.4.1 Cell culture Several different types of pancreatic beta cell lines as well as other types of cell lines were investigated during preliminary technique development (Table 1). 42 Table 1. Cell lines resources and cultures. Cell line Species and tissue Culture condition Insulin secretion (Y/N) Seeding density (cell/cm 2 ) Other INS1- E[112] Rat insulinoma RPMI-1640, 5%FBS, 1 mM sodium pyruvate, 2 mM glutamine, 10 mM HEPES, 0.05 mM beta-ME Yes 70,000 Pierre’s lab. Passages 54-96 INS1-E Rat insulinoma AddexBio Optimized RPMI-1640, 10%FBS, 0.05 mM beta-ME Yes 70,000 From AddexBio MIN-6[113] M. musculus insulinoma AddexBio Optimized DMEM, 15%FBS, 0.05 mM beta-ME Yes 70,000 From AddexBio 1.1B4[114] Human Pancreatic carcinoma of duct cell and islet hybrid RPMI 1640, 2 mM L- Glutamine, 10% FBS No 30,000 HEK293T Human embryonic kidneys DMEM-high glucose, 10% FBS, Pen/Strep No 20,000 Pen/Strep, 100 U/mL penicillin and 100 µg streptomycin; FBS, fetal bovine serum. Cells were routinely cultured and passed in T25 cell culture flasks at 37℃ supplemented with 5% CO2. Before imaging, cells were seeded on 8-well Chambered Coverglass with poly-L-lysine (PLL) coating to enhance cell attachment and adhesion. Treatment or imaging were applied at 70-80% cell confluency after full attachment. 2.4.2 Rodent islet isolation and dispersion Animal models were provided by Butler’s group at UCLA. The animals were anesthetized using isoflurane and sacrificed animals were used to harvest the liver and pancreas. Wild-type (WT) 43 male mice aged between 8-10 weeks pancreas were processed by injecting Liberase/DNAse through the bile duct and pancreas connect tissue detached by incubating pancreas at 37℃ for 17 minutes. Islets were further selected from tissue mixture by a minimum of three rounds of manual islet picking and then cultured in 11.1 mM glucose RPMI 1648 medium supplemented with Pen- Strep (Invitrogen) and 10% fetal bovine serum. Mouse islets were picked on ice and dispersed by Accutase in a 37℃ water bath for 5 minutes with gentle trituration every 3 minutes. Dispersed mouse islets cells were spun and resuspended in culture medium and then seeded on PLL and Laminin double-coated grid slides. Treatment and imaging were performed three days after seeding for full attachment. 2.4.3 FLIM imaging For confocal and FLIM imaging, a Leica TCS SP8 was used. Water immersion objective 40×/1.4NA was used for islets image acquisition. Water objective 40×/1.0 NA was used for dispersed islet cells, cell lines, and fixed cell image acquisition. NAD(P)H FLIM signal was collected with 740 nm 2-photon excitation and 440-500 nm emission. Excitation intensity was adjusted to yield 100 photon/pixel for z-stacks. For monolayers, while increasing the view field was attempted to capture more cells in one figure, unfortunately, it was found still to be unsuitable for FLIM imaging with insufficient photon count. So, we used 40× objectives to focus on single cells and imaged multiple positions with a spiral scanning at the beginning to keep track of cell positions. 44 2.4.4 FLIM data analysis Desirable z planes were selected and a 5×5 median filter was applied by LAXS-lifetime software. FLIM G and S coordinate data were further processed and analyzed using customized threshold setting and manual ROI in Matlab (Matlab code is provided in Appendix I). The G values were calculated by the median of points G axis distribution. 2.4.5 Seahorse INS-1E cells were seeded in Seahorse XF Cell Culture microplates 3 days prior to experiment. On the experiment day, the cells were starved in 1.1 mM Krebs–Ringer phosphate -Seahorse (KRBH- S) buffer (111 mM NaCl, 4.8 mM KCl, 1.2 mM KH2PO4, 1.2 mM MgSO4, 10 mM HEPES, 2.3 mM CaCl2) for 30 minutes. Then glucose and compounds were injected sequentially as shown in the experimental profile (Figure 13B) by Agilent Seahorse XFe 96 Analyzer. 2.4.6 Glucose stimulation assay Before glucose stimulation, cells fasted under 1.1 mM (INS-1E) or 4 mM (mouse dispersed cells) glucose Krebs–Ringer phosphate buffer (KRBH) buffer (pH 7.4, 111 mM NaCl, 25 mM NaHCO3, 4.8 mM KCl, 1.2 mM KH2PO4, 1.2 mM MgSO4, 10 mM HEPES, 2.3 mM CaCl2 and 0.1% bovine serum albumin (BSA)) for 1 hour in the microscopy incubator to reach temperature and pH balance. Glucose was added to a final concentration of 25 mM (INS-1E) or 16 mM (mouse dispersed cells) for 2 hours. 10 µM Ex-4 was added together with glucose stimulation. 45 2.4.7 Immunofluorescence The cell was fixed in 4% paraformaldehyde at room temperature for 20 minutes followed with a phosphate buffered saline rinse. Slides were soaked with 0.4% TritonX-100 for 30 minutes and blocked by 3% BSA with 0.2% Triton X-100. The primary antibodies were prepared in the cocktail and treated at 4℃ overnight. A cocktail of corresponding secondary antibody was used to detect the primary antibody. Slides were mounted with Prolong Antifade Glass with NucBlue for counterstaining and mounting. 46 Chapter 3: Glucose-stimulated metabolism in healthy vs. diabetic rodent islets 3.1 Background: metabolic reprogramming in diabetic islets 3.1.1 Human IAPP vs. mouse IAPP As mentioned in Chapter 1, human IAPP form toxic oligomers that induce ER stress and hypoxia [75]. But in mice, natural IAPP fibrils are not found. The protein sequence of mouse IAPP has a proline at position 28 instead of serine as in human IAPP which is crucial for fibril formation (Figure 19). The [Pro28] mutation of human IAPP disrupts oligomerization and fibril structure [115]. This result is partially aligned with the pathogenesis of diabetes in different species, as a mouse only develops diabetes under genetic mutation induced obesity or hyperglycemia. Rodent IAPP is soluble and nontoxic, which cannot induce ER stress and hypoxia [116]. Thus, we induced human IAPP expression in mouse beta cells to mimic human diabetes pathogenesis. Figure 19. Protein sequences of IAPP between 20 -29 amino acids in different species. Adapted from ‘Islet Amyloid Polypeptide (IAPP) Transgenic Rodents as Models for Type 2 Diabetes’ [117]. 47 3.1.2 Cell stress leads to metabolic reprogramming Butler et al. [75] showed that the glycolysis-related genes, such as HIF1α, PDK4 and LDHC, are up-regulated in the misfolded protein hIAPP T2D model. IAPP toxic oligomers induced ER stress and elevated cytosolic calcium concentration. In response to high cytosolic Ca 2+ concentration and ER stress, beta cells activated the HIF1α/PFKFB3 pathway, which is well known as the inducing factor of anaerobic metabolism and replication in tumors [118]. Glycolysis, through the LDH pathway, which is a suppressed gene in beta cells, was found to be activated in hIAPP mouse islets. The increased activity of LDH directs the metabolic flux to glycolysis via a Warburg-like effect in β cells [74, 75]. Thus, the shift to glycolysis leads to a drop in OX PHOS and ATP production in response to glucose stimulation. In Chapter 2, a high concentration of glucose enhancing OX PHOS was verified in monolayer beta cells by FLIM imaging. Distinct alpha and beta cells phasor-FLIM signatures were also observed. As a result, we hypothesize further that the same effect of glucose-stimulated OX PHOS can be seen in a whole islet. Meanwhile, in hIAPP transgenic islets, it is expected to see downregulated OX PHOS levels and an even higher ratio of glycolysis. Here, 2-photon microscopy is exploited to investigate autofluorescence deep into intact islets. 48 3.2 Results 3.2.1 Glucose enhanced beta cell oxidative phosphorylation in islets. Around 80% of cells in a mouse islet are beta cells, located mainly in the islet core [119]. Alpha cells and other types of cells are mostly located at the periphery of the islets. This well-organized morphology enables the identification of the beta cells more effectively with the FLIM signature. After high glucose stimulation, most cells shift from light blue to red (Figure 20) in the NADH lifetime map, indicating an increased ratio of OX PHOS. Beta cells, located at the central part of islets, are less OX PHOS (red) than the others under low glucose concentration. This verified the hypothesis that healthy beta cells prefer to produce ATP by OX PHOS as the most efficient way of utilizing a carbon source. Figure 20. Beta cells upregulated OX PHOS level in response to glucose stimulation. Slice view of non-diabetic rodent islet under low (4 mM, left) and high (16 mM, right) glucose. White dash, alpha cells. Purple outline, undefined cell type. Scale bar, 20 µm. 49 3.2.2 Alpha and beta cells also have the opposite response to glucose stimulation in islets Alpha cells (Figure 20, white dash) shift from green to bright blue indicating a downregulation of OX PHOS taken place under high glucose, while beta cells upregulate the OX PHOS level. This is well-aligned with our observation of dispersed islets in Chapter 2. We applied image segmentation to isolate two groups of cells to analyze the median phasor G value in 3 different islets. Due to poor resolution after a depth of 50 µm into the microtissue, we selected slices between 10-40 µm depth for segmentation. The G value of beta cells decreased in high glucose conditions, indicating a transition to a higher bound-NADH ratio and enhanced OX PHOS level (Figure 21). In alpha cells, G values shifts in the opposite direction, suggesting the suppression of OX PHOS activity under high glucose conditions. . 50 Figure 21. Alpha and beta cells phasor plot of G value in non-diabetic (ND) mouse islets. (n=3 islets for each cell Type. *P<0.5, **P<0.05, the unpaired two-tail t-test was used of all comparisons) 3.2.3 Human IAPP stress can suppress beta cell OX PHOS NADH-FLIM imaging was then applied to hIAPP transgenic mouse islets. First, the architecture of hIAPP mouse islets is quite different from the WT. The alpha cells, which are at higher OX PHOS levels under low glucose condition, make up over 15% of the cell population (Figure 22A, white dash). Additionally, alpha cells are not just located at the outer layer of the beta cell core but intrude into the beta cell network. In Figure 22B, the same islet was fixed and immunostained with glucagon and amyloid antibodies. After fixation, the cells were not the exact same shape nor at the same position, but it was clear that beta cells were present at the outer layer of the alpha cell groups (Figure 22B, yellow arrow). This observation is consistent with a previous study that the architecture of islets were disrupted in hIAPP rat, in which alpha and beta cell groups distributed irregularly [120, 121]. 51 After being treated with high glucose, beta cells shifted from yellow-green to green-blue, indicating an increased rate of glycolysis (Figure 22A). This verified the hypothesis that hIAPP expression can cause cell stress and metabolic reprogramming. Segmentation was also applied to quantify the median value of G in alpha and beta cells (Figure 23). Both alpha and beta cells showed a significant increase in G value, which indicates an increase in glycolysis level. 52 Figure 22. FLIM signature shifts to glycolysis under glucose stimulation in hIAPP mouse islets. A. Slice view of hIAPP transgenic rodent islet under low (4 mM) and high (16 mM) glucose. B. Immunostaining of whole islets. Red, glucagon. Yellow, IAPP. Yellow arrow, IAPP positive beta cells. Purple outline, undefined cell. Scale bars, 20 µm. 53 Figure 23. Phasor plot of G values for alpha and beta cells in hIAPP transgenic mouse islets. (ND, non-diabetic. WT, wild Type. n=3 islets for each cell Type. **P<0.05, the unpaired two-tail t-test was used for all comparisons). Taken together with the non-diabetic islets response, the relative OX PHOS level ∆G (= Glow – Ghigh) was further compared (Figure 24). In non-diabetic beta cells, ∆G is positive, representing increased OX PHOS after glucose treatment. In contrast, ∆G is negative in hIAPP beta cells, indicating a transition to a higher anaerobic glycolysis metabolism mechanism. Alpha cells in hIAPP mouse islets showed a slight but insignificant decreased level of glycolysis with glucose stimulation. 54 Figure 24. Relative OX PHOS level change of alpha and beta cells in non-diabetic (ND) and hIAPP transgenic mouse islets. (n=3 islets for each cell type. ns, not significant. *P<0.5, the unpaired two-tail t-test was used of all comparisons) 3.2.4 Beta cells form the “cell hub” to synchronize response of all beta cells It was also noticed that, in the WT islet, there were a few beta cells that responded first, around 30 minutes, with increased OX PHOS levels (Figure 25, outlined in white). Beta cells around these leaders responded to the metabolism shift later and more gradually before all beta cells reached a similar OX PHOS level after 2 hours. This is the first time it has been possible to monitor this “cell hub” [57] in a label free platform. However, in the hIAPP model, the transition stage of the cell hub effect was not captured as non-diabetic between 30-120 minutes. Most of the beta cells had responded to glucose stimulation by 60 minutes. Therefore, beta cells in hIAPP islets may have lost their hierarchy or they synchronize too rapidly to be captured after 30 minutes compared to the WT (Figure 25C). 55 56 Figure 25. Hub cells synchronized all beta cells OX PHOS response in healthy islets but not in hIAPP islets. A. Slice view of a time series of a non-diabetic mouse islet before and after glucose stimulation. B. Time series of the activation of the network of hub and follower cells. C. Slice view of a hIAPP mouse islets before and after glucose stimulation. Scale bars, 20 µm. White outline, hub cells. Grey outline, follower cells. Arrow, alpha cell. 3.2.5 Identification of different cell types and islet structures by FLIM and SHG From both WT and hIAPP transgenic islets, other cell types characterized by a distinct autofluorescence signature were also found. In Figure 20 and Figure 22, a cell type was found (purple line) that had an extremely short average lifetime with a deep blue color, occupying around 1% of the overall cell population. These cells have a higher autofluorescence intensity at both the NADH channel (440-500 nm) and the long-wavelength channel (560-620 nm) (Figure 26). These cells were also glucagon and IAPP negative (Figure 22B). Based on the distribution of these cells, they were mainly located in the core area of the islets. It is unlikely that these indeterminate cells are delta cells, as these have been consistently shown to be located in the periphery of the rodent islets. Figure 26. The autofluorescence signal of hIAPP islets at two channels 440-500 nm (left) and 560-620 nm (right). Scale bar, 20 µm. 57 The 3D structure of the whole islet was further investigated by combing second harmonic generation (SHG) imaging with FLIM. SHG microscopy enables us to visualize highly ordered structures, such as collagen, in live tissue. In Figure 27, lots of collagen (violet) are entangled around the core area of the islet. Most of the extreme glycolytic cells are located next to collagen, indicating a possible functional relationship between them. Figure 27. 3D projection of overlapped FLIM image and SHG signal. Scale bar, 20 µm. 3.3 Discussion This work extends the findings in Chapter 2 by investigating beta cell function in the context of whole islets, taking the islet microenvironment and cell-cell interactions into consideration. NADH-FLIM was successfully applied onto the whole islet in vitro. There, alpha and beta cells were found to shift towards higher glycolysis and OX PHOS, respectively, under high glucose conditions in the same way as the dispersed monolayer islets. 58 Moreover, in isolated islets, beta cells near the core area were not directly exposed to glucose stimulation due to a lack of vascularization. If glucose-sensing only depends on glucose concentration, we should observe a depth gradient in the activation of beta cells, starting from the exterior layer to the core. However, these results reveal that several cells in the middle layer actually maximize their energy production first, not those on the islet surface. This indicates that factors other than glucose concentration play a role in regulating glucose response in beta cell networks. Gap junction between islet cells has been reported to be crucial maintaining insulin secretion dynamics and glucose homeostasis [49]. That is to say, linked electrical activity and hormone exchange between cells are also regulating glucose response in beta cell network. Beta cell heterogeneity was recently highlighted as an important feature in maintaining islet function [44, 46, 47]. “Hub” cells were characterized as a pacemaker with preceding activity in Ca 2+ signaling. Here, a similar effect was observed where a group of beta cells exhibited an accelerated OX PHOS shift under high glucose stimulation followed by waves of the surrounding beta cells reaching a universal response throughout the islet. Hence, it is possible that this observation is the same kind of pacemaker “hub cells”. The time scale of the Ca 2+ hub effect is much shorter than the OX PHOS response. Ca 2+ increasing in the cytoplasm, induced by KATP channel closure and membrane depolarization, activates NADH malate-aspartate shuttle, which is crucial in maintaining the NAD + pool in the cytoplasm under low LDH activity in beta cells [122, 123]. Spatial diffusion of an increased ratio of bound- 59 NAD(P)H in the phasor-FLIM signature could be an indirect indicator of Ca 2+ synchronization in the beta cell network. Thus, the OX PHOS hub effect is likely to reflect the downstream Ca 2+ response and the second phase of insulin secretion. Loss of such an effect in hIAPP islets might be caused by the downregulation of malate dehydrogenase (MDH) and diminished cytosolic glucose-responsive Ca 2+ [75]. In Figure 25A, these hub beta cells seem to have higher OX PHOS levels at lower glucose conditions, revealing a possible distinct metabolic ground state. Meanwhile, Johnson et al. [57]’s study also showed more hyperpolarized mitochondria in hubs within the first 15 minutes of glucose stimulation. This finding aligns with our observation of higher OX PHOS levels in the first 30 minutes in hubs. Thus, the synchronization of the OX PHOS level in follower cells between 30- 120 minutes may represent transcriptional and translational regulation in follower cells for efficient energy production and the second phase of insulin secretion. The identification of the "hub cell" effect reveals the power of FLIM, monitoring metabolism alterations, and its ability to detect non-invasively, in 3D and in vitro, cell heterogeneity in whole islets. Another type of cell was detected with an extremely high level of glycolysis and strong autofluorescence at a longer wavelength (Figure 22 purple outline and Figure 27 blue cells). An intriguing explanation for these cells might be that they are macrophages carrying lots of lipid droplets, which contain broad-spectrum lipo-pigment [124]. The collagen that surrounds them is the supportive matrix component of the vascular membrane [125, 126] and it has been suggested that this collagen structure may represent the original blood vessel distribution before islet isolation. 60 It is possible that macrophages can invade the islets through blood vessels due to local inflammation, or due to this artifact of the islet isolation procedure. Even though this study provides an efficient and effective way to identify and characterize the metabolism of beta cell subtype, we are still unable to define the hub signature at the molecular level. Two of the big questions that remain are: how do these waves of OX PHOS activation contribute to two-phase insulin secretion? And, what leads to the failure of hub cell synchronization in diabetes development? To answer these questions, single-cell analysis, such as single-cell RNA sequencing and single-cell proteomics, should be recruited in parallel with metabolic imaging. It will help to understand what the unique gene or gene expression pattern is that determines the identity and the fate of hub beta cells. At the same time, the studies were all focused on the first 2 hours of glucose response. The influence of long-term hyperglycemia or hormone circadian rhythm was not taken into consideration. Meanwhile, despite the preservation of the mouse islets architecture and alpha-beta cell interaction, it was not possible to mimic the complete microenvironment of human beta cells. In isolated islets, most of the blood vessel structure and nerve fibers were killed. This meant the influence of hormone, oxygen, and glucose supply from blood and stimulation of nerve fibers were excluded from the investigation. Thus, human islet metabolism in vitro is investigated in Chapter 4 using FLIM imaging and these concerns are then addressed in Chapter 5. 61 3.4 Method and Material 3.4.1 Rodent islet culture and immobilization Animal models were provided by Butler’s group at UCLA. hIAPP overexpressing (HIP) male mice were generated by encompassing the hIAPP sequence under insulin promotor. WT and HIP mice aged between 8-10 weeks pancreas were processed by injecting Liberase/DNAse through the bile duct and pancreas connect tissue detached by incubating pancreas at 37℃ for 17 minutes. Islets were further selected and cultured in 11.1mM glucose RPMI 1648 medium supplemented with Pen-Strep (Invitrogen) and10% fetal bovine serum. Mouse islets were seeded on poly-L-lysine and Laminin double-coated grind slides. For whole islet imaging, islets with diameters between 80- 100um were picked. Islets were cultured for 2 days for attachment before imaging. 3.4.2 FLIM imaging and SHG imaging For confocal and FLIM imaging, a Leica TCS SP8 was used. A water immersion objective 40×/1.4NA was used for islets image acquisition. The NAD(P)H FLIM signal was collected with 740 nm 2-photon excitation and 460-500 nm emission. Other wavelength bands were optimized for detecting different species. Excitation intensity and frame repeats were adjusted and compensated to yield at least 100 photon/pixel in different z-stacks. Z step size varied from 1.5 µm to 2 µm, depending on islet size. Islet NAD(P)H FLIM data were collected in 256 × 256 pixel format at 100 Hz. All imaging was performed at 37℃, in a 5% CO2 water bath incubator to maintain islet activity. 62 SHG imaging was performed by the same instrument in sequential to FLIM imaging. The SHG signal was collected using 870 nm 2-photon excitation and 430-440 nm emission. 3.4.3 Data analysis The phasor FLIM data were analyzed in Leica LAX/FCS software and MATLAB.[127] The MATLAB code is provided in Appendix I. The raw lifetime data was filtered by 7 photon count and a median of 5. Thus, pixels under 7 photon count were filtered out in the phasor plot calculations. The lifetime of one pixel is the median value of the 5 × 5 pixels surrounding it. Then for each pixel in the image, the lifetime exponential decay was plotted in the phasor half sphere by Fourier transformation. Corresponding coordinates G and S of each figure were exported for segmentation and statistical analysis in MATLAB. The lifetime map and SHG signal of the figure was highlighted by rainbow and violet cursor on the phasor graph, representing the lifetime of each pixel. Segmentation was performed manually in 10 selected slices based on intensity, lifetime, relative position and immunostaining results. The phasor distribution, G, of the region of interest (ROI) was analyzed in MATLAB by calculating the median of all G values in the ROI pixels’ phasor cluster. ∆G = G (low glucose)-G (high glucose) was analyzed within the same islet. 63 Chapter 4: Understanding human islet metabolism in vitro 4.1 Background: Comparing human with rodent islets 4.1.1 Islet structure Human islet morphology is quite different from the highly ordered mouse islets. They form multiple smaller alpha-beta-delta cell compartments within one islet [53]. Even though human islets are still around 65% beta cells, they do not form a beta cell core in the islet center. In human islets, there are more intra-islets vessels invading the core area, and some beta cells have direct interaction with blood vessels [128, 129]. Because of the structural differences between human and mouse islets, knowledge of one cannot simply be applied to the other. Here, human islets were acquired from adult human cadavers through the Integrative Islet Distribution Program; donors were otherwise healthy or had type 2 diabetes prior to death. In some instances, samples may have required up to a few days after harvest for transportation prior to imaging, which may have also affected islet cell health and stress levels. Since the isolation procedure removed blood vessels within the human islets, their structure and function may have been altered during the experimental attachment steps (Method, 4.4.1). In Johnson et al. [57]’s research of human islets, the hub cell effect was sub-regional and compartmentalized. Therefore, it is reasonable to speculate the hub cell effect is partially determined by beta cell - beta cell interaction (i.e. between neighboring cells) and that, if there is a rearrangement of the islet structure, the cell hub effect may not be detected in FLIM imaging. 64 4.1.2 Aging and waste accumulation Aging and lipid metabolism are two significant factors in human islet development and pathogenesis. In the mouse study, samples from WT mice between 8-10 weeks old were used, which are quite young and healthy. Therefore, aging and lipid toxicity were not a consideration. However, our human islet donors, whether diabetic or not, were not young, so effects of aging and lipid metabolism must be considered. As discussed in the last chapter, human IAPP oligomer toxicity has similarities to age-related neurodegenerative diseases, such as Alzheimer's and Huntington’s disease [18, 130-133]. The accumulation of waste products in beta cells leads to continuous ER stress and HIF1α pathway activation [75, 116]. Although aging can cause beta cell dysfunction at multiple levels, such as mitochondria dysfunction and aggregated protein accumulation, aging is not the only trigger of diabetes pathogenesis. Thus, it is quite crucial to understand how aging effects may influence diabetes development and how therapies can be designed to reduce beta cell failure in the inevitable aging process. Lipofuscin is another age-related aggregate in cells, including beta cells [134-136]. The lipofuscin body is derived from autophagy, which is recruited for removing cell waste and unfunctional organelles. It remains unclear why lipofuscin is not removed by the lysosomes. Nonetheless, as it accumulates during the aging process and can be enhanced by cell stress, lipofuscin accumulation can be used as a potential indicator of cell age and cell stress levels. 65 4.1.3 Autofluorescence in live tissue Lipofuscin is also an auto-fluorescent substance in live cells. A previous study has shown the high fluorescent intensity of lipofuscin in the human islet, which may interfere and bleed through the NADH detection channel at 440-500 nm [137]. In Figure 28, lipofuscin (lipo-pigments) has a relatively high excitation (left) and emission (right) band. This may be due to the complex components in the lipofuscin body. Our 2-photon excitation wavelength at 740 nm can excite lipofuscin and NADH at the same time. The peak of lipofuscin is at 550 nm, but the emission spectrum starts from 450 – 650 nm, which means the high concentration of lipofuscin may still present a strong fluorescent intensity in the NADH detection range. Besides lipofuscin, there are many other auto-fluorescent substances shown in Figure 28. Flavins are an energy-related metabolite. The NADH signal is usually isolated from flavins by setting the detector range no larger than 500 nm. Fortunately, flavin concentration is quite low in beta cells compared with NADH. Since Flavin auto-fluorescence almost overlaps with the lipofuscin spectrum, we cannot separate their signals. 66 Figure 28. Excitation (left) and emission (right) spectrum of common auto-fluorescent substances in live tissue [138]. Reprinted with the permission of ‘NADH Autofluorescence—A Marker on its Way to Boost Bioenergetic Research’ by Einem B.V. Cytometry, 95: 34-46. Published by John Wiley and Sons. © 2018 International Society for Advancement of Cytometry. In this chapter, the NADH-FLIM signature in human pancreatic islets are investigated and strategies to analyze the NADH-FLIM signal in the presence of lipofuscin are discussed. Additionally, hyperspectral imaging is incorporated to further explore lipofuscin fluorescent characters and study the auto-fluorescent species in human islets. 4.2 Results 4.2.1 Glucose enhances human islet oxidative phosphorylation In human islets, which may be acquired by older deceased donors, there is likelihood of concurrent expression of another auto-fluorescent species - lipofuscin, which, as explained above, is related 67 to aging [135]. Lipofuscin can also be excited at 740 nm with emission peaks at 550 – 600 nm. Therefore, a strong lipofuscin signal could leak to the 400 – 500 nm channel of NAD(P)H, but can potentially be filtered out by its high intensity. In mouse islets, there is less signal expected in the 560 - 620 nm channel, since smaller quantities of lipofuscin and flavins exist in the relatively younger mouse samples (Figure 29). Figure 29. The auto-fluorescence signal of mouse islets in two channels, 440 – 500 nm (left) and 560 – 620 nm (right). Scale bar, 20 µm. In contrast, many locations with quite high intensity in the 560 - 620 nm channel are observed in human islets, which are likely lipofuscins (Figure 30). A similar distribution of these points appear in the 440 - 500 nm channel and are much stronger than signals from other areas. This verified the hypothesis that human donor islets in our study have an excess amount of lipofuscin accumulation attributable to donor age and sample stress. 68 Figure 30. The auto-fluorescence signal of human islets in two channels, 440 – 500 nm (left) and 560 – 620 nm (right). Scale bar, 20 µm. In Figure 31, which shows a 3D projection of non-diabetic human islets, there is still a color shift from blueish green to reddish-green, indicating an overall increase in islet OX PHOS level. There is a cluster of cells (blue arrow) lacking NADH signal under low glucose, dark and blurry. Under high glucose, it responded with increased NADH intensity, bright red, possibly indicating the existence of a cell hub in human islets. However, we noticed that the lipofuscin body (green) can influence the quantification of phasor distribution due to its strong intensity. If the median value of the G axis is still used to quantify phasor distribution, it is highly possible that this is a lipofuscin signature instead of NADH. Thus, better strategies are needed to analyze and quantify the FLIM- signature of islet metabolism. 69 Figure 31. Human islets 3D view of FLIM signal under low (LG, 4 mM) and high (HG, 16 mM) glucose. Scale bar, XX µm. 4.2.2 Diabetic islets oxidative phosphorylation are suppressed Due to auto-fluorescent signal interference from lipofuscin and limitations of human donor sample availability, a more qualitative analysis strategy was applied in the present human islet study. Here, the shift of phasor was measured on the phasor plot directly, as described in Method 4.4.2. Non- diabetic human islets showed enhanced OX PHOS levels in response to high glucose, while the T2D islets downregulated their OX PHOSs response, indicating a relative increase in glycolysis and metabolism pathway remodeling (Figure 32). 70 Figure 32. OX PHOS change in human nondiabetic islets and T2D islets from two patients. 4.2.3 K + channel opener restored insulin release and increased oxidative phosphorylation In Chapter 2, the role of the diabetes drug Ex-4 in metabolic pathway regulation and remodeling was investigated. As expected, Ex-4-promoted insulin secretion was independent of mitochondria- linked energy production and could upregulate LDH-linked glycolysis. Here another investigational drug, the K + channel opener, NN414[139], which has been shown restore insulin secretion in beta cells, is investigated. The hypothesis is that NN414 can regulate beta cells in the opposite way to Ex-4 and enhance OX PHOS; while Ex-4 administration causes X to happen, NN414 administration causes Y. 71 Figure 33. Slice view of human islets under 16 mM Glucose and 30 µM NN414. A. Lifetime map rainbow color-coded, including an extra violet mask which indicates high bound NAD(P)H ratio in these pixels. B. Phasor plot of the islets, with the points of longer lifetime circled in purple. Scale bar, 10 µm. In the phasor plot of Figure 33B, a second peak at a longer NADH lifetime was found that formed 40 minutes after NN414 treatment. This peak is shown in Figure 33B with a violet circle. The corresponding pixels are mainly located at the cytosolic part of the cells, excluding the lipofuscin body. This result verified the hypothesis that NN414 functions in promoting OX PHOS in human islets. 4.2.4 Autofluorescence species in human islets To better understand and analyze the FLIM data in human islets, it was necessary to investigate the character of lipofuscin. Firstly, the mitochondria membrane potential dye 72 tetramethylrhodamine methyl ester (TMRM) was used to label mitochondria and examine the spatial relationship of the lipofuscin body and mitochondrial network. In Figure 34, the cell was stimulated at 740 nm and emission was measured in three channels, 400 – 500 nm, 500 – 600 nm, and 650 – 700 nm. In the first channel, it was possible to detect both NADH and lipofuscin signals. A vague mitochondria network contributed by NADH and strong bright spots of lipofuscin were visible. In the second channel, the mitochondria network stained by TMRM is more apparent. Due to the significantly high intensity of the signal from the TMRM dye, it was difficult to distinguish the lipofuscin structure in the 500 – 600 nm or 650 – 750 nm channel. Looking at the overlapping and maximum projection of the three channels (Figure 34D and F), the entangled TMRM-mitochondria (red) and lipofuscin (green) structures are both apparent and highlight how separating the signals would be difficult. 73 Figure 34. FLIM imaging of a single beta cell stained by TMRM and stimulated at 740 nm with emission measured at three different channels. A. NADH channel 400 – 500 nm. B. TMRM channel 500 – 600 nm. C. Long wavelength channel 650 – 750 nm. D. 74 Overlap figure of 3 channels. E. Bright field image showing cell morphology. F. Maximum projection of 3 channels. Scale bar, 5 µm. Since TMRM and lipofuscin have a distinct lifetime on the phasor plot, the TMRM and lipofuscin signals were masked in red and green, respectively, as shown in the phasor plot in Figure 35. From the 3D structure view of the lifetime mask (Figure 35A), it is clear that the lipofuscin is buried in the middle of the mitochondrial network. However, it is still unclear if lipofuscin is structurally connected to the mitochondria due to experimental resolution limits. Figure 35. A. 3D projection of a single beta cell lifetime color-coding map. B. Phasor plot of 2 channels 400 – 500 nm and 500- 600 nm highlighted by green and red circle. Red, TMRM. Green, Autofluorescence. Scale bar, 5 um. Next, all the fluorescent species in human islets were further explored by hyperspectral imaging and Hyper-Spectra Phasors (HySP) analysis. In HySP we also used a Fourier transformation to denoise and unmix different fluorophore signatures based on their spectral distribution. Here, five 75 common auto-fluorescent species were labeled on the phasor plot. Strong signals could be seen aligning with the free and NADH markers (Figure 36A, yellow and red circle). This finding confirmed the presence of NADH in islet tissue. A signal was also found in the third quadrant (Figure 36, black arrow) of the phasor plot, which was contributed by lipofuscin. When zooming into an area of lipofuscin clustering, the lipofuscin signal increased as in Figure 36B. When the islets were further treated with potassium cyanide, most of the bound-NADH signal (red) was killed due to ETC failure (Figure 36C). Meanwhile, the lipofuscin signal was not affected by the ETC failure, indicating its function is independent from short-term metabolic reprogramming. 76 Figure 36. HySP phasor plot of auto-fluorescence in the human islet. A. 4 mM Glucose. B. 4 mM Glucose in higher magnification. C. KCN treated. Tracer: yellow, free NADH; red, bound NADH; green, retinol; light green, retinol acid; blue, elastin; black, lipofuscin. 4.3 Discussion This chapter undertook an investigation beta cell metabolism in human islets from human cadaveric donors. Even though an increase in overall NADH lifetime was observed in islets under 77 high glucose stimulation, additional data is needed before a full interpretation of the NADH-FLIM data will be possible. Firstly, it was noticed that lipofuscin had a much stronger intensity in the 440 – 500 nm channel and, therefore, it overrides the signal of NADH in human islets. The phasor distribution of lipofuscin was also quite close to the NADH metabolic trajectory. When the same excitation laser power was applied to human samples, pixels with lipofuscin reached more than 200 photon count/pixel. However, the NADH signal was quite low; 2-40 photon count/pixel. The low photon counts in these data may lead to inaccurate lifetime distribution on the phasor plot. Since the phasor of lipofuscin was also close to the NADH signal, it was hard to analyze the NADH-FLIM signature. As a result, it may be necessary to increase the laser power and scanning repeats to enhance the photon counts of the NADH signal. In this case, both lipofuscin and NADH signal is elevated. But with efficient and effective amount of NADH photon count, it would be more accurate to fit and unmix NADH lifetime exponential decay in FLIM. As mentioned previously, masking the FLIM image with a maximum photon count filter, such as 7- 50 counts, was attempted. While this may effectively remove the lipofuscin signal when setting the even lower maximum value, the intensity of the lipofuscin signal might not be consistent across the z stack, or even within the same figure. The lipofuscin bodies in these studies had various sizes ranging from 2 – 4 µm, which may include a cavity in the middle. Thus, the intensity of the lipofuscin body also varied depending on its content. Meanwhile, from the TMRM staining of dispersed human islets, it was clear that lipofuscin was wrapped up by the mitochondrial network. 78 However, the distance between the mitochondria and lipofuscin is still unclear at this resolution. As the pixel size in these figures cannot resolve the boundary of lipofuscin, the photon count filter may hide considerable NADH signal from adjacent mitochondria in the lipofuscin signal. Since mitochondrial NADH is the key NADH signal source, these results may underestimate OX PHOS contribution to FLIM signature changes. The main question then is, can the lipofuscin signal be separated from NADH in FLIM imaging? Or indeed, can FLIM imaging be used to monitor metabolic states in beta cells? One hypothesis is that the lipofuscin lifetime will not be altered by metabolism pathway reprogramming. Since lipofuscin is a derivative product of autophagosome and is related to the slow aging process, it can be assumed that it has a quite a stable concentration and lifetime in the short term. If this is true, it is only necessary to increase the laser power and scanning repeats in FLIM imaging to reach effective photon counts of NADH. All the shifting in the phasor plot is expected to reflect the contributions of the free versus bound NADH. This hypothesis was partially verified by incorporating hyperspectral imaging. As the lipofuscin emission peak was different from NADH, the hybrid signal could be easily separated in HySP. These results showed that lipofuscin had a distinct HySP signature from free- and bound-NADH. When potassium cyanide was applied to disrupt ETC and OX PHOS, only bound NADH signals changed in response. This demonstrated that the lipofuscin concentration is not directly affected by mitochondrial function. Previous research showed that lipofuscin formation can be accelerated by oxidative stress, but will only increase less than 5% within a week in normal culture conditions 79 [136, 140, 141]. Thus, intensity changes of lipofuscin could be temporarily ignored and it was reasonable to assume a consistent fractional contribution of lipofuscin in a phasor plot. As there is no evidence of lifetime stability of lipofuscin in biosamples, we still need to further demonstrate that the lifetime of lipofuscin will not change under glucose or other drug treatment. Since lipofuscin formation is related to oxidative stress, the downstream antioxidation defense of OX PHOS may still influence its signal. Therefore, another strategy is proposed here to address these questions in the future, which is a combination of FLIM imaging and simplified hyperspectral imaging. Four equidistant channels could be set during FLIM imaging, such as 450 – 500 nm, 500 – 550 nm, 550 – 600 nm, and 600 – 650 nm, as a simplification of the 64 channels employed in hyperspectral imaging. This way, the same concept utilized in HySp can be employed to separate the NADH and lipofuscin signals in the 450 – 500 nm channel, which corresponds with the lipofuscin signal in the longer wavelengths channel. In addition to disturbing signal of lipofuscin, due to the data quality as in Figure 30, 400 – 500nm channel, it was barely possible to characterize cell boundaries by NADH intensity in the human islets. Since heterogeneity of the basal FLIM signature and lifetime change in different regions of islets is still detectible, it should be possible to distinguish alpha and beta cells in the whole islet. All in all, better imaging strategies and higher quality data are still needed for understanding human islets metabolism. 80 4.4 Method and Material 4.4.1 Human islet culture and immobilization Human islets were cultivated in suspension in 5% CO2 in 5.5 mM glucose RPMI 1640 medium supplemented with Pen-Strep (Invitrogen) and 10% FBS. Human islets and dispersed cells were seeded on Grid-500 coverslip pre-coated with the HTB-9 matrix. 4.4.2 FLIM data analysis Here, the same imaging settings are used as in Chapter 3. The center of the phasor was picked manually on LAX FLIM/FCS software. The free and bound NADH tracer was set on the semicircle, then the shift of the phasor’s center was calculated by its distance to the free and bound NADH tracer. We set the average shift distance of non-diabetic sample as 100% and normalized other data to it. 4.4.3 Hyper spectrum imaging Hyperspectral imaging was collected using a Zeiss Confocal Laser Scanning Microscope 780 Inverted, with 2-photon excitation at 740 nm. The emission signal was collected by a 40× water objective in the HySP model with 64 channels. The z-stack was 90 slices with a 0.5 µm step size. Data was processed in HySP with a filter of 3. 81 Chapter 5: Live cell imaging and multi-source data integration for in vivo study 5.1 Integrating FLIM data with other data types In the previous chapters, how FLIM imaging was used to investigate pancreatic beta cells and islets were discussed. Though it is a powerful technique for studying live tissue metabolism non- invasively, it still has limitations in many respects. Firstly, the auto-fluorescent FLIM imaging, such as the NADH-FLIM used here, is restricted by the auto-fluorescence’s spectrum and abundance. Besides NADH, retinol, retinol acid and flavin are all common auto-fluorescent substances in live tissues [138]. In different tissues, the abundance of each species may vary. For example, retinol and retinol acid are commonly detected in liver and eye tissue, and flavin is also a proton donor on the ETC [101, 142]. Meanwhile, oxidized lipid derivatives may also have fluorescence and be related to oxidative stress [143]. Arachidonic acids are one of the lipid derivatives that have been characterized for its spectrum and dominant auto- fluorescence in liver tissue [144]. However, as Figure 28 showed, fluorescent lipid derivatives may have a very broad spectra, due to the complexity of their lipid’s types. This can be a double-edged sword. On one side, it is beneficial for evaluating the oxidative stress level and other aspects of tissue. However, on the other side, it may interfere with the auto-fluorescent target and adds complexity to the systematic analysis of tissue. As in the findings in Chapter 4, lipofuscin is a species-dependent and age-related auto-fluorescent substance that was not expected in a preliminary study. Therefore, for NADH-FLIM, it is challenging to comprehensively understand 82 how these different auto-fluorescent species contribute to readouts and to isolate the key information from this complex system. Secondly, NAD(P)H is involved in many metabolic reactions at the same time. As discussed in Chapter 2, NAD(P)H is produced mainly in glycolysis and the TCA cycle and is reduced on mitochondria ETC. Based on the proteomic analysis of pancreatic tissue and islets, dihydrodiol dehydrogenase, NADH dehydrogenase (Complex I), PDH, malate dehydrogenase, isocitrate dehydrogenase, hydroxypyruvate reductase, and GAPDH are the most abundant NAD(P)H binding proteins (Figure 37) [145]. Here it was assumed that the upregulation of LDH was the principal contributor to the metabolism reprogramming in the metabolic trajectory. However, it may be possible to miss many other moderate contributors to metabolism pathways hiding behind the FLIM signature. 83 Figure 37. NADH and NADPH involved metabolic pathways [84]. Reprinted by permission of Elsevier from ‘Investigating mitochondrial redox state using NADH and NADPH autofluorescence’ by Blacker, T.S. Free Radic Biol Med, 2016. 100: p. 53- 65. Published by Elsevier Ltd. Finally, the NADH-FLIM imaging presented herein was limited by instrument capabilities. The Leica SP8 FLACON is one of the FAST-FLIM microscopes that can acquire the whole islet FLIM image within 15 minutes. The current two-phase insulin secretion theory states that the first phase of secretion occurs in the first 15 minutes and the second phase lasts for 1 to 2 hours. Thus, in this analysis, the temporal resolution is insufficient in NADH-FLIM. On the other hand, since NADH is an intrinsic fluorophore, its intensity is limited to metabolic activity. Even though we incorporated a 2-photon excitation laser to minimize the focal volume in the samples to decrease out-of-focus photon exposure and bleaching, repeat scanning is still required to acquire the 84 effective number of photons for analysis. This approach inevitably caused photobleaching and limits the time-lapse imaging. Even though there are many limitations of FLIM imaging, it still has advantages for studying pancreatic islets and cell metabolism. As discussed in Chapter 2, it is a non-invasive, label-free imaging technique that can monitor multiple intrinsic fluorophore states and subcellular structures together. Thus, if combining this technique with other imaging techniques or data types, it may be helpful to systematically understand FLIM data and islet metabolism. The preliminary attempts here to combine FLIM imaging and SHG were successful (Section 3.2.7). The mapping of cell locations together with collagen distribution were demonstrated. In the dispersed islets, the lipofuscin and mitochondria network were localized (Section 4.2.4), however, due to the limitations of light microscopy, it was not possible to determine if two structures were connected or not. When a GAPDH inhibitor was applied to INS-1E cells together with glucose, an increase in the OX PHOS level of beta cells was observed along with the suppression of the OX PHOS response to glucose by asparagine (Section 2.2.1). The GAPDH inhibitor and asparagine were predicted by metabolomic analysis in Prof. Graham’s lab (USC), showing the ability to inhibit and enhance insulin secretion separately (Data not shown). The Pancreatic Beta Cell (PBC) Consortium proposed to integrate multiscale imaging and modeling to build the whole-cell model to understand the nanoscale cell structure and dynamics (Figure 38). Recent research has utilized correlative super-resolution (SR) and electron 85 microscopy (EM) to reveal the nanoscale cell structure in pancreatic islets [146]. Cryo-SIM was able to preserve simultaneously the frozen structure and fluorescent labeling of cell organelles so that ion scanning EM could be applied to the same sample. One of the challenges in nanoscale modeling is how to integrate different data types and link the data from various resolutions. In order to study cell dynamics, techniques that have lower cell toxicity are the most important component in a correlative study. Meanwhile, to incorporate static imaging data from two adjacent scales, it is necessary to develop an integrated experimental platform and data-driven computational modeling to fill in the gap between two distinct data types. Thus, here several correlative imaging and modeling studies are proposed as future directions for research. Figure 38. PBC Consortium whole cell model – imaging across scales and nanoscale modeling. Image courtesy of Kate White. 86 5.1.1 Cell hub prediction In Chapter 3, the cell hub effect of the OX PHOS response in the mouse islet was observed. From this, a few questions emerge that need to be answered. First, what are the defining characteristics of a hub cell? Previous research has shown that these cells have the features of metabolic and transcriptionally immature cells [56, 147, 148]. The authors investigated several genes related to maturation and revealed high glucokinase and low Pdk1 expression in cell hubs. However, these features cannot be applied as markers for targeting and predicting the hub beta cell. In this thesis, it is proposed to use single cell metabolomics and proteomics to analyze hub cell features. As the hub cells and follower cells can be identified, these cells can be isolated and subjected to single- cell metabolomics and proteomics in order to seek markers of the cells. Current laser capture microdissection can be applied following FLIM imaging after being fast frozen to maintain the spatial distribution [149, 150]. The biggest challenge in correlating these two techniques is relocating the hub cells after they are frozen for analysis in a laser capture microdissection microscope. Additional cell membrane dye and other endogenous fluorescent markers may make it easier to code and track hub cells. For example, we could integrate modeling to simulate possible morphological changes of cell membranes and the relative location of cell nuclei (Figure 39). In this way, we could capture and classify cells by the cell code to further enrich the proteomic or metabolomic samples, which can decrease the biased false-positive results from a single-cell sample. Another strategy using the laser particle to label and sort cells has been reported to overcome the difficulties in tracking and analyzing single cells in live tissues[151]. Comparing the transcriptomic, proteomic and metabolomic profiles of hub cell and follower cells, we could systematically understand the differences of metabolism regulation and gene expression in two 87 cell types. This can benefit our understanding of how to protect and preserve the hub cell function in diabetes development. Figure 39. Scheme of predicting and aligning hub cell location after morphology change caused by sample preparation. Red, alpha cell. Light green, hub beta cell. Dark green, follower beta cell. Black arrow, interaction parameter. Grey dash, distance parameter. The second question is whether the beta cell hub formation is stable or oscillating over time? Since most research has been done in favor of using isolated islets with only a few cycles of GSIS investigated, it is still unclear if the hub cells maintain their leadership all the time or if they take turns as affected by other factors [147, 148]. Hub cell identity may relate to the architecture of 88 islets and be determined by the cell-cell interactions, or it may be genetically determined during islet formation. To address this question, we still need to perform long-term imaging for tracking the hub cell function over longer periods (i.e. days). If hub cell formation is related to cell-cell interactions and hormone supply in the islet microenvironment, we may also need to perform in vivo imaging to provide blood supply and other supporting cells. In vivo imaging is discussed in the next section. In isolated islets, the analysis of the distribution of hub cells is proposed to combine FLIM images, SHG images, and 3D structure reconstitution first to learn about correlations between hub cell location and islet architecture (for example, if the hub cell forms in the core of the beta cell group, if it is adjacent to alpha cells, or if it is attached to vessel collagen). This would enable the evaluation of the influence of hormone and energy supply to the islet microenvironment on beta cell development and will further help to develop protections to hub cells against immune system attacking in diabetes mellitus [147]. 5.1.2 Metabolic pathway modeling Leben et al. [85]’s research of bound-NADH lifetime showed the possibility of mapping the subcellular location of the NADH binding enzyme. The analysis was quite limited to a few types of purified enzymes in a live cell. Some key enzymes in mitochondria, such as NADH dehydrogenase and isocitrate dehydrogenase, were not measured due to the difficulty of protein purification. However, the mitochondrial activity is quite crucial for maintaining cell viability and all the metabolic steps are tightly linked. Since it is quite hard to uncouple OX PHOS and the TCA cycle spatially and functionally, here, the integration of FLIM imaging, glucose isotope labeling, metabolomics data, phosphoproteomic data, and metabolism modeling is proposed to stimulate and predict metabolic regulation in beta cell lines. The preliminary results presented in this thesis 89 showed that the GAPDH inhibitor and asparagine can shift the FLIM signature in the direction as predicted in metabolomics and metabolism models undertaken by collaborators of Graham’s lab. This makes it possible to systematically analyze enzyme contributions to the bound-NADH FLIM signature. We could incorporate metabolomics and metabolism modeling to analyze pathway contributions to metabolism under different conditions and stress. Then the phasor FLIM signature (G and S distribution) could be integrated as additional vectors in metabolism modeling. As LDH over PDH activity was assumed to determine the relative OX PHOS level in beta cells, we could regulate the ratio of two pathways by additional drug or induced LDH production. Metabolomic analysis at multiple time points and modeling simulation could quantify the energy flux ratio into two pathways. Then it will be possible to investigate how this ratio affects the G and S axis of the FLIM signature shift, based on the percentage contribution of two enzymes. Phosphoproteomics can also contribute to understanding enzyme activity at different points in time. Thus, more quantitative and systematic analysis of metabolism in FLIM imaging will be necessary. This may require collecting as much of the signal from the low abundance species in order to avoid biased phasor plots caused by lack of photons. Since repetitive collecting could cause strong photobleaching in this case, we may only collect FLIM imaging at one time points in each sample before it causes phototoxicity during the metabolomic sample collection. 5.1.3 Mitochondria network Previous research has shown the mitochondrial network is crucial for maintaining beta cell function, and mitochondria fission could reflect cell stress conditions [110, 152, 153]. Since the FLIM imaging could be used to monitor the morphology of mitochondria structure, but only at low resolution, the integration of FLIM imaging with other techniques is proposed here to model 90 mitochondria network movement in response to glucose in beta cells and islets. In this preliminary research, a few other techniques were investigated, such as confocal imaging of TMRM stained mitochondria (Section 4.2.3). This approach revealed the complicated network in single layer cells. TMRM, as a potential dye, can show different membrane potential across the mitochondria network [154]. The limitation of its application in tissue study is that its gradient intensity across deep tissue may influence membrane potential analysis in different cells. Holotomography is another invasive and less toxic technique that can monitor the mitochondria network (Figure 40). It measures the reflection index of cell structure, which represents the density of targets. However, it is also clear that holotomography is limited to a single-layered cell studies. Figure 40. Holotomography imaging. Color coding by reflection index. Scale bar, 10 µm. 91 In super-resolution imaging, we still have to face the same question caused by endogenous fluorophore expression; can the endogenous labeling affect protein function [155]? Since the limitation of light microscopy is determined by optical properties, the integration of optical microscopy data and tomography data is proposed to simulate the mitochondria network movement. Optical microscopy data could provide the dynamic information of mitochondria backbones. Super-resolution and tomography data, including x-ray tomography and electron tomography, can provide static structure information of mitochondria at higher resolution. For example, we could investigate the lipofuscin and mitochondria interaction surface by electron tomography and study the dynamic correlation of the two structures. As we modeled the mitochondrial dynamics based on the structural information from tomography data, light imaging data can be applied as restraints in modeling and used as a final validation of nanoscale models. 5.2 In vivo study of islets is essential for understanding the microenvironment effect in diabetes This current research was all preformed on isolated islets, dispersed islet cells, and beta cell lines. In the study of monolayer cells and dispersed cells, a 3D spatial environment is lacking for single cells, as most of the cells have easy accessibility to the fluid environment. The cell may lose its polarity and identity when hormone and cell-cell interaction is missing [51, 156, 157]. In isolated islets, the cell-cell interaction was kept within islets. But during isolation preparation, Liberase would destroy the connection between islets and blood vessels when it cleans the connective tissue. As a result, in the isolated islets, only hormone contributions directly from adjacent alpha and delta cells were considered. A lack of vascularization in isolated islets also influences the oxygen supply 92 to the central of islets. In the physiological condition, the oxygen tension of islets is about 40-80 mmHg [158-160]. Without a blood oxygen supply, the core oxygen tension will drop along a gradient and cause central necrosis in the hypoxic core. In this study, we selected islets around 80- 120 µm in diameter, which may have oxygen tension from 160 mmHg at the surface to 40 mmHg near the core under 21% oxygen and 5% CO2 experimental conditions [158]. The oxygen level and selected islet size reached the minimum requirement to avoid hypoxic core formation, but it also changed the oxygen supply direction through the islets. Besides blood vessels, nerve and connective tissue are also mostly removed after isolation. Thus, next the FLIM signature of an islet in vivo is evaluated for overcoming the limitation of isolated islet. The pancreas is hiding deep in the abdomen of humans and rodents, which makes it difficult to continuously monitor its function in the way you can with a xenograft tumor under the skin of mice. Zebrafish, another model organism, is transparent and showed heterogeneous glucose regulation during development.[52, 161] This model is good for studying islet development because of its easy accessibility and genetic manipulation. However, adult zebrafish have more autofluorescent lipo-pigment expression. Thus, it is not an ideal animal model. In mouse islet studies, an abdominal imaging window has been employed by pulling the pancreas out or close the skin and setting up a temporary window for long-term imaging of the pancreas tissue.[162] Even though this technique preserved the pancreas and islet microenvironment, it carries the risk of infection of the islets after surgery and imaging. Therefore, here another strategy is proposed to study islet function in vivo – transplant isolated islets to the anterior chamber (AC) of the mouse eye[124, 137]. This technique has been developed to study islets in vivo and monitor vascularization. In addition, the eye is an immune-privileged area, in which the immune and 93 inflammatory response is locally limited. So, there is less immune rejection happening after transplantation. As a result, it also benefits the study of human islets in the mouse model, as minor immune rejections will not cause severe cell death after transplantation. Meanwhile, islets in the eye chamber have been used as “reporter islets” for obesity and diabetes research, for it has similar blood glucose and hormone fluctuation as the islets in the pancreas. In future research, we can apply FLIM imaging to islets in the AC and investigate the FLIM signature change over long-term treatment. Here, it will be necessary to employ an upright Leica DIVE microscope for better accessibility to the mouse eye. Firstly, how healthy and diabetic islets respond to long-term hyperglycemia can be investigated in a mouse (Figure 41A). By transplanting WT and hIAPP mouse islets to the hIAPP mouse, we expect islets can be exposed to long-term high blood glucose. A week after transplantation, vascularization will occur. The hypothesis proposed here is that long-term hyperglycemia can gradually suppress glucose-stimulated OX PHOS response in both WT and hIAPP islets. The mouse will fast overnight and will be stimulated with a glucose injection to monitor its FLIM signature shift over the first 2 hours. One possible outcome for the hIAPP islets is severe beta cell loss due to amyloid stress of hIAPP. The suppression of the OX PHOS shift in WT islets may be less severe than hIAPP islets since there is no amyloid stress in WT islets to work as a combined factor with hyperglycemia. At the same time, evaluating the effect of drugs on diabetic islets would be of interest. One of the most interesting drugs is the SGLT2 inhibitor as it aims to lower glucose levels without interfering with insulin and hormone metabolism in the body. The SGLT2 inhibitor can decrease blood 94 glucose levels by inhibiting reabsorption of sugar in the kidney. Although there are no effective ways to regenerate functional beta cells after diabetic pathogeneses, is it possible to slow down beta cell loss and maintain or restore part of beta cell stress by only lowering blood glucose? The hypothesis of this research is that restraint of sustained hyperglycemia by SGLT2 inhibition in the setting of beta cells stress delays loss of glucose-induced insulin secretion by the preservation of OX PHOS in response to an increase in glucose. The glucose-stimulated OX PHOS response in both WT and hIAPP islets will be monitored in the hyperglycemia mouse model. Then we will further deplete hyperglycemia with SGLT2 inhibitor to mimic the treatment of Jardiance in T2D patients (Figure 41B). In both groups, the expectation is to see diminished suppression of OX PHOS after SGLT2 inhibitor treatment, compared with the hyperglycemia condition in the first 3 weeks and the no treatment controls in Figure 41a. One possible outcome of hIAPP islets is slower cell mass loss and restored beta cell function. Figure 41. Longitudinal hyperglycemia timeline scheme. After overnight fasting, FLIM imaging was conducted right after glucose injection every week. (a) 12 weeks of timecourse imaging over the same islets under a continuous control diet. (b) SGLT2 inhibitor treatment was added after 4 weeks of imaging. In these in vivo FLIM imaging experiments, we can better evaluate islet function and vascularization at the same time. It will make it possible to mimic glucose metabolism in the … Transplantation * * * * * * * * * * * * * Week 1 Week 12 Week 16 … Transplantation * * * * * * * * * * * * * SGLT2 inhibitor treatment Week 12 Week 16 a. b. Standard diet Overnight fasting and FLIM timecourse imaging * 95 whole-body system for understanding diabetes pathogenesis systematically. We can further use human islets, diabetic and non-diabetic, in the same experimental setting. It will also facilitate drug evaluation on its effect on islets function, not just blood glucose level. 5.3 Islet FLIM signature in high throughput drug screening platform As discussed in Section 5.2, the FLIM signature of mouse “reporter islets”, which can be transplanted to the eye chamber, may be used to evaluate diabetes drugs and their effect on islet function. Some of the diabetes drugs, such as the SGLT2 inhibitor and insulin, are not directly acting on beta cells or islets. As a result, even though they could lower blood glucose levels or increase insulin sensitivity, we are not quite clear about their long-term effect on islets. Since the FLIM signature of islets reflects the relative healthiness of beta cells and alpha cells, it is possible to apply this as an evaluation index during drug screening. For drugs that act directly on beta cells or islets, such as GLP-1R agonists, it could be possible to perform in vitro screening in isolated islets first to evaluate drug effect before moving to a long- term mouse study. By immobilizing isolated islets on to 48-96 well plates, we could rapidly monitor the FLIM signature of these islets under various drug treatments with multiple repeats. To fulfill this platform, we may need to optimize the microscope settings and imaging strategies to collect the FLIM signals quickly and reliably. Then we could rank the drug effects by its ability to maintain/promote OX PHOS level in islets. 96 These drug evaluation platforms, in vitro and in vivo, can help us better understand diabetes drugs and their function systematically. It can also benefit us to develop a new treatment to control diabetes and preserve islet function during disease development. Moreover, it encourages us to seek new drugs and treatments to restore islets function and cure diabetes. 5.4 Experiment Summary In this thesis, many advanced technologies and strategies, including FLIM imaging, have been applied to the primary pancreatic beta cells and islets study for the first time. While these techniques have been successfully applied to many other fields of studies, it must be noticed that different tissue samples may have distinct and unique metabolism characters due to their corresponding function. Meanwhile, cell lines are also quite different from its corresponding tissue type, while they are sharing the similar function. These differences can affect the final readout of the assays. As a result, even though previous studies in other tissues might be informative enough, preliminary studies of corresponding cell lines and cultured tissue are essential and valuable before in vivo application. For example, glucose metabolism in pancreatic beta cells is unique for its insulin secretion function. The low activity of hexokinase and lactate dehydrogenase leads to enhanced oxidative phosphorylation in mitochondria and effective energy production [163]. In contrast, stem cells and cancer cells prefer lactate dehydrogenase guiding anaerobic glycolysis for glucose metabolism. Thus, after glucose stimulation, the FLIM signature of neuron stem cells moves to the glycolysis direction [90]. The distinct response of beta cells encouraged us to further investigate the detailed 97 and unique metabolism pathway in two cell types. By combining metabolomic and proteomic analysis, PDH, LDH and Complex I were targeted as the key modulators of NADH-FLIM signature in beta cells. The study of beta cell lines can be misleading, as cell lines were derived from cancer cells or stem cells to maintain replication character. In Chapter 2, the INS-1E cell lines showed high glycolysis activity in the nucleus, aligning with its active pentose phosphate pathway for replication (Figure 11C). Thus, in the study of primary tissue samples, it’s crucial to analyze preliminary data from a dialectic view and integrate different omics data for comprehensive understanding. Besides the preliminary test of cell lines, initial adjustment of imaging parameters is also challenging in applying FLIM imaging to primary cells, especially for endogenous fluorescent markers like NADH. Fast data acquisition speed, high image resolution, and low photobleaching and phototoxicity are mutually exclusive. Even though higher image resolution could present subcellular change in the whole islet tissue, it would cost more time to acquire the 3D structure. It is not desirable to capture metabolism change happening within a few minutes. As a result, in the study for glucose-stimulated insulin secretion, the subcellular resolution was sacrificed to be able to image the whole islet in 15 minutes. At the same time, the time intervals are limited to reduce laser exposure caused by phototoxicity. Even though it would miss the intermediate transition state in between, the cell healthiness and function would be well preserved. 98 5.5 Conclusions By applying FLIM imaging to investigate the change of pancreatic islet metabolism in vitro, this thesis has shown the capability of non-invasively monitoring pathological metabolism reprogramming in diabetes. It can be concluded that the NADH-FLIM signature reflects the aerobatic glycolysis and OX PHOS activities in pancreatic islets and can be used as label-free aids to identify specific types of cells. The research illustrated the heterogeneity of beta cells in response to glucose, which provides new strategies to target and monitor hub beta cells in islets. The results of human islets FLIM imaging indicate the potential application to clinical diagnose and pharmaceutical development. As FLIM imaging provides a non-invasive and label-free platform to characterize islet metabolism in vitro, further application into in vivo research is needed to determine islets function and long- term diabetes pathological development under the physiological environment. Meanwhile, the technique showed great potential in multiscale data integration and modeling as it reflects both temporal and spatial metabolites response in islets systematically. While it is still challenging to cure diabetes, the research provides new insight into diabetes prevention and treatment in the near future. 99 References 1. Kharroubi, A.T. and H.M. Darwish, Diabetes mellitus: The epidemic of the century. World journal of diabetes, 2015. 6(6): p. 850-867. 2. Vijan, S., In the clinic. Type 2 diabetes. Ann Intern Med, 2015. 162(5): p. ITC1-16. 3. King, G.L. and M. Brownlee, The cellular and molecular mechanisms of diabetic complications. Endocrinology and Metabolism Clinics, 1996. 25(2): p. 255-270. 4. Seissler, J., et al., Immunological heterogeneity in type I diabetes: presence of distinct autoantibody patterns in patients with acute onset and slowly progressive disease. Diabetologia, 1998. 41(8): p. 891-897. 5. Atkinson, M.A. and G.S. Eisenbarth, Type 1 diabetes: new perspectives on disease pathogenesis and treatment. The Lancet, 2001. 358(9277): p. 221-229. 6. LaGasse, J.M., et al., Successful prospective prediction of type 1 diabetes in schoolchildren through multiple defined autoantibodies: an 8-year follow-up of the Washington State Diabetes Prediction Study. Diabetes care, 2002. 25(3): p. 505-511. 7. Imagawa, A., et al., A novel subtype of type 1 diabetes mellitus characterized by a rapid onset and an absence of diabetes-related antibodies. New England journal of medicine, 2000. 342(5): p. 301- 307. 8. Nishida, Y., et al., Antibody-validated proteins in inflamed islets of fulminant type 1 diabetes profiled by laser-capture microdissection followed by mass spectrometry. PLoS One, 2014. 9(10): p. e107664. 9. Imagawa, A., et al., A proposal of three distinct subtypes of type 1 diabetes mellitus based on clinical and pathological evidence. Annals of Medicine, 2000. 32(8): p. 539-543. 10. Bekris, L.M., T.J. Kavanagh, and A. Lernmark, Targeting type 1 diabetes before and at the clinical onset of disease. Endocr Metab Immune Disord Drug Targets, 2006. 6(1): p. 103-24. 11. Knip, M., et al., Environmental triggers and determinants of type 1 diabetes. Diabetes, 2005. 54 Suppl 2: p. S125-S136. 12. Shapiro, A.J., et al., Islet transplantation in seven patients with type 1 diabetes mellitus using a glucocorticoid-free immunosuppressive regimen. New England Journal of Medicine, 2000. 343(4): p. 230-238. 13. Swisa, A., B. Glaser, and Y. Dor, Metabolic Stress and Compromised Identity of Pancreatic Beta Cells. Front Genet, 2017. 8: p. 21. 14. Bensellam, M., J.C. Jonas, and D.R. Laybutt, Mechanisms of β-cell dedifferentiation in diabetes: recent findings and future research directions. J Endocrinol, 2018. 236(2): p. R109-R143. 100 15. Eldin, W.S., M. Emara, and A. Shoker, Prediabetes: a must to recognise disease state. International Journal of Clinical Practice, 2008. 62(4): p. 642-648. 16. Sangrós, F.J., et al., Association of General and Abdominal Obesity With Hypertension, Dyslipidemia and Prediabetes in the PREDAPS Study. Rev Esp Cardiol (Engl Ed), 2018. 71(3): p. 170-177. 17. Hayden, M.R., et al., Type 2 diabetes mellitus as a conformational disease. JOP, 2005. 6(4): p. 287- 302. 18. Bharadwaj, P., et al., The Link between Type 2 Diabetes and Neurodegeneration: Roles for Amyloid-β, Amylin, and Tau Proteins. J Alzheimers Dis, 2017. 59(2): p. 421-432. 19. Janciauskiene, S. and B. Ahrén, Fibrillar islet amyloid polypeptide differentially affects oxidative mechanisms and lipoprotein uptake in correlation with cytotoxicity in two insulin-producing cell lines. Biochem Biophys Res Commun, 2000. 267(2): p. 619-25. 20. Astorga, G. and J. Bedoya, [A comparison of fenbufen and acetyl-salicylic acid in the treatment of osteoarthritis (author's transl)]. Rev Med Chil, 1977. 105(6): p. 371-4. 21. Hohmeier, H.E., et al., Inflammatory mechanisms in diabetes: lessons from the beta-cell. Int J Obes Relat Metab Disord, 2003. 27 Suppl 3: p. S12-6. 22. Donath, M.Y. and S.E. Shoelson, Type 2 diabetes as an inflammatory disease. Nat Rev Immunol, 2011. 11(2): p. 98-107. 23. Sarles, H., Chronic pancreatitis and diabetes. (0950-351X (Print)). 24. Sandler, S., A. Andersson, and C. Hellerström, Inhibitory effects of interleukin 1 on insulin secretion, insulin biosynthesis, and oxidative metabolism of isolated rat pancreatic islets. Endocrinology, 1987. 121(4): p. 1424-31. 25. Aljada, A., et al., Tumor necrosis factor-[alpha] inhibits insulin-induced increase in endothelial nitric oxide synthase and reduces insulin receptor content and phosphorylation in human aortic endothelial cells. Metabolism-Clinical and Experimental, 2002. 51(4): p. 487-491. 26. O'Neill, C.M., et al., Circulating levels of IL-1B+ IL-6 cause ER stress and dysfunction in islets from prediabetic male mice. Endocrinology, 2013. 154(9): p. 3077-3088. 27. Janikiewicz, J., et al., Islet β-cell failure in type 2 diabetes--Within the network of toxic lipids. Biochem Biophys Res Commun, 2015. 460(3): p. 491-6. 28. Ježek, P., et al., Fatty Acid-Stimulated Insulin Secretion vs. Lipotoxicity. Molecules, 2018. 23(6). 29. Hou, Z.Q., et al., Involvement of chronic stresses in rat islet and INS-1 cell glucotoxicity induced by intermittent high glucose. Mol Cell Endocrinol, 2008. 291(1-2): p. 71-8. 30. Li, N., M. Karaca, and P. Maechler, Upregulation of UCP2 in beta-cells confers partial protection against both oxidative stress and glucotoxicity. Redox Biol, 2017. 13: p. 541-549. 101 31. Leighton, B. and G.J. Cooper, Pancreatic amylin and calcitonin gene-related peptide cause resistance to insulin in skeletal muscle in vitro. Nature, 1988. 335(6191): p. 632-5. 32. Tfayli, H., et al., Phenotypic type 2 diabetes in obese youth: insulin sensitivity and secretion in islet cell antibody-negative versus -positive patients. Diabetes, 2009. 58(3): p. 738-44. 33. Ramirez, R., et al., Effects of high-carbohydrate or high-fat diet on carbohydrate metabolism and insulin secretion in the normal rat. Diabetes Res, 1990. 15(4): p. 179-83. 34. Abbasi, A., et al., Role of HDL cholesterol and estimates of HDL particle composition in future development of type 2 diabetes in the general population: the PREVEND study. The Journal of Clinical Endocrinology & Metabolism, 2013. 98(8): p. E1352-E1359. 35. Takahashi, K., et al., Impaired oxidative endoplasmic reticulum stress response caused by deficiency of thyroid hormone receptor α. J Biol Chem, 2014. 289(18): p. 12485-93. 36. Hatanaka, M., et al., Chronic high fat feeding restricts islet mRNA translation initiation independently of ER stress via DNA damage and p53 activation. Sci Rep, 2017. 7(1): p. 3758. 37. Lu, B., et al., Impaired β-cell glucokinase as an underlying mechanism in diet-induced diabetes. Dis Model Mech, 2018. 11(6). 38. Woodland, D.C., et al., Short-term high-fat feeding induces islet macrophage infiltration and β-cell replication independently of insulin resistance in mice. Am J Physiol Endocrinol Metab, 2016. 311(4): p. E763-E771. 39. Li, G., et al., Multifunctional in vivo imaging of pancreatic islets during diabetes development. J Cell Sci, 2016. 129(14): p. 2865-75. 40. Jones, B.J., T. Tan, and S.R. Bloom, Minireview: Glucagon in stress and energy homeostasis. Endocrinology, 2012. 153(3): p. 1049-1054. 41. Allister, E.M., et al., UCP2 Regulates the Glucagon Response to Fasting and Starvation. Diabetes, 2013. 62(5): p. 1623. 42. Christensen, A.A. and M. Gannon, The Beta Cell in Type 2 Diabetes. Curr Diab Rep, 2019. 19(9): p. 81. 43. Brereton, M.F., et al., Alpha-, Delta- and PP-cells: Are They the Architectural Cornerstones of Islet Structure and Co-ordination? The journal of histochemistry and cytochemistry : official journal of the Histochemistry Society, 2015. 63(8): p. 575-591. 44. Bedoya, F.J., et al., Radiometric oil well assay for glucokinase in microscopic structures. Anal Biochem, 1985. 144(2): p. 504-13. 45. Yang, C., F. Galivo, and C. Dorrell, Is a β cell a β cell? Curr Opin Endocrinol Diabetes Obes, 2017. 24(2): p. 92-97. 102 46. Gutierrez, G.D., J. Gromada, and L. Sussel, Heterogeneity of the pancreatic beta cell. Frontiers in genetics, 2017. 8: p. 22. 47. Avrahami, D., et al., Beta cell heterogeneity: an evolving concept. Diabetologia, 2017. 60(8): p. 1363-1369. 48. Da Silva Xavier, G. and G.A. Rutter, Metabolic and Functional Heterogeneity in Pancreatic β Cells. J Mol Biol, 2019. 49. Dwulet, J.M., et al., How Heterogeneity in Glucokinase and Gap-Junction Coupling Determines the Islet [Ca. Biophys J, 2019. 50. Van Schravendijk, C., R. Kiekens, and D. Pipeleers, Pancreatic beta cell heterogeneity in glucose- induced insulin secretion. Journal of Biological Chemistry, 1992. 267(30): p. 21344-21348. 51. van der Meulen, T., et al., Virgin Beta Cells Persist throughout Life at a Neogenic Niche within Pancreatic Islets. Cell Metab, 2017. 25(4): p. 911-926.e6. 52. Zhao, J., et al., In vivo imaging of β-cell function reveals glucose-mediated heterogeneity of β-cell functional development. eLife, 2019. 8: p. e41540. 53. Arrojo e Drigo, R., et al., New insights into the architecture of the islet of Langerhans: a focused cross-species assessment. Diabetologia, 2015. 58(10): p. 2218-2228. 54. Burg, A.R. and H.M. Tse, Redox-Sensitive Innate Immune Pathways During Macrophage Activation in Type 1 Diabetes. Antioxid Redox Signal, 2018. 29(14): p. 1373-1398. 55. Jain, R. and E. Lammert, Cell-cell interactions in the endocrine pancreas. Diabetes Obes Metab, 2009. 11 Suppl 4: p. 159-67. 56. Johnston, N.R., et al., Beta Cell Hubs Dictate Pancreatic Islet Responses to Glucose. Cell Metab, 2016. 24(3): p. 389-401. 57. Johnston, N.R., et al., Beta Cell Hubs Dictate Pancreatic Islet Responses to Glucose. Cell Metab, 2016. 24(3): p. 389-401. 58. Matschinsky, F.M. and D.F. Wilson, The Central Role of Glucokinase in Glucose Homeostasis: A Perspective 50 Years After Demonstrating the Presence of the Enzyme in Islets of Langerhans. Front Physiol, 2019. 10: p. 148. 59. Bedoya, F.J., et al., The glucokinase glucose sensor in human pancreatic islet tissue. Diabetes, 1986. 35(1): p. 61-7. 60. Bedoya, F.J., J.C. Oberholtzer, and F.M. Matschinsky, Glucokinase in B-cell-depleted islets of Langerhans. J Histochem Cytochem, 1987. 35(10): p. 1089-93. 61. Johnson, D., et al., Glucose-Dependent Modulation of Insulin Secretion and Intracellular Calcium Ions by GKA50, a Glucokinase Activator. Diabetes, 2007. 56(6): p. 1694. 103 62. Hou, J.C., L. Min, and J.E. Pessin. 63. Nesher, R. and E. Cerasi, Modeling phasic insulin release: immediate and time-dependent effects of glucose. Diabetes, 2002. 51(suppl 1): p. S53-S59. 64. Bratanova-Tochkova, T.K., et al., Triggering and Augmentation Mechanisms, Granule Pools, and Biphasic Insulin Secretion. Diabetes, 2002. 51(suppl 1): p. S83. 65. Hao, M., et al., Regulation of two insulin granule populations within the reserve pool by distinct calcium sources. Journal of Cell Science, 2005. 118(24): p. 5873. 66. Kalwat, M.A. and M.H. Cobb, Mechanisms of the amplifying pathway of insulin secretion in the β cell. Pharmacology & Therapeutics, 2017. 179: p. 17-30. 67. Ashcroft, S.J., L.C. Weerasinghe, and P.J. Randle, Interrelationship of islet metabolism, adenosine triphosphate content and insulin release. Biochem J, 1973. 132(2): p. 223-31. 68. Sener, A. and W.J. Malaisse, Stimulation by D-glucose of mitochondrial oxidative events in islet cells. Biochem J, 1987. 246(1): p. 89-95. 69. Jijakli, H., et al., Relevance of lactate dehydrogenase activity to the control of oxidative glycolysis in pancreatic islet B-cells. Arch Biochem Biophys, 1996. 327(2): p. 260-4. 70. Tamarit-Rodriguez, J., et al., Lactate production in pancreatic islets. Diabetes, 1998. 47(8): p. 1219- 23. 71. Ainscow, E.K., C. Zhao, and G.A. Rutter, Acute overexpression of lactate dehydrogenase-A perturbs beta-cell mitochondrial metabolism and insulin secretion. Diabetes, 2000. 49(7): p. 1149- 55. 72. Prentki, M., F.M. Matschinsky, and S.M. Madiraju, Metabolic signaling in fuel-induced insulin secretion. Cell metabolism, 2013. 18(2): p. 162-185. 73. Wu, J., et al., Sources and implications of NADH/NAD(+) redox imbalance in diabetes and its complications. Diabetes Metab Syndr Obes, 2016. 9: p. 145-53. 74. Cantley, J. and T.J. Biden, Sweet and sour β-cells: ROS and Hif1α induce Warburg-like lactate production during type 2 diabetes. Diabetes, 2013. 62(6): p. 1823-1825. 75. Montemurro, C., et al., IAPP toxicity activates HIF1α/PFKFB3 signaling delaying β-cell loss at the expense of β-cell function. Nat Commun, 2019. 10(1): p. 2679. 76. Lytrivi, M., M. Igoillo-Esteve, and M. Cnop, Inflammatory stress in islet β-cells: therapeutic implications for type 2 diabetes? Curr Opin Pharmacol, 2018. 43: p. 40-45. 77. Dysli, C., et al., Fluorescence lifetime imaging ophthalmoscopy. Progress in Retinal and Eye Research, 2017. 60: p. 120-143. 104 78. Taylor, D.L. and Y.-l. Wang, Fluorescence Microscopy of Living Cells in Culture, Part B: Quantitative Fluorescence Microscopy-Imaging and Spectroscopy. Vol. 30. 1989: Academic Press. 79. Becker, W., Fluorescence lifetime imaging by multi-dimensional time correlated single photon counting. Medical Photonics, 2015. 27: p. 41-61. 80. Liao, S.-C., Y. Sun, and U. Coskun, FLIM analysis using the phasor plots. 2015. 81. Stephens, D.J. and V.J. Allan, Light Microscopy Techniques for Live Cell Imaging. Science, 2003. 300(5616): p. 82. 82. So, P.T., et al., Two-photon excitation fluorescence microscopy. Annual review of biomedical engineering, 2000. 2(1): p. 399-429. 83. Hage, C.H., et al., A readily usable two-photon fluorescence lifetime microendoscope. J Biophotonics, 2019. 12(5): p. e201800276. 84. Blacker, T.S. and M.R. Duchen, Investigating mitochondrial redox state using NADH and NADPH autofluorescence. Free Radic Biol Med, 2016. 100: p. 53-65. 85. Leben, R., et al., Systematic Enzyme Mapping of Cellular Metabolism by Phasor-Analyzed Label- Free NAD(P)H Fluorescence Lifetime Imaging. Int J Mol Sci, 2019. 20(22). 86. Lakowicz, J.R., et al., Fluorescence lifetime imaging of free and protein-bound NADH. Proc Natl Acad Sci U S A, 1992. 89(4): p. 1271-5. 87. Skala, M.C., et al., In vivo multiphoton fluorescence lifetime imaging of protein-bound and free nicotinamide adenine dinucleotide in normal and precancerous epithelia. J Biomed Opt, 2007. 12(2): p. 024014. 88. Sharick, J.T., et al., Protein-bound NAD(P)H Lifetime is Sensitive to Multiple Fates of Glucose Carbon. Sci Rep, 2018. 8(1): p. 5456. 89. Stringari, C., et al., Metabolic trajectory of cellular differentiation in small intestine by Phasor Fluorescence Lifetime Microscopy of NADH. Sci Rep, 2012. 2: p. 568. 90. Stringari, C., et al., Phasor fluorescence lifetime microscopy of free and protein-bound NADH reveals neural stem cell differentiation potential. PLoS One, 2012. 7(11): p. e48014. 91. Ranjit, S., et al., Determination of the metabolic index using the fluorescence lifetime of free and bound nicotinamide adenine dinucleotide using the phasor approach. J Biophotonics, 2019. 12(11): p. e201900156. 92. Ranjit, S., et al., Multicomponent Analysis of Phasor Plot in a Single Pixel to Calculate Changes of Metabolic Trajectory in Biological Systems. J Phys Chem A, 2019. 123(45): p. 9865-9873. 93. Gómez, C.A., et al., Cerebral metabolism in a mouse model of Alzheimer's disease characterized by two-photon fluorescence lifetime microscopy of intrinsic NADH. Neurophotonics, 2018. 5(4): p. 045008. 105 94. Chacko, J.V. and K.W. Eliceiri, NAD(P)H fluorescence lifetime measurements in fixed biological tissues. Methods Appl Fluoresc, 2019. 7(4): p. 044005. 95. Bird, D.K., et al., Metabolic mapping of MCF10A human breast cells via multiphoton fluorescence lifetime imaging of the coenzyme NADH. Cancer Res, 2005. 65(19): p. 8766-73. 96. Drozdowicz-Tomsia, K., et al., Multiphoton fluorescence lifetime imaging microscopy reveals free-to-bound NADH ratio changes associated with metabolic inhibition. J Biomed Opt, 2014. 19(8): p. 086016. 97. Blacker, T.S., et al., Metabolic Profiling of Live Cancer Tissues Using NAD(P)H Fluorescence Lifetime Imaging. Methods Mol Biol, 2019. 1928: p. 365-387. 98. Provenzano, P.P., K.W. Eliceiri, and P.J. Keely, Multiphoton microscopy and fluorescence lifetime imaging microscopy (FLIM) to monitor metastasis and the tumor microenvironment. Clin Exp Metastasis, 2009. 26(4): p. 357-70. 99. Stringari, C., et al., Label-free separation of human embryonic stem cells and their differentiating progenies by phasor fluorescence lifetime microscopy. J Biomed Opt, 2012. 17(4): p. 046012. 100. Meleshina, A.V., et al., Probing metabolic states of differentiating stem cells using two-photon FLIM. Sci Rep, 2016. 6: p. 21853. 101. Meleshina, A.V., et al., Two-photon FLIM of NAD(P)H and FAD in mesenchymal stem cells undergoing either osteogenic or chondrogenic differentiation. Stem Cell Res Ther, 2017. 8(1): p. 15. 102. Corbet, C., Stem Cell Metabolism in Cancer and Healthy Tissues: Pyruvate in the Limelight. Frontiers in Pharmacology, 2018. 8(958). 103. Wang, S.F., et al., 7-Ketocholesterol induces P-glycoprotein through PI3K/mTOR signaling in hepatoma cells. Biochem Pharmacol, 2013. 86(4): p. 548-60. 104. Knudsen, J.G., et al., Dysregulation of Glucagon Secretion by Hyperglycemia-Induced Sodium- Dependent Reduction of ATP Production. Cell Metabolism, 2019. 29(2): p. 430-442.e4. 105. Huang, C., L. Yuan, and S. Cao, Endogenous GLP-1 as a key self-defense molecule against lipotoxicity in pancreatic islets. Int J Mol Med, 2015. 36(1): p. 173-85. 106. Malaisse, W.J., Regulation, perturbation, and correction of metabolic events in pancreatic islets. Acta Diabetol, 1996. 33(3): p. 173-9. 107. Carlessi, R., et al., GLP-1 receptor signalling promotes β-cell glucose metabolism via mTOR- dependent HIF-1α activation. Scientific reports, 2017. 7(1): p. 1-13. 108. Kim, J.-w., et al., HIF-1-mediated expression of pyruvate dehydrogenase kinase: A metabolic switch required for cellular adaptation to hypoxia. Cell Metabolism, 2006. 3(3): p. 177-185. 106 109. Yeung, S., J. Pan, and M.-H. Lee, Roles of p53, MYC and HIF-1 in regulating glycolysis—the seventh hallmark of cancer. Cellular and Molecular Life Sciences, 2008. 65(24): p. 3981. 110. Westermann, B., Bioenergetic role of mitochondrial fusion and fission. Biochimica et Biophysica Acta (BBA) - Bioenergetics, 2012. 1817(10): p. 1833-1838. 111. Torres, G., et al., Glucagon-like peptide-1 inhibits vascular smooth muscle cell dedifferentiation through mitochondrial dynamics regulation. Biochemical Pharmacology, 2016. 104: p. 52-61. 112. Merglen, A., et al., Glucose sensitivity and metabolism-secretion coupling studied during two-year continuous culture in INS-1E insulinoma cells. Endocrinology, 2004. 145(2): p. 667-78. 113. Ishihara, H., et al., Pancreatic beta cell line MIN6 exhibits characteristics of glucose metabolism and glucose-stimulated insulin secretion similar to those of normal islets. Diabetologia, 1993. 36(11): p. 1139-45. 114. Vasu, S., et al., Mechanisms of toxicity by proinflammatory cytokines in a novel human pancreatic beta cell line, 1.1B4. Biochim Biophys Acta, 2014. 1840(1): p. 136-45. 115. Westermark, P., et al., Islet amyloid polypeptide: pinpointing amino acid residues linked to amyloid fibril formation. Proceedings of the National Academy of Sciences, 1990. 87(13): p. 5036-5040. 116. Press, M., et al., Protein aggregates and proteostasis in aging: Amylin and β-cell function. Mech Ageing Dev, 2019. 177: p. 46-54. 117. Matveyenko, A.V. and P.C. Butler, Islet amyloid polypeptide (IAPP) transgenic rodents as models for type 2 diabetes. ILAR journal, 2006. 47(3): p. 225-233. 118. Ferrer, C.M., et al., O-GlcNAcylation regulates cancer metabolism and survival stress signaling via regulation of the HIF-1 pathway. Molecular cell, 2014. 54(5): p. 820-831. 119. Arrojo e Drigo, R., et al., New insights into the architecture of the islet of Langerhans: a focused cross-species assessment. Diabetologia, 2015. 58(10): p. 2218-28. 120. Butler, A.E., et al., β-cell deficit and increased β-cell apoptosis in humans with type 2 diabetes. Diabetes, 2003. 52(1): p. 102-110. 121. Butler, A.E., et al., Diabetes due to a progressive defect in β-cell mass in rats transgenic for human islet amyloid polypeptide (HIP rat): a new model for type 2 diabetes. Diabetes, 2004. 53(6): p. 1509-1516. 122. Tan, C., et al., Role of NADH shuttles in glucose-induced insulin secretion from fetal beta-cells. Diabetes, 2002. 51(10): p. 2989-96. 123. Bender, K., et al., The importance of redox shuttles to pancreatic beta-cell energy metabolism and function. Biochem Soc Trans, 2006. 34(Pt 5): p. 811-4. 124. Li, G., et al., Multifunctional in vivo imaging of pancreatic islets during diabetes development. J Cell Sci, 2016. 129(14): p. 2865-2875. 107 125. Ko, J.H., et al., Collagen esterification enhances the function and survival of pancreatic β cells in 2D and 3D culture systems. Biochem Biophys Res Commun, 2015. 463(4): p. 1084-90. 126. Llacua, L.A., B.J. de Haan, and P. de Vos, Laminin and collagen IV inclusion in immunoisolating microcapsules reduces cytokine‐mediated cell death in human pancreatic islets. Journal of tissue engineering and regenerative medicine, 2018. 12(2): p. 460-467. 127. Alvarez, L.A., et al., SP8 FALCON: a novel concept in fluorescence lifetime imaging enabling video-rate confocal FLIM. 2019, NATURE PUBLISHING GROUP MACMILLAN BUILDING, 4 CRINAN ST, LONDON N1 9XW, ENGLAND. 128. Bosco, D., et al., Unique Arrangement of α- and β-Cells in Human Islets of Langerhans. Diabetes, 2010. 59(5): p. 1202. 129. Caicedo, A., Paracrine and autocrine interactions in the human islet: More than meets the eye. Seminars in Cell & Developmental Biology, 2013. 24(1): p. 11-21. 130. Tucker, H.M., et al., Human amylin induces "apoptotic" pattern of gene expression concomitant with cortical neuronal apoptosis. J Neurochem, 1998. 71(2): p. 506-16. 131. Grizzanti, J., R. Corrigan, and G. Casadesus, Neuroprotective Effects of Amylin Analogues on Alzheimer's Disease Pathogenesis and Cognition. J Alzheimers Dis, 2018. 66(1): p. 11-23. 132. Dong, Y., et al., Reversibility of Age-related Oxidized Free NADH Redox States in Alzheimer's Disease Neurons by Imposed External Cys/CySS Redox Shifts. Sci Rep, 2019. 9(1): p. 11274. 133. Dong, Y., M.A. Digman, and G.J. Brewer, Age- and AD-related redox state of NADH in subcellular compartments by fluorescence lifetime imaging microscopy. Geroscience, 2019. 41(1): p. 51-67. 134. Terman, A., Garbage catastrophe theory of aging: imperfect removal of oxidative damage? Redox Report, 2001. 6(1): p. 15-26. 135. Gregg, T., et al., Pancreatic β-Cells From Mice Offset Age-Associated Mitochondrial Deficiency With Reduced KATP Channel Activity. Diabetes, 2016. 65(9): p. 2700-10. 136. Cnop, M., et al., The long lifespan and low turnover of human islet beta cells estimated by mathematical modelling of lipofuscin accumulation. Diabetologia, 2010. 53(2): p. 321-30. 137. Speier, S., et al., Noninvasive high-resolution in vivo imaging of cell biology in the anterior chamber of the mouse eye. Nat Protoc, 2008. 3(8): p. 1278-86. 138. Schaefer, P.M., et al., NADH Autofluorescence-A Marker on its Way to Boost Bioenergetic Research. Cytometry A, 2019. 95(1): p. 34-46. 139. Ritzel, R.A., et al., Induction of beta-cell rest by a Kir6.2/SUR1-selective K(ATP)-channel opener preserves beta-cell insulin stores and insulin secretion in human islets cultured at high (11 mM) glucose. J Clin Endocrinol Metab, 2004. 89(2): p. 795-805. 108 140. Schweitzer, D., et al., Towards metabolic mapping of the human retina. Microsc Res Tech, 2007. 70(5): p. 410-9. 141. Nilsson, S.E.G., et al., Aging of cultured retinal pigment epithelial cells: oxidative reactions, lipofuscin formation and blue light damage. Documenta Ophthalmologica, 2003. 106(1): p. 13-16. 142. Trasino, S., et al., Obesity Leads to Tissue, but not Serum Vitamin A Deficiency. Scientific Reports, 2015. 5: p. 15893. 143. Croce, A.C., et al., Fatty liver oxidative events monitored by autofluorescence optical diagnosis: Comparison between subnormothermic machine perfusion and conventional cold storage preservation. Hepatology Research, 2017. 47(7): p. 668-682. 144. Croce, A.C. and G. Bottiroli, Lipids: Evergreen autofluorescent biomarkers for the liver functional profiling. European journal of histochemistry : EJH, 2017. 61(2): p. 2808-2808. 145. Waanders, L.F., et al., Quantitative proteomic analysis of single pancreatic islets. Proceedings of the National Academy of Sciences, 2009. 106(45): p. 18902-18907. 146. Hoffman, D.P., et al., Correlative three-dimensional super-resolution and block-face electron microscopy of whole vitreously frozen cells. Science, 2020. 367(6475): p. eaaz5357. 147. Lei, C.L., et al., Beta-cell hubs maintain Ca. Islets, 2018. 10(4): p. 151-167. 148. Scarl, R.T., et al., Intact pancreatic islets and dispersed beta-cells both generate intracellular calcium oscillations but differ in their responsiveness to glucose. Cell Calcium, 2019. 83: p. 102081. 149. Nakamura, N., et al., Laser Capture Microdissection for Analysis of Single Cells, in Single Cell Diagnostics: Methods and Protocols, A. Thornhill, Editor. 2007, Humana Press: Totowa, NJ. p. 11- 18. 150. Marselli, L., et al., Gene expression profiles of Beta-cell enriched tissue obtained by laser capture microdissection from subjects with type 2 diabetes. PLoS One, 2010. 5(7): p. e11499. 151. Martino, N., et al., Wavelength-encoded laser particles for massively-multiplexed cell tagging. bioRxiv, 2018: p. 465104. 152. Lu, H., et al., Molecular and metabolic evidence for mitochondrial defects associated with beta- cell dysfunction in a mouse model of type 2 diabetes. Diabetes, 2010. 59(2): p. 448-59. 153. Wang, E., et al., Amylin Treatment Reduces Neuroinflammation and Ameliorates Abnormal Patterns of Gene Expression in the Cerebral Cortex of an Alzheimer's Disease Mouse Model. J Alzheimers Dis, 2017. 56(1): p. 47-61. 154. Gerstenberger, J.P., P. Occhipinti, and A.S. Gladfelter, Heterogeneity in Mitochondrial Morphology and Membrane Potential Is Independent of the Nuclear Division Cycle in Multinucleate Fungal Cells. Eukaryotic Cell, 2012. 11(3): p. 353. 109 155. Godin, Antoine G., B. Lounis, and L. Cognet, Super-resolution Microscopy Approaches for Live Cell Imaging. Biophysical Journal, 2014. 107(8): p. 1777-1784. 156. Cortijo, C., et al., Planar Cell Polarity Controls Pancreatic Beta Cell Differentiation and Glucose Homeostasis. Cell Reports, 2012. 2(6): p. 1593-1606. 157. Rutter, G.A. and D.J. Hodson, Beta cell connectivity in pancreatic islets: a type 2 diabetes target? Cellular and Molecular Life Sciences, 2015. 72(3): p. 453-467. 158. Komatsu, H., et al., Oxygen environment and islet size are the primary limiting factors of isolated pancreatic islet survival. (1932-6203 (Electronic)). 159. Carlsson, P.O., et al., Markedly decreased oxygen tension in transplanted rat pancreatic islets irrespective of the implantation site. (0012-1797 (Print)). 160. Hutton, J.C. and W.J. Malaisse, Dynamics of O2 consumption in rat pancreatic islets. Diabetologia, 1980. 18(5): p. 395-405. 161. Jacobs, H.M., et al., Embryonic exposure to Mono(2-ethylhexyl) phthalate (MEHP) disrupts pancreatic organogenesis in zebrafish (Danio rerio). Chemosphere, 2018. 195: p. 498-507. 162. Reissaus, C.A., et al., A Versatile, Portable Intravital Microscopy Platform for Studying Beta-cell Biology In Vivo. Scientific Reports, 2019. 9(1): p. 8449. 163. Sekine, N., et al., Low lactate dehydrogenase and high mitochondrial glycerol phosphate dehydrogenase in pancreatic beta-cells. Potential role in nutrient sensing. Journal of Biological Chemistry, 1994. 269(7): p. 4895-4902. 110 Appendix I Matlab code in collaboration with Peiyu Wang. %% Islet Analyzing with Thresholding % 10/3/2019 % Since there are things that are not NADH which are really bright. We are % doing a thresholding that only analyzes pixels that have photon count % below a certain value. clear all; close all; %% Data Read In channel_No = 2; % how many channels are used. har_num = 1; % how many harmonics are used. if channel_No<3 channel_con = '%01d'; else channel_con = '%02d'; end imageFile = dir(fullfile(pwd,'*.tif')); z_stacks = numel(imageFile)/(har_num*4*channel_No); %2 harmonics, 4 channel for each acquisition channels ref_stack = cell(z_stacks,channel_No); 111 for z = 1: z_stacks % Adjusting the read in names according to the number of z stacks. if z_stacks == 1 current_z = '*.tif'; elseif z_stacks < 11 current_z = num2str(z-1.','z%01d'); else current_z = num2str(z-1.','z%02d'); end current_filename = []; for k = 1: numel(imageFile) if contains(imageFile(k).name,current_z) current_filename = [current_filename;imageFile(k).name]; end end for j = 1: channel_No if har_num == 2 for h = 1: size(current_filename,1) if contains(current_filename(h,:),['ch' num2str((j-1)*4,channel_con) '.tif']); ref_int = imread(fullfile(pwd,current_filename(h,:))); end if contains(current_filename(h,:),['ch' num2str((j-1)*4+2,channel_con) '.tif']) && contains(current_filename(h,:),'h1_','IgnoreCase',true); G = imread(fullfile(pwd,current_filename(h,:))); G = standardPhase(G); end 112 if contains(current_filename(h,:),['ch' num2str((j-1)*4+3,channel_con) '.tif'])&& contains(current_filename(h,:),'h1_','IgnoreCase',true); S = imread(fullfile(pwd,current_filename(h,:))); S = standardPhase(S); end if contains(current_filename(h,:),['ch' num2str((j-1)*4+2,channel_con) '.tif'])&& contains(current_filename(h,:),'h2_','IgnoreCase',true); G2 = imread(fullfile(pwd,current_filename(h,:))); G2 = standardPhase(G2); end if contains(current_filename(h,:),['ch' num2str((j-1)*4+3,channel_con) '.tif'])&& contains(current_filename(h,:),'h2_','IgnoreCase',true); S2 = imread(fullfile(pwd,current_filename(h,:))); S2 = standardPhase(S2); end end current_ref = struct('int',ref_int,'G', G, 'S', S, 'G2',G2,'S2', S2); else for h = 1: size(current_filename,1) if contains(current_filename(h,:),['ch' num2str((j-1)*4,channel_con) '.tif']); ref_int = imread(fullfile(pwd,current_filename(h,:))); end if contains(current_filename(h,:),['ch' num2str((j-1)*4+2,channel_con) '.tif']); G = imread(fullfile(pwd,current_filename(h,:))); G = standardPhase(G); end if contains(current_filename(h,:),['ch' num2str((j-1)*4+3,channel_con) '.tif']); S = imread(fullfile(pwd,current_filename(h,:))); S = standardPhase(S); end end 113 current_ref = struct('int',ref_int,'G', G, 'S', S); end ref_stack{z,j} = current_ref; end end file_name = "Islet"; save(file_name,'ref_stack') %% Analyze the FLIM signal analyzed_ch = 1; % Inputting which channel needs to be analyzed; thresh_max = 20; % Inputting the maximum thresh hold; thresh_min = 2; % This is putting the all z plane into one vector for each channel; group_vector = cell(1,channel_No); thresh_vector = cell(1,channel_No); for j = 1:size(ref_stack,2) % js is the channel number; int = [];G = [];S = [];% for i = 1: size(ref_stack,1) % i is the z number; int = cat(1,int,ref_stack{i,j}.int(:)); G = cat(1,G,ref_stack{i,j}.G(:)); S = cat(1,S,ref_stack{i,j}.S(:)); end current_ref = struct('int',int,'G', G,'S', S); group_vector{1,j} = current_ref; 114 int_thresh = int((int<thresh_max & int>thresh_min)); G_thresh = G((int<thresh_max) & (int>thresh_min)); S_thresh = S((int<thresh_max & int>thresh_min)); thresh_ref = struct('int',int_thresh,'G', G_thresh,'S', S_thresh); thresh_vector{1,j} = thresh_ref; end figure set(gcf, 'units','normalized','outerposition',[0 0 1 1]); subplot(1,2,1) plotPhasorFast(group_vector{1,analyzed_ch}); title("Original Phasor Plot") subplot(1,2,2) plotPhasorFast(thresh_vector{1,analyzed_ch}); title("Thresholded phasor Plot") z = 3; %% Change the number to the layer that you want to image; figure set(gcf, 'units','normalized','outerposition',[0 0 1 1]); subplot(1,2,1) imagesc(ref_stack{z,1}.int) axis image; colorbar; colormap jet;caxis([0 60]); title("Original NADH Image") subplot(1,2,2) thresh_img = ref_stack{z,1}.int; thresh_img(ref_stack{z,1}.int>thresh_max) = 0; 115 imagesc(thresh_img) axis image; colorbar; colormap jet;caxis([0 60]); title("Thresholded NADH Image") %% G_filt_thresh = G_thresh(G_thresh>1.53e-05); S_filt_thresh = S_thresh(S_thresh>1.53e-05); [G_cen_org,S_cen_org] = findCenPhasor(group_vector{1,analyzed_ch}); [G_cen_thresh,S_cen_thresh] = findCenPhasor(thresh_vector{1,analyzed_ch}); G_precentile_org = quantile(group_vector{1,analyzed_ch}.G(group_vector{1,analyzed_ch}.G>1.53e-5),3); G_25_org = G_precentile_org(1); G_75_org = G_precentile_org(3); S_precentile_org = quantile(group_vector{1,analyzed_ch}.S(group_vector{1,analyzed_ch}.S>1.53e-5),3); S_25_org = S_precentile_org(1); S_75_org = S_precentile_org(3); G_precentile_thresh = quantile(thresh_vector{1,analyzed_ch}.G(thresh_vector{1,analyzed_ch}.G>1.53e-5),3); G_25_thresh = G_precentile_thresh(1); G_75_thresh = G_precentile_thresh(3); S_precentile_thresh = quantile(thresh_vector{1,analyzed_ch}. S(thresh_vector{1,analyzed_ch}.S >1.53e-5),3); 116 S_25_thresh = S_precentile_thresh(1); S_75_thresh = S_precentile_thresh(3); %% figure set(gcf, 'units','normalized','outerposition',[0 0 1 1]); subplot(1,2,1) errorbar(G_cen_org,S_cen_org,S_cen_org - S_25_org,S_cen_org - S_75_org, ... G_cen_org - G_25_org,G_cen_org - G_75_org,'ro'); title("Original Phasor Distribution Percentile") hold on; plotUnitCircle subplot(1,2,2) errorbar(G_cen_thresh,S_cen_thresh,S_cen_thresh - S_25_thresh, S_cen_thresh - S_75_thresh,... G_cen_thresh - G_25_thresh,G_cen_thresh - G_75_thresh,'o'); title("Thresholded Phasor Distribution Percentile") axis([0 1 0 1]) hold on; plotUnitCircle %% Gaussian Fitting: bin_vect = [1/512:1/512:1-1/512]; [org_counts,org_centers] = hist(group_vector{1,analyzed_ch}.G,bin_vect); figure subplot(1,2,1) 117 title('Bar Plot of G, Before Thresholding') bar(org_centers(2:end),org_counts(2:end)); f_org = fit(org_centers(2:end)',org_counts(2:end)', 'gauss2') cen_org1 = f_org.b1;%First Center of fit cen_org2 = f_org.b2;%Second Center of fit subplot(1,2,2) plot(f_org,org_centers(2:end),org_counts(2:end)); % (org_centers(2:end)* org_counts(2:end)')/sum(org_counts(2:end)) %% [thresh_counts,thresh_centers] = hist(thresh_vector{1,analyzed_ch}.G,bin_vect); figure subplot(1,2,1) bar(thresh_centers(2:end),thresh_counts(2:end)); title('Bar Plot of G, After Thresholding') f_thresh = fit(thresh_centers(2:end)',thresh_counts(2:end)', 'gauss2') cen_thresh1 = f_thresh.b1; %First Center of fit cen_thresh2 = f_thresh.b2; %Second Center of fit subplot(1,2,2) plot(f_thresh,thresh_centers(2:end),thresh_counts(2:end)); %% Functions: 118 %% Function: Plot Intensity, First Harmonic. %Peiyu Wang % 03/20/2019 function plotPhasorFast(org_ref) map_res = 1024; phasor_his = zeros(map_res,map_res); for i = 1:size(org_ref.int,1) for j = 1:size(org_ref.int,2) G_index = floor((org_ref.G(i,j)-1.526e-05)*map_res/2+map_res/2+1); %function floor is doing the binning for you. S_index = floor((org_ref.S(i,j)-1.526e-05)*map_res/2+map_res/2+1); if G_index < 1; G_index = 1; end if S_index < 1; S_index = 1; end if G_index > map_res; G_index = map_res; end if S_index > map_res; S_index = map_res; end phasor_his(S_index,G_index) = phasor_his(S_index,G_index)+1; end end % because the pixel value at (0,0) is too high, we change that to 0; [max_val,max_Idx] = max(phasor_his(:)); phasor_his(max_Idx) = 0; phasor_his = flip(phasor_his); imagesc(phasor_his); colormap jet; 119 colorbar; axis image; caxis([0 30]) x_circle = [map_res/2:map_res]; y_circle_pos = map_res/2-floor(sqrt((map_res/4)^2-((x_circle-map_res/2)-map_res/4).^2)); y_circle_neg = map_res/2+floor(sqrt((map_res/4)^2-((x_circle-map_res/2)-map_res/4).^2)); hold on; plot(x_circle,[y_circle_pos;y_circle_neg],'k','LineWidth',1) axis([map_res/2 map_res map_res/5 map_res/2]) xticks([map_res/2:map_res/2^4:map_res]); xticklabels({'0','0.125','0.25','0.375','0.5','0.625','0.75','0.875','1'}); yticks([0:map_res/2^4:map_res/2]); yticklabels({'1','0.875','0.75','0.625','0.5','0.375','0.25','0.125','0'}); end %% Function: Finding the center plot of the phasors function [G_cen, S_cen] = findCenPhasor(org_ref) G_cen = mean(org_ref.G( org_ref.G >= 1.53e-04)); S_cen = mean(org_ref.S( org_ref.S >= 1.53e-04)); end %% Functions function sta_phase = standardPhase(org_phase) %G and S vales were scaled from -1 ~ +1 to 0 ~ (2^16-1), 32767.5 is 0; sta_phase = (double(org_phase)-32767.5)/32767.5; end %% Function: Plot Unit Circle 120 function plotUnitCircle uni_x = [0:1/255:1]; uni_y1 = sqrt(0.25-(uni_x-0.5).^2); uni_y2 = -uni_y1; plot(uni_x,uni_y1,'k',uni_x,uni_y2,'k','HandleVisibility','off'); axis image axis([0 1 0 0.7]) grid on xlabel('G') ylabel('S') end

Abstract (if available)

Abstract Both type 1 and 2 diabetes (T1D and T2D) are characterized by a progressive loss of beta cell function. In both T1D and T2D, impaired glucose-stimulated insulin secretion precedes diabetes onset, which declines further after diabetes and is attributed to glucose toxicity superimposed on beta cell stress. In healthy individuals, the rate of insulin secretion is tightly regulated by blood glucose concentration, although there may be heterogeneity between beta cells for the threshold glucose concentration at which they are recruited to secrete insulin. Glucose readily enters beta cells through membrane resident glucose transporter proteins and the rate of glycolysis is regulated by glucokinase with a Km in the mid physiological range. All pyruvate generated by glycolysis enters the TCA cycle generating ATP that prompts insulin exocytosis. In diabetes, metabolism pathways are reprogrammed, and glycolytic lactate production is upregulated. Glycolysis generates free nicotinamide adenine dinucleotide (NADH) while the TCA cycle generates NADH bound to complexes of the electron transport chain. By applying the phasor approach of fluorescent lifetime microscopy (FLIM), this thesis investigates and establishes strategies to quantify the relative abundance of free and bound NADH in living pancreatic islets and thus the relative rate of glycolysis and oxidative phosphorylation (OX PHOS) in response to glucose in individual cells within islets. For the first time, it has been possible to characterize alpha and beta cell identity and function non-invasively in dispersed and isolated islets enabling the observation of both the suppressed OX PHOS activity in the diabetic mouse model and beta cell hub signaling in healthy islets. The same approach was further applied to human islets in vitro. Finally, islet metabolism was successfully monitored and evaluated by live imaging of FLIM. The findings presented in this thesis will make it possible to non-invasively detect diabetes pathological changes in pharmaceutical and clinical studies in the near future.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Modeling biochemical components and systems through data integration and digital arts approaches

PDF

Machine learning and image processing of fluorescence lifetime imaging microscopy enables tracking and analysis of subcellular metabolism

PDF

Transplantation and genetics of human pancreatic islets in diabetes: Approaches in translational medicine and statistics

PDF

Structure – dynamics – function analysis of class A GPCRs and exploration of chemical space using integrative computational approaches

PDF

Dissecting metabolic changes in muscle stem cells during activation

PDF

Pten regulates beta-cell regeneration intrinsically and independently of development

PDF

Computer aided inhibitor design for fungal bromodomains

PDF

Pten deletion in adult pancreatic beta-cells induces cell proliferation and G1/S cell cycle progression

PDF

System biology approaches to cancer and diabetes metabolism

PDF

Protein kinase Bα (AKT1) but not protein kinase Bβ (AKT2) controls pancreatic β cell growth and survival

PDF

Ectopic fat and adipose tissue inflammation in overweight and obese African Americans and Hispanics

PDF

Prenatal and lifestyle predictors of metabolic health and neurocognition during childhood

PDF

Pro-regenerative effects of fasting/fasting mimicking diet (FMD) on hematopoietic system and pancreatic islet

PDF

IPS and CNS cell models of peroxisomal disorders

PDF

Renin-angiotensin system modulation for the prevention and treatment of metabolic dysfunction

PDF

Metabolic profiling of single hematopoietic stem cells for developing novel ex vivo culture strategies

PDF

Molecular signature of aggressive disease and clonal diversity revealed by single-cell copy number analysis of prostate cancer cells across multiple disease states

PDF

Differential role of two coactivators, CCAR1 and CARM1, for dysregulated beta-catenin activity in colorectal cancer cell growth and gene expression

PDF

The role of endoplasmic reticulum proteins GRP78 and IP3R1 in regulation of glucose homeostasis and acute pancreatitis

PDF

From risk mitochondrial and metabolic phenotype towards a precision medicine approach for Alzheimer's disease

Asset Metadata

Creator Wang, Zhongying (author)

Core Title Non-invasive live-cell imaging for monitoring and evaluating pancreatic islet and beta cell metabolism

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Molecular Biology

Publication Date 05/08/2021

Defense Date 03/04/2020

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag FLIM,metabolism,OAI-PMH Harvest,pancreatic islets

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Stevens, Raymond (committee chair), Cherezov, Vadim (committee member), Fraser, Scott (committee member), Katritch, Vsevolod (committee member)

Creator Email zhongyiw@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-302204

Unique identifier UC11663518

Identifier etd-WangZhongy-8484.pdf (filename),usctheses-c89-302204 (legacy record id)

Legacy Identifier etd-WangZhongy-8484.pdf

Dmrecord 302204

Document Type Dissertation

Rights Wang, Zhongying

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA