Diabetes mellitus—Progress and opportunities in the evolving epidemic

Diabetes, a complex multisystem metabolic disorder characterized by hyperglycemia, leads to complications that reduce quality of life and increase mortality. Diabetes pathophysiology includes dysfunction of beta cells, adipose tissue, skeletal muscle and liver. Type 1 diabetes (T1D) results from immune-mediated beta cell destruction. The more prevalent Type 2 diabetes (T2D) is a heterogenous disorder characterized by varying degrees of beta cell dysfunction in concert with insulin resistance. The strong association between obesity and type 2 diabetes involves pathways regulated by the central nervous system governing food intake and energy expenditure, integrating inputs from peripheral organs and the environment. The risk of developing diabetes or its complications represents interactions between genetic susceptibility and environmental factors including availability of nutritious food and other social determinants of health. This perspective reviews recent advances in understanding the pathophysiology and treatment of diabetes and its complications, that could alter the course of this prevalent disorder.

This section will review lessons learned from human genetics approaches that seek to inform diabetes pathophysiology, and review the pathophysiology of T1D. Specific contributions of the brain and nervous system, adipose tissue, skeletal muscle and liver, to T2D pathophysiology particularly in humans, will be reviewed. We will then discuss the importance of environmental factors that play an important role in diabetes pathogenesis. Genetics of T2D At the end of the last century, our understanding of the genetic landscape for T2D, although not universally accepted, centered on the notion that only a handful of loci, each with a significant impact on an individual’s risk for diabetes, would in concert with environmental risk factors, determine whether an individual developed diabetes. Two decades later, powered by hypothesis-free large-scale genome-wide association studies (GWAS), the genetic landscape now comprises of hundreds of variants, the vast majority with very small effect sizes 8 , 9 . Most T2D-associated variants do not directly alter protein function (i.e. change an amino acid) but rather alter their abundance by modifying regulatory elements in non-coding genomic sequences, which control gene expression 8 , 9 . Many of these elements work in temporal and spatial dependent manners, meaning they give rise to effects on gene expression in precise cell types and at defined developmental time points 8 , 9 . The greatest existing challenge, and potential opportunity is to map these regulatory signals to relevant genes, often called “effector transcripts”, which mediate their influence on diabetes risk. Their identification holds important clues not only into the mechanisms by which glucose homeostasis, diabetes progression and risk of complications are altered in people with diabetes but also the potential to identify safe and effective targets for therapeutic development. The emergence of single cell resolution multi-omic datasets which provide information on whether a gene is expressed in specific cell types, whether chromatin is accessible to transcription factors and whether promoters are in contact with enhancers provides a powerful strategy for connecting diabetes associated variants to their effector transcripts 10 . When these data are coupled with high-throughput cellular phenotyping efforts which alter the expression of hundreds or thousands of genes, the disease relevance of altered gene expression, linked to variants can be assessed at scale 11 . Although each of these signals provides an opportunity for biological insight into the underlying pathophysiology of diabetes unlike in monogenic forms of diabetes there is currently no direct path to precision diagnostics or medicine. There has been considerable interest in overlaying genetic data for cardiometabolic and glycemic traits with those derived for T2D risk 12 , 13 . Shared signals provide important clues regarding underlying tissues and mechanisms through which variants alter the risk for diabetes or its complications. For example, genetic signals that are shared between T2D and proinsulin levels point to a mechanism of action in the pancreatic islet. Both “hard” and “soft” clustering approaches have been deployed by researchers to identify common processes, called “clusters”, which are defective in T2D (e.g. insulin action, beta cell function, dyslipidemia) and their clinical utility is being closely evaluated 8 , 13 . Since most genetic studies have been performed in European populations, efforts are urgently needed to perform similar studies in more diverse populations to prevent health disparities arising from limited access to genetically informed diabetes care that addresses diabetes heterogeneity.

T1D accounts for 5–10% of all diabetes cases and results from autoimmune-mediated destruction of pancreatic beta cells. The year 2021 marked the 100th anniversary of the discovery of insulin, an event that transformed T1D from a once fatal diagnosis into a chronic health condition. Over the ensuing 100 years, knowledge gains have facilitated remarkable advances in diabetes management, as well as the recent approvals of the first disease-modifying therapy and the first cell-based therapy for T1D 25 . However, despite these remarkable achievements, only about 20% of individuals with T1D are able to achieve optimal glycemic control 26 , and life expectancy for those with T1D remains 8–17.7 years shorter than those without diabetes, depending upon age at diagnosis 27 , 28 . We will briefly summarize current understanding of T1D pathophysiology, to set the stage for subsequent discussion of how this knowledge has informed novel strategies for disease prevention and reversal. GWAS have identified over 60 loci that contribute to T1D genetic risk, showing that T1D is highly heritable 29 . The ability to identify T1D genetic risk has facilitated a variety of natural history studies, including birth cohorts assembled through newborn screening and cross-sectional cohorts assembled through targeted autoantibody screening of affected families. Longitudinal assessment of these cohorts has provided insights into environmental associations, potential disease triggers, the trajectory of islet autoimmunity, and the identification of metabolic and immunologic phenotypes during disease evolution 30 – 33 . One of the most important observations informing the natural history of T1D came from a combined analysis of four birth cohorts from the U.S. and Europe, which demonstrated that the presence of two or more islet autoantibodies led to a >80% risk of developing clinical T1D over 15 years of follow-up 34 . In 2005, this observation formed the basis for a new disease staging system, where Stage 1 T1D is defined by the presence of two or more autoantibodies, Stage 2 T1D is defined as multiple autoantibody positivity and dsyglycemia, and Stage 3 T1D is defined by overt hyperglycemia based on American Diabetes Association standards 35 .

A major driver of T2D is obesity and increased adipose tissue mass 59 . Adipocytes are distinct from other cells in their ability to store lipids. Up to 80% of white adipocyte tissue mass can be composed of lipid droplets, an organelle containing a phospholipid monolayer and a core of triglycerides and cholesterol esters. The energy storage capacity of adipocytes allows them to play a central role in communicating energy availability as an endocrine organ. Disruption of energy homeostasis by caloric excess leads to insulin resistance in adipocytes; these cells expand and swell reaching maximal capacity by hypertrophic growth which induces tissue hypoxia. Adipocyte hypertrophy increases surface-to-volume ratio that correlates with adipocyte insulin resistance, and reduced production of the insulin-sensitizing adipokine adiponectin. Adipocyte hypertrophy also increases inflammatory cytokine production leading to increased infiltration of proinflammatory immune cells and systemic inflammation 60 . Metabolically, adipose tissue insulin resistance increases lipolysis elevating circulating FFAs 61 . Thus adipose tissue expansion is not only a manifestation of tissue-specific insulin resistance, but also a driver of systemic insulin resistance by altering adipokine release, promulgating inflammatory cytokines and increasing FFA delivery to other organs. Increased circulating FFAs induce insulin resistance in adipocytes, liver, and skeletal muscle in part through increased production of diacylglycerides (DAGs) and ceramides 61 – 64 ( Figure 3 ). Increased levels of DAGs and ceramides in human plasma are observed in prediabetes, and have been proposed as a diagnostic marker of metabolic health 65 . DAGs and ceramides directly drive insulin resistance through the activation of phosphatases. Mechanistic experiments in mice and cells demonstrated that DAGs bind protein kinase C ε (PKCε) isozymes at the plasma membrane, leading to inhibitory phosphorylation of the insulin receptor that limits its kinase activity and impairs insulin signaling 66 . This regulation is stereospecific, with sn-1,2 DAGs having higher affinity for PKCε and localization to the plasma membrane, while sn-1,3 and sn-2,3 DAGs have higher localization to the lipid droplet and the endoplasmic reticulum (ER). Excess FFAs in T2D also increase the production of ceramides especially long chain C16 and C18 67 . Ceramides activate protein kinase C ζ (PKCζ) for inhibitory phosphorylation of AKT or protein phosphatase 2A (PP2A) to remove activating phosphorylation of AKT, leading to impaired insulin signaling 68 . Induction of insulin resistance is specific to ceramides, as other sphingolipids such as dihydroceramides or sphingomyelin fail to induce insulin resistance or to inhibit lipolysis in mouse models 69 . As DAG and ceramide levels increase in skeletal muscle and liver, selective insulin resistance exacerbates ectopic lipid deposition, further accelerating diabetes pathophysiology. Another way in which obesogenic adipose tissue drives insulin resistance is through decreased release of fatty acid esters of hydroxy fatty acids (FAHFAs) 70 . FAHFAs are a class of complex lipids with an ester linkage of two fatty acids that have been shown to improve insulin sensitivity and are decreased with T2D. FAHFAs exert their activity in part by binding to G-protein coupled receptors in key metabolic tissues to regulate insulin sensitivity, adipogenesis, and energy expenditure in mice 71 . Recent work in mice, demonstrated that adipose triglyceride lipase (ATGL) that regulates lipolysis may act a synthase for FAHFAs, providing a potential link between FAHFAs and lipolysis in T2D through altered ATGL function 72 . The convergence of multiple adipose-derived signals in the obesogenic state, drives a feed-forward cycle that worsens systemic insulin resistance. Adipocyte insulin resistance, characterized by impaired glucose uptake has been mechanistically linked to altered release of insulin sensitizing adipokines and complex lipids, which impact insulin action elsewhere. These changes particularly in liver and skeletal muscle are quantitatively the major drivers of increased glucose levels in diabetes 73 . Skeletal muscle insulin resistance is characterized by an early reduction in insulin-mediated glycogen synthesis 74 , impaired insulin-mediated GLUT4 translocation to the plasma membrane and reduced glucose oxidation 61 . Over time the skeletal muscle insulin resistance contributes to skeletal muscle atrophy, diminished exercise capacity, and reduced mitochondrial mass and bioenergetics 75 . Hepatic insulin resistance manifests primarily as increased hepatic glucose production secondary to impaired suppression by insulin of gluconeogenic genes while promoting lipid accumulation. Hyperglycemia per se, also exacerbates insulin resistance in adipose tissue, skeletal muscle, and liver through increased flux of glucose through the hexosamine biosynthesis pathway to gener

Individuals with T2D often show excess hepatic lipids in imaging tests. What is less clear is which of these individuals will progress to clinically meaningful liver disease, or the time course or inciting factors that initiate this progression. Similarly, GWAS in subjects with T2D and MASLD have exposed common risk alleles in genes that regulate body weight (i.e., FTO ) or hepatic lipid accumulation (i.e., PNPLA3 , TM6SF2 , APOB ) 102 , 103 , but these same risks do not translate well to prediction algorithms of disease progression to MASH. Thus, patients with T2D and MASLD may have hepatocytes that intrinsically store but cannot handle excess lipid without cellular injury and/or non-genomic risks that determine hepatocyte-NPC communication, that culminate in inflammation and fibrosis. To distinguish between these hypotheses, investigators have relied on modeling MASH in mice. Traditional high-fat diets (HFD) induce insulin resistance, liver steatosis and modest inflammation, but not fibrosis 104 . Once popular methionine-choline deficient (MCD) diets that induce liver injury have largely fallen out of favor due to significant anorexia and progressive weight loss. While HFD-MCD hybrid diets were eventually developed 105 , many of these diets fail to mimic obesity and insulin resistance that characterizes human disease 106 . To fill the resultant gap, protocols were developed combining fructose-containing drinking water with diets rich in saturated fat, sucrose and sufficient cholesterol to “humanize” the model, as commonly used strains of mice only absorb a small portion of dietary cholesterol 107 – 109 . These nutrient-dense diets result in obesity, insulin resistance, and all three cardinal features of MASH – hepatic steatosis, inflammation and fibrosis – and eventually hepatocellular carcinoma (HCC) 110 , and thus represent the current state-of-the-art in MASH modeling 111 and are particularly important to mimic comorbid T2D and MASH. This innovation has enabled better understanding of how hepatocyte-NPC interactions determine heterogeneity in disease trajectory. For example, while hepatocyte insulin resistance has long been considered causal to the fasting hyperglycemia that often heralds T2D 112 , recent studies have re-positioned hepatocytes also as causal determinants, not simple bystanders, in liver inflammation and fibrosis. For example, hepatocytes show a surprisingly large endocrine contribution to NPC infiltration and activation, through elaboration of chemotactic 113 and fibrogenic 114 cytokines, even in the absence of detectable hepatocyte injury 114 . Upstream determinants of this hepatocyte response include processes that are increased in individuals with T2D and MASLD, such as re-activated Notch signaling 115 , 116 that increases both HGP 117 and DNL 118 . Understanding dynamics of these hepatocyte signals may have translational implications, given the ability to target hepatocyte pathways with relative specificity, using GalNAc-modified anti-sense oligonucleotides or siRNA, and potentially, in vivo base editing. Despite recent advances, many open questions remain. Key directions for the field include: Role of hyperinsulinemia and non-hormonal factors in co-incident T2D/MASLD: Insulin resistance in T2D prompts compensatory hyperinsulinemia 112 . Data from humans show a positive relationship between plasma insulin levels and hepatic DNL 119 , corroborating animal studies suggesting that inappropriate timing of insulin action may be causal to MASLD. Intriguingly, blocking insulin secretion with octreotide decreased DNL markers and liver triglyceride in rats 120 . This concept is now being tested in non-diabetic individuals using diazoxide ( NCT05729282 ); whether these results will extrapolate to individuals with T2D is unknown. Other hormones (i.e. glucagon) clearly contribute as well, not only by forcing glycogen breakdown but also by reducing hepatic lipids. Similarly, fructose 121 and cholesterol 109 may affect hepatic lipid production. Finally, whether non-nutrient and non-hormone determinants of HGP, such as sympathetic outflow to the liver 122 , similarly co-regulate lipid production is less well-understood.

Despite greater understanding of pro-inflammatory and fibrotic pathways in liver, relatively less attention has been paid to how fibrosis is cleared and how hepatocyte pathways may affect fibrosis resolution. We speculate the existence of commensurate “fibrosis-off” signals for all the recently discovered hepatocyte-determined “fibrosis-on” signals, and that a systematic approach for discovery of regression pathways will have similar impact on liver fibrosis as current work in vascular lesion resolution in atherosclerosis 125 , with possible translational implications. Novel therapeutic targets may be of particular value in individuals with T2D, who are partially resistant to the weight loss and downstream hepatic benefits, of incretin therapy.

Recent studies found convergent gain-of-function somatic mutations in FOXO1 that appear to be clonally selected in liver biopsies of patients with MASLD/MASH 128 . Conceptually similar work identified mutations in other metabolic pathways in mouse models 129 . These data suggest the attractive hypothesis that chronic lipid overload may lead to genetic alterations to protect from further injury. Equally intriguing, this finding also represents a plausible mechanism to explain the epidemiologic associations between MASLD and incident T2D.

Diabetes is a global pandemic, impacting 500 million lives worldwide that disproportionately burdens low and middle-income populations and countries 2 . Social drivers (determinants) of health (SDoH) are the conditions in which people are born, grow, live, work, and age 131 – 133 . SDoH are shaped by power, money and resources and are responsible for greater than 60–70% of health and deleteriously impact T2D 131 – 133 . The SDoH can be considered within the socioecological model where social factors, community and interpersonal relationships influence health behaviors 6 . The SDoH include economic stability (employment, income, expenses, debt, etc.), neighborhood and physical environment (housing, transportation, safety, parks, pollution, geography, etc.), education (literacy, language, etc.), food (food security, nutrition security, access to healthy options, etc.), community and social context (social integration, support systems, community engagement, discrimination, stress, etc.), and the healthcare system (insurance, provider access and availability, linguistic and cultural competency, quality of care, etc.) 131 , 132 . SDoH have both population level components (e.g. the food system) and individual level components (e.g. food insecurity) with the individual level components being referred to as non-medical health-related social needs 132 . The development and control of T2D is uniquely sensitive to SDoH due to the multifaceted effect of SDoH on dysglycemia, from lifestyle behaviors (poor nutritional intake, physical inactivity, sleep insufficiency, stress, etc.) to molecular mechanisms governed by inflammation, hypothalamic-pituitary-adrenal (HPA) axis activation, sympathetic nervous system activation, gut microbial dysbiosis, epigenetic modification, and mitochondrial dysfunction. The impact of SDoH on inequity in T2D outcomes among people and populations has been recently reviewed 131 , 132 . Here we review mechanisms linking SDoH and T2D development and progression using food insecurity and air pollution as two exemplars, and discuss future directions to advance the field. Food and Nutrition Insecurity: Food security is defined as “access by all people at all times to enough food for an active, healthy life,” while nutrition security was recently defined as, “a condition of having equitable and stable availability, access, affordability, and utilization of foods and beverages that promote well-being and prevent and treat disease” 134 , 135 . Thus, nutrition security encompasses food security, dietary quality and SDoH. Community and individual level food insecurity are associated with diabetes incidence, prevalence, and poorer control leading to worse long-term outcomes 134 . Food insecurity drives its deleterious impact on diabetes through: 1) diet and nutrition; and 2) stress. Food insecurity is associated with lower fruit and vegetable intake, along with increased processed foods, refined carbohydrates, saturated fats, added sugars, and unhealthy snacks, leading to worse overall diet quality 134 . The overconsumption of these ultra-processed and calorically dense foods and underconsumption of whole grains, fish, nuts and legumes, essential components of the Mediterranean and American Heart Association’s Life’s Essential 8 diet are linked to inflammation in the short-term through oxidative stress and in the long-term lead to adipose tissue expansion, with resultant adipokine mediated inflammation 136 , 137 . Food insecurity has been associated with increased inflammation (elevated C-reactive protein [CRP] and white blood cell count) 138 , allostatic load (neuroendocrine and inflammatory components including serum DHEA-S and urinary cortisol [HPA axis] and urinary epinephrine and norepinephrine [SNS]), and dietary inflammatory index 139 . Inflammation mediates the association of food insecurity with insulin resistance in diabetes 140 and supplemental nutrition assistance moderates the association of food insecurity with inflammation 141 . Food insecurity has also been linked with reductions in gut microbial diversity in a limited set of studies, which could contribute to perturbed inflammatory responses 142 , 143 . Food insecurity stress impacts compensatory behaviors (time and effort to secure food) and inflammation (toxic stress activates inflammatory pathways) 134 , 144 . The multidimensional complexity of food and nutrition security makes it difficult to establish animal models to interrogate mechanisms, but investigators have recently used unpredictability in the timing and amount of food to recapitulate food insecurity 145 , 146 . This protocol led to changes in food intake with a heightened attraction to palatable food, weight gain and impaired coping mechanisms and memory 145 , 146 . Future studies are warranted to determine the best model to interrogate the pathophysiological pathways of food insecurity to determine precise molecular mechanisms.

Consistent with the DOHaD paradigm, SDoH including food insecurity and air pollution, impacts future development of diabetes in offspring from the periconceptional period through infancy 181 . Some of the best evidence for poor maternal nutrition impacting offspring comes from famines. Adults born across many periods of famine have greater glucose intolerance and risk of T2D as adults 181 . Maternal exposure to air pollution during preconception and gestation has been shown to significantly impair beta cell function and size in adult male offspring in C57Bl/6J mice 182 . These effects are thought to be driven by epigenetic changes including DNA methylation/demethylation, histone modifications, microRNAs and long non-coding RNAs 181 . In summary as exemplars, nutrition security and particulate matter, although contextually different, share common pathophysiological impacts on T2D including oxidative stress, inflammation and HPA axis activation ( Figure 4 ). SDoH impact T2D development and progression through direct and indirect pathophysiological effects from environmental, biological and social factors 133 , 183 . Further understanding of varying SDoH pathophysiology in exposome-based approaches is critical for developing novel interventions and policies to address SDoH and advance equity in T2D prevention and management 132 , 144 . Cross-disciplinary team science with diverse human participants and novel animal models capitalizing on expertise across the translational research continuum are key to determining precise mechanistic insights 184 , 185 .

Diabetes increases the risk of heart failure independently of the increased risk of CAD 207 , 208 . A large number of studies in animal models have identified mechanisms that impair cardiomyocyte and coronary microvascular function and have been extensively reviewed 207 – 209 . These mechanisms include carbotoxicity (lipotoxicity and glucotoxicity), oxidative stress, impaired mitochondrial bioenergetics, mitochondrial uncoupling, impaired myocardial excitation-contraction coupling, and activation of pro-fibrotic pathways ( Figure 5 ). Additionally, activation of hypertrophic signaling pathways results in part from selective insulin resistance, whereby hyperinsulinemia activates hypertrophic and lipotoxic pathways. Recent studies in humans who have received heart transplantations have corroborated these findings 210 , 211 . By leveraging the cardiac biopsy samples obtained post-transplantation, independent groups have now confirmed that within months of cardiac transplantation, normal donor hearts that were transplanted into recipients who develop diabetes exhibit evidence of triglyceride overload and accumulation of toxic lipids such as ceramides. In addition there is clear evidence of mitochondrial respiratory insufficiency, oxidative stress and inflammation. Intriguingly individuals who were treated with metformin exhibited attenuation of these changes 210 . To underscore how rapidly the heart maladapts to more subtle changes in the metabolic milieu, mitochondrial oxidative defects were also observed in transplant recipients with pre-diabetes relative to those who remained non-diabetic 211 . Thus the myocardium in the context of dysregulated glucose metabolism can be likened to a canary in a coal mine. These changes develop rapidly and set the stage for long-term myocardial maladaptation to additional stressors such as ischemia or hypertrophy 188 .

Diabetic Kidney Disease (DKD) is characterized by albuminuria and a reduced estimated glomerular filtration rate (eGFR) 214 . Roughly 40% of patients with diabetes will develop DKD, making DKD a leading cause of end-stage kidney disease. Despite the decline in cardiovascular diseases in the general population and to a certain extent in people with diabetes 215 , the prevalence of DKD has only minimally decreased. This highlights the urgent need to better understand its pathophysiology and to identify new therapies that can slow its progression. Renal Vascular Dysfunction and DKD: The kidney has a unique circulation characterized by a double capillary system. The incoming renal artery (afferent) gives rise to the glomerular capillaries. The outgoing vessel from the glomerulus (efferent artery), still carrying arterial blood then becomes the peritubular capillary system 216 . One of the earliest features of DKD is glomerular hyperfiltration 217 . Systemic hyperglycemia can cause increased proximal tubule sodium reabsorption (as glucose is transported into tubular cells by a sodium-coupled transport mechanism), resulting in reduced sodium and chloride delivery to the macula densa, which is falsely sensed as reduced circulating volume. Consequently, the glomerulus responds by increasing the filtration rate (hyperfiltration) by an angiotensin-mediated constriction of the glomerular efferent artery. This mechanism is described as tubuloglomerular feedback. DKD is a primary microvascular complication of diabetes. Endothelial cells express the insulin receptor and the insulin responsive glucose transporters. Hyperinsulinemia and hyperglycemia increases flux into the polyol pathway, increasing reactive oxygen species production, and inducing the expression of adhesion molecules 218 , 219 . In the kidney, the glycocalyx network surrounding glomerular endothelial cells plays a pivotal role, and the loss of this glycocalyx correlates with albuminuria 220 . Impaired angiogenesis is another crucial aspect of diabetic complications. Within the glomerulus, podocytes serve as an important source of VEGFA, which is essential for the health of glomerular endothelial cells 221 . Animal studies indicate that the initial stage of DKD is characterized by increased glomerular VEGF levels, but in the later stages, VEGF levels are lower contributing to the loss of glomerular and peritubular capillaries 222 . Both VEGF and insulin regulate cellular Akt levels and downstream endothelial nitic oxide synthase. Nitric oxide, an important regulator of vascular smooth muscle tone, also modulates the contractility of mesangial cells in the glomerulus 214 , 223 .

Important contributions of genetics to DKD were suggested by the familial aggregation of the disease. Large genetic consortia including GENIE (Genetics of nephropathy, international effort) identified genetic variations in the COL4A3 gene, which was found to be protective against DKD 230 . Although comprehensive eGFR GWAS investigations have identified numerous loci associated with eGFR, subsequent analyses have shown minimal differences in the genetic architecture of eGFR in diabetic and non-diabetic cohorts 227 . The intriguing “metabolic memory” phenomenon, where historical glycemic control casts shadows on subsequent kidney disease susceptibility, has brought the role of epigenetics in DKD to the forefront 231 . Several underlying mechanisms have been posited, with DNA methylation featuring prominently. Results from the Diabetes Control and Complication Trial (DCCT) have emphasized the role of methylation variations in this enigmatic “metabolic memory” effect. Methylation differences in blood cells, particularly within the TXNIP locus, are correlated with DKD trajectory 232 . Moreover, the methylation landscape in human kidney specimens reveals significant differences in healthy and DKD kidneys, supporting the potential role of epigenetics in DKD progression 233 , 234 . Histone modifications, another facet of epigenetics, involve processes like acetylation and methylation, which govern chromatin accessibility and the transcriptional readiness of DNA. Patterns of histone modifications, notably H3K9 and H3K4, were strongly associated with DKD in the DCCT cohort 235 . It is noteworthy that multiple epigenome modifying enzymes ranging from HDACs to Sirtuins, are now implicated in the development DKD, fibrosis, inflammation, and cellular injury. Recent single cell studies of DKD and control human kidneys detected cell-specific epigenetic changes, that impact chromatin accessibility in DKD 236 . These shifts suggest a potential preprogramming of kidney cells, modulating their responsiveness to external influences, thereby potentially dictating the course of DKD 236 . Various lipid species have been identified as biomarkers or causal factors in DKD. Circulating acylcarnitines, which are intermediates in lipid metabolism linked to insulin resistance, inversely correlate with eGFR 237 . Analysis of blood samples from CKD patients revealed an abundance of specific fatty acids, with β-oxidation efficiency markers decreasing as CKD progressed. Phospholipid species also undergo dynamic changes, with associations drawn between phosphatidylcholine and eGFR decline. Intriguingly, individuals with DKD exhibited increased urinary lysophosphatidylcholine levels as kidney function declined. The role of different metabolites in DKD development remains poorly understood. Most importantly it is difficult to distinguish between changes that cause DKD and those observed as a consequence of the disease. The contribution of novel therapeutics to DKD prevention will be discussed in the section on advances in therapies.

In the last 20 years, there has been a transition in the epidemiology of diabetes 3 . Whilst T2D incidence continues to rise globally, the presentation is now occurring at earlier ages and the burden of the disease rests in low- and middle-income countries (LMIC) with an estimated 4 out of 5 people living with T2D from these regions 3 . The study of T2D across ancestry groups has revealed considerable disease heterogeneity. For example, ketosis-prone T2D in African-Caribbean people 257 , the relatively lean Asian T2D phenotype 258 and higher risk for T2D in south Asian individuals relative to people of white ancestry 259 . Coupled with this recognition has been the analysis of carefully curated longitudinal population studies in people with T2D from European and other ancestries, which have revealed significant disease heterogeneity at presentation that can be linked to outcomes such as DKD or the need for insulin treatment 260 . This heterogeneity has catalyzed precision medicine approaches in diabetes 261 to leverage better outcomes according to sub-phenotype with the aim of tailoring diagnostics or therapeutics to subgroups of populations sharing similar characteristics. The focus on precision medicine approaches in T2D is anchored in the success of monogenic diabetes as an exemplar, which has proven that identification of the specific molecular mechanisms underpinning diabetes can lead to precise diabetes treatment. For example, mutations in the glucokinase gene, require no medical treatment as affected individuals demonstrate no significant increase in lifetime risk of microvascular or macrovascular complications despite lifelong fasting hyperglycemia 262 , whereas mutations in the transcription factor gene HNF1A , can be managed with low-dose sulfonylurea therapy 263 , to achieve superior glycemic control compared to standard care 264 . However, it is not just target-based therapeutics that makes monogenic diabetes a successful front-runner in the diabetes precision medicine space. The implementation of clinical pathways to enable genetic diagnosis in people with suspected monogenic diabetes, has demonstrated that patient stratification through use of biomarker and clinical data, and provision of DNA-based diagnostics, can be integrated into clinical care across different health systems. This advance illustrates how ‘omics” or complex datasets that may be necessary for precision medicine could be integrated into real-world clinics rather than merely being a fanciful future prospect. Precision medicine has been defined as an approach that tailors diagnostics or therapeutics to subgroups of populations sharing similar characteristics, thereby improving accuracy in medical decisions and health recommendations 261 . While precision medicine and personalized medicine are often used interchangeably, the latter extends the definition by incorporating a subjective approach that customizes treatment to align with an individual’s preferences, circumstances, and capabilities 265 . Precision tools are the instruments with which the more nuanced approach (based on objective data) can be taken. For diabetes subclassification these tools can be considered in terms of complexity 266 . At a rudimentary level, simple clinical features or other objective data have been used in isolation to identify subpopulations with similar characteristics. For example, younger age-at-onset of T2D is associated with a shortened life expectancy 267 and a rapid progression to cardiovascular complications. Earlier age at diagnosis is often observed in East Asian and South Asian populations and has been associated with worse beta cell function at diagnosis which appears in part to be linked to genotype 268 . However, identifying sub-populations sharing similar characteristics does not itself fulfil the central tenet of precision medicine; tailored treatment is also needed to make the sub-classification meaningful 261 . Such an approach has been exemplified in the first randomized study (with crossover) of precision treatment for T2D 269 . The study demonstrated that using dichotomous BMI or eGFR cut-offs as stratification tools, predicted greater reduction in HbA1c in people with T2D. Participants with obesity (BMI > 30 kg/m 2 ) exhibited improved glycemic outcomes when treated with pioglitazone compared to sitagliptin 269 . Additionally, those with lower eGFR (60–90 ml/min/1.73 m 2 ) demonstrated a greater reduction in HbA1c levels in response to sitagliptin versus canagliflozin. A similar study conducted in New Zealand showed that the presence of obesity and/or hypertriglyceridemia predicted a greater reduction in HbA1c with pioglitazone than vildagliptin 270 . A separate approach to diabetes subclassifications has stemmed from the integration of several clinical and biomarker variables and/or genetic data using machine learning or complex mathematical algorithms 266 . A clustering approach to classification was pioneered in a study

Here we provide an overview of the major advances in therapeutics for type 2 diabetes, emphasizing the actions of new medicines to reduce glucose, while preventing weight gain, and improving cardiorenal outcomes. Three classes of glucose-lowering medicines were introduced for the treatment T2D in the last 20 years, starting with GLP-1RAs, followed by DPP-4 inhibitors, and SGLT-2 inhibitors. These new medicines enabled control of glucose without weight gain and with a very low risk of hypoglycemia. DPP-4 inhibitors have few adverse events (AEs), are generally administered as once daily tablets, can easily be combined with metformin, and are well suited for treatment of individuals not requiring simultaneous reduction of cardiovascular risk. In contrast, SGLT-2 inhibitors, also administered as a once daily tablet, reduce rates of hospitalization for heart failure and CKD in people with or without T2D. As a result, SGLT2 inhibitors are indicated for reduction of CVD and CKD in people with T2D. Importantly, although the glucose lowering efficacy of SGLT2 inhibitors is diminished in people with a reduced eGFR <60ml/min/1.73m 2 , these agents still exert nephroprotective effects in people with or without T2D, even when administered to individuals with an eGFR as low as 20–25 ml/min/1.73m 2 281 . The dual sodium-glucose co-transporter-1 and -2 inhibitor sotagliflozin does not consistently reduce rates of renal outcomes in people with pre-existing renal impairment and T1D or T2D, although this result might have related to study design. However, sotagliflozin rapidly reduced rates of re-hospitalization for heart failure and cardiovascular death in subjects with T2D with a history of recent worsening heart failure 282 . The SGLT2 inhibitor class of medicine is now established for cardiorenal protection in people with and without T2D, with less extensive innovation expected in these classes beyond several ongoing trials exploring possible new indications. SGLT2 inhibitors and sotagliflozin have been extensively studied in people with T1D, and ongoing trials are exploring the extent to which the benefits may be safely captured, while mitigating the risks by using new technologies to identify and forestall the risk of ketoacidosis. GLP-1 was originally identified as an insulin-stimulating hormone, with subsequent actions encompassing reduction of glucagon secretion and gastric emptying, supporting its development for the treatment of T2D 48 . Subsequent preclinical studies in 1996 identified that intracerebroventricular administration of GLP-1 inhibited food intake, leading to weight loss. Exenatide, a naturally occurring GLP-1RA isolated from the venom of the lizard Heloderma suspectum, was the first GLP-1RA approved for the treatment of T2D in 2005. Exenatide was first developed as a twice daily injectable medicine, followed a few years later by the introduction of lixisenatide, a once daily short-acting GLP-1RA, and liraglutide, an acylated long-acting human GLP-1RA suitable for once daily administration 48 . The AEs associated with GLP-1RAs are predominantly gastrointestinal (GI), principally nausea, vomiting, diarrhea, constipation, and gallstones or gallbladder inflammation 48 . These AEs are most notable at the time of drug initiation and dose up-titration. Persistent GI AEs compromising food and water intake may lead to dehydration, and rarely, acute kidney injury, highlighting the importance of maintaining adequate hydration. Gallbladder events including cholecystitis and cholelithiasis have been reported with GLP-1RAs. The incidence of the gastrointestinal AEs wane over time in the majority of subjects; however, in some individuals, the AEs persist, necessitating treatment discontinuation. Exenatide once weekly, was formulated by incorporating synthetic exenatide into microspheres and injected subcutaneously once a week, enabling sustained delivery of exenatide for the treatment of T2D 283 . It was approved as the world’s first once weekly medicine for people with T2D in 2012. Dulaglutide, a once weekly GLP-1RA, containing a DPP-4-resistant GLP-1 peptide covalently attached to a human IgG4-Fc heavy chain via a small peptide linker, was approved for the treatment of T2D in 2014. Semaglutide first developed as a small acylated peptide GLP-1RA suitable for once-weekly administration was approved for T2D in 2017. An oral once daily version of semaglutide, co-formulated with an absorption enhancer sodium N-(8-[2-hydroxybenzoyl]amino)caprylate, enabling transcellular absorption of semaglutide across the gastric mucosa 284 was approved in 2019. The efficacy of oral versus. injectable semaglutide is proportional to the plasma levels achieved, with somewhat greater bioavailability evident in women, and individuals with a lower BMI 284 . Each iteration of novel GLP-1RAs has achieved greater efficacy both for reduction of HbA1c, and secondarily, for weight loss. Observations of weight loss in people treated with GLP-1