Insulin physiology and metabolic control: current concepts and perspectives

Review design and search strategy

This study was conducted as a narrative review with a structured literature search and integrative evidence synthesis. This design was selected because the review addresses multiple interconnected physiological and pathophysiological domains - from insulin biosynthesis and secretion to its clearance and peripheral resistance - rather than a single, predefined intervention, exposure, or clinical outcome. Methodological rigor was strengthened by considering the quality domains proposed by the Scale for the Assessment of Narrative Review Articles (SANRA), particularly explicit description of the search, appropriate referencing, and critical appraisal of the level and relevance of the evidence.

Electronic searches were conducted in PubMed/MEDLINE, Scopus, and Web of Science as primary bibliographic databases, supplemented by ScienceDirect, SAGE Journals, and Taylor & Francis Online to broaden retrieval of full-text articles. The principal search period was 1 January 2016 to 31 July 2026, with no language restriction. Seminal studies published before 2016 were retained when necessary to support established physiological or mechanistic concepts. Reference lists of key reviews and eligible primary studies were manually screened to identify additional relevant publications.

Search terms combined controlled vocabulary (including MeSH terms where available) with free-text keywords, covering insulin biosynthesis and processing, β-cell identity and dedifferentiation, glucose-stimulated secretion and KATP-channel signaling, incretins, hepatic, renal, and peripheral insulin clearance, CEACAM1 and IDE, insulin signaling and resistance, adipokines and ectopic lipid deposition, gut microbiota and bile acid-incretin signaling, hyperinsulinemia, cellular senescence, mitochondrial dysfunction, metabolic aging, and precision-medicine therapeutic strategies. Because the review spans mechanistically distinct domains, targeted supplementary searches were also performed for section-specific questions, avoiding reliance on a single master query that could reduce retrieval sensitivity across heterogeneous areas of insulin physiology [Supplementary Table 1].

Eligibility, selection, and evidence synthesis

Peer-reviewed publications were considered eligible if they provided direct evidence on normal insulin physiology, dysregulation at any point of its life cycle, or clinically relevant consequences for T2D and related metabolic disorders, including studies in humans, animal models, isolated tissues, primary cells, or organoids, as well as observational, interventional, and translational studies. Systematic reviews, meta-analyses, consensus statements, and authoritative reviews were used for higher-order synthesis, contextualization, and identification of landmark primary studies.

Studies unrelated to insulin physiology or metabolic control were excluded, along with those providing insufficient methodological detail, conference abstracts, non-peer-reviewed grey literature, editorials without original data, and duplicate reports. Titles and abstracts were screened, duplicate records were removed before screening, and potentially eligible records were assessed at full text. When several publications addressed the same mechanism, preference was given to the most recent and methodologically informative evidence, while retaining older landmark studies when necessary for historical or conceptual context.

For each selected source, the study model or population, the pathway or tissue examined, the principal mechanistic finding, translational or clinical relevance, and pertinent limitations were extracted. Evidence was organized according to the sequential and interconnected stages of the insulin life cycle addressed in the manuscript, and was synthesized narratively - not quantitatively - given the heterogeneity of study designs, experimental models, and biological endpoints. Interpretation of each finding considered study design, reproducibility across independent studies, concordance between experimental and human data, and clinical relevance; findings derived predominantly from animal or in vitro models were presented as mechanistic or hypothesis-generating evidence, distinguished from observations directly supported by human studies.

This review was not conducted as an intervention-focused systematic review or meta-analysis, and no quantitative pooling of effect estimates was planned. The structured search strategy, explicit eligibility criteria, domain-based organization of evidence, and critical distinction between preclinical and human evidence support a rigorous narrative synthesis appropriate to the broad physiological and translational scope of the article.

Preproinsulin synthesis and endoplasmic reticulum processing

Pancreatic β-cells are highly specialized cells that produce and secrete insulin in response to hyperglycemia. The insulin gene (INS), located on human chromosome 11p15.5, encodes preproinsulin, a 110-amino-acid precursor whose transcription depends on the synergistic binding of key transcription factors to conserved promoter elements (RI/RIPE and C1/A1). The master regulator of β-cell identity is the pancreatic and duodenal homeobox 1 (PDX-1), acting alongside V-maf musculoaponeurotic fibrosarcoma oncogene homolog A (MAFA), neurogenic differentiation 1 (NEUROD1), and NK6 homeobox 1 (NKX6.1), factors that respond dynamically to glucose and nutrient stimuli through chromatin remodeling, including forkhead box A2 (FOXA2)-mediated H3K4me1 marks^[9-11]. PDX-1 physically engages NEUROD1, while MAFA transcription at mature stages depends on upstream NKX6.1 activation. Together, the PDX-1/NEUROD1/MAFA triad can reprogram non-β-cells, including fibroblasts, induced pluripotent stem cells (iPSCs), and stem cells, into glucose-responsive insulin producers, establishing a foundation for β-cell replacement therapies^[12-16].

In T2D, oxidative stress generated by hyperglycemia and lipotoxicity induces post-translational modifications (phosphorylation and ubiquitination) in PDX-1 and NKX6.1, reducing their binding to the MAFA promoter and triggering the activation of progenitor state genes [Neurogenin-3 (NEUROG3) and SRY of box 9 (SOX9)]^[14]. The result is β-cell dedifferentiation with loss of glucose responsiveness but without significant apoptosis^[16,17]. This process has been confirmed in two independent models. Nkx6.1^−/− mice fail to mature properly and recapitulate the human T2D β-cell profile, characterized by reduced MAFA expression, elevated immature markers, and impaired insulin secretion. In T2D patient-derived iPSCs, stress exposure triggers epigenetic silencing of MAFA and PDX-1 through promoter methylation and H3K27me3 deposition, reactivating a juvenile islet epigenome^[13,18-22]. Single-cell CRISPR screens further confirm the therapeutic potential of these factors in terms of regeneration^[13,23-26].

Insulin biosynthesis begins with translation of INS mRNA into preproinsulin, which is cotranslationally inserted into the endoplasmic reticulum (ER) lumen, where signal peptide cleavage generates proinsulin (86 amino acids). Three disulfide bonds (A6-A11, A7-B7, A20-B19) are then formed between cysteine residues, catalyzed by protein disulfide isomerase (PDI) and ER oxidoreductase (Ero1β), establishing the native insulin conformation. Proinsulin then traffics through the Golgi apparatus into immature secretory granules, where proprotein convertases (PC)1/3 and PC2 cleave two dibasic sites, followed by carboxypeptidase E (CPE)-mediated dipeptide trimming to yield mature insulin and equimolar C-peptide^[27]. Mature insulin is stored as zinc-stabilized hexameric insulin crystals in the acidic (pH ~5.5) granules. Secretion of equimolar C-peptide to insulin makes it a clinically invaluable surrogate marker of endogenous insulin secretion^[28-31].

Whether C-peptide exerts biological activity beyond its role as a biosynthetic by-product remains an unresolved controversy. Wahren et al. have shown that C-peptide binds to cell-surface receptors and activates intracellular signaling cascades, including Gi-protein-coupled pathways, protein kinase C epsilon (PKCε), and mitogen-activated protein kinase (MAPK), producing effects on renal tubular function, peripheral nerve conduction velocity, and microvascular flow in preclinical models and phase II clinical trials in patients with type 1 diabetes (T1D)^[32].

Although preclinical data and small clinical trials suggest that C-peptide may be bioactive, a viewpoint review concluded that despite 30 years of research, its specific receptor and mechanism of action have not been identified. Whereas C-peptide shows therapeutic potential for vascular and nervous damage in diabetes, there is insufficient evidence to support its therapeutic use^[33].

Proinsulin misfolding has emerged as a primary trigger of ER stress in T2D progression^[28,34]. In Akita mutants fed a high-fat diet, hyperactivation of the inositol-requiring enzyme 1 alpha/spliced X-Box binding protein 1 (IRE1α/XBP1s) pathway depletes PDI/Ero1β and drives C/EBP homologous protein (CHOP)-mediated apoptosis independently of glucotoxicity^[34,35]. Single-cell RNA sequencing identified stressed β-cell clusters with elevated binding immunoglobulin protein/glucose-regulated protein 78 kDa (BiP/GRP78) and reduced PDI levels correlating with impaired insulin secretion^[36]. The β-cell dedicates approximately 50%-60% of its total protein synthesis capacity to insulin, demanding substantial constitutive ER folding. Accordingly, the unfolded protein response (UPR), orchestrated by IRE1α, PKR-like endoplasmic reticulum kinase (PERK), and activating transcription factor 6 (ATF6), functions not merely as a stress pathway but as a constitutively active adaptive mechanism^[30].

An unresolved debate concerns the inflection point at which the UPR in pancreatic β-cells shifts from a homeostatic program to a maladaptive state that drives β-cell dysfunction^[37]. Under glucose stimulation, IRE1α-mediated XBP1 splicing enhances ER folding capacity and promotes insulin secretion, while phosphorylation of PERK-mediated eukaryotic translation initiation factor 2 alpha (eIF2α) transiently reduces protein synthesis to protect the folding environment. These responses are essential for β-cell function^[38].

Under sustained demand, as in obesity-associated insulin resistance, chronic IRE1α hyperactivation has been observed to degrade insulin mRNA via regulated IRE1-dependent decay (RIDD), promote c-Jun N-terminal Kinase (JNK)-mediated insulin receptor substrates 1 (IRS1) serine phosphorylation, and initiate CHOP-mediated pro-apoptotic signaling^[39].

In human β-cells, it remains unclear whether this transition reflects a quantitative threshold of IRE1α activation, a qualitative switch in the oligomerization state of IRE1α, or cell-subtype-specific differences in UPR buffering capacity^[31]. Thus, selectively modulating XBP1s without amplifying RIDD or CHOP could constitute an attractive strategy. However, it has not yet been achieved pharmacologically in human islets^[40].

A study by Yau et al. further showed that mitoribosome insufficiency in β-cells impairs oxidative phosphorylation and amplifies UPR-mediated failure reminiscent of T2D islet pathology. Together, this illustrates the co-regulation of mitochondrial and ER quality control^[41].

Several molecular regulators fine-tune ER processing. ER membrane protein complex subunit 10 (EMC10) modulates PDI activity to prevent proinsulin aggregation^[42]. Peroxisome proliferator-activated receptor beta/delta (PPARβ/δ) agonists (GW501516) counteract stress-induced PDI inhibition^[43]. Additionally, chemical chaperones [such as 4-phenylbutyric acid (PBA) and α-lipoic acid (α-LA)] mitigate ER overload by enhancing PDI/Ero1β flux and silencing PERK/activating transcription factor 4 (ATF4)^[31]. CPE is non-redundant in granule maturation. For instance, β-cell-specific CPE deletion (βCpeKO) causes near-total loss of mature granules and elevates proinsulin. It also causes mitochondrial dysfunction and MAFA downregulation in response to high-glucose challenge. Moreover, human CPE mutations are associated with obesity and diabetes. For instance, insulin regulates CPE translation via PDX-1/sterol regulatory element-binding protein 1 (SREBP-1)/eukaryotic translation initiation factor 4 gamma 1 (eIF4G1), identifying maturation defects as early, translationally rescuable stressors^[44].

Of note, murine β-cells express two INS (Ins1 and Ins2), whereas humans possess a single INS gene with greater dependence on NKX6.1, PDX-1, and MAFA for glucose-responsive expression. This difference is critically relevant when extrapolating findings from rodent knockout models to human physiology.

Transcriptional regulation and epigenetic control

The transcription factors PDX-1, NKX6.1, and MAFA form the core regulatory triad governing β-cell identity and INS gene expression. PDX-1 acts as the master regulator of pancreatic development and mature β-cell identity, while MAFA, a late-stage differentiation factor sensitive to oxidative stress, glucotoxicity, and lipotoxicity, represents a key early target of β-cell failure^[16,17]. The clinical relevance of this triad is underscored by Genome-Wide Association Studies (GWAS) data identifying numerous T2D risk loci in proximity to these transcription factors and their targets^[41].

Single-cell RNA sequencing has revealed remarkable heterogeneity within the β-cell population. A landmark study by Dror et al. demonstrated that epigenetic dosage defines at least two major β-cell subtypes with distinct functional and stress-response characteristics^[45]. Additionally, Rutter et al. described subpopulations with differential connectivity within islet electrical networks^[46]. The functionally dominant hub β-cells may bear disproportionate vulnerability to metabolic stress, though whether analogous heterogeneity characterizes human islets to the same degree remains an active area of investigation.

Epigenetic regulation through DNA methylation and histone modification adds a further layer of control. Using isolated human islets exposed to high glucose ex vivo, Hall et al. demonstrated that acute exposure to 19 mM glucose for 48 h alters DNA methylation at CpG sites proximal to PDX1, linking hyperglycemia to early epigenetic reprogramming of the β-cell. Complementing this, comparative studies between islets from donors with type 2 diabetes and non-diabetic donors - Dayeh et al., and more recently Rönn et al. - have identified stable methylation alterations in T2D risk genes, including CDKN1A and SLC2A2, reinforcing the idea that chronic metabolic context remodels the human islet methylome^[47-49]. This work warrants careful interpretation: the conditions represent acute, supraphysiological glucose exposure ex vivo, and whether these methylation changes are causally upstream of β-cell dysfunction or a secondary epigenetic response to an already-dysfunctional metabolic state remains unclear. Longitudinal studies tracking methylation dynamics before and after clinical T2D onset are needed to establish causality. Complementary support comes from genome-wide methylation profiling of T2D islets, which identifies hypermethylation of PDX-1 and MAFA promoters as a consistent T2D epigenomic feature^[14,50].

An integrative framework: the biosynthesis-quality-identity axis

The mechanisms previously described, including INS transcription, ER folding, granule maturation, and epigenetic maintenance of β-cell identity, are conventionally presented as parallel or sequential processes. We propose that they are better understood as a single homeostatic circuit: the BQI axis [Figure 1].

Figure 1. The BQI axis in pancreatic β-cells under physiological and T2D conditions. (Upper panel) Under physiological glucose stimulation, three interconnected nodes sustain β-cell homeostasis. (1) The PDX-1/MAFA/NKX6.1 TF triad drives INS gene expression; (2) ER quality-control machinery - PDI, Ero1β, PC1/3 and PC2, and CPE - ensures correct proinsulin processing, while a constitutively active but self-limiting UPR involving IRE1α/XBP1s, PERK/eIF2α, and ATF6 maintains folding homeostasis; (3) The β-cell identity node encompasses both the mature secretory phenotype and its epigenetic foundation - the PDX-1 and MAFA gene promoters maintained in an active, demethylated state decorated with H3K4me1 and H3K27ac marks. Autocrine insulin signaling through IRS2/PI3K/AKT (green dashed line) reinforces transcription factor binding and epigenetic maintenance, closing a positive homeostatic feedback loop that sustains all three nodes. (Lower panel) Chronic glucotoxicity and lipotoxicity disrupt the BQI circuit at three nodes. (1) Oxidative stress induces post-translational modifications of PDX-1 and NKX6.1, reducing INS transcription; (2) Excess proinsulin synthesis saturates PDI/Ero1β, generating misfolded aggregates and chronically activating IRE1α; IRE1β hyperactivation degrades INS mRNA via RIDD and activates JNK, further impairing IRS2 signaling (red dashed arrow, RIDD/JNK); (3) Loss of autocrine insulin input, combined with oxidative promoter modifications, drives progressive hypermethylation of PDX-1 and MAFA loci, silencing TF expression and reactivating progenitor-state genes (NEUROG3 and SOX9) - the TF reprogramming node (red box). This node carries no asterisk because it is the downstream pathological consequence of epigenome loss at node *3, not an independent therapeutic target. A negative feedback loop (red dashed arrow) illustrates that reduced INS output alleviates ER biosynthetic load but consolidates the dedifferentiated state. Candidate therapeutic targets are indicated by a red asterisk (*) badge in the upper-right corner of the relevant node. (*1) ER quality-control node, upper panel: chemical chaperones (PBA and α-LA) restore PDI/Ero1β folding flux and prevent proinsulin aggregation; PPARβ/δ agonists (GW501516) counteract stress-induced PDI inhibition and stabilize insulin receptor β-subunit conformation; (*2) ER overload node, lower panel: PPARβ/δ agonists act directly in the glucolipotoxic environment to restore PDI activity and reduce misfolded proinsulin burden; (*3) β-cell identity node, upper panel: epigenetic interventions - DNA methyltransferase (DNMT) inhibitors and H3K27me3 demethylases - reactivate PDX-1 and MAFA promoters hypermethylated during T2D progression. The green bidirectional dashed arrow in the right margin of the figure connects node *3 (upper panel) to the TF reprogramming node (lower panel), indicating that pharmacological restoration of the β-cell epigenome at *3 is the upstream intervention that reverse the downstream transcriptional collapse. Green dashed arrows, positive feedback (physiological); red dashed arrows, maladaptive loops (T2D). BQI: Biosynthesis-quality-identity; T2D: type 2 diabetes; INS: insulin; PDX1: pancreatic and duodenal homeobox 1; MAFA: MAF bZIP transcription factor A; NKX6.1: NK6 homeobox 1; TF: transcription factor; ER: endoplasmic reticulum; PDI: protein disulfide isomerase; Ero1β: endoplasmic reticulum oxidoreductase 1 beta; PC1/3: proprotein convertase 1/3; PC2: proprotein convertase 2; CPE: carboxypeptidase E; UPR: unfolded protein response; IRE1α: inositol-requiring enzyme 1 alpha; XBP1s: spliced X-box binding protein 1; PERK: protein kinase R-like endoplasmic reticulum kinase; eIF2α: eukaryotic translation initiation factor 2 alpha; ATF6: activating transcription factor 6; H3K4me1: histone H3 lysine 4 monomethylation; H3K27ac: histone H3 lysine 27 acetylation; IRS2: insulin receptor substrate 2; PI3K: phosphoinositide 3-kinase; AKT: protein kinase B; RIDD: regulated IRE1-dependent decay; JNK: c-Jun N-terminal kinase; NEUROG3: neurogenin 3; SOX9: SRY-box transcription factor 9; PBA: 4-phenylbutyric acid; α-LA: alpha-lipoic acid; PPARβ/δ: peroxisome proliferator-activated receptor beta/delta; DNMT: DNA methyltransferase; H3K27me3: histone H3 lysine 27 trimethylation.

Under physiological conditions, the PDX-1/MAFA/NKX6.1 triad drives INS transcription at a level matched to the β-cell’s ER folding capacity, maintained by PDI, Ero1β, and a tonically active but self-limiting UPR. Correct proinsulin processing by PC1/3, PC2, and CPE completes the production of mature insulin granules. Equimolar C-peptide release serves as a proxy for this entire upstream process. Mature secretory output reinforces transcription factor binding to the INS promoter through autocrine insulin signaling via IRS2/phosphoinositide 3-kinase (PI3K)/serine/threonine protein kinase AKT [also referred to as protein kinase B (PKB)], closing a positive feedback loop that stabilizes β-cell identity. The proinsulin/insulin ratio has been shown to be elevated in preclinical models of β-cell functional immaturity^[51,52].

Chronic glucotoxicity and lipotoxicity disrupt this circuit at multiple nodes simultaneously: (i) oxidative stress modifies PDX-1 and NKX6.1 post-translationally, reducing INS transcription; (ii) excess proinsulin synthesis saturates PDI/Ero1β capacity, generating misfolded aggregates that constitutively activate IRE1α; (iii) chronic IRE1α activation degrades INS mRNA via RIDD and activates JNK, further impairing IRS2 signaling; (iv) loss of autocrine insulin input accelerates PDX-1 nuclear exclusion and MAFA promoter methylation, deepening the dedifferentiated state. The net result is a self-reinforcing collapse of the BQI axis in which dedifferentiation, far from representing a catastrophic failure, may constitute a regulated, homeostatic response that reduces ER biosynthetic load at the cost of secretory function.

This framework has three therapeutic implications not immediately apparent from a sequential model: (a) interventions targeting only glucose control may be insufficient if the epigenetic silencing of PDX-1/MAFA has already been established; (b) chemical chaperones (PBA, α-LA) and PDI activators (PPARβ/δ agonists) that restore ER quality control address a node upstream of transcription factor loss; and (c) complete restoration of β-cell function in advanced T2D may require simultaneous targeting of the epigenetic (demethylation of PDX-1/MAFA promoters), proteostatic (UPR rebalancing), and metabolic (reduction of glucolipotoxic load) arms of the BQI circuit - a multi-target strategy currently untested in human clinical trials. For clinical translation notes of this section, see Supplementary Materials.

The K_ATP channel paradigm: classical model of glucose-stimulated insulin secretion

Glucose-stimulated insulin secretion (GSIS) follows a well-defined metabolic and electrophysiological cascade. The process begins with the entry of glucose into the β-cell cytoplasm through specific transporters. In rodents, this function relies primarily on glucose transporter (GLUT) 2, a high-K_m transporter. In human β-cells, a combination of GLUT2 and the low-K_m transporter GLUT1 are responsible for glucose uptake. This difference reflects a more restrictive glycemic homeostatic control in humans compared to rodents. Notably, loss of GLUT2 from the β-cell plasma membrane has been reported as a marker of functional immaturity associated with constitutive insulin secretion^[53-55] [Figure 2].

Figure 2. Paracrine, endocrine, and intracellular regulation of insulin secretion in pancreatic β-cells. The diagram distinguishes three tiers of regulatory input converging on the β-cell (yellow). Within the islet, α-cells provide paracrine stimulation via glucagon acting on GCGR to activate the cAMP-PKA-EPAC2 axis; under conditions of metabolic stress, α-cells also generate GLP-1* through alternative proglucagon processing by PC1/3, which acts on β-cell GLP-1 receptors. δ-cells exert tonic paracrine inhibition through somatostatin, which suppresses cAMP production and hyperpolarizes the β-cell membrane via SSTR2/5-coupled GIRK channels. Systemic endocrine signals include gut-derived incretins GLP-1 and GIP, which amplify GSIS through Gs-coupled receptor signaling; epinephrine, released from the adrenal medulla, inhibits secretion via α2-adrenergic receptors by suppressing cAMP and promoting KATP channel opening. Gap junction coupling through CX36 among hub and follower β-cells coordinates calcium oscillations across the islet syncytium, while primary cilia contribute to glucose sensing and intracellular Ca²⁺ coordination. Within the β-cell, glucose enters via GLUT1/2 and is phosphorylated by GCK, initiating glycolytic and mitochondrial flux that raises the cytoplasmic ATP/ADP ratio. Subplasmalemmal ATP is additionally supplied by PEP-dependent PK activity. Together these signals drive KATP channel closure (KIR6.2/SUR1), membrane depolarization, VDCC activation, and Ca²⁺ influx, the triggering pathway. An amplifying pathway, acting independently of KATP, enhances exocytosis through MCF (NADPH, glutamate, citrate) and cAMP-PKA-EPAC2 signaling. Insulin granule mobilization requires sequential cytoskeletal remodeling: kinesin-driven long-range transport along microtubules followed by myosin V-mediated short-range delivery along actin filaments. Granules from the RRP are released rapidly in the first secretory phase; RP granules sustain the prolonged second phase. The resulting biphasic, pulsatile insulin output is delivered to the portal circulation. Dashed green arrows, stimulatory; dashed red arrows, inhibitory. GLUT1/2: Glucose transporter 1/2; GCK: glucokinase; KATP: ATP-sensitive potassium channel; PEP: phosphoenolpyruvate; PK: pyruvate kinase; tATP: total ATP; ATP: adenosine triphosphate; ADP: adenosine diphosphate; NADPH: reduced nicotinamide adenine dinucleotide phosphate; KIR6.2: potassium inwardly rectifying channel subfamily J member 11; SUR1: sulfonylurea receptor 1; VDCC: voltage-dependent calcium channel; cAMP: cyclic adenosine monophosphate; PKA: protein kinase A; EPAC2: exchange protein directly activated by cAMP 2; RRP: readily releasable pool; RP: reserve pool; AR: adrenergic receptor; SSTR2/5: somatostatin receptor 2/5; GIRK: G protein-gated inwardly rectifying potassium channel; GLP-1: glucagon-like peptide-1; GIP: glucose-dependent insulinotropic polypeptide; GLP-1R: glucagon-like peptide-1 receptor; GPR: G protein-coupled receptor; GCGR: glucagon receptor; PC1/3: proprotein convertase 1/3; CX36: connexin 36; GSIS: glucose-stimulated insulin secretion; MCF: mitochondria-derived metabolic coupling factors.

Once inside the cell, glucose is phosphorylated by glucokinase (GCK, hexokinase IV), regarded as the true “glucose sensor” of the β-cell and the rate-limiting step of its metabolic flux. This phosphorylation represents the activating signal that triggers glycolysis and mitochondrial oxidative phosphorylation, thereby initiating the metabolic cascade leading to insulin granule exocytosis^[56]. The net result of this metabolism is an increase in the cytoplasmic ATP/ADP ratio, which acts as the key effector signal on the ATP-sensitive potassium channel (K_ATP channel).

The K_ATP channel is composed of four inward rectifier potassium channel 6.2 (KIR6.2) pore-forming subunits and four sulfonylurea receptor 1 (SUR1) regulatory subunits. The rise in the ATP/ADP ratio triggers closure of this channel, producing plasma membrane depolarization, activation of voltage-dependent calcium channels (VDCCs), and a consequent massive influx of Ca²⁺ into the cell^[30,56,57]. This calcium influx drives exocytosis of insulin-containing secretory granules through a biphasic process, determined by the existence of two distinct granule pools: the readily releasable pool (RRP) and the reserve pool (RP). RRP granules are rapidly released within the first 5-10 min following glucose elevation, giving rise to the first phase of secretion, followed by a more prolonged second phase that can extend for hours.

This so-called triggering pathway is complemented by an amplifying pathway that enhances insulin secretion beyond what calcium influx alone can achieve, through metabolic coupling factors including glutamate, reduced nicotinamide adenine dinucleotide phosphate (NADPH), long-chain acyl-CoAs, and malonyl-CoA [Figure 2]. A computational model by Deepa Maheshvare et al. systematically quantified the contributions of key enzymatic steps - particularly phosphofructokinase (PFK) activity and the ATP/ADP-dependent triggering step - to GSIS, confirming that the flux balance between glycolysis and mitochondrial oxidation is the primary determinant of secretory amplitude^[56]. That model also highlighted that K_ATP-channel-independent mechanisms contribute substantially to secretory regulation.

Canonical insulin secretion mechanism under question

The discovery of the structure and function of the β-cell K_ATP channel unquestionably revolutionized our understanding of β-cell physiology. From a therapeutic standpoint, patients harboring mutations that lead to K_ATP inactivation and consequent loss of insulin secretory capacity have greatly benefited from sulfonylureas, which pharmacologically close the channel and restore insulin release. However, the central tenet of the canonical model, namely, that K_ATP closure depends on an increased mitochondrial ATP/ADP ratio, has come under scrutiny. Merrins et al. have recently proposed that this classical framework is incomplete and potentially incorrect. Their work provides genetic evidence that phosphoenolpyruvate (PEP) exerts metabolic control over pyruvate kinase (PK)-dependent ATP/ADP generation at the plasma membrane, which in turn governs K_ATP closure, calcium signaling, and insulin secretion^[58] [Figure 2]. The authors identify three fundamental shortcomings of the dogmatic model: first, it fails to account for mitochondrial thermodynamics, which dictate how mitochondrial ATP production is accelerated in response to increased workload driven by rising ADP levels; second, it disregards metabolic compartmentation; and third, it neglects temporal compartmentation, a feature intrinsically linked to oscillatory metabolism^[58].

In response to these limitations, Merrins et al. proposed a revised model built around four main integrated metabolic cycles: (1) spatial and temporal dimensions of metabolic signaling; (2) the functional relationships among cellular compartments, including the mitochondria, cytosol, and the subplasmalemmal space; (3) both “on-switch” signals that trigger secretion and “off-switch” signals that sustain metabolic and secretory oscillations; and (4) the redundancy among certain pathways, particularly redox components, which may help reconcile divergent findings reported in the literature^[58]. Complementing this framework, Foster et al. provided direct genetic evidence through β-cell-specific deletion of the Pkm1 and Pkm2 isoforms of PK in mice, demonstrating that these isoforms are essential nutrient sensors linking PEP-driven ATP/ADP generation at the plasma membrane to K_ATP channel regulation and downstream calcium signaling^[59] [Figure 2].

Efficient insulin secretion depends on intercellular islet communication. Hub β-cells (1%-10% of the population), which are characterized by high GCK and reduced PDX-1/NKX6.1 expression, act as pacemakers coordinating calcium dynamics and secretory activity of follower cells^[60]. Loss of this connectivity, observed in T2D mouse models and following metabolic/inflammatory insults, contributes to the secretory dysfunction observed in diabetes [Figure 2]. Whether these events happen in the human pancreas requires further research.

Amplifying pathways and mitochondrial coupling factors

Recent studies have significantly deepened our understanding of the amplifying signals that drive GSIS. Mitochondria play a central role by generating metabolic coupling factors like glutamate, aspartate, citrate, and the NADPH redox state. These signals link mitochondrial activity directly to the exocytosis of secretory granules, working independently of the K_ATP channel pathway. Notably, pyruvate cycling pathways produce NADPH, which can directly modulate exocytosis by activating glutaredoxin and thioredoxin systems^[57] [Figure 2].

One of the most fascinating aspects of β-cell biology is how these cells selectively prevent the expression of genes that would create metabolic “futile cycles” incompatible with proper glucose sensing. For instance, lactate dehydrogenase A (LDHA) and monocarboxylate transporter 1 (MCT1), which would allow lactate to act as a secretagogue and blur glucose specificity, are actively silenced in mature β-cells^[57]. When this “disallowance” breaks down, it becomes an early marker of β-cell dedifferentiation and happens before any measurable secretory defects can be detected. This has been confirmed in human islets from donors with T2D, assigning it clinical relevance^[30,57].

Incretin hormones add another layer to this regulatory network. Glucagon-like peptide-1 (GLP-1) and glucose-dependent insulinotropic polypeptide (GIP) are released from enteroendocrine L-cells and K-cells in the small intestine in response to nutrient intake. They bind to glucagon-like peptide-1 receptor (GLP-1R) and glucose-dependent insulinotropic polypeptide receptor (GIPR) receptors on β-cells, amplifying GSIS through the cyclic adenosine monophosphate (cAMP)-protein kinase A (PKA)-exchange protein directly activated by cAMP 2 (EPAC2) signaling pathway^[30]. The physiological amplification provided by incretins accounts for a substantial 50%-70% of postprandial insulin secretion - the “incretin effect” [Figure 2]. One of the early features of T2D is the loss of this incretin effect, primarily due to impaired β-cell responsiveness to GLP-1 rather than reduced GLP-1 secretion. This mechanistic insight explains why GLP-1 receptor agonists (GLP-1RAs) have been clinically successful in treating T2D^[61,62].

However, the native half-lives of GLP-1 and GIP are extremely short because dipeptidyl peptidase-4 (DPP-4) rapidly inactivates both hormones. For this reason, DPP-4 inhibitors are used therapeutically in T2D, as they prolong circulating GLP-1 and GIP levels, thereby lowering postprandial glycemia and improving overall glycemic control. Their effectiveness is also influenced by gastric emptying. In parallel, gut microbiota-derived metabolites such as short-chain fatty acids (SCFAs), secondary bile acids, and indole can directly stimulate incretin release from L-cells. This has led to growing interest in microbiota-targeted approaches, including probiotics, prebiotics, and dietary interventions, as potential adjuncts to improve glycemic control and support gut barrier function^[63].

The limited translation from preclinical to clinical research reveals a significant gap in our understanding of insulin secretion dynamics. In rodent models, pulsatile secretion patterns and strong β-cell synchronization are well-documented. Clinical studies confirm pulsatility of secretion in patients, but the effects of β-cell heterogeneity on secretion patterns remain much less characterized^[64-66]. The critical gap here is whether the impressive single-cell findings in the laboratory actually translate to the complex whole-organ secretion dynamics occurring in human patients. This uncertainty about how cellular heterogeneity impacts overall insulin output represents a major challenge in bridging preclinical mechanistic insights with clinical validation.

Hormonal regulation of insulin secretion

To avert hypoglycemia, β-cell secretion is subject to tonic inhibition from multiple sources. Pancreatic δ-cells produce somatostatin that activates somatostatin receptors (SSTR2/5) to suppress cAMP and hyperpolarize the cell membrane via G protein-gated inwardly rectifying potassium (GIRK) channels. Epinephrine produced by adrenal glands acts via α2-adrenergic receptors to inhibit cAMP while opening K_ATP channels and eventually antagonizing insulin secretion^[67-69] [Figure 2].

K _ATP-independent amplification pathways sustain insulin release. Glucagon from α-cells acts via GCGR-cAMP-PKA-EPAC2 signaling in granule priming and exocytosis, and enhances INS transcription via GCGR-PKA-cAMP response element-binding protein (CREB)^[70]. Glucagon and GLP-1 both derive from preproglucagon, processed by PC2 in α-cells (glucagon) or PC1/3 in L-cells (GLP-1)^[3]. GLP-1 and GIP amplify GSIS via Gs-coupled receptors. Under metabolic stress, α-cells also generate GLP-1 that augments β-cell GLP-1Ra signaling^[71] [Figure 2].

Tirzepatide, a dual agonist of the GIPR and GLP-1R, is a highly effective therapy for T2D and obesity. In mouse models, tirzepatide-stimulated insulin secretion is mediated predominantly through GLP-1R, consistent with its lower potency at the GIPR. In contrast, studies in human islets show that GIPR antagonism consistently attenuates the insulinotropic response to tirzepatide, indicating a physiologically relevant role for GIPR in human tissue. Furthermore, tirzepatide increases both glucagon and somatostatin secretion in human islets, supporting the notion that it modulates islet hormone release through coordinated activation of both incretin receptors^[72].

Other mechanisms underlying insulin secretion

Beyond the classical mechanisms, new molecular targets have been identified for the proper function of pancreatic β-cells. One layer of regulation relies on molecular motors, specialized proteins responsible for the intracellular transport of insulin granules. Following their formation on the trans-Golgi network, these granules need to be transported to the cell membrane for regulated exocytosis. The interlinked and dynamic cellular networks of tubulin and actin cytoskeletons are critical for insulin granule transport and secretion^[73] [Figure 2].

Zhu et al. showed that microtubules are remodeled by glucose in β-cells from mice and humans. Furthermore, microtubules restrict GSIS in mice, acting as a “rheostat” for the β-cell. In fact, dysfunctional β-cells in diabetic mice display abnormally high microtubule density^[74]. In the first phase of insulin secretion, actin filaments depolymerize into actin monomers, allowing the rapid release of insulin granules in RRP. In the second phase, actin filaments repolymerize into a new actin network that facilitates the transport of insulin granules to and release of insulin from the plasma membrane^[75]. Thus, GSIS requires the coordination of both cytoskeleton networks: the transportation of insulin granules first involves long-range movement driven by kinesins on microtubules, with the subsequent handoff to myosin V for the short-range movement on actin filaments to the plasma membrane^[75].

The primary cilium in β-cells regulates insulin secretion. These immotile organelles - confined between acinar cells and islet vasculature - function as sensory hubs coordinating glucose sensing, calcium influx for GSIS, and intercellular communication^[76]. They also form axo-ciliary synapses with nerve fibers for neuroendocrine regulation^[77]. Loss of primary cilia impairs β-cell differentiation, reduces insulin content, and disrupts glucose homeostasis, directly linking ciliary integrity to diabetes etiology^[78,79] [Figure 2].

Pulsatile insulin secretion and paracrine islet regulation

Insulin is secreted in coordinated ultradian pulses (every 3-5 min) arising from synchronized β-cell calcium and metabolic oscillations across the islet. These pulses amplify hepatic insulin extraction and suppress glucagon secretion^[80]. Disruption of pulsatility - elevated basal insulin with blunted peaks - is an early T2D defect exacerbated by hyperinsulinemia itself^[8]. Islet paracrine regulation by glucagon (α-cells) and somatostatin (δ-cells) is a critical secretory control dimension disrupted in T2D^[46].

Gap junction coupling through Connexin 36 (CX36) coordinates electrical activity across the islet syncytium, enabling the coherent calcium oscillations that underlie pulsatile secretion^[46]. Diminished CX36 expression reduces oscillatory coherence and impairs GSIS in mouse models [Figure 2]. However, direct evidence in human islets for analogous changes in T2D remains limited, representing a gap between preclinical mechanistic insight and clinical validation.

Recent advances have transformed our understanding of insulin secretion, revealing that pulsatile release and oscillatory control depend on mechanisms far more complex than previously thought. As introduced previously, β-cell heterogeneity - including functionally specialized hub cells - directly governs the synchronization of calcium oscillations across subpopulations that is critical for coordinated pulsatile secretion^[81,83].

The temporal component has emerged as another fundamental pillar: circadian rhythms critically regulate glucose metabolism and insulin secretion, adding a chronological dimension to secretion control that modulates the amplitude and frequency of ultradian pulses (every 3-5 min)^[83]. This is complemented by the role of incretin hormones like GLP-1, which regulate glucose homeostasis and insulin secretion, with gut microbiota modulating their production and creating a gut-islet link that affects the response to insulin pulses^[84].

In T2D, β-cell functional deterioration occurs in multiple progressive phases - reduction in cell number, functional exhaustion, and dedifferentiation. As noted above, loss of pulsatility is an early defect exacerbated by hyperinsulinemia itself. Islet paracrine regulation (glucagon from α-cells and somatostatin from δ-cells) is likewise disrupted in T2D, altering the secretory control that maintains coherent calcium oscillations^[85-87].

It is crucial to recognize that abnormal insulin secretion is not exclusive to diabetes. In Cushing’s syndrome and paragangliomas, excess glucocorticoids or catecholamines generate insulin resistance and altered secretion with glucose intolerance without overt diabetes^[88-90]. In obesity and metabolic syndrome, compensatory hypersecretion maintains normal glucose but signals existing resistance. For instance, in offspring of T2D parents, hyperinsulinemia precedes diabetes and predicts future risk^[91-93]. These conditions demonstrate that abnormal secretion can occur independently of overt diabetes, serving as an early marker of metabolic dysregulation and identifying individuals at risk of progression to T2D. This underscores the loss of insulin pulsatility as a central event in the transition from metabolic compensation toward manifesting diabetes.

Abnormal insulin secretion in non-diabetic conditions: mechanisms and clinical diagnostic value

Dysregulated insulin secretion is a major but often overlooked feature of non-diabetic conditions. Although these disorders arise from diverse pathophysiological mechanisms, they share abnormal insulin release, ranging from insufficient secretion to hyperinsulinemic hypoglycemia (HH). HH occurs when insulin secretion is inappropriately elevated despite low blood glucose levels, reflecting a loss of normal glucose-dependent regulation of β-cell function^[94].

Maturity-Onset Diabetes of the Young (MODY) due to hepatocyte nuclear factor 1-alpha (HNF1A) or HNF4A mutations (MODY3 and MODY1) causes impaired GSIS through transcriptional network instability, mitochondrial dysfunction, and exocytotic defects. Inheritance is autosomal dominant, and the phenotype includes early-onset hyperglycemia with preserved insulin sensitivity. Clinical recognition is essential, often characterized by a marked sulfonylurea sensitivity that allows targeted therapy and differentiation from T1D and T2D^[95,96] [Table 1].

GCK-MODY (MODY 2) results from a shift in the glucose-sensing threshold of the β-cell, leading to a higher glycemic set-point for insulin secretion without intrinsic secretory failure. This mechanism reflects altered enzymatic activity of GCK, the key glucose sensor in β-cells. The condition is caused by heterozygous inactivating mutations in the GCK gene. Phenotypically, it presents as mild, stable fasting hyperglycemia from early life, with little progression and minimal risk of complications. From a clinical perspective, correct diagnosis is essential, as most patients do not require pharmacological treatment, and unnecessary therapy can be avoided^[97] [Table 1].

Neonatal diabetes from activating K_ATP channel mutations in potassium inwardly rectifying channel subfamily J member 11 (KCNJ11) or in ATP binding cassette subfamily C member 8 (ABCC8) causes persistent channel opening, electrically silencing the β-cell and preventing insulin release. The disease is manifested by the onset of diabetes by 6 months of age. Most patients respond to high-dose sulfonylureas, often enabling insulin therapy discontinuation^[98,99] [Table 1].

Congenital hyperinsulinism (CHI) is caused by loss-of-function mutations in ABCC8/KCNJ11 (most common) or GCK, glutamate dehydrogenase 1 (GLUD1), short-chain 3-hydroxyacyl-CoA dehydrogenase (SCHAD), uncoupling protein 2 (UCP2), HNF1A, or HNF4A, defining distinct molecular subtypes. The phenotype is severe neonatal/infantile hypoglycemia with suppressed ketogenesis and neurological injury risk. Diffuse K_ATP-related forms are often diazoxide-unresponsive and may require near-total pancreatectomy; focal forms (somatic loss of heterozygosity at 11p15) can be cured by targeted resection^[100,101] [Table 1].

While congenital forms highlight genetic causes of insulin hypersecretion early in life, acquired focal lesions represent an important cause of HH in adults.

Pancreatic insulinoma causes autonomous insulin hypersecretion independent of glucose levels, producing recurrent fasting hypoglycemia with neuroglycopenic symptoms. Most insulinomas are sporadic neuroendocrine tumors, with rare cases being associated with multiple endocrine neoplasia type 1 (MEN1). Diagnosis relies on Whipple’s triad plus inappropriately elevated insulin (≥ 3 mIU/mL), C-peptide (≥ 0.6 ng/mL), and proinsulin (≥ 5 pmol/L) during hypoglycemia with suppressed β-hydroxybutyrate^[102,103].

From a clinical diagnostic standpoint, endogenous HH can be systematically evaluated using a supervised 72-h fast or mixed-meal testing, which allows documentation of inappropriate insulin secretion under controlled conditions^[94]. Localization studies are then essential to identify focal causes, particularly insulinoma, and include imaging modalities such as endoscopic ultrasound, pancreatic protocol computed tomography (CT), and functional imaging with ⁶⁸Ga-DOTA-exendin-4 PET/CT^[104]. Establishing the underlying molecular or structural cause is critical for guiding treatment decisions - ranging from medical therapy with diazoxide (K_ATP channel activator) or octreotide (activates somatostatin receptor) to surgical intervention - and for enabling genetic counseling in inherited conditions such as CHI.

Post-bariatric hypoglycemia (PBH) following Roux-en-Y gastric bypass arises primarily from an exaggerated response to incretin, mainly GLP-1. GLP-1R blockade studies and protein-load secretion tests demonstrated that this is triggered by rapid gastric emptying/nutrient delivery after bypass anatomy, which stimulates excessive insulin secretion and produces postprandial hyperinsulinemia and hypoglycemia. In contrast, β-cell hyperplasia or nesidioblastosis - the histological correlate of islet mass expansion - shows inconsistent findings. While some studies find no increase in β-cell number, others document hypertrophy/hyperplasia in resected specimens. Even in patients with positive calcium-stimulation tests, histopathology confirms nesidioblastosis in only a subset of cases, with hypoglycemia frequently recurring after partial pancreatectomy. Consequently, incretin-mediated β-cell hyperfunction is accepted as the central and best-supported mechanism, while the causal role of β-cell hyperplasia/nesidioblastosis remains debated and is likely secondary or not universal^[105-107] [Table 1].

Non-insulinoma pancreatogenous hypoglycemia syndrome (NIPHS) occurs in patients without prior surgery and has been histologically associated with nesidioblastosis in some cases. The hyperinsulinemia in NIPHS arises from an intrinsic, diffuse functional abnormality of pancreatic β-cells rather than from a discrete insulin-secreting tumor. For instance, β-cells show inappropriate, dysregulated insulin release that is not adequately suppressed as glucose falls, leading to postprandial hypoglycemia driven by nutrient-stimulated insulin secretion rather than continuous fasting hypersecretion. This diffuse, non-neoplastic β-cell dysfunction distinguishes NIPHS from insulinoma, where hyperinsulinemia stems from autonomous secretion by a focal clonal tumor largely independent of meal timing. Differentiation from insulinoma is critical to avoid unnecessary pancreatectomy. Its management includes dietary modification and, in severe cases, pharmacologic agents such as diazoxide, somatostatin analogs (octreotide), acarbose, or calcium channel blockers, with surgery reserved for refractory cases^[108,109] [Table 1].

In polycystic ovary syndrome (PCOS), hyperinsulinemia acts as a central endocrine driver through amplification of insulin-androgen feedback loops, contributing to both metabolic and reproductive dysfunction. This mechanism may occur independently or in parallel with classical insulin resistance. Although PCOS is genetically complex and polygenic, multiple susceptibility loci influence insulin signaling and androgen synthesis pathways. The phenotype includes hyperandrogenism, oligomenorrhea, and clinical signs such as acanthosis nigricans, often accompanied by compensatory hyperinsulinemia. Clinically, understanding the primary role of hyperinsulinemia in some phenotypes supports targeted interventions aimed at reducing insulin levels, which can improve both metabolic and reproductive outcomes^[110] [Table 1].

Pancreatic cancer-associated β-cell dysfunction arises from impaired GSIS due to tumor-induced endocrine disruption. Proposed mechanisms include the release of tumor-derived exosomes and induction of ER stress, which impair β-cell function. While not driven by classical germline mutations, this condition involves tumor-related molecular signaling pathways affecting glucose metabolism. Clinically, the disease is characterized by new-onset dysglycemia, often preceding the diagnosis of pancreatic cancer. This association is highly relevant, as recent-onset hyperglycemia in adults may serve as an early warning sign of pancreatic malignancy and warrants further evaluation in appropriate contexts^[111] [Table 1].

Insulin autoimmune syndrome is caused by antibody-mediated disruption of insulin kinetics, leading to delayed insulin clearance and dissociation. Insulin-binding autoantibodies sequester circulating insulin and release it unpredictably, producing fluctuations in its bioavailability. This condition is not linked to a single genetic mutation but is associated with autoimmune predisposition and, in some populations, specific human leukocyte antigen (HLA) haplotypes. The phenotype consists of recurrent, often postprandial hypoglycemia with markedly elevated total insulin levels and positive anti-insulin antibodies in the absence of exogenous insulin use. Clinically, accurate diagnosis is essential to avoid misclassification as insulinoma or factitious hypoglycemia, and management focuses on dietary measures and immunomodulatory strategies when necessary^{[103,112,113]} [Table 1]. For clinical translation notes of this section, see Supplementary Materials.

Portal delivery and first-pass extraction

Insulin exits β-cell islet microvasculature into the portal venous system, reaching the liver first via sinusoidal fenestrae. It has a short half-life of 4-6 min and is cleared at 700-1,000 mL/min^[116]. Under physiological conditions, 50%-80% is cleared during the first hepatic passage, establishing a 2-3-fold portal-to-peripheral gradient. Kidneys handle 25%-40% of insulin disposal. This gradient suppresses hepatic glucose production while protecting against systemic hypoglycemia.

Insulin clearance, together with insulin secretion, regulates the physiologic level of insulin^[5,117-119]. Hepatic insulin clearance occurs via receptor-mediated endocytosis: insulin receptor (INSR) activation triggers carcinoembryonic antigen-related cell adhesion molecule 1 (CEACAM1) phosphorylation, recruiting the insulin-INSR complex to clathrin-coated pits via adaptor protein 2 (AP2)^[80,120]. Internalized insulin is degraded by IDE, PDI, and cathepsin D (CTSD), while the INSR is recycled. CEACAM1 coordinates receptor recycling and insulin degradation through fatty acid synthase (FASN) and src homology region 2-containing protein tyrosine phosphatase 2 (SHP2) interactions^[5,116] [Figure 3A].

Figure 3. Hepatic insulin clearance and the clearance-resistance loop. (A) Hepatic insulin clearance. Portal delivery, receptor-mediated endocytosis, and consequences of impaired clearance. Under physiological conditions, insulin secreted in a biphasic, pulsatile pattern by pancreatic β-cells enters the portal venous system, establishing a portal-to-peripheral insulin gradient. Upon reaching the liver, 50%-80% of portal insulin is extracted during the first pass via receptor-mediated endocytosis in hepatocytes. Insulin binding to the INSR activates tyrosine kinase signaling, which phosphorylates CEACAM1; the phosphorylated CEACAM1-INSR-insulin complex is recruited to clathrin-coated pits via AP2 and internalized into acidified endosomes, where insulin is degraded by IDE, CTSD, and PDI, while the INSR is recycled to the plasma membrane. The residual ~50% of insulin reaches the peripheral circulation, where muscle, adipose tissue, brain, and erythrocytes contribute to disposal; the kidneys handle an additional 25%-40% through megalin-mediated reabsorption and IDE-dependent degradation in proximal tubule cells. Physiological hepatic clearance supports suppression of HGP, lipogenesis regulation, and glycogen synthesis. Conditions that impair hepatic clearance - including downregulation of CEACAM1, hepatic steatosis, obesity, and high-fat feeding - disrupt this axis and promote chronic hyperinsulinemia, peripheral insulin resistance, and MASLD. Subcutaneous insulin administration bypasses the portal route, abolishing the portal-to-peripheral gradient and resulting in relative hepatic under-insulinization alongside peripheral hyperinsulinemia; (B) The clearance-resistance loop. Pathological feedback between hepatic IR and impaired insulin clearance. Hepatic insulin resistance reduces INSR kinase activity and CEACAM1 phosphorylation, impairing receptor-mediated endocytosis and lowering hepatic insulin clearance. The resulting chronic hyperinsulinemia desensitizes and promotes INSR downregulation, further worsening hepatic insulin signaling and completing a self-reinforcing feedback cycle. Each node of this loop independently contributes to a common set of metabolic consequences (central panel). This clearance-resistance loop may represent an underappreciated and early driver of metabolic deterioration, operating independently of primary defects in β-cell secretory capacity. CEACAM1: Carcinoembryonic antigen-related cell adhesion molecule 1; INSR: insulin receptor; AP2: adaptor protein 2; IDE: insulin-degrading enzyme; CTSD: cathepsin D; PDI: protein disulfide isomerase; HGP: hepatic glucose production; MASLD: metabolic dysfunction-associated steatotic liver disease; IR: insulin resistance; T2D: type 2 diabetes.

Renal and peripheral clearance are integrated components of the insulin lifecycle. Proximal tubule cells remove insulin via megalin (LRP2)-mediated reabsorption and peritubular INSR-mediated uptake; urinary excretion accounts for less than 1% of the filtered load^{[116,119,121]}. Chronic kidney disease prolongs insulin half-life, increasing hypoglycemia risk. Obesity may partially compensate by increasing renal plasma flow.

The renal contribution to endogenous insulin clearance is debated. Clamp studies show clearance correlates with lean mass and insulin sensitivity, not GFR^[122]. Peripheral tissues (muscle, adipose, brain, erythrocytes) may account for 15%-20% of disposal via INSR-mediated internalization^[123]. Clearance is lower in African American populations, potentially contributing to T2D susceptibility^[124-127]. High-fat feeding suppresses hepatic clearance by up to 30%, underscoring it as a regulated adaptive process^[128]. Dysregulated clearance promotes hyperinsulinemia, steatosis, and insulin resistance independently of β-cell output^{[5,8,80,125,129,130]}. Recent studies on metabolic dysfunction-associated steatotic liver disease (MASLD)/metabolic dysfunction-associated fatty liver disease (MAFLD) and obesity further support that excess energy intake and hepatic steatosis impair hepatic insulin clearance and promote chronic hyperinsulinemia^[117,131].

Contrary to the traditional liver-centric model, these data support an integrated system implicating kidneys and peripheral tissues in the active regulation of insulin clearance. Characterizing these distinct pathways offers a framework for patient phenotyping and targeted treatment allocation.

Subcutaneous insulin delivery and loss of the portal‑to‑peripheral gradient

As noted above, under physiological conditions, pancreatic insulin secretion creates a portal-to-peripheral insulin gradient, exposing the liver to insulin concentrations three to four times higher than those in peripheral tissues. This gradient is essential for the liver to effectively suppress hepatic glucose production while peripheral tissues maintain normal glucose uptake^[132].

Subcutaneous insulin administration fundamentally abolishes this physiological distribution. When insulin is injected subcutaneously, it enters the peripheral circulation directly, resulting in relative hepatic insulin deficiency and peripheral hyperinsulinemia^[133]. This unphysiological distribution has been consistently demonstrated across multiple studies.

Targeting insulin specifically to the liver corrected metabolic defects caused by peripheral delivery. When insulin was infused into the portal vein, glucose uptake was equally divided between liver and muscle. However, when insulin was delivered peripherally, muscle glucose uptake was four-fold greater than liver uptake, with liver glucose uptake being less than half of normal^[134]. These defects could not be corrected by simply adjusting the peripheral insulin dose.

Comparative studies of portal vs. peripheral venous administration confirm that the portal route is superior for maintaining appropriate insulin concentrations and controlling hepatic glucose output. In peripherally infused animals, basal hepatic glucose output was significantly higher than in portally infused animals. Similarly, peripheral venous administration resulted in inadequate suppression of hepatic glucose production at basal insulin rates, while higher peripheral doses increased peripheral glucose uptake and hypoglycemia risk, a “double jeopardy” for insulin-treated patients^[135].

Current subcutaneously injected insulin products result in hyperinsulin exposure in peripheral tissues (skeletal muscle and adipose) while almost completely avoiding the liver^[136]. This increases the risk of hypoglycemic episodes and causes weight gain due to the unphysiological distribution.

The clinical implications are significant: under-insulinization of the liver at basal rates leads to increased hepatic glucose output and residual hyperglycemia, while higher peripheral doses increase the risk of hypoglycemia. This mechanism contributes to persistent hyperglycemia and diabetic complications despite insulin therapy. The physiological distribution is lost when insulin is injected subcutaneously, causing impaired regulation of hepatic glucose metabolism^[133,134].

Development of hepato-preferential insulin analogs and intraperitoneal (IP) delivery systems aims to restore the portal-peripheral gradient and improve hepatic glucose control. Three therapeutic strategies aim to restore preferential hepatic exposure to insulin:

Hepato-Preferential Analogs (insulin-327): insulin-327 is a 22-carbon fatty diacid-acetylated insulin analog (Novo Nordisk) designed to be hepato-preferential. The molecule cannot cross the continuous capillary endothelium of muscle but freely accesses fenestrated hepatic sinusoids, physiologically replicating the portal gradient. In canine models, insulin-327 corrected deficits in hepatic glucose metabolism induced by peripheral insulin^[134]. However, despite promising preclinical data, insulin-327 remains a preclinical model compound and lacks published human trials.

Oral insulin as an indirect hepato-preferential strategy: The oral route physiologically replicates the first hepatic pass: insulin is absorbed in the intestine and reaches the liver before systemic circulation^[137]. ORMD-0801 (Oramed) is the first oral insulin analog in clinical trials. Oramed demonstrated reduction of hepatic steatosis and fibrosis in patients with T2D + metabolic dysfunction-associated steatohepatitis (MASH), a unique therapeutic convergence between hepato-preference and metabolic liver disease^[138].

IP administration systems: IP administration restores a gradient similar to the portal route: insulin is absorbed in the peritoneal cavity and reaches the liver before distributing to systemic circulation^[139]. This allows the liver to be exposed to higher insulin concentrations, similar to physiological pancreatic secretion. Pharmacokinetic modeling confirms that this space functions as a “virtual compartment” with rapid kinetics (min) compared to slow subcutaneous absorption (1-2 h)^[140,141]. The advantage of this strategy is driven by three key advances: (1) mature continuous glucose monitoring (CGM) systems^[139,142]; (2) thermostable insulin formulations^[143,144] and (3) more robust control algorithms for closed-loop systems^[140,141].

Therapeutic modulation of insulin clearance: opportunities and challenges from CEACAM1 to IDE

Studies in genetically modified mice have firmly established CEACAM1 as the central mediator of receptor-dependent hepatic insulin clearance. Liver-specific inactivation or global deletion of CEACAM1 leads to hyperinsulinemia, peripheral insulin resistance, and hepatic steatosis - phenocopying many features of metabolic syndrome and MASLD^[79,117]. Conversely, forced hepatic overexpression of CEACAM1 curtails diet-induced insulin resistance^[145]. These findings have been replicated across multiple mouse model systems, providing strong preclinical evidence. Direct human data on CEACAM1 polymorphisms and insulin clearance are more limited, though reduced hepatic CEACAM1 expression has been reported in patients with MASLD and MASH, especially as fibrosis progresses^[146,147]. Moreover, lower circulating levels of CEACAM1 have been reported in Portuguese patients with dysregulated glucose homeostasis and with high fatty liver index and, reciprocally, reduced insulin clearance^[148].

Mechanistically, the bidirectional relationship between insulin clearance and insulin resistance creates a pathological feedback loop: hepatic insulin resistance reduces CEACAM1 phosphorylation and insulin endocytosis, lowering clearance and raising systemic insulin concentrations. The resultant hyperinsulinemia desensitizes INSR and promotes receptor downregulation, further impairing signaling^[8,117]. This “clearance-resistance loop” may be an underappreciated driver of metabolic deterioration in early T2D and MASLD^[8]. Hyperinsulinemia also limits the pulsatility of insulin release and reduces brown adipogenesis, thereby decreasing energy expenditure and adding further metabolic burden^[80] [Figure 3B].

The evidence supporting CEACAM1 as a therapeutic target is biologically compelling, but the translation into clinical tools remains incomplete. The available evidence falls into three different tiers:

Molecular control of hepatic gluconeogenesis by insulin: the aquaporin-9 pathway

The liver occupies a central position in glucose metabolism through glycerol gluconeogenesis, accounting for approximately 5%-10% of hepatic glucose output under fasting conditions and rising substantially in insulin-resistant states. The essential step of glycerol entry into hepatocytes (both in rodents and humans) requires facilitated transport across the plasma membrane. This step is primarily mediated by aquaporin-9 (AQP9), the principal aquaglyceroporin expressed in hepatocytes. AQP9 is a rate-limiting step in glycerol gluconeogenesis, making it an important metabolic gatekeeper^[170-172].

AQP9 expression is negatively regulated by insulin: insulin-mediated PI3K/AKT signaling phosphorylates forkhead box O1 (FoxO1), which is then sequestered from the Aqp9 gene promoter, suppressing AQP9 transcription^[171,173]. This regulatory mechanism is physiologically coherent: postprandial insulin elevation suppresses glycerol entry and gluconeogenesis simultaneously, contributing to the postprandial switch from glucose production to glucose uptake. In insulin-resistant states, where FOXO1 is inadequately suppressed, AQP9 expression rises, amplifying glycerol-driven gluconeogenesis and contributing to fasting hyperglycemia. AQP9 downregulation in MASLD, as documented in human liver biopsies^[174], appears to represent a late-stage compensatory mechanism to limit lipid accumulation and glycerol-driven glucose production.

The adipose tissue provides the circulating glycerol pool through lipolysis, mediated by adipocyte aquaglyceroporins AQP3 and AQP7, which facilitate glycerol efflux from adipocytes and are regulated by both insulin (which suppresses lipolysis) and leptin (which activates sympathetic innervation of white adipose tissue to promote lipolysis)^[173]. This adipose-liver aquaglyceroporin axis integrates lipolytic regulation with hepatic glucose production, representing an important inter-organ communication pathway disrupted across multiple points in obesity and T2D^[173,174]. The regulation of aquaglyceroporins by insulin and leptin through the PI3K/AKT/mammalian target of rapamycin (mTOR) pathway provides a mechanistic link between two major hormonal axes and hepatic gluconeogenic substrate supply^[173]. For clinical translation notes of this section, see Supplementary Materials.

Molecular architecture of INSR proximal signaling

Insulin action in peripheral tissues is initiated by binding to the INSR, a heterotetrameric tyrosine kinase receptor composed of two extracellular α-subunits and two β-subunits. The β-subunits span the plasma membrane once and contain tyrosine phosphorylation sites and tyrosine kinase domains in their cytoplasmic tails. INSR is expressed as two isoforms, INSR-A and INSR-B, differing by the presence or absence of a 12-amino-acid sequence encoded by exon 11 segment in the C-terminus domain of the α-subunits. INSR-B predominates in metabolic tissues and mediates the classical metabolic actions of insulin. INSR-A, with higher affinity for insulin and insulin-like growth factor 2 (IGF-2), mediates mitogenic signaling. CEACAM1 preferentially promotes insulin endocytosis via INSR-A, which helps protect the expression of the high-affinity isoform on the hepatocyte surface^[175]. The clinical relevance of INSR isoform imbalance in disease states is an active area of research^[80,117,119].

Insulin binding to α-subunits initiates signaling by triggering conformational changes and trans-autophosphorylation of the β-subunit tyrosine residues (Tyr1158, Tyr1162 and Tyr1163). This activation of intracellular IRS1-4 generates critical docking sites for SH2-domain proteins, effectively bifurcating the signaling cascade into two canonical branches: the mitogenic Ras/MAPK arm and the metabolic PI3K/AKT pathway. Downstream of this metabolic branch, phosphorylated IRS recruits the p85 regulatory subunit of PI3K, activating the p110 catalytic subunit to generate phosphatidylinositol (3,4,5)-trisphosphate (PIP3) at the inner leaflet of the plasma membrane. PIP3 subsequently recruits phosphoinositide-dependent kinase 1 (PDK1) and mammalian target of rapamycin complex 2 (mTORC2), which phosphorylate and activate AKT^[176-178].

Insulin action in the liver

In the liver, the metabolic PI3K/AKT pathway predominates, whereas the mitogenic Ras/MAPK branch plays a less significant role in hepatocytes under normal physiologic conditions^[47]. There is general agreement that the acute suppression of hepatic glucose production by insulin is mediated by both a direct and an indirect effect on hepatocytes^[117,179]. Portal insulin suppresses hepatic glucose production while promoting storage. AKT2 phosphorylates glycogen synthase kinase-3 (GSK3α/β), relieving the inhibition of glycogen synthase 2 (GYS2)^[180-182]. Concurrent activation of protein phosphatase 1 (PP1) further dephosphorylates and activates glycogen synthase, accounting for ~30% of postprandial glucose disposal. AKT excludes FOXO1 from the nucleus, curtailing gluconeogenic transcription [phosphoenolpyruvate carboxykinase (PCK1) and glucose-6-phosphatase (G6PC1)] and reducing CREB/peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC-1α) activity. Glycogenolysis decreases as AKT activates phosphodiesterase 3B (PDE3B) to degrade cAMP, counteracting the effects of glucagon^[92,183,184]. Insulin also induces glycolysis by upregulating GCK via SREBP-1c and by elevating fructose-2,6-bisphosphate through PFK-2, which stimulates PFK-1 allosterically^[2,185]. A complementary insulin-regulated gluconeogenic substrate pathway involves glycerol entry through AQP9. Insulin-mediated FOXO1 nuclear exclusion suppresses AQP9 transcription, thereby restricting glycerol-driven gluconeogenesis. Dysregulation of this axis in insulin-resistant states is discussed herein.

Insulin, particularly at high levels, drives de novo lipogenesis (DNL) as the primary anabolic cue. AKT activates mTORC1 to process and translocate SREBP-1c, a master transcriptional regulator of lipogenic genes. This induces the expression of FASN, acetyl-CoA carboxylase (ACC), and stearoyl-CoA desaturase (SCD1). Additionally, carbohydrate-enriched diets elevate glucose-6-phosphate levels, acting synergistically by causing nuclear translocation of carbohydrate response element-binding protein (ChREBP) to induce the transcription of lipogenic genes^{[2,46,186-188]}. Insulin stabilizes apolipoprotein (Apo)B acutely to promote very low-density lipoprotein (VLDL) assembly. However, in obesity and T2D, “selective hepatic insulin resistance” preserves lipogenic signaling while impairing FOXO1 suppression, yielding hyperglycemia alongside steatosis. Impaired adipose lipolysis causes an excess of free fatty acids (FFAs) to flood the liver, overwhelming oxidation and fueling re-esterification and inflammation; ultimately causing progression of MASLD^[8,189-191].

Despite the lipogenic action of insulin, the liver is physiologically protected against the high portal insulin concentrations. This protection operates in part through CEACAM1/FASN coupling: upon acute insulin pulses, CEACAM1 phosphorylation enhances receptor-mediated insulin internalization and targeting to endosomal compartments. Eventually, phosphorylated CEACAM1 recruits FASN from perinuclear regions, pulling it away from the internalized insulin-INSR complex in late endosomes to suppress FASN lipogenic activity. This results in the simultaneous dissociation of insulin from its receptor to undergo degradation as INSR-A recycles back to the membrane. This dual action of CEACAM1 - coordinating insulin clearance and lipogenesis - is intact even in early MASLD and challenges the concept of “selective hepatic insulin resistance” as an intrinsic signaling bifurcation, suggesting instead that it may be a consequence of impaired clearance (see controversy 1).

Insulin action in skeletal muscle

Skeletal muscle accounts for 60%-80% of postprandial glucose disposal, making it the dominant site of insulin-stimulated glucose uptake. In the basal state, GLUT4 glucose transporters are stored in intracellular vesicles (GSVs) and are retained by phosphorylated AS160 [TBC1 Domain Family Member 4 (TBC1D4)], which maintains Ras-related proteins Rab8A (RAB8A)/RAB10/RAB14 in their inactive GDP-bound form. Insulin activates the INSR-IRS1-PI3K-AKT2 cascade, which phosphorylates AS160 at sites such as Thr642, thereby relieving its inhibitory activity. This allows Rab to engage with the soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) machinery [vesicle-associated membrane protein 2 (VAMP2), syntaxin4 (STX4), and synaptosomal-associated protein 23 (SNAP23)], driving GLUT4 to translocate to the plasma membrane and t-tubules. This mechanism amplifies glucose transport capacity by a factor of 10-40^[2,192-194].

Upon entry, glucose is phosphorylated by hexokinase II (HKII), which traps it intracellularly to fuel glycogen synthesis and oxidation. AKT2 inactivates GSK3 and activates PP1, which dephosphorylates GYS1. This accounts for 30%-35% of postprandial glucose disposal^[195-197]. Insulin further enhances glycolysis and oxidation by activating pyruvate dehydrogenase phosphatase 1 (PDP1) and channeling pyruvate into the Krebs cycle^[198].

In addition to governing carbohydrate metabolism, insulin regulates protein turnover. AKT2 phosphorylates tuberous sclerosis complex (TSC)1/2, thereby activating mTORC1 and ras homolog enriched in brain (RHEB), which drive translation via the phosphorylation of eukaryotic translation initiation factor 4E-binding protein 1 (4EBP1) (releasing eIF4E) and ribosomal protein S6 kinase β-1 (S6K1)^[199]. At the same time, nuclear exclusion of FOXO3 suppresses the expression of components of the ubiquitin-proteasome system, such as atrogin-1 (FBXO32) and muscle RING finger 1 (TRIM63/MuRF1), thereby reducing proteolysis^[200].

Insulin action in adipose tissue

Adipose tissue is an active insulin target with multiple metabolic roles. Insulin powerfully suppresses lipolysis in adipocytes by activating PDE3B, reducing intracellular cAMP and inhibiting hormone-sensitive lipase (HSL) and adipose triglyceride lipase (ATGL). This anti-lipolytic action is among the most insulin-sensitive of all insulin responses, occurring at lower insulin concentrations than those required for glucose uptake. The consequence is that in insulin-resistant states, unsuppressed lipolysis maintains elevated plasma FFA flux, which in turn propagates insulin resistance in liver and muscle, a “lipotoxic feedforward” cycle central to metabolic disease progression^[131,201].

Adipose tissue exhibits high insulin sensitivity, primarily channeling triglycerides into storage while curbing lipolysis. Insulin activates the IRS1/2-PI3K-AKT signaling pathway, which suppresses HSL activity. This is achieved through AKT-mediated activation of PDE3B, which hydrolyses cAMP and reduces PKA signaling. This reduces HSL phosphorylation at activating sites (Ser563/660) and perilipin-1 (PLIN1) at Ser517, thereby preventing comparative gene identification-58 (CGI-58) from being recruited to ATGL. This antilipolytic brake represents the body’s most sensitive insulin response (half-maximal at ~10-15 pM) and lowers circulating FFAs to indirectly enhance muscle glucose uptake and restrain hepatic gluconeogenesis^[2,202-204].

Insulin also drives lipogenesis by triggering GLUT4 translocation to facilitate glucose uptake and re-esterification of glycerol-3-phosphate. Insulin also induces lipoprotein lipase (LPL) on capillary endothelial cells to hydrolyze triglyceride-rich lipoproteins, thereby increasing the influx of FFAs. Although less prominent than in the liver, insulin also upregulates enzymes involved in DNL via SREBP-1c. This consolidates postprandial lipids in adipocytes and prevents their deposition elsewhere in the body^[202,205]. Insulin also suppresses the efflux of glycerol from adipocytes: the aquaglyceroporins AQP3 and AQP7, which facilitate glycerol exit from adipocytes, are regulated by insulin through the PI3K/AKT/mTOR pathway, thereby coupling adipose insulin action to the provision of gluconeogenic substrate for the liver^{[173,206,207]}.

The dynamic secretion and actions of adipokines, including adiponectin, leptin, resistin, tumor necrosis factor alpha (TNF-α), and interleukin 6 (IL-6), play key roles in regulating systemic insulin sensitivity. Failure of adipose tissue to suppress lipolysis appropriately constitutes a critical dysfunction that amplifies systemic metabolic disease. Unchecked FFA efflux floods the liver and muscle, driving ectopic lipid accumulation and lipotoxicity, while reduced secretion of protective adipokines, particularly adiponectin, exacerbates insulin resistance. Elucidating how adipose tissue dysfunction initiates and sustains insulin resistance, as well as the endocrine and paracrine pathways by which it regulates metabolic homeostasis, is therefore crucial.

Insulin action in the central nervous system

The brain is now recognized as a critical regulator of energy homeostasis through insulin signaling. In the hypothalamus, particularly in the arcuate and ventromedial nuclei, insulin activates the PI3K-AKT pathway in agouti-related neuropeptide (AgRP)/neuropeptide Y (NPY) neurons, thereby curbing food intake, stimulating proopiomelanocortin (POMC) neurons, and enhancing satiety. Administering insulin intracerebroventricularly in rodents confirms reduced feeding and improved peripheral sensitivity. Beyond appetite regulation, hypothalamic insulin signaling via vagal circuits suppresses hepatic gluconeogenesis independently of direct liver effects. Insulin also modulates dopaminergic reward circuits by influencing the activity of the dopamine transporter, with central resistance potentially fueling overconsumption in obesity^{[162,208-210]}.

INSR are expressed throughout the central nervous system, particularly in the hypothalamus, hippocampus, and striatum. Central insulin signaling suppresses hepatic glucose output through autonomic pathways, modulates appetite and energy expenditure, and influences peripheral insulin sensitivity, effects that have been demonstrated in human intranasal insulin studies^[211]. Impaired central insulin signaling has been proposed as a mechanism linking insulin resistance to cognitive decline and the increased Alzheimer’s disease risk in T2D, though this connection requires further causal establishment in humans^[212].

Mechanisms of insulin resistance: classical and novel perspectives

Insulin resistance, the impaired ability of target tissues to respond to normal insulin concentrations, is the central feature of T2D, obesity, and the metabolic syndrome. The classical model attributed insulin resistance primarily to serine phosphorylation of IRS1, reducing its affinity for the p85 PI3K subunit. This serine phosphorylation is driven by inflammatory kinases [JNK and the inhibitor of nuclear factor kappa-B kinase subunit β (IKKβ)] activated by saturated fatty acids, ceramides, diacylglycerols (DAGs), and pro-inflammatory cytokines including TNF-α and IL-6^{[178,202,213]}.

The lipid overflow hypothesis states that ectopic lipid accumulation in skeletal muscle and liver, particularly DAG and ceramide species arising from excess saturated fatty acid flux, disrupts insulin signaling through activation of PKCε (liver) and PKCθ (muscle), which phosphorylate IRS1/2 at inhibitory serine residues, reducing PI3K/AKT activation. This model is well supported by human metabolic studies, including isotope tracer studies and liver biopsy data correlating intrahepatic DAG content with hepatic insulin resistance independent of total liver fat^[8,118,214].

A key paradox of hepatic insulin resistance deserves explicit attention: in MASLD, insulin fails to suppress gluconeogenesis (through the PI3K/AKT/FOXO1 axis) but continues to stimulate DNL (through mTORC1/SREBP-1c). This “selective” or “branch-specific” insulin resistance produces the clinical paradox of fasting hyperglycemia and concurrent hepatic lipid accumulation under the same hyperinsulinemic state. Recent evidence has attributed this selectivity to the differential sensitivity of IRS1 vs. IRS2 to inhibitory phosphorylation. This clinically important distinction has been documented in both rodent models and human liver biopsies from MASLD patients^[8,118,214].

Muscle insulin resistance, an early indicator of the progression of T2D, entails impairment of IRS1 phosphorylation through serine modification by kinases such as IKKβ, JNK, and PKCθ. Intramyocellular lipid accumulation, particularly of DAGs and ceramides resulting from incomplete fatty acid beta-oxidation, activates PKCθ and protein phosphatase 2A (PP2A), thereby disrupting INSR signaling. Mitochondrial dysfunction exacerbates this process by reducing oxidative capacity, increasing reactive oxygen species (ROS), and hindering β-oxidation. This ultimately blunts the AS160-RAB-SNARE axis and reduces the availability of GLUT4 on the cell surface. Ectopic intramyocellular lipid accumulation, particularly ceramides derived from saturated fatty acids, impairs AKT activation through PP2A-mediated dephosphorylation. A growing body of evidence has shown that exercise restores insulin-stimulated GLUT4 translocation in insulin-resistant skeletal muscle in humans. This occurs through AMP-activated protein kinase (AMPK)-dependent and Ras-related C3 botulinum toxin substrate 1 (Rac1)-dependent mechanisms that bypass the defective PI3K/AKT pathway to provide a mechanistic rationale for exercise as a cornerstone of T2D management^{[192,215-219]}.

Across insulin-resistant tissues, three convergent pathological processes amplify and perpetuate the signaling defects described above. Chronic low-grade inflammation, initiated by macrophage infiltration of adipose tissue and propagated systemically via nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) and inflammasome activation, generates interleukin 1-beta (IL-1β), TNF-α, and IL-6, which directly impair IRS1 signaling in liver, muscle, and β-cells^[220-222]. Oxidative stress, arising from mitochondrial dysfunction in insulin-resistant tissues, activates JNK and p38 MAPK through ROS-mediated mechanisms and can directly damage INSR and IRS1 through oxidative modification^[221-223]. ER stress, induced by lipid overloading and misfolded protein accumulation, activates the UPR through the PERK-eIF2α-ATF4 branch, suppressing IRS1 expression and promoting resistance through suppressor of cytokine signaling (SOCS)1/3 induction^[221-223]. These three mechanisms are most granularly characterized in rodent models of diet-induced obesity. Human validation, from prospective inflammatory biomarker data, adipose biopsy crown-like structure quantification, and skeletal muscle ceramide and oxidative capacity measurements, confirms their operation but has not achieved comparable mechanistic resolution.

Integrated insulin-glucagon regulation of peripheral insulin action

Following a mixed meal, portal insulin levels increase 10-fold within 15-20 min to suppress hepatic glucose output, primarily by reducing gluconeogenesis and glycogenolysis. Peak peripheral disposal occurs at 60-90 min, driven by GLUT4 translocation in skeletal muscle. Meanwhile, anti-lipolysis in adipose tissue curbs competing FFAs to prioritize glucose oxidation. Concurrent glucagon suppression by portal insulin and co-secreted zinc further silences hepatic production, ensuring efficient nutrient storage^[224,225].

In normal non-obese individuals, insulin levels drop to 2-10 μIU/mL during fasting, shifting the insulin/glucagon molar ratio towards glucagon dominance. Hepatic glucose output of ~7 g/h sustains euglycemia through glucagon-driven gluconeogenesis and glycogenolysis. Adipose tissue lipolysis provides FFAs and glycerol for hepatic triglyceride esterification and substrate provision. Ketogenesis emerges during prolonged fasting or ketogenic conditions^[226-231].

Below ~3.8 mmol/L of glucose, the counterregulatory hierarchy unfolds: insulin secretion halts while glucagon rises (at the ~2.5-3.5 mmol/L threshold), followed by epinephrine (at ~3.2 mmol/L), with growth hormone and cortisol lagging behind (at ~3.0 mmol/L). This culminates in neuroglycopenia below ~2.8 mmol/L^[232,233]. In long-standing T1D and T2D, impaired glucagon responses due to lost islet paracrine cues or autonomic neuropathy heighten the risk of hypoglycemia^[234-237].

Current controversies in peripheral insulin signaling

An integrated network model of peripheral insulin sensitivity

A tissue-by-tissue description of peripheral insulin sensitivity can give the misleading impression that each tissue operates as an independent, parallel module. The evidence and controversies discussed above support the view that insulin sensitivity across tissues constitutes a hierarchical, context-dependent system. Within this framework, dysfunction at any node can propagate through the organism via relative sensitivity thresholds, inter-organ crosstalk, and temporal dynamics^{[195,253,254]}. This network can be described by four organizing principles:

Physiological insulin responses follow hierarchical tissue sensitivity thresholds

Insulin-responsive tissues exhibit varying hormone sensitivity. Adipose anti-lipolysis is half-maximally activated at approximately ~10-15 pM, which is reached under basal conditions^[255,256]. Suppression of hepatic glucose output requires ~30-60 pM^[257,258], while maximal skeletal muscle GLUT4 translocation necessitates around ~340-720 pM^[259]. This hierarchy ensures that adipose anti-lipolysis acts as a substrate gatekeeper, suppressing FFA-mediated interference before muscle glucose disposal. Disruption of this process initiates cascades where excess FFA activates lipotoxic pathways in the liver and muscle, propagating upstream resistance. This suggests that adipose insulin resistance is temporally and mechanistically upstream of hepatic and muscular resistance in common forms of T2D, consistent with longitudinal data from high-risk cohorts^[195] [Figure 4].

Figure 4. Hierarchical network model of peripheral insulin sensitivity and failure modes. The diagram summarizes four principles: a hierarchical threshold of tissue sensitivity to insulin, in which adipose tissue responds first, followed by liver and then muscle; inter-organ communication centered on the liver as a metabolic hub; a constitutive portal-pulsatile insulin architecture with higher portal than peripheral insulin concentrations; and three major failure modes of the network - adipose, hepatic, and muscular - that converge on hyperglycemia and glucose intolerance. This framework links physiological organization to the rationale for combination therapies that target multiple nodes of the network. FFAs: Free fatty acids; VLDL: very-low-density lipoprotein.

Adipose tissue as an active endocrine organ

Adipose tissue was long regarded as a passive lipid storage depot, but the discovery of leptin and adiponectin established it as a dynamic endocrine organ. Adipose tissue secretes biologically active molecules, termed adipokines, plus growth factors, inflammatory cytokines, lipid mediators, and extracellular vesicles that exert autocrine, paracrine, and endocrine effects on glucose metabolism, lipid homeostasis, vascular biology, inflammation, and angiogenesis^[264].

With adipose expansion and dysfunction, secretion shifts from anti-inflammatory adipokines (adiponectin, omentin-1, adipolin, vaspin) that enhance insulin sensitivity to a pro-inflammatory profile. Expansion beyond lipid-storage capacity causes adipocyte hypertrophy, hypoxia, stress, and infiltration by pro-inflammatory M1-polarized macrophages, the hallmark of obese dysfunctional adipose tissue. This profile, dominated by leptin, resistin, visfatin, TNF-α, IL-6, monocyte chemoattractant protein-1 (MCP-1), and plasminogen activator inhibitor-1 (PAI-1), impairs insulin signaling and promotes atherogenesis^[265,266].

Adipose tissue expandability varies between individuals, determining thresholds for overflow and ectopic fat deposition, with implications for T2D risk stratification beyond body mass index (BMI)-based classification. High expandability may preserve insulin sensitivity, whereas limited subcutaneous expandability favors ectopic fat accumulation at lower adiposity levels^[267,268].

Key adipokines: mechanisms and clinical relevance

Leptin, produced mainly by subcutaneous white adipose tissue in proportion to fat mass, signals through leptin receptors (LEPR) (especially hypothalamic LEPRb) to suppress appetite, increase energy expenditure, and activate the sympathetic nervous system. In lipolysis, leptin has a direct effect: chronic signaling increases sympathetic tone in white adipose tissue, promoting lipolysis and reducing adipose mass without changes in food intake in rodents and humans. Leptin also regulates aquaglyceroporins: insulin and leptin coordinate AQP3 and AQP7 expression in adipocytes through the PI3K/AKT/mTOR pathway, linking these hormones to glycerol efflux. In obesity, despite high circulating leptin, central leptin resistance develops through reduced LEPRb signaling and SOCS3-mediated inhibition, blunting anti-obesity signals^[269].

Adiponectin, the most abundant circulating adipokine, has insulin-sensitizing, anti-inflammatory, and cardioprotective effects. Through adiponectin receptors (AdipoR1 and AdipoR2), it activates AMPK and PPARα in muscle and liver, promoting fatty acid oxidation, suppressing hepatic gluconeogenesis, and enhancing glucose uptake. Circulating adiponectin is reduced in obesity and inversely correlates with visceral fat, insulin resistance, and T2D risk. The AdipoR agonist (AdipoRon) and derivatives are in preclinical and early clinical development^[270].

Resistin (from adipose macrophages) promotes hepatic insulin resistance via NF-κB/SOCS3. TNF-α inhibits insulin signaling through IRS1 serine phosphorylation via JNK/IKKβ. IL-6 has concentration-dependent effects: high levels worsen hepatic insulin resistance; low levels may enhance muscle insulin sensitivity via AMPK^[271]. Emerging adipokines [fibroblast growth factor 21 (FGF21), neuregulin-4 (Nrg4), lipocalin-2, omentin-1] extend this network linking adipose biology to β-cell function and hepatic metabolism. This adipose-islet axis - where adiponectin, FGF21, and omentin-1 support GSIS while resistin and pro-inflammatory cytokines impair it - is underappreciated in T2D progression and deserves further investigation^[264].

Angiogenesis, growth factors, and adipose remodeling

Dysfunctional adipose tissue produces growth factors, including vascular endothelial growth factor (VEGF), hepatocyte growth factor (HGF), and platelet-derived growth factor (PDGF), that modulate vascularization, remodeling, and immune recruitment. Hypoxia in expanding adipose tissue stimulates hypoxia-inducible factor 1-α (HIF-1α)-driven VEGF expression; despite angiogenic signaling, vascularization in obesity is often insufficient, sustaining chronic hypoxia and adipokine dysregulation. Visceral adipose tissue, with more macrophages and proximity to the portal circulation, has greater metabolic consequences than subcutaneous fat, with direct clinical implications for cardiovascular risk stratification and therapeutic lifestyle intervention^[272]. For clinical translation notes of this section, see Supplementary Materials.

Microbiota composition and insulin secretion

The gut microbiome, with 38 trillion cells and vast genetic diversity, profoundly affects host metabolism, immunity, and endocrine signaling. It influences the insulin life cycle through mechanisms affecting β-cell function, incretin secretion, hepatic metabolism, and insulin sensitivity^[273].

SCFAs - butyrate, propionate, and acetate - from microbial fiber fermentation are primary metabolic regulators. SCFAs stimulate L-cell GLP-1 and peptide YY (PYY) secretion via G-protein-coupled receptor 41 (GPR41) and GPR43, enhancing the incretin effect and β-cell GSIS^[273]. Butyrate improves β-cell function via AMPK phosphorylation and increases GLP-1. Propionate and butyrate also suppress adipocyte lipolysis and lipogenesis while increasing insulin-stimulated glucose uptake^[274]. High-fiber diets support this by increasing GLP-1 and improving glycemia in T2D patients^[275].

Specific taxa also influence insulin action. Akkermansia muciniphila, low in T2D and obesity, promotes barrier integrity and reduces metabolic endotoxemia; supplementation improves insulin sensitivity in metabolic syndrome^[276]. Mendelian randomization by Sun et al. provided causal evidence linking genera like Flavonifractor and Clostridiaceae to T2D risk^[277], overcoming limitations of cross-sectional studies.

Metabolic endotoxemia and insulin resistance

The gut microbiome promotes systemic insulin resistance via metabolic endotoxemia, where lipopolysaccharide (LPS) enters circulation due to impaired intestinal barrier function. LPS activates toll-like receptor 4 (TLR4) on macrophages, hepatocytes, and adipocytes. This triggers NF-κB-mediated inflammatory signaling that converges on IRS1 serine phosphorylation and JNK/IKKβ pathways^[274]. Metabolic endotoxemia occurs in obese humans, with plasma LPS levels correlating with insulin resistance, HOMA-IR, and T2D risk^[278].

Bile acid metabolism is another microbiome-metabolic axis. Bacterial hydrolases and dehydroxylases modify primary bile acids into secondary bile acids. These activate the G-protein-coupled receptor 5 (TGR5) on enteroendocrine L-cells, stimulating GLP-1 secretion, and the nuclear receptor farnesoid X receptor (FXR), modulating glucose and lipid metabolism^[279]. T2D is linked to altered secondary bile acid profiles, potentially reducing incretin secretion^[280]. RYGB-induced T2D improvement is partly attributed to altered bile acid profiles and increased GLP-1 secretion driven by microbiome changes^[281].

Microbiome-based therapeutic strategies

Probiotic supplementation with Lactobacillus and Bifidobacterium species showed modest improvements in glycated hemoglobin and β-cell function in randomized T2D trials^[282]. Patients receiving probiotics had lower glycated hemoglobin and improved β-cell function than metformin alone^[283]. Fecal microbiota transplantation (FMT) from lean donors to individuals with T2D was tested in a randomized trial by Wu et al., reporting reversal of insulin resistance at 12 weeks, correlating with engraftment of Akkermansia muciniphila and Bifidobacterium^[284]. However, FMT for T2D is investigational and not recommended outside clinical trials. For clinical translation notes of this section, see Supplementary Materials.

A temporal framework for T2D progression

The mechanisms described above - β-cell biosynthesis, insulin secretion, portal delivery and hepatic clearance, and peripheral signaling - do not fail simultaneously in T2D. Prospective studies of high-risk individuals document a stereotyped temporal sequence, the recognition of which has fundamental implications for disease staging and therapeutic timing.

Peripheral insulin resistance develops first, driven by ectopic lipid accumulation, sedentary behavior, and genetic susceptibility. β-Cells compensate through insulin hypersecretion. Concurrently, hepatic steatosis reduces CEACAM1-mediated insulin endocytosis, lowering its clearance and raising systemic insulin - partially masking the emerging secretory deficit. Progressive glucolipotoxicity triggers β-cell dedifferentiation with loss of GSIS. When secretory compensation fails despite reduced clearance, fasting hyperglycemia and clinical T2D manifest^[152].

This sequence reframes the natural history of T2D as a cascade of partially compensated failure rather than a linear decline, and identifies distinct intervention windows at each stage: insulin sensitization and lifestyle modification in the resistance-dominant phase; clearance restoration and β-cell identity preservation in the compensatory hypersecretion phase; and β-cell rescue or replacement strategies once dedifferentiation is advanced [Figure 5].

Figure 5. Temporal dynamics of β-cell failure in T2D. Predisposing factors (ectopic lipid, inactivity, genetic susceptibility) lead to peripheral insulin resistance (Stage 1) with reduced PI3K/AKT signaling in muscle and adipose tissue, driving compensatory β-cell hypersecretion (Stage 2) and reduced hepatic insulin clearance (CEACAM1), which raises systemic insulin and can mask secretory defects. Progressive β-cell stress causes loss of identity programs (MAFA, PDX-1, NKX6.1, FOXO1) and re-expression of progenitor/dedifferentiation markers (NEUROG3, SOX9) with disrupted hub β-cell signaling (CX36, GCK, first-phase GSIS lost), producing irreversible epigenetic remodeling and dedifferentiation (Stage 3) that precedes overt decompensated secretory failure and fasting hyperglycemia (Stage 4). Biomarkers (proinsulin/insulin and C-peptide/insulin ratios) and therapeutic windows (insulin sensitizers, GLP-1Ra, hepatic clearance restoration, β-cell rescue/replacement) are indicated. T2D: Type 2 diabetes; PI3K: phosphoinositide 3-kinase; AKT: protein kinase B; CEACAM1: carcinoembryonic antigen-related cell adhesion molecule 1; MAFA: MAF bZIP transcription factor A; PDX1: pancreatic and duodenal homeobox 1; NKX6.1: NK6 homeobox 1; FOXO1: forkhead box O1; NEUROG3: neurogenin 3; SOX9: SRY-box transcription factor 9; BQI: biosynthesis-quality-identity; CX36: connexin 36; GCK: glucokinase; Ca²⁺: calcium ion; GSIS: glucose-stimulated insulin secretion; HICI: hepatic insulin clearance index; IVGTT: intravenous glucose tolerance test; GLP-1Ra: glucagon-like peptide-1 receptor agonist; ALDH1A3: aldehyde dehydrogenase 1 family member A3; Rx: treatment.

Dedifferentiation as the dominant mechanism of β-cell failure

Classical models attributed progressive β-cell loss in T2D primarily to apoptosis driven by glucotoxicity, lipotoxicity, inflammatory cytokines, and IAPP deposition^[285]. A paradigm shift emerged from work by Son et al.: rather than dying, a substantial proportion of β-cells undergo dedifferentiation, reverting to a progenitor-like state that is hormone-negative but chromogranin A-positive. Human T2D pancreatic specimens confirm enrichment of these “empty” cells, supporting dedifferentiation rather than apoptosis as the dominant mechanism of apparent β-cell loss^[14,286].

The molecular basis of this process, characterized by loss of MAFA, PDX-1, NKX6.1, and FOXO1 expression, reactivation of progenitor genes such as NEUROG3 and SOX9, and epigenetic silencing via hypermethylation of the PDX-1 and MAFA promoters, was previously described as failure of the BQI circuit. Here, that same process is contextualized within the temporal disease arc: dedifferentiation represents the downstream consequence of sustained glucolipotoxic load that the BQI circuit can no longer buffer.

Critically, the severity of metabolic stress determines reversibility. Under moderate glucolipotoxicity, dedifferentiation remains limited and pharmacologically reversible, a state amenable to GLP-1Ra, intensive glucose control, and caloric restriction^[14]. Under severe or prolonged stress, irreversible epigenetic remodeling, mitochondrial dysfunction, and increased apoptotic loading narrow the therapeutic window^[287]. Aldehyde dehydrogenase 1 isoform A3 (ALDH1A3) has emerged as a functional biomarker of the dedifferentiated β-cell state in both human and rodent T2D; its pharmacological inhibition restores β-cell identity markers and improves glycemia in diabetic mice, representing a potential staging and therapeutic target^[288].

β-cell subpopulation vulnerability and hub cell loss

The β-cell population is not homogeneous. Single-cell RNA sequencing of human pancreatic islets has identified functionally distinct subpopulations with differential roles in coordinating islet secretory dynamics. A minority of “hub” or “leader” β-cells, characterized by high CX36 expression and specific metabolic gene signatures, orchestrates islet-wide oscillatory calcium dynamics and bears disproportionate responsibility for coordinated pulsatile secretion. These cells are identifiable by reduced PDX-1 and NKX6.1 expression alongside elevated GCK, a paradoxical combination that confers functional dominance but also heightened vulnerability to metabolic stress-induced dedifferentiation^[60,289,290].

Loss of hub cells therefore has an outsized impact on secretory capacity relative to their fractional representation in the islet. This hierarchical architecture explains why early functional deficits in T2D, loss of first-phase insulin secretion and disrupted pulsatility, may precede significant reductions in total β-cell mass^[290,291]. Additionally, single-cell transcriptomic analyses have revealed that endocrine cell alternative splicing is broadly dysregulated in T2D, adding a post-transcriptional dimension to the dedifferentiation landscape that is not captured by transcription factor immunostaining alone^[292].

Disease staging and the therapeutic window

The temporal model outlined above offers a framework for staging T2D that goes beyond glycemic thresholds. The proinsulin/insulin ratio and the loss of first-phase GSIS, detectable by hyperglycemic clamp or intravenous glucose tolerance test, serve as clinical indicators of advancing dedifferentiation. An increasing proinsulin/insulin ratio, in the context of preserved fasting glucose, signals early BQI circuit failure and identifies individuals who may benefit most from interventions targeting β-cell identity before irreversible epigenetic changes occur.

The C-peptide/insulin molar ratio (HICI) provides a complementary staging dimension: a declining HICI in the context of normoglycemia indicates that hyperinsulinemia is sustained partly by reduced clearance rather than pure hypersecretion, a distinction with therapeutic implications for the choice between insulin secretion-reducing vs. clearance-restoring strategies^[80,293,294].

Together, these biomarkers define a staging axis that ranges from compensated resistance with preserved β-cell identity, through dedifferentiation onset with clearance adaptation, to decompensated T2D with irreversible secretory failure. This axis may be more mechanistically informative than glycated hemoglobin A1c (HbA1c) alone and better aligned with the pathophysiological sequence described in this manuscript. For clinical translation notes of this section, see Supplementary Materials.

Therapeutic strategies targeting the insulin life cycle

The insulin life cycle framework provides a rational basis for organizing therapeutic targets. Different drug classes act at different phases: incretins enhance secretion; metformin reduces hepatic glucose production; SGLT2i lower glycemia by renal glucose excretion; insulin sensitizers (thiazolidinediones) improve peripheral signaling; GLP-1RAs additionally reduce appetite and body weight; and insulin replacement corrects absolute secretory deficiency^[61,62].

GLP-1RAs: from insulin lifecycle modulation to precision therapeutics

GLP-1RAs have transformed T2D management by acting at multiple nodes of the insulin life cycle. By augmenting postprandial insulin secretion in a glucose-dependent manner, suppressing glucagon, delaying gastric emptying, and reducing appetite through central pathways, GLP-1RAs act on several nodes of metabolic regulation simultaneously^{[61,62,295,296]}. Their clinical benefits also extend to cardiovascular and renal outcomes, while heterogeneity in glycemic response appears to depend in part on residual β-cell function, with C-peptide emerging as a potential predictive biomarker^[61,297].

SGLT2i: a novel mechanism with pleiotropic benefits

SGLT2i reduce hyperglycemia by blocking renal glucose reabsorption, causing glucosuria independent of insulin. Lowered glucose and insulin, in turn, reduce β-cell demand, potentially slowing dedifferentiation. Cardiorenal benefits across multiple outcome trials involve mechanisms beyond glycemia: osmotic diuresis, natriuresis, reduced intraglomerular pressure, and direct cardiac energetic effects^[298-302]. SGLT2i also reduce hepatic fat and improve transaminases in MASLD, consistent with reduced hepatic insulin resistance^[118].

Precision medicine and disease subtyping

A major shift in T2D management philosophy is the recognition of T2D as a heterogeneous syndrome. The cluster-based classification proposed by Ahlqvist et al.^[263] defined five T2D clusters based on six variables [glutamic acid decarboxylase antibodies (GADA), age at diagnosis, BMI, HbA1c, HOMA-B, HOMA-IR]: severe autoimmune diabetes (SAID), severe insulin-deficient diabetes (SIDD), severe insulin-resistant diabetes (SIRD), mild obesity-related diabetes (MOD), and mild age-related diabetes (MARD). These clusters showed differential risks of diabetic complications and potentially different optimal therapies.

A systematic review^[61] found that reduced insulin secretion markers predicted a lesser glycemic response to GLP-1Ra, and reduced renal function predicted a lesser response to SGLT2i. A subsequent Bayesian causal forest analysis^[62] validated an individualized SGLT2i vs. GLP-1Ra selection algorithm, showing predicted-response-based targeting improved short-term tolerability and long-term microvascular risk. These analyses represent early steps toward clinical precision medicine in T2D^[61,62], though prospective validation in diverse populations is required.

Emerging targets: β-cell regeneration and clearance optimization

Several emerging therapeutic strategies address upstream life-cycle defects. ALDH1A3 inhibition to reverse β-cell dedifferentiation has shown promise in rodent T2D models^[288] but awaits clinical translation. Stem cell-derived islet transplantation achieved insulin independence in a single T2D patient in a first-in-human pilot study, representing a promising but preliminary proof-of-concept for treating end-stage secretory deficiency; larger controlled trials are needed to confirm efficacy and safety.^[303,304]. Therapeutic targeting of CEACAM1 or IDE to optimize hepatic insulin clearance remains at the preclinical stage^[80,159]. Microbiome modulation through precision dietary interventions, probiotic formulations, and postbiotic therapies targeting the SCFA-GLP-1-insulin axis represents a low-risk adjunctive strategy with growing human evidence^{[284,305,306]}. For clinical translation notes of this section, see Supplementary Materials.

Reframing T2D as a disease of metabolic aging

A fundamental but underappreciated dimension of the insulin lifecycle is its temporal relationship with aging. β-Cell deterioration, hepatic steatosis, peripheral insulin resistance, and chronic inflammation share mechanistic overlap with hallmarks of biological aging. The Najjar laboratory has recently shown that reduction of hepatic CEACAM1 levels and impaired insulin clearance mediate age-related hepatic steatohepatitis and fibrosis and that protecting CEACAM1 prevents the metabolic dysfunction and liver injury associated with aging and bestows a survival advantage^[307]. Reframing T2D as a disease of metabolic aging shifts therapeutic thinking toward upstream, disease-modifying intervention targeting aging processes that may simultaneously prevent the full spectrum of age-related metabolic dysfunction^{[101,308-310]}.

Senescent cells accumulate in metabolically stressed tissues (pancreatic islets, adipose tissue, liver, muscle) and are characterized by irreversible cell cycle arrest, apoptosis resistance, and secretion of a senescence-associated secretory phenotype (SASP), pro-inflammatory cytokines including IL-1β, IL-6, TNF-α, and MCP-1^[310]. The SASP creates a pro-inflammatory milieu that impairs β-cell function, deepens insulin resistance, promotes hepatic steatosis, and disrupts peripheral insulin signaling.

Cellular senescence as an upstream therapeutic target

Pancreatic β-cells are particularly susceptible to stress-induced senescence from glucotoxicity, lipotoxicity, inflammatory cytokines, and oxidative stress, which induce DNA damage and senescence mediated by p16INK4a (protein of 16 kilodaltons/inhibitor of cyclin-dependent kinase 4a) and p21CIP1 (protein of 21 kilodaltons/CDK-interacting protein 1)^[310,311]. Senescence may precede β-cell dedifferentiation or apoptosis, positioning it as a temporally earlier therapeutic target^[310,312]. The convergence between cellular senescence, SASP, and inflammatory pathways (NF-κB, JNK, IKKβ) establishes senescence as a mechanistic hub linking aging to the entire insulin life cycle^[313].

Senotherapeutics include senolytics (which eliminate senescent cells: dasatinib/quercetin, navitoclax, fisetin) and senomorphics [which suppress SASP: janus kinase (JAK)1/2 inhibitors, rapamycin, metformin]^[310,314]. While preclinical models demonstrate metabolic improvements, human evidence remains limited to small early-phase trials focused on safety and biomarker modulation rather than glycemic efficacy^{[310,315,316]}. Senotherapeutics should be positioned as a conceptual framework to guide future precision-based clinical trials rather than advocating near-term clinical adoption.

Sirtuin activators and NAD+ biology: bridging geroscience and diabetology

The sirtuin family (SIRT1-7) interfaces energy sensing, aging, and metabolic regulation. SIRT1 activates PGC-1α (mitochondrial biogenesis master regulator) and suppresses NF-κB inflammatory signaling. SIRT3 maintains oxidative phosphorylation and suppresses mitochondrial ROS, directly countering two of the principal drivers of β-cell failure and peripheral insulin resistance^[317-319]. Declining NAD+ with aging impairs sirtuin activity, contributing to senescence-inflammation and metabolic deterioration. NAD+ precursors (nicotinamide riboside, NMN) and sirtuin activators (SRT2104, resveratrol) have shown modest improvements in insulin sensitivity and inflammatory markers in early human trials, though larger trials with mechanistic endpoints are required to establish clinical significance^[320-322].

Geroscience applications in precision risk stratification

The advent of high-resolution precision medicine tools has created an unprecedented opportunity to implement the shift from reactive to preventive diabetes medicine. Polygenic risk scores (PRS) for T2D achieve sufficient discriminatory power to stratify risk decades before hyperglycemia. Variants clustering around insulin secretion loci (TCF7L2, KCNJ11) identify different biological subtypes than those around insulin resistance loci (IRS1, PPARγ, FTO), enabling biologically informed stratification to direct prevention toward individuals likely to benefit from specific interventions^[323,324].

Multi-cancer early detection (MCED) tests and epigenetic clocks illustrate the feasibility of population-level biological aging biomarkers. Epigenetic clocks provide quantitative surrogates of aging rate correlating with T2D incidence and mortality, potentially serving as actionable endpoints in senotherapy trials^[310]. Together, PRS and biological aging metrics enable stratification by metabolic phenotype and aging rate, facilitating individualized intervention timing and target selection^[310].

Integrating geroscience with diabetology: toward future precision-based and disease-modifying strategies

Current T2D management is reactive: treatment begins after hyperglycemia, targeting downstream manifestations of a years-long process. The geroscience-informed vision is substantively different: identify high-risk individuals through PRS and aging biomarkers, characterize the dominant life-cycle defect through molecular phenotyping, and deploy senescence-targeted or sirtuin-activating interventions to interrupt the causal chain between accelerated metabolic aging and clinical T2D before irreversible damage occurs^{[118,310,325]}.

Cellular senescence can be understood as a cross-cutting upstream mechanism degrading all insulin life cycle phases: SASP-mediated ER stress impairs β-cell biosynthesis; senescent β-cells lose secretory identity; hepatic senescence reduces insulin clearance; SASP cytokines drive inflammatory insulin resistance; adipose senescence amplifies dysfunctional adipokine secretion; and senescence-associated mitochondrial dysfunction accelerates insulin degradation. Cellular senescence is thus a unifying upstream regulator of insulin life cycle deterioration in metabolic aging^[310,326].

Translating geroscience into practice faces challenges: identifying optimal targets, establishing therapeutic windows (senescence serves physiological roles), developing biomarkers tracking target engagement, and designing adequately powered trials^[327]. For clinical translation notes of this section, see Supplementary Materials.