Type 2 diabetes – unmet need, unresolved pathogenesis, mTORC1-centric paradigm

The classical pathophysiology paradigm of the pre-diabetes stage of T2D maintains that peripheral insulin resistance comes first, followed by compensatory hyperinsulinemia due to increase in beta cells mass and activity [48]. Indeed, peripheral insulin resistance has been verified in normoglycemic first-degree relatives of diabetes parents, having a very high life-time risk of developing diabetes [49]. However, these individuals present as well with fasting hyperinsulinemia, making it impossible to decide between insulin resistance and hyperactive beta cells as primary driver. Also, claiming resistance as the primary driver calls for a humoral and/or neuronal agent(s) that may mediate between peripheral resistance and beta cells mass and function during the early normoglycemic normolipemic pre-diabetes stage. However, in spite of repeated efforts, no such mediators have yet been definitively identified.

Alternatively, one may argue for beta cells hyperactivity as a primary driver, at least under conditions of nutrient excess, resulting in primary hyperinsulinemia followed by downregulation of peripheral insulin receptors and/or their signaling pathway [50]. Indeed, acute or chronic increase in plasma insulin in healthy normoglycemic subjects results in significant decrease in insulin-stimulated glucose disposal [51, 52], implying that primary hyperinsulinemia may drive insulin resistance. Also, a variety of high-fat diet (HFD) feeding studies in rats and mice demonstrate fasting hyperinsulinemia at the earliest feeding stage prior to any detectable increase in plasma glucose as a surrogate for insulin resistance (50). However, downregulation of insulin receptors by 90% may still allow for adequate insulin signaling due to spare insulin receptors [53], implying that peripheral insulin resistance must further involve post receptor defects [54, 55]. However, no agents have yet been identified which may drive post receptor defects due to hyperinsulinemia. Hence, the question of which comes first in driving the pathogenesis of the pre-diabetes stage of T2D still remains unresolved.

Insulin binding to the insulin receptor (IR) results in forming an IR / insulin receptor substrate (IRS) node attached to the plasma membrane. Phosphorylation of IRS tyrosines by the IR tyrosine kinase results in binding sites for the PI3K and its activation, followed by phosphorylating phosphatidylinositol 4,5-bisphosphate (PIP2) to yield the membrane bound phosphatidylinositol 3,4,5-triphosphate (PIP3). PIP3 binds and activates PDK1 which phosphorylates Akt/PKB(Thr308). Further phosphorylation of Akt/PKB(Ser473) by PIP3-activated mTORC2 results in the fully activated phospho-Akt(Thr308, Ser473). Signal termination is achieved by the PIP3 phosphatase PTEN or the protein phosphatases PP2A and PHLPP which dephosphorylate Akt(Thr308) and Akt(Ser473), respectively. The IR-Akt transduction pathway controls carbo-lipid metabolism in response to insulin (Fig. 3). Thus, phosphorylation of glycogen synthase kinase 3 (GSK3) by Akt results in its suppression and in activation of liver and muscle glycogen synthase; phosphorylation of AS160/TBC1D4 by Akt results in muscle and adipose glucose uptake by translocating GLUT4 to the plasma membrane; and phosphorylation of the transcription factor FOXO1(Thr24,Ser256) by Akt results in its cytosolic sequestration and transcriptional suppression of hepatic gluconeogenesis and adipose fat lipolysis. Hence disruption of the IR-Akt transduction pathway resists insulin action and results in unrestrained hyperglycemia and adipose lipolysis [56].

In addition to the IR/IRS node that drives Akt activation, binding of insulin to IR results in generating the IR/Shc/Grb2/SOS node that drives the activation of the Ras/RAF/MEK/Erk1,2/p90RSK transduction pathway. However, Erk1,2/p90RSK are considered to transduce mitogenic signaling, in contrast to Akt/PKB which controls carbo-lipid metabolism (Fig. 3).

Whereas the IR-Akt transduction pathway may account for insulin resistance in the glycemic context of T2D, other disease aspects of T2D surprisingly present full response to insulin, implying “selective insulin resistance” [57]. The first example of an “insulin paradox” was concerned with hepatic lipogenesis, namely, de novo fatty acids biosynthesis followed by their esterification to yield lipids. Lipogenesis partially accounts for the NAFLD disease of T2D, and proved to be fully responsive to insulin in T2D animal models and patients, in face of resistance to insulin in the glycemic context [57]. Another example has been realized in studying VLDL production in human subjects or animal models. Since insulin suppresses VLDL production [58], IR knockout predicts dyslipidemia. However, IR knockout results in hyperglycemia as expected, but the dyslipidemia of T2D is avoided [59, 60], implying an apparent selective role of the IR in controlling hyperglycemia as contrasted with dyslipidemia.

Several explanations were offered to solve the gluconeogenesis / lipogenesis insulin paradox. These have argued for insulin-independent lipid synthesis [61, 62], or hepatic / adipose selective sensitivity to insulin [63, 64], or extrahepatic insulin effects (hypothalamic, adipose, pancreatic alpha-cell) [65], or selective / differential activity of Akt in phosphorylating its downstream substrates [66], or hepatic zone specificity of carbo-lipid metabolism [67]. However, all were limited to the glycemic-steatosis paradox, while failing to realize that, except of glycemic control, response to insulin is essentially the rule in T2D, rather than the exception. Indeed, the diabesity of T2D reflects insulin-responsive body weight gain, in face of resistance to insulin in the glycemic and lipolysis context [68‐70]. Also, the hypertension disease of T2D reflects insulin-responsive sympathetic activity [71, 72], renal sodium reabsorption [73] and endothelial vasoconstriction [74]. Similarly, T2D hyperuricemia reflects suppression by insulin of renal uric acid clearance [75]. Also, the macrovascular disease of T2D reflects insulin-responsive dyslipidemia [59], hypertension [70‐74], diabetic cardiomyopathy [76], vascular smooth muscle cells (VSMC) proliferation, endothelial dysfunction and prothrombosis [77]. Of note, the same tissue may present resistance to insulin in the glycemic context together with response to insulin in a non-glycemic aspect. Hence, the interplay between response and resistance to insulin in shaping the pathogenesis of T2D still remains unresolved, calling for an insulin-dependent, context-dependent, resistance-response unifying paradigm.

Insulin resistance has been proposed to be driven by ‘mitochondrial dysfunction’ with decrease in mitochondria content, size, biogenesis, electron flux and oxidative phosphorylation, and increased oxidative stress [78, 79]. Indeed, muscle insulin resistance with concomitant mitochondrial dysfunction has been verified in diabetes patients, obese prediabetic subjects, insulin-resistant lean non-diabetic off-springs of T2D parents, insulin-resistant non-diabetic elderly and T2D animal models, implying a putative causal relationship [62]. Mitochondrial dysfunction has been proposed to result in suppression of fatty acid oxidation, followed by their esterification into intra-hepatic and intra-myocellular lipids (IMCL) [80]. Intracellular diacylglycerols and oxidative stress are proposed to activate novel PKCs (nPKC) and/or JNK respectively, resulting in serine/threonine phosphorylation of IRS1. Serine/threonine phosphorylation of IRS1 suppresses IRS tyrosine phosphorylation by the IR tyrosine kinase, resulting in disruption of the IR-Akt transduction pathway [62]. Indeed, hepatic steatosis is associated with insulin resistance [81], whereas increased beta-oxidation results in hepatic sensitivity to insulin [82]. However, the IMCL narrative still leaves unresolved the athlete paradox, whereby aerobic exercise is reported to result in increase in IMCL while counteracting muscle insulin resistance [83]. Moreover, the mitochondrial dysfunction narrative still leaves unresolved the insulin sensitizers’ riddle, whereby anti-diabetic agents that promote sensitization to insulin e.g., biguanides / metformin, thiazolidinedioones / pioglitazone, high-dose aspirin, berberine and others, are all reported to act as mitochondrial complex I inhibitors [84], implying that ‘mitochondrial dysfunction’ is more of a solution rather than a problem in the T2D context [85]. Also, mitochondrial ROS is reported by some (but refuted by others) to enhance insulin activity and/or counteract resistance to insulin, rather than exacerbate resistance due to putative oxidative stress [86‐89]. Hence, the upstream driver(s) and respective downstream targets that drive insulin resistance by disrupting the IR-Akt transduction pathway still remain unresolved.

T2D is considered to follow a two-stage pre-diabetes / diabetes sequence, affixed by a demarcating plasma glucose value of 126 mg/dL, with annual conversion rate of 5–10% [1, 47]. The pre-diabetes stage is characterized by an increase in beta cells mass and activity, resulting in hyperinsulinemia which offsets peripheral insulin resistance. The pre-diabetes stage is proposed to be followed by progressive beta cells failure and an established diabetes status [1] (Fig. 2), ascribed to low innate beta cells mass [90] and/or stunned beta cells [91]. Beta cells failure may culminate in de-differentiation, mitochondrial dysfunction, ROS production, unfolded protein response (UPR) / endoplasmic reticulum (ER) stress and islet amyloid polypeptide (IAPP) overexpression, followed by glucolipotoxicity and apoptosis [92, 93]. The two-stage paradigm implies an apparent turning point in T2D pathogenesis and treatment policy. However, concomitantly with beta cells hyperactivity, the pre-diabetes stage presents loss of the first-phase of insulin secretion [94] and about 50% loss of beta cell number due to apoptosis [95, 96]. Hence, the pre-diabetes stage presents functional-structural failures which are carried over to, and progress during the hyperglycemic diabetes stage. Thus, similar to peripheral insulin resistance which prevails throughout the course of T2D, beta cells dysfunction and loss progress throughout the pre-diabetes / diabetes continuum, calling for a pathogenic paradigm that may account for beta cells concomitant hyperactivity and failure.

The Mammalian target of rapamycin complex I (mTORC1) controls growth and metabolism in response to nutrients, energy and redox status. The mTORC1 complex consists of the mTOR kinase, the core subunits Raptor and mLST8, and the PRAS40 and DEPTOR inhibitory subunits. mTORC1 activation is enabled by two converging arms, namely, mTORC1 attraction to the lysosomal surface, and its activation by the constitutively-attached lysosomal Rheb. These two activation arms are independent, and each is affected by specific upstream effectors.

Binding of mTORC1 to the lysosomal surface via its Raptor subunit is mediated by the lysosomal Rag.GTPase heterodimers, RagA/B and RagC/D. Binding of mTORC1 requires the RagA/B.GTP and RagC/D.GDP conformation, being determined by lysosomal Ragulator, acting as specific GTP exchange factor (GEF) of Rags, and by lysosomal Gator1 and Folliculin-FNIP2 which function as GTPase activation proteins (GAP) for RagA/B and RagC/D, respectively. The Rag arm is affected by a variety of cytosolic and intra-lysosomal amino acids (e.g., leucine, arginine, glutamine) which bind to specific amino acid receptors (e.g., Sestrin, Castor1, SLC38A9) that interact, directly or indirectly, with the GAPs and/or GEFs that determine Rags conformation. Of importance, the Rag arm may also be activated by glucose [100]. Activation by glucose is mediated by binding of glucose-derived aldolase-bound fructose 1,6-bisphosphate (FBP) to Ragulator, resulting in RagA/B.GTP and its association with mTORC1 [101, 102]. Hence, the Rag arm mediates mTORC1 activation by amino acids and/or glucose, while suppressing mTORC1 activity in their absence (Fig. 4).

Once attracted to the lysosome, mTORC1 may be activated by lysosomal Rheb.GTP. Rheb may cycle between its active GTP and inactive GDP forms. The GTPase activity of Rheb is modulated by the Tuberous sclerosis complex (TSC), whereby its TSC2 subunit acts as a specific GAP of Rheb. The Rheb arm is activated upon suppressing TSC GAP and/or by suppressing TSC lysosomal attachment. TSC activity may be inhibited by specific phosphorylation of TSC2(Ser939,981,1130,1132,Thr1462) by Akt, resulting in a canonical transduction pathway that leads from insulin/IR to Rheb.GTP and mTORC1 activation [103]. mTORC1 activation by Akt is further complemented by phosphorylation of PRAS40(Thr246) by Akt, resulting in de-repressing mTOR [104]. Most importantly, TSC activity may similarly be inhibited by specific phosphorylation of TSC2(Ser540,644,1798) by activated Erk1,2 and/or p90RSK [105], resulting in an alternative transduction pathway that leads from insulin/IR to Rheb.GTP and mTORC1 activation. The two alternative pathways imply a functional redundancy of the Akt and the Erk1,2/RSK effectors in mediating mTORC1 activation by insulin/IR as well as by other growth factors / RTKs. Also, TSC activity may further be inhibited by specific phosphorylation of TSC2(Ser487,511) by activated IkapaB kinase (IKK) [106], resulting in mTORC1 activation by pro-inflammatory TNFa/TNFR, IL1/IL1R and LPS/TLR4. In contrast, TSC GAP may be activated by specific phosphorylation of TSC2(Ser1345,Thr1227) by AMPK / GSK3beta [107], resulting in suppressing mTORC1 activity by metabolic stress. Suppression of mTORC1 activity by AMPK is further complemented by phosphorylation of Raptor(Ser792) by AMPK [108], resulting in disrupting mTORC1 composition. Similarly to AMPK, multiple different metabolic stresses e.g., hypoxic, deoxy glucose, hyperosmotic, pH, may promote lysosomal TSC2 and/or increase its stability, resulting in inhibition of mTORC1 activity [109, 110]. Hence, mTORC1 activation is enabled by two converging hits, its lysosomal binding driven by nutrients (e.g., glucose, amino acids), and its activation by lysosomal Rheb.GTP driven by growth factors (e.g., insulin) (Fig. 4). In line with that, TSC1,2 knockout and constitutive RagA.GTP result in constitutive activation of mTORC1 under fasting conditions.

mTORC1 controls growth and metabolism by phosphorylating and/or affecting its downstream targets S6K1, 4EBP, CRTC2, lipin, ATF4, HIF1a, PPARg, PPARa, ULK1, TFEB and others [97, 99]. Phosphorylation of S6K1 and 4EBP results in ribosome biogenesis and in initiating CAP-dependent mRNA translation. Phosphorylation of CRTC2 and lipin results in activating the transcription factor SREBP and in driving lipogenesis and lipid synthesis. Phosphorylation of ATF4 and S6K1 results in purine and pyrimidine biosynthesis, respectively. Activation of HIF1a and SREBP transcription factors results in enhancing glycolysis and the pentose shunt. PPARg activation by mTORC1 promotes adipogenesis, while PPARa inhibition by mTORC1 suppresses fatty acid oxidation and ketogenesis. mTORC1 controls G1/S transition and G2/M progression. Most importantly, phosphorylation of the ULK1 kinase, Atg13 and the TFEB transcription factor by mTORC1 blocks autophagy and lysosome biogenesis. Not all immediate downstream targets of mTORC1 have presently been verified. However, downstream targets of mTORC1 may be inferred by their response to mTORC1 inhibitors (below).

Some mTORC1 activities in the T2D context are transduced by interacting with mTORC2. The mTORC2 complex consists of the mTOR kinase, mLST8, Rictor (instead of mTORC1 Raptor) mSin1, Deptor and Protor [97, 99]. mTORC2 phosphorylates and activates Akt(Ser473), the AGC protein kinases (PKA, PKG, PKC) and Serum and glucocorticoid kinase (SGK). mTORC2 activity is inhibited by phosphorylation of its mSin1 and Rictor subunits by S6K1, thus forming a negative feedback loop whereby activation of mTORC1 by Akt results in inhibition of Akt due to suppression mTORC2 activity by S6K1.

mTORC1 activity may be blocked by rapamycin and respective rapalogs. Rapamycin forms a complex with the FK506-binding protein (FKBP12), and the complex binds to the FRB domain of mTOR, resulting in allosteric inhibition of mTORC1. Hence, response to rapamycin may imply an apparent involvement of mTORC1. However, chronic exposure to rapamycin may also inhibit mTORC2, resulting in a false positive inference of mTORC1 involvement [111]. Also, not all mTORC1 downstream targets are affected by rapamycin (e.g., 4EBP), and lack of response to rapamycin may result in a false negative inference of mTORC1 involvement [112]. mTORC1 and mTORC2 may both be inhibited by mTOR kinase inhibitors (e.g., Torin).

Peripheral resistance to insulin in the glycemic context is proposed to be driven by disruption of the IR-Akt transduction pathway by hyperactive mTORC1 and its downstream S6K1 in liver, muscle and adipose tissue, namely, the main organs that control glucose production and its utilization. Thus, phosphorylation of IRS1(Ser307, 1101) by hyperactive S6K1, and phosphorylation of IRS1(Ser636/639, 422) by hyperactive mTORC1, result in suppressing IRS tyrosines phosphorylation by the IR tyrosine kinase, followed by IRS ubiquitination and degradation [113‐115]. Also, phosphorylation of GRB10 by hyperactive mTORC1 results in disrupting IR/IRS by phospho-GRB10 [116, 117]. The IR-Akt transduction pathway is further disrupted by inhibition of Akt(Ser473) phosphorylation by mTORC2, due to suppression of mTORC2 kinase activity by hyperactive S6K1 [118, 119] (Fig. 5). Disruption of the IR-Akt pathway by hyperactive mTORC1/S6K1 results in liver and muscle glycogenolysis, liver gluconeogenesis, GLUT4 sequestration and unrestrained hyperglycemia. In line with that, genetic deletion of S6K1 protects mice from HFD-induced diabetes [120, 121]. Hence, resistance to insulin in the glycemic context is proposed to be congruent with mTORC1/S6K1 hyper activation.

Disruption of the IR-Akt transduction pathway by hyperactive mTORC1 may still allow for sustained hyper activation of mTORC1 by insulin, being mediated by the IR/Ras/Raf/MEK/Erk/p90RSK/TSC/Rheb/mTORC1 transduction pathway, implying redundancy of IR-Akt and IR-Erk/RSK in activating mTORC1 [122, 123] (Figs. 4, 5). Moreover, the reciprocal relationship between the IR-Akt and the IR-Erk/RSK pathways, due to inhibitory phosphorylation of Raf(Ser259) by activated Akt [124, 125], implies enhancement of the IR-Erk/RSK activity upon inhibiting the IR-Akt pathway by hyperactive mTORC1. Hence, in face of resistance to insulin in the glycemic context, insulin-driven Erk/RSK may transduce a variety of mTORC1-mediated disease aspects of T2D (e.g., beta cell failure, obesity, NAFLD, dyslipidemia, hypertension, diabetes macro- and micro-vascular disease) as outlined below (Fig. 5). Indeed, IR knockout results in hyperglycemia, but also in protecting from non-glycemic diseases of T2D [59], implying an obligatory role for insulin and IR in driving the non-glycemic diseases of T2D.

Diabetic retinopathy consists of non-symptomatic non-proliferative first stage retinopathy, followed by second stage proliferative neovascularization. The proliferative stage consists of extensive angiogenesis of leakage-prone blood vessels [163] driven by HIF1a-induced VEGF. HIF1a may be induced under normoxic conditions by hyperactive mTORC1 [164, 165].

Diabetic peripheral neuropathy is driven by decrease in density of small un-myelinated or thinly-myelinated intra epidermal nerve fibers (IENF) which originate in the dorsal nerve root ganglia and mediate pain, temperature sensation or autonomic functions. Hyperactive mTORC1 is reported to interfere with synaptic plasticity and to induce chronic neuropathy [166, 167]. In line with that, suppression of mTORC1 activity is reported to result in anti-nociceptive effects in experimental models of inflammatory and neuropathic pain [166‐168].