Insulin Receptors and Intracellular Ca 2 + Form a Double-Negative Regulatory Feedback Loop Controlling Insulin Sensitivity

Since the discovery of insulin and insulin receptors (IR) in the brain in 1978, numerous studies have revealed a fundamental role of IR in the central nervous system and its implication in regulating synaptic plasticity, long-term potentiation and depression, neuroprotection, learning and memory, and energy balance. Central insulin resistance has been found in diverse brain disorders including Alzheimer’s disease (AD). Impaired insulin signaling in AD is evident in the activation states of IR and downstream signaling molecules. This is mediated by Aβ oligomer-evoked Ca2+ influx by activating N-methyl-Daspartate receptors (NMDARs) with Aβ oligomers directly, or indirectly through Aβ-induced release of glutamate, an endogenous NMDAR ligand. In the present opinion article, we highlight evidence that IR and free intracellular Ca2+ concentration [Ca]i form a doublenegative regulatory feedback loop controlling insulin sensitivity, in which mitochondria play a key role, being involved in adenosine triphosphate (ATP) synthesis and IR activation. We found recently that the glutamate-evoked rise in [Ca]i inhibits activation of IR and, vice versa, insulin-induced activation of IR inhibits the glutamate-evoked rise in [Ca]i. In theory, such a double-negative feedback loop generates bistability. Thus, a stable steady state could exist with high [Ca]i and nonactive IR, or with active IR and low [Ca]i, but no stable steady state is possible with both high [Ca]i and active IR. Such a circuit could toggle between a high [Ca]i state and an active IR state in response to glutamate and insulin, respectively. This model predicts that any condition leading to an increase of [Ca]i may trigger central insulin resistance and explains why central insulin resistance is implicated in the pathogenesis of AD, with which glutamate excitotoxicity is a comorbid condition. The model also predicts that Open Peer Review


Introduction
Since the discovery of insulin 1 and insulin receptors (IR) 2 in the brain in 1978, numerous studies have revealed a fundamental role of IR in the central nervous system (CNS). IR-mediated signaling is implicated in the regulation of diverse functions in the CNS, including synaptic plasticity, long-term potentiation and depression, neuroprotection, learning and memory, and energy balance 3 . Central insulin resistance has been found in neurodegenerative diseases such as Alzheimer's disease (AD) 4,5 and Parkinson's disease (PD) 6 , stroke, and traumatic brain injury (TBI) 7 . Impaired insulin signaling in AD is evident in the activation states of IR and downstream signaling molecules 5 . Compared with control cases, insulin in AD brains induced 24-58% less activation at the level of IR and 90% less activation of insulin receptor substrate 1 (IRS-1) 5 . It has been presumed 5 that the inhibition of IR activation is mediated by Aβ oligomer-triggered Ca 2+ influx, in part by activating N-methyl-D-aspartate receptors (NMDARs) 8 , followed by a rise in Akt1 pS 473 9 , which can inhibit insulin-induced IR activation through Thr phosphorylation of the IR β subunit 10 . Aβ oligomers may activate the NMDAR-gated Ca 2+ influx directly 11 or indirectly through the intermediate release of glutamate, a ligand of NMDAR 11-15 . This suggests that the rise in intracellular free Ca 2+ concentration ([Ca 2+ ] i ), evoked by either Aβ oligomers or glutamate, leads to dysfunctional activation of IR in AD. In the present opinion article, we highlight evidence that IR and [Ca 2+ ] i form a double-negative regulatory feedback loop controlling insulin sensitivity, and mitochondria have a key role in this feedback loop, being involved in adenosine triphosphate (ATP) synthesis and IR activation.

Glutamate-evoked rise in [Ca 2+ ] i causes inhibition of IR signaling
Glutamate serves as the major excitatory neurotransmitter in the CNS. Its excessive accumulation in a synaptic cleft can trigger excitotoxicity, a pathologic process leading to neuronal cell death. Glutamate-induced activation of the NMDARgated Ca 2+ influx is generally considered central to the development of excitotoxicity 16 . Prolonged glutamate exposure causes a rapid initial increase in the [Ca 2+ ] i , followed by a larger secondary [Ca 2+ ] i increase concomitant with a decrease in the mitochondrial inner membrane potential (ΔΨ m ) 17-19 . We recently found that on Ca 2+ -induced mitochondrial depolarization, insulin induced 48% less activation of IR (assessed by pY 1150/1151 ) compared with control 20 . Earlier, we showed that a decrease in ΔΨ m can abrogate IR activation 18 , since the ΔΨ m -dependent mitochondrial signal at complex II is involved in the activation of IR in neurons 21-23 . Thus, the glutamate-evoked increase in [Ca 2+ ] i , followed by the drop in ΔΨ m , leads to the inhibition of insulin-induced activation of IR.

Insulin prevents glutamate-evoked rise in [Ca 2+ ] i
Normally, the NMDAR-gated Ca 2+ influx is counterbalanced with Ca 2+ efflux, which is governed by plasma membrane Ca 2+ ATPase and the Na + /Ca 2+ exchanger (NCX) 24,25 . NCX-mediated Ca 2+ efflux is also ATP-dependent, since NCX exchanges one Ca 2+ for three Na + , and the three Na + are then pumped out by the Na + /K + ATPase at the expense of one ATP. In excitotoxicity, prolonged stimulation with glutamate leads to ATP depletion and an abnormal rise in [Ca 2+ ] i , since the massive Ca 2+ influx is no longer counterbalanced by Ca 2+ efflux 26 . Therefore, maintenance of ATP production is crucial for preventing the rise in [Ca 2+ ] i in excitotoxicity. We found recently that pre-treatment with insulin prevents neurons from glutamate-evoked ATP depletion due to its protective effect on spare respiratory capacity (SRC), a measure that relates to the amount of extra ATP that can be produced via oxidative phosphorylation in case of increased energy demand 19 . The effect of insulin on SRC relates to its action on mitochondrial metabolism. It has long been known that the tricarboxylic acid cycle is the intracellular site of insulin action and that insulin acutely stimulates succinate oxidation at mitochondrial complex II 26,27 . Succinate oxidation at mitochondrial complex II has been identified recently as the main source of SRC 28 . In line with this, insulin prevented the glutamateevoked rise in [Ca 2+ ] i in our experiments with glutamate excitotoxicity 19 .

IR and [Ca 2+ ] i form a double-negative feedback loop controlling insulin sensitivity
Collectively, this evidence suggests that a double-negative regulatory feedback loop exists between IR and [Ca 2+ ] i . The glutamate-evoked rise in [Ca 2+ ] i inhibits activation of IR and, vice versa, insulin-induced activation of IR inhibits the glutamate-evoked rise in [Ca 2+ ] i (Figure 1a).
In theory, a double-negative feedback loop generates bistability 29 . Thus, a stable steady state could exist with high [Ca 2+ ] i and nonactive IR (Figure 1b), or with active IR and low [Ca 2+ ] i ( Figure 1c), but no stable steady state is possible with both high [Ca 2+ ] i and active IR. Such a circuit could toggle between a high [Ca 2+ ] i state and an active IR state in response to glutamate and insulin, respectively. This double-negative feedback loop model predicts that any condition leading to an increase in [Ca 2+ ] i may trigger insulin resistance. It appears to explain why central insulin resistance is implicated in the pathogenesis of disorders such as AD 4,5 , PD 6 , stroke, and TBI 7 , with which glutamate excitotoxicity is a comorbid condition 30 . The model also predicts that any intervention aiming to prevent Ca 2+ influx of or enhance efflux of Ca 2+ from neurons, thereby maintaining low [Ca 2+ ] i , may be useful for treating central insulin resistance. Given that Ca 2+ efflux is ATP-dependent, any intervention directed to enhance ATP production in neurons may be especially useful to improve insulin sensitivity in the brain.

Data availability
No data are associated with this article. expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.
Although insulin receptors are ubiquitously expressed in different cell types of the CNS, these cell types rely on different mechanisms for ATP production. Glial cells predominantly use glycolytic pathways in the cytoplasm whereas neurons rely on oxidative phosphorylation in the mitochondria. (For review see Pellerin and Magistretti, 2012 2 ). Accordingly, the cellular localization of ATP production should be taken into consideration. Furthermore, do the authors believe different cell type specific mechanisms exist that their model should take into consideration? In several places, the authors discuss the role of amyloid on activation of NMDA receptors and the subsequent intracellular Calcium increase. However, the authors have not taken into account how amyloid binding can also elicit glutamate release from either α7nAChRs (see Hascup  Additionally, the manuscript could be improved if a discussion on how their model might vary across different CNS disorders such as AD, Parkinson's TBI, etc. in relation to healthy functional activity.
The authors neglect to take into account factors that also play a role in modulating insulin signaling. These include: Inflammation. This is prominent in numerous disease states and can negatively impact insulin signaling, while exacerbating mechanisms associated with multiple neurodegenerative disorders.
○ Inhibitory feedback regulation. Insulin signaling is tightly controlled to prevent perturbations in metabolism as well as control the specificity of the signal on multiple downstream effectors. Several phosphatases are responsible for this, not just at the receptor, but also on effector enzymes.The current model does not take into consideration this tightly controlled feedback loop. Minor concerns: The glutamate pathway in Figure 1 should have a different color scheme to make it easier to differentiate from Calcium concentration.  system as a whole. Insulin resistance is the term that currently applied to any of the biological actions of insulin and, therefore, is too broad to be discussed in terms of models that may predict the system behavior. The proposed model in the opinion article is not oversimplified, but takes into consideration only link between [Ca 2+ ]i and the stage of activation of insulin receptor kinase (i.e Tyr1150/1151 phosphorylation), the earliest step in insulin action that precedes all other signaling events and effects of insulin.
In our opinion article we selected only two measurable parameters, namely [Ca 2+ ]i and insulin receptor activation state defined as Tyr1150/1151 phosphorylation, but not downstream molecules of IR signaling pathway such as IRS-1 or others. Therefore, our opinion relates only to the activation of insulin receptor, and not to downstream molecules or events. This approach relates directly to insulin sensitivity, since the activation of the receptor with insulin is the only stage at which insulin sensitivity can be measured directly. In our studies [references 19 and 20] that underlie the our opinion we used glia-free cortical neurons and directly measured ATP levels.
Our opinion relates to link between [Ca 2+ ]i and insulin receptor activation state, and amyloid-NMDA relationship are out of scope of our opinion.
Our opinion is limited to only insulin receptor activation, but not to more broad "insulin signaling". According to current knowledge, inflammation affects insulin signaling, but downstream of insulin receptor at IRS-1 level. This is out of scope of our opinion.
Our opinion does not relate to any signaling events downstream of insulin receptor, such phosphatase action and receptor internalization.
The insulin regulated internalization of insulin receptors has been shown to require autophosphorylation of all three regulatory tyrosines 1146, 1150, and 1151. Carpentier JL et al. Two steps of insulin receptor internalization depend on different domains of the betasubunit. J Cell Biol. 1993 Sep;122(6):1243-52. doi: 10.1083/jcb.122.6.1243. Thus, the internalization occurs after the activation of IR and, therefore, is out of the scope of our opinion.
According to the model, insulin inhibits rise of [Ca 2+ ]i independently of the insulin source. Therefore, during periods of insulin transport to the brain, the rise of [Ca 2+ ]i , e.g. glutamate-evoked, would be diminished, independently on whether it comes from pancreas or administered intranasally.
We did not propose the same mechanism for IGF-1 in our opinion aticle, since we have no supportive evidence. However, it is likely that there is a link between [Ca 2+ ]i and activation state of IGFR1.
In a conclusion, the model proposed in the opinion article relates only to relationship beteen two measurable parameters, namely intracellular Ca2+ concentrations and insulin receptor activation state. Therefore, all other downstream elements of insulin signaling system and factors affecting the downstream elements are out of scope of this opinion. The opinion is completely based on our experimental results previously published, references 19 and 20 of the opinion article.