Mitochondrial H2O2 as an enable signal for triggering autophosphorylation of insulin receptor in neurons

Background Insulin receptors are widely distributed in the brain, where they play roles in synaptic function, memory formation, and neuroprotection. Autophosphorylation of the receptor in response to insulin stimulation is a critical step in receptor activation. In neurons, insulin stimulation leads to a rise in mitochondrial H2O2 production, which plays a role in receptor autophosphorylation. However, the kinetic characteristics of the H2O2 signal and its functional relationships with the insulin receptor during the autophosphorylation process in neurons remain unexplored to date. Results Experiments were carried out in culture of rat cerebellar granule neurons. Kinetic study showed that the insulin-induced H2O2 signal precedes receptor autophosphorylation and represents a single spike with a peak at 5–10 s and duration of less than 30 s. Mitochondrial complexes II and, to a lesser extent, I are involved in generation of the H2O2 signal. The mechanism by which insulin triggers the H2O2 signal involves modulation of succinate dehydrogenase activity. Insulin dose–response for receptor autophosphorylation is well described by hyperbolic function (Hill coefficient, nH, of 1.1±0.1; R2=0.99). N-acetylcysteine (NAC), a scavenger of H2O2, dose-dependently inhibited receptor autophosphorylation. The observed dose response is highly sigmoidal (Hill coefficient, nH, of 8.0±2.3; R2=0.97), signifying that insulin receptor autophosphorylation is highly ultrasensitive to the H2O2 signal. These results suggest that autophosphorylation occurred as a gradual response to increasing insulin concentrations, only if the H2O2 signal exceeded a certain threshold. Both insulin-stimulated receptor autophosphorylation and H2O2 generation were inhibited by pertussis toxin, suggesting that a pertussis toxin-sensitive G protein may link the insulin receptor to the H2O2-generating system in neurons during the autophosphorylation process. Conclusions In this study, we demonstrated for the first time that the receptor autophosphorylation occurs only if mitochondrial H2O2 signal exceeds a certain threshold. This finding provides novel insights into the mechanisms underlying neuronal response to insulin. The neuronal insulin receptor is activated if two conditions are met: 1) insulin binds to the receptor, and 2) the H2O2 signal surpasses a certain threshold, thus, enabling receptor autophosphorylation in all-or-nothing manner. Although the physiological rationale for this control remains to be determined, we propose that malfunction of mitochondrial H2O2 signaling may lead to the development of cerebral insulin resistance.

Results: Experiments were carried out in culture of rat cerebellar granule neurons. Kinetic study showed that the insulin-induced H 2 O 2 signal precedes receptor autophosphorylation and represents a single spike with a peak at 5-10 s and duration of less than 30 s. Mitochondrial complexes II and, to a lesser extent, I are involved in generation of the H 2 O 2 signal. The mechanism by which insulin triggers the H 2 O 2 signal involves modulation of succinate dehydrogenase activity. Insulin dose-response for receptor autophosphorylation is well described by hyperbolic function (Hill coefficient, n H , of 1.1±0.1; R 2 =0.99). N-acetylcysteine (NAC), a scavenger of H 2 O 2 , dose-dependently inhibited receptor autophosphorylation. The observed dose response is highly sigmoidal (Hill coefficient, n H , of 8.0±2.3; R 2 =0.97), signifying that insulin receptor autophosphorylation is highly ultrasensitive to the H 2 O 2 signal. These results suggest that autophosphorylation occurred as a gradual response to increasing insulin concentrations, only if the H 2 O 2 signal exceeded a certain threshold. Both insulin-stimulated receptor autophosphorylation and H 2 O 2 generation were inhibited by pertussis toxin, suggesting that a pertussis toxin-sensitive G protein may link the insulin receptor to the H 2 O 2 -generating system in neurons during the autophosphorylation process.

Conclusions:
In this study, we demonstrated for the first time that the receptor autophosphorylation occurs only if mitochondrial H 2 O 2 signal exceeds a certain threshold. This finding provides novel insights into the mechanisms underlying neuronal response to insulin. The neuronal insulin receptor is activated if two conditions are met: 1) insulin binds to the receptor, and 2) the H 2 O 2 signal surpasses a certain threshold, thus, enabling receptor autophosphorylation in all-or-nothing manner. Although the physiological rationale for this control remains to be determined, we propose that malfunction of mitochondrial H 2 O 2 signaling may lead to the development of cerebral insulin resistance.

Background
Insulin receptor is a member of the receptor tyrosine kinase family. Upon insulin binding to the extracellular αsubunits, the receptor undergoes rapid autophosphorylation at three specific tyrosine residues within the activation loop of the cytoplasmic β-subunits [1,2], resulting in more than a 200-fold increase in receptor tyrosine kinase activity [3]. Therefore, the autophosphorylated receptor is regarded as fully activated [4]. Research conducted over 30 years ago revealed that cells generate hydrogen peroxide (H 2 O 2 ) in response to insulin stimulation [5,6]. Evidence from several studies supports the hypothesis that the main role of insulin-induced H 2 O 2 is inhibition of protein tyrosine phosphatases (PTPs), which otherwise dephosphorylate the autophosphorylated insulin receptor [7][8][9]. According to this theory, H 2 O 2 prolongs the duration of time for which the insulin receptor remains active, rather than directly influence receptor activation. Additionally, exogenous H 2 O 2 has been shown to facilitate receptor autophosphorylation in immunoprecipitates of the insulin receptor in the presence of phosphate donors [10,11]. The obvious independence of this effect on intracellular PTPs suggests that H 2 O 2 also participates in insulin receptor activation. Insulin receptors are widely distributed in the brain, where they play roles in synaptic function, memory formation, and neuroprotection [12][13][14]. The neuron-specific isoform A is the predominant insulin receptor type in the brain. Isoform A is generated from alternative splicing and differs from its peripheral counterpart (isoform B) in some notable respects, such as higher affinity for insulin and absence of negative cooperativity in insulin binding [15,16]. Earlier studies by our group demonstrated that neurons generate H 2 O 2 in response to insulin stimulation [17]. This H 2 O 2 is derived from the mitochondrial respiratory chain and plays a role in insulin receptor autophosphorylation. However, the kinetic characteristics of the H 2 O 2 signal and its functional relationships with the insulin receptor during autophosphorylation in neurons remain to be clarified. In the current investigation, these issues have been explored as an extension of our previous study.

Results and discussion
Insulin dose-response for receptor autophosphorylation is well described by hyperbolic function We characterized insulin-stimulated receptor autophosphorylation in a primary culture of rat cerebellar granule neurons (CGN). The insulin dose-response curve for the autophosphorylation process is depicted in Figure 1A. Fitting the curve to the Hill equation generated ED 50 of 16.3±2.2 nM and Hill coefficient, n H , of 1.1±0.1 (R 2 =0.99), indicating that this process in neurons is described by a classic hyperbolic function. These results suggest that the insulin dose-response for receptor autophosphorylation in neurons is gradual and not switch-like.

The H 2 O 2 signal precedes receptor autophosphorylation during insulin stimulation
To determine the temporal relationship between receptor autophosphorylation and H 2 O 2 generation during insulin stimulation, we compared the kinetics of insulininduced autophosphorylation and H 2 O 2 production in CGN. The time-course of receptor autophosphorylation   Figure 1 Relationship between H 2 O 2 signal and receptor autophosphorylation in neurons stimulated with insulin. (A) Insulin doseresponse for receptor autophosphorylation in CGN exposed to insulin for 10 min (black triangles, mean±SEM of 5 to 9 cultures, *P<0.05 vs. control). (B) Time course of receptor autophosphorylation in CGN exposed to 100 nM insulin (black triangles, mean±SEM of 3 to 4 cultures, *P<0.05 vs. baseline). (C) N-acetylcysteine dose-response for receptor autophosphorylation in CGN exposed to 100 nM insulin for 10 min (black triangles, mean±SEM of 3 to 7 cultures, *P<0.05 vs. 100 nM insulin). (D) Left Y axis: time courses of H 2 O 2 efflux from CGN exposed to vehicle (white squares, mean of 3 culture dishes) or 100 nM insulin (red squares, mean of 10 culture dishes). Right Y axis: first time derivative (rate) of H 2 O 2 efflux from CGN exposed to vehicle (black line) or 100 nM insulin (red line). (E) Areas under curves (AUC) for 30-s periods of H 2 O 2 efflux from CGN exposed to vehicle (white columns, mean±SEM, n=3) or 100 nM insulin (red columns, mean±SEM, n=10, *P<0.05 vs. control). (F) Time courses for insulin-stimulated H 2 O 2 efflux and receptor autophosphorylation. Left Y axis: time course of receptor autophosphorylation in CGN exposed to 100 nM insulin (black triangles, mean±SEM of 3 to 4 cultures). Right axis: first time derivative (rate) of H 2 O 2 efflux from CGN exposed to 100 nM insulin (red line, mean of 10 culture dishes).
in response to insulin stimulation is depicted in Figure 1B. Autophosphorylation peaked at 10 min and dissipated by 30% at 45 min of stimulation. Figure 1D presents the kinetics of the insulin-induced H 2 O 2 signal. H 2 O 2 efflux from neurons into the incubation medium was used as a measure of the signal, given that H 2 O 2 penetrates readily across cellular membranes with an estimated time of gradient formation within 1 s [18]. Our data showed that insulin stimulation evokes a transient single H 2 O 2 spike with a peak at 5-10 s and duration of less than 30 s. We observed a significant difference between the areas under curves (AUCs) calculated for the H 2 O 2 signal in insulin-and vehicle-exposed neurons during the 30 s of insulin stimulation (P<0.001), while no difference was observed at baseline and periods following this time ( Figure 1E). A comparison of the kinetic curves for the insulin-induced H 2 O 2 signal and receptor autophosphorylation revealed that the H 2 O 2 signal precedes autophosphorylation ( Figure 1F). Notably, at 20 s of stimulation, when the H 2 O 2 signal was 95% complete, receptor autophosphorylation was still in progress.

The H 2 O 2 signal generates strong ultrasensitivity in insulin-induced receptor autophosphorylation
To address the specific function of the H 2 O 2 signal in insulin-stimulated receptor autophosphorylation, we investigated the effects of increasing concentrations of Nacetylcysteine (NAC), a scavenger of hydrogen peroxide, on receptor autophosphorylation in neurons stimulated with insulin. NAC dose-dependently and completely (at concentrations ≥4 mM) inhibited receptor autophosphorylation in CGN cultures exposed to 100 nM insulin ( Figure 1C). Approximation of the experimental doseresponse data with the Hill function generated IC 50 of 3.7±0.2 mM and a Hill coefficient, n H , of 8.0±2.3 (R 2 =0.97), signifying that the observed dose response is highly sigmoidal and considerably steeper than a hyperbolic curve (n H =1). Therefore, even a small increase in the H 2 O 2 scavenger dose above the threshold results in complete abrogation of receptor autophosphorylation. Given that a steep dose-response with a Hill coefficient n H >1 is defined as ultrasensitive [19,20], the results suggest that the insulin-induced H 2 O 2 signal generates strong ultrasensitivity in receptor autophosphorylation. Autophosphorylation only occurs when the H 2 O 2 signal has surpassed a certain threshold. Conversely, if the H 2 O 2 signal does not reach this threshold, no autophosphorylation occurs, even in response to the highest insulin dose (100 nM). Therefore, H 2 O 2 appears to function as an enabling signal that permits insulin receptor autophosphorylation in an all-or-nothing manner. This mode of switching between two modes of action is often referred to as a decision-making process. In this context, H 2 O 2 signal above the threshold serves as a decision-making step of the neuron to permit activation of the insulin receptor.

Succinate dehydrogenase of mitochondrial complex II regulates the insulin-induced H 2 O 2 signal
To extend our previous finding that the mitochondrial electron transport chain (ETC.) provides the source of insulin-induced H 2 O 2 in neurons [17], we investigated the roles of ETC. complexes I and II in this process. Malonate, a competitive inhibitor of succinate dehydrogenase of complex II, dose-dependently and completely (at concentrations ≥ 4 mM) inhibited receptor autophosphorylation stimulated by 100 nM insulin in neurons, as shown in Figure 2A. Approximation of the malonate dose-response curve for autophosphorylation with the Hill function generated an IC 50 of 2.0±0.3 mM and Hill coefficient, n H , of 3.4 (R 2 =0.92), indicating that the dose response is a highly sigmoidal, and therefore, even a small change in succinate dehydrogenase activity around a certain threshold can have a dramatic effect on insulin receptor autophosphorylation. At concentrations of malonate that completely inhibited receptor autophosphorylation, the insulin-induced H 2 O 2 signal was abolished ( Figure 2B,C). Since the inhibitory effects of malonate (on both autophosphorylation and H 2 O 2 generation) were so complete, we conclude that the reducing equivalents for generation of H 2 O 2 do not originate from sources other than the succinate dehydrogenase reaction. The results collectively suggest that activation of succinate dehydrogenase is a possible mechanism underlying insulin-mediated switching of the mitochondrial H 2 O 2 signal critical for receptor autophosphorylation.
As shown in Figure 2B, the rate of H 2 O 2 generation exhibits biphasic behavior, showing a fast initial increase followed by a rapid decrease to the baseline steady-state level while insulin stimulation is sustained. In keeping with common theory, such a response can be generated by an incoherent feed-forward loop (IFFL) in which insulin modulates the H 2 O 2 rate via at least two intermediate pathways with opposite functions of "fast activation" and "delayed inhibition". We speculate that fast activation and delayed inhibition of succinate dehydrogenase generates the biphasic response. In this scheme, delayed inhibition of succinate dehydrogenase is attributed to the downstream metabolites of succinate oxidation, oxaloacetate [21] and H 2 O 2 [22][23][24]. These metabolites act in orchestrated manner to yield optimal inhibition, since oxaloacetate binds the oxidized form of succinate dehydrogenase more effectively than the reduced form [25] and oxidative conditions induce more significant inhibition of succinate dehydrogenase by oxaloacetate [26][27][28][29]. Given that the transition times between the active and inactive states of succinate dehydrogenase [26,30] correspond well to the timing of the insulin-induced biphasic response, we propose that succinate dehydrogenase has sufficient regulatory capacity to generate the transient H 2 O 2 signal in response to insulin stimulation.
To clarify the factors influencing generation of the insulin-induced H 2 O 2 signal, we briefly reviewed current knowledge on succinate-supported H 2 O 2 generation in isolated mitochondria. Succinate promotes the highest rates of H 2 O 2 production among all respiratory substrates [31][32][33][34][35]. Most mitochondrial H 2 O 2 is produced from superoxide generated by reduction of molecular oxygen in the electron transport chain (ETC.) [36,37]. Although several ETC. sites may generate superoxide in mitochondria respiring on succinate, the majority is produced at complex II [38] and at complex I during reverse electron transport (RET) from complex II [32][33][34][35][36], where complex I-associated superoxide production is sensitive to modulation of the mitochondrial membrane potential, ΔΨ m [32,39]. Notably, RET-associated superoxide generation is observed at the highest nonphysiological succinate concentrations at which superoxide production at complex II is suppressed, whereas generation at complex II occurs only within the range of lower physiologically relevant succinate levels [38]. At succinate concentrations favoring RET-associated H 2 O 2 production, generation of H 2 O 2 depends on the metabolic state of mitochondria and changes in parallel with succinate dehydrogenase activity. The highest H 2 O 2 rates [37] and succinate dehydrogenase activity [27,28] are observed at metabolic state 4. On transition to state 3, H 2 O 2 production rates drop rapidly [34,39] and succinate dehydrogenase activity decreases [27,28]. Protonophores completely inhibit H 2 O 2 production [33,39] and deactivate succinate dehydrogenase to the lowest observable activity [27]. In summary, the documented literature suggests that ΔΨ m and activity of mitochondrial complexes I/II are the   Mitochondrial complex I is involved in control of insulininduced receptor autophosphorylation, but to a lower extent than complex II To investigate whether ETC. complex I functions in the control of insulin-stimulated receptor autophosphorylation, we examined the effects of rotenone on receptor autophosphorylation in neurons stimulated with 100 nM insulin. Rotenone is a selective inhibitor of the ubiquinone reduction site at mitochondrial complex I. Rotenone inhibited insulin-stimulated receptor autophosphorylation, as shown in Figure 2D. The effect of rotenone was significant, but incomplete, and even at the highest concentration, autophosphorylation was reduced by less than 50%. Fitting the dose-response curve generated a Hill coefficient, n H , of 0.7 (R 2 =0.83), indicating that the observed dose response is gradual and not switch-like.
Our results suggest that ETC. complex I is involved in control of insulin receptor autophosphorylation, presumably as part of the mitochondrial machinery of RETassociated H 2 O 2 generation. However, its role does not appear to be as important as that of ETC. complex II.

Insulin-induced receptor autophosphorylation is sensitive to mitochondrial depolarization
To address whether receptor autophosphorylation is sensitive to ΔΨ m , we compared the effects of insulin and other additives on autophosphorylation and ΔΨ m in neurons. Rhodamine 123 was used as a measure of mitochondrial ΔΨ m [40]. Insulin did not have a significant effect on Rhodamine 123 fluorescence in neurons. Malonate induced a dose-dependent increase in Rhodamine 123 fluorescence, signifying that inhibition of succinate dehydrogenase results in mitochondrial depolarization (Figure 2A). Fitting the fluorescence curve to the Hill equation led to ED 50 of 5.2±0.1 mM and Hill coefficient, n H , of 5.4 (R 2 =0.99). Comparison of the malonate doseresponse curves for autophosphorylation and fluorescence revealed that at a malonate concentration of 2 mM (inducing 50% inhibition of autophosphorylation), mitochondria were depolarized by less than 1%. This finding suggests that ΔΨ m in neurons is ultrasensitive to succinate dehydrogenase activity and mitochondrial depolarization is not a causative factor for the inhibitory effect of malonate on receptor autophosphorylation. Figure 2D shows that the inhibitory effect of rotenone on autophosphorylation is not accompanied by mitochondrial depolarization. The protonophore, FCCP, known to dissipate the transmembrane proton gradient, evoked an increase in Rhodamine 123 fluorescence and inhibited insulin-stimulated recep-tor autophosphorylation in a dose-dependent manner ( Figure 2E). Fitting the FCCP dose-response curves for fluorescence and autophosphorylation gave ED 50 of 0.030±0.003 (R 2 =0.99) and IC 50 of 0.07±0.02 (R 2 =0.98), respectively. At a FCCP concentration of 0.07 μM (inducing 50% inhibition of autophosphorylation), >95% mitochondrial depolarization was observed. These data suggest that FCCP-induced mitochondrial depolarization leads to inhibition of insulin-stimulated receptor autophosphorylation. Given that protonophores deactivate succinate dehydrogenase to the lowest observable activity [27], we propose that the inhibitory effect of depolarization on autophosphorylation is mediated via deactivation of succinate dehydrogenase. Taken together, these results suggest that insulin receptor autophosphorylation is sensitive to factors inducing mitochondrial depolarization, while modulation of ΔΨm is not implicated in the mechanism of insulin-triggered receptor autophosphorylation.

Nox is not involved in control of insulin-induced receptor autophosphorylation in neurons
A number of previous studies using non-neuronal cells have assigned the insulin-induced H 2 O 2 signal to activation of NADPH oxidase (Nox) that is sensitive to inhibition by diphenyleneiodonium (DPI) [41][42][43]. Accordingly, we investigated the effect of DPI on receptor autophosphorylation in neurons exposed to 100 nM insulin. DPI treatment at a concentration of 10 μM significantly inhibited insulin-stimulated receptor autophosphorylation. Moreover, the inhibitory action of DPI was dose-dependent ( Figure 2F). Since selective mitochondrial inhibitors that do not inhibit Nox completely abrogate both insulin-induced H 2 O 2 generation and receptor autophosphorylation, the results of the DPI experiment cannot be interpreted to conclude that Nox is a source of the insulin-induced H 2 O 2 signal in neurons. Although the nonspecific flavoprotein inhibitor, DPI, is often claimed to be a specific inhibitor of Nox, compelling evidence shows that DPI inhibits many targets [44], including ROS generation at the mitochondrial complex I [45,46] and succinate-supported H 2 O 2 generation in rat brain mitochondria [33]. In this context, data from the DPI experiment do not contradict our main conclusion that the mitochondrial ETC. is the only source of insulin-induced H 2 O 2 in neurons. Another argument against a role of Nox in neuronal insulin receptor autophosphorylation is that the kinetics of the insulininduced H 2 O 2 signal is unusually fast, compared to that previously observed for Nox in non-neuronal cells [8,43]. The neuronal H 2 O 2 signal is a single spike with a sharp peak at 5-10 s and duration of less than 30 s. In contrast, the insulin-induced oxidant signal in adipocytes reaches a peak at 5 min and begins to dissipate by 10 min [8]. In HepG2 cells, ROS generation is optimal at about 10 min and dissipates after 30 min of insulin stimulation [43]. This significant difference in the timing of H 2 O 2 signals indicates that the source of insulininduced H 2 O 2 in neurons differs from that in adipocytes and HepG2 cells, which is Nox [8,43]. Our results collectively suggest that Nox is not involved in H 2 O 2 signaling that controls insulin-stimulated receptor autophosphorylation in neurons. Data from the present study raise the issue of whether the mitochondrial origin of the insulin-induced H 2 O 2 signal is a neuron-specific phenomenon. Although abundant literature is available on insulin-induced H 2 O 2 generation in cells, limited studies have focused on the origin of insulin-induced H 2 O 2 implicated in insulin receptor autophosphorylation. In experiments with 3T3-L1 adipocytes [41,42] and HepG2 cells [43], the source of H 2 O 2 was assigned to activated Nox on the basis of data obtained with DPI only. Strong evidence from experiments with dominant-negative Nox4 constructs and Nox4 siRNA further suggested that Nox4 is the source of the insulin-induced H 2 O 2 signal in adipocytes [47]. The results from the present study, together with previous data [17], indicate that mitochondrial ETC. is the only source of insulin-induced H 2 O 2 in neurons. In summary, owing to limited data, it is not currently possible to address whether mitochondria represent the only neuronal source of insulin-induced H 2 O 2 implicated in receptor activation.

The insulin-induced H 2 O 2 signal and receptor autophosphorylation in neurons are pertussis toxin-sensitive
A pertussis toxin-sensitive G protein links the insulin receptor to the insulin-induced H 2 O 2 signal in nonneuronal cells [48]. To determine whether G protein signaling is involved in generation of insulin-induced H 2 O 2 in neurons, we examined the effects of pertussis toxin (PTX), an inhibitor of Gi/0 protein-receptor coupling, on H 2 O 2 generation and receptor autophosphorylation in CGN exposed to 100 nM insulin. PTX treatment completely inhibited insulin-stimulated receptor autophosphorylation at the highest dose of 2 mg/L, and this suppression was dose-dependent ( Figure 3A). Approximation of the curve to the Hill equation gave IC 50 of 0.16±0.06 mg/L and Hill coefficient, n H , of 1.7±2.0 (R 2 =0.96), implying that insulin-induced receptor autophosphorylation is ultrasensitive to PTX. PTX (2 mg/L) completely eliminated the insulin-induced H 2 O 2 signal ( Figure 3B). Thus, at the dose at which PTX completely abrogated the H 2 O 2 signal (2 mg/L), receptor autophosphorylation was abolished, as shown in Figure 3C. Based on these results, we suggest that a PTX-sensitive G protein links the insulin receptor to the H 2 O 2 -generating system in neurons during generation of the insulin-induced H 2 O 2 signal critical for insulin receptor autophosphorylation.
Although our data show for the first time that a PTXsensitive G protein is possibly involved in activation of the insulin receptor in CNS, the involvement of a PTXsensitive G protein in insulin receptor activation in peripheral tissues is widely documented. In mice, Gαi2 deficiency results in impaired tyrosine phosphorylation of the insulin receptor substrate, IRS1, and frank insulin resistance in peripheral tissues [49], whereas targeted expression of a constitutively active form of Gαi2 enhances insulin action through amplifying tyrosine phosphorylation of the insulin receptor and IRS1 [50][51][52]. In human adipocytes, insulin recruits Gαi2 to activate Nox-dependent H 2 O 2 generation [48] and regulate insulin receptor autophosphorylation [53]. Although no data supporting the involvement of Gαi2 in brain insulin signaling are available, Gαi2, and especially its short-lived splice variant, sGi2, are widely distributed throughout rat and monkey brain, where sGi2 is detected in both axons and dendrites at presynaptic and postsynaptic sites [54]. Gαi2 operates largely in plasma membranes, while sGi2 is localized in a variety of subcellular locations, including mitochondria [54][55][56]. Our present results are generally in keeping with current knowledge on the amplification role of pertussis toxinsensitive Gi protein in insulin receptor activation. However, the specific G protein isoform implicated in neuronal insulin receptor activation remains to be determined.

Functional association of the insulin receptor and mitochondria during receptor activation in neurons
It is possible to draw some tentative conclusions regarding the functional relationship between the insulin receptor and mitochondria during receptor activation in neurons ( Figure 4). Insulin stimulation induces receptor autophosphorylation, which reaches a peak at 10 min. Upon autophosphorylation, the receptor becomes fully activated and initiates signaling to the inside of the neuron. At times preceding autophosphorylation, insulin induces a transient H 2 O 2 signal, which plays a permissive role in activation of the insulin receptor. Autophosphorylation only occurs once the H 2 O 2 signal has surpassed a certain threshold. Under conditions where the H 2 O 2 signal does not reach this threshold, no autophosphorylation occurs, even in response to the highest insulin dose. In this context, H 2 O 2 signal above the threshold serves as the neuron's decision to activate the insulin receptor. The insulin-induced H 2 O 2 signal is derived from mitochondria. Succinate dehydrogenase in complex II plays a key role in control of H 2 O 2 generation. An unknown pertussis toxin-sensitive G protein links the insulin receptor to the mitochondrial H 2 O 2generating system during H 2 O 2 signal activation.
The current study provides novel insights into the mechanisms underlying neuronal response to insulin. The insulin receptor is activated if two conditions are met: 1) insulin binds to the receptor, and 2) the H 2 O 2 signal exceeds a certain threshold, enabling receptor autophosphorylation. Generally, this mode of control means that activation of the neuronal insulin receptor is conditional on mitochondrial functioning. Moreover, receptor activation may be conditional on neural activity. Neural activity evokes insulin release from synaptosomes within nerve endings into the synaptic cleft [57,58] and induces mitochondrial migration to dendritic spines [59], which are commonly poor in mitochondria.  Therefore, periods of synaptic activity favor insulin receptor activation in the postsynaptic density of dendritic spines where the two above conditions are met.  [60][61][62]. Mice overexpressing Gpx1 are insulin-resistant, obese, and along with hyperinsulinemia, display a 70% reduction in insulin-stimulated phosphorylation of insulin receptors, compared to wild-type control mice [63]. In contrast, mice lacking Gpx1 are protected from high fat diet-induced insulin resistance, while administration of the H 2 O 2 scavenger, NAC, renders them more insulin-resistant [64]. Brain insulin resistance is an early and common feature of Alzheimer's disease (AD). Compared to age-matched normal brains, AD brains are characterized by diminished tyrosine phosphorylation of the insulin receptor and its substrate, IRS1 [65], and significant overexpression of glutathione peroxidase [66,67] and peroxiredoxins (Prx1 and Prx2) [68][69][70][71]. Elevated antioxidant activity in early AD is considered the result of compensatory response to oxidative stress [72,73]. From this viewpoint, oxidative stress seems to be another factor that interferes with the insulin-induced H 2 O 2 signal and consequently induces central insulin resistance.

Conclusions
In

Insulin receptor autophosphorylation assay
Amounts of the autophosphorylated β-subunit of the insulin receptor were measured with the PhosphoDetect™ insulin receptor (pTyr1162/1163) ELISA kit (Calbiochem) suitable for studies with the rat insulin receptor. CGN cultures were incubated in Hepes-buffered salt solution (145 mM NaCl, 5.6 mM KCl, 1.8 mM СаCl 2 , 1 mM MgCl 2 , 20 mM HEPES, and 0.5 mM glucose) at pH 7.4 for 30 min, followed by exposure to insulin (10 min, 37°C) or no insulin (control) in the presence or absence of other additives (preincubation for 5-30 min before insulin). The experiment was terminated by removing the medium, washing with ice-cold PBS, and adding 200 μL/dish cell lysis buffer (Biosource) supplemented with 1 mM PMSF, 50 mM protease inhibitor set III (Sigma), and 2 mM sodium orthovanadate as the tyrosine phosphatase inhibitor on ice for 10 min. Lysates were centrifuged at 12,000 rpm at 4°C for 12 min. In each CGN lysate, amounts of autophosphorylated receptors were measured as described by the manufacturer. For each experiment, values were normalized to the total amounts of insulin receptor β-subunit (IR) measured using the insulin receptor (β-subunit) ELISA kit (Calbiochem) and normalized to scale between 0 and 1 using Equation 1: where Y i is an original receptor autophosphorylation value, Normalized (Y i ) the normalized autophosphorylation value, Y CTR the mean autophosphorylation value in neurons exposed to vehicle, and Y INS the mean autophosphorylation value in neurons exposed to 100 nM insulin.

Monitoring of Rhodamine 123 fluorescence
Mitochondrial depolarization changes in individual neurons within CGN cultures in response to additives were measured via fluorescence of Rhodamine 123, as described earlier [75]. CGN cultures were equilibrated with 10 μg/ml Rhodamine 123 in Hepes-buffered salt solution for 10 min at 20°C. Cells were washed with HBSS before the experiment (excitation, 488 nm; emission, 530 nm). CGN cultures were exposed to insulin, rotenone, malonate or FCCP for 10 min, and fluorescence measured. Data were expressed as the F/F 0 ratio of the fluorescence signal measured after exposure of neurons to additives to that measured at baseline.

Curve fitting
Experimental data were fitted by non-linear regression to the Hill equation to provide the parameter values in Equation 2 : where Y is a parameter value, X a variable, K the concentration of the variable producing half the effect, and n H the Hill coefficient.

Statistical analysis
Data were analyzed for statistical significance with oneway analysis of variance (ANOVA). Values are presented as means ± SEM. Differences were considered significant at P<0.05.