Inhibition of mitochondrial complex 1 by the S6K1 inhibitor PF-4708671 partly contributes to its glucose metabolic effects in muscle and liver cells

mTOR complex 1 (mTORC1) and p70 S6 kinase (S6K1) are both involved in the development of obesity-linked insulin resistance. Recently, we showed that the S6K1 inhibitor PF-4708671 (PF) increases insulin sensitivity. However, we also reported that PF can increase glucose metabolism even in the absence of insulin in muscle and hepatic cells. Here we further explored the potential mechanisms by which PF increases glucose metabolism in muscle and liver cells independent of insulin. Time course experiments revealed that PF induces AMP-activated protein kinase (AMPK) activation before inhibiting S6K1. However, PF-induced glucose uptake was not prevented in primary muscle cells from AMPK α1/2 double KO (dKO) mice. Moreover, PF-mediated suppression of hepatic glucose production was maintained in hepatocytes derived from AMPK α1/2-dKO mice. Remarkably, PF could still reduce glucose production and activate AMPK in hepatocytes from S6K1/2 dKO mice. Mechanistically, bioenergetics experiments revealed that PF reduces mitochondrial complex I activity in both muscle and hepatic cells. The stimulatory effect of PF on glucose uptake was partially reduced by expression of the Saccharomyces cerevisiae NADH:ubiquinone oxidoreductase in L6 cells. These results indicate that PF-mediated S6K1 inhibition is not required for its effect on insulin-independent glucose metabolism and AMPK activation. We conclude that, although PF rapidly activates AMPK, its ability to acutely increase glucose uptake and suppress glucose production does not require AMPK activation. Unexpectedly, PF rapidly inhibits mitochondrial complex I activity, a mechanism that partially underlies PF's effect on glucose metabolism.

Mechanistic target of rapamycin (mTOR) 2 complex 1 (mTORC1) and its downstream effector p70 ribosomal S6 kinase (S6K1) are important modulators of energy homeostasis (1,2). This pathway regulates cell proliferation, growth, and protein synthesis, which are tightly regulated depending on substrate availability (3). Furthermore, the mTORC1/S6K1 pathway also acts as an energy sensor by sensing amino acids, lipids, and energy status and triggers a negative feedback loop toward the PI3K/Akt pathway to terminate its activation after insulin or growth factor stimulation (4-7). However, this negative feedback loop is overactivated by nutrient satiation in metabolic diseases such as obesity, contributing to the development of insulin resistance and type 2 diabetes (7)(8)(9). Therefore, targeting downstream effectors of mTORC1, such as S6K1, might be a promising strategy to improve the metabolic syndrome.
We have shown previously that pharmacologic inhibition of S6K1 using PF-4708671 (PF) improved glucose tolerance in mice fed a high-fat diet by restoring insulin-induced Akt phosphorylation in obese mice (10). Furthermore, in vitro studies using L6 myocytes and FAO hepatoma cells revealed the ability of PF to increase basal muscle glucose uptake and decrease basal hepatic glucose production (10). These results were obtained in the absence of insulin stimulation, suggesting an additional mechanism of this compound to modulate energy homeostasis independent of enhancing insulin signaling. Previously, several studies have shown a role of S6K in regulation of AMP-activated protein kinase (AMPK), which is known to be important for energy homeostasis. Indeed, S6K has been shown to phosphorylate AMPK on Ser-491 in neurons, inhibiting its activity (11). In addition, selective S6K1/2 deletion in muscle is associated with an increase in AMPK activity because of an increase in the AMP/ATP ratio (12), suggesting that S6K may participate in inhibition of AMPK by either direct and/or indirect mechanisms. However, a previous study using mouse embryonic fibroblasts derived from S6K1/2 KO mice showed that PF activates AMPK independent of S6K1 inhibition (13). AMPK is an important regulator of energy homeostasis by  cro ARTICLE sensing the energy level via the AMP/ATP ratio, thus activating catabolic processes to provide energy to cells. AMPK activation induces glucose uptake, fatty acid oxidation, and autophagy and inhibits cell growth and protein synthesis (14 -16). However, it remains unknown whether the insulinindependent effect of PF on glucose metabolism is linked to either S6K1 inhibition or AMPK activation in skeletal muscle and liver, the principal sites of glucose disposal.
In this paper, we explored the mechanisms behind the changes in glucose metabolism in muscle and hepatic cells upon S6K1 inhibition independent of insulin stimulation. We evaluated the contribution of S6K1, AMPK, and mitochondrial function to PF-4708671-mediated glucose uptake in muscle cells and decreased basal glucose production in hepatocytes. In both muscle cells and hepatocytes, we observed that PF-4708671 increases both AMPK Thr-172 and ACC Ser-79 phosphorylation, reflecting an increase in AMPK activity. We next used primary muscle and liver cells derived from AMPK ␣1/2 dKO and S6K1/2 dKO mice to determine the role of both AMPK and S6K in the glucose metabolic effect of PF-470867. These experiments showed that neither AMPK nor S6K is required for the acute effect of the compound on glucose metabolism. On the other hand, we found that PF-4708671 reduces mitochondrial complex I activity, which is partly responsible for its glucose-enhancing effect in muscle cells.

PF-4708671 rapidly increased AMPK phosphorylation before inhibiting S6K1
We have shown previously that S6K1 is a key regulator of glucose disposal in both muscle and liver (7,17) and that the S6K1 inhibitor PF increases both insulin-dependent and insulin-independent glucose metabolism in muscle and hepatic cells (10). However, the effect of PF on AMPK activation remains unexplored in metabolically relevant cellular models. As shown in Fig. 1, A and B, dose-response experiments showed that acute (2-h) PF treatment (1-20 M) increases phosphorylation of AMPK on Thr-172 in both L6 myocytes and FAO hepatocytes. The drug also enhanced phosphorylation of the AMPK substrate ACC on Ser-79. These effects were associated with a dosedependent inhibition of S6 phosphorylation on Ser-240/44, suggesting that S6K1 inhibition by PF leads to AMPK activation.
To confirm that S6K1 inhibition is required for AMPK activation, we next performed time course experiments (Fig. 1, C  and D). Unexpectedly, AMPK Thr-172 and ACC Ser-79 phosphorylation were already robustly increased 5 min after PF treatment, whereas S6 phosphorylation was only found to be reduced at least 30 min after PF exposure in L6 cells (Fig. 1C) and FAO cells (Fig. 1D). These results indicate that AMPK activation by PF precedes S6K1 inhibition and further suggest that rapid AMPK activation may contribute to PF action on glucose metabolism in muscle and liver cells.

PF-4708671 increased glucose uptake in muscle cells and decreased hepatic glucose production independent of AMPK
Because AMPK is a key regulator of muscle glucose uptake (18), we next determined whether the induction of glucose uptake by PF was AMPK-dependent. We used differentiated muscle cells from WT mice or mice lacking both AMPK␣1 and AMPK␣2 catalytic subunits (AMPK ␣1/2 dKO muscle cells). After 5 and 48 h of treatment with PF, AMPK phosphorylation on Thr-172 and ACC phosphorylation on Ser-79 were increased in differentiated WT myocytes but, as expected, not in AMPK ␣1/2 dKO muscle cells ( Fig. 2A). In both AMPK WT and dKO muscle cells, PF treatment decreased basal and insulinstimulated S6 phosphorylation in both WT and AMPK ␣1/2 dKO muscle cells. PF treatment of both WT and AMPK ␣1/2 dKO muscle cells for 5 and 48 h increased glucose uptake compared with vehicle-treated cells (Fig. 2B). However, lack of AMPK activation did not reduce the stimulatory effects of PF treatment on muscle glucose uptake. These data demonstrate that AMPK is dispensable for the stimulatory effect of PF on muscle glucose uptake.
To further investigate the role of AMPK in PF-induced suppression of hepatic glucose production (HGP), we next isolated hepatocytes from WT mice or mice lacking both AMPK␣1 and AMPK␣2 catalytic subunits (AMPK ␣1/2 dKO hepatocytes). As expected, Western blot analysis confirmed that AMPK ␣1/2 dKO hepatocytes did not express AMPK (Fig. 2C). In WT hepatocytes, PF increased both AMPK phosphorylation on Thr-172 and ACC phosphorylation on Ser-79 without (Fig. 2C, left panel) or with (Fig. 2C, right panel) glucagon incubation. In addition, PF dose-dependently inhibited S6 phosphorylation in both AMPK WT and ␣1/2 dKO hepatocytes. As a positive control, we used metformin which increased AMPK and ACC phosphorylation with or without glucagon (Fig. 2C). Interestingly, metformin did not affect S6 phosphorylation in WT hepatocytes but decreased S6 phosphorylation in AMPK ␣1/2 double KO hepatocytes (Fig. 2C).
We next determined HGP in primary hepatocytes under basal and glucagon-stimulated conditions (Fig. 2D). As reported previously (19), AMPK is not required for basal or glucagon-induced HGP. Inhibition of S6K1 with PF (20 M) was found to reduce basal and glucagon-stimulated HGP in AMPK WT, and this response was amplified in ␣1/2 dKO hepatocytes, especially in the glucagon-stimulated state. Similar results were obtained using metformin, which was found to blunt HGP even more in ␣1/2 dKO than in WT hepatocytes, in agreement with a previous report (19). These results indicate that AMPK is also dispensable for the inhibitory effect of PF on hepatic glucose production.

PF-4708671 decreased basal hepatic production independent of S6K1
To determine whether PF-mediated inhibition of HGP was S6K1-dependent, we next isolated hepatocytes from S6K1/2 double KO mice. As anticipated, S6K1 expression was absent in S6K1/2 dKO hepatocytes, and PF decreased S6 phosphorylation in WT hepatic cells (Fig. 3A). In addition, PF increased AMPK and ACC phosphorylation in both S6K1/2 WT and double KO hepatocytes, suggesting that S6K1 is dispensable for PF-induced AMPK activation (Fig. 3A). Similarly, metformin strongly increased AMPK and ACC phosphorylation in both WT and S6K1/2 dKO hepatocytes but did not affect S6 phosphorylation in WT hepatocytes (Fig. 3A). In both WT and S6K1/2 dKO hepatocytes, PF (20 M) decreased glucagon-stim-PF regulates glucose metabolism independent of insulin ulated HGP (Fig. 3B). Metformin also decreased HGP with no differences between S6K1/2 WT and double KO hepatocytes. Thus, neither AMPK nor S6K is required for the ability of PF to blunt glucose production in hepatocytes.

PF-4708671 decreased mitochondrial respiration and mitochondrial spare respiratory capacity in muscle cells and hepatocytes
Because metformin has been shown previously to control glucose production through regulation of the hepatic energy state and inhibition of mitochondrial respiration (19 -21), we next investigated whether PF could exert its glucose metabolic effects through a similar mechanism. Using a Seahorse XF 24 system to determine mitochondrial respiratory parameters (Fig. 4A), we next examined the effect of PF on oxygen consumption in L6 myotubes and FAO hepatocytes. Cells were first treated with PF (1-20 M) for 400 min to determine the maximal effect of the compound on basal respiration and whether this effect could be sustained over time. In L6 cells, there was a trend for PF to reduce basal mitochondrial res- PF regulates glucose metabolism independent of insulin piration (Fig. 4, B and C), which was more apparent in FAO cells, as well as respiration associated with ATP production at 20 M (Fig. 4, D and E). In addition, PF lowered maximal respiration and the spare respiratory capacity in both L6 and FAO cells. These results indicate that PF reduced mitochondrial respiratory capacity in both cell types.

PF-4708671 reduced mitochondrial complex 1 activity
We then assessed whether this decrease in maximal respiratory capacity was due to a decrease in mitochondrial complex I and/or II activity. Cells were permeabilized to allow ADP, pyruvate (Pyr), malate (Mal) and succinate to pass through the plasma membrane and to reach mitochondria. L6 cells were acutely treated with different concentrations of PF (i.e. the drug was added just before the assay), and the plate was added to the Seahorse XF 24 system. Addition of Pyr/Mal in the presence of ADP increased mitochondrial complex I activity, which was dose-dependently inhibited by PF (Fig. 5A). Mitochondrial complex II activity was measured by adding the complex I inhibitor rotenone first, followed by succinate. No significant effect of PF treatment on mitochondrial complex II activity was observed (Fig. 5B). Similar results were obtained in FAO cells. Indeed, PF reduced mitochondrial complex I activity in hepatic cells but failed to affect mitochondrial complex II activity (Fig.  5, C and D). Taken together, these results show that PF acts mainly on mitochondrial complex 1 to regulate mitochondrial function in both muscle and hepatic cells.

PF-induced glucose uptake is partially dependent on mitochondrial complex 1 activity
To determine the role of mitochondrial complex I in mediating glucose uptake in L6 myocytes, we stably expressed Saccharomyces cerevisiae NADH:ubiquinone oxidoreductase (NDI1) in these cells, which can maintain respiration even in the presence of complex I inhibitors (22)(23)(24)(25). Expression of NDI1 in muscle cells (Fig. 6A) was found to significantly reduce PF-mediated glucose uptake without interfering with insulininduced glucose disposal in these cells (Fig. 6B). When taking

PF regulates glucose metabolism independent of insulin
into account the changes in basal glucose uptake in NDI1-expressing cells to calculate the specific effect of PF on this parameter, we concluded that about 45% of the drug effect may be accounted for by inhibition of mitochondrial complex 1 activity in muscle cells (Fig. 6C). These results demonstrate that PF increases insulin-independent glucose uptake at least in part by inhibition of mitochondrial complex 1 activity in muscle cells but that other unknown mechanisms are also involved.

Discussion
We recently reported that pharmacologic inhibition of S6K1 by PF for only 1 week improves glucose tolerance in obese mice fed a high-fat diet and increases insulin-induced Akt phosphorylation in liver, muscle, and white adipose tissue (10). Interestingly, these studies also revealed that PF enhances basal glucose uptake in myocytes and decreases basal glucose production in hepatocytes, suggesting that the S6K1 inhibitor also modulates glucose metabolism through insulin-independent pathways. This is consistent with other studies in C2C12 cells, where transfection of a siRNA targeting S6K1 also enhanced basal glucose uptake (26). In this study, we found that PF rapidly modulated glucose metabolism in both L6 myotubes and FAO hepatocytes in association with increased phosphorylation of AMPK and its downstream effector ACC.
However, following acute PF treatment, we observed an increase in both AMPK Thr-172 and ACC Ser-79 phosphorylation, a well-known substrate of active AMPK. The time course experiment performed in PF-treated cells revealed that AMPK is activated rapidly (Thr-172 AMPK and Ser-79 ACC phosphorylation), faster than PF-mediating S6K1 inhibition. Despite this rapid increase in AMPK activation, we showed that PF-regulating glucose homeostasis is independent of

PF regulates glucose metabolism independent of insulin
AMPK, ruling out a potential role of AMPK on PF-stimulating glucose uptake and decreasing hepatic glucose production. These results are also supported by other studies showing that AMPK deficiency failed to impact hepatic glucose production and the suppressive effect of metformin on this process (19, 27).
PF has been suggested previously to activate AMPK by inhibiting mitochondrial complex I of the respiratory chain independent of S6K1 in mouse embryonic fibroblasts (13). Using muscle cells and hepatocytes, we also observed that PF acutely decreased mitochondrial complex I activity, suggesting that the drug modulates glucose metabolism in these cells through inhibition of mitochondrial function. Using S6K1/2 dKO hepatic cells, we further showed that PF inhibits basal or glucagonstimulated hepatic glucose production independent of S6K. These results suggest that PF action on hepatic glucose production is dispensable for S6K inhibition and thus might be entirely due to a higher energy demand and metabolism rewiring toward replenishing acetyl-CoA pools caused by lower mitochondrial complex I activity.

PF regulates glucose metabolism independent of insulin
Glucose uptake can be attributed to glucose transport, glucose oxidation, and glycogen synthesis. Our respirometry analysis revealed that PF decreased oxidative phosphorylation, suggesting that a reduction in mitochondrial ATP production might have induced glucose transport for ATP production from glycolysis. Reduction of mitochondrial complex 1 activity explains, in part, the insulin-independent action of PF on glucose metabolism. In L6 muscle cells, the ability of PF to enhance glucose uptake was partially reversed by stable NDI1 expression, indicating that the decrease of mitochondrial complex 1 activity contributes to the stimulatory effect of the compound on muscle glucose uptake. This conclusion is also supported by previous studies showing that inhibition of mitochondrial complex 1 with a very low dose of rotenone or siRNA-mediated silencing of NADH:ubiquinone oxidoreductase subunit A13, a subunit of complex I, can reduce hepatic glucose production and stimulate glucose consumption in muscle cells indepen-dent of AMPK (28). In addition, other studies have reported similar mechanisms. Indeed, berberine has been shown to enhance glucose utilization and glycolysis by reducing mitochondrial glucose oxidation in L6 myotubes. Altogether, our results suggest that the energy demand generated by inhibition of mitochondrial complex 1 activity is enough to drive glucose uptake in muscle and to decrease hepatic glucose production in PF-treated cells, independent of AMPK activation. However, the finding that stable expression of NDI1 only partially blunted PF-induced glucose uptake in muscle cells suggests that other mechanisms, such as S6K1 inhibition, may also contribute to the insulin-independent metabolic effect of the drug.
This dual effect of PF-4708671 might be an interesting advantage to treat type 2 diabetes more efficiently. We found that part of the action of PF-4708671 is by inhibiting mitochondrial complex 1, which is also a well-known target for the most prescribed type 2 diabetes drug, metformin (20, 29). These On the day of the experiment, cells were washed twice with mannitol/sucrose buffer containing 4 mM ADP and treated with the indicated PF-4708671 concentrations before cell permeabilization with the XF plasma membrane permeabilizer (1 nM) from Seahorse Bioscience. Then the plate was inserted into the XF 24-well apparatus. A and C, mitochondrial complex I and II activity was assessed by successively adding Pyr/Mal (10 mM/1 mM), rotenone (Rot, 1 M), succinate (10 mM), and antimycin A (AA, 1 M) in differentiated L6 cells (A) and FAO hepatocytes (C). B-D, mitochondrial complex I activity was calculated from OCR differences between pyruvate/malate and rotenone. Mitochondrial complex II activity was calculated from OCR differences between succinate and antimycin A. *, p Ͻ 0.05; **, p Ͻ 0.01 versus vehicle-treated cells as indicated. Shown is the mean Ϯ S.D. of four to five independent experiments.

PF regulates glucose metabolism independent of insulin
results suggest that PF might reproduce many beneficial effects mediated by metformin in addition to those associated with S6K1 inhibition, such as direct inhibition of the negative feedback loop regulating insulin signaling. Metformin's ability to blunt hepatic glucose production is also independent of AMPK and, rather, mediated by inhibition of mitochondrial complex 1 (20) and mitochondrial glycerol phosphate dehydrogenase (19, 30).

Conclusions
PF-4708671 was characterized as a selective S6K1 inhibitor, but our study and others demonstrate that it also has S6K1independent effects, notably by reducing mitochondrial complex I activity. This inhibition led to rapid but transient induction of AMPK activation and ACC inhibition, followed by S6K1 inhibition in muscle cells and hepatocytes. Although S6K1 inhibition by this drug restored insulin signaling in cells and tissues from obese insulin-resistant mice (10), our data reveal that PF also regulates glucose metabolism independent of AMPK by reducing mitochondrial complex 1 activity. These results suggest that PF might be an interesting therapeutic tool because it can both reverse S6K1-mediated insulin resistance (10) and improve glucose control through inhibition of mitochondrial respiration, combining two key mechanisms that are well-recognized for the treatment of type 2 diabetes.

NDI1 expression in L6 cells
The pMXS-NDI1 (NADH:ubiquinone oxidoreductase) plasmid was a gift from David Sabatini (Addgene plasmid 72876), and the pMXS-control vector was kindly provided by Kivanc Birsoy (Rockefeller University). Plasmids were amplified as described previously in Phoenix-Eco cells (ATCC, CRL-3214) (28). Undifferentiated rat L6 myoblasts were infected with the amplified retrovirus and selected with 5 g/ml of blasticidin to generate L6-pMXS and L6-NDI1 stably transfected myoblasts. Both L6 cell lines were then differentiated into myotubes, as described in the previous section, creating cell lines that stably overexpressed NDI1 or the control vector. NDI1 overexpres-

Western blotting
Western blots were performed as described previously (36). Briefly, equal amounts of proteins were separated by SDS-PAGE (9% (w/v)) and transferred onto nitrocellulose membranes. Membranes were blocked in 5% (w/v) milk diluted in 50 mM Tris (pH 7.4), 0.1% (v/v) Tween (TBS-T) and incubated overnight at 4°C with the respective antibodies diluted in 3% (w/v) BSA in TBS-T.

Glucose uptake
L6 cells were serum-deprived for 5 h prior to the experiments, and 100 nmol/liter insulin was used to stimulate the cells during the last hour of deprivation. L6 cells were incubated for 8 min in HEPES-buffered saline containing 10 mol/liter unlabeled 2-deoxy-glucose and 10 mol/l D-2-deoxy-[ 3 H]glucose (0.5 Ci/ml). The reaction was terminated by washing three times with ice-cold 0.9% NaCl (w/v). Cell-associated radioactivity was determined by lysing the cells with 0.05 N NaOH, followed by liquid scintillation counting and normalization to protein concentration.

Glucose production
FAO cells were incubated for 16 h in serum-free medium with or without the indicated concentration of insulin and PF-4708671 (10 mol/liter). The cells were washed three times with PBS and incubated with phenol-and glucose-free DMEM supplemented with 20 mmol/liter sodium L-lactate and 2 mmol/liter sodium pyruvate for 5 h with or without the indicated concentration of insulin and PF-4708671 (10 mol/liter). Cell supernatants were collected, and glucose concentration was measured with the Amplex-Red glucose assay kit (Invitrogen) according to the manufacturer's instructions. Cells were lysed with 50 mmol/liter NaOH, and protein concentration was determined using a BCA protein assay kit for normalization. Primary mouse hepatocytes on 6-well plates (2.5 ϫ 10 5 cells/ well) were maintained in M199 medium containing antibiotics and 100 nM dexamethasone for 16 h prior to measurement of glucose production. Hepatocytes were washed once with PBS, and glucose production was determined after an 8-h incubation period in glucose-free DMEM containing lactate/pyruvate (10:1 mM) and 100 nM dexamethasone alone or with glucagon 25 nM (Sigma-Aldrich) and with various doses of PF-4708671. Glucose concentration values were corrected by protein content.

XF metabolic assay
Oxygen consumption rate (OCR) measurements were made using the Seahorse XF24 Extracellular Flux Analyzer. For mitochondrial metabolic parameters, we used successively different concentrations (1-20 M) of acute PF-4708671, oligomycin (1 M), carbonyl cyanide p-trifluoromethoxyphenylhydrazone (2 M), and rotenone/antimycin (1 M). For determination of complex I and complex II activity, cells were preincubated in mannitol/sucrose buffer free of any substrate but containing 4 mM ADP for 30 min, followed by addition of XF plasma membrane permeabilizer agent (1 nM) just before adding the plate to the XF24 instrument. Pyruvate/malate was injected into the cells, followed by rotenone (1 M), succinate (10 mM), and antimycin (1 M).

Quantification and statistical analyses
One-and two-way analysis of variance with Newman-Keuls post hoc test was performed with Sigma Plot version 12 (Systat Software, San Jose, CA); p Ͻ 0.05 was considered significant. Standard deviation is represented in the graphs.