In in vitro studies class-I PI3Ks (phosphoinositide 3-kinases), class-II PI3Ks and mTOR (mammalian target of rapamycin) have all been described as having roles in the regulation of glucose metabolism. The relative role each plays in the normal signalling processes regulating glucose metabolism in vivo is less clear. Knockout and knockin mouse models have provided some evidence that the class-I PI3K isoforms p110α, p110β, and to a lesser extent p110γ, are necessary for processes regulating glucose metabolism and appetite. However, in these models the PI3K activity is chronically reduced. Therefore we analysed the effects of acutely inhibiting PI3K isoforms alone, or PI3K and mTOR, on glucose metabolism and food intake. In the present study impairments in glucose tolerance, insulin tolerance and increased hepatic glucose output were observed in mice treated with the pan-PI3K/mTOR inhibitors PI-103 and NVP-BEZ235. The finding that ZSTK474 has similar effects indicates that these effects are due to inhibition of PI3K rather than mTOR. The p110α-selective inhibitors PIK75 and A66 also induced these phenotypes, but inhibitors of p110β, p110δ or p110γ induced only minor effects. These drugs caused no significant effects on BMR (basal metabolic rate), O2 consumption or water intake, but BEZ235, PI-103 and PIK75 did cause a small reduction in food consumption. Surprisingly, pan-PI3K inhibitors or p110α inhibitors caused reductions in animal movement, although the cause of this is not clear. Taken together these studies provide pharmacological evidence to support a pre-eminent role for the p110α isoform of PI3K in pathways acutely regulating glucose metabolism.
PI3Ks (phosphoinositide 3-kinases) are a family of eight enzymes that are capable of phosphorylating the D3 position of the inositol head group of phosphoinositides. Although all of these enzymes share a high degree of sequence similarity in the kinase domain, there are significant differences in other domains, and so the PI3Ks have been divided into three classes based on structural similarities . The catalytic domain of the PI3K family also shares a high degree of homology with a family of five serine kinases that are referred to as the PIKKs (phosphoinositide kinase-related kinases) . This family includes mTOR (mammalian target of rapamycin) and ATM (ataxia telangiectasia mutated) .
There is a significant body of evidence to indicate that various forms of PI3K play roles in the regulation of glucose metabolism. Class-II PI3Ks are activated by insulin and have also been implicated in mediating insulin-induced increases in glucose uptake [3,4]. The class-III PI3K is not regulated directly by insulin levels, but is regulated by changes in cellular glucose levels . Of the PIKKs, mTOR [6,7] and ATM  have been implicated in regulating pathways involved in glucose metabolism. The class-IB PI3Ks may play a role in regulating insulin secretion in vitro  and in vivo . However, the role of class-IA PI3Ks in mediating the effects of insulin on glucose metabolism has been investigated most extensively . A number of approaches have been used to define the role of specific isoforms of class-IA PI3K in the regulation of glucose metabolism. Overexpression of p110α or p110β is sufficient to induce GLUT-4 (glucose transporter type 4) translocation and glucose uptake in vitro [12–16]. However, high-level expression of PI3Ks does not prove that a particular PI3K isoform is involved, as forced overexpression of p110 causes not only large increases in PtdIns(3,4,5)P3, but also in the other D3 inositides, so it is possible that the effects seen are due to the increase in PtdIns3P, PtdIns(3,4)P2 and PtdIns(3,5)P2 . Global gene KOs (knockouts) of p110α and a KI (knockin) that creates a kinase dead allele of p110α are embryonically lethal, and data on insulin action have only been obtained from studies of heterozygous mice  or tissue-specific PI3K KO models [19,20]. These studies have provided evidence for impairments in glucose metabolism when levels of p110α are chronically reduced. KI mice have also been created in which the kinase activity of p110β is ablated and mice homozygous for this mutation have minor defects in glucose metabolism, implying a role for the catalytic activity of p110β in pathways regulating glucose metabolism [20,21]. However, long-term gene knockdown can cause developmental problems in key glucoregulatory tissues that could contribute to the defects in glucose metabolism, and the results of studies with seemingly similar PI3K KO models do not always produce similar effects on glucose metabolism .
Pharmacological inhibitors offer a more direct means of studying the role of the catalytic functions of the PI3K enzymes . A wide range of small molecule inhibitors targeting class-I PI3K isoforms and mTOR have been developed [2,23,24]. A number of these are selective for particular class-I PI3K isoforms and/or mTOR [25–31]. Some of these inhibitors have been used in a limited range of in vitro studies of insulin action [26,32,33], but there is very little data available on the in vivo effect of these inhibitors on glucose metabolism .
In the present study we have investigated the effects of a range of inhibitors with varying specificity for class-I PI3K isoforms and mTOR on whole-body glucose metabolism in mice. The present study supports a major role for the p110α isoform of PI3K in maintaining glucose homoeostasis in vivo. Surprisingly the data also demonstrate that animals treated with a pan-PI3K inhibitor or p110α inhibitors display a marked reduction in movement.
The GTT (glucose tolerance test), ITT (insulin tolerance test) and PTT (pyruvate tolerance test) studies used male CD1 mice. Metabolic cage studies used male C57Bl/6 mice that were mass and percentage of fat matched into groups using the EchoMRI-100 quantitative magnetic resonance system (Echo Medical Systems). The light/dark cycle was 12 h in all cases and all animals were fed on standard laboratory chow (Harland Teklad). All animal experiments were approved by the Animal Ethics Committees' of Auckland University in New Zealand and the Agency for Science, Technology and Research (A*STAR) Biomedical Science Institutes in Singapore.
Compounds and reagents
The study used ZSTK474 (pan-PI3K inhibitor) , PI-103 (pan-PI3K/mTOR inhibitor) , BEZ235 (pan-PI3K inhibitor/mTOR) , PIK75 (p110α inhibitor) , A66 (p110α-selective inhibitor) , TGX221 (p110β-selective inhibitor) , IC87114 (p110δ-selective inhibitor)  and AS252424 (claimed to be a p110γ-selective inhibitor) . These were synthesized in-house as described previously [28,32] or obtained from Symansis. All compounds were greater than 99% pure by HPLC analysis and NMR data indicated that they were the correct molecules. Unless otherwise stated, other reagents were purchased from Sigma Chemicals.
GTT, ITT and PTT
GTTs, ITTs and PTTs, as well as determinations of insulin levels, were performed as described previously , except that male CD1 mice were used instead of rats. For GTTs and PTTs the mice were starved overnight and for the ITT food was withdrawn 2 h prior to the start of the experiments. Drugs were dosed intraperitoneally 1 h after the end of the dark cycle and 1 h prior to the intraperitoneal dosing with glucose or pyruvate (2 g/kg of body mass) or insulin (0.75 unit/kg of body mass).
Metabolic cage studies
Oxymax/CLAMS (Columbus Instruments) was used to quantify oxygen consumption (V̇O2), CO2 production (V̇CO2), BMR (basal metabolic rate), food intake, water intake and animal movement as described previously . BMR was expressed as a function of lean body mass as recommended in a previous study . All data were normalized to total lean mass using the EchoMRI-100 quantitative magnetic resonance system as described previously . Animals were acclimatized for 24 h in cages and the data were collected over the following 24 h.
Analysis of drug levels
Pharmacological kinetics studies were undertaken in fed CD1 male mice (30 g body mass). Animals were administered with the stated PI3K inhibitors via oral gavage or intraperitoneal injection, and terminal blood samples were collected in EDTA blood collection tubes at 15 min, and 1, 2, 4, 6 and 24 h post-drug exposure. All drugs were dissolved in DMSO. Blood was centrifuged (2000 g for 10 min and 4°C) and plasma isolated for drug quantification. Drug quantification was undertaken using LC-MS/MS (liquid chromatography tandem MS). Briefly, 300 μl of 100% methanol was added to 100 μl of plasma. The samples were gently mixed and centrifuged (2000 g for 10 min and 4°C). The supernatant was removed and 50 μl was added into vials for LC-MS/MS. The ion-source type was ESI (electrospray ionization) with the following conditions: spray voltage (5500 V), sheath gas pressure (50 units), ion sweep gas pressure (0.0 unit), auxillary gas pressure (2 units), capillary temperature [370°C and the capillary offset at 35 V. HPLC kinetex columns were used (100 mm × 3 mm, 2.6u C18(2)-HST; Phenomenex]. The run method was isocratic 10% (0.1% formic acid in water) and 90% methanol. The flow rate was 0.2 ml/min. Retention times were 2.64 min (PI-103), 2.76 min (TGX221) and 2.35 min (IC87114). Unknown concentrations were determined from the standard curve and internal standard.
We have reported previously pharamacokinetic data for BEZ235 and A66 . In the present paper we report pharmacokinetic data for PI-103, TGX221 and IC87114 following oral or intraperitoneal injection (Supplementary Figure S1 at http://www.BiochemJ.org/bj/442/bj4420161add.htm). These studies established that an intraperitoneal dose of 10 mg/kg of body mass gave suitable blood concentrations of drug for short-term metabolic studies.
Effect of inhibitors on whole-body glucose metabolism
The results of the present study show that the pan-PI3K/mTOR inhibitors PI-103 and BEZ235, and the pan-PI3K inhibitor ZSTK474 severely impaired whole-body glucose metabolism in mice (Figures 1–4). The finding that the drugs induced severe impairments in insulin tolerance (Figure 1) suggests they are causing insulin resistance at the level of one or all of the major insulin target tissues, i.e. muscle, liver or fat. The finding that they all increased production of glucose from pyruvate in a PTT (Figure 2) indicates that gluconeogenesis is increased and provides evidence that insulin action in the liver is impaired. Further evidence that the drugs induce insulin resistance comes from the GTT results which show that all three of these pan-PI3K inhibitors induced significant impairments in the ability of the mice to dispose of a glucose load (Figure 3). Of the isoform-selective class-IA PI3K inhibitors, PIK75 and A66 induced significant impairments in the ITT and GTT, and an increase in glucose production during a PTT (Figures 1–3), with TGX221 and IC87114 having only minor effects. AS252424 caused a significant increase in hepatic glucose production (Figure 2H) and a trend towards an impairment in insulin tolerance (Figure 1H). AS252424 was originally described as a p110γ-selective inhibitor, but the findings above lead us to re-evaluate this and we find that it inhibits p110γ with an IC50 value of 17 nM (compared with 30 nM reported in ) and p110α with an IC50 value of 80 nM (compared with 935 nM reported in ). Therefore in vivo this inhibitor is likely to be cross-reacting with p110α.
Acute effect of PI3K inhibitors on insulin tolerance
Acute effect of PI3K inhibitors on hepatic glucose output
Acute effect of PI3K inhibitors on glucose tolerance
Acute effects of PI3K inhibitors on insulin levels during the GTT
One possible explanation for defects in glucose metabolism could be an inhibitory effect on insulin release as such effects have been reported previously in vitro . However, insulin levels did not decrease in the drug-treated animals during the GTT (Figure 4). In fact insulin levels rose in the case of the pan-PI3K inhibitors and PIK75 and A66, in line with the impaired glucose tolerance as would be expected in an insulin-resistant state. Therefore, although a small effect on insulin release can not be ruled out, the drugs certainly don't completely block insulin release.
We were also interested to investigate whether acute administration of these PI3K inhibitors might affect energy expenditure and so we performed metabolic cage studies. These studies did not find any changes in BMR or oxygen consumption (Figure 5). Neither were there major changes in water consumption. However, BEZ235 induced significant reductions in food intake in both the light and dark cycle, whereas PI-103 and PIK75 caused significant decreases in food intake during the light cycle (Figure 5).
Effect of PI3K inhibitors on oxygen consumption, BMR, food consumption and water intake
During the metabolic cage studies, data were also obtained on animal movement. Surprisingly this showed that a number of the inhibitors induced a reduction in movement and that this was an acute effect of the drugs (Figure 6). The reduction in movement was mainly due to a reduction in Z-movement (i.e. up–down movement) (Figure 6). It is notable that the pan-PI3K inhibitors PI-103 and BEZ235, and both of the p110α-selective inhibitors, were the inhibitors that caused the largest effects.
Effects of PI3K inhibitors on movement
The present study shows that the pan-PI3K/mTOR inhibitors PI-103 and BEZ235 have dramatic effects on whole-body glucose metabolism. This extends the findings of Knight et al.  who demonstrated that PI-103 induced impairments in insulin tolerance. The present study also shows that PIK75 caused a serious impairment of glucose metabolism in mice. This also extends the findings of Knight et al.  who only looked at insulin tolerance. They concluded that this was evidence for an important role for p110α in regulating glucose metabolism in vivo. However, PIK75 is a suboptimal inhibitor to use for such studies as it has a number of off-target effects, including inhibition of p110γ and a number of protein kinases. However, the effects of PI-103 and BEZ235 are most likely not to be due to inhibition of mTOR as ZSTK474, which inhibits class-I PI3K isoforms, but not mTOR, has very similar effects. Furthermore, it is unlikely to be due to inhibition of class-II PI3Ks as PI-103 and PIK75 do not inhibit these isoforms . Using a number of different inhibitors with different profiles against protein kinases also guards against the possibility that the effect of the drugs might be due to off-target effects. Furthermore, we find PI-103, BEZ235 and ZSTK474 (Supplementary Table S1 at http://www.BiochemJ.org/bj/442/bj4420161add.htm) and A66  have very low levels of off-target activity.
The present study is the first to examine the effect of a selective p110α inhibitor (A66) on glucose metabolism in vivo. We find that A66 impairs all measures of in vivo insulin action, almost to the same level as the pan-PI3K inhibitors. This provides strong pharmacological evidence that p110α is the most important isoform in the pathways acutely regulating glucose metabolism, and that functional redundancy between PI3K isoforms is unlikely to be a major feature of major pathways regulating glucose metabolism in vivo . The effects of A66 on glucose metabolism are a phenocopy of mice heterozygous for global expression of a kinase-dead form of p110α . However, even though A66 is inhibiting p110α globally, the results of the present study are also remarkably similar to those seen in mice in which the Pik3ca gene had been deleted either acutely or chronically only in liver . Taken together with our PTT results this suggests that a major site of action of the p110α in regulating the effects of insulin on glucose metabolism is in liver.
An area where our studies do not correlate with genetic studies is with regard to p110β inhibition. Two previous studies have analysed the role of p110β in glucose metabolism using genetic models. One of these was a KI model, which created a kinase-dead form of p110β , whereas the other ablated p110β specifically in liver . Both of these models showed impairments of glucose tolerance and insulin tolerance, as well increased hepatic glucose output, which is a similar phenotype to that seen in our studies with pan-PI3K inhibitors. However, we only see minor changes in glucose metabolism in animals treated with TGX221 and these do not achieve statistical significance. This is supported by the studies of Knight et al.  who found that the p110β inhibitor TGX115 did not affect insulin tolerance in mice. One explanation could be that the defects in glucose metabolism seen in the genetic studies  may be caused by long-term effects of the loss of p110β function, which are not seen with acute inhibition of the catalytic activity of the enzyme. Another explanation would be that our results support a non-catalytic role for the p110β in pathways controlling metabolism in the liver, as has previously been suggested .
The finding of the present study that some of the drugs induce a small reduction in food intake differs from previous studies in genetic mouse models [18,38] and our own studies in which isoform selective PI3K inhibitors were directly injected into the brain . Those studies have indicated that a reduction in p110α and p110β activity in the brain actually leads to increased food intake rather than a decrease. It is not clear why the drugs in the present study did not induce a similar effect, but it may be related to the fact that they were administered peripherally and so they may not be crossing the blood–brain barrier to a sufficient extent to achieve such effects. Also, the reduced food intake does not necessarily mean a reduced appetite as the reduction in movement may be preventing the animals from eating.
The reduction in movement seen in mice treated with pan-PI3K inhibitors or the p110α-selective inhibitors is interesting. A similar reduction in movement was observed in mice in which the p110α gene had been deleted in the liver . One interpretation of this would be that p110α plays some previously unsuspected role in regulating movement, but it is also possible that it is a side effect of the off-target actions of the drugs. Further studies will be required to resolve this issue.
In summary, the results of the present study provide strong pharmacological evidence to support the contention that p110α activity is necessary for the pathways regulating glucose metabolism in vivo. The results also show that acute dosing with pan-PI3K and p110α inhibitors have effects on food intake and animal movement, indicating that the these effects should be monitored in human clinical trials using PI3K inhibitors.
ataxia telangiectasia mutated
basal metabolic rate
glucose tolerance test
insulin tolerance test
liquid chromatography tandem MS
mammalian target of rapamycin
PI3K catalytic α polypeptide
phosphoinositide kinase-related kinase
pyruvate tolerance test
All authors contributed to the experimental design. Gordon Rewcastle and Jackie Kendall synthesized the chemicals used. Greg Smith performed the experiments. Peter Shepherd wrote the paper.
We thank Woo-Jeong Lee (Department of Molecular Medicine and Pathology, University of Auckland, Medical School, Auckland, New Zealand) for PI3K assays.
This work was supported by the Maurice Wilkins Centre for Molecular Biodiscovery (to P.R.S.), the Health Research Council of New Zealand [grant numbers 06/062A and 09/388 (to P.R.S. and G.W.R.)], A*STAR Biomedical Research Council (to W.H.) and a joint research grant of A*STAR-HRC of New Zealand [grant number A*STAR-NZ HRC JGC 10-023 (to P.R.S. and W.H.)]. G.S. is funded by a Foundation for Research, Science and Technology post-doctoral research fellowship. P.R.S. is a founder of Symansis.