Direct Effects of Caffeine and Theophylline on p110δ and Other Phosphoinositide 3-Kinases

We investigated the effects of methylxanthines on enzymatic activity of phosphoinositide 3-kinases (PI3Ks). We found that caffeine inhibits the in vitro lipid kinase of class I PI3Ks (IC50 = 75 μm for p110δ, 400 μm for p110α and p110β, and 1 mm for p110γ), and theophylline has similar effects (IC50 = 75 μm for p110δ, 300 μm for p110α, and 800 μm for p110β and p110γ) and also inhibits the α isoform of class II PI3K (PI3K-C2α) (IC50 ≈ 400 μm). However, four other xanthine derivatives tested (3-isobutyl-1-methylxanthine, 3-propylxanthine, alloxazine, and PD116948 (8-cyclopentyl-1,3-dipropylxanthine)) were an order of magnitude less effective. Surprisingly the triazoloquinazoline CGS15943 (9-chloro-2-(2-furyl)(1,2,d)triazolo(1,5-c)quinazolin-5-amine) also selectively inhibits p110δ (IC50 < 10 μm). Caffeine and theophylline also inhibit the intrinsic protein kinase activity of the class IA PI3Ks and DNA-dependent protein kinase, although with a much lower potency than that for the lipid kinase (IC50 ≈ 10 mm for p110α , 3 mm for p110β, and 10 mm for DNA-dependent protein kinase). In CHO-IR cells and rat soleus muscle, theophylline and caffeine block the ability of insulin to stimulate protein kinase B with IC50 values similar to those for inhibition of PI3K activity, whereas insulin stimulation of ERK1 or ERK2 was not inhibited at concentrations up to 10 mm. Theophylline and caffeine also blocked insulin stimulation of glucose transport in CHO-IR cells. These results demonstrate that these methylxanthines are direct inhibitors of PI3K lipid kinase activity but are distinctly less effective against serine kinase activity and thus could be of potential use in dissecting these two distinct kinase activities. Theophylline, caffeine, and CGS15943 may be of particular use in dissecting the specific role of the p110δ lipid kinase. Finally, we conclude that inhibition of PI3K (p110δ in particular) is likely explain some of the physiological and pharmacological properties of caffeine and theophylline.

Caffeine and theophylline are naturally occurring methylxanthine compounds that can be found in micromolar concen-trations in human circulation as a result of dietary intake or pharmacological use. These compounds have been the subject of intense study to determine how they act at physiological concentrations, and a number of effects have been ascribed to these compounds at such concentrations including stimulation of muscle contraction levels (1), anti-inflammatory and immunomodulatory effects (2), alterations in glucose metabolism (3)(4)(5)(6)(7)(8), attenuation of the antilipolytic effect of insulin (3), and induction of apoptosis (9,10). Several mechanisms of action have been identified for these methylxanthines, and these can explain some of the pleiotropic effects these compounds have on cells at their physiologically achievable concentrations. These include their abilities to directly inhibit phosphodiesterases and thus increase cellular cAMP levels, to directly antagonize adenosine receptors, and to cause increases in cytosolic Ca 2ϩ levels (11). While often overlooked, it has been known for more than 25 years that these methylxanthines also have inhibitory effects on phosphoinositide metabolism (12)(13)(14)(15), although the molecular basis of this effect has not been established. In particular, nothing is known about the role of methylxanthines in regulating phosphoinositide 3-kinase activity.
PI3Ks 1 are a family of enzymes that phosphorylate the D3 position of the inositol ring of phosphoinositides, and these enzyme activities are essential in a wide range of cellular processes (16 -18). The PI3Ks are subdivided into four classes (IA, IB, II, and III) that possess lipid kinase activity. The class I enzymes are thought to largely produce phosphatidylinositol 3,4,5-trisphosphate in vivo, which triggers signaling pathways via pleckstrin homology domain-containing molecules, whereas the class III enzymes exclusively produce phosphatidylinositol 3-phosphate, which regulates vesicle trafficking via proteins containing FYVE finger domains (19). The in vivo lipid product of the class II enzymes remains to be defined, but in vitro it is capable of producing phosphatidylinositol 3-phosphate and phosphatidylinositol 3,4 bisphosphate. Class IA PI3Ks also have a protein kinase activity that has been reported to phosphorylate exogenous substrates including insulin receptor substrate 1 (20) and phosphodiesterase 3b (21). The best established function of this class IA protein kinase activity is an autoregulatory role involving phosphorylation of Ser 608 of the p85␣ adapter subunit, which in turn inhibits the lipid kinase activity of the catalytic subunit (22)(23)(24). There is little autophosphorylation on p110␣ itself, but with p110␤, the opposite is observed (it has a greater tendency to autophosphorylate itself, with less phosphorylation of p85). A number of PI3K-related kinases have been identified. These share a high degree of homology in the kinase domain with the lipid kinases, but they only act as protein kinases. Enzymes in the latter group include DNA-dependent protein kinase (DNA-PK), ataxia telangiectasia mutated (ATM), ATM-and Rad3-related kinase (ATR), and mammalian target of rapamycin (mTOR) (16 -18). Recent studies have found that caffeine inhibits ATM (IC 50 ϭ 0.2 mM), ATR (IC 50 ϭ 1.1 mM), and mTOR (IC 50 ϭ 0.4 mM), although DNA-PK is relatively resistant, with an IC 50 of ϳ10 mM (25,26). Furthermore, it has been reported that 5 mM theophylline inhibits the activity of mammalian target of rapamycin (27). However, there have been no studies on the effect of the methylxanthines on members of the PI3K family capable of lipid kinase activity. We have studied this possibility and found that whereas some methylxanthines do not inhibit PI3K activity until millimolar concentrations are used, caffeine and theophylline inhibit the lipid kinase activity of PI3K, particularly that of p110␦, at concentrations that are within the pharmacological and physiological range.

EXPERIMENTAL PROCEDURES
Materials-Polyclonal anti-p85 antibody, raised against the N-terminal Src homology 2 domain, was provided by Prof. Kenneth Siddle (University of Cambridge). PI3K-C2␣ antibodies were raised to a glutathione S-transferase fusion protein corresponding to amino acids 2-337 of the human PI3K-C2␣ sequence. Purified DNA-PK was from Promega. Recombinant protein kinase CK2 and recombinant 4EBP1 (PHAS-I) were from Calbiochem. Phospho-Akt (Ser 473 ), phospho-Akt (Thr 308 ), phospho-p42/44 MAPK (Thr 202 /Tyr 204 ) E10, Akt, and p42/44 MAPK antibodies were purchased from Cell Signaling Technology. Recombinant baculoviruses were provided by Prof. M. D. Waterfield (Ludwig Institute for Cancer Research, University College London, London, UK). All other chemicals, including xanthine derivatives, were obtained from Sigma, and they were of the highest grade available. Recombinant p110␥ was the gift of Dr. Roger Williams.
For experiments on PI3K expressed in mammalian cells, CHO-IR cells were serum-starved for 16 h before insulin stimulation and then treated with 100 nM insulin for 10 min at 37°C. Cells were washed once in phosphate-buffered saline and lysed in a buffer containing 50 mM Tris-HCl, pH 7.4, 5 mM EDTA, 150 mM NaCl, 50 mM NaF, and 1% Triton X-100 supplemented with 2 g/ml aprotinin, 1 M pepstatin, 1 ng/ml leupeptin, 2 mM phenylmethylsulfonyl fluoride, and 1 mM sodium orthovanadate. Immunoprecipitations were performed from the Tritonsoluble fraction using the indicated antibodies diluted 1:100. Immune complexes were collected with protein A-agarose beads, washed three times with lysis buffer, followed by three washes with kinase assay buffer, and then processed for lipid kinase assays.
Preparation, incubation, and protein extraction of rat soleus muscles were performed as described previously (28).
Immunoblotting using the phosphospecific antibodies was performed according to the manufacturer's instructions. Visualization was performed with ECL. Images were analyzed with a Fuji LAS-1000 Luminescent Image Analyzer and Fuji Image Gauge software.
Protein Kinase and PI3K Lipid Kinase Assays-PI3K autophosphorylation assays were performed as described previously (22,23). For protein kinase CK2 assays, 2 units of recombinant CK2 were used per assay. For DNA-PK assay, 2 units of purified DNA-PK and 500 ng of linearized double-stranded plasmid DNA were used per assay. Kinase assays were performed in a buffer consisting of 50 mM Tris-HCl, pH 7.4, 50 mM NaCl. 1 g of recombinant 4EBP1 was used as a substrate. Reactions were initiated by adding a mix of ATP (final concentration of 100 M plus 0.5 Ci of [␥-32 P]ATP/assay) and MgCl 2 /MnCl 2 (final concentration, 10 mM each), incubated for 20 min at 25°C, and terminated by the addition of 4ϫ electrophoresis sample buffer. The reaction prod-ucts were analyzed by SDS-PAGE and autoradiography.
PI3K lipid kinase assays were performed in a total volume of 50 l in a buffer containing 50 mM HEPES, pH 7.4, 100 mM NaCl, 1 mM dithiothreitol, 5 mM MgCl 2 , and 100 M ATP (plus 0.1 Ci of [␥-32 P]ATP/ assay) using 200 g/ml phosphatidylinositol as a substrate. Reactions were incubated at 25°C for 20 min and terminated by the addition of 100 l of 0.1 M HCl and 200 l of chloroform:methanol (1:1). The mixture was vortexed, and the phases were separated by centrifugation at 10,000 ϫ g for 2 min. The aqueous phase was discarded, and the lower organic phase was washed with 80 l of methanol:1 M HCl (1:1). After centrifugation, the aqueous phase was again discarded, and the lower organic phase was evaporated to dryness. Reaction products were resuspended in 30 l of chloroform:methanol (4:1) and spotted onto thin layer Silica Gel 60 plates (Merck) pretreated with 1% oxalic acid and 1 mM EDTA in water:methanol (6:4). TLC plates were developed in chloroform:methanol:4 M ammonia (9:7:4). Images of radiolabeled protein and lipid products were analyzed using a Fuji FLA-2000 phosphorimager and Fuji Image Gauge software.
Glucose Transport Assay-CHO-IR cells were starved overnight on plain Ham's F-12. The cells were preincubated for 30 min at 37°C with the indicated concentrations of either caffeine or theophylline in Krebs-Ringer-HEPES buffer with bovine serum albumin at pH 7.4 (136 mM NaCl, 4.7 mM KCl, 1.25 mM MgSO 4 , 1.2 mM CaCl 2 , 2 mg/ml bovine serum albumin, 20 mM HEPES, pH 7.4). The cells were then stimulated with 100 nM insulin for 30 min at 37°C. After 20 min of stimulation, the assay was initiated by adding 100 mM 2-deoxy-D-glucose containing 0.5 Ci of 2-deoxy-D-[ 3 H]glucose and stopped 10 min later by washing the cells three times with ice-cold phosphate-buffered saline. The cells were then solubilized in 0.05% SDS, and the incorporated radioactivity was determined by scintillation counting.
To look at the effect of these compounds on endogenous PI3K activities, the effects of caffeine and theophylline were tested on lipid kinase activity in class IA PI3K p85 adapter subunit and PI3K-C2␣ immunoprecipitates from CHO-IR cells. Both caffeine and theophylline inhibited PI3K activity in these immunoprecipitates with a similar potency (Fig. 2).
We next wanted to investigate whether the core structure of the xanthines could be useful in deriving more potent and potentially isoform-specific inhibitors of PI3K. None of the four other xanthine derivatives used (3-isobutyl-1-methylxanthine, alloxazine, PD116948, and 3-propylxanthine) were as effective as caffeine and theophylline, showing that there is specificity in the inhibitory effect of methylxanthines toward PI3K (Table I). However, surprisingly, CGS15943, a triazoloquinazoline compound somewhat distantly related to the xanthines, inhibits p110␦ in vitro with an IC 50 of Ͻ10 M and also inhibits p110␣ and p110␤, albeit with an IC 50 in excess of 100 M. Although this compound has also been described as an inhibitor of the human adenosine A3 receptor (30), it does suggest that this compound could be useful in dissecting the specific role of p110␦ in cells and in developing even more potent PI3K inhibitors.
Previous studies have shown that caffeine inhibits the protein kinase activity of the PI3K-related kinases mammalian target of rapamycin, ATM, and ATM-and Rad3-related kinase with a potency similar to that with which it inhibits the lipid kinase activity of the class IA and class II PI3Ks. Therefore, we have investigated whether the protein kinase activity of the class IA PI3Ks is also inhibited by theophylline. We found that the serine kinase activities of p110␣ and p110␤ are both inhibited by theophylline, but with a significantly lower potency compared with that observed for the lipid kinase activities (Fig.  3A). There is also a clear difference in the effect of theophylline on the serine kinase activities of the two class IA isoforms, with p110␤ autophosphorylation being more sensitive to inhibition (IC 50 Ϸ 3 mM) than the p110␣-mediated phosphorylation of p85␣ (IC 50 Ϸ 10 mM) (Fig. 3A). The fact that there is differential inhibition of the protein and lipid kinases suggests that the mechanism by which the methylxanthines inhibit PI3Ks is likely to differ from that used by the classical PI3K inhibitors wortmannin and LY294002. We also found that theophylline and caffeine inhibit DNA-PK, but with an IC 50 of Ͼ3 mM.
Other PI3K inhibitors have been described as having direct inhibitory effects on some protein kinases. Therefore, the effect of methylxanthines on other protein kinases was also investigated. The ability of insulin to stimulate phosphorylation of ERK1 and ERK2 was not inhibited by theophylline at concentrations as high as 10 mM (Fig. 4A). The protein kinase CK2 is known to be potently inhibited by another PI3K inhibitor, LY294002 (31), but we found that concentrations of caffeine or theophylline as high as 5 mM had no effect on CK2 activity (Fig.  4B). This suggests that these methylxanthines are not promiscuous inhibitors of protein kinases.
Our findings suggested that the physiological actions of PI3K may be mediated in part by their ability to inhibit PI3K. Activation of PKB/Akt is a very convenient physiological read-out of the activation of the PI3K pathway. PKB is activated by phosphorylation of Thr 308 in the activation segment and Ser 473 in the hydrophobic motif. Phosphorylation of Thr 308 is carried out by PDK1, whereas Ser 473 is phosphorylated by a distinct yet unidentified kinase termed PDK2 (32). Full activation of the enzyme requires phosphorylation of both sites, and the molecular mechanism was demonstrated recently by analysis of the crystal structures of the unphosphorylated and Thr 308phosphorylated states of the PKB kinase domain (33). Therefore, we assessed the effect of caffeine and theophylline on insulin-stimulated phosphorylation of Thr 308 and Ser 473 . Insulin-stimulated phosphorylation of Thr 308 was found to be inhibited by both caffeine and theophylline (Fig. 5A), and this could correlate with impaired PDK1 activation caused by PI3K inhibition. Ser 473 phosphorylation was also inhibited with the same potency as Thr 308 by both caffeine and theophylline (Fig.  5B). Both the theophylline and caffeine treatment of CHO-IR cells resulted in strong inhibition of insulin-stimulated PKB Ser 473 phosphorylation, with 50% inhibition occurring at 0.5 mM for theophylline and 1 mM for caffeine in CHO-IR cells (Fig.  5C). The inhibitory effect on PKB activation is not restricted to cell culture models because insulin stimulation of Ser 473 PKB phosphorylation was inhibited by theophylline and caffeine in intact rat soleus muscle with even higher potency than that seen in CHO-IR cells, i.e. with an IC 50 in the vicinity of 0.1 mM (Fig. 6). Insulin stimulation of glucose transport is another cellular process requiring activation of PI3K (17), and we found that both caffeine and theophylline block insulin-stimulated glucose transport in CHO-IR cells (Fig. 7). However, the inhibitory effects on glucose transport require a higher concentration of theophylline and caffeine than that required for inhibition of PKB activation. DISCUSSION Our results identify methylxanthines as a novel class of PI3K inhibitor with the potential for isoform selectivity toward p110␦, and this is the first description of an inhibitor with selectivity for this isoform. This finding is of direct physiological relevance because methylxanthines such as caffeine and theophylline are found in a variety of foodstuffs and beverages and have also been widely used as pharmacological agents. Concentrations of the methylxanthines reach relatively high levels in the circulation before their effects are manifest. Caffeine concentrations can reach 50 -100 M in humans, whereas in the case of theophylline, an effective pharmacological con-centration in humans is around 100 M, and concentrations can reach 500 M before being highly toxic (2,7,11). Additionally, these agents have also been used extensively as experimental reagents for in vitro experiments on cells and tissues. Therefore, there has been a great deal of interest in understanding the mechanisms by which these agents are exerting their effects in this micromolar concentration range. The most potent effect described for methylxanthines is the ability to antagonize adenosine receptors because the K i for this effect is in the low micromolar range (11). Inhibition of phosphodiesterases and regulation of calcium metabolism are two other properties widely implicated in the effects of caffeine and theophylline. However, the inhibitory effects of methylxanthines on phosphodiesterases have an IC 50 in the high micromolar range, whereas the effects on calcium metabolism have an IC 50 in the low millimolar range (11). This means that the inhibitory effects of caffeine and theophylline on PI3Ks described here are at least as potent as the effects on phosphodiesterases and ryanodine receptors, and the effect on p110␦ is in the same range as the antagonistic effect on adenosine receptors. There-  fore, the results of the current study suggest that inhibition of PI3K is likely to contribute to the effects of the methylxanthines observed at physiological concentrations. For example, inhibition of PI3K would be expected to block insulin-mediated glucose metabolism (17), and a severely insulin-resistant state has been found in cases of pharmacological toxicity of theophylline (2).
The inhibition of PI3K by methylxanthines could have beneficial therapeutic effects in at least three scenarios. Firstly, it has been reported that knockout of class IB PI3K activity reduces ADP-induced platelet aggregation (34), indicating that inhibition of PI3K may be useful in treating thrombosis. Sec- ondly, caffeine has been reported to sensitize cells to apoptosis, and theophylline has been described to promote apoptosis (9,10,25), indicating that these agents may be useful in treating cancers directly or potentiating the actions of cytotoxic drugs. In part, the proapoptotic effects are likely to work through inhibition of ATM-and Rad3-related kinase and ATM (25,26). However, our finding that methylxanthines also inhibit class IA PI3Ks with a similar potency identifies a mechanism that could act directly to increase the rate of apoptosis because class IA PI3Ks play an important role in preventing apoptosis (19). In this context, promiscuous inhibitors affecting PI3K, ATM, and ATM-and Rad3-related kinase could be useful in treating cancer (35). Finally, the p110␦ isoform of class IA PI3Ks is found in a limited subpopulation of cells such as leukocytes and melanoma cells (36), and a specific inhibitor of this isoform might prevent inflammation or cell migration (37). However, long-term inhibition of p110␦ is likely to be undesirable because animals in which the p110␦ gene is deleted show reduced B-and T-cell function and also develop inflammatory bowel disease (38).
There are currently only a limited number of suitably specific inhibitors available for the study of PI3K, and these are lacking in isoform selectivity. Isoform-selective inhibitors are useful in a range of ways. For example, these would allow quantitation of the relative contribution of the different class IA isoforms in signaling complexes and dissection of the role of the different isoforms in particular cellular processes, and ultimately, once the role of different isoforms is clearly understood, isoformselective inhibitors are likely to be of use as therapeutic agents for treating diseases such as cancer and thrombosis. Two PI3K inhibitors have been widely used: LY294002, derived from quercetin (39), and the fungal metabolite wortmannin (40). These are relatively specific inhibitors of lipid kinase because in a test of a panel of over 25 protein kinases, wortmannin only cross-reacted with smooth muscle myosin light chain kinase at concentrations required to inhibit PI3K, whereas LY294002 shows some inhibition of protein kinase CK2 and glycogen synthase kinase-3 at low doses (31). However, LY294002 and wortmannin in general lack selectivity against different classes and/or isoforms of PI3K, with only the ␣ isoform of the class II PI3K being resistant to wortmannin and LY294002 (41,42). This means that methylxanthines may be of use due to the fact that compared with traditional PI3K inhibitors, the methylxanthines differ in their inhibitory profile toward different classes of PI3K. For example, it is notable that the ␣ isoform of the class II PI3K and class IA PI3Ks are inhibited by theophylline and caffeine at similar concentrations. Furthermore, DNA-PK is relatively resistant to inhibition by these methylxanthines, whereas the class IA PI3Ks and DNA-PK are similarly sensitive to inhibition by classical class IA PI3K inhibitors LY294002 and wortmannin (43). Additionally, caffeine, theophylline and, in particular, CGS15943 show selectivity toward p110␦. It was reported previously that p110␣, p110␤, and p110␦ are all equally sensitive to wortmannin and LY294002 (36), so the methylxanthines are likely to be of use in dissecting the role of p110␦ in the limited range of cell types (e.g. leukocytes) in which it is expressed (36). Finally, the protein kinase activities of class IA PI3K are less sensitive than the lipid kinase activity to inhibition by caffeine and theophylline. Therefore the differential inhibitory properties of the methylxanthines are likely to be of use in ongoing studies to define the roles of the different PI3K family members.
Our results also indicate that inhibition of PI3K is not a general property of all xanthines or all molecules containing the purine ring because at least four other xanthines tested were significantly less effective at inhibiting PI3K, and it is known that adenosine has little effect on PI3K activity, although it does potently inhibit some forms of phosphoinositide 4-kinase (44).
The pleiotropic effects of caffeine and theophylline on systems, including the antagonistic effects on adenosine receptors, inhibition of phosphodiesterases, and stimulation of calcium mobilization, mean that it is more difficult to prove that these compounds are acting to inhibit PI3Ks in vivo. However, our findings that both theophylline and caffeine block insulin-induced activation of PKB and glucose uptake at concentrations similar to those required to inhibit PI3K in vitro provide strong evidence that these methylxanthines are able to inhibit PI3K in cells at physiologically relevant concentrations. It seems unlikely that that these effects are mediated by the other known activities of these compounds. For example, the effect of methylxanthines to raise intracellular calcium levels, as seen most dramatically in muscle (1), would actually increase glucose transport rather than inhibiting it and in fact may explain why stimulation of glucose transport is slightly less sensitive to attenuation by the methylxanthines in our experiments. Methylxanthines also inhibit phosphodiesterases, causing a rise in cAMP, but it has previously been shown that such a rise would not inhibit insulin stimulation of PKB or glucose transport (27). Furthermore, in adipocytes, the concentrations of theophylline required to inhibit glucose transport differ significantly from the concentrations required to stimulate the cAMP-mediated increase in lipolysis via increases in cAMP (3). Finally, the effect of methylxanthines to antagonize adenosine actions could potentially have effects on insulin signaling because adenosine can potentiate insulin-induced activation of PKB and stimulation of glucose transport in a range of tissues (45,46). However, inhibition of glucose transport by theophylline is not fully mimicked by the 8-phenyltheophylline, which is a pure adenosine receptor antagonist (6). Furthermore, insulin still stimulates PKB and glucose transport in the absence of adenosine (45,46), so the adenosine receptor antagonism could only explain part of the inhibitory effects of theophylline and caffeine on insulin stimulation of cellular events.
In conclusion, we find that caffeine and theophylline inhibit both class I and class II PI3Ks, with the most potent effects seen against p110␦. The finding that these methylxanthines are less effective against the protein kinase activity of the class IA PI3Ks provides a novel mechanism for dissecting the roles of the protein and lipid kinase activities of these enzymes. The finding that the inhibitory effects of theophylline are within a physiologically achievable concentration range suggests that methylxanthines or more potent derivatives of them could be used to treat conditions such as inflammation, cancer, and thrombosis in which suppression of PI3K would be predicted to have therapeutic benefit, but long-term administration may also contribute to the development of inflammatory bowel disease.