Antiproliferative Plant and Synthetic Polyphenolics Are Specific Inhibitors of Vertebrate Inositol-1,4,5-trisphosphate 3-Kinases and Inositol Polyphosphate Multikinase*

Inositol-1,4,5-trisphosphate 3-kinases (IP3K) A, B, and C as well as inositol polyphosphate multikinase (IPMK) catalyze the first step in the formation of the higher phosphorylated inositols InsP5 and InsP6 by metabolizing Ins(1,4,5)P3 to Ins(1,3,4,5)P4. In order to clarify the special role of these InsP3 phosphorylating enzymes and of subsequent anabolic inositol phosphate reactions, a search was conducted for potent enzyme inhibitors starting with a fully active IP3K-A catalytic domain. Seven polyphenolic compounds could be identified as potent inhibitors with IC50 < 200 nm (IC50 given): ellagic acid (36 nm), gossypol (58 nm), (–)-epicatechin-3-gallate (94 nm), (–)-epigallocatechin-3-gallate (EGCG, 120 nm), aurintricarboxylic acid (ATA, 150 nm), hypericin (170 nm), and quercetin (180 nm). All inhibitors displayed a mixed-type inhibition with respect to ATP and a non-competitive inhibition with respect to Ins(1,4,5)P3. Examination of these inhibitors toward IP3K-A, -B, and -C and IPMK from mammals revealed that ATA potently inhibits all kinases while the other inhibitors do not markedly affect IPMK but differentially inhibit IP3K isoforms. We identified chlorogenic acid as a specific IPMK inhibitor whereas the flavonoids myricetin, 3′,4′,7,8-tetrahydroxyflavone and EGCG inhibit preferentially IP3K-A and IP3K-C. Mutagenesis studies revealed that both the calmodulin binding and the InsP3 binding domain in IP3K are involved in inhibitor binding. Their absence in IPMK and the presence of a unique insertion in IPMK were found to be important for selectivity differences from IP3K. The fact that all identified IP3K and IPMK inhibitors have been reported as antiproliferative agents and that IP3Ks or IPMK often are the best binding targets deserves further investigation concerning their antitumor potential.

Inositol-1,4,5-trisphosphate 3-kinases (IP3K) A, B, and C as well as inositol polyphosphate multikinase (IPMK) catalyze the first step in the formation of the higher phosphorylated inositols InsP 5 and InsP 6 by metabolizing Ins(1,4,5)P 3 to Ins(1,3,4,5)P 4 . In order to clarify the special role of these InsP 3 phosphorylating enzymes and of subsequent anabolic inositol phosphate reactions, a search was conducted for potent enzyme inhibitors starting with a fully active IP3K-A catalytic domain. Seven polyphenolic compounds could be identified as potent inhibitors with IC 50 < 200 nM (IC 50 given): ellagic acid (36 nM), gossypol (

nM), (؊)-epicatechin-3-gallate (94 nM), (؊)-epigallocatechin-3-gallate (EGCG, 120 nM), aurintricarboxylic acid (ATA, 150 nM), hypericin (170 nM), and quercetin (180 nM). All inhibitors displayed a mixed-type inhibition with respect to ATP and a noncompetitive inhibition with respect to Ins(1,4,5)P 3 . Examination of these inhibitors toward IP3K-A, -B, and -C and IPMK from mammals revealed that ATA potently inhibits all kinases while the other inhibitors do not markedly affect IPMK but differentially inhibit IP3K
isoforms. We identified chlorogenic acid as a specific IPMK inhibitor whereas the flavonoids myricetin, 3,4,7,8-tetrahydroxyflavone and EGCG inhibit preferentially IP3K-A and IP3K-C. Mutagenesis studies revealed that both the calmodulin binding and the InsP 3 binding domain in IP3K are involved in inhibitor binding. Their absence in IPMK and the presence of a unique insertion in IPMK were found to be important for selectivity differences from IP3K. The fact that all identified IP3K and IPMK inhibitors have been reported as antiproliferative agents and that IP3Ks or IPMK often are the best binding targets deserves further investigation concerning their antitumor potential.
To ease studies of the importance of higher InsPs for cell proliferation or other cellular functions in vivo, one possible approach is to block their biosynthesis by pharmacological inhibition of IP3K isoforms and IPMK. Other approaches are knock-out or RNAi experiments shutting down enzymes generating Ins(1,3,4,5)P 4 or higher InsPs. An individual knock-out of IP3K-A in mice resulted in a very weak CNS phenotype in the former case (34). IP3K-B knockout revealed an essential role of the enzyme or its products for T-cell immunity (35). That deletion of the broadly expressed IP3K-B did not generate further phenotypes may be caused by rescue phenomena based on the fact that a total of four enzymes are able to convert Ins(1,4,5)P 3 to Ins(1,3,4,5)P 4 . These considerations led us to start a search for pharmacological IP3K and IPMK inhibitors. We first examined herbal and synthetic compounds belonging to the flavonoids, anthraquinones, coumarins, triphenylmeth-* This work was supported by Grants Ma989 and GRK 336 (to G. W. M.) from the Deutsche Forschungsgemeinschaft. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed. anes, tyrphostins, perylenequinones, and staurosporine derivatives for their ability to inhibit a catalytic domain of avian IP3K-A. This enzyme form, when expressed in bacteria, exhibits V max and K m values indiscernible from those of enzyme purified from native tissue (36) and recent crystal structures of the catalytic domain of IP3K-A have revealed intact folding (37,38). The most potent inhibitors found by this prescreen were analyzed in detail for their mechanism of enzyme inhibition of avian IP3K-A and subsequently tested for their inhibitory effect on all three recombinant mammalian IP3K isoforms and human IPMK. Selectivity profiles derived from inhibitors identified here indeed provide promising lead structures for both broad spectrum and isoform specific inhibitors. The intriguing finding that all inhibitors identified in our search have reported antiproliferative effects suggests a correlation between cell growth and IP3K and/or IPMK action.

EXPERIMENTAL PROCEDURES
Materials-The enzymes SpeI and NdeI were purchased from NEB (Leusden, Netherlands), NheI, XhoI, and Eco31I from MBI (St. Leon-Roth, Germany), Pfu polymerase from Stratagene (La Jolla, CA). The plasmid pZErO 2.1 was obtained from Invitrogen, pGEM T-easy from Promega (Madison, WI), expression vector pET 17b from Novagen (Madison, WI), and expression vector pASK IBA 3 from IBA GmbH (Göttingen, Germany). Escherichia coli strain BL21 (DE3) was obtained from Novagen. Primers were purchased from MWG (Ebersberg, Germany). Quercetin and ellagic acid were obtained from Fluka (Buchs,  Germany), amentoflavone from Roth (Karlsruhe, Germany), hypericin from Calbiochem (Bad Soden, Germany), phenylmethylsulfonyl fluoride, gossypol, ECG, and EGCG from Sigma, and ATA from Aldrich (Steinheim, Germany). All other tested inhibitor compounds were also purchased from the suppliers mentioned above at highest purity available (Ͼ95%). Me 2 SO was of "Uvasol" quality and obtained from Merck (Darmstadt, Germany). L-Lactate dehydrogenase from pig muscle, pyruvate kinase from rabbit muscle, lysozyme, DNase I, NADH, and ATP (special vanadate-free quality) were purchased from Roche Applied Science. CaM was self-prepared (13). Other chemicals were of the highest quality available. Optical enzyme assays were performed in UV/vis-spectral photometers, types Lambda 2, 5, or 20 (PerkinElmer Life Sciences, Ü berlingen, Germany). Semimicro quartz or polystyrene cuvettes with an optical path of 10 mm were employed. Highly amplified digitally converted recordings of M NADH versus time were evaluated.
Cloning, Expression, and Purification of GgIP3K-A Isoform-The full-length cDNA coding for chicken IP3K-A was cloned from chicken erythrocytes (36). Subsequently, a recombinant fragment comprising the catalytic domain and the CaM binding domain (GgIP3K-A CaM/cat ) was expressed in E. coli BL21(DE3)cells and purified by phosphocellulose and CaM affinity chromatography as described (36).
Cloning, Expression, and Purification of HsIP3K-A Isoform-A fragment comprising the catalytic domain including the CaM binding domain (amino acids 165-462: HsIP3K CaM/cat ) of human IP3K-A was amplified by PCR from EST clone 4792595 with the following primers: Sense 5Ј-GGTCTCGAATGAAAAACCACTGGCAGAAGATC-3Ј; antisense 5Ј-GGTCTCCGCGCTTCTCTCAGCCAGGCTGGC-3Ј. The PCR product, cloned into the pGEM T-Easy vector, was cleaved with Eco31I and re-ligated into pASK-IBA3 vector. The HsIP3K-A fragment was bacterially expressed and purified as described for GgIP3K-A (36).
Cloning, Expression, and Purification of HsIP3K-B Isoform-The cDNA for human IP3K-B, HsIP3K-B, was cloned from poly(A ϩ ) RNA prepared from TF1 human myeloic precursor cells (gift from Stefan Horn, University Hospital Hamburg) by RT-PCR (sense primer: 5Ј-GGGACCACTAGTGTGGAGGCGGGAATTCCTTCTGGC-3Ј; antisense primer: 5Ј-GTGGGCCTCGAGGGCGAGTGGGGCATCCTGGGACAT-3Ј), based on the published sequences for this isoform (5). RT-PCR was performed on 1 g of RNA using the Titan TM One Tube-PCR system from Roche Applied Science according to the manufacturer's instructions. The resulting 1467-bp PCR product was cut with SpeI and XhoI and ligated into vector pZErO 2.1, cut with the same enzymes. A fragment representing the CaM-binding and the catalytic domains of HsIP3K-B (HsIP3K-B CaM/cat ) was amplified by PCR (sense primer: 5Ј-CTGGACCATATGTCAGCTTTCCTGCATACCCTGGAC-3Ј; antisense primer: 5Ј-AGATGCGCTAGCCTAGGCGAGTGGGGCATCCTGGGAC-3Ј). This fragment, was cut with NdeI and NheI and ligated into the same sites of the plasmid pET17b. The IP3K-B fragment was bacteri-ally expressed and purified as described for GgIP3K-A (36).
Cloning, Expression, and Purification of RnIP3K-C Isoform-The full-length cDNA coding for rat IP3K-C, RnIP3K-C, was cloned from a rat cDNA library (13). Subsequently, two different recombinant fragments of rat IP3K-C, one comprising the catalytic domain and the calmodulin binding domain (RnIP3K-C CaM/cat ), the other comprising only the catalytic domain (RnIP3K-C cat ), were expressed in E. coli BL21(DE3)RIL cells and purified by phosphocellulose (13). HsIP3K-C CaM/cat was further purified by CaM affinity chromatography (36).
Mutation of GgIP3K-A and HsIPMK-The HsIPMK deletion mutant, HsIPMK⌬ (amino acids 266 -371), and the point mutants GgIP3K-A K268E and GgIP3K-A K272D were created by using PCR-based site-directed QuikChange mutagenesis (39). GgIP3K-A CaM/cat or HsIPMK (see above) cDNA were used as templates and specific oligonucleotides were designed (see Table I). Expression in E. coli and purification was performed as described for wild-type GgIP3K-A CaM/cat and HsIPMK (16,36).
IP3K and IPMK Enzyme Inhibition Assay-The inhibitors were dissolved in Me 2 SO at a concentration of 3-10 mM under light exclusion and stored at Ϫ20°C. Prior to use, aliquots were diluted to concentrations of 0.05-1 mM in Me 2 SO. The enzyme activities of IP3K isoforms and HsIPMK were measured by an optical assay coupling ADP formation to NADH consumption via pyruvate kinase and lactate dehydrogenase reactions and using a wavelength of 339 nm. IP3K assays were routinely performed without Ca 2ϩ -CaM. The final assay mixture was: 0.2 mM NADH, 1 mM phosphoenol pyruvate, 10 mM triethanolamine-HCl, 30 mM KCl, 1 mM dithiothreitol, 500 M ATP, 5 units/ml L-lactate dehydrogenase, 2.5 units/ml pyruvate kinase, pH 7.5. The final volume of the assay was 800 l. GgIP3K-A, HsIP3K-A, HsIP3K-B, RnIP3K-C, or IPMK were added to the mixture (preincubated at 30°C for 10 min) to a final concentration of between 3 and 20 nM. The mixture was further incubated at 30°C for 10 min. After determining the low basal rate of ATP consumption without InsP 3 (due to coupling enzyme contamination with ATPase), maximal enzyme activity was measured by adding 25 M Ins(1,4,5)P 3 . The enzyme concentration was adjusted so that the activity persisted at V max for Ͼ5 min. Potential inhibitors were added stepwise in this period in 1-5-l volumes until maximum inhibition or a concentration of 100 M was reached. The final concentration of Me 2 SO was kept below 3% (v/v). This concentration did not affect IP3K or IPMK activity. To preclude time dependent irreversible effects and a potential photodecomposition of inhibitors, multiple assays with varying initial concentrations of inhibitors were also performed.
Assay for the Influence of Inhibitors on the K m for ATP-Measurements were carried out with GgIP3K-A as described above with the concentration of Ins(1,4,5)P 3 kept at 25 M (saturation); the concentration of inhibitor was varied as well as the initial concentration of ATP, the latter of which was changed from 25 to 800 M.
Assay for the Influence of Inhibitors on the K m for Ins (1,4,5)P 3 -Measurements were carried out with GgIP3K-A as described above with the concentration of ATP kept constant at 500 M (close to saturation); that of Ins(1,4,5)P 3 was varied between 10 M and Ͻ0.1 M by means of single transient recordings (40). In the range of apparent half-maximal substrate concentration (about 1 M), 9 M of the product, Ins(1,3,4,5)P 4 , was present, acting as a competitive type of product inhibitor. Although the derived K m Ј values thus were higher than the true K m (ϳ0.1 M), the assay was adequate to rapidly analyze influence of inhibitors on the K m for Ins(1,4,5)P 3 .
Test of Influence of Inhibitors on Coupling Enzymes-The standard reaction mixture was incubated at 30°C for 10 min in the presence and absence of inhibitor but without adding IP3K. Subsequently, ADP was added to a final concentration of 10 M and the observed rapid consumption of NADH was recorded. In no case was any of the coupling assays detectably inhibited by the tested inhibitors or by their solvent.
Reversal of Enzyme Inhibition by Addition of Triton X-100 and Ca 2ϩ -CaM-The standard reaction mixture was preincubated at 30°C for 10 min, and enzyme inhibition was brought about with inhibitor concentrations up to 5-fold above IC 50 . After 5 or 20 min, neutral CaCl 2 (final concentration 50 M) and CaM (final concentration 4 M), or Triton X-100 (final concentration 0.2%) were added from concentrated stock solutions of 1 mM, 100 M, and 10% (w/v), respectively, and the reaction was followed for another 5 min.
Test of Enzyme Aggregation-IP3K and IPMK inhibition assays with ATA were performed according to the standard conditions but IC 50 values were derived for low (3 nM GgIP3K-A or 9 nM IPMK) and high (9 nM GgIP3K-A or 27 nM IPMK) enzyme concentrations. Assay for an Activating or Inactivating Influence of Photometer Light on Photosensitive Inhibitors-For each compound, four reaction mixtures were incubated at 30°C for 15 min under irradiation with photometer light ( ϭ 339 nm), two in the presence and two in the absence of a concentration of inhibitor about 3-fold above its IC 50. In each pair of mixtures, one only contained Ins(1,4,5)P 3 while the other only contained ATP. After pre-illumination, the assay was started either with 500 M ATP or 25 M Ins(1,4,5)P 3 , respectively. In the pair of reaction mixtures still lacking inhibitor the latter was supplemented together with the second substrate employed to start the reaction.
Test of Covalent Inhibitor Binding by MALDI-Re-TOF MS-GgIP3K-A (final concentration 15 M) was mixed with inhibitors (ATA or THF; final concentration 30 M). These mixtures were incubated (buffer: 10 mM triethanolamine-HCl pH 7.5, 1 mM dithiothreitol, 0.1% Triton X-100, 250 mM NaCl) for 20 min and diluted 1:50 with H 2 O. After another 20 min of incubation, 1 l of this sample was applied onto the MALDI target and allowed to air-dry. Once dry, 1 l of a saturated solution of sinapinic acid (matrix) in acetonitrile/water (1:1 with 0.1% (v/v) trifluoroacetic acid) was pipetted directly onto the sample surface. The solvent was allowed to evaporate and MALDI-Re-TOF MS analysis was performed using the Bruker reflex IV system. Mass spectra (arbitrary intensity versus m/z data, z ϭ 1) were numerically smoothed, baseline subtracted, and submitted to a peak analysis employing the program Peak Fit 4.11 (SPSS Inc). For baseline subtraction, the best D2 function at a tolerance level of 4.5% was employed, smoothing was by the Fast Fourier Transform method with a smoothing factor of 12.5%, and fitting of peaks (maxima, widths, and areas) was by the residuals method employing spectroscopy type peak functions of the Voigt type with area-based optimization at an amplitude threshold of 4.5%.
Test on Topoisomerase II Inhibition-Inhibition of Topo II was assayed by the Topoisomerase II Drug Screening Kit obtained by Topo-GEN. Activity was visualized by the change of linking number in superhelical plasmid DNA. Assays (20 l) containing 0.25 g of plasmid DNA and 0.05 to 50 M inhibitor in 20 l of 120 mM KCl, 10 mM, MgCl 2 , 0.5 mM ATP, 0.5 mM dithiothreitol, and 50 mM Tris-HCl, pH 8.0 (final concentrations) preincubated at 37°C in the dark were started by adding 4 units of Topo II. Me 2 SO, the solvent of inhibitor stocks, was Յ 1% (v/v) in the assay, which had no effect on Topo II activity. After 30 min at 37°C in the dark, the reaction was stopped and subsequent agarose gel electrophoresis of extracted DNA was done according to the manufacturer's suggestion.

RESULTS
Screening for Inhibitors-On the basis of published inhibitors of PI3K, phosphodiesterases, and protein kinases (41-44) a series of substances (listed in Table II) belonging to the flavonoid group were first tested for inhibition of avian IP3K-A. The most potent inhibitors among the flavonols and flavones were quercetin (IC 50 , 180 nM), 3Ј,4Ј,7,8-tetrahydroxyflavone (THF; IC 50 , 290 nM), and myricetin (IC 50 , 540 nM). Starting from the molecular structures of these compounds, further flavonols were tested which differed from the latter with respect to the position and number of hydroxyl groups at the chromone ring system and at the B ring (Table II). Catechin and epicatechin, belonging to the flavan-3-ol group, inhibited IP3K to a low extent (IC 50 , Ͼ100 M). In contrast, gallic acid substituted catechin derivatives like ECG and EGCG (Table II) showed a strong inhibitory effect on IP3K (IC 50 , 94 and 120 nM, respectively). The isoflavones daidzein and genistein (Table II), potent tyrosine kinase inhibitors (45,46), were only very weak inhibitors of GgIP3K-A (IC 50 , 131 and 116 M, respectively). Because of their known inhibitory effect on protein kinases (47), pseudohypericin (Table III) and hypericin (Table IV) were also tested. Hypericin reduced the enzyme activity of IP3K very effectively already in absence of illumination (IC 50 , 170 nM). Irradiation of enzyme plus hypericin just with lab light (15 min) reduced the apparent IC 50 to 75 nM. Photometer light ( ϭ 339 nm) did not enhance the inhibitory potential of this inhibitor. Pseudohypericin and calphostin C, both structural perylenequinone analogues of hypericin (47)(48)(49)(50), showed a markedly higher IC 50 of 2.4 M and 1 M, respectively, than hypericin (Table III, assays done in the dark). Anthraquinones, reported inhibitors of Topo II, were also analyzed (Table III); quinalizarin displayed the most potent inhibition (IC 50 , 2.5 M). Adriamycin (Table III), having been reported by indirect evidence as a potential inhibitor of IP3K (51), did not strongly affect the enzyme activity of IP3K neither in the glycoside form (IC 50 , Ͼ200 M) nor in the aglycone form (IC 50 , 62 M). Based on the apparent structure-function relationship derived for flavonoids, the triterpene gossypol, the phenolic bis-lactone ellagic acid, and the triphenylmethane analogue ATA (Table IV), were analyzed. All three are highly potent inhibitors, the former two being the most potent inhibitors of IP3K identified so far (gossypol, 58 nM; ellagic acid, 36 nM; ATA, 150 nM). Since most of the identified inhibitors structurally exhibit a certain bihandedness of ring structures (see Tables II-IV) we also tested biflavonoids. In fact, the weak inhibitor apigenin (IC 50 , 8.3 M), when dimerized to the related biflavonoid amentoflavone (52), exhibited a 10-fold increased potency (IC 50 , 0.5 M, see Table II). The protein kinase inhibitor rottlerin (53), distantly resembling complex flavonoids, also inhibited IP3K with IC 50 of 1.21 M (Table III). For the most potent inhibitors (Table IV), representative dose response relationships observed at saturating Ins(1,4,5)P 3 and close to maximal ATP concentration are plotted in Fig. 1, A and B. A 100% inhibition was found for hypericin, gossypol, ECG, EGCG, and ATA. In contrast, ellagic acid, quercetin, myricetin, and THF inhibited the enzyme only by about 80% at a concentration of 100 M.
Inhibition of Mammalian IP3K Isoforms and Human IPMK-The most potent inhibitors of avian IP3K-A were now tested for their effects on mammalian HsIP3K-A, HsIP3K-B, RnIP3K-C, and IPMK. All inhibitors identified in the extensive screening with GgIP3K-A also reduced the enzyme activity of mammalian IP3K isoforms, but the inhibitor selectivity of mammalian IP3K isoforms A and C differed strongly from that of isoform B (Table V and Fig. 1C). The flavonoids myricetin, THF, and EGCG had a markedly stronger effect on isoforms A and C than on isoform B. Ellagic acid was the only inhibitor showing preferred selectivity for isoform B. Many of the potent IP3K inhibitors (EGCG, ECG, quercetin, THF, myricetin, hypericin) did not affect IPMK activity at up to 100 M. Obviously, only acidic polyphenolic compounds containing carboxylic acid or other acidic side groups are potent inhibitors of this InsP kinase. The most potent inhibitors were (IC 50 values given) ATA (44 nM) Ͼ rose bengal (620 nM) Ͼ chlorogenic acid (1.15 M) Ͼ ellagic acid (1.37 M). Gossypol lacking an acidic group showed also a weak IPMK inhibition (IC 50 , 3.4 M). Chlorogenic acid was the only substance exclusively inhibiting IPMK but not IP3K.
Inhibition of Topoisomerase II by IP3K/IPMK Inhibitors-Since for some of the identified IP3K/IPMK inhibitors a strong inhibition of Topo II has been reported in the literature (see discussion below), but reported assay conditions were strongly differing, we submitted all identified potent inhibitors to a standardized assay of Topo II inhibition. These data (Table V) revealed that for all inhibitors except for EGCG and ECG the IC 50 for Topo II inhibition was distinctly above the IC 50 for IP3K inhibition.
Partial Irreversibility of Inhibition by IP3K Inhibitors-Triton X-100 was identified as a substance antagonizing the effect of all inhibitors identified with GgIP3K-A. When 0.2% (w/v) Triton X-100 was added to assays containing the respective inhibitors at ՆIC 50 , the previous inhibition of GgIP3K-A was immediately reversed by an extent between 49 and 100%, depending on the inhibitor and the time when Triton X-100 was For optimal visualization, the inhibitory potential, defined as 1/IC 50 with dimension 1/M is depicted. The data presented are the means Ϯ S.E. of IC 50 determinations from three individually measured dose response curves. For data see Table V. added (data not shown). When the detergent was added immediately or 5 min after inhibitor addition, the reversal of inhibition was almost complete (75-100%) for all tested inhibitors (quercetin, myricetin, amentoflavone, ECG, EGCG, ellagic acid, gossypol, hypericin in the dark, ATA). When it was added 20 min after inhibitor addition, the degree of reactivation was distinctly smaller for all of the employed inhibitors except for amentoflavone exhibiting full reversibility. Depending on the inhibitor employed, now up to 51% of the previous inhibition was not reversed by Triton (data not shown). Difference spectroscopy analyses (data not shown) performed with solutions of inhibitors (20 or 50 M) mixed with an increasing concentration Triton X-100 (0 -0.2% w/v) revealed that the reversal of inhibition could mainly be attributed to a direct complexing of inhibitors by Triton X-100 (about 1:1 stoichiometry, K d ,app for Triton ranging from 0.012% (ϳ184 M) to 0.034% (ϳ521 M)) and a decrease of the free inhibitor concentration by this scavenging reaction.
IP3K and IPMK Inhibitors Do Not Induce Enzyme Aggregation-For a number of proteins of pharmacological interest it has been shown that certain inhibitors mainly cause inactivating irreversible protein aggregation instead of a specific inhibition of ligand binding. Such an action is accompanied by decreasing IC 50 values with increasing protein concentration. The observed partial or full reversal of IP3K inhibition by Triton X-100 could be an indication of such inhibitory enzyme aggregation. Therefore, we investigated the dependence of IC 50 values of GgIP3K-A and IPMK for the common inhibitor ATA on enzyme concentration. As shown in Fig. 2A, neither of these two enzymes revealed a decrease of its IC 50 with increasing protein concentration. Instead, a slight increase of IC 50 with increasing enzyme concentration was found as predicted for unchanged binding affinity. Inactivating enzyme aggregation caused by inhibitors of GgIP3K-A and IPMK thus could be ruled out.
Effect of Ca 2ϩ -CaM on IP3K and IPMK Inhibition-Ca 2ϩ -CaM (4 M), when added to GgIP3K-A after 5 min or 20 min of inhibition also partially reversed the previous inhibition (by 10 -28%, the same inhibitors were employed as in the Triton X-100 reactivation experiments, data not shown). This Ca 2ϩ -CaM-induced reactivation never reached the extent observed with Triton X-100 (see above). Again the duration of previous inhibition negatively influenced the degree of reactivation, with the exception of amentoflavone, whose inhibition was fully reversible. Reactivation by Ca 2ϩ -CaM never brought the enzyme activity to the level measured for uninhibited Ca 2ϩ -CaMactivated IP3K-A (about 150% of the activity of IP3K-A alone, see Ref. 7). Other IP3K isoforms exhibited similar partial reactivation by Ca 2ϩ -CaM (data not shown). These data allowed no discrimination between a direct scavenging of inhibitor by Ca 2ϩ -CaM and a "normal type" of Ca 2ϩ -CaM activation of inhibited enzyme.
Irreversible Binding of Inhibitors to IP3K-In order to further clarify the reason of the slow partial irreversible inhibition, MS experiments were performed which could reveal a covalent inhibitor binding or a quasi-irreversible non-covalent "tight binding." Under both conditions, mass spectrometry should reveal increased masses of the enzyme-inhibitor complex previously diluted 50-fold in order to allow dissociation of reversibly bound inhibitor. The analysis revealed that ATA and, to a larger extent, THF in fact caused a mass increase of part of the enzyme protein (16 -35%) by a mass corresponding to a 1:1 or 2:1 stoichiometry of ligand bound to enzyme (Fig. 2B). Thus, the slowly arising irreversible part of enzyme inhibition most likely is due to covalent binding of inhibitors or a tight binding, which is not reversed under MS conditions. Insufficient purity of IPMK precluded corresponding experiments with IPMK. Kinetic Analysis of the Reversible Part of Inhibition of IP3K-The reversible type of interaction of inhibitors with the enzyme was analyzed in detail in GgIP3K-A. In order to perform these analyses at close to steady state conditions uninfluenced by the slow irreversible inhibition (above), only short time enzyme assays (up to 5 min) were employed (see "Experimental Procedures" for details). All potent inhibitors increased the K m for ATP (uninhibited 75 M, inhibited see Table VI). However, none of the inhibitors behaved purely competitive with ATP, but all showed linear mixed-type inhibition with respect to this substrate (54). For inhibition by ellagic acid and hypericin, double reciprocal plots are depicted in Fig. 3. The K m for Ins(1,4,5)P 3 determined in absence of Ins(1,3,4,5)P 4 is about 0.1 M (7). To simplify K m determinations for Ins(1,4,5)P 3 in presence of inhibitors, single transient assays (40) were performed. In these transients, product inhibition by Ins(1,3,4,5)P 4 accumulating during the reaction led to higher apparent K m Ј values for Ins(1,4,5)P 3 (about 1 M, see Ref. 7). But these apparent K m values still reveal a competition of inhibitor with (1,4,5)P 3 binding. As compiled in Table VI, none of the most potent inhibitors showed any influence on the K m Ј value for Ins(1,4,5)P 3 . In Fig. 3 the 1/v versus 1/[InsP 3 ] data for hypericin and ellagic acid are also plotted. The derived linear non-competitive inhibition with respect to InsP 3 proofs that none of the inhibitors apparently interferes directly with the binding of Ins(1,4,5)P 3 to the catalytic site.
Search for Inhibitor Binding Sites-Because the addition of a molar excess of Ca 2ϩ -CaM could partially antagonize the effect of inhibitors on our recombinant IP3K isoforms, all containing a functional CaM binding domain and exhibiting normal activation by Ca 2ϩ -CaM (see Refs. 7,13,36), we investigated whether the CaM binding domain of IP3Ks is involved in inhibitor binding. For these analyses, we compared the inhibitory effect of the potent IP3K-A and IP3K-C inhibitor THF on RnIP3K-C CaM/cat (an enzyme comprising the Ca 2ϩ -CaM binding domain and the catalytic domain) with RnIP3K-C cat (an enzyme comprising only the catalytic domain). The latter enzyme revealed a lower affinity for InP 3 in comparison to the former one and a significantly higher specific activity, the K m for ATP was almost unchanged (see Table VII and Ref. 13). Under conditions where both enzyme forms were saturated with Ins(1,4,5)P 3 , RnIP3K-C CaM/cat in fact showed a 3-fold lower IC 50 value for THF than RnIP3K-C cat . The maximum degree of inhibition was unchanged. This result taken together with the finding that Ca 2ϩ -CaM partly reverses IP3K inhibition indicates a facilitating but not essential involvement of the CaM binding domain in inhibitor binding. That Ca 2ϩ -CaM apparently competes with this interaction and not directly binds and thus scavenges inhibitors can also be deduced from the ineffectiveness of Ca 2ϩ -CaM in de-inhibiting IPMK, which lacks a CaM binding domain (Ref. 16, see Fig. 4A). Because the deletion of the CaM binding domain in IP3K only weakened but not abolished inhibitor binding, there must be further subdomains in the IP3K catalytic domain involved in inhibitor binding. In a search for amino acids involved in inhibitor binding we mutated several amino acids within the catalytic domain of GgIP3K-A CaM/cat and analyzed their inhibition by THF. Two of the obtained point mutants revealed a drastically reduced inhibition by THF (Table VII). Substitution of Lys 268 with Glu led to a 45-fold increase in the IC 50 and substitution of Lys 272 with Asp even led to a 260-fold increase. On the other hand, kinetic parameters of the enzyme with respect to both substrates were nearly unchanged (Table VII). Thus, by substituting either one of theses two lysines in a basic segment involved in InsP 3 and InsP 4 binding (55) and located in the IP-binding lobe (37,38) we have created inhibitor-resistant IP3K-A with full enzymatic activity and normal substrate affinity. As shown in the sequence alignment of IP3K isoforms and IPMK in Fig. 4B, the corresponding segment is missing in HsIPMK but present in all isoforms of IP3K. Both the absence of the CaM binding domain (above) and the missing "InsP 3 binding core domain" (55) may thus contribute to the observed differences in inhibitor selectivity of IPMK as compared with IP3K (see Fig. 1C).
In contrast to IP3K, IPMK harbors a unique basic segment of 105 amino acids, which is inserted between the SSLL motif and the DFG motif (56). In the three-dimensional structure of IP3K-A (37, 38) this insert is extending a surface loop between ␤-strands ␤4 c and ␤5 c in the C-lobe, and functionally it was

FIG. 2. Reaction characteristics of IP3K and IPMK inhibitors.
A, test of inhibitor-induced enzyme aggregation. Inhibition assays with ATA under same conditions as in Fig. 1 were performed at two different enzyme concentrations: 3 nM and 9 nM for GgIP3K-A and 9 nM and 27 nM for HsIPMK, respectively. B, mass spectrometric analyses of inhibited and uninhibited GgIP3K-A. Enzyme was incubated with molar excess of the inhibitors ATA or THF or without inhibitor for 20 min (see "Experimental Procedures"), and, after a 50-fold dilution in water and another 20 min of incubation, MALDI-Re-TOF-MS analysis was performed using the Bruker reflex IV system. Pre-smoothed raw arbitrary intensity versus m/z data (z ϭ 1) are depicted. Masses determined by PeakFit TM analysis (see "Experimental Procedures") and calculated theoretical masses are given in the figure. The vertical line indicates the mass of free GgIP3K-A.
shown to be responsible for nuclear localization of IPMK (Ref. 16, see Fig. 4A). As the enzyme activity of IPMK is preferentially inhibited by acidic polyphenolic compounds, this segment might be involved in inhibitor binding. Therefore we deleted the whole segment. A fully active enzyme, termed IPMK⌬, with almost unchanged kinetic parameters (only K m for ATP and V max were about doubled, see Table VII) resulted. Its IC 50 values for most of the potent IPMK inhibitors, namely ATA, rose bengal, and chlorogenic acid, showed only slight differences as compared with wild-type enzyme. But surprisingly, gossypol, the weak wild-type IPMK-inhibitor (IC 50 , 3.4 M) but strong IP3K inhibitor (IC 50 , 0.06 -0.34 M) revealed a 5-fold lower IC 50 in IPMK⌬ of 0.7 M (Table V). Still more intrigu-ingly, THF and EGCG, inhibitors which did not significantly inhibit wild-type IPMK up to 100 M, became almost as potent inhibitors in IPMK⌬ (IC 50 , 0.52 M and 0.39 M, respectively) as they are for IP3K-A and IP3K-C (IC 50 , 0.18 -0.29 M and 0.12-0.21 M, respectively, Table V). DISCUSSION Structural Features of Potent Inhibitors-With one exception, namely the common inhibitor ATA, we found that certain polyphenolic compounds inhibit IP3K and IPMK in a specific, almost mutually exclusive manner. IP3K isoforms are inhibited most effectively by polyphenolic substances containing multiple (Ն2) aromatic ring systems with more than three   Table VI) displayed a linear mixed-type inhibition with respect to ATP and a linear non-competitive type of inhibition with respect to Ins(1,4,5)P 3 . Inhibition parameters derived from these assays are accumulated in Table VI. phenolic hydroxyls and one or more carbonyl groups, the phenol rings thereby forming a "bi-handed" structure, optimally represented by several flavonoids. IPMK is best inhibited by acidic triphenylmethanes or similar acidic polyphenolic substances. For avian IP3K-A a comparison of structure and inhibitory effects of flavonoids indicates that the degree of inhibition of IP3K is influenced by the position and number of hydroxyl residues of the B ring (Table II). If hydroxyl groups are missing or substituted by amino groups, the inhibition is reduced. The same is true if the B ring is replaced by a long chain substituent, e.g. in stigmatellin (Table II). Crucial for a strong inhibitory effect of all flavonoids tested seems to be the hydroxyl group at position 3 of the pyrone ring (Table II), since its absence appreciably increases IC 50 values. Also, a movement of the B ring from the flavone to the isoflavone position as in genistein increases the IC 50 by a factor of about 100. Similarly, saturation of the C2-C3 bond at the pyrone ring has a markedly negative influence on inhibition. Another essential structural requirement for a potent flavonoid inhibitor of IP3K is the presence of the carbonyl function at the pyrone ring flanked by one or more hydroxyl groups. According to three-dimensional molecular models (built with the help of MOBY v. 1.5) some of the complex ring systems of potent inhibitors (gossypol, flavonoids, ATA) are strongly distorted with angles of up to 60°between the planes of the ring moieties leading to distorted bihanded structures. Propeller-like symmetry in such bihanded molecular structures that is represented by gossypol, ellagic acid, and hypericin appears to favor strong interaction with the enzyme.
Mechanism of Interaction of Inhibitors with IP3K-MS data (Fig. 2B) as well as the slowly developing irreversible inhibition indicate that after a rapid pre-equilibrium there is a subsequent slow covalent or quasi irreversible "tight binding" of the most potent inhibitors. Since all inhibitors exhibiting such slowly irreversible inhibition contain carbonyl or can mesomerically form such group, Schiff base formation with the enzyme is likely, presumably via ⑀-NH 2 groups of lysines. Inhibitors lacking this feature generally exhibited lower inhibitory potency. Slowly irreversible covalent inhibitor actions often lead to a decrease of the apparent IC 50 values with time. If such covalent interaction is due to photoactivation, fully irreversible inhibition can be brought about by a short artificial illumination. Such effects have been described for hypericin (57). We found that already a 15-min irradiation by lab light could photoactivate this IP3K inhibitor and reduce IC 50 by about 3-fold. Enzyme aggregation which could also explain an irreversible inhibition was ruled out by measurements of the concentration dependence of IC 50 values both for IP3K and IPMK.
The reversible part of the inhibition reaction revealed a mixed-

FIG. 4. Differences in the amino acid sequence of IP3K and IPMK and sites involved in inhibitor binding in GgIP3K-A.
A, schematic alignment of the homologous sequences of HsIPMK and GgIP3K-A. Segments strongly conserved between these two enzymes are in dark gray and designated with segment names. The CaM binding domain present in all IP3K isoforms but absent in IPMK is marked in black. Bright gray bars represent the other two regions divergent between IP3K and IPMK, namely the InsP 3 binding core domain unique for all IP3Ks and a polybasic and hydrophobic insert unique in IPMK (16). B, identification of sites of GgIP3K-A involved in inhibitor binding. Partial sequences from the catalytic domains of mammalian IP3K isoforms and HsIPMK comprising the PDGK motif (underlined) being essential for the catalytic activity of all PDKG-type InsP kinases (56) and the InsP 3 binding core domain downstream of this motif (55) are aligned. Amino acids in GgIP3K-A (bold) shown by mutagenesis to be involved in inhibitor binding (Lys 268 and Lys 272 ) are boxed.
type inhibition with respect to ATP and a non-competitive one with respect to Ins(1,4,5)P 3 . Ligands with such complex type of inhibition of protein or substrate kinases were discussed to be advantageous in searches for highly selective inhibitors whereas pure binding competitors of ATP, a substrate broadly used by many cellular enzymes, are less suited to selectively act against only one out of these many enzymes (58).
Inhibitory Selectivity Profiles for Different IP3K Isoforms and IPMK-THF, myricetin and EGCG showed a higher affinity for isoforms A and C in comparison to isoform B. Ellagic acid is the only inhibitor preferring isoform B whereas all other IP3K inhibitors show no preferred selectivity (see Fig. 1C and Table  V). Our screen of commercial compounds provides a promising starting framework of lead structures for a development of even better inhibitors with (i) common selectivity for all IP3K isoforms, (ii) selectivity for isoforms A and C, and (iii) selectivity for isoform B. The fact that no inhibitors markedly differing in selectivity for IP3K-A and IP3K-C could be discovered may be the consequence of the very close homology of their catalytic domains. The inhibitor spectrum of IPMK markedly differs from that of IP3K isoforms. Only acidic polyphenolic compounds potently inhibit this enzyme. Except for gossypol, all other potent IP3K inhibitors lacking an acidic function do not significantly influence IPMK activity. A correlation exists between negative charge of the phenolic substances and IPMK inhibition. ATA, which contains three carboxylic groups, is the best inhibitor; polyphenols containing fewer acidic groups (ellagic acid, rose Bengal, and chlorogenic acid, see Table IV and Fig. 5) are weaker inhibitors, followed by gossypol completely lacking an acidic group. ATA, the best inhibitor of IPMK is also an excellent inhibitor of all IP3K isoforms and thus is the best candidate for a common inhibitor generally preventing Ins(1,4,5)P 3 phosphorylation in cells. Chlorogenic acid, showing no effect on IP3K activity likely because of the lack of the second aromatic ring structure may be useful as a lead to more specific IPMK inhibitors (see Fig. 5).
Structural Basis of Inhibitor Selectivity-The markedly differing inhibitor selectivity of IP3K and IPMK reflects different inhibitor binding sites within these two homologous enzymes (see the structural sketch in Fig. 4A). One important position in the IP3K catalytic domain for inhibitor binding seems to be located within the InsP 3 binding segment, since substitution of one of two closely neighboring lysine residues downstream of the PDKG motif (Ref. 56, see Fig. 4) caused a 45-260-fold decrease of the inhibitory effect. No corresponding site is present in IPMK (see Fig. 4B). The location of corresponding residues in the three-dimensional structure of IP3K-A (37,38) indicates that they are located in the IP lobe but not directly involved in InsP 3 /InsP 4 binding. An interaction of inhibitor with these side chains thus may induce a major conformational change modifying both the IP lobe structure and the nucleotide binding site. The Ca 2ϩ -CaM binding domain, which is unique to all vertebrate IP3K isoforms, is positively, but not essentially, involved in IP3K-inhibitor binding since (i) Ca 2ϩ -CaM addition partially reverses IP3K inhibition and (ii) deletion of the Ca 2ϩ -CaM binding domain causes a weakening of inhibitor action on IP3K but does not abolish it. In IPMK lacking this domain Ca 2ϩ -CaM in fact did not antagonize inhibition. The unique IPMK specific insert of 105 amino acids rich in basic and hydrophobic residues (Fig. 4A) apparently protects IPMK from the inhibitory action of most strong IP3K inhibitors, i.e. is kind of an inhibitor resistance domain, while its deletion renders IPMK more IP3K-like in its inhibitor selectivity. Its steric loca-   tion near the nucleotide binding domain (insert 2 region) in the IP3K-A three-dimensional structure (37) is compatible with this protective role. All three described structural differences between IP3K and IPMK together thus may contribute to the markedly different inhibitor selectivity of these two enzyme families. Alternative Targets of IP3K and IPMK-Since many of the identified IP3K and IPMK inhibitors influence other cellular enzymes or metabolic systems, a detailed examination of reported IC 50 or K i values for these effects obtained in vitro was performed. In most cases, the IC 50 for IP3K and IPMK inhibition is perceptibly lower than for other enzymes (Table VIII). Only two other enzyme families, DNA polymerase and Topo II are inhibited by some of our identified inhibitors with comparable efficiency. Quercetin, EGCG, and ECG inhibit DNA polymerases ␣ and ␤ as potently as IP3K-A and -C. ATA and, according to our analyses, ECG and EGCG are comparably potent Topo II inhibitors (see Tables V and VIII). EGCG has a further potent inhibitory effect on prolyl-endopeptidase (66).
Cellular Effects of Inhibitors of IP3K and IPMK-An actual literature search (see references in Table IX) revealed that most discovered potent IP3K and IPMK inhibitors have been reported to exert antiproliferative effects on cultured cells (designated "in vitro") or in animal experiments/tumor treatment studies (designated "in vivo"). An ongoing screen for further inhibitors of IP3K and IPMK up to now only revealed compounds also inhibiting cell growth in cell lines of non-transformed and transformed phenotype. 2 THF is one of these novel compounds. IP3K and/or IPMK should therefore be put onto the list of potential targets of natural and synthetic antiproliferative drugs and future studies have to show whether IP3K or IPMK are key players of proliferative signal transduction. It will also be of high interest to know how plant cells manage both inhibitory polyphenol biosynthesis and the action of their IPMKs (18).