F. Erdődi, Department of Medical Chemistry, Medical and Health Science Center, University of Debrecen, Nagyerdei krt. 98, H-4032 Debrecen, Hungary Fax: +36 52 412566 Tel: +36 52 412345 E-mail: email@example.com
Protein phosphatase-1 (PP1) and protein phosphatase-2A (PP2A) are responsible for the dephosphorylation of the majority of phosphoserine/threonine residues in cells. In this study, we show that (–)-epigallocatechin-3-gallate (EGCG) and 1,2,3,4,6-penta-O-galloyl-β-d-glucose (PGG), polyphenolic constituents of green tea and tannins, inhibit the activity of the PP1 recombinant δ-isoform of the PP1 catalytic subunit and the native PP1 catalytic subunit (PP1c) with IC50 values of 0.47–1.35 μm and 0.26–0.4 μm, respectively. EGCG and PGG inhibit PP2Ac less potently, with IC50 values of 15 and 6.6 μm, respectively. The structure–inhibitory potency relationships of catechin derivatives suggests that the galloyl group may play a major role in phosphatase inhibition. The interaction of EGCG and PGG with PP1c was characterized by NMR and surface plasmon resonance-based binding techniques. Competitive binding assays and molecular modeling suggest that EGCG docks at the hydrophobic groove close to the catalytic center of PP1c, partially overlapping with the binding surface of microcystin-LR or okadaic acid. This hydrophobic interaction is further stabilized by hydrogen bonding via hydroxyl/oxo groups of EGCG to PP1c residues. Comparative docking shows that EGCG binds to PP2Ac in a similar manner, but in a distinct pose. Long-term treatment (24 h) with these compounds and other catechins suppresses the viability of HeLa cells with a relative effectiveness reminiscent of their in vitro PP1c-inhibitory potencies. The above data imply that the phosphatase-inhibitory features of these polyphenols may be implicated in the wide spectrum of their physiological influence.
hexahistidine-tagged recombinant protein phosphatase-1 catalytic subunit δ-isoform
protein phosphatase-2A catalytic subunit
surface plasmon resonance
saturation transfer difference
The biologically active constituents of tannins and green tea are water-soluble polyphenolic compounds that are known to evoke a wide range of physiological responses [1,2]. The common elements in the structures of these compounds are the polyol units (e.g. d-glucose or flavanol) coupled to one or more polyphenols, such as gallic acid and/or other polyphenolic rings (Fig. 1). 1,2,3,4,6-Penta-O-galloyl-β-d-glucose (PGG) is one of the main constituents of Aleppo oak tannin , and it is also a major component of gallotannins . (–)-Epigallocatechin-3-gallate (EGCG) is the most abundant polyphenol in green tea , which contains other catechin components (see Fig. 1 for structures), such as (–)-epicatechin-3-gallate (ECG), (–)-epigallocatechin (EGC), (–)-epicatechin (EC), and (–)-gallocatechin-3-gallate (GCG). EGCG has been thought to be responsible for the chemopreventive properties of this beverage, and it was shown to induce cell cycle arrest and apoptosis in many cancer cells without affecting normal cells . Previous studies established that EGCG may exert its antitumor effect by affecting various protein kinases in distinct signaling pathways , implying a role in the regulation of the phosphorylation level of cellular proteins. Several kinases important in the regulation of cell proliferation and survival [e.g. extracellular signal-related kinase (ERK)1/2, Akt/protein kinase B, p38 mitogen-activated protein kinase and cyclin-dependent kinase 2] are directly inhibited by EGCG . PGG exerted similar biological effects to EGCG with respect to changing the phosphorylation status of proteins [8–10], decreasing the viability of malignant cells, and causing cell cycle arrest [2,11,12]. On the other hand, the effects of catechin derivatives and PGG on the protein dephosphorylation processes, and therefore their influences on the protein phosphatases, have remained largely unexplored.
We reported earlier that gallotannin inhibited the activity of protein phosphatase-1 (PP1) and protein phosphatase-2A (PP2A) catalytic subunits (PP1c and PP2Ac) in in vitro assays and exhibited partial selectivity towards PP1c . Short-term (30 min) incubation of A549 cells in the presence of gallotannin alone or in combination with cytokines increased the phosphorylation level of c-Jun NH2-terminal kinase (JNK) substantially, and that of ERK1/2 and p38 to a lesser extent. Concomitantly, an increase in the phosphorylation level of the respective transcription factors (c-Jun, CREB, and ATF-2) phosphorylated by these kinases was also apparent . Long-term incubation of cells with EGCG (2–24 h) had contrasting effects on PP1 and PP2A expression and activity. EGCG acted synergistically with ibuprofen to increase PP1c expression and activity in prostate cancer cell lines, whereas EGCG alone was without effect . In contrast, PP2Ac expression and activity were decreased by 24 h of EGCG treatment in JB6 cells, and these changes resulted in elevated p53 phosphorylation and positively influenced p53-dependent apoptosis . It was also shown that the 67-kDa laminin receptor and eukaryotic translation elongation factor 1A ensured responsiveness of tumor cells to EGCG through the induction of the dephosphorylation of myosin phosphatase target subunit-1 (MYPT1) at the inhibitory phospho-Thr696 site, thereby activating myosin phosphatase (MP) . MP is a PP1-type holoenzyme consisting of PP1c, the 130/133-kDa MYPT1 subunit, and a 20-kDa protein of unknown function . The above studies  suggested that MP activation may play a central role in the cell death-inducing effect of EGCG on cancer cell lines.
The present study showed that PGG and EGCG preferentially inhibit PP1c as compared with PP2Ac; therefore, we examined the structural bases for the interactions of these molecules with PP1c. PP1 is present in cells in holoenzyme forms, in which PP1c is associated with regulatory and/or targeting subunits [18,19], which are implicated in the cellular localization and substrate specificity of PP1c and are often involved in the regulation of phosphatase activity. The activities of PP1 and PP2A, as well as those of a few related enzymes (protein phosphatase-4 and protein phosphatase-5) are potently inhibited by natural toxins such as okadaic acid (OA), microcystin-LR (MC-LR), calyculin-A (CLA) and tautomycin (TM) at nanomolar concentrations [20,21]. The crystal structures of PP1c in complex with MC-LR , OA , CLA  or TM  have been determined. These studies revealed that the catalytic center of PP1c lies at the bifurcation point of a Y-shaped groove consisting of C-terminal, acidic and hydrophobic ‘arms’ that are involved in binding of the substrate and inhibitor molecules. The toxin inhibitors bind to the hydrophobic groove, the catalytic site, and part of the β12–β13 loop in PP1c, and these interactions contribute to their inhibitory potency. In our present study NMR and surface plasmon resonance (SPR)-based measurements proved direct binding of PGG and EGCG to PP1c. EGCG docks at the surface of the hydrophobic groove close to the catalytic center of PP1c, overlapping the binding surface of toxin inhibitors. EGCG and GCG (the two epimers) and ECG (lacking a hydroxyl group from the polyphenolic ring) suppress PP1c activity with similar effectiveness, whereas deletion of the galloyl ring (EGC or EC) decreases the inhibitory potency dramatically. PGG and EGCG and its derivatives suppress the viability of HeLa cells with similar effectiveness to that of their phosphatase inhibitory potencies, implying that they may also act as regulators of the phosphorylation level of key proteins involved in the mediation of cell survival.
Effect of polyphenolic compounds on the phosphatase activity of PP1c and PP2Ac
PGG is a constituent of gallotannins, so our earlier observation that gallotannins inhibited PP1c and PP2Ac  prompted us to examine whether this compound itself could be responsible for the phosphatase-inhibitory effect. To this end, we performed phosphatase activity measurements with PP1c or PP2Ac in the presence of different concentrations of PGG (Fig. 2A–C), using phosphorylated 20-kDa light chain of smooth muscle myosin (MLC20) as substrate. Two different PP1c preparations, native PP1c prepared from rabbit skeletal muscle, and hexahistidine-tagged recombinant PP1c δ-isoform (PP1cδ) expressed in Escherichia coli and purified via Ni2+–agarose chromatography , were applied in these experiments. Note that the purified phosphatase preparations were diluted for the assays without BSA, which is generally used at 0.3–1 mg·mL−1 for stabilization of the diluted enzyme, because it is known that the polyphenolic compounds also bind to BSA [27,28], and it would therefore decrease the effective concentrations of the inhibitors. The results shown in Fig. 2A,B indicate that PGG effectively inhibited both native and recombinant PP1c, with similar IC50 values of 0.4 and 0.26 μm, respectively (see also Table 1 for a summary of IC50 values). PGG proved to be much less inhibitory (Fig. 2C) on PP2Ac (IC50 = 6.6 μm). These data suggest that PGG is a partially selective inhibitor of PP1, and that at micromolar concentrations it can distinguish between PP1c and PP2Ac activity in vitro. The inhibitory action of PGG may arise from its metal ion-complexing ability, resulting from its polyphenolic nature. To refute this hypothesis, the activity of PP1cδ was tested in the presence of 1 mm MnCl2. Addition of Mn2+ did not significantly influence the inhibitory effect of PGG on PP1c (data are not shown), suggesting that the inhibitory effect is not attributable to the extraction of the metal ions from the enzyme. In addition to PGG, the influence of Aleppo tannin (a natural tannin) and gallic acid (one of the building blocks of PGG) on the phosphatase activity of both PP1c forms and PP2Ac was also tested (Fig. 2A–C). Aleppo tannin decreased the activity of PP1c and PP1cδ (IC50 = 0.10 μm and IC50 = 0.11 μm) even more effectively than PGG, whereas gallic acid was much less inhibitory (IC50 > 30–100 μm). It is interesting to note that the largest difference (∼ 100-fold) in sensitivity to inhibition of PP1c and PP2Ac was found in the case of Aleppo tannin.
Table 1. IC50 values of polyphenolic compounds for the inhibition of PP1c, PP1cδ, and PP2Ac. IC50 values are mean ± standard error of the mean (n = 3–5).
0.40 ± 0.07
0.26 ± 0.03
6.60 ± 1.30
1.35 ± 0.20
0.47 ± 0.10
15.00 ± 2.50
1.10 ± 0.10
1.15 ± 0.19
1.60 ± 0.18
1.50 ± 0.20
0.10 ± 0.02
0.11 ± 0.02
10.50 ± 1.70
The effects of flavanol ring-based polyphenols (EGCG and its derivatives) on the activity of PP1c, PP1cδ and PP2Ac were also assessed (Fig. 2A–C). The IC50 values for phosphatase inhibition are summarized in Table 1. The epimers EGCG and GCG suppressed the activities of PP1c and PP1cδ similarly, whereas GCG appeared to inhibit PP2Ac less potently than did EGCG. ECG, which lacks one hydroxyl from the polyphenol ring (attached at position 2 of the flavanol ring), had a slightly decreased inhibitory potency for the PP1c forms (IC50 = 1.5–1.6 μm) as compared with EGCG and GCG, but it inhibited PP2Ac slightly (IC50 > 100 μm). EGC does not include the galloyl group (attached at position 3 by an ester bond; Fig. 1), and, in addition, EC lacks a hydroxyl group from the polyphenol ring (attached at position 2). These structural changes decrease the inhibitory potency dramatically, as an IC50 of > 100 μm was estimated for both EGC and EC for all assayed phosphatase forms. The above data imply that, in EGCG, the galloyl group, and to a lesser extent the polyphenolic ring and its hydroxyls, may play an important role in suppressing phosphatase activity. The wide spectrum of the biological effects exerted by these polyphenolic compounds suggest multiple targets within the cells that may compete for binding with the protein phosphatases. Figure 2D shows that PGG and EGCG suppressed the phosphatase activity of HeLa cell lysate with IC50 values of 6.2 and 10 μm, respectively. These data imply that PGG and EGCG are able to inhibit not only the catalytic subunits but also the holoenzyme forms of PP1 and PP2A, and their inhibitory potencies are exerted in the presence of other cellular proteins.
Interaction of PGG and EGCG with PP1cδ as revealed by SPR binding experiments
To study the molecular basis of the inhibition of PP1c by polyphenolic compounds, the binding of PGG and EGCG to PP1cδ was investigated by SPR-based binding techniques. It was previously shown that this technique was suitable for proving direct interactions of α-amylase with PGG or with different tannin mixtures, as well as for comparing their relative strengths of binding to the enzyme . In our experiments, PP1cδ was immobilized on CM5 chips by covalent coupling in the presence of OA in order to prevent extensive modification of the enzyme at side chains involved in the binding of phosphatase inhibitors . PGG or EGCG was run over the PP1c surfaces at 1–50 μm (Fig. 3A,B). The sensograms obtained indicate the binding of both PGG and EGCG to PP1cδ. Initial kinetic evaluation of the interactions with biaevaluation 3.1 software (Biacore AB, Uppsala, Sweden) using a 1 : 1 binding model gave rate constants for the association (ka) and dissociation (kd) that suggested relatively slow association of both PGG (ka = 65.5 ± 32.5 m−1·s−1) and EGCG (ka = 479.6 ± 204.7 m−1·s−1) with PP1c, and slow dissociation of the PP1c–PGG [kd = (2.5 ± 0.4) × 10−4·s−1] and the PP1c–EGCG [kd = (3.4 ± 1.5) × 10−4·s−1] complexes. The association constants (Ka) for the formation of the PP1c–PGG and PP1c–EGCG complexes were (2.2 ± 1.1) × 105m−1 and (1.4 ± 0.6) × 106m−1, respectively. Although the above analysis of the binding parameters resulted in only slight deviation of the experimental and the fitted data, the presumed 1 : 1 binding model was seriously challenged, as the molecular ratios of the PGG and EGCG bound to the surface of PP1c at the highest injected concentration (50 μm) were ∼ 11 and ∼ 12 mol/mol, respectively (Fig. 3C). These data raise the possibility that several binding sites for both molecules may exist on the surface of PP1c, although the possibility of stacking of the bound molecules cannot be excluded either, according to a previous report .
PGG was shown to interact with α-amylase via its aromatic (hydrophobic) rings to cause its inhibitory action on enzyme activity . A similar mechanism could be involved in the inhibition of PP1 by PGG and EGCG, as the active site of PP1 is situated at the bifurcation point of a Y-shaped groove, on the surface of which one arm represents a hydrophobic groove . This latter groove is implicated in the inhibition of PP1 by inhibitory toxins, such as MC-LR and OA [22,23]. To determine the effect of MC-LR on the interaction of PP1 with PGG or EGCG, competition assays were carried out (Fig. 3D,E). First, PGG (Fig. 3D) or EGCG (Fig. 3E) was injected over the PP1c surface, and the association and dissociation phases were recorded. After regeneration of the surfaces, MC-LR (1 μm) was injected on the PP1cδ-covered surface until saturation, but even at this stage the bound MC-LR/PP1c molar ratio was ∼ 0.40–0.45 mol/mol, indicating that only some of the immobilized PP1 molecules are available for binding of MC-LR. Then, the binding of PGG or EGCG on the MC-LR-covered surface was measured and compared with the values obtained in the absence of the toxin (Fig. 3D,E). MC-LR couples covalently to PP1c, so the PP1c–MC-LR complex does not dissociate on injection of PGG or EGCG. Binding of PGG and EGCG to the surface was decreased significantly when PP1cδ was presaturated with MC-LR (Fig. 3D,E), but was not completely abolished. The formation of PP1cδ–PGG and PP1cδ–EGCG complexes on the MC-LR-covered surfaces became slower, whereas their dissociation was faster. One of the major regions where MC-LR binds at the catalytic surface of PP1 is the hydrophobic groove; therefore, the above results suggest that PGG and EGCG also interact with this region of the enzyme. The residual binding of PGG and EGCG with altered kinetics may raise the possibility again of the existence of additional interaction site(s) of these effectors.
To further clarify the significance of the MC-LR-binding surface in the interactions of polyphenols with PP1c, we carried out SPR-based competitive binding assays with a different experimental setup. We synthesized biotin-coupled MC-LR (MC-LR–biotin) as described earlier ; it was immobilized on a streptavidin-coupled sensorchip, and PP1cδ was then injected over these surfaces (Fig. 3F–H). Owing to the high affinity of binding of PP1c to MC-LR–biotin, the MC-LR–biotin–PP1c complex dissociated slowly, reaching a plateau where no significant dissociation of the complex occurred. This MC-LR–biotin–PP1c complex could be regarded as a PP1c surface on which only the MC-LR-binding site was occupied. Injection of PGG (Fig. 3F) or EGCG (Fig. 3G) at this stage did not result in any change in the resonance signal, indicating that, when the MC-LR-binding surface was occupied, interaction of PGG or EGCG with PP1c was excluded. In contrast, Fig. 3H shows that association of PP1c with MC-LR–biotin left other interaction site(s) functional, as binding of Flag–MYPT1 at the regulatory subunit binding site took place to a significant extent. These experiments proved straightforwardly that the polyphenols bind to PP1c at a single site, possibly at part of the hydrophobic groove occupied by MC-LR and other inhibitory toxins.
Binding of EGCG and PGG to PP1cδ revealed by saturation transfer difference (STD) NMR measurements
The interaction of EGCG or PGG with PP1cδ was also characterized by STD studies with NMR spectroscopy [33–35]. This technique may help to uncover the structural regions in the ligands (i.e. EGCG and PGG) that are involved in the interaction with PP1c. 1H-NMR spectra of EGCG dissolved in dimethylsulfoxide-d6 (dimethylsulfoxide was added to improve the solubility of the inhibitor) and transferred into D2O acetate buffer (final concentration of EGCG of 4.3 mm) were recorded in both the absence (Fig. 4A) and the presence of 9–10 μm PP1cδ previously dialyzed into the same buffer (Fig. 4B). On comparison of these spectra, striking differences could be observed. In the absence of protein, only the resonances of the aromatic and flavanol ring protons of EGCG could be detected, whereas those of the hydroxyl protons were completely missing, owing to their exchange with D2O. In contrast, in the presence of PP1cδ, weak, residual OH resonances could be identified, implying that the hydroxyl protons are only partially exchanged with D2O, owing to their possible involvement in intermolecular hydrogen bonds with PP1cδ. The formation of a PP1cδ–EGCG complex, i.e. the exchange of EGCG between the free and PP1cδ-bound forms, was further corroborated by the significant broadening of 1H resonances observed when EGCG was added to the PP1cδ sample. The STD 1H-NMR spectrum shown in Fig. 4C identifies those hydrogens of EGCG that are involved in the interaction with PP1c. The observed STD resonances unambiguously reveal close contact between the aromatic ring hydrogens of EGCG and PP1cδ. In Fig. 4D,E, the 1H-NMR and STD spectra, respectively, of PGG in the presence of PP1cδ are shown. The characteristics of these spectra, such as detectable OH resonances, exchange-mediated line-broadening, and STD effects observed on the aromatic ring protons, suggest that binding of PGG to PP1c also involves the aromatic rings and OH groups of PGG in the formation of the complex.
Molecular modeling of the PP1c–EGCG and PP2Ac–EGCG complexes
To further characterize the structure of the PP1c–EGCG complex, molecular modeling studies were performed, on the basis of the coordinates of the X-ray structure of the PP1cγ–OA crystals . In addition, comparative docking studies for the PP2Ac–EGCG complex were carried out, on the basis of the X-ray structure of the PP2Ac–OA crystals . The calculated lowest energy structure of the PP1c–EGCG complex (with a yellow carbon skeleton) overlaid with the X-ray (violet) and the modeling-based (black) structures of the PP1c–OA complex are shown in Fig. 5A. According to the results of modeling, the phenolic ring of EGCG docks at a binding site of the hydrophobic groove overlapping with the binding surface of OA, whereas the galloyl group and the flavanol ring reside at the extension of the hydrophobic groove of PP1c. Figure 5B shows an enlarged docked structure of the PP1c–EGCG complex. One-dimensional transferred NOE experiments  suggested that the conformation of EGCG was altered significantly (data not shown); that is, the molecule assumed a much more compact structure upon binding to PP1c. As a substantiation, proton–proton distances corresponding to ∼ 5 Å in EGCG in complex with PP1c are indicated by yellow lines in Fig. 5B. Note that the relevant distances in the free EGCG are significantly longer than 5 Å, suggesting an extended time-averaged structure of the free inhibitor, which was also supported by molecular dynamics simulations (data not shown). Figure 5C illustrates the major interactions between the amino acid side chains of PP1c and EGCG predicted by the docking model. It suggests that Ile130, Ile133, Trp206 and Val195 contribute significantly to the hydrophobic environment for the binding of the aromatic rings of the polyphenolic and the galloyl units. This hydrophobic interaction is further stabilized via hydrogen bonds (indicated by black lines) formed by the hydroxyl groups of the galloyl and flavanol rings as well as the carbonyl group of the ester bond toward peptide bonds and to the Trp206, Cys127 and Cys202 side chains of PP1c. In addition, the side chains of Ser129 and Asp194 as well as the carbonyl group of Pro196 are close enough to EGCG to form hydrogen bonds, even though their positions do not meet the criteria required for the perfect hydrogen bond. Docking of a second EGCG molecule to the most stable PP1c–EGCG complex was also attempted, and it identified secondary binding sites at the C-terminal groove and in the proximity of the hydrophobic groove with substantially lower affinity than that of the primary EGCG-binding site (data are not shown).
The EGCG molecule could also be docked to the hydrophobic groove of PP2Ac, although the pose of the docked EGCG was different from those observed at PP1c (Fig. 5D). In the PP2Ac–EGCG complex, the hydrophobic groove is formed by Ile123, Val126, Val189, and Trp200. EGCG forms hydrogen bonds with the side chains of Tyr127, Gln122, His191, Cys196, and Trp200, as well as with the peptide carbonyl atoms of Pro190, His191, and Gly215. The sequence and structural similarities between PP1c and PP2Ac are remarkable, especially for their active site regions  (Fig. 5C,D). Nevertheless, the position of the Ala182–Gly193 loop in PP2Ac is slightly different from that of the corresponding Arg187–Gly199 loop in PP1c (Fig. 5E). This difference in loop positions and the very different side chain positions reduce the size of the hydrophobic groove in PP2Ac, explaining the distinct pose of EGCG binding in PP2Ac as compared with that in PP1c.
The above data are consistent with the structure–inhibitory potency relationship suggested by the activity assays, namely that the galloyl unit itself, the aromatic rings with their hydrophobic character and the polyphenolic hydroxyls may play important roles in the interaction of EGCG with PP1c.
Effects of PGG, EGCG and other catechins on the viability of HeLa cells
It is well established that PGG and EGCG are able to suppress the viability of various malignant cell types [2,5,11,12]. It was of interest to examine how this effect of the polyphenolic compound relates to their structure and phosphatase inhibitory potencies. Figure 6 shows that, after 24 h of incubation with PGG and EGCG and its derivatives the viability of HeLa cells was suppressed differentially. PGG, EGCG and GCG decreased cell survival potently, concentration-dependently, and with similar effectiveness. ECG and EGC were much less effective, and EC was without effect. The above order of potency of the polyphenolic compounds in inducing cell death corresponds to their relative PP1 inhibitory effectiveness obtained with in vitro phosphatase activity assays. PGG and EGCG decreased the viability of THP-1 cells and HeLa cells similarly (data are not shown).
Our present study has identified PGG and EGCG as novel protein phosphatase inhibitors. PGG and EGCG appear to be more selective for PP1 than PP2A; therefore, they can be used to distinguish between PP1 and PP2A activities. The structure–activity relationships of EGCG and its derivatives suggest that the galloyl group and the polyphenolic ring attached to the flavanol unit are important determinants of the inhibitory potency. Consistent with these conclusions, NMR measurements indicated the involvement of the aromatic polyphenolic rings as well as the phenolic hydroxyls in both PGG and EGCG in the interactions with PP1c. This hydrophobic interaction pattern was further supported by SPR-based interaction analysis, which also revealed competitive binding of polyphenols (PGG or EGCG) and the phosphatase-inhibitory toxin MC-LR to PP1c. The crystal structures of PP1c–MC-LR  and PP1c–OA  established that these toxin inhibitors occupy, among other surfaces, part of the hydrophobic groove in PP1c. The hypothetical structure of PP1c in complex with EGCG was obtained by molecular modeling, which confirmed that EGCG docks at the hydrophobic groove and that its interaction with PP1c is stabilized by numerous hydrogen bonds formed via the phenolic hydroxyls as well as the ester bond carbonyl towards the closely positioned peptide bonds and amino acid side chains.
It is apparent that EGCG and PGG are less potent inhibitors of PP1c (IC50 ∼ 0.20–0.47 μm) than the toxin inhibitors (IC50 ∼ 0.1–50 nm) [20,21], and that they form less stable adducts with PP1c (Ka ∼ 105–106) than does OA (Ka ∼ 109) . A model of the PP1c–EGCG complex suggests that EGCG does not have a phosphate mimic to interact with the catalytic site as do toxins, and that its major binding surface is the hydrophobic groove. The same side chains (Ile130 and Trp206) of PP1c are involved in the hydrophobic interaction with EGCG as with MC-LR or OA. In addition, the question of whether the binding surfaces of toxins and polyphenols overlap only partially was also raised. On the basis of the PP1c–EGCG model, it is hypothesized that the flavanol and the galloyl units make contacts with the extension of the hydrophobic groove interacting with residues (Asp194, Val195, Pro196, and Asp197) that are not involved in the binding of MC-LR or OA, but are involved (Val195 and Pro196) in the interaction with CLA  and TM . Consistent with this, binding of the C1–C14 unit in TM to the extension of the hydrophobic groove of PP1c was also thought to stabilize the interaction and increase inhibitory potency [25,38]. The partial overlap between the binding sites and the high PGG/PP1c and EGCG/PP1c molar ratios on the surfaces may explain why preadsorption of MC-LR on the PP1c surface suppressed substantially, but not entirely, binding of either PGG or EGCG to PP1c, and suggest multiple binding sites for both PGG and EGCG on PP1c. However, on a ‘homogeneous’ PP1c surface, where PP1c is not coupled covalently and only the MC-LR-binding regions are occupied (MC-LR–biotin–PP1c adduct), no binding of polyphenols is observed, suggesting that the overlapping binding surfaces of MC-LR and polyphenols are essential to the interaction of PGG or EGCG with PP1c. These findings are also consistent with the conclusion that PGG or EGCG binds at a single site on PP1c. However, an obvious enigma then remains to be clarified: how binding of such an extremely high molar amount of PGG or EGCG over the PP1c on the surface is possible. These high EGCG/PP1 and PGG/PP1c binding ratios may be attributable to self-association of the binding molecules with each other, i.e. stacking of EGCG or PGG at a single binding site, a mechanism suggested by previous studies on the interaction of EGCG with human salivary histatins . It is assumed that binding of the polyphenols to PP1c creates a favorable platform for either EGCG or PGG for stacking at much lower concentrations that would be required for self-association of these molecules in the absence of an interacting protein.
The greater stability of PP1c–toxin complexes is attributable to multisite interactions of MC-LR, OA, CLA and TM with the enzyme, which include the catalytic center and the β12–β13 loop beside the hydrophobic groove. The more potent inhibition of PP2Ac by OA than of PP1c is believed to occur via a stronger interaction of the toxin with the β12–β13 loop  in PP2A than in PP1c. In contrast, the preference of PGG and EGCG for inhibition of PP1c over PP2Ac is less well understood. Sequence alignment of PP1c and PP2Ac [22,36] suggests a high degree of identity of the amino acids in those regions that form the hydrophobic groove. The residues important for the hydrophobic interactions are the same (i.e. Ile, Trp, and Val) in both PP1c and PP2Ac. In contrast, a few amino acids involved in hydrogen bonding to the phenolic hydroxyls or the ester bond oxygen of EGCG in the PP1c–EGCG complex are substituted (Cys127→Ser120; Ser129→Gln122; Asp194→Glu188; and Asp197→His191) in PP2Ac. These amino acids in PP2Ac are similar to those in PP1c with respect to hydrogen-bonding abilities, and Gln122 and His191 are indeed involved in the stabilization of EGCG binding to PP2Ac by hydrogen bonds (Fig. 5D). However, there is a remarkable difference between the PP1c–EGCG and the PP2Ac–EGCG complexes in the pose of EGCG at the hydrophobic groove. Thus, it is assumed that truncation of the hydrophobic groove by the Ala182–Gly193 loop in PP2Ac does not allow EGCG to bind to the extension of this groove, as occurs in PP1c. In conclusion, the EGCG molecule binds to the hydrophobic groove of both PP1c and PP2Ac with high probability, but the binding poses are significantly different.
PGG and EGCG and its derivatives decrease the viability of malignant cells with relative potencies reminiscent of the structure–activity relationships determined in in vitro phosphatase assays. These data raise the possibility that there is a correlation between the phosphatase-inhibitory ability and the cell death-inducing effect of the polyphenols. PGG and EGCG have been reported to influence the phosphorylation level of ERK1/2, JNK and p38 kinase [7,39] and of key proteins (Akt kinase, p53, and retinoblastoma protein) involved in the execution and regulation of apoptosis [8–10,15]. These studies provided controversial data on the changes in phosphorylation level of these proteins in different cell types upon short or long incubation periods with the effectors. For instance, EGCG treatment stimulated [39,40], inhibited  or had no effect on  JNK phosphorylation at the activating sites (Thr183 and Tyr185) in different cell types. In contrast, PGG (after long-term incubation) was shown to decrease JNK and ERK phosphorylation, whereas it increased p38 kinase phosphorylation, in endothelial cells . Our present findings that PGG and EGCG inhibit phosphatase activity in cell lysate imply that they suppress the activity of the PP1 and PP2A holoenzymes and may also influence protein phosphatase activity (with a preference for PP1) within cells. These data add novel aspects to the evaluation of changes in the phosphorylation levels of intracellular proteins by polyphenolic compounds. Dissection of the phosphorylation processes in which PGG or EGCG is involved require carefully designed experimental approaches, as an interplay of multiple effects (i.e. kinase inhibition/activation and phosphatase inhibition/activation) of these compounds should be assumed.
In summary, PGG and EGCG, two major biologically active ingredients of tannins and green tea, are identified in this study as reasonably selective inhibitors of PP1. The significance of these findings is that, although PGG and EGCG are less potent inhibitors of PP1c than natural toxins, they are much less toxic and therefore may serve as possible lead compounds for the development of more potent and selective PP1c-inhibitory drugs. In addition, the data presented here, by identifying PP1 as one of the possible intracellular targets of PGG and EGCG, may contribute to our understanding of the molecular basis of the cellular actions of these compounds.
Chemicals and vendors were as follows: [32P]ATP[γP] was from Hungarian Isotope Institute (Budapest, Hungary); l-glutamine, MEM and fetal bovine serum were from Gibco (Gaithersburg, MD, USA); PGG was provided by G. Gyémánt and L. Kandra (Department of Chemistry, University of Debrecen) or purchased from Sigma (St Louis, MO, USA); MC-LR was a gift from C. Máthé (Department of Botany, University of Debrecen); EGCG, EGC, EC, GCG, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-tetrazolium bromide (MTT), poly(l-lysine)–agarose and the anti-Flag M2 affinity column were from Sigma; the Precision Plus Protein Dual Color standard was from Bio-Rad (Hercules, CA, USA); enhanced chemiluminescence reagent was from Pierce (Rockford, IL, USA); the CM5 sensorchip, the Sensorchip SA, N-ethyl-N′-dimethylaminopropyl-carbodiimide, N-hydroxysuccinimide and the Amine Coupling Kit were from Biacore AB (Uppsala, Sweden). MC-LR–biotin adduct was prepared and purified as previously described . Heparin–Sepharose, Ni2+–Sepharose, MonoQ HR 5/5 and Superdex 200 HR 10/30 were from GE Healthcare Biosciences AB (Uppsala, Sweden). All other chemicals used were of the highest purity commercially available.
The proteins used are described in the relevant references: MLC20 was prepared from turkey gizzard, and phosphorylated by myosin light chain kinase in the presence of [32P]ATP[γP] ; Flag–MYPT1 was expressed in tsA201 cells, and purified on an anti-Flag M2 affinity column according to the manufacturer’s instructions.
Protein phosphatase preparation
PP1c and PP2Ac were generated by the acetone precipitation procedure as previously described . PP1c was separated from PP2Ac by chromatography on heparin–Sepharose, and PP1c was further purified on an FPLC-MonoQ column and gel filtration on a Superdex 200 HR column . PP2Ac from the heparin–Sepharose flow-through fractions was subjected to chromatography on poly(l-lysine)–agarose and then to gel filtration on a Superdex 200 HR column. PP1c and PP2Ac represented 80–90% of the total protein content of the final preparations as determined by SDS/PAGE, and their specific activities determined with 1 μm [32P]MLC20 were 1.89 and 0.61 μmol·min−1·mg−1, respectively. Hexahistidine-tagged recombinant PP1cδ was expressed in E. coli and purified to homogeneity with Ni2+–agarose chromatography  followed by heparin–Sepharose chromatography, and its specific activity was 0.97 μmol·min−1·mg−1.
Assay of protein phosphatases
Purified skeletal muscle PP1c (0.7 nm), PP1cδ (1 nm), skeletal muscle PP2Ac (1.6 nm) or HeLa cell lysate (750-fold dilution, 2.4 μg·mL−1) was assayed with 1 μm [32P]MLC20 at 30 °C in 20 mm Tris/HCl (pH 7.4) and 0.1% 2-mercaptoethanol. The phosphatase samples were diluted in 20 mm Tris/HCl (pH 7.4) and 0.1% 2-mercaptoethanol. Different concentrations of polyphenolic compounds were preincubated with the phosphatase for 10 min, and the reaction was then initiated by addition of the substrate. After 10 min of incubation the reaction was terminated by the addition of 200 μL of 10% trichloroacetic acid and 200 μL of 6 mg·mL−1 BSA. The precipitated proteins were collected by centrifugation at 8000 g for 2 min, and the released 32Pi was determined from the supernatant (380 μL) in a scintillation counter .
SPR binding experiments
Kinetic parameters of the interactions of PP1 with PGG or EGCG were analyzed by SPR-based binding experiments with the Biacore 3000 instrument. PP1cδ was immobilized covalently on a CM5 sensor chip by the amine-coupling method, as recommended by the manufacturer. Flow cells were activated by the injection of a solution containing 50 mmN-hydroxysuccinimide and 200 mmN-ethyl-N′-dimethylaminopropyl-carbodiimide. PP1cδ was diluted to 5 μm in 10 mm sodium acetate (pH 5.5), 1 mm dithiothreitol, 2 mm MnCl2, and 5 μm OA (to prevent immobilization through the inhibitor-binding site), and injected over the surface for 7 min at a flow rate of 10 μL·min−1. The remaining reactive groups were blocked with a 7-min injection of 1 m ethanolamine/HCl (pH 8.5). PGG and EGCG (referred to as the analytes) were injected over the surface at different concentrations in running buffer containing 25 mm Tris/HCl (pH 7.4), 150 mm NaCl, 1 mm dithiothreitol, 2 mm MnCl2, 0.05% Surfactant P20, and 5% dimethylsulfoxide. The association phase of the interactions between PP1cδ and the polyphenolic molecules were monitored for 7 min, and the dissociation phase in running buffer without the analyte was monitored for 6–7 min. Binding to the immobilized PP1cδ was monitored as a sensogram, where response unit (RU) values were plotted against time. The control surface (flow cell blocked by ethanolamine) was treated identically to the protein surface to determine unspecific binding, which was subtracted from the data obtained with the protein surface. The kinetic parameters were derived from the sensograms with biaevaluation 3.1 software (Biacore AB, Uppsala, Sweden). The molar amounts of immobilized PP1cδ and surface-bound PGG or EGCG were calculated on the basis that an increase of 1000 RU in the resonance signal corresponded to the binding of 1 ng of protein or analyte molecule on the surface, also taking into account that the surface of one flow cell was 1.2 mm2 and its volume was 20 nL. The sensorchip was regenerated after each binding assay by a brief injection (10 μL·min−1 flow rate for 1 min) of 10 mm glycine/HCl (pH 2.1). In cases of incomplete regeneration, this step was repeated with 1 m ethanolamine (pH 8.5). During competition assays, the regenerated PP1cδ-covered surface was saturated with 1 μm MC-LR, and this was followed by the injection of 10 μm PGG or 5 μm EGCG. For assay of binding of PGG or EGCG to MC-LR–biotin–PP1c, the adduct was first formed on a streptavidin-coupled sensorchip capturing MC-LR–biotin (350–475 RU), and then PP1cδ was injected over this surface to achieve a PP1cδ immobilization level of 750–2500 RU. After washing of the chip with running buffer and stabilization of the resonance signal to an approximately constant level, PGG, EGCG or Flag–MYPT1 was run over these surfaces and binding was recorded.
All NMR experiments were performed with a Bruker Avance DRX 500 spectrometer operating at 500 MHz, and equipped with a 5-mm z-gradient multinuclear proton detection probe head. Both 1H-NMR and STD spectra were recorded at 298 K on samples containing 1 mg of EGCG or PGG dissolved in 7 μL of dimethylsulfoxide-d6 and transferred to 500 μL of D2O acetate buffer. The samples for STD measurements contained 9–10 μm PP1cδ. The duration of the hard 90° proton pulse was 11 μs. The 1H-STD spectra were recorded with 2-s semiselective irradiation of the PP1cδ methyl resonances at 0.8 p.p.m. by a train of Gaussian 90° pulses of 50 ms each. In the reference experiment, the off-resonance frequency was set to − 15 000 Hz. Selective 1D transferred NOE experiments were performed  with a mixing time of 150 ms (data not shown). Selective excitation was achieved with a Gaussian 180° pulse of duration 35 ms incorporated into a DPFGSE sequence. All spectra were processed with topspin 2.1.
The receptors (PP1c or PP2Ac) were built on the basis of the X-ray structure of OA-bound PP1cγ (Protein Data Bank ID: 1jk7 ) or PP2Ac (Protein Data Bank ID: 2IE4 ), keeping only the protein molecule, two manganese ions and a water molecule in close proximity to the metal ions. In order to include all of the substrate-binding (C-terminal, acidic, and hydrophobic) grooves and their proximities in the searchable area, a 36 × 36 × 41 Å grid size was chosen. The grid points (spacing at 0.375 Å) and the atom-specific affinity at these points were calculated with autogrid 4 software (The Scripps Research Institute, La Jolla, CA, USA). A Lamarckian genetic algorithm  with local search in flexible ligand-rigid receptor docking calculations was used as implemented in autodock 4 (The Scripps Research Institute, La Jolla, CA, USA). Both autogrid 4 and autodock 4 are included in the autodock suite of docking tools . The number of hybrid genetic algorithm with local search calculations, the number of individuals in the population and the maximum number of energy evaluations were set to 150, 150, and 75 000 000, respectively. The other parameters were kept at their default values [autodock Version 4.2, User Guide (Automated Docking of Flexible Ligands to Flexible Receptors), available at http://autodock.scripps.edu/]. The poses obtained were clustered according to 2-Å rmsd criteria. The EGCG and OA (the latter for validation purposes) molecules only were docked because the large number of freely rotatable bonds in the PGG molecule would not allow the performance of a feasible docking study. Docking of the EGCG molecule to a slightly modified receptor with two more water molecules at the metal-binding site, and leaving the grid size and all the other docking parameters unchanged, yielded essentially the same results. The docking results were visualized with the chimera modeling and visualization software package .
Cell cultures, treatments, and viability assays
HeLa (cervical carcinoma) and THP-1 (monocytic leukemic) cells were obtained from the European Collection of Cell Cultures, and cultured according to the supplier’s recommendations. Prior to treatments, cells were incubated with serum-free medium for 16 h. To investigate the influence of polyphenolic compounds on cell viability, cells were treated with 0–100 μm PGG, EGCG, GCG, ECG, EGC or EC for 24 h, and this was followed by the addition of MTT, which can be converted by viable cells into a colored formazan product that can be measured spectrophotometrically at 562 nm. The absorbance (A) was linearly related to the number of viable cells.
We are grateful to G. Gyémánt and L. Kandra (Department of Chemistry, University of Debrecen) for providing an initial sample of PGG for the experiments. The authors are indebted to Á. Németh and E. Kovács for their excellent technical assistance. This work was supported by grants from the Hungarian Scientific Research Fund (OTKA K68416, NK-68578, and CNK 80709. Financial support of the TÁMOP 4.2.2.-08/1-2008-0019 DERMINOVA, TÁMOP-4.2.2./B-10/1-2010-0024 and TÁMOP-4.2.1./B-09/1/KONV-2010-0007 projects is greatly appreciated.