Selective Inhibition of MAPKK Wis1 in the Stress-activated MAPK Cascade of Schizosaccharomyces pombe by Novel Berberine Derivatives*

Intracellular molecular targets of novel berberine derivatives, HWY 289 and HWY 336, were identified by a screen of a variety of mutants in fission yeast Schizosaccharomyces pombe. HWY 289 and HWY 336 completely inhibited the proliferation of wild type as well as various mutant fission yeast cells (minimal inhibitory concentrations were 29.52 μm for HWY 289 and 11.83 μm for HWY 336), but did not affect the proliferation of Wis1 mitogen-activated protein kinase kinase (MAPKK) deletion mutants. In addition, HWY 289 with an IC50 value of 7.3 μm or HWY 336 with IC50 of 5.7 μm specifically inhibited in vitro kinase activities of purified Wis1, whereas either compound did not affect the activities of other kinases in the mitogen-activated protein kinase (MAPK) cascades of fission yeast. These genetic and biochemical results demonstrate the high degree of specificity of HWY 289 and HWY 336 to MAPKK Wis1 and suggest that the cytotoxicity of these compounds is not simply due to the inhibition of Wis1 kinase activity. High salt wash experiments have shown that strong noncovalent binding occurs between Wis1 and either HWY 289 or HWY 336. The preincubation of Wis1 kinase with ATP did not affect the inhibition of Wis1 by HWY 289 and HWY 336, but when Wis1 was preincubated with MBP, a protein substrate, Wis1 kinase activity was no longer inhibited. These observations demonstrate that HWY 289/HWY 336 do inhibit Wis1 kinase, not by binding to the ATP-binding site but by disturbing the binding of substrate to the kinase. Target validation of the complex of HWY 289/HWY 336 and Wis1 kinase will provide important clues for the mechanism of specific cytotoxicity of these compounds in S. pombe. On a broader aspect, it would create an initiative to further modify and develop compounds that selectively inhibit kinases and cause cytotoxicity in various MAPK cascades including those of mammals.

a specific signal; yet, activation of a single MAPK can be affected by various signals. In addition, specific MAPK signaling pathways interact with one another to form a complex network (13). Therefore, to better understand signal transduction in MAPK cascades, we must answer questions about how various signals are integrated to activate a single response and how signaling specificity is maintained in stress-specific and tissuespecific manners. Synthetic inhibitory components of several MAPK cascades have proven excellent tools for studying MAPK signaling specificity mechanisms in mammalian cells. PD 098059 (2-(2Ј-amino-3Ј-methoxyphenyl)-oxanaphthalen-4-one) and U0126 (1,4-diamino-2,3-dicyano-1,4-bis(2-minophenylthio)butadiene) are specific inhibitors for MEK1 and MEK2 (14,15). The pyridinyl imidazoles SB 203580 and SB 202190 inhibit the MAPK subfamilies, stress-activated protein kinases (SAPKs) 2a and 2b, and p38 MAPK isoforms (16 -20).
Three MAP kinase cascades have been elucidated in the fission yeast S. pombe. MAPK Spk1, MAPKK Byr1, MAPKKK Byr2, and a small GTPase, Ras1, comprise a cascade required for mating and sporulation (21)(22)(23). The integrity of the Wis1-Spc1 pathway is required for survival under varied stress conditions such as high osmolarity and oxidative stress. MAPKK Wis1 is activated by MAPKKKs Win1 and Wis4, which in turn activates MAPK Spc1 following osmotic, heat, or oxidative stress (24 -26). Another stress-activated MAP kinase cascade, consisting of MAPKKK Mkh1, MAPKK Skh1, and MAPK Spm1, regulates morphogenesis and cell wall integrity in relation to environmental conditions (27)(28)(29). There are highly conserved functional similarities between MKK4/7-SAPK and MKK3/6-p38 pathways of mammalians and the Wis-Spc1 pathway of fission yeast. Similarities are also observed between the mammalian Ras-Raf-MEK1-ERK pathway and the fission yeast Ras1-Byr2-Byr1-Spk1 pathway. These striking resemblances to mammalian cells and the relative simplicities of the cascades, combined with genetic amenability, make fission yeast an ideal system for studying the mechanisms of a drug action by chemical control of gene function. In this report, we show that specific berberine derivatives selectively inhibit the kinase activity of Wis1 MAPKK in S. pombe and suggest the applicability of these berberine derivatives to the development of selective inhibitors for kinases in divergent stress-activated MAPK cascades of mammalian cells.
Minimal Inhibitory Concentration (MIC) Determination and Reversibility Test on Fission Yeast-Each stock solution of berberine chloride (1 mg/ml), HWY 289 (1 mg/ml), and HWY 336 (1 mg/ml) was prepared in DMSO. The MIC of each compound for the wild type of S. pombe was determined by transferring cells grown to midlog phase (5 ϫ 10 6 cells/ ml) in yeast extract supplemented with adenine and uracil into the fresh media containing different concentrations of HWY 289, HWY 336, and berberine, respectively. Cells were incubated for 12 h with shaking and counted every 3 h for 24 h to monitor the growth rate. DMSO, the compound solvent, had no visible effect on cell growth. The reversible effect of HWY 289 and HWY 336 on the growth of S. pombe was examined by counting cell numbers in drug-free media every 3 h for 6 h after incubating with MIC of each drug for 18 h. The same number of cells incubated with each drug were collected and washed three times with double-distilled H 2 O and media before switching to drug-free media. Each MIC of HWY 289 and HWY336 was also tested on a range of mutants of S. pombe for 12 h with shaking and counted every 3 h for 24 h to monitor the growth rate. Cell morphologies were observed at ϫ100 using a light microscope (Axioscop; Zeiss).
Purification and Assay of Protein Kinases-The expressions of 6ϫHis/HA-tagged Win1, Skh1, Spm1, Byr1, and GST-tagged Wis1 and Spc1 were respectively induced from the pREP1 nmt1 promoter by incubating log phase KGY246 cells transformed with each plasmid in the absence of thiamine for 18 h at 29°C (32). Before harvest, cells expressing tagged Win1, Skh1, Spm1, Wis1, and Spc1 kinases were exposed to 0.6 M KCl for 20 min to activate the kinase by osmotic stress, and cells expressing Byr1 were incubated in nitrogen-free media for 6 h to activate the kinase by nitrogen starvation. Cells were disrupted by glass beads in HB buffer (25 mM HEPES, pH 7.5, 60 mM ␤-glycerophosphate, 15 mM EGTA, 0.1 mM sodium vanadate, 0.1% Triton-X, 1 mM phenylmethylsulfonyl fluoride) after washing once with PBS containing 50 mM NaF and 1 mM NaN 3 , and the lysate was incubated with 3.2 g of HA antibody (Roche Molecular Biochemicals) for 2 h at 4°C followed with 80 l of protein A-Sepharose (Sigma) for 90 min at 4°C or 45 mg/ml GST beads (Sigma) for 2 h at 4°C, respectively. The beads were washed three times with 1.0 ml of HA buffer containing 100 mM NaCl and twice with kinase buffer (25 mM HEPES, pH 7.5, 60 mM ␤-glycerophosphate, 15 mM EGTA, 0.1 mM sodium vanadate, 0.1% Triton, 1 mM phenylmethylsulfonyl fluoride) prior to kinase assays. The kinase assay was performed for 30 min at 30°C by the addition of 1 mg/ml MBP (Sigma), 1 mM ATP, and 20 Ci of [␥-32 P]ATP. Phosphorylated MBP The competition of HWY 289 and HWY 336 with protein substrate or ATP for Wis1 binding was resolved by comparing the kinase activity of GST bead-purified Wis1, preincubated with 1 mg/ml MBP or 1 mM ATP, respectively, prior to the addition of HWY 289 or HWY 336, to that of purified Wis1 kinase, pretreated with each HWY compound before the addition of 1 mg/ml MBP or 1 mM ATP.

RESULTS
In Vivo Effects of Novel Berberine Derivatives on the Growth of S. pombe-The novel berberine derivatives HWY 289 and HWY 336 were prepared in the lead berberine optimization process to search for compounds having potent anti-fungal activity and better pharmacological profiles. The synthesis of these berberine derivatives was started with commercially available berberine chloride and processed as presented in Fig.  1. Each desired derivative was verified by 1 H NMR (300 MHz, CDCl 3 ) spectroscopic analysis.
To utilize fission yeast S. pombe for studying the targets and biological actions of berberine derivatives, we first examined the cytotoxic effects of the lead compound berberine chloride and the novel berberine derivatives HWY 289 and HWY336 on the proliferation of fission yeast. Berberine slowed down but did not completely inhibit the proliferation of wild type S. pombe ( Fig. 2A). However, as expected from the highly effective anti-Candida activities of HWY 289 and HWY 336 over berberine (1), these compounds completely blocked the proliferation of wild type fission yeast ( Fig. 2A). The MIC on the wild type S. pombe was 29.52 M for HWY 289 and 11.83 M for HWY 336 (Table II). We also examined whether the inhibitory effect of these compounds on the proliferation of fission yeast was reversible. The cells treated with HWY 289 and HWY 336 could not resume active proliferation even after several washes with media, suggesting that the effects of these compounds on the growth of S. pombe were due to very tight binding to targets (Fig. 2B). In addition to growth inhibition, the wild type fission yeast cells were more sensitive to HWY 289 and HWY 336 when osmotic stress (0.6 M KCl) was applied to these cells and were less sensitive in isotonic media containing 1.2 M sorbitol (data not shown). These observations suggest that the berberine derivatives HWY 289 and HWY 336 might target the stress-response pathways, particularly those involved in regulating osmolarity. We then examined the effects of HWY 289 and HWY 336 on various mutant cells in which the function of each gene related to osmolarity control was destroyed, including mutants of ion and metabolite transport and mutants of MAP kinase cascades for osmotic stress signaling. The effect of berberine chloride on these mutants was also examined as a control for specificity of these compounds.
Three S. pombe mutations, cyh3, cyh4, and pma1, have been reported for their defects in metabolite transport. The genes for cyh3 and cyh4 have not been cloned, but these mutants exhibited reduced membrane ergosterol and reduced uptake of amino acids (33,34). pma1 encodes an H ϩ -ATPase in the  plasma membrane for metabolite transport, and its mutant exhibited multiple drug resistance (35). When the effects of HWY 289, HWY 336, and berberine chloride on these mutants were compared with those on wild type S. pombe, cyh3, cyh4, and pma1 mutant cells were as sensitive as wild type cells (data not shown). As described in the introduction, two pathways among three stress-activated MAP kinase cascades have been elucidated for osmotic stress signaling in S. pombe. The effects of HWY 289 and HWY 336 were examined in each deletion mutant of the components in two MAP kinase cascades for osmotic stress signaling including win1, wis4, wis1, spc1, mkh1, skh1, and spm1. The effects of HWY 289 and HWY 336 on byr2 and byr1 mutants of another MAP kinase cascade were also observed as controls for specificity. While HWY 289 and HWY 336 completely blocked the growth of other kinase mutants and wild type cells, they did not inhibit the proliferation of wis1 deletion mutant cells (Fig. 3, A and B; data not shown). Previous genetic studies have reported that wis1 as well as its direct upstream activators, win1 and wis4, and the downstream target spc1 are required for survival in high osmolarity conditions, but their deletion mutants are not lethal (24 -26). Our results showed that HWY 289 and HWY 336 did not inhibit the growth of wis1 deletion mutant cells but did block the proliferation of win1, wis4, and spc1 deletion mutants (Fig.  3, A and B). These observations also demonstrate that the effect of HWY 289/HWY 336 on wild type is different from the phenotype of a wis1 null mutation. The specific effect of HWY 289 and HWY 336 on the growth of wis1 deletion mutant cells and the discrepancy between the effect of these compounds on wild type and the phenotype of wis1 deletion mutant strongly suggest that Wis1 kinase could be a direct target of these compounds, although these compounds may not inhibit cell growth by simply blocking Wis1 kinase activity. In contrast to HWY 289 and HWY 336, berberine and the solvent Me 2 SO showed the same effect on these mutants as on wild type cells (Fig. 3, C and D).
Since HWY 289 and HWY 336 could not inhibit the proliferation of wis1 deletion mutant cells specifically, the effect of HWY 289 and HWY 336 on the S. pombe mutants of other kinases including Cdc2 and polo kinase Plo1 was also examined. Other kinase mutants were as sensitive to HWY 289 and HWY 336 as wild type fission yeast cells (data not shown), suggesting the high specificity of HWY 289 and HWY 336 on wis1 deletion mutant cells. The specific effect of HWY 289 and HWY 336 on the growth of wis1 deletion mutant cells led us to analyze the precise effect of these compounds on the kinase activity of each component of the Wis1-Spc1 MAP kinase cascade.
Selective Inhibition of Wis1 Kinase by HWY 289 and HWY 336 -Based on in vivo observations, we first investigated whether HWY 289 and HWY 336 could directly inhibit the kinase activity of Wis1 MAPKK at each MIC concentration, respectively. We also examined the effect of HWY 289 and HWY 336 on the kinase activity of MAPKKK Win1 and MAPK Spc1 in the Wis1 MAPK cascade and of MAPKKs Skh1 and Byr1 and MAPK Spm1 in other MAPK cascades. The His 6 / hemagglutinin-tagged Win1, Skh1, Spm1, and Byr1 and GSTtagged Wis1 and Spc1, each of which is expressed in wild-type KGY246, were activated upon exposure to 0.6 M KCl or by nitrogen starvation and were purified using tag-specific beads or antibodies. The activity of each purified kinase was measured by MBP phosphorylation after preincubation with HWY 289 and HWY 336. The effect of the lead compound berberine chloride on Wis1 kinase was also examined. As shown in Fig. 4, A and B, Wis1 kinase activity was completely blocked by incubation with HWY 289 and HWY 336, and neither the phospho-rylation of substrate MBP nor the autophosphorylation of Wis1 kinase was observed. On the other hand, the kinase activities of Win1 and Spc1 were not altered by incubation with HWY 289 and HWY 336 (Fig. 4D). The presence of berberine chloride did not show any effect on Wis1 kinase activity, which is consistent with in vivo assays (Fig. 4C). These results demonstrated that HWY 289 and HWY 336 selectively block Wis1 kinase activity but have no effect on the kinase activities of Win1 and Spc1. We also investigated the effects of HWY 289 and HWY 336 on the kinase activity of other MAPKKs, Skh1 and Byr1, to examine whether the kinase-inhibitory actions of HWY 289 and HWY 336 are specifically aimed at Wis1 kinase, but not at other MAPKKs, as suggested by the in vivo mutant analyses. HWY 289 and HWY 336 did not inhibit the kinase activity of Skh1 and Byr1 (Fig. 4D and data not shown). As expected from in vivo studies, these compounds do not affect the activity of MAPK Spm1 (Fig. 4D). These findings are consistent with the in vivo results and indicate that the novel berberine derivatives HWY 289 and HWY 336 selectively bind and inhibit MAPKK Wis1.
The selective inhibition of Wis1 kinase by HWY 289 and HWY 336 led us to determine the concentration of these compounds for 50% inhibition of Wis1 kinase activity (IC 50 ) by performing in vitro kinase assays in the presence of increasing concentrations of HWY 289 or HWY 336 (Fig. 5, A and B). Wis1 kinase activity was partially inhibited at 5 M and inhibited completely at 10 M of both HWY 289 and HWY 336 (Fig. 5, A  and B). The IC 50 of Wis1 kinase was 7.3 M for HWY 289 and 5.7 M for HWY 336, as listed in Table III.
Noncovalent but Strong Binding of HWY 289 and HWY 336 to Wis1-As shown previously in Fig. 2B, cells treated with HWY 289 and HWY 336 did not resume active proliferation even after several washes with media, suggesting that the effects of these compounds on the growth of S. pombe were due to very tight bindings to targets. Therefore, we inspected whether the specific binding and inhibition of Wis1 kinase by HWY 289 and HWY 336 is irreversible. The activity of purified Wis1 kinase bound to beads was completely inhibited by 10 M HWY 289 or HWY 336 when incubated with each compound for 10 min before kinase assays (Fig. 5). Wis1 kinase activity was not restored despite several washes of each compound with kinase buffer prior to kinase assays (Fig. 6, A and B, lanes 2). Wis1 kinase activity was examined after Wis1 kinase, preincubated with each compound, was washed three times with buffers of increasing NaCl concentrations in order to determine the reversibility of Wis1 binding to these compounds. The activity of Wis1 kinase preincubated with HWY 289 or HWY 336 was restored when washed with buffer containing 500 mM NaCl, but was not with buffer containing 100 mM NaCl (Fig. 6,  A and B). These results indicate that binding of HWY 289 and HWY 336 to Wis1 kinase is strong enough to withstand wash- ing by a buffer containing 100 mM NaCl, but the binding is reversible, probably by noncovalent interactions. These results also explain why washing with media alone failed to reverse HWY 289 and HWY 336 inhibition of S. pombe proliferation in vivo (Fig. 2B).
Competition of HWY 289 and HWY 336 with Protein Substrate for Wis1 Binding-From their structures and strong binding to Wis1 kinase, we expected that HWY 289 and HWY 336 might inhibit the kinase activity of Wis1 by binding to an ATP-binding site as many other known chemical kinase inhibitors do (36,37). Therefore, we first examined whether HWY 289 and HWY 336 competed with ATP to bind to Wis1 kinase. Purified Wis1 kinase was preincubated with ATP before incubation with HWY 289 or HWY 336, and its kinase activity was compared with that of Wis1 preincubated with HWY 289 or HWY 336 prior to adding ATP (Fig. 7, A and B, lanes 2 and  lanes 4). However, Wis1-inhibitory action of HWY 289 or HWY 336 was not altered by the relative order of incubation with ATP and these compounds (Fig. 7, A and B, lanes 4), demonstrating that the bindings of ATP and these compounds to Wis1 are independent. These findings suggest that HWY 289 and HWY 336 do not bind to the ATP-binding site of Wis1 and in fact bind to Wis1 regardless of ATP binding to inhibit kinase activity.
Since HWY 289 and HWY 336 do not compete with ATP for binding to the Wis1 kinase, we further analyzed the inhibitory action of these compounds on Wis1 by examining whether these compounds compete with protein substrate for binding to Wis1. In the previous kinase assays described in Figs. 4 and 5, the purified Wis1 kinase was preincubated with each berberine derivative before the substrate MBP. In this experiment, Wis1 kinase was preincubated with MBP prior to HWY 289 or HWY 336. Surprisingly, after preincubating with MBP, concentrations of HWY 289 and HWY 336 that had completely inhibited Wis1 no longer inhibited Wis1 kinase (Fig. 7, A and B, lanes 3). Conversely, these compounds inhibited Wis1 kinase activity only when added before the substrate (Fig. 7, A and B, lanes 2). These observations strongly suggest that HWY 289 and HWY 336 compete with protein substrates for binding to the Wis1 kinase or at least prevent the access of substrates to their binding targets on the kinase. Taken together, HWY 289 and HWY 336 do not inhibit kinase activity by binding to Wis1 at ATP-binding sites but rather by competing for specific protein substrate binding sites. This behavior is contrary to many other known kinase inhibitors that block kinase activity by binding to ATP-binding sites. DISCUSSION A panel of natural and synthetic compounds can be screened for increased or decreased cytotoxicity against one or more mutants compared with wild type cells, to identify targets of known compounds or conversely to select useful compounds from a chemical library. This approach exploits yeast mutations in genes for physiological importance including cell cycle regulation and signal transduction; these mutants should be viable but show different sensitivities to test compounds when compared with wild type cells. Such studies with the budding yeast S. cerevisiae have provided invaluable insights into the actions of diverse drugs and compounds with quite specific activities in both mammals and fungi (10,11). For example, the targets of rapamycin, cyclosporin, and FK506 for cell cycle block were first identified in S. cerevisiae, which block T-lymphocyte function and are used as immunosuppressants (7,8,12). Selective protein kinase inhibitors were also developed with S. cerevisiae by creating additional trisubstituted purine analogues and screening purine libraries for potent inhibitory activities against Cdc28p and other yeast kinases (36,37).
In this paper, we identify a target of newly modified berberine derivatives with anti-Candida activity by using a range of mutants in fission yeast S. pombe and by verifying direct inhibition of target activity by these compounds. The novel berberine derivatives synthesized by modifying berberine, HWY 289, and HWY 336, blocked the proliferation of wild type and other mutants of S. pombe except the wis1 deletion mutant. These compounds also specifically inhibited Wis1 MAPKK by binding with strong affinity, while the lead compound berberine did not affect its activity. However, these compounds appeared not to  influence other kinases in MAPK cascades of S. pombe in vitro. Previous genetic studies have reported that wis1 as well as its direct upstream activators, win1 and wis4, and the downstream target spc1 are required for survival in high osmolarity conditions, but their deletion mutants are not lethal (24 -26). Our genetic data showed that HWY 289 and HWY 336 did not inhibit the proliferation of wis1 deletion mutant but did block the proliferation of win1, wis4, and spc1 deletion mutants, demonstrating the specific effect of HWY 289 and HWY 336 on the growth of wis1 deletion mutant but not on the deletion mutants of its upstream regulators and its downstream target. These observations also represent the difference between the effect of these compounds on wild type and the phenotype of a wis1 null mutation. Taken together, these results strongly suggest that HWY 289 and HWY 336 do not show cytotoxicity by simply inhibiting Wis1 kinase activity. One possibility that would account for both genetic and biochemical data of the specific inhibition of Wis1 kinase by these berberine derivatives could be that HWY 289 or HWY 336 tightly binds to Wis1 and recruits another target protein required for the viability of S. pombe by forming a ternary complex. This hypothesis is analogous to the mechanisms of cytotoxicity of FK506 and rapamycin. FK506 forms a complex with highly conserved prolylisomerase immunophilin FKBP12, which in turn targets calcineurin, a Ca 2ϩ -calmodulin-regulated serine/threoninespecific protein phosphatase (8,38,39). Calcineneurin has been reported as a conserved target of the FK506-FKBP12 complex from yeast and T-cells of human to pathogenic fungi (40). Rapamycin also forms a complex with prolylisomerase immunophilin FKBP12, but rapamycin-FKBP12 complex selectively binds to TOR kinases conserved from yeast to human to inhibit TOR-dependent signaling pathways (7,41,42). As we observed that wis1 deletion mutant cells are viable and insensitive to HWY 289 and HWY 336, S. cerevisiae mutants lacking FKBP12 are viable and are resistant both to FK506 and rapamycin (7,8). In our future study, identification of a specific target for the berberine derivative-Wis1 complex that is responsible for cytotoxicity would be necessary to validate this possibility.
HWY 289 and HWY 336 were first screened for anti-fungal activity in C. albicans, and their binding target identified in fission yeast S. pombe was Wis1 MAPKK. The binding target identified in fission yeast can easily explain why these compounds inhibited the hyphal development, since the mutations in C. albicans of MAPK signaling components cause defects in hyphal development and are avirulent (43)(44)(45). HWY 289 and HWY 336 might bind and inhibit a Wis1 homologue, Hst7 MAPKK, in the MAPK cascade for hyphal formation and virulence of C. albicans. However, the anti-fungal effect of HWY 289 and HWY 336 not only on hyphae development form but also on yeast growth cannot be fully explained by their inhibition of a Wis1 homologue in C. albicans, since its deletion does not inhibit yeast growth. The anti-Candida effect of these compounds can be explained by the hypothesis mentioned above that these berberine derivatives bind to the Wis1 homologue of C. albicans and recruit the conserved target protein required for viability by forming a ternary complex.
The stress-activated Wis1-Spc1 MAP kinase cascade is highly conserved in higher eukaryotes. Five different MAP kinase cascades have been identified in mammalian cells, and the stress-activated MKK4/7-c-Jun N-terminal kinase (JNK) and MKK3/6-p38 cascades are likely to correspond to the Wis1-Spc1 pathway of S. pombe (13). No selective inhibitor for MKK3/6 and MKK4/7 has been discovered yet. Since HWY 289 and HWY 336 block the proliferation of S. pombe except wis1 deletion mutant and specifically inhibit the activity of Wis1 kinase, it would be interesting to investigate their cytotoxic effect on mammalian cells and their specific inhibition of MKK4/7 and MKK3/6. According to our observations, HWY 289 and HWY 336 selectively and strongly bind to Wis1 kinase and block its kinase activity. Since HWY 289 and HWY 336 bound to Wis1 kinase could be washed out by high concentrations of salt in vitro, the strong binding of HWY 289 and HWY 336 to Wis1 kinase is probably due to noncovalent interactions that are essentially irreversible in physiological conditions. Surprisingly, converse to many other known kinase inhibitors that are targeted to the ATP-binding site, HWY 289 and HWY 336 do not compete with ATP but rather with protein substrates for specific binding to Wis1. Since many kinase inhibitors identified are targeted to the ATP-binding sites, the design of highly specific inhibitors for kinases has been difficult (36,37). With further modifications, novel berberine derivatives, such as HWY 289 and HWY 336, that show high specificity and do not target the ATPbinding site could be valuable as kinase inhibitors of high specificity. Specific inhibitors for each kinase in distinct MAPK cascades could be developed by various modifications to berberine. We tried to map the possible binding site of HWY 289 and HWY 336 by a three-dimensional modeling of Wis1 kinase. By considering the structures of HWY 289, HWY 336, and Wis1 kinase as well as the competition of these compounds with protein substrates for binding, we speculate that an extended pocket in the active site of Wis1 kinase would be a plausible binding site for HWY 289 and HWY 336 (Fig. 8). To understand the structural basis of HWY 289 and HWY 336 selectivity, we also compared deduced three-dimensional structures of each kinase in the MAP kinase cascade of S. pombe including Wis1, Win1, and Spc1 (data not shown). Each consists of a very conserved structure except an opening formed by an extended pocket near the peptide substrate-binding site of each kinase. Our speculation of an extended pocket of Wis1 kinase for a plausible binding site of HWY 289 and HWY 336 would be supported by the observations in mammalian MAP kinases. p38 and ERK2 have similar three-dimensional structures, but their relative openness of the pocket near the substrate-binding site is responsible for inhibitor selectivity (46). This specu- FIG. 8. A possible model for a binding of HWY 289 to Wis1 kinase. A three-dimensional structure of Wis1 kinase was deduced by comparative modeling with x-ray crystallography structures and the sequence and position of amino acids of CDK2, which shows high sequence identity (30.3%) with Wis1. In a three-dimensional modeling of Wis1, each putative domain for substrate and ATP binding, and active phosphorylation is designated. An extended pocket in the active site of the Wis1 kinase is inferred as the most plausible binding site for HWY 289. lation might also be supported by our unpublished findings that several other derivatives synthesized by further modifications of berberine specifically inhibit the activity of other kinases in S. pombe MAPK cascades other than Wis1. 2 Further investigation into the specific binding modes of berberine derivatives to kinases will provide clues about how to develop specific inhibitors against distinct members of MAP kinase cascades by additional modification.