Introduction

Top Summary Introduction Experimental Procedures Results Discussion References

The phosphodiesterase (PDE) isozymes found in rod and cone photoreceptor cells of the retina belong to a large family of cyclic nucleotide PDEs that catalyze cAMP and cGMP hydrolysis. The photoreceptor PDE [constituting the photoreceptor phosphodiesterase (PDE6) subfamily] is the central effector enzyme involved in the pathway of visual excitation in vertebrate photoreceptors (Pfister et al., 1993; Pugh and Lamb, 1993; Takemoto et al., 1993; Helmreich and Hofmann, 1996). In the dark-adapted state, the rod PDE holoenzyme consists of a catalytic heterodimer (P) tightly associated with two inhibitory  subunits (P); this ₂ holoenzyme is the nonactivated state of the enzyme. The carboxyl terminus of the P subunit is thought to directly bind to the active site and inhibit cGMP hydrolysis in a competitive manner (Granovsky et al., 1997). Activation of the visual pigment rhodopsin by illumination results in activation of the retinal G protein transducin. The activated form of transducin relieves the inhibitory constraint of the P subunit and activates PDE >300-fold in vitro (Arshavsky et al., 1992). Subsecond decreases in cGMP levels resulting from light-activated PDE6 directly regulate the cyclic nucleotide-gated ion channel, leading to the photoreceptor receptor potential (Koutalos and Yau, 1993; Yarfitz and Hurley, 1994; Palczewski and Saari, 1997). In addition to the catalytic sites, the  and  subunits of PDE6 contain noncatalytic cGMP binding sites, which can bind 2 mol of cGMP/mol of PDE holoenzyme with high affinity (Yamazaki et al., 1980; Gillespie and Beavo, 1989b; Cote and Brunnock, 1993). Activation of PDE by transducin reduces the binding affinity of cGMP at the noncatalytic sites (Yamazaki et al., 1982; Cote et al., 1994).

The PDE6 subfamily is most closely related to the PDE5 (cGMP binding, cGMP-specific PDE) isozyme, as judged by several criteria: 1) similarity of the amino acid sequence of the catalytic subunits (McAllister-Lucas et al., 1993); 2) strong preference for cGMP over cAMP as a substrate (Baehr et al., 1979; Thomas et al., 1990); 3) presence of a putative PDE5-associated, low-molecular-weight protein immunologically related to the PDE6 P subunit (Lochhead et al., 1997); and 4) pharmacological properties (Gillespie and Beavo, 1989a).

In the only pharmacological comparison of the efficacy of different PDE inhibitors to block PDE6 catalysis, Gillespie and Beavo (1989a) reported that two inhibitors that preferentially act on PDE5 (i.e., zaprinast and dipyridamole) were also good inhibitors of PDE6. Surprisingly, these two drugs could also activate PDE6 catalysis under certain conditions (Gillespie and Beavo, 1989a). However, this study was unable to determine the molecular basis by which these drugs exerted both inhibitory and stimulatory effects on PDE6.

The interest in PDEs as molecular targets of drug action has grown with the development of isozyme-selective PDE inhibitors that offer potent inhibition of selected isozymes without the side effects attributed to nonselective inhibitors such as theophylline. PDE5-selective inhibitors show promise as therapeutic agents in several areas, including pulmonary and cardiovascular diseases (Sybertz et al., 1995). However, little is known about the potency, selectivity, and mechanism of action of PDE5-selective inhibitors with PDE6. We have chosen to examine E4021 (Takase et al., 1993, 1994; Saeki et al., 1995), a PDE5 inhibitor that is significantly more potent and selective than zaprinast (an early PDE5 inhibitor) and comparable in its PDE inhibitory properties to sildenafil (Viagra; Ballard et al., 1998), a drug recently approved for the treatment of male erectile dysfunction.

In this article, we report on the mechanism of action of E4021 on both the nonactivated and activated forms of rod PDE6 because both states are relevant to understanding how PDE5-selective inhibitors may alter signal transduction pathways in photoreceptor cells. Our results are consistent with a single site of action of E4021 at the catalytic site of the enzyme. However, the effects of E4021 on PDE6 depend on whether the P subunits are associated with the catalytic P dimer. We conclude that the mutually exclusive binding of drug, P, and cGMP at the catalytic site must be accounted for when evaluating the potency and selectivity of PDE5-targeted drugs to alter PDE6 activity in retinal photoreceptor cells.

	Experimental Procedures

Top Summary Introduction Experimental Procedures Results Discussion References

Materials. E4021 was a gift of Eisai Co., Ltd. (Tokyo, Japan), and zaprinast was a gift from Rhone-Poulenc Rorer (Dagenham, UK). Stock solutions of the PDE inhibitors were prepared as follows: 100 mM E4021 was prepared fresh daily in 1-methyl-2-pyrrolidinone; 400 mM zaprinast was dissolved in 1-methyl-2-pyrrolidinone and stored for up to 3 months at 20°C; and 600 mM 3-isobutyl-1-methylxanthine (IBMX) was prepared in 1-methyl-2-pyrrolidinone and stored at 20°C.

Preparation of PDE. Purified frog rod outer segments (ROS) were isolated in the dark (using infrared illumination) using modifications of a method described previously (Cote and Brunnock, 1993). Briefly, frog retinas were removed from enucleated eyes, and ROS were detached from the retinas by gentle shaking into a Ringer's solution containing 5% Percoll. The ROS were then purified by centrifugation in a discontinuous Percoll gradient consisting of 5%, 30%, and 60% Percoll. After dilution of the Percoll with Ringer's and brief centrifugation of the ROS (1 min at 3000g), the ROS were resuspended in the appropriate assay medium and homogenized at 4°C using a nylon pestle and glass mortar. No structure was detectable by phase-contrast microscopy after homogenization. Endogenous nucleotides were depleted by incubating the homogenized ROS at room temperature for 30 min; this treatment has been shown to remove >95% of the endogenous, bound cGMP (Cote and Brunnock, 1993).

PDE Activity Assay. Nucleotide-depleted PDE was diluted to 45 µl with PDE activity assay medium. Unless indicated, PDE was preincubated with the inhibitor for 15 min at 21°C before initiating the PDE activity assay with the addition of 5 µl of the cGMP substrate. The reaction was stopped by quenching 10-µl samples in 50 µl of 100 mM HCl at various times. To quantify the amount of cyclic nucleotide that was hydrolyzed, each sample was neutralized with 50 µl of 100 mM Trizma base, treated with 10 µl of 12.5 mg/ml snake venom (to convert the 5'-nucleoside monophosphate to the nucleoside plus inorganic phosphate), and analyzed using either a radiotracer assay followed by separation of nucleotides by diethylaminoethyl anion exchange chromatography (Kincaid and Manganiello, 1988) or a colorimetric assay to quantify the amount of inorganic phosphate (Artemyev et al., 1998). Total substrate hydrolysis was <30% in all experiments. The calculation of the rate of cyclic nucleotide hydrolysis was based on three to six individual time points during which the time course remained linear.

Filter Binding Assay of cGMP Binding to Noncatalytic Sites of PDE. Nucleotide-depleted PDE holoenzyme was diluted to 7.5 nM with binding assay medium and incubated overnight at 4°C with 10 mM EDTA to prevent hydrolysis of [³H]cGMP used in the binding assay. [EDTA inhibits hydrolytic activity by chelating divalent cations, which are a required cofactor for catalysis (Srivastava et al., 1995). The long incubation period for effective inhibition probably reflects slow dissociation of divalent cations from the active site due to high-affinity binding of P to this region (Granovsky et al., 1997). The effects of EDTA on PDE function were limited to inhibition of the catalytic site and were fully reversible on the addition of excess Mg²⁺.] The PDE was then mixed with the indicated concentration of PDE inhibitor, and preincubated for 15 min at 21°C. The sample was cooled to 4°C, and the binding assay was initiated with [³H]cGMP (final PDE concentration, 3 nM). After incubating for 1 h at 4°C, 50-µl samples were filtered in duplicate on prewet nitrocellulose filters (HAWP 025; Millipore) using a vacuum manifold system. PDE activity measurements showed that 2% of the cGMP was hydrolyzed during the filter binding assay in the absence of PDE inhibitor.

Fluorescent Assay of P-1-83BC Binding to PDE. The ability of E4021 to displace P bound to P was measured with a fluorescence assay essentially as described previously (Granovsky et al., 1997; Artemyev et al., 1998). Briefly, a mutant of P was constructed in which the four C-terminal residues were replaced with a cysteine; this mutant was then labeled with 3-(bromoacetyl)-7-diethylaminocoumarin to form P-1-83BC. P-1-83BC (10 nM) was mixed with 3.6 nM trypsin-activated PDE. To perform the fluorescence measurements, the trypsin-activated PDE was separated from the ROS membranes by centrifugation (180,000g for 5 min at 21°C) before addition to the cuvette.

Data Analysis. Dose-response curves at a single substrate concentration and in the absence of drug-induced stimulation were fit to the Hill equation using nonlinear least-squares analysis. The IC₅₀ value determined in this way was related to the drug inhibition constant (K_I) by the following equation (Cheng and Prusoff, 1973): K_I = IC₅₀/(1 + [S]/K_M), where [S] is the substrate concentration, K_M is the Michaelis constant, and the mechanism of inhibition is assumed to be competitive. For nonactivated PDE holoenzyme, a K_M value for cGMP (95 µM) was verified in our laboratory to be the same as that reported earlier (Dumke et al., 1994); for trypsin-activated PDE, the K_M for cGMP hydrolysis was experimentally determined to be 21.8 ± 2.2 µM (mean ± S.E.M., n = 7). Kinetic analyses over a range of cGMP and E4021 concentrations were used to more accurately determine the K_I value using a Dixon plot (Segel, 1975). Analysis of fluorescence data was performed as described previously (Granovsky et al., 1997). All experiments were repeated at least three times, and average values are reported as the mean ± S.E.M.

	Results

Top Summary Introduction Experimental Procedures Results Discussion References

E4021 Is a Simple Competitive Inhibitor of Activated PDE6. We first examined the potency of E4021 to inhibit the activated form of frog rod PDE6 in which the P subunits have been removed by limited proteolysis. We found that E4021 is able to effectively inhibit catalysis at an IC₅₀ value of 70 nM (Fig. 1). This value is 20- and 3000-fold lower than the IC₅₀ values for zaprinast (a prototype PDE5 inhibitor) and IBMX (a nonselective PDE inhibitor), respectively (Fig. 1). With the IC₅₀ data in Fig.1 and knowledge of the K_M value for trypsin-activated PDE (21.8 ± 2.1 µM), we can calculate K_I values of 1.6 nM for E4021, 32 nM for zaprinast, and 4.7 µM for IBMX. These results indicate that E4021 is the most potent inhibitor of PDE6 identified to date.

View larger version (29K): [in this window] [in a new window]   Fig. 1.   E4021 is a potent inhibitor of the photoreceptor PDE. Activated PDE (0.1 nM final concentration) was prepared as described in Experimental Procedures and mixed with the indicated concentrations of E4021 (), zaprinast (), or IBMX () for 15 min before the addition of 1.0 mM cGMP. Samples were quenched, and PDE activity was determined with a colorimetric assay. Results were normalized to the rate in the absence of drug (34 ± 15 pmol of cGMP hydrolyzed/s) for three separate experiments. Lines are a fit of data to the Hill equation with the following coefficients: E4021: IC50 = 73 ± 5.0 nM, n = 0.9 ± 0.1, r2 = 0.998; zaprinast: IC50 = 1.5 ± 0.1 µM, n = 1.0 ± 0.1, r2 = 0.999; and IBMX: IC50 = 220 ± 30 µM, n = 1.2 ± 0.2, r2 = 0.994.

View larger version (32K): [in this window] [in a new window]   Fig. 2.   Dixon plot shows that E4021 is a simple competitive inhibitor of activated PDE. PDE activity assays were conducted with 50 pM activated PDE and at cGMP concentrations of 89 (), 31 (), 21 (), 15 (), and 10 µM () and with the indicated amounts of E4021 added (see Experimental Procedures). The reciprocal of the initial rate of cGMP hydrolysis is plotted versus the final E4021 concentration. Point of intersection in this individual experiment was KI = 2.0 nM; the average KI value for five separate determinations was 1.7 ± 0.3 nM. Inset, slope of each line at a fixed cGMP concentration was determined by regression analysis and plotted as a function of the reciprocal substrate concentration. Intersection of the line on the y-intercept (0.16 ± 0.14) was not statistically different from zero at the 95% level of confidence, consistent with a simple competitive inhibition mechanism.

PDE Holoenzyme Is Inhibited by E4021 in a Complex Manner. Because PDE is normally associated with its P subunits in vivo, we next tested the relative potency of E4021 to inhibit PDE holoenzyme (₂) that was not activated. The hydrolytic activity of this PDE preparation is low compared with the maximum rate attainable for fully activated PDE; the basal rate reflects the equilibrium binding of the P subunits to the catalytic P dimer (Hurley and Stryer, 1982; Wensel and Stryer, 1986). The PDE holoenzyme preparation was incubated with inhibitor 15 min to permit the inhibitor to attain binding equilibrium at the catalytic site. After preincubation with the drug, the hydrolytic activity of PDE was determined by the addition of 200 µM cGMP. Figure 3 shows that even though the cGMP concentration (200 µM) is 5-fold less than that shown in Fig. 1, higher concentrations of PDE5-selective drugs were needed to inhibit 50% of the PDE activity of the nonactivated holoenzyme compared with the activated catalytic dimer. In contrast, IBMX displayed a similar potency of inhibition for nonactivated holoenzyme and activated PDE. To quantify the difference in the ability of E4021 to inhibit nonactivated and activated PDE, a Dixon plot was also constructed for PDE holoenzyme at several cGMP and E4021 concentrations. The lines intersected at a single point above the x-axis with a K_I value of 70 nM; the slope replot was linear and intersected the y-axis at the origin (data not shown). These results are consistent with E4021 acting as a pure competitive inhibitor of cGMP hydrolysis at the active site. However, the presence of the P subunits associated with the P catalytic dimer reduce the inhibitory potency of the drug ~40-fold compared with the activated P dimer.

View larger version (30K): [in this window] [in a new window]   Fig. 3.   Complex inhibition of nonactivated PDE by inhibitors. PDE holoenzyme (12 nM) was incubated for 15 min with the indicated concentrations of E4021 (), zaprinast (), or IBMX () as described in Experimental Procedures. Hydrolytic activity of PDE was determined after addition of 200 µM cGMP. PDE activity is reported as the percent of the rate observed in the absence of inhibitor: E4021, 90 ± 18 cGMP hydrolyzed/PDE holoenzyme/s (n = 5); zaprinast, 90 ± 17 cGMP hydrolyzed/PDE holoenzyme/s (n = 3); and IBMX, 120 ± 46 cGMP hydrolyzed/PDE holoenzyme/s (n = 3). The lines are fits to the Hill equation, with the following parameters: E4021: IC50 = 230 ± 30 nM, n = 1.8 ± 0.4, r2 = 0.992; zaprinast: IC50 = 2.4 ± 0.4 µM, n = 1.3 ± 0.2, r2 = 0.996; and IBMX: IC50 = 65 ± 12 µM, n = 0.7 ± 0.1, r2 = 0.993.

View larger version (25K): [in this window] [in a new window]   Fig. 4.   E4021 stimulates the activity of PDE holoenzyme when high levels of cGMP are used. The indicated concentrations of E4021 were preincubated for 15 min with 12 nM PDE; then 0.2 mM (), 1.0 mM (), or 10 mM (), cGMP was added, and hydrolytic activity was determined. The y-axis is normalized to the rate of hydrolysis in the absence of E4021 and corresponds to 90 ± 18 (0.2 mM cGMP; n = 5), 122 ± 20 (1 mM cGMP; n = 3), or 150 ± 15 (10 mM cGMP; n = 3) cGMP hydrolyzed/PDE/s. The 0.2 mM cGMP line was fit using the Hill equation (IC50 = 230 ± 30 nM, n = 1.8 ± 0.4, r2 = 0.992); the other curves did not fit a simple competitive displacement model. The dashed line is the dose-response function of activated PDE with 1 mM cGMP as substrate, taken from Fig. 1. *PDE activities that were significantly greater than 100% at the 95% level of confidence.

E4021 Interacts Very Weakly with Noncatalytic cGMP Binding Sites on PDE Holoenzyme. Previous work has shown that high-affinity binding of cGMP to the noncatalytic binding sites on PDE requires that P be associated with the catalytic P dimer (Yamazaki et al., 1980; Cote et al., 1994); we verified for this study that trypsin-activated PDE (in which the P subunits have been proteolyzed) was unable to bind [³H]cGMP (<0.2 mol of cGMP/mol of P; M.R. D'Amours and R.H. Cote, unpublished observations). We then reasoned that the differences in how E4021 acts on the activated P dimer (Fig. 1) versus the nonactivated PDE holoenzyme (Figs. 3 and 4) might be related to the occupancy of the noncatalytic sites by cGMP in the latter case.

View larger version (29K): [in this window] [in a new window]   Fig. 5.   PDE inhibitors weakly displace [3H]cGMP from noncatalytic binding sites on PDE holoenzyme. EDTA-treated PDE (final concentration 3.0 nM) was mixed with the indicated concentration of inhibitor for 15 min at 21°C before cooling to 4°C and the subsequent addition of 16 nM [3H]cGMP. After a 60-min incubation period at 4°C, the amount of bound [3H]cGMP was determined with a filter binding assay (see Experimental Procedures). The data from three separate experiments was first normalized to the extent of cGMP binding in the absence of drug (0.8 ± 0.1 cGMP bound/PDE holoenzyme). The dashed line represents the predicted displacement curve on the addition of unlabeled cGMP, assuming a KD value of 10 nM for cGMP binding to EDTA-treated PDE holoenzyme (M. R. D'Amours and R. H. Cote, unpublished observations) and simple competitive inhibition.

E4021 Is a Competitive Inhibitor of P Binding to Catalytic Site on PDE. Having ruled out the noncatalytic cGMP binding sites on PDE as the site by which E4021 can activate PDE holoenzyme, we considered other potential drug binding sites that might account for the anomalous behavior of E4021. Although all of the unusual effects of E4021 are exhibited only when P is present, we consider it very unlikely that E4021 can bind directly to P because the P subunit lacks sufficient secondary and tertiary structure (Berger et al., 1997) to form a high-affinity binding pocket for E4021 or other PDE5-selective drugs. Rather, it seems most likely that E4021 occupies a high-affinity binding site on P that also is a site of interaction with the P subunit. Although a novel drug-binding site on PDE cannot be excluded, we first considered whether the effects of E4021 on PDE could be ascribed to high-affinity binding at the active site of the enzyme in direct competition with P binding there.

View larger version (17K): [in this window] [in a new window]   Fig. 6.   E4021 displaces a fluorescently labeled mutant of P at the active site of PDE. Activated PDE (3.6 nM) lacking endogenous P subunits was mixed with 10 nM P-1-83BC (intrinsic fluorescence before adding PDE: F/F0 = 1.2), and the fluorescence enhancement was allowed to reach equilibrium. The indicated concentrations of E4021 were added to the cuvette, and the change in fluorescence was recorded. Data points represent the mean ± S.D. of three experiments. The line is the fit to a Hill equation (IC50 = 14.6 ± 1.6 nM, n = 1.1 ± 0.1; r2 = 0.99).

	Discussion

Top Summary Introduction Experimental Procedures Results Discussion References

Our results demonstrate that E4021 is the most potent inhibitor of PDE6 studied to date. However, it appears that although PDE5-selective inhibitors may show good discrimination of PDE5 from most other PDE isoforms, there is likely to be poor discrimination of these drugs between PDE5 and PDE6 isozymes. Our results also demonstrate that the efficacy and mechanism of drug inhibition depend on the subunit composition and state of activation of PDE6. Activated PDE (lacking bound P) is inhibited by E4021 (Fig. 1) or zaprinast (Gillespie and Beavo, 1989a) (Fig. 1) in a simple competitive manner, whereas the nonactivated PDE holoenzyme (containing bound P) is affected by PDE5-selective inhibitors in a complex manner that depends on several factors. Our results rule out a major effect of E4021 on the noncatalytic sites on PDE and support a mechanism of action in which E4021, P, and cGMP all compete in a mutually exclusive manner at the active site.

Comparison of Potency and Selectivity of E4021 with Other PDE5-Selective Inhibitors. With a K_I value of 1.7 nM for the activated enzyme, E4021 is the most effective pharmacological inhibitor of PDE6 thus far reported. By comparison, older PDE5-selective drugs such as zaprinast and dipyridamole are >10-fold less potent in inhibiting the activated PDE6. Recently, another PDE5-selective drug, sildenafil, was reported to inhibit nonactivated bovine PDE6 with a K_I value of 30 nM (Ballard et al., 1998), ~2-fold lower than the K_I value for E4021 inhibition of nonactivated frog PDE6 (see Results). Based on the 40-fold increase in potency of E4021 with activated PDE, we predict that sildenafil may also act much more potently (K_I ~ 1 nM) on the activated form of PDE6.

Mutually Exclusive Competitive Binding of E4021 and P at Active Site Can Account for Complex Effects of E4021 on PDE Holoenzyme. The high affinity with which E4021 can both inhibit cGMP hydrolysis at the active site of PDE and displace a fluorescently labeled P mutant (Fig. 6) supports the idea that E4021 exerts its complex effects on PDE holoenzyme solely through direct binding of the drug at the active site. We describe how the K_I shift and the apparent stimulation of hydrolysis by E4021 can be explained by a model in which P, E4021, and cyclic nucleotides compete for binding to the active site of PDE.