Journal list menu

Volume 267, Issue 9 p. 2760-2767
Free Access

Phosphorylation of phosphodiesterase-5 by cyclic nucleotide-dependent protein kinase alters its catalytic and allosteric cGMP-binding activities

Jackie D. Corbin

Jackie D. Corbin

Department of Molecular Physiology and Biophysics, Vanderbilt University School of Medicine, Nashville, TN, USA

Search for more papers by this author
Illarion V. Turko

Illarion V. Turko

Department of Molecular Physiology and Biophysics, Vanderbilt University School of Medicine, Nashville, TN, USA

Search for more papers by this author
Alfreda Beasley

Alfreda Beasley

Department of Molecular Physiology and Biophysics, Vanderbilt University School of Medicine, Nashville, TN, USA

Search for more papers by this author
Sharron H. Francis

Sharron H. Francis

Department of Molecular Physiology and Biophysics, Vanderbilt University School of Medicine, Nashville, TN, USA

Search for more papers by this author
First published: 25 December 2001
Citations: 212
J. D. Corbin, Department of Molecular Physiology and Biophysics, 702 Light Hall, Vanderbilt University School of Medicine, 21st and Garland Avenues, Nashville, TN 37232-0615, USA. Fax: + 1 615 343 3794, Tel.: + 1 615 322 4384, E-mail: [email protected]


In addition to its cGMP-selective catalytic site, cGMP-binding cGMP-specific phosphodiesterase (PDE5) contains two allosteric cGMP-binding sites and at least one phosphorylation site (Ser92) on each subunit [Thomas, M.K., Francis, S.H. & Corbin, J.D. (1990) J. Biol. Chem.265, 14971–14978]. In the present study, prior incubation of recombinant bovine PDE5 with a phosphorylation reaction mixture [cGMP-dependent protein kinase (PKG) or catalytic subunit of cAMP-dependent protein kinase (PKA), MgATP, cGMP, 3-isobutyl-1-methylxanthine], shown earlier to produce Ser92 phosphorylation, caused a 50–70% increase in enzyme activity and also increased the affinity of cGMP binding to the allosteric cGMP-binding sites. Both effects were associated with increases in its phosphate content up to 0.6 mol per PDE5 subunit. Omission of any one of the preincubation components caused loss of stimulation of catalytic activity. Addition of the phosphorylation reaction mixture to a crude bovine lung extract, which contains PDE5, also produced a significant increase in cGMP PDE catalytic activity. The increase in recombinant PDE5 catalytic activity brought about by phosphorylation was time-dependent and was obtained with 0.2–0.5 μm PKG subunit, which is approximately the cellular level of this enzyme in vascular smooth muscle. Significantly greater stimulation was observed using cGMP substrate concentrations below the Km value for PDE5, although stimulation was also seen at high cGMP concentrations. Considerably higher concentration of the catalytic subunit of PKA than of PKG was required for activation. There was no detectable difference between phosphorylated and unphosphorylated PDE5 in median inhibitory concentration for the PDE5 inhibitors, sildenafil, or zaprinast 3-isobutyl-1-methylxanthine. Phosphorylation reduced the cGMP concentration required for half-maximum binding to the allosteric cGMP-binding sites from 0.13 to 0.03 μm. The mechanism by which phosphorylation of PDE5 by PKG could be involved in physiological negative-feedback regulation of cGMP levels is discussed.


  • PDE
  • phosphodiesterase
  • PKA
  • cAMP-dependent protein kinase
  • PKG
  • cGMP-dependent protein kinase
  • IBMX
  • 3-isobutyl-1-methylxanthine
  • Cyclic nucleotides mediate the regulation of myriad physiological processes, and the levels of these nucleotides are altered by signals that modulate activities of adenylate and guanylate cyclases or cyclic nucleotide phosphodiesterases (PDEs). Phosphodiesterases have emerged as important targets of regulation involving three different schemes: substrate availability, extracellular signals that alter intracellular signaling, and feedback regulation [1]. The latter two may be initiated by various processes including protein phosphorylation and binding of cGMP to allosteric regulatory sites on some PDEs [1–8]. The cGMP-binding cGMP-specific PDE (PDE5) is one of at least 11 families of mammalian PDEs (PDE1–PDE11), and this enzyme is the target of sildenafil for treatment of erectile dysfunction [9]. PDE5 contains both a cyclic nucleotide-dependent protein kinase-catalyzed phosphorylation site and two allosteric cGMP-binding sites in the regulatory domain of each subunit [1,10–12]. Purified PDE5 has been used to demonstrate that interaction of cGMP analogs, including PDE inhibitors, at the catalytic site stimulates cGMP binding at the allosteric sites, which exposes a site (Ser92) for phosphorylation by cGMP-dependent protein kinase (PKG) [11,13]. PDE5 has been shown to be phosphorylated in immunoprecipitates of extracts of intact vascular smooth muscle cells treated with cGMP-elevating agents, and this correlated with increased PDE activity in the immunoprecipitates [14]. However, the increased activity was not shown to be PDE5-specific, and, to date, a functional change resulting from phosphorylation of purified native or recombinant PDE5 at Ser92 has not been demonstrated, although a comprehensive study of this subject has not been carried out. The present investigation shows that phosphorylation of isolated PDE5 by PKG or PKA not only causes activation of catalysis, but also increases the affinity of allosteric cGMP binding of this enzyme. The possible mechanism of PDE5 activation and the likelihood that the overall process represents negative-feedback control of cellular cGMP levels are discussed.

    Experimental procedures


    [8-3H]cGMP was purchased from Amersham Life Science Inc. cGMP, cAMP, Crotalus atrox snake venom, 3-isobutyl-1-methylxanthine (IBMX) and zaprinast were from Sigma. Zaprinast was dissolved in 100% dimethyl sulfoxide and then diluted in aqueous buffer for the experiments. [γ-32P]ATP was obtained from NEN–DuPont. Sildenafil citrate was kindly provided by Pfizer Central Research (Sandwich, Kent, UK). It was dissolved in 100% dimethyl sulfoxide and diluted in aqueous buffer for the experiments. Microcystin-LR was from Alexis Biochemicals (San Diego, CA, USA). Phosphoamino acids and histone VIII-S were purchased from Sigma.

    Preincubation of PDE5 in a phosphorylation reaction mixture

    The COMPLETE phosphorylation reaction mixture contained 0.28 μm recombinant bovine PDE5 holoenzyme, 5 mm potassium phosphate (pH 6.8), 0.5 mm EDTA, 12.5 mm 2-mercaptoethanol, 2 mm magnesium acetate, 0.2 mm ATP, 50 μm cGMP, 0.25 mm IBMX, 0.4 mg·mL−1 BSA, 0.4 μm microcystin, and 1.1 μm purified bovine lung PKG-Iα in a total volume of 50 μL. In certain instances, purified catalytic subunit of PKA at the indicated concentrations was substituted for PKG. Partially purified recombinant PDE5 was diluted in KPEM [10 mm potassium phosphate (pH 6.8), 1 mm EDTA, 15 mm 2-mercaptoethanol] containing 1 μm microcystin and 1 mg·mL−1 BSA immediately before addition to the reaction mixture. Individual components of the phosphorylation mixture were sometimes omitted as indicated. The reaction was started by adding PKG and PDE5 in rapid succession and placing tubes in a water bath at 30 °C. After incubation for 1 h, tubes were placed on ice and diluted 250 times in cold KPEM containing 1 μm microcystin to minimize carry-over of potentially interfering reaction components into subsequent assays. A 10-μL portion of diluted enzyme was assayed for PDE5 activity using 0.4 μm[3H]cGMP as substrate, unless otherwise indicated, in a final volume of 250 μL as described below. Where indicated, the volume of the reaction mixture was increased to 450 μL and, after preincubation, Sephadex G-25 (superfine) column (0.9 × 25 cm) chromatography instead of dilution was used to minimize carry-over of reaction components into subsequent assays. The entire reaction mixture was applied to the column, which was equilibrated in KPEM, and 1-mL fractions were collected. Bradford protein assay was used to locate the peak fraction (≈ fraction 7) to be used for assays of PDE catalytic activity and cGMP-binding activity.

    In some experiments, ATP in the phosphorylation reaction mixture was spiked with [γ-32P]ATP (≈ 1000 c.p.m.·pmol−1) for determination of incorporation of phosphate into PDE5. After various preincubation times, 15-μL aliquots were taken for SDS/PAGE (10% gels); the gels were stained with Coomassie Brilliant Blue, dried, and subjected to radioautography. The band corresponding to PDE5, which migrates at Mr 92 000, was excised from the gel and counted. The stoichiometry of phosphorylation of PDE5 was calculated by assuming a specific enzyme activity of 5 μmol·min−1·mg−1[10]. The PKG-Iα band, which autophosphorylates during the preincubation, was also excised from the gel and counted, and its stoichiometry of phosphorylation was calculated by assuming a specific enzyme activity of 2.6 μmol·min−1·mg−1[15]. A correction factor for recovery of 32P-labeled protein was obtained in a separate experiment in which PKG-Iα was autophosphorylated using [γ-32P]ATP, and the radioactivity of the PKG-Iα band excised from an identical SDS/polyacrylamide gel was compared with that obtained by spotting the reaction mixture on to P31 paper, washing and counting. The recovery of radioactivity in the PKG-Iα band (62%) was assumed to be the same for PDE5.

    Determination of the ratio of radioactive phosphoserine to phosphothreonine incorporated into PDE5

    In some experiments, after phosphorylation of PDE5 for 20 min in the presence of [32P]ATP as described above, replicate samples were subjected to SDS/PAGE and transferred to a nitrocellulose membrane (ProBlott from Applied Biosystems). Radioactive PDE5 bands were identified by radioautography, excised from the membrane, and suspended in 100 μL 6 m HCl in a Microfuge tube (up to six replicate bands per 100 μL HCl). The sample was partially hydrolyzed by placing it in a heating block at 110 °C for 4 h. The liquid sample was then aspirated, and non-radioactive phosphoserine (0.43 nmol) and phosphothreonine (0.19 nmol) were added as internal standards. The sample was dried and subjected to routine amino-acid analysis using the Waters AccQ.Tag system (C18 column) in the Vanderbilt Peptide Sequencing and Amino acid Analysis Shared Resource. Fractions of volume 1 mL (1-min) were collected during chromatography, the Cerenkov radiation was measured, and the peaks were compared with the peaks of unlabeled phosphoserine (9.30 min) and phosphothreonine (13.24 min). In a separate experiment, standard unlabeled phosphotyrosine chromatographed at 23.45 min. In addition to the free 32Pi peak, only radioactive phosphoserine was detected. As control, autophosphorylated PKG bands were also excised from the membrane and analyzed. After incubation for either 5 min or 60 min, radioactive phosphoserine comprised about 33% and radioactive phosphothreonine about 66% of the total radioactive phosphate detected in PKG for these two amino acids.

    Preparation of rat lung extract

    Lungs were removed from five euthanized Sprague–Dawley rats and homogenized for 30 s in 4 mL·g−1 KPEM using a Cuisinart Mini-Mate Plus grinder. The homogenate was centrifuged at 10 000 g for 30 min, and the supernatant was used for the experiments. Where indicated, the supernatant fraction (16 mL) was applied to a DEAE-Sephacel column (0.9 × 3 cm) equilibrated with KPEM. The column was washed with 50 mL KPEM containing 50 mm NaCl and then developed using a linear gradient of NaCl from 50 to 300 mm (30 mL). Fractions of volume 1 mL were collected, and PDE activity was determined in 5 μL aliquots of each fraction using 0.4 μm cGMP as substrate and a 5-min incubation as described below. Only one main peak of PDE activity was resolved, which was eluted in the position of PDE5. All fractions were inhibited more than 95% by 40 nm sildenafil.

    PDE assay

    PDE activity was determined by slight modifications of the method previously described [16] in a reaction mixture containing 40 mm Mops, pH 7.5, 0.5 mm EGTA, 15 mm magnesium acetate, 0.17 mg·mL−1 BSA, 30 μm cAMP, diluted enzyme, and 0.4 μm[3H]cGMP (unless indicated otherwise) at a specific radioactivity of ≈ 2 × 106 c.p.m.·nmol−1. The indicated values are concentrations in the final assay volume of 250 μL. Reactions at 30 °C were initiated by addition of the PDE5 and were terminated at 15 min by addition of 20 μL of a stop mix containing 143 mm EDTA, 60 mm theophylline, 30 mm cAMP, 30 mm cGMP, and 286 mm Tris/HCl, pH 7.5. Crotalus atrox 5′-nucleotidase (200 μg) was then added to each tube, which was then incubated at 30 °C for 10 min before termination of the reaction by addition of 1 mL of an ice-cold solution of 0.1 mm adenosine, 0.1 mm guanosine and 15 mm EDTA. Samples were then applied to QAE-Sephadex columns (1.5 mL) equilibrated in 0.02 m ammonium formate, pH 7.5, and 4 mL of this buffer was used to wash the columns. Eluates were combined with 10 mL aqueous scintillant and counted.

    cGMP-binding assay

    To 50 μL binding mixture (KPEM, 0.1 mm IBMX, 0.5 mg·mL−1 histone VIII-S) was added 10 μL of diluted stock [3H]cGMP (60 µm), and the binding reaction was started by adding 20 μL undiluted peak protein fraction from Sephadex G-25 chromatography. After 45 min incubation at 0 °C, 1 mL cold KPE was added to each sample before filtration using a 0.45-μM Millipore membrane. Time courses indicated that binding equilibrium was achieved in 20–30 min of incubation. Filter membranes were washed with cold KPE (3 × 2 mL) before being placed in slots of a scintillation counting vial box and dried for 10 min in a drying oven. They were then placed in nonaqueous scintillant and counted.

    PKG and PKA

    Bovine lung PKG-Iα was purified to homogeneity as described [17]. Free catalytic subunit of PKA was purified to apparent homogeneity from bovine heart by a previously published method [18]. Most preparations of highly purified catalytic subunit still contained trace contaminating cGMP PDE activity. To avoid a false-positive PDE activation signal caused by this contamination, after incubation of the phosphorylation reaction mixture, the mixture was diluted 250-fold before measurement of cGMP PDE activity (final dilution in PDE assay = 6250-fold). There was no significant carry-over of contaminating cGMP PDE activity into the PDE assay after this dilution.


    Wild-type bovine PDE5 was expressed in Sf9 cells and purified [19]. Culture medium (≈ 250 mL) was fractionated by sequential ammonium sulfate precipitation (25–45% saturated ammonium sulfate instead of the 25–40% originally described for this step). After centrifugation, the pellet was resuspended in 30 mL 10 mm sodium phosphate, pH 7.2, and centrifuged at 48 000 g for 30 min at 4 °C. The supernatant was loaded on a hydroxyapatite (Bio-Rad) column (1.5 × 15 cm) equilibrated with 10 mm sodium phosphate, pH 7.2. The column was washed with 100 mL 70 mm sodium phosphate and then eluted with 120 mm sodium phosphate at a flow rate of 5 mL·h−1. The pool containing PDE5 activity was diluted with 6 vol. ice-cold water and concentrated to ≈ 1 mL using an Amicon filtration cell equipped with a PM-30 membrane. All purification steps were performed at 4 °C. The final preparation, which was ≈ 5% pure by SDS/PAGE analysis, was stored in small aliquots in 20% glycerol at 70 °C.


    Aliquots of volume 15 µL were boiled in 10% SDS/2 m 2-mercaptoethanol/bromphenol blue (1 mg·mL−1) and subjected to SDS/PAGE (10% gels).


    Effect of phosphorylation of PDE5 on its catalytic properties

    Partially purified recombinant PDE5 was preincubated for 1 h at 30 °C with a mixture of various components (see Experimental procedures) that have been established to produce phosphorylation of this enzyme [11]. Under similar conditions used here, we detected only phosphorylation of Ser92 in PDE5 by PKG in our earlier studies. There are three consensus phosphorylation sites (R/KR/KXS/TX) for PKG in PDE5 (-RKGTR-41, -RKISA-93, and -KRLTD-630), and only the one containing Ser92 has serine as phosphate acceptor. When radioactive phosphoserine vs. phosphothreonine was measured after phosphorylation in the presence of [32P]ATP in the present studies, only radioactive phosphoserine was detected. The method would have detected at least a 10% contribution by radioactive phosphothreonine.

    After phosphorylation, carry-over effects of reaction components were minimized by diluting the reaction mixture 250-fold before determination of PDE activity using 0.4 μm cGMP as substrate. It can be seen in Fig. 1 that the COMPLETE phosphorylation mixture produced 1.5- to 1.7-fold higher PDE activity than was obtained when PKG, ATP, cGMP, or IBMX was omitted from the incubation. The requirement of each of these components for increased PDE activity would be expected if enzyme phosphorylation caused the increase. PKG in the presence of ATP is known to be an effective catalyst for PDE5 phosphorylation [11]. PKG is activated by cGMP, and this nucleotide also binds to allosteric sites on PDE5 as well, which exposes a phosphorylation site (Ser92) on this enzyme for PKG [11]. In addition to protecting cGMP from degradation by inhibiting PDE activity, binding of IBMX to the catalytic domain of PDE5 stimulates cGMP binding to its allosteric sites [10]. Three different preparations of recombinant PDE5 were used to obtain the results shown in Fig. 1. When Sephadex G-25 chromatography was used instead of dilution of the preincubation mixture to minimize carry-over effects of reactants into the PDE assay, the stimulatory effect of the COMPLETE system (+ PKG) was 1.5-fold and 1.6-fold in two separate experiments (not shown).

    Details are in the caption following the image

    Effect of preincubating PDE5 with a phosphorylation reaction mixture on PDE activity: consequences of omission of individual components of the reaction mixture. The composition of the COMPLETE phosphorylation reaction mixture, including enzyme concentrations and conditions of the preincubation are described in Experimental procedures. Various components of the COMPLETE system were omitted as indicated. After preincubation, the reaction mixture was diluted 250-fold for determination of PDE activity using 0.4 μm cGMP as substrate. Results are mean ± SEM (n = 4, each of the four determinations performed in triplicate).

    When a crude rat lung extract was substituted for recombinant PDE5 as a source of this enzyme and incubated with the phosphorylation reaction mixture for 60 min, similar results to those described above were obtained (not shown). The addition of the COMPLETE mixture caused a 1.36 ± 0.07-fold (n = 8) increase in PDE activity, but omission of either ATP, cGMP or IBMX caused loss of this effect. That the PDE activity in the extract was mainly due to PDE5 was suggested by the finding that it was inhibited 92% by 40 nm sildenafil. Furthermore, DEAE-Sephacel chromatography of the extract resolved one main peak of PDE activity using the same assay conditions. This peak was eluted in the position of PDE5. It should be emphasized that all of the PDE assay reaction mixtures included a final concentration of 30 μm cAMP to minimize the contribution of prominent PDEs that utilize either cGMP or cAMP, such as some isoforms of PDE1, PDE2, and PDE3.

    The incubation time course of effect of the COMPLETE phosphorylation reaction mixture on PDE activity and phosphorylation of recombinant PDE5 is shown in Fig. 2. The PDE5 was rapidly phosphorylated (see inset) and activated under these conditions, and there was a positive correlation between these two events. The radioactive Coomassie Blue-stained PDE5 band was excised from the gel and counted. The results indicated that the enzyme incorporated a substantial amount of phosphate (0.6 mol phosphate per mol subunit in 60 min). Because PKG was also included in the phosphorylation incubation, and this enzyme catalyzes autophosphorylation, it was used to verify the methodology. PKG incorporated 2 mol phosphate per subunit in 60 min, which is approximately the value reported previously using similar incubation conditions [20].

    Details are in the caption following the image

    Time course of phosphorylation and activation of PDE5 by incubation in a phosphorylation reaction mixture containing PKG. The experiment was performed as described for the COMPLETE system in Fig. 1 (plus PKG), including the use of the same concentrations of enzymes and other factors. Incubations were also performed (minus PKG) at each time point, and the fold activation by PKG was calculated. Each point represents the mean ± SEM from nine determinations of PDE activity (□) and the mean of two separate determinations for phosphate incorporation (▪), where each determination was performed in duplicate. A radioautograph of SDS/PAGE of the phosphorylated PDE5 band at each time point is shown in the inset.

    A concentration of PKG subunit as low as 0.2 μm in the phosphorylation reaction produced a significant increase in PDE activity of recombinant PDE5 (Fig. 3). This compared favorably with the intracellular concentration of PKG (0.3–0.5 μm) estimated to be present in vascular smooth muscle cells [21]. It can be seen that free catalytic subunit of PKA also produced an increase in PDE activity, albeit at considerably higher concentrations. At the highest level of PKA (5.5 μm), the increase in PDE activity was similar to that produced by 0.4 μm PKG. These results are consistent with our earlier finding that PKA is an ≈ 10-fold weaker catalyst for phosphorylation of PDE5, and that each kinase phosphorylates the same residue (Ser92) [11]. However, as PKA subunit is present at 0.3–0.4 μm in vascular smooth muscle cells [21], this enzyme as well as PKG could catalyze physiological PDE5 phosphorylation and activation. Cellular PKA-catalyzed phosphorylation of PDE5 would presumably require elevation of both cAMP and cGMP, as the latter is needed to bind to the allosteric sites of PDE5 and expose its phosphorylation site [11].

    Details are in the caption following the image

    Effect of PKG or PKA concentration in the phosphorylation reaction mixture on PDE5 activity. The experiment was performed as described for the COMPLETE system in Fig. 1 except that various amounts of either PKG or PKA were used. Purified catalytic subunit was used as the source of PKA. Each point represents the mean ± SEM from four determinations for PKG and three determinations for PKA, where each determination was performed in duplicate.

    After incubation in the phosphorylation reaction mixture in the presence (‘phosphorylated’) and absence (‘unphosphorylated’) of PKG, the effect of varying the cGMP substrate concentration in the PDE assay on recombinant PDE5 enzyme activity was examined. It can be seen in Fig. 4 that the fold effect of PKG was slightly higher at the lowest concentrations of substrate tested (0.1 μm and 0.24 μm), and steadily declined as the cGMP concentration was elevated to 9.6 μm. Stimulation was 1.74 ± 0.10-fold (n = 9) at 0.1 μm, 1.86 ± 0.18-fold (n = 5) at 0.24 μm, 1.65 ± 0.07-fold (n = 12) at 0.4 μm, 1.28 ± 0.04-fold (n = 9) at 1.6 μm, 1.15 ± 0.03-fold (n = 9) at 3.2 μm, 1.10 ± 0.04-fold (n = 6) at 5.6 μm, and 1.13 ± 0.07-fold (n = 9) at 9.6 μm substrate concentration. Phosphorylation caused a modest decrease in the cGMP Km (from 0.96 to 0.58 μm), as calculated by Lineweaver–Burk analysis using the values in Fig. 4. The lack of a large effect on the maximum rate of catalysis as seen in Fig. 4 implied that the observed increase in activity of PDE5 was not due to an increase in enzyme stability as the result of phosphorylation, or to artifacts such as the presence of PDE activity in the kinase preparations. Even though phosphorylation of PDE5 exhibited a greater increase in catalytic rate when concentrations of cGMP used as substrate were below the Km value, this does not exclude an effect of phosphorylation on Vmax as the enzyme could display complicated kinetics because of cGMP interaction with multiple sites. Activation of other PDEs (PDE3, PDE4) by phosphorylation results from an increase in Vmax[22,23]. Whether the enhancement in catalytic activity of PDE5 is due to a direct effect of the phosphate introduced into the enzyme or indirectly to a phosphate-induced increase in cGMP binding at the allosteric sites (see below), which then affects the catalytic site, could not be ascertained. Activation of PDE2 by cGMP binding at the allosteric sites is effected by an increase in affinity for substrate [24].

    Details are in the caption following the image

    Effect of cGMP concentration in the PDE assay on activity of unphosphorylated and phosphorylated PDE5. The experiment was performed as described in Fig. 1 using the COMPLETE phosphorylation mixture, including the concentrations of enzymes and other factors, except for the absence or presence of PKG. The concentration of [3H]cGMP used in the PDE assay was varied as indicated. Number of determinations and statistical analyses are given in the text.

    Because most characterized PDE inhibitors are believed to act by interacting with the catalytic site of the various PDEs [1], the effect of the potent and PDE5-selective inhibitor, sildenafil, on phosphorylated (presence of PKG) and unphosphorylated (absence of PKG) recombinant PDE5 was examined. As seen in Fig. 5, there was no detectable difference in the sildenafil inhibitory potency on these two forms of PDE5. Using the same conditions as those in Fig. 5, the sensitivity to zaprinast was also not altered by phosphorylation (see inset). This was also the case for IBMX, which had a median inhibitory concentration of 2.5 μm in the absence or presence of phosphorylation (not shown). Thus, the molecular alteration in the catalytic site of PDE5 produced by phosphorylation does not improve interactions with PDE inhibitors of high, medium, or low potency. This may be explained by the fact that cGMP does not perform precisely the same interactions as PDE inhibitors with the catalytic site of PDE5 [19,25]. It should be emphasized that PDE inhibitors were tested here only at a cGMP substrate concentration (0.4 μm) found to be near-physiological in smooth muscle [21]. It could not be ruled out that phosphorylation could affect PDE inhibitor sensitivity using lower substrate concentrations or other conditions.

    Details are in the caption following the image

    Effect of sildenafil or zaprinast on PDE activity of unphosphorylated and phosphorylated PDE5. The experiment was performed as described in Fig. 1 using the COMPLETE phosphorylation mixture, including the concentrations of enzymes and other factors, except for the absence or presence of PKG. PDE activity was assayed using 0.4 μm cGMP as substrate. Results were typical of those obtained in three separate experiments, where each data point was performed in duplicate. The mean ± SD increase in PDE activity produced by PKG in these experiments was 1.66 ± 0.07-fold. Inset, effect of zaprinast using the same conditions.

    When we first reported phosphorylation of native PDE5 by PKG, we stated in the text, without showing any data, that significant activation was not observed [11], and it was speculated that factor(s) required to observe activation could have been missing. However, the finding of activation using recombinant PDE5 suggests that a specific non-covalent factor is not required. There are many other possible reasons, some quite trivial, for not observing an effect in the earlier study. The present comprehensive investigation clearly establishes that phosphorylation causes PDE5 activation.

    The results of the present study are different in some respects from those of Burns et al. [26], who reported that incubation of PDE5 with PKA produced activation of PDE catalytic activity. First, these workers utilized partially purified guinea pig lung PDE5 as the source of the enzyme and a different source of purified PKA catalytic subunit as catalyst for phosphorylation and activation in vitro. Although we mainly used PKG to phosphorylate and activate recombinant bovine PDE5 in the present study, PKA catalytic subunit was also effective. Another distinguishing feature was that activation in our hands required the presence of a PDE inhibitor (IBMX) and cGMP. Lastly, we showed here that phosphorylation of PDE5 did not significantly alter its sensitivity to PDE inhibitors (sildenafil, zaprinast or IBMX), while sensitivity of the guinea pig enzyme to a PDE inhibitor (zaprinast) was markedly decreased by phosphorylation. These notable differences in results from the two laboratories suggest that different processes were responsible for the effects even though activation of PDE5 was observed by both laboratories. It does not seem likely that different residues were phosphorylated in the two studies because Ser92 is phosphorylated by both PKG and PKA [11,12]. The use of recombinant PDE5 expressed in Sf9 cells in the present study suggests that specific low-molecular-mass peptides are not required for the activation observed. Whether or not the disparate results are due to the use of different enzyme sources or experimental conditions should be considered.

    Effect of phosphorylation on cGMP-binding activity of PDE5

    The cGMP-binding activity of PDE5 was determined using both phosphorylated and unphosphorylated enzymes, which were prepared as described in Fig. 1 with two exceptions. PKA (catalytic subunit) was used instead of PKG as catalyst for phosphorylation because the cGMP-binding activity of PKG would interfere in measurements of cGMP-binding activity of PDE5. Sephadex G-25 chromatography was used instead of dilution after preincubation to minimize carry-over effects of preincubation reaction components into subsequent assays. This procedure maintained an adequate PDE5 concentration for the cGMP-binding assay. As seen in Fig. 6, cGMP-binding activity of PDE5 was increased by phosphorylation, and the effect was particularly evident at lower concentrations of cGMP in the binding reaction. From Hill plots, the cGMP concentration required for half-maximum binding was calculated to be 0.13 ± 0.02 μm (n = 5) in the absence of phosphorylation and 0.03 ± 0.01 μm (n = 5) in the presence of phosphorylation. We previously reported this value to be 0.2 μm for native unphosphorylated PDE5 [10]. Hill constants (0.65 ± 0.11 for phosphorylated PDE5 and 0.67 ± 0.09 for unphosphorylated PDE5) were not significantly different. Low values for these constants suggested either the existence of negative co-operativity or the presence of multiple-affinity binding sites that are not positively co-operative for each enzyme.

    Details are in the caption following the image

    Effect of preincubating PDE5 with a phosphorylation reaction mixture on cGMP-binding activity. The preincubation was performed as described for the COMPLETE system in Fig. 1, including use of the same concentrations of factors, except that 3 μm PKA (catalytic subunit) instead of PKG was used as catalyst for phosphorylation, and Sephadex G-25 chromatography (Experimental procedures) instead of dilution was employed to remove potentially interfering factors in the preincubation mixture from phosphorylated or unphosphorylated PDE5 before the cGMP-binding assay. The peak protein fraction was used for the experiment in each case. After 30-fold dilution, PDE assays were carried out to ensure that > 1.5-fold increase in enzyme activity was produced by phosphorylation. The binding assay was performed as described in Experimental procedures. Results are normalized for protein (mainly BSA) in the fraction used and are typical of those in nine experiments performed. PHOSPHORYLATED and UNPHOSPHORYLATED represent the presence and absence of PKA.


    For efficient regulation of cyclic nucleotide pathways to occur, increases in levels of cAMP or cGMP cannot be excessive. Maximum stimulation of most of these pathways is brought about by only a threefold to fourfold elevation in the respective cyclic nucleotide [1]. This phenomenon is probably explained by the following. Because of the relatively high affinities of cAMP and cGMP for the PKA and PKG, each nucleotide is almost entirely in a protein kinase-bound form in cells. This means that regulation of these kinases is virtually stoichiometric with respect to intracellular cyclic nucleotide concentration. Even without addition of cyclic nucleotide-elevating agents to most isolated tissues, about 25–35% of cyclic nucleotide-binding sites of either kinase should be filled [27], which would require a threefold to fourfold elevation of cyclic nucleotide concentration to fill the remaining sites of the kinase. Higher elevations would be supersaturating for PKA or PKG and would challenge cellular processes that inactivate or counter the cyclic nucleotide signaling pathways. Although the above explanation is consistent with results so far, it is also possible that other processes, such as saturation of PKA or PKG substrate(s)-phosphorylation sites, could limit the maximum effect of stimulation of cyclic nucleotide pathways.

    In certain instances, short-term negative feedback processes have evolved to prevent excessive accumulation of cyclic nucleotides and allow for efficient termination of cyclic nucleotide pathways. For the cAMP pathway, it has been shown conclusively that such feedback processes exist, and these processes often involve phosphorylation and stimulation of the catalytic activities of cAMP PDEs [1,4,28,29]. The presence of this mode of regulation was first suggested by the finding that cAMP-elevating agents cause activation of PDE3 in adipocytes [30]. Direct evidence that this activation represents a negative-feedback process came from the finding that treatment of these cells with cAMP analogs that activate PKA decreases the cAMP level [31]. Purified PDE3 is phosphorylated and activated by PKA [4], which probably represents the underlying molecular mechanism for this negative-feedback process. PDE3 also hydrolyzes cGMP in vitro, but whether or not PDE3 is phosphorylated and activated by PKG has not been demonstrated. Phosphorylation and activation of PDE4 by PKA is also implicated in negative feedback regulation of cAMP pathways [29], and the induction of this PDE, as well as other PDEs, after long-term cAMP elevation could serve a similar purpose [32,33]. The maximum activation of either PDE3 or PDE4 in vitro by PKA-catalyzed phosphorylation is usually 1.5–2-fold [29,34], which is similar to the degree of activation of PDE5 by phosphorylation reported here. These small effects may be sufficient in cells considering the narrow physiological window of cyclic nucleotide levels.

    It is proposed that cGMP pathways as well as cAMP pathways are controlled by negative-feedback processes. In intact cells, either of these nucleotides usually declines after transient elevation by an extracellular stimulant even though the stimulant is not removed [14,28]. If phosphorylation and activation of PDE5 by PKG exemplifies such a process in cGMP pathways, then this process resembles those involving PDE3 or PDE4 in cAMP pathways. PKA directly phosphorylates PDE3 or PDE4 to activate these enzymes. The activation path for PDE5 is apparently more convoluted. We have presented evidence that cGMP or its analog, including PDE inhibitors, first occupies the catalytic site of PDE5, which then facilitates binding of cGMP to the allosteric sites [11,35]. We have also shown that the latter event causes exposure of a PDE5 phosphorylation site (Ser92) to PKG, and cGMP concomitantly activates PKG [11,36]. According to these lines of evidence and the results presented here, the order of cellular events would be cGMP elevation, occupation of the catalytic site, occupation of the allosteric sites, enzyme phosphorylation and activation, and lowering of cGMP levels.

    Athough we do not yet know the physiological consequence of stimulation of cGMP binding to PDE5 by phosphorylation, several possibilities emerge. Although phosphorylation of PDE5 could directly increase catalytic activity, enhanced cGMP binding to the allosteric sites could also directly increase catalytic activity; therefore, these two effects could act in concert to cause activation. That allosteric cGMP binding stimulates catalysis might be predicted from the principle of reciprocity [37], as interaction of cGMP at the catalytic site stimulates cGMP binding at the allosteric sites [10,35]. There is a precedent for stimulation of a PDE by allosteric cGMP binding. Activation of the homologous enzyme, PDE2, is effected by cGMP binding to its allosteric sites, although this occurs in the absence of phosphorylation [38]. Increased cGMP binding could sustain phosphorylation, thereby providing, in a feed-forward manner, an enhanced effect of phosphorylation on catalytic activity, assuming that phosphorylation directly stimulates catalytic activity. For example, increased cGMP binding could render PDE5 more resistant to phosphoprotein phosphatase action. The net result of either of the aforementioned possibilities would be that enhancement of cGMP binding to PDE5 by phosphorylation stimulates catalysis.

    Feedback regulation remains to be studied in intact cells, but this is likely to be very difficult because: (a) two enzymes (PDE5 and PKG) of the pathway require cGMP for modulation, confounding the use of cGMP analogs as discussed below; (b) reliable and specific PKG inhibitors for use in intact cells are lacking [39]; (c) PKA as well as PKG may catalyze phosphorylation of PDE5 in intact cells. It should be emphasized that the overall mechanism of PDE5 activation would provide a high degree of cGMP specificity if this nucleotide were elevated in cells. Even though it is possible that cAMP could lower cGMP levels through PKA activation or PKG cross-activation, this would require elevation of both cAMP and cGMP because the latter is required to expose the phosphorylation site of PDE5. An approach that utilized cAMP analogs to activate PKA was successful in demonstrating negative feedback regulation of the cAMP pathway, i.e. cellular cAMP levels were lowered by these treatments [28]. However, the putative cGMP feedback mechanism discussed here involves cGMP interactions at several sites: PDE5 catalytic site, PDE5 allosteric sites, and PKG allosteric sites. This makes selection of cGMP analogs that might be useful for this type of study difficult. Results obtained using a different approach with rat aorta vascular smooth muscle cells suggest that the PDE5 activation system is operational [14]. Treatment of these cells with atrial natriuretic factor to elevate cGMP levels increased incorporation of phosphate into immunoprecipitated PDE5, and this effect was associated with an increase in PDE activity in the immunoprecipitate. Bakre & Visweswariah [40] have reported that elevation of cGMP in T84 cells by heat-stable enterotoxin treatment causes an increase in PDE5 activity in extracts of these cells.


    We are grateful to Steve Ballard and Pfizer Central Research, Sandwich, UK, for helpful advice and for the gift of sildenafil. We thank Eric Howard for the amino-acid analyses. This work was supported by National Institutes of Health (DK40029), American Heart Association, American Heart Association Southeast Affiliate, and E. Bronson Ingram Cancer Center at Vanderbilt University.


  • Enzymes: cyclic nucleotide phosphodiesterase (EC; cAMP-dependent protein kinase (EC; cGMP-dependent protein kinase (EC