New Insights from the Structure-Function Analysis of the Catalytic Region of Human Platelet Phosphodiesterase 3A

Human phosphodiesterase 3A (PDE3A) degrades cAMP, the major inhibitor of platelet function, thus potentiating platelet function. Of the 11 human PDEs, only PDE3A and 3B have 44-amino acid inserts in the catalytic domain. Their function is not clear. Incubating Sp-adenosine-3′,5′-cyclic-S-(4-bromo-2,3-di-oxobutyl) monophosphorothioate (Sp-cAMPS-BDB) with PDE3A irreversibly inactivates the enzyme. High pressure liquid chromatography (HPLC) analysis of a tryptic digest yielded an octapeptide within the insert of PDE3A ((K)T806YNVTDDK813), suggesting that a substrate-binding site exists within the insert. Because Sp-cAMPS-BDB reacts with nucleophilic residues, mutants Y807A, D811A, and D812A were produced. Sp-cAMPS-BDB inactivates D811A and D812A but not Y807A. A docking model showed that Tyr807 is 3.3 angstroms from the reactive carbon, whereas Asp811 and Asp812 are >15 angstroms away from Sp-cAMPS-BDB. Y807A has an altered Km but no change in kcat. Activity of wild type but not Y807A is inhibited by an anti-insert antibody. These data suggest that Tyr807 is modified by Sp-cAMPS-BDB and involved in substrate binding. Because the homologous amino acid in PDE3B is Cys792, we prepared the mutant Y807C and found that its Km and kcat were similar to the wild type. Moreover, Sp-cAMPS-BDB irreversibly inactivates Y807C with similar kinetics to wild type, suggesting that the tyrosine may, like the cysteine, serve as a H donor. Kinetic analyses of nine additional insert mutants reveal that H782A, T810A, Y814A, and C816S exhibit an altered kcat but not Km, indicating that catalysis is modulated. We document a new functional role for the insert in which substrate binding may produce a conformational change. This change would allow the substrate to bind to Tyr807 and other amino acids in the insert to interact with residues important for catalysis in the active site cleft.

The anti-platelet drugs aspirin and clopidogrel have proven efficacy in secondary prevention of stroke, myocardial infarction, and peripheral vascular reocclusion (1,2). Aspirin inhibits cyclooxygenase, thereby decreasing synthesis of thromboxane A2. Clopidogrel, a P2Y12 antagonist, blocks the ability of ADP to inhibit stimulated adenylate cyclase. However, despite prophylaxis with these anti-platelet drugs, reocclusion of coronary arteries occurs in 20 -30% of patients after thrombolytic therapy or angioplasty probably because of the inability of these drugs to inhibit thrombin-induced platelet activation (3,4). At low concentrations of thrombin, platelet aggregation depends in part on ADP and thromboxane A2, which are released by platelets and exert autocrine-mediated enhancement. At high concentrations of thrombin, platelets are aggregated and activated by pathways independent of both ADP and thromboxane A2. In contrast, elevation of intracellular cAMP produces potent inhibition of all pathways of platelet activation including increase in intracellular Ca 2ϩ , shape change, aggregation, secretion, and the effects of phospholipases A 2 and C, as well as their responses of platelets to thrombin.
Cyclic nucleotide PDE3A 4 is the most abundant cAMP PDE in platelets. PDE3A hydrolyzes cAMP resulting in lowering the intracellular cAMP levels, which in turn potentiates platelet activation. Drugs that inhibit PDE3A raise cAMP levels in platelets, thereby increasing the phosphorylation of proteins by cAMP-and cGMP-dependent protein kinases (5). Currently two PDE3A competitive inhibitors cilostazol and milrinone have respectively been used for treating patients with intermittent claudication and acute congestive heart failure (6,7). Unfortunately cilostazol is contraindicated in patients with congestive heart failure, and milrinone is associated with undesirable cardiac arrhythmias. Examination of the inhibitory mechanism of PDE3A is important to exploit other ways of inhibiting this enzyme to minimize side effects.
The available PDE family crystal structures known to date are those of the catalytic domains cAMP-PDE (PDE4B2B and PDE4D) (8,9), cGMP-PDE (PDE5A and PDE9A) (10,11), and dual cAMP/cGMP-PDE (PDE1B and PDE3B) (12,13). The overall crystal structures of the catalytic domains of these PDEs contain a compact structure consisting of 16 ␣-helices. Each PDE has three subdomains with a deep hydrophobic pocket at the interface and two conserved metal-binding sites within that pocket. The hydrogen bond network of the neighboring residues His 948 and Trp 1072 in PDE3B (His 956 and Trp 1085 in PDE3A) serves to orient the absolutely conserved residue Gln 988 (Gln 1001 in PDE3A) to accept or donate hydrogen bonds to the purine ring, thereby determining the nucleotide recognition specificity of the enzyme (13). In PDE3B, residues Phe 991 (Phe 1004 in PDE3A) and Ile 955 (Ile 967 in PDE3A) on each side of the purine ring and Tyr 960 and Pro 941 (Try 973 and Pro 954 in PDE3A) form the hydrophobic clamp (13). Residues His 741 , His 821 , Asp 822 , and Asp 937 (His 756 , His 836 , Asp 837 , and Asp 950 in PDE3A) and one water molecule in PDE3B are involved in the first metal Mg 2ϩ binding. The second Mg 2ϩ forms hydrogen bonds with Asp 822 and five water molecules (13). Water molecules coordinated to the metal ions may act as the nucleophile in the hydrolysis reaction to mediate catalysis.
The presence of a 44-amino acid insert within the catalytic domain is a unique feature of the PDE3 gene family. In the PDE3B crystal structure, the 44-amino acid insertion (Pro 758 -Cys 801 ; Fig. 1) is located between helices 6 and 7 (13). This insertion lacks any clear secondary structure organization, and residues 767-781 are not visible in the electron density maps. This insert in human PDE3A is comprised of amino acid residues 773-816 (Fig. 1). Within the insert there is 38.6% identity between PDE3A and PDE3B including conserved triplets at the N terminus, C terminus, and the middle of the insert. Tang et al. (14) showed that removal of this insert from PDE3A resulted in the complete loss of activity. Double mutants, P773A/G774A and Y814A/G815A, from each of the N and C termini of the PDE3A insert, which are ␤ turns, display markedly reduced activity (14). However, knowledge of the role of the 44-amino acid insert of PDE3A in the regulation of enzyme activity or interactions with substrate and/or inhibitor is incomplete.
Previously, we have synthesized a nonhydrolyzable, reactive substrate analog, Sp-cAMPS-BDB, which irreversibly inactivates PDE3A in a time-dependent fashion with K I ϭ 10.1 Ϯ 1.7 M and k max ϭ 0.0116 Ϯ 0.0004 min Ϫ1 (15) . We have demonstrated that Sp-cAMPS-BDB targets both the cAMP-and cGMP-binding sites but favors the cAMP site. The protection studies indicate effectiveness of protectants in decreasing rate of inactivation by Sp-cAMPS-BDB is: Sp-cAMPS (K d ϭ 24 M) Ͼ Rp-cGMPS (1360), Sp-cGMPS (1460) Ͼ GMP (4250), AMP (10600), Rp-cAMPS (22170 M). Sp-cAMPS-BDB has proven to be an effective active site-directed affinity label for PDE3A.
In this paper, we describe specific incorporation of PDE3A by a reactive substrate analog, Sp-cAMPS-BDB, isolation of a peptide in the unique insert of PDE3A, and construction of mutant enzymes that identify the amino acid targeted by Sp-cAMPS-BDB. In addition, the role of the insert was further explored by kinetic analyses of nine additional insert mutants. The results define a new functional mechanism by which binding of cAMP to the flexible loop of platelet PDE3A may induce a local conformational change that allows interaction with catalytic residues.
Measurement of the Incorporation of Sp-cAMPS-BDB into PDE3A-PDE3A was incubated with 100 M Sp-cAMPS-BDB in a 50 mM Hepes buffer at pH 7.3 containing 20 mM MES, 10 mM MgCl 2 , and 0.5 M NaCl. At various times of incubation (0, 20, 30, 40, 60, and 80 min, respectively), the aliquots were removed, and the residual enzyme activity of PDE3A was determined to correlate with the incorporation (see "Enzyme Activity Assay"). At each time interval, 100 mM [ 3 H]NaBH 4 (dissolved in 20 mM NaOH) was added consecutively to reach a final concentration of 2 mM at 4°C for a total of 1.5 h.
[ 3 H]NaBH 4 reduces the two oxygens of the diketo group from Sp-cAMPS-BDB to two [ 3 H] hydroxyl groups. The excess [ 3 H]NaBH 4 and the free Sp-cAMPS-BDB were removed by four consecutive centrifugations using Microcon centrifugal devices (Millipore, Billerica, MA) at 14,000 ϫ g for 20 min. Aliquots were removed from the retentate to measure the protein concentration using the Coomassie Plus protein assay. The amount of Sp-cAMPS-BDB incorporated into PDE3A from reduction of the affinity labeled enzyme by [ 3 H]NaBH 4 was calculated by measuring the radioactive ( 3 H) content by using a Beckman Coulter liquid scintillation analyzer (model LS6500; Fullerton, CA). Control samples were tested using a similar procedure with the pretreatment of cold NaBH 4 with Sp-cAMPS-BDB prior to the addition of enzyme.
Trypsin Digestion of the Sp-cAMPS-BDB-modified Enzyme-PDE3A (0.8 mg) was incubated with 100 M Sp-cAMPS-BDB at 25°C for 3 h (ϳ10% residual activity remained). The incubated mixture was treated twice with 100 mM [ 3 H]NaBH 4 for a total of 1.5 h (final concentration, 2 mM), followed by a carboxylation of free SH groups with 10 mM N-ethylmaleimide for 10 min. After removal of the excess reagents by centrifugation using Microcon centrifugal devices, the modified enzyme was digested at 37°C by 2 consecutive additions of 5% (w/w) tosylphenylalanyl chloromethyl ketone-treated bovine pancreatic trypsin for a total of 2 h.
Purification and Determination of the Sequence of Modified Peptide-The radioactive tryptic digest was lyophilized, redissolved in 250 l of 0.1% trifluoroacetic acid, and applied to an HPLC system using a reverse phase Vydac (Hesperia, CA) C 18 column (0.46 ϫ 25 cm). Separation was conducted at the elution rate of 1 ml/min using solvent A (0.1% trifluoroacetic acid in water) for the first 10 min, followed by a linear gradient from solvent A to 45% solvent B (0.1% trifluoroacetic acid in acetonitrile) for 220 min, a linear gradient from 45% solvent B to 100% Solvent B for 20 min, and solvent B for 10 min, successively. The eluent was monitored at 220 nm. Fractions of 1 ml were collected, from which 400 l was counted for radioactivity. The amino acid sequence of isolated radioactive peptides was determined using an automated gas phase peptide sequence analyzer from Applied Biosystems (model 470A; Foster City, CA) equipped with an on-line phenylthiohydantoin analyzer (model 120) and computer (model 900A). The sequencing results were used to identify the location of the modified peptide in the active site of the catalytic region of PDE3A. This process was repeated twice with identical results.
Construction and Purification of PDE3A Mutants-A deletion mutant of PDE3A cDNA coding for the amino acid residues 665-1141 (16) was subcloned into a pENTER-TOPO vector (Invitrogen) to produce two sites for linear recombination. PDE3A insert mutants H782A, H796A, H798A, S804A, K805A, Y807A, Y807C, T810A, D811A, D812A, Y814A, G815A, and C816S were constructed using a QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). All of the mutants were confirmed by nucleotide sequence analysis (Sidney Kimmel Nucleic Acid Facility, Thomas Jefferson University, Philadelphia, PA). Recombinant mutant baculoviruses were produced by linear combination using BaculoDirect Transfection kit (Invitrogen). Expression of the catalytic region (residues 665-1141) of PDE3A wild type and mutant enzymes using a baculovirus/insect cell Sf9 system and protein purification using a ProBond Nickel resin column has been previously described (17,18).
Protein Concentration Determination-Protein concentration of the purified enzymes and purified anti-insert antibody were determined using Coomassie Plus protein assay reagent using bovine serum albumin as standard. The absorbance at 595 nm was measured using a Bio-Tek automatic microplate reader equipped with KC4 module for data analysis (Bio-Tek Instruments, Inc., Winooski, VT).
Western Blot Analysis-The PDE3A wild type and mutants were separated on 10% Bis-Tris gel electrophoresis purchased from Invitrogen. The proteins were transferred to a polyvinylidene difluoride membrane using the Xcell II module at a constant voltage of 30 volts for 1 h at room temperature for Western blotting. The membranes were processed using the Chromogenic WesternBreeze system and probed with anti-in-sert PDE3A antibody (see effects of anti-insert antibody) to detect the presence of PDE3A.
Enzyme Activity Assay-PDE3A activity was measured by the amount of cAMP hydrolyzed as previously described (19). Enzyme was added to a buffer containing 50 mM Tris-HCl, pH 7.8, 10 mM MgCl 2 , and 0.8 M [ 3 H]cAMP. Reaction mixtures both with and without enzymes were incubated at 30°C for 15 min. Catalysis was terminated by serial addition of 0.2 M of ZnSO 4 and 0.2 M Ba(OH) 2 , which precipitates AMP but not cAMP. Samples were vortexed and centrifuged at 10,000 ϫ g for 5 min. The BaSO 4 pellets containing the [ 3 H]5Ј-AMP precipitant were discarded. Aliquots of supernatants containing unreacted [ 3 H]cAMP were removed and counted in a Beckman Coulter liquid scintillation analyzer. Enzyme activity was measured by comparing the amount of cAMP hydrolyzed in PDE3A containing samples to no enzyme controls. These data were then used to calculate enzyme specific activity in nmol of cAMP hydrolyzed per mg of protein per min.
Kinetic Constants Determination-The rates (nmol/s) of cAMP hydrolysis for the PDE3A wild type and mutant enzymes were determined using various concentrations of substrate cAMP from 0.02 to 14 M. The values of K m and V max for each of the enzymes were determined by Michaelis-Menten equation as calculated by Enzyme Kinetics Module 1.1 software (Systat Software, Point Richmond, CA). The k cat (s Ϫ1 ) was obtained by dividing V max (nmol/s) by the molar enzyme concentration (nmol). Effect of Anti-insert Antibody on Enzyme Activity-A rabbit polyclonal antibody against the synthetic peptide 802 VFSK-TYNVTDDKYGC 816 , the C-terminal 15 amino acids of the PDE3A insert (Fig. 1), which also contain the octapeptide, was prepared by Sigma Genosys and designated as an anti-insert antibody. PDE3A, and mutants Y807A and Y807C were incubated respectively with various concentrations of the anti-insert antibody to a enzyme to antibody ratio of 1.3, 2.0, or 4.0 for 1 h at 37°C. After incubation, enzyme activity was determined according to the "Enzyme Activity Assay" procedure. The activity of PDE3A wild type, Y807A, and Y807C without antibody was set as 100% activity. The preimmune IgG was used as a control to compare the activity of wild type, Y807A, and Y807C. All of the experiments were performed in triplicate.

Reaction of Sp-cAMPS-BDB with Mutant Enzymes-Purified
Molecular Modeling-A homology model of PDE3A based on the crystal structure of PDE4B2B has been published (8). However, the model did not contain the additional 44-amino acid insert found in PDE3A. We have now refined the PDE3A model using the recently published PDE3B structures (13) that contain the 44-amino acid insert unique to PDE3. Sybyl 6.91 FlexX docking module (Tripos) was then used to dock the affinity label Sp-cAMPS-BDB to PDE3A. Because mutant Y807A affected the K m , Tyr 807 was included in the defined cAMP-binding pocket to construct the model. Residues involved in cAMP binding (17,18) were used as a defined cAMP binding pocket (Tyr 807 , Asn 845 , Glu 866 , Glu 971 , Phe 972 , and Phe 1004 ). This docking model was utilized to illustrate and further evaluate the kinetic results obtained from the mutants of insert amino acids of PDE3A.

RESULTS
Incorporation of Sp-cAMPS-BDB into PDE3A Is Time-dependent-To quantify the amount of the affinity label Sp-cAMPS-BDB incorporated into PDE3A, the enzyme (0.38 mg/ml) was incubated with 100 M Sp-cAMPS-BDB at pH 7.3, as described under "Experimental Procedures." Fig. 2 (left  panel) shows that the incorporation of PDE3A by Sp-cAMPS-BDB is linear as a function of time. The addition of [ 3 H]NaBH 4 to an incubation mixture of enzyme and Sp-cAMPS-BDB stops the reaction by reducing the diketo group of Sp-cAMPS-BDB to a [ 3 H]diol group. Fig. 2 (right panel) shows that the residual enzymatic activity is inversely proportional to the incorporation. At 80 min, 0.86 mol of Sp-cAMPS-BDB was incorporated for each mol of enzyme which corresponded 19% of residual enzymatic activity or 81% inactivation. Thus, 1.08 mol of Sp-cAMPS-BDB was required to inactivate each mole of enzyme indicating a stoichiometry close to 1.0 of the affinity label and the enzyme.
The Isolated Sp-cAMPS-BDB-modified Peptide in PDE3A Is Located in the Unique 44-Amino Acid Insert-PDE3A (11 nmol) was incubated with 100 M Sp-cAMPS-BDB for 3 h and treated with [ 3 H]NaBH 4 as described under "Experimental Procedures." The modified enzyme was digested by trypsin for 2 h as described under "Experimental Procedures." Fig. 3 (solid line) shows that on the reverse phase HPLC separation of the tryptic digest, most of the peptides elute between 0 and 160 min (0 and 30% solvent B). Two major radioactive peaks were observed as shown in Fig. 3 (dashed line, labeled I and II).
The amino acid sequence of the purified peptides (Fig. 3, peaks I and II) was determined by Edman degradation using an automated gas phase sequencer. Peak I contains small peptides (data not shown). The amino acid sequence of the peptide from peak II exhibits a single octapeptide, assigned as 806 TYNVT-DDK 813 within the unique 44-amino acid insert of PDE3A (Fig.  1). This peptide results from enzyme cleavage after Lys 805 and Lys 813 , consistent with the specificity of the trypsin recognition sites. The yield of each phenylthiohydantoin-derivative was recorded and ranged from 40 to 20 pmol (data not shown). As expected, the yield decreases as the cycle number increases. Peptide 806 -813 is located C-terminal of the first metal-binding motif, 752  To evaluate the effect of mutations on the reaction with Sp-cAMPS-BDB, the mutant enzymes were incubated with the affinity label, and their activity was tested as a function of time. Fig. 4 (A-D) shows the results of reaction of wild type and mutant enzymes, Y807A,  D811A, and D812A, respectively, with Sp-cAMPS-BDB. Sp-cAMPS-BDB irreversibly inactivates both mutants D811A and D812A exhibiting saturation kinetics (Fig. 4, H-I). The k max values for D811A and D812A are 0.005 Ϯ 0.0002 and 0.003 Ϯ 0.0001 min Ϫ1 , and the K I values are 29.9 Ϯ 2.9 and 24.9 Ϯ 2.5 M, respectively. The K I values of both D811A and D812A is 2.5-3-fold larger than that of wild type (K I ϭ 10.1 Ϯ 1.7 M; Fig.  4F). The k max values of D811A and D812A are one-half and one-third, respectively, that of wild type (k max ϭ 0.0116 Ϯ 0.0004 min Ϫ1 ; Fig. 4F). These relatively minor changes in kinetics indicate that residues Asp 811 and Asp 812 are not the modified amino acid of the wild type enzyme that reacts with Sp-cAMPS-BDB. In contrast, Y807A is not inactivated by Sp-cAMPS-BDB (50 -400 M; Fig. 4, B and G), identifying Tyr 807 as the amino acid modified by Sp-cAMPS-BDB.

Docking Model of Sp-cAMPS-BDB into PDE3A Supports Tyr 807 as the Amino Acid Modified by the Affinity Label-
The catalytic domain of PDE3A, including the unique 44-amino acid "insert," was modeled using Sybyl Composer based on the crystal structure of PDE3B (1SO2 and 1SOJ) (13). FlexX docking module (Sybyl 6.91) was then used to dock Sp-cAMPS-BDB into the PDE3A model with a defined active site pocket of Tyr 807 , Asn 845 , Glu 866 , Glu 971 , Phe 972 , and Phe 1004 . Molecular modeling of the "insert" region, based on the crystal structure of PDE3B, suggests that this region is a flexible loop (Fig. 5). Based on the docking model of Sp-cAMPS-BDB into the PDE3A model, Tyr 807 (green) is most likely to be the amino acid modified by Sp-cAMPS-BDB, because the reactive carbon C-9 of the affinity label is 3.3 Å from the hydroxyl oxygen of Tyr 807 , whereas the carboxyl oxygens of Asp 811 and Asp 812 are more than 15 Å away from the reactive carbon of the affinity label (Fig. 5). These results further support the inactivation data that Tyr 807 is the amino acid modified by Sp-cAMPS-BDB.
Residue Tyr 807 in PDE3A Is Involved in Substrate cAMP Binding- Table 1 shows the kinetic characteristics of the mutant enzymes D811A and D812A. The K m values for both D811A and D812A are similar to that of the wild type. The k cat values of D811A and D812A are similar to that of wild type and suggest that single alanine mutation of the residues, Asp 811 and Asp 812 does not affect the enzyme catalytic activity. We further studied the mutant Y807A. The K m of Y807A is 6.79 Ϯ 0.83 M, which is 30-fold greater than that of the wild type PDE3A. This indicates that Tyr 807 is involved in cAMP binding. The k cat value of the mutant Y807A was similar to the wild type (Table 1).
Mutant Y807C Mimics the Wild Type PDE3A-The amino acid corresponding to Tyr 807 in the second member of the PDE3 gene family PDE3B is Cys 792 (20). We hypothesized that the cysteine 792 might serve as a hydrogen donor similar to tyrosine 807. Therefore, we produced the mutant Y807C in PDE3A. Both the K m (0.16 Ϯ 0.01 M) and k cat (105 s Ϫ1 ) of PDE3A mutant Y807C were similar to the wild type. To test the hypothesis that the thiol group mimics the phenolic group, we performed the inactivation studies of the mutant Y807C using Sp-cAMPS-BDB. This mutant, Y807C, is irreversibly inactivated by the affinity label Sp-cAMPS-BDB in a time-dependent manner exhibiting a K I of 18.0 Ϯ 2.7 M and k max of 0.004 Ϯ 0.0002 min Ϫ1 (Fig. 4, E and J). These values of Y807C are very close to the wild type.
Mutant Y807A Is Not Inhibited by the Anti-insert Antibody-The anti-insert antibody was raised against the 15 amino acids ( 802 VFSKTYNVTDDKYGC 816 ) located at the C-terminal end of the insert, within which the octapeptide at positions 806 -813 is identified by the affinity label. When the anti-insert antibody was added prior to the PDE3A activity assay, the wild type decreased in activity to 84, 55, and 32% proportional to antibody concentration (Fig. 6A). Under the same conditions, PDE3B, the other gene product of the PDE3 gene family does not decrease in enzymatic activity when the anti-insert antibody is added (data not shown), indicating the specificity of the neutralization by antibody to PDE3A. Two mutants were made at residue 807 position to compare with the wild type tyrosine: Y807A eliminates the phenolic group, and Y807C substitutes a thiol group mimicking the corresponding residue Cys 792 of PDE3B. When the anti-insert antibody was preincubated with the enzyme prior to the activity assay, the mutant Y807A did not decrease in activity at any of the antibody concentrations (Fig. 6B). Similarly, when the Y807C was preincubated with varying concentrations of anti-insert antibody, the enzyme activity did not decrease as a function of antibody concentrations (Fig. 6C). The failure of the anti-insert antibody to inhibit both the inactive mutant Y807A and the active mutant Y807C indicates that tyrosine residue is a critical part of the epitope of the antibody. Aromatic amino acids are frequently highly antigenic. The results indicate that the concentration of the antibody used did not block the active site.
The K m of the wild type enzyme was similar in the presence or absence of the anti-insert antibody (K m ϭ 0.203 versus 0.197 M). The k cat is decreased 2.2-fold (k cat ϭ 158 versus 70.6 s Ϫ1 , respectively), which is consistent with the decrease of 3.6-fold in enzymatic activity in the presence of 0.053 M anti-insert IgG (Fig. 6A). The preimmune IgG, which does not contain the epitope to interact with the enzyme, did not inhibit the activity of wild type, Y807A, and Y807C (data not shown). These data support the finding that residue Tyr 807 is the amino acid modified by the affinity label Sp-cAMPS-BDB and is consistent with the markedly increased K m of the mutant Y807A.
Four of the Nine Additional Insert Mutants Are Involved in Catalytic Activity-Based on the conservation of the amino acid sequence (Fig. 1) and molecular model of PDE3A (Fig. 5), eight additional amino acid residues His 782 , His 796 , His 798 , Ser 804 , Lys 805 , Thr 810 , Tyr 814 , and Gly 815 at the unique 44amino acid insert were chosen to mutate to alanine. In addition, Cys 816 was mutated to a serine, because as previously reported mutant C816A completely abrogated the enzyme activity (21). Each of the mutants exhibited a single band on SDS gel electrophoresis (data not shown). Table 1 shows that mutants H796A, H798A, S804A, K805A, and G815A have no marked changes in kinetic constants compared with wild type PDE3A. In contrast, mutants H782A, T810A, Y814A, and C816S show a significant 5-29-fold decrease in k cat but no major changes in K m as wild type PDE3A. Because these four amino acid residues are distant from those involved in catalysis in the enzyme, the insert amino acid residues His 782 , Thr 810 , Tyr 814 , and Cys 816 are important for PDE3A catalytic activity only after substrate binding.

TABLE 1 Kinetic parameters of PDE3A insert mutants
The mutants are listed in the order of their position in the sequence. The results are the means of three sets of experiments, and for the K m values the S.E. is indicated. The data shown in bold indicate statistically significant changes (Student's t test, p Ͻ 0.01) in kinetic parameters of the mutant enzyme when compares that with the wild type. The Functional Role of the Insert Is Supported by the Conformational Similarity of Mutants and Wild Type-It is possible that mutation of a single amino acid could extensively alter the protein conformation and that this change would result in the loss of enzymatic function. Therefore the immunoreactivity of all the mutants was analyzed to ensure that the conformation was similar between the wild type and mutants. Each mutant was detected as a single band in Western blot with the polyclonal anti-insert antibody (Fig. 7). In addition, the 12 mutants had insignificant changes of K m , with exception of Y807A ( Table 1). The CD spectra of both PDE3A wild type and mutant Y807A were measured to test whether the secondary structures were similar. The CD spectra of both wild type and mutant Y807A are virtually superimposable and display a negative trough at 209 nm ([] WT ϭ Ϫ5040 and [] Y870A ϭ Ϫ4890 deg cm 2 dmol Ϫ1 , respectively (data not shown). The overall similarity in the far-UV CD spectra of the mutant Y807A and wild type PDE3A indicates that the mutant Y807A maintains a native secondary structure. The Western blot results and the CD spectral data suggest that all of the mutants have overall conformation similar to that of the wild type PDE3A.

DISCUSSION
Reactive purine nucleotide analogs have been used as affinity labels to probe nucleotide binding sites (22)(23)(24). We have described the use of the cAMP affinity analog 8-[(4bromo-2,3-dioxobutyl)thio]-adenosine 3Ј,5Ј-cyclic monophosphate (8-BDB-TcAMP) in studies to identify important amino acids within the active site of PDEs. 8-BDB-TcAMP irreversibly inactivated PDE2A (25), PDE3A (26), and PDE4A (27). In the case of PDE4A, a peptide containing the residue modified by 8-BDB-TcAMP was isolated, and the amino acid sequence was identified. However, the utility of 8-BDB-TcAMP was limited because it inactivates PDEs only at millimolar concentrations, because of continuous hydrolysis to the 5Ј-AMP derivative by the enzymes under investigation. We reported the synthesis of a new nonhydrolyzable reactive cAMP derivative, Sp-cAMPS-BDB, which contains both reactive bromoketo and dioxo groups (15). The bromoketo group can form covalent bonds with the nucleophilic side chains of many amino acids including cysteine, aspartate, glutamate, histidine, tyrosine, and lysine, whereas the dioxo provides the ability to react with arginine residues.
We here demonstrate that Sp-cAMPS-BDB acts as an affinity label of PDE3A. The Sp-cAMPS-BDB is a substrate analog of cAMP with the reactive bromodioxobutyl group at the phosphorothioate ester. An octapeptide ( 806 TYNVTDDK 813 ) in PDE3A has been identified by tryptic digest, peptide isolation, and N-terminal amino acid sequencing. Three nucleophilic amino acid residues in the octapeptide were selected to produce mutants (Y807A, D811A, and D812A) to identify the target residue reacting with Sp-cAMPS-BDB. Both mutants D811A and D812A were inactivated by Sp-cAMPS-BDB, whereas Y807A is not inactivated by the affinity label. Furthermore, we showed that Tyr 807 exhibits a large change in K m and that this amino acid is close (3.3 Å) to the reactive carbon of the affinity label in a docking model based on the crystal structure of PDE3B. Tyr 807 , although it is present in the 44-amino acid insert, is functionally part of the cAMP-binding site. The other two amino acids, D811A and D812A, in the octapeptide capable of reacting with the affinity label are more than 15 Å from the reactive carbon of Sp-cAMPS-BDB and have similar k cat /K m to  wild type. Thus, Tyr 807 is the amino acid modified by the affinity label rather than Asp 811 or Asp 812 .
Because the function of the unique 44-amino acid insert in PDE3A has not been fully understood, we have produced nine additional insert mutants to explore its function based on the affinity labeling data and alignment analysis of the PDE3 gene family. Four mutants (H782A, T810A, Y814A, and C816S) exhibit a 5-21-fold decrease in k cat without a significant change in K m , and the other five mutants (H796A, H798A, S804A, K805A, and G815A) have similar k cat /K m to wild type. These results indicate that the flexible 44-amino acid insert of human platelet PDE3A predominately regulates the catalysis of the substrate cAMP.
To further explore the function of the 44-amino acid insert, we produced an antibody to a peptide containing the 15 amino acids at the C-terminal end of the insert. In the presence of the anti-insert antibody, the ability of PDE3A and mutant H798A to hydrolyze cAMP was reduced, whereas the apparent K m remains unchanged. These results support the conclusion that one of the important functions of this insert is to regulate the enzyme catalytic activity. The exception to this is Tyr 807 , which is the closest amino acid to the docked affinity label Sp-cAMPS-BDB. Not only is this amino acid close enough to bind the affinity label but also appears to be the only residue of the 12 mutants tested to affect the K m . Therefore Tyr 807 is a participant in the substrate-binding site. Because the K m is increased 30-fold compared with the wild type, it was especially important to test whether a major conformational change had occurred in Y807A. The CD spectra of both wild type and mutant Y807A are almost superimposable, indicating that the mutant Y807A maintains a native conformation (data not shown).
When we assessed Y807A with the anti-insert antibodies, we found no inhibition at any of the concentrations tested in contrast to the behavior of the wild type enzyme. These results indicate that although Tyr 807 does not affect the k cat ; it regulates the catalytic activity directly, presumably because of conformational change upon substrate binding, and directly by influencing the substrate binding. However, this is not related to the loss of activity because the active mutant Y807C is also not inhibited by the antibody. These results suggest that Tyr 807 is a critical part of the epitope of antibody, but other epitopes exist in the loops.
We speculate that Cys 792 in PDE3B might play the same role as Tyr 807 in PDE3A. The corresponding PDE3A mutant and its functional group reacts similarly to Tyr. The kinetic constants K m and k cat of Y807C were similar to that of wild type. Furthermore, Y807C is irreversibly inactivated by the affinity label Sp-cAMPS-BDB in a time-dependent manner to a similar extent when compared with the wild type. It is likely that the phenolic group of PDE3A and the thiol group of PDE3B both function as hydrogen donors in the interaction with substrate cAMP.
The 44-amino acid insert shown in the molecular model constitutes a flexible loop exposed on the surface of the enzyme (Fig. 5). Although the unique insert lies in the first conserved metal-binding motif 752 HNRIH 756 X 24 -26 E 825 in the active site cleft from the primary structure of the enzyme, the homology model based on crystalline PDE3B indicates that the flexible loop of the insert is distant from the active site cleft in the model. These kinetic analyses and molecular modeling data imply that upon substrate binding, this surface flexible insert may undergo substantial local conformational change. We hypothesize that the flexible insert flips into the active site cleft to regulate the substrate binding and catalytic activity. Further studies are underway to document any conformational change associated with substrate binding.
A precedent for local conformational change of a loop has been shown in the reaction of trypsin with ␣ 1 antitrypsin (28,29). When trypsin cleaves the reactive center loop of ␣ 1 antitrypsin, the cleaved reactive center loop undergoes a large local conformational change and zips into a groove of ␤-sheet of the molecule with the translocation of trypsin to the other pole of ␣ 1 antitrypsin.
In the case of aspartic peptidases, the variation in flap conformations observed in x-ray studies of free and inhibitorbound enzymes indicates that the flaps in the free enzyme are flexible in solution. For example, binding of pepstatin to cathepsin D induced small structural changes in the flap region that contains the ␤-hairpin structure from residues 72-87 (30). Residues 79 and 80 at the tip of the flap moved in toward the inhibitor by about 1.7 Å, and the flexibility of this ␤-bend decreases because of electrostatic interaction of His 77 of the flap with the C terminus of the inhibitor. Similar changes in conformation upon inhibitor binding have also been shown in the pepstatin-bound form of both rhizopuspepsin and penicillopepsin (31,32). The closing of the flap over the inhibitor substrate serves to remove the peptide bond of the substrate from effective contact with solvent.
In conclusion, use of the nonhydrolyzable affinity label Sp-cAMPS-BDB and structural analysis have allowed us to identify a new cAMP-binding amino acid (Tyr 807 ) in the 44-amino acid insert that forms a flexible loop unique for the PDE3 gene family. These results challenged us to produce nine additional insert mutants that defined the role of His 782 , Thr 810 , Tyr 814 , and Cys 816 as amino acids interacting with the residues involved in catalysis and/or metal binding. The identical behavior of the mutant Y807C to the wild type suggests that this tyrosine residue may be functioning as a H donor. The presence of a similar loop in PDE3B with a cysteine instead of tyrosine in PDE3A at the homologous position suggests that a similar mechanism may be involved with PDE3B substrate binding. The affinity labeling results and the kinetic data from the mutants suggest a functional role of the insert and provide a new strategy for structure-based inhibitor design to develop new specific inhibitors for PDE3A.