Trypanosome Cyclic Nucleotide Phosphodiesterase 2B Binds cAMP through Its GAF-A Domain*

of sleeping sickness in humans and livestock, expresses at least three cAMP-specific class I phosphodiesterases (PDEs), all of which are essential for survival of the parasite. These PDEs have either one or two N-terminal GAF domains, which in other proteins function as signaling domains. However, neither the functional roles nor ligands for these domains in trypanosome PDEs are known. The present study shows that TbPDE2B, which contains two tandem GAF domains, binds cAMP with high affinity through its GAF-A domain. A purified recombinant N terminus (cid:1) GAF-A domain binds cAMP with an affinity ( K i ) of (cid:1) 16 n M . It also binds cGMP but with a 15-fold lower affinity of (cid:1) 275 n M . The TbPDE2B holoenzyme has a somewhat lower affinity ( (cid:1) 55 n M ) for cAMP but a greatly lower affinity ( (cid:1) 10 (cid:2) M ) for cGMP. This suggests that both the selectivity and affinity for a ligand can be determined not only by the nature of the binding domain but also by the adjacent domains. Addi-tionally, binding of cAMP to the holoenzyme showed positive cooperativity, with a Hill coefficient value of 1.75. However, binding of cGMP to the holoenzyme did not show any cooperativity, suggesting differences in the conformational changes Homology Modeling— A homology model of the TbPDE2B GAF-A domain was constructed using Swiss-Model (26, 27) based on the crystal structure of the mmPDE2B GAF-B domain (1MC0.pdb) using a Clust- alW (biology workbench) alignment of TbPDE2B GAF-A on mmPDE2B GAF-B as a template. Images were generated using the Swiss Protein Data Bank viewer and further refined using Pov-Ray. Other Methods— Total protein concentration was determined using the Bradford assay with bovine serum albumin (Pierce) as the standard. Protein levels in Western blots (with an anti-V5 antibody) were nor- malized using ImageJ (NIH) to measure relative densities of the detected bands.

Cyclic nucleotide phosphodiesterases (PDEs) 1 regulate cAMP and cGMP signaling pathways by controlling the intracellular levels of cyclic nucleotides. In higher eukaryotes, PDEs are known to regulate a variety of processes including visual transduction, olfaction, control of metabolic activities, insulin secretion, fertility, and a host of other functions. The precise cellular profile of PDE expression is thought to influence the response of a tissue or organism to cyclic nucleotides.
In mammals, there are 11 distinct Class I PDE families (1, 2) that have conserved catalytic domains but different regulatory domains. These 11 mammalian PDE family members each have different kinetic and substrate characteristics, inhibitor profiles, amino acid sequences, and regulatory ligands (3). Other organisms such as Saccharomyces cerevisiae have a second class of phosphodiesterases, termed Class II PDEs, that have a different evolutionary origin and catalytic domains that are not homologous to mammalian PDEs (4). Five of the 11 mammalian Class I PDE family members contain regulatory segments with one or two tandem GAF domains (5). GAF domains are a very large family of small molecule-binding domains recently found to be present in nearly all organisms (5). Thus far, the only high affinity ligand identified for a PDE GAF domain has been cGMP. Binding of cGMP to the GAF domains of PDE5 (6) and PDE2 (7) is an important mechanism for regulating these enzymes. cGMP also tightly binds the photoreceptor PDE6 (8), where its precise role is being investigated (9). Recently, cAMP was found to bind and activate the GAF domain of the cyaB1 adenylyl cyclase from the cyanobacterium Anabaena genus (10). Significant insights into the mechanism of cyclic nucleotide binding to these GAF domains was provided by the recent structure determination of the murine PDE2A regulatory GAF domains in complex with cGMP (11). A follow-up study also revealed a number of critical contacts enabling nucleotide binding and subtle discrimination between cAMP and cGMP (12). Two recent reviews discuss the evolutionary conservation, properties, and possible functions of these GAF domains in cyclic nucleotide phosphodiesterases and other proteins (13,14).
In parasites of the Trypanosomatidae family, cAMP is an important mediator of cell transformation and proliferation. Intracellular levels of cAMP are known to vary during different life cycle stages of trypanosomes (15). In addition, a densitysensing mechanism that signals cell cycle arrest in trypanosomes and leads to differentiation appears to be via the cAMP pathway (16). The precise roles of PDEs in these parasites are still largely unexplored (17). However, it is known that the differentiation of Trypanosoma brucei from bloodstream forms to short, stumpy forms can be inhibited by non-selective PDE inhibitors (18), demonstrating the importance of PDEs in these organisms. Recently, three different Class I cAMP-specific PDEs, TbPDE2A, -B, and -C, were identified in T. brucei (19 -21). All three were found to be essential for proliferation of the bloodstream form of T. brucei (21). These three also contained either one or two tandem GAF domains (19 -21). TbPDE2B intriguingly did not appear to bind cGMP (19 -21), despite containing a majority of the "critical residues" defined for binding cGMP in mammalian PDE2 (11). Neither a guanylyl cy-clase nor a function for cGMP has yet been identified in trypanosomes. Considering the challenges being faced in present chemotherapy for human sleeping sickness (22) and the established success of PDE inhibitors as drugs in various human diseases, TbPDE2A, -B, and -C are attractive antitrypanosomal drug targets. This provides much incentive for understanding their regulation and the function of their GAF domains. The current study reports high affinity cAMP binding to the TbPDE2B GAF-A domain, characterizes the binding, and provides data that suggest this cAMP binding is important for regulation of the trypanosome PDE activity.
Cell Culture-Human embryonic kidney (HEK)293 or 293T cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 100 units/ml penicillin, and 100 g/ml streptomycin at 37°C in a humidified 5% CO 2 atmosphere. Sf9 cells were grown at 27°C in Grace's insect cell medium (Invitrogen) supplemented with 10% fetal bovine serum.
Expression and Purification of Bacterially Expressed Protein-The N terminus ϩ GAF-A domains were expressed in Rosetta(DE3)pLysS (Novagen) cells grown in Luria broth with 75 g/ml carbenicillin. The cells were grown at 37°C up to an A 600 of 0.4 -0.6, induced with 0.3 mM isopropyl 1-thio-␤-D-galactopyranoside, and harvested after 22 h of growth at 16°C. The cells were resuspended and lysed in lysis buffer (phosphate-buffered saline (Invitrogen) plus 10% glycerol, 2 mM phenylmethylsulfonyl fluoride, 5 mM ␤-mercaptoethanol) by microfluidization (10,000 p.s.i.) (Microfluidics, Newton, MA), and centrifuged at 16,000 ϫ g for 30 min. The supernatant was purified using a TALON metal affinity resin (Clontech) and eluted with lysis buffer containing 100 mM imidazole. The imidazole was removed using a PD10 buffer exchange column, replacing it with the original lysis buffer.
Site-directed Mutagenesis-The T317A point mutant was made using holoenzyme TbPDE2B pcDNA3.1V5his as the template. The mutation was made using the QuikChange protocol from Stratagene, with Pfu Turbo polymerase (Stratagene). All plasmids subjected to mutagenesis and subcloning were sequenced to ensure the presence of the desired mutation. Primers 5Ј-GCTACTGGGTACCGTGCAAAGAC-GATACTCTGC-3Ј and 5Ј-GCAGAGTATCGTCTTTGCACGGTAC-CCAGTAGC-3Ј were used.
Transient Transfection of HEK293T Cells-Plasmids with the TbPDE2B constructs subcloned into the pcDNA3.1-V5-his vector (Invitrogen) were purified using a Qiagen plasmid maxi kit (Qiagen). Cells plated on 100-mm plates were transiently transfected with DNA using the Lipofectamine 2000 (Invitrogen) method according to the manufacturer's protocol. Protein expression was verified by Western blot anal-ysis of cell lysates using an anti-V5 antibody (Invitrogen). Cells were harvested for experiments after 40 -48 h of growth, washed in Dulbecco's phosphate buffered saline (Invitrogen) and resuspended in homogenization buffer (Dulbecco's phosphate-buffered saline, pH 7.5, 1 mM phenylmethylsulfonyl fluoride, 3 mM EDTA, 1 mM dithiothreitol). The cells were lysed using a Virsonic 100 sonicator for 4 -5 s and centrifuged at 16,000 ϫ g for 20 min at 4°C. The supernatants were used for all further assays.

Construction of a Baculovirus Expression Vector System for TbPDE2B Expression in Sf9 Cells and Purification-Recombinant
TbPDE2B holoenzyme-expressing Baculoviruses were constructed using Baculogold DNA (Pharmingen) co-transfecting Baculogold DNA with the TbPDE2B-pAcGHLT-A vector into Sf9 cells using the manufacturer's transfection set and the manufacturer's protocol (Pharmingen/BD Biosciences). After 5 days, the supernatants from the experimental co-transfection plates were collected, and co-transfection efficiencies were estimated using end point dilution assays. The virus was amplified twice to generate high titer stock, and the viral titer was determined by plaque assays. Sf9 cells were infected with the virus at a multiplicity of infection ϭ 3. Maximum protein expression was seen after 3 days, when the cells were harvested. The cells were resuspended and lysed in lysis buffer (phosphate-buffered saline (Invitrogen) plus 10% glycerol, 2 mM phenylmethylsulfonyl fluoride, 2 mM dithiothreitol, pH 7.5) by microfluidization (10,000 p.s.i.) (Microfluidics) and centrifuged at 16,000 ϫ g for 30 min. The supernatant was subsequently purified on a glutathione uniflow resin (BD Biosciences) column (using the glutathione S-transferase epitope tag on TbPDE2B), and TbPDE2B was eluted using 10 mM reduced glutathione in the same buffer. The protein exhibited normal wild-type enzymatic activity and was stored at Ϫ20°C for at least 3 weeks without any significant loss of activity.
Immunoprecipitation-Transfected and control (untransfected) cell supernatants were first cleared by incubation with 15 l of protein G-agarose beads at 4°C for 45 min (with gentle mixing). These extracts were subsequently centrifuged at 16,000 ϫ g for 5 min and the supernatants transferred to different tubes. Cleared supernatants were incubated with the anti-V5 antibody (Invitrogen) and 30 l of protein G-agarose beads for 2-3 h in a volume of 1 ml (with gentle mixing). After three washes in homogenization buffer, the immunoprecipitates were resuspended in homogenization or binding buffer or prepared in SDS sample buffer for SDS-PAGE resolution and Western blot analysis.
cNMP Competition Binding Assays and IC 50 Curves-To determine an IC 50 value for cAMP or cGMP displacement of 3 [H]cAMP from the purified N terminus ϩ GAF-A domain, binding assays were conducted in a total volume of 10 ml of binding buffer. Ten pmol of [ 3 H]cAMP (1 nM) and ϳ6 pmol of purified bacterial N terminus ϩ GAF-A protein (0.6 nM protein) were used/assay point, with increasing concentrations of cold cNMP (as indicated). The large 10-ml volume was used to keep the absolute concentrations of [ 3 H]cNMP sufficiently low (because binding affinities for cAMP were ϳ16 nM), therefore retaining sufficient counts. Following an incubation of 20 -30 min on ice, ammonium sulfate was added to a final concentration of either 1 or 3 M (as described under "Results"). For cNMP binding to the holoenzyme, the assay was done in a concentration of 1 ml of binding buffer (10 nM radioligand, 6 nM protein). These conditions for the assay were selected keeping in mind the assumptions and possible ambiguities of competitive binding curves, as described by Motulsky and Christopoulos (23) (i.e.". . .if the concentration of labeled ligand greatly exceeds the equilibrium dissociation constant (K d or K i ), the curve is ambiguous, as an infinite number of curves defined by different K d values are almost identical. . ."). Hence, care was taken to ensure that the [ 3 H]cAMP concentration was well below the K d , that the protein concentration was less than the [ 3 H]cAMP concentration, and that there were sufficient total counts to measure accurately. Therefore, under the conditions of assay, the equilibrium dissociation constant (K i ) obtained from the IC 50 values (using the Cheng and Prusoff equation (24)) closely reflected the IC 50 value itself. More detailed explanations are available in the online manuals on the Prism website resource library (www.graphpad.com). The solution was filtered onto a nitrocellulose filter (Millipore), washed twice with the ammonium sulfate solution, dissolved in scintillation fluid (Filter-Count/PerkinElmer Life Sciences), and the counts bound meas-cAMP-binding GAF Domain PDE from T. brucei ured in a scintillation counter. Non-linear regression analysis of the data was done using Prism version 4.0 (GraphPad Software, San Diego, CA) to obtain the IC 50 and K i values and was presented as the mean Ϯ S.D. from four independent experiments. To obtain Hill coefficients, the binding data were re-expressed as a percentage of [ 3 H]cAMP displaced by unlabeled ligand (100% binding or 0% displaced with no cold ligand present) and analyzed using a sigmoidal dose response curve in Prism, with a variable slope equation, from which the Hill slope was obtained.
Determination of Binding Equilibrium-To determine the time necessary for cNMP binding to reach equilibrium (for both the purified holoenzyme as well as the N terminus ϩ GAF-A construct), binding assays were done as described earlier with 10 nM [ 3 H]cAMP and 6 nM purified protein. The samples were filtered onto the nitrocellulose membrane after 1, 5, 10, 20, 30, 40, 60, and 180 min. Subsequent wash, binding, and detection procedures were as described above.
Phosphodiesterase Assay-PDE assays were carried out using the method of Hansen et al. (25).C. H. Assays were performed using transfected cell supernatants for 10 min at 30°C with 0.1-100 M cAMP as substrate and 15 M isobutylmethylxanthine (to inhibit non-trypanosomatid PDEs in the cell supernatants). Assays were initiated by adding protein to the substrate-buffer mix. The reaction volume was 250 l (40 mM MOPS, pH 7.5, 2.0 mM EGTA, 15 mM magnesium acetate, 0.2 mg/ml bovine serum albumin, and 100,000 cycles/min of [ 3 H]cAMP). TbPDE2B was diluted to produce Ͻ30% total cAMP hydrolysis at any given substrate concentration. Calculations were done using Microsoft Excel and plotted in Prism version 4.0 (GraphPad Software, San Diego, CA) and presented as the average of at least three experiments. K m and apparent V max values were obtained from Prism by analyzing the data using the Michaelis-Menten equation in the program.
Homology Modeling-A homology model of the TbPDE2B GAF-A domain was constructed using Swiss-Model (26,27) based on the crystal structure of the mmPDE2B GAF-B domain (1MC0.pdb) using a Clust-alW (biology workbench) alignment of TbPDE2B GAF-A on mmPDE2B GAF-B as a template. Images were generated using the Swiss Protein Data Bank viewer and further refined using Pov-Ray.
Other Methods-Total protein concentration was determined using the Bradford assay with bovine serum albumin (Pierce) as the standard. Protein levels in Western blots (with an anti-V5 antibody) were normalized using ImageJ (NIH) to measure the relative densities of the detected bands.

TbPDE2B Binds cAMP through Its GAF-A Domain-
HEK293T cells were transiently transfected with a full-length native (wt) TbPDE2B expression plasmid or a TbPDE2B catalytic domain fragment plasmid. Both contained a C-terminal V5 epitope tag (Fig. 1A) allowing immunoprecipitation of the expressed protein using an anti-V5 antibody. Given the high sequence identity of TbPDE2B to mammalian PDE2A, especially within the GAF-A domain, it seemed likely that the TbPDE2B GAF-A domain should bind cyclic nucleotides. How- cAMP-binding GAF Domain PDE from T. brucei ever, an earlier study (20) had not been able to detect cGMP binding. Yet this did not rule out binding of cAMP, which is an important signaling molecule in trypanosomids, nor was the binding to cGMP quantitatively determined in the previous study. To test this possibility, a Millipore filter-binding assay was carried out with [ 3 H]cAMP and immunoprecipitated proteins from control (untransfected), catalytic domain, or TbPDE2B holoenzyme transfected cells. This was done in the presence of at least 5 mM EDTA (which inhibits enzymatic activity by chelating the Mg 2ϩ ion required for hydrolysis) due to the absence of other effective inhibitors for this cAMP-specific PDE. [ 3 H]cAMP bound to immunoprecipitated TbPDE2B but not to immunoprecipitates of control lysates or the catalytic domain alone, ruling out the possibility that the radioactivity detected was because of cAMP binding to the catalytic domain (Fig. 1B).
Given the primary amino acid sequence similarity of the TbPDE2B GAF-A to the PDE2A GAF-B domain (19) as well as the CyaB1 GAF domain of the Anabaena cyclase, the GAF-A domain seemed to be the region most likely to bind cAMP. To test this hypothesis, the N terminus ϩ GAF-A region (Fig. 1A) was cloned, expressed in HEK293T cells, and used in binding studies. The immunoprecipitated protein clearly bound cAMP (Fig. 1B). Cell supernatants of all three expressed proteins were stable and showed no significant visible degradation (as detected by Western blot analysis), even after 3 days of storage at 4°C. For quantification of relative levels of expression, the expressed wt TbPDE2B, catalytic domain fragment and the N terminus ϩ GAF-A fragment were visualized by Western analysis and the relative intensities measured by densitometry (Fig. 1C). It was not possible to determine the binding capacity, if any, of TbPDE2B GAF-B, because recombinant GAF-B in bacteria was insoluble. Similarly, constructs expressing either GAF-B or the GAF-B ϩ catalytic domain in HEK293 cells were unstable and appeared to be degraded immediately upon lysis (data not shown).
cAMP Binds to Recombinant N terminus ϩ GAF-A Domain with Nanomolar Affinity-To obtain enough purified protein to easily determine the binding affinity of cAMP to the GAF-A domain, the N terminus ϩ GAF-A domain was cloned into pET28a (Fig. 1A) and expressed in RosettaDE3 Escherichia coli cells. Protein from the high speed supernatant fraction of the cell lysate was affinity-purified on a TALON affinity resin ( Fig.  2A) and immediately desalted on a PD10 column. Purified protein was then used in [ 3 H]cAMP competition binding assays to obtain an IC 50 value, with increasing concentrations of cold cAMP as the competitive ligand. The IC 50 values for cAMP displacing [ 3 H]cAMP were obtained by analyzing the competitive binding data by non-linear regression and fitting it to a one-site competition equation using Prism software. The actual affinities (K i values) were calculated as explained under "Experimental Procedures." 2 Binding of cAMP to the GAF-A do- 2 The IC 50 of a ligand for a receptor depends on three factors: the affinity of the receptor for the competing drug (defined by the equilibrium dissociation constant K i ), the concentration of the radioligand, and cAMP-binding GAF Domain PDE from T. brucei main was found to be of high affinity, with an IC 50 value of 17.6 Ϯ 1.8 nM (Fig. 2B). This high affinity cAMP binding is similar to that seen for cGMP binding to the mammalian PDEs 2, 5, and 6.
cGMP Binds to the TbPDE2B GAF-A Domain with Lower Affinity than cAMP-Given the high affinity binding for cAMP, as well as the close similarity of the TbPDE2B GAF-A domain to the mmPDE2A GAF-B domain, it seemed reasonable to expect at least some binding of cGMP to the GAF-A domain. Indeed, with the availability of larger amounts of purified bacterially expressed N terminus ϩ GAF-A protein, binding of [ 3 H]cGMP could be detected (Fig. 2C). This purified protein was then used in [ 3 H]cAMP competition binding assays to determine an IC 50 for cGMP binding by competition analysis using increasing cold cGMP as the competitive ligand. 2 cGMP was found to have at least 15-fold lower affinity for the GAF-A domain (compared with cAMP), with an IC 50 value of 289 Ϯ 5.4 nM (Fig. 2D). This discrimination in cyclic nucleotide affinity is just the opposite of the relative differences in the affinities for cGMP and cAMP of the mammalian PDE2A GAF-B domain (12).
Binding of cNMP to the Purified TbPDE2B Holoenzyme-Though the GAF-A domain of TbPDE2B bound cAMP with a high affinity, and cGMP with lower affinity, it was of interest to understand whether this phenomenon is representative of the TbPDE2B holoenzyme. To determine this, a baculovirus expression vector able to express the full-length TbPDE2B enzyme was constructed, baculovirus produced, TbPDE2B expressed in infected Sf9 insect cells, and purified (Fig. 3A). This purified TbPDE2B protein was subsequently used in competition binding studies (as described under "Experimental Procedures") to determine cAMP and cGMP affinities. It was found that cAMP bound the TbPDE2B holoenzyme with an IC 50 of the affinity of the radioligand for the receptor (K d ). The affinity of the ligand (cNMP) can thus be calculated using the equation of Cheng and Prusoff (24), which states that the equilibrium dissociation constant of the ligand  cAMP-binding GAF Domain PDE from T. brucei 62.5 Ϯ 3.6 nM, which was about 3-fold higher than the GAF-A domain alone (Fig. 3B). However, and perhaps more interestingly, cGMP had a much higher IC 50 of 12.2 Ϯ 2.3 M for TbPDE2B. Thus the holoenzyme was able to discriminate more effectively for cAMP, as opposed to cGMP, than the GAF-A domain alone. The data were analyzed using Prism software and did not indicate a second high affinity site for cyclic nucleotide binding. The actual affinities of cNMP for the N terminus ϩ GAF-A domain or the TbPDE2B holoenzyme were calculated from the IC 50 using the Cheng and Prusoff equation (as explained earlier (24)). The equilibrium dissociation constants (K i ) and IC 50 values for cNMP binding to the N terminus ϩ GAF-A domain or to the TbPDE2B holoenzyme are shown in Table I.
The binding data were also reanalyzed in Prism by fitting it to a sigmoidal dose response curve with a variable slope (Fig.  3C), as explained under "Experimental Procedures." Binding of cAMP to the holoenzyme showed positive cooperativity, with a Hill slope of 1.75 Ϯ 0.1. However, binding of cGMP to the holoenzyme showed no cooperativity, with a Hill slope of 0.85 Ϯ 0.1. Thus it seems that this enzyme is not only selective for cAMP, but the nature of the interactions of these two cyclic nucleotides with the protein appear to be different.

The N terminus ϩ GAF-A Domain Alone Reaches cAMP Binding Equilibrium Faster than the TbPDE2B Holoenzyme-
The time to reach binding equilibrium for [ 3 H]cAMP was determined as described under "Experimental Procedures." We found that, at low [ 3 H]cAMP concentrations of 10 nM and low temperature (4°C) and protein concentration of 6 nM, the purified N terminus ϩ GAF-A domain reached cAMP binding equilibrium in Ͻ5 min (Fig. 4). However, the holoenzyme took ϳ45 min to reach binding equilibrium (Fig. 4). These conditions were subsequently used in the binding assays to the GAF domain or to the holoenzyme, respectively. As expected for bimolecular reactions, binding equilibrium was reached more rapidly at higher cAMP concentrations.
Effect of Ammonium Sulfate on cNMP-binding Stoichiometry-It was found that the concentration of ammonium sulfate used in the dilution/wash buffer was a factor for the amount of cNMP bound to the GAF domain. When high concentrations (3 M) of ammonium sulfate were used, an apparent binding stoichiometry of ϳ0.6 mol of cNMP bound/mol of (monomeric) protein was obtained; however, with 1 M ammonium sulfate, an apparent binding stoichiometry of ϳ 0.2 mol of cNMP bound/ mol of protein was seen (data not shown). Effects of ammonium sulfate on cGMP binding to other PDE GAF domains have been observed previously and its use in cNMP binding assays studied in detail (28,29), although the magnitude of the effect seems to depend on the particular protein being investigated. It should be noted that the increased binding appears not to be due to greater loss of bound protein from the filter at the lower ammonium sulfate concentration, as multiple filters or multiple passes of the filtrate through several fresh filters did not increase binding. Therefore, the ammonium sulfate is altering some property of the GAF domain containing protein itself. The advantages and concerns of using ammonium sulfate in cGMP binding assays for other PDEs has been discussed extensively earlier (28 -31). It also should be noted that the K i values determined for cAMP binding were similar whether 1 or 3 M ammonium sulfate was used, indicating that only the maximal amount of cAMP retained on a filter, and not binding affinity, was affected.
A Point Mutation (T317A) in a Predicted Binding Site Results in Loss of cAMP Binding-Given the high similarity of this GAF-A domain with the cGMP-binding GAF-B domain of mammalian PDE2A, critical binding residues could potentially be predicted using the crystal structure of murine PDE2A GAF-B (11) as a model. Nine of the eleven "contact" residues of mmPDE2A GAF-B are well conserved in the TbPDE2B GAF-A domain as seen in a sequence alignment with mmPDE2A and the cyaB1 GAF-B domains (Fig. 5A), suggesting the likelihood of a cyclic nucleotide binding pocket in GAF-A (32). Using this information, as well as a homology model (Fig. 5B) of the TbPDE2B GAF-A domain based on the murine PDE2A GAF-B, the conserved Thr-317 in TbPDE2B (the equivalent of Thr-492 in mmPDE2A) was mutated into an alanine. In mmPDE2A, Thr-492 stabilizes cyclic nucleotide binding within the pocket through a critical contact with the ribose sugar of cGMP and was found to be essential for cGMP binding (12). In the TbPDE2B GAF-A model, the equivalent conserved Thr-317 appeared to function in a similar way. This region is also the most highly conserved region in the model. Therefore, the TbPDE2B T317A mutant was made and expressed in HEK293T cells. Binding assays were performed after immunoprecipitation of native and mutant proteins. We found that under our binding assay conditions, using either 1 or 3 M ammonium sulfate, the T317A mutation resulted in complete loss of detectable cAMP binding to the protein (Fig. 5C). The same result was also seen with a purified, bacterially expressed TbPDE2B N terminus ϩ GAF-A T317A mutant (not shown). No binding could be detected for cGMP as well. This suggests that the cAMP-binding pocket of the TbPDE2B GAF-A domain is likely to have an overall structure similar to that of mmPDE2A GAF-B but with selectivity toward cAMP. It also suggests that Thr-317 forms a stabilizing contact with cAMP, thereby allowing cAMP to remain bound within the binding pocket. Additionally, because the point mutant T317A in GAF-A abolished all detectable cAMP binding in the holoenzyme, the presence of a second high affinity cAMP-binding site in TbPDE2B GAF-B seemed unlikely. However, the presence of a lower affinity site in GAF-B cannot be ruled out.
Binding Site Point Mutation Results in Decreased Enzymatic Activity at Low Substrate Concentrations-This observed loss in cAMP binding of the T317A mutant allowed us to test the hypothesis that the GAF-A domain might act as a regulator of  cAMP-binding GAF Domain PDE from T. brucei 3777 enzymatic activity through its ability to bind cAMP. When tested for activity, we found that the K m of the recombinant T317A mutant holoenzyme was ϳ4-fold higher than that of the native (wt) TbPDE2B holoenzyme (17.45 Ϯ 1.7 versus 4.4 Ϯ 0.8 M, respectively) (Fig. 6A). This strongly suggests that the diminished cAMP binding to the GAF domain results in a decreased catalytic efficiency at the active site at low cAMP concentrations. Because the mutant protein is catalytically active with no significant change in the V max (Fig. 6A), the data also indicate that the point mutation did not affect the global conformation or overall stability of the protein. The recombinant isolated catalytic domain had an even higher K m (Ն44 M) than the mutated holoenzyme, along with a possibly decreased maximal activity. Together these data suggest that cAMP binding to the GAF-A domain allows the full catalytic activity of the TbPDE2B holoenzyme to be expressed. Perhaps more importantly, binding of cAMP to the GAF-A domain allows more catalytic activity at lower cAMP concentrations in the cell. TbPDE2B is highly selective for cAMP as substrate, and cGMP did not affect cAMP-PDE activity when tested between 1 and 50 M cGMP (data not shown), consistent with earlier reports (19 -21). Finally, no cooperative activity on its own hydrolysis was found for cAMP in the kinetic studies. A Hill coefficient of 1.0 was calculated from the kinetic curves for cAMP degradation. The reported Hill coefficient for a mammalian cGMP-stimulated PDE2 is 1.3 (7), which agrees with the maximal predicted theoretical n value of 1.36 for dimeric enzymes showing this kind of regulation (33). DISCUSSION The data presented in this paper shows that cAMP binds to the TbPDE2B GAF-A domain with a high affinity. This is the first report of high affinity cAMP binding to any cyclic nucleotide phosphodiesterase GAF domain. The nucleotide binding is similar to what has been observed in the CyaB1 cyclase or mammalian PDEs 2 and 5 (with cGMP). Both the GAF-A domain alone and the holoenzyme bound cAMP with high affinity, with the GAF-A domain having a somewhat higher affinity than the holoenzyme. However, the holoenzyme shows much greater selectivity for cAMP binding over cGMP binding.
Importantly, there is an increase in the K m of the T317A mutant enzyme, which does not bind cAMP in its GAF domain. The K m increase is even higher for the recombinant catalytic domain (devoid of GAF domains), with an additional possible decrease in V max . These differences in kinetic properties indicate that there is likely to be communication between the GAF-A domain and the catalytic domain. Physiologically, this would suggest that cAMP binding to the intact GAF-A domain can increase the catalytic activity of TbPDE2B by increasing its affinity for substrate. For example, at 1 M cAMP, the binding of cAMP to the GAF-A domain allows at least 10 times greater activity to be expressed compared with the isolated catalytic domain. This higher activity in turn could allow for greater feedback control over cAMP levels. A similar decrease in K m after cNMP binding and activation of the enzyme has been reported in activation studies on mammalian PDE5 (6). The exact molecular mechanisms by which this activation happens, however, remain unclear.
Although cGMP could bind to the GAF-A domain, it did so with much lower affinity than cAMP. This result was even more dramatic in the purified holoenzyme. The presence of the additional domains did not greatly alter the affinity of cAMP binding but significantly decreased the affinity for cGMP binding. Thus it appears that TbPDE2B has fine-tuned cyclic nucleotide selectivity preferentially for cAMP by evolving, not just the binding domain, but the regions around it as well. Additionally, positive cooperativity is seen for cAMP binding to the protein and not for cGMP binding. This suggests that the conformational changes caused by cyclic nucleotide binding to a GAF domain could possibly differ depending on the nucleotide. TbPDE2B, similar to many other PDEs, appears to be dimeric (from gel filtration and light scattering experiments, not shown), and it is conceivable that the cooperativity for cAMP binding exists between the GAF-A domains in the two subunits that compose the dimer. Therefore, it is plausible that the initial nucleotide interaction is at the catalytic site, with the GAF site being denied nucleotide access until the catalytic site is occupied. Such an interaction might then allow the GAF site to be exposed and allow the nucleotide to access the high affinity GAF site. This would be consistent with the binding and kinetic data obtained, because cGMP is not hydrolyzed by this PDE, nor does it affect cAMP hydrolysis by this highly cAMP-specific enzyme. This would also predict that cAMP binding to the holoenzyme would be slower than to the GAF-A alone, as is seen in Fig. 4. Finally, such a model could also fit with the differential effects of ammonium sulfate. At present, it is unclear whether the two nucleotides do cause different conformational changes within the GAF domains or if another mechanism is used to affect enzyme activity. It is also not clear whether any of the mammalian GAF domain-containing PDEs may also share a similar activation mechanism. Ultimately, the structures of the cyclic nucleotide bound and unbound forms of the enzyme should provide a better understanding of these regulatory mechanisms for TbPDE2B and for other GAF domain-containing PDEs.

cAMP-binding GAF Domain PDE from T. brucei
The only other known cAMP-stimulated PDE (without GAF domains) is found in Dictyostelium (34), another lower eukaryote. In Dictyostelium, cAMP levels regulate chemotaxis and aggregation (35), and a cAMP-stimulated PDE was found to be important in controlling the level of aggregation (34). A similar role therefore may exist for TbPDE2B in trypanosomes. This selectivity for cAMP binding to the TbPDE2B GAF domain (compared with cGMP) and the fact that cGMP does not affect TbPDE2B activity is consistent with the absence of reports for a role of cGMP in T. brucei or any other member of the Trypanosomatidae family. A guanylyl cyclase is yet to be identified in these parasites, and there is no significant cGMP hydrolysis in trypanosome lysates. Given that cAMP levels need to be finely regulated for the survival of the parasite, a cAMP-binding GAF domain able to control PDE activity may be one way to execute this exquisite regulation. Other effects of cNMP binding to GAF domains are known in other PDEs, such as effects on phosphorylation of predicted cyclic nucleotide-dependent protein kinase phosphorylation sites (36). Such effects on TbPDE2B are still unexplored. Cyclic nucleotide-regulated protein kinases have been identified in trypanosomes (37,38), although their physiological roles, expression patterns, natural targets, and mechanisms of action are unknown. cAMP binding may also be a mechanism to allow or prevent other proteins from interacting with TbPDE2B.
Finally, the data presented, indicating that cAMP can regulate catalytic activity by binding to a GAF domain, suggests that the GAF domain cyclic nucleotide-binding site, in addition to the catalytic site of TbPDE2s, may provide a new target for novel anti-trypanosomal drugs.