Biological and Structural Characterization of Trypanosoma cruzi Phosphodiesterase C and Implications for Design of Parasite Selective Inhibitors*

Background: Cyclic nucleotide specific phosphodiesterases (PDEs) are essential enzymes in many parasitic protozoa and represent important new drug targets. Results: The crystallographic structure and enzymatic properties of Trypanosoma cruzi phosphodiesterase C (TcrPDEC) have been determined. Conclusion: A parasite-specific pocket next to the TcrPDEC active site might allow designing parasite-specific inhibitors. Significance: The findings highlight the potential of PDEs as anti-parasite drug targets. Trypanosoma cruzi phosphodiesterase C (TcrPDEC) is a potential new drug target for the treatment of Chagas disease but has not been well studied. This study reports the enzymatic properties of various kinetoplastid PDECs and the crystal structures of the unliganded TcrPDEC1 catalytic domain and its complex with an inhibitor. Mutations of PDEC during the course of evolution led to inactivation of PDEC in Trypanosoma brucei/Trypanosoma evansi/Trypanosoma congolense, whereas the enzyme is active in all other kinetoplastids. The TcrPDEC1 catalytic domain hydrolyzes both cAMP and cGMP with a Km of 23.8 μm and a kcat of 31 s−1 for cAMP and a Km of 99.1 μm and a kcat of 17 s−1 for cGMP, thus confirming its dual specificity. The crystal structures show that the N-terminal fragment wraps around the TcrPDEC catalytic domain and may thus regulate its enzymatic activity via direct interactions with the active site residues. A PDE5 selective inhibitor that has an IC50 of 230 nm for TcrPDEC1 binds to TcrPDEC1 in an orientation opposite to that of sildenafil. This observation, together with the screen of the inhibitory potency of human PDE inhibitors against TcrPDEC, implies that the scaffold of some human PDE inhibitors might be used as the starting model for design of parasite PDE inhibitors. The structural study also identified a unique parasite pocket that neighbors the active site and may thus be valuable for the design of parasite-specific inhibitors.


that the scaffold of some human PDE inhibitors might be used as the starting model for design of parasite PDE inhibitors. The structural study also identified a unique parasite pocket that neighbors the active site and may thus be valuable for the design of parasitespecific inhibitors.
Chagas disease, also known as American trypanosomiasis, is a tropical parasitic disease that currently affects ϳ10 million people in Latin America (1,2). The disease is caused by Trypanosoma cruzi, a kinetoplastid protozoan that is delivered by blood-sucking bedbugs such as Rhodnius prolixus (3). After an acute phase of infection, with a mortality rate of ϳ10%, the infection can persist without overt clinical symptoms for 20 -30 years. Although a considerable proportion of the cases never progress beyond this initial stage, many patients finally succumb to the slow but inexorable destruction of the musculature of heart, intestines, and esophagus (2).
In the kinetoplastids, including T. cruzi, differentiation and cell cycle regulation are under the control of cyclic nucleotide signaling (4 -6), in which phosphodiesterases (PDEs) 4 are the sole enzymes that decompose cyclic nucleotides. PDE activity is also involved in the osmoregulation of cell volume after hypoosmotic shock (7,8). Disruption of cyclic nucleotide signaling has been shown to eliminate Trypanosoma brucei infection from the host (9). In this context, PDE inhibitors might constitute a new class of potent drug candidates for protozoal infections. In T. cruzi, several cyclic nucleotide-specific PDEs have already been characterized (10 -14), and a number of TcrPDE inhibitors have been reported (15).
The genomes of kinetoplastids, including T. cruzi, encode four PDE families, termed TcrPDEA, B, C, and D (16) (Fig. 1), in contrast to the human genome, which contains 21 PDE genes categorized into 11 families. TcrPDEA, C, and D each consist of a single member, whereas TcrPDEB contains two highly similar and tandemly arranged genes. TcrPDEA is a cAMP-specific PDE with a high K m and is not inhibited by isobutylmethylxanthine, papaverine, or theophylline (10). The highly similar TcrPDEB1 and TcrPDEB2 are cAMP-specific and contain two GAF domains upstream of their catalytic domains (11,14,17). TcrPDEC contains a FYVE-type domain (18) at its N terminus and was shown to hydrolyze both cAMP and cGMP with similar K m values (13). This observation is supported by recent work of Docampo and co-workers (15) but is in contrast to the claim that TcrPDEC is cAMP-specific (11). TcrPDEC activity has been implicated in osmoregulation (8,12,15). TcrPDED and its orthologs in other kinetoplastids have so far not been experimentally investigated.
Although Chagas disease and its causative agent T. cruzi were discovered ϳ100 years ago (2), there are only a handful of clinically useful drugs at present, and no vaccine is available (19 -22). Worse still, all current drugs cause severe side effects, and their use is hampered by drug resistance (23), emphasizing an urgent need for the discovery of new and more effective drugs for the treatment of Chagas disease (24). In this study, we report the enzymatic properties of various kinetoplastid PDECs, and the crystal structures of the unliganded TcrPDEC1 and its complex with the inhibitor wyq16 (5-(2-ethoxy-5-(N-isopropyloxycarbonyl-aminosulfonyl)-1-methyl-3-propyl-1Hpyrazolo[4,3-d]pyrimidin-7(6H)-one) (Fig. 2). These results provide guidelines for the design of TcrPDEC inhibitors for the treatment of Chagas disease.
For expression in S. cerevisiae, the entire PDEC open reading frames were subcloned from the TA vectors into two variants of the yeast expression vector pLT1 (13) using the restriction enzymes SalI/BamHI or NdeI/BglII (sites are indicated by bold letters in the primer sequences). One pLT1 variant directs the expression of the kinetoplastid protein alone, whereas the other adds an N-terminal hemagglutinin tag to allow detection of the recombinant protein. Transformation of the constructs into the PDE-deficient S. cerevisiae strain PP5 (MATa leu2-3 leu2-112 ura3-52 his3-532 his4 cam pde1::URA3 pde2::HIS3) was carried out as described previously (25).
Yeast Cell Lysis-Yeast cell lysates were prepared as previously described (26). Briefly, yeast was grown in SC-Leu medium to end log phase, collected by centrifugation, and resuspended in the original volume of prewarmed YPD. The  culture was incubated for an additional 3.5 h to maximize protein expression. The cells were then washed twice in H 2 O, and the pelleted cell was snap-frozen in liquid nitrogen and stored at Ϫ70°C. The pellet was thawed on ice and resuspended in an equal volume of ice-cold extraction buffer (50 mM Hepes, pH 7.5, 100 mM NaCl, 1ϫ Complete protease inhibitor mixture without EDTA (Roche Applied Science)). Cell lysates were produced by mechanical disruption with glass beads (0.45-0.50 mm) using a FastPrep FP120 high speed reciprocating device (four times for 45 s at setting 4). The lysate was cleared by centrifugation and frozen for later use.
Subcloning of Catalytic Domain TcrPDEC1 in Escherichia coli and Protein Purification-A pair of primers that contain the restriction sites NheI and EcoRI was designed for the subcloning. The DNA sequence encoding the catalytic domain of TcrPDEC1 (amino acids 270 -614) was amplified by PCR. The amplified TcrPDEC1 fragment and the expression vector pET28a were digested with the restriction enzymes, purified by agarose gel, and ligated with T4 DNA ligase. The resultant plasmid pET-PDEC1 was transferred into E. coli strain BL21(CodonPlus) (Agilent Technologies) for expression. The E. coli cell carrying pET-PDEC1 was grown in LB medium at 37°C to an absorption of A 600 ϭ 0.7, and then 0.1 mM isopropyl ␤-D-thiogalactopyranoside was added to further grow at 15°C for 24 h. The harvested cells were suspended in a lysis buffer (3 ml/gram) consisting of 20 mM Tris-HCl, pH 7.5, 300 mM NaCl, and 15 mM imidazole and disrupted by French Press with 1500 p.s.i. After centrifugation at 15,000 rpm for 20 min, the supernatant was loaded into a nickel-nitrilotriacetic acid column (Qiagen). The nickel-nitrilotriacetic acid column was washed with ϳ300 ml of lysis buffer and then ϳ100 ml of buffer of 20 mM Tris-HCl, pH 8,0, 50 mM NaCl, 15 mM imidazole. The TcrPDEC1 protein was eluted by using 20 mM Tris-HCl, pH 7.5, 50 mM NaCl, and 250 mM imidazole. The N-terminal His 6 -tagged TcrPDEC1 was cleaved by thrombin, and the protein was further purified by Q-Sepharose with three buffers of 20 mM Tris-HCl, pH 7.5, 1 mM 2-mercaptoethanol, 1 mM EDTA, and NaCl at 0.1, 0.2, and 0.3 M. Finally, a Sephacryl S300 column (GE Healthcare) was used to remove aggregated protein and to switch to a buffer consisting of 20 mM Tris-HCl, pH 7.5, 1 mM 2-mercaptoethanol, 1 mM EDTA, 50 mM NaCl. A typical purification yielded Ͼ10 mg of TcrPDEC1 with a purity Ͼ95% from a 2-liter cell culture. Protein concentrations were calculated from extinction coefficients by the program ProtParam, and an A 280 of 0.701 M Ϫ1 cm Ϫ1 was equal to 1 mg/ml of TcrPDEC1 (residues 270 -614).
Selenomethionyl-substituted TcrPDEC1 was prepared using the protocol described by Doublie (27). The E. coli cell was cultured in the M9 minimal medium that was supplemented with the following amino acids: 100 mg/ml Lys, 100 mg/ml Thr, 100 mg/ml Phe, 50 mg/ml Leu, 50 mg/ml Ile, 50 mg/ml Val, and 50 mg/ml selenomethionine. The protocol for purification of selenomethionyl-substituted TcrPDEC1 was the same as that used for purification of the wild-type TcrPDEC1.
Enzymatic Properties-Enzymatic activities were assayed using [ 3 H]cAMP and [ 3 H]cGMP as substrates, as previously reported (28). The catalytic domain of TcrPDEC1 was incubated at 24°C for 15 min with a reaction mixture of 20 mM Tris⅐HCl, pH 7.  able enzyme concentration. Nine concentrations of cAMP or cGMP in a range of 0.04 -800 M were used to obtain the kinetic parameters. Enzymatic properties were analyzed by steady state kinetics, as shown in the form of Michealis-Menten's equation (29). The nonlinear regression (GraphPad Prism 5.0) was applied to obtain the values of K m , V max , and k cat . For measurement of IC 50 , eight concentrations of inhibitors were used at a substrate concentration of Ͻ1/10 K m and a suitable enzyme concentration. All of the measurements were repeated three times.
For screening potential inhibitors, a set of human PDE inhibitors was dissolved in Me 2 SO as 10 mM stocks and used at a concentration of 100 M in the PDE assay. The assay was conducted with lysates of yeast expressing full-length TcrPDEC1 as described (26). 1 M cAMP was used as substrate, and no more than 20% substrate was hydrolyzed in all reactions. The Me 2 SO concentration in the final assay solutions never exceeded 1%, and appropriate control reactions with Me 2 SO alone were always included.
Crystallization and Structure Determination of TcrPDEC1-The catalytic domain of the unliganded TcrPDEC1 (270 -614) and its complex with wyq16 was crystallized by vapor diffusion against a well buffer of 20% PEG 3350, 0.4 M sodium formate, 0.2 M guanidine, 0.1 M MES, pH 6.5, at 4°C. The hanging drops contain 2 l of protein (10 mg/ml) and 2 l of well buffer. The inhibitor complex was prepared by incubation of TcrPDEC1 with 2 mM wyq16 (0.1 M stock in Me 2 SO) for 1 h on ice before crystallization. Diffraction data were collected on Beamline X29 at Brookhaven National Laboratory (Table 1) and processed by the program HKL (30). The crystal of the unliganded TcrPDEC1 has the space group P4 1 2 1 2 with cell dimensions of a ϭ b ϭ 130.3 Å and c ϭ 388.9 Å. The TcrPDEC1-wyq16 complex was crystallized in the same space group but slightly different cell dimensions of a ϭ b ϭ 131.1 Å and c ϭ 394.9 Å.
The structure of TcrPDEC1 was solved by the single-wavelength anomalous diffraction method (31), using a selenomethionine derivative of the TcrPDEC1 catalytic domain. The initial positions of selenium atoms were located by program SHELX-97 (32) and refined to yield a figure of merit of 0.43. The map was subsequently modified by program PHENIX (33) and clearly revealed the protein trace (supplemental Fig. S1). The structure of TcrPDEC1 in complex with inhibitor wyq16 was solved by the molecular replacement program AMoRe (34), using the native TcrPDEC1 structure as the initial model. The phases from the molecular replacement were improved by the density modification package of CCP4. The structures were rebuilt by the program O (35) and refined by the program CNS (36).  Because these residues are critical for binding of the metal ion and substrate, the mutations would presumably inactivate the PDEs. To verify this prediction, the PDECs of T. brucei and T. congolense (predicted to be inactive) and of T. vivax, T. cruzi, and L. major (predicted to be active) were expressed in a PDEdeficient strain of S. cerevisiae. Complementation of PDE-deficient S. cerevisiae restores its heat shock resistance and represents a very sensitive assay for cAMP-PDE activity (25). As predicted by sequence analysis, the PDECs of T. cruzi, L. major, and T. vivax were catalytically active and able to complement the PDE deficiency (Fig. 3A). In contrast, the PDEC of neither T. brucei nor T. congolense was active in this assay (Fig. 3A), although the expression of their recombinant proteins was ascertained by immunoblotting (Fig. 3, B and C). Phylogenetic analysis demonstrated that the mutational inactivation of PDEC occurred rather late in the evolution of the trypanosomes, after the separation of the T. brucei/T. congolense/ T. evansi clade from the remainder of the trypanosomatids (Fig.  3D) (38). The fact that the open reading frames of TbrPDEC, TviPDEC, and TevPDEC encode full-length but catalytically inactive proteins implies that these proteins play a role unrelated to their PDE activity. To determine whether inactive PDECs are expressed, transcription of PDEC was studied in T. brucei. Deep sequencing of mRNAs of various stages of the T. brucei life cycle (39) revealed that stable transcripts of the TbrPDEC gene (Tb927.3.3070) are present in all life cycle stages and that they are processed at one major and a few minor splice sites. The overall abundance of TbrPDEC transcripts in T. brucei (ϳ50/10 6 total transcripts) is not significantly different between different life cycle stages (bloodstream versus procyclic forms). It is also similar to that of the enzymatically active TcrPDEC of T. cruzi (ϳ30/10 6 total transcripts). 7,8 Northern blot analysis indicates a transient 2-fold increase in mRNA levels during transformation from bloodstream to procyclic forms, in agreement with a recent microarray analysis that demonstrated a similar up-regulation of Tb927.3.3070 during differentiation from short stumpy to procyclic forms (40). The expression of the TbrPDEC protein has also been established in a proteomics approach with procyclic T. brucei (41) Table 2). These enzymatic parameters are consistent with the K m values of 31.6 and 78.2 M for cAMP and cGMP, which were reported earlier for full-length TcrPDEC1 (13). It is interesting to note that the k cat of the TcrPDEC1 catalytic domain is 7.5-90 times the k cat values of other reported PDEs, although the catalytic efficacy of TcrPDEC1 is still comparable with other PDEs because of its weak apparent association constant ( Table 2).

RESULTS AND DISCUSSION
The potency of various inhibitors of human PDEs was assayed against full-length TcrPDEC1 (Fig. 4 and Table 3). It was found that some inhibitors such as etazolate and trequinsin are very potent (Table 3 for IC 50 values), and thus their scaffolds might be taken as starting models for the design of TcrPDEC selective inhibitors.
Structure of TcrPDEC1 Catalytic Domain-The structures of the unliganded TcrPDEC1 catalytic domain (traceable residues 277-609) and its complex with inhibitor wyq16 were determined at medium resolution ( Table 1). The crystallographic asymmetric unit contains eight molecules that are associated into an octamer or two tetramers (Fig. 5, A and B). The oligomerization of the TcrPDEC1 catalytic domain is different from those of human PDEs and LmjPDEB1 (42,43). It is not clear whether the tight tetrameric or octameric association of the TcrPDEC1 catalytic domain is biologically relevant.
A monomer of the TcrPDEC1 catalytic domain comprises 16 ␣-helices with two divalent metal ions bound at the bottom of the active site (Fig. 5, C and D). The ␣-helices are arranged in a similar topology as those of the catalytic domains of human PDEs and LmjPDEB1 (42,43). The nature of the divalent metal ions was not identified, but zinc and magnesium were used in the structure refinement, as it has been done in the structures of human PDEs (42). The metal ion assignment appears to be reasonable and is supported by the comparable B-factors of the metal ions with the protein residues (Table 1) and proper electron density in the unbiased maps of (F o Ϫ F c ) and (2F o Ϫ F c ). 7 K. Gunasekera, personal communication. 8 T. Ochsenreiter, personal communication. The zinc ion chelates with His-372, His-409, Asp-410, and Asp-521, plus two water molecules in an octahedral configuration. The magnesium ion coordinates with Asp-410 and five water molecules to form an octahedron. An unexpected observation of this study is that the N-terminal fragment of TcrPDEC1 stretches from the back to the front of the molecule and approaches the substrate-binding pocket of the enzyme (Fig. 5,  C and D). This observation implies that the N terminus might play a role in regulation of the enzymatic activity of TcrPDEC1 via a direct interaction with the active site. This argument is consistent with the report that a fragment of the upstream conserved region of human PDE4 interacts with the inhibitors at the active site (44). However, the corresponding loop of PDE4, which links the upstream conserved region fragment to the back of the PDE4 domain, is disordered, in contrast to the observable conformation of our N-terminal stretch. Structural superposition of the TcrPDEC1-wyq16 complex over human PDE4D, human PDE5A, and LmjPDEB1 yielded RMS deviations of 1.45, 1.38, and 1.59 Å for the C␣ atoms of 269, 287, and 247 comparable residues, respectively (Fig. 5, E and F), indicating their overall similarity. However, there are some subtle but significant differences between them. TcrPDEC1 lacks the short helix H4 but has an extra helix before H7. The long helix H14 is split into two helices in TcrPDEC1 because of a missing amino acid before Val-537. The biological meaning of these differences is not clear but might simply reflect the consequence of the amino acid sequence variation. The most significant change is associated with the M-loop (residues 544 -564) at the active sites of the PDEs, which showed differences of several Angstroms between their C␣ atoms (Fig.  5, B, E, and F). These conformational changes do not appear to result from inhibitor binding, as shown by the comparable conformations and positions of the M-loops between the unliganded and wyq16-bound structures. However, it is not clear whether the M-loop changes represent the native conformation differences between TcrPDEC and other PDEs or the consequence of the crystal packing because some residues of the M-loop such as Pro-558 and Glu-561 are involved in formation of the tetramer in the crystal (Fig. 5B). On the other hand, the H-loop at the active site of TcrPDEC favorably compares with that of human PDE4 but not PDE5. Because most human PDE families have an H-loop conformation similar to PDE4 and PDE5 uniquely shows multiple conformations (42,45,46), the H-loop might not be an attractive target for the design of TcrPDEC inhibitors.
Inhibitor Binding-The PDE5 inhibitor wyq16 (Fig. 2) shares the ethoxyphenyl and pyrazolopyrimidinone groups with silde-  respectively. The electron density maps clearly define the conformation and position of the ethoxyphenyl and pyrazolopyrimidinone groups of wyq16 (supplemental Fig. S3). However, the isopropyloxycarbonylamino tail seems more flexible, as shown by the weaker density and higher B-factors. wyq16 binds to the TcrPDEC1 active site with two conserved characteristics: the NE2 atom of the invariant Gln-570 forms a hydrogen bond with O1 of wyq16, and Phe-573 stacks against the pyrazolopyrimidinone ring of wyq16 (Fig. 6). In addition, wyq16 forms van der Waals contacts with residues Tyr-367, Thr-417, Met-482, Leu-496, Ile-538, Phe-542, Phe-560, Ser-569, and Phe-577. An unexpected finding is that the common pyrazolopyrimidinone rings of wyq16 and sildenafil orient oppositely. As a result, the hydrogen bonding patterns are completely different, although the pyrazolopyrimidinone rings of both molecules stack against the phenylalanine (Phe-573 in TcrPDEC1). The side chain of the invariant glutamine forms two hydrogen bonds with the pyrazolopyrimidinone ring of sildenafil in PDE5A (45), but only one with wyq16 (Fig. 6B). This difference might be due to the different conformations of the invariant glutamine, in which the glutamine side chain of TcrPDEC1 has an opposite orientation to that of PDE5 (Fig. 6) but the same as that of PDE4. Potential Parasite Pocket-Our earlier work revealed that LmjPDEB1 has a small pocket neighboring the active site (originally designated as the L-pocket) (43). The TcrPDEC1 structures further showed that the pocket is unique and universal for the kinetoplastid PDEs, and thus the term "P-pocket" appears  to be more appropriate. The P-pocket is made up of residues from the M-loop and helix H14, including Leu-539, Ala-543, Gly-559, Ser-564, Val-566, and Ser-569 of TcrPDEC1 (Fig. 7A).
Two key residues, Gly-559 and Ser-569, serve as the gatekeepers guarding the access to the P-pocket. In most other kinetoplastid PDE families, the residues corresponding to Gly-559 and Ser-569 are methionine/phenylalanine and glycine (Fig.  7B). Therefore, the gate to the P-pocket of all the kinetoplastid PDEs is sterically open and accessible. However, in human PDEs, this pocket is either blocked by two large gating residues or disappears because of variations in sequence, as shown by the sequence alignment (Fig. 7B). Thus, the P-pocket might be an invaluable structural element for developing parasite-specific PDE inhibitors.
Hint on Use of Some Human PDE Inhibitors as Starting Scaffolds for Design of Parasite PDE Inhibitors-The kinetic and structural studies imply that certain inhibitors of human PDEs could be used as the starting framework for the design of parasite-selective inhibitors. This argument is supported by the inhibitory effects of some human PDE inhibitors on TcrPDEC ( Fig. 4 and Table 3). For example, etazolate inhibits human PDE4 and TcrPDEC with the IC 50 values of 2 and 0.7 M, respectively, for the full-length enzymes (13) and of 1.1 and 0.3 M for the PDE4D2 and TcrPDEC catalytic domains. Moreover, the pyrazolopyrimidinone ring of wyq16 binds to TcrPDEC in an opposite orientation to the same ring of sildenafil in PDE5A (45). Although the IC 50 of wyq16 is only 230 nM, the pyrazolopyrimidinone ring might be used as a potential starting scaffold for TcrPDEC inhibitors because of the opposite orientations of both the nine-membered ring and the side chain of the invariant glutamine. Finally, the structure shows that the pyzarole of wyq16 points toward the P-pocket. A suitable substitution on the pyzarole ring would extend the inhibitors to the P-pocket and may thus improve their affinity and selectivity.
Implication of Inactive Parasite PDECs-The finding that mutations during the course of evolution inactivated the PDECs of T. brucei, T. evansi, and T. congolense, has come as a surprise. Interestingly, such a case has also been reported for other enzymes of eukaryotes, including protein kinases (47), protein phosphatases (48), and proteases (49). For example, the EGF receptor kinase HER3 is mutationally locked in an inactive conformation, despite the fact that it can still bind ATP and coordinate Mg 2ϩ . However, HER3, rather than serving as a kinase, plays an important role in controlling the activity of other EGF receptors upon ligand binding (50). In another example, inactive homologs of the rhomboid membrane proteases play an important role in controlling the activity of the active enzymes (49).
A systematic survey of metazoan enzymes has identified many inactive forms of other enzymes, but PDE homologs (51). Thus, the kinetoplastid PDECs represent the first case of a PDE family, in which active and inactive homologs coexist and are stably expressed. The inactive TbrPDEC gene is regulated and expressed in T. brucei at similar levels as the active TcrPDEC in T. cruzi, suggesting that the inactive PDECs may still fulfill a cellular function unrelated to cyclic nucleotide hydrolytic activity. Nevertheless, the function of the active enzyme in species of L. major and T. cruzi remains to be established in detail. Of equal interest will be the elucidation of whether maintenance of the inactive PDECs in T. brucei, T. congolense, and T. evansi is due to a selective pressure exerted by their life style.