Guanylyl Cyclase Activity Associated with Putative Bifunctional Integral Membrane Proteins in Plasmodium falciparum *

We report here that guanylyl cyclase activity is associated with two large integral membrane proteins (PfGCα and PfGCβ) in the human malaria parasite Plasmodium falciparum. Unusually, the proteins appear to be bifunctional; their amino-terminal regions have strong similarity with P-type ATPases, and the sequence and structure of the carboxyl-terminal regions conform to that of G protein-dependent adenylyl cyclases, with two sets of six transmembrane sequences, each followed by a catalytic domain (C1 and C2). However, amino acids that are enzymatically important and present in the C2 domain of mammalian adenylyl cyclases are located in the C1 domain of the P. falciparum proteins and vice versa. In addition, certain key residues in these domains are more characteristic of guanylyl cyclases. Consistent with this, guanylyl cyclase activity was obtained following expression of the catalytic domains of PfGCβ inEscherichia coli. In P. falciparum, expression of both genes was detectable in the sexual but not the asexual blood stages of the life cycle, and PfGCα was localized to the parasite/parasitophorous vacuole membrane region of gametocytes. The profound structural differences identified between mammalian and parasite guanylyl cyclases suggest that aspects of this signaling pathway may be mechanistically distinct.

The life cycle of the human malaria parasite Plasmodium falciparum is complex with several stages that differ at both the morphological and biochemical levels. Mosquito transmitted sporozoites migrate to the liver where they undergo asexual multiplication within hepatocytes. Liver forms are then released into the bloodstream where they invade red blood cells. A small proportion of the erythrocytic forms develop into gametocytes, precursors of male and female gametes. When gametocytes are taken up by a mosquito during a bloodmeal, they must first emerge from the red blood cells as gametes before fertilization occurs in the midgut, prior to completion of the insect stages of the parasite life cycle. The emergence of eight motile male gametes from a red blood cell is known as exflagellation. Recent work (1) has identified xanthurenic acid, a metabolite of tryptophan, as a putative in vivo gametocyteactivating factor. In both P. falciparum and Plasmodium berghei there is evidence that cyclic GMP may be involved in the development and emergence of male gametes (2). The parasite signaling pathways involved in these processes remain to be elucidated.
Two key components of signal transduction pathways are adenylyl cyclase (AC) 1 and guanylyl cyclase (GC), enzymes that catalyze the conversion of ATP and GTP to cAMP and cGMP, respectively. Membrane-associated mammalian ACs are activated indirectly following interaction of a ligand with a separate receptor. This in turn binds to an intracellular heterotrimeric GTP-binding protein (G protein). Subunits of the activated G protein then interact with AC stimulating the synthesis of cAMP, which activates cAMP-dependent protein kinase (PKA). PKA phosphorylates a number of proteins; its final destination is often the nucleus where it activates transcription factors, thereby changing the pattern of gene expression. In mammals, nine distinct membrane-localized ACs have so far been characterized (3). They conform to the same basic structure comprising two sets of six hydrophobic transmembrane domains, each of which is followed by a cytoplasmic catalytic region termed C1 and C2, respectively. The aminoterminal portions (C1 a and C2 a ) are homologous with each other and between species (4). Guanylyl cyclase exists in two main forms. The receptor form (with a single membrane-spanning domain) binds directly to an extracellular ligand leading to activation and production of intracellular cGMP. The soluble, cytosolic form of the enzyme is activated by nitric oxide in * This work was supported by a project grant from the Wellcome Trust (to D. A. B. and J. M. K.), a Wellcome Trust University Award (to D. A. B.), and a WHO/UNDP/World Bank special program award (to D. C. W.). Preliminary sequence data for P. falciparum chromosome 11 were obtained from the Insitute for Genomic Research website. Sequencing of chromosome 11 was part of the International Malaria Genome Sequencing Project and was supported by award from the NIAID, National Institutes of Health. Sequence data for P. falciparum chromosome 13 were obtained from the Sanger Center website. Sequencing of P. falciparum chromosome 13 was accomplished as part of the Malaria Genome Project with support by the Wellcome Trust. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
b These authors contributed equally to this work. the presence of heme (reviewed in Ref. 5). GCs have a variety of roles in higher organisms, for example in photoreceptor signal transduction (reviewed in Ref. 6). GCs have also been identified in several lower organisms including protozoans such as Dictyostelium (7), Paramecium, and Tetrahymena (8,9). In sea urchins, a membrane-bound GC found on the surface of sperm can act as a receptor for peptides released by the eggs, which influence sperm chemotaxis (reviewed in Ref. 10). cGMP often exerts its function by binding to and activating cGMP-dependent protein kinase, which phosphorylates and regulates the activity of a number of specific proteins.
As part of an investigation of the role of cyclic nucleotides in signal transduction in the malaria parasite, we have isolated two genes from P. falciparum that encode integral membrane proteins. These proteins are unusual in that they appear to be bifunctional. The amino-terminal regions have strong similarity to P-type ATPases, and the carboxyl-terminal regions, including domains with GC activity, conform to a structure normally associated with G protein-dependent ACs.
Isolation of Nucleic Acids-Genomic DNA was isolated from 10 9 mixed asexual erythrocytic stage parasites according to standard procedures (14). Briefly, the parasites were first released from red blood cells by treatment with 0.1% saponin in phosphate-buffered saline. The cell pellet was then washed twice with phosphate-buffered saline and incubated in a solution containing 100 g ml Ϫ1 proteinase K, 0.5% SDS, 100 mM NaCl, 10 mM Tris, pH 8.0, 25 mM EDTA at 37°C for 16 h. DNA was further purified by phenol extraction and ethanol precipitation. Total RNA was isolated from 5 ϫ 10 8 gametocytes (mainly stages III-V) or mixed asexual erythrocytic stage parasites by lysis with 4 M guanidine thiocyanate and CsCl centrifugation (15).
Amplification of PfGC␣ Sequences by PCR-A 220-bp fragment of the PfGC␣ gene was isolated from genomic DNA (strain K1) by PCR using the degenerate primers DC1 (sense) and MS2 (antisense), which were derived from conserved regions of AC/GC genes from diverse species. The reactions (50 l) were carried out in 10 mM Tris, pH 8.8, 50 mM KCl, 0.5-5 mM MgCl 2 , and 1% Triton X-100 with 1 M of each primer, 200 M dNTP mix, 1 unit of Biotaq enzyme (Bioline) and 1-50 ng of template DNA. To minimize nonspecific amplification products, the "touchdown" method (16) of thermal cycling was used with the following conditions: first cycle, 94°C (2.5 min), 41°C (1 min), and 72°C (1 min); and second cycle, 94°C (45 s), 40°C (1 min), and 72°C (1 min). The annealing temperature was then decreased by 1°C each two cycles until it reached 35°C. 25 cycles were performed under these conditions. A final cycle was then carried out at 94°C (45 s), 35°C (1 min), and 72°C (10 min). Inverse PCR (17) was used to extend the available PfGC␣ sequence after library screening (below) failed to isolate the full-length gene. Based on Southern blot data, an AluI digest of genomic DNA (1 g, strain K1) was circularized with T4 DNA ligase, and an overlapping fragment was obtained by inverse PCR using primers IPCR1 and IPCR2.
Construction of a Genomic DNA Library-A GEM-12 library was constructed with P. falciparum (strain K1) genomic DNA using a Promega kit according to the manufacturer's instructions. Briefly, a Sau3a partial digest of genomic DNA was end-filled and ligated with bluntended XhoI cut GEM-12 arms. After ligation, the DNA was packaged in vitro using a Packagene Extract (Promega) and plated on Escherichia coli strain LE392. The library (containing 2.7 ϫ 10 5 clones) was screened with the original 220-bp PCR product using standard procedures (14), and a genomic clone (1) containing part of the PfGC␣ gene was isolated.
Construction and Screening of Vectorette Libraries-Vectorette libraries were constructed as described previously (18) using 1 g of genomic DNA (clone 3D7A) that had been digested with one of several restriction enzymes. This DNA (200 ng) was ligated into blunt-ended Vectorette DNA (Cambridge Research Biochemicals). Based on known sequence and Southern blot information, PCR was performed using specific primers (AC1-AC9 for PfGC␣, pvec1-3 for PfGC␤) and the Vectorette I primer (Cambridge Research Biochemicals). With this procedure we were able to obtain the remainder of the PfGC␣ coding sequence and the cyclase-encoding domain of PfGC␤. In the PCR reactions, the polymerase was added to the tubes in the cycling block when the reaction temperature was at least 80°C. This "hot start" and the use of annealing temperatures in the range 60 -70°C were important for the success of the reaction.
Oligonucleotides Nucleic Acid Analysis-To isolate chromosome-sized DNA, asexual blood stage parasites were first released from red blood cells by saponin lysis and then washed in TSE buffer (20 mM Tris, pH 8.0, 100 mM NaCl, 50 mM EDTA). The pellet was resuspended in 9 volumes of TSE and then mixed with an equal volume of 1.6% agarose in TSE and pipetted into a gel mold (Bio-Rad). The agarose blocks (10 8 -10 9 parasites/block) were incubated in 1% sarcosinate, 0.5 M EDTA, pH 8.0, 2 mg ml Ϫ1 proteinase K for 48 h at 50°C and then stored at 4°C. Chromosome separations were performed with 1% agarose gels using a contourclamped homogeneous field electrophoresis system (CHEF DRII, Bio-Rad). The gels were run in 0.5% TBE buffer at 100 V for 96 h with a pulse time of 360 s ramped to 800 s. Typically, hybridizations were performed at 42°C in 6ϫ SSC and 50% formamide with washes in 1ϫ SSC at 50°C. For lower stringency experiments, hybridizations were at 50°C in 5ϫ SSC with washes in 2ϫ SSC at room temperature. Northern blots were carried out using standard conditions (15).
Production of Antibodies against the C2 Catalytic Domain of PfGC␣-Oligonucleotide primers 5CAT and 3CAT were designed to amplify the carboxyl-terminal region of PfGC␣, Ser 3914 -Leu 4226 ), incorporating the C2 catalytic domain. EcoRI restriction sites were included in the primers to facilitate insertion of the fragment in-frame with the Schistosoma japonicum glutathione S-transferase gene of the pGEX1T plasmid (Amersham Pharmacia Biotech). Expression of the fusion protein was induced in a 500-ml mid-log phase E. coli (strain JM109) culture (A 600 ϭ 0.4 -0.5) by the addition of isopropyl-1-thio-␤-D-galactopyranoside to a final concentration of 1 mM for 2 h. Cells were resuspended in phosphate-buffered saline /1% Triton X-100 and sonicated on ice. The soluble fusion protein was bound to 2 ml of glutathione-agarose (Sigma) on a Poly-Prep column (Bio-Rad) according to the manufacturer's instructions. Bound protein was eluted by addition of 5 mM reduced glutathione and dissolved in 0.5 ml of normal saline at a final concentration of 2 mg ml Ϫ1 . This was mixed with 0.5 ml of Ribi Adjuvant System (Sigma) resuspended in 1 ml of normal saline. The mixture was homogenized by vortexing, and a New Zealand White rabbit was immunized at six intradermal sites (50 g each), two intramuscular sites (200 g each), and two subcutaneous sites (100 g each). Similar booster injections were given at 3-4-week intervals. Preimmune serum was taken prior to immunization.
Immunoelectron Microscopy-Percoll-purified gametocytes that had been stimulated to undergo gametogenesis were fixed for 30 min with 2% paraformaldehyde and 0.1% glutaraldehyde in phosphate-buffered saline, dehydrated in ethanol, embedded in LR White (London Resin Co.), and polymerized at 50°C for 48 h. Ultrathin sections were collected on piloform-coated nickel grids and incubated with primary antibody and gold-conjugated secondary antibody as described previously (19). The sections were silver enhanced using an Amersham Pharmacia Biotech IntenSE M silver enhancement kit and then stained with a saturated uranyl acetate solution in 30% methanol followed by Reynolds lead citrate solution. The samples were analyzed with a JEOL 1200EX transmission electron microscope at 80 kV and photographed using Agfa Scienta EM film.
Sequence Analysis-DNA sequencing was performed using an ABI PRISM 377 DNA sequencer. The reactions were performed with a dye terminator cycle sequencing Ready Reaction Kit (Perkin-Elmer). Pre-liminary sequence data for P. falciparum chromosome 11 were obtained from the Institute for Genomic Research website. Sequencing of chromosome 11 was part of the International Malaria Genome Sequencing Project and was supported by an award from the NIAID, National Institutes of Health. Sequence data for P. falciparum chromosome 13 were obtained from the Sanger Center website. Sequencing of P. falciparum chromosome 13 was accomplished as part of the Malaria Genome Project with support by the Wellcome Trust.
Adenylyl/Guanylyl Cyclase Assays-AC and GC activity was assayed in reaction volumes of 40 l. 30-l aliquots of fusion protein (2-5 g) were added to 10 l of 4ϫ reaction buffer (40 mM Tris, pH 8.0, containing either ATP or GTP at 4 mM and 8 mM MgCl 2 or MnCl 2 ) and incubated at 30°C. Reactions were stopped by addition of 10 l of 0.2 M EDTA, pH 8.0, followed by boiling for 2 min. For the t ϭ 0 time points, EDTA was added before the protein. cAMP or cGMP concentrations were measured in 96-well plates using a competition assay (20) with purified regulatory subunit of beef heart cAMP-dependent protein kinase as the cAMP-binding protein (PKA-R) or highly specific anti-cGMP antibodies as the cGMP binding protein (21)

Isolation of PfGC␣, a Gene That Encodes a Protein with
Distinct Cyclase and ATPase Domains-Degenerate oligonucleotide primers were designed based on conserved regions in the catalytic domains of AC/GC from other species (see "Experimental Procedures"). Using these, a 220-bp fragment was amplified by PCR from P. falciparum genomic DNA. Data base searches suggested that the product was derived from a purine nucleotide cyclase gene homologue. This fragment was used to screen a genomic DNA library, and a clone was isolated. Sequence analysis of a 5.5-kilobase BamHI-EcoRV fragment identified a 1.8-kilobase open reading frame (ORF) corresponding to the 3Ј-end of a cyclase-like gene, with a putative stop codon and adjacent 3Ј noncoding sequence. Inverse and Vectorette PCR techniques (see "Experimental Procedures") were used to extend the sequence toward the 5Ј-end of the gene. This generated an uninterrupted ORF of approximately 10 kilobases. A putative start codon was then identified following searches of the Institute for Genomic Research P. falciparum data base. Verification of the gene structure was obtained by sequencing the corresponding region ( Fig. 1), amplified from genomic DNA. In addition the entire sequence of PfGC␣ has been confirmed following release of data from the Institute for Genomic Research data base, which became available during the preparation of this manuscript.
The nucleotide sequence of PfGC␣ comprises a single ORF of 12,678 bp (these sequence data have been submitted to the GenBank TM data base under accession number AJ245435). The A/T content (75.6%) and codon usage conform to that of all P. falciparum genes reported to date. The designated termination codon (TAA) of PfGC␣ occurs in a position analogous to that in mammalian-type AC (4, 23-25) and is followed immediately by an extremely A/T-rich region (95%). Within the ORF, two regions of unusually high A/T content were identified (nucleotides 7350 -7530 and nucleotides 9060 -9450). To investigate the possibility of introns at these positions and elsewhere in the gene, a series of RT/PCR reactions were carried out. There were no differences in the sizes of the products obtained from RNA (DNase I-treated) and genomic DNA (data not shown). In negative controls where the reaction was performed in the absence of RT, no products were amplified. Using this approach it was shown that PfGC␣ was uninterrupted by introns.
Analysis of the PfGC␣ sequence revealed an unexpected finding; the gene had the potential to encode a bifunctional protein in which the amino-terminal domain had high similarity to P-type ATPases (26), and the carboxyl terminus was a structural homologue of G protein-dependent adenylyl cyclases. The nucleotide sequence adjacent to the putative start codon is aaaaATG, which is similar to the consensus sequence upstream of many P. falciparum genes (27). The sequence immediately upstream of this is preceded by a highly A/T-rich sequence interrupted by stop codons in all three reading frames.
Isolation of PfGC␤-A P. falciparum GEM-12 library was screened at low stringency with a probe corresponding to the PfGC␣ C2 catalytic domain (see "Experimental Procedures"). A clone containing 1.3 kilobase of sequence with similarity to PfGC␣ and encompassing an in-frame stop codon was isolated. This gene was designated PfGC␤. The remainder of the cyclase encoding region (an additional 2000 nucleotides) was obtained from Vectorette libraries (see "Experimental Procedures"). Upstream, an additional open reading frame with a high degree of similarity to P-type ATPases was discovered. The sequence of this region was obtained from the P. falciparum genome project (Sanger Center). The cyclase and the ATPase coding regions were found to be separated by 241 bp of A/T-rich sequence. This was shown to be an intron, and both coding regions were found to form a contiguous ORF. Sequence analysis of a series of RT/PCR products obtained using primers specific to both the cyclase and ATPase domains demonstrated the presence of 12 introns in the PfGC␤ sequence. These were all confined to the ATPase encoding domain (Fig. 1). The fully spliced transcript has the potential to encode a bifunctional protein of 3122 amino acids (accession number AJ249165). The 5Ј-most exon contains an A/T-rich sequence upstream of the designated start codon that has stop codons in all three reading frames. Sequencing of RT/PCR products has confirmed that this A/T-rich sequence forms part of the mature mRNA. The predicted stop codon (position 9367) is located upstream of a region with an extremely A/T-rich composition that has stop codons in all three reading frames. The positions of the designated start and stop codons are consistent with those found in other genes encoding similar P-type ATPases and cyclases, respectively. The sequence of the cyclase-encoding region has been confirmed by chromosome 13 sequence data that were released (Sanger Center) during preparation of this manuscript. Using the Genestream Align global alignment program the sequences of PfGC␣ and PfGC␤ were found to share 22% identity in their cyclase domains and 19% in their ATPase domains.
Interestingly, the aggregation-specific adenylyl cyclase (ACA) of the protozoan Dictyostelium discoideum has a similar sequence insertion at the same position within the C1 catalytic domain (29). There is a second, shorter insertion in PfGC␣ C1 (Tyr 3186 -Thr 3225 ; Fig. 2) that corresponds (in terms of both position and size) to an insert found in all of the trypanosomatid ACs sequenced to date. No such inserts are present in the cyclase catalytic domains of PfGC␤. Alignment of PfGC␣ C1 with PfGC␣ C2 revealed 23% sequence identity, whereas alignment of PfGC␤ C1 and PfGC␤ C2 gave 34% identity.
The cyclase catalytic domains of both PfGC␣ and PfGC␤ are each preceded by highly hydrophobic regions that in membrane-associated mammalian ACs have been predicted to form two sets of six transmembrane domains. The PfGC␣ and PfGC␤ sequences are also compatible with this type of organization (Fig. 3).
Features of the ATPase Domains-The putative ATPase domains of both proteins conform to the P-type (or E1-E2) family of cation transporting ATPases (30). The model for this family of enzymes predicts 10 transmembrane domains arranged as a set of four toward the amino terminus and a set of six near the carboxyl terminus. The amino acid sequences of both PfGC␣ and PfGC␤ are consistent with this type of structure (Fig. 3). For both PfGC␣ and PfGC␤ the relatedness to P-type ATPases from diverse species is concentrated in two regions and is summarized below. The amino-terminal domain of PfGC␣ has 26 -32% identity and 48 -52% similarity over 250 -300 amino acids; for the carboxyl terminus the figures are 28 -30% identity and 48 -50% similarity. For PfGC␤, the amino-terminal domain has 23-25% identity and 43-46% similarity; for the carboxyl terminus the figures are 21-23% identity and 46 -48% similarity. The membrane-spanning regions of the P-type ATPase domains are more highly conserved than are those of the cyclase domains. This might be expected given the potential functional significance of these transmembrane regions. The consensus phosphorylation site of P-type ATPases (DKTGTLT) is present in PfGC␣ (Asp 756 -Thr 762 ) but not in PfGC␤. This signature sequence is involved in formation of an aspartyl phosphate intermediate during ATP hydrolysis (31).
Expression and Subcellular Localization-Southern analysis of genomic DNA with a probe derived from the region of PfGC␣ corresponding to the C2 domain produced a hybridization pattern consistent with a single copy gene (data not shown). In accordance with this, when P. falciparum (clones 3D7A and T996) chromosomes separated by CHEFE and transferred to nylon membranes were analyzed (the probe corresponded to nucleotides 10,788 -11,289), a single band identified as chromosome 11 was detected (Fig. 4). PfGC␤ was localized to chromosome 13. When northern blots of RNA derived from sexual and asexual blood stage parasites were probed with PfGC␣, a transcript of approximately 12,000 -14,000 nucleotides was detected that was specific to the sexual blood stages (Fig. 4). The sexual stage RNA had been prepared from Percoll-purified gametocytes (mainly stages III-V), and the asexual stage RNA was prepared from an asynchronous culture containing ring, trophozoite, and schizont stages. The same blot was hybridized with a probe derived from the ATPase-encoding domain of PfGC␣ (nucleic acids 5,901-6,630), and a band of equal size was detected (Fig. 4). A northern blot was also hybridized with a probe corresponding to the cyclase region of PfGC␤, and a slightly smaller transcript was identified (Fig. 4). The northern blots were hybridized with other probes (data not shown) to demonstrate both the integrity of the RNA preparations and the equivalence of loading as described previously (32,33). The probes were derived from genes encoding sexual (Pfs16) and asexual blood stage specific proteins (Msp1) and another (cal-modulin) expressed in similar amounts in both stages.
A rabbit polyclonal antiserum was raised to a recombinant protein corresponding to the carboxyl terminus of PfGC␣ (amino acids Ser 3914 -Leu 4226 , incorporating the C2 catalytic domain) fused to glutathione S-transferase (see "Experimental Procedures"). The serum was used in immunoelectron microscopic examination of sections of red blood cells infected with P. falciparum. Antibody binding was visualized by using a colloidal gold-conjugated secondary antibody. Sexual erythrocytic stage parasites (gametocytes) were strongly labeled in the plasma/parasitophorous vacuole membrane region of the parasite (Fig. 5). Only very low levels of gold particles were observed in the gametocyte cytoplasm. Little or no labeling was present on uninfected red blood cells or asexual blood stage parasites (Fig.  5). These sections also contained a small proportion of developing gametes at various stages of emergence from the surrounding red blood cell membrane and parasitophorous vacuole membrane. No gold particles were observed in these forms. Preimmune serum from the rabbit showed little or no labeling at the same concentration as the test serum. An antiserum raised to a control glutathione S-transferase fusion protein (derived from Schistosoma mansoni) also gave no significant labeling.
Guanylyl Cyclase and Adenylyl Cyclase Enzyme Assays-The regions of PfGC␣ encoding both the C1 (Gly 2999 -Val 3329 ) and C2 (Gln 3956 -Lys 4163 ) domains and the regions of PfGC␤ encoding both C1 (Val 1533 -Tyr 1758 ) and C2 (Glu 2946 -Ala 3122 ) were expressed in E. coli strain TP610 (AC-deficient) using an amino-terminal hexahistidine tag system. The selection of these regions was based on the scheme used for expression/crystallization of the rat type II AC C2 domain (25). Coomassie Bluestained SDS-polyacrylamide gels and Western blots (using a monoclonal antibody that reacts with a vector-derived epitope at the amino terminus of fusion protein) showed distinct bands at the predicted molecular masses (47 kDa for PfGC␣ C1 , 32 kDa for PfGC␣ C2 , 34 kDa for PfGC␤ C1 , and 28 kDa for PfGC␤ C2 ) of the fusion proteins (data not shown). The fusion proteins, purified by affinity chromatography on Ni 2ϩ -nitrilotriacetic acid Sepharose columns, were tested for AC or GC activities. Levels of cAMP/cGMP were measured by a competition assay using 3 H-labeled cAMP or cGMP (see "Experimental Procedures").
The purified fusion proteins were first assayed for AC activ-  (49), the C1 domains of both mammalian type V adenylyl cyclase (VC1, (50)) the Dictyostelium ACA (29), and the C2 domain of rat type II adenylyl cyclase (IIC2). The known secondary structure of the mammalian type V C1 domain (from a crystallographic study (36)) is indicated by thick lines (␣ helix) and arrows (␤ sheet) beneath the sequence. The alignment illustrates the reversal of the C1 and C2 domains of the Plasmodium proteins in terms of key functional residues compared with the C1 and C2 domains of mammalian adenylyl cyclases. The residues that determine purine binding specificity are marked with a plus symbol; green residues are indicative of adenine binding, and red residues are indicative of guanine binding. Other functionally important residues, determined by mutagenesis and structural studies (discussed in the text) are indicated and have been shown as follows. Light blue r, ribose binding; light blue p, phosphate binding; dark blue m, divalent cation binding. The amino acid residue numbers of each sequence are given on the right. The alignment was produced using the PILEUP program in the GCG package (51).
ity. No AC activity was obtained using the ␣C1, ␣C2, ␤C1, or ␤C2 proteins derived from these constructs either separately or in combination (Fig. 6). The catalytic domain of a receptor-type AC (ADC-1) from T. cruzi was purified and assayed in parallel.
This fusion protein gave a high level of AC activity which was maximized in the presence of Mn 2ϩ ions (34). We were also able to demonstrate functional complementation of the AC-deficient E. coli strain (TP610) with the trypanosome-derived construct,  1, 3, and 5) and 3D7A (lanes 2, 4, and 6). Electrophoresis conditions were as described in "Experimental Procedures." Lanes 1 and 2 show an ethidium bromide-stained agarose gel. After transfer to a nylon membrane the DNA was hybridized with PfGC␣ (lanes 3 and 4) or RESA-2 (lanes 5 and 6). RESA-2 has been localized previously to chromosome 11 (53). Similar results were obtained with the P. falciparum K1 strain-(data not shown). but this was not achievable with the Plasmodium-derived constructs (data not shown).
We then tested the fusion proteins for GC activity. Fig. 7A shows that the C1 and C2 domains from the PfGC␣ gene showed no significant activity over isolates from control bacteria either when tested individually or in combination with each other. However, the fusion proteins derived from the PfGC␤ gene showed moderate levels of activity when tested individually, with ␤C1 being somewhat more active than ␤C2. When used in combination, however, considerably elevated levels of activity were obtained (Fig. 7B). Following a brief lag period, cGMP synthesis increased rapidly over the first 10 min before decreasing. To test whether the transient nature of this activity was due to the low stability of the enzyme, we incubated the C1/C2 mixture at 30°C for 25 min prior to the assay; this resulted in complete loss of activity (Fig. 7C). Addition of forskolin, Ca 2ϩ , or Ca 2ϩ /calmodulin to the reaction mixtures did not affect the level of GC activity (data not shown). In all cases activity was dependent on the presence of Mn 2ϩ ions. If the divalent cation was replaced with Mg 2ϩ , no activity could be detected. Proteins from bacteria transformed with control pTrcHis plasmid showed no activity. DISCUSSION We have isolated two genes from P. falciparum that encode proteins with cyclase-like domains similar in structure to the membrane-localized adenylyl cyclases of higher eukaryotes. Unusually, the catalytic regions of at least one of these enzymes has guanylyl cyclase activity. During the preparation of this manuscript, a GC with this type of structure was also described in Paramecium (9). An additional remarkable feature found in both the Plasmodium and Paramecium proteins is that they each have an amino-terminal domain with high similarity to P-type ATPases in terms of sequence and predicted structure. GCs have hitherto fallen into two subclasses: (i) membrane-associated receptor molecules that form homodimers each with a single transmembrane domain and (ii) soluble molecules that form heterodimers. PfGC␣ and PfGC␤ therefore fall into a third subclass that is membrane-associated (with 12 putative membrane-spanning domains) with a pair of catalytic domains. As in mammalian ACs, their activity is probably mediated by interaction of the C1 and C2 domains.
Northern analysis of mRNA from the erythrocytic stages of P. falciparum also indicated that expression of both genes is restricted to gametocytes (Fig. 4). Immunoelectron microscopy localized PfGC␣ to the plasma/parasitophorous vacuole membrane region of this stage of the life cycle (Fig. 5). In red blood cells, malaria parasites are surrounded by three membranes: the host cell membrane, the parasitophorous vacuole membrane, and the parasite plasma membrane. The latter two membranes are closely opposed, and it is not clear from our electronmicrographs in which membrane PfGC␣ is located (Fig.  5). However, it can be inferred from the sequence and by comparison with other cyclases that both PfGC␣ and PfGC␤ are integral membrane proteins. The parasites used for the immunoelectron microscopy study had been stimulated to undergo gametogenesis (see "Experimental Procedures"). The micrographs show parasites at various stages of emergence from the red blood cell and parasitophorous vacuole membranes. No labeling is associated with extracellular gametes, suggesting that PfGC␣ has a pre-emergence role.
Structural Comparison of the PfGC␣ and PfGC␤ Catalytic Domains with Those of Mammalian Adenylyl Cyclases-The active site of mammalian AC is formed by the interaction of the C1 and C2 domains (35). Recent crystallographic and mutagenesis studies have given insight into the nature of the catalytic site and have identified residues involved in ATP binding (3, 25, 36 -38). Based on these studies, models for the analogous binding of GTP within the active site of GC have been derived. From sequence alignment, it can be inferred that the cyclase catalytic domains of both PfGC␣ and PfGC␤ are different from those of mammalian G protein-dependent ACs in that the C1 and C2 domains appear to be structurally reversed. Several catalytically important residues that are present in the mammalian C2 domain are found only in the C1 domain of the Plasmodium molecules (Fig. 2). Likewise, mechanistically important residues present in the C1 domain of the mammalian enzyme are located in the C2 domains of PfGC␣ and PfGC␤. For example, in mammalian AC it has been predicted that Arg 398 (type I AC), which approaches the ␤ phosphate of ATP, and Arg 1011 , which interacts with the ␣ phosphate, are localized in the C1 and C2 domains, respectively (36,39). The equivalent residues in PfGC␣ (Arg 4070 and Arg 3268 ) and PfGC␤ (Arg 3063 and Arg 1696 ) are reversed with respect to their location ( Fig. 2). Similarly, three residues in mammalian AC that interact with the ribose moiety of ATP, Thr 315 (type I C1), Asp 354 (type I C1), and Asn 1007 (type I C2), are localized in the opposite catalytic domain in the Plasmodium enzymes (Fig. 2). The equivalent positions in PfGC␣ are Thr 3973 (C2), Asp 4019 (C2), and Asn 3264 (C1), and in PfGC␤ they are Thr 2973 (C2), Asp 3012 (C2), and Asn 1692 (C1). In type I mammalian AC, three negatively charged residues (Asp 354 , Asp 310 , and Glu 432 ) localized in the C1 domain are thought to be involved in Mg 2ϩ binding, an essential requirement for enzyme activity (Refs. 3 and 39 and Fig. 2). The corresponding negatively charged residues in PfGC␣ (Asp 4019 , Asp 3975 , and Glu 4104 ) and PfGC␤ (Asp 3012 , Asp 2968 , and Glu 3097 ) are all in the C2 domain.
Purine Binding Specificity-The highly conserved regions of mammalian-type ACs (C1a and C2a) are very similar to the corresponding regions of the catalytic domains of GCs. Mutagenesis and crystallographic studies have shown that a small number of residues are involved in conferring the purine binding specificity of the two enzymes (36,37,39,40). Two separate studies have demonstrated that the purine specificity of GC can be altered by changing key residues to their counterparts in AC (E925K and C997D of retinal guanylyl cyclase 1, retGC-1 (39); R592Q of soluble GC␣ 1 ; E473K and C541D of soluble GC␤ 1 (37)). This results in conversion to an enzyme with an AC activity. These studies indicate a direct interaction between Glu 925 of GC and the N1 and N2 positions of the guanine ring. In the C1 catalytic domains of both PfGC␣ and PfGC␤, the corresponding position is occupied by a glutamic acid residue (Glu 3058 and Glu 1591 , respectively) as would be expected in a GC, and not lysine as in AC. However, the Plasmodium enzymes are distinct from their mammalian counterparts in that residues in the PfGC␣ and PfGC␤ positions that correspond to the Cys 997 (retGC-1), thought to be important for guanine binding specificity, are both alanines (Ala 3257 and Ala 1785 , respectively; Fig. 2). In the Paramecium GC, the corresponding position is occupied by a serine (9). It has been proposed (36) that the glutamic acid residue (Glu 473 , GC␤) is probably the most important residue for determining GTP binding specific-FIG. 6. Adenylyl cyclase enzyme activity assays. The graphs show time courses of cAMP production by the recombinant catalytic domains derived from T. cruzi ADC-1 and PfGC␣ and PfGC␤. The data represent the means Ϯ S.E. of three independent experiments. 2-5 g of affinity purified His 6 -tagged fusion proteins were incubated at 30°C for the indicated time periods with 2 mM ATP and 4 mM Mn 2ϩ and assayed for cAMP as described under "Experimental Procedures." Control assays were performed on proteins purified from cultures containing plasmid (pTrcHisC, Invitrogen) with no insert. A, the catalytic domains of PfGC␣ (␣C1 and ␣C2) were assayed both separately and in combination (␣C1/␣C2). B, the catalytic domains of PfGC␤ (␤C1 and ␤C2) were assayed separately and in combination (␤C1/␤C2). C, the catalytic domain of ADC-1 derived from T. cruzi (34) was cloned into pTrcHisC, purified, assayed in parallel, and used as an AC positive control.
FIG. 7. Guanylyl cyclase enzyme activity assays. The graphs show time courses of cGMP production by the recombinant catalytic domains derived from PfGC␣ and PfGC␤. The data represent the means Ϯ S.E. of three independent experiments. 2-5 g of affinity purified His 6 -tagged fusion proteins were incubated at 30°C for the indicated time periods with 2 mM GTP and 4 mM Mn 2ϩ and assayed for cGMP as described under "Experimental Procedures." Control assays were performed on proteins purified from cultures containing plasmid (pTrcHisC, Invitrogen) with no insert. A, the catalytic domains of PfGC␣ (␣C1 and ␣C2) were assayed both separately and in combination (␣C1/␣C2). B, the catalytic domains of PfGC␤ (␤C1 and ␤C2) were assayed separately and in combination (␤C1/␤C2). C, a further time course experiment showing the effect of preincubation of ␤C1/␤C2 at 30°C for 25 min.
ity. This was inferred from studies in which mutated cyclases lacking a glutamic acid at this position were shown to have no GC activity. Modelling studies (36,37,39) have also indicated that an arginine residue (Arg 539 , GC␤, and Arg 592 , GC␣) is important for GTP binding. The arginine residue at this position is conserved in GCs, and the corresponding positions in the C2 domains of both PfGC␣ and PfGC␤ are also occupied by arginines (Arg 4089 and Arg 3082 , respectively).
Enzyme Activity of PfGC␣ and PfGC␤-Recombinant proteins expressed in E. coli corresponding to the C1 and C2 catalytic domains of PfGC␣ and PfGC␤ were tested for both AC and GC activity. No AC activity was detectable in any of the four malaria-derived recombinant proteins. We were, however, able to demonstrate GC activity in the C1 and C2 domains of PfGC␤ both separately and in combination. The PfGC␤ C1/C2 mixture showed the highest level of cGMP synthesis. No GC activity was detectable in the C1 or C2 domains of PfGC␣ either separately or in combination. It is not clear why activity was undetectable with PfGC␣, but it is possible that in the case of the C1 domain, the asparagine-rich insert that interrupts the domain may interfere with catalysis in the context of this construct. When the respective C1 and C2 domains of the PfGC␣ and PfGC␤ were mixed with their heterologous partner (␣C1/␤C2 and ␤C1/␣C2), there was evidence of some inhibition of activity of the PfGC␤ domains (data not shown). A similar phenomenon of inhibition between two ACs from Leishmania donovani has been reported (41). In this case activity could also be detected in only one of the cyclases.
In the present study, we have demonstrated that although the C1 and C2 domains of PfGC␤ synergize to synthesize cGMP, the individual domains also display activity (Fig. 7). Basal levels of AC activity have been reported in both the C1 and C2 catalytic domains of mammalian AC in the absence of specific activators (35). By analogy, the GC activity that we have demonstrated in this study probably represents only basal activity, because no activator has yet been identified. It is also possible that other parts of the PfGC␣ and PfGC␤ proteins, not included in the expressed fusion proteins, may have a regulatory role and could be required for optimal activity.
The Association of Ca 2ϩ and Guanylyl Cyclase Activity-Of the P-type ATPases, it is the Ca 2ϩ pump isoforms that have the highest similarity with the amino-terminal domains of PfGC␣ and PfGC␤. This suggests the possibility that Ca 2ϩ may be involved in the regulation of GC activity in P. falciparum. Ca 2ϩ -dependent GC activity has been demonstrated in a number of systems. For example, in Paramecium there is a causal association between Ca 2ϩ influx and an increase in intracellular cGMP levels (reviewed in Ref. 42). This Ca 2ϩ -dependent cGMP production is vital to the control of ciliary movement and locomotion. In mammals, retinal GC activity, involved in light activation of photoreceptors, is closely associated with Ca 2ϩ levels, and it has been shown that specific calcium-binding proteins regulate this activity (Ref. 43; reviewed in Ref. 6). The intimate association of GC activity with a putative Ca 2ϩ pump in PfGC␣ and PfGC␤ is intriguing in light of some earlier work on Plasmodium (2). This has implicated both Ca 2ϩ and cGMP in the control of exflagellation. Ca 2ϩ antagonists (but not Ca 2ϩ channel inhibitors) strongly inhibit exflagellation, whereas it is enhanced by cGMP and agents that increase levels of cGMP. It will therefore be of considerable interest to investigate whether the linkage of GC activity with a putative Ca 2ϩ pump ATPase in a single polypeptide is of functional significance.
Evolutionary Significance-It can be presumed that the "progenitor PfGC" protein arose from the fusion of two genes, one encoding a P-type ATPase and the second encoding a cyclase. This is not without precedence. There are at least two other examples in Plasmodium of bifunctional proteins that occur as distinct molecules in other organisms. These are the dihydrofolate reductase-thymidylate synthase (44,45) and dihydropteroate synthetase/pyrophosphokinase (46,47). The most likely scenario is that the initial fusion event was followed by gene duplication creating two copies that diverged to become PfGC␣ and PfGC␤. The fusion event appears to have occurred in a common ancestor of Plasmodium and Paramecium. This is consistent with classification of the ciliates with apicomplexans and dinoflagellates in the Alveolata (9,48).
An additional feature of the two Plasmodium genes is the presence of 12 introns in the ATPase encoding domain of PfGC␤, compared with none in the corresponding region of PfGC␣ (Fig. 1). From a functional viewpoint, it is not immediately obvious why PfGC␣ and PfGC␤, two proteins with the same potential activity, should be expressed during the sexual stages of the life cycle. One possibility is that they are expressed differentially during this complex life cycle phase. Another is that they are co-expressed but that they respond to different activators. A greater understanding of this signaling pathway and the manner in which it is regulated is of importance, particularly in the context of novel chemotherapeutic targets. The structural differences that we have identified between Plasmodium GC and its mammalian counterparts may be exploitable for the development of drugs capable of blocking transmission of malaria. The recent development of genetic manipulation techniques applicable to malaria parasites will allow this possibility to be explored further.