Multiple Splice Variants Encode a Novel Adenylyl Cyclase of Possible Plastid Origin Expressed in the Sexual Stage of the Malaria Parasite Plasmodium falciparum

We report the characterization of an unusual adenylyl cyclase gene from Plasmodium falciparum , here designated PfAC (cid:1) . The level of mRNA expression is maximum during development of gametocytes (the sexual blood stage of the parasite life cycle). The gene is highly inter-rupted by 22 introns, and reverse transcriptase-PCR analysis revealed that there are multiple mRNA splice variants. One intron has three alternative 3 (cid:1) -splice sites that confer the potential to encode distinct forms of the enzyme using alternative start codons. Deduced amino acid sequences predict membrane-spanning regions, the number of which can vary between two and six depend-ing on the splice variant. Expression of a synthetic form of two of these variants in Xenopus oocytes and in Dictyostelium adenylyl cyclase-deficient mutants, confirms that PfAC (cid:1) is a functional adenylyl cyclase. These results identify a novel mechanism in P. falciparum

Adenylyl cyclase (AC) 1 catalyzes synthesis of the signaling molecule cAMP from ATP. The most widely studied form of this enzyme, found in mammals and all higher eukaryotes, has a tandem pair of catalytic domains each preceded by a set of 6 transmembrane domains. The enzyme is activated via heterotrimeric G proteins after binding of an extracellular ligand to a G protein-coupled receptor (1). A soluble form of AC (also containing a pair of catalytic domains) has also been identified in mammals (2). It is G protein-independent and activated by bicarbonate ions (3). In lower eukaryotes and prokaryotes, the overall architecture of ACs varies between species. In all class III enzymes, however (the universal class of purine nucleotide cyclases, which includes guanylyl cyclases (GCs); Ref. 4) there are a number of conserved amino acid residues within the catalytic domain that are vital for enzyme activity (5). Receptor-type ACs have been identified that possess a single transmembrane domain and a single catalytic domain (reminiscent of mammalian GCs) and are activated directly rather than by heterotrimeric G proteins (e.g. in trypanosomes, Refs. 6 and 7). Many of the bacterial ACs have a modular structure with distinct non-catalytic regulatory domains. The filamentous cyanobacterium Anabaena cylindrica has an unusual AC (cyaB1) with two membrane-spanning segments. The enzyme also has additional functional regions including a PAS domain (a ubiquitous small molecule receptor) and an allosteric cyclic nucleotide-binding domain (8,9).
In mammals, the membrane-associated ACs (types I-IX) have roles in numerous biological processes including glycogen metabolism (10), olfaction (11), and nerve cell communication (12). Some bacterial pathogen exotoxins comprise ACs (e.g. from Bacillus anthracis, the causative agent of anthrax; Ref. 13). In lower eukaryotes, the list of confirmed physiological roles for AC is growing, and cAMP levels are known to control development and differentiation in many species. In the protozoan Dictyostelium discoideum, cAMP signaling is extremely complex. One of the ACs (ACA) controls chemotactic aggregation of amoebae under starvation conditions (14), and another (ACG) regulates spore germination (15). In the fission yeast Schizosaccharomyces pombe, AC is thought to be important in regulating sexual development (16). In the ciliate Paramecium, synthesis of cyclic nucleotides is coupled with ion currents (17); AC activity is associated with the ciliary membranes and is involved in locomotion. Synthesis of cAMP is stimulated by hyperpolarization of ciliary membranes, and this is inhibited by K ϩ channel blockers suggestive of a single unit acting as both a cyclase and an ion channel (18).
In the human malaria parasite Plasmodium falciparum, cyclic nucleotide signaling pathways have been implicated in sexual differentiation. Sexual blood stage parasites mediate disease transmission via Anopheles mosquitoes (whereas asexual blood stage parasites are responsible for disease pathology). Studies with phosphodiesterase inhibitors (19) have suggested that cGMP levels enhance a key step in male gametogenesis (exflagellation). Recently we have characterized two unusual bifunctional, membrane-bound GCs that are expressed during the sexual development of P. falciparum (20). We have also measured GC activity in mature gametocyte membranes, which is enhanced by the addition of a mosquito-derived factor (xanthurenic acid) that stimulates exflagellation (21). Several studies have provided evidence that the cAMP signaling pathway might be involved in development of the sexual blood stage of the life cycle (gametocytogenesis). Addition of cAMP to P. falciparum cultures at high parasitaemias led to greatly increased levels of gametocytes (22). In another study, basal levels of AC activity were found to be equivalent in a gametocyte producer and non-producer strain; however, cAMP-dependent histone II-A kinase activity was significantly higher in the gametocyte producers (23). AC activity in P. falciparum has been found to be distinct from that of the human red blood cell (24), although the enzymes responsible have not been characterized. Here we describe the identification of a gene encoding an unusual, functional AC from P. falciparum that may have been acquired by lateral gene transfer early in the evolution of this parasite.
Gametocytes were Percoll-purified using a modification of the method described by Carter et al. (27). Percoll step gradients (30,45,54,60, and 90% diluted with RPMI 1640) were prepared in 15-ml polypropylene centrifuge tubes. To harvest late gametocytes for exflagellation the cultures were pelleted and layered onto Percoll gradients made up in suspended activation medium (SAM) containing RPMI 1640, 25 mM Hepes, 1 l CaCl 2 , and 1% decarbonated serum, pH 7.4 (28) and washed three times in warm (37°C) SAM. After washing, an equal volume of cold (4°C) 100% human serum was added, giving about a 100 l per pellet in total. They were left at room temperature until a sample spotted out onto a slide showed that exflagellation was underway. This normally occurred after 15 min. After 20 min 1 ml of Tri reagent (Sigma) was added, and they were stored at Ϫ80°C until processing. Asexual blood stage parasites were synchronized by sorbitol treatment (29).
RNA Isolation and Northern Blotting-The RNA was resuspended in deionized formamide, gel-fractionated, and Northern blotted according to standard procedures (30,31). Blots (BrightStar TM Plus Nylon membrane, Ambion) were hybridized overnight with radiolabeled probes and visualized using phosphor screens (Kodak) scanned on a Storm Phosphorimager (Amersham Biosciences).
In Vitro mRNA Expression of a Synthetic PfAC␣-A DNA oligomer corresponding to variant 3 (Met 1 -Leu 895 ) of the PfAC␣ sequence was synthesized commercially (Bionexus Inc.) according to the preferred codon usage of Xenopus laevis, cloned into the pCR-Blunt vector (Invitrogen), and sequenced using an ABI 377 automated sequencer. The synthetic gene was then cloned (a BglII site had been incorporated at both ends) into an X. laevis expression vector pSP64T (incorporating a 5Ј strong Kozak consensus (CACC)), which contains 5Ј-and 3Ј-untranslated X. laevis ␤-globulin sequences (32). The plasmid construct (with the confirmed orientation) was linearized with SmaI, treated with proteinase K, phenol/chloroform extracted, and precipitated with sodium acetate. Capped PfAC␣ cRNA was transcribed using the mMES-SAGE mMACHINE SP6 (Ambion, Austin, TX). After completion of the reaction, template DNA was removed using DNase I (37°C, 15 min). RNA was precipitated with LiCl, washed with 70% ethanol, dried, and resuspended in DEPC-treated water.
Expression in X. laevis Oocytes and Measurements of cAMP Levels-X. laevis oocytes were harvested, and connective tissue removed with collagenase treatment (2 mg ml Ϫ1 for 2 h on a shaker) (33). Stages V to VI oocytes (Dumont, 1972) were selected and microinjected with cRNA (5-35 ng) encoding PfAC␣ or with a comparable amount of DEPC-treated water (ϳ30 nl). Oocytes were incubated in Barth's solution (33) at 19°C for 3 days. Quantification of in vivo accumulation of cAMP in oocytes was performed using an immunoassay kit (Alexis) according to the manufacturer's instructions. Briefly, 15 oocytes were selected in triplicate and homogenized in a microcentrifuge tube containing 270 l of 10% trichloroacetic acid. The homogenate was then extracted three times with 5 volumes of water-saturated ether. Residual ether was removed by heating to 70°C for 20 min, and the samples were stored at Ϫ80°C until the assay was performed.
Expression in Dictyostelium ACA-/ACG-null Mutants-Synthetic forms of variants 1 and 3 of PfAC␣ were expressed as yellow fluorescent protein (YFP) fusions to facilitate detection and localization by Western blotting. The open reading frame (ORF) of variants 1 and 3 were amplified from the pCR-Blunt vector using oligonucleotides incorporating a BamHI and XhoI site. The BamHI/XhoI-digested fragment was cloned into the similarly digested vector pB17S, which placed the PfAC␣ variants downstream of the constitutive actin15 promoter and in-frame with the YFP ORF at the C terminus, to generate vectors PfAC␣1Y and PfAC␣3Y. An AC-deficient Dictyostelium cell line (aca Ϫ / acg Ϫ ) was transformed with PfAC␣1Y and PfAC␣3Y by electroporation, and the transformants were selected by growth in HL5 medium in the presence of 100 g ml Ϫ1 of G418 (34).
Western Analysis-Dictyostelium cells were resuspended to 2 ϫ 10 7 cells ml Ϫ1 in 10 mM potassium phosphate buffer, pH 6.2 (KK2), and the lysates size-fractionated on 10% polyacrylamide gels. The proteins were transferred to nitrocellulose membranes, which were incubated overnight at 4°C with a 1:1000 diluted anti-GFP mouse monoclonal antibody (Roche Applied Science). Detection was performed with the Supersignal chemiluminescence kit (Pierce) according to the manufacturer's instructions, using 1:20,000 diluted horseradish peroxidaseconjugated goat anti-mouse IgG (Promega) as secondary antibody.
cAMP Accumulation in Intact Cells-Cells were harvested from growth medium, resuspended in KK2 buffer at a concentration of 5 ϫ 10 7 cells ml Ϫ1 , and shaken at 120 rpm for 10 min at 22°C. Aliquots (25 l) of cell suspension were made up to a final volume of 30 l in microtiter plate wells. Reactions were initiated by addition of 5 mM dithiothreitol (DTT) and terminated by addition of 30 l of 3.5% perchloric acid. Lysates were neutralized with KHCO 3 , and cAMP levels were determined using the isotope dilution assay (35).
Adenylyl Cyclase Assay in Cell Lysates-Cells were resuspended at a concentration of 5 ϫ 10 7 cells ml Ϫ1 in ice-cold lysis buffer (250 mM sucrose in 10 mM Tris-Cl, pH 8.0) and lysed through nucleopore filters (pore size, 3 m). Aliquots (10 l) of cell lysate were added to 5 l of divalent cation solution (Mg 2ϩ , Mn 2ϩ , or water) and incubated at 0°C for 5 min. Reactions were initiated by addition of 5 l of assay mixture (2 mM ATP, 0.8 mM 3-isobutyl-1-methylxanthine (IBMX) and 40 mM DTT in lysis buffer), and the samples were transferred to a 22°C water bath. The reaction was terminated by adding 10 l of 0.4 M EDTA, pH 8.0 and by boiling the samples for 1 min (36). cAMP levels were assayed as above.
Phylogenetic Analysis-Searches were performed using the WU-Blast2 Tool (www.ebi.ac.uk) to identify proteins in the SWISS-PROT Protein Knowledgebase with high sequence identity to either the single catalytic domain of PfAC␣ or the proposed C2 domain of PfAC␤. An alignment was compiled of the catalytic domains from representative prokaryote and eukaryote AC or GC sequences. Where proteins contained paired catalytic domains (e.g. PfAC␤), only the C2 domain was included in the alignment. An initial alignment was performed using the MegAlign program (DNAstar), which was modified manually. Phylogenetic trees were constructed using the Fitch-Margoliash and Least-Squares Distance Methods program FITCH, after protein distances had been calculated using PROTDIST. Node robustness was assessed by 100 bootstrap replications (data set for bootstrap analysis was generated with SEQBOOT). PROTDIST, FITCH and SEQBOOT are part of the PHYLIP package (37). Maximum parsimony (PROTPARS), and neighbor joining (NEIGHBOR) methods were also used to generate phylogenies to confirm the results obtained by the above method.

Identification of a Gene Encoding a Putative AC and the
Presence of mRNA Splice Variants-Conserved motifs within the catalytic domain of ACs from diverse eukaryotic species were used in BLAST searches of P. falciparum preliminary sequence data (Sanger Centre, TIGR, and Stanford University). Initially a single match was found in the chromosome 14 data base (TIGR). However, it was clear that the various motifs were non-contiguous and that the sequence was highly interrupted by introns. RT-PCR was performed on total RNA extracted from mixed blood stage parasites (clone 3D7, see "Experimental Procedures") to determine the number and position of introns in the sequence. Examination of the upstream and downstream sequence data (TIGR chromosome 14 shotgun, unfinished sequence data) for potential splice sites and extensive RT-PCR analysis revealed a total of 22 exons and 21 introns (Fig. 1) predicting a protein size of 81.2 kDa. Initial RT-PCR analysis indicated that the first in-frame start codon is in exon 5. This suggests that introns 1-4 are situated in the 5Ј-untranslated region of the gene. However, further RT-PCR analysis of the 5Ј-region revealed the presence of mRNA molecules corresponding to additional mRNA splice variants. The first in-frame stop codon is in exon 22 indicating the predicted position of the C terminus of the protein.
The majority of the RT-PCR products examined conformed to the above arrangement (variant 1). However, further analysis demonstrated that intron 3 contains alternative 3Ј-splice sites ( Fig. 1B and Fig. 2A). In two instances this results in two small introns (rather than a single large one) and an additional exon (exon 3A, Fig. 1, A and B). Also, there are two versions of the small intron 3 that give rise to different deduced N-terminal protein sequences (variants 2 and 3, Fig. 1, A and B). Intriguingly, in variant 3 all exons are contiguous resulting in a full-length protein of 108 kDa. Variant 2 predicts a protein of intermediate size (98.2 kDa, Fig. 1C). As shown in Fig. 1C, the protein sequences encoded by variants 1, 2, and 3 have differing numbers of predicted transmembrane domains (2, 5, and 6, respectively). The nucleotide sequences toward the 5Ј-end of variants 1-3 (near to their predicted translation start codons) obtained by sequencing RT-PCR products are shown in Fig. 2A. The precise positions of the introns and the variant intron junctions are shown in relation to the exons. Fig. 2B shows the differences in N-terminal amino acid sequence encoded by the three variants.
Relationship of PfAC␣ to a Second AC from P. falciparum-The majority of the single catalytic domain of PfAC␣ is encoded by exons 16 and 17 toward the C terminus. Surprisingly this sequence is most closely related to an AC from the cyanobacterium Trichodesmium erythraeum (28% identity and 49% similarity over 224 amino acids) as determined by a BLAST search (38). In addition to PfAC␣, we have also identified a second putative AC (here designated PfAC␤) sequence on chromosome 8 in the P. falciparum genome data (plasmodb.org/). PfAC␤ also has a high degree of relatedness to cyanobacterial ACs. However, it has a double catalytic domain (C1 and C2) at the N terminus, which is topologically more similar to the soluble AC of mammals than to the membrane-bound forms. Fig. 3 shows an alignment of the catalytic domains of PfAC␣ and PfAC␤ (C2) with those of the C1 (type V) and C2 (type II) domains of mammalian AC for which a crystal structure of the heterodimer is available (39,40). The initial alignment was performed using the ClustalW program (41) but was modified manually to introduce gaps (where possible) between the areas of secondary structure of the mammalian ACs. The catalytic domain of the homodimeric, human retinal GC is included in the alignment to emphasize the diagnostic residues that distinguish ACs from GCs. The catalytic domain of a cyanobacterial cyclase (T. erythraeum), and the C2 catalytic domain of the human soluble AC are also included for comparison. The majority of the residues involved in catalysis in both ACs and GCs are conserved in the P. falciparum homologues (see legend to Fig. 3). The critical residue, however, which determines purine specificity in nucleotide cyclases, is a lysine in both PfAC␣ (Lys 576 ) and PfAC␤ (Lys 164 ) and in all known ACs. A glutamic acid residue is present at the equivalent position in all GCs. The sequence data therefore suggest strongly that both PfAC␣ and PfAC␤ are ACs. Remarkably, a highly conserved aspartic acid residue in the purine-binding pocket (see Fig. 3) of ACs is replaced by a serine or a threonine residue in PfAC␣ and PfAC␤, respectively. This feature is characteristic of prokaryotic forms of AC (see below).
Phylogenetic Analysis of PfAC␣ and PfAC␤-Searches were performed with the catalytic domains of both PfAC␣ and PfAC␤ to determine the most closely related sequences in the data bases. Although both sequences show an unexpectedly high sequence similarity with certain bacterial ACs, it is clear that the two Plasmodium proteins are not closely related to each other. PfAC␣ has only a single catalytic domain containing all the residues required for enzyme activity, whereas PfAC␤ has two catalytic domains (C1 and C2) each containing specific motifs that would be required for activity. This profound difference between the catalytic domains is reflected in a shared sequence identity of only 19% (using the C2 domain of PfAC␤) and 22% (using the C1 domain of PfAC␤) at the amino acid level. This is reinforced by the grouping of the proteins into two distinct lineages of ACs in a phylogenetic analysis (Fig. 4A). In this analysis of the catalytic domains, PfAC␣ clusters strongly with putative ACs from other apicomplexan species (bootstrap value of 100%) but also with ACs from prokaryotes including cyanobacteria, spirochaetes and certain proteobacterial species (bootstrap value 50%). By contrast PfAC␤ groups strongly with all known examples of soluble ACs (bootstrap value of 100%). This grouping of PfAC␤ reflects a shared double catalytic domain. The group of soluble ACs includes the mammalian bicarbonate sensor (3) together with hypothetical proteins from the mosquito Anopheles gambiae, Chloroflexus aurantiacus (a green non-sulfur bacteria), and the Dictyostelium sGCA, which is in fact a GC (42). The phylogenetic analysis also shows that this class of cyclases, despite having a double catalytic domain, is quite distinct from the membrane-bound G protein-dependent ACs.
A key amino acid residue in the active site, which determines substrate specificity is also indicated in Fig. 4A. Membranebound G protein-dependent ACs all have an aspartic acid residue in this position, whereas ACs in the 2 prokaryotic clades have a serine/threonine residue. Analysis suggests that this is a defining feature of these 2 classes of cyclases (discussed further below). The alanine residue in the Dictyostelium sequence at this position probably reflects that it is a functional GC. Fig. 4B shows an alignment of ACs from various species showing the two non-contiguous segments of the enzyme active site that are involved in purine binding, and highlighting the residues that define substrate specificity. In all cases the crucial lysine residue (replaced by arginine in Chloroflexus) is present, but the key aspartic acid residue invariant in all membrane-bound G protein-dependent ACs is replaced by a serine or threonine residue in both the soluble ACs (with the paired catalytic domains C1 and C2, like PfAC␤) and the ACs with a single catalytic domain (like PfAC␣). This substitution may define a distinct mechanistic feature shared by these forms of AC.

Expression of a Synthetic PfAC␣ in Xenopus Oocytes and Dictyostelium AC-deficient Mutants Demonstrate That It Is a
Functional AC-P. falciparum proteins are notoriously difficult to express in heterologous systems, and problems have been overcome in some cases by changing the extremely A/T-rich codon bias (43,44) by in vitro resynthesis of the gene. The entire coding region of the full-length PfAC␣ (variant 3; accession number AY191005) was resynthesized according to the codon bias of X. laevis to facilitate functional expression in oocytes. This system has been used successfully to express P. falciparum proteins with multiple transmembrane domains (33). The verified sequence was then cloned into the X. laevis expression plasmid (pSP64T). RNA was synthesized by in vitro transcription under the control of the SP6 promoter (see "Experimental Procedures"). Oocytes were injected with either DEPC-treated distilled water, mRNA encoding full-length PfAC␣ (variant 3) or a negative control protein PfHT, a hexose transporter (33). After incubation of oocytes for 3 days, in vivo accumulation of cAMP was measured. Fig. 5 shows a representative experiment indicating a ϳ3-fold increase in cAMP levels in oocytes that had been injected with mRNA encoding PfAC␣ compared with the negative controls (p Ͻ 0.008 for PfAC␣ compared with either PfHT or DEPC, p ϭ 0.97 PfHT compared with DEPC, Student's t test). Each experiment (performed three times in triplicate) resulted in a 2.5-3-fold increase in cAMP levels. These results therefore demonstrate that PfAC␣ can catalyze the synthesis of cAMP in the Xenopus oocyte system and indicate that it is a functional AC.
An additional study was initiated in parallel to express variants 1 and 3 in the protozoan D. discoideum (Fig. 6). A mutant cell line in which the endogenous cyclases ACA and ACG had been deleted (aca Ϫ /acg Ϫ ) was used in these experiments (34). The other Dictyostelium adenylyl cyclase, ACB, shows almost no activity during early development (36). PfAC␣ fragments were expressed as YFP fusions (see "Experimental Procedures"). To confirm that the fusion proteins were of the expected size, lysates of transformed cells were separated by SDS-PAGE, immunoblotted with anti-GFP antibodies (that also detect YFP). Bands of ϳ110 and 130 kDa were detected in cells transformed with variant 1 and variant 3, respectively (Fig. 6A). This agrees well with the predicted sizes of the variant proteins (81 and 108 kDa) plus the YFP component (27 kDa). Dictyostelium cells rapidly secrete most of the cAMP that they produce and cAMP production by intact cells can be monitored by adding DTT, an inhibitor of the extracellular phosphodiesterase, to the medium. PfAC␣, variant 1 is structurally homologous to Dictyostelium ACG, which is stimulated by high osmolarity. We therefore tested cAMP production by variants 1 and 3 in the presence and absence of high osmolarity provided by 200 milliosmolar sorbitol. Both variant 1 and variant 3 transformed cells accumulated significant amounts of cAMP over a 30-min period, whereas the untransformed aca Ϫ /acg Ϫ  (39). The secondary structure of the C1 domain is shown below the sequence in green (arrows represent ␤-sheets and bars ␣-helices). An AC from the cyanobacterium T. erythraeum (Trich, ZP_00074939) is included in the alignment as it is currently the most closely related sequence available in the data base. The C2 catalytic domain of the human soluble AC (HsC2, NP_060887) is also included as it is related to PfGC␤. A human retinal GC (GC, NP_000171) highlights the residues diagnostic of AC (green) and GC (red) cyclases. Apart from purine specificity(s), other functional residues (determined by mutagenesis and crystallography studies) are indicated above or below the appropriate mammalian AC sequence: Mg 2ϩ -coordination (m), phosphate-binding (p), ribose-binding (r), adeninebinding (a), and maintenance of the catalytic transition state (t). To emphasize conserved areas, the sequences are highlighted in gray where at least five of the sequences are identical at that position. The sequences are numbered both to the left and right. parent cell line accumulated no cAMP (Fig. 6B). Sorbitol reduced rather than stimulated cAMP production, which indicates that neither variant is activated by high osmolarity. The cAMP production by variant 3 was somewhat higher than variant 1, but this may reflect different expression levels of the constructs, rather than differences in intrinsic catalytic activity.
AC activity was also measured directly in cell lysates provided with ATP and different concentrations of Mg 2ϩ and Mn 2ϩ ions. Again, no activity could be detected in the aca Ϫ /acg Ϫ parent. Both variant 1 and variant 3 showed highest activity in the presence of Mn 2ϩ ions, with maximum activity at 1 and 3 mM Mn 2ϩ , respectively (Fig. 6C). There was very little activity with Mg 2ϩ as a cofactor. This demonstrates that the measured activity cannot be caused by ACB, which shows highest activity with Mg 2ϩ as a cofactor (34).
PfAC␣ mRNA Expression Is Maximum in the Sexual Stage of the P. falciparum Life Cycle- Fig. 7 shows the results of a Northern blot containing total RNA preparations from sorbitolsynchronized asexual blood stage parasites and GlcNAc-synchronized sexual blood stage parasites. Panel A shows the ethidium bromide-stained agarose gel from which the Northern blot was derived. The blot was hybridized with a probe corresponding to the catalytic domain of PfAC␣ (Fig. 7B). A transcript of ϳ4,500 nucleotides was detected in the total RNA preparations from sexual blood stage parasites and was maximal during early/mid-stage gametocyte development (stage II-III: 2-3 and 3-5 days old, respectively). The blot was reprobed with Pfs16, a highly expressed sexual stage-specific gene (45), to test for the presence of contaminating sexual stage parasites in the asexual preparations (Fig. 7C). This indicated that the asexual stage preparations also contain low levels of gametocytes, which could account for faint bands in these tracks when probed with PfAC␣. (39). The presence of a serine/threonine in this position in the cyanobacterial enzyme (9) is thought to play the same role but in the context of the single amino acid insertion that occurs in this region. The species (accession number in brackets) from which the cyclase sequences derive used in this alignment are abbreviated as follows: A. cylindrica, An (BAA13998); T. erythraeum, Trich (ZP_00074939); P. falciparum, Pfa (AY191005) and PfbC2 (NP_704518); Chloroflexus aurantiacus, ChlC2 (ZP_00018205); A. gambiae, AgC2 (EAA10271); human soluble AC, HsC2 (NP_060887); D. discoideum, DsgcC2 (AAK92097); and human G protein-dependent AC, HIIC2 (Q08462). In those cyclases with a pair of catalytic domains, the C2 domain is used in the alignment and indicated in the abbreviation. and have an aspartic acid residue in the crucial position, which determines purine binding specificity. By contrast, PfAC␣ (shown in bold) groups strongly with cyclases from other apicomplexans, but also with prokaryotic cyclases, which have a serine/threonine residue at this position (bootstrap value of 50%). The corresponding residue at this position is shown to the right of each species. The soluble ACs form a third distinct group (bootstrap value of 100%), which contains the PfAC␤ protein (shown in bold). Accession numbers for the proteins used in this analysis are as follows: Mycobacterium, NP_338294; P. falciparum ␤, NP_704518; Chloroflexus aurantiacus, ZP_00018205; D. discoideum soluble GC (Sgc), AAK92097; A. gambiae soluble adenylyl cyclase (S), EEA10271; Rattus norvegicus soluble AC (S), AF081941; human soluble adenylyl cyclase (S), NP_060887; Takifugu rubripes VI, AAB96362; C. elegans, NP_504553; Rattus norvegicus II, P26769; human type IX, AAC24201; X. laevis, CAA87082; Mus musculus I, AAC29478; Mus musculus VI, AAA37182; D. melanogaster, NP_511156; A. cylindrica, P43524; D. discoideum, AAD50121; Spirulina platensis, BAA22996; Legionella pneumophila, AAM00644; Magnetococcus, ZP_00042813; T. erythraeum ZP_00074939; Sinorhizobium melioti, NP_438008; Nostoc punctiforme, ZP_00108886; P. falciparum ␣, AY191005; Toxoplasma gondii (from ToxoDB) TGG_7615-1-12238-13605; Eimeria tenella (from The Sanger Centre) eimer-2574h12.qlk and Leptospira interrogans, NP_712717; Treponema pallidum, NP_218926. B, an alignment of the sequences in the 2 regions of the catalytic domain, which have been shown to determine purine binding specificity of ACs and GCs. The secondary structure of the rat C2 domain derived from a crystal structure (40) is shown below the sequence in orange. The invariant lysine is colored green and marked beneath with an asterisk. The aspartic acid residue of conventional eukaryotic ACs or serine/threonine residues of the cyanobacterial AC are colored in purple and light blue, respectively and marked below with a ϩ. The associated single amino acid insertion in this region in the latter form is indicated by a OE symbol. The aspartic acid residue acts as a hydrogen acceptor interacting with the N 6 atom of the adenine ring

Identification of Two Genes Encoding Putative P. falciparum
ACs-Searches for an AC gene in annotations of the recently completed P. falciparum genome sequence gave no matches (46). However, our own searches have revealed the presence of two distinct AC genes. The biochemical activity of one of these (PfAC␣) has now been verified and is the main focus of this study. The large number of introns in PfAC␣ made it particularly difficult to identify, and this has probably led to the lack of functional assignment. For example, the conserved cyclase motifs span four separate exons (exons [15][16][17][18]. The existence of PfAC␤ was probably obscured by the presence of an insert of ϳ120 amino acids, which interrupts the C1 catalytic domain. Otherwise the protein aligns well with known cyclases and contains all the necessary amino acid residues for enzyme activity. The position of this insert (between ␤3 and ␣3) corresponds to a shorter insert found in PfAC␣ and also in the two P. falciparum GCs (20). A similar insert that occurs in all kinetoplastid ACs (between ␣3 and ␤4) has been postulated to have a regulatory role, since it has been shown to bind DTT in vitro leading to activation of the enzyme (7).
The Occurrence of Splice Variants Encoding PfAC␣-More than 90% of the examined RT-PCR products conformed to variant 1 in which the first in-frame start codon occurs in exon 5. The presence of introns in the 5Ј-untranslated region has previously been reported in Plasmodium berghei. Here alternative splicing of a mRNA in distinct life cycle stages gives rise to two distinct transcripts, one of which has introns in the 5Ј-untranslated region (47). The role of introns in the 5Ј-untranslated region is not known. All three variants of PfAC␣ are predicted to be integral membrane proteins, consistent with our measurement of native AC activity in gametocyte membrane fractions (data not shown). Variant 1 potentially encodes a short form of the enzyme with just two transmembrane domains; a pattern found in, for example, the germinationspecific ACG of D. discoideum (an osmosensor; Ref. 15). However, unlike ACG, PfAC␣ is not activated by high osmolarity. Variants 2 and 3 are predicted to encode enzymes with five and six transmembrane domains, respectively. It is possible that the three variants of PfAC␣ might, through their differing predicted architectures, respond to different environmental signals each resulting in an appropriate change in intracellular levels of cAMP required for cellular function.
Comparison of the Catalytic Domain of PfAC␣ and PfAC␤ with Those of Other Cyclases-Crystallography (39,40), mutagenesis (5,48), and modeling studies (49) have determined the amino acid residues responsible for catalytic activity and substrate binding in both AC and GC. Furthermore it has also been shown that substrate specificity can be altered by substitution of two residues in the purine-binding pocket (50 -52). One of these positions is occupied by a lysine in all known ACs and a glutamic acid in all known GCs. In both PfAC␣ and PfAC␤ this position contains a lysine. The second position, which determines substrate specificity, is occupied by an invariant aspartic acid residue in all G protein-dependent ACs and most of the remaining eukaryotic ACs. This aspartate is not present in the Plasmodium ACs. Instead, PfAC␣ has a serine and PfAC␤ a threonine at this position. Prior to this study, this substitution had only been reported in the cyanobacterium Anabaena (9). A more extensive analysis of sequence data bases reveals that this unusual feature is shared with other bacterial species and some lower and indeed higher eukaryotes; for example, a Dictyostelium AC (rAC, Ref. 53), and the mammalian soluble AC (2). Both of these cyclases were previously noted to have an unexpected bacterial relationship, but this critical substitution, which we suggest defines these two prokaryotic classes of AC, was not identified. This substi-  (1), asexual schizont (2), early (3), mid (4), and late (5) stage gametocytes (see "Experimental Procedures") and mature gametocytes stimulated to undergo exflagellation (6). B, a Northern blot hybridized with a radiolabeled 1-kb fragment derived from cDNA corresponding to the AC catalytic domain (Gln 517 -Val 762 ) of PfAC␣. The position of the transcript (ϳ4.5 kb) is indicated with an arrow. The blot was derived from the gel shown in panel A. C, the same blot but stripped and re-hybridized with a probe derived from a gametocyte-specific gene (Pfs16). tution also correlates with the presence of a single amino acid insert in this highly conserved region. These two associated changes are thought to reflect the structural constraints required for substrate (ATP) binding in this unique, but functional form of the enzyme. This threonine residue in the Anabaena cyclase is essential for activity (9). The biochemical significance of this substitution is not known but a reduced affinity for ATP has been observed in the mammalian soluble AC (2). All other amino acid positions in the cyclase catalytic domain (39,40), which bind the substrate (adenine, ribose, and phosphate moieties), Mg 2ϩ ions, or are essential for catalysis (maintenance of the transition state) are conserved in the PfAC␣ and PfAC␤ sequences.
In G protein-dependent ACs, the catalytic site is formed by the interaction of a C1 and C2 heterodimer resulting in a double pocket, part of which is for ATP binding and the other structurally related part binds forskolin (a non-physiological activator). The motifs required for substrate binding and catalysis are contributed by the C1 and C2 domain, and therefore both are required for activity. In ACs from most lower organisms (e.g. Dictyostelium ACG, Ref. 14 and Trypanosoma cruzi AC Ref. 6) a homodimer is formed and gives rise to a pair of identical catalytic sites, which accommodate two substrate molecules (49). All the motifs required for substrate binding and catalytic activity are therefore present in a single catalytic domain, but a dimer is required to form two identical active sites. PfAC␣ has the latter conformation and is therefore likely to form a homodimer. PfAC␤ on the other hand, has two catalytic domains and is related to the mammalian soluble AC rather than the G protein-dependent isoforms.
Developmental Regulation of PfAC␣-Northern blot analysis indicates that blood stage expression of PfAC␣ is confined to the sexual forms (gametocytes) and is maximal at stage II-III (54) of gametocyte development. PfAC␣ transcripts were also detected in gametocytes, which had been stimulated to undergo gametogenesis, suggesting that cAMP may have a role in subsequent mosquito stages as well as gametocyte development in the human.
The cAMP signaling pathway has been implicated in the initiation of sexual commitment in P. falciparum (22,23). On the basis of temporal expression of mRNA, the present study suggests that PfAC␣ may have a role in sexual development rather than triggering differentiation because we did not detect expression in the asexual blood stages (from which the sexual stages arise). A previous study has reported native AC enzyme activity in P. falciparum blood stage parasite preparations (24). The properties of this enzyme were distinct from that of the host enzyme. For example, their experiments suggested that the enzyme was G protein-independent and showed a marked preference for Mn 2ϩ over Mg 2ϩ . These findings are consistent with both the predicted structure and properties of PfAC␣ expressed in D. discoideum.
Phylogenetic Relationships of P. falciparum ACs-There is a low level of relatedness between the two P. falciparum ACs indicating that they are unlikely to have arisen by a gene duplication event. PfAC␣ is most closely related to single domain ACs from certain cyanobacterial (e.g. Trichodesmium and Anabaena) and other bacterial species (e.g. spirochaetes). By contrast, the twin catalytic domain structure of PfAC␤ places it clearly in a novel class of soluble cyclases, which are themselves characterized by a high level of similarity with bacterial ACs. Prior to this study, these soluble cyclases had only been found in mammals and Dictyostelium. However a newly deposited sequence from the green non-sulfur bacteria Chloroflexus shows the highest levels of identity with PfAC␤. This type of soluble cyclase has not been found in other species including Drosophila melanogaster and Caenorhabditis elegans despite completion of genome sequencing projects, although intriguingly it has now been detected in the malaria parasite, its mammalian host, and its insect vector, A. gambiae.
The relationship of PfAC␣ to the ACs of certain bacteria, including cyanobacteria, was unexpected. It is therefore possible that PfAC␣ may be the product of lateral gene transfer, derived from the non-photosynthetic plastid that is present in Plasmodium (termed the apicoplast as it is present all apicoplexans). The apicoplast (a vestigial chloroplast) is thought to be derived from an ancestral symbiotic cyanobacterium (55)(56)(57). A number of chromosomally encoded proteins are subsequently targeted to the apicoplast and probably originated from this organelle (46). This hypothesis is supported by examination of unfinished sequence data from other apicomplexans (e.g. Toxoplasma; toxodb.org/ToxoDB.shtml and Eimeria; www.sanger.ac.uk); their genomes also appear to contain genes encoding this prokaryotic form of AC. It is also conceivable that the soluble lineage (which includes PfAC␤) found in mammals arose by lateral gene transfer. The alternative hypothesis is one of selective gene loss from various species including D. melanogaster, for which some evidence has been presented (58).
In summary, we have described the first functional AC in Plasmodium. Identification of this gene and demonstration of its biochemical activity will be an important step in investigating its biological role in gametocyte development. In addition, the discovery of multiple mRNA splice variants and a second, distantly related isoform provides insight into the complexity of this signaling pathway in the malaria parasite.