Crystal Structures Explain Functional Differences in the Two Actin Depolymerization Factors of the Malaria Parasite*

Apicomplexan parasites, such as the malaria-causing Plasmodium, utilize an actin-based motor for motility and host cell invasion. The actin filaments of these parasites are unusually short, and actin polymerization is under strict control of a small set of regulatory proteins, which are poorly conserved with their mammalian orthologs. Actin depolymerization factors (ADFs) are among the most important actin regulators, affecting the rates of filament turnover in a multifaceted manner. Plasmodium has two ADFs that display low sequence homology with each other and with the higher eukaryotic family members. Here, we show that ADF2, like canonical ADF proteins but unlike ADF1, binds to both globular and filamentous actin, severing filaments and inducing nucleotide exchange on the actin monomer. The crystal structure of Plasmodium ADF1 shows major differences from the ADF consensus, explaining the lack of F-actin binding. Plasmodium ADF2 structurally resembles the canonical members of the ADF/cofilin family.

regulatory proteins, which are poorly conserved compared with their yeast and mammalian orthologs and which have somewhat different biochemical properties (9).
Actin depolymerization factors (ADFs) 4 /cofilins are among the most important regulators of microfilament dynamics. The conventional members of this family have a dual role in destabilizing filamentous (F-)actin, leading to accelerated filament turnover. Upon binding to F-actin, they destabilize the monomer-monomer interactions within the filament, leading to higher off-rates at pointed ends and also filament severing (10,11) and, thus, accelerated depolymerization from an increased number of pointed ends (12). When bound to ADP-bound globular (G-)actin, ADFs inhibit polymerization by decreasing the rate of nucleotide exchange from ADP to ATP (12)(13)(14)(15)(16)(17). At high concentrations, ADF/cofilins have also been reported to nucleate actin filaments (13). ADFs are regulated by post-translational modifications, such as phosphorylation (18) and oxidation (19), respond to changes in pH (20 -23), and bind to membrane phosphoinositols (21,24).
Mammals have three ADF/cofilins, which are highly similar in structure and function (25) but have different expression patterns in different cell types and developmental stages (26,27). Protozoa generally have a single ADF isoform, but Plasmodium species have two. ADF1 lacks the conserved F-actin-binding motifs in its sequence and, unlike other ADFs, binds only G-actin and stimulates nucleotide exchange (28). ADF2 resembles, at the sequence level, more the conventional ADFs. Toxoplasma gondii, a related parasite, has only one ADF (29), which resembles Plasmodium ADF1 (30). Thus, the primary function of these two proteins may be to sequester actin monomers in a polymerizable form.

EXPERIMENTAL PROCEDURES
Protein Structure Determination-Recombinant PfADF1 and PbADF2 were expressed, purified, and crystallized as described before (34). Diffraction data were collected at 100 K to 2.0-and 2.1-Å resolution on PfADF1 and PbADF2 crystals, respectively, as already described (34) ( Table 1). The structures of PbADF2 and PfADF1 were solved by molecular replacement, using three different search models as separate ensembles in the program Phaser (35). The previously solved 2.0-Å structure of ADF1 from Arabidopsis thaliana (1F7S) (32), the 2.3-Å structure of actophorin from Acanthamoeba castellanii (1AHQ) (33), and the 1.72-Å structure of cofilin from Schizosaccharomyces pombe (2I2Q) (13) were used as search models for both PbADF2 and PfADF1. Notably, no obvious molecular replacement solutions were found for PfADF1 when using the structure of PbADF2 as a search model. The structures were refined using TLS (36) parameters derived from TLSMD (37) in Refmac5 (38) and Phenix (39) to R work /R free factors of 0.187/0.230 and 0.187/0.246, for PfADF1 and PbADF2 respectively, with good geometry (Table 1). COOT (40) was used for manual model building. The validity of the models was checked using Procheck (41), SFCHECK (42), and Mol-Probity (43). The coordinates and x-ray diffraction data have been submitted to the Protein Data Bank under the accession codes 2XF1 and 2XFA.
Sequence Analysis and Homology Modeling-A structurebased sequence alignment was performed for the known ADF/ cofilins using SSM (44). Homology models of T. gondii (Tg)ADF and P. falciparum actin were built using SWISS-MODEL (45), based on sequence alignments with PfADF1 and rabbit actin, respectively. Rabbit and P. falciparum actins are 82% identical, and TgADF and PfADF1 are 36% identical.
To model the possible binding of Plasmodium ADFs on Gand F-actin, PfADF1 and PbADF2 were superimposed on the mouse twinfilin C-terminal ADF homology domain (Twf-C)rabbit actin complex (31), which in turn was superimposed on the recently published fiber diffraction F-actin model (46), the Holmes F-actin model (47), and the ADF-decorated F-actin model based on the electron microscopy reconstruction by Galkin et al. (31,48) for comparison. Finally, the models of the Plasmodium ADFs bound to the 162°and 167°twist actin filament were subjected to energy minimization by the program YASARA (49).
The interaction between ADFs and G-actin was studied using native PAGE, as has been done before for PfADF1 (28). The protein of interest and ADP-actin were preincubated for 30 min at room temperature and separated on a 7.5% acrylamide gel containing 80 mM Tris-HCl, pH 8.4, 0.5 mM ADP, 0.1 mM CaCl 2 , and 20 M MgCl 2 at 277 K, 5 W for 1 h, using 50 mM Tris-HCl, pH 8.4, 30 mM N,N-bis(2-hydroxyethyl)glycine, 0.1 mM ADP, 0.1 mM CaCl 2 , 20 M MgCl 2 , and 1 mM DTT as the running buffer. The identity of the proteins was confirmed by excising the bands from the gel and separating the proteins by SDS-PAGE using the NuPAGE MES buffer system (Invitrogen), followed by silver staining, using the SilverQuest TM Silver Staining kit (Invitrogen) according to the manufacturer's manual.
For cosedimentation and sequestering assays, 4 M ATPactin was induced to polymerize by adding 1 mM MgCl 2 and 0.15 M KCl, in the presence of 0 -10 M ADFs, and incubated at room temperature for 2-3 h. After ultracentrifugation at 150,000 ϫ g for 60 min at 293 K, equal amounts of the supernatants and pellets were analyzed by SDS-PAGE and Coomassie staining.
To measure the effect of PbADF2 and PfADF1 on the kinetics of actin polymerization, 4 M PbADF2 or PfADF1 was added to 4 M human platelet actin including 5% pyrene-actin (Cytoskeleton) in the polymerizing buffer (G-buffer including 1 mM MgCl 2 and 0.1 M KCl), and the change in fluorescence was followed for 15 min using an excitation wavelength of 360 nm and emission wavelength of 407 nm.
The rate of nucleotide exchange on actin was measured by monitoring the increase in fluorescence upon the exchange of ADP to 1,N 6 -etheno-ATP (⑀-ATP; Jena Bioscience, Jena, Germany) in G-actin at room temperature. PbADF2 or PfADF1, at concentrations ranging from 4 to 20 M, was added to 4 M actin in a reaction volume of 150 l, and the reaction was started by the addition of 40 M ⑀-ATP. The reaction was monitored for 15 min using an excitation wavelength of 360 nm and emission wavelength of 410 nm with a Tecan Infinite 200 fluorescence plate reader.
Severing/Polymerization Assay-Severing activity of PfADF1 and PbADF2 on tetramethylrhodamine isothiocyanate (TRITC)-phalloidin-labeled actin filaments was observed by laser scanning microscopy. Briefly, polymerization of 2.88 M actin (Cytoskeleton) was induced by adding 1 mM MgCl 2 and where I hi is the intensity of the ith measurement of reflection h and ͗I h ͘ is the mean intensity for the reflection h.
0.15 M KCl. After 30 min of polymerization at room temperature, TRITC-labeled phalloidin (Invitrogen) was added, in an equimolar ratio, to fluorescently label and protect the F-actin against spontaneous depolymerization. Immediately before the start of the experiment, the actin solution was diluted by a factor of 1/5000 in 25 mM imidazole, 1 mM EGTA, 4 mM MgCl 2 , 25 mM KCl, pH 7.4, and incubated for 15 min on coverslides in the presence of 1/10 ratio of either PfADF1 or PbADF2. Samples were coated on coverslides, and images were obtained using a Zeiss Axiovert 200M microscope. Analysis of the actin filament length was done using ImageJ software.

Plasmodium Actin Depolymerization Factor 2 Binds to Both Monomeric and Filamentous Actin-Plasmodium
ADFs share 29 and 38% sequence identity with their closest homologs, for which structures are known (A. castellanii actophorin for PfADF1 and A. thaliana ADF1 for PbADF2), and 28% with each other (Fig. 1). Previously, it has been shown that PfADF1 only binds G-actin and functions as a nucleotide exchange factor, accelerating the rate of exchange from ADP to ATP on G-actin (28). Here, we set out to probe the biochemical functions of the second Plasmodium ADF isoform, PbADF2.
To establish that our recombinant PbADF2 protein is able to bind actin, we carried out native gel electrophoresis under the conditions that had previously been used to characterize PfADF1 (28). A complex between human platelet ADP-actin and PbADF2 migrated as a smear, which consistently was present only when mixing the two proteins and the intensity of which was seen to grow in a concentration-dependent manner, at an intermediate position between the PbADF2 and actin bands ( Fig. 2A).
To study the effect of PbADF2 on nucleotide exchange on G-actin, we monitored the increase in fluorescence upon the exchange from ADP to ⑀-ATP. Unlike conventional ADF/cofilins and similar to PfADF1 (28), a clear concentration-dependent acceleration of the nucleotide exchange on human platelet G-actin could be seen for PbADF2 (Fig. 2D).
Unlike most other ADFs, PfADF1 does not bind F-actin but binds monomers (28). To assess the binding of PbADF2 to F-actin, cosedimentation assays were performed. PbADF2 cosedimented with F-actin in a concentration-dependent manner (Fig. 2C). Control experiments confirmed that Saccharomyces cerevisiae cofilin bound to F-actin, whereas PfADF1 did not. In addition, yeast cofilin diminished the amount of F-actin in the high speed pellets, likely due to both F-actin depolymerizing and monomer sequestering activity. Also, PfADF1 reduced the amount of F-actin, without associating with the filaments. PbADF2, however, had no visible effect on the amount of F-actin in the pellet fractions.
To test the effect of PbADF2 on the kinetics of actin polymerization, a polymerization assay using fluorescent pyrene-actin was performed. At 4 M human platelet actin concentration and a 1:1 ratio of actin to PbADF2, actin polymerization was almost completely inhibited (Fig. 2B). As already reported (28), PfADF1 did not show any significant effect on actin polymerization. To investigate further the roles of the two Plasmodium ADFs in the regulation of actin dynamics, we visualized the effects of PfADF1 and PbADF2 on TRITC-phalloidin-labeled F-actin using a laser scanning microscope (Fig. 3). The presence of PbADF2 resulted in a shorter persistence length of actin filaments and a larger number of shorter filaments observed within each frame. We conclude that PbADF2 severs actin filaments. PfADF1 affected filament length distribution only modestly, and importantly, long filaments were still observed. This is consistent with the previously identified monomer sequestering activity of PfADF1.
Crystal Structures of Plasmodium Actin Depolymerization Factors 1 and 2-To gain insight into the structural basis for the differential functions of the apicomplexan ADFs, we have determined the crystal structures of the two Plasmodium ADF isoforms. Datasets to 2.0 and 2.1 Å resolution were collected on PfADF1 and PbADF2 crystals, respectively (34). Both structures were solved by molecular replacement, using a combination of multiple search models. PfADF1 crystallized in the space group P3 2 21: with one molecule in the asymmetric unit, and the PbADF2 crystals belonged to the space group P2 1 2 1 2, with two molecules in the asymmetric unit.
Both PfADF1 and PbADF2 structures resemble the classical ADF/cofilin family members. The root mean square deviation for C ␣ atoms, when superimposed on A. thaliana ADF1, is 1.6 Å for PfADF1 and 1.2 Å for PbADF2, and the root mean square deviation, when superimposing the C ␣ atoms of the two Plasmodium ADFs on each other, is 1.8 Å. The core of the ADF/ cofilin structures is formed by a six-stranded ␤-sheet, sandwiched between two ␣-helices (Fig. 4, A-C). However, there are significant deviations in the Plasmodium ADFs both from the canonical ADF/cofilin fold as well as from each other, in particular in the postulated F-and G-actin-binding regions (Figs. 1, 4C, and 5).
The largest differences in the two apicomplexan ADFs between each other and compared with other ADF/cofilin family members appear in the C-terminal half of the proteins, starting from ␤-strands 4 and 5 and the hairpin loop connecting them, all the way to the C terminus, which is short in PfADF1 compared with other ADF/cofilins (Figs. 1, 4, and 5). The short ␤-strand located just before the C-terminal helix, which is also involved in G-actin binding, is missing in PfADF1 but conserved in PbADF2. In PbADF2, the C terminus is longer and well defined in the electron density maps. The C-terminal ␣-helix 4 forms a continuous long helix and does not have any gaps or kinks, such as those seen in e.g. human cofilin (23) and A. castellanii actophorin (33) (Figs. 1, 4, and 5). ␣-Helix 3, which forms a central G-actin-binding motif, is highly conserved in all ADFs but has some small but significant differences in both Plasmodium ADFs, as discussed in detail below.
Most members of the ADF/cofilin family contain conserved cysteines and are often regulated by oxidation. PfADF1 has four cysteine residues, and two of these, Cys-24 and Cys-34, are oxidized in our crystal structure. Although the exact positions of these cysteines are not conserved, Cys-34 in PfADF1 occupies a similar position as Cys-39 in human cofilin, which has been suggested to form both intramolecular and intermolecular disulfide bridges, which abolish the ability of cofilin to bind to actin (50). To verify their compositions, bands were excised from the gel (boxed) and separated by SDS-PAGE, followed by silver staining (right panel). B, effect of PbADF2 on the kinetics of actin polymerization was tested by measuring the change in fluorescence upon incorporation of 5% pyrene actin into growing actin polymers. PbADF2 added at a 1:1 ratio to human platelet actin (4 M) almost completely inhibits polymerization. C, ADF binding to F-actin was probed using a cosedimentation assay. After preincubation of actin (4 M) in filament-stabilizing buffer in the presence of ADF/cofilin proteins, as indicated, the samples were subjected to ultracentrifugation, and the pellets were dissolved in sample buffer and separated by SDS-PAGE, followed by Coomassie staining. Yeast cofilin and PbADF2 cosedimented with F-actin, whereas PfADF1 did not. Yeast cofilin and PfADF1 also reduced the amount of F-actin in the pellet, whereas PbADF2 did not. D, the effect of PbADF2 on the exchange of G-actin-bound ADP for free ⑀-ATP was monitored by measuring ⑀-ATP fluorescence upon incorporation into actin. 4 M human platelet actin was used in the assay. Addition of PbADF2 increased the rate of nucleotide exchange in a concentration-dependent manner. AUGUST 12, 2011 • VOLUME 286 • NUMBER 32

JOURNAL OF BIOLOGICAL CHEMISTRY 28259
ADF/cofilin activities and their subcellular distribution are in part regulated by interactions with membrane phospholipids (51). There are four sulfate ions bound in the PfADF1 crystal structure. Three of these are located in close proximity to each other and the N terminus, possibly mimicking the three phosphates of phosphatidylinositol bisphosphates (Fig. 4E). One of the sulfates is coordinated by Arg-6 and binds close to the N terminus of the first ␣-helix, the second is bound between Lys-101 and Arg-6, and Lys-100 is in the direct vicinity. The third sulfate ion is coordinated by Asn-16 and a water molecule, which forms a hydrogen bond to Arg-21. These residues are rather conserved, more so in cofilins than ADFs. In PbADF2, instead of Arg, there is an Asn at position 6, but Lys-36 points to the same direction and could be involved in a similar interaction. The positions of Lys-100 and Lys-101 are occupied by Arg-114 and Lys-115 in PbADF2, and Arg-21 and Asn-16 are conserved in both Plasmodium ADFs. However, the crystallization buffer of PbADF2 contained neither sulfate nor phosphate, and there are no ions bound to these sites on PbADF2.
Actin-binding Sites Are Only Partially Conserved in Plasmodium Actin Depolymerization Factors-ADFs bind to the "lower cleft" of the groove between subdomains 1 and 3 in G-actin (Fig. 5A). This site can be denoted as a "hot spot" on actin because most monomer-binding proteins use this platform for binding. Different proteins, however, have opposite effects on nucleotide exchange, either opening or closing the nucleotidebinding groove on the opposite side of the molecule, between subdomains 2 and 4. It is not understood what triggers these long distance conformational changes.
Three major sites of the ADF homology domain are responsible for the interaction with G-actin (31). These are (i) the N terminus, which interacts with the C terminus of actin; (ii) the N-terminal part of the long ␣-helix 3, which inserts in a groove between actin subdomains 1 and 3; and (iii) the ␤-strand/loop preceding the C-terminal ␣-helix 4, which interacts via two salt bridges with subdomain 3 of actin. We have compared these regions in the Plasmodium ADFs with each other and with their homologs from other phyla. In the N terminus, Ser-3, which is a target for phosphorylation (52), is part of a consensus sequence formed by this Ser, a Gly, and a hydrophobic residue. This motif is completely conserved in all apicomplexan ADFs. In our crystal structures, this region lacks secondary structure and is flexible, but it seems that the hydrophobic interactions with actin at this region would be conserved in the apicomplexan ADFs.
The second motif responsible for actin binding is the long ␣-helix 3, which is present in both Plasmodium ADFs. At the N terminus of this helix, there are two conserved basic residues, which point out to opposite directions and, in the mouse Twf-C-rabbit actin complex (31), interact with main chain and side chain oxygen atoms of actin. Both of these residues are argin- ines in PfADF1 and both of them lysines in PbADF2, and therefore, these interactions seem conserved in both parasite ADFs. Surrounding these basic residues, there is a patch of hydrophobic residues, including a fully conserved Met, which interact in the mouse Twf-C-rabbit actin complex with actin. Some of these hydrophobic residues, in particular Ile-266 and Leu-271 (Twf-C numbering), are not conserved in the Plasmodium ADFs.
Spatially close by, there is also the loop connecting ␤-strand 5 and ␣-helix 3, which interacts with the central part of the ADF-binding platform on actin, which can be seen as a kind of hinge region for the domain movements of the actin monomer. There are two fully conserved proline residues (Pro-332 and Pro-333) in this region in actin, which are stacked against Tyr-262 (Twf-C), which seems to be accurately positioned by the fully conserved Pro-260 (Twf-C). In PbADF2, this proline residue is Pro-94, and the residue corresponding to Tyr-262 (Twf-C) is Leu-96, which also could stack against the two pro-lines in actin. In PfADF1, this proline residue is not conserved and is replaced by an arginine, and the hydrophobic residue corresponding to Tyr-262 (Twf-C) or Leu-96 (PbADF2) is missing, and this position is occupied by a serine.
The largest differences between Plasmodium and other ADFs are in the C-terminal parts, which are involved in interactions with both G-and F-actin. ␤-Strands 4 and 5 and the hairpin loop connecting them form a central F-actin-binding motif in ADF/cofilins. This motif is present in PbADF2, but the strands are much shorter in PfADF1, and the hairpin structure protruding from the core of other ADF/cofilins is virtually missing in PfADF1 and, judging from the sequences, also other apicomplexan ADFs (Figs. 4 and 5). In PbADF2, as in other ADF/cofilins, the long ␣-helix 3 is connected to the C-terminal helix by ␤-strand 6. This strand is virtually absent in PfADF1, and this stretch forms a loop region (Fig. 5A). As already predicted from the sequence, the C terminus of PfADF1 is shorter than in other ADF/cofilins, and thus, the C-terminal ␣-helix 4 is also very short, comprising only one full The N and C termini as well as all ␣-helices and ␤-strands are labeled. B, stereo view of PbADF2. The N and C termini, all ␣-helices and ␤-strands, as well as the F-actin binding F-loop are labeled. C, superposition of PfADF1 (orange) and PbADF2 (green) crystal structures. The N and C termini are labeled, as are the actin-binding ␣-helix 3 and the F-actin-binding ␤-hairpin motif, which is present only in PbADF2. Note also the virtually missing ␤-strand 6 just before the C-terminal helix in PfADF1. D, homology model of TgADF (lilac) superimposed on the crystal structure of PfADF1 (orange). The N and C termini as well as ␣-helices 2 and 3 are labeled. Note the predicted differences between PfADF1 and TgADF in ␣-helix 2, the following loop, and ␤-strand 4. E, surface representation of PfADF1 with sulfate ions shown in sticks. Three of the four sulfates are bound in a cluster close to the N terminus. The fourth one is at the opposite side of the molecule. The N and C termini as well as ␣-helix 1 are labeled.
turn. In the mouse Twf-C-rabbit actin complex (31), Lys-294 and Glu-296 in ␤-strand 6 interact with actin. In PbADF2, the corresponding residues, Lys-121 and Glu-124, present in ␤-strand 6, could be engaged in similar interactions. However, in PfADF1, where this strand is practically missing, there are also no charged residues in the corresponding loop region.
At present, there is no high resolution structure of ADFs bound to F-actin available. However, based on low resolution electron microscope reconstructions, models for the binding mode have been suggested (31,48,53). We have constructed a model of both Plasmodium ADFs bound to the side of the filament, based on the mouse Twf-C-rabbit actin monomer crystal structure (31) modeled on the so-called 162°and 167°twist filament structures. Our models illustrate that the hairpin motif between ␤-strands 4 and 5 and the C-terminal ␣-helix, in PbADF2, would insert into the filament, presumably leading into destabilization of the monomer-monomer interactions, as suggested earlier (31). In PfADF1, where these structural protrusions are missing, such interactions would not be possible (Fig. 5, B and C).
Homology Model of TgADF-We have constructed a homology model of TgADF, based on our crystal structure of PfADF1 (Fig. 4D). For most part, the main chain traces of the TgADF model and PfADF1 are identical. Significant differences are predicted in ␣-helix 2 and ␤-strand 5, which both are much shorter in TgADF. The actin-binding sites of TgADF resemble those of PfADF1. The two basic G-actin-interacting residues at the N terminus of ␣-helix 3 are Arg and Lys in TgADF. As in PfADF1, the last ␤-strand 6 is missing, and the C-terminal helix is short.

DISCUSSION
Apicomplexan parasites depend on rapid turnover of extremely short actin filaments for their gliding motility and host cell invasion. At the sequence level, the major actin isoform of Plasmodium is highly conserved with other eukaryotic actins. Yet, its biochemical properties are quite different; it polymerizes weakly, forming only short (ϳ100-nm) "stubs" (6 -8). Actin polymerization in Apicomplexa is under control of few regulatory proteins (9). Many of these are poorly conserved in both structure and function with their yeast and mammalian counterparts. Because of the crucial role of actin for the survival and pathogenicity of the malaria parasite and its relatives, an understanding of the regulation of their actin filament dynamics and of the structures of the regulatory proteins is an important challenge.
ADFs bind to both G-and F-actin and function in different ways to enhance the rate of filament turnover. In mammals, several ADF/cofilins are differentially expressed and regulated (51). Many protozoans have a single gene encoding ADF, and unlike other Apicomplexa, Plasmodium species have two ADFs, which as we show here, differ in both structure and function. PfADF1 displays atypical biochemical activities for an ADF/cofilin family member. It does not bind to F-actin, and, when bound to G-actin, it accelerates nucleotide exchange from ADP to ATP (28). Here, we show that Plasmodium ADF2, unlike ADF1, binds to both G-and F-actin but, surprisingly, like the major isoform, ADF2 also has a stimulatory effect on nucleotide exchange. In addition, ADF2 but not ADF1 seems to work in severing actin filaments. Our crystal structures now provide a structural basis for the different F-actin-binding properties of the two Plasmodium ADFs (Fig. 5). However, from the sequences and structures of ADF/cofilin family proteins, the determinants FIGURE 5. Modeling of Plasmodium ADFs on G-and F-actin. A, of the three major G-actin-binding sites, the N terminus and the basic residues in ␣-helix 3 are conserved in both PfADF1 (orange) and PbADF2 (green). The Plasmodium ADF structures were superimposed on the mouse Twf-C (gray) in complex with actin (blue) (31). The C-terminal actin-binding motif is conserved in PbADF2 but not in PfADF1. The actin subdomains 1-4 are labeled, as are the N and C termini and ␣-helix 3 of the ADFs. The nucleotide bound in the cleft between actin subdomains 2 and 4 is shown as sticks. B, PfADF1 (orange) and PbADF2 (green) are modeled to the side of a 167°twist actin filament. The model is based on the mouse Twf-C in complex with a rabbit actin monomer (31) and the electron micrographic reconstruction of an actin filament by Oda et al. (46). Each of the actin monomers is shown in a different color, and the direction of the filament is indicated by the ϩ and Ϫ for the barbed and pointed ends, respectively. The C terminus and the F-actin-binding loop in PbADF2 are labeled. In PbADF2, both the C-terminal ␣-helix and the F-loop could penetrate into the filament and destabilize subdomain 4 and its interactions with subdomain 3 of the next actin monomer. In PfADF1, these regions are shorter and would not cause such a destabilizing effect. C, Plasmodium ADFs are modeled on a 162°twist actin filament. Labels are as in B. Again, it can be seen that the F-loop and C-terminal helix of PbADF2, but not PfADF1, can insert deep into the filament structure.
of nucleotide exchange activity are unknown, and it is not clear why the Plasmodium ADFs stimulate nucleotide exchange, whereas the vast majority of ADF/cofilins inhibit nucleotide exchange on actin monomers.
It was recently reported that TgADF, which is structurally most similar to PfADF1 and does not bind to F-actin, inhibits nucleotide exchange on G-actin and functions primarily to sequester actin monomers (30). Thus, it seems that TgADF is functionally not as divergent from the classical ADF/cofilins as PfADF1, although its sequence is clearly more closely related to PfADF1 than to PbADF2 or their mammalian homologs. Also, PfADF1 seems to sequester actin monomers in a readily polymerizable form, whereas PbADF2 may rather act as a filament-severing factor. Here, we show subtle differences in the presumed actin-binding interface of Plasmodium ADFs compared with each other and especially with other ADFs, including the other apicomplexan ones, which may be responsible for the observed biochemical differences.
Still, it is not clear why Plasmodium needs two ADF isoforms, and further genetic and biochemical characterization of the parasite is needed to elucidate this question. Taking into account that Apicomplexa have only very short and unstable actin filaments, there may be no general need for depolymerization factors. However, short actin filaments necessitate efficient cycling of actin monomers back to the growing end of filaments during motility and invasion as well as other actin-dependent processes. Plasmodium and other Apicomplexa have only few actin regulators. Therefore, particularly ADF1 may have evolved as a second, differentially regulated nucleotide exchange factor, in addition to profilin.
The life cycle of Plasmodium requires two very different host organisms; a mammal and an arthropod. It has been suggested that ADF2 is specific for the sexual stages of the parasite within the mosquito host, but parasites lacking ADF2 have relatively mild defects at later stages (54). Also, the degree of conservation between the two ADFs in different Plasmodium species underlines the fact that ADF1 is the major, and essential, isoform, and ADF2 seems to play a minor role during the life cycle of the parasite. Thus, it is likely that certain actin functions, possibly only during specific stages of the life cycle, function more efficiently with a conventional ADF, which is able to quickly sever and depolymerize actin filaments. This may not be the case for other Apicomplexa, such as Toxoplasma, which only need mammalian hosts.
Apicomplexa are eukaryotic, unicellular organisms, which are evolutionarily distant from higher eukaryotes. Therefore, proteins from these parasites are interesting also from the evolutionary point of view. It is quite remarkable how conserved actins are throughout evolution, even apicomplexan actins sharing Ͼ80% sequence identity with mammalian ones. The actin-binding regulatory proteins, however, are strikingly poorly conserved between Apicomplexa and other phyla, raising the question as to whether many of them have evolved via divergent or convergent evolution. Judging by the present state of knowledge, parasite actin itself possesses all of the most important features required for its dynamical behavior, and the actin-binding proteins have evolved to enhance and fine tune these properties. Profilin, for exam-ple, is structurally very divergent in Apicomplexa (55). Yet in the parasites, it seems to fulfill all of the key functionalities of canonical profilins. ADFs, on the other hand, have a fairly well conserved structure, but have adopted at least partially different functions in Plasmodium compared with higher eukaryotes.