A unique profilin-actin interface is important for malaria parasite motility

Profilin is an actin monomer binding protein that provides ATP-actin for incorporation into actin filaments. In contrast to higher eukaryotic cells with their large filamentous actin structures, apicomplexan parasites typically contain only short and highly dynamic microfilaments. In apicomplexans, profilin appears to be the main monomer-sequestering protein. Compared to classical profilins, apicomplexan profilins contain an additional arm-like β-hairpin motif, which we show here to be critically involved in actin binding. Through comparative analysis using two profilin mutants, we reveal this motif to be implicated in gliding motility of Plasmodium berghei sporozoites, the rapidly migrating forms of a rodent malaria parasite transmitted by mosquitoes. Force measurements on migrating sporozoites and molecular dynamics simulations indicate that the interaction between actin and profilin fine-tunes gliding motility. Our data suggest that evolutionary pressure to achieve efficient high-speed gliding has resulted in a unique profilin-actin interface in these parasites.

Profilin, one of the core actin regulators, has been suggested to be essential in the blood stage, where transfection and selection of transgenic parasites is performed [33]. Indeed, conditional down-regulation of profilin, likely leading to a defect in merozoite invasion of red blood cells, suggested that it is important for efficient blood stage growth [34], while overexpression had no effect on the blood stage [35]. Canonical profilins bind monomeric ADP-actin, catalyze ADP-to-ATP exchange, and enhance actin polymerization in conjunction with the actin nucleators of the formin family [36]. Importantly, apicomplexan formins contain only rudimentary or no profilin-binding motifs, and both P. falciparum and T. gondii profilins appear to sequester actin monomers also in the presence of formins [37,38]. Furthermore, T. gondii profilin inhibits, rather than accelerates, nucleotide exchange on actin monomers [38,39]. Thus, it was suggested that T. gondii profilin functions mainly as an actin monomer sequestering protein, possibly limiting actin polymerization in vivo [38].
Profilin may have evolved via gene fusion from two smaller proteins [40]. Subsequent divergent evolution seems to have resulted in two structurally very different branches: the classical profilins in most organisms as well as the apicomplexan ones with a large sequence insertion at the border of the two small ancestral proteins [33,40]. This insertion, a unique arm-like ß-hairpin domain seems optimally positioned to participate in actin binding [33,39]. However, a possible functional contribution of this motif in both actin binding and parasite motility remained uncharacterized so far. To probe this potential interaction and to investigate the structure-function relationships of profilin across the Plasmodium life cycle, we generated a series of transgenic parasites, featuring two mutations at the tip of the ß-hairpin. We further expressed the corresponding proteins in vitro and show that the arm-like domain contributes critically to the actin-binding capacity of Plasmodium profilin. Interestingly, this additional motif appears important only during the rapid motility of sporozoites, where a weakened profilin-actin interaction leads to a decrease in their capacity to generate force. Strikingly, one mutant still retained continuous gliding motility, while the other one lost this ability, although for both mutants, the capacity to produce force was similarly weak. Molecular dynamics simulations suggested differences in the actin-binding mode of the two mutants, which appears to be at the core of the different gliding phenotypes. Taken together, our data show how sporozoites fine-tune actin dynamics to achieve gliding motility.

Results
Profilin is expressed during the entire Plasmodium life cycle and localized throughout the parasite P. berghei profilin has previously been shown by RT-PCR analysis to be expressed across the entire parasite life cycle [33]. To investigate profilin protein abundance and localization across the life cycle, we endogenously tagged profilin with mCherry (S2 Fig). Since the profilin gene consists of 4 exons and 3 introns, we measured expression of profilin-mCherry constructs with and without introns (Pfn -i mCh and Pfn +i mCh), as these could influence gene expression [41,42]. The fluorescent parasites progressed through the life cycle in a manner largely comparable to wild-type parasites, suggesting that the mCherry fusion did not detectably impair parasite viability (Table 1). Western blot analysis showed that profilin-mCherry migrated at the expected size of the fusion protein (S3A Fig). Imaging parasites at the ring, trophozoite, and sporozoite stages revealed uniform profilin localization in the cytosol and the nucleus, while in gametocytes and ookinetes the profilin-mCherry fluorescence was more pronounced in the nucleus (S3B Fig). Profilin localization is consistent with the localization of GFP-actin, which was also found in the cytosol and the nucleus [43]. Addition of actin filament stabilizing (jasplakinolide) or depolymerizing (cytochalasin D) compounds did not alter the localization of profilin, as expected for an actin monomer binding protein. Quantitative analysis of images from different stages showed that the profilin-mCherry expression levels varied slightly across the stages. The only significant differences in expression depending on the presence or absence of introns occurred in the blood stages ( Fig 1A). Intriguingly, the parasites lacking introns in the profilin gene grew slightly more slowly in the blood stage (Table 1) and reached slightly larger cell sizes of up to 45 μm 3 (S3C and S3D Fig). The fluorescence intensity of Pfn +i mCh and Pfn -i mCh motile ookinetes also appeared slightly different (p = 0.06), yet the Pfn expression in sporozoites was not affected (Fig 1A and 1B). Curiously, while the speed of motile ookinetes was not affected (Fig 1C), the speed of sporozoites expressing mCherry-tagged profilin both with and without introns was significantly reduced (Fig 1D). This suggests steric hindrance by the mCherry tag that only has an effect on motile sporozoites. The speed of sporozoites could be restored to wild type levels in parasites that expressed a longer linker between profilin and mCherry (Pfn // mCh) ( Fig 1D). However, in these parasites, the mCherry tag was largely cleaved off, leading to a majority of non-tagged profilin ( Fig 1E). In all tagged lines, Pfn expression was under the control of the endogenous promoter and the 3'UTR from PbDHFR, suggesting that the 3'UTR did not influence the level of profilin expression (S2A Fig). These data highlight the importance of carefully evaluating the impact of large fluorescent tags on protein functionality.

P. falciparum profilin can functionally complement P. berghei profilin
As human profilin has previously been shown to partially compensate for P. berghei profilin depletion [33], we hypothesized that P. falciparum profilin should also compensate for P. berghei profilin depletion. We reasoned that this would enable detailed mutational analysis in a more straightforward fashion (see methods) and with the protein of a disease-relevant parasite species that has already been expressed and analyzed in vitro [33]. Considering also the lack of striking differences in expression in the presence or absence of introns, we thus generated a replacement plasmid using P. falciparum cDNA and obtained a parasite line that showed no difference to wild type P. berghei parasites across the life cycle (S2 Fig, Table 1), suggesting that the two profilin orthologues are functionally interchangeable. Analysis of parasite lines expressing mCherry-tagged profilin with or without introns or different variants of P. falciparum profilin. Median blood-stage multiplication rates were determined after injecting either 100 or 5000 blood stages intravenously into four C57BL/6 mice respectively. Oocyst numbers (mean) were obtained by counting the cysts of at least 50 infected mosquito midguts (MG) on day 12. MG and salivary gland (SG) sporozoite (spz) numbers Molecular dynamics simulations suggest involvement of the Plasmodium-specific motif in the profilin-actin interaction The structure of Pf Pfn suggested that the parasite-specific hairpin and in particular the acidic residues at the tip of this loop might be involved in interactions with actin, presenting a unique actin-profilin interface [33]. The profilin-binding surface in Pf actin 1 is conserved with  We first analyzed the electrostatic properties of the modelled interface between Plasmodium actin and P. falciparum profilin (Pf Pfn). The complex can be superimposed on the rabbit actin-human profilin complex (PDB ID: 2PAV) [44], but there are remarkably large differences in the two profilin-actin complexes. In addition to the arm-like β-hairpin insertion in Pf Pfn (Fig 2A), the lack of the short α-helix 3 and the much longer β-strands 4 and 5 in Pf Pfn cause significant differences at the interface. As hypothesized before, the acidic residues 64 EDE 66 , conserved across Plasmodium spp., at the tip of the parasite-specific arm ( Fig 2B) are in close contact with a positively charged surface patch on subdomain 3 of actin in our model, suggesting the possibility to form a strong ionic (salt-bridge) interaction (Fig 2A, S1 Movie). The actin residues contributing to this positive patch include K284, R291, K292, Y295 and T325 (numbering as per Pf actin 1) and are conserved between Plasmodium and vertebrate skeletal muscle α-actin, suggesting that Pf Pfn can engage in similar interactions with both actins.
To get more insight into how the acidic residues of the arm motif in Pf Pfn contribute to actin binding, we simulated actin-profilin complexes with two mutant versions, in which the residues 64 EDE 66 were exchanged to QNQ or AAA (Fig 2B, S2 Movie). These mutations were chosen in order to weaken (QNQ) or abolish (AAA) the interactions of the acidic arm motif with actin. We analyzed the inter-protein hydrogen-bond occupancy during simulations for three interaction regions of the protein-protein interfaces: the arm-specific interaction (shown in red in Fig  . At the tip of the arm motif, E66 of Pf Pfn makes two strong hydrogen bonds with R291 of actin with the highest occupancy (181% out of a possible 200% occupancy for 2 hydrogen bonds, individually 90% and 91%) of all the hydrogen bonds at the actin-profilin interface ( Fig  2C). In the QNQ mutant, Q66 lacks the negative charge and therefore loses the salt-bridge nature of this interaction (S1 Movie). Nevertheless, Q66 of the QNQ mutant profilin can form a single hydrogen bond with actin-R291 (52% occupancy) (Fig 2C). On the other hand, the AAA mutant as predicted is unable to make any arm-mediated ionic or hydrogen-bonding interactions at the actin-profilin interface (Fig 2C and 2D, S2 Movie).
In contrast to the profilin arm-mediated interactions, the hydrogen bonds in the vicinity of the Pfn arm are preserved in the two mutants. Surprisingly, the hydrogen-bonding residuepairs significantly differ at the interface formed by β-strand 4 of Pf Pfn (Fig 2C), and the AAA mutant has the most divergent hydrogen-bonding pattern in this region. For example, while E111 of Pf Pfn makes strong hydrogen bonds (103%) with actin R373, these have lower occupancy (80%) in the QNQ mutant and are absent in the AAA mutant. However, in the AAA mutant, E111 forms hydrogen bonds with K114 (79%) and H372 (29%) in actin. The loss of these hydrogen bonds in the QNQ mutant is compensated by an additional stronger hydrogen bond: E107 with actin-K114 (56%) (S6 Fig). The weakening of the interaction with actin for the AAA mutant, due to the lacking hydrogen bonds involving the tip of the β-hairpin loop, is also visualized by molecular dynamics simulations, during which the arm region in the AAA mutant is destabilized and moves away from actin ( Fig 2D, S2 Movie). This destabilization further weakens the interactions at the conventional actin-profilin interface involving β-strand 4 ( Fig 2C, blue bars).
We finally computed the relative free energy values (lacking the full entropic contribution) for the wild type and mutant Pf Pfn-actin complexes ( Table 2). This suggests that wild type Pf Pfn binds most strongly to actin (MM-PBSA free energy: -64.2 ± 11.2 kcal/mol; similar values and the same trends were obtained with the MM-GBSA method). The QNQ mutant has a less favorable binding free energy (-44.2 ± 12.8 kcal/mol), which is in line with its weaker H-bond interactions. The AAA mutant is the least energetically stable (-34.4 ± 11.1 kcal/mol) of the complexes, which may be due to the instability of the arm region which further destabilizes other interactions at the actin-profilin interface.

Mutations in the Plasmodium-specific arm domain of profilin lead to impaired actin binding
To investigate whether the differences suggested by the simulations can lead to biochemically detectable differences in actin binding of the mutants, we next characterized the effect of recombinant P. falciparum profilin and both mutants on the polymerization kinetics of vertebrate skeletal muscle α-actin. Like wild-type Pf Pfn, both QNQ and AAA mutants could be purified as monomeric, soluble, and folded proteins (S7A and S7B Fig). Whereas wild-type Pf Pfn drastically reduced the rate of canonical actin polymerization, both mutants showed a significantly reduced effect (Fig 3). The weaker binding of the mutants to actin was also seen in the lack of increased initial fluorescence (S5C and S5D Fig). We next assessed, whether the same would be true for the biologically relevant binding partner, Pf actin 1. Using native PAGE, complex formation was observed for wild type but not for the mutant profilins ( Fig  4A). However, intriguingly, the effects of Pf Pfn on canonical and parasite actin polymerization kinetics were different. While Pf Pfn significantly reduced the rate of elongation for both (classic actin-profilin interface encompassing β-strands 4 and 5). Subdomains 1-4 are labelled in Roman numerals and the Apicomplexa-specific actin-profilin interface is indicated with a red arrow on the right. (B) Amino acid sequence of the arm domain from P. berghei and P. falciparum profilins showing the AAA and QNQ mutations. (C) Analysis of the hydrogen bonds during 100-150 ns of molecular dynamics simulation formed by the residues indicated with the matching colors in A. The graph shows the occupancy of hydrogen bonds between the indicated amino acid residue pairs for the Pf Pfn and the QNQ and AAA mutants. Note that the AAA mutant shows no interaction at the tip of the hairpin. K100, Q102, E107 and E111 were previously predicted to interact with actin [33]. Values higher than 100% occupancy are due to formation of more than one hydrogen bond. (D) Ensemble structures of Plasmodium profilin in complex with Plasmodium actin during 150 ns of molecular dynamics simulations. All trajectories are aligned to the backbone of the WT complex (frame 1). Red highlights the starting and blue the final structure. Note the large mobility of the arm-like hairpin in the AAA mutant (yellow double-arrow).
https://doi.org/10.1371/journal.ppat.1006412.g002 actins, it had little effect on the steady-state of α-actin (Fig 3A), but a clear monomer-sequestering activity on P. falciparum actin 1 (Fig 4A and 4B). The mutated profilins showed a weakened capacity for reducing the rate of polymerization as well as monomer sequestering (Figs 3 and 4). However, there was no clear difference in the effects of the QNQ and AAA mutants, especially on parasite actin (Fig 4), suggesting that even the relatively subtle differences in the binding of the QNQ mutant compared to the wild type Pf Pfn are sufficient for significantly reducing the binding affinity to actin.

Parasites with mutations in the Plasmodium-specific arm domain of profilin show aberrant sporozoite motility
To address whether the mutations would lead to a detectable effect in vivo, we generated parasites expressing the respective P. falciparum profilins in place of the single endogenous P. berghei profilin gene (S2 Fig) and compared them to wild type P. berghei parasites. Expression of the QNQ and AAA mutants in the parasite had no detrimental effect on life cycle progression (Table 1). Whereas only minor differences in the speed of wild type and transgenic ookinetes were observed (Fig 5A-5C), a striking effect was seen in sporozoites ( Fig 5D-5I). We found no difference between the wild type and QNQ sporozoites in their capacity to glide on a flat substrate, but the AAA mutants largely failed to move progressively (Fig 5D-5G). Curiously, Pf Pfn-expressing sporozoites moved faster but less persistently than wild type P. berghei sporozoites, and this was also seen for the QNQ mutant ( Fig 5G and 5H). The differences in speed were even more obvious when instantaneous speeds were compared ( Fig 5I). Interestingly, also the AAA mutant ookinetes moved at lower speeds than the QNQ mutant ookinetes ( Fig  5B). Together with the biochemical data and the molecular dynamics simulations, these data on sporozoite motility confirm that the parasite-specific arm in Pf Pfn is crucial for actin binding and, thereby, sporozoite motility. It is interesting that both simulations and motility assays show differences between the QNQ and AAA mutants, but these are not obvious from the biochemical assays, suggesting that the subtle differences in the two mutants are more complicated than a simple effect on binding affinity. Sporozoites expressing profilin mutants produce lower forces Mutations of the unique profilin arm clearly emphasize the importance of this region in facilitating efficient sporozoite motility. However, the phenotypic characterization described so far, does not yet lead to a functional understanding of how profilin-actin interactions modulate parasite motility. To gain further functional understanding, we employed a recently-established laser trap assay [20]. This allowed us to probe at defined forces the capacity of sporozoites to pull a micron-sized bead from an optical trap and to translocate it to the rear of the cell (Fig 6). We presume that this pulling force corresponds to the force that sporozoites exert on their surrounding environment to move forward. To measure this force, a bead was trapped in the laser focus and pushed onto the sporozoite surface, where it was captured by the sporozoite (  Fig 6A). Subsequently, the sporozoite pulled the bead backwards, presumably due to the myosin-based translocation of actin filaments anchored via transmembrane adhesins [20]. This translocation backwards thus most likely corresponds to the retrograde flow of actin filaments observed in fibroblasts and other cells and also assumed to occur in sporozoites [20,30,45]. We probed our set of parasites with this assay and scored whether, at a given optical force, the sporozoite pulled the bead out of the trap or not. In total, over 1,200 sporozoites were probed from 4 different parasite lines at 3 different optical forces (Fig 6B-6D). In this very sensitive assay, no difference was found between wild type P. berghei parasites and P. berghei parasites expressing P. falciparum profilin. At an optical force of 70 pN, sporozoites from both lines were able to pull approximately 70% of the beads out of the trap, while at 130 pN, this percentage dropped to approximately 35% (Fig 6B, [20]) and at 190 pN to 20% [20]. Both the AAA and the QNQ mutant were less efficient in pulling the bead from the trap at all forces (Fig 6B). At 70 pN, both mutant lines only managed to pull approximately 20% of beads out of the trap, and at 130 pN, this dropped further to approximately 10%.

Profilin mutants show enhanced retrograde flow rates
To probe whether the mutations affect retrograde flow, we used the same laser trap setup. However, in these experiments the force of the trap was set to approximately 10 pN, which was sufficiently small for the sporozoite to instantly move the bead towards the rear (Fig 6A). In all parasite lines, the beads were transported backwards with an increasing speed until a peak or plateau was reached ( Fig 6C). To analyze potential differences in the transport speeds, we compared the maximum speed of over 170 sporozoites across the four different parasite lines. We found no difference between the wild-type P. berghei and the line expressing wild type P. falciparum profilin (Fig 6C and 6D). However, parasites expressing the AAA and QNQ mutant profilins moved the beads at an increased speed, i.e. showed a faster retrograde flow (Fig 6C and  6D). Thus, while the QNQ mutant seems impaired in actin binding and force generation to a similar degree as the AAA mutant, there are subtle differences that lead to differences in the ability of the mutants to support fast, directional motility. The molecular dynamics simulations suggest this could be due to differences in the off rates of the complexes, especially in the situation in vivo, where other proteins compete with profilin for actin binding.

A unique actin-profilin interface important for motility
Profilins as ancient actin-binding proteins have typically low sequence identity but share a highly conserved 3D fold. Apicomplexan profilins differ from all other profilins in containing a unique β-hairpin extension. The exact length and sequence of this insertion is not fully conserved between different Apicomplexa, but they all share the acidic nature of the tip of this extension. Here, we show that mutations in this unique motif in Plasmodium diminished actin monomer sequestration in vitro (Figs 3 and 4) and led to impaired sporozoite motility ( Fig  5G-5I), while only slightly affecting ookinete motility (Fig 5B and 5C). This differential effect was already visible in the construction of profilin-mCherry fusions, which slightly impaired sporozoite motility while they had no effect on ookinete motility (Fig 1C and 1D). The defects were more accentuated in the mutant parasite lines analyzed. This, together with previous data [31] suggests that for the evolution of the core motility machinery, the sporozoite presents the major constraining stage of the parasite, as it appears the most vulnerable to subtle changes. This is likely due to the fact that the sporozoite has to migrate extremely fast, while passing several biologically very different tissue barriers. Profilin in malaria parasite motility

Mutants reveal a stepwise loss of actin-profilin interaction
Our data revealed an interesting dichotomy: while both QNQ and AAA mutants showed similar differences compared to wild type in our biochemical and biophysical assays, only the AAA mutant showed a defect in motility. The molecular dynamics simulations might provide an answer for this intriguing difference. They revealed a salt bridge with two hydrogen bonds that contributes to stabilizing the interaction between the arm-motif of profilin and actin (Fig 2, S1 Movie). Interestingly the salt bridge is lost in both mutants, whereas one of the two hydrogen bonds is retained in the QNQ mutant but both hydrogen bonds are lost in the AAA mutant. This suggests that the biochemical and biophysical assays were sensitive to the differences due to the loss of the salt bridge and one of the hydrogen bonds, whereas the motility assay revealed the difference upon losing the last hydrogen bond of the arm-motif. This mechanistic insight would have not been revealed by simply studying defects in gliding motility and shows the power of the combination of biochemical with biophysical and computational assays. The data suggests a threshold, such that a successively lower affinity between profilin and actin leads first to the loss of force generation (concomitant with an increase in retrograde flow speed) followed by a loss in the capacity to glide on a 2D substrate (Table 3). Further disruption of actin dynamics or the organization of actin filaments would then lead to additional problems such as salivary gland invasion [30,31,35].
How is force generated for motility?
We had previously hypothesized that integration of retrograde flow and force production produces motility [20]. However, the near-normal motile behavior of the QNQ mutant could not have been predicted through these two parameters alone. This leads us to postulate that an as yet unknown factor plays a role in producing sporozoite motility (Fig 7). This factor could be a molecule, a molecular complex or the dynamic arrangement of proteins in complexes. Considering the complexity of the interplay between many factors to generate motility, we favor the latter hypothesis. Interesting in this respect and for the future direction of research into parasite motility that goes beyond the identification of molecular players are two recent studies on T. gondii Table 3. Summary of results, including previous work, revealing a possible thresholding phenomenon during sporozoite motility. tachyzoites. The first study adapted our laser tweezer approach to T. gondii tachyzoites [47]. This revealed surprising differences compared to sporozoites. Most strikingly, retrograde flow in tachyzoites occurred after a delay of approximately one minute, while it was instantaneous in sporozoites. Also, directed force production only occurred after this initial phase in tachyzoites. This clearly suggests different kinetics to assemble a directional motor between the two parasites. A second study investigating the essential role of many of the key proteins involved in motility showed that parasites can still be motile, albeit at a much reduced rate, in the absence of core motility proteins, including actin [48]. This showed that even in the absence of actin, retrograde flow still moved beads towards the rear of the tachyzoite. Yet, retrograde flow occurred in a manner of minutes and not seconds as observed with P. berghei sporozoites. This suggests a much more complex picture and the existence of important differences between apicomplexan parasites that are currently not fully appreciated.

Actin dynamics influences force generation in sporozoites
Sporozoite motility relies on a complex interaction of the formation and turnover of adhesion sites as well as the generation of force [19][20][21]49]. It is currently not clear how this complex interplay is orchestrated. Investigating motility thus relies on the determination of many different factors, such as the percentage of motile sporozoites, and their average and instantaneous speeds (Fig 5, Table 3). Previously, we have shown that low concentrations of jasplakinolide, which likely leads to longer or less dynamic actin filaments, cause higher retrograde flow speed but lower force [20] and less motility ( [46], S1B Fig, Table 3). Cytochalasin D, which in contrast disrupts actin filaments, led to lower force and less motility, yet had only mild effects on retrograde flow ( [20], S1B Fig, Table 3). Thus, longer and more stable filaments would lead to a higher retrograde flow, which is the case for both of our mutants (Fig 6D). Both longer filaments and disruption Actin modulators such as profilin (this study) or coronin [30] influence the state of actin by either promoting or preventing filament formation, filament crosslinking and filament orientation. The actin state influences retrograde flow and force generation as revealed by the use of inhibitors or mutants. Motility is the outcome of a non-trivial interplay between retrograde flow and force as these are not perfectly inversely correlated. Hence we stipulate the existence of another unknown factor that influences the actin state to produce optimal motility. The system is robust within a certain range (red dashed line), while unbalancing one or several of these factors leads to aberrant motility.
https://doi.org/10.1371/journal.ppat.1006412.g007 of filaments would lead to lower forces, which again is the case for both mutants (Fig 6B). Thus, the length of filaments and their dynamics do not scale linearly with the loss of force. Rather, maximum force generation likely requires optimized filament length, dynamics, and orientation, which are achieved through the interplay of several proteins in vivo [18,20,30,50].
Although both mutants have weakened actin-binding affinity, the QNQ mutant seems to be above a threshold required for gliding. It seems likely that the AAA mutant releases actin monomers faster, so that in vivo, other monomer-binding proteins have the chance to bind actin. These could be for example the ADFs, which promote filament formation by catalyzing nucleotide exchange [28,51]. This also shows that sporozoite motility in vivo is robust enough to tolerate small changes in binding affinities of individual proteins, and that our in vitro assays are sensitive enough to now understand the fine-tuning of this incredibly efficient and fast machinery.

Concluding remarks
In summary, we present the first study using site-directed mutagenesis of a constitutively expressed protein involved in malaria parasite motility that uncovers a new type of interaction between actin and profilin. We provide conclusive evidence from a combination of biochemical, biophysical, computational and molecular genetic studies that this actin-profilin interface is important for Plasmodium sporozoite migration. We suggest that sporozoite entry into salivary glands is the key constraint for the evolution of the Plasmodium motility machinery.

Ethics statement
All animal experiments were performed according to the German Animal Welfare Act (Tierschutzgesetz) and were approved by the responsible German authorities (Regierungspräsidium Karlsruhe, numbers G-3/11, G-134/14 and G-310/14).

Animal work
Generation of parasite lines and infections of mosquitoes was performed using female NMRI mice (Janvier). Monitoring of parasite prepatency and blood stage growth rates was performed using female C57BL/6 mice (Charles River).

Recombinant protein production and biochemical work
A mutant Pf Pfn construct in which the sequence 64 EDE 66 was replaced by QNQ was obtained by gene synthesis from Mr. Gene and sub cloned into pETM-11 using NcoI and XhoI restriction sites. This was used as a backbone to generate the AAA mutant using site-directed mutagenesis with the QuikChange Lightning mutagenesis kit (Agilent Technologies) with primer pair 12. Wild type [33] and mutant Pf Pfns were expressed and purified [37] and the correct folding of the mutants verified using synchrotron CD spectroscopy, as described before [40] (S7 Fig). Pig (Sus scrofa) skeletal muscle α-actin was purified as before [37,52]. Pf actin 1 was expressed in Spodoptera frugiperda Sf21 cells (Invitrogen) as described earlier [37]. After expression the cells were harvested and suspended in lysis buffer (10 mM HEPES pH 7.5, 250 mM NaCl, 5 mM CaCl 2 , 15 mM imidazole, 1 mM ATP and 7 mM β-mercaptoethanol). Cells were lysed by sonication on ice, and the lysate was clarified by a 1-h centrifugation at 30600 g. The supernatant was mixed with 1 ml of HisPur Ni-NTA resin (Thermo Fisher Scientific), incubated for 1 h at 4˚C, and the resin was washed extensively with lysis buffer 1 with 3 mM β-mercaptoethanol, the same buffer with 500 mM NaCl and 25 mM imidazole, and with G-buffer (10 mM HEPES pH 7.5, 0.2 mM CaCl 2 , 0.5 mM ATP and 3 mM β-mercaptoethanol). Finally, the protein was eluted with G-buffer containing 300 mM imidazole and imidazole removed using a PD-10 desalting column (GE Healthcare). The His-tag was cleaved by a 1 h incubation at 22˚C with TEV protease, after which the protein was passed through a Ni-NTA column and supplemented with 300 mM ammonium acetate. Finally, the protein was concentrated and applied to a Superdex 200 10/300 GL column (GE Healthcare) in GF-buffer (Gbuffer containing 300 mM ammonium acetate and 0.5 mM TCEP-HCl instead of β-mercaptoethanol) for final purification.
The effects of wild type and mutant profilins on α-actin polymerization kinetics were measured using fluorescence spectroscopy. 4 μM actin, of which 5% pyrene-labelled, in 10 mM Tris-HCl, pH 7.5, 0.5 mM ATP, 0.2 mM CaCl 2 , 1 mM DTT was induced to polymerize by the addition of polymerizing buffer to final concentrations of 0.1 M KCl, 0.1 mM MgCl 2 , 5 mM Tris-HCl, pH 7.5, 0.2 mM ATP in the presence of 0-20 μM wild-type Pf Pfn or the QNQ and AAA mutants. Polymerization was followed for 1 h by measuring the increase in fluorescence signal upon incorporation of pyrene-labelled actin into growing filaments using a Tecan M1000 Pro plate reader at 25˚C with excitation and emission wavelengths of 365 and 407 nm, respectively.
The effects of Pf and mutant profilins on Pf actin 1 polymerization were studied using fluorescence spectroscopy as in the case of α-actin but using approximately 1 and 2% of pyrene-labelled Pf actin 1 in the case of the wild-type Pf profilin concentration series and Pf, QNQ and AAA profilin comparison experiments, respectively. Polymerization of 4 μM Pf actin 1 alone and in the presence of 1-64 μM Pf profilin or 16 μM Pf, QNQ or AAA profilin (in duplicates) was induced by adding polymerizing buffer to final concentrations of 50 mM KCl, 4 mM MgCl 2 , and 1 mM EGTA. Polymerization was followed for 2 h using the parameters described above.
All polymerization curves were set to start from zero fluorescence intensity, and the initial polymerization rates were determined as the slopes of linear fits to the polymerization data from 300 to 600 s for α-actin and 600 to 1200 s for Pf actin 1. The relative initial polymerization rates were obtained by dividing the initial polymerization rate values with the initial polymerization rate of Pf actin 1 alone.
For co-sedimentation experiments 100 μl of each Pf actin 1 polymerization sample were recovered from the 96-well plate. Samples were centrifuged 1 h at 25˚C and 100000 rpm, using a TLA-100 rotor (Beckman Coulter), and the resulting supernatants and pellets were separated. The supernatants were mixed with 25 μl of 5x SDS-PAGE sample buffer (250 mM Tris pH 6.8, 10% SDS, 50% glycerol, 0.02% bromophenol blue, 1.43 M β-mercaptoethanol), and the pellets were resuspended in 125 μl of PBS pH 7.4 supplemented with 1x SDS-PAGE sample buffer. Samples were incubated 5 min at 95˚C and 10 μl of each sample was analyzed on 4-20% SDS-PAGE gels. The protein bands were visualized with PageBlue stain (Thermo Fisher Scientific). Gels were imaged using the ChemiDoc XR S+ system and protein band intensities were determined with the Image Lab 3.0 software (both from Bio-Rad). For each supernatant and pellet pair, the total intensity of Pf actin 1 was set to 100% and relative amounts of actin 1 in supernatants and pellets were presented as percentages of that.
To study the Pf actin 1-profilin interaction on native PAGE, samples containing 5.6 μM Pf actin 1 alone and together with increasing concentrations (1-256 μM) of Pf, QNQ, and AAA profilins were prepared. In addition, samples with 32 μM profilins alone were included. The samples were incubated for 15 min at 22˚C and supplemented with native PAGE loading dye (6% glycerol, 50 mM Tris pH 8.0, 0.004% bromophenol blue, final concentrations). 12.5 μl of each sample were analyzed on a 4-20% MiniProtean TGX gel (Bio-Rad) using 25 mM Tris, 200 mM glycine, 0.5 mM ATP, 0.1 mM CaCl 2 pH 8.3 as the running buffer and constant 150 V for 1.5 h at 4˚C. Finally, the gels were stained with PageBlue protein staining solution (Thermo Scientific).

Small-angle X-ray scattering
For SAXS analysis, wild-type Pf Pfn and α-actin were mixed in a 1:1 molar ratio, incubated on ice for 2 h, concentrated to 500 μl using a centrifugal filter, and gel filtered using a Superdex S200 10/300 GL column, equilibrated in 5 mM Tris-HCl (pH 7.5), 0.2 mM CaCl 2 , 0.2 mM ATP, and 2 mM DTT. The peak fractions containing the complex were pooled and concentrated to 2.5 mg/ml. SAXS data were collected on beamline X33 at EMBL/DESY, Hamburg (Germany) at concentrations 2.5 and 1.25 mg/ml. Analysis of the data was carried out using the ATSAS package [53]. Ab initio models were built using DAMMIF [54] and GASBOR [55]. The Pfn-actin complex had an R g value of 2.4 nm and D max of 8 nm, compared to the respective values of 2.5 and 8.1 for BSA and 1.4 and 4.7 for Pf Pfn alone. Actin alone in our experience polymerizes during the SAXS measurement and does not give meaningful values for comparison with globular proteins.

Molecular dynamics simulations
Preparation of structures: The crystal structures of P. falciparum actin 1 (PDB ID: 4CBU, 1.3 Å resolution) [25] and P. falciparum profilin (PDB ID: 2JKF, 2.31 Å resolution) [33] were retrieved from the RCSB-PDB database [56]. These structures were aligned to the crystal structure of rabbit alpha skeletal muscle actin co-crystallized with human profilin (PDB ID: 2PAV, 2.31 Å resolution) [57]. The thus obtained P. falciparum actin-profilin complex was prepared for simulations using the Protein Preparation Wizard module of Schrodinger (version 2016r1). In brief, the complex was preprocessed to assign bond orders, to add missing hydrogen atoms and to add missing side chains. Co-crystallized waters were kept in the complex structure. PROPKA [58] was used to predict the protonation states at pH 7.0 of the titratable residues. Missing residues were modelled using the Prime module in the Schrodinger software. Note that the P. falciparum and P. berghei actin only differ in three amino acid residues E3D, D5E and V11I which are distant from the conventional actin-profilin interface.
To model P. berghei profilin, the P. berghei profilin sequence was retrieved from the Plas-moDB database [59] and modelled using PRIME software using the P. falciparum profilin structure (75.9% identical to P. berghei profilin) as the template structure. The residues at the tip of the arm of P. falciparum profilin, 64 EDE 66, were mutated using Maestro (Schrodinger) to generate the two mutants: 64 AAA 66 and 64 QNQ 66. Molecular dynamics simulations: The modelled protein complexes were prepared for allatom molecular dynamics simulations using the tleap program in the AMBER molecular dynamics package version 14 (http://ambermd.org/) [60]. ATP parameters were taken from the AMBER parameter database (http://research.bmh.manchester.ac.uk/bryce/amber) [61]. GAFF [62] and ff14SB [60] parameters were assigned to the ligand and protein, respectively. Non-bonded interactions were cut off at 8 Å and PME was applied. The systems were solvated using the TIP3P water model [63] in a truncated octahedral box. K + and Clions were added to obtain an ionic strength of 50 mM and the systems were neutralized using Na + counterions. A 2-step minimization was performed on each system as follows: 1000 steps of minimization while keeping restraints (force constant 100 kcal/mol Å 2 ) on the solute (protein and ligands) (first 500 steps of steepest descent, next 500 steps of conjugate gradient) followed by all-atom minimization (first 1500 steps of steepest descent, next 1000 steps of conjugate gradient). The minimized systems were gradually heated (0 to 298 K in 80 picoseconds) using the canonical ensemble (NVT) at each temperature point. In the next step, the pre-heated systems were equilibrated in an isothermal-isobaric ensemble (NPT) at 298 K. Berendsen temperature coupling and a constant pressure of 1 atm with isotropic molecule-based scaling was used in the equilibration. The SHAKE algorithm [64] was applied to constrain all covalent bonds containing hydrogen atoms and a time step of 2 fs was used. All systems were simulated with periodic boundary conditions in the NPT ensemble for 150 ns. The analysis of the MD trajectories was carried out with the CPPTRAJ module of AMBER 14. VMD (version 1.9.2), Chimera (version 1.10) and Pymol (version 1.8.2.3) were used for visualization.
Binding free energy calculations: The molecular mechanics energies combined with the Poisson Boltzmann or generalized Born and surface area continuum solvation (MM/PBSA and MM/GBSA) energies were used to estimate the actin-profilin binding free energy. The snapshots were retrieved at an interval of 1 ns from the last 50 ns of the MD trajectories (between 100 and 150 ns). Because the current study involves the comparison of similar systems, we did not explicitly calculate entropic contributions to the binding free energy, assuming they are similar in all cases. Therefore, the calculated energies do not correspond to the absolute free energies but can be used to compare similar systems.

Molecular cloning and parasite generation
All vectors used in this work are based on the b3D+ vector [65]. We modified the vector for homologous recombination in the profilin (PBANKA_0833000) locus on chromosome 8 as follows. The P. berghei profilin 5' upstream region (871 bp) was amplified from P. berghei ANKA WT genomic DNA using primer combination 5 (see S1 Table) and subsequently inserted into b3D+ via SacII and NotI digestion and ligation (S2 Fig). The profilin 3' downstream region (805 bp) was amplified with primer combination 6 and inserted using ClaI and KpnI. For the Pfn-mCh lines, the P. berghei profilin gene either from gDNA (+i) or cDNA (-i) was fused to mCherry using overlap extension PCR with primer combinations 7 and 8. The fused genes connected through four glycines as a linker were cloned using NotI and BamHI.
The third tagged line Pfn -i // mCh in which the mCherry tag was cleaved to at least 75%, was created with a longer linker (amino acids: AAAASRTSAAAA; this sequence includes the amino acids encoded by the XbaI and SpeI restriction sites). P. berghei wild type profilin was amplified with primer combination 9 and cloned with NotI and XbaI. The mCherry gene was amplified with primer combination 10 and cloned using SpeI and BamHI. As a control, the Heussler group (Bern University, Switzerland) kindly provided us with a parasite expressing cytoplasmic mCherry under the ef1α promoter [66]. P. falciparum wild type and mutant (QNQ & AAA) profilins were amplified from the respective pETM-11 vectors used for protein expression with primer combination 11 and cloned into b3D+ using NotI and XbaI.

Parasite transfection and sporozoite generation
All transfections and generation of clonal parasite lines were performed as previously described [67]. All vectors were linearized with SacII and KpnI prior to transfection. A PCR to probe correct integration was performed after limiting dilution cloning (S2 Fig). Anopheles stephensi mosquitoes were infected with clonal lines as follows: Infected mouse blood was injected intraperitoneally (IP) into a naïve NMRI mouse. At a parasitemia of >1% the blood was harvested and 10-20 million parasites were injected IP into two or three naïve mice. Three to four days later, mice were anesthetized and positioned on top of a mosquito cage so as to allow mosquitoes to feed. Mosquitoes were analyzed for oocyst numbers on day 12 and for midgut and salivary gland sporozoites on days 17-23.

Assessment of parasite function across the life cycle
We determined parasite growth rates by injecting 100 or 5000 infected red blood cells into each of four C57BL/6 mice respectively. Parasitemia was monitored daily from day 3 on. We calculated parasite growth rate as explained before [68,69]. Oocyst numbers were determined by extracting midguts of infected mosquitoes on day 12 after infection. Midguts were stained for 20 min using 0.1% mercurochrome solution [70]. Stained oocysts were counted using a 10x objective in at least 50 infected midguts.
Imaging was performed using an inverted Axiovert 200 M microscope (Zeiss, Göttingen). Blood stages were imaged by applying a drop of blood from the tail vein of an infected mouse, adding 1 μg / ml Hoechst 33342 DNA dye and diluting it with PBS. The mixture was covered with a cover slip and imaged with differential interference contrast (DIC) and fluorescence microscopy using the same exposure times and objective lens (63x, N.A. 1.4).

Quantification of sporozoite and ookinete motility
Ookinete cultures were prepared as previously described [72]. Imaging of motile ookinetes was performed by applying a drop of the culture onto a glass slide and covering it with a cover slip. DIC images were acquired at 0.05 Hz for 15 min. Salivary glands of infected mosquitoes were isolated between days 17 and 25 after infection. They were kept in RPMI (supplemented with 50 000 units / l penicillin and 50 mg /l streptomycin) containing 3% bovine serum albumin (BSA, Roth Ltd) and transferred to a 96 well plate (Nunc MicroWell 96 well optical bottom plates, Sigma) for imaging. The plate was centrifuged for 5 min at 500 g to settle the sporozoites. DIC images were acquired at 0.33 Hz for 5 min. Sporozoite and ookinete speeds were analyzed using the ImageJ plug-in 'Manual tracking' [73].

Retrograde flow experiments
The retrograde flow experiments were performed on the self-built laser trap setup described in [20]. In brief, beads (PC-S-2.0, streptavidin-polystyrene microparticles 1.5-1.9 μm, 1% w/v, Kisker) were held with minimal laser power by a stationary laser trap. Consequently the stage and thereby the self-built open flow cell with the gliding sporozoites in it, were moved towards an optically trapped bead. The bead was then positioned onto the front end of the sporozoite. When sporozoite and bead made contact, the sporozoite pulled the bead out of the focus of the laser and translocated the bead to rear of the cell. This was imaged with a frame rate of 100 images per second. The speeds of the transported beads were tracked using MATLAB routines.

Force measurements
The force measurement experiments were performed as described in detail in [20]. Beads were captured in the center of the trap-this time with defined forces (70 pN, 130 pN and 190 pN) -were brought in close proximity with the sporozoite until they touched the beads. Sporozoites were challenged to displace the bead from the focus of the trap. and both mutant (blue) profilins. The purity of the samples is shown on a Coomassie-stained SDS-PAGE gel in the inset. The samples are: 1) wild type Pf Pfn, 2) the QNQ mutant, 3) the AAA mutant, 4) molecular-weight standards. The molecular weights of the relevant standards in kDa are shown on the right. (B) The folding of the mutant profilins (blue curves) was shown to be identical to that of the wild-type protein (green) using CD spectroscopy. (C) Sedimentation of 4 μM Pf actin 1 alone and in the presence of 1-64 μM Pf profilin. (D) Sedimentation of 4 μM Pf actin 1 alone and in the presence of 16 μM Pf, QNQ and AAA profilins. Samples were analyzed on 4-20% SDS-PAGE gels and protein bands were visualized with PageBlue stain (Thermo Fisher Scientific). S denotes supernatant and P pellet. Quantification from duplicate gels is presented in Fig 3B. (TIF) S1 Table. Primer pairs with sequences used for cloning and verification of integration. Respective restriction enzyme motifs (lowercase) and binding site descriptions are indicated. Start codons are coloured green and stop codons red. Blue shows the glycine and orange the alanine linker. (PDF) S1 Movie. 360˚rotation of the models of the P. falciparum profilin-actin I complex from molecular dynamics simulations corresponding to Fig 2A. The first rotation shows actin in gold and profilin in grey; the actin-profilin complex is shown with different contact residues highlighted in red (tip of hairpin), green (arm-neighbor interactions) and blue (classic actinprofilin interface). The second rotation shows surface charges in actin that are highlighted in blue (basic) and red (acidic). Note the tip of the arm-like β-hairpin (acidic amino acids in red) contacting a basic region of actin subdomain 3.