Crystal structure of the PbCPαα homodimer
To gain structural insight into the unusual homodimerization of PbCPαα, we crystallized and determined the structure of a C-terminally truncated PbCPαα homodimer (PbCPααΔC20) to 2.2-Å resolution (Fig. 1 and Table S1). The two PbCPαΔC20 subunits form a structure that is less compact and intertwined than the canonical heterodimer (Fig. 1a, b), verifying earlier small-angle X-ray scattering (SAXS) and thermal stability analyses . The quaternary structure closely resembles the mushroom shape of metazoan CPs, despite the low sequence conservation (Fig. S1). Albeit being a homodimer, the structure is not completely symmetric. Thus, we will refer to chain A of the crystal structure as PbCPα1ΔC20 and chain B as PbCPα2ΔC20. Superposition of the two subunits reveals that structural deviations are mainly present at the dimer interface, possibly to accommodate complementary intersubunit interactions (Fig. 1c). In canonical CP heterodimers, the stalk domain of the β-subunit pivots relative to the core structure, to allow a denser packing. In PbCPαΔC20, the presence of two identical stalk domains disrupts the formation of a compact structure (Fig. 1d).
PbCPα contains a sequence insertion, not present in metazoan CPs or PbCPβ (Fig. S1). Similar insertions are found in several other apicomplexan ABPs . Their functions are often unknown but could be involved in parasite-specific protein-protein interactions or immune evasion [14, 45]. The 23-residue Plasmodium CPα-specific insert is located between the globule and cap b-sheets (Ser136-Ala159). In PbCPα1ΔC20, this loop protrudes from the structure (Fig. 1a) making space for the long H5 helix (Ile243-Arg281), which would be followed by the α-tentacle in the full-length protein. The base of the insert is stabilized by a partial disulfide bond between Cys158 and Cys179 in the PbCPα1ΔC20 subunit but not in PbCPα2ΔC20. PbCPα2ΔC20 is notably more disordered, due to the reduced number of intersubunit and crystal contacts and has a higher average B-factor than PbCPα1ΔC20 (Fig. S1, S2), similarly to the crystal structure of CapZαβ . Consequently, the Plasmodium-specific insert is not visible in PbCPα2ΔC20. The H5 helix of PbCPα1ΔC20 is broken at His267 and turns back towards the same subunit in a b-hairpin-like structure (Leu268-Leu286), instead of extending into PbCPα2ΔC20, resulting in very different C termini in the subunits (Fig. 1a, c). Interestingly, the first short β-strand of this hairpin motif ends with Ala270, which is the previously identified C-terminal degradation point in full-length PbCPαα . This implies that the C-terminal hairpin-fold of PbCPα1ΔC20 might be biologically relevant and not an artifact arising from the α-tentacle-truncated PbCPααΔC20 construct.
Homodimeric PbCP assembly lacks canonical interaction sites
The observed structural differences result in a non-canonical dimeric CP structure and an example of a rare asymmetric homodimer among proteins in general . It is unclear, however, how these deviations would impact the possible barbed-end binding mode of PbCPααΔC20. To understand whether a canonical binding mode is compatible with homodimeric CPs, we used molecular docking to generate a model of a PbCPααΔC20-capped PfActI filament (Fig. 2a). The expression of Arp1 in P. berghei  could implicate the existence of PbCP-capped Arp1 filaments in Plasmodium, in a similar manner as the metazoan dynactin . Therefore, we used the cryo-EM structure of CapZαβ-capped Arp1 filament in Sus scrofa dynactin  as a basis for modeling.
The electrostatic potential surfaces at the actin filament barbed end (Fig. 2b) and CPs (Fig. 2c) reveal the absence of canonical charged interaction points in the Plasmodium proteins (Fig. S4), even though residues involved in forming these interaction surfaces are conserved (Fig. S1, S3, S5). In PfActI, flanking residues mask the apparent charge of the patch (Fig. 2b, S4a). Differences in surface electrostatic potential  could also play a large role in the divergent nucleation and polymerization properties of PfActI . Residues of the basic patch in PbCPααΔC20 are more buried in the structure, not in close proximity to each other, and the slight asymmetry of the homodimer does not compensate significantly for the loss of contributing residues of PbCPβ. The putative basic patch spans a larger area in the PbCPαβ model, only loosely resembling the arrangement in CapZαβ (Fig. S4b). Possibly due to the absence of CARMIL proteins in Plasmodium , the positively charged CARMIL-binding pocket of canonical CPs  is less prevalent in the PbCPαβ model (Fig. S4b) and absent in PbCPααΔC20 (Fig. 2c). Upon binding the barbed end, the β-tentacle of CapZαβ locks the complex by binding the hydrophobic pocket of the terminal actin subunit [39, 40]. In our model, the Plasmodium-specific insert of PbCPα1ΔC20 is close to the expected position of the β-tentacle (Fig. 2d). This insert shares many possible hydrophobic or electrostatic interaction points with the β-tentacle. These residues are mainly conserved across Plasmodium spp. (Fig. S1), suggesting a possible role for the insert in barbed-end binding in the absence of the β-tentacle. Mapping per-residue sequence conservation of Plasmodium CPα sequences on the structure of PbCPααΔC20 reveals that core residues, especially those involved in the intersubunit surface, are highly conserved, suggesting that CPαα homodimers likely exist in other Plasmodium spp. Canonical binding partner interaction points present in CapZαβ lack localized sequence conservation in PbCPααΔC20 (Fig. 2e, S1).
The differences in the CARMIL-binding pocket instigated us to take a deeper look into other CP regulators in Plasmodium. While the majority of these are absent from Plasmodium , we cannot exclude the possibility of other, so far uncharacterized, regulators. Our search for V-1/myotrophin homologs and the CP-binding and uncapping motif of CARMIL proteins among known Plasmodium transcripts, or for the S100B interaction motif in Plasmodium CPs resulted in no clear hits. To date, PIP2is the only described direct CP regulator in P. knowlesi with HSC70 mentioned as a binding partner uninvolved in the capping activity. Residues involved in canonical interactions with protein regulators are generally not conserved in Plasmodium CPs (Fig. S1, S3), with the exception of several basic residues (K278 and K282 in PbCPα, R233 in PbCPβ), which are involved in canonical F-actin interaction and PIP2 binding [39, 40, 51].
Conservation of the CP domain structure
Canonical CP subunits share a strikingly similar fold, despite low sequence identity. Compared to the heterodimeric CPs, the homodimeric PbCPααΔC20 has an increased surface area but a significantly decreased dimer interface area (Table S2). Expectedly, the canonical CP heterodimers are more similar to each other than to the Plasmodium CP homodimer (Table S3). Despite the low sequence identity, both PbCPααΔC20 subunits are strikingly similar to canonical CPα isoforms (Table S4). This confirms our previous findings  and is further substantiated by the identifiable hits of CP-related CATH domains (1.20.1290.20 and 126.96.36.199) in the PbCPααΔC20 structure. The identified domains in PbCPααΔC20 superimpose well with canonical CP subunits (Table S5), suggesting the importance of conserved structural elements in the actin filament capping function among broad taxa, despite the lack of sequence conservation. Indeed, during the erythrocytic stages of P. berghei, where PbCPβ is phenotypically silent and homodimeric PbCPαα takes over functionally, PbCPα can be complemented by its PfCPα cognate despite the fact that they only share 52% sequence identity .
Plasmodium CPs form similar-shaped homo- and heterodimers in solution
Our previous SAXS studies  agree well with the crystal structure suggesting that the homodimeric structure is a native conformation and not an artifact resulting from crystal packing or the absence of the α-tentacle domains. To gain further insight into the homo- and heterodimerization of PbCPs, we carried out further SAXS and homology modeling studies based on the crystal structure of PbCPααΔC20.
PbCPααΔC20, PbCPαα, and PbCPαb all form similar pseudo-symmetric dimers in solution (Fig. 3). The SAXS data indicate folded, globular proteins containing flexible parts, with PbCPαβ being somewhat more compact than the homodimers (Fig. 3a, b). Interparticle distances are similar for all PbCPs, with the heterodimer showing slight bimodality (Fig. 3c). Modeling into the SAXS data suggests that PbCPαβ forms a more canonical tight structure, whereas both subunits of the homodimers adopt a loose arrangement, similar to the PbCPααΔC20 crystal structure (Fig. 3d). The orientation of the Plasmodium-specific insert and the tentacle domains of both subunits cannot be reliably deduced from the SAXS data, indicating that they are disordered in solution, which is characteristic of the tentacle domains in canonical CPs as well [35, 38, 52].
PbCPs control actin polymerization in a non-canonical manner
CPs typically block the barbed end, increasing its apparent critical concentration (Ccapp) to that of the pointed end . PbCPs increase the supernatant fraction of polymerized PfActI in pelleting assays , which could be explained either by filament shortening or by limited depolymerization due to the increase of Ccapp. However, in dilution series of pyrene-labeled PfActI filaments, none of the PbCPs affected the fluorescence signalof PfActI, even at high stoichiometric ratios (Fig. 4a). Similar behavior was observed with heterologous α-actin (Fig. 4b), despite PbCPs being able to modulate α-actin  or β-actin polymerization [22, 23]. Furthermore, gelsolin, which is a major barbed end capper  absent from Plasmodium spp. , does not affect the Ccapp of PfActI filaments (Fig. S6) or PfActI depolymerization dynamics . The decreased amount of pelletable actin in the presence of PbCPs and the lack of effect on the fluorescence signal in the pyrene assay are likely due to a decrease in filament length to oligomers that do not sediment. Another possibility would be CP binding to the barbed end of actin in either monomeric or dimeric form, thus increasing it fluorescence signal similarly to polymerization.
In higher eukaryotes, CPs nucleate filaments, abolishing the lag phase, but block subunit exchange at the barbed end , reducing both the initial elongation velocity and steady-state filament mass. Contrary to expectations, PbCPs increased PfActI elongation rates and considerably raised the steady-state fluorescence levels, indicating elevated amounts of actin in non-monomeric forms (Fig. 5). The effect of PbCPs on nucleation was ambiguous, due to the nature of PfActI polymerization curves, which lack a pronounced lag phase . The heterodimeric CPs did not modulate the steady-state fluorescence levels of PfActI to the extent seen with the PbCP homodimers, which could also be observed as slope differences in the Cc assays (Fig. 4). Interestingly, all PbCPs seem to block the barbed end of α-actin, similarly to CapZαβ (Fig. S7). However, they do not display any notable nucleation activity, and unlike heterodimeric CPs, the PbCP homodimers increase the initial fluorescence levels. PbCPααΔC20 displays an only slightly diminished capping effect, despite the absence of the tentacle domains [39, 40]. Despite its in vivo indispensability  the reduced importance of the Plasmodium α-tentacle domain in PbCPs has been suggested before .
Major differences between the homo- and heterodimers are seen in the steady-state PfActI fluorescence levels (Fig. 5a). We also measured length distributions of PfActI filaments polymerized with and without homo- and heterodimeric CPs from negative-stained electron micrographs (Fig. 6). PfActI alone forms mainly short, non-helical structures with a sporadic presence of long helical filaments . Upon incubation with CPs, the long filaments were completely removed (Fig. 6a) and the shorter species became more abundant. The average filament length is reduced by half (Fig. 6b, c) as also seen for canonical actins [22, 23, 44]. In line with the polymerization assays, the reduction in filament length is more pronounced with the homodimeric PbCPs, and the tentacle domain is at least partly dispensable for the PbCP function.
PfActI filaments are highly dynamic and unstable [5–7] with high disassociation rates at both ends . Upon dilution below their Ccapp, filaments decompose rapidly, albeit slower than α-actin if filament end concentration is taken into account . To investigate the barbed-end blocking efficiency of PbCPs, we followed the disassembly of fluorescently labeled actin filaments. In contrast to canonical CPs , both homo- and heterodimeric PbCPs facilitate the depolymerization of both PfActI and α-actin filaments (Fig. 7, S8). The increased velocity can be partly attributed to an increased filament end concentration , caused by shortening of the filaments (Fig. 6). However, this may not be the only explanation. Interestingly, CapZαβ also increased the rate of PfActI depolymerization (Fig. S8). This is in line with CapZαβ increasing the initial elongation rate of PfActI in the polymerization assays (Fig. 5a).