Single particle cryo-EM analysis of Rickettsia conorii Sca2 reveals a formin-like core

Abstract

Spotted fever group Rickettsia undergo actin-based motility inside infected eukaryotic cells using Sca2 (surface cell antigen 2): an ~1800 amino-acid monomeric autotransporter protein that is surface-attached to the bacterium and responsible for the assembly of long unbranched actin tails. Sca2 is the only known functional mimic of eukaryotic formins, yet it shares no sequence similarities to the latter. Using structural and biochemical approaches we have previously shown that Sca2 uses a novel actin assembly mechanism. The first ~400 amino acids fold into helix-loop-helix repeats that form a crescent shape reminiscent of a formin FH2 monomer. Additionally, the N- and C- terminal halves of Sca2 display intramolecular interaction in an end-to-end manner and cooperate for actin assembly, mimicking a formin FH2 dimer. Towards a better structural understanding of this mechanism, we performed single-particle cryo-electron microscopy analysis of Sca2. While high-resolution structural details remain elusive, our model confirms the presence of a formin-like core: Sca2 indeed forms a doughnut shape, similar in diameter to a formin FH2 dimer and can accommodate two actin subunits. Extra electron density, thought to be contributed by the C-terminal repeat domain (CRD), covering one side is also observed. This structural analysis allows us to propose an updated model where nucleation proceeds by encircling two actin subunits, and elongation proceeds either by a formin-like mechanism that necessitates conformational changes in the observed Sca2 model, or via an insertional mechanism akin to that observed in the ParMRC system.

Graphical Abstract

Introduction

Actin is a globular protein (G-actin) that can self-associate, or polymerize, into filaments (F-actin), and its dynamic remodeling is crucial in many processes essential to cellular survival, such as division, motility, endocytosis, and intracellular trafficking [1]. As such, it constitutes a major target for intracellular pathogens [2–5]. Some of these pathogens activate actin assembly on their surfaces to drive their motility inside infected eukaryotic cells via an actin comet tail mechanism [6, 7]. Amongst these are pathogens that encode effector proteins that directly bind and assemble actin [8, 9]. Since these effector proteins have evolved to functionally mimic eukaryotic actin assembly factors using different sequences and folds, understanding the structural basis of how they assemble actin would allow us to better understand fundamentals of actin assembly [9]. Structural knowledge of such proteins could also facilitate the development of anti-bacterial agents. Here, we focus on surface cell antigen 2 (Sca2) from spotted fever group Rickettsia, the only known pathogen effector protein, to date, that is a functional mimic of eukaryotic formins [10–13].

Rickettsia are obligate intracellular bacteria, transferred to humans by arthropod vectors such as ticks and fleas, and cause mild and severe human disease such as rickettsiosis and spotted fever [14, 15]. Spotted fever group Rickettsia invade mammalian cells via a zipper mechanism [3, 16], escape into the cytosol, and undergo actin-based motility [17]. During early infection stages, they rely on RickA and the host Arp2/3 complex for actin-based motility, forming short and curved tails similar in appearance to those observed in Listeria and Shigella, which also utilize the host Arp2/3 [18]. Later during infection, Sca2 was shown to be responsible and required for actin-based motility [11, 19], producing actin tails that appear long, unbranched, and independent of RickA or host Arp2/3 [20].

Sca2 is a functional mimic of formins: a family of eukaryotic actin assembly factors that nucleate and processively elongate actin filaments, tracking the barbed, or growing, end of the filament [21–23]. All formins contain various N-terminal domains important for autoinhibition, activation, signaling, and localization, whilst their C-termini harbor two conserved formin homology (FH) domains (FH1 and FH2) that are sufficient to account for their actin assembly activities. The FH2 domain forms a doughnut shape as a dimer to encircle two actin subunits, where each FH2 monomer contributes half of the ‘doughnut’ shape that forms the barbed-end tracking unit. FH2 domains dimerize through a head-to-tail interaction. For elongation, actin subunits can either be added directly to the barbed end, or through the FH1 domain. The FH1 domain is a flexible region that contains one or more proline-rich domains (PRDs) that can recruit profilin-actin to add to the barbed end.

Autotransporters, or bacterial type Va secretion system proteins, such as Sca2 (Figure 1A), are assembled on the outer leaflet of the Rickettsial outer membrane. They are first transported through the inner membrane to the periplasmic space using a signal sequence (residues 1-33), which is then cleaved. Transport through the inner membrane is facilitated by the Sec machinery. Then, the transmembrane C-terminal autotransporter domain (residues 1516-1795) forms a pore in the outer membrane, assisted by the BAM complex, through which the rest of the protein (passenger domain, residues 34-1515) is translocated, folds, and remains attached, exposed to the surface [24, 25].

Figure 1. Sample preparation and overall structure of Sca2.

(A) The domain organization of Rickettsia conorii Sca2 (Uniprot Q92JF7). The expressed Sca2 construct (residues 34-1515) used in this study is indicated. SS: signal sequence, NRD: N-terminal repeat domain, PRD: Proline-rich domain, CRD: C-terminal repeat domain, AC: Auto chaperone, TD: Translocator domain. (B) SDS-PAGE of Sca2 elutions following Ni-NTA and StrepTrap affinity purification. (C) Glycerol gradient chromatogram and SDS-PAGE analysis of the boxed fractions. (D) Representative negative stain TEM micrograph of Sca2. (E) Representative cryo-EM micrograph of Sca2. (F) Cryo-EM density map of Sca2 shown from three views, and the domains of Sca2 rigid-body fit into the density.

While unrelated in sequence and structure, Sca2 functionally mimics formins in that it can assemble unbranched filaments, remain associated with the growing barbed end, compete with capping protein, and use profilin-actin for efficient elongation [11]. This is achieved through an entirely different mechanism, however. Sca2 is a monomer which binds two actin subunits with nanomolar affinity and displays intramolecular interaction and cooperativity between its N- and C- termini for actin assembly [12, 13]. Structurally, the high-resolution crystal structure of Rickettsia conorii Sca2 residues 34-400 N-terminal repeat domain (NRD), previously determined by our group [12], shows a novel helix-loop-helix repeat that folds into a crescent shape very similar to that of a formin FH2 monomer. Taking into consideration the fact that: 1-Sca2 is a monomeric autotransporter, where it’s N-terminal passenger domain, after passing through its C-terminal translocator domain, has to fold on the outside of a bacterium surface without an energy source [26], 2- Autotransporters typically adopt simple folds, 3- Residues 868-1515 are suggested to be helical based on circular dichroism measurements [12], and 4-Experimental data supporting an end-to-end intramolecular interaction within monomeric Sca2 [13], we hypothesize that the fold observed in the NRD is also present in the remainder of the passenger domain to form a doughnut or formin-like core that can encircle two actin subunits [12, 13].

Toward testing this model, we sought to obtain a high-resolution structure of the full-length passenger domain of Sca2. Since other than the NRD, which we previously crystallized, Sca2 had proven recalcitrant to crystallization, we performed single-particle cryo-electron microscopy (cryo-EM) analysis on Sca2. Despite being unable to obtain a high-resolution structure, the electron density in the 3D reconstruction from the cryo-EM data strongly supports the presence of a formin-like core: A clear doughnut-like, or circular, density is observed with a hollow center. Interestingly, however, the circular fold is covered from one side. Using our previously determined crystal structure of NRD (residues 34-400) [12], and AlphaFold [27] models of regions 401-670 and the C-terminal repeat domain (CRD) residues 1090-1355, the most likely Sca2 domain configuration (see Methods) is one where the NRD and region 401-670 form the circular fold, with the CRD accounting for the electron density covering one opening. This model suggests a nucleation mechanism where Sca2 can encircle two actin subunits in a manner similar to formins, whereas at least two possibilities for an elongation mechanism are plausible. Elongation may proceed either by a formin-like mechanism which would require conformational changes in the observed Sca2 model, or by an insertional type elongation mechanism analogous to that of actin-like ParM filament elongation by ParR [28].

Materials and Methods

Expression and purification of Rickettsia conorii Sca2

The Sca2 gene sequence from Rickettsia conorii (Uniprot Q92JF7) was used in this study. A codon-optimized Sca2 full-length passenger domain (Sca residues 34-1515, referred to as Sca2 in this study) was cloned into a pET28a plasmid, with an engineered TEV site between the N-terminus of the gene and the His tag, and an engineered C-terminal Strep tag (Genscript Inc.). Sca2 was expressed in Escherichia coli Arctic Express DE3 cells grown in Terrific Broth media to an optical density (OD₆₀₀) of 1, then induced with isopropyl β-D-1-thiogalactopyranoside (IPTG) for 24 h at 10° C. Cells were resuspended in Lysis buffer (20 mM HEPES (pH 7.5), 200 mM NaCl, 20 mM imidazole, 1 complete protease inhibitor tablet (Roche)) and lysed using a microfluidizer (MicroFluidics Corporation, Newton, MA). After centrifugation to remove cellular debris, the protein was purified by Ni-NTA affinity chromatography and eluted with Lysis buffer containing 300 mM imidazole. The eluted protein was further purified using a StrepTrap HP column (Cytiva), and eluted with Lysis buffer containing 40 mM biotin. Protein purity was confirmed by SDS-PAGE (Figure 1B). Concentration was determined by A₂₈₀ using an estimated molecular weight of 171 kDa and extinction coefficient of 97,530 M⁻¹ cm⁻¹ calculated with Protparam (https://web.expasy.org/protparam/).

Cryo-EM sample preparation and data collection

Sca2 protein for EM analyses was isolated by glycerol gradient centrifugation (Figure 1C). Samples were sedimented at 40,000 rpm for ~16 h in a 5 to 30% glycerol gradient at 4° C using a Beckman SW 60 Ti rotor. Glycerol gradients were prepared using a Gradient Master device (BioComp Instruments), and when cross-linking was used, 0.125% (v/v) glutaraldehyde was added to the 30% glycerol solution before gradient preparation. After gradient fractionation, the glutaraldehyde cross-linker was quenched by the addition of glycine-HCl buffer (pH 7.5) to a final concentration of 40 mM. Peak gradient fractions of Sca2, as determined by UV traces of cross-linked samples and SDS-PAGE analysis of non-cross-linked samples, were pooled and concentrated to ~10 mg/ml before flash freezing in 30 μl aliquots. Sample homogeneity was first determined by negative stain electron microscopy (Figure 1D). To prepare grids, samples were thawed and dialyzed in glycerol-free buffer (20 mM HEPES (pH 7.5), 75 mM NaCl), diluted to 2 mg/ml and plunge frozen. Samples of Sca2 (3 μl) were applied to glow-discharged (1 min, easiGlow, Pelco) R1.2/1.3 200-mesh Quantifoil holey carbon grids (Electron Microscopy Sciences) just after the addition of decyl maltose neopentyl glycol (DMNG) to a final concentration of 0.00034% (1/10^th the critical micelle concentration). The grids were blotted for 2.5 s at force 10 with Whatman 41 filter paper and flash-frozen by plunging into liquid ethane using a Vitrobot Mark IV.

Samples of Sca2 with actin monomers were prepared as above, except a five-fold molar excess of actin over Sca2 was applied to the glycerol gradient to promote saturation of the Sca2 actin-binding sites. Peak gradient fractions appeared to have a saturating amount of actin on Sca2 (Figure S1A).

Cryo-EM datasets were collected automatically by Latitude (Gatan) on an FEI Titan Krios transmission electron microscope operating at 300 kV and equipped with a K3 (Gatan) direct electron detector with an energy quantum filter (Gatan). Images were taken at a nominal magnification of 105,000x in counting mode with a pixel size of 0.873 Å. A total of 8397 images (Sca2 sample) and 6897 images (Sca2 with actin sample) were collected at a defocus range of −1.0 to −2.5 μm. The exposure time was 3.0 s divided into 40 frames, at a dose rate of 12.76 e⁻ sec⁻¹ pixel⁻¹ resulting in a nominal dose of 50 e⁻/Ų.

Cryo-EM data processing

The Sca2 dataset was imported to Relion (version 3.1.3) for motion-correction using MotionCorr2 [29] and for contrast transfer function (CTF) correction using CTFFIND4 [30]. The 8,397 micrographs were sorted based on CTF correction, and micrographs with a maximum resolution fit >8 Å were excluded, resulting in a subset of 8,115 micrographs. The general model within the program crYOLO [31] was used to automatically pick particles. A total of 1,626,411 particles were picked, and two rounds of reference-free 2D classification were performed to remove particles that lacked clear features, resulting in a subset of 889,697 particles. The reconstruction with cryoSPARC [32] resulted in a ~12Å-resolution map. Additional rounds of refinement were unable to improve the resolution of the map, suggesting a large amount of flexibility within the molecule preventing high-resolution alignment. Data processing of the Sca2-actin dataset was similar to that of the Sca2 dataset. From 6,897 micrographs, a total of 1,548,912 particles were picked using crYOLO and subsequently processed as described above.

Model building and refinement

Both our structure of the NRD and an AlphaFold-generated model of Sca2 regions 401-670 and CRD (residues 1090-1355) were used in model building. Individual domains were fit as rigid bodies using the program Phenix. Taking into consideration several factors, including the positioning of domains relative to one another within the polypeptide chain of Sca2, functional coordination among domains established in our previous studies [12, 13], and the model-to-map correlation coefficient from Phenix, we generated the most likely model shown in Figure 1F.

Results and Discussion

Structure determination and overall structure of Sca2

Using negative-stain TEM screening, we found it was necessary to use glycerol gradient fixation, or GraFix, and the addition of a mild crosslinker to prepare the Sca2 protein sample before freezing grids [33]. Data for 1,626,411 particles of Sca2 were collected from 8,115 micrographs. The best 3D ab initio model obtained from this data is a low-resolution map of ~ 12 Å. Our previous NRD crystal structure (residues 34-400) and AlphaFold-predicted [27] structures of residues 401-670 and the CRD (residues 1090-1355) were used for rigid-body fitting into the density map, yielding a model that is consistent with both the structural data from this study along with our previous biochemical data (Figure 1F). The fit of the domains into the map was best when the domains were positioned as shown in the model. Residues 671-1089, comprising the two proline-rich domains (PRDs) and the Middle domain, are unaccounted for in the electron density map. The low-resolution data and missing region are presumably due to the unexpected flexibility of the molecule.

The Sca2 model obtained here is consistent with our previous biochemical data [12, 13]. We had previously shown that the two halves of the Sca2 passenger domain, namely, Sca_34-670 and Sca_868-1515 can be co-purified when co-expressed and can reproduce the actin polymerization activity of the full-length passenger domain Sca_34-1515 (Residues 67-867 appeared to be dispensable for this intramolecular interaction) [12]. Using analytical size exclusion chromatography (SEC), we showed that deletions at either end of these Sca2 constructs abolish this intramolecular interaction, supporting a model where the two halves of Sca2 interact in a circular, head-to-tail manner. The two halves of Sca2 interact with high affinity as determined by isothermal titration calorimetry (ITC) titration of Sca_34-670 into Sca_868-1515, suggesting that this interaction is stable [13]. Furthermore, using ITC we demonstrate that the full-length passenger domain of Sca2 binds two actin subunits with high affinity [13]. The structural model of Sca2 suggested by the single-particle cryo-EM analysis reveals the presence of a circular shape with a hollow center and a diameter of ~ 74 Å (Figure 1F and Figure 2A). What is peculiar, however, is that this circular shape seems to be formed by the N-terminal half residues 34-670, with the CRD (residues 1090-1355) contributing to the extra electron density that is covering one opening of the circle. This gives rise to an unexpected overall shape of an open circle somewhat covered on one side. The exact position of the missing residues 671-1089 relative to this observed structure is difficult to pinpoint, but our previous data where both PRDs, located at 671-702 and 1075-1087 (Figure 1A), respectively, appear to synergize for profilin-dependent actin elongation suggests that they are near the barbed end of the elongating actin filament [12].

Figure 2. Sca2 comparison with formins.

Sca2 from our cryo-EM analysis compared with crystal structures of various formins. (A) Top- Sca2 model, with NRD residues 34-400 in red, residues 401-670 in yellow, and CRD residues 1090-1355 shown in green. Bottom- Sca2 model showing only residues 34-670, with CRD residues hidden. (B-E) Crystal structures of different formins. FH2 monomers are colored red or yellow for a straightforward comparison, and in (B-C) actin is colored in blue (B) Top- FMNL3 bound to actin (PDB 4EAH [37]). Bottom- Same as top, actins hidden. (C) Top- Bni1p bound to actin (PDB 1Y64 [35]). Bottom- Same as top, actins hidden. (D) DAAM1 (PDB 2Z6E [36]). (E) Bni1p (PDB 1UX5 [34]).

To address the flexibility of the Middle domain, which we predicted occurred due to the absence of bound actin monomers, we attempted to obtain structural information of Sca2 complexed with monomeric actin. We had already shown that Sca2 binds at least 2 actin subunits with nanomolar affinity, and that a construct comprising residues 869-1060 (part of the region 671-1089 missing in the electron density map) contributes to binding one of the actin subunits [13]. Based on this data, we hypothesized that the binding of Sca2 to actin could lead to a more stable, less flexible molecule, which would aid high-resolution structure determination. This attempt was unsuccessful; despite a pure and homogenous GraFix-purified Sca2-actin complex (Figure S1) and good cryo-EM data collection, the data does not yield high-resolution maps during reconstruction and model building.

Insights into actin assembly by Sca2

The circular fold in the Sca2 model displays a diameter of ~ 74 Å (Figure 2A). Interestingly, this is a very similar diameter to that observed for an FH2 dimer in previously studied crystal structures of formins (Figure 2B–E) [34–37]. Previously we had shown that the Sca2 NRD binds actin on its inner concave surface, and displays a similar surface charge distribution as a formin FH2 [12]. Taking this into account, along with the position of the NRD in the Sca2 model and the observed diameter of ~ 74 Å, this would place two actin subunits on the inside of the observed circular fold, supporting our previous model that Sca2 can encircle at least two actin subunits[13].

What is contrary to the previous model, however, is that the circular fold observed here is formed by the N-terminal half (residues 34-670), and not an intramolecular interaction between the NRD and CRD. Furthermore, although the PRDs of Sca2 were expected to be semi flexible, the Middle domain of Sca2 binds one actin in the absence of the PRDs and is predicted to be mostly helical by secondary structure prediction and AlphaFold [13]. Both the PRDs and the Middle domain are absent from the Sca2 model. Finally, there is extra density observed covering one end of the circular fold, proposed to encompass the CRD, with only the other end open and able to accommodate actin subunits. Altogether, this represents a mechanistic conundrum: How does Sca2 elongate an actin filament and remain associated with the growing barbed end during elongation?

Parallels with Formins

In comparing with Formins, the FH2 dimer encircles the barbed end, and utilizes its semi flexible FH1 domain consisting of one or a few PRDs to bring in actin subunits via PRD-profilin-actin to elongate the filament. Upon formation of a ring-like intermediate, profilin is released from the barbed end, and thought to subsequently be released from the formin PRD (PRDs have a 10-fold lower affinity for profilin compared with profilin-actin [38]). While PRD mutations are somewhat tolerated in formins [39, 40], mutating either of the two PRDs in Sca2 has the same effect as a double PRD mutant in that these mutants equally fail to elongate [12]. Also, the PRDs are found on two separate FH1 domains in formins, contrary to Sca2 which harbors them in the same polypeptide chain. This suggests that the two PRDs cooperate in bringing in actin subunits for elongation, and that they are physically close to the growing barbed end. Although the PRDs are unaccounted for and it is difficult to predict their precise placement, the Sca2 model does not contradict these findings. Formin FH2 dimers are thought to track the barbed end by a two-state model involving conformational changes that bring one FH2 monomer closer to the terminal subunit exposing a region that is able to bind an incoming actin subunit (stair stepping model). This is made possible by the presence of a flexible linker within each FH2 monomer allowing for conformational flexibility within the FH2 dimer, and also by virtue of an FH2 dimer adopting a doughnut-like fold which is open from both ends. In the Sca2 model, one opening of the circular fold is covered by the CRD. At this point, it is unclear what conformational changes are involved in accommodating incoming actin subunits from profilin-actin, and how the circular fold translocates along the growing barbed end. Our structural data is a direct visualization of one stable conformation, and elongation may involve conformational changes and intermediates involving movement of the CRD and exposing the opening, which would allow for elongation in a formin-like manner.

Parallels with bacterial ParMRC

The controlled processive elongation of a filament containing a protein that tracks the growing end and protects it from capping is not exclusive to eukaryotic actins. Some interesting parallels can be drawn from examination of an analogous system: the ParMRC [28].

The ParMRC system, Par for partitioning, is part of a larger family of partitioning systems that rely on a polar actin-like filament (ParM) and an adaptor protein (ParR) that binds plasmid DNA on a centromeric region (ParC). It is responsible for the segregation of low copy-number plasmids after replication and during cell division into newly replicated daughter cells to ensure that each daughter cell faithfully receives a copy of the plasmid. ParM filaments are left-handed and undergo dynamic instability, akin to microtubules, and are stabilized upon ParR binding their barbed ends [41, 42]. To segregate plasmids to opposite ends of a dividing cell, ParM filaments bundle, without any additional cofactors, in an antiparallel manner, with their barbed ends pointing in opposite directions towards the edges of the dividing cell. ParR is able to mediate processive elongation of ParM filaments, while simultaneously binding ParC sites (10 base-pair repeat sequence on plasmid DNA) thus bringing along the plasmid [43].

ParR is a ~ 100 amino acid protein that adopts an RHH fold (Ribbon-Helix-Helix) found in DNA-binding proteins. It contains a disordered C-terminus that forms a helix and binds the barbed end of ParM in a site analogous to the hydrophobic-binding cleft in actin between subdomains I and III. ParR binds ParC sites as a dimer, and forms a larger 20-mer oligomer, or 10 ParR dimers. The assembly of this 20-mer ParR oligomer binds both ParC sites and the barbed end, and adopts an alpha-solenoid like shape, somewhat encircling the barbed end of ParM [44]. Similar to Sca2 and formins, ParR caps the barbed end of ParM, protecting it from depolymerization. Additionally, it promotes processive elongation of the actin-like ParM filament. Since the barbed-end tracking unit is a ParR 20-mer containing 20 sites that can bind ParM subunits (although not simultaneously) on the barbed end, with 2 of them already occupying the two terminal ParM subunits in the filament, the free sites that are close to the barbed end capture and bring in additional ParM subunits to the barbed end. Thus, it is thought that elongation proceeds through an insertional polymerization mechanism [28, 43, 45, 46].

Although Formins separate barbed-end binding and tracking (FH2 dimer) from capturing additional subunits for elongation (FH1 PRDs), both Formins (Figure 3B) and ParR (Figure 3C) bind the barbed-end using an open circular doughnut-like shape. Our current data suggests that Sca2 has one side of its circular fold inaccessible to actin subunits by the CRD. As such, another viable possibility for Sca2 elongation of actin may be through an insertional mechanism, in a manner analogous to ParR: The open side of the Sca2 circular fold formed by residues 34-670 may bind the two terminal subunits on the barbed end, with the CRD covering the other side of the circular fold. Region 671-1089, which was not observed in the electron density during reconstruction of the Sca2 model, contributes actin subunits to the barbed end through its two PRDs and profilin-actin, with region 869-1060 contributing to binding one actin subunit in some manner. During elongation, the barbed end tracking unit, formed by residues 34-670 and the CRD, translocates as more actin subunits are incorporated into the growing filament end (Figure 3A).

Figure 3. Proposed models for actin assembly by Sca2, Formins, and ParMRC.

(A) Sca2 nucleates actin by encircling two (or more) actin monomers, and elongation could proceed by an insertional mechanism, using its PRDs and profilin-actin. (B) Formins nucleate actin by encircling two (or more) actin monomers, and elongate by the addition of actin subunits from profilin-actin via their PRDs in their flexible FH1 domains. (C) ParR binds both the barbed end of the actin-like ParM (via it’s C-terminus) and ParC (via it’s N-terminus). The free ParM binding sites on unoccupied ParR are involved in bringing in ParM monomers to the barbed end for processive elongation. In the elongation model, ParRC is shown as a single semi-circular shape for simplicity. See text for details.

Conclusions

We used single-particle cryo-EM analysis to obtain structural information about the actin assembly factor from spotted fever group Rickettsia: Sca2. Perhaps due to the unexpected flexibility or another feature of this molecule, a high-resolution 3D model was unattainable. The data did allow for a low-resolution 3D reconstruction, however, and several important conclusions can be made from this data. First, the presence of a doughnut-like shape formed by an intramolecular interaction within the Sca2 monomer is confirmed. One opening of the circle appears covered, though. Using our previous NRD crystal structure and AlphaFold models of residues 401-670 and 1090-1355, we propose that the circle is formed by the residues 34-670, with the CRD contributing to the residues that cover one side of the circle. Second, region 671-1089, consisting of the two PRDs and Middle domain, are not present in the 3D reconstruction, owing to the flexibility of the PRDs. This flexibility may be important for the cooperativity of the two PRDs during elongation. Finally, the current structural data allows for updating the proposed model of Sca2 nucleation and elongation: The doughnut shape (residues 34-670) along with the CRD form the barbed end tracking unit which can encircle two or three actin subunits, with region 671-1089 responsible for bringing in actin subunits via profilin-actin for elongation. One possibility for elongation by Sca2, since one side of the doughnut is covered, is that it proceeds via an insertional mechanism (Figure 3A), analogous to elongation of ParM filaments by ParR (Figure 3C) or the recently reported elongation of actin by Vibrio VopL/F [47]. Another is through not-yet-observed conformational intermediates that expose the opening covered by the CRD in the structure observed in this study. The conformational changes associated with incorporation of actin subunits on the growing filament end and translocation of Sca2 to track the barbed end, are not yet clear. Further work using EM methods to obtain higher-resolution information about Sca2 bound to profilin-actin and barbed ends will provide valuable molecular details about actin assembly by Sca2, and, in more general terms, a novel formin-like mechanism for assembly of long unbranched actin tails.

• Sca2 adopts a formin-like circular fold with similar diameter.

• Sca2 proline-rich domains are, like formins, in a flexible region of the protein.

• The Sca2 structure allows for the proposal of an updated model of actin assembly.

Data availability

The density map and molecular model are available upon request.