The full-length espin 1 gene (GenBank accession No. NG_015866) was inserted downstream of the T7 promoter into the plasmid vector pET-30a (+). The resultant plasmid DNA constructs were transformed into Escherichia coli BL21 (DE3) star cell lines, and transformants were selected on LB agar medium (Luria Bertani Broth High Salt; Kisanbio, Seoul, Republic of Korea, Bacto Agar; Bacton, Dickinson and Company, Franklin Lakes, NJ) supplemented with 50 μg/ml kanamycin (Table 1). Transformants were grown in LB medium containing 50 μg/ml kanamycin at various temperatures with shaking at 180 rpm until the optical density (OD₆₀₀) reached 0.6–0.8 (4 h at 37°C). Protein expression was induced by adding IPTG at a final concentration of 1 mM for various induction periods (4 h at 37°C, 16 h at 18°C and 1 week at 4°C respectively). After IPTG induction, the cells were centrifuged for 30 min at 4°C with a speed of 5,000 x g (Beckman Coulter Inc., United States) to remove the LB medium, and the pellet was resuspended in lysis buffer solution containing 100 mM NaCl, 1 mM EDTA, 1 mM TCEP, 10% glycerol, and 50 mM Tris-HCl at pH 8.5. The resuspended pellet was sonicated for 15 min at 20% amplitude with 2 s intervals between each sonication pulse (Sonics, United States). The lysate was centrifuged for 30 min at 4°C with a speed of 20,000 x g (Beckman Coulter Inc., United States). The supernatant was collected and loaded onto a Ni-NTA column (GE Healthcare, United Kingdom), which was pre-packed and pre-equilibrated with lysis buffer containing 5 mM imidazole. Full-length recombinant espin 1 was eluted stepwise with 10 ml of 50, 100, and 200 mM imidazole gradients, and each eluted fraction was analyzed on a 10% SDS-PAGE gel.

Purification of gelsolin protein from bovine serum (Pel-Freez Biologicals, United States) was performed according to the previous study (Kurokawa et al., 1990), which was subjected to dialysis in a buffer consisting of 50 mM Na-acetate, 2 mM MgCl2, 1 mM EGTA, 1.1 mM CaCl2, 0.1 mM MgATP, and 30 mM HEPES at pH 7.5 then and stored at −80°C. Myosin S1 from scallop (Placopecten magellanicus) striated adductor muscles was prepared by standard procedure described in (Jung and Craig, 2008; Jung et al., 2008). Purified myosin S1 was further dialyzed in a same buffer used in actin polymerization. Skeletal actin from rabbit skeletal muscle was a generous gift from M. Ikebe laboratory and stored at −80°C before use.

The automatic ÄKTA pure chromatography system and the Superdex 200™ 10/300 GL column (GE Healthcare, United States) were used for SEC. Prior to the sample injection, the Superdex 200™ 10/300 GL column was connected to the automatic ÄKTA pure chromatography system, washed with filtered distilled water and equilibrated with espin 1 buffer containing 50 mM sodium acetate, 2 mM MgCl₂, 1 mM EGTA, and 30 mM HEPES at pH 7.5. The concentrated sample containing espin 1 was filtered through a 0.2 μm syringe filter and then carefully injected into the 0.5 ml capillary loop of the automated chromatography system, using a 1 ml syringe. SEC was performed at 0.4 ml/min, each 1 ml fraction was collected into a fraction collector and later detected at 280 nm UV wavelength by Ultraviolet (UV) detector in real time. The fractions corresponding to the UV peaks were analyzed using SDS-PAGE.

Skeletal actin was dissolved in a buffer consisting of 50 mM Na-acetate, 2 mM MgCl₂, 1 mM EGTA, 1.1 mM CaCl₂, 0.1 mM MgATP, and 30 mM HEPES at pH 7.5 and incubated to polymerize for 30 min at 25°C. This sample (5 μl) was applied to a glow-discharged, carbon-coated grid at 25°C, followed by negative staining with 1% uranyl acetate. The grids were examined using a Technai 10 (FEI, U.S) TEM operated at 100 kV, where the images were recorded using an UltraScan 1000 CCD camera (Gatan, U.S) at magnifications of ×7,000, ×17,000, and ×34,000 i.e., 1.47, 0.62, and 0.62 nm/pixel, respectively. The instrumentation was used at Kangwon Center for Systems Imaging and the images were further compiled and arranged using Adobe Photoshop CS6 (Adobe Inc., U.S). For the Ni-NTA nanogold labeling, based on the affinity of the his₆-tag for Ni-NTA, 5 nm Ni-NTA nanogold (Nanoprobes, Inc. NY, United States) was used to label the N-terminal his₆-tag of recombinant espin 1. Samples (F-actin or espin 1) were first loaded onto a carbon coated TEM grid to incubate for 1 min at 25°C. Gold drops (20 μl of gold solution) were applied onto a clean parafilm, and the grids with samples were turned upside down and allowed to float on the gold drops for 1 min incubation at 25°C. After the incubation, the grid was washed three times with buffer drops to remove any non-specific binding. Gold-labeled specimens were then examined from the negative stained TEM images.

Micrographs from TEM at a magnification of ×34,000 (0.32 nm/pixel) were subjected to single-particle analysis using the SPIDER program suite (System for Processing Image Data from Electron microscopy and Related fields; Health Research Inc., Rensselaer, NY, United States) according to procedures described previously (Burgess et al., 2004) in a 100 × 100 pixel window basis, with a globular shape and high frequency. A total of 1,384 particle sets were processed based on the “K-mean clustering” algorithm (Frank et al., 1996; Frank, 2009). All 2D classes were generated using SPIDER.

F-actin, myosin S1, gelsolin, and espin 1 at concentrations of 4 μM, 4 μM, 2 μM, and 1 μM, respectively, were incubated for 30 min at 25°C in a buffer solution containing 50 mM Na-acetate, 2 mM MgCl₂, 1 mM EGTA, 1.1 mM CaCl₂, 0.1 mM MgATP, and 30 mM HEPES at pH 7.5. Each mixture of F-actin, gelsolin, and espin 1 was centrifuged for 30 min at 4°C with a speed of 20,000 x g (Hanil, South Korea), followed by removal of the supernatant. An equal amount of the supernatant was compensated with fresh buffer in the pellet. The supernatant and pellet were analyzed using SDS-PAGE and TEM imaging.

Previous studies have shown that anionic surfactants such as sarkosyl detergents, are essential for heterologous protein expression of human espin 1, which render the synthesized proteins insoluble (Chen et al., 1999). The solubilizing effects of sarkosyl detergents are highly dependent on their concentration; the effects increase the viscosity of the samples, resulting in difficulties in further purification processes (Tao et al., 2010). In this study, a sarkosyl-free approach was used to express soluble espin 1 for the structural analysis. Among the several non-denaturing purification approaches that increase the solubility of recombinant proteins, the low-temperature incubation method was used. To determine the optimized temperature condition on the solubility of recombinant human espin 1, various temperatures ranging from 4°C to 37°C were screened. There were no significant differences in the solubility at 18°C through 37°C induction, but 4°C induction with long period of incubation (i.e., one week) showed notable changes in solubility of espin 1 (Figure 1A). Additional refinement steps were applied using affinity- and size exclusion chromatography; the SEC profile was observed as a single peak and SDS-PAGE results showed a significant major band at 100 kDa, indicating an espin 1 monomer (Figure 2). The functional actin bundling activity of the purified recombinant espin 1 was further verified using TEM (Supplementary Figure S1).

To visualize and confirm the appearance of purified espin 1, two-dimensional Class average (2D classification) (Figure 3A) and Ni-NTA nanogold labeling (Figure 3B) were done using SPIDER. The 2D class averages of full-length espin 1 suggested that the overall size of the particles was approximately 15 nm, corresponding to the monomeric size of the protein. This was further confirmed by nanogold labeling (Figure 3B; Supplementary Figure S2). Moreover, non-reducing PAGE showed a single band at 100 kDa, implying that purified espin 1 is in monomeric form (Figure 3C). To determine the actin bundling property of espin 1, Ni-NTA nanogold-labeled espin 1 was co-incubated with an actin filament (F-actin) (Figures 3D–F; Supplementary Figure S3). In the experiment, gold labeling was clearly observed along with F-actin, which represents espin 1 monomers binding the actin filaments (Figure 3E). Close-up views of the actin–espin 1 complex indicated that the espin 1 monomers cross-linked the two actin filaments (Figure 3E,F). Taken together, these results suggest that espin 1 monomers can induce actin bundling without multimerization.

Further studies were conducted to gain more detailed insights into the bundling mechanisms of espin 1. A co-sedimentation assay was performed with gelsolin, which is known to have an actin-severing activity by binding at actin subdomains 1 and 3 (Figure 4) (McLaughlin et al., 1993; Chumnarnsilpa et al., 2009). When espin 1 interacted with actin, both actin and espin 1 bands were found in the pellet, implying actin bundling formation (Figure 4A). In contrast, when gelsolin was treated with actin, the bands relevant to gelsolin were only found in the supernatant portion due to the actin severing activity (Figure 4A). Interestingly, when actin was mixed with both gelsolin and espin 1, gelsolin was found only in the supernatant regardless of the protein addition order, whereas espin 1 was found in pellets with actin (Figures 4A,B). Similar results were observed in the TEM-based structural analysis (Figures 4C–H). First, the actin-bundling activity of espin 1 and the severing effect of gelsolin were confirmed in this experiment (Figures 4D,E). Consistent with the results of previous gel experiments, actin bundling was observed in the actin–espin 1-gelsolin mixture, regardless of the order of addition of each protein (Figures 4F–H). This may be due to the actin binding of espin 1, preventing the access of gelsolin to actin or protecting actin from severing by espin 1 actin bundling.

To define a more conclusive espin 1 binding subdomain of F-actin, a co-sedimentation assay was performed with myosin S1, which is well-known to bind actin subdomain 1, following a concept similar to the gelsolin treatment experiment (Holmes et al., 2004; Lorenz and Holmes, 2010). Upon centrifugation, both proteins moved to supernatant, indicating a non-polymerized form (Figure 5A). When these proteins were treated with F-actin, the gel band shifted to a pellet band, indicating that each protein was bound to actin filaments (Figure 5B). Furthermore, when F-actin was treated with espin 1 and myosin S1, the SDS-PAGE showed co-localized bands of espin 1, myosin S1, and actin in the pellet, suggesting that these two proteins do not share the same binding site (Figure 5C). Myosin S1 is a well-known actin-binding part of striated muscle myosin containing myosin head which binds to actin subdomain 1 (Holmes et al., 2004). N-terminal ankyrin repeat (AR) of espin 1 is also known to interact with only Tail homology domain (THDI) of myosin 3 and there is no interaction is revealed with C-terminal actin binding domain and myosin head region (Liu et al., 2016). Thus, myosin S1 cannot bind espin 1 but only interact with F-actin indicating one of the espin 1 binding sites of actin is its subdomain 3, but not 1. The gel band thicknesses of espin 1 and myosin S1 showed differences in migration, which may be due to steric hindrance of the actin filament configuration. It is difficult for myosin S1 to bind to the core of the actin bundle after the latter is built by espin 1, and also for myosin S1 decorated actin filaments to form compact actin bundles that interfere with the binding of espin 1.