Acidianus filamentous virus 1 coat proteins display a helical fold spanning the filamentous archaeal viruses lineage

Two MCPs Form the AFV1 Core.

Only two proteins from the AFV1 virion could be identified by SDS gels and a trypsin peptide mass fingerprint (Fig. 1B), AFV1–132 (15 kDa, pI = 9.4) and AFV1–140 (15.7 kDa, pI = 5.4), which likely form the core of the filamentous virion. The proteins share 31% sequence identity but have no significant sequence similarity with any other proteins in public sequence databases. The first 50 residues of these proteins share 46% sequence identity and are rich in basic residues, while the sequence identity of the last 40 residues decreases to 14%. Furthermore, the last helix of AFV1–140 displays a nonambiguous amphiphilic character (Fig. 1C).

We expressed both proteins in Escherichia coli (see Materials and Methods). Significant shifts were observed in electrophoretic mobility shift assays (EMSA) in which genomic 760-bp dsDNA was incubated with AFV1–132 or AFV1–140. This assay demonstrates that both MCPs are DNA-binding proteins that likely interact with the viral genome (Fig. 1D). Furthermore, in vitro experiments mixing Lambda DNA fragments with AFV1–132 or AFV1–140 gave rise to long filaments, as observed by electron microscopy (Fig. 1 E and F). These filaments seem rather flexible, as manifested by straight and bent segments. The variation in both filaments' appearance in terms of thickness might be due to unspecific DNA-protein interactions as expected for a whole genome packaging proteins. Although these filaments are not directly relevant to the virion structure, they indicate that both AFV1–132 and AFV1–140 may participate in nucleocapsid assembly.

Crystal Structure of the AFV1–132 Protein.

Large orthorhombic crystals (250 × 31 × 15 μm) containing one cleaved monomer of AFV1–132 protein per asymmetric unit [Vm = 1.3 ų Da⁻¹; (28)] were obtained after 1 week in crystallization plates (Table 1). We collected two highly redundant datasets at the Cu Kα wavelength (λ = 1.54 Å) and solved the structure by Sulfur-SAD using an anomalous signal from three methionines (Met-74, Met-89, and Met-98), one cysteine (Cys-62), and one solvent-bound chloride ion. The structural model was refined at a resolution of 1.95 Å with R and R_free values of 0.158 and 0.198, respectively (Table 1), and counts for 100% in the most favored regions of the Ramachandran chart (29). The electron density map obtained for AFV1–132 contains residues between Tyr-51 and the C-terminal Tyr-132. Tight packing of the monomers and mass spectrometry measurements confirmed that AFV1–132 was cleaved during crystallization. Unexpectedly, ΔN_AFV1–132 and the C-terminal region (residues 52–134) of the basic major coat protein of the rudivirus SIRV (SIRV-134) were found to have an identical four-helix bundle fold (r.m.s deviation ≈1.4 Å for all Cα atoms) despite a very low 13% sequence identity detected only using structure-based alignment (Fig. 2 A and B). We were unable to produce crystals of full-length AFV1–132. These helical domains are stabilized by an extensive network of hydrophobic interactions in the protein core. Long helices ranging from 15 (α1 and α4) to 24 residues (α2) in length are connected with short loops of approximately three residues. The two longest helices, α2 and α3, are almost parallel with each other, whereas the two shortest, α1 and α4, are slightly rotated relative to each other and are twisted at a large angle (≈45°) relative to α2 (Fig. 2A). The C terminus of SIRV-134 exhibits a supplementary extended stretch of eight residues, in contrast with ΔN_AFV1–132 that ends with an α-helix. A Dali search for similar structures did not return any significant hits (PDB: 2ot3, Z-score: 7.7 to a subregion of the RAB5 GDP/GTP exchange factor). Therefore, the helical domains of SIRV-134 and AFV1–132 should be considered independently folded structures. Finally, both proteins have a large positive region probably involved in DNA-binding (Fig. 2C).

An important characteristic shared by AFV1–132 and SIRV-134 is the disorder of their N-termini. In SIRV-134, this disorder has been demonstrated directly by NMR (27), while in AFV1–132, we deduced the disordered character of the N terminus due to its rapid cleavage both in solution and during crystallization. Cleavage of AFV1–132 also occurs at the exact place where disorder begins in SIRV-134.

Crystal Structure of the AFV1–140 Protein.

Thin orthorhombic crystals of selenomethionine-labeled AFV1–140, containing four molecules per asymmetric unit (Vm = 2.21 ų Da⁻¹), grew in a few days in crystallization plates. The structure was solved by SAD at the Se K-edge (λ = 0.979 Å) and was refined at a resolution of 2.3 Å with R and Rfree values of 0.18 and 0.25, respectively. The final model accounted for 96.9 and 3.1% of residues being in the most favored and generously allowed regions of the Ramachandran chart, respectively (29). Native AFV1–140 crystallized under the same conditions as selenomethionine-labeled AFV1–140 but diffracted X-rays at a low resolution. Therefore, we used the Se-SAD structure of AFV1–140 for the rest of the description.

AFV1–140 encompasses five α-helices and has an elongated shape with overall dimensions of 82.5 Å × 23.5 Å × 26 Å (Fig. 3A). The first 14 residues (predicted as a positively charged helix) are not visible in the electron density map. At the N terminus, the first helix α'1 (15–24) is connected to the long 53 Å helix α'2 (32–67) via a seven-residue flexible region. ΔN_AFV1–132 and the C-terminal domain of AFV1–140 have very close structures (r.m.s deviation of 1.54 Å for all Cα atoms; Fig. 3A) despite their low sequence identity (23%) and their different hydrophilic nature. Even though some N-terminal residues are missing, electrostatic surface properties clearly identify the first helix as the prominent positive area (Fig. 3B). AFV1–140 crystals are made up of one bone-like tetramer (Fig. 3C) per asymmetric unit. This structure can be described as a dimer of dimers. Two monomers (A and C) associate with each other head-to-tail through a 1,398-Ų area of extensive van der Waals interactions, a network of hydrogen bonds, and one salt bridge that locks the α'1 helix (Lys-19-N^ζ) to the C-terminal domain (Asp-73-O^δ) at each extremity. The same dimer BD stacks onto the other such that the extended structures connecting α'1 and α'2 helices in each monomer cross each other and hug the terminal domain (Fig. 3D). This assemblage is stabilized by hydrophobic interactions involving eight detergent molecules of n-octyl-β-d-glucopyranoside (β-OG), which are essential for purification and crystallization. Four of these molecules are buried at the C-terminal domains (191 Ų buried surface area for each), and four are buried along the α'2 helix each of domain that is involved in the buried surface area of 135.8 Ų, 238.8 Ų, and 53.6 Ų with monomers A, B, and β-OG_143C, respectively (Fig. 3D). In contrast to AFV1–132, a large part of the N terminus of AFV1–140 is ordered, with the exception of the first 14 residues. Indeed, the packing of the AFV1–140 may contribute to the stability of the N terminus. This arrangement of dimer of dimers packs each N terminus domain against the three others in the crystal packing. This arrangement also forms well-defined detergent pockets. It should be noticed that AFV1–140 alone is monomeric in solution at low concentration (<0.1 mg/mL, the upper concentration limit in the absence of detergent). When β-OG is added, the protein can be concentrated up to 5 mg/mL. However, analysis of this solution using MALLS/UV/RI (see Materials and Methods) yields a broad unsymmetrical peak, revealing the presence of several species in solution, possibly from monomer to tetramers. Indeed, crystallization conditions may have displaced the equilibrium toward tetramers. Such associations are probably not directly relevant to biological oligomeric states, since positively charged helices are not accessible and in opposite directions. Nevertheless, as it is always the case for viral capsid proteins, AFV1–140 as well as AFV1–132 must likely be self-associated in some way to build the virion architecture.

Production of AFV1–132 and AFV1–140.

AFV1–132 and AFV1–140 were cloned into the Gateway™ destination vector pETG-20A (Dr. Arie Geerlof, EMBL Hamburg). The final construct encoded a thioredoxin fusion protein containing an N-terminal hexahistidine tag followed by a TEV protease recognition site upstream of the protein of interest. Fusion proteins were overexpressed in the E. coli Rosetta(DE3) pLysS strain (Novagen) grown in Terrific Broth (Gibco) at 37 °C for 4 h and at 25 °C overnight for AFV1–132 and AFV1–140, respectively (33). Overexpressed proteins were purified as follows: (i) using nickel affinity chromatography (His-Trap 5 mL column, GE Healthcare) with a step-gradient of imidazole; (ii) performing TEV protease cleavage; (iii) recovering cleaved proteins in the flowthrough of a second nickel affinity chromatography; and (iv) performing a preparative size-exclusion chromatography on a Superdex 200 HR26/60 in Tris 10 mM, pH 8, and NaCl 300 mM and on a Superdex 200 HR10/30 in Tris 10 mM, pH 8, NaCl 300 mM, and 30 mM n-octyl-β-d-glucopyranoside (Anagrade, Anatrace) for AFV1–132 and AFV1–140, respectively. Samples of AFV1–140 with n-octyl-β-d-glucopyranoside were analyzed using analytical size exclusion chromatography (SEC) on a HPLC system (Waters) with on-line multi-angle laser light scattering, absorbance and refractive index detectors (MALLS/UV/RI) (Wyatt Technology). The SEC was performed on a Shodex KW-403 4F column with Tris 10 mM, pH 8, NaCl 300 mM, and 30 mM n-octyl-β-d-glucopyranoside as the eluent. Selenomethionine-labeled AFV1–140 was prepared following standard procedures in the minimum medium M9 by blocking the methionine biosynthesis pathway. Expression, purification, and characterization of the SeMet-labeled AFV1–140 were carried out using the same protocols as for native protein.

Electron Microscopy of the MCPs and DNA.

AFV1–140 and DNA filaments were generated using a solution containing 13 mM CaCl₂, 13 mM MgCl₂, 1.3 mM NaN₃, and 83 ng/mL Lambda DNA cut with EcoRI and HindIII, and 1 mg/mL AFV1–140. The solution was adjusted to a pH of 3 with 5% H₂SO₄ and incubated at 78 °C for 7 days, and filaments were negatively stained with 2% uranyl acetate. Filaments of AFV1–132 and DNA were generated mixing 50 ng/μL Lambda DNA cut with EcoRI and HindIII and 3 mg/mL AFV1–132. Filaments were then incubated at 60 °C in a 10 mM Hepes buffer with a pH of 7 that contained 12.8 mM CaCl₂, 12.8 mM MgCl₂, and 1.28 mM NaN₃. Filaments were negatively stained with 1.5% phosphotungstic acid and were observed after 4 days of incubation on a 200 kV electron microscope at magnification, ×20,000.

Electrophoretic Mobility Shift Assay.

PCR-generated 760-bp dsDNA fragments from the AFV1 genome were independently incubated with AFV1–132 and AFV1–140 in TBE buffer (90 mM Tris-borate, pH 8, and 2 mM EDTA) supplemented with 40% glycerol at increasing dsDNA/protein molar ratios of 1/0, 1/4, 1/10, 1/50, 1/100, 1/200, 1/500, and 1/1,000. After 15 min incubation at 55 °C and then overnight incubation at room temperature, reaction products were visualized by ethidium bromide staining after electrophoresis for 8 h at 1 V/cm through 1.0% agarose gels run in TBE buffer.

Crystal Structures.

Crystallization of AFV1–132 was performed at 20 °C using the hanging-drop vapor-diffusion technique. Diffraction quality crystals (250 × 31 × 15 μm) were obtained after 1 week by mixing 1 μL protein at 12 mg/mL with 1 μL reservoir solution containing 12% PEG 8000 and 25 mM Na₂HPO₄/KH₂PO₄ with a final pH of 6. MALDI-TOF performed on crystals returned a molecular mass for the cleaved protein of ≈9,000 Da. After cryo-cooling in reservoir solution supplemented with 10% glycerol (vol/vol), a SAD experiment was performed in-house at the Cu Kα wavelength (λ = 1.54 Å) using a Bruker Axis Microstar rotating anode. Two SAD datasets were collected from a single crystal at different χ angles using a MAR345dtb image plate detector. Data were integrated with MOSFLM and scaled with SCALA (34). The PHENIX software suite was used to solve the structure (35). HYSS readily identified five peaks arising from three methionines, one cysteine, and one solvent-bound chloride ion. Initial phases were calculated with PHASER and were improved using RESOLVE (36). The structural model was automatically built with Arp/wArp 6.0 (37) and then refined at 1.95 Å using REFMAC5 (38).Crystallization of selenomethionine-labeled AFV1–140 was performed at 20 °C using the sitting-drop vapor-diffusion technique at 20 °C and a nanodrop-dispensing robot (Cartesian Inc.) in 96-well Greiner crystallization plates. Crystals grew by mixing 100 nL protein at 4.3 mg/mL with 100 nL of 17% PEG 550 MME and 0.1 M sodium cacodylate with a final pH of 5. A SAD experiment was performed on beamline ID14–4 (ESRF) at the Se K-edge (λ = 0.979 Å). One SAD dataset was collected at 100 K from a single crystal using an ADSC Q4 detector. Data were processed using MOSFLM/SCALA (34). Selenium substructure (16 sites) was solved using SHELXD and initial phases were calculated after density modification using SHELXE (39). Solvent flattening, noncrystallographic symmetry (NCS) averaging and partial model building were performed using RESOLVE (36, 40). This model was manually completed using COOT (41) and was refined at a resolution of 2.3 Å using REFMAC5 (38) and by applying three TLS groups (Translation/Libration/Screw rotation) (42) per monomer with R and Rfree values of 0.18 and 0.25, respectively (Table 1). The final model accounted for 96.9% and 3.1% of residues being in most favored and generously allowed regions of the Ramachandran chart, respectively (29). Native AFV1–140 crystallized in the same condition as the selenomethionine-labeled protein. One dataset was collected at the ESRF beamline ID14–2 using an ADSC Q4 detector. Native structure was solved by molecular replacement.