Fig. 2.
Surface properties of the putative membrane-binding face of the BDV-M tetramer. (A) The electrostatic surface potential reveals this face to be highly basic (areas colored in white, red, and blue denote neutral, negative, and positive potential contoured at 0, − 3, and +3 kT/e, respectively). View is rotated 180° around the horizontal axis relative to Fig. 1C. (B) Side view of the tetramer, obtained by rotating (A) by 90° about a horizontal axis, depicting BDV-M approaching a phospholipid membrane (Lower). The alternating surface charges (acidic/basic) of the tetramer edges would facilitate lateral assembly to form planar arrays.
In the crystal lattice, each tetramer is positioned face-to-face with a second tetramer via the putative membrane binding surface to form a weakly-associated crystallographic octamer. A positively-charged patch around Arg-98 of 1 tetramer juxtaposes the negatively-charged patch near the C terminus of the opposing tetramer. Because of the disk-like shape of the BDV-M tetramer, distances between the tetramers range from 3.95 Å (Pro 142 O–Arg 98 NE), through 6–8 Å at the edges, up to 20–25 Å between their centers, resulting in large solvent channels with dimensions 35 × 35 × 35 Å. Interactions between the 2 tetramers are mediated by solvent molecules, including SO42− ions. This finding is consistent with experimental results showing that higher oligomeric states of the M-protein appear after long incubation at high salt concentration and ultracentrifugation studies confirming that an equilibrium exists between tetramers and octamers of BDV-M (16). Because the membrane binding face is not accessible within the crystallographic octamer, it is unlikely that it is able to bind to lipid bilayers.
Electron Density Reveals Nucleotide Binding to BDV-M.
During the structure determination, additional electron density was found near to the C terminus of helix α4, close to the 4-fold axis of the BDV-M tetramer. The density could be interpreted as a pyrimidine mononucleotide, which we have assigned as cytidine-5′-monophosphate (Fig. 3). The base is sandwiched in a pocket between the side chains of Phe-37 (from loop β1–β2) and His-112 (of helix α4) at a distance of 3.4 Å to each (Fig. 3A); such a parallel ring stacking between bases and aromatic residues is typical for protein–nucleic acid interactions (28). A series of hydrogen bonds to the base are observed: between the main-chain atoms of Phe-37 and the base ring of the nucleotide, from the side chain of Gln-36, from the side chain of Asn-115# of a symmetry-related molecule and a water-mediated hydrogen bond to Asn-39 N, and the side chain of Asn 115# (Fig. 3A). In addition, the O2′ atom of a riboxynucleotide sugar moiety would hydrogen-bond to the side chain of Gln-36. The network of interactions observed for the cytosine base would be similar for uracil, with an extra hydrogen bond for the uracil N3 hydrogen. Binding of thymine is also conceivable, although the methyl group would make close contacts to the side chain of Asn-115#. In contrast to these pyrimidine bases, neither the size of the pocket nor the hydrogen-bonding pattern is suited to the binding of the bulky purine bases adenine and guanine. Surprisingly, RNA binding occurs on a face of the monomer opposite to that corresponding to the trinucleotide binding site in VP40 octamers (18) (Fig. S1 and Fig. S3).
Fig. 3.
The nucleotide binding site in BDV-M. (A) Experimental electron density (omit map contoured at 3 σ) with refined monoribonucleotide cytidine-5′-monophosphate at the interface between 2 monomers (view and colors as in Fig. 1C). The pyrimidine ring is sandwiched between the side chains of Phe 37-and His-112 and is held in position through hydrogen bonds to the main chain of Phe-37 and the carboxamide group of the neighboring Asn-115 (distances in Å). In addition, the O2′ atom of a riboxynucleotide sugar moiety would hydrogen-bond to the side chain of Gln 36. (B) Electrostatic potential surface of BDV-M (contour levels as in Fig. 2) with refined RNA trinucleotide (CCC). Note the alternative positions of the phosphate groups in the crystallographically 4-fold averaged structure.
Binding of Oligonucleotides to BDV-M.
In previous work, UV spectra of BDV-M oligomeric fractions showed 2 maxima with a ratio OD280 to OD260 of ≈1.1 (16), suggesting the presence of nucleic acid. Although the density described above is clear, showing that a pyrimidine base is bound to the heterologously produced BDV-M near the tetramer axis, the limited resolution and high overall B value of the data does not allow differentiation between RNA and DNA. Furthermore, the 4-fold averaging imposed by the crystallographic symmetry precludes identification of longer oligonucleotides that do not follow this symmetry. The electrostatic distribution at the protein surface suggests, however, that BDV-M could also bind larger nucleic acid fragments (Fig. 4).
Fig. 4.
Heterologously-expressed BDV-M binds ssRNA oligonucleotides. (A) Urea-PAGE of isolated nucleic acid from purified BDV-M tetramer. Left lane, defined 14-nt polyadenine marker; lane 1, isolated RNA labeled at the 5′ terminus using PNK; lane 2, as lane 1 after RNase T1 digestion; lane 3, isolated RNA labeled at the 3′-terminus using PAP; lane 4, as lane 3 after RNase T1 digestion; lane 5, polyadenine marker. These results prove the copurification of heterologously expressed BDV-M tetramer with RNA oligonucleotides of ≈16 nt. (B) Schematic depiction of possible binding mode of ssRNA. A distinctive basic patch along the tetramer diagonals could accommodate the polyphosphate backbone (yellow arrows), such that an incoming chain (3′-end bottom left) ends with the observed bound nucleotide (center). Assuming a specificity for cytidine/uridine, there are 3 possible exit routes for the 5′-end: (i) at the top left, with 1 pyrimidine base bound near the tetramer axis; (ii) at the top right, with 2 nt; and (iii) at the bottom right, with 3 central bases contributing to specificity. It is not possible to distinguish between them because of the 4-fold crystallographic symmetry.
Freshly-prepared BDV-M was analyzed to investigate both the type and length of oligonucleotides bound to the protein. Nucleic acids isolated from BDV-M were alternatively radioactively-labeled at the 5′ and 3′ ends and analyzed by using urea-PAGE (Fig. 4A). 5′ Labeling using T4 polynucleotide kinase (PNK), which can label both DNA and RNA nucleotides, shows a strong band corresponding to a 16-nt fragment. Digestion with RNase T1 followed by urea-PAGE separation reveals degradation of the nucleic acid fragments, providing evidence that the nucleic acid in question is indeed RNA. 3′ Labeling using polyadenylate polymerase (PAP), specific for ssRNA nucleotides, reveals 13- and 16-nt fragments. RNase treatment of the sample also results in degradation of the nucleic acid fraction, although the ensuing fragments are larger than for the digested 5′-labeled PNK sample. Pretreatment of the BDV-M protein fraction with either RNase A or the dual specific endonuclease Benzonase before nucleic acid isolation gave equivalent results, showing that the RNA oligonucleotide is protected by the M-protein. Conversely, attempts to remove the nucleic acid via denaturation of BDV-M and subsequent in vitro folding were unsuccessful until now, raising the possibility that RNA binding plays a structural role in the integrity of the tetramer. Treatment of both the 3′- and 5′-labeled isolated nucleic acid samples with RNase A yielded almost complete digestion to mononucleotides. Thus, both labeling strategies provide evidence that BDV-M expressed heterologously in Escherichia colibinds and protects ssRNA oligonucleotides with a length of ≈16 bases or more.
Combining the crystallographic and in vitro data, it is likely that the BDV-M tetramer binds RNA oligonucleotides containing 1, 2, or 3 consecutive pyrimidine bases; 4 bases cannot be bound simultaneously because of steric restraints (Fig. 4B). For a better understanding and more realistic view of RNA bound to the BDV-M tetramer, we modeled and refined a cytidine trinucleotide by linking 3 symmetry-related CMP molecules (Fig. 3B) in 2 possible orientations: 3′–5′ and 5′–3′. Only 1 orientation of the trinucleotide could be refined to fit the electron density map with reasonable geometry and protein contacts without increasing the crystallographic R factors. Conformational parameters for the refined mononucleotide and trinucleotide are presented in Table S1. Refinement as a trinucleotide alters the nucleotide conformation compared with the mononucleotide model slightly: although the pyrimidine bases barely change their positions, locations of the phosphate groups and sugar moieties change as a consequence of the covalent linkage of the mononucleotides. The largest shifts are observed for the 5′ nucleotide, whose pyrimidine ring is shifted ≈0.5 Å and free phosphate group moved 2.2 Å compared with the mononucleotide refinement. Similar results are obtained upon refinement of a dinucleotide.
The refined triribonucleotide model fits well to the (4-fold crystallographically averaged) electron density and allows tentative interpretation of weak residual electron density close to each terminal phosphate group as belonging to preceding and following nucleotides. Inspection of the BDV-M electrostatic potential indicates a possible course for longer polynucleotide chains along the surface of the tetramer (Fig. 4B), corroborated by the locations of bound sulfate ions in the crystal. For minimal strain in the ssRNA near the tetramer axis, a single BDV-M tetramer should bind an oligoribonucleotide segment containing 2 consecutive pyrimidine bases, although other binding modes (1 nucleotide or 3 consecutive bases) are also feasible (Fig. 4B). It is possible that further RNA specificity pockets are present on the surface, but are undetec