Crystal structure of the fibre head domain of bovine adenovirus 4, a ruminant atadenovirus

Sequence analysis suggested that residues 414–535 might form the C-terminal head domain. Therefore, expression vectors for the full-length sequence (1–535), a C-terminal fragment with part of the predicted shaft domain (residues 248–535) and the putative fibre head domain alone (amino acids 414–535) were constructed. All three constructs expressed the expected protein, but soluble protein was only obtained for the putative fibre head domain. Expression was carried out at low temperature and the protein was purified by metal affinity chromatography and ion exchange chromatography as described in the Methods section. In cation exchange chromatography, the protein eluted in five different peaks (Fig. 2). N-terminal sequence analysis by Edman degradation of protein from the five peaks yielded the same sequence (GSSHH), suggesting the N-terminal six-histidine tags to be intact and the N-terminal Met to be removed upon expression in E. coli. Mass spectrometry on protein from the five peaks suggested adducts of around 180 Dalton, which could correspond to spontaneous alpha-N-6-phosphogluconoylation of the six-histidine tag, as observed for other proteins with the same N-terminal six-histidine tag [46]. Multiple peaks eluting from cation chromatography were also observed for a different protein expressed in E. coli with the same N-terminal six-histidine tag [46]. Careful inspection of the gel in Fig. 2b suggests the presence of two bands. The presence of two bands may well be caused by differential alpha-N-6-phosphogluconoylation. The fact that the protein is a trimer may lead to mixed species and thus the generation of the multiple peaks observed. The fact that the earlier peaks contain more adduct is consistent with the cation exchange chromatography behaviour, because the adduct would be expected to remove a positive charge and lead the protein to elute at lower salt concentrations. Protein from the three major peaks b, c and d was concentrated separately to 12 mg/ml (peaks a and e were considered to contain too little protein for successful crystallization). Around 13 mg of purified protein in total was obtained per litre of expression culture. Adenovirus fibre heads eluting in several peaks from ion exchange chromatography has been observed before [24, 47].

Well-diffracting crystals were obtained from all of the three major cation exchange elution peaks at 21 °C, by sitting drop vapour diffusion from precipitant solutions containing 20 % poly-ethylene glycol 3350 and either 0.2 M potassium thiocyanate or 0.2 M sodium isothiocyanate. Crystals appeared after three days and grew to their maximal size in about three weeks. They were found to belong to space group P1, with one protein trimer in the unit cell, leading to a solvent content of 44 % and a Matthews coefficient of 2.2 [48]. A calculated self-rotation function confirmed the presence of non-crystallographic three-fold symmetry in the crystal. We were not successful in solving the structure by molecular replacement, although the availability of a high-quality isomorphous derivative dataset meant we did not pursue this method extensively. High-resolution isomorphous datasets were collected from the native and derivative crystals, and the derivative dataset contained high-quality anomalous signal, which allowed automated structure solution by single isomorphous replacement with anomalous dispersion. The final refined model contains residues 420–534 from each of the three protein chains in the trimer, plus ordered solvent atoms. No reliable density was observed for the N-terminal purification tag, for residues 414–419 or for the C-terminal glutamine (amino acid 535), suggesting that these are disordered. No differences were observed between crystallization success, crystal shape or form, for protein obtained from the 3 central peaks; which is consistent with that they presumably only differ in the adduct on the disordered N-terminal purification tag, for which there is room in the crystal lattice. The non-observed residues 414–419 may form a linker between the shaft and head domain in the intact fibre protein. Data collection, phasing and refinement statistics are shown in Table 1.

The structure of the BAdV-4 fibre head domain is composed of three monomers, associated into a compact trimer (Fig. 3). Each monomer consists of a beta-sandwich of two beta-sheets (ABCJ- and GHID-sheets; Fig. 3a), in the same topology (Fig. 3c) as other adenovirus fibre heads (the topology of the HAdV-5 fibre head is shown as an example in Fig. 3d) [7, 47, 49]. The same topology is also shared with reovirus fibre heads [50, 51], the bacteriophage PRD1 fibre head [52] and lactobacillus phage receptor-binding domains [53, 54], as previously described [23]. The loop between strands D and G is longer than the others and contains an alpha helix (residues 471–476), instead of the E- and F-strands that many mastadenovirus fibre head domains have in this loop (Fig. 3d) [49]. The A-, B-, G- and H-strands are relatively long (eight, ten, nine and 13 residues, respectively), while the C-, D-, I- and J-strands are shorter (six, five, six and six residues, respectively). The AB-, BC-, GH- and HI-loops are short beta-turns of four amino acids each, while the CD-, DG- and IJ-loops are longer (twelve, 16 and 14 residues, respectively). Similar as observed for other adenovirus fibre heads, the CD- and IJ-loops are located at the top of the fibre head, whereas the DG-loop is located on the side. All are potentially involved in receptor interactions. The CD- and IJ-loops run parallel to each other and contain negatively charged residues, forming electronegative patches on the protein surface (Fig. 4a). The CD- and IJ-loops are also involved in inter-monomer contacts with the J-strand of the neighbouring monomer.

The two beta-sheets of the BAdV-4 fibre head monomer face each other at an angle of around 120°. The ABCJ-sheet is mostly on the “inside”, and only a part of the AB beta-hairpin is solvent-accessible. The rest of the AB beta-hairpin and the whole C- and J-strands are buried and extensively involved in inter-monomer contacts. In contrast, most of the GHID-sheet is solvent-accessible (Fig. 3b). Together with the alpha-helix, the two beta-sheets from each monomer form a compact globular trimeric beta-propellor.

Mastadenovirus fibre heads are larger than atadenovirus fibre heads, as previously described [23], mainly due to the surface loops being longer. The AB-, CD-, DG-, HI- and IJ-loops are all shorter in the BAdV-4 fibre head than in the prototype mastadenovirus fibre head structure, HAdV-5. Exceptions are strands G and H and the BC-loop, which are longer in the BAdV-4 fibre head. The low similarity of the structures makes it impossible to make meaningful speculations about receptor binding.

The BAdV-4 fibre head shares only 15 % sequence identity with the other atadenovirus fibre head (from SnAdV-1) for which a structure is known [23]. Nevertheless, the structure is very similar (Fig. 5). This is surprising considering that today one of these viruses exists in cattle (after a supposed host switch from reptiles) [31, 55] and the other still in snakes, with which they presumably continuously co-evolved. The two trimers can be superposed with an overall Z-score of 13 and a root mean square deviation (r.m.s.d.) of around 2 Å. When monomers are superposed, the r.m.s.d. is only slightly lower (1.8 Å), indicating that the relative orientation of the monomers in the trimer is also very similar. The topologies of both fibre heads are identical, with a conserved beta-sandwich motif and an alpha-helix in the DG-loop. Both fibre heads share the same 120° angle between two beta-sheets, but differences are observed in the length and the conformation of the loops.

The CD- and IJ-loops of the BAdV-4 fibre head are longer than those in the SnAdV-1 fibre head. In contrast, the DG-loop of the BAdV-4 fibre head is two amino acids shorter than its SnAdV-1 counterpart. A noticeable difference is observed in the length of the G- and H-strands, which are quite a bit longer in the BAdV-4 fibre head (9 vs. 5 for the G-strand and 13 vs. 8 residues for the H-strand). As mentioned before, the CD-, IJ- and GH-loops are all located on the top of the molecule, while the DG-loop is on the side, making all of them potentially important for receptor interaction. In a structure-based alignment (Fig. 5c), the lack of sequence similarity becomes even more evident, except for the very C-terminal part of the protein (end of the H-strand and the I- and J-strands and the HI-turn). It may be speculated that this conserved part of the protein is important for initiation of protein folding and has therefore diverged less. When calculated electrostatic surfaces are considered, the BAdV-4 fibre head has distinct electro-negatively charged patches on its surface (Fig. 4a), while the SnAdV-1 surface is more electro-positive (Fig. 4b) [23]. This may have implications for receptor-binding and suggest they do not bind the same receptor, although the nature of the receptors for neither of the viruses is known.

From previous observations in other adenoviruses, the BAdV-4 fibre head was expected to be a stable protein. However, unlike other head domains such as that of the fowl adenovirus 1 long fibre [56] and of the SnAdV-1 fibre [57], this trimer is not denaturant-stable at room temperature (Fig. 2b). Its melting temperature, as measured by a thermofluor assay, was estimated to be 67 °C, which is relatively high. In the same assay, no unfolding transition was observed for the SnAdV-1 fibre head below 95 °C, confirming the SnAdV-1 fibre head is more stable. Each monomer has a total solvent accessible surface area of 6.6 × 10³ Å², out of which 0.62 × 10³ Å² (9 %) gets buried upon trimer formation. In the SnAdV-1 fibre head, about 1.5 times more surface area is buried [23]. No salt bridges are formed upon trimeric assembly for the BAdV-4 fibre head, while in the SnAdV-1 fibre head a strong bidentate salt bridge between Arg304 and Glu333 of neighbouring monomers is present, which may explain why the SnAdV-1 fibre head is more stable than the BAdV-4 fibre head. Table 2 lists the identified interactions, from which it can be concluded that the SnAdV-1 fibre head has more hydrophobic, aromatic and ionic interactions between monomers, while the number of inter-monomer hydrogen bonds is the same.

Ovine adenovirus 7 (OAdV-7), another ruminant atadenovirus, is intensively tested as a basis for a gene therapy vector [16], but the structure of its fibre head is unknown. Our BAdV-4 fibre head structure is likely similar to that of the OAdV-7 fibre head [27], with which it shares 44 % sequence identity (51 out of 116 amino acids; Fig. 5d). It should now be possible to build a reliable structural model of the OAdV-7 fibre head in order to identify surface residues possibly involved in receptor binding and mutating them. Our cloning, expression and crystallization strategy may also be used to obtain crystals for the OAdV-7 fibre head, and the structure determined experimentally by molecular replacement using data collected on these crystals.