This study advances our understanding of the secondary, tertiary, and quaternary structures of δAg that are essential for HDV replication. Previous studies have identified a dimerization domain (CCD) spanning amino acids 12-60 (). In particular, the crystal structure of a synthetic 12-60 peptide revealed an anti-parallel coiled-coil dimer in which over 90% of the residues (45 out of 49) are α-helical (Zuccola et al., 1998
) consistent with earlier CD data (Rozzelle et al., 1995
). From our CD measurements on full-length δAg we deduced that the remainder of the protein contains very little additional secondary structure (). We arrived at a similar conclusion using the recently developed meta-predictor PONDR-Fit (Xue et al., 2010
) which combines six neural network programs trained on a large number of regions of known protein disorder. Analysis of the δAg sequence studied here () as well as for all 8 clades of HDV () predicted high levels of disorder (P-Fit scores >0.75) throughout the protein, except for parts of the CCD region. There was no evidence for additional ordered regions outside the CCD that are conserved between clades. PONDR-VLXT, one of the six components of PONDR-Fit, predicts somewhat higher levels of order, both in the CCD and two or three segments in the C-terminal half of δAg (especially for clades 5-8). This algorithm recognizes short regions experimentally known to be disordered that become structured when they are bound to other proteins (Obradovic et al., 2005
). This supports the possibility that δAg multimerization promotes helix formation in the CCD while isolated segments within the initially disordered C-terminal region are poised to become ordered upon interaction with other cellular proteins.
Intrinsic disorder is increasingly recognized for its relevance in understanding the intra- and inter-molecular interactions of important proteins (Dunker et al., 2008
). In the case of viral proteins that are expressed in cells in relatively large amounts, disordered regions may increase the capacity to structurally adapt for engagement in a wide variety of homo- and heteromultimeric interactions (Goh et al., 2008a
); during HDV replication δAg interacts with more than 100 different host proteins (Cao et al., 2009
). It is interesting to compare δAg with apoptin, a protein of chicken anemia virus. Like δAg this protein is small (125 amino acids), basic, essential for replication, binds nucleic acids, and has not yielded to crystal-structure determination. Apoptin forms homomultimers of 30-40 molecules and is known to interact with several different host proteins (Leliveld et al., 2003a
; Leliveld et al., 2003b
; Los et al., 2009
; Teodoro et al., 2004
). Comparison of the PONDR-Fit profiles of apoptin with that of δAg () show somewhat higher degrees of order centered around residue 45 with flanking regions of predicted disorder, similar to the patterns observed for the N-terminal 120 residues of δAg. (Incidentally, the N-terminal 69 amino acids of apoptin have been demonstrated to account for its multimerization behavior (Leliveld et al., 2003b
).) However, a much higher fraction of the residues in δAg are predicted to be highly disordered suggesting that intrinsic disorder might be even more relevant to understanding the properties and functions of δAg in the HDV life cycle.
Qualitative insight into δAg quaternary structure was obtained by denaturing gel electrophoresis. Prior cross-linking revealed the existence of multimers formed both by the purified protein and that present in virus-like particles. With the purified δAg the complexes were detected when the proteins were at 0.2 and 2 μM during cross-linking ( and data not shown). Much higher concentrations are present in cells during HDV replication, suggesting that multimers should also arise in vivo
; for example, the observed accumulation of 2.4 million copies of δAg in a nucleus of ~9 μm diameter represents an average concentration of 24 μM (Chang et al., 2005
Relative to protein standards the majority of cross-linked multimers appeared to contain 6 - 8 molecules of δAg (). This confirms and extends previous mass spectrometry studies in which octamers were detected although not quantified (Zuccola et al., 1998
). In addition, the hydrodynamic dimensions observed in our dynamic light scattering measurements, were consistent with an oligomer of ≤12 subunits. The δAg multimers we observed were remarkably stable even in the presence of high NaCl or urea concentrations (Fig. S2
The multimers of full-length δAg did bind nucleic acids but without specificity; HDV RNAs with and without rod-like folding and of different lengths were bound, as were non-HDV RNAs, and double- and single-stranded DNAs (Figs. , and S3
). This binding is consistent with simple electrostatic interactions. The δAg has a predicted net charge of +12 at neutral pH (Kuo et al., 1988
) and multimers will have a greater total charge and are thus expected to have enhanced affinity for nucleic acids that have one negative charge per nucleotide. Similar electrostatic interactions with nucleic acids have been reported for the apoptin mentioned earlier, which is also basic and forms large oligomers (Leliveld et al., 2003a
). We also found that nucleic acid binding was achieved when the protein was first cross-linked (), indicating that δAg rearrangement was not necessary during binding. Observations that the binding was reversed with increasing concentrations of NaCl () or vanadyl ribonucleosides (VRC)(), agrees with the interpretation of an electrostatic interaction. The findings are also consistent with our earlier studies on natural HDV ribonucleoprotein complexes where VRC released the HDV RNA but left the δAg complexes intact (Dingle et al., 1998b
). However, unlike the natural situation, the purified δAg did not demonstrate specificity for rod-like HDV RNA.
The ability of δAg to bind non-specifically to nucleic acids might explain its behavior as a chaperone. In a series of papers Wang et al
. showed that δAg would facilitate strand exchange reactions between short DNA oligonucleotides (Wang et al., 2003
). We confirmed their results including evidence that the chaperone effect can be achieved with N-terminal and C-terminal fragments of the δAg lacking the oligomerization domain (G. Moraleda, H.J. Netter, and J.M. Taylor, unpublished observations). Thus, the chaperone activity of δAg appears to be independent of its ability to oligomerize. It seems to be a property of a positively charged protein independent of multimerization.
The present findings of non-specific binding of δAg to nucleic acids also appear to disagree with earlier studies from this lab by Chao et al
. which indicated that a delta antigen would bind to HDV RNAs but only if they had the ability to form rod-like folding structures (Chao et al., 1991
). However, the assay used in the previous study was quite different: the protein tested was a fusion protein and the detection method was a northwestern. The present findings also disagree with recent studies that employed a δAg with an N-terminal fusion and a C-terminal deletion in a mobility shift assay and concluded that binding to HDV RNA required a minimum of ~311 nt of rod-like folding (Defenbaugh et al., 2009
; Lin et al.). However, it should be noted that when a non-truncated form of fusion protein was used, no specificity was detected in the RNA binding assay.
Further in vitro
experiments are thus needed to determine whether or not we can identify features of the full-length protein and/or the RNA substrates that will produce specific interactions. Recent in vitro
studies with the full-length core protein of HBV have also encountered RNA interactions that are efficient but not specific for viral RNA (Porterfield et al., 2010
). However, it remains to be seen how and to what extent the in vivo
interactions can be specific for viral RNA. Further work is needed to gain a more detailed understanding of the mechanism of in vivo
self-associations of δAg and its interactions with HDV RNAs as well as host proteins and nucleic acids. We have reported that in vivo
δAg is present in large complexes, even in the absence of HDV RNA genome replication or when HDV RNAs are present but released by treatment with VRC (Dingle et al., 1998b
) or RNase (Chang et al., 2008
; Dingle et al., 1998a
). In addition, a recent study reported the identification by mass spectrometry of more than 100 host proteins that interact with δAg and began to test how these might contribute to supporting HDV replication (Cao et al., 2009
). Sorting out which are the relevant associations is non-trivial because millions of copies of δAg are produced per cell during HDV replication, most of which are in the nucleus at concentrations exceeding 24 μM (Chang et al., 2005
), much higher than the 0.2 - 2 μM used for the in vitro
studies reported here. A further in vivo
complication is that in the presence or absence of HDV RNA, δAg is found in either the nucleoplasm or nucleolus, respectively (Han et al., 2009
). That is, the same high concentrations of δAg can assume different intranuclear associations depending upon whether or not HDV RNA species have been allowed to replicate.