HDV Produces Two Forms of HDAg from the Same Gene

Early analyses of HDV proteins showed that there are two electrophoretic forms of HDAg, but it was not clear how these forms differed biochemically and functionally.^5-85-8 (These forms were sometimes referred to by their apparent molecular weights, p-24 and p-27; they are denoted here as HDAg-S and HDAg-L for short and long, respectively.) Following the cloning of HDV cDNAs,^9,10 a series of studies illuminated the functional roles of HDAg-S and HDAg-L in HDV replication. Taylor's group showed that while HDAg-S is required for replication of HDV RNA, HDAg-L inhibits replication.,^11,12 Subsequent studies from several laboratories showed that HDAg-L interacts with the envelope protein of the helper virus, HBV, and is required for the formation of HDV particles.,^13-15

Despite these advances, the mechanisms leading to the formation of HDAg-S and HDAg-L were not clear. Cloning and sequencing of the genome in 1986 indicated heterogeneity at several positions in the 1679 nt genome;⁹ most of this heterogeneity involved single base transitions (A vs. G and C vs. U). Although the significance was not fully understood at the time, heterogeneity at one position, in particular, proved to be important. This position is near the end of the single conserved open reading frame, which is in the antigenome sequence (HDV is a negative strand RNA virus). In some clones this position was found to be adenosine, which is part of the UAG (amber) stop codon of HDAg-S; in other clones it was found to be guanosine, which changes the codon to UGG (W, tryptophan) and thereby extends the open reading frame by 19 amino acids.^9,16Expression of protein from clones that contained either the UAG or UGG sequence showed that the former encoded HDAg-S and the latter HDAg-L.^16,17Still, it was not clear how this heterogeneity fit into the HDV replication scheme.

An important advance came when HDV infection was initiated in chimpanzees by transfection with a cloned HDV cDNA expression construct that could initiate HDV replication.¹⁸ Remarkably, while the transfected genome encoded only HDAg-S, both HDAg-S and HDAg-L were detected in the liver and in HDV particles isolated from the serum of the infected chimpanzee. ¹⁸ This result was further extended by analysis of HDV replication in cultured cells: careful examination of western blots indicated that HDAg-L also became detectable several days after transfection of cultured cells with a cDNA clone that initiated HDV RNA replication and that encoded only HDAg-S.¹⁹ No HDAg-L was detected when cells were transfected with an expression construct for HDAg-S that did not produce replicating HDV RNA. Thus, the appearance of HDAg-L was linked to HDV replication.

Analysis of HDV RNA isolated from the serum of the transfected chimpanzee and from transfected cultured cells showed that heterogeneity appeared at the position corresponding to the adenosine in the UAG stop codon for HDAg-S.¹⁹ This finding recalled the heterogeneity observed at this position during the sequencing of the original isolates of HDV and led to the suggestion that this heterogeneity was responsible for the appearance of HDAg-L. Thus, during the course of HDV replication, some genomes encoding HDAg-S are converted, or edited, to encode HDAg-L. Because of the different functions of HDAg-S and HDAg-L, editing is part of a classic switch from viral RNA replication to packaging, and plays a central role in the HDV replication cycle.

HDV Antigenomic RNA Is Edited by the Host RNA Adenosine Deaminase ADAR1

One difficulty encountered in establishing the mechanism of editing at the amber/W site was identifying the RNA substrate: assays performed on replicating RNAs could not definitively determine whether the substrate for editing was the genome or the antigenome, or even whether editing was the result of cotranscriptional misincorporation. Initial attempts led to the erroneous suggestion that the genomic RNA might be the substrate, in which case editing would occur as a U to C transition.^20,21However, the creation of nonreplicating RNA expression constructs that could exclusively produce either genomic or antigenomic RNA in transfected cells led to the unambiguous conclusion that editing occurs on the antigenome RNA.²² This result was further supported by analysis of editing on in vitro-transcribed RNAs mixed with nuclear extracts: only antigenomic RNA was edited at the amber/W site.²² This observation indicated that HDV editing occurs post-transcriptionally, and is not the result of transcriptional misincorporation.

Because HDAg is encoded on the HDV antigenome (HDV is a negative strand RNA virus) editing on the antigenome is consistent with A to I editing by RNA adenosine deaminase,¹⁹ an enzymatic activity which is present in nuclear extracts from numerous metazoan species and can edit adenosines in double-stranded RNAs. Subsequently, it was shown that RNA adenosine deaminase (ADAR) from Xenopus laevis can edit the amber/W site in the HDV antigenome with considerable specificity in vitro.²³ Because the HDV amber/W site was edited with high specificity in vitro using just purified ADAR, no additional factors aside from HDV RNA and RNA adenosine deaminase are required for amber/W site editing to occur.

In mammalian cells two related genes, ADAR1 and ADAR2, have been identified that encode proteins capable of editing adenosine in RNA.^24,27 These proteins contain a catalytic deaminase domain along with 3 or 2, respectively, copies of dsRNA binding motifs (DRBMs). Both genes are essential for viability in mice.^28,29Substrates for ADAR1 and ADAR2 include a number of host premRNAs,^30,31 including the glutamate receptor subunit B (gluR-B) Q/R and R/G sites, the gluR-5 and gluR-6 Q/R sites, a splice site in ADAR2, and several sites in the serotonin receptor 5HT2CR.³¹ Both ADAR1 and ADAR2 can edit HDV RNA at the amber/W site in transfected cultured cells.^32,34While ADAR1 and ADAR2 are expressed in a variety of tissues, levels of ADAR2 are highest in brain, and the level of ADAR1 mRNA expression is considerably higher than ADAR2 in liver. Knockdown experiments using siRNA have shown that the short form of ADAR1, which is localized in the nucleus, is responsible for amber/W site editing during HDV replication.^35,36

In the current model for RNA editing in the HDV replication cycle (Fig. 2), the amber/W site in full-length antigenome RNA is edited from A to I by the host RNA adenosine deaminase ADAR1. The edited antigenome serves as template for the synthesis of genome RNA that then contains C at the position complementary to the amber/W site in the antigenome. This genome subsequently serves as template for transcription of mRNA that contains the UGG tryptophan codon and therefore encodes HDAg-L. It is important to note that, in this model, HDV mRNA is not edited directly, unlike cellular mRNA substrates for RNA adenosine deamination. Rather, editing occurs on the full-length antigenome, which is a replication intermediate. Consistent with this model, sequences required for forming the structure required for editing (see below) are more than 300nt downstream of the polyadenylation and ribozyme sites, and are not included in the mRNA sequence. Furthermore, analysis of RNA in viral particles indicates that genome RNAs contain the expected C at the position complementary to the amber/W site.

Figure 2

The role of RNA editing in the HDV replication cycle. (Adapted from Polson et al 1996). 1, Synthesis of mRNA encoding HDAg-S; 2, translation of HDAg-S, which is required for RNA replication; 3, replication of full-length antigenomic and genomic RNA; 4, (more...)

Cis Elements Required for Editing

ADAR1 and ADAR2 exhibit distinct but overlapping substrate specificities. Although some substrates, such as the gluRB Q/R site, show a clear preference for one ADAR, many substrates, including the HDV amber/W site, can be edited by both. Inspection of the predicted structures of known editing sites reveals several common features (Fig. 3). All substrates include at least 6 contiguous base pairs around the editing site, and many substrates contain more. Base-pairing on the 5' side of sites varies between 2 and 5 base-pairs. On the 3' side base-pairing is greater, in most cases extending for at least about 20 base-pairs that are disrupted by mismatches, bulges and internal loops. The role of these disruptions may be to position the ADAR protein via the double-stranded RNA binding domains such that the deaminase domain is positioned correctly at the editing site.^31,37-39 In an apparent contradiction to this idea, some studies have suggested that extensive base-pairing 3' of editing sites may not be essential for efficient editing. Sato and Lazinski³⁴ showed that a minimal substrate that was derived from the HDV amber/W site and that contained only 8 base-pairs could be efficiently edited, and Herbert et al⁴⁰ showed that ADAR1 could efficiently edit even when the three DRBMs were removed. Both of these studies examined editing in transfected cells expressing high levels of ADAR, and could indicate that the deaminase domain itself possesses some RNA binding activity that can be effective under certain conditions.

Figure 3

Predicted RNA secondary structure around known sites for RNA adenosine deamination. The structure shown for the gluRB Q/R site was obtained using the RNA secondary structure prediction program mfold,56 and is different from another proposed structure.57 (more...)

It is interesting to note that the base-pairing on the 3' side of the HDV amber/W site is more frequently disrupted by bulges and mismatches than other editing substrates. Mutations that improve base-pairing in this region in the HDV genotype I RNA lead to increased editing and markedly decreased viral RNA replication (Sato and Lazinski, personal communication, Jayan and Casey, unpublished data). There are two likely reasons for the presence of these disruptions 3' of the HDV amber/W site. First, the more extensive base-pairing found in other editing substrates could interfere with HDV RNA replication (either because RNA structures required for viral RNA replication cannot tolerate extensive base-pairing, or because the base-pairing may trigger cellular antiviral responses to dsRNA segments). Second, the disruptions may be important for preventing excessive editing (Lazinski and Sato, personal communication; Jayan and Casey, unpublished data); that is, the HDV editing site (at least for genotype I) may be selected to be sub-optimal in order to prevent the rapid accumulation of too much HDAg-L, which would inhibit viral RNA replication.

The structure in the immediate vicinity of edited adenosines varies somewhat, but in most substrates the target adenosine occurs as either an A-U pair or an A-C mismatch pair. Mutational analysis of some substrates, including the HDV genotype I amber/W site, indicates that editing levels are much higher when the adenosine occurs as an A-C mismatch rather than an A-U pair.^{20,23,33,41,42} Moreover, at least for HDV genotype I, any change in the position opposite the amber/W site (deletion, or substitution by A or G), led to markedly reduced editing levels.²⁰

Despite recognition of the common features among editing sites noted above, the sequence and structural determinants for highly specific editing are still not well understood. Only a handful of substrates for highly specific editing have been identified to date in mammals, and it is anticipated that many more remain to be identified. Knowledge of sequence and structural requirements for editing will likely facilitate the prediction of potential adenosine deamination editing sites from analysis of genomic sequences. Moreover, it is reasonable to expect that differences in editing levels among different substrate adenosines are due in part to variations in the structure of the RNA in the vicinity of the editing sites. Understanding the effects of structural variations will contribute to our understanding of how this important posttranscriptional regulatory mechanism is modulated. Thus, defining the structural determinants for editing remains an important goal; it is likely that the HDV amber/W site will be a useful tool in this endeavor.

Genotype Variations and Amber/W Editing

The sequence/structures around the amber/W site and the C-terminal sequences specific to HDAg-L are distinguishing features of the HDV genotypes.^43-48The genotype-specific variations in these two functional elements may be the result of the connection between amber/W editing and HDAg-L function, and might reflect different requirements for HDAg-L during the course of replication in the three genotypes.

In HDV genotype I the structure required for editing at the amber/W site is part of the unbranched rod structure characteristic of HDV RNA.²⁰The 8 Watson-Crick base-pairs flanking the amber/W site and the A-C mismatch pair involving the amber/W site are highly conserved among over 50 genotype I sequences.^244,49 Indeed, the only variation is the occasional substitution of the A-U pair 2 positions 5' of the amber/W site by a G-C pair. Site-directed mutagenesis studies have shown that both the base-pairing and the A-C mismatch are required for editing.²⁰

In the unbranched rod structure formed by HDV genotype III RNA the base-pairing in the immediate vicinity of the amber/W site is disrupted such that this structure does not function as a substrate for amber/W editing.⁴⁵ Rather, editing in genotype III requires an alternative structure that creates better base-pairing in the immediate vicinity of the amber/W site (Fig. 4).⁴⁵ Remarkably, this structure, termed the double hairpin, differs from the unbranched rod structure by nearly 80 base-pairs. The mechanisms by which the genotype III double hairpin RNA structure is formed and rearranges to the unbranched rod (which is required for RNA replication)⁴⁵ remain to be determined, but it appears that, as for viroid RNAs, RNA structural dynamics play an important role in HDV replication, not only in the activity of the highly structured ribozyme, but in other processes as well.

Figure 4

Schematic of RNA structures required for amber/W site editing in HDV genotypes I and III. The location of the amber/W site is indicated by the star. Sequences involved in forming the base-paired secondary structure essential for editing are indicated (more...)

The structure required for editing in HDV genotype II has not yet been determined. However, comparative analysis of the predicted secondary structure in the vicinity of the genotype II amber/W site in the unbranched rod reveals a conserved structure more similar to the genotype I structure than that in the genotype III unbranched rod, but still more disrupted than the genotype I structure.^46,48 This slightly disrupted structure, if it is indeed used as the substrate for genotype II amber/W editing, may be consistent with the observation that in transfected cells, editing levels were lower for replicating genotype II RNA than for genotype I RNA.⁴⁶

Inspection of the predicted structures around the amber/W sites in the three genotypes indicates that the genotype II and III amber/W sites vary from the type I site at positions that have been shown to be essential for editing in type I. For example, the A-C mismatch pair that involves the amber/W adenosine and which is highly conserved among genotype I isolates, occurs as an A-U pair in genotype III; when introduced by site-specific mutagenesis into a genotype I genome, this specific change substantially reduces editing and virus production.20,50 Perhaps the variations at the genotype II and III sites can be explained by compensatory effects, such as changes elsewhere in the editing site region, including sequences/structures 3' of the editing site, or differences in the mechanisms/processes by which HDV regulates editing during replication. Variations among the amber/W sites in the three HDV genotypes may provide a valuable opportunity for analyzing the structural determinants for RNA editing and evaluating the effects of different sequences/structures on editing at the HDV amber/W site.

Specificity of Editing

ADAR1 and ADAR2 can extensively edit long (≥50 base-pairs) double-stranded RNAs, in which up to 50% of adenosines may be deaminated. The role of this activity in cells is not clear, but the fact that dsRNA is a target and that one form of ADAR1 is induced by interferon²⁷ has led to the suggestion that editing of dsRNA may be part of the cellular response to virus infection. Clearly promiscuous editing such as occurs on dsRNA could be deleterious to virus replication. Indeed, spurious editing on HDV RNA by overexpressed ADAR1 and ADAR2 led to the production of protein variants that inhibited replication.³² Whether interferon treatment increases editing at the amber/W site, or elsewhere on the HDV RNA, remains to be determined.

Even though HDV RNA exhibits significant base-pairing in the unbranched rod structure, promiscuous editing does not typically occur during HDV infection; the amber/W site is edited 600-fold more efficiently than the other 337 adenosines in the RNA.⁵¹ It is likely that the primary and secondary structure of the HDV RNA have evolved to avoid undesirable (for the virus) editing at sites other than amber/W. Guanosine is by far the most common 5' neighbor a. based on HDV genotype I prototype, accession no. X04451 for adenosine in both the HDV genome and antigenome, and the ratios of observed to expected occurrences for the dinucleotides GA and UC (which would be GA in the complementary strand) are higher than for any other dinucleotides (Table 1). This bias may be due, in part, to selection for sequences that place non-amber/W adenosines in contexts that are less likely to be edited: analysis of editing on dsRNAs has indicated that adenosines flanked by a 5' guanosine are much less likely to be deaminated than other adenosines.³⁸ As for secondary structure, base-pairing in the HDV RNA unbranched rod structure is interrupted by frequent bulges, internal loops and mismatches, which have been shown to restrict editing on artificial dsRNA substrates.^37,39,52

Table 1

Dinucleotide frequencies in HDV RNA^a.

Regulation of Editing

HDV must regulate both the rate and the extent of editing at the amber/W site because HDAg-L, which is produced as a result of editing, is necessary for virion production but inhibits viral RNA replication. Varying the efficiency of editing at the amber/W site, either by altering levels of ADAR expression or by the introduction of mutations near the amber/W site, can affect HDV replication, virus production, or both.^32,45 Premature editing at the amber/W site results in reduced levels of RNA replication and reduced production of viable virions because edited antigenomes encode HDAg-L, which is a trans-dominant inhibitor of HDV RNA replication.^12,50,53 Insufficient editing can lead to increased intracellular HDV RNA replication, but inhibits virion production.⁴⁵ In addition to the rate, the extent of editing must also be controlled because the mechanism of editing in the HDV replication scheme (Fig. 2) produces genomes encoding HDAg-L. These genomes are packaged but are not likely to be infectious because HDAg-L does not support HDV RNA replication. Thus, the kinetics and extent of editing are likely regulated during HDV replication to maximize the rate and amount of infectious virus produced.

Control mechanisms for editing rely on several viral components and functions, including: RNA structure, HDAg, and viral RNA replication. HDV does not appear to regulate editing by affecting ADAR1 expression because ADAR1 levels are unaffected by HDV replication.³⁵Some of the control mechanisms may be described as passive, in that they are not affected by (or responsive to) the level of editing. This category includes the secondary structure of the RNA around the amber/W site. As mentioned above, the disruptions in base-pairing 3' of the amber/W site in HDV genotype I create a sub-optimal substrate for editing. Mutations that increase base-pairing in this region increase editing, but severely reduce replication and virion production (Sato and Lazinski, personal communication; Jayan and Casey, unpublished). It is not yet known whether the structures in the vicinity of the amber/W sites of genotypes II and III are also sub-optimal. One potential dilemma for the virus that is posed by using a sub-optimal structure to limit editing efficiency is that the specificity of editing is likely to be compromised because the specificity is determined by the ratio of the efficiency of editing at the amber/W site to the efficiency of editing at other “nonspecific” sites. The danger for the virus of nonspecific editing is the production of additional genomes defective for replication, or even the creation of dominant negative HDAg-S mutants.³² Thus, there may be limits as to how much amber/W editing can be restricted by using sub-optimal structures. HDV does appear to have a mechanism for minimizing the effects of editing at nonamber/W sites: in one study of HDV replicating HDV in transfected cells, all nonamber/W changes that occurred during replication were found on genomes that were also edited at the amber/W site.⁵¹

HDV genotype I uses an additional mechanism to slow down editing early in the replication cycle. For this genotype, HDAg-S is a strong inhibitor of editing (Fig. 5). While editing on replicating RNA 2-3 days post-transfection is nearly undetectable, up to 40% of nonreplicating RNAs produced in transfected cells in the absence of HDAg are edited. However, cotransfection of an HDAg-S expression construct leads to markedly reduced levels of editing on nonreplicating RNAs, most likely by binding to HDV RNA and preventing access of ADAR1.⁵¹ The levels of HDAg-S required for this inhibition are similar to those seen in cells replicating HDV RNA. Thus, it appears that HDAg-S prevents the rapid accumulation of editing early in the HDV genotype I replication cycle.

Figure 5

Schematic of the regulation of editing in HDV genotype I and III by HDAg. Left, genotype I. Both HDAg-S and HDAg-L effectively inhibit editing, most likely by binding the RNA near the editing site and limiting access of ADAR1. Right, genotype III. HDAg-L, (more...)

On the other hand, for HDV genotype III, HDAg-S is not an effective inhibitor of editing and likely does not play a direct role in limiting editing levels.⁵⁴ Rather, HDV genotype III uses the distribution of the RNA between at least two conformations to restrict editing.^45,54 Only RNA molecules that adopt the double hairpin structure can be edited (Fig. 4). However, the majority of the genotype III RNA appears to assume the unbranched rod conformation, which is not a substrate for editing.^45,54 Thus, while the amber/W site itself in genotype III RNA can be edited with efficiency similar to the genotype I site, editing levels in nonreplicating genotype III RNAs are much lower because most of the RNA assumes the unbranched rod conformation, which is not a substrate for editing.⁴⁵ The introduction of mutations in the genotype III RNA that shift the distribution of the RNA to the double hairpin structure increases editing to levels comparable with those seen with nonreplicating genotype I RNA.⁴⁵

It is not clear whether the inhibition of editing in genotype I or the conformational control of editing in genotype III is influenced by levels of editing and/or replication. Possibly, the HDAg:RNA ratio may vary during replication and thereby affect HDV genotype I editing rates. Likewise, for genotype III, if the RNA transcription rate varies during HDV genotype III RNA replication, such variations could affect editing by altering the distribution of the RNA between the double hairpin and unbranched rod structure.

Other mechanisms to control editing are by their nature responsive to editing levels. In HDV genotype III, HDAg-L is a much better inhibitor of editing than is HDAg-S, and is likely involved in a negative feedback loop to limit editing levels (Fig. 5).⁵⁴ This regulatory behavior requires the double hairpin structure peculiar to HDV genotype III amber/W editing. The reason for the differential effects of HDAg-S and HDAg-L is not yet clear; the hairpin on the 3' side of the amber/W site plays an essential role,⁵⁴ and could interfere with HDAg-S binding near the amber/W site. Genotype I does not use the same mechanism to control editing levels because genotype I HDAg-S and HDAg-L do not exhibit differential effects on editing (Cheng and Casey, unpublished data). Rather, HDV genotype I may limit the extent of editing via the inhibitory effect of HDAg-L on RNA replication, which is required for editing to occur on the antigenome and for the synthesis of edited genomic copies (Sato and Lazinski, personal communication).