Two lines of evidence support that the enzyme catalyzing elongation of PSTVd and closely-related viroids is RNA polymerase II: i) nanomolar concentrations of the fungal toxin α-amanitin that typically inhibit this enzyme also block in vivo and in vitro transcription of (+) and (-) viroid strands, and ii) viroid (+) and (−) strands have been retrieved with a monoclonal antibody against the carboxy-terminal domain of the largest subunit of RNA polymerase II [4].

Regarding the transcription initiation sites, this question has been recently tackled by priming a nuclear extract from a non-infected cell culture of the host plant Solanum tuberosum with the PSTVd mc (+) RNA. Following purification by affinity chromatography, analysis by primer extension of the (-) strands synthesized de novo revealed a single start site located in the hairpin loop of the left terminal domain of the rod-like secondary structure proposed for PSTVd mc (+) RNA [18]. Further analyses in planta of site-directed mutants are consistent with this hairpin loop being involved in infectivity and perhaps replication [18], as also is a genome-wide mutational analysis [17]. However, additional studies are needed to confirm this site as well as to identify the transcription initiation site of (+) strands, which remains unknown. The possibility exists that, like other RNA polymerase II transcripts, the 5’ termini of (+) and (-) RNAs of nuclear viroids could be capped in vivo, a signature that would mark unambiguously their initiation sites.

Early studies implicated the CCR in processing of the oligomeric (+) strands of the family Pospiviroidae either through hairpin I, a metastable motif that can be formed by the upper CCR strand and flanking nucleotides during thermal denaturation [19], or through a thermodynamically stable double-stranded structure that can be alternatively assumed by the same sequences in oligomeric RNAs. Data supporting this view include infectivity bioassays with different PSTVd DNA and RNA constructs [20], with longer-than-unit clones of Hop stunt viroid (HSVd) [21], and with constructs of Citrus exocortis viroid (CEVd) containing sequence repetitions and point mutations in the upper CCR strand [22]. A critical reassessment of these data led to a model involving the double-stranded structure in cleavage, although the model did not predict the mechanism of cleavage-ligation or specify the exact processing site [23]. On the other hand, in vitro and thermodynamic analyses of the products obtained by incubating a potato nuclear extract with a full-length PSTVd RNA containing a 17-nt repeat of the upper CCR strand led to propose that cleavage of (+) strands is driven by a multibranched structure with a hairpin —different from hairpin I— capped by a GAAA tetraloop conserved in members of the genus Pospiviroid. This multibranched structure subsequently switches to an extended conformation with a loop E (see below) that promotes ligation [24]. Yet, the processing complex formed in vitro may not mimic the corresponding complex in vivo. Moreover, other members of the family Pospiviroidae cannot form the hairpin capped by a GAAA tetraloop [4], and alternative processing sites in the lower CCR strand, or outside this region, have been observed for several members of this family [25].

This question has been recently examined in vivo [26], using a system based in transgenic A. thaliana expressing dimeric (+) transcripts, dt (+) RNAs, of CEVd, HSVd, and Apple scar skin viroid (ASSVd) [27] of the genera Pospiviroid, Hostuviroid, and Apscaviroid, respectively, within the family Pospiviroidae [4]. This system circumvents most of these limitations. It is an in vivo system in which processing is correct: transgenically expressed dimeric transcripts of typical members of the family Pospiviroidae are cleaved to the ml forms and then ligated to the infectious mc RNAs, whereas the complementary dt (-) RNAs are not [26,27], thus reproducing the situation observed in typical hosts. However, in contrast to typical hosts in which turnover of the longer-than-unit (+) replicative intermediates is difficult to follow because of their low accumulation and diverse size, the A. thaliana-based system provides a constant supply of a size-specific replicative-like intermediate that can be easily quantified as well as its processing products. Moreover, the ml and mc (+) RNAs can be assumed to come essentially from processing of the transgenically expressed dt (+) RNA. Therefore, the effects of specific mutations in the primary transcript on cleavage and ligation can be evaluated —regardless of whether the resulting products are infectious or not — and it is even possible to identify mutations affecting only ligation (see below).

Results with the A. thaliana-based system have mapped the cleavage site of CEVd (+) strands at the upper CCR strand [26], in a position equivalent to that inferred for PSTVd with an in vitro system [24]. However, the RNA motif directing cleavage in vivo does not seem to be the GAAA-capped hairpin proposed previously [24], but the hairpin I/double-stranded structure [26]. The first argument supporting this view is that whereas the cleavage sites of HSVd and ASSVd (+) strands also map at equivalent positions in a similar hairpin I/double-stranded structure, these viroids cannot form the GAAA-capped hairpin. In contrast, examination of the hairpin I/double-stranded structure reveals some appealing features. Hairpin I is composed by a tetraloop, a 3-bp stem, an internal symmetric loop of 1–3 nt in each strand that presumably interact by non-Watson-Crick base pairs [28], and a 9–10-bp stem that can be interrupted by a 1-nt symmetric or asymmetric internal loop [22,25] (Figure 2). Remarkably, these structural features are conserved in the type species of the five genera composing the family Pospiviroidae and additionally: i) the capping tetraloop is palindromic itself, and ii) the two central positions of the tetraloop and the central base pair of the 3-bp stem are phylogenetically conserved (Figure 2) [25]. As a consequence, a long double-stranded structure with a GC-rich central region of 10 bp containing the cleavage sites can be alternatively assumed by the same sequences in a di- or oligomeric RNA (Figure 3). The second argument supporting the hairpin I/double-stranded structure as the RNA motif directing cleavage derives from the effects on this reaction of different CEVd mutants expressed transgenically in A. thaliana affecting differentially this motif versus the GAAA-capped hairpin [26].

Altogether, the results obtained with the A. thaliana-based system and with some experimental hosts [26] indicate that the substrate for cleavage in vivo of all members of the family Pospiviroidae is the double-stranded structure proposed previously [23], with hairpin I playing a role in promoting the adoption of this structure (see below). Moreover, the cleavage sites in the double-stranded structure leave two 3’-protruding nucleotides in each strand (Figure 3), the characteristic signature of RNase III enzymes [29,30]. The participation of an enzyme of this class, of which there are at least seven in A. thaliana [31], is consistent with the nuclear location of some of them, which additionally have preference for substrates with a strong secondary structure resembling that of viroids. Going one step further, if an RNase III indeed catalyzes cleavage of the oligomeric (+) RNAs of the family Pospiviroidae, the resulting products should have 5’-phosphomonoester and free 3’-hydroxyl termini. Characterization of the ml (+) RNAs from A. thaliana transgenically expressing dt CEVd (+) RNAs has shown that this is actually the case [32]. The adoption in vivo of the double-stranded structure with a GC-rich central region containing the cleavage sites could be promoted by hairpin I because prior work with PSTVd has mapped a dimerization domain at this hairpin [28]. This situation resembles that observed previously in retroviruses in which dimerization, a critical step of their infectious cycle, is mediated by a hairpin with a palindromic loop that can dimerize via a kissing loop interaction [33]. During transcription of oligomeric (+) RNAs of the family Pospiviroidae, a kissing loop interaction between the palindromic tetraloops of two consecutive hairpin I motifs might similarly start intramolecular dimerization, with their stems then forming a longer interstrand duplex [32] (Figure 3).

Initial in vitro assays showed that incubation with wheat germ RNA ligase converts the monomeric linear (+) PSTVd RNA isolated from infected tissue into its bona fide circular counterpart, suggesting that the reaction might be mediated by an enzyme of this class [34], which demands 5’-hydroxyl and 2’,3’-cyclic phosphodiester termini [35]. Although subsequent experiments with the fungal RNase T1 revealed that it can process in vitro linear oligomeric (+) PSTVd RNAs into infectious monomeric circular RNA [36], it is unlikely that an RNAse could catalyze in vivo both cleavage and ligation. Moving into a more physiological context, incubation with a potato nuclear extract of a monomeric (+) PSTVd RNA with a short repeat of the CCR upper strand produced the infectious monomeric circular form, leading to the proposal that the enzymatic cleavage and ligation of the PSTVd (+) strand is driven by a switch from a branched structure containing a GAAA-capped hairpin (see above) to an extended conformation with an E loop [24]. Loop E, a UV-sensitive motif of RNA tertiary structure that is conserved in PSTVd and members of its genus [4] and exists in vitro [37] and in vivo [38,39], has also been involved in host specificity [40], pathogenesis [41], and transcription [42]. A structural model based on isostericity matrix and mutagenic analyses has been derived recently for PSTVd loop E [42]; this model can be extended to other closely-related viroids like CEVd.

With the aim of circumventing the intrinsic limitations of in vitro approaches, the most important of which is to what extent results from these approaches can be extrapolated to the in vivo situation, the A. thaliana-based system [27] described in the previous section for studying in vivo cleavage of the oligomeric (+) strands has also been used to examine the requisites for proper ligation of the resulting ml (+) RNAs. Data obtained with this system indicate that the substrate for this reaction in the genus Pospiviroid is the extended conformation containing loop E [26] (Figure 3). Therefore, whereas cleavage is only dependent on the upper CCR strand and flanking nucleotides, ligation is dependent on nucleotides of both CCR strands that encompass those of loop E and others adjacent. Because within the family Pospiviroidae loop E is only formed in the genera Pospiviroid and Cocadviroid, other genera of this family must have alternative motifs playing a similar role in ligation. Potential candidates are the extended conformation of the CCR with a bulged-U helix conserved in all members of the genera Pospiviroid, Hostuviroid and Cocadviroid (Figure 3D), and similar structures in the other genera of this family. More recent data, obtained by rapid amplification of 5’ and 3’ cDNA ends and in vitro ligation assays with the T4 RNA ligase 1 and A. thaliana tRNA ligase, have shown that the ml CEVd (+) RNA resulting from cleavage of a dimeric transcript transgenically expressed in A. thaliana indeed contains the 5’-phosphomonoester and 3’-hydroxyl termini expected from cleavage mediated by an RNase III [32]. Therefore, unless the termini are later modified, these results suggest the existence of a second plant RNA ligase that would mediate joining of the same 5’-P and 3’-OH termini as those required by the T4 RNA ligases 1 and 2 [43,44].

Two different RNA polymerases have been reported in plastids, the plastid-encoded polymerase (PEP) with a multisubunit structure similar to the Escherichia coli enzyme, and the single-unit nuclear-encoded polymerase (NEP) resembling phage RNA polymerases. The available evidence supports the involvement of the second in replication of chloroplastic viroids. First, adding micromolar concentrations of the toxin tagetin to chloroplastic preparations from ASBVd-infected leaves prevents in vitro transcription of representative chloroplastic genes without essentially affecting synthesis of ASBVd strands [51]. And second, accumulation of a typical NEP transcript and of PLMVd strands is particularly high in areas displaying peach calico (an albinism induced by specific PLMVd variants), in which development of proplastids into chloroplasts and processing of chloroplastic rRNA precursors (and translation of plastid-encoded proteins like PEP) is impaired [52]. These latter data also suggest that other chloroplastic proteins mediating replication (see below) are also nuclear-encoded.

The question of the transcription initiation sites has been undertaken by in vitro capping with [α³²P]-GTP and guanylyl-transferase, which specifically labels the free 5’-triphosphate group characteristic of chloroplastic primary transcripts. This approach, combined with RNase protection assays, has identified the initiation sites of ASBVd ml (+) and (−) RNAs isolated from infected avocado at similar (A+U)-rich terminal loops in their predicted quasi-rod-like secondary structures [53]. The same methodology, but combined with RNA ligase-mediated rapid amplification of cDNA ends to enhance its sensitivity, has mapped the initiation sites of PLMVd ml (+) and (−) RNAs isolated from infected peach at similar double-stranded motifs of 6 to 7 bp that also contain the conserved GUC triplet preceding the self-cleavage site in both polarity strands; within their predicted branched secondary structures, the motifs are located at the base of similar long hairpins that presumably contain the NEP promoters [54]. These PLMVd initiation sites, which have been later confirmed [55], are different from those proposed previously on the basis of primer-extensions of the 5’ termini of PLMVd subgenomic RNAs isolated from infected peach, and of in vitro transcriptions with truncated PLMVd RNAs and the RNA polymerase of E. coli [56]. This discrepancy most likely derives from the difficulties in reconstituting in vitro a bona fide initiation complex reproducing the in vivo situation, particularly with a eubacterial RNA polymerase very different from NEP. Therefore, the structural motifs containing the transcription initiation sites are clearly distinct in the two viroids: terminal loops in ASBVd and double-stranded motifs in PLMVd.

Apart from the viroid discovery itself, the finding °first in ASBVd and then in the other members of the family Avsunviroidae [12,50]° that the strands of both polarities self-cleave through hammerhead ribozymes has attracted much attention on these small RNA replicons, which are now regarded as remnants of the RNA world postulated to have preceded the present world on Earth based on DNA and proteins [57]. Previous descriptions of the hammerheads found in viroids and in other small RNAs, and of the evidence supporting that they mediate self-cleavage in vivo of the multimeric viroid replicative intermediates wherein they are embedded, have been reviewed elsewhere [12,50]. Here we will focus on recent advances, which once again, have provided some surprises.

Hammerhead structures are formed by a central core of conserved sequences surrounded by three double-stranded stems (I, II and III) with loose sequence requirements and usually capped by short loops (1, 2 and 3). X-ray crystallography has unveiled that the actual conformation of these catalytic motifs does not resemble a hammerhead, as suggested by the initial bidimensional representation, but rather a Y wherein stems III and II are almost colinear (Figure 4). Data obtained with model hammerheads acting in trans (an artificial design with opened loops 1 and 2 that facilitates kinetic analysis in protein-free media) showed that efficient cleavage in vitro requires Mg²⁺ concentrations of 5-10 mM, an order of magnitude higher than that existing in vivo. However, re-examinations in vitro and in vivo have demonstrated that natural cis-acting hammerheads with intact loops 1 and 2 self-cleave much faster (retaining activity at 0.5-1 mM Mg²⁺), and that modifications of loops 1 and 2 cause a severe reduction of their catalytic activity [58,59], thus supporting that loop-loop tertiary interactions play a key role in the folding and catalytic activity of natural hammerheads. The recent X-ray crystal structures of two full-length natural hammerheads indeed show that the tertiary contacts, despite being distal from the active site, promote the adoption of a catalytically active conformation by the central core [60,61]. Moreover, applying a combination of NMR (nuclear magnetic resonance) spectroscopy, site-directed mutagenesis and kinetic and infectivity analyses to the (+) and (-) hammerheads of CChMVd has revealed that, in both hammerheads, loop 1 is a heptanucleotide hairpin loop containing an exposed U at its 5’ side and an extrahelical U at its 3’-side critical for the catalytic activity of the ribozyme in vitro and for viroid infectivity in vivo, whereas loop 2 has a key opened A at its 3’-side. These structural features promote a specific loop-loop interaction motif across the major groove, the structural features of which °base pairing between the 5’ pyrimidine of loop 1 and the 3’ purine of loop 2, and interaction of the extrahelical pyrimidine of loop 1 with loop 2 (Figure 4)° are likely shared by a significant fraction of natural hammerheads [62].

Two additional points regarding hammerhead function in the natural context deserve a comment. First, the loop-loop tertiary interactions could be further stabilized in vivo by chloroplastic proteins acting as RNA chaperones [63]. Further dissection of the processing requirements in the family Avsunviroidae may be tackled with a system based in transplastomic Chlamydomonas reinhardtii expressing (+) and (-) dimeric transcripts; despite the absence of viroid RNA-RNA transcription, this system can be used to identify the cellular factors facilitating cleavage (and ligation, see below) [64]. And second, the differential effects in co- and post-transcriptional self-cleavage of mutations affecting the trinucleotide preceding the self-cleavage site of (+) and (-) ELVd hammerheads suggest that: i) natural hammerheads have been evolutionarily selected to function co-transcriptionally, and ii) the trinucleotide AUC preceding the self-cleavage site is most likely excluded in the majority of natural hammerheads because it promotes the transient adoption of catalytically-inactive metastable structures during transcription [65]. Therefore, hammerhead-mediated catalysis appears closely associated with RNA folding taking place during transcription.

First, considering that the 5’-OH and 2’, 3’-cyclic phosphodiester termini resulting from hammerhead-mediated self-cleavage are those typically required by the wheat germ tRNA ligase, whose homologue from A. thaliana is the only plant RNA ligase well-characterized biochemically and genetically [66], and considering also its low substrate specificity, an enzyme of this class appears an excellent candidate for catalyzing RNA ligation in chloroplastic viroids. However, pre-tRNA splicing in yeast has been assumed to occur in the nucleus and, in consonance with this view, immunofluorescence and immune electron microscopy detected the tRNA ligase in this organelle [67]. This long-held paradigm has been questioned by the finding that yeast pre-tRNA splicing takes place in the cytoplasm [68]. Most importantly, in A. thaliana and Oryza sativa two of the splicing enzymes, the tRNA ligase and 2’-phosphotransferase (which catalyzes removal the 2’-phosphomonoester group generated by the tRNA ligase), have transit signals at their N-termini and are predominantly targeted to chloroplasts and proplastids, respectively [69]. Therefore, the plant tRNA ligase now fulfills all criteria for mediating circularization of members of the family Avsunviroidae. Moreover, in one viroid-like satellite RNA the junction resulting from ligation of the two termini generated by hammerhead-mediated self-cleavage contains the expected signature of a tRNA ligase: a 2’-phosphomonoester, 3’, 5’-phosphodiester group [70]. Encapsidation of this viroid-like satellite RNA by the coat protein of its helper virus might have precluded removal of the 2’-phosphomonoester. In this same direction we have observed in some PLMVd and CChMVd variants a deletion of the nucleotide immediately after the self-cleavage site that strongly reduces in vitro self-cleavage and infectivity. We suspect that this deletion may have been introduced during reverse transcription, because it was detected only in the plus-polarity strand, and that its ultimate cause could be the presence in a fraction of the viroid population of a 2’-phosphomonoester at the nucleotide preceding the self-cleavage/ligation site, which would lead to a jumping of the reverse transcriptase [71-73]. Another possibility is the 2’, 5’-phosphodiester linkage formed in the in vitro self-ligation of PLMVd, assuming that this atypical bond is indeed present in vivo (see below).

Second, the hammerhead ribozyme may mediate not only self-cleavage but also ligation generating a 3’, 5’-phosphodiester linkage. This is indeed the case for another small ribozyme, the hairpin ribozyme, which is found embedded in certain viroid-like satellite RNAs [74]. When compared with the hammerhead ribozyme, the RNA ligase activity of the hairpin ribozyme is much higher. However, recent data show that the ligase activity is considerably increased in hammerheads wherein the tertiary stabilizing interaction between loops 1 and 2 is preserved [75,76]. Therefore, it is still possible that replication of chloroplastic viroids might be an RNA-mediated mechanism only requesting a host RNA polymerase.

Third, the reaction could also be autocatalytic (self-ligation) but operating through a different pathway. This hypothesis is based on the observation that the monomeric linear PLMVd RNAs resulting from hammerhead-mediated self-cleavage can self-ligate in vitro mostly producing 2’, 5’- instead of the conventional 3’, 5’-phosphodiester bonds [77], and on the proposal that these atypical bonds exist in circular PLMVd RNAs isolated from infected tissue and impede their in vitro self-cleavage [78]. However, this may not well be the physiological mechanism because: (i) in vitro self-ligation of PLMVd demands concentrations of Mg⁺² around 100 mM, which are much higher than those existing in vivo, ii) in vitro self-ligation has also been observed in PSTVd [23] and in HDV RNA, in this latter case also generating 2’, 5’-phosphodiester bonds [79], while neither PSTVd nor HDV RNA appear to follow this route in vivo [26,32,80], and iii) there are no previous reports on the existence of natural RNAs with 2’,5’-phosphodiester bonds serving as transcription templates; indeed, this atypical linkage blocks elongation catalyzed by the reverse transcriptase in vitro [77] and presumably by the NEP in vivo, while we have not observed prominent stops at the ligation site in our primer extension experiments using ASBVd circular forms isolated from infected tissue [81].