HDV-like self-cleaving ribozymes

Abstract

HDV ribozymes catalyze their own scission from the transcript during rolling circle replication of the hepatitis delta virus. In vitro selection of self-cleaving ribozymes from a human genomic library revealed an HDV-like ribozyme in the second intron of the human CPEB3 gene and recent results suggest that this RNA affects episodic memory performance. Bioinformatic searches based on the secondary structure of the HDV/CPEB3 fold yielded numerous functional ribozymes in a wide variety of organisms. Genomic mapping of these RNAs suggested several biological roles, one of which is the 5′ processing of non-LTR retrotransposons. The family of HDV-like ribozymes thus continues to grow in numbers and biological importance.

Introduction

Hepatitis delta virus (HDV) is a small [∼1,700-nucleotide (nt)] single-stranded RNA virus first isolated from human hepatocytes infected with hepatitis B virus.¹ HDV harbors two structurally related self-cleaving ribozymes in its genome, one in the genomic and one in the complementary, antigenomic strand.^2–6 Like other small self-cleaving ribozymes, these RNAs catalyze a transesterification reaction, promoting a nucleophilic attack by a 2′ hydroxyl on the adjacent phosphate and yield both a 2′–3′ cyclic phosphate and a liberated 5′ hydroxyl.³

HDV Replication

The HDV antigenomic ribozyme was identified by in vitro transcription of cloned HDV genome.³ This method mimics self-scission during the viral rolling circle replication, a mechanism similar to plant viroid, virusiod and satellite RNA replication.³ During rolling circle replication, several copies of the genome are produced in a concatameric form. Each transcript is cleaved to produce monomer-length genomes for subsequent ligation and viral assembly steps.^7,8 The genomic RNA is a template for the synthesis of the concatemers of antigenomic RNA, which is then self-cleaved by its cis-ribozyme followed by host-aided ligation to generate circular, monomer length molecules.^9,10 The circular “antigenomic” RNAs then serve as the templates for the genomic RNA synthesis followed by the similar self-cleaving and ligation processes.

The genomic ribozyme was shown to self-cleave by two groups: Taylor, Dinter-Gottlieb and coworkers discovered the activity in the liver RNA of infected chimpanzees,⁴ while Lai and colleagues revealed that the monomer circular RNA was found in monkey kidney cell line transfected with linear dimer cDNA.² The cleavage site for genomic ribozyme is between positions 685 and 686, whereas in the antigenomic RNA it is between positions 900 and 901.⁴ Both ribozymes require divalent metal ions, such as Mg²⁺, Mn²⁺ or Ca²⁺, for efficient catalysis and exhibit a drastically lower activity in monovalent ions.^2,11

The cleavage site of the antigenomic ribozyme is 33 nts downstream of the polyadenylation site of the mRNA that encodes HDV's only protein, the delta antigen.³ The sequence between the two processing positions, polyadenylation and ribozyme scission, has been shown to affect both the ribozyme cleavage rate and the extent to which the ribozymes self-cleave, presumably affecting the balance between the translation-competent mRNA and the replication intermediate RNA.¹²

Structure of the Ribozymes

The structures of the HDV ribozymes consist of five paired (P) regions that form two coaxial stacks (P1 stacks on P1.1 and P4, while P2 stacks on P3), which are linked by single-stranded joining (J) strands J1/2 and J4/2 (Fig. 1A). Crystal structures of inhibited precursor and product states show that the genomic ribozyme folds into a nested double pseudoknot that constrains the overall structure and forms the active site of the ribozyme (Fig. 1B).^13–15 The ribozyme cleaves the RNA backbone at the base of the P1 helix, typically at a guanosine residue, and the active site is formed by the P3 helix, and L3 and J4/2 strands.^13–15

Figure 1.

HDV-family of self-cleaving ribozymes. (A) Secondary structures of the antigenomic and genomic HDV ribozymes. (B) Crystal structure of the self-cleaved form of the genomic ribozyme. (C) Human CPEB3 ribozyme. Secondary structure elements are distinguished by color. Single-stranded regions are shown in black, except J4/2, which is shown in pink (active-site cytosine is labeled by its position with respect to the cleavage site). Black lines with arrowheads indicate connection and direction of strands. Products of self-scission are shown; cleavage site corresponds to the 5′ guanosine residue.

A single nucleotide upstream and about seventy nucleotides (nts) downstream of the cleavage site are sufficient for self-scission.^5,16 Both HDV ribozymes have seven base-pairs in P1 and three in P3, while the P2 and P4 helices tend to show more length variation, mispairings and insertions. Moreover, the P4 helix is not absolutely required for activity, although it does seem to stabilize the active ribozyme.^17,18 Only a few positions in the sequence are invariant, including an active-site cytosine residue, a guano-sine-pyrmidine base pair at the base of the P1 helix, an adenosine in J4/2 that forms an A-minor interaction with P3, and a G•U reverse wobble pair in L3 that forms the base of the P3 helix.^13–15 Both the 5′ leader sequence, which is liberated during the reaction, and the 3′ tail can affect the kinetics of the overall reaction, depending on whether they disrupt or facilitate correct folding of the active ribozyme structure.^5,16,19–23

Construction of trans-Cleaving Ribozymes from cis-Cleaving Forms

HDV ribozymes can be converted into trans-active forms by bisection of the J1/2 and/or L4 regions.^8,24 Although splitting the ribozymes in the L4 loop results in more extensive base pairing, which increases binding specificity and allows incorporation of modified nucleotides in the active site,²⁵ the J1/2-bisected molecule preserves the ribozyme core and recognizes a target strand exclusive of any catalytic components. The design of J1/2-split molecule has led to ribozymes that could cleave target RNAs in vitro and in vivo.^26–29

In vitro Selection of an HDV-Like Ribozyme from a Human Genomic Library

The human cytoplasmic polyadenylation element-binding protein 3 (CPEB3) ribozyme was the first HDV-like ribozyme discovered outside the virus.³⁰ The ribozyme was identified using an in vitro selection from a circularized human genomic library. Concatameric RNA transcripts were purified, incubated in the presence of Mg²⁺ to promote self-scission, and self-cleaved dimers were isolated. The dimers contained at least one copy of the unknown genomic sequence flanked by primer sequences in proper orientation, so that the selected pool could be reverse transcribed, amplified and re-circularized for subsequent rounds of selection. Four ribozymes were found in the human genome, of which CPEB3 is the best characterized. A single copy of the ribozyme resides in the second intron of CPEB3 gene. EST and rapid amplification of 5′ complementary DNA ends (5′ RACE) analyses of mammalian transcripts indicated that the ribozyme is expressed and cleaved in vivo at the same position as in in vitro.

The biochemical properties of the CPEB3 ribozyme, mapping of the mammalian variants, and mutational analysis supported a model in which the CPEB3 ribozyme folds into the nested doublepseudoknot secondary structure of the HDV ribozymes (Fig. 1C).³⁰ The human CPEB3 ribozyme contains a single nucleotide polymorphism at position 36, which results in either a G•U or a G-C base-pair at the base of P1. Ribozymes containing the G-C pair self-cleave about three times faster than when a G•U wobble pair is present in this location.³⁰ A recent association study revealed that individuals homozygous for the faster variant of the ribozyme perform poorer in episodic memory tests, particularly when the presented material has emotional valence.³¹ Given that the CPEB3 gene expression is inducible by kainate in murine hippo-campus,³² the connection of the CPEB3 ribozyme to memory performance is not surprising. However, self-cleaving ribozymes had not previously been linked to any neurobiological process, therefore the role of the ribozyme in memory performance warrants further investigation.

The biological role of the CPEB3 ribozyme may arise from co-transcriptional processing of the CPEB3 pre-mRNA. If the ribozyme self-cleaves before the next exon is synthesized and tagged for splicing, it could prevent processing of the premRNA. Initial in vitro analysis indicated slow CPEB3 cleavage rate of ∼0.7 h⁻¹ at physiological-like conditions. Assuming that self-scission proceeds at the same rate in vivo, the ribozymes would likely disrupt only a small fraction of the pre-mRNA if the gene is transcribed at typical rates and splicing occurs co-transcritpionally.³⁰ This estimate is based on the distance between the ribozyme and the next CPEB3 exon (∼12 kb) and the Pol II transcription rate of 1,000–2,000 nt/min, which together result in ∼6–12 min between the time the ribozyme and the next exon are synthesized.^30,33 However, significantly faster wild-type CPEB3 cleavage was detected when a residue in a 5′ flanking sequence was mutated, indicating that in the absence of alternative secondary structures, the ribozyme is capable of fast self-scission.³⁴ Furthermore, the co-transcriptional self-cleavage rate of the human ribozyme is faster than that observed in the standard assay, which uses ribozymes prepared via a denaturation step, indicating that the ribozyme may be able to fold into its active form and self-cleave rapidly in vivo.³⁴ Together these recent findings suggest that the mammalian CPEB3 ribozymes can self-cleave fast co-transcriptionally and can affect the stability of the CPEB3 pre-mRNA on timescales relevant to gene expression.

Sequence variation among the CPEB3 ribozymes.

The CPEB3 ribozyme is highly conserved among mammals and was not initially found in other genomes (Table 1). Sequence variation supports the HDV-like secondary structure, with several notable exceptions. For example, the human, sloth, armadillo and dolphin ribozymes, carry a C·A mismatch at the top of the P1 helix, while other mammals possess a fully base-paired P1 region. Some species contain an adenosine at the first position of P1, forming an A1-U36 base pair and four organisms (Sus scrofa, Canis familiaris, Felis catus and Ailuropoda melanoleuca) possess a pyrmidine at the first position, and a purine in the second position that might serve as the actual cleavage site (Table 1).

Table 1.

HDV, CPEB3 and CPEB3-like ribozymes aligned by structural element

Several mammals likely contain inactive ribozymes because the active-site cytosine is substituted with other nucleotides, such as adenosine in Dipodomys ordsii or uridine in Felis catus, Dasypus novemcinctus and Myotis lucifugus. Similarly, the CPEB3 ribozyme from Bos taurus is likely less active or inactive because the base of the P3 helix contains an A·C mismatch that would weaken the ribozyme core.

CPEB3 and HDV ribozymes form the same secondary structures, but their sequences are dissimilar. Active variants of the three ribozymes, combined with sequences isolated from in vitro selection experiments based on HDV ribozymes, suggest that there are only six invariant positions in the ribozyme.^30,35–37 Because the helical regions dominate the structure and require base co-variation, but not specific sequences, alignment-based searches are only partially successful at uncovering new HDV-like ribozymes. Recently, however, non-mammalian ribozymes similar to the CPEB3 sequence have been identified in Strongylocentrotus purpuratus, Branchiostoma floridae and Gadus morhua, although the S. purpuratus and B. floridae ribozymes were first identified using structure-based searches (Table 1).³⁸

Structure-Based Searches

The discovery of HDV-like ribozymes in mammalian genomes suggested that self-cleaving ribozymes are more widely distributed than was previously thought. As mentioned in the previous section, sequence searches did not initially reveal similar ribozymes in non-mammalian genomes. On the other hand, a bioinformatic search based on the secondary structure of the HDV ribozymes led to the discovery of many more ribozymes (Table 2 and Fig. 2).³⁸ The motif search was performed using the RNABOB software by scanning genomic databases for sequences capable of forming the prescribed secondary structure, while placing the conserved nucleotides at the correct positions.^39,40 A single tertiary contact, an A-minor interaction between J4/2 and P3, was also included among the invariant positions. Figure 2A shows a typical structure descriptor for HDV-like ribozymes.

Table 2.

Segmental alignment and in vitro activity of HDV-like ribozymes from structure- and sequence-based searches

*All rate constants marked with asterisk are presented here for the first time. All others have been published previously (ref. ³⁸).

Figure 2.

The consensus secondary structure for HDV-like ribozymes (A). P1 through P4 represent base-paired regions where co-variation is required; J1/2, J4/2 and L4 are single-stranded regions. N, any nucleotide; R, purine; Y, pyrimidine; H, adenine, cytidine or uracyl. Dashed lines represent variable-length sequence insertions. Solid lines represent direct connections without nucleotide insertions. (B) drz-Agam-2-2, one of the largest and (C) drz-Hdis-1, the smallest known HDV-like ribozymes.

Nomenclature of the new ribozymes.

The multitude of new ribozymes required new systematic nomenclature that could identify the ribozymes with the species of origin as well as any family of sequences found within that organism. The proposed names of individual ribozymes were divided into parts representing the ribozyme type (drz, delta-like ribozyme), organism name, in analogy to the rules established for restriction endonucleases, based on the binomial nomenclature (Agam, Anopheles gambiae), and sequence family and subfamily number in consecutive order (e.g., drz-Agam-2-1, drz-Agam-2-2). The viral sequences were named using the common acronym for the virus (drz-CIV-1).³⁸

African mosquito.

Several ribozymes that belong to two main families and about five subfamilies were identified in African mosquito A. gambiae. The ribozymes were grouped according to the length of the secondary structure and conservation of core sequences. In the smaller, drz-Agam-1 family, the J4/2 region contains a second cytosine next to the active-site cytosine that permits residual self-cleavage activity even when the catalytic cytosine has been mutated, suggesting that the J4/2 strand is flexible enough to swap residues in the active site of the ribozyme.³⁸ The larger, drz-Agam-2 class is the first family of naturally occurring HDV-like ribozymes with extended J1/2 and P4 regions discovered. The J1/2 region of these ribozymes is predicted to form a base-paired structure that stabilizes the fold of the core in part because it appears to extend the P2 helix by several additional base-pairs (Fig. 2B).

The cleavage activity of the drz-Agam ribozymes is likely regulated in vivo, as 5′ RACE and RT-qPCR experiments performed on RNA extracts from different developmental stages and sexes of the mosquito showed both differential expression and cleavage extent.³⁸ Expression was found to vary between sexes and life-stages of the insect, with the highest expression but lowest cleaved steady-state fraction observed in adult male mosquitoes. Self-cleaved forms of the ribozyme are readily detected in both pupal and larval stages of the insect, which show lower overall expression of the motif relative to the adults. The differential scission of the ribozyme is particularly intriguing because such striking difference in self-cleavage activity had not been observed for other ribozymes. At this point we can only speculate as to the regulatory mechanism, which may involve cofactors that either bind directly to the ribozyme or promote formation of alternative secondary structures in its flanking regions resulting in misfolding of the ribozyme core.

One of the newly discovered ribozymes, drz-Agam-2-2 mapped to an RTE retrotransposon, a non-long-terminal-repeat (non-LTR) genetic element. RTEs are autonomous mobile elements that typically insert into other transcription units, and do not harbor promoter sequences. The drz-Agam-2-2 ribozyme cleavage site maps to the 5′ terminus of the RTE, suggesting that the ribozyme's function is to liberate the retrotransposon RNA from the upstream transcript.

Ribozymes with sequence similarity to drz-Agam-1.

A search for sequences similar to drz-Agam-1 ribozymes revealed many more ribozymes in insects, fungi, plants, round worms and a unicellular marine eukaryote (Table 2). These sequences were not initially tested for in vitro activity, but our current analysis of many of the sequences shows that they are robust ribozymes. The previously unpublished rate constants measured at standard conditions are presented in Table 2.

Ribozymes identified from insects include drz-Afun-1 from the malaria mosquito Anopheles funestus. Efficient self-scission is only observed in higher concentrations of Mg²⁺, which may result from a weaker, 6-bp P1 region that contains a C-C mismatch. Other insect ribozymes similar to drz-Agam-1 include drz-Rpr-1 from the triatomid bug Rhodinus prolixus, a vector of Chagas disease parasite Trypanosoma cruzi, where it maps to the beginning of many ESTs isolated from the insect's central nervous system (e.g., FG545964); and drz-Apis-1 from the pea aphid Acyrthosiphon pisum.

Ribozymes found in fungi include Neurospora crassa drz-Ncra-1 ribozyme, which cleaves with a rate constant of 58 h⁻¹ and whose in vivo expression is supported by an EST (GE968064); drz-Tvir-1 from the mycoporasidic fungus Trichoderma virens (rate constant of 8.9 h⁻¹); drz-Gmon-1 from the rice bakanae disease fungus Gibberella moniliformis; and drz-Tatr-1 from the cold tolerant fungus Trichoderma atroviride. Other fungi harboring HDV-like ribozymes include the Darling's disease fungus Ajellomyces capsultus, where the ribozyme maps to an RNase H gene (EST XM_001543659); filamentous fungi Podospora anserina and Talaromyces stipitatus; invasive fungal infection Neosartorya fischeri; and plant pathognic fungi Magnaporthe grisea, Phaeosphaeria nodorum SN15 and Alternaria brassicicola.

The first plant HDV-like ribozyme with confirmed in vitro activity is drz-Hann-1 from sunflower, Helianthus annus. Its in vivo expression is supported by an EST (DY925396); however, in vitro cleavage was only detectable at elevated concentrations of Mg²⁺ (50 mM), perhaps due to a U-U mismatch in P1.

Marine organisms harbor the smallest HDV-like ribozyme identified to date, drz-Hdis-1 from the Pacific abalone Haliotis discus (58 nts, Fig. 2C). The P4 helix is replaced by just five nts and the cleavage rate constant is 0.049 h⁻¹ (at 37°C and 10 mM Mg²⁺). The marine diplonemid Diplonema papillatum harbors a robust ribozyme that bisects an RNA transcript containing an upstream splice-leader sequence and downstream 5S ribosomal RNA, suggesting that the ribozyme may be involved in processing of non-coding RNAs.

Many of these ribozymes contain non-canonical core sequences, including in the J4/2 region and mispaired positions in helical elements (Table 2), providing further information about sequence variation that will allow future refinement of structure-based searches for more dissimilar ribozymes of this family.

Purple sea urchin.

There are four distinct HDV-like families found in purple sea urchin, S. purpuratus, distinguished by the sequences of P3, L3, P1.1 and J4/2 regions. All S. purpuratus ribozyme families contain multiple copies dispersed throughout the genome. The drz-Spur-1 ribozyme is the most abundant family, with 22 copies appearing in the genome at loci that include the first exon of non-LTR-like reverse transcriptase, an intronexon junction of an ependymin-related protein precursor, and the 3′ UTR of a predicted transmembrane protein containing HEAT repeats. In one case, four copies of the ribozyme appear within a ∼2,500-nt region of the genome. Some of these ribozymes are expressed and self-cleaved in vivo, as indicated by EST CD324081. Sequence-based searches seeded with the drz-Spur ribozymes revealed a number of additional putative ribozymes, some of which have ESTs corresponding to the self-cleaved RNAs. The in vitro activity of several of these new ribozymes has now been demonstrated (Table 2).

Other metazoan ribozymes.

There are two families of HDV-like ribozymes in the lancelet B. floridae and one in the sea lamprey Petromyzon marinus. One of the two copies of Bflo-1 maps to a peptidyl arginine deaminase gene and the drz-Pmar-1 ribozyme maps to a zinc finger domain, where it is expressed in vivo. The drz-Pmar-1 sequence revealed a similar ribozyme, drz-Lmen-1, in coelacanth Latimeria menadoensis. This small ribozyme exhibits strong temperature dependence (Fig. 3), a common feature of HDV-like ribozymes; however, its (previously unpublished) activity is independent of Mg²⁺ between 1 and 10 mM, suggesting that its affinity for Mg²⁺ is higher than in other ribozymes of this family.

Figure 3.

Secondary structure and in vitro activity of the coelecanth drz-Lmen-1 ribozyme. (A) Secondary structure of drz-Lmen-1. (B) Denaturing PAGE of ribozyme self-scission at 10 mM Mg²⁺, 37°C. (C) Graph of the in vitro ribozyme self-cleavage activity at 1 mM Mg²⁺ and indicated temperature.

In the nematode Caenorhabditis japonica, a single family of HDV-like ribozymes appears in many copies, representing a significant fraction of the organism's genome. The ribozymes are conserved in the 3′ flanking region but differ in the upstream sequence, suggesting that the ribozyme spread throughout the genome via retrotransposition of the cleaved RNA. There are 32 copies of a single ribozyme family in another nematode, Pristionchus pacificus, all of which are intergenic, but both upstream and downstream regions of the ribozymes show sequence conservation. The ribozymes can be grouped into two subfamilies, which are very similar in sequence, but their in vitro cleavage rate spans an order of magnitude in 10 mM Mg²⁺ at 37°C.

Bacterial and viral ribozymes.

The only bacterial HDV-like ribozyme found to date maps 106 nts upstream of a phosphoglucosamine mutase (GlmM) and 38 nts downstream of a cytochrome d ubiquinol oxidase genes of a human gut bacterium Faecalibacterium prausnitzii. The two genes may be part of a polycistronic mRNA processed by the ribozyme, because the genome contains another copy of the GlmM gene preceded by the ribozyme; however, this second copy of the ribozyme lacks the first six nts and is thus likely inactive. This copy of the ribozyme-GlmS gene may be the result of a retrotransposition event, during which the first six nts of the ribozyme were lost. Initial analysis indicated that the ribozyme self-cleaves slowly, with a rate constant of 0.46 h⁻¹, but expansion of the leader sequence to 46 nts resulted in a rate constant of 54 h⁻¹, perhaps due to the formation of a stable structure that prevents misfolding of the ribozyme.

The ribozyme of the insect virus Chilo iridescent virus (CIV, also known as invertebrate iridescent virus 6) maps 144 nt upstream of the DNA-dependent RNA polymerase (largest subunit) start codon and 120 nt downstream of a DNA binding protein gene. The cleavage site is close to the 5′ terminus of the RNAP transcript. The ribozyme may thus serve to inactivate the mRNA translation by cleaving off the 5′ cap once the expression of this immediate-early viral gene is no longer necessary, or in cap-independent translation, where the ribozyme may act to promote translation initiation.

R2 Retrotransposition in Drosophila

Following the discovery of the A. gambiae ribozyme at the 5′ of the RTE element, similar large HDV-like ribozymes were found at the 5′ termini of R2 elements in several Drosophila species.^38,41 The R2 element is one of the most extensively studied non-LTR retrotransposons. As in the case of the RTE retrotransposon, an essential step in the R2 retrotransposition cycle is co-transcriptional processing of the element from its upstream transcript. The 5′ co-transcriptional cleavage was observed more than 25 years ago⁴² and the 5′ UTRs of Drosophila R2 elements have been shown to contain regions of high base co-variation;⁴³ however, it wasn't until last year that the processing was attributed to a self-cleaving ribozyme whose base-paired elements consist of the conserved segments.⁴¹

Conclusion

The sequences of the similarly folded HDV-like ribozymes exhibit great diversity. Did these ribozymes arise independently multiple times or is the diversity a product of multiple horizontal gene transfers and subsequent variation in the base-paired and peripheral regions? A single origin of HDV-like ribozymes is based on two lines of evidence: (1) the ribozymes are informationally complex, therefore it is unlikely that their nested double-pseudoknot fold would evolve independently in genomic sequences multiple times; and (2) despite numerous in vitro selections of self-cleaving ribozymes from random pools,⁴⁴ HDV-like RNA has never been identified, even when the sample pool is sufficiently diverse (∼10¹⁶) that, statistically, such sequences could arise. These findings are in stark contrast with hammerhead ribozymes, which have a less complex fold, are also widely distributed in nature, and have been isolated several times in vitro.^45–48

Even though it is not yet clear how the HDV-like ribozymes evolved, the fact that they appear in such a widespread manner and varying biological roles, including retrotransposition and viral genome processing, suggests that they may be involved in many processes. Further analysis will surely reveal mechanisms by which these ribozymes are regulated and new pathways that are affected by their scission.