Intracellular Cleavage and Ligation of Hepatitis ... - Journal of Virology

JOURNAL OF VIROLOGY, Feb. 1995, p. 1190–1200 0022-538X/95/$04.0010 Copyright q 1995, American Society for Microbiology

Vol. 69, No. 2

Intracellular Cleavage and Ligation of Hepatitis Delta Virus Genomic RNA: Regulation of Ribozyme Activity by cis-Acting Sequences and Host Factors DAVID W. LAZINSKI

AND

JOHN M. TAYLOR*

Fox Chase Cancer Center, Philadelphia, Pennsylvania 19111-2497 Received 6 September 1994/Accepted 10 November 1994

During replication, a ribozyme within the genomic RNA of hepatitis delta virus cleaves multimeric precursors to release a unit-length linear intermediate. Intramolecular ligation of this intermediate produces the circular genomic RNA. Although one copy of the ribozyme is reconstituted by such ligation, it does not subsequently cleave and destroy the circular conformation. We have identified cis-acting attenuator sequences that prevent self-cleavage of the circular product by base pairing with and inactivating the ribozyme. Furthermore, we have shown that during the initial processing of the multimeric precursor RNA, host-specific factors activate the ribozyme by preventing its association with the attenuator sequences. Thus, we demonstrate a novel switching mechanism that regulates ribozyme activity inside the cell. an intramolecular ligation reaction so that a circular product results. However, the same target site and ribozyme sequences that function in the replicative precursor are reconstituted within this final circular product. Thus, if the circular conformation is to remain intact, the activity of the reconstituted ribozyme must be attenuated. In vitro, the HDV genomic ribozyme functions as an 85nucleotide (nt) contiguous sequence that uses Mg21 ions for cleavage and generates 59 hydroxyl and 29, 39 cyclic phosphate termini (11, 20). Although the so-called hammerhead and paperclip ribozymes from plant subviral RNAs produce the same termini (27), the structure of the HDV ribozyme is quite different from these and other examples. Figure 1A displays the proposed pseudoknot-like structure of the genomic ribozyme as adapted from Been (1). Evidence from both his laboratory and at least three additional laboratories supports the validity of the proposed structure (9, 15, 21, 28). However, during infection, this 85-nt sequence functions in the context of the whole 1.7-kb genome. The single-stranded genome is thought to fold into an unbranched rod-like structure in which roughly 70% of its nucleotides are paired (3, 30). The ribozyme is situated close to one end of this rod-like structure. In fact, addition of 132 nt of distal sequence is sufficient to allow potential rod-like pairing of the whole 85-nt ribozyme (Fig. 1B). The formation of such a structure should be incompatible with self-cleavage and, consistent with this possibility, ribozyme-containing RNAs that include the distal sequences display drastically reduced activity in vitro (11, 31). Thus, it has been proposed that in the circular genome, the ribozyme is inhibited by its interaction with distal sequences in the rod-like structure (14). If these distal sequences could be shown to inhibit ribozyme activity within the circular product, a second problem would exist. How would the initial self-cleavage of the linear precursor ever be achieved? It has been speculated that, in the infected cell, self-cleavage occurs during transcription of the nascent precursor, prior to the synthesis of the distal sequences (14). In this way, the ribozyme could fold and function in the absence of the inhibitory rod-like structure. The model predicts that, once synthesized, the distal sequences constitutively inhibit self-cleavage. Thus, the multimeric species observed at relatively low levels during HDV replication are predicted to

cis-acting ribozymes are catalytic RNAs that promote the site-specific cleavage and, in some cases, subsequent ligation of precursor RNA molecules (27). Although ribozymes from certain subviral pathogenic RNAs, as well as from the group I and group II self-splicing introns, each possess unique structures and cleave their substrates via different biochemical mechanisms, all have some common properties. These ribozymes tightly coordinate a divalent metal ion which is used in catalysis, and the specificity of reactions is governed by base-pairing interactions that guide the substrate into the ribozyme’s catalytic core. When a ribozyme resides within an intron, it is ultimately excised from the final product and any activity that it might subsequently possess does not threaten the integrity of the product. Such ribozymes can function constitutively, since there is no requirement for a regulatory mechanism to modulate their activity. In contrast, when a ribozyme remains included within a processed product, some mechanism must operate to ensure that the ribozyme is initially active for processing but is later inactive so that it does not recleave and destroy the product. In this study, we investigated such a ribozyme in the intracellular processing of human hepatitis delta virus (HDV) RNA. The circular, single-stranded genomic RNA of HDV contains a cis-acting ribozyme (11) that provides an essential function in the propagative cycle of the virus (17). During replication, the antigenome, a circular intermediate of complementary sequence, serves as the template for genome synthesis. Recent evidence indicates that the host RNA polymerase II (Pol II) is fully capable of providing such RNAdirected synthesis, at least in vitro (7, 16). Genome replication is thought to occur by a rolling-circle mechanism in which synthesis proceeds multiple times around the circular template. The ribozyme prevents the accumulation of very long multimeric intermediates by specifically cleaving the genomic RNA at a unique site (11). Each terminus of the resulting linear unit-length genomic RNA is generated from a separate self-cleavage event. These termini are then the substrates for * Corresponding author. Mailing address: Fox Chase Cancer Center, 7701 Burhome Ave., Philadelphia, PA 19111-2497. Phone: (215) 728-2436. Fax: (215) 728-3616. Electronic mail address: jm_taylor@ w470.fccc.edu. 1190

VOL. 69, 1995

HDV RIBOZYME REGULATION

1191

FIG. 1. Proposed alternate structures involving the HDV genomic ribozyme. (A) Pseudoknot structure of the active genomic ribozyme as adapted from Been (1). Self-cleavage occurs at the 59 end of the 85-nt sequence between the U and G residues, as indicated in boldface print. The residues that were targeted for interaction with the synthetic attenuator are shown in boldface italics. (B) Region of the rod-like structure that includes a portion of the ribozyme and its partially complementary sequences. This structure was predicted by the RNAFOLD program (32). An equivalent phosphodiester linkage between the U and G residues, which is cleaved by the ribozyme, is thought to be restored upon ligation. Again, those residues are shown in boldface print. Boldface italics are used to highlight the guide region which was proposed to be used for the ligation of the termini that reside on the opposite side of the rod-like structure.

be unprocessed by-products, rather than true intermediates. Although this model is plausible, its validity had not been investigated. Our understanding of the ligation of the termini of processed HDV replication intermediates is also incomplete. In an in vitro reaction, no ligation was observed following the incubation of two RNA fragments, one containing the 29, 39 cyclic phosphate and the other containing the 59 hydroxyl terminus as generated from self-cleavage. However, ligation was observed, albeit with very slow kinetics under nonphysiological conditions, upon the addition of a third RNA (26). This RNA included the partially complementary sequences distal to the ribozyme that could form the rod-like structure after annealing to the substrate RNAs. Two hypotheses were suggested to explain the requirement for the ribozyme distal sequences. First, they were thought to bind to and inactivate the ribozyme, so as to stabilize the ligated product. Second, these sequences, which are partially complementary to the regions that flank the two termini, were proposed to function as a guide (i.e., splint) RNA on which the two termini could be brought into physical proximity (Fig. 1B). Again however, neither of these possible roles had been further examined, either in vitro or in vivo. A more complete understanding of the mechanisms which govern cleavage and ligation requires an investigation of these reactions, not only in vitro but also as they occur within the host cell. Recently, we analyzed the processed products obtained after transcription of HDV RNA precursors from cDNA templates in mammalian cells (13). We found that no more than a 348-nt contiguous region from the HDV genome

is required for both cleavage and ligation and that these processes can occur efficiently in the absence of the only HDVencoded protein, the delta antigen. However, the possibility that fewer than 348 nt might be sufficient for both processes was not tested. Here, we have continued and expanded upon that study to address the following questions. (i) What are the minimal sequence requirements for the generation of cleaved and ligated species inside the cell? (ii) What additional sequences, if any, are needed to attenuate the activity of the reconstituted ribozyme within the circular product? (iii) Does the host polymerase, or any other host factor, participate in the regulation of the cleavage and ligation reactions, and how might such regulation be achieved? MATERIALS AND METHODS RNA expression vectors. The vector used to express the m1 RNA in vitro was designated pDL610. This plasmid was derived from pGem4Z (Promega) in which sequences between the PstI and blunted EcoO109I sites were replaced with HDV sequences, according to the numbering of kuo et al. (10), starting at position 656 and continuing to position 966, followed by sequences originating at position 619 and ending at position 785. The deletion between positions 966 and 619 was created by excising the SalI-StyI fragment and subsequently filling in and ligating the resulting cohesive overhangs. The m2-, m3-, and m4-expressing vectors, pDL613, pDL612, and pDL611, respectively, were produced by deleting sequences from pDL610. The precise endpoints of each deletion are indicated in Fig. 3. The restriction sites used to generate these deletions were EcoO109IRsrII, RsrII-EaeI, and XbaI-SalI, respectively. The m5-expressing vector, pDL614, was constructed by annealing the oligonucleotide 59 TCGAATTCGC CATTAAATGTTGCCC 39 to a second oligonucleotide, 39 TAAGCGGTA ATTTACAACGGGAGCT 59, and inserting this duplex into the unique SalI restriction site of pDL611. The same duplex was inserted in inverted orientation

1192

LAZINSKI AND TAYLOR

J. VIROL.

FIG. 2. A possible pathway for the processing of precursor into circular HDV RNA following transcription from cDNA. The precursor RNA has two ribozymes which ultimately cleave it into L, M, and R fragments. The sites of self-cleavage are depicted by black circles, the minimal ribozyme sequences are shown as shaded rectangles, and the sequences distal and partially complementary to the ribozyme (i.e., attenuator) are shown as unshaded rectangles. All reactions are considered to be essentially unidirectional except for the presumed interconversion of linear and circular species (MNMc).

into the same site within pDL611 to generate the m6-expressing plasmid, pDL615. The corresponding cytomegalovirus (CMV) promoter-containing vectors, pDL616 to pDL621, were then generated from pDL610 to pDL615, respectively. In each case, the PvuII-HindIII fragment containing the T7 promoter, was excised and replaced with a StuI-HindIII CMV promoter-containing fragment which was derived from psTat (kindly provided by L. Tiley and B. Cullen, Duke University Medical Center). For expression with T7 RNA polymerase (T7 RNAP) in Escherichia coli and in Huh7 cells, the Tf transcription terminator was inserted downstream of HDV sequences in pDL610 to pDL615 to produce pDL625 to pDL630, respectively. In each case, sequences immediately downstream of the last HDV nucleotide, continuing to the unique ScaI site, were replaced by a blunted HindIII-ScaI fragment derived from pGemex2 (Promega). The two plasmids that express T7 RNAP in mammalian cells, pAR3126 and pAR3132, were kindly provided by J. Dunn, Brookhaven Laboratories (6). Transfections, sample harvest, and Northern analysis. For all experiments, the human hepatoma cell line Huh7 was grown on 35-mm tissue culture plates with Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum and transfected at near confluence with calcium-precipitated DNA. In all transfections, the secreted alkaline phosphatase expression vector was included so that efficiency could be monitored as previously described (2, 13). Three days following transfection, cells were harvested and total RNA was recovered by a modification of the acid guanidine-thiocyanate-phenol-chloroform method (4). For Northern (RNA) analysis, samples were resuspended in a loading buffer that contained 90% formamide, 10 mM Tris (pH 7.9), and 1 mM EDTA and were then subjected to polyacrylamide electrophoresis in the presence of 10 M urea. For Fig. 4, 20% of the total sample was loaded per lane, while for Fig. 8, 1% was loaded. In all experiments, pDL616, digested with MspI and end-labeled with 32P, was used as the DNA marker. Following electrophoresis, RNA was transferred electrophoretically onto charged nylon membrane (Zetaprobe GT; Bio-Rad), cross-linked with UV light, and then hybridized and washed as previously described (3). For all experiments, a uniformly labeled RNA probe complementary to position 686 to 785 of the HDV genome was used. A Fuji Bas 1000 radioimager was used to quantitate both the signal strength and distance traveled for each RNA species. In vitro transcription and self-cleavage reactions. To produce uniformly labeled precursors, 1 mg of plasmid linearized with SspI was added to a T7 RNAP reaction mixture under standard conditions except that incubation was at room temperature to minimize self-cleavage during transcription. These 25-ml reaction mixtures contained 500 mM ATP, GTP, and CTP; 10 mM UTP, 0.5 mM a-32Plabeled UTP (800 Ci/mmol), 40 mM Tris-HCl, 25 mM NaCl, 8 mM MgCl2, 2 mM spermidine, 5 mM dithiothreitol, 52 U of RNasin (Promega), and 10 U of T7 RNAP (Promega). Reactions were stopped after 30 min, and RNA was extracted by the acid guanidine-thiocyanate-phenol-chloroform method (4). For all samples shown in Fig. 6 except lane 1f of Fig. 6B, RNAs were resuspended in a buffer containing 40 mM NaCl, 20 mM Tris-HCl (pH 7.9), and 4 mM MgCl2, while the buffer used for the sample in lane 1f additionally contained formamide at a final proportion of 66%. For Fig. 6A, samples were incubated at 378C for 30 min to allow self-cleavage and the reactions were then stopped by the addition of EDTA

and formamide to final concentrations of 10 mM and 90% respectively. For Fig. 6B, samples were heated to 958C for 1 min, cooled in a dry ice-ethanol bath for 30 s, and then incubated at 378C for 10 min. After this process was repeated for a total of five cycles, the reactions were again stopped by the addition of EDTA and formamide to final concentrations of 10 mM and 90%, respectively. RNA expression in E. coli. The strain BL21:DE3 was used to express HDV RNA. This strain contains an integrated copy of the gene that expresses T7 RNAP under the control of the inducible lac promoter. Tight repression of this promoter was obtained by transformation of the cells with the Kanr, pACYC 177 derivative, pREP 4 plasmid, which expresses the lacIq repressor at high levels. After transformation with the compatible Ampr pDL625 to pDL630 plasmids, cells were grown in Luria broth at 378C to an optical density at 600 nm of 0.25 and then induced for 15 min by the addition of 5 mM isopropyl-b-D-thiogalactopyranoside (IPTG). One milliliter of culture was then centrifuged for 1 min in a microcentrifuge, and the resulting pellet was quickly resuspended at 08C with 0.1 ml of a buffer containing 50 mM glucose, 10 mM EDTA, and 25 mM Tris-HCl (pH 7.9). RNAs were then isolated by a modification of the acid guanidine-thiocyanate-phenol-chloroform method (4). For Fig. 7, 1% of the total sample was loaded per lane.

RESULTS Experimental design. We employed a genetic approach by studying the ability of six mutant HDV RNAs to undergo processing following synthesis from a cDNA template. Figure 2 depicts a possible pathway by which such processing could yield circular HDV RNA. This diagram shows a primary transcript that has defined 59 and 39 ends and contains two copies of the genomic ribozyme. The corresponding two self-cleavage sites define the leftward (L), middle (M), and rightward (R) regions that constitute the unprocessed precursor RNA (LMR). A single self-cleavage event is expected to produce one of two potential intermediates, MR or LM, each of which, after a second self-cleavage event, should release the unitlength linear product M. Ligation of the M termini generates the final circular product, Mc. This last step is depicted as a bidirectional event. In the reverse direction, the ribozyme cleaves Mc to create M. Thus, maintenance of the circular product could be achieved by inhibition of the ribozyme in the product. Although M and Mc are of the same length, they can still be resolved as two distinct species after polyacrylamide gel electrophoresis under denaturing conditions. Circular RNAs are

VOL. 69, 1995


1193

FIG. 3. Predicted circular RNA products obtained after processing of six mutant HDV precursors. Regions of base pairing are indicated with dashes within the rod-like structure. The self-cleavage site is shown as a black circle, a shaded rectangle indicates the position of the ribozyme, and the attenuator is depicted as an unshaded rectangle. Dotted lines denote the sequences deleted in each mutant, and the number at each deletion junction corresponds to that position on the HDV genome according to the numbering of Kuo et al. (10). attf designates the synthetic attenuator inserted in a forward and functional orientation, while attr refers to the same sequence inserted in reverse orientation.

known to migrate more slowly than linear RNAs during such conditions, and the difference is enhanced as the percentage of polyacrylamide is increased. For this reason, electrophoresis on gels of both 4 and 6% polyacrylamide was used as a method to identify circular species (13). We first examined the importance of sequences distal to the genomic ribozyme for the generation and maintenance of circular products. Toward this end, we used a deleted construct with 348 nt of unique HDV sequence (13). Figure 3 depicts the circular, rod-like structure of that mutant, m1, as well as five additional mutants derived from it, m2 to m6. In m2, a region starting immediately after the minimal ribozyme sequence and continuing through part of the partially complementary sequences distal to the ribozyme was deleted. This construct retained a portion of the rod-like structure that could potentially interact with the 59 portion of the ribozyme and, in addition, the region partially complementary to sequences flanking the site of ligation which has been proposed to function as a guide RNA (Fig. 1B). In m3, the sequences starting immediately 39 to the deletion junction of m2 and continuing to the nucleotide whose phosphate linkage is cleaved by the ribozyme were excised. In this case, only the 39 region of the ribozyme could be involved in rod-like structure, and the putative guide sequence was removed. In m4, the potential for involvement of any part of the ribozyme in rod-like structure was eliminated, as was the putative guide region. Attenuator sequences stabilize the circular products. The four mutant RNAs m1 to m4 were tested for their ability to exist in the circular conformation within the mammalian cell. The corresponding mutant cDNAs were inserted into an expression vector under the control of the CMV immediate-early promoter. The termination of transcription was not specified in these vectors. Therefore, all R-containing RNAs (LMR, MR, and R) should contain heterologous 39 ends and are not expected to exist as discrete species of defined lengths. RNA samples harvested after transfection of the human hepatoma cell line Huh7 were subjected to Northern analysis, with a probe complementary to the ribozyme sequences. Since the probe used was not complementary to L, it should be possible to detect only LM, M, and Mc as species of defined lengths. The steady-state levels of mutant RNAs m1 to m6, as expressed and processed in Huh7 cells, are shown in Fig. 4. In this and subsequent figures, we will follow a convention in

which the lane number in a figure indicates the mutant under analysis; i.e., m3 will always appear in lane 3. The identities of the linear species were deduced from their sizes compared with those of linear DNA markers. For instance, the predicted sizes for the LM and M species generated from m3 are 307 and 185 nt, respectively. Consistent with this, the assigned band for LM on the 4% gel migrated slightly faster than the 319-nt DNA marker while M migrated slightly faster than the 190-nt marker (Fig. 4A, lane 3). The same relationships were also observed with the 6% gel (Fig. 4B, lane 3). In contrast, the band assigned as the 185-nt circular RNA, Mc, displayed a retarded mobility compared with linear markers on the 4% gel (migrating between the 242- and 190-nt markers), and mobility was further retarded on the 6% gel (migrating between the 319- and 242-nt markers). As an independent method to confirm that bands were correctly assigned, the logarithms of the predicted lengths of all M and Mc species from Fig. 4A and B were plotted against their migrations (Fig. 5). The resulting curves behaved as would be expected theoretically; i.e., all four curves were linear and had very similar slopes. Furthermore, the two lines generated with the linear M species for each gel percentage nearly superimposed, while the line generated with the mobilities of the Mc species on the 4% gel was shifted closer to origin of the mobility axis. This shift was more pronounced with the line generated from Mc run on the 6% gel. Quantitation of the data shown in Fig. 4 revealed that the processed RNAs of m1, m2, and m3 were detected predominantly as circular species (81, 81, and 87%, respectively). Thus, the putative guide region, which is absent in m3, is not required for intracellular circle formation. In contrast to the first three mutants, m4 was almost exclusively linear (.99%). However, a relatively longer exposure of Fig. 4 did reveal low amounts (0.8%) of a band whose mobility was consistent with that expected for an m4 circular species. This designation was confirmed in additional experiments (data not shown). Our interpretation of these results is as follows. Inside the cell, there may be an equilibrium between the unit-length linear and circular species M and Mc, respectively. If an RNA, such as m1 to m3, contains sequences which can base pair with, and thereby attenuate the ribozyme in, Mc, then the equilibrium is driven toward the accumulation of circular species. If however, as in m4, such attenuator sequences are completely absent, the ribozyme reconstituted by ligation promptly recleaves the cir-

1194

LAZINSKI AND TAYLOR

J. VIROL.

FIG. 4. Northern analyses of RNAs expressed in Huh7 cells by the host RNA Pol II. Mutants m1 to m6 appear in lanes 1 to 6, respectively. The letters used to identify each RNA species correspond to those used in Fig. 2 and the text. DNase I treatment of the RNA samples was omitted because the Mg21-containing reaction buffer might activate the HDV ribozyme, potentially causing an underestimation of ligated species. Lane 1* of panel B was given 1/12th the exposure time that lane 1 from the same panel received, so that the Mc band could be distinguished from the template DNA signal immediately above it. The predicted sizes of m1 to m6 are 482, 381, 319, 305, 330, and 330 nt, respectively, for the LM species and 348, 247, 185, 171, 196, and 196 nt, respectively, for the M and Mc RNAs. Sizes of linear DNA markers (in nucleotides) are indicated at the left.

cular product and the equilibrium is shifted in favor of linear molecules. Circular RNA accumulated in the presence of a synthetic attenuator sequence. In such an interpretation, the attenuator sequences possess no catalytic activity and function solely by inactivating the ribozyme through base pairing. To test this, we examined whether a heterologous sequence that was designed to base pair with the ribozyme could provide the same function as the natural attenuator. We inserted into m4 a 25-nt oligomer that included two stretches, of 10 and 11 nt, which were complementary to two predominantly single-stranded regions within the proposed pseudoknot structure of the ribozyme. The ribozyme nucleotides which were the target for interaction

FIG. 5. Semilog plot of length versus migration of RNA species from Fig. 4. The linear species of m1 to m6 and the circular species of m1 to m3 and m5 were plotted. The linear regression (r) values ranged from 0.993 to 0.999 for the four lines.

with this synthetic attenuator sequence are shown in boldface italics in Fig. 1A. The resulting mutant is shown as m5 in Fig. 3. The synthetic attenuator in m5 differs from the natural attenuator sequences in that the former contains two small interrupted stretches of perfect complementarity with the ribozyme while in the latter case, there is an average of 76% complementarity over a contiguous 86-nt region. In addition, the synthetic attenuator was analyzed for its similarity to all sequences on the HDV genome with the computer algorithm Bestfit of the Genetics Computer Group package. In this analysis, the synthetic attenuator displayed no homology with the natural sequences it was designed to replace. As a negative control, the complement of the synthetic attenuator sequence was inserted into m4, so as to generate m6. In this case, the inserted sequence was not complementary to the ribozyme and was not expected to function as an attenuator. As can be seen in Fig. 4, lanes 5 and 6, 32% of the processed product of m5 was circular but when the attenuator was inverted in m6, this value was reduced 100-fold, to 0.3%. These results support the inference that any sequence capable of binding to and inactivating the ribozyme could promote the accumulation of circular products. In the mammalian cell, attenuator-mediated inhibition of the ribozyme is suppressed following synthesis. According to the above interpretation, both the natural and synthetic attenuator sequences in m1 to m3 and m5 inhibited the ribozyme in the circular products and yet they could not have interfered with its function in the RNA precursors or the M species would never have been generated. This apparent paradox could be reconciled by a model in which the ribozyme promptly cleaves the nascent precursor prior to the synthesis of the inhibitory sequences. In such a model, the attenuator sequences, if present, should constitutively inhibit self-cleavage. To test this, we inserted an additional copy of the natural attenuator immediately upstream of the HDV sequences in an m1 expression vector. In this case, even though synthesis of the attenu-

VOL. 69, 1995


1195

FIG. 6. Expression of RNAs in vitro with T7 RNAP. (A) Results obtained after RNAs were subjected to a single self-cleavage reaction; (B) results obtained after RNAs were subjected to five cycles of heat denaturation and self-cleavage. In lane 1f of panel B, cycled heat denaturation was performed in the presence of 66% formamide. Band assignments were made on the basis of observed mobilities compared with DNA size markers. In all cases, there was no more than a 5% discrepancy between the observed and predicted sizes. The predicted sizes of m1 to m6 are 677, 576, 514, 500, 525, and 525 nt, respectively, for the LMR species; 623, 522, 460, 446, 471, and 471 nt, respectively, for the MR species; and 402, 301, 239, 225, 250, and 250 nt, respectively, for the LM species. The predicted sizes for the M species of m1 to m6 are the same as those shown in Fig. 4. All mutants are predicted to generate a 275-nt R fragment and a 54-nt L fragment. Sizes of linear DNA markers (in nucleotides) are indicated at the left.

ator sequence preceded that of the ribozyme, it did not interfere with self-cleavage since both M and Mc were produced efficiently (data not shown). Thus, we proceeded to test the alternative hypothesis that, inside the mammalian cell, attenuator sequences are transcribed prior to self-cleavage but somehow the attenuator-ribozyme interaction is transiently prevented. In vitro, self-cleavage of the precursor is inhibited by attenuator sequences. We next tested whether, in vitro, the attenuator-ribozyme interaction in the precursor might be similarly prevented. T7 RNAP was used to generate uniformly labeled RNA precursors of m1 to m6. In this case, the template DNAs were cleaved by digestion with a restriction endonuclease so that the resulting primary transcripts would have both defined 59 and 39 termini. Thus, following self-cleavage of the primary transcripts, it was theoretically possible to detect all seven RNA species (LMR, MR, LM, L, R, M, and Mc) depicted in Fig. 2. The results after electrophoresis on denaturing 4% polyacrylamide gels are shown in Fig. 6A. Of the seven possible predicted species, all but Mc were readily detected as products of the two mutants that lacked attenuator sequences, m4 and m6 (Fig. 6A, lanes 4 and 6). Mc was not expected, since this species was barely detected with these mutants in mammalian cells (Fig. 4). However, in contrast to what was observed in mammalian cells, virtually no M species were detected in vitro following synthesis of the four mutants that contained attenuators (m1 to m3 and m5). From the quantitation of the molar abundance of LMR, MR, LM, and M in Fig. 6A (Table 1), we determined that at least 30-fold more M was produced from mutants that lacked attenuators than from those that contained such sequences. We concluded that in vitro, unlike in vivo, the attenuator interferes with the processing of precursor RNAs.

Attenuation of self-cleavage was relieved upon heat denaturation. Although the presence of attenuator sequences prevented the generation of doubly cleaved products (M), the two singly cleaved intermediates, LM and MR, were produced far more efficiently (Fig. 6A and Table 1). Since each precursor molecule (LMR) contained two copies of the ribozyme but only one copy of the attenuator sequence (Fig. 2), our interpretation is that only one of the two ribozymes in a given precursor molecule could be inhibited by the attenuator. For example, if in one molecule the attenuator bound to and inactivated the upstream ribozyme, then the downstream ribozyme could function to produce LM. Conversely, in another molecule, when the attenuator paired with the downstream ribozyme, then the upstream ribozyme might function to create

TABLE 1. Relative proportions of precursors, intermediates, and products generated in the experiment reported in Fig. 6 RNA species

Treatment

Amt. of RNA species (mol%) generated for indicated RNA mutant m1

m2

m3

m4

m5

m6

LMR

1 cleavage 5 cycles

14.6 1.4

41.7 1.6

65.2 5.1

29.1 0.2

22.0 0.3

25.7 0.1

MR

1 cleavage 5 cycles

27.1 24.8

10.2 10.7

26.0 73.9

5.6 0.1

11.5 4.6

2.8 0.2

LM

1 cleavage 5 cycles

57.7 71.4

47.4 83.0

8.5 8.9

34.5 1.1

65.8 17.2

46.7 0.4

M

1 cleavage 5 cycles

0.6 2.3

0.8 4.7

0.3 12.2

30.8 98.5

0.7 77.9

24.8 99.3

1196

LAZINSKI AND TAYLOR

J. VIROL.

FIG. 7. Northern analyses of RNAs expressed in E. coli cells by T7 RNAP. The predicted sizes of RNAs are the same as those shown in Fig. 6 except that all R-containing species are expected to be 26 nt shorter since, in this case, the 39 end was specified by the Tf transcription terminator. Sizes of linear DNA markers (in nucleotides) are indicated at the left.

MR. According to this interpretation, if MR were now subjected to heat, then the interaction between the attenuator sequences and downstream ribozyme might be disrupted. Upon renaturation, in at least some of the molecules the attenuator might pair with the upstream ribozyme of MR so that the downstream ribozyme would now be active. With repeated cycles of denaturation-renaturation, it should be possible to generate increasing amounts of M (and R) from the MR intermediate. This possibility was tested by subjecting RNA transcripts to cycles of denaturation at 958C followed by renaturation and then incubation at 378C under standard cleavage reaction conditions. After five cycles, the extent of cleavage was greatly enhanced (Fig. 6B). Fewer of the precursors and intermediates (LMR, MR, and LM) and more of the final products (L, M, and R) were observed with all six RNAs than when a single cleavage reaction was performed (compare Fig. 6B with 6A and see Table 1). This is consistent with previous results which showed that thermal cycling stimulates self-cleavage, even in the absence of an attenuator (31). The results shown in Fig. 6B also provided further evidence that band assignments were correct, since they were in accord with the predicted precursorproduct relationship between the individual species. As was predicted, cycled denaturation greatly increased the yield of the doubly cleaved product M obtained from attenuator-containing mutants (see Table 1). With m5, which contained the synthetic attenuator, thermal cycling increased the yield of M about 100-fold while, with m6, which lacked a functional attenuator, there was only a 4-fold increase. Furthermore, the amount of M obtained in the presence of an attenuator sequence was inversely related to the extent of possible base pairing in the rod-like structure. For instance, m5 had only 21 bp and yielded the highest amount of M (78%); m3 had 56 bp and yielded 12.2% M; m2 had 74 bp and yielded 4.7% M; while m1, which contained 128 bp in the predicted rod-like structure, yielded only 2.3% M on a molar basis. This low yield of M obtained from m1 may have been a consequence of incomplete denaturation of the rod-like structure. Since the HDV ribozyme can function in the presence of denaturants (24), 66% formamide was included in the reaction mixture to enhance melting during the thermal

cycling procedure. Under these conditions, m1 yielded 31% doubly cleaved product (Fig. 6B, lane 1f). We therefore concluded that, in vitro, inhibition by the attenuator can be alleviated if the attenuator-ribozyme interaction is transiently denatured. In E. coli, the attenuator sequences interfered with processing of the precursor. Almost no doubly cleaved products from attenuator-containing precursors were obtained in vitro unless the extreme conditions of heat and denaturants were employed. Thus, some essential component capable of preventing the attenuator-ribozyme interaction in mammalian cells was absent from our in vitro reaction. This component might be an ion or cofactor common to all cells or could be a specific factor unique to mammalian cells. As an initial attempt to distinguish between these possibilities, we tested whether the same six mutants could be properly processed in an E. coli strain that expresses T7 RNAP. In this case, precise 39 ends were specified by inserting a transcription terminator from bacteriophage T7 (Tf) downstream of HDV sequences in the expression vectors. The results of Northern analysis of the six mutant RNAs expressed in E. coli are presented in Fig. 7. With m1, numerous unexpected species were observed migrating between MR and R (Fig. 7, lanes 1). These were interpreted to be degradation products generated by the heterologous host. Therefore, no conclusions could be made from the results obtained with this mutant. However, with the other five mutants, significantly fewer degradation products were observed. M was generated from the precursors of the two mutants that lacked attenuator sequences, m4 and m6, but could not be detected with m2, m3, and m5, mutants that contained attenuator sequences. This is best exemplified by comparing m5 and m6, which are of the same length but contained the synthetic attenuator in either a functional or inverted orientation. None of the bands observed on the 4% gel displayed altered mobility in relation to their mobilities on the 6% gel (compare Fig. 7A with 7B). From this we concluded that no circular species were detected. Overall, the results obtained upon expression in E. coli mirrored those obtained after expression in vitro. We concluded that regulation of the attenuator-ribozyme interaction in precursor molecules did not occur in E. coli. Thus, factors present within mammalian

VOL. 69, 1995


1197

FIG. 8. Northern analyses of RNAs expressed in Huh7 cells by T7 RNAP. In lanes 1 to 6, cells were cotransfected with pAR3132, a simian virus 40-based vector that expresses a version of the T7 RNAP which contains a nuclear localization signal (6). In lane 12, this vector was replaced with unmodified pSVL, while in lane 1c, it was replaced with pAR3126, a vector that expresses a cytoplasmic form of T7 RNAP. The predicted sizes of RNAs are the same as for those shown in Fig. 7; sizes of linear DNA markers (in nucleotides) are indicated at the left.

cells, but absent from E. coli, must be responsible for this phenomenon. Processing of RNA in mammalian cells occurred without a requirement for synthesis by the host polymerase. The incomplete processing of T7 RNAP-generated precursors expressed in E. coli might reflect a requirement for synthesis by the mammalian RNA polymerase II. This Pol transcribed precursors which were efficiently processed (Fig. 4) and is thought to be responsible for RNA-directed synthesis during HDV replication (7, 16). Thus, it is possible that either Pol II or one of its accessory factors might transiently destabilize the ribozymeattenuator interaction in the nascent transcript. Alternatively, such destabilization could be achieved by other mammalian cell-specific factors after synthesis of the precursor RNA is complete. In this case, the identity of the polymerase responsible for synthesis would be irrelevant. To distinguish between these two models, the same vectors used to express m1 to m6 in E. coli were introduced into Huh7 cells along with a second vector that directs the synthesis of T7 RNAP (6). Since expression in this system is independent of Pol II, the production of M and Mc would provide evidence in support of the latter model. The results from Northern analysis of RNA produced in these transfected cells are shown in Fig. 8. Products in lanes 12, 1c, and 1, were from cells that were cotransfected with the m1 expression vector as well as a second vector which expressed either no T7 RNAP, a form of T7 RNAP that localizes in the cytoplasm, or a form of T7 RNAP that localizes in the nucleus, respectively. The absence of signal in the control lane, 12, confirmed that RNA synthesis in all other lanes was dependent on T7 RNAP. Surprisingly, only a twofold difference in the amount of total RNA and no difference in the overall pattern of processing were observed when cytoplasmic and nuclear T7 RNAP expression were compared (lanes 1c and 1). In each case, significant amounts of M, and especially Mc, were generated. The other three attenuator-containing mutants, m2, m3 and m5, behaved similarly. In contrast, the two mutants that lacked attenuator sequences, m4 and m6, yielded no detectable circular species. Overall, the results obtained after

expression with T7 RNAP were very similar to those observed after expression by Pol II (compare Fig. 4 and 8). Two exceptions were that (i) during synthesis by T7 RNAP, the 39 end of the primary transcript was specified and this allowed detection of the R-containing species, and (ii) approximately 100-fold more synthesis was obtained with T7 RNAP expression than with that expressed by the host Pol II enzyme after initiation from the CMV promoter. We conclude that processing of HDV precursors can occur without a requirement for synthesis by the host polymerase. Thus, suppression of the attenuatorribozyme interaction in the precursor must have been dependent on the activity of other host-specific factors. The unattenuated ribozyme cleaved the circular product. Thus far, we have provided evidence in support of a model for ribozyme regulation in which attenuator sequences stabilize the circular product by inactivating the ribozyme. Implicit in this model is the assumption that, in the absence of such sequences, the ribozyme would cleave Mc to generate the linear molecule M. We next set out to directly test the validity of this assumption. m1 RNA was isolated from Huh7 after expression by T7 RNAP and then subjected to Northern analysis as shown in Fig. 9. In lane A, 85% of the doubly cleaved product of this untreated RNA sample was circular [i.e., amount of Mc/(amount of Mc 1 amount of M) 5 0.85]. This RNA was then incubated overnight in hybridization buffer, precipitated, resuspended in self-cleavage buffer, and subjected to either one or five cycles of self-cleavage (Fig. 9, lanes B and D, respectively). In each case, following such treatment, Mc remained the majority species (67 and 63%, respectively). Since no substantial reduction in the proportion of circular molecules was observed, we concluded that when attenuated, the ribozyme was unable to cleave Mc so as to generate M. We next tested whether the ribozyme might be capable of cleaving Mc to M if the ribozyme-attenuator interaction was disrupted. This was accomplished by subjecting RNA samples to the same treatments as were performed in lanes B and D except that during the hybridization step, an antisense RNA complementary to the attenuator sequences was included. Upon hybridization of the attenuator sequences with the anti-

1198

LAZINSKI AND TAYLOR

FIG. 9. Northern analysis of the self-cleavage of circular m1 RNA following disruption of the ribozyme-attenuator interaction. All lanes were loaded with 0.2 mg of total RNA isolated from Huh7 cells 3 days after cotransfection with m1 and T7 RNAP expression vectors. In lane A, the RNA was untreated. In lanes B to E, RNA was incubated overnight in hybridization buffer (5 Prime-3 Prime, Inc.) at 558C, precipitated with ethanol, and then resuspended in the same selfcleavage buffer used for Fig. 6. Samples from lanes B and C were subjected to a single self-cleavage reaction, as for Fig. 6A, while those in lanes D and E were cycled through five cleavage reactions, as for Fig. 6B. For samples from lanes C and E, 0.02 mg of RNA complimentary to the attenuator (att; positions 781 to 962 of the antigenomic strand) were included during the hybridization step. Sizes of linear DNA markers (in nucleotides) are indicated at the left.

sense RNA, the ribozyme, freed from interaction with the attenuator, should be able to fold into an active conformation. Consistent with this, after a single self-cleavage reaction at physiological temperature and conditions, most of Mc was converted to M and only 11% of the molecules remained circular (Fig. 9, lane C). Moreover, following five cleavage cycles, more than 98% of the circular molecules were linearized (Fig. 9, lane E). We therefore concluded that the circular product is a substrate for ribozyme cleavage and that the ribozyme must be attenuated if the circular conformation is to be preserved. DISCUSSION Our studies with six mutant HDV RNAs provide new information regarding the minimal sequences required for cleavage and ligation to occur within the mammalian cell. Precursors from both m2 and m3 were extensively processed to the circular form (Fig. 4), yet these mutants had only 84 nt of common sequence. This sequence corresponds to the minimal genomic ribozyme as previously defined in vitro (20). We therefore conclude that no additional sequences play an obligatory role in the formation of the circular product inside the cell. However, we also observed that certain additional sequences, distal to the ribozyme, were needed for the stable maintenance of this circular product. We propose that these attenuator sequences base pair with and thereby inactivate the ribozyme so that subsequent recleavage of the circular product cannot occur. Consistent with this proposal, we found that the attenuator-ribozyme interaction prevented self-cleavage of circular molecules in vitro; however, when this interaction was pre-

J. VIROL.

vented, virtually all circular molecules were cleaved to the linear form (Fig. 9). Although the natural attenuator sequences that reside on the portion of the rod-like structure opposite to the ribozyme include nt 815 to 900, we observed that effective attenuation could be obtained with different subfragments of this region (e.g., m2 and m3). In addition, a small synthetic attenuator of dissimilar sequence could function in place of the natural sequences. Thus, attenuator sequences inactivate the ribozyme through simple base pairing and possess no catalytic properties per se. Contrary to prior speculation based upon in vitro studies (26), mutants m3 and m5 demonstrate that circle formation can proceed efficiently within the mammalian cell in the apparent absence of a guide sequence that could bring the 59 hydroxyl and 29, 39 cyclic phosphate termini into physical proximity (Fig. 4). In other systems in which ribozymes are known to promote ligation, essential guided interactions occur between the ribozyme and its substrates (27). Our finding that similar interactions are not required for HDV RNA ligation might reflect the involvement of a host ligase. Guideless ligation is more readily accommodated by such a model, since an RNA ligase is expected to bind to both substrate termini and bring them into proximity within its catalytic core. Thus, in this case, the protein would provide the guide function. Our inability to detect intramolecular ligation in vitro with the doubly cleaved M species of Fig. 6B, even following prolonged incubation under a variety of reaction conditions (data not shown), is also consistent with a possible requirement for a host ligase. Although the potential involvement of a host ligase in the ligation of HDV RNA has not been widely entertained, there is precedence from a related plant pathogen in support of such a possibility (8). Nevertheless, an unguided intramolecular selfligation mechanism remains possible, and the results obtained here underscore the importance of further study to determine the means by which HDV ligation occurs within the cell. It was previously speculated that, when present in a transcript, attenuator sequences constitutively inhibit the adjacent ribozyme. Thus, if self-cleavage is to occur, it must happen during transcription and prior to the synthesis of the distal attenuator sequences (14). However, we have obtained two lines of evidence which are inconsistent with that speculation. First, with a construct in which transcription of the attenuator sequences actually precedes, rather than follows, that of the ribozyme, no inhibition of self-cleavage was observed (data not shown). Second, the processing of RNA precursors generated by T7 RNAP in E. coli was incomplete (Fig. 7) and yet the same precursors synthesized by the same polymerase in mammalian cells were processed to yield M and Mc (Fig. 8). Thus, factors present in mammalian cells but absent from E. coli suppressed attenuator-mediated inhibition. For such suppression to be accomplished during transcription and prior to the synthesis of the attenuator sequences, the mammalian factors would have to interact with the heterologous bacteriophage T7 RNAP. We therefore propose the alternate and more likely possibility that host factors operate posttranscriptionally to transiently destabilize the attenuator-ribozyme interaction in the multimeric precursor. The proposed involvement of host factors in ribozyme-mediated catalysis is not without precedent. Ribozymes of certain group I and group II self-splicing introns display little or no intrinsic activity in vitro, but this activity is greatly stimulated in vivo by host factors (12, 19). In these cases, however, selfcleavage of the intron is not regulated. Rather, the host factors act constitutively to stimulate ribozyme activity. In contrast, the HDV ribozyme is truly regulated in that it is switched on in the

VOL. 69, 1995

precursor but switched off in the circular product. Unlike the aforementioned self-splicing introns, in the absence of attenuator sequences the HDV ribozyme is fully active without the need for host factors. However, as shown here, such activity is unregulated so that the ribozyme is switched on in the product and the circular conformation is promptly destroyed (Fig. 4). In the presence of attenuator sequences, but in the absence of host factors, again regulation is abolished and the ribozyme is switched off in the precursor so that processing is incomplete (Fig. 7). Regulation is restored only when both host factors and attenuator sequences are provided (Fig. 8). Although the identity of the proposed regulatory host factor(s) remains unknown, results obtained here do provide information about potential candidates. (i) Three different polymerases, the host Pol II as well as nuclear and cytoplasmic versions of T7 RNAP, each synthesized precursors which were substrates for such host factors (Fig. 4 and 8). (ii) For the latter two cases, expression was 100-fold higher than with the former. Thus, the putative factors may be present at abundant levels in both the nucleus and cytoplasm. (iii) In addition, by transiently denaturing the attenuator-ribozyme complex in vitro, we could overcome the constitutive inhibition of the ribozyme and obtain doubly cleaved products (Fig. 6B). It therefore seems plausible that host factors might similarly act to transiently destabilize base pairing of the attenuator with the ribozyme in the precursor RNA. There are numerous precedents for the proposed destabilization of RNA structures by host factors. These so-called RNA chaperones can initiate profound conformational changes in RNA and mediate the interconversion between alternative structures (23). The heterogeneous nuclear ribonucleoprotein (hnRNP) A1 is perhaps the best studied RNA chaperone. In vitro, this protein can promote the dissociation and reassociation of both double-stranded DNA and RNA (22), and such properties have even been exploited to stimulate the turnover rate for cleavage of a target molecule by a synthetic ribozyme (29). In vivo, hnRNP A1 is one of a family of proteins that interact with nascent Pol II transcripts to form hnRNP complexes (5) and have been shown to effect the processing of certain spliced messages (18). This very abundant protein predominantly localizes to the nucleus but is also observed at significant levels in the cytoplasm (5). Thus, the abundance and intracellular localization of hnRNP A1 are consistent with its possible role in HDV processing. A number of other proteins that possess similar RNA chaperone activity have been recently identified (23). Any one of these, or some presently unidentified factor, might, singularly or in concert, modulate the ribozyme-attenuator interaction. If RNA chaperones actually do function to destabilize the ribozyme-attenuator interaction, how could this be achieved specifically with the precursor and not with the circular product? A possible explanation is that RNA chaperones might bind to the precursor RNA but, after processing is complete, permanently dissociate from the product so that the ribozyme would remain attenuated. This model is consistent with a potential role for hnRNP proteins since, at least in relation to Pol II messages, these proteins similarly bind to precursors but ultimately dissociate from processed products (5). It is also consistent with the natural life cycle of HDV. The delta antigen is known to bind to processed HDV RNA and form an RNP structure (25). In such an RNP, the genome might no longer be accessible to RNA chaperones. Thus, although it is not required for the processing reactions to occur (13), the delta antigen could nevertheless assist in displacing putative RNA chaperones from the circular product during a natural infection. Further studies will be required both to identify the


1199

host factors that regulate HDV self-cleavage and to determine their precise mechanisms of action. ACKNOWLEDGMENTS D.W.L. was supported, in part, by fellowship 1F32 AI08637-01 from the National Institutes for Health. J.M.T. was supported by grants CA-06927, RR-05539, and AI-26522 from the National Institutes for Health and by an appropriation from the Commonwealth of Pennsylvania. We thank A. Skalka, B. Mason, C. Seeger, and R. Katz for their valuable critical reading of the manuscript.

ADDENDUM IN PROOF We have recently completed an analysis of the RNA processing of mutants 1 to 6 (Fig. 3) after expression by bacteriophage T7 RNA polymerase in yeast cells. The results obtained were similar to those obtained after expression in E. coli (Fig. 7). Although at least some doubly cleaved product, M, was observed with the two mutants that lack attenuator sequences, m4 and m6, no such product was detected with the remaining mutants that contained attenuators. Consistent with this finding, the circular species, Mc, was not observed with any of the six mutants. From such results we conclude that yeast cells lack a functional homolog to the factor(s) that activates HDV processing in mammalian cells. Moreover, even in the absence of attenuator sequences, the ribozymes of m4 and m6 processed less efficiently in yeast cells than in E. coli (Fig. 7) or in vitro (Fig. 6). Thus, yeast cells may contain RNA binding proteins that partially interfere with the HDV genomic ribozyme selfcleavage reaction. REFERENCES 1. Been, M. D. 1994. cis- and trans-acting ribozymes from a human pathogen, hepatitis delta virus. Trends Biochem. Sci. 19:251–256. 2. Berger, J., J. Hauber, R. Hauber, R. Geiger, and B. R. Cullen. 1988. Secreted alkaline phosphatase: a powerful new indicator of gene expression in eucaryotic cells. Gene 66:1–10. 3. Chen, P.-J., G. Kalpana, J. Goldberg, W. Mason, B. Werner, J. Gerin, and J. Taylor. 1986. Structure and replication of the genome of hepatitis d virus. Proc. Natl. Acad. Sci. USA 83:8774–8778. 4. Chomczynski, P., and N. Sacchi. 1987. Single-step method of RNA isolation by acid guanidinium-thiocyanate-phenol-chloroform extraction. Anal. Biochem. 162:156–159. 5. Dreyfuss, G., M. J. Matunis, S. Pinol-Roma, and C. G. Burd. 1993. hnRNP proteins and the biogenesis of mRNA. Annu. Rev. Biochem. 62:289–321. 6. Dunn, J. J., B. Krippl, K. E. Bernstein, H. Westphal, and F. W. Studier. 1988. Targeting bacteriophage T7 RNA polymerase to the mammalian cell nucleus. Gene 68:259–266. 7. Fu, T.-B., and J. Taylor. 1993. The RNAs of hepatitis delta virus are copied by RNA polymerase II in nuclear homogenates. J. Virol. 67:6965–6972. 8. Kiberstis, P. A., J. Haseloff, and D. Zimmern. 1985. 29 phosphomonoester, 39-59 phophodiester bond at a unique site in a circular viral RNA. EMBO J. 4:817–827. 9. Kumar, P. K., K. Taira, and S. Nishikawa. 1994. Chemical probing studies of variants of the genomic hepatitis delta virus ribozyme by primer extension analysis. Biochemistry 33:583–592. 10. Kuo, M. Y. P., J. Goldberg, L. Coates, W. Mason, J. Gerin, and J. Taylor. 1988. Molecular cloning of hepatitis delta virus RNA from an infected woodchuck liver: sequence, structure, and applications. J. Virol. 62:1855– 1861. 11. Kuo, M. Y.-P., L. Sharmeen, G. Dinter-Gottlieb, and J. Taylor. 1988. Characterization of self-cleaving RNA sequences on the genome and antigenome of human hepatitis delta virus. J. Virol. 62:4439–4444. 12. Lambowitz, A. M., and M. Belfort. 1993. Introns as mobile genetic elements. Annu. Rev. Biochem. 62:587–622. 13. Lazinski, D. W., and J. M. Taylor. 1994. Expression of hepatitis delta virus RNA deletions: cis and trans requirements for self-cleavage, ligation, and RNA packaging. J. Virol. 68:2879–2888. 14. Lazinski, D. W., and J. M. Taylor. 1994. Recent developments in hepatitis delta virus research. Adv. Virus Res. 43:187–231. 15. Lee, B.-S., H.-N. Wu, and T.-H. Huang. 1993. The catalytic domain of human hepatitis delta virus RNA. FEBS Lett. 324:296–300. 16. Macnaughton, T. B., E. J. Gowans, S. P. McNamara, and C. J. Burrell. 1991.

1200

17. 18. 19. 20. 21. 22.

23. 24.

LAZINSKI AND TAYLOR

Hepatitis delta antigen is necessary for access of hepatitis delta virus RNA to the cell transcriptional machinery but is not part of the transcriptional complex. Virology 184:387–390. Macnaughton, T. B., Y.-J. Wang, and M. M. C. Lai. 1993. Replication of hepatitis delta virus RNA: effect of mutations of the autocatalytic cleavage sites. J. Virol. 67:2228–2234. Mayeda, A., and A. R. Krainer. 1992. Regulation of alternative pre-mRNA splicing by hnRNP A1 and splicing factor SF2. Cell 68:365–375. Mohr, G., M. G. Caprara, Q. Guo, and A. M. Lambowitz. 1994. A tyrosyltRNA synthetase can function similarly to an RNA structure in the Tetrahymena ribozyme. Nature (London) 370:147–150. Perrotta, A. T., and M. D. Been. 1990. The self-cleaving domain from the genomic RNA of hepatitis delta virus: sequence requirements and the effects of denaturant. Nucleic Acids Res. 18:6821–6827. Perrotta, A. T., and M. D. Been. 1993. Assessment of disparate structural features in three models of the hepatitis delta virus ribozyme. Nucleic Acids Res. 21:3959–3965. Pontius, B. W., and P. Berg. 1990. Renaturation of complementary DNA strands mediated by purified mammalian heterogenous nuclear ribonucleoprotein A1 protein: implications for a mechanism for rapid molecular assembly. Proc. Natl. Acad. Sci. USA 87:8403–8407. Portman, D. S., and G. Dreyfuss. 1994. RNA annealing activities in HeLa nuclei. EMBO J. 13:213–221. Rosenstein, S. P., and M. D. Been. 1991. Self-cleavage of hepatitis delta virus

J. VIROL.

25. 26.

27. 28.

29. 30.

31.

32.

genomic strand RNA is enhanced under partially denaturing conditions. Biochemistry 29:8011–8016. Ryu, W.-S., H. J. Netter, M. Bayer, and J. Taylor. 1993. Ribonucleoprotein complexes of hepatitis delta virus. J. Virol. 67:3281–3287. Sharmeen, L., M. Y.-P. Kuo, and J. Taylor. 1989. Self-ligating RNA sequences on the antigenome of human hepatitis delta virus. J. Virol. 63:1428– 1430. Symons, R. H. 1994. Ribozymes. Curr. Opin. Struc. Biol. 4:322–330. Thill, G., M. Vasseur, and N. K. Tanner. 1993. Structural and sequence elements required for the self-cleaving activity of the hepatitis delta virus ribozyme. Biochemistry 32:4254–4262. Tsuchihashi, Z., M. Khosla, and D. Herschlag. 1993. Protein enhancement of hammerhead ribozyme catalysis. Science 262:99–102. Wang, K.-S., Q.-L. Choo, A. J. Weiner, J.-H. Ou, C. Najarian, R. M. Thayer, G. T. Mullenbach, K. J. Denniston, J. L. Gerin, and M. Houghton. 1986. Structure, sequence and expression of the hepatitis delta viral genome. Nature (London) 323:508–513. Wu, H.-N., and M. M. C. Lai. 1990. RNA conformational requirements of self-cleavage of hepatitis delta virus RNA. Mol. Cell. Biol. 10:5575– 5579. Zuker, M., and P. Stiegler. 1981. Optimal computer folding of large RNA sequences using thermodynamics and auxiliary information. Nucleic Acids Res. 9:133–148.