Role of an Essential Triloop Hairpin and Flanking Structures in the 3′ Untranslated Region of Alfalfa Mosaic Virus RNA in In Vitro Transcription

Abstract

The minus-strand promoter of Alfalfa mosaic virus (AMV), a tripartite plant virus belonging to the family Bromoviridae, is located within the 3′-terminal 145 nucleotides (nt), which can adopt a tRNA-like structure (TLS). This contrasts with the subgenomic promoter for RNA4 synthesis, which requires ∼40 nt and forms a single triloop hairpin. Detailed analysis of the minus-strand promoter now shows that a similar triloop hairpin, hairpin E (hpE), is crucial for minus-strand synthesis. The loop sequence of hpE appeared to not be essential for RNA synthesis, whereas the identity and base-pairing capability of bases below the triloop were indeed essential. Reducing the size of the bulge loop of hpE triggered transcription from an internal site similar to the process of subgenomic transcription. Similar effects were observed when deleting (part of) the TLS, suggesting that tertiary contacts between hpE and the TLS prevent internal initiation. The data indicate that the minus-strand promoter hpE and the subgenomic promoter hairpin are equivalent in binding the viral polymerase. We propose that the major role of the TLS is to enforce the initiation of transcription by polymerase at the very 3′ end of the genome.

The family Bromoviridae consists of five genera of plant viruses with a tripartite RNA genome. RNA1 and RNA2 encode the viral subunits of the replicase. RNA3 is dicistronic and codes for a movement protein that is required for cell-to-cell movement and a coat protein (CP) needed for cell-to-cell and long-distance transport. CP is translated from a subgenomic messenger, RNA4, that is coterminal with the 3′ 800 to 1,000 nucleotides (nt) of RNA3. Some members of the Bromoviridae family produce a second subgenomic RNA (sgRNA) which is derived from RNA2. sgRNA synthesis is thought to occur by internal transcription on the minus strand of RNA3 (or RNA2). It has been suggested that a hairpin structure is required for subgenomic promoter (sgp) activity in all Bromoviridae (10). This hypothesis was experimentally verified for three genera within this family (4, 8, 9).

Within this family, bromo- and cucumoviruses possess a tRNA-like structure (TLS) that can be charged with tyrosine whereas the RNAs of Alfalfa mosaic virus (AMV) and ilarviruses cannot be charged with an amino acid, although their 3′ ends were recently shown to adopt a putative TLS (13). The 3′ untranslated region (UTR) of the genomic RNAs of Olive latent virus type 2, the type species of the fifth genus, Oleavirus, can be folded into a TLS similar to that of the bromovirus RNAs (13). A unique property of the 3′ UTR of the Alfamovirus and Ilarvirus RNAs is their ability to adopt two mutually exclusive conformations: one recognized by the CP and one, a pseudoknotted conformation, recognized by the viral replicase. The latter conformer resembles the TLS of Brome mosaic virus (BMV) RNA and, for AMV, was shown to be essential for viral replication. The binding sites for CP in AMV RNAs have been characterized extensively in the past (reviewed in reference 2). The requirements for binding of the polymerase are less well defined: a minimal promoter element was delimited to the 3′-terminal 145 nt of RNA3 (18).

In the present study, we performed an extensive mutation analysis of the 3′ UTR of RNA3 to identify a putative polymerase binding site. We have identified a triloop hairpin, hairpin E (hpE), as the most crucial element in minus-strand synthesis in vitro. The TLS domain, although harboring the initiation site for replication, by itself showed no template activity, nor did it compete with the full 3′ UTR for the replicase. Interestingly, we observed that severe deletions in the TLS caused transcription to initiate at a position upstream of hpE. This mode of transcription is similar to the action of the sgp. This promoter region was recently characterized in our laboratory and shown to require the formation of a similar triloop hairpin (8). We propose that (i) the AMV subgenomic and minus-strand promoters are basically the same and that (ii) the role of the TLS is to ensure that transcription initiates at the very 3′ terminus.

MATERIALS AND METHODS

Construction of 5′ deletion mutants C, D, E, F, and G.

Oligonucleotides used in this study are listed in Table 1. PCR was performed on plasmid 3kWT, which contains a full-length copy of AMV RNA3 and has a KpnI restriction site engineered directly downstream of the stop codon of the CP gene (17). Construct G, referred to as WT in this paper, was generated by PCR with primers T7G and WT2. Likewise, to obtain construct F, primers T7F and WT2 were used, and so on. Conditions for amplification were as follows: 40 cycles at 94°C for 30 min, 45°C for 45 min, and 72°C for 1 min in a 50-μl volume containing 5 mM MgSO₄, 0.2 mM deoxynucleoside triphosphates, 500 nM of each primer, 5 ng of (plasmid) DNA, and 1 U of Vent DNA polymerase (New England Biolabs, Beverly, Mass.).

TABLE 1.

Oligonucleotides used in this study

Oligonucleotide	Sequencea (restriction site)
T7G	AATTTAATACGACTCACTATAGGGTACCCCATTAATTTGG (KpnI)
BIO358	CTACCTGCAGCATCCCTTAGGGGCATTC (PstI)
WT2	GCATCCCTTAGGGGCATTCAT
HP2DEL	GCATCCCTTAGGGGCATTAGCATTTATATATGTGCGTTAG
HP3REV	AGCAATATGAAGTCGATCCTATC
HP4REV	CATACCTTGACCTTAATCCACC
HP5REV	GCATTAAATGACTTTAGCATCCCAAAT
HP3DEL	GATCGACTTCATATTGCTATGCTCATGCAAAACTG
HP4DEL	GGGTGGATTAAGGTCAAGGTATTGCTTATATATGTGCTAACG
HP5DEL	GATGCTAAAGTCATTTAATGCTAAGGTATGAAGTCCTATTCG
T7C	AATTTAATACGACTCACTATAGGTTATATATGTGCTAACGC
T7D	AATTTAATACGACTCACTATAGGGTATGAAGTCCTATTCG
T7E	AATTTAATACGACTCACTATAGGTGACCTCCACTGGGTGGA
T7F	AATTTAATACGACTCACTATAGGGATGCTAAAGTCATTTAATGC
ENSI	GGATGCTAAAGTCATTTAATGCATGCATCCACTGGGTGGATTAATGCATAGGTATGAAGTCCTATTCG
ET1	CATGGTGTGGGTGGATTAATGCA
ET2	TTAATCCACCCACACCATGCATG
ET3	CATGGTGTGGCACCATTAATGCA
ET4	TTAATGGTGCCACACCATGCATG
ET5	CATCCACTGGCTGGATTAATGCA
ET6	TTAATCCAGCCAGTGGATGCATG
ET7	CATCCAGTGGCTGGATTAATGCA
ET8	TTAATCCAGCCACTGGATGCATG
EL1	CATCCACTTTGTGGATTAATGCA
EL2	TTAATCCACAAAGTGGATGCATG
EL3	CATCCACAATGTGGATTAATGCA
EL4	TTAATCCACATTGTGGATGCATG
EB0	GGACTTCATACCTTGACCTCCACCCAGTGGAGG
EB1	GGACTTCATACCTTGACCTTCCACCCAGTGGAGG
EB3	GGACTTCATACCTTGACCTTATCCACCCAGTGGAGG
EBFOR	AGGTCAAGGTATGAAGTCC
3L	GCACgatTtTaAAGCAATATGAAGTCGATCC
3R	GCATTAaAatcGTGCGTTAGCACATATATAAGC
4R	CGAATAGGACTTCATACCTTGACCTTAACCACCCAG
DEL13	GCATTCATGCAGTTTTGC
T7G1	GTAATACGACTCACTATAGGGTACCCCATTAATTTGGGATGtTAAAGTCATTTAATGCTGACCTCCACTGGGTGG (KpnI)
T7G2	GTAATACGACTCACTATAGGGTACCCCATTAATTTGGGATGCTAAAGTtATTTAATGCTGACCTCCACTGGGTGG (KpnI)
EREV	TTGACCTTAATCCACCCAGTGGAGGTC

^aT7 promoter sequences are shown in boldface, mutated bases are lowercase, and restriction sites are underlined.

Construction of hairpin deletion mutants.

ΔB was generated in a single PCR on 3kWT with primer combination T7G-HP2DEL. ΔC, ΔD, and ΔE were obtained as follows. First, a 5′ fragment was synthesized by PCR on 3kWT, using primer pairs T7G-HP3REV, T7G-HP4REV, and T7G-HP5REV, respectively. Second, a 3′ fragment was generated on 3kWT, using with primer pairs HP3DEL-BIO358, HP4DEL-BIO358, and HP5DEL-BIO358, respectively. Subsequently, corresponding 5′ and 3′ fragments were gel purified and PCR amplified using primers T7G and BIO358. The resulting fragments were cloned in KpnI/PstI-digested 3kWT and checked by DNA sequencing. Mutant 3kENSI was also obtained by this two-step PCR method. First, two separate PCR fragments were generated on 3kWT DNA with primer pairs T7G-HP5REV and ENSI-BIO358. The two fragments were fused in a second PCR with primers T7G and BIO358. The resulting fragment was cloned in KpnI/PstI-digested in 3kWT. Mutant plasmids 3kET1, 3kET2, 3kET3, 3kET4, 3kEL1, and 3kEL2 were obtained by cloning the following oligonucleotide pairs into SphI/NsiI-digested 3kENSI: ET1-ET2, ET3-ET4, ET5-ET6, ET7-ET8, EL1-EL2, and EL3-EL4, respectively. Mutants 3kEB0, 3kEB1, and 3kEB3 were synthesized via the overlapping PCR technique on 3kWT DNA. Primer pairs T7G-EB0, T7G-EB1, and T7G-EB3 were used to generate the respective 5′ fragments, and primer pair EBFOR-BIO358 was used for generation of the 3′ fragment. Products were cloned in KpnI/PstI-digested 3kWT. Templates for transcription by T7 RNA polymerase were obtained by PCR amplification of these mutant plasmids with primers T7G and WT2.

Construction of 3′ deletions.

The 3′ deletion mutants 3L, 3L/Dra, 3L/ET2, and 3L/NS2 were obtained by PCR on 3kWT, 3kΔD (17), 3kET2, and 3kNS2, respectively, with primers T7G and 3L. Plasmid 3kNS2 was obtained after religating NsiI-digested 3kENSI. Δ35 and Δ94 were obtained by PCR on 3kWT with primer pairs T7G-3R and T7G-4R, respectively. Mutants Δ13, EB3/13, EB1/13, and EB0/13 were synthesized by PCR with primers T7G and DEL13 on plasmids 3kWT, 3kEB3, 3kEB1, and 3kEB0, respectively. Constructs GFE, GFE1, and GFE2 were synthesized by PCR on 3kWT with primer pairs T7G-EREV, T7G1-EREV, and T7G2-EREV, respectively. EB3/100, EB1/100, and EB0/100 were synthesized by PCR on plasmids 3kEB3, 3kEB1, and 3kEB0, respectively, with primer combinations T7G-EB3, T7G-EB1, and T7G-EB0, respectively.

Synthesis of transcripts.

PCR fragments were purified by chloroform extraction, followed by precipitation with 1 volume of 5 M ammonium acetate and 2.5 volumes of isopropanol. After washing with 70% ethanol and drying, pellets were dissolved in 30 μl of sterile water. Approximately 200 ng of PCR fragment was used for in vitro transcription with T7 RNA polymerase (12). After transcription, reactions were DNase treated and phenol extracted; the RNA was precipitated twice from ammonium acetate-isopropanol at room temperature to remove unincorporated nucleotides and short oligonucleotides. Pellets were washed with 70% ethanol, dried, and finally dissolved in sterile water. Quantitation of RNAs was done by UV spectrometry. RNA templates were checked for integrity on ethidium bromide-stained 2 to 3% agarose gels. Note that this method does not discriminate between n+1 and n+2 products that may have been produced by T7 transcription. However, in all experiments with templates that had either 5′ or internal deletions, the 3′ ends were identical, meaning that the observed differences were due to the mutations introduced. In the case of 3′ deletions, differences in the level of internal initiation and not 3′-terminal initiation were measured.

RdRp assays.

In vitro transcriptions were carried out in 25-μl reactions containing 50 mM Tris-HCl (pH 8.0); 10 mM MgCl₂; 0.5 mM each of ATP, CTP, and GTP; 5 U of RNase inhibitor (Gibco-BRL); 2.5 μl of AMV RNA-dependent RNA polymerase (RdRp) isolated as described previously (15); 10 μCi of [α-³²P]UTP (800 Ci/mmol; ICN); and 10 pmol of template RNA (unless stated otherwise). After 10 min of incubation at room temperature, 2.5 μl of 10 mM UTP was added and the reaction was continued at 28°C for 30 min. The reactions were extracted with phenol-chloroform, and the RNA was precipitated from ammonium acetate-isopropanol. Pellets were dissolved in 10 μl of S1 buffer (50 mM sodium acetate [pH 4.6], 200 mM NaCl, 2 mM ZnSO₄) and incubated with 40 U of S1 nuclease at room temperature for 30 min. After addition of 5 μl of loading buffer (98% formamide, 0.2% bromophenol blue, 0.2% xylene cyanol), the samples were electrophoresed in 4 or 8% acrylamide (19:1, acrylamide:bisacrylamide) gels in 1× TBE (89 mM Tris, 89 mM boric acid, 2 mM EDTA). Gels were dried and exposed to X-ray film for a duration that allowed visualization of all relevant products. This sometimes led to artifacts such as band merging. Band quantitation, however, was done with a phosphorimager and Quantity One software (Bio-Rad) under conditions where band merging did not occur.

RESULTS

hpE is crucial for minus-strand synthesis.

Recently, it was demonstrated that the 3′ UTR of AMV RNAs can adopt an alternative conformation which resembles the TLS that is present at the 3′ end of bromo- and cucumovirus RNAs. The AMV TLS is required for replication of viral RNA in vitro and in vivo and is also recognized by the yeast CCA-adding enzyme (13). The precise elements in the 3′ UTR that are involved in the binding of viral RdRp have not been determined. To delimit the putative RdRp binding site, a set of nested deletions was constructed by removing complete hairpins one by one from the 5′ end of the 3′ UTR. The deletion endpoints of these mutants are indicated in Fig. 1. Transcript G (180 nt), which comprises the entire 3′ UTR of RNA3, will be referred to as WT throughout this paper. Figure 1 shows that removal of hpG (lane 2) or both hpF and hpG (lane 3) still yielded efficient templates which had 50 and 100% of the WT activity, respectively (transcript WT, lane 1). However, when the deletions included hpE (transcript D) or hpE and hpD (transcript C), template activity was completely lost (Fig. 1, lanes 4 and 5, respectively).

FIG. 1.

(Top) Secondary structure of the 3′ UTR of RNA3. Hairpins are labeled from B to G. A′ indicates the pseudoknot that is formed between loop D and stem A (13). The 5′ endpoints of the various 3′-coterminal deletion mutants are indicated by arrowheads. In the final transcripts, three additional G residues are present at the 5′ ends of mutants G, F, and D and two are present at the 5′ ends of mutants E and C. (Bottom) Effect of 5′ truncations of the 3′ UTR on minus-strand synthesis in vitro. Templates (5 pmol each) were as follows: lane 1, WT; lane 2, F; lane 3, E; lane 4, D; lane 5, C. Lanes 6 through 13 show competition assays. Prior to the addition of AMV polymerase, WT template (5 pmol) was mixed with 5 and 25 pmol of F (lanes 7 and 6, respectively), 5 and 25 pmol of E (lanes 9 and 8, respectively), 5 and 25 pmol of D (lanes 11 and 10, respectively), and 5 and 25 pmol of C (lanes 13 and 12, respectively). Reactions were carried out in the presence of [³²P]UTP, and the products were separated by electrophoresis in a 4% polyacrylamide gel. The position of the WT 180-bp product is indicated by the arrowhead. An autoradiogram of the dried gel is shown.

Competition assays showed that mutants C and D were also incapable of competing with WT template for AMV RdRp. When added at a molar excess that was 5-fold (Fig. 1, lanes 10 and 12) or even up to 15-fold (data not shown), transcripts C and D had no detrimental effect whatsoever on WT template activity. In fact, a slight increase of WT template activity was observed: 140% in the case of transcript C and 125% in the case of transcript D. In contrast, transcripts E and F efficiently competed with WT: at a fivefold molar excess, they reduced transcription from the WT template to 27% (lane 8) and 29% (lane 6), respectively. These data indicate that the 3′-terminal 112 nt (transcript D) are not sufficient for template activity or for binding of the viral polymerase and that the presence of hpE is essential for minus-strand synthesis.

A similar conclusion was reached in experiments involving individual hairpin deletion mutants (Fig. 2, top). Deleting hpB, except for the bottom base pair, resulted in a template that retained 37% of WT activity (Fig. 2, lane 2). Also, mutants ΔC and ΔD still retained 80 and 53%, respectively, of transcriptional activity compared to that of WT (lanes 3 and 4). However, no products could be detected with mutant ΔE (lane 5), illustrating once more the crucial role of hpE in minus-strand synthesis.

FIG. 2.

(Top) Position of single hairpin deletions in the 3′ UTR. Deleted regions are indicated by stippled boxes. (Bottom) Autoradiogram showing the template activity of WT (lane 1) and of the hairpin deletion mutants ΔB, ΔC, ΔD, and ΔE. Lanes 6 through 9 are competition assays. Prior to the addition of AMV polymerase, WT template (10 pmol) was mixed in a 1-to-1 ratio with ΔB, ΔC, ΔD, and ΔE. The relative activity levels, which are indicated below each lane number, are the averages from at least two independent experiments. The standard deviation is shown below the average value. For further details, see the legend to Fig. 1.

Although mutants ΔB, ΔC, and especially ΔD, which lacks the pseudoknot structure that was previously shown to be important for minus-strand synthesis, retained substantial template activity, they could not compete efficiently with a WT template. When ΔB, ΔC, and ΔD were mixed at a 1-to-1 ratio with WT, their transcription went down to 10, 14, and 9%, respectively, of WT (Fig. 2, lanes 6 through 8). We speculate that these deletion mutants, although possessing an RdRp binding site (hpE), are less efficient in directing RdRp to its initiation site due to a distorted tertiary structure. We note that the activity of WT in these competition assays increased about 1.4-fold, as was also observed with transcripts C and D (see above). In the presence of ΔE, the activity of WT was neither decreased nor increased (Fig. 2, compare lanes 1 and 9).

Sequence and structure requirements for hpE.

To test the role of several residues in hpE on minus-strand synthesis, we first introduced a few restriction sites into cDNA3 at positions that correspond with the lower part of stem E. The resulting mutant, ENSI, differs from the wild-type stem by having four different base pairs (Fig. 3). The template activity of this mutant was 48% of that of the WT template (Fig. 3, lanes 1 and 2). This reduced activity could have resulted from the lower stability of the stem due to its lower G+C content. The next set of mutations was constructed in ENSI. Disruption of the top 4 bp (mutant ET1) abolished minus-strand synthesis completely (Fig. 3, lane 3). However, restoring base pairing (ET2) by making compensating changes in the top of ET1 did not restore its template activity (Fig. 3, lane 4). These two mutants were also incapable of competing with WT (data not shown). This suggests that the identity of one or more base pairs in this region is essential for binding the replicase or contributes to the positioning of functional base moieties involved in binding. Opening up the upper base pair (ET3) resulted in a sixfold reduction in transcription (Fig. 3, lane 6); restoring base pairing (ET4) restored transcription to 75% of ENSI (lane 5). This suggests that the loss of template activity of mutant ET2 is not due to inversion of the loop closing the C·G base pair but probably to changing of the identity of the other 3 bp (see also Discussion).

FIG. 3.

Mutations in hpE and their effect on minus-strand synthesis in vitro. Mutations were made in the full-length 3′ UTR (180 nt). In mutant ENSI, base changes with respect to WT are indicated in boldface. Mutants ET1 through ET4, EL1, and EL2 were made in the ENSI construct. Only the relevant regions are shown; base changes with respect to ENSI are printed in bold. For further details, see the legend to Fig. 2.

Next, we investigated the role of the triloop sequence in transcription. Converting the UGG loop sequence in the ENSI construct to UUU (EL1), AAU (EL2), or AAA (data not shown) had no dramatic effect on minus-strand synthesis (Fig. 3, lanes 7 and 8). Deletion of the UGG sequence from the loop in the WT 3′ UTR (ΔDRA) reduced transcription to a mere 3% (reference 18 and data not shown). This mutation likely disrupts the entire upper part of the helix as in ET1.

Comparison of the bulge loop of hpE between different isolates of alfamovirus RNAs revealed that its size is invariably 4 nt, whereas its sequence may differ (R. C. L. Olsthoorn, unpublished data). Reducing the bulge loop size to 3 nt (EB3) in the WT construct did not strongly affect RNA synthesis (Fig. 4, compare lanes 1 and 2). More dramatic effects were observed when the bulge loop was replaced with a single A-bulge (EB1) or when it was deleted entirely (EB0). Synthesis of the full-length RNA was reduced to less than 10% of the WT level (Fig. 4, lanes 3 and 4). Several shorter-than-full-length products were visible, most notably, two products with estimated sizes of 25 and 30 bp. (Note that all products on the gel are double stranded, so their sizes are indicated in base pairs.) Their total relative accumulation amounted to approximately 180% (EB1) and 290% (EB0) of the full-length product directed by the WT template. Whether the shorter products were really derived from internal initiation or merely from premature termination is examined below.

FIG. 4.

Role of bulge size of hpE on minus-strand synthesis in vitro. Mutations were introduced into the full-length 3′ UTR (180 nt). Products were separated on a native 8% acrylamide gel. Sizes of the two short products in lanes 3 and 4 were estimated by comparison of their migration distance to that of bromophenol blue. Internal product levels were calculated as described in the legend to Fig. 8. For further details, see the legend to Fig. 2.

Cryptic initiation sites.

To check whether short RNAs were also generated with other mutants, their products were reexamined by 8% polyacrylamide gel electrophoresis. For ΔB, ΔC, and ΔD, this revealed the presence of two faster-migrating products with estimated sizes of 25 and 30 bp that were similar to those obtained with EB0 and EB1 (Fig. 5, lanes 2 through 4). Products of this size were also observed with mutant 3L, which lacks the 3′-terminal 55 bases (Fig. 5, lane 5), but were virtually absent in reactions with a WT template (lane 1). The same doublet was also produced in reactions with templates lacking the 3′-terminal 94 or 106 nt (data not shown). Since the migration of the doublet appeared to be independent of the size of the 3′ truncation, the doublet must originate from internal transcription upstream of nt 107. The fact that the same two fragments were produced by completely different sets of mutants argues against the presence of potentially contaminating short templates that might be responsible for these products (see also Materials and Methods).

FIG. 5.

(Left) Deletions in and truncations of the TLS activate cryptic initiation. Arrows on the left of the 8% polyacrylamide gels indicate the position and size of the 180-bp product obtained with WT template, a likely 125-bp product obtained from 3′-end transcription on 3L template, and cryptic initiation products of approximately 25 and 30 bp. (Right) The structure of mutant 3L. Mutated bases are underlined. Hooked arrows show putative transcription start sites. Initiation at residues C141, C150, and C157 would result in products of 40, 31, and 24 bp, respectively.

We note that 3L also yields a product of about 125 bp which could have originated from transcription initiation 3′ of hpE (see arrow in Fig. 5). This type of initiation is independent of the presence of hpE and emerges only when long single-stranded regions are available in the template (Olsthoorn, unpublished). This phenomenon has been studied previously by other laboratories using different plant viral RdRps (5, 19).

Mapping internal initiation sites.

We envisaged three possible initiation sites 5′ of hpE that could give rise to products of 25 to 30 bp: C141 (numbering starting at the 3′ end), C150, and C157, resulting in products of 40, 31, and 24 nt, respectively (Fig. 5, 3L). C141 could be ruled out on the basis of the following experiment. Deleting hpF, including C150, and the 5′ AUGC motif, including C157, completely abolished internal initiation (data not shown). If C141 were used as the start site, a product of 24 nt would be synthesized by the hpF deletion mutant.

The other potential transcription start sites, C157 and C150, were investigated by mutating them separately into U in a GFE construct (Fig. 6, left). GFE gave rise to the expected doublet (Fig. 6, lane 3). Mutation of C157 to U (GFE1) reduced synthesis of both products approximately threefold (lane 1); however, mutation of C150 to U (GFE2) abolished synthesis of both products (lane 2). This means that C157 is not used as an initiation site and that initiation at C150 results in two products, the smaller of which may be a premature termination product of the longer one. Premature termination could be due to formation of the complement of hpG in the nascent strand. In combination with the low stability of the duplex between the template and nascent strand in this region, this might facilitate strand release analogous to rho-independent transcription termination.

FIG. 6.

Mapping of the internal initiation site 5′ of hpE. (Left) Structure of template GFE corresponding to the 5′ 68 nt of the 3′ UTR. The GFE1 mutation of C157 to U and the GFE2 mutation of C150 to U are shown. (Right) Products obtained by in vitro transcription with AMV RdRp of GFE1, GFE2, and GFE were electrophoresed in an 8% acrylamide gel. The two products migrated at the same position as the short products obtained with ΔB, ΔC, ΔD and 3L (Fig. 5) and EB0 and EB1 (Fig. 4). The template activity of GFE was taken as 100%. For further details, see the legend to Fig. 2.

Internal initiation requires a functional hpE.

Shorter products were not observed with ΔE as a template (not shown), indicating that internal initiation requires the presence of hpE. To verify this, mutations were made in hpE in the 3L construct. Deletion of the UGG motif from the loop of hpE (3L/Dra), a mutation that reduced minus-strand synthesis from a WT template to 3%, strongly reduced internal initiation in the 3L context without interfering with the terminal transcription (Fig. 7, compare lanes 1 and 4). Likewise, reversing the top 4 bp (3L/ET2), which was shown above to abolish transcription in the full-length 3′ UTR, silenced internal but not terminal transcription (Fig. 7, lane 2). Finally, removing the AUGC sequence between hpE and hpF (3L/NS2) even enhanced internal initiation (lane 3). This shows once more that C141 is not used as a start site and that the distance between hpE and the initiation site may vary. Thus, internal initiation is not an unspecific reaction but rather relies on specific sequence and/or structure elements in hpE.

FIG. 7.

Analysis of the role of hpE in internal initiation. (Left) Structure of the 3L construct and three mutants thereof. 3L/ET2 and 3L/NS2 are derived from mutant ENSI and thus also have different bottom base pairs (indicated by lightface lettering). The internal transcription start site (int) is marked by the solid hooked arrow; the terminal transcription start site (3′) is marked by the stippled hooked arrow. (Right) Autoradiogram of an 8% acrylamide gel showing products of in vitro transcription reactions with AMV RdRp and templates 3L, 3L/ET2, 3L/NS2, and 3L/Dra. int, products originating from internal transcription at C150; 3′, terminal transcription products. The template activity of 3L was taken as 100%. For further details, see the legend to Fig. 2.

When is internal initiation triggered?

The above experiments showed that internal initiation was triggered by deletions in the bulge of hpE or the TLS domain. We assumed that both the bulge and elements from the TLS contributed to the tertiary structure of the minus-strand promoter. Therefore, distortions in the three-dimensional structure rather than inactivation of the 3′-end initiation would be expected to activate internal transcription. Indeed, mutation of the terminal AUGC to AUAA, though abolishing 3′-end initiation on the full-length 3′ UTR minus-strand template, was not sufficient to trigger internal initiation (data not shown). Deletion of the terminal 13 nt (Fig. 8A) also did not activate internal transcription (Fig. 8B, lane 3) even though this mutation prevented pseudoknot formation. When the deletion was extended to 35 nt (Fig. 8A), internal transcription was eventually triggered (Fig. 8B, lane 2). The total relative accumulation of these products was about 130% compared to that of the full-length product obtained with WT (Fig. 8B, lane 1). This percentage did not change when the deletion was extended to 55 nt (mutant 3L [Fig. 5, lane 5]). Deletion of 94 nt (Fig. 8A) or 100 nt (data not shown), i.e., almost the entire TLS, raised internal transcription to 365% (Fig. 8C, WTΔ100). These data suggest that several elements in the TLS domain contribute to the suppression of internal initiation.

FIG. 8.

(A) Position of various 3′ truncations of the 3′ UTR and their effect on internal initiation (see panel C). The Δ13 and Δ100 deletions were also made in combination with mutations in the bulge of hpE: EB3/13, EB1/13, and EB0/13 and EB3/100, EB1/100, and EB0/100, respectively. Flags denote the 3′ ends of the mutant RNAs. For clarity, the linear conformation of the 3′ UTR is shown. (B) Autoradiogram of an 8% acrylamide gel showing products obtained by in vitro transcription with AMV RdRp and templates WT, Δ35, Δ13, and EB3/13. int, products originating from internal transcription at C150; 3′, terminal transcription products. (C) Percentage of internal initiation as a function of bulge size of hpE and downstream sequences. The bulge mutations were engineered in templates WT (full-length 3′ UTR), Δ13, and Δ100. Percentages were calculated by measuring the incorporation of radioactive UTP in the ∼26- and 31-bp products and correcting for the number of A residues present in the template upstream of C150. This number is 8 for the 31-bp product and presumably 7 for the product of ∼26 bp. Note that this uncertainty does not influence the ratios of internal initiation among the 12 constructs themselves but does introduce a small but constant error in all percentages relative to the full-length product (180 bp) synthesized on the WT template (taken as 100%). The mean value obtained from at least two experiments was plotted. Error bars indicate the standard deviations.

The contribution of the bulge of hpE became apparent in the following set of mutations. The 13-nt deletion was sufficient to trigger internal transcription with mutant EB3, which has a single A deletion in the bulge of hpE (Fig. 8B, lane 4). As indicated in Fig. 8C, the 13-nt deletion activated internal initiation more than threefold compared with the full-length EB3 template. This suggests that the pseudoknot—disrupted by the 13-nt deletion—does play a role in suppressing internal initiation. Further deletion of the entire TLS (Δ100) raised internal initiation to 290% for the EB3 mutant (Fig. 8C). The results with EB1 and EB0 mutants were somewhat different. These mutants already showed high levels of internal initiation of 180 and 290%, respectively, on the full-length 3′ UTR template (Fig. 4, lanes 3 and 4). For the EB1 mutant, the 13-nt deletion slightly increased internal transcription, which was further enhanced by removal of the TLS (Fig. 8C, EB1Δ100). For mutant EB0, deletions of 13 nt or more did not significantly change the level of internal initiation, which varied around 300% (Fig. 8C).

Thus, mutations in the bulge of hpE do affect internal initiation when introduced in the WT template but not when introduced in the Δ100 template. This Δ100 template, which lacks the entire TLS, shows high levels of internal initiation even when the bulge is intact. From this, we can conclude that the bulge of hpE is not a recognition site per se for the RdRp but rather is likely involved in maintaining the right orientation of hpE with respect to the 3′ end via interactions with the TLS.

DISCUSSION

By deletion analysis of the 3′ UTR of AMV RNA3, we have identified a single stem-loop structure, hpE, whose presence is absolutely required to allow minus-strand synthesis in vitro. hpE is contained within the 3′-terminal 145 nt that are common between RNA1, RNA2, and RNA3. The 145-nt region was previously shown to constitute the core promoter sequence for minus-strand synthesis, whereas the upstream sequence in the 3′ UTR of RNA3, hpF and hpG, acted as an enhancer for replication in vitro and in vivo (18). In these experiments, mutations were introduced into the full-length RNA3. It is possible that hpF and hpG serve as spacer elements to allow binding of the RdRp to or proper folding of hpE in the presence of potentially interfering upstream sequences. In our transcripts, which lack the upstream 1,962 nt, hpF and hpG have apparently become superfluous since the template activity of a transcript lacking these hairpins was not affected.

The structure of hpE is conserved in all AMV isolates sequenced so far and features a 10-bp stem, a 4-nt bulge, and a UGG triloop. We have shown here that the UGG sequence is not essential for in vitro transcription and could be mutated to AAU, UUU, or AAA without significant loss in activity. This is rather surprising considering the strong conservation of the UGG loop sequence. It is likely that, in vivo, the loop sequence does play a role since the AAA mutant was not infectious to tobacco plants (Olsthoorn, unpublished). In addition, we have demonstrated that disruption, but also reversal of the upper 4 bp, was detrimental for transcription whereas replacing 4 of the 6 lower bp still allowed about 40% transcription. The identity of base pairs below the bulge therefore appears less important than that of base pairs above the bulge. This is partly reflected in the sequences of RNA1 and RNA2, which have the same 4 upper bp as RNA3 but both differ in 2 bp in the lower part of hpE. Although the effect of the current mutations has not been assessed in vivo, the data altogether suggest that bases below, rather than in, the triloop are crucial for recognition by the AMV RdRp.

A similar stem-loop structure can be found at the homologous position in the 3′ UTR of the RNAs of ilarviruses that are closely related to AMV. In Prune dwarf virus (PDV) RNA1 and RNA3, a triloop hairpin that strongly resembles hpE, having in common 8 out of 10 bp, can be folded (Fig. 9B). In Apple mosaic virus (ApMV) RNA3 and Prunus necrotic ringspot virus (PNRSV) RNA3, a pentaloop hairpin is present at this location in the 3′ UTR (Fig. 9C and D). The PNRSV 3′ UTR was recently shown to be recognized by the AMV RdRp in vitro and, when substituted for the 3′ UTR of AMV RNA3, supported replication of this chimeric RNA3 in tobacco plants (1). Interestingly, the 4 bp below the pentaloop of PNRSV are identical to those in hpE of AMV. Also, in PDV and ApMV RNAs, the top 3 to 4 bp are strictly conserved (Fig. 9). Possibly, the identity or structure of these 3 to 4 bp is the primary recognition site for AMV RdRp and likely for PNRSV, ApMV, and PDV RdRps as well. This would explain why mutant ET2, in which these base pairs were mirrored, is not active in transcription.

FIG. 9.

Comparison of hpE with similar hairpins in the 3′ UTR of the RNA3 of viruses from the family Bromoviridae. (A) AMV; (B) PDV (GenBank accession no. L28145); (C) ApMV (GenBank accession no. U15608); (D) PNRSV strain PV32 (GenBank accession no. Y07568); (E) BMV (GenBank accession no. V00099); (F) CMV (GenBank accession no. D10538). Structures of PDV, ApMV, and PNRSV are supported by phylogenetic data (Olsthoorn, unpublished). BMV and CMV hairpins were taken from reference 16.

Recently, similar structures required for minus-strand synthesis have been identified in the 3′ UTR of BMV (3) and Cucumber mosaic virus (CMV) RNAs (16). Stem-loop structure C (SLC) of BMV consists of 11 bp, a triloop, and a 4-nt bulge, thereby resembling the AMV hpE (Fig. 9E). SLC of several isolates of CMV consists of 13 bp, a 5-nt bulge, and a triloop (Fig. 9F), but variants with a pentaloop do also exist (16). Although the BMV replicase can recognize the SLC of CMV strain Fny and vice versa (16), there is little sequence similarity between the two hairpins. The only conserved sequence between BMV and CMV stem-loop structures is the top C·G base pair and the 5′ A of the loop. Interestingly, it has been shown that the BMV RdRp specifically recognizes the 5′-most A in the AUA triloop which is involved in a so-called clamped adenine motif (11). This contrasts with the AMV RdRp, which appears insensitive to the loop sequence but probably recognizes specific base pairs in the stem. It is intriguing that related viruses have evolved different strategies to recognize their minus-strand promoter hairpins.

hpE shares many features with a recently identified structure that was shown to be necessary for sgRNA synthesis by the same AMV RdRp (Fig. 10, left). The sgp hairpin also consists of a 10-bp stem interrupted by a 3′ bulge and features a trinucleotide loop (8). The orientation of both structures seems to be fundamentally different: the sgp hairpin is located 3′ of the initiating nucleotide whereas the minus-strand promoter hairpin is located 5′ of the initiation site. So what determines where the replicase will start? Are subtle differences in these two hairpins responsible for positioning the replicase, or do flanking structures dictate the orientation of the replicase or the initiation site?

FIG. 10.

Comparison of the triloop hairpins involved in subgenomic and minus-strand RNA synthesis. The hooked arrows indicate the transcription initiation sites. The TLS is schematically shown as a cloverleaf.

One could envisage that base composition of the loop or bulge size is important in this respect. We have seen above that loop mutants AAA, UUU, and AAU, the latter one representing the loop sequence of the sgp hairpin, did not direct initiation to sites upstream of hpE. Interestingly, reducing the bulge loop size to 1 or 0 nt could trigger the production of short RNAs that were shown to originate from transcription at C150 located upstream of hpE. This effect is not simply due to a higher semblance to the sgp hairpin. In the absence of the TLS, wild-type and bulge mutants of hpE directed similar levels of internal initiation (Fig. 8C). We postulate that tertiary interactions between hpE and the TLS are required for efficient minus-strand synthesis and that disturbance of these interactions results in the activation of cryptic initiation sites. The bulge loop either may be directly involved in such interactions or may lend sufficient flexibility to hpE to make these contacts.

An attractive candidate for interaction with hpE is the bulge loop (5′ AUCG 3′) of hpD (Fig. 1). Previously, it was proposed that the latter loop may be the equivalent of the T loop of tRNAs (13). The T loop-DHU loop interaction gives tRNA its typical L-shape and has also been proposed to occur in the TLS of Tobacco mosaic virus RNA (7). A putative DHU loop equivalent in AMV RNA could be the bulge loop of hpE. In this regard, it is interesting to note that sequence differences in the 3′ UTR of AMV RNA3 isolates are usually restricted to the bulge loops of hpE and hpD, as if a mutation in one loop is compensated by a mutation in the other. However, so far, we have not been able to demonstrate an interaction between these bulge loops.

The AMV TLS cannot be aminoacylated (reference 13 and references therein), and its presence may have another reason, i.e., enforcing the RdRp to initiate at the very 3′ ends of genomic RNAs. It is has been suggested that the aminoacylatable TLSs of BMV and Tobacco mosaic virus play a similar role in positioning the 3′ end close to the catalytically active site of the RdRp (6, 14). In accordance with a structural function of TLSs, aminoacylation is not a prerequisite for promoter activity but possibly a side effect.