Stem-Loop III in the 5′ Untranslated Region Is a cis-Acting Element in Bovine Coronavirus Defective Interfering RNA Replication

Virus and cells.

A DI RNA-free stock of the Mebus strain of BCoV at 4.5 × 10⁸ PFU/ml was prepared and used as a helper virus as previously described (5-7, 53). The human rectal tumor cell line HRT-18 (49) was used as previously described (5-7, 16-19, 37, 53).

Plasmid constructs.

Construction of pGEM3Zf(−)-based pDrep1 (Promega) has been previously described (6). To construct the plasmid derivatives of pDrep1, overlap PCR mutagenesis (20) was done by using oligonucleotides listed in Table 1. The strategy used for constructing each of the derivatives was the same as that for pSLIII-mut1R except for the differences noted. For making pSLIII-mut1R, NdeI(+), SL3R(−), and pDrep1 DNA were used in the first PCR, HpaI(+), GEM3Zf(−)NdeI(−), and pDrep1 DNA were used in the second reaction, and GEM3Zf(−)NdeI(−) and NdeI(+) and the products of the first two reactions were used in the third reaction to make a 1,183-nt product that was trimmed to 693 nt with restriction endonucleases NgoMIV and AvrII and was cloned into NgoMIV/AvrII-linearized pDrep1. For making pSLIII-mut1L, SL3L(−) replaced SL3R(−) in the first reaction. For making pSLIII-mut1L/R, SL3L(−) replaced SL3R(−) and pSLIII-mut1R DNA replaced pDrep1 DNA in the first reaction, and pSLIII-mut1R DNA replaced pDrep1 DNA in the second reaction. For making pSLIII-mut2L, SRST3L(−) replaced SL3R(−) in the first reaction, and SRST3L(+) replaced HpaI(+) in the second reaction. For making pSLIII-mut2R, SRST3R(×) replaced SL3R(-) in the first reaction, and SRST3R(+) replaced HpaI(+) in the second reaction. For making pSLIII-mut2L/R, SRST3double(−) replaced SL3R(−) and pSLIII-2R DNA replaced pDrep1 DNA in the first reaction, and SRST3double(+) replaced HpaI(+) and pSLIII-mut2R DNA replaced pDrep1 DNA in the second reaction. For making pSLIII-mutG109→A, SRG109T(−) replaced SL3R(−) in the first reaction, and SRG109T(+) replaced HpaI(+) in the second reaction. For making pSLIII-mutG110→A, SRG110T(−) replaced SL3R(−) in the first reaction, and SRG110T(+) replaced HpaI(+) in the second reaction. For making pSLIII-mut5, SRAU99(−) replaced SL3R(−) in the first reaction, and SRAU99(+) replaced HpaI(+) in the second reaction. For making pSLIII-mut6, SR81(−) replaced SL3R(−) in the first reaction, and SR81(+) replaced HpaI(+) in the second reaction. For making pSLIII-mut3L, SRT101C(−) replaced SL3R(−) in the first reaction, and SRT101C(+) replaced HpaI(+) in the second reaction. For making pSLIII-mut4L/R, SRUTRSTOP(−) replaced SL3R(−) in the first reaction, and SRUTRSTOP(+) replaced HpaI(+) in the second reaction.

Enzyme structure probing of RNA.

Enzyme structure probing of RNA was carried out as previously described (37). Briefly, for in vitro synthesis of RNA, 10 μg of XbaI-linearized mung bean nuclease-blunt-ended pDrep1 DNA in a 200-μl reaction volume was transcribed at 37°C for 1 h by using 80 U of T7 RNA polymerase (Promega) to yield approximately 40 μg of a 595-nt-long vector-free transcript. RNA was treated with 10 U of RNase-free DNase (Promega) at 37°C for 30 min, extracted with phenol-chloroform and chloroform, chromatographed through a Biospin 6 column (Bio-Rad), spectrophotometrically quantitated, and stored in water at −20°C. For RNase treatments, 40 μg of RNA was heat denatured at 65°C for 3 min and renatured by slow cooling (0.5 h) to 35°C in a 400-μl reaction volume containing 30 mM Tris HCl (pH 7.5)-20 mM MgCl₂-300 mM KCl, and aliquots containing 2 μg of sample RNA and 5 μg of yeast tRNA were incubated in 100 μl of the same buffer and 0.05, 0.5, 1, 2, or 4 U of RNase CV1 (Pharmacia), 0.05, 0.5, 1, 5, or 10 U of RNase T₁ (GIBCO), or 0.05, 1, 2, 3, or 4 U of RNase T₂ (GIBCO). RNase digestion was carried out at 25°C for 15 min and was terminated by the addition of 150 μl of 0.5 M sodium acetate, after which the RNA was extracted with phenol-chloroform and chloroform and was ethanol precipitated. Digested RNA preparations were used in primer extension reactions with 5′-end-labeled plus-strand-binding oligonucleotide SRDM(+) (Fig. 1). Undigested RNA was used with the same primer in dideoxynucleotidyl sequencing reactions to generate a sequence marker. Products were analyzed on a DNA sequencing gel of 6% polyacrylamide.

Northern assay for DI RNA replication.

The Northern assay was performed essentially as previously described (5-7, 19, 53). Briefly, cells at 80% confluency (∼10⁶ cells per dish) in 35-mm-diameter dishes were infected with BCoV at a multiplicity of 10 PFU per cell and were transfected with 500 ng of DI RNA transcript in Lipofectin (Invitrogen) at 1 h postinfection (hpi). For passage of progeny virus, supernatant fluids were harvested at 48 hpi and 500 μl was used to infect freshly confluent cells in a 35-mm-diameter dish. RNA was harvested at the designated times and at 48 h after virus passages 1 and 2 (VP1 and -2, respectively). Total RNA extracted by the NP-40-proteinase K method (yielding approximately 10 μg per dish) was stored as an ethanol precipitate, and precisely one fourth of the preparation was used per lane for formaldehyde-agarose gel electrophoresis. Approximately 1 ng of transcript was loaded per lane as a size marker. Northern blots were probed with reporter-specific oligonucleotide porcine transmissible gastroenteritis virus (TGEV) 8(+) 5′-end labeled with ³²P to specific activities of 1.5 × 10⁵ to 3.5 × 10⁵ cpm/pmol (Cerenkov counts). Blots were read with a Packard InstantImager Autoradiography System for quantitation and were exposed to Kodak XAR-5 film for 6 h to 5 days at −80°C for imaging. Images were recorded by digital scanning of the film. In every experiment in which the replication of mutants of pDrep1 RNA was examined, wild-type (wt) pDrep1 RNA was examined in parallel.

Sequence analysis of the stem-loop III region in progeny DI RNAs.

For direct sequencing of asymmetrically amplified cDNA, the procedure described by Hofmann et al. (17) for analysis of plus-strand RNA was used. For this sequencing, oligonucleotides TGEV8b(+) and SRleader(−) were used for reverse transcription-PCR (RT-PCR) with RNA extracted at 48 hpi from cells infected with the VP1 and, in some cases, with VP2, and 5′end-radiolabeled oligonucleotide SRDM(+), HpaI(+), or SRleader(−) was used for sequencing. When mixed sequences were revealed by this method, RT-PCR fragments were cloned with the TOPO XL PCR cloning kit (Invitrogen) and were sequenced.

Synthetic oligonucleotides and accession numbers.

Oligonucleotides used in this study are described in Table 1, and GenBank accession numbers for the sequences studied are M62375, M16620, and M30612 for BCoV-Mebus, AF523843 for human coronavirus (HCoV)-OC43, AF523844 for human enteric coronavirus (HECoV)-4408, AF523845 for porcine hemagglutinating encephalitis virus (HEV)-67N, AF523846 for equine coronavirus (ECoV)-NC99, NC001846 for mouse hepatitis virus (MHV)-2, M27198 for MHV-A59, M55148 for MHV-JHM, 234093 for TGEV-Purdue 116, 69721 for HCoV-229E, and M42 for avian infectious bronchitis virus (IBV)-Beaudette.

A predicted stem-loop III with an associated short ORF in the BCoV 5′ UTR shows phylogenetic conservation among group 2 coronaviruses.

Earlier studies that used the Tinoco algorithm (48) had predicted a potential stem-loop III with a helix of 7 bp and a loop of 6 nt mapping at nt 97 through 116 in the 210-nt 5′ UTR of BCoV, although enzyme structure probing had indicated the stem was probably 8 bp and the loop was 4 nt in length (7). An AUG start codon for an intra-5′ UTR ORF of 24 nt was found within the stem beginning at base 100. In the same study, stem-loops I and II were identified as components of a required 5′-terminal cis-acting element for BCoV DI RNA replication. Structure predictions, however, have failed to show a consistent pattern for stem-loops I and II among the group 2 coronaviruses (7 and data not shown), suggesting that perhaps a specific sequence rather than higher-order structures in the 5′-terminal leader region is important for this cis-acting function. To further evaluate the higher-order nature of stem-loop III in BCoV, the genomic 5′-terminal sequences were analyzed by the mfold program of Zuker (http://bioinfo.math.rpi.edu/∼zukerm/) (32, 54), and a stem-loop with a stem of 8 bp, a loop of 4 nt, and a free energy value of −9.1 kcal/mol was predicted (Fig. 1A). In the minus strand, a stem-loop with an internal loop, a terminal hexaloop, and a free energy of −3.0 kcal/mol was also predicted (Fig. 2). The UGUG loop in the positive strand does not belong to any identified family of especially stable tetraloops (UNCG, GNRA, CUNG) (52), suggesting that it may not contribute extraordinarily to stem-loop stability.

5′ UTR sequences for 12 coronaviruses have now been reported, enabling comparisons of predicted higher-order structures (see references associated with the GenBank accession numbers given above). For the eight members in group 2, BCoV-Mebus, HCoV-OC43, HECoV-4408, HEV-67N, ECoV-NC99, MHV-2, MHV-A59, and MHV-JHM, an mfold-predicted stem-loop III with small variations was found (stem-loop III is identical in BCoV, HCoV-OC43, and HEV) (Fig. 1A and Table 2). That is, there was predicted (i) a stem-loop III in the positive strand with a stem (with or without an internal loop) of 7 or 8 nt, a terminal tetraloop, and a net negative free energy within the range of −11.5 to −5.1 kcal/mol, (ii) a stem-loop in the negative strand with a net negative free energy within the range of −9.3 to −3.0 kcal/mol, and (iii) an AUG start codon for an ORF of 24 nt beginning at the third or fourth base position in the stem of the positive strand. The stem begins at base 96 for ECoV, 97 for MHV-A59 and MHV-2, and 101 for MHV-JHM, and an internal loop occurs at the fourth base position in the stem for all MHV strains and at the fifth base in ECoV. Although there is no evidence for a conserved primary sequence among coronavirus groups 1 to 3 in the region of stem-loop III, mfold analyses of groups 1 and 3 revealed a possible stem-loop III homolog in the same relative position within the 5′ UTR (see Discussion). Interestingly, all possess a closely associated AUG start codon for a short intra-5′ UTR ORF (Table 2). Thus, from folding predictions and evidence of phylogenetic conservation it was postulated that stem-loop III in BCoV DI RNA is a higher-order structure with possible function in RNA replication.

Enzyme structure probing supports the existence of stem-loop III in BCoV RNA.

Previous enzyme structure probing of the 5′-terminal 262-nt region of the BCoV genome had revealed, along with stem-loops I (nt 11 through 42) and II (nt 51 through 84), a stem-loop III with an 8-base stem and a tetraloop (7). Since RNA secondary structures can be significantly influenced by flanking sequences, evidence for stem-loop III was reexamined in the context of a 595-nt transcript. For this, 5′-terminal transcripts of XbaI-linearized pDrep1 DNA were made and were subjected to digestion with either single-strand-specific RNase T₁ (data not shown) or T₂ or with double-strand-specific RNase CV1, and the products were analyzed by reverse transcriptase extensions of 5′-end-labeled primer SRDM(+) (Fig. 1B, lanes 1 to 4 and 9 to 13). Radiolabeled products of an RNA sequencing reaction were analyzed in parallel to determine sites of RNase digestion (Fig. 1B, lanes 5 to 8). The results show single- and double-stranded regions in agreement with an mfold-predicted structure (Fig. 1C).

Integrity of stem-loop III is required for BCoV DI RNA replication.

To examine the importance of the helical region in stem-loop III for DI RNA replication, site-directed mutations in pDrep1 predicted to disrupt and then restore base pairing with compensatory changes were made, and transcripts were tested for replication. A stem was deemed required for replication if helix-disrupting mutations led to loss of accumulation (as determined by the absence of mutant progeny molecules in VP1 RNA) and structure-restoring compensatory mutations led to some level of restored accumulation (as determined by the presence of the double mutant progeny in VP1 RNA). Analyses focused on DI RNA progeny in VP1, since (i) passaged (i.e., packaged) molecules should be free of long-lived transfected mutant transcripts that would otherwise appear as progeny in RT-PCR assays and (ii) Northern assays alone do not discriminate between replicating mutant and replicating (recombination-derived revertant) wt molecules. Here it was assumed that stem-loop III does not function as a packaging signal (see Discussion). In the first of two sets of mutants examined, pSLIII-mut1L and pSLIII-mut1R were made in which four of the eight base pairs were disrupted in each of the left and right arms leading to stem-loop III free energy values of −2.2 and −1.7 kcal/mol in the positive strands, respectively, and no predicted structures in the negative strands (Fig. 2). By Northern analysis, 6% or less of wt levels of RNA had accumulated by 24 hpi and less than 1% by 48 hpt and by 48 hpi in VP1, but sequence analysis of asymmetrically amplified cDNA from VP1 RNA showed the products of both mutants to be revertants to wt, and thus the mutants had not replicated. By contrast, accumulation of progeny from double mutant pSLIII-mut1L/R for which the free energy value had become −8.8 kcal/mol in the positive strand and −3.0 kcal/mol in the negative strand was 14% of wt levels in VP1, and asymmetrically amplified cDNA from progeny molecules in VP1 as well as in cDNA clones of these showed progeny to be a mixture of wt and double mutant sequences (Fig. 2). Of eight clones sequenced, two were wt and six were the SLIII-mut1L/R mutant. cDNA from VP2 also showed a mixture of wt and mutant sequences, indicating continued replication of the double mutant through a second virus passage. These results together indicate that the helical region of stem-loop III either in the positive or negative strand or both is a requirement for DI RNA replication. A lower rate of accumulation in the double mutant than in wt, however, suggested that factors other than the helical nature of the stem, perhaps the stability of the stem, the nucleotide sequence within the stem, the size and location of the stem-loop-associated ORF, or the ORF product, might also be important for maximal accumulation.

To test for the requirement of the stem under conditions that would minimally disrupt the short ORF, a second set of three mutants was examined in which the short ORF product differed from the wt by no more than three amino acids (Fig. 3). In pSLIII-mut2L, 5 of the 8 bases in the left arm were changed, leaving only those in the wt 100AUG codon unchanged and increasing the free energy value in the positive strand to +0.9 kcal/mol and leading to no predicted structure in the negative strand. The changes altered only the second amino acid (L→G), yielding a potential oligopeptide product of MGVGVDFS. In pSLIII-mut2R, 5 nt in the right arm were changed, maintaining the 111CGU base pairing partner with 100AUG and increasing the free energy in the positive strand to +2.3 kcal/mol and also leading to no predicted structure in the negative strand and potentially encoding an oligopeptide product of MLVSVLFS. In the double mutant pSLIII-mut2L/R, the calculated stem-loop III free energy was −10.0 kcal/mol in the positive strand and −3.2 kcal/mol in the negative strand, and the potential ORF product was MGVSVLFS. Northern analyses of progeny from pSLIII-mut2L and pSLIII-mut2R showed an accumulation of <1% of wt levels at 24 and 48 hpt and in VP1, and sequence analyses of asymmetrically amplified cDNAs showed progeny to be all revertant wt molecules, indicating that the mutants had not replicated. On the other hand, accumulation of progeny from the double mutant pSLIII-mut2L/R was >26% of wt levels at 24 and 48 hpt and >2% in VP1, and all progeny in VP1 had retained the mutant sequences. Thus, these results are consistent with those of the experiment described in Fig. 2 and suggest that stem-loop III either in the positive or negative strand or both is required for replication. A less-than-wt level of accumulation for the double mutant, however, would suggest again that perhaps the nucleotide sequence within the stem or possibly the amino acid sequence of the short ORF product were not optimal for maximal accumulation.

The notion that the nucleotide sequence within the stem or the amino acid sequence of the ORF product are important factors in accumulation in the context of an otherwise stable stem was supported by results from mutants in which single base changes were made in the upper part of the stem. In mutant pSLIII-mutG109→A, a stem of 8 nt was predicted for which the free energy of stem-loop III became −9.0 kcal/mol in the positive strand and −4.2 kcal/mol in the negative strand, and the potential product became MLVSVDFS; in pSLIII-mutG110→A, a stem of 7 nt was predicted for which the free energy of stem-loop III became −4.1 kcal/mole in the positive strand and −1.0 in the negative strand, and the potential ORF product became MLVDVDFS (Fig. 4A and B). For these, although only mutant sequences were found in VP1 DI RNA, accumulation was still less than maximal in that only 47 and 9% of wt levels, respectively, were found in VP1. Thus, to evaluate the role of stem-loop III integrity in mutants that contained nucleotide changes but still produced a potential wt short ORF product, double mutants were made by changing bases in codon wobble positions such that a wt ORF product was retained but that the stem was either maintained or disrupted. For these, pSLIII-mut5 was made in which stem III was maintained in both the positive (−8.71 kcal/mol) and negative (−2.6 kcal/mol) strands, and pSLIII-mut6 was made in which no structure was predicted in either strand of stem-loop III (Fig. 4C). Accumulation of pSLIII-mut5 progeny was no less than that of wt DI RNA at all time points, and the sequences of VP1 progeny were all mutant molecules, indicating that the stable stem mutants potentially producing a wt ORF product replicated at least as well as wt stem-loop III. On the other hand, accumulation of pSLIII-mut6 in which no stem-loop III structure was predicted was at less than wt levels, and only 15% of wt levels were produced in VP1 for which all progeny had wt sequence. These results together demonstrate that the integrity of stem-loop III is important for replication independent of any effect of the short ORF product and that at least some nucleotide changes in the stem are well tolerated.

Replication of BCoV DI RNA shows a strong preference for the presence of a stem-loop III-associated short ORF.

An association of an AUG-initiated short ORF with stem-loop III (or its presumed homolog) in all coronaviruses (see Discussion) led us to hypothesize that the short ORF plays a role as a cis-acting element in replication. To test whether it does, the DI RNA was modified in ways that were predicted to either diminish or prevent translation of the ORF but minimally perturb the integrity of stem-loop III. In the first modification, pSLIII-mut3L was made in which 100AUG was changed to 100AcG, a threonine codon known to function inefficiently as a translation start codon (38) (Fig. 5A). In this construct it was predicted by mfold that base pairing in the positive strand would be maintained between the AcG codon and the cognate wt triplet 111CGU and that the negative free energy of stem-loop III would increase to −10.9 kcal/mol in the positive strand and increase approximately threefold to −9.6 kcal/mol in the negative strand. By Northern analysis, accumulation of transcripts in VP1 was 18% of wt, and the sequence of the replicating progeny in VP1 was a mixture of both wt and mutant, indicating that some of these mutants had replicated. Progeny in VP2 were all wt. These results indicated that transcripts with an AcG-initiated short ORF can be replicated but with less efficiency than wt DI RNA and that an AUG-initiated ORF is preferred.

To test the importance of the short ORF by another approach, pSLIII-mut4L/R was made in which the 100AUG was changed to a 100uaG stop codon and its cognate base pairing triplet was changed from 111CGU to 111Cua to maintain full base pairing in the stem (Fig. 5B). In pSLIII-mut4L/R, the mfold-predicted stem-loop III had a free energy of −9.3 kcal/mol in the positive strand and −7.4 kcal/mol in the negative strand, and it had no predicted ORF product. Whereas Northern analysis showed an accumulation of <1% of wt after VP1, sequence analysis of asymmetrically amplified cDNA of progeny at VP1 revealed a mixture of 100A(U,a)G and 111C(G,u)(U,a). Thus, the cDNA products were cloned and sequenced to resolve the makeup of progeny molecules. Of 13 clones, 5 were wt sequence, 3 showed 100AUG but with a cognate base-pairing 111CGa triplet (potentially encoding the peptide MLVGEDFS), and 5 showed 100AaG (known not to function as a start codon [38]) with a cognate 111CGa triplet. No clones had the sequence of the original mutations of 100uaG or 111Cua. All progeny in VP2 were wt as determined from sequencing asymmetrically amplified cDNA. These results, therefore, established that molecules with the non-start codon AaG had replicated poorly and those with a uaG non-start codon had not replicated, indicating, as did the experimental results with mutant pSLIII-mut3L (Fig. 5A), that there is a preference by the DI RNA for a 100AUG-initiated short ORF in replicating molecules.