The RNA structures engaged in replication and transcription of the A59 strain of mouse hepatitis virus

Coronaviruses belong to the Order Nidovirales, whose members utilize a unique discontinuous RNA synthetic process to produce a 3' co-terminal, nested set of mRNA molecules, each with the same 5' leader sequence that is found only at the 5' end of the genome (Snijder & Meulenberg, 1998 ). Discontinuous transcription, rather than splicing, most likely is the mechanism responsible for generating the subgenomic mRNA (Spaan et al., 1988 ). The proposal that discontinuous transcription occurred during positive-strand synthesis, the leader-primed transcription model, was based on finding only genome-length negative strands and double-stranded cores (RF) of the viral replicative intermediates (RI) in cells infected with the A59 strain of mouse hepatitis virus (MHV) (Baric et al., 1983 ; Lai et al., 1982b ). The RF cores were stated as having an unusual migration during gel electrophoresis because they migrated faster than the genome. Subsequently, we found transcriptionally active RNA intermediates containing a complementary (negative strand) copy of each of the subgenomic mRNAs in coronavirus-infected cells (Sawicki et al., 1990 ; Sawicki & Sawicki, 1995 ). The synthesis in coronavirus-infected cells of negative-strand templates corresponding to the subgenomic mRNAs was first reported by David Brians laboratory (Sethna et al., 1989 ) with swine transmissible gastroenteritis virus. Others have also demonstrated that MHV A59 (Baric & Yount, 2000 ; Schaad & Baric, 1994 ) and equine arteritis virus (EAV) (den Boon et al., 1996 ), which is a member of the Arteriviridae, produced a series of RF cores following RNase treatment of infected cell RNA.

We proposed (Sawicki & Sawicki, 1995 ) a different model, in which the discontinuous step occurred during negative-strand synthesis. This model also proposed that there were two classes of RNA intermediates in MHV-infected cells. One had genome-length templates engaged in replicating the genome or in the production of subgenome-length negative-strand templates, i.e. anti-subgenomes. The other had anti-subgenomes being transcribed into subgenomic mRNA. Because the RNA intermediates active in transcribing subgenomic mRNA are not replicating, we call them TIs and TFs, for transcriptive intermediate and form RNA, respectively. We reserve the terms RI and RF for viral RNA involved in replication.

RI/TIs and RF/TFs would differ in their relative proportion of single-stranded and double-stranded character. Native RF/TFs would be completely or nearly completely double-stranded as a result of having only one or a few polymerases engaged in transcription on each template. Thus, positive and negative strands in native RF/TFs would be mostly equal in length. Native RFs are soluble in high salt solutions such as 2 M LiCl or 1 M NaCl, as are tRNA and DNA (Ammann et al., 1964 ; Montagnier & Sanders, 1963 ). RF gives a defined T_m with a sharp melting point (Ammann et al., 1964 ; Bishop & Koch, 1967 ). The biological significance of native RFs in poliovirus and RNA phages includes functions as short-lived intermediates and as end-products that accumulate when RNA synthesis ceases and the last positive strand is not released (reviewed in Koch & Koch, 1985 ). On the other hand, RI was a multi-stranded intermediate active in positive-strand synthesis. Poliovirus RI contained an average of three, or as many as eight to ten, nascent, single-stranded RNA tails (reviewed in Koch & Koch, 1985 ; Richards et al., 1984 ) and one poly(A) sequence (Yogo & Wimmer, 1975 ). This high degree of single-strandedness makes the RI insoluble in high salt solutions, as are messenger and ribosomal RNA (Baltimore, 1966 ; Bishop et al., 1969 ; Erikson & Gordon, 1966 ; Fenwick et al., 1964 ). Digestion of RI with controlled levels of RNase degraded the single-stranded regions composed of nascent chains and left an RF core now soluble in 2 M LiCl (reviewed in Koch & Koch, 1985 ). While in vivo RI was shown to be predominantly single-stranded (Richards et al., 1984 ), isolated poliovirus RF cores contained an intact template strand and short to almost full-length nascent chains base-paired along the length of the template (Baltimore, 1968 ). Inhibiting transcription initiation converted poliovirus RI to RF cores.

We undertook an analysis of RNA structures formed in MHV-infected cells to identify the number and abundance of RI/TIs and native RF/TFs and to develop methods to purify individual RI/TIs. A second goal was to attempt to explain the initial failure to find TF cores after RNase treatment and gel filtration chromatography of MHV RI/TIs.

Virus and cells.
Seventeen clone 1 (17Cl-1) mouse fibroblast cells and the A59 strain of MHV (Sturman & Takemoto, 1972 ) were originally provided by L. Sturman (Wadsworth Center, Albany, NY, USA). 17Cl-1 cells were grown in Dulbeccos minimum essential medium (DMEM) supplemented with 6% foetal bovine serum (FBS) and 5% tryptose phosphate broth, and high titre stocks of MHV were prepared using low pH medium (Sawicki & Sawicki, 1986b ).

Labelling of viral RNA and isolation of high-salt-soluble and -insoluble RNA species.
17Cl-1 cells (5x100 mm Petri dishes with 1015x10⁶ cells per dish) were infected with MHV at 50100 p.f.u. per cell at 37 °C. Following infection, the cells were labelled with [³H]uridine or [³²P]orthophosphate. For [³H]uridine, the cells were incubated in medium containing 200 µCi/ml of [³H]uridine and 20 µg/ml of dactinomycin in DMEM supplemented with 6% FBS. For ³²P, they were labelled with 200 µCi/ml of [³²P]orthophosphate and 20 µg/ml of dactinomycin in phosphate-reduced MEM supplemented with dialysed FBS. Cells were solubilized with 5% lithium dodecyl sulfate and 200 µg of proteinase K/ml and deproteinized by extraction with low pH phenol (pH 4·3) followed by chloroform. The RNA was collected by ethanol precipitation. For separation of high-salt-soluble and -insoluble species, the aqueous phase obtained by phenol and chloroform extraction was adjusted to 2 M LiCl, placed on ice for 18 h, and centrifuged at 10000 r.p.m. (15000 g) for 1 h to obtain the supernatant (LiCl-soluble) and precipitated (LiCl-insoluble) fractions (Baltimore, 1968 ; Sawicki & Gomatos, 1976 ). LiCl-soluble RNA was collected by ethanol precipitation.

Ribonuclease protection assays to measure negative-strand synthesis.
Viral RNA was denatured by heating at 100 °C in 1 mM EDTA, quick cooled at 0 °C, and reannealed in the presence of an excess of virion positive strands and 0·4 M NaCl at 68 °C for 30 min followed by 25 °C for 30 min, as described (Sawicki & Sawicki, 1986a ). One half of each sample was analysed directly for acid-precipitable radioactivity; the other half was digested with 5 µg/ml of RNase A in 0·4 M NaCl for 30 min at 37 °C before acid-precipitation.

Chromatography.
Columns (1·5x90 cm) of Sepharose 2B or Sephacryl S-1000 (Pharmacia) were equilibrated in 0·1 M NaCl, 10 mM TrisHCl, pH 7·4, 20 mM EDTA and 0·2% SDS, as described earlier for Sepharose 2B (Sawicki & Gomatos, 1976 ). Briefly, RNA to be chromatographed was dissolved in the column buffer and applied to the column in small volumes. The buffer head was reapplied and the column allowed to run under the recommended pressure and at a flow rate of 1013 ml/h. Fractions of 1·01·3 ml (Sepharose 2B total RNA and LiCl-soluble RNA) or 0·65 ml (Sepharose 2B LiCl-insoluble RNA) or 0·5 ml (Sephacryl S-1000) were collected. The RNA was precipitated in the presence of 50100 µg of carrier yeast tRNA with ethanol.

Ribonuclease digestion.
For determination of the ribonuclease-resistance of RI/TIs and native RF/TFs and fractions, RNA samples were digested with RNase T1 (Ambion) at 30 units per sample in 0·3 M NaCl at 30 °C for 30 min or with RNase A (affinity purified; Ambion) at 5 µg/ml in 0·3 M NaCl, 30 mM sodium citrate (2x SSC) at 37 °C for 30 min. Trichloroacetic acid was added to 5%, and the precipitated RNA collected on glass fibre filters and counted by liquid scintillation spectroscopy. We routinely used 30 units of RNase T1 to digest ∼30 µg of RNA, the amount in 1x10⁶ cells.

Velocity sedimentation and agarose gel electrophoresis.
Solubilized cell extracts were passed through a 27 gauge needle and layered onto 1530% sucrose gradients and centrifuged at 20 °C in a Beckmann SW28 rotor at 28000 r.p.m. for 18 h, which pelleted the 60S viral genome RNA. Gel electrophoresis was in 0·8% or 1% agarose in TBE buffer with 0·2% SDS for 370 V-h, after which the gels were washed in water to remove excess SDS, stained with ethidium bromide, and processed for fluorography.

Chromatography on oligo(dT) beads.
Oligo(dT)₂₅ beads were obtained from Dynal (Dynabeads) or from Novagen (Magnetight particles) and were used according to the instructions of the manufacturers. Immediately before use, the beads were washed in binding buffer twice before addition of the sample. Usually 100 µl of cell lysate or RNA in water was heated for 2 min at 75 °C and then added to 200 µl of washed oligo(dT)₂₅ beads resuspended in 100 µl of binding buffer. After rotating at 25 °C for 5 min, the magnet was applied and the supernatant (unbound fraction) was removed. The beads were washed twice with washing buffer (wash fractions) before incubation in 1020 µl of 10 mM TrisHCl, pH 7·4, and heating at 75 °C for 2 min to elute the poly(A)⁺ RNA or bound fraction.

Gel filtration chromatography and velocity sedimentation
MHV RI/TIs are separable by velocity sedimentation. MHV-infected cells were labelled with [³H]uridine from 2·5 to 6 h post-infection (p.i.) and the RNA applied to a 1530% sucrose gradient. Fig. 1(A) shows the distribution of labelled, viral single-stranded RNA (the genome RNA was pelleted) and Fig. 1(B) shows the distribution of the RI/TIs after conversion to RF/TFs with RNase T1. RI/TIs represented about 5% of the viral RNA and, while separable from the smaller TIs, the RI overlapped TIs for RNA-2 and -3 and the TIs for RNA-2 and -3 overlapped TIs for RNA-4, -5, -6 and -7. For the gel shown in Fig. 1(C), the experiment was repeated using ³²P and labelling from 1·5 to 5 h p.i. and fractions were combined into four pools, numbered starting from the bottom of the gradient: pool 1 (fractions 14) was enriched in the RI; pool 2 (fractions 69) in RI and TIs II and III; pool 3 (fractions 1013) in TIs II and III; and pool 4 (fractions 1619) in the smaller TIs IV, V, VI and VII. After conversion to RF/TFs, pools 2 and 4 were subjected to chromatography on Sephacryl S-1000 columns (Fig. 2A). Sephacryl S-1000 is a hydrophobic, rigid, allyl dextran/N-N'-methylenebisacrylamide matrix with a 20 kbp DNA exclusion limit. RF/TFs in pool 2 were mostly excluded by the column or at the boundary of the excluded/included region and consisted of RF cores and some TF-II and TF-III cores (Fig. 2B). Only RF was found in the earliest fractions to exit the column, providing some separation of the RF from the smaller TF species. Pool 4 (Fig. 2C) contained low amounts of the RF and of TF-II and TF-III, and large amounts of the four smallest TF cores, which eluted very narrowly and by size and overlapped one another.

(75K): Fig. 1. Separation of MHV RI/TIs by velocity sedimentation. After low pH phenol and chloroform extraction, RNA labelled with 200 µCi/ml [3H]uridine from 2·5 to 6 h p.i. was layered onto 15% to 30% sucrose gradients and centrifuged at 28000 r.p.m. for 18 h at 20 °C. Fractions (numbered from the bottom to the top of the gradient) of about 1 ml were collected. (A) Sedimentation of the viral single-stranded RNA. 5% of each gradient fraction was analysed by electrophoresis on a 1% agarose gel that was stained with ethidium bromide to locate the ribosomal RNA. (B) RNA (95%) in each gradient fraction was precipitated with ethanol and digested with RNase T1 to remove single-stranded mRNAs and convert RI/TIs to their double-stranded RF/TF cores that are readily visualized by electrophoresis on a 1% agarose gel. (C) From identical velocity sedimentation conditions reported above for [3H]uridine, 32P-labelled RNA, from cells labelled 1·5 to 5 h p.i., in pool 1 (fractions 14), pool 2 (fraction 69), pool 3 (fractions 1013) and pool 4 (fractions 1619), was precipitated with ethanol. After RNase T1 digestion, a small sample of each was analysed by electrophoresis.

(37K): Fig. 2. Sephacryl S-1000 gel filtration chromatography of RNase-T1-derived RF/TFs. (A) Distribution of radiolabelled RF/TFs. Pools 2 and 4 from Fig. 1(C) were individually applied after RNase T1 treatment to Sephacryl S-1000 columns as described in Methods. Fractions of 0·5 ml were collected and 20 µl samples were acid-precipitated for scintillation spectroscopy. Presence of individual RF/TFs core species in pool 2, fractions 81100 (B) and pool 4, fractions 80130 (C) was determined. The RNA in each fraction was collected by ethanol precipitation and electrophoresed on agarose gels.

Identification of seven native RF/TF RNA species
Native RF/TFs are recovered in the 2 M LiCl-soluble fraction whereas RI/TIs are mostly single-stranded and are precipitated by 2 M LiCl. When chromatographed on Sepharose 2B directly, LiCl-soluble RNA fractionates into three peaks, the largest of which is excluded and in the void volume (Fig. 3A). Fractions across the excluded and included (fractionated) regions were combined into eight pools that were analysed by gel electrophoresis without nuclease treatment (Fig. 3B). Pool 1 contained the entire native RF and native TFs II and III, and small amounts of the native TFs IV, V and VI. Pool 2, the region between the excluded and included fractions, contained mostly TFs IV, V, VI and VII. Pool 3 contained mostly TFs V, VI and VII. Pool 4 was enriched in TF-VII. Labelled RNA in pools 68 was much smaller than mRNA 7, migrating off the gel during electrophoresis. Most likely, this represented short oligoribonucleotides created by blocking transcription with dactinomycin. Thus, MHV-infected cells contained seven species of native RF/TFs.

(51K): Fig. 3. Sepharose 2B gel filtration chromatography of MHV LiCl-soluble, native RF/TFs and LiCl-insoluble RI/TIs. MHV-infected cells were labelled from 1 to 5·5 h p.i. with 200 µCi/ml [3H]uridine, extracted with low pH phenol and chloroform, and the aqueous phase was adjusted to 2 M LiCl and incubated overnight on ice. (A) The LiCl-soluble, native RF/TFs were precipitated with ethanol and subjected to chromatography on Sepharose 2B, as described in Methods. 20 µl (∼2%) of each fraction was TCA-precipitated for scintillation spectroscopy. (B) Fractions of the LiCl-soluble material were pooled and a sample was electrophoresed directly on agarose gels. (C) The LiCl-insoluble single-stranded RNA and RI/TIs were pelleted directly by centrifugation (15000 g for 60 min), resuspended and applied to a Sepharose 2B column, as described in Methods. 10 µl (∼2%) of each fraction was analysed for acid-precipitable radioactivity. (D) Distribution of the LiCl-insoluble, mostly single-stranded RNA species. A small part of every second or third fraction was subjected to agarose gel electrophoresis. Fractions 6065 contained the viral RI that migrates near the top of the gel. Most of the genomes were degraded during sample preparation, but small quantities of intact genomes were present in fractions 6065 and they migrated faster than the RI. Fractions were pooled for further analysis as shown in (E). (E) Distribution on Sepharose 2B of the LiCl-insoluble RI/TIs. The fractions from the chromatograph shown in (C) were collected into 20 pools, starting with fraction 56 of the excluded fractions (fractions 5666). Pool 1 contained the first seven fractions, pool 2 the next three fractions, and thereafter each pool contained the next five fractions. The RNA in each pool was collected by ethanol precipitation and a sample was treated with RNase T1. The resultant RF/TF cores were separated by electrophoresis. Only 18 lanes are shown in (E); no radioactive bands were present in pools 19 and 20.

Fig. 3(C) shows the results of the chromatography on Sepharose 2B of the RNA that was precipitated by 2 M LiCl. At least 13 times more viral RNA was found in the excluded volume from the LiCl-insoluble fraction compared to the LiCl-soluble fraction (Fig. 3A) and the RNA in the included volume eluted broadly. Starting at fraction 56 and ending at fraction 155, every second or third fraction was analysed by gel electrophoresis (Fig. 3D). 28S and 18S ribosomal RNA (detected by ethidium bromide staining) was in fractions 90145 and 120160, respectively. Genome (RNA-1) started to elute in the excluded volume and just after RI, which was visible at the top of the gel in fractions 6065 (Fig. 3D). Eluting later and according to their relative size was single-stranded subgenomic mRNA-2 to -7 (RNA-2 to -7). As observed earlier (Sawicki & Sawicki, 1990 ), the RI migrated slower than RNA-1 on agarose gels.

About 100 fractions were collected into 20 pools of about five fractions each and numbered in the order of elution. RNA in pools 1 to 18 was treated with RNase T1 and analysed on gels to identify the location of RI/TIs (Fig. 3E). Pool 1 (fractions 5662) contained most of the RI, TIs II and III, and pools 67 (fractions 8185 and 8690) were enriched in TIs IV, V, VI and VII. The presence of large amounts of RNA in pool 1, and possibly larger fragments derived from RNA-1 by RNase T1 digestion, may account for the smear of labelled material. We found the seven species of RI/TIs fractionated similarly to native RF/TFs on Sepharose 2B and each eluted ahead of its single-stranded RNA counterpart. To approximate the relative numbers of RI/TIs to native RF/TFs, radiolabelled RNA recovered as RNase A-resistant cores after Sepharose 2B for each was calculated. The 2 M LiCl-insoluble RI/TIs accounted for 67% of all labelled, double-stranded RNA accumulating in infected cells early (1·55·5 h p.i.). This is the period when rates of syntheses of both negative-strand and positive-stranded RNA were increasing and viral RNA was accumulating exponentially (Sawicki & Sawicki, 1986a ).

Ribonuclease-sensitivity of coronavirus RI/TI and RF/TF species
Essentially all of the [³H]uridine incorporated into the LiCl-soluble native RF/TFs was acid-precipitable after digestion with RNase T1 or A (Table 1). In contrast, RNase A digestion of LiCl-insoluble, Sepharose 2B-fractionated RNA showed RNA in pool 1 was mostly double-stranded (49%), and RNA in pools 520 was mostly single-stranded, 1·5% or less was resistant to RNase (Table 1). Our next experiment showed both native form and intermediate structures, while double-stranded, can be degraded with excessive RNase A. We treated the purified, LiCl-soluble, native RF/TFs (Fig. 3B, pools 1 and 3 were combined to restore the population of seven species) and LiCl-insoluble RI/TIs (Fig. 3E, pools 1 and 5 were combined together) with RNase T1 or A, using several concentrations of each. RNase T1 at concentrations of 10100 units per sample converted RI/TIs to RF/TF cores and left intact all seven native RF/TFs as judged by their electrophoretic migration in agarose gels (Fig. 4A; data not shown). However, treatment with as little as 1 µg/ml of RNase A destroyed the intact structure of larger, native RF/TFs species and higher amounts also degraded the smaller TFs (Fig. 4B). The RI/TIs were also degraded by these concentrations of RNase A (Fig. 4C). For the volumes used, a concentration of 1 µg/ml represents on a µg basis a 400:1 RNA to RNase ratio.

Table 1. Ribonuclease resistance of coronavirus RI and RF RNA

(60K): Fig. 4. Sensitivity of native RF/TFs and RI/TIs to RNase A. (A) The LiCl-soluble, Sepharose 2B pools 14 (Fig. 3B) were digested individually with RNase T1 at 10 units per sample in 0·3 M NaCl for 30 min at 30 °C, SDS was added and the samples were separated by electrophoresis on 1% agarose gels. (B) Native RF/TFs were reconstituted as a population by combining pools 1 and 3 (Fig. 3B). An equal amount was digested with each of the indicated concentrations of RNase A in 2x SSC at 37 °C for 30 min. In addition to native RF/TFs from 5·7x106 cells, each sample also contained 10 µg of tRNA that was added as a carrier during the ethanol precipitation step and formed the bulk of substrate RNA. After digestion, SDS and proteinase K was added and the samples were subjected to electrophoresis on a 1% agarose gel. (C) A sample of LiCl-insoluble viral RI/TIs, obtained by combining pools 1 and 5 (Fig. 3E), was treated with RNase A identically to that described for (B).

Presence of poly(A) in coronavirus RI/TI and RF/TF populations
Preparations of LiCl-soluble and -insoluble RNA were selected on oligo(dT)₂₅ beads to monitor for the presence of poly(A) sequences that were not base-paired with a complementary poly(U) sequence (Fig. 5). Presence of a free poly(A) sequence on nascent strands in the RI/TIs was suggested because all seven species of RI/TIs bound (lane 5), in addition to all the single-stranded mRNA species (lane 2). Although it appears here that the RI did not bind to oligo(dT)₂₅ beads (lane 5), other experiments using cell lysates directly, without phenolchloroform extraction, demonstrated that almost all the RI, in addition to the TIs, bound. Essentially all of the poly(A) RNA that bound to and was eluted from oligo(dT)₂₅ rebound when challenged with a second round of poly(A) selection; none of the unbound RNA bound when incubated with fresh oligo(dT)₂₅ beads (data not shown). Most if not all of the native RF/TFs had a free poly(A) sequence long enough to bind oligo(dT)₂₅ (Fig. 5, lane 8); all of the smallest, and decreasing amounts of the RF/TFs of increasing sizes bound. In some experiments, as for the RI, only about 10% of native RF was observed to bind oligo(dT)₂₅. This indicated that the RF poly(A) sequence was not free or the extreme size of the 32 kb double-stranded RF interfered with stable binding to short oligo(dT)₂₅ sequences under the assay conditions. In contrast to the inconsistency of recovery of RI/RF by oligo(dT)₂₅ beads, we observed consistent binding by all of the TIs/TFs.

(78K): Fig. 5. MHV RI/TIs and native RF/TFs bind oligo(dT)25. The 2 M LiCl-insoluble single-stranded RNA (lanes 13, enriched in mRNA 47, is RNA from pools 13+14 of Fig. 3E), the 2 M LiCl-insoluble RI/TIs (lanes 46 is RNA from pools 1+5 of Fig. 3E) and 2 M LiCl-soluble RNA (lanes 79 is RNA from pools 1+3 in Fig. 3B) were analysed for their ability to bind to magnetic beads coated with oligo(dT)25. Input RNA (lanes 1, 4 and 7) was equivalent in amount to that applied to the beads and found either as bound (lanes 2, 5 and 8) or not-bound (lanes 3, 6, and 9) fractions. RNA in lanes 13 was analysed directly; that in lanes 46 and 79 was treated with RNase T1 before electrophoresis on 1% agarose gels. Lanes 13 contained RNA from one-half as many cells as lanes 46, and lanes 46 were exposed for twice as long as lanes 13 (a final fourfold difference).

Kinetics of labelling of MHV RNA synthetic structures
If TIs were merely dead-end products of the synthesis of negative strands as was suggested by Jeong & Makino (1994) , radiolabelled uridylate would accumulate continuously in these molecules as seen for poliovirus RF (Baltimore, 1968 ). On the other hand, if they were active intermediates in subgenomic mRNA synthesis, their labelling would resemble the labelling of RI. Fig. 6 shows the results of this type of analysis. Infected cells were pulse-labelled for 290 min with [³H]uridine at 6·5 h p.i., when less than 1015% of the [³H]uridine that is incorporated into RI/TIs is in negative strands. The RF/TFs RNA derived by RNase T1 from the RI and TIs was separated by electrophoresis on gels and the radiolabel incorporated into each species at each time-point was determined by cutting out and counting the specific gel area that contained each RF/TFs. The kinetics of labelling into all of the TFs cores showed the expected properties for true transcription intermediates (Fig. 6). They rapidly incorporated radiolabelled uridylate into growing chains at rates indistinguishable from that of RF, and the amount of incorporation reached a maximum (saturated) after about 30 min of labelling. Thus, the six species of MHV TIs, in addition to the RI, are authentic transcription intermediates.

(18K): Fig. 6. Kinetics of incorporation into MHV RI/RF and subgenomic TI/TF RNA size classes. Infected cells were labelled at 6·5 h p.i. with [3H]uridine for 290 min periods, as indicated. Cell lysates were treated with RNase T1 to obtain the RF/TF cores, which were separated by electrophoresis on 1% agarose gels. The area of the gels containing each RF/TF species (, RF; •, TF-II; , TF-III; , TF-IV; , TF-V; , TF-VI; , TF-VII) was cut out and its amount of radiolabelled RNA was determined by scintillation spectroscopy.

The second type of analysis is shown in Table 2. If subgenomic mRNA was copied into negative strands that were not templates for subgenomic mRNA synthesis, the TI negative strands would be enriched in [³H]uridine. Infected cells were labelled for 30 min periods from 1 to 6 h p.i., the large (18S40S) and small (5S18S) RI and TIs were obtained by velocity sedimentation (Fig. 1), and the percentage of labelled negative strands in their purified, double-stranded cores was determined in ribonuclease-protection assays. The same relative amounts of labelled negative-strand RNA, newly made during the pulse period, was present in populations enriched in the genome or in the four smaller TI species (Table 2). Early in infection, each 30 min labelling period showed large amounts of negative-strand synthesis and about equal accumulation in both size populations of RI/TIs. A value of 3335% for labelled RF/TF RNA that was negative-stranded indicated that about 70% of the total negative strands were made during this 30 min period, as expected during the exponentially increasing phase of replication. This decreased to 1214% at 5·56 h p.i., a time when the rate of overall positive-strand synthesis became maximal (Sawicki & Sawicki, 1986a , 1990 ). The results indicated that the TIs were not formed by a single round of negative-strand synthesis on subgenomic mRNA templates (dead-end synthesis) but were transcription intermediates active in the synthesis of subgenomic mRNAs.

Table 2. Percentage of newly synthesized negative strands in TIs IVVII is the same as in RI RNA

Seven species of MHV RF/TFs and of RI/TIs were readily isolated and analysed by velocity sedimentation or gel filtration chromatography. Comparison of gel filtration on Sephacryl S-1000 (Fig. 2) and Sepharose 2B (Fig. 3) demonstrated that they chromatographed close to their predicted sizes based on the fractionation properties of each matrix. Only Sephacryl S-1000 had all TFs/TIs species within its fractionation range and excluded mainly RI/RF. Not even Sephacryl S-1000, however, was able to eliminate overlapping of the two mid-sized TI/TF RNA molecules with each other, and of the four smallest TIs/TFs with each other. Thus, electrophoresis on agarose gels remains the best method to purify individual size classes of RI/TIs and RF/TFs, which was the method we used previously (Sawicki & Sawicki, 1990 ). By combining sedimentation on sucrose gradients and gel filtration, larger RI and TIs II and III were separated from the smaller TIs IV, V, VI and VII.

At the time of maximum RNA synthesis rates, the multi-stranded RI and TIs were the vast majority of replication and transcription intermediates. Less than 30% were present as native RF/TFs. This is comparable to other positive-stranded RNA animal and bacterial viruses (reviewed in Koch & Koch, 1985 ). RI/TIs represent replication and transcription intermediates on which viral RNA-dependent RNA polymerases are repeatedly initiating positive-strand synthesis. The relative abundance of each RI/TIs or native RF/TFs species was similar and proportional to the relative abundance of the viral positive strands. Native RF was originally defined by its property of solubility in high salt, a property shared by both native RF and TFs of MHV. That this indeed reflected their completely or nearly completely double-stranded nature was confirmed by finding essentially all of the labelled MHV native RF/TF RNA was acid-precipitable after RNase A digestion. No single-stranded labelled RNA was found when this population was analysed directly by electrophoresis. In contrast, high-salt-insoluble viral RNA contained genomes and subgenomic mRNA as well as RI/TIs and only 3% of it was resistant to RNase A. Viral genome and subgenomic mRNA and positive-strand components of RI/TIs possessed a poly(A) sequence capable of binding to oligo(dT) sequences, showing that coronavirus RI/TIs resemble RIs formed by other positive-stranded RNA viruses (Ammann et al., 1964 ; Baltimore, 1966 , 1968 ; Erikson & Gordon, 1966 ; Koch & Koch, 1985 ; Montagnier & Sanders, 1963 ; Richards et al., 1984 ; Sawicki & Gomatos, 1976 ; Yogo et al., 1977 ). Binding of native MHV RF/TF species to oligo(dT)₂₅ suggests that if a polyuridylate is present at the 5' ends of MHV negative strands, it is shorter than the poly(A) in its positive-strand counterpart. Bovine coronavirus was reported to have a poly(U) of about 926 nt, compared to a positive-strand poly(A) of 100130 nt (Hofmann & Brian, 1991 ).

Initial failure to find subgenomic TIs led to the leader-primed transcription model (Baric et al., 1983 ). The present results indicate this failure was due most likely to technical error and not to any aberrant behaviour of these molecules on gel filtration chromatography. Each of seven RI/TIs and native or core RF/TF species fractionated as predicted on each of the two different matrices. In addition to detecting and characterizing seven species of RI/TIs and RF/TFs in experiments duplicating that of earlier investigators, it was also important to attempt to explain their failure to find TI/TFs. While MHV A59 RI and RF are resistant to concentrations of RNase A of 0·33 µg/ml (Sawicki & Sawicki, 1990 ), the concentrations of ribonuclease A of 10 µg/ml (Baric et al., 1983 ) and 20 µg/ml (Lai et al., 1982b ) used would have degraded RI/TIs and native RF/TFs, leaving only short fragments of the original structures. Moreover, this and other studies (Baric & Yount, 2000 ; Sawicki & Sawicki, 1990 ) found native RF/TFs and double-stranded cores of RI/TIs migrated on gels slower than RNA-1, not faster as was claimed (Baric et al., 1983 ). This means that not even authentic RF was recovered in earlier studies (Baric et al., 1983 ; Lai et al., 1982b ).

Another issue has to do with recovery of particular RNase T1 oligonucleotides assigned to viral mRNA and found in fractions thought to contain only RI (Baric et al., 1983 ; Lai et al., 1982b ). Finding RNase T1-resistant oligonucleotides #10 and #19 in RI-containing fractions after Sepharose 2B chromatography led others (Baric et al., 1983 ) to claim the genome-length RI was utilized for the synthesis of subgenomic mRNA. At the time, the authors (Baric et al., 1983 ) expressed surprise at not finding oligonucleotide #3a that was unique to subgenomic mRNA-5 and oligonucleotide #19a that was unique to subgenomic mRNA-6 (Lai et al., 1982a , 1983 ). Subgenomic mRNA-6 is almost as abundant as subgenomic mRNA-7. We can explain their results in the light of our finding the TIs for subgenomic mRNA-2 and -3, together with the RI, in the excluded fraction after Sepharose 2B chromatography. Oligonucleotide #10, which is derived from the leader sequence present only once at the 5' end of the genome and of each subgenomic mRNA, and oligonucleotide #19, present in sequences from the leaderbody junction regions of the genome and mRNA-2, -3 and -7, were found in the excluded fraction of Sepharose 2B (Baric et al., 1983 ). We found this fraction actually contained the genome and some mRNA-2 and -3, in addition to the RI and TIs II and III. This readily accounts for the presence of oligonucleotides #10 and #19. Absence of oligonucleotides #3a and #19a is explained by our finding TIs for subgenomic mRNA-5 and -6, respectively, in the included volume, not in the excluded volume. If genome-length, RI negative strands were being used to synthesize subgenomic mRNA-6, oligonucleotide #19a would have been present in the excluded column volume. Rather than favouring leader-primed transcription, the results of Baric et al. (1983) actually argued against it.

Is leader-priming used at any time during nidovirus replication? Our results and those of Baric & Yount (2000) for MHV and van Marle et al. (1999) for EAV would argue not. Recently, a study suggested leader-primed transcription occurred immediately after infection. An et al. (1998) detected subgenomic defective interfering (DI) mRNA but not its negative-strand template at the very earliest times after infection, although shortly thereafter subgenomic DI negative strands became detectable. Because there are about 100 times fewer negative-strand templates compared to its products, it would be difficult to rule out the possibility that formation of the negative-strand template for the subgenomic DI did not precede formation of the subgenomic mRNA. An et al. (1998) did in fact confirm that negative-strand templates for subgenomic DI mRNA were detectable very early after infection and at levels reflective of subgenomic DI mRNA levels. If leader-primed transcription was required to produce subgenomic DI mRNA at early but not later times, we are left with the question of why two, redundant mechanisms, one requiring a primer and one not, are used to produce subgenomic positive strands?

MHV TIs behaved during metabolic labelling as authentic transcription intermediates. Kinetics of their labelling with [³H]uridine were similar to that observed for the RI. Also, kinetics of negative-strand synthesis for smaller TIs were the same as for the RI, and all populations of RI/TIs incorporated [³H]uridine mostly into positive strands late in infection when negative-strand, but not positive-strand, synthesis was declining. Kinetic labelling experiments similar to those reported in this study have been published by others (Baric & Yount, 2000 ; Schaad & Baric, 1994 ). These support our hypothesis that complementary negative strands function as templates for subgenomic mRNA synthesis during coronavirus infection. The ability to isolate the full set of seven viral RI/TIs and seven native RF/TFs from infected cells and to explain the initial failures (Baric et al., 1983 ; Lai et al., 1982b ) to find them invalidates the original basis for proposing the leader-primed model. Furthermore, the initial failure (Lai et al., 1982b ) to find MHV subgenome-length negative strands was most likely also due to a technical error. The authors used as a probe to detect negative strands ³²P viral RNA obtained from infected cells grown in [³²P]orthophosphate. Such a probe would not have sufficiently high specific activity to detect negative strands by Northern blot. Contrary to their claim, and with our results, it is now clear that authentic negative strands were not detected.

At this time, our model (Sawicki & Sawicki, 1995 ) identifying the discontinuous transcription event as the step or process generating the 5' nested set of negative strands that serve as templates for subgenomic mRNA synthesis best explains all available experimental data. Because subgenomic negative strands in coronavirus-infected cells contain anti-leader sequences (Sawicki & Sawicki, 1995 ; Sethna et al., 1991 ), they directly serve to template the 3' nested set of viral mRNA. However, with MHV, in order for positive strands to serve as templates for negative strands, i.e. to replicate, sequences downstream of the 5' genomic leader sequence are required (Masters et al., 1994 ). Therefore, coronavirus subgenomic mRNA and their negative-strand templates do not replicate. The exact mechanism by which coronaviruses, and the related arteriviruses, generate a 5' nested set of negative-strand templates that are complementary copies of the subgenomic mRNA remains to be elucidated. Whatever the unique and intriguing mechanism, it is used probably by all members of the Nidovirales (Snijder & Meulenberg, 1998 ). It would determine and regulate relative numbers of each of the variously sized negative-strand templates. With the Nidovirales, the synthesis of genomes and subgenomic mRNAs would be determined by the number of each negative-strand template formed during infection. Details of this mechanism and identification of the RNA sequences/structures and essential viral proteins are only now being investigated at the molecular level (van Dinten et al., 2000 ; van Marle et al., 1999 ). The endeavour will benefit greatly from availability of infectious clones for arteriviruses (van Dinten et al., 1997 ) and more recently for coronaviruses (Almazan et al., 2000 ; V. Thiel, J. Herold, B. Schelle & S. G. Siddell, unpublished data).

Support for these studies was from the National Institutes of Health, awarded to S.G.S. (AI-28506) and D.L.S. (AI-15123).