Genetic characterization of the coxsackievirus B2 3' untranslated region

Coxsackievirus B2 (CVB2) is a human pathogen that causes a broad spectrum of disease ranging from summer grippe to far more serious infections, such as myocarditis and meningitis (Melnick, 1996 ). CVB2 is a positive-stranded RNA virus that is closely related to other group B coxsackieviruses and echoviruses, which are all members of the Human enterovirus B subgroup of the genus Enterovirus, family Picornaviridae (King et al., 2000 ). The genome of the CVB2 prototype strain Ohio-1 (CVB2O) is 7411 nt and contains a single open reading frame that is flanked by both 5' and 3' untranslated regions (UTR), and the 3' terminus of the genome is polyadenylated (Polacek et al., 1999 ). The UTRs of enteroviruses have complex secondary structures that are important for several events in the virus life-cycle. The structural folds of the 5'UTR may be divided into two main functional elements (Rohll et al., 1994 ). The first element enables initiation of positive-strand RNA synthesis and has been shown in the case of poliovirus to consist of a cloverleaf-like structure at the utmost 5' end. The virus precursor 3CD and the cellular poly(rC)-binding protein (PCBP) bind to this 5' structure to form a multicomponent ribonucleoprotein (RNP) complex (Andino et al., 1993 ; Harris et al., 1994 ; Parsley et al., 1997 ). The second structural element is an internal ribosome entry site, which catalyses cap-independent initiation of translation, and is involved in binding host cell translation initiation factors and RNP (Belsham & Sonenberg, 2000 ; Jang et al., 1988 ; Pelletier et al., 1988 ; Trono et al., 1988 ). It has been suggested recently that the RNP complex (by the specific interaction of PCBP and the viral polymerase 3D with the 5' cloverleaf structure) can determine whether or not the RNA is to be used as a template for translation or for RNA synthesis (Gamarnik & Andino, 1998 , 2000 ).

Although enteroviruses and rhinoviruses have similar 5'UTRs, the 3'UTR is more divergent in terms of its length and structure. Within the enterovirus genus, the 3'UTR is known to fold into two main structures: a double hairpin structure (poliovirus-like viruses) with domains X and Y, and a triple hairpin structure (coxsackievirus B-like viruses) with domains X, Y and Z. Both types of structural folds are joined together by an S domain consisting of parts of the poly(A) tail and a U-rich region upstream of the stop codon. Together, these domains constitute an intramolecular secondary and tertiary structure that participates in the initiation of negative-strand RNA synthesis and is referred to as oriR (Melchers et al., 1997 ; Mirmomeni et al., 1997 ; Pilipenko et al., 1992 , 1996 ; Wang et al., 1999 ). Negative-strand RNA is synthesized by the primer-dependent viral RNA polymerase 3D, which uses a uridylylated form of the genome-linked protein VPg as its primer (Paul et al., 1998 ; Toyoda et al., 1987 ). It has been suggested recently that nucleotide structures within the polyprotein-coding regions of poliovirus and rhinovirus might also be a part of the replication machinery (Goodfellow et al., 2000 ; McKnight & Lemon, 1996 ).

In this study, we investigated the genetic characteristics of the 3'UTR of CVB2O. Previously, we generated an infectious clone of CVB2O by RTPCR using CVB3-derived primers (Lindberg et al., 1997 ). However, subsequent determination of the wild-type (wt) sequence (Polacek et al., 1999 ) showed the presence of point mutations and a five nucleotide extension at the 3'UTR prior to the poly(A) tail compared with the constructed clone and other enteroviruses. In order to investigate the importance of these additional nucleotides, virus derived from the cDNA clone was propagated in cell culture for a number of passages and the 3'UTR genotype of the virus progeny was analysed. Our results show that the clone-derived virus progeny compensate for the 5 nt deletion and the point mutations that are found in the original clone by forming more energetically favourable secondary structures at the 3'UTR, as predicted by the calculation of conformational free energy. In addition, we observed that, during the first round of passages in cell culture, the poly(A) tail of the cDNA clone was extended from the initial 17 A residues to more than 40 A residues in the virus progeny. During further passages, the poly(A) tail increased to about 100 A residues; this seems to be a more favourable length for CVB2O under these conditions.

Molecular characterization of an infectious CVB2O clone.
The strategy for cloning CVB2O has been described previously (Lindberg et al., 1992 ). Briefly, the CVB2O genome was amplified by a long-distance PCR method using 5' and 3' end primers derived from a coxsackievirus B3 full-length sequence (Klump et al., 1990 ). The PCR amplicon was then cloned into the phagemid vector pCR-Script Direct SK(+) (Stratagene) using the AscI and NotI cloning sites. One of the infectious clones, pCVB2O-8 (referred to previously as pCVB2Ov-A8G; Lindberg et al., 1997 ), was selected and the complete nucleotide sequence was determined by primer walking, as described for the prototype strains of CVB2O, coxsackievirus B5 and echovirus 5 (Lindberg et al., 1998 ; Lindberg & Polacek, 2000 ; Polacek et al., 1999 ). The length of the poly(A) tail in pCVB2O-8 was determined by sequencing to be 17 A residues. Comparison of the cloned and wt sequences of CVB2O is shown in Table 1.

Table 1. Sequence differences between wt CVB2O and pCVB2O-8

Lipofection and passages in green monkey kidney (GMK) cells.
pCVB2O-8 (6 µg) was lipofected into T25 flasks containing GMK cells (∼60% confluent) using Lipofectamine (Life Technologies). When cytopathic effects (CPE) were observed, the flasks were frozen and then freezethawed three times. Supernatant (1 ml) was used to infect new T25 flasks (∼75% confluent) with GMK cells, which were frozen when CPE were apparent and then treated as above for subsequent passages. Lipofection (L) was performed in four parallel experiments termed A, B, C and D (L_A, L_B, L_C and L_D), where L_A0 refers to the lipofection stage, L_A1 to passage one and L_A2 to passage two, etc. Virus progeny from L_AC were studied for a further seven passages (L_A0L_A7, L_B0L_B7 and L_C0L_C7) and L_D for a further 12 passages (L_D0L_D12) (Fig. 1). The titres of progeny virus were determined as TCID₅₀ (Reed & Muench, 1938 ).

(13K): Fig. 1. Experimental overview of the passages of cloned and uncloned CVB2O. A wt CVB2O clone, pCVB2O-8, was lipofected in four parallel experiments (LAD) and propagated for up to 12 passages in GMK cells. As virus progeny from passage LB7 consisted of a mixed population, a separate analysis of isolated clones was performed. The filled hexagon indicates wt CVB2O, open squares indicate cloned CVB2O and open hexagons indicate lipofected virus progeny.

Sequence analysis of the 3D and 3'UTR regions of revertants.
Viral RNA from each of the different passages was extracted from freezethawed supernatants by differentiated centrifugation and SDS/proteinase K treatment, as described previously (Lindberg et al., 1997 ). Thereafter, RNA was reverse-transcribed with Superscript II (Life Technologies) at 45 °C for 2 h using NotdT25 as the primer (Table 2). The RNA template was hydrolysed in 0·2 M aqueous NaOH for 20 min at 37 °C and the newly synthesized single-stranded cDNA was precipitated with ethanol and dissolved in Tris/EDTA buffer. The sequence corresponding to the 3' end was amplified by PCR with the primer pair dT26V and CB-43bio (Table 2) and then purified using the QIAquick gel extraction kit (QIAGEN). The sequence of the PCR product was determined with the dRhodamine Terminator Cycle Sequencing kit (Perkin Elmer) using the primer CB-47 (Table 2). Sequencing of the 3' end of the virus progeny from each passage was performed in at least two independent reactions on an ABI Prism 310 Genetic Analyser (Perkin Elmer). Additional sequencing was performed to extend to the active-site cleft of the viral 3D polymerase. This region was amplified by PCR using the primers EV-106 and EV-136, purified as above and sequenced with the primer EV-177 (Table 2).

Table 2. Oligonucleotides used for cDNA synthesis, PCR, RACEPAT and DNA sequencing

Cloning and sequence analysis of L_B7 3'UTR.
Purified RNA from passage L_B7 containing an apparent heterogeneous mixture of genotypes, as detected by sequencing, was reverse-transcribed with the primer NotdT25 (as above) and then amplified by PCR using the primers NotdT25 and CB-43bio (Table 2). The PCR product was then cut with NotI/NsiI and ligated into pCVB2O-8, which had been digested previously with the same enzymes. The construct was then transformed into competent E. coli strain DH5α. PCR with an increased initial denaturation time of 2 min was performed directly on 20 individual colonies using the primers NotdT25 and CB-47. PCR products were purified with the QIAquick gel extraction kit and sequenced with the primer CB-44, as described above.

One-step growth curves and plaque assay.
Uncloned CVB2O and viruses from passage seven (L_AD7) were analysed for growth properties by infecting GMK cells with 0·1 TCID₅₀ per cell (in duplicate). Virus was allowed to attach to the cell surface for 1 h. Unbound virus was then removed and cells were washed twice with Dulbeccos modified Eagles medium (DMEM) supplemented with penicillin, streptomycin and L-glutamine. Serum-free medium was then added and samples were incubated at 37 °C in 7·5% CO₂ for 0, 6, 12, 24, 48 and 72 h prior to freezing at -20 °C. All samples were titrated on GMK cells after three cycles of freezethawing. The plaque morphology of uncloned CVB2O and clone-derived virus propagated for seven passages was determined using plaque assays stained with crystal violet (Hierholzer & Killington, 1996 ).

Rapid amplification of cDNA endspoly(A) test (RACE-PAT).
The length of the viral poly(A) tail from viruses from different passage number was estimated by PCR amplification of the proximal 3' end using the RACEPAT method (Sallés et al., 1999 ). RACEPAT is based on PCR amplification using an oligo(dT) primer and a target-specific primer situated close to the poly(A) tail. Synthesized cDNA from L_B0L_B3 and L_B7 was amplified by PCR with the primer pair CB-44 and NotdT25. pCVB2O-8 containing the 17 A residue poly(A) tail was used as the control. PCR amplicons were analysed on a 2·5% agarose gel.

Genetic analysis of wt and cloned CVB2O 3'UTR
We had previously constructed an infectious CVB2O clone, pCVB2O-8 (Lindberg et al., 1997 ). The nucleotide sequence of pCVB2O-8 was determined and compared to the wt CVB2O sequence (Polacek et al., 1999 ) (differences shown in Table 1). The uncloned CVB2O genome is 7411 nt, which upon alignment with the 3'UTR of other enteroviruses shows an extension of five nucleotides (AGGAG) prior to the poly(A) tail (Fig. 2). The only other enterovirus with a similar feature is coxsackievirus A9, which has three additional nucleotides in the 3'UTR (Chang et al., 1989 ). The computer-predicted secondary structure of the CVB2O 3'UTR (Fig. 3A), determined using the STAR (Abrahams et al., 1990 ) and MFOLD (Zuker et al., 1999 ) programs, correlates with the experimentally determined structures reported for other viruses of the coxsackievirus B-like subgroup (Mirmomeni et al., 1997 ; Pilipenko et al., 1992 , 1996 ; Wang et al., 1999 ). The three characteristic hairpin domains, X, Y and Z, together with the partly poly(A) tail constituting domain S are all present in the predicted folding structure (Fig. 3A; nomenclature refers to Pilipenko et al., 1992 ). The X and Y domains consist of 8 and 12 bp, respectively, and are linked together by a 6 bp tertiary kissing interaction that forms the K domain as seen for other enteroviruses (Melchers et al., 2000 ; Wang et al., 1999 ). The interplay between the X and Y domains seems to be of vital importance to the virus, as covariation of their length is shown to be able to restore temperature sensitive (Ts) mutants to wt-like phenotypes (Melchers et al., 2000 ). For coxsackievirus B3, it has been demonstrated that not only the length but also the specific base pairing of the 3'UTR are a prerequisite for a functional oriR (Melchers et al., 2000 ). In the Y domain, the presence of two CG base pairs at the two distal positions is required to obtain a cis-acting replicative element as well as a UA base pair at the most distal position for domain X. These essential genotypic features correspond to all of the CVB2 3'UTR sequences analysed in this study (Fig. 3).

(73K): Fig. 2. Alignment of the 3' end of selected serotypes within the enterovirus genus. The complete 3'UTR has been aligned with ClustalX and the depicted region comprises the last 50 nt prior to the poly(A) tail. The alignment shows the protruding end of CVB2O and the conserved GUAAA sequence of human enteroviruses (shaded box). Serotypes and strains used in the alignment are as follows: CVB1, Japan (M16560); CVB2, Ohio-1 (AF085363); CVB3, Nancy (M33854); CVB4, JVB Benschoten (X05690); CVB5, Faukner (AF114383); CVB6, Schmitt (AF114384); CVA9, Griggs (D00627); Echo1, Farouk (AF029859); Echo5, Noyce (AF083069); Echo9, Barty (X92886); Echo11 (X80059); Echo12, Travis (X79047); PV1, Mahoney (J02281); CVA16, G10 (U05876); CVA21, Coe (D00538); EV-70, J670/71 (D00820); BEV-1, SL305 (AF123433); PEV-8, V13 (AJ001391); PEV-9, UKG/410/73 (Y14459). Asterisks indicate the conserved nucleotides in all serotypes.

(23K): Fig. 3. Predicted secondary and tertiary structures of the 3'UTR of uncloned and cloned CVB2O, including pseudorevertants generated from the lipofected clone pCVB2O-8. (A) Uncloned wt CVB2O, ΔG=-42·2 kcal/mol; (B) clone pCVB2O-8, ΔG=-30·8 kcal/mol; (C) pCVB2O-8 pseudorevertant, ΔG=-37·9 kcal/mol; (D) pCVB2O-8 pseudorevertant, ΔG=-36·4 kcal/mol. Free energy values were calculated by the MFOLD RNA secondary structure prediction program and the nomenclature of stems was adopted from Pilipenko et al. (1992) . Nucleotide positions are indicated and open arrows indicate the five protruding nucleotides of uncloned CVB2O. Filled arrows indicate the introduced point mutations and triangles indicate the site for the deletion in the clone. Shaded arrows show the compensatory mutations in the pseudorevertants.

In order to investigate the importance of the extension at the 3'UTR, we used the infectious clone pCVB2O-8, in which the last five nucleotides of the 3'UTR are deleted. pCVB2O-8 also contains two primer-induced point mutations upstream of the deletion (Table 3). The predicted 3'UTR structure of pCVB2O-8 shows an unfavourable conformation within the X-stem and contains two bulges that are not present in the wt strain (Fig. 3B). One disruption is caused by the 5 nt deletion and the other by the first primer-induced mutation. The second point mutation, which is located in a motif conserved among human enteroviruses, is situated in the bridging sequence between the X and Y stems and is not known to interact with any other part of the 3'UTR. The free energy of the wt and the pCVB2O-8 3'UTR structures was estimated using the MFOLD program and was found to be -42·2 and -30·8 kcal/mol at 37 °C, respectively. The considerable difference in energy status confirms the predicted secondary structure of pCVB2O-8 to be unfavourable.

Table 3. Sequences of the proximal 3'UTR of wt CVB2O, clone pCVB2O-8 and adapted revertants obtained after seven passages in GMK cells (LAD7)

Evolution and analysis of adapted revertants derived from pCVB2O-8
To study the significance of the five additional nucleotides preceding the poly(A) tail of the CVB2O 3'UTR, we used the combined point/deletion mutant clone pCVB2O-8 with disrupted folding. pCVB2O-8 was lipofected into GMK cells in four parallel experiments (L_AD) and the virus progeny were monitored and analysed for up to 12 passages (Fig. 1), each showing typical enterovirus CPE. Virus populations obtained from the different passages were collected, amplified by RTPCR and sequenced. The generated sequence was derived from a pool of genotypes where the consensus sequence describes the most frequently occurring nucleotide in each position. In order to determine if the virus population compensates for the introduced sequence alterations and how quickly these compensations occur, we have followed two experimental series for multiple passages (L_B03, 7 and L_C3, 7). During the lipofection stage and the first passage (L_B01), only pCVB2O-8-derived sequence could be detected. In the second passage (L_B2), revertant populations began to emerge and, during the third passage (L_B3 and L_C3), a predominant revertant genotype was demonstrated; this revertant was dominant in further passages (until L_B7 and L_C7). In agreement with these results and assuming that stable revertants had been established, we chose to study further the seventh passage of the four lipofection experiments (L_A7, L_B7, L_C7 and L_D7). A strong bias of selection that favoured either of the two main revertant genotypes was observed in all four experiments, where both genotypes compensate for the mutations introduced into the original pCVB2O-8 clone. The first lipofection experiment, L_A7, generated predominant revertants with a second-site mutation (₇₄₀₄C→G) to compensate for the primer-introduced mutation in the X-stem (Table 3; Fig. 3C). As can be seen in Fig. 3(C), this compensation closes the upper bulge of the X-stem and the free energy of this 3'UTR genotype was estimated to be -37·9 kcal/mol. In passages L_C7 and L_D7, revertants were found to compensate for the lower bulge of the X-stem (by ₇₄₀₈A→G), which was due to the 5 nt deletion (Fig. 3D). This structure showed a slightly lower free energy of -36·4 kcal/mol compared with the first revertant. The last revertants (serial L_B7) consisted of a heterogeneous pool and showed no single predominant population, although it had been propagated for seven passages (Table 3). The heterogeneity was evident following sequence analysis, illustrating that a mixture of compensatory mutations was present in L_A,CD7 (data not shown). In order to investigate whether this heterogeneity was due to viruses with one or both compensating mutations, individual virus clones were analysed. The 3'UTR of L_B7 RNA was amplified by RTPCR and inserted into pCVB2O-8. The 3' end of 20 individual clones was sequenced, revealing that 50% were the same type of revertant as those found in L_A7 (₇₄₀₄C→G) and 40% were the same as those found in L_C7 and L_D7 (₇₄₀₈A→G) (Table 3). No dual revertants (compensating for both mismatches in the X-stem) were isolated, although two unique clones, R₅ and R₇, were found (Table 3). The R₅ revertant had a ₇₄₀₄C→G compensatory mutation combined with a restoration of the introduced upstream mutation in the bridging sequence between the X and Y stems, while the R₇ revertant had a ₇₄₀₈A→G compensatory mutation but contained an additional deletion of ₇₄₀₆G. The calculated free energies of these revertants were -38·4 and -35·6 kcal/mol for R₅ and R₇, respectively. These clones were not included in further studies.

Nucleotide analyses show the evolution of a clone-derived virus with an unfavourable structure into a virus population with structurally ameliorated revertants. Growth properties of these revertants (L_AD7) were determined in comparison to the wt strain by a one-step growth curve over a 72 h period (Fig. 4). Plaque assay comparisons show that the revertants produced the same plaque size phenotype as the wt (data not shown). The overall growth rates in GMK cells were similar for the wt and the four revertant virus populations. This indicates that the compensatory mutations described in L_AD7 passages produce wt-like viruses.

(20K): Fig. 4. One-step growth curve of wt CVB2O and pCVB2O-8 revertants propagated for seven passages (LAD7). GMK monolayers were infected at timepoint -1 h and samples were collected at 0, 6, 12, 24, 48 and 72 h after the removal of unbound virus. All determinations of TCID50±SD were in quadruplicate.

Analysis of further compensatory mutations
The stability of the 3'UTR of the predominant genotypes was examined by following one of the experimental series (L_D) for a total of 12 passages. Nucleotide sequencing showed that L_D12 contained the same proximal 3'UTR sequence as L_D7, but the bridging sequence between the X and Y stems contained a mixed population. This variable coincided with the second primer-induced mutation and the population contained both clone-derived sequence (₇₃₈₃GGAAA₇₃₈₇) and wt sequence (₇₃₈₃GUAAA₇₃₈₇).

Extension of the pCVB2O-8 poly(A) tail
An additional observation made during the sequence analysis of the adapted revertants was the increase in the length of the poly(A) tail during subsequent passages. In the mutant clone pCVB2O-8, the poly(A) tail was immediately extended from 17 to about 40 A residues, as detected in L0. During sequence analysis of the passages, a continual increase in the poly(A) tail was detected. Since DNA sequencing of long homologous sequences has its limitations, the RACEPAT method was used to estimate the approximate length of the tail from each of the different passages (Sallés et al., 1999 ). The last 240 nt of the 3'UTR together with the poly(A) tail from viruses L_B03 and L_B7 were amplified by RACEPAT using pCVB2O-8 as a control (Fig. 5). The amplified clone appears as a sharp band, while the amplicons derived from progeny viruses appear as broad bands, indicating an increasing poly(A) tail. It can be concluded that the broadened bands arise from a heterogeneous population with different poly(A) tail lengths and, in part, from unspecific binding throughout the poly(A) tail. It is clear, however, that an increase in the poly(A) tail length is already detected in lipofected cells (L0) and that this increase proceeds until the poly(A) tail approaches optimal length, which is estimated in this study to be about 100 nt (Fig. 5).

(135K): Fig. 5. RACEPAT amplification of the poly(A) tail of virus progeny from different passages. Lanes M, size marker pBR322 cut with HinfI; C, control pCVB2O-8 with a 17 A poly(A) tail; 1, lipofected pCVB2O (LB0); 2, pCVB2O after one passage (LB1); 3, pCVB2O after two passages (LB2); 4, pCVB2O after three passages (LB3); and 5, pCVB2O after seven passages (LB7).

The purpose of this study was to investigate the importance of an unusual 5 nt extension in the 3'UTR of CVB2O. We have shown that deletion of these nucleotides in combination with point mutations affecting the 3'UTR secondary structure is compensated for in adapted viruses with wt-like growth properties. Both the 5' and the 3'UTRs possess complex secondary structures, although of different natures, implying different strategies in the initiation of replication. For human rhinovirus type 14, it has been demonstrated that there are several features of the 3'UTR that are critical for genome replication; a loop sequence, the stability of the stem and the position of the stem adjacent to the poly(A) tail (Rohll et al., 1995 ). These general features might be important for all picornaviruses to achieve efficient negative-strand RNA synthesis, although inefficient but still functional replication can be detected in poliovirus and human rhinovirus when the complete 3'UTR has been deleted (Todd et al., 1997 ). This indicates that enteroviruses are able to compensate for the loss of the 3'UTR but replication efficiency is severely impaired. Nevertheless, the predicted secondary structure of the 3'UTR is restored in revertants by point mutations and insertions, as shown in this report and by others, suggesting that a complete 3'UTR is advantageous for the virus (Melchers et al., 2000 ; Mirmomeni et al., 1997 ).

In this study, we have used a combined 3'UTR point/deletion mutant of CVB2O (pCVB2O-8) with a predicted unfavourable structure and containing two mismatches in the X-stem. Propagation of pCVB2O-8 in cell culture generated revertants with more favourable RNA secondary structures compared with the initial genotype. The adapted revertants contained mutations located in the X-stem and could be divided in two types: (i) type 1 compensates for the point mutation and (ii) type 2 compensates for the structural effect of the deletion. These results indicate that the stability of the X-stem is essential for CVB2O replication, since the virus restores either of the two structurally unfavourable mutations in all cases. However, no viruses were isolated that compensated for both of the mismatches in the X-stem. Among the revertants, no preference could be observed for the restoration of either the lower or the upper part of the X-stem, since both unfavourable positions were compensated for equally. In fact, the two genotypes were simultaneously maintained in equal amounts in one of the lipofection experiments (L_B). Wt CVB2O has a perfectly matching X-stem but revertants contain a single mismatch in this structure, as seen for coxsackievirus B3 and B4, where a single mismatch is present in the X-stem (Melchers et al., 1997 ; Pilipenko et al., 1992 ). The mismatches or internal loops reported are always symmetrical and hence exert a relatively mild destabilizing effect. Compared with our data, this could indicate that a certain level of stability is required for virus function but that a lower match throughout the X-stem is accepted and may even be favourable. The single ₇₃₉₅CG₇₄₀₃ base pair present at the top of the X-stem in both the clone and the revertant type 2 (Fig. 3 B, D) might be formed by coaxial stacking of the X and K domains (Melchers et al., 2000 ). Our data suggest that CVB2 does not accept more than one internal loop and that the position of the loop within the stem is of secondary importance.

The first revertant populations occurred after two passages in cell culture and, after seven passages, a predominant revertant genotype was present showing growth rates and plaque formation equivalent to the wt. The more advantageous predicted secondary structure of the 3'UTR seems to be favoured, as the emerging virus population with compensatory mutations rapidly out-competes the initial clone-derived population.

The second introduced point mutation disrupted a sequence that is completely conserved among human enteroviruses. This sequence, ₇₃₈₃GUAAA₇₃₈₇, is present as a bridge between the X and Y stems (Fig. 2). These nucleotides appear to be rather exposed in the predicted three-dimensional structure of coxsackievirus B3 (Melchers et al., 1997 ) but they have not yet been reported to interact with RNA or proteins. The introduced GGAAA mutation was compensated for in only 1 of 20 individually analysed virus clones from passage seven (R₅) (Table 3) and it is not until further passage that a detectable population compensating for this mutation emerges. Since the GUAAA motif is completely conserved among human enteroviruses and partly in the bovine and porcine enteroviruses (Fig. 2), it is surprising that restoration of this highly conserved sequence seems to be less important than the other introduced mutations.

Mutational changes in the 3'UTR may also induce single compensatory coding changes in the active site of the viral 3D polymerases, as has been shown for rhinovirus and poliovirus (Meredith et al., 1999 ). In this experimental system, no sequence alterations in the active-site cleft of 3D (L_AD7) were observed, indicating that the deletions/mutations introduced at the 3'UTR are not involved in the immediate interaction with the active site cleft of 3D in CVB2O (data not shown).

The increase in the size of the CVB2O 3'UTR might be considered to be only four nucleotides, since the first of the five residues reported in this study is an A residue, as in the following poly(A) tail. However, since it is followed by additional nucleotides in both CVB2O and CVA9, we chose to consider it as a genomic extension present prior to the poly(A) tail.

During our study of the 3'UTR, we observed an increase in the length of the poly(A) tail of the virus progeny after lipofection of the cDNA clone. In our experiments, we used a cDNA clone containing a 3' terminus with 17 A residues: this is in the lower range of the number of A residues considered to be necessary for picornavirus infectivity (Cui & Porter, 1995 ; Spector & Baltimore, 1974 ). Sequencing of the virus progeny from the lipofected cells (L0) revealed an increase in the length of the poly(A) tail from 17 A residues to more than 40. Using a PCR-based method of analysis, the poly(A) tail was shown to increase further during the first number of passages to a maximal length of about 100120 residues, which was then maintained during subsequent passages; this corresponds to the previously reported poly(A) tail length of polioviruses (Spector & Baltimore, 1975 ; Yogo & Wimmer, 1972 ). The poly(A) tail is required for infectivity and is also involved in the spatial organization of the 3'UTR by partly forming a portion of the S-stem and therefore constitutes an important part of the genome. Since no cellular adenylation signals (Zhao et al., 1999 ) were found in the 3'UTR, it must be concluded that picornaviruses might use an alternative polyadenylation mechanism to the one that is known today.

The terminal sequence and folding of the wt CVB2O 3'UTR seems to form a rigid structure involving the five additional nucleotides. The variable base pairing in the S-stem seems to be adequate for the secondary structure of this domain and no mutations emerged in this region, under our experimental conditions. Instead, the main differences were observed in the X-stem. Introduced mutations in the CVB2O 3'UTR were neutralized by compensatory mutations that restored at least part of the X-stem. These compensations are, in this experimental model, point mutations of nucleotides in the X-stem or nucleotides constituting the first part of the poly(A) tail. Revertants (adapted and with wt-like propagation properties) used only one of these compensatory strategies. As the 3'UTR is believed to be involved in the synthesis of negative-strand RNA, further studies of the CVB2 3'UTR should be related to the efficiency of replication of the complementary RNA strand.

We are grateful to Willem Melchers for valuable help with the STAR program and Ian Nicholls for reviewing this manuscript. This research has been supported with grants from the University of Kalmar.