INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The neurovirulence of a virus depends on a variety of genetic determinants. In picornaviruses, naked icosahedral viruses with a single-stranded 7.5- to 8-kb RNA genome of positive polarity (25), a set of such determinants maps to the 5' untranslated region (5UTR) of the RNA and is related to the translational control (26, 27; reviewed in references 2 and 3). Numerous studies demonstrated that modifications of these determinants may result in different levels of attenuation of viral pathogenicity. The major aim of the present report is to document that certain point mutations within a translational cis-acting element in the 5UTR of a picornavirus, Theiler's murine encephalomyelitis virus (TMEV), not only lower the viral virulence quantitatively but may result in distinct clinical patterns.

Under natural conditions, TMEV normally causes asymptomatic enteric infection. However, upon intracerebral inoculation of mice, it exhibits strong neuropathogenicity. Depending on the strain, the infection results in either an acute paralytic and lethal encephalomyelitis (e.g., in mice infected with the GDVII strain) or a persistent demyelinating disease (BeAn or DA strains), with this difference being attributed to point mutations in viral capsid proteins (references 1, 10, 16, and 18 and references therein). TMEV is widely used as a model pathogen to study a variety of aspects of viral neurovirulence as well as pathogenesis of human diseases, e.g., multiple sclerosis (7, 17, 20, 28).

The reproduction of TMEV, like that of other picornaviruses, involves 5' cap-independent internal initiation of translation promoted by a cis-acting element, the internal ribosome entry site (IRES) (reviewed in references 2, 9, and 29). This highly structured (21) element, which is several hundred nucleotides long, is located within the viral RNA 5UTR. The ribosome (or its 40S subunit) is believed to bind the IRES and to form a productive contact with a downstream segment, the starting window, of the RNA template (23). In TMEV, the starting window comprises about 12 nucleotides (nt) and lies about 17 nt downstream of the IRES (23). After contacting the starting window, the ribosome begins translation of the viral polyprotein from an initiator AUG that lies within the window and has a favorable context (purine residues at positions 3 and +4 relative to the first nucleotide of the AUG).

To define the properties of the starting window, AUG-lacking insertions between the TMEV (GDVII) IRES and the initiator codon have been engineered (23). These insertions, while displacing the latter from the starting window, did not markedly affect viral reproduction in BHK-21 cells or translation of its RNA in some cell-free systems. The ribosomes appeared to scan the template from the starting window until a downstream good-context AUG was encountered (23). On the other hand, the same insertions resulted in a dramatic loss of neurovirulence (24). The viruses isolated from the brains of rare paralyzed animals exhibited virulent phenotypes and contained either deletions or AUG-generating point mutations, both returning an AUG to the starting window (24). Neurovirulence was nearly fully or partially restored when an AUG in a favorable or unfavorable context, respectively, was introduced into the starting window (24). While the good-context AUG slightly extended the open reading frame (ORF) of the viral polyprotein by adding nine in-frame codons, the poor-context AUGs of these mutants could not serve for polyprotein initiation because an in-frame termination codon was placed between this triplet and the polyprotein ORF. In the course of that study, it was noticed that the clinical patterns elicited by mutants with different poor AUG contexts may differ from one another. This notion, if correct, should have important implications. It may suggest that point mutations within a control element of the 5UTR may modulate clinical signs of a neurologic disease. Furthermore, laboratory characterization of pathogenic strains may in this case require exact knowledge of the primary structure of a certain region of the 5UTR. Taking into the account these considerations, we constructed a novel set of GDVII TMEV mutants having different unfavorable contexts of the AUG within the starting window followed by a stop codon. These mutations could be expected to affect the formation of the ribosome-template productive complex, whereas the viral polyprotein should be initiated at its cognate site. The mutants proved to be similar to each other and to the wild-type (wt) virus in their in vitro phenotypes and were all attenuated to a certain extent. Remarkably, these mutants differed markedly from each other in the clinical signs of the disease they induced, even though the paralytogenic doses (doses causing paralysis in 50% of the animals [PD₅₀s]) for some of these mutants were almost the same. Thus, point mutations within the starting window outside the polyprotein ORF may differently affect the morbidity and mortality caused by a viral infection and may result in distinct attenuation phenotypes.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Mice. BALB/c, CBA, and C57/BL mice were obtained from the Animal House of the Russian Academy of Medical Sciences (Stolbovaya, Moscow Region, Russia); SJL/J mice were from the Institut Pasteur (Paris, France).

Construction of mutant viruses. Plasmid GD-43 and the corresponding mutant virus (mutant 43) have been described previously (23). New mutants were engineered by oligonucleotide-directed mutagenesis with the pUGD/I.27-40 intermediate construct, and the viruses were recovered from transcript RNA-transfected BHK-21 cells as described previously (23). In vitro and in vivo viral phenotypic properties were determined as described previously (24).

Construction and translation of luciferase-expressing plasmids. The pGLV intermediate construct was generated by ligation of a HindIII/BamHI fragment of GD-2 (23), containing a pUC vector sequence and nt 1 to 1047 of the GDVII cDNA under the control of a T7 promoter, to a BamHI/HindIII fragment of pBLC (a plasmid kindly donated by T. D. K. Brown), containing the luciferase cDNA sequence (nt 22 to 1841) and a polylinker. PCR-amplified TMEV DNA fragments (nt 779 to 1122) were derived from the wt and mutant plasmids. KpnI-digested fragments (nt 937 to 1122) were ligated to pGLV, which was linearized with BamHI (at position 1047), filled in with the Klenow enzyme, and digested with KpnI (at position 937). The resulting pLG constructs contained, under control of the T7 promoter, the entire wt or mutant GDVII 5UTRs and an ORF starting at the TMEV initiator AUG₁₀₆₈ followed by 28 codons, mostly of viral origin, fused in frame to the luciferase gene.

Characterization of viral genomes in mouse brains. The presence of the viral genome in the central nervous system (CNS) was evaluated by dot blot hybridization. Anesthetized mice were perfused through the left ventricle with 20 ml of phosphate-buffered saline (PBS), and their brains and spinal cords were immediately removed. Total RNA was extracted (6) and quantified spectrophotometrically. Four serial fivefold dilutions of total RNA from the brain and spinal cord of each mouse were dot blotted on Hybond C-extra filters (Amersham), starting with 10 µg per dot, and hybridized with a ³²P-labeled DNA probe corresponding to the viral sequence from nt 5437 to 6914 generated by random priming. Hybridization of the same material with a -actin probe was carried out as a control.

Histopathology and immunofluorescence. Anesthetized mice were perfused with 20 ml of PBS followed by 20 ml of 4% paraformaldehyde in PBS. Dissection of the CNS, postfixation, and paraffin embedding were performed as described previously (4). Sections of brains and longitudinal sections of spinal cords were examined for histopathology by hematoxylin staining. Viral antigens were detected with a rabbit anticapsid serum, a secondary biotinylated goat anti-rabbit immunoglobulin, and the LSAB peroxidase kit (DAKO, Carpinteria, Calif.).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Genome structures of the mutants and their in vitro properties. The structure of the relevant region of the TMEV (GDVII strain) genome is presented in Fig. 1. All of the mutants used here contained a nonviral 27-nt insertion between the IRES and the initiator AUG₁₀₆₈, which by itself did not appreciably change the in vitro phenotype of the virus (23) but modulated its neurovirulence, depending on whether an AUG was present within the insert (24). The absence of AUG (mutant 40 [Fig. 1]) was accompanied by a nearly complete loss of virulence (log₁₀ PD₅₀ = 7.7), whereas the presence of AUG in an optimal context (aauAUGg; in mutant 44) resulted in no marked change compared to that of the wt GDVII (log₁₀ PD₅₀s of 3.7 and 2.5, respectively); a poor AUG context (uauAUGu; mutant 43) resulted in an intermediate level of attenuation (log₁₀ PD₅₀ = 5.2) (24). The suboptimal-context AUG within the starting window of the mutants could not serve to initiate the viral polyprotein, even inefficiently, because a terminator UGA was present 3 nt downstream.

View larger version (26K): [in this window] [in a new window]   FIG. 1.   Genome structures of GDVII and its derivatives. Dots in the GDVII sequence are for alignment. AUGs are underlined. For the mutants with the 27-nt inserts, the AUG and its variable contexts within the starting window are in boldface and the stop codon that follows is in italics.

Neurovirulence and clinical patterns. When tested for virulence by intracerebral inoculation of BALB/c mice, all of the mutants except mutant 83 could be assigned to a single set, having PD₅₀ values 10² to 10³ higher than that of the wt GDVII (Table 1). Despite this similarity, these mutants could be considered to belong to two groups that were dramatically different in the pattern of clinical signs they induced. Once paralyzed, a mouse inoculated with mutant 63 (cuuAUGg) was prone to develop, like a wt-infected mouse, a severe paralytic disease (e.g., tetraplegy) and had a very high chance of dying within 1 to 2 weeks postinoculation (p.i.) (generally 2 to 3 days after the appearance of the first clinical signs) (Fig. 2). Rare survivors did not recover from the neurologic disease during the observation period (7 to 10 weeks) (Fig. 3). A similar, although slightly milder, clinical pattern was characteristic of mice inoculated with mutant 65 (uauAUGg) (Fig. 3). On the other hand, the disease caused by the representatives of the second group of mutants, mutants 43 (uauAUGu) and 64 (uggAUGc), was less severe: the overwhelming majority (85%) of the diseased mice did survive, and a significant proportion of them appeared to completely recover by the end of the seventh week (Fig. 2 and 3). The rare deaths occurred only after many days of disease and was likely caused by revertants (see below).

View larger version (47K): [in this window] [in a new window]   FIG. 2.   Time course of clinical signs after intracerebral inoculation of mice with mutants 43 and 63 at 108 and 108.5 TCD50, respectively. Twenty-eight and 24 animals were infected simultaneously with mutants 43 and 63, respectively, and monitored individually. The numbers of mice that were paralyzed, died, and developed the second wave of paralytic symptoms are shown as gray, black, and white areas, respectively. Numbers on the x axis are days p.i.

View larger version (43K): [in this window] [in a new window]   FIG. 3.   Dependence of the outcome of infection by 7 to 10 weeks p.i. on the time of the appearance of the first clinical signs. The proportions of dead (black bars), irrecoverably paralyzed (gray), and fully recovered (white) mice are shown. Data were collected from several experiments. The numbers of animals that got sick during the first (1 w), second (2 w), and third (3 w) weeks, respectively, were as follows: 72, 70, and 26 (mutant 43); 66, 34, and 8 (mutant 63); 38, 35 and 10 (mutant 64); and 29, 9, 0 (mutant 65).

Viral antigen and RNA accumulation. The variation in the clinical patterns induced by the mutants correlated well with the viral antigen accumulation in the CNS. At days 5 to 6 p.i., numerous antigen-containing cells, apparently neurons, could be seen in the gray matter of the brains and spinal cords of GDVII- and mutant 63-infected SJL/J mice, whereas such cells were much less abundant in the diseased mice inoculated with mutants 43 and 64 (Fig. 4). The viral antigen disappeared nearly completely by 8 to 11 days. Very few, if any, antigen-containing cells could be detected in the white matter of the surviving mice at any time of infection with any of the viruses investigated.

View larger version (145K): [in this window] [in a new window]   FIG. 4.   Histological findings for SJL/J mice after inoculation with 105 PFU of wt and mutant viruses. Longitudinal sections of the spinal cords of the infected animals were prepared on day 5 for GDVII (A) on day 6 for mutants 63 (B), 64 (C), and 43 (D), and on day 22 for mutant 43 (E and F). Viral antigens were detected by the immunoperoxidase assay with a polyclonal anti-TMEV rabbit serum (brown clusters). Virus-infected cells were localized in the gray matter and were more abundant in the GDVII- and mutant 63-infected animals than in the mutant 43 and 64 infections. In the case of mutants 43 and 64 (but not in the case of GDVII and mutant 63), there were marked signs of inflammation in the white matter (arrows) even at day 22 p.i., when no viral antigen was detectable (E and F). Magnification, ×312.

View larger version (39K): [in this window] [in a new window]   FIG. 5.   Viral RNA in the CNSs of infected mice as assayed by dot blot hybridization. GDVII- and mutant-infected mice were sacrificed on days 5 and 6 p.i., respectively. Total cytoplasmic RNA was extracted from the brains and spinal cords of infected mice, loaded on the membrane (in the amounts [in micrograms] indicated at the top) and hybridized with 32CTP-labeled PCR-amplified, virus-specific DNA. The numbers on the left correspond to individual mice.

Pathological findings. Early upon infection with GDVII or mutant 63, there was a weak inflammatory reaction in the gray matter and meningea and nearly no such reaction in the white matter (Fig. 4). However, with mutants 64 and, especially, 43 there were marked signs of inflammation in the white matter (Fig. 4), even though it contained very little, if any, detectable viral antigen. In the gray matter, the distributions of the antigen-containing cells and inflammatory loci mostly did not coincide with each other either. Marked inflammation in the white matter of mutant 43-infected mice could be seen at day 22 (when the antigen was no longer detectable even in the gray matter) (Fig. 4) and at day 45, although at this time it was less pronounced (not shown). A qualitatively similar inflammatory reaction was also characteristic of mutant 64-infected mice (not shown). Comparable, although somewhat milder, patterns of distribution of antigen-containing cells and inflammation were found also in the CNSs of the mutant-infected BALB/c mice (not shown).

Genome structures of the viruses in the CNSs of diseased mice. Sequencing of the relevant region of the viral genome revealed no additional mutations (other than those engineered) in the RNAs of the more virulent mutants 63 and 65 in the severely paralyzed or dead animals (Table 2). In the case of mutants 43 and 64, mice with a typical (mild) clinical pattern also harbored the parental viral genome, whereas pseudoreversions (deletions) returning the cognate initiator AUG to the starting window were found in the CNSs of the mice exhibiting more grave clinical signs or dead animals, as revealed by sequencing (Table 2) and RT-PCR length polymorphism (not shown).

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The results described above demonstrate that the target cells in the CNS, presumably neurons, may discriminate between different nonoptimal contexts of an upstream AUG within the starting window, even though no marked discrimination could be demonstrated in either BHK-21 cells or some cell-free translation systems. Most likely, this discrimination occurs at the level of translation initiation, but the mechanism of the AUG context recognition remains unknown. If the poor-context AUG within the insert was still used to some extent as an initiator codon, only a dipeptide could have been formed due to the neighborhood of a terminator UGA. As is generally accepted (5, 8, 15), initiation at an upstream AUG should interfere with the expression of the polyprotein ORF. Accordingly, the closer to the optimal is the context of the upstream AUG within the insert, the less efficient is the viral reproduction that could be expected. The very high level of attenuation of mutant 83 (aauAUGu), which has a purine (A) at the 3 position, complied with this expectation, since this position is considered to have a dominant effect among other positions of the AUG context and the A₃ residue endows the context with the highest strength (13).

The properties of other mutants, however, appeared to contradict this rule. Mutants with a 3 pyrimidine but having a purine (G) residue in the +4 position of the upstream AUG elicited a more severe clinical pattern than the mutants with a less favorable pyrimidine (C or U) residue. This fact suggests that upstream AUGs with a stronger context may, contrary to expectations, exert less interference with the polyprotein translation. Although the mechanism ensuring such an apparently paradoxical sensitivity of the translation initiation machinery of relevant CNS cells to the nucleotide in the +4 position of an upstream AUG is unknown, two hypothetical explanations may be put forward. Either the noninitiating AUGs with a G₊₄-containing context within the starting window facilitate formation of a productive ribosome-template complex (23), or such a context promotes synthesis of the dipeptide followed by reinitiation at the cognate initiator AUG, as is known to occur in some cases after translation of an upstream ORF (12, 14, 19).

Whatever the molecular mechanism, the finding of distinct attenuation phenotypes caused by changing the context of an upstream AUG within the translational starting window was an important and unexpected observation. Especially intriguing were the apparently separate effects of appropriate mutations on the morbidity (as judged by PD₅₀s) and mortality rates caused by the mutant-induced infections. We are not aware of a similar phenotypic effect of any other mutations in any other neurovirulent viruses. Although the difference between the clinical phenotypes of the mutants could partly be attributed to a likely slower growth and/or spread of the mutants with a +4 pyrimidine (mutants 43 and 64) compared to G₊₄-containing mutants (mutants 63 and 65), several lines of indirect evidence suggested that some qualitative rather than merely quantitative distinctions in the virus-host interaction were also involved. Thus, the probability of apparent recovery of mutant 43- and 64-inoculated mice did not decrease with a shorter incubation period, suggesting that a relatively benign progression of the infection was an inherent property of these viruses. The enhanced immune response could hardly fully explain the relative mildness and reversibility of neurologic symptoms induced by the viruses, since the rare emergence of more virulent revertants resulted in severe, irrecoverable paralytic disease. The significant inflammatory reaction in the white matter of mutant 43- and 64-infected mice in the absence of detectable viral antigen (an intriguing and enigmatic finding) could well contribute to the clinical manifestations and might, for example, be responsible for the disproportionately low PD₅₀s and the second wave of paralytic symptoms observed in some cases.

Another important aspect of this study concerned the absence of persistence, which is characteristic of BeAn or DA strains, in mice infected with attenuated, nonlethal mutants of the GDVII strain of TMEV, a notion discussed in more detail elsewhere (11, 18). This finding is in line with previous results demonstrating that the relevant capacities of BeAn and DA strains of this virus are due largely, if not exclusively, to point mutations in the viral capsid (1, 10, 16). Thus, the same virus can induce quite different diseases depending on minute changes in coding (GDVII versus BeAn and DA strains) and noncoding (GDVII and mutants 63 and 65 versus mutants 43 and 64) parts of its genome. Such a variability in the mortality, morbidity, and clinical patterns elicited by virus variants differing from one another by only a few point mutations spread over the viral genome is certainly not a unique property of TMEV. It can likely be extrapolated to different viruses that are pathogenic for humans and should be taken into consideration in studies on viral pathogenesis as well as in laboratory identification of viruses.