Dual Stem Loops within the Poliovirus Internal Ribosomal Entry Site Control Neurovirulence

Abstract

In the human central nervous system, susceptibility to poliovirus (PV) infection is largely confined to a specific subpopulation of neuronal cells. PV tropism is likely to be determined by cell-external components such as the PV receptor CD155, as well as cell-internal constraints such as the availability of a suitable microenvironment for virus propagation. We reported previously that the exchange of the cognate internal ribosomal entry site (IRES) within the 5′ nontranslated region of PV with its counterpart from human rhinovirus type 2 (HRV2) can eliminate the neuropathogenic phenotype in a transgenic mouse model for poliomyelitis without diminishing the growth properties in HeLa cells. We now show that attenuation of neurovirulence of PV/HRV2 chimeras is not confined to CD155 transgenic mice but is evident also after intraspinal inoculation into Cynomolgus monkeys. We have dissected the PV and HRV2 IRES elements to determine those structures responsible for neurovirulence (or attenuation) of these chimeric viruses. We report that two adjacent stem loop structures within the IRES cooperatively determine neuropathogenicity.

Tissue tropism and pathogenic properties of a virus are determined largely by the cellular receptor(s) and an intracellular milieu suitable for viral proliferation. Poliovirus (PV), the prototype of the genus Enterovirus of the family Picornaviridae, has a markedly restricted tissue tropism that is limited to unidentified cells within the gastrointestinal tract and to motor neurons in the spinal cord and brainstem. Neurological disease, culminating in flaccid paralysis, evolves with the progressive destruction of motor neurons by PV in the spinal cord anterior horn, a condition called poliomyelitis (4, 31).

Studies of the attenuated PV vaccine strains of Sabin (32) have revealed that, remarkably, some of the attenuating mutations map to a specific region in the 5′ nontranslated region (5′NTR) of the viral genome (6, 16, 26). This region was subsequently identified as domain V of the internal ribosomal entry site (IRES) of PV (reviewed in reference 41).

IRES elements are 400-nucleotide-(nt)-long segments within the 5′NTR of picornavirus genomic RNAs that allow initiation of protein synthesis independently of a 5′ end (14, 27) (Fig. 1). These genetic entities, which are a signature not only of picornaviruses (40) but also of hepatitis C virus (39), have been recognized by their function, not by their structure. Indeed, IRES elements of different viruses may have only minor, if any, homology (28, 29), yet they are interchangeable from virus to virus, leading to novel chimeric infectious agents (2, 7, 22). The mechanism by which IRES elements function remains obscure.

FIG. 1.

PV1/HRV2 chimeras featuring heterologous or synthetic IRESes of various composition. The cognate IRES of PV1(M) (A) was exchanged with its counterpart from HRV2 as described previously [B; PV1(RIPO) (7)]. All viruses feature the wt PV1(M) ORF with the exception of C [PV1(RIPOS)], which was engineered to contain all mutations within the capsid encoding region (P1) specifying PV1(S). The cloverleaf (5′NTR domain I) was derived from PV1(M) in all viruses. All featured IRESes in a PV context were tested for their propensity to mediate neurological disease after intracerebral (i.c.) inoculation into CD155 tg mice (third column). A horizontal bar indicates that intracerebral inoculation of 10⁹ PFU of the variant in question did not elicit any neurological symptoms in CD155 tg mice.

The attenuating point mutations in the IRESes of the Sabin vaccine strains (nt 480, 481, and 472 in types 1, 2, and 3, respectively; see Fig. 4A) each map to domain V of the IRES. Their presence has been correlated with reduced translational efficiencies of the corresponding viral mRNAs compared to wild-type (wt) templates (36–38). It was suggested that this may be the mechanism by which these mutations confer an attenuating (att) phenotype to the Sabin strains (40). Interestingly, infection of neuroblastoma cells could differentiate between wt and att phenotypes (1, 20). These studies led to the notion that domain V of the PV IRES is a major determinant of neurovirulence, but the mechanism remained unsolved. The contribution to att of these IRES mutations relative to mutations downstream of the IRES (the open reading frame [ORF] of the polyprotein) has been recently a matter of debate; however, the importance of the viral capsid in the expression of the att phenotype of the Sabin strains has been stressed (5, 23).

FIG. 4.

Construction of PV1(M)/HRV2 intradomain chimeras featuring heterologous uppermost regions of domains V and VI linked to the stem structures originating from HRV2 and PV1(M), respectively. Sequences from PV1(M) are shown on a white background; structures derived from HRV2 are boxed in light gray. (A) The 5′NTR of PV1(RIPO) with heterologous distal stem loop structures V and VI derived from PV1(M) yield PV1(rpp). The nucleotide numbering of the PV1(M) IRES was applied to the corresponding HRV2 sequence. Asterisks indicate the positions of single point mutations implicated in the att phenotype of the Sabin vaccine strains (see text). A KpnI endonuclease restriction site introduced for cloning purposes is outlined by a white box. Each dark gray box delineates the position of a silent AUG triplet conserved among all enterovirus and rhinovirus strains. Recombinant IRES A [PV1(rpp)] features the authentic initiating AUG triplet of PV1(M) separated from domain VI by a 154-nt stretch. In construct B, deletion of the 154-nt spacer separating the IRES from the initiating AUG triplet in PV1(rpp) yielded PV1(rpr), creating an AUG codon within an optimal Kozak context (…AnnAUGG…) characteristic of HRV2. For construct C, a synthetic IRES converse to B [PV1(rpr)] was generated by exchanging those nucleotides within the upper stem loop regions of IRES domains V and VI of PV1(M) outlined by light gray boxes with their counterparts from HRV2, yielding PV1(prr).

We previously studied a chimeric virus, PV1(RIPO), that possesses the genotype of PV type 1 (Mahoney) [PV1(M)] except that the cognate IRES plus spacer sequence (40) was replaced by the corresponding elements of human rhinovirus type 2 (HRV2) (Fig. 1B). PV1(RIPO) replicates in HeLa cells with PV1(M) kinetics but is unable to proliferate in human neuroblastoma SK-N-MC cells (7). We have also reported (7) that PV1(RIPO) has lost the neurovirulent phenotype, characteristic of its PV1(M) progenitor, in mice transgenic for the human PV receptor (CD155 tg mice [18]) despite the fact that the chimera grows in skeletal muscle of these transgenic animals or in mouse L cells expressing CD155 (7).

Lack of replication of PV1(RIPO) in cells of neuronal origin may offer an opportunity to construct derivatives of this or of related chimeric viruses that could be used as vectors to target motor neurons in human gene therapy. It was of great interest, therefore, to test PV1(RIPO) for its neurovirulence phenotype in nonhuman primates under conditions at which the attenuated PV Sabin vaccine strains are analyzed. Moreover, we have studied the genetic determinants responsible for the att phenotype of PV1(RIPO), aiming ultimately at elucidating the mechanism of tissue tropism restriction induced by IRES elements. The data have revealed an unexpected complexity of the structural basis of IRES-dependent pathogenicity. Previous studies of the molecular genetics of IRES elements have made extensive use of gene expression systems measuring IRES-mediated reporter gene expression either in cell-free translation or in tissue culture cells. Our studies of IRES genetics included the analysis of the influence of specific IRES sequences on the pathogenic phenotype in experimental animals.

MATERIALS AND METHODS

Construction of PV/HRV2 IRES chimeras.

IRES chimeras were constructed by using a cloning cassette to exchange the cognate IRES of PV1(M) with heterologous IRES elements of various compositions (7). The following primers were used to generate composite IRES elements recombining stem loops of HRV2 with those of PV1(M) (nucleotide sequence numbering is that assigned to the PV1(M) IRES sequence (17)]: 1, ccgaattcaacttagaagtttttcacaaag; 2, gggaattcagacgcacaaaaccaag; 3, ccggatcctccggcccctgaatgcg; 4, ccggatcctccggcccctgaatgtgg; 5, cc ggatccttatgtagctcaatagg; 6, ccggatccaaagcgagcacacggggc; 7, ccggtaccgcttatggtgacaatcacag; 8, gcggtaccgcttatggtgacaatatatac; 9, gcggtaccaataaaataaaaggaaacacggacacc; 10, ccggtacctaaaggaaaaagtgaaaca; 11, cctgagctcccattatgatacaattgtctg; 12, cctgagctcccatggtgccaatatatatattg; 13, ggccaatcactggtttgtgaccaccagctgcagggttaagg; 14, ggccagtgattggccagtcgtaatgagc; 15, gggagctcccatgataacaatctgtg; 16, ggttacgtgctcttgctccgaggttggg; and 17, ggcacgtaacccaatgtgtatcttgtcgtaacgcgc.

For PV1(R2-4,6), primers 1 and 6 were used to generate a PCR fragment encompassing domains II to IV of the HRV2 IRES, which was ligated to a PCR fragment obtained from PV1(M) by using primers 3 and 9 corresponding to PV1(M) domain V and to a PCR product from HRV2, yielding domain VI, with the use of primers 8 and 12. For PV1(R2-5), primers 1 and 10 were used to generate a fragment containing domains II to V of HRV2 that was ligated to domain VI of PV1(M), PCR synthesized with the use of primers 7 and 11. PV1(R5-6) was generated by ligating a PCR product from PV1(M) by using primers 2 and 5 with a PCR product encompassing domains V and VI from HRV2 generated with primers 4 and 12. A PCR product from PV1(M) generated by using primers 2 and 9 was ligated to domain VI of HRV2 obtained by PCR with primers 8 and 12. PV1(rpr) and PV1(rpp) were cloned as follows. An IRES fragment encoding nt 484 to 508 (the tip of domain V) from PV1(M) on top of the ascending stem of domain V derived from HRV2 was generated by using HRV2 cDNA as a template with primers 1 and 13. This fragment was ligated to a PCR product resulting from priming with primers 14 and 10, yielding the descending branch of stem loop V and poly(U) tract from HRV2 [with nt 484 to 508 from PV1(M)]. Domain VI was derived from PV1(M) by PCR with primers 7 and 11 or with primer 15. Use of primer 15 in the latter reaction yielded PV1(rpr), deleting the long terminal stretch of the IRES absent among rhinoviruses and putting the initiating AUG in a rhinovirus context. Primer 11 resulted in PV1(rpp), conserving the terminal IRES stretch and the cognate PV1(M) AUG (see Fig. 4). PV1(prr) was generated by PCR with primers 2 and 16 from template PV1(M) cDNA to yield IRES domains II to V (ascending loop) and primers 17 and 12 from template PV1(R6) cDNA to yield IRES domains V (descending loop) from PV1(M) and domain VI from HRV2. The resulting chimera PV1(prr) was the reverse of PV1(rpr), featuring the upper loop of domain V (nt 484 to 508) and domain VI derived from HRV2 in a PV1(M) background (see Fig. 4). Each chimeric IRES construct was subjected to MFOLD analysis to determine the predicted secondary structure. We found that overall IRES structure predicted for enteroviruses and rhinoviruses alike was not affected by the intergeneric exchange of intact stem loop domains or subfragments thereof described in this study.

All composite IRESes were inserted into a cloning cassette yielding full-length viral cDNA, which was then linearized and used as a template for in vitro transcription (7). Viral RNA thus obtained was transfected into HeLa cells to produce infectious viral particles (7).

Virus propagation, cell culture, and growth curve assays.

Viral recombinants were grown and isolated as described previously (7). One-step growth curves in HeLa cells and SK-N-MC neuroblastoma cells were established as follows. Cell monolayers were washed with Dulbecco’s minimal essential medium and overlaid with a solution containing the recombinant virus to be tested at an multiplicity of infection of 10. After the dishes were rocked for 30 min at room temperature, the cells were thoroughly washed to remove unbound virus and placed at 37°C. Sample dishes were removed for quantitative assessment of viral reproduction at the specified intervals. Dishes were subjected to three consecutive freeze-thaw cycles, and the lysate was analyzed with a plaque assay as described before (7).

Neurovirulence assays in experimental animals.

Neurovirulence testing in CD155 tg mice has been described before (7). For monkey neurovirulence assays, virus strains to be tested were plaque assayed on primary cultured Cynomolgus monkey kidney cells. Cynomolgus monkeys were inoculated intraspinally with 10⁶ 50% cell culture infectious doses per ml. Monkeys were sacrificed 17 days after intraspinal inoculation of virus, and the extent and distribution of spinal histopathology were assessed in a series of sections in a manner described before (26). Lesion scores were determined by established procedures (16, 43).

RESULTS

Neurovirulence of PV1(RIPO) and PV1(RIPOS) in Cynomolgus monkeys.

Following the standardized protocols of testing PV Sabin strains (oral PV vaccines) for neurovirulence, we analyzed PV1(RIPO) (Fig. 1B) for its att phenotype by intraspinal inoculation of Cynomolgus monkeys (16, 26, 43). In addition, to assess the possible contribution of the Sabin capsid to attenuation in the PV1(RIPO) background, we constructed a chimeric virus related to PV1(RIPO) in which the coding region for the capsid proteins (P1) was that of PV1(Sabin) [PV1(RIPOS) (Fig. 1C)]. Both PV1(RIPO) and PV1(RIPOS) were highly attenuated, as histopathological analyses revealed lesion scores comparable to those typically associated with PV1(S) (Table 1). Apart from transient minor manifestations in the form of paresis of the foot in two animals, neurological disease did not appear in monkeys that had received intraspinal inocula of either virus (see Materials and Methods). No real differences in the level of attenuation between the two strains were apparent, an observation suggesting that a high degree of attenuation of neurovirulence is attained by the HRV2 IRES alone, without a contribution of the Sabin capsid. These results show that the attenuation of neurovirulence associated with PV carrying an HRV2 IRES, originally observed in CD155 tg mice, extends to primates.

TABLE 1.

Neurovirulence testing in Cynomolgus monkeys

Monkey no.	Virus strain tested	Lesion score	Clinical observation
1	PV1(RIPOS)	0.60	No paralysis
2		0.70	No paralysis
3		0.62	Partial paralysisa
4		0.60	Partial paralysisa
5	PV1(RIPO)	0.70	No paralysis
6		0.0	No paralysis
7		0.40	No paralysis
	PV1(S)	0.92b
	PV1(M)	2.48c

^aTransient paresis of the foot monolaterally was noticed as a consequence of virus inoculation. 

^bMean value from a series of treated animals (16). 

^cMean value from a series of treated animals (26). 

Exchanges of domains V and VI of PV and HRV2 IRES elements.

Replacement of IRES domains V and VI in PV1(RIPO) with the corresponding domains of PV1(M) yielded a viable virus with a chimeric IRES element [PV1(R2-4) (Fig. 1D)] whose neurovirulence in CD155 tg mice was restored (reference 7 and Fig. 1D). PV1(R2-4) grew with wt kinetics in HeLa and SK-N-MC tissue culture cells (data not shown). The converse recombinant virus PV1(R5-6) (Fig. 1H) expressed the expected phenotype: it was attenuated in transgenic mice (Fig. 1H) and failed to replicate efficiently in neuroblastoma cells (Fig. 2A) but retained efficient growth in HeLa cells (Fig. 2B). These observations suggested that domains V and VI may function independently of domains II to IV or, alternatively, that hypothetical interactions between domains II to IV and domains V and VI have been conserved among IRES elements of enteroviruses and rhinoviruses.

FIG. 2.

One-step growth curves of intergeneric IRES domain recombinants shown in Fig. 1. Growth curves were established as described before (7) in SK-N-MC cells (A) and HeLa cells (B). IRES constructs with intergeneric combined domains V and VI, viruses PV1(R2-4) (+), PV1(R2-4,6) (□), PV1(R2-5) (▴), PV1(R5) (■), PV1(R5-6) (▵), and PV1(R6) (×), failed to drive efficient PV replication in SK-N-MC cells, similar to the degree of growth restriction observed with the intact HRV2 IRES [PV1(RIPO) (○)]. Extensive domain swapping barely affected the ability of the recombinant IRES elements to function efficiently in HeLa cells [compare recombinant growth curves with that for PV1(M): (•)]. p.i., postinfection.

All of the single point mutations identified in the IRESes of the three Sabin strains that lower neurovirulence map to domain V (see Fig. 4A). It was previously concluded, therefore, that domain V alone may carry the determinants to yield neurovirulence or attenuation (41). We have tested this hypothesis in the context of our intergeneric IRES constructs by exchanging domain V of PV1(RIPO) from HRV2 to the PV1(M) sequence [PV1(R2-4,6) (Fig. 1E)]. Surprisingly, PV1(R2-4,6) did not express a neurovirulent phenotype in transgenic mice (Fig. 1E). Moreover, although growing well in HeLa cells, yielding titers >80% of wt PV1(M) titers (Fig. 2B), PV1(R2-4,6) did not replicate in SK-N-MC cells to titers >0.5% of PV1(M) titers (Fig. 2A). The converse recombinant virus, carrying a PV1(M) IRES of which domain V was changed to that of HRV2 [PV1(R5) (Fig. 1F)], showed a phenotype very similar to that of PV1(R2-4,6) (Fig. 1F and 2). These results can be interpreted to mean that domain V cannot be the sole determinant of neurovirulence or attenuation. We therefore conclude that a coordinate action between domains V and VI is likely to effect growth properties in cells of neuronal derivation and neuropathogenicity.

Analysis of chimeric IRESes suggested that heterologous structural elements of domains V and VI originating from either PV1(M) or HRV2 may be unable to cooperate in an as yet unknown manner in cells of neuronal origin, thereby severely impairing viral growth. This hypothesis was supported by the phenotypes of the converse chimeric viruses PV1(R2-5) and PV1(R6), whose IRESes differed by the nature of domains VI (Fig. 1G and I, respectively). Whereas these exchanges did not affect significantly the growth properties in HeLa cells (Fig. 2B), they greatly decreased replication in neuroblastoma cells (Fig. 2A). Fittingly, neither of these two viruses was neurovirulent in CD155 tg mice (Fig. 1).

Fine mapping of host range determinants in IRES domains V and VI.

Overall, the secondary structures of the IRESes of PV1(M) and HRV2 are very similar (21, 29, 34). Closer inspection of domains V and VI of the PV1(M) and HRV2, however, revealed significant differences. First, sequence alignments of these domains, shown in Fig. 3, reveal a high degree of homology in the lower stem regions of these domains, whereas the upper parts of the domains, comprising the region of and surrounding the loops, are quite heterologous (for example, nt 482 to 510 and nt 597 to 616 [Fig. 3]). Second, stem loop VI of HRV2 is slightly longer than that of PV1(M), the former containing a peculiar AU repeat in the extended part of the upper stem (compare Fig. 4B and C). Third, in HRV2, initiation of polyprotein synthesis commences at an AUG just 33 nt downstream of the silent yet absolutely conserved AUG codon, whereas this spacer in PV is 154 nt long (discussed in reference 41).

FIG. 3.

Sequence alignment of IRES domains V and VI of PV1(M) (17) and HRV2 (33). GenBank accession numbers for the sequences used are J02281 for PV1(M) and X02316 for HRV2. The stem loop structures indicated by the superimposed lines are based on published secondary structure calculations, biochemical probing, computational methods, and comparative sequence analysis (21, 29, 34). Shaded areas delineate segments of homology between PV1(M) and HRV2. Note that extensive homology is present between the IRES domains of these virus species in areas forming the ascending and descending lower stems of both domains V (nt 448 to 482 and 510 to 562) and VI (nt 581 to 596), whereas the upper loop regions of domains V (nt 483 to 509) and VI (nt 597 to 606) are characterized by sequence disparity. Roman numerals indicate the major loop structures forming the tips of domains V and VI, respectively (see Fig. 4 for secondary structure). Despite a high degree of conservation of secondary structure in the IRES nt 100 to 580, the architecture of domain VI of HRV2 is different from that of PV1(M) (indicated by upper and lower domain sketches). The AUG codon initiating the HRV2 polyprotein is shown in white letters on black background. The 154-nt-long spacer region between the PV IRES and initiating AUG has been omitted (41). Sequence alignments were generated by using GeneBee-Net (version 1.0. (release 1.1) software.

Based on the results of domain exchanges, we entertained the possibility that sequences within the uppermost regions of domains V and VI may coordinately influence neuropathogenicity. To test this hypothesis, we constructed two genetic variants of PV1(RIPO) that feature PV-specific sequences in these regions [PV1(rpp) and PV1(rpr) (Fig. 4A and B, respectively)]. These viruses differ only in the length of the spacer: PV1(rpp) carries the 154-nt spacer naturally occurring downstream of PV domain VI, which in PV1(rpr) is shortened to 33 nt as in HRV2.

The growth phenotypes of these constructs are shown in Fig. 5. Both recombinant viruses replicated in HeLa cells with growth kinetics similar to those of PV1(M) and PV1(RIPO) (Fig. 5B). Interestingly, exchange of residues within the upper regions of domains V and VI from HRV2 to PV1(M) genotypes restored the tissue tropism characteristic for PV1(M), as both PV1(rpp) and PV1(rpr) replicated in SK-N-MC cells nearly as well as PV1(M) (Fig. 5A). The similarity of growth phenotypes of PV1(rpp) and PV1(rpr) suggested that the nature of the spacer sequence between the silent AUG (Fig. 4) and initiating AUG had little, if any, effect on the growth phenotype. In contrast, a converse viral recombinant, containing HRV2-derived sequences in the upper loop regions of domains V and VI of the PV IRES [construct PV1(prr) (Fig. 5)], suppressed replication in SK-N-MC cells. Exchange of the unconserved nucleotides within certain regions of the PV IRES domains V and VI (Fig. 4C) with their HRV2 counterparts was sufficient to reduce growth kinetics in SK-N-MC cells to the level found for PV1(RIPO) (Fig. 5).

FIG. 5.

Growth characteristics of PV1(rpr) (□), PV1(rpp) (■), and PV1(prr) (×) in HeLa cells and SK-N-MC cells compared to one-step growth curves of PV1(M) (•) and PV1(RIPO) (○). Growth curves established in SK-N-MC neuroblastoma cells (A) and HeLa cells (B) demonstrate cell-specific growth capacities exhibited by different IRES recombinants. All viruses replicated with equal efficiency in HeLa cells. Propagation of PV1(RIPO) in SK-N-MC cells is exceedingly poor. Replacing nucleotide sequences of PV1(M) within the upper loop regions of domains V and VI in an HRV2 background [PV1(rpp) and PV1(rpr)] restored neurovirulence, indicated by their wt growth characteristics in cells of neuronal derivation. The converse recombinant virus of PV1(rpr), containing HRV2 sequences in the uppermost regions of domain V and VI within a PV1(M) IRES [PV1(prr)], was unable to efficiently replicate in neuroblastoma cell lines, underlining the critical importance for cooperative function of stem loops V and VI in the determination of a neurovirulent phenotype of PV. p.i., postinfection.

We then tested PV1(rpp), PV1(rpr), and PV1(prr) in CD155 tg mice for neurovirulence. As anticipated from the growth properties in human neuroblastoma cells, both PV1(rpp) and PV1(rpr) expressed the neurovirulence phenotype, the 50% lethal dose (LD₅₀) being slightly larger than with the parental wt virus, regardless of whether the isolates were administered intravenously or intracerebrally (Table 2). Accordingly, PV1(prr), carrying HRV2 sequences within critical regions of domains V/VI, lacked neurovirulent potential (Table 2). We conclude that the determinants for tropism for cells of neuronal origin, and neurovirulence in experimental animals, map to two specific regions in the IRES domains V and VI of PV.

TABLE 2.

Neurovirulent indices of PV/HRV2 chimeras in CD155 tg mice

Virus strain	LD50 (log10 PFU) after:
	Intravenous administration	Intracerebral administration
PV1(M)	4.0	2.2
PV1(RIPO)	—a	—
PV1(rpr)	5.0	3.2
PV1(rpp)	4.4	3.0
PV1(prr)	—	—

^aInoculation of 10⁹ PFU did not lead to clinical symptoms. 

DISCUSSION

IRES elements are arguably the most complex recognition signals known to be present on viral RNA genomes (41). Between 300 to 400 nt long, they contain many more bits of structural information than would seem necessary to attract ribosomal subunits and direct them to the proper initiation codon of translation (42). Although much work has been published since their discovery (14, 15, 24, 27), the mechanism of internal initiation of translation mediated by IRESes remains largely obscure (13). One of the perplexing properties of IRES elements is that different IRES elements function in very similar fashions despite an apparent lack of structural homology. Such differences are particularly striking when the PV IRES is compared with the IRES of encephalomyocarditis virus or hepatitis C virus. The assertion of similar functions in diverse genetic backgrounds, however, has been based solely on assays with artificial dicistronic mRNAs first constructed by Jang et al. (14), with dicistronic (24) or chimeric viruses (2), that were expressed either by cell-free translation or after transfection in tumor cell lines (13). Tissue-dependent restriction of IRES function in experimental animals, on the other hand, has been observed so far only with the PV chimera PV1(RIPO) in CD155 tg mice (7).

The phenotype of impaired growth of PV(RIPO) in human neuroblastoma cells or in CD155 tg mice is surprising because the IRES elements of PV and HRV2 are very closely related in structure. PV and HRV2, two picornaviruses belonging to related genera Enterovirus and Rhinovirus, respectively, have very similar genotypes and almost identical strategies of replication (11, 30). Indeed, among the IRESes of all picornaviruses, those of entero- and rhinoviruses have been classified into a single group, called type 1 IRESes (41).

The host cell restriction of PV1(RIPO) is apparent not only in CD155 tg mice or in a human neuronal tumor cell line (SK-N-MC) but also in primates, using standard protocols for poliovirus neurovirulence (Table 1). This is an important result because it shows that the mouse model is adequate to map IRES components influencing the neurovirulent phenotype. PV1(RIPO) is a chimera consisting of wt genetic elements of two picornaviruses. Remarkably, PV1(RIPO) proliferates in HeLa cells with the kinetics of the highly neurovirulent strain PV1(M), yet in the highly sensitive test of spinal cord damage in nonhuman primates (lesion scores), it has scored better than the currently used attenuated vaccine strain Sabin type 1.

Our analysis of genetic determinants within the HRV2 IRES responsible for the phenotype of PV1(RIPO) led to the construction of PV/HRV2 IRES chimeras. To our knowledge, this approach has not been taken before, and it was risky because the strategy required changes of primary sequences of the parental cis-acting RNA elements. Such changes could have abolished IRES function. For example, our original attempts to insert restriction sites into a single-stranded sequence (UCCUCC) linking domains IV and V of poliovirus yielded nonviable viruses (data not shown). While our studies were in progress, we learned from the work by Belov et al. (3) that for unknown reasons, the integrity of this sequence is essential for PV IRES function. Furthermore, it seems likely that different domains within an IRES element form tertiary interactions with each other (19). As an assay for the function of IRES chimeras in viral replication in HeLa cells, we have used a carcinoma cell line most commonly used for proliferation of PV. The focus of the study, however, was on a correlation between IRES structure and neurovirulence.

Most of the PVs with chimeric IRESes depicted in Fig. 1 and 4 replicated in HeLa cells with kinetics similar to that of wt PV at physiological temperature, and they yielded viral progeny in the range of 90 to 20% of PV1(M) (Fig. 2). These results demonstrated that exchanges of IRES domains V and/or VI between PV and HRV2 did not yield lethal phenotypes, an observation suggesting that the sequences that differ in these domains between PV and HRV2 (Fig. 3) do not participate in essential tertiary interactions with other upstream domains.

The spacer sequence between the conserved AUG (Fig. 4) and the initiating AUG had little, if any, influence on replication in tissue culture cells or on the neurovirulence phenotype, regardless of whether assayed in the context of a PV or HRV2 domain VI (Fig. 5). This observation, reported also previously for different viruses (35), was important for our neurovirulence tests, since we constructed chimeric viruses that initiated polyprotein synthesis 154 nt downstream of the conserved AUG, as in PV, or 33 nt downstream from the conserved AUG, as in rhinovirus.

Of all the IRES chimeras shown in Fig. 1, only construct D was neurovirulent in CD155 tg mice. This came as a surprise, as we had anticipated that the attenuating locus of PV1(RIPO) may reside in domain V, just as in the attenuated Sabin strains of PV vaccines where attenuating mutations have been mapped to domain V only (Fig. 4 and reference 40). However, recombinant virus E was also highly attenuated. On the other hand, a PV IRES of which domain V was exchanged to that of HRV2 (construct F) was also highly attenuated, just as IRES hybrid I, in which only domain VI was exchanged to HRV2. These observations strongly suggested that both domains V and VI of PV are required to produce a neurovirulent phenotype in CD155 tg mice. It appears, therefore, that specific sequences within domains V and VI are coordinately responsible for the neurovirulent phenotype of PV.

Comparison of the primary sequences of domains V and VI of PV1(M) and HRV2 reveals characteristic differences (Fig. 3). These appear to cluster predominantly to nucleotides in the upper portion of domains V and VI and to the length of domain VI. Starting with PV1(RIPO), the exchange of 13 nt in HRV2 domain V to corresponding nucleotides of PV1(M), together with the construction of a domain VI resembling that of PV1(M), yielded a derivative PV1(rpr) (Fig. 4B) with a tissue tropism phenotype similar to that of PV1(M) (Fig. 5). PV1(rpr) was highly neurovirulent in CD155 tg mice (Table 2). A similar recombinant virus in which the spacer was extended to that of PV [construct PV1(rpp) (Fig. 4A)] had properties similar to those of PV1(rpr) (Fig. 5; Table 2), an observation confirming that the spacer is of little consequence to the observed phenotypes.

As the exchange of few nucleotides in HRV2 domains V and VI with those of PV1(M) produced efficient replication in SK-N-MC cells and neurovirulence in CD155 tg mice, the reverse recombinant yielded the opposite phenotypes. PV1(prr) (Fig. 4C), in which the corresponding nucleotides of the PV1(M) V/VI domains were exchanged with those of HRV2, was attenuated, expressing a replication phenotype in SK-N-MC cells and neurovirulence potential in CD155 tg mice equal to those of PV1(RIPO). These data leave no doubt that neurovirulence of PV1(RIPO) can be restored through specific regions of domains V and VI.

What is the mechanism leading to the observed phenotypes? Preliminary evidence suggests that PV1(RIPO) RNA can be translated in cell extracts of SK cells, albeit with efficiency three- to fourfold lower than that for PV1(M) (24a). A mutation in domain V of PV1(M) was recently reported to result in reduced translation efficiency in a neuronal cell extract (9), but this study was not followed up with neurovirulence tests in experimental animals. RNA of a luciferase-expressing derivative of PV1(RIPO) (in which the P1 coding region was exchanged with the luciferase ORF) yielded a luciferase signal on transfection into SK-N-MC cells. Again, the signal was reduced in comparison to PV1(M) (24a). Thus, diminished translational efficiency in cells of neuronal origin may not be the sole reason for the observed phenotype of PV1(RIPO).

The requirement of domains V/VI for neurovirulence in the chimeric viruses seems to indicate that these domains are engaged in IRES function under the specific circumstances of proliferation in the central nervous system. This conclusion is supported by reports that the removal of the entire domain VI resulted in attenuation in monkeys (12) or in mice (35). The latter experiments used a specific PV1(M) constructed to produce neurovirulence in wt mice (25). It is important to note that even construct I was attenuated. Thus, the presence of HRV2 domain VI diminished expression in SK cells and abrogated neurovirulence in CD155 tg mice. Chimeric IRESes, therefore, may express phenotypes distinct from those observed in wt mice with some deletion or insertion mutants of PV1 (35).

It is possible that domains V and VI bind specifically factors involved in IRES function and that these factors are either modified or in short supply in neuronal cells. Synergy of domains V and VI in binding of cellular proteins in cell extracts has been observed (8), but whether these effects relate to the observation of the remarkable tissue tropism reported here remains to be seen.