Evolution of the Sabin Vaccine into Pathogenic Derivatives without Appreciable Changes in Antigenic Properties: Need for Improvement of Current Poliovirus Surveillance

Abstract

The Sabin oral polio vaccine (OPV) may evolve into pathogenic viruses, causing sporadic cases and outbreaks of poliomyelitis. Such vaccine-derived polioviruses (VDPV) generally exhibit altered antigenicity. The current paradigm to distinguish VDPV from OPV and wild polioviruses is to characterize primarily those poliovirus isolates that demonstrate deviations from OPV in antigenic and genetic intratypic differentiation (ITD) tests. Here we report on two independent cases of poliomyelitis caused by VDPVs with “Sabin-like” properties in several ITD assays. The results suggest the existence of diverse pathways of OPV evolution and necessitate improvement of poliovirus surveillance, which currently potentially misses this class of VDPV.

Viruses constituting the Sabin oral polio vaccine (OPV) are inherently genetically unstable (2, 3). Upon reproduction in vaccinees and their contacts, they tend to lose attenuating mutations. At a certain step in their evolution, vaccine-derived polioviruses (VDPV) have virtually no phenotypic distinctions from wild polioviruses and are able to circulate, especially in underimmunized communities, causing sporadic cases of the disease and outbreaks (3, 14). Such VDPV pose a serious challenge to the Global Polio Eradication Initiative (23) because the real eradication of polioviruses should obviously include not only “true” wild viruses but also VDPV (9, 12). Since it is unlikely that the usage of OPV (inevitably producing new VDPV) will be discontinued soon, i.e., until its replacement by inactivated polio vaccine (9, 12), the identification of VDPV is a key issue for the eradication program.

Currently the World Health Organization classifies as VDPV those Sabin vaccine derivatives in which the genome segment encoding the capsid protein VP1 deviates from the parental vaccine strain by at least 1% of nucleotides (14). It is assumed that such isolates have already independently evolved for at least 1 year, a time believed to be sufficient for the loss of attenuating mutations and acquisition of pathogenic properties. Although this criterion seems arbitrary, since isolates with <1% divergence in the VP1-coding region could be as transmissible and paralytogenic as the officially recognized VDPV (25), it satisfactorily serves programmatic purposes. To identify VDPV, the World Health Organization recommends further characterization primarily of those poliovirus isolates that exhibit deviations from OPV in at least one standard intratypic differentiation (ITD) test, genetic (e.g., PCR with Sabin-specific primers) or antigenic (24). Retention of “Sabin-like” properties in these assays is believed to indicate that the isolate in question is not a VDPV. The results reported here demonstrate that this is not necessarily true.

A type 2 poliovirus, BLR-319-07-28 (hereinafter referred to as PV2/Bel), was isolated from stools of 3-year-old boy, an inhabitant of the Gomel region, Belarus, collected on May 2007 on day 6 after the onset of acute flaccid paralysis, initially diagnosed as Guillain-Barre syndrome. The child had been vaccinated with two doses of inactivated polio vaccine followed by three doses of OPV, with the last dose given 9 months before the paralytic disease. Serum antibodies against poliovirus measured in June 2007 had titers of 1:12 for serotype 2 and <1:8 for serotypes 1 and 3. The child had a history of several respiratory diseases, with the last episode of pneumonia in April-May 2007. An immunological investigation (June 2007) indicated that the patient was immunocompromised (<320 mg/dl of immunoglobulin G [IgG], <25 mg/dl of IgM, and <25 mg/dl of IgA). In July 2007, gamma-globulin replacement therapy was started. No polioviruses were isolated from stools 1 month or more after the development of paralysis.

Another type 2 poliovirus, RUS 08-063-034-001 (PV2/Rus), originated from stools of a 6-month-old nonvaccinated boy in an orphanage of the town of Kuybyshev, Novosibirsk region, Russia. The samples were taken on the next day after the patient presented signs of acute flaccid paralysis in February 2008. The level of serum antibodies against type 2 poliovirus increased from 1:256 to 1:1,024 on days 2 and 16, respectively, and no (<1:8) antibodies against poliovirus types 1 and 3 were detected. There were no obvious abnormalities in the immune status of the patient. Stool samples collected on days 15 after the onset of paralysis or later were poliovirus negative.

Both isolates did not markedly deviate from the Sabin-2 virus in the standard ITD tests, enzyme-linked immunosorbent assay with polyclonal cross-adsorbed antisera and PCR with Sabin-specific primers (24). PV2/Bel was also subjected to the two other ITD assays, neutralization with monoclonal antibodies (11) and restriction fragment length polymorphism analysis of the segment encoding a portion of VP1 (nucleotides 2404 to 2883) (4), which also did not reveal any significant differences from the Sabin-2 virus. The two patients had residual paralysis 60 days after the onset of the disease, and the cases were classified as vaccine-associated paralytic poliomyelitis. Full-length viral RNA sequencing revealed that PV2/Rus was a multiple (Sabin-2/Sabin-3/Sabin-2/Sabin-1) recombinant isolate (Fig. 1; Table 1) and that both PV2/Bel and PV2/Rus underwent a marked evolution (Fig. 1). The deviations of their nucleotide sequences in the VP1-coding regions from that of Sabin-2 were 1.88% (17 mutations) and 1.44% (13 mutations), respectively. According to the above-mentioned currently adopted borderline, >1% substitutions in the VP1-coding region (14), both strains should be classified as VDPV. Since PV2/Bel was isolated from an immunodeficient patient, it could be designated VDPV, whereas PV2/Rus, having an unknown (“ambiguous”) history, was an aVDPV (14).

FIG. 1.

Nucleotide and amino acid differences of isolate PV2/Rus (A) or PV2/Bel (B) from its parental OPV strain. While exhibiting serotype 2 antigenicity, PV2/Rus is a triple intertypic recombinant and also contains, in addition to the Sabin-2-derived sequences (gray), sequences originating from Sabin-1 (dashed) or Sabin-3 (dotted). For the coordinates of crossover regions and the extent of divergence of each genomic segment from the respective OPV parent, see Table 1. Nucleotide numbering is given according to the Sabin-2 sequence. Mutations known to eliminate the attenuated phenotype of Sabin-2 are in italics and underlined.

TABLE 1.

Genomic structure of PV-2/Rus

Segment no.	Coordinatesa	% Divergence fromb:			Origin
		Sabin-1	Sabin-2	Sabin-3
1	1-4305	24.12	0.74	25.05	Sabin-2
2	4306-4527	17.94	15.70	0	Sabin-3
3	4528-5022	18.18	1.01	12.12	Sabin-2
4	5023-7439	0.50	13.56	12.86	Sabin-1

^aNucleotide numbering is given according to the Sabin-2 sequence. The crossovers were assumed to map to the middle of the regions of nucleotide identity of two parent strains (4303 to 4307, 4519 to 4535, and 5017 to 5027).

^bBold italic numbers indicate minimal divergence from the respective OPV strains.

Importantly, the two known major signatures of Sabin-2 attenuation (14) were lost in both isolates: A-481 in the 5′-untranslated region (5′-UTR) was replaced by G, and Ile-143 of VP1 was replaced, respectively, by Asn and Thr in PV2/Bel and PV2/Rus (Fig. 1). Such replacements are very characteristic of type 2 VDPV (15, 22) and are expected to markedly increase neurovirulence (14). In addition, stabilization of the secondary structures of the 5′-UTR internal ribosome entry site (IRES) appeared to occur. Indeed, several U-G base pairs were converted into U-A pairs (positions 275/403 and 491/500) or into C-G pairs (positions 189/217 and 278/398) in one or both isolates (Fig. 2). Although the physiological impact of these changes is unknown, two independent lines of evidence suggest that they might enhance fitness and perhaps virulence: (i) nearly all of them are not infrequently found in other vaccine-related polioviruses (1, 6, 17, 21, 22, 29), and (ii) viral fitness and virulence are known to be dependent on the stability of the secondary structure of the IRES (2). Likewise, amino acid replacements in one or both isolates at positions VP1-19, VP1-77, VP1-101, VP1-231, 2C-3, 3C-50, and 3C-60 corresponded to acquisition of residues found in wild serotype 2 polioviruses or other VDPV (1, 21, 22, 29), suggesting that at least some of these replacements might also confer a selective advantage.

FIG. 2.

Nucleotide substitutions in domains III to V of the 5′-UTR genomic region (18) of isolates PV2/Bel (dotted arrows) and PV2/Rus (arrows with diamond). For mutations in the stems, the opposing (base-paired) nucleotides are also shown.

On the other hand, the sequencing results were fully compatible with the outcome of genetic ITD (no mutations of the relevant nucleotides) and with conservation of the “Sabin-like” antigenicity of the isolates. In polioviruses, the major epitopes are thought to involve predominantly amino acid residues 91 to 102 of VP1 (antigenic site 1 or AgS1), residues 221 to 226 of VP1, 164 to 170 and 270 of VP2 (AgS2), residues 58 to 60 and 71 to 73 of VP3 (AgS3), and residues 72 of VP2 and 76 of VP3 (AgS4) (16). Of these positions, only Ala-101 was changed to Thr in PV2/Rus, but mutation at this position was previously shown not to alter the Sabin-like enzyme-linked immunosorbent assay response (27).

As already mentioned, a poliovirus isolate considered to be a wild type or a VDPV candidate (and therefore destined for more-detailed characterization by sequencing) must differ from the homotypic Sabin strain in at least one of the ITD assays (24, 26). By this criterion, both PV2/Bel and PV2/Rus would have been considered not worth further investigation. Retrospectively, it becomes obvious that VDPV with Sabin-like antigenicity is not unique. Indeed, such VDPV have also been detected by others (19, 20, 26), though they did not appear to attract special attention. It may be noted that VDPV with VP1 sequences nearly identical to that of PV2/Rus were isolated from three healthy contacts of the relevant patient (two of these isolates contained two additional identical synonymous mutations), confirming transmissibility of these viruses.

A significant parameter of OPV derivatives is their “age,” i.e., the time elapsed after the onset of their independent evolution from the vaccine. This time may provide an estimate of the number of people involved in the spread of an isolate. The “age” is usually calculated on the basis of the extent of genetic divergence from the vaccine strains (13, 14). Accumulation of synonymous mutations is considered to be the most reliable measure, since their appearance is assumed to be random, being restricted perhaps only by the secondary structure-based cis-acting RNA elements and biases in the codon or codon pair usage (5, 10). Nevertheless, we observed a striking irregularity in the distribution of synonymous substitutions along the genome (Fig. 1), a feature which appears to be not infrequent among VDPV and which poses problems for precise evaluation of the viral “age.” In some cases, such unevenness could be explained by intratypic recombination (8, 28). This explanation seems to be especially realistic for PV2/Rus, whose history involved multiple recombination events apparently occurring between partners of different “ages.” Indeed, its serotype 3-derived RNA segment was identical to the relevant segments of the parental Sabin strain (Fig. 1; Table 1). However, PV2/Bel was not a recombinant, supporting the notion that mechanisms other than recombination might also be contributing factors. Uncertainties of the causes of uneven accumulation of synonymous mutations notwithstanding, the extent of this accumulation in the VP1-coding region is widely used and has proved to be a relatively reliable measure of the VDPV “age.” Assuming the rate of 3 × 10⁻² synonymous substitutions per synonymous site per year (8), PV2/Bel and PV2/Rus were estimated to be ∼1.6 and ∼1.2 years “old,” respectively. If PV2/Bel iVDPV could likely have replicated only within a single immunodeficient host, PV2/Rus, being isolated from a 6-month-old unvaccinated baby, had certainly passed through more than one person.

Our findings also suggest the existence of distinct pathways of evolution of Sabin viruses. As is well known, the OPV is genetically unstable due to the acquisition, during its original selection, of mutations that decreased not only neurovirulence but also general fitness (2). Evolution of Sabin strains during their multiplication in the vaccine recipients and during person-to-person spread is accompanied by a more or less rapid loss of fitness-decreasing attenuating mutations. In PV2/Bel and PV2/Rus, such deattenuation was exemplified by the replacement of vaccine-specific residues by those either typical of wild polioviruses or infrequently found in other VDPV. Among mutations rendering OPV strains less neurovirulent and less fit are those impairing early steps of the virus/cell interaction. Since the receptor-recognizing regions of the capsid proteins overlap the regions comprising immunodominant viral epitopes, acquisition of certain attenuating mutations was accompanied by vaccine-specific alterations in antigenic properties. Thus, the appearance of “non-Sabin-like” antigenic characteristics typical of most VDPV is believed to be due to two independent processes, immune pressure and loss of fitness-decreasing capsid mutations (27). However, the examples of PV2/Bel and PV2/Rus demonstrate that elimination of antigenicity-changing and fitness-decreasing capsid mutations is not an obligatory component of extended OPV evolution. It is tempting to speculate that the lack of appropriate changes in the capsid proteins was compensated by accumulation of other adaptive mutations, which are yet to be defined. Stabilization of the secondary structure of the IRES and other presumably fitness-increasing replacements mentioned above could possibly contribute to this distinct evolutionary pathway.

In conclusion, we may state that the repeated occurrence of highly diverged pathogenic VDPV, which retained Sabin-like antigenic properties, presents a serious challenge to the current poliovirus surveillance strategy. New tools, for example, oligonucleotide microarrays (7) or specially designed real-time PCR (26), should be adapted for the poliovirus ITD and, if found reliable, should urgently become an obligatory part of the routine surveillance. Peculiarity of antigenic evolution (or rather a lack of it) of the isolates also contributes to a deeper understanding of factors ensuring acquisition of pathogenic properties by vaccine polioviruses.

Nucleotide sequence accession numbers.

The nucleotide sequence data reported in this article are available from the GenBank nucleotide sequence database under accession no. FJ517648 and FJ517649.