INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Polioviruses are among the most rapidly evolving viruses known. The most important mechanism for the rapid evolution of RNA viruses is a high rate of base misincorporation by the viral RNA polymerases, which generally lack 3'5' exonuclease proofreading mechanisms (8, 9). Base misincorporation rates in the range of 10⁵ to 10⁴ per base pair per replication have been reported for numerous poliovirus alleles (7, 9, 44). The high mutability of poliovirus genomes, coupled with the frequent occurrence of genetic bottlenecks (8) during poliovirus replication in the human intestine (21, 31), results in exceptionally rapid rates of evolution (~10² substitutions per site per year) during person-to-person transmission (J. Jorba, L. De, J. Kim, and O. M. Kew, Abstr. 18th Annu. Meet. Am. Soc. Virol., p. 148, 1999). Most of this evolution appears to be random genetic drift, as >80% of nucleotide substitutions within the coding region generate synonymous codons (14, 31, 36, 40; Jorba et al., Abstr. 18th Annu. Meet. Am. Soc. Virol., 1999).

A second important mechanism for generating sequence diversity among picornaviruses is recombination (1, 25, 38, 44), first described with polioviruses (18). Because the frequency of genetic exchange is correlated with the extent of sequence homology among templates (24, 41), recombination occurs among closely related templates at high frequencies but can be unambiguously demonstrated only under well-controlled experimental conditions (4, 24). Natural recombination is most readily recognized when genetic exchanges occur between templates lacking close sequence similarities. Natural recombination in polioviruses was first recognized when viruses having chimeric noncapsid sequences were isolated from children exposed to the trivalent oral poliovirus vaccine (OPV) (3, 20, 23). Although most of the vaccine-derived recombinants were produced by genetic exchange only among the three vaccine strains (10), some recombinants arose from exchange with other enteroviruses, including wild polioviruses (15, 16, 27). Because OPV strains rarely spread beyond the close contacts of vaccine recipients (2), recombinant OPV-derived isolates generally have distinct crossover patterns, indicating that the different recombinant genomes are the products of independent evolutionary events (3, 10, 16, 23, 30).

Recombinants have also been detected among wild polioviruses (36, 49), some of which have spread extensively within human communities (49; H. G. A. M. van der Avoort, personal communication). The key epidemiologic properties of wild poliovirus recombinants, such as association with paralytic poliomyelitis (polio) and the capacity for extensive circulation, are essentially indistinguishable from those of their wild parents (49). Indeed, for wild polioviruses, the distinction between recombinant and nonrecombinant genomes appears to be arbitrary.

In this report, we describe a group of type 1 wild-vaccine strain recombinants that had circulated widely in China from 1991 to 1993. The recombinants were first isolated from patients with paralytic polio in northern China in April 1991, following the peak polio epidemic years of 1989 to 1990. All of the wild-vaccine recombinants appear to be derived from a single infection, as each had a homologous 367-nucleotide (nt) block of sequences, spanning the VP1/2A junction, that was derived from the Sabin type 1 OPV strain (LSc 2ab). Because no substitutions were found in the Sabin 1-derived sequences of the earliest recombinant isolates, it is likely that the recombinants were detected shortly after the occurrence of the original recombination event. By analysis of the pathways of evolution of the VP1 and 2A genes (nt 2480 to 3832) of the recombinant isolates, in combination with the epidemiologic data, we followed the spread of the recombinant virus over a 30-month period from northern China to much of the rest of the country.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Viruses. Type 1 polioviruses (Table 1) were isolated by standard methods (45) in the laboratories of the Provincial Epidemic Prevention Stations of China. Isolates were initially characterized by neutralization with hyperimmune antisera, followed by hybridization with Sabin strain-specific (5) and wild genotype-specific (6, 50) RNA probes and partial (150-nt) genomic sequencing across the junction of the VP1 and 2A genes (49). Viruses were propagated in RD cell (human rhabdomyosarcoma cell line; ATCC CCL136) monolayers for extraction of poliovirus RNA.

Extraction of RNA and reverse transcription-PCR. Viral RNA was extracted from freeze-thaw lysates of infected RD cell culture supernatants using Trizol LS reagent (GIBCO-BRL, Gaithersburg, Md.). In vitro reverse transcription and PCR amplification were modified from methods described previously (47). Two pairs of PCR primers, H7S (sense; positions 2402 to 2421, 5'-TTTGTGTCAGCATGTAACGA-3') plus H2A (antisense; 3155 to 3174, 5'-GCTGCACCGTAAAGCGAATC-3') and H5S (3083 to 3102, 5'-TCGAACGCCTACTCACACTT-3') plus H1A (3884 to 3903, 5'-TTGTCTCCAATTTGCTGAGT-3'), were prepared to prime the amplification of two overlapping sequencing templates that together span the complete VP1 and 2A genes.

Nucleotide sequencing of PCR-amplified genome segments. Amplified PCR products were purified with a QIAQuick PCR purification kit (QIAGEN, Chatsworth, Calif.) and used directly for sequencing with the dideoxynucleotide method, using an ABI Prism BigDye Terminator Cycle Sequencing Ready Reaction kit (Perkin-Elmer Applied Biosystems, Foster City, Calif.) and an automated DNA sequencer (model 377A; Perkin-Elmer Applied Biosystems). The four PCR primers and four additional sequencing primers (H4S [2717 to 2736, 5'-TTCTTTGCGCGGGGTGCATG-3'], H3A [2753 to 2772, 5'-GTGGTAGCCGAATTGTCCAC-3'], H6S [3410 to 3429, 5'-GTGTACACTGCAGGTTACAA-3'], and Q8A [3485 to 3504, 5'-AAGAGGTCTCTATTCCACAT-3']) (36) were used for sequencing the PCR amplicons. To minimize possible sequence ambiguities arising from base misincorporation during in vitro amplification, duplicate amplicons were independently amplified from the RNA templates, and the sense and antisense polarity sequences were determined on separate amplicons.

Phylogenetic analysis. Evolutionary trees were constructed from the VP1 and 2A sequences (nt 2480 to 3832) of the 34 recombinant isolates using the DNA Maximum Likelihood program (11), applying the HKY model of nucleotide substitution (17), contained in the PHYLIP 3.572c program package. Input sequences were randomized seven times, and 19,657 alternative trees (among the 7.3 × 10⁴⁵ possible rooted trees for 34 sequences [12]) were compared by heuristic search. The tree with the highest log likelihood score was presented (12). For comparison, trees were also constructed by 10,000 steps of quartet puzzling under a maximum-likelihood algorithm using the program PUZZLE 4.0 (39), by a maximum-parsimony method (FITCH) (12), and by neighbor joining (12, 37). Trees were displayed using the program TreeView (34).

Estimation of the evolution rate of the VP1 and 2A genes. Rates of fixation of synonymous base substitutions into the VP1 and 2A genes were estimated from the sequence differences among 21 recombinant isolates for which the dates of specimen collection or onset of paralysis were available. The number of substitutions at synonymous sites (K_s) was computed according to the methods of Li et al. (28, 29) as modified by Pamilo and Bianchi (35), implemented in the program DIVERGE of the Wisconsin Package version 9.1 (Genetics Computer Group, Madison, Wis.). This method applies the Kimura two-parameter model (22) to correct for multiple substitutions at a site and to account for differences in substitution rates for transitions and transversions (estimated from the data set to be 8.0:1). K_s values from the root sequence (that of 3645/HB91) were plotted as a function of the time of specimen collection (time zero [24 April 1991]). The date of sampling for isolate 3648/HB91 was also set to time zero (actual sample date, 19 April 1991). When only the date of onset was available, the date of specimen collection was assumed to be 15 days after the onset date (the average observed interval between these dates when both dates were known). The rates of accumulation of synonymous substitutions were estimated by linear regression through the origin.

Numbering of nucleotide and amino acid positions. The coding sequences of Sabin 1, 3645/HB91, and 3653/HE91 were colinear, but each differed by a small number of insertions or deletions in their 5' untranslated regions (33; H.-M. Liu, unpublished results). To facilitate comparisons, numbering of the VP1-2A nucleotide positions of all isolates followed that described for Sabin 1 (33). Amino acid positions were indicated by the name of the viral protein and numbered consecutively from residue 1 of each protein. Substitutions were indicated by the following convention: original residue-position-substitution residue(s). For example, VP1:I018M,V indicates a substitution of isoleucine by methionine or valine at amino acid position 18 of VP1.

Estimation of the time of occurrence of the recombination event. The time of occurrence of the recombination between the indigenous type 1 wild poliovirus and the Sabin type 1 vaccine strain was estimated by comparing the differences in the non-vaccine-derived VP1 sequences (nt 2480 to 3270) of the earliest recombinant isolates (3645/HB91, 3646/HB91, and 3648/HB91) and the most closely related nonrecombinant type 1 wild isolate, 3653/HE91. A maximum-likelihood point estimate for the time of divergence (t_div) for the isolates based on the rate of evolution of synonymous VP1 nucleotides was calculated for each pair using the relationship.
where nt_syn is the number of synonymous VP1 nucleotide differences between the two isolates, r_syn is the rate of accumulation of synonymous substitutions in the interval compared (measured at 3.69 × 10² substitutions per synonymous site per year, or 1.01 × 10⁴ substitutions per synonymous site per day), nt_{syn, total} is the total number of synonymous sites (274) in the interval compared, and t is the time interval between the dates of sample collection for the two isolates. The 95% confidence interval for the point estimate is approximated by the equation
where z = 1.96.

Nucleotide sequence accession numbers. The nucleotide sequences of polioviruses determined in this study were submitted to the GenBank library under accession numbers AF11950 to AF11983.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Epidemiologic background. Following the large nationwide polio outbreak in 1989 (4,623 cases) and 1990 (5,065 cases), China strengthened its routine OPV immunization activities (nationwide coverage was estimated to be >90% by March 1991) and initiated a program of supplemental OPV immunization through national immunization days. Supplemental immunization began in the winter of 1990 to 1991, with the highest rates of OPV coverage achieved in the northern provinces Hebei, Henan, Shandong, and Hubei, where two full rounds of immunization were conducted in December 1990 and January 1991 (46). In response, polio cases fell rapidly, from 2,009 in 1991 to 2 in 1994 (43). In 1990, the patterns of poliovirus circulation in China were complex, with at least three indigenous type 1 genotypes and one indigenous type 3 genotype still endemic to the country (19, 26, 49). By 1994, only two cases associated with wild polioviruses were found; they occurred in widely separated provinces (Xinjiang and Hubei) and were associated with different type 1 genotypes (19).

Genetic relationships of an early recombinant to other polioviruses. The complete VP1 and 2A sequences were determined for an early wild-vaccine recombinant, 3645/HB91, isolated from a specimen obtained from a polio patient in Hebei province on 24 April 1991 (Table 1). The sequences were unusual because they contained a stretch of 367 nt spanning the VP1/2A junction that perfectly matched those of Sabin 1. The first 791 VP1 nucleotides of the recombinant closely matched (787 of 791, 99.5%) (Fig. 1; Table 2) those of isolate 3653/HE91, obtained from a patient in neighboring Henan province whose paralytic symptoms first appeared on 15 October 1991 (Table 1). Isolate 3653/HE91 is a representative of a type 1 wild poliovirus genotype that was widely distributed in China up to 1994 (49). The last 195 2A nucleotides of 3645/HB91 were unrelated to those of Sabin 1 (76.9% sequence similarity) or 3653/HE91 (73.8% sequence similarity) (Fig. 1; Table 2).

View larger version (91K): [in this window] [in a new window]   FIG. 1.   Alignment of VP1 and 2A nucleotide sequences of a type 1 wild poliovirus isolate (3653/HE91) representative of the genotype associated with the major outbreak in China in 1989 to 1991, an early recombinant isolate of type 1 poliovirus (3645/HB91) having 367 nt of sequence derived from Sabin 1, and the Sabin 1 vaccine strain. Genomic intervals derived from nonvaccine viruses are shaded. Sabin 1 nucleotide positions are numbered according to Nomoto et al. (33); those of the clinical isolates are numbered similarly for comparability. Boldface letters identify sequences that encode NAg1 (nt 2750 to 2785), NAg2 (nt 3140 to 3157), and NAg3 (nt 3338 to 3355) (32).

Sequence relationships among recombinant isolates. Sequences of the VP1 and 2A genes were determined for 33 additional vaccine-wild recombinants isolated in 10 different Chinese provinces from 1991 to 1993. All shared the 367-nt block of Sabin 1-derived sequence found in 3645/HB91. VP1 and 2A sequences of two other early isolates, 3646/HB91 and 3648/HB91, contained no base substitutions in the Sabin 1-derived interval and differed from those of 3645/HB91 only by a C2590T transition. The sequence data indicated a close link between the three corresponding cases, which occurred within 22 days of each other in communities separated by 150 to 300 km (Table 1). Apart from 3646/HB91 and 3648/HB91, all isolates were much more divergent from 3645/HB91, having from 2 to 13 substitutions in the Sabin 1-derived interval and from 9 to 36 substitutions in the flanking sequences.

View larger version (32K): [in this window] [in a new window]   FIG. 2.   Maximum-likelihood tree of sequence relationships in the complete VP1 and 2A genes (nt 2480 to 3832) among 34 type 1 wild-vaccine poliovirus recombinants isolated in China from 1991 to 1993. The tree was rooted to the sequence of isolate 3645/HB91. Clusters of related sequences are indicated by brackets labeled.

Substitution patterns within the VP1 and 2A genes and proteins. Using the maximum-likelihood tree in Fig. 2 for reference, and assuming that each substitution along any given branch occurred only once, we propose that the tree represents the pattern of fixation of a total of 369 nucleotide substitutions from the root sequence. Most (314 of 369, 85%) of the substitutions generated synonymous codons. Of the estimated 472 synonymous sites in the VP1 and 2A genes (35), 256 (54%) had no substitutions, 133 (28%) had one substitution, 66 (14%) had two substitutions, 13 (2.8%) had three substitutions, 3 (0.6%; at positions 2905, 2914, and 3802) had four substitutions, and 1 (0.2%; at position 2941) had five substitutions (Fig. 3). When more than one substitution occurred independently at a site (i.e., along separate branches of the tree), they usually were parallel mutations. However, reverse mutations to the residue originally found in 3645/HB91 occurred six times at five sites (at positions 2537, 2560, 2815, 3556 [two independent reversions], and 3682), and sequential nonreverse mutations (all involving a transversion followed by a transition) occurred at three sites (at positions 3140, 3346, and 3532). Overall, 89% (328 of 369) of all nucleotide substitutions were transitions.

View larger version (30K): [in this window] [in a new window]   FIG. 3.   Cumulative nucleotide (lower bars) and amino acid (upper bars) substitutions into the VP1 and 2A sequences of recombinant isolates along all observed lineages. The reference sequence (set at zero substitutions) was that of 3645/HB91. Nucleotide and amino acid substitutions were based on the maximum-likelihood tree and the assignments shown in Fig. 2. Sabin 1-derived sequences are shown in gray. The secondary structural domains of VP1 (arrows,  strands; rectangles,  helices), determined from the crystal structure (13), are aligned above the VP1 cumulative substitution map.

Rates of evolution of VP1 and 2A genes. The rates of fixation of nucleotide substitutions into synonymous sites were estimated for different intervals within the VP1 and 2A genes for the 21 recombinant isolates for which records were available on the dates of onset or specimen collection (Table 1). The reference sequence (assigned zero substitutions) was that of 3645/HB91, and time zero was 24 April 1991. The overall synonymous substitution rate for the complete VP1 and 2A genetic interval was (3.73 ± 0.52) × 10² substitutions per synonymous site per year (Fig. 4A). The synonymous substitution rate for the VP1 interval derived from wild type 1 poliovirus was (3.69 ± 0.64) × 10² (Fig. 4B), a rate similar to that [(3.17 ± 0.67) × 10²] for the Sabin 1-derived interval (Fig. 4C). Interestingly, a higher synonymous substitution rate [(5.23 ± 1.16) × 10²] was found for sequences encoding the non-Sabin 1-derived 2A sequences (Fig. 4D). The synonymous substitution rate for the complete VP1 gene [(3.45 ± 0.57) × 10²] (data not shown), representing the combined rates for the wild and Sabin 1-derived sequences, are very similar to the rates of VP1 substitution that have been reported previously (14, 21, 31; Jorba et al., Abstr., 18th Annu. Meet. Am. Soc. Virol., 1999). A slightly higher synonymous substitution rate [(4.40 ± 0.70) × 10²] was observed for the complete 2A gene (data not shown).

View larger version (14K): [in this window] [in a new window]   FIG. 4.   Estimation of rates of fixation of synonymous substitutions into different intervals of the VP1 and 2A genes of 21 type 1 vaccine-wild recombinants. Abscissa, time that sample was taken for each isolate (time zero, time of sampling for isolate 3645/HB91 [24 April 1991]); ordinate, Ks (number of synonymous substitutions per synonymous site; zero substitutions, sequences of 3645/HB91). Linear regression lines estimating synonymous substitution rates are shown for the complete VP1-2A interval (nt 2480 to 3832) (A), wild type 1 poliovirus-derived VP1 sequences (nt 2480 to 3270) (B), Sabin 1-derived VP1 and 2A sequences (nt 3271 to 3637) (C), and non-Sabin 1-derived 2A sequences (nt 3638 to 3832) (D).

Estimated time of occurrence of the recombinant event. The close sequence match (only four mismatches in the first 791 nt in VP1) (Fig. 1) between the early recombinant isolate, 3645/HB91 (case onset date, 15 April 1991), and the nonrecombinant isolate, 3653/HE91 (case onset date, 14 October 1991), suggests that the recombinant viruses were detected soon after the occurrence of the recombination event. This view is reinforced by the observation that no substitutions were found in the Sabin 1-derived sequences of 3645/HB91, 3646/HB91, and 3648/HB91. We estimated the time of divergence between 3645/HB91 and 3653/HE91 from the observed number sequence differences at synonymous sites (3) in the first 791 nt of VP1 and the average rate of evolution at synonymous sites in that interval (assumed to be 3.69 × 10² per year [Fig. 4B]). The maximum-likelihood point estimate of the time of divergence was calculated to be ~49 days after the occurrence of the case associated with 3645/HB91 (around 3 June, a date inconsistent with the epidemiologic record). However, the 95% confidence interval for the point estimate was from ~56 days before (around 19 February) to ~85 days after onset of the 15 April case. Similar maximum-likelihood calculations based on the sequence differences between 3653/HE91 and 3646/HB91 and 3648/HB91 yielded point estimates for the dates of divergence of 26 May and 13 May, respectively (also inconsistent with the epidemiologic record), but earliest dates of divergence were calculated from the 95% confidence intervals to be around 3 February and 21 January. By comparing the calculated range of divergence times with the epidemiologic record, we estimate that the recombination event most likely occurred between mid-January and late March 1991 (the case associated with recombinant isolate 3646/HB91 had the earliest onset date, 30 March 1991 [Table 1]). While this is about the time of the first mass immunization campaigns in Hebei and Henan, the exact source of the vaccine exposure is unknown, as routine OPV coverage was high in those provinces in 1991 (46).

Geographic distribution of recombinant lineages. The pathways of divergence from the root VP1 and 2A sequence shown in the tree (Fig. 2) also trace the spread of the recombinant lineages within China (Fig. 5). The earliest isolates (from March and April cases [Table 1]), having the closest sequence similarity to 3645/HB91 (Fig. 2), were from the northern provinces of Hebei (root cluster) and Henan (cluster A1). Recombinant lineages had spread to the neighboring provinces of Nei Mongol and Shanxi (cluster F1) by June 1991 and to the southern province of Fujian (early G1 cluster isolates) by the following December. By the end of 1991, recombinant polioviruses were already circulating by multiple chains of transmission. For example, 1991 isolates in Henan were associated with clusters A1 and D, and 1991 isolates from Nei Mongol were associated with clusters E and F1 (Fig. 2).

View larger version (33K): [in this window] [in a new window]   FIG. 5.   Distribution and inferred transmission pattern of the natural type 1 intratypic wild-vaccine recombinants in China from 1991 to 1993. Independent lineages are shown as arrows and labeled as in Fig. 2. The communities in southern Hebei and northern Henan where the earliest recombinant isolates (root and A1 lineages) were detected are enclosed in the ellipse. Province abbreviations: AN, Anhui; FJ, Fujian; GD, Guangdong; GS, Gansu; GX, Guangxi; GZ, Guizhou; HA, Hainan; HB, Hebei; HE, Henan; HL, Heilongjiang; HN, Hunan; HU, Hubei; JL, Jilin; JS, Jiangsu; JX, Jiangxi; LN, Liaoning; NM, Nei Mongol; NX, Ningxia; QH, Qinghai; SA, Shaanxi; SC, Sichuan; SD, Shandong; SX, Shanxi; XJ, Xinjiang; XZ, Xizang (Tibet); YN, Yunnan; ZJ, Zhejiang.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The findings of this study highlight the remarkable speed at which type 1 poliovirus lineages, derived from a single source, can spread in a large, polio-endemic country with a mobile population. All of the wild-vaccine recombinants described here appear to be derived from a common ancestor closely related to 3645/HB91. The ancestral recombinant most likely arose during the mixed infection of one person, possibly a child living in the vicinity of southern Hebei province in early 1991, who had concurrent exposure to indigenous wild type 1 poliovirus and OPV. The progeny of this one infection quickly radiated along multiple independent chains of transmission over a wide geographic area, involving communities separated by as much as 2,200 km, in provinces with a combined population of over 450 million. Recombinant isolates were first detected in the northern provinces Hebei, Henan, Nei Mongol, and Shanxi in the spring of 1991; by the following year, they were found in the southern provinces Fujian, Guangdong, Hainan, and Yunnan and in the central province Sichuan. Transmission of the recombinant probably extended to other provinces in China but was not detected because of gaps in surveillance. Further spread of the recombinant may have been limited because most families with young children in China do not travel outside the country. The recombinants found in eastern and central China were not detected elsewhere, as the contemporary type 1 poliovirus isolates from Xinjiang province in western China, and from the bordering countries of Kyrgyzstan, Tajikistan, Pakistan, India, Nepal, Thailand, and Vietnam, were members of other, unrelated genotypes (49).

The rapid spread of the recombinant lineages in China does not imply that the recombinants were any more transmissible than the preexisting lineages. If the transmissibilities of cocirculating lineages are nearly equivalent, then the main determinants of transmission would be various, essentially unpredictable, epidemiologic factors. For example, if the recombination event occurred early in the poliovirus season (before spring), the recombinant might have a greater chance to contribute more lineages to the following high-transmission season. Widespread population movements, which peak during the winter months in China, would favor wide dissemination of potentially new founder strains before the start of the next polio season.

It is uncertain whether the recombinants could have completely displaced the previous nonrecombinant genotypes within China, because their appearance and spread coincided with the implementation of intensified supplemental nationwide polio immunization campaigns. The last reported case associated with the recombinant virus occurred in late 1993, and the last reported case associated with any indigenous wild type 1 poliovirus in China occurred in early 1994 (43). The mass immunization campaigns of 1991 to 1994 were highly effective in stopping all wild poliovirus circulation in China, first in the temperate provinces of the north, then in the more tropical south, and finally in the west (48, 50).

The appearance of recombinant lineages, and their detection soon after the recombination event, allowed us to readily follow their transmission above the complex background of multiple cocirculating type 1 lineages in China (26, 49). All of the recombinants carried a well-defined genetic marker, derived from a discrete natural event, in the 367-nt block of sequence derived from Sabin 1. This extended marker is very stable to reversion as a complete block (however, it could be lost by subsequent exchange with other templates), and thus it differs importantly from individual variable nucleotide positions, where the patterns of substitution can be approximately reconstructed using statistical models (11, 39). Moreover, because the marker was recently derived from a vaccine strain with a standardized sequence (33), it was possible to make a reasonable estimate about the timing of the recombination event.

Several alternative recombination pathways could account for the VP1 and 2A sequence patterns of 3645/HB91. The simplest recombination mechanism would be the insertion by double crossover of Sabin 1-derived sequences into the genome of an indigenous wild type 1 poliovirus. This mechanism might appear to be ruled out because the nonrecombinant wild isolate (3653/HE91) with the highest VP1 sequence match (99.5%) to 3645/HB91 in the non-vaccine-derived interval had a low sequence match (77.4%) to 3645/HB91 in the non-vaccine-derived 2A interval. However, the non-vaccine-derived 2A sequences of 3653/HE91 were obtained by recombination with a third virus (Liu, unpublished). In contrast, non-vaccine-derived sequences of a 1989 epidemic isolate from Jiangxi, 1340/JX89, matched those of 3645/HB91 in both the VP1 (95.8% match) and 2A (88.1% match) intervals. Although these sequence similarities indicate that 1340/JX89 and 3645/HB91 had a recent common ancestry, divergence of the capsid sequences apparently occurred later than divergence of the 2A sequences (assuming that the evolution rates of the VP1 and 2A regions are similar and constant). Thus, the evolutionary relationships between capsid and noncapsid domains of wild polioviruses appear to be very complex.

The recombinants described here are unusual because they involve a crossover within the capsid region of the poliovirus genome. The best-characterized poliovirus recombinants are those isolated from persons exposed to trivalent OPV. Intertypic recombinants excreted by recent vaccinees generally have crossovers restricted to the noncapsid region (3, 10, 16, 23, 30). Intertypic recombinants with crossovers in the capsid region are apparently very rare, possibly because of structural incompatibilities (13). If this view is correct, the only poliovirus recombinants generated by a single crossover within the capsid region likely to retain high replicative fitness would be those derived by exchanges between polioviruses of the same serotype. However, homotypic natural infections involving different poliovirus genotypes are presumably rare, except for those involving wild and vaccine-derived polioviruses, as apparently occurred in northern China in 1991.

Recombination occurs frequently during poliovirus evolution (16). Many different wild poliovirus recombinants have been found in China and other countries. As one example, additional crossovers were detected in the noncapsid regions of several of the vaccine-wild recombinants that were isolated from 1992 and 1993 cases (H.-M. Liu, D.-P. Zheng, L.-B. Zhang, O. M. Kew, and M. A. Pallansch, Abstr. 19th Annu. Meet. Am. Soc. Virol., p. 114, 2000). Indeed, possibly the only sequences in poliovirus genomes that are uniquely derived from other polioviruses are those encoding the capsid proteins.