During May–December 2019, using standardized procedures of the Global Polio Laboratory Network (https://polioeradication.org), we detected VDPV-2s in fecal samples of 19 children with acute flaccid paralysis (AFP). Positive samples came from 10 districts across the country, including Bangui, the capital city (Figure 1). In addition, we detected 49 VDPV-2s in fecal samples from healthy children living in the vicinity of the children with poliomyelitis; 17 were detected in environmental samples. In December 2017, routine environmental surveillance was implemented in CAR, restricted to 4 sampling sites in Bangui; 6 additional sites were gradually opened in 2019 (Figure 1). Compared with the vaccine Sabin-2 strain (reference strain), CAR VDPV-2s had 6–20 nt differences in the viral capsid protein 1 (VP1)–encoding region (903 nt), above the threshold used to discriminate VDPV-2s from Sabin-2 (>6 mutations within the VP1-encoding sequence). Given the evolutionary rate of this genomic region (≈10⁻² nucleotide changes/site/year [5]), this range suggests that VDPVs had been circulating in CAR from a few months to a couple of years before detection.

Phylogenetic analysis based on VP1-encoding regions showed that the CAR VDPV-2s fell into different lineages (Figure 2, panel A; Appendix Figure). Although the low number of nucleotide differences in young VDPVs makes precise marking of the boundaries of phylogenetic clusters challenging, we identified >12 main branches in this phylogram (Figure 2, panel A, branches A–L), indicating the concomitant emergence of multiple VDPV lineages. Branches A and J gathered sequences of VDPVs sampled from districts located hundreds of kilometers apart (Figure 1), which suggests active circulation of these lineages in the country; by contrast, some lineages (F, I, K, L) were detected only 1 time. No isolates from patients with AFP were of lineages D, F, G, I, K, L; however, determining whether some AFP cases were missed or, alternately, whether surveillance managed to uncover VDPV lineages before they caused poliomyelitis, is not possible. Environmental surveillance is expected to detect poliovirus circulation before it causes the first poliomyelitis case, but the alert system is efficient only if the surveillance is dense enough to cover the entire country, a goal that is difficult to reach in CAR because of the political troubles.

Among the 70 CAR VDPV-2s for which genomes have been fully sequenced through gene walking and Sanger sequencing, only 4 (branches G and L, from healthy children) were free of recombination events and feature a global nucleotide divergence <1% compared with Sabin-2. The 66 other CAR VDPV-2 genomes comprised sequences derived from Sabin-2 and from other nonpolio enteroviruses in 12 recombinant patterns; polio/nonpolio breakpoints were within the 2A, 2B, 2C, 3A, 3C, and 3D-encoding regions (Figure 2, panel B; Appendix Figure). In the nonpolio region, the unique representative of recombinant pattern 5 (member of VP1 branch B) shared recent common ancestors through recombination with the genomes of patterns 3 and 6: it was closely related to genomes of pattern 6 from the 2A through the 3C genomic regions and to the genomes of pattern 3 downstream (Figure 2, panel C). Pattern 4 also shared a recent common ancestor with pattern 3, from which it diverged only near the 3′ extremity of the genome (Figure 2, panel B). Similarly, the genomes of patterns 7 and 8 were closely related from the 2B region through the middle of the 3C region and substantially diverged downstream. Genomic mosaicism is a common trait found in enterovirus ecosystems because of frequent recombination exchanges between cocirculating enteroviruses, including the poliovirus vaccine strains. Thus, VDPVs generally feature genomes resulting from multiple recombination events (6). Three VP1 branches (A, B, and D) contained various recombinant patterns (Figure 2, panel A); reciprocally, 2 recombinant patterns (3 and 11) were each found in several VP1 branches (Figure 2, panel B), thereby illustrating how recombination can make different segments of the enterovirus genome evolve independently (7).

Although VDPV-2s commonly harbor a recombinant nonpolio 5′ untranslated region (UTR), all CAR VDPVs had a 5′ UTR from the vaccine Sabin-2 strain. Nonetheless, an A→G reversion was found in all genomes at nt position 481, which harbors one of the major determinants of attenuation of the Sabin-2 strain (8). A second major determinant of Sabin-2, located within the VP1-encoding region (nt position 2909), had also reverted (U→C, isoleucine-to-threonine) in all CAR VDPVs.

The origin of the CAR VDPV-2s remains unknown. In April 2016, a switch from use of the trivalent OPV to the bivalent OPV, which contains the Sabin-1 and Sabin-3 attenuated strains (but not Sabin-2), was synchronized globally (9). The low nucleotide divergence observed within the VP1-encoding sequence between the CAR VDPV-2s and Sabin-2 makes the hypothesis of silent circulation of Sabin-2–derived strains originating from the trivalent OPV over >3 years unlikely. More likely, the CAR VDPV-2s may derive from the Sabin-2 strain contained in the monovalent OPV that was used to control a 2017–2018 VDPV-2 outbreak in the Democratic Republic of the Congo, which borders CAR (10). Population movements across the border between the 2 countries could have allowed introduction of Sabin-2–derived viruses into CAR, a country in which most children born after the global vaccine switch have no immunity against serotype 2. The silent circulation of these viruses for several months was probably rendered possible by the difficulties of implementing efficient surveillance in some regions of CAR because of the civil war that has been ongoing in the country since 2012.

We show that the CAR VDPV-2 outbreak resulted from several independent emergence events, involving recombinant genomes with no recombination in the 5′ UTR. Beyond the situation in CAR, 2019 was a dark year for the Global Polio Eradication Initiative; VDPV-2 outbreaks surged in many countries in Africa (11). OPV of serotype 2 remains the best tool to stop VDPV-2 outbreaks, but it also constitutes the seed for emergence of VDPVs. The pending release of a novel OPV that contains a genetically stabilized serotype 2 strain less prone to reversion is expected to put an end to this vicious cycle (12).

Phylogenetic tree for type 2 vaccine-derived polioviruses, Central African Republic, 2019.