We examined the sequence diversity of 8 geographically and temporally variable BRZ-VACV isolates. We compared molecular sequence data from 3 genes and a variable region of the poxvirus genome (Table 1) among BRZ-VACV isolates, available vaccine strains related to those used during the eradication campaign in Brazil, and other VACVs isolated from domestic animals (including endemic buffalopox virus in India [15–17] and horsepox virus [HSPV] from Asia [18]).

Sequences were collected from GenBank (Tables 1 and 2). Some isolates had limited sequence data for the 4 genes examined. Although all 4 genes are not available from Lister (B19R gene is not present in the Lister genome), the Lister isolate was included because of its importance in the history of vaccination in Brazil (3,6). Sequences from 3 genes were manually aligned and trimmed to include only the regions available for all isolates (VACV-COP A56R 161210–162095, E3L 51465–50929, B19R 178242–179173). Nucleotide identities were calculated by using BioEdit (www.mbio.ncsu.edu/bioedit/bioedit.html) (Technical Appendix). PAUP* version 4 (Sinauer Associates; Sunderland, MA, USA) and MrBayes 3.1 (http://mrbayes.csit.fsu.edu) were used to analyze a matrix of the 3 concatenated genes (19,20). Parsimony analysis was performed on 1,000 random addition replicates. Appropriate models of sequence evolution were determined by Modeltest (21) with the Akaike information criterion (22,23). The identified model was then applied to maximum likelihood and Bayesian analyses. In MrBayes, the Markov chain Monte Carlo searches explored 4 chains for 1 million generations, sampling every 100 generations. Sample points during the first 50,000 generations were discarded as burn-in, before which the chain reached stationarity (fixed condition). To account for any considerable rate variation between genes, a second Bayesian analysis was run wherein each gene was partitioned and allowed to vary independently (MrBayes command: prset ratepr = variable). DNA sequence data for the fourth coding region, rpo132-ATI-p4c-A27 (A-type inclusion body gene), were not included in the phylogenetic analysis; however, published digestion profiles from relevant strains were compared.

The resultant phylogenetic tree (Figure 2A) depicts several highly supported clusters including monophyletic VACV, MPXV, VARV, and camelpox virus (CMLV) assemblages. VARV and CMLV isolates are depicted as sister taxa with high support. Brazilian isolates fall into 2 well-supported monophyletic groups. Group 1 comprises Araçatuba virus (ARAV), Cantagalo virus (CTGV), Guarani P2 virus (GP2V), and Passatempo virus (PSTV); group 2 comprises GP1V, Belo Horizonte virus (VBH), BeAn 58058 virus (BAV), SPAn232 virus (SAV), and VACV-Western Reserve (WR). Results of all analyses—parsimony, maximum likelihood, and Bayesian analysis (data not shown)—all generated 2 distinct Brazilian clades and indicated a close relationship between group 2 and VACV-WR. Neither group of BRZ-VACV was directly linked with VACV-Oswaldo Cruz Institute (IOC) or VACV-Lister in any analysis. Both BRZ-VACV groups are represented by 4 isolates each. VACV-Acambis3000 and VACV-modified vaccinia Ankara (MVA) group together with high support (98, Figure 2A), which reflects their derivation from vaccinia isolate Ankara (both are virus isolates of MVA). Isolates with known origins from Dryvax (Wyeth Laboratories, Marietta, PA, USA) (VACV-3737, -Acambis3, -Acambis2000, -DUKE) did not group together consistently. This could be a result of numerous passages, the history of Dryvax as a nonclonal vaccine strain, or both.

Genetic comparisons of the analyzed VACV strains can be seen in the Technical Appendix. For BRZ-VACV at all 3 loci, the highest identity values occurred within each phylogenetic group. The E3L sequence of PSTV is the only exception. The PSTV E3L sequence is identical to those of GP1 and WR and showed the next highest identity level with the VBH sequence. This finding may indicate recombination in this portion of the genome between isolates from these 2 groups. Additional genomic data and analysis are needed.

The sequence diversity of the orthopoxvirus hemagglutinin (HA) gene has made it a potential marker for molecular diagnostics and phylogenetics (24). In particular, an 18-nt deletion within the gene has been proposed as an identifier of BRZ-VACV strains (8,11,25). A recent article describes 2 approaches for identifying BRZ-VACV–like isolates (25), both of which rely on presence of the 18-nt deletion. Figure 3 shows that while some Brazilian strains (ARAV, CTGV, GP2V, PSTV) share this deletion, others (SAV, BAV, GP1V and VBH) do not and would not be detected by these methods. Within the variably deleted region (Figure 3, amino acids 245–255), VACV isolates can be grouped into 4 types; Brazilian isolates fall into 2 of these. Rabbitpox virus from Utrecht has a unique deletion of only 2 residues (254, 255); VACV-3737 and VACV-Malbran share a 5–amino acid deletion (245–249). The VACV-3737 isolate was “plaque-purified from a vaccinia lesion following vaccination with Dryvax” (GenBank,accession no. DQ377945); VACV Malbran was used in vaccination programs in Argentina between 1937 and 1970 (27). GP1V and VBH share the same amino acid sequence (ADLYDTYNDND; 245–255), identical to that of VACV-WR, VACV-Lister, and several other strains. The remaining isolates from Brazil share the familiar deletion of 6 amino acids (250–255) described previously. This deletion is also present in the VACV-IOC vaccine isolate (absent in Lister and VACV-WR). The variation in this region demonstrates that the deletion itself is not representative of all Brazilian isolates and therefore is not useful as an identifier of Brazilian VACV.

This same deletion has been recognized as a possible shared derived trait between VAVC-IOC and BRZ-VACV. Although this scenario is indeed possible for the CTGV-like BRZ-VACV of group 1, those isolates of group 2 are unlikely to be derived from the IOC vaccine strain because this strain does not carry the insertion. Furthermore, another variable gap occurs in this region of HA among isolates derived from VACV-Lister and Dryvax (New York City Board of Health [NYCBH]). Some clones present the deletion; others do not. The sequence can be found directly adjacent upstream of the variable deletion of the group 1 Brazilian isolates (Figure 3). Perhaps this region is nonessential and able to easily withstand such variation without detrimental effects on the utility of the HA protein or the fitness of the virus. Regardless, this deletion clearly cannot be relied upon to identify isolates directly related to Lister or NYCBH strains. A single shared deletion is only 1 character toward determining phylogenetic relatedness among strains. If smallpox vaccines are not clones but rather pools of virions that vary at the molecular level, assuming relatedness based on a single molecular variation is even more difficult. Our phylogenetic analyses of 302 informative characters, including variation across the HA gene, does not indicate a close relationship between VACV-IOC and BRZ-VACV. The consensus tree (Figure 2A) depicts the IOC vaccine strain in an unresolved basal polytomy within the vaccinia cluster and is thus inconclusive as to the relatedness of VACV-IOC and BRZ-VACV.

Finally, published data indicate a large amount of variability in the region that includes the A-type inclusion body gene of the BRZ-VACV isolates (5,9,11–13). Figure 3B illustrates the arrangement of open reading frames in this region (rpo132-ATI-p4c-A27L) for CPXV Brighton Red, VACV Copenhagen, VACV-WR, and BRZ-VACV. The formation of A-type inclusion bodies is restricted to cells infected with CPXV, ectromelia virus, and raccoonpox virus. The C-terminus of the gene encoding the ATI protein is highly conserved among orthopoxviruses because it overlaps with the C-terminus of the RNA polymerase (132) coding region (9,13). However, the region between this conserved sequence and the N-terminus of the P4c precursor shows considerable variation among orthopoxvirus strains (26). Deletions and nucleotide changes generate variable coding regions and distinguishable digest patterns in this portion of the genome. Three profiles of this region have been discerned among BRZ-VACV; 2 are differentiated by distinct digestion patterns (6,7,11,13). The third state is the near complete absence of the A26L gene except for the last 112 nt (9,13). According to da Fonseca et al., SAV, a Brazilian VACV isolate, shows an A26L digestion profile identical to that of the VACV-WR strain (profile 1) (6). ARAV, PSTV, and GP2V show a digestion profile compatible with but not identical to that of the SAV, VACV-WR, and Lister strains (profile 2) (7,11,13). In BAV, VBH, and GP1V, a major portion of the A26L gene is missing (profile 3) (5,9,13). The open reading frame is small and is probably not expressed, although this has not been verified by Northern or Western blot. Damaso and colleagues demonstrated that A26L is present in CTGV; however, the digestion profile has not been established. Western blot analysis detected a 94-kDa protein typical of VACV-WR and VACV-IOC (8). These 3 profiles correspond to 3 clades recovered in the phylogenetic tree (Figure 2A). These findings constitute a major distinction between CTGV (also SAV, ARAV, PSTV, and GP2V, which exhibit profiles 1 and 2) and the other Brazilian isolates BAV, VBH, and GP1V (profile 3). Furthermore, the unusual deletion presented by BAV, VBH, and GP1V is remarkably different from the corresponding regions of VACV-IOC, -WR, and -Lister, strains implicated in the release of VACV in Brazil.

Although eradication had been achieved in Europe by 1953 and in Central and North America by 1951, Brazil did not conduct a nationwide vaccination campaign during the 1950s. However, Brazilian people in cities and towns were vaccinated when outbreaks were reported by local authorities (2,3). Not until 1962 did Brazil, working with the Pan American Health Organization, launch a national campaign against smallpox (3). The intensified global WHO program prompted a renewed national program in Brazil in 1966. Fenner et al. (3) present a sample questionnaire that the smallpox eradication unit circulated in 1967 to all vaccine producers in countries accessible to WHO. In Brazil, this included 4 laboratories in different regions: the IOC, the Butantan Institute in São Paulo, the Institute for Biological Research in Porto Alegre, and an institute in the town of Recife, Pernambuco State (3). Records identify the seed strains used at these institutes to be “Paris,” Lister, and “Lederle” (28). “Paris” appears to be a reference to the original animal vaccine imported from the Chambon Institute in 1887. The Institute for Biological Research in Porto Alegre received from London the Lister strain, which was originally developed at the Lister Institute, England. Recife and Butantan both indicated use of a strain obtained from Wyeth’s Lederle Laboratory in the United States. The NYCBH strain was the source of Lederle and forerunner of Dryvax, the live vaccine maintained in the United States since the 1970s, and the research vaccine strain VACV-WR (3). These records confirm information that “Paris,” Lister, and NYCBH were used in Brazil between 1968 and 1971 (2,3). It has been suggested that VACV-WR was used as a vaccine in Brazil as well; however, such use is now thought to be unlikely (6,7). The strain was derived by serial passage in mice infected intracranially and selected for neural virulence to mimic encephalitis, a rare side effect of the vaccine in humans (29,30). When available vaccine supplies dropped to critical levels in 1970, Brazil requested and received reserve vaccine from Argentina at least once (3). The Malbran strain was used in vaccination programs in Argentina until 1970, after which the Lister strain was adopted following WHO recommendations (2,3,27). Vaccine production at the Butantan Institute converted to the Lister strain late in 1970 (3), and Butantan became the official distributor for Brazil. Therefore, at least 4 strains of vaccine might have been distributed within Brazil during the WHO eradication campaign.