On the origin of smallpox: Correlating variola phylogenics with historical smallpox records

The availability of genome sequences from a collection of VARV isolates with a broad geographic distribution (Fig. 1) was used for the evolutionary study of VARV (4). VARV, a member of the genus Orthopoxvirus, contains a single linear double-stranded DNA genome of 186 kb that encodes most enzymes for its propagation (5). The VARV isolates studied demonstrated a low mutation rate because epidemiologically linked isolates showed little or no sequence changes in samples with collection times up to 1 year apart [supporting information (SI) Table 3]. This low mutation rate can be clearly seen in our genetically identical samples from Bangladesh collected between 1974 and 1975. These samples certainly represent long complex transmission chains from one of the most densely populated regions in the world but have identical or nearly identical nucleotide sequences. The related Orthopoxvirus monkeypox virus similarly shows single or no nucleotide changes among isolates from three different hosts in a recent United States outbreak (6). Within and across species of orthopoxviruses, the genomes possess nucleotide identities of >99.6% among VARV isolates and 98% among VARV and congeneric taxa such as taterapox virus (TATV) (4) (West Africa) and camelpox virus (CMLV) (7) (Central Asia) (Table 1), two species with the highest sequence similarity to VARV (4). The large size and slow nucleotide substitution rate allow the VARV genomic DNA to retain phylogenetically informative mutations while lowering the incidence of homoplastic mutations (i.e., fewer character-state reversals). The diversity of 47 recently sequenced VARV genomes was largely associated with the geographic origin of the isolates (ref. 4 and this study), and mutations had minimal association with temporal variation across isolates. These observations demonstrate that there was not a single pandemic strain (SI Text, Note 1) but rather that the genetic clades of these analyses are representative of old and perhaps ancient locally endemic strains.

We developed a unique method of using concatenated SNP (cSNP) matrices for VARV phylogenetic analysis. The cSNPs, defined as a polymorphic nucleotide flanked by a fixed number of conserved nucleotides, provided a less complex representation of the whole genomic sequence by excluding insertion/deletions (indels) (SI Text, Note 2), and the topology generated in this study (Fig. 2) agreed with the analyses of the viral genome in ref. 4. SNPs constitute ≈89% of the total genomic sequence diversity among the most diverse VARV isolates (Table 1). The cSNPs of VARV isolates alone and VARV isolates combined with CMLV and TATV captured ≈79% and 42% of the total SNP diversity of VARV, respectively (Table 1). Phylogenetic analyses of cSNPs within VARV and among VARV, CMLV, and TATV resulted in a cladogram with a very low homoplasy index (0.039) (see Methods) and well supported clades (Fig. 2 B). The low incidence of homoplastic events indicated that the resultant VARV and VARV–CMLV–TATV phylogenies were not likely influenced by convergent evolution (8, 9). The VARV cSNPs, using different numbers of conserved flanking nucleotides (SI Text, Note 3), generated different numbers of SNPs but resulted in congruent topologies (Fig. 2 and SI Fig. 4). The low homoplasy indices (9, 10) suggest that DNA recombination was not likely influencing the topology generated by these analyses, although recombination occurs in laboratory conditions (11).

Our rooted analyses of VARV resulted in the description of two primary clades that diverged from a common ancestral VARV (Fig. 2). Primary clade I (P-I) comprises Asian VARV major and East, Central, and South African VARV isolates. Asian VARV was traditionally known as VARV major with a CFR of 20–30% (2). The non-West African (African) isolates examined in this study had a 30-year temporal representation from South, Central, and Eastern Africa and were collected from regions with both low (1%) and high (10–15%) CFR smallpox (4), including the last naturally occurring smallpox case in Somalia (African minor). Excluding isolates with poorly characterized origins of importation into Europe (Fig. 2), the South, Central, and East African (African) isolates formed a related monophyletic cluster with isolates of VARV major from Bangladesh. This analysis suggests that African minor is likely a variant of Asia VARV major, which is consistent with their similar biological properties (12). Thus, the traditional nomenclature separating smallpox into strains of variola major and variola minor based exclusively on CFRs is not supported by our phylogeny.

VARV primary clade II (P-II) comprises two subclades. One encompasses the biologically distinctive South American isolates traditionally known as alastrim minor, and the other is composed of West African isolates. Limited historical medical documents indicate that alastrim was described by indigenous people in central Brazil as a disease with pustular rash similar to smallpox but with a markedly low fatality rate (13 –15). The first published description of this clinically distinct mild type of smallpox was in the Caribbean region by Anderson in the 1860s (16), and the disease was later exported to Florida and subsequently elsewhere in North America (17). During the smallpox eradication program, the CFR of smallpox in West Africa was described to be intermediate (≈10%) between variola major and alastrim based on age-adjusted CFRs, and the isolates in this region were described with intermediate biological characteristics with respect to the alastrim minor and VARV major isolates (12, 18). Our analyses show that West Africa VARV shared a relatively recent common ancestor with alastrim minor (compared with its relationship to P-I) (Fig. 2).

We calculated estimates of evolutionary rates and time to most recent common ancestry on multiple evolutionary clocks. Initially, the year of isolation for each isolate was used under both strict and relaxed molecular clocks as in Bird 2007 (19, 20) (Table 2). These analyses result in a prediction that the time to the most recent common ancestor (t_MRCA) of the two primary VARV clades is 207/231 years before present (YBP), and the radiation of P-I VARV major is 83/92 YBP, using strict/relaxed clocks, respectively (Table 2 and SI Datasets 1 and 2), which implies a very recent origin for the VARV isolates in this study. Both of these calculations would suggest that the VARV major isolates are from a recent pandemic sweep from East Asia (the basal subclade), which replaced all original local endemic VARV strains within the last ≈90 years. However, this contradicts the geographically (as opposed to temporal) associated topology of VARV isolates in this study. Additionally, VARV of South Africa (African minor), which was endemic in that region in the late 19th century, was first officially reported in 1904 (21, 22), and our isolates were obtained in 1965. These historical records would not agree with a 90-year (very recent) diversification of the isolates used in our phylogeny. Additionally, this analysis estimates the divergence of VARV from the rodent isolates (TATV) of West Africa at ≈1,374 (relaxed clock)/1,899 (strict clock) YBP, which is the approximate time that endemic smallpox was initially recorded in East Asia. The large geographic separation of Asian VARV and the West African enzootic TATV (occurring in a West African naked-soled gerbil) would not support this timeline. The slowly evolving VARV ancestor would likely need more time to become an efficient and exclusively human pathogen; therefore, the above results based on the entire set of isolation dates most likely underestimate the evolutionary history of VARV.

The historical smallpox records provide two calibration points for the coalescent analyses. The radiation of VARV P-I appears to begin in East Asia (basal node), which is congruent with the location of the earliest recorded smallpox cases. The basal position of the Asian isolates is supported by the presence of 42 pleisomorphic (maintained from an ancestral virus) SNPs between these isolates and members of P-II (SI Fig. 5). If the radiation of P-I represented the gradual establishment of smallpox endemics within the Old World following the smallpox historical records, then the VARV P-I radiation began as early as 1,600 YBP (400 A.D.) as described in Fenner et al.'s discussion (2) of the earliest (Chinese) descriptions of smallpox. Using this date as a “prior” in the Bayesian Evolutionary Analysis Sampling Trees (BEAST) software (http://beast.bio.ed.ac.uk) analyses under a strict molecular clock yields a divergence date of ≈6,300 YBP for P-I/P-II divergence (Table 2 and SI Dataset 3). These results support an ancient origin of VARV and would suggest that smallpox endemics started in East Asia and spread through the Middle East and India and then to Africa. This hypothesis agrees with both the resultant tree topology and the early smallpox history in East Asia and India and historic records that describe the importations of smallpox to South Africa in 1713, when highly virulent VARV strains were imported from the Indian continent (23) and later spread to Central Africa (24). As an alternate possibility, we used the smallpox historic records of South Africa as a calibration point, ≈250 YBP, as the “prior” divergence date (BEAST) for the South, Central, and East African (African) isolates subclade. The entire topology yields a divergence date of P-I/P-II at ≈1,400 (Table 2 and SI Dataset 4) to 2,300 YBP (SI Table 4).

If the P-I radiation represents the smallpox endemic history in the Old World, it explains the historic puzzle regarding the absence of a description of smallpox in the literature of the ancient Greek and Roman civilizations (1). It is probable that the smallpox epidemics in this region occurred at a time later than those in ancient China and India. The “Athens plague” in 430 B.C. described by Thucydides, once considered to be smallpox (3), would contradict this theory, but it has recently been determined to be typhoid fever (25). During the reign of Ramses V, Egypt was in a civil war and was attacked by enemies from the north (3); if the pustular eruption of Ramses V was from smallpox, it could represent a smallpox outbreak from imported cases because of war rather than regional endemic disease. This hypothesis is supported by the fact that only three mummies in that period had similar lesions (2).

The smallpox history/VARV topology-based analyses show that the divergence of alastrim minor and West African variola began at least 800 years ago (Table 2 and SI Table 4), predating the previous hypothesis that this divergence coincides with the beginning of the slave trade. Under the more ancient origin hypothesis, this divergence could have occurred after a speciation event in either the New World (based on the calibration from the earliest smallpox historic records; Fig. 3 A) or West Africa (based on the calibration from the South African smallpox historic records; Fig. 3 B). These scenarios would suggest that there were unidentified variola minor/alastrim isolates that existed in either Africa or the New World much earlier than the discovery of alastrim minor (P-II); thus, this taxon has a longer history than has been described.

Although the well documented importation of VARV major into the Americas had a devastating effect on many of the native populations (2), the records of rapid disease spread with great mortality might, in part, be attributable to the combined effect of the high population density, lack of basic care for victims, and potential coinfections with other Old World diseases (26, 27). Historically, there are few records of smallpox in West Africa, and Henige (28) argued that the first VARV major smallpox outbreak in the New World did not originate from West Africa, which would refute the hypothesis of an early smallpox history in West Africa. It is possible that after an ancient introduction, alastrim evolved independently in isolated regions within the large geographic area of the New World. In fact, Phelan's study showed that the native populations of the Ecuador Highlands did not decrease during the initial Spanish rule (attributed to have introduced VARV major to the New World) (29), suggesting that these populations could have been more resistant to smallpox, perhaps by previous contact with alastrim. In Trinidad, a mild smallpox outbreak in 1902 was documented to have originated from this region (30). If this diversification occurred in the New World, then the primary isolation event would likely have resulted from the movement of an ancestral VARV with early humans into that region (Fig. 4A). The absence of SNP signatures suggestive of mixing between P-I and P-II (SI Figs. 6 and 7), the consistent and significantly different clinical characteristics of alastrim from traditional VARV major of P-I (31), could be explained by such a significant geographical separation.

If the split of P-I/P-II occurred within the Old World, P-II likely originated in West Africa. The two P-II subclades would then have diverged ≈800–1,100 YBP (Table 2 and SI Table 4) on the African continent. The geographic distributions of these events are difficult to determine because of a lack of written history in these regions. However, the presence and diversification of other closely related Orthopoxviruses such as monkeypox and TATV could support an African origin for ancestral VARVs. If the isolation and diversification of P-II occurred in Africa, then, based on our phylogeny, there is an undescribed African representative from the P-II subclade, which was introduced into the New World perhaps hundreds of years after the diversification event, possibly during the time of the slave trade between countries in the Old and New World (Fig. 3 B).

An inadequate understanding of the history of these cultures and the relatively milder clinical symptoms (with respect to variola major) contribute to the absence of a more extensive historic record of smallpox disease(s) in either the New World or sub-Saharan Africa. The historical human population densities in both of these regions were lower than the regions with endemic variola major. In cases of milder disease, individuals would have been more mobile when infectious and would have had a greater number of contacts and thus perpetuated the disease's spread. In support of this hypothesis, the mild smallpox outbreaks in Somalia persisted for long periods among small nomadic bands and showed a significantly different transmission pattern compared with that of VARV major (2). Although it is unknown whether the practice of variolation/inoculation was used by the indigenous peoples of Africa or the Americas, the longevity of VARV within convalescent scabs (up to a decade) (32 –34) would provide another mechanism for the repeated introduction of a smallpox strain within a geographic community either by inadvertent contact with scab materials or directly through variolation practices.

Possible clues regarding the adaptation of VARV to humans can be found in the close relationship between VARV and TATV/CMLV (4). TATV is associated with a terrestrial rodent native to West Africa (35). Our coalescent analyses indicate that the divergence between VARV and TATV occurred from 16,000 YBP (Table 2 and SI Dataset 4, based on the smallpox historic records of South Africa) to 68,000 YBP (Table 2 and SI Dataset 3, based on the earliest recorded smallpox history in East Asia). Thus, like the related zoonotic orthopoxviruses with rodent reservoirs (6, 36, 37), VARV may have evolved from an enzootic pathogen of African rodents and subsequently spread out of Africa.

The genomic sequences of VARV isolates selected for this study were taken from the World Health Organization Collaborating Center repository and, although they display a low level of sequence divergence, are representative of the greatest geographic and temporal diversity within the collection. The cSNP-derived phylogeny of these isolates produces a topology consistent with historical human migratory patterns, population expansion, and documented smallpox epidemics. The depicted topology is supportive of a hypothesis (Fig. 3) that two clades of variola diverged from an ancestral virus and subsequently evolved in geographically discrete human populations. Although the reason for the different human pathogenesis of alastrim and VARV major is unclear, the genetic structure of alastrim minor and its topology with the other VARV isolates add critical pieces to the smallpox evolutionary puzzle. This proposed phylogeny of VARV extends the evolutionary time of smallpox relative to that of previous theories.