Ancient Chinese medical books suggest that an anthrax-like disease has been present in China for more than 5,000 years and that by 500–600 A.D. the epidemiology and symptoms of anthrax had been described [1]. A 1995 report from China described the results of an anthrax surveillance and control project in 10 provinces in China between 1990–1994 [2]. Stations in these 10 provinces (Sichuan, Tibet, Inner Mongolia, Xinjiang, Qinghai, Gansu, Guangxi, Guihou, Yunnan and Hunan) reported 72 outbreaks and 8,988 human cases of anthrax. These results, which are indicative of a long history and significant levels of contamination in these specific areas, are the reason for concern by the Chinese Institute of Epidemiology and Microbiology [2].

The population structure of Bacillus anthracis has only recently begun to be resolved with specific geographical patterns spread across areas mostly inhabited by man and his animals. Higher genetic resolution within B. anthracis has resulted from two molecular typing approaches: An ongoing comparative, single nucleotide polymorphism (SNP) analysis of diverse isolates that describes a conserved, clonally derived basal tree, [3] and a multiple locus variable number tandem repeat analysis (MLVA) system that provides improved resolution among individual isolates [4–7]. This process for molecular typing has now been applied to the study of isolates from China.

An archival collection of 191 B. anthracis isolates from China [collection dates from 1947–1983, except isolates A0034 (1993) and A0038 (1997)] was obtained and used in this study (see Methods and Additional file 1). This collection contained an unusual subset of 122 B. anthracis isolates recovered from soil, including 107 isolates collected between 1981/1982 in Xinjiang province. This province is located in the western most tip of China and was one of the 10 regions surveyed in the study conducted from 1990–1994. The remaining isolates originated from many regions across the whole of China. This report focuses on the molecular genotyping of these 191 isolates. Our goal was to determine the nature and distribution of genotypes found in China and to establish phylogenetic relationships between these isolates and those found elsewhere in the world.

Canonical SNP analysis

The original comparative analysis of 5 B. anthracis whole genome sequences examined the status of ~1,000 single nucleotide polymorphisms (SNPs) in 26 diverse isolates [3]. This study revealed an extremely conserved phylogenetic tree with only one homoplastic character in ~26,000 measurements. These results prompted the hypothesis that a few strategically placed "canonical SNPs" could replace the 1,000 assays and still describe an accurate SNP based tree. This idea was confirmed in a study using 13 canonical SNPs (canSNP) to examine 1,000 world-wide isolates of B. anthracis [5]. Figure 1 illustrates this original canSNP tree and is used here to define important nomenclature and terminology.

The basic tree is now defined by 7 sequenced genomes that form 7 sub-branches or sub-lineages ending in "stars" in Figure 1. Each of these sub-lineages is designated by the nomenclature from the whole genome sequence site in Genbank, e.g. A.Br Ames, A.Br.WNA (for western North America), and A.Br.Vollum. The relative position of each canSNP is indicated by vertical script and a small arrow and is arbitrarily defined, e.g., as A.Br.001 where A refers to the major subgroup and 001 is the first canSNP (see the A.Br.Ames sub-lineage in Figure 1, also [5]). In this case the derived A.Br.001 SNP defines all isolates that are on the same branch as the sequenced Ames strain. In addition to these 7 sub-lineages the analysis of 26 diverse isolates uncovered 5 nodes or sub-groups along the branches of this tree. Four of these nodes are in the major A Branch and one is in the B Branch (see "circles" in Figure 1). These nodes are defined by the two canSNPs on either side of the node position, e.g. A.Br.001/002 or A.Br.008/009. All of the initial 1,000 isolates in the Van Ert study [5] were placed into one and only one of these 12 sub-lineages or sub-groups.

Human anthrax has been an old and continuous problem in many rural regions in China where as much as six percent of environmental samples have been found to be contaminated with B. anthracis [2, 2]. An archival collection of 191 B. anthracis isolates was obtained from China and canonical SNP typing indicated that only 5 of the 12 worldwide sub-lineages/sub-groups of this pathogen were represented in this collection. One striking feature of the distribution of these B. anthracis isolates within this country was the discovery that three of the five canSNPs sub-lineages/groups (A.Br008/009, A.BrAust.94, and A.Br.Vollum) are predominantly found in the western most Chinese province of Xinjiang. The previous observation [5] that these three sub-lineages/sub-groups are prominent genotypes in India, Pakistan, Turkey and most of Europe suggest a likely transmission pattern for anthrax along the ancient trade route known as the Silk Road [11] that extended from Europe, the Middle East, portions of Asia and into Xinjiang province and the whole of China, Figure 2.

More specifically, 107 isolates were recovered from "soil samples" between 1981–1982 from unspecified sites relatively close to the city of Kashi in this province. Kashi (also Kashgar, Kaxgar, Kǝxkǝr) was a major "oasis" crossroads city along the ancient Silk Road and dates back more than 2,000 years [11]. Consistent with the idea that the life cycle of B. anthracis can be maintained by viable spores in previously contaminated areas, the later 1990–1994 surveillance project in China described three regions in Xinjiang Province where severe anthrax outbreaks had previously occurred [2]. Two of these towns, Zepu and Atushi, are located approximately 144 and 33 kilometers respectively from the city of Kashi. In the 1990–1994 study, Zepu recorded 24 villages with 202 human infections and Atushi recorded 4 villages with 81 human infections.

Despite a clear correlation between canSNP genotypes from the A radiation and the spectrum of isolates found across the Trans-Eurasian continents, there is one set of genotypes in Europe that are clearly missing in China. These are representatives from the B branch that appear to be prevalent in several European states including at least 27 B2 isolates from France and isolates identified in both the B2 and B1 branches from Croatia, Germany, Poland, Italy, Norway and Slovakia [5, 6, 12]. It is not obvious why examples of the B branch are limited mostly to Africa, this region of Europe and a small location in California, USA. Aside from sampling issues the B branch does not appear to have participated in the world-wide, dynamic radiation that has characterized the A branch [5].

Additional analyses with the rapidly evolving MLVA markers suggest that establishment in China of two of these sub-groups/sub-lineages, A.Br.Aust94 and A.Br.Vollum, resulted from relatively recent events (Figure 3a and 3b). In both of these instances, a sizeable number of isolates (44 and 15, respectively) are clustered into only three different MLVA15 genotypes (Nei's Diversity Indices = 0.031 and 0.038 respectively, Figure 2). Although these results may reflect a certain sampling bias, the MLVA comparison to other worldwide isolates from this branch indicates that the A.Br.Aust94 sub-lineage in China is most closely related to isolates recovered from the large 1997 outbreak in Victoria, Australia (data not shown). The precise origin and time-scale for this exchange is not certain but relatively recent exchanges between the Far East and Australia appear to have originated from India [13], which could represent a common ancestor or an intermediate step in the transmission route.

By direct contrast the MLVA analysis of 49 isolates belonging to the A.Br.008/009 sub-group revealed a more complex pattern with 14 different MLVA15 genotypes (Nei Diversity Index = 0.143, Figures 1 and 3c). This is a remarkable finding because it indicates that a variety of MLVA genotypes are persisting in the different soils from which the A.Br.008/009 isolates were recovered. These results are an indication that A.Br.008/009, a major sub-group in Europe and Asia [5], has had an extensive history in China. It is difficult to determine the precise origins of the A.Br.008/009 subgroup (e.g. China versus Europe) at this point because rapidly evolving MLVA markers are subject to homoplasy and potentially inaccurate phylogenetic reconstructions. These issues can eventually be resolved using additional whole genome sequencing and phylogenetic inference to more accurately predict the origins of the A.Br.008/009 sub-group.

The Ames sub-lineage appears to have descended from the A.Br.001/002 sub-group, a sub-group that has 106 isolates in our worldwide collection [5]. Seventy-four of these accessions were isolated from outbreaks in China and the remaining 32 isolates were recovered in the UK, other parts of Europe, North America and other parts of Asia. The large number of MLVA15 genotypes (n = 32) among the 74 Chinese isolates and a wide distribution throughout the country indicates that the A.Br.001/002 sub-group is a major part of the B. anthracis population structure in this region (Figure 5a). This sub-group also appears to be basal to the Ames sub-lineage, indicating that 8 isolates from China and 11 isolates from Texas may share common ancestors that originated in China (Figure 5b and [10]).

How then did the Ames lineage come to Texas and why is this lineage not found in Europe? This is still not known and subject to considerable speculation. By several accounts, it is believed that anthrax was introduced into the Gulf Coast states (Louisiana and Texas) by early settlers from Europe. Stein [14, 15] indicates that the first recorded episodes of anthrax in livestock in Louisiana occurred in 1835, 1851 and 1884; and in Texas in 1860 and 1880. By 1916, when a first national survey was conducted to obtain nation-wide information on the incidence of anthrax, Texas already had 41 counties reporting infections. A composite of outbreaks compiled after the 4^th National Survey by the U.S. Department of Agriculture between 1916–1944 (Figure 6) indicates three major outbreak pockets: one in California, one in the Dakotas/Nebraska and the third along the coastal regions of Texas and Louisiana [15].

An important feature of the outbreaks in Texas is that the "modern" outbreaks have occurred repeatedly in many of the same counties depicted in this historical map (Figure 6 and USDA Report: Epizootiology and Ecology of Anthrax: http://www.aphis.usda.gov/vs/ceah/cei/taf/emerginganimalhealthissues_files/anthrax.pdf). A culture-confirmed study between 1974–2000 indicated that 179 isolates were spread across 39 Texas counties (counties outlined in yellow) that are in general agreement with the dispersal patterns observed in the early national surveys depicted in Figure 6. The one significant difference is a shift from the historical outbreaks in the coastal regions to counties more central and southwesterly in "modern" times. Similarly, culture-confirmed isolates from a 2001 outbreak in Val Verde, Edwards, Real, Kinney and Uvalde counties in southwest Texas are similar to outbreaks in 2006 and 2007 when 4 Ames-like isolates were recovered from Real, Kinney, and Uvalde county [9].

It appears that B. anthracis was introduced into the Gulf Coast, probably by early European settlers or traders through New Orleans and/or Galveston during the early to mid 1800s. The disease became established along the coastal regions and then became endemic to the regions of Texas where cattle and other susceptible animals are currently farmed. Are these B. anthracis, Ames-like genotypes from the Big Bend region (Real, Kinney, Uvalde counties) of Texas representative of the ancestral isolates brought to the Gulf Coast? Van Ert et al. [5] used synonymous SNP surveys to estimate the divergence times between the major groups of B. anthracis and these estimates suggest that the Western North American and the Ames lineages shared common ancestors between 2,825 and 5,651 years ago. Extrapolating to the much shorter SNP distances between the most recent Chinese isolate (A0728) and the recent Texas isolates on the Ames sub-lineage would approximate that these two shared a common ancestor between 145 to 290 years ago. These estimates would be consistent with the hypothesis that an Ames-like isolate was introduced into the Galveston and/or New Orleans area in the early to middle 1800s.

This relatively recent expansion is in direct contrast to analyses of the Western North American (WNA) sub-lineage that appears to have an ancient and significantly longer evolutionary presence in North America; this group stretches from the central regions of Canada and into North and South Dakota (Figure 6; [16]). Phylogenetic reconstruction of > 250 Western North American isolates indicates that the more ancestral isolates of this sub-lineage are found in the upper reaches of central Canada and portrays a migration pattern where the youngest isolates are found in cattle outbreaks in North/South Dakota and Nebraska. Kenefic, Pearson et al. [16] suggest that the ancestral isolates may have entered the North American continent via the Beringian straights 13,000 years ago.

A recent ecological niche model suggests that natural anthrax outbreaks are "concentrated in a narrow corridor from southwest Texas northward into the Dakotas and Minnesota" [17]. This model indicates that conditions like vegetation, precipitation and altitude along this corridor are suited for maintaining naturally occurring anthrax outbreaks in livestock and wildlife. Although historical records provide evidence that validate this model, there is a molecular and genotyping anomaly: there does not appear to be a direct epidemiological link between the "younger" Ames-like cluster and the Western North American lineage. Despite nearly 100 years of monitoring since the first national outbreak tabulations [15], there is still a clear physical division between the Ames-like isolates to the south and the Western North American lineage to the north (Figure 6). This gap is not obvious until the spatial patterns are examined in hindsight of the genetic discontinuity. These observations probably reflect the awareness and controls that were being observed for anthrax outbreaks as the US entered the 20^th century.

Limited sample analysis of isolates from the Texas/Louisiana coastline prevents any conclusions about the overall dominance of the Ames sub-lineage in this area and we also cannot exclude the possibility that there are other sub-groups/sub-lineages that might have been imported and even become transiently established along the Texas/Louisiana Gulf region during this same time frame.

Despite containing only 5 of the initial 12 canSNP genotypes used to define a collection of world-wide isolates [5], the analysis of 191 Chinese B. anthracis isolates reveals an interesting impact on global distribution. The major diversity in these isolates is concentrated in the western province of Xinjiang and especially the city of Kashi, the hub of the Silk Road around the Taklimakan Desert into and out of China. These results reinforce the idea that this Silk Road region was central to the spread of anthrax between the trans-Eurasian continents.

In addition to the three distinct sub-groups found in the western Xinjiang province, the central and eastern regions of China are dominated by a different, highly diverse, canSNP sub-group, A.Br.001/002. This sub-group is a major presence in relationship to our world-wide collection since 70% of all the isolates and most of the diversity for this sub-group were in this Chinese collection. These results suggest that the A.Br.001/002 cluster may have originated in China. Finally, the Ames and Ames-like strains in Texas are descended from common ancestors in Inner Mongolia in China as an extension of this sub-group. It is curious that this lineage would become established in Texas, and perhaps Louisiana, and not in Europe. This leaves behind a missing historical gap within the phylogeography of the Ames lineage.