Previous studies 620 have shown high levels of variation between the genomes of different strains of obligate intracellular bacteria. To compare the levels found in A. marginale to other genomes, similar comparisons were made for organisms meeting the following criteria: organisms with 1) a single chromosome, 2) more than one sequenced strain, and 3) assembled and finished genome sequences deposited in Genbank, including free-living, facultative intracellular, and obligate intracellular bacteria (Figure 1). Single-factor analysis of variance (ANOVA) finds no significant differences in the level of variation between obligate intracellular, facultative, and free-living bacteria. The number of SNPs ranged from 0.00% to 8.00% of the larger genome, with significant intraspecies and intragenera variation.

The Anaplasma marginale Florida strain genome is composed of a single 1,202,435 bp circular chromosome predicted to contain 942 coding sequences (CDS) (Table 1). Similar to most other previously sequenced Anaplasmataceae, there are no plasmids and no identifiable insertion sequences. Compared to the previously sequenced St. Maries strain genome 11, there are seven fewer CDSs despite the larger genome size, due primarily to differences in split open reading frames (ORFs) and annotation differences. The high degree of synteny between these two strains is disrupted by two inversions; one approximately 30 kb long is flanked by repeat elements (msp3 pseudogenes), while the other is a single gene flanked by short duplicated hypothetical genes.

Split ORFs were first described in the Rickettsia conori genome 21, and are postulated to represent genes that are in the first stage of reductive evolution. The idea that these ORFs have split recently is consistent with the findings in Anaplasma, as different ORFs are split in the two completely sequenced strains. The four split ORFs annotated in the St. Maries genome (mutL, murC, aatA, and aspS) 11 are intact in the Florida genome, and two tandem genes annotated as hypothetical in the St. Maries genome (AM574 and AM576) are fused in the Florida genome (AMF_437). Only one split ORF, petA, is found in the Florida genome. Four small ORFs in the St. Maries genome (AM380, AM395, AM974, and AM976), ranging in size from 204 bp to 378 bp are not present in the Florida genome. These ORFs are flanked by repetitive DNA sequences, and appear to be missing due to recombination events.

The msp2 superfamily is a group of related A. marginale genes encoding surface proteins 11. Msp2 encodes a highly antigenic protein that varies over time during infection by gene conversion of functional pseudogenes into a single expression site, to create new antigenic variants capable of evading the existing immune response. Compared to the St. Maries genome, the Florida genome has one additional msp2 functional pseudogene. Of the eight Florida msp2 functional pseudogenes, four are identical to those in the St. Maries genome. The Florida genome has two sets of duplicated functional pseudogenes, TTV 4F15/TTV 1O6 and KAV 4F15/KAV 1F20 (Figure 2); while St. Maries was found to have duplicated functional pseudogenes, this was not noted in a functional pseudogene-targeted examination of other strains 22. Florida has a set of duplicated functional pseudogenes in the same genome positions as St. Maries (2/3H1 in St. Maries, and KAV 4F15/KAV 1F20 in Florida). As obligate intracellular bacteria are not thought to undergo lateral gene transfer, identical functional pseudogenes indicates the sequence is either evolutionarily conserved or has been selected independently in both strains due to a fitness advantage. Interestingly, both copies in Florida have a change encoding 15 amino acids at the 5' end of the hypervariable region compared to their St. Maries counterparts; either both strains duplicated a functional pseudogene after this change occurred in an ancestral strain, or both copies in one of the strains acquired identical changes after the ancestral strain duplicated the original functional pseudogene. In contrast, only two of the seven MSP3 functional pseudogenes are identical between Florida and St. Maries (msp3 C/msp3-1, and msp3 4L1/msp3 6). The omp1-15 genes are present in both genomes, with a high degree of conservation between the predicted amino acid sequences (85.3–100% identity) as previously reported 17.

The aaap gene was first recognized and characterized as an Anaplasma appendage associated protein 23. Subsequently, additional related genes were identified that appear to be tandemly-duplicated copies that have diverged to have relatively low levels of sequence identity (Table 2). There is expansion of this locus in the Florida strain relative to the St. Maries strain, with a duplicated copy of the aaap gene. Because of the repetitive nature of this gene family, these sequences tend to be missing from pyrosequenced genome assemblies; therefore, we examined the status of this locus in several the strains via Southern analysis, revealing that this locus is highly plastic both within and between strains (Figure 3).

An additional two transmissible strains (Virginia and Puerto Rico) and one non-transmissible strain (Mississippi) were subjected to genome-scale pyrosequencing 24 (454 Life Sciences, Branford, CT), which provided at least 96% genome coverage when compared to either Florida or St. Maries (Table 3). Most of the missing sequences corresponded to repetitive regions (such as msp2 and msp3 pseudogenes and the aaap locus) (Figure 4) [also see additional files 1 and 2], and reflects the limitations of assembling short sequence reads (averaging approximately 250 bp per read) without additional scaffolding. No new genes were detected in the pyrosequenced contigs of any of the strains.

Global comparison of all strains with the Florida strain revealed 20,028 total sites with a single nucleotide polymorphism (SNP) in at least one of the compared strains. Of these, 511 (2.6%) were different in the Florida genome and identical in the other four strains, and 13,316 (66.5%) were unique to one of the four strains (Figure 5). The remaining 30.9% of SNPs represent those SNPs relative to Florida that were present in two or three of the strains. There were 9,609 SNPs between the Florida and St. Maries strains, comprising 0.80% of the larger Florida genome. The SNPs were distributed evenly throughout the genome, which is similar to both Ehrlichia ruminantium and Rickettsia bellii [see additional file 3], and are proportionally distributed throughout coding and non-coding regions. The numbers of polymorphisms in the Puerto Rico, Virginia, and Mississippi strains (2,729, 3,868, and 6,773, respectively) are minimums, as the gaps are regions predicted to have significant numbers of SNPs. When the genome size was corrected for the gaps in coverage, the SNP rates for the Puerto Rico, Virginia, and Mississippi genomes were 0.32%, 0.46%, and 0.73% of the Florida genome, respectively.

All animal experiments described in this paper were approved by the Washington State University Institutional Animal Care and Use Committee (IACUC), with approval number 3386.

The Florida strain [GenBank: CP001079] of A. marginale was originally isolated from a pool of blood samples collected from cattle in 1955 163132. The Mississippi strain [GenBank: ABOP00000000] was isolated from an acute clinical case of anaplasmosis 2533. Both of these strains are virulent, and are not transmissible by D. andersoni ticks. The Virginia strain [GenBank: ABOR00000000] was isolated from a cow in Southern Virginia in 1972 34. The Puerto Rico strain [GenBank: ABOQ00000000] was received as a frozen stabilate after isolation from cattle in Puerto Rico in 1985 3536. Both the Virginia and Puerto Rico strains are virulent, and are transmissible by D. andersoni. While passage histories are not well documented for these strains, all strains have been passaged multiple times in cattle since isolation. The Florida strain has the longest passage history, being passed continuously since isolation, and the Puerto Rico strain has only been passaged once since coming to our laboratory.

Blood stabilates from the Florida, Mississippi, Puerto Rico, and Virginia strains were inoculated into splenectomized calves, which were shown to be free of A. marginale infection via competitive enzyme-linked immunoabsorbent assay (cELISA) 37. Blood samples were taken at peak parasitemia, washed seven times in phosphate buffered saline (137 mM NaCl, 10 mM Phosphate, and 2.7 mM KCl), and centrifuged at 1,500 × g for 10 minutes with the removal of the buffy coat after each spin. Erythrocytes not used immediately were diluted 1:1 in PBS and frozen for later use.

After thawing, lysed erythrocytes were passed over a column containing loosely-packed CF-11 cellulose (Sigma-Aldrich Corporation, St. Louis, MO). The column eluate was washed repeatedly with PBS and centrifuged at 19,000 g for 20 minutes until all remaining hemoglobin was removed, leaving a pellet of A. marginale initial bodies and erythrocyte membranes.

The bacterial preparation was embedded into 1% agarose blocks (A-9539, Sigma Chemical Co., St. Louis, MO), and cells within the blocks were lysed using proteinase K and SDS 38. A. marginale genomic DNA was partially digested with either HindIII or MboI, size selected on pulse-field gels, ligated into the pBELOBAC11 vector, and electroporated into Escherichia coli strain DH10B (Amplicon Express, Pullman, WA). A total of 3,072 clones (1,536 clones from each restriction enzyme) were arrayed into 384 well plates. The average insert size of the clones was 120 kb.

For the Florida strain, a BAC-based clone-by-clone strategy was adopted. BAC clones were screened using digoxigenin (DIG)-labeled (Roche Applied Science) probes to bovine genomic DNA and several A. marginale genes (including msp1α, msp1β, msp2, msp3, msp4, msp5, dnaK, recA, groEL, and sodB). Selected clones were end-sequenced and a minimum tiling path was constructed based on comparison with the previously-sequenced St. Maries strain. Sequencing of BACs, assembly of completed sequences, and genome annotation were as described 11.

Genomic DNA from the Mississippi, Puerto Rico, and Virginia strains was extracted from isolated bacteria (prepared as described above) using the Puregene Blood kit (Qiagen Corporation, Valencia, CA). DNA was then sequenced on a Genome Sequencer 20 instrument (454 Life Sciences Corporation, Branford, CT), using a pyrosequencing protocol 24. The Newbler program was used with its default settings to assemble the sequence and to compare all contigs to the completed Florida and St. Maries genomes, which revealed the location of gaps in coverage. High-quality variations were called when four reads, each with at least 20 base pairs flanking the polymorphic site, contain the difference, with at least one read in each direction. The nucleotide sequences of assembled contigs were compared to the St. Maries genome using BLASTn. Any contigs without hits better than 1e^-10 were then compared to the bovine whole genome shotgun sequence database, to screen for bovine DNA contamination. Any contigs with no hits to bovine sequence were then compared to the nt database. Large contigs assembled by Newbler were compared to the Florida genome using MUMmer v3.1 39 after filling gaps in the assembly with the corresponding sequence from the Florida strain.

MUMmer v3.1 was used to compare the completed St. Maries and Florida genomes and the contigs of the Mississippi, Puerto Rico, and Virginia strains, as described 5. Output from the SNP detection algorithm (show-snps) was processed using custom scripts (written with AutoIt v3.2.2.0) to determine the number of SNPs per ORF. Show-snps output was also processed in Excel (Microsoft Corporation, Redmond, WA) to graph the location of SNPs throughout the genome. FASTA sequences from single-chromosome genomes with multiple strains sequenced were downloaded from Genbank [see supplementary table 2 for the genomes compared, their sizes, and Genbank accession numbers]. All strains for a given species were compared to each other using MUMmer 3.1, as described above. The number of SNPs per comparison was then divided by the larger of the two compared genomes to yield the percent SNPs per genome. For species with more than two strains sequenced, all percentages were averaged to give the mean and standard deviation. The phylogenetic tree was inferred using the Maximum Parsimony method 40 of MEGA4 41 comparing concatenated sequences from groEL, groES, atpA, and recA. The bootstrap consensus tree is inferred from 1000 replicates, and branches corresponding to partitions reproduced in less than 50% bootstrap replicates are collapsed. There were a total of 458 positions in the final dataset, out of which 379 were parsimony informative.

Genomic DNA from all five strains was digested with XbaI and HindIII (New England Biolabs Corporation, Ipswich, MA), as these enzymes cut within the conserved flanking genes. Resultant fragments were separated on a 0.8% agarose gel, and subsequently transferred to a charged nylon membrane and crosslinked with a Stratalinker UV apparatus (Stratagene Inc., La Jolla, CA) per the manufacturer's directions. The blots were prehybridized at 42°C for at least two hours in Dig Easy Hyb buffer (Roche Corporation, Indianapolis, IN). Digoxigenin-labeled probes to aaap were produced using the PCR DIG Probe Synthesis Kit (Roche Corporation) and hybridized overnight at 42°C in DIG Easy Hyb buffer. The membrane was washed three times for 15 minutes in 2 × SSC and 0.1% SDS, with the first two washes at room temperature and the third at 65°C. A final wash was performed in 0.2 × SSC and 0.1% SDS at 65°C. Chemiluminescent detection of the probes was performed using the DIG Wash and Block Buffer Kit (Roche Corporation) per the manufacturer's directions.