The most conserved of these subunits are VirB3, VirB8 and VirB10, with few differences between strains. VirB3 has been linked in Agrobacterium tumefaciens with pilus assembly and substrate translocation 32 33 . It is absolutely conserved between strains with no amino acid changes and conforms to the typical VirB3 structure. Two alpha-helical domains for insertion into the cytoplasmic membrane are strongly predicted by TMpred. VirB8, proposed to function as a nucleation factor during the assembly of T4SS 34 35 , is also well conserved, particularly VirB8-1 in the polycistronic transcription locus (one amino acid change between all strains). VirB10, proposed to function as a scaffold across the entire cell envelope 36 , is also generally well-conserved with the exception of one ruminant strain, ApNorLamb-V1, which has 31 amino acid substitutions with respect to ApHZ (data not shown). However, all A. phagocytophilum VirB10^′s, including ApNorLamb-V1, have two strongly predicted transmembrane domains, which supports their function as membrane scaffolding subunits in these organisms.

Of these inner membrane channel subunits, the data on VirB6 are the most interesting. All VirB6 subunits that have been described possess a highly hydrophobic membrane domain including five or more predicted transmembrane domains 20 . Some VirB6 proteins also have an extended C-terminal hydrophilic domain that has been proposed to protrude through the T4SS into the target cell, or may be proteolytically released from the N-terminal domain and then translocated into the target cell. Evidence has been obtained for surface exposure of extended VirB6 in some Rickettsiales 37 . Of all the membrane channel subunits, the most sequence diversity between A. phagocytophilum strains was in the four VirB6 paralogs (Figure 1). Although there were no amino acid changes in the VirB6-1, VirB6-2 and VirB6-3 paralogs between human, dog and rodent strains, the ruminant and horse strains had numerous substitutions throughout each molecule, agreeing with the closer evolutionary relationship between strains infecting humans and dogs in the U.S. (Figure 2). Furthermore, major differences in repeat number and sequence were found in the C-terminal repeat region of VirB6-3 (yellow boxes in Figure 3A and Additional file 1: Figure S1) in ruminant and horse strains, with the horse strain showing the least variability from ApHZ.

The only amino acid differences detected between the human, dog and rodent strains were in the VirB6-4 subunit. VirB6-4 in these strains contains four repeat regions (R1-R4 in Figure 3A) and variability in repeat number, order and sequence were found mainly in R3 and R4 (Additional file 2: Figure S2). Within R1 (Figure 3A), the only difference detected was in ApDog2 which had 4 and 1 partial of 231 bp repeat units (data not shown), compared to 3 and 1 partial repeats in the ApDog1, ApJM and ApHZ virB6-4 R1. Optical mapping of the Dog1 genome and comparison with ApHZ suggested that the sequence obtained previously for the human HZ strain virB6-4 was incorrect (Figure 3B). This was confirmed by PCR and sequencing, and mapped specifically to the 3^′-most R4 region (Figure 3C). Because of its size and unusual composition it was only possible to resolve this sequence using the long read-length Pacific Biosciences technology (see Methods). The corrected virB6-4 R4 of ApHZ, totaling 6.89 kb, differed from the original by 5.88 kb of additional sequence composed exclusively of 84 bp [type 1, a and b (T1a, T1b); light/dark blue boxes, respectively, in Figure 3A]and 162 bp [type 2, a and b (T2a, T2b); light/dark orange boxes, respectively, in Figure 3A] repeat units, giving a complex repeat structure containing 53 and 1 partial repeat units compared to 8 and 1 partial in the original sequence. Further, the 5^′- and 3^′-most 2.7 kb of this complex structure are identical in sequence, and the 3^′-most 1.15 kb of each of these segments is repeated again in the center of R4 (Figure 3A and Additional file 2: Figure S2). Although the possibility exists that the ApHZ population from which we isolated gDNA differs within the virB6-4 R3/R4 repeat regions from the population used to generate CP000235, the fact that all strains investigated herein presented expansive R3/R4 regions (Figure 3C) would contradict that. Instead, it is more plausible that the existence of 2.7 kb of identical repeats at the ends of the ApHZ R4 may have lead to the excision of most of its sequence during construction/propagation of those libraries. Interestingly, virB6-4 R3 and R4 were identical both in size and sequence in the Dog1 and rodent strains despite differing markedly from the HZ and Dog2 strain regions (Additional file 2: Figure S2A). Within R3, these strains had 2 additional 405 bp repeats compared to ApHZ and one more compared to the Dog2 strain. However, differences between strains were most dramatic within R4. Not only was this region in ApDog1/ApJM 2.87 kb larger than in ApHZ bringing the total number of repeats to 81 and 1 partial, but intriguingly, the repeat pattern was completely unrelated to that in the HZ strain. Also, the Dog1 and rodent strain R4 lacked T1b repeat units, while having a third type 2 repeat variant, namely T2c, which differed from T2b by 1 SNP and a 12 bp deletion (Additional file 2: Figure S2). Partial analysis of the ApDog2 454 reads spanning R4 (estimated at ~8 kb by PCR; Figure 3C) showed that the order of the 5^′- and 3^′-most three repeat units differed from either the HZ or Dog1/rodent strain R4 repeat patterns (Additional file 2: Figure S2A). Notably, our preliminary analyses of the horse and ruminant 454 reads suggest the absence of distinct R3 and R4 regions in virB6-4 in these strains. Rather, the few repeat units identified to date appear to be a combination of R3 and R4 repeats (data not shown). It is also unclear if the ~17 kb and ~25 kb PCR products generated with primers AB1393/1466 in ApVar-1 and ApNorLamb-V2, respectively (Figure 3C), are composed mainly of repeats, or alternatively if a fifth virB-6 gene paralog exists in these strains. Taken together, the data presented here clearly demonstrate the extreme variability of the T4SS VirB6-4 subunit among A. phagocytophilum strains. Although the differences between the more closely related human, dog and rodent US strains were mainly within repeat-laden regions, the fact that an extensive, distinct repeat pattern was maintained in two strains would speak against the possibility that the variability may be attributed solely to the highly recombinogenic nature of such structures. Worth noting, Camp Ripley, where the infected jumping mouse was captured (2001) is only ~20 miles away from the city of Baxter, MN, where Dog1 resides. Although there are no records of where this dog may have actually acquired the infection, it presented with severe clinical disease in 2007.

The unusual structure and likely antigenicity of the C-terminal region of the A. phagocytophilum VirB6-4^′s is apparent in hydrophobicity plots (Figure 4). What specific properties these distinct repeat patterns may confer onto each strain awaits functional analysis of these proteins in A. phagocytophilum. The corrected VirB6-4 translated protein had a predicted molecular weight of 470,695 Da containing 4,322 amino acid residues compared to molecular weights of 90,742, 103,204 and 158,321 Da for the HZ strain VirB6-1, VirB6-2 and VirB6-3, respectively. Interestingly, the predicted acidity of the VirB6^′s also increased from VirB6-1 to VirB6-4 (pI’s of 8.4, 6.8, 5.1 and 4.0 for the ApHZ VirB6-1, VirB6-2, VirB6-3 and VirB6-4, respectively). The ApDog1/ApJM VirB6-4 polypeptides had a predicted molecular weight of 603,529 Da containing 5,550 amino acids, and a pI of 3.96. Despite these dissimilarities, at least eight transmembrane segments were predicted for all VirB6 paralogs.

Several other T4SS subunits contribute to the secretion channel across the periplasm and outer membrane. VirB7 subunits are typically small lipoproteins that may stabilize VirB9 38 39 . In A. phagocytophilum strains a putative VirB7 is absolutely conserved between strains and may be lipid modified through an N-terminal cysteine on the mature molecule. VirB9 is hydrophilic and also localizes to the periplasm and outer membrane. In A. tumefaciens the C-terminal region of VirB9 is part of the outer membrane protein channel and is surface accessible 40 . There is also evidence for surface exposure of VirB9 in Ehrlichia chaffeensis and A. phagocytophilum 18 19 41 . VirB9-1, which is encoded on the polycistronic virB8 1 virD4 transcript 31 , has a strongly predicted signal peptide and two transmembrane helices. Of all the potentially exposed components of the T4SS, VirB9 of A. phagocytophilum appears to be the least diverse among strains. There are some amino acid substitutions in ruminant and horse strains (2–6 total compared to ApHZ) but in the other strains VirB9^′s are unchanged (data not shown).

Unlike VirB9^′s, VirB2^′s are the most diverse of all T4SS subunits in A. phagocytophilum, in terms of both copy number and sequence. VirB2 proteins are typically constituents of pili and of the secretion channel and their diversity in Anaplasma suggests the possibility of exposed, antigenically variable structures. In A. marginale, VirB2 is expressed together with the major outer membrane protein MSP3 on a sequence-variable polycistronic transcript 25 42 . The mechanism of expression in A. phagocytophilum is not known. VirB2^′s of other genera are typically small hydrophobic proteins with a long signal peptide sequence and two hydrophobic alpha helices for integration into the cytoplasmic membrane. This also appears to be the case for A. phagocytophilum. The VirB2 paralogs in the different strains are predicted to have two hydrophobic alpha-helices of lengths 22+/−3 and 20+/−0.2 amino acids and signal peptides of length 27+/−2 amino acids. This is true despite their sequence diversity (Figure 5). As with many other T4SS components, the ruminant and horse strains are more distant taxonomically in VirB2 sequence compared to VirB2^′s of human and dog strains. Alignment of all VirB2 paralogs and orthologs shows that sequence diversity is primarily localized to two hypervariable regions either preceding an N-terminal cysteine or close to the C-terminus (Figure 6). This is similar to the hypervariable regions found among VirB2 paralogs of A. marginale 25 .

ATPases are typically used in T4SS to energize substrate transfer and have been found in every T4SS described. In gram-negative bacteria these are typically integral membrane proteins encoded by genes residing upstream of virB2 (encoding pilin). This is true for all strains of A. phagocytophilum and it has been suggested that this arrangement of multiple virB2 paralogs and virB4 2 may allow assembly of an antigenically variable surface organelle 20 . The energetic subunit itself, VirB4-2, is however, well conserved between strains. The most distant taxonomic relationship was found between human and ruminant strains (29 total amino acid substitutions in ApNorLamb-V1 compared to ApHZ, Figure 7). The other energetic subunit, VirB11, was also well-conserved between strains (6 amino acid substitutions between ApNorLamb-V1 and ApHZ; data not shown).

Type 4 coupling proteins such as VirD4 are ATPases that function in substrate recognition and translocation using the T4SS. They are associated with most effector translocator systems. They typically possess a minimum of two N-terminal transmembrane domains. Often most heterogeneity exists in these N-terminal regions 20 . The A. phagocytophilum VirD4^′s conform somewhat to this stereotype with three strongly predicted N-terminal transmembrane segments. As with the other ATPases of the A. phagocytophilum T4SS, there is little variation in VirD4, a total of 17 amino acid substitutions of which 4 are N-terminal but more (12) are C-terminal. Again, the evolutionary relationships among VirD4 sequences position the ruminant and horse strains more distantly to the U.S. dog, human and rodent strains (Figure 8).

The A. phagocytophilum U.S. strains HZ (human-origin, NY), MRK (horse-origin, CA), JM (rodent-origin, MN) and Dog1 (dog-origin, MN) were propagated in HL-60 cells in RPMI-1640 medium (Thermo Fisher Scientific, Inc., Waltham, MA) supplemented with final 10% heat-inactivated fetal bovine serum (Thermo Scientific) and 4 mM L-glutamine (Lonza, Rockland, ME), and in the absence of antibiotics. ApHZ and ApMRK have been described previously 15 44 . The ApJM strain (CR01-1258) originated from a meadow jumping mouse (Zapus hudsonius) trapped at Camp Ripley, MN 45 . The ApDog1 strain originated from the blood of a dog from Baxter, MN naturally infected with A. phagocytophilum, as evidenced by the detection of distinctive morulae in a diagnostic blood sample, and sequencing of the Expression Site-linked msp2/p44 gene. Briefly, whole blood was collected from the animal with EDTA as an anticoagulant. The buffy coat layer was collected after low-speed centrifugation of the whole-blood, washed in 1x phosphate buffered saline (PBS, Hyclone, cat. no. SH30256.01), then added to a culture of uninfected HL-60 cells. The culture was left undisturbed for 3 days, after which morulae began to appear. The ApDog2 strain also originated from a MN dog and was passaged to and maintained in the Ixodes scapularis ISE6 tick cell line as described 46 . The Ap variant 1 CRT35 strain (tick-origin, MN), maintained in ISE6 cells, has been described 47 . For DNA isolation, cultures were maintained until 90-100% of cells were infected with mature morulae. Cells were pelleted by centrifugation at 2500 x g for 20 min at 4°C. Pellets were gently resuspended in 1.5 ml cold PBS, transferred to screw-cap microfuge tubes, and centrifuged at 1500 x g for 20 min at 4°C. Supernatants were removed and the cell pellets stored at −80°C until further use.

Two naturally occurring Norwegian lamb A. phagocytophilum strains differing in the 16S rRNA gene and degree of virulence were used to experimentally infect lambs raised in an indoor environment with barriers against tick entry and tick infestation. Lamb 00186 was infected with the more virulent variant 1 (identical to GenBank M73220) and lamb 0054 with variant 2 (identical to GenBank AF336220) 48 , to be referred to as ApNorLamb-V1 and -V2 from here on. Infections were monitored by microscopy and blood was harvested at maximum parasitemia. To purify buffy coats containing the infected neutrophils, approximately 2.5 l of Na-citrated blood was collected from each animal. The blood was transferred to 1 l centrifuge bottles and centrifuged at 2,500-3,000 x g in a swing-out bucket rotor for 30 min at 4°C. After removing most of the plasma layer, the buffy coat layer was collected with minimal contamination of red blood cells. The cells were diluted 1:3 with PBS, mixed gently and centrifuged at 1,500x g for 20 min at 4°C. Following three PBS washes, supernatants were removed and the cell pellets stored at −80°C. The experimental study in sheep was approved by the Norwegian Animal Research Authority.

For the HZ, JM, Dog1, MRK and NorLamb-V1 and -V2 strains, intact, host cell-free organisms with minimal host cell gDNA/RNA contamination were purified from frozen PBS pellets of infected cells prepared as above. Samples and reagents were maintained on ice throughout the entire procedure, and all centrifugations performed at 4°C. Following a quick thaw, host cells were disrupted by vigorous vortexing for 5 min. An equal volume of PBS was added and vortexing continued for 3 min. Cellular debris was removed by centrifugation at 200 x g for 15 min. After removing most of the supernatants to fresh tubes, these were passed several times through a 31 G needle and saved on ice. Pellets were resuspended well in final 500 μl PBS then passed serially through 22 G, 25 G, 28 G and, when possible, 31 G needles attached to a 1 ml syringe. 3–5 volumes PBS were added and mixed by vortexing. Debris was removed by centrifugation at 200 x g for 10 min. Supernatants were pooled to those from the previous centrifugation step. RNaseA was added to a final 250–300 μg/ml and the samples incubated 45–60 min at 37°C. Samples were centrifuged at 21,000 x g for 30 min and the supernatants removed completely. Pellets were resuspended in 50–100 μl PBS each and transferred to fresh tubes. To ensure homogeneity of the suspension, initially a drawn-out 10 μl pipette tip was used to disrupt the pellet by swirling followed by up/down pipetting and gentle vortexing, before switching to a larger tip. The sample was further homogenized by several passes through a 28-31 G needle. PBS was added to final 500–700 μl and DNaseI to final 250 μg/ml. Following 45–60 min incubation at 37°C the samples were centrifuged at 21,000 x g for 30 min. Pellets were homogenized as above and the DNaseI treatment repeated. EDTA (pH 8.0) was added to final 25 mM and the samples centrifuged as above. Tubes were washed twice with PBS without disturbing the pellets and residual PBS was removed after 3 min centrifugation at 21,000 x g. Pellets were homogenized as above in 600–800 μl RPMI culture medium (containing 10% fetal bovine serum) added incrementally and transferred to a 50 ml tube. Culture medium was added to a final volume of 6 ml before passage through a pre-wet, 2 μm pore-size, 25 mm, GMF-150 glass microfiber syringe filter (Puradisc 25GD; Whatman Inc., Florham Park, NJ). The filter was washed 3-4x with culture medium. Washes were pooled to the filtrate and centrifuged at 22,000 x g for 30 min. The pellets, comprised of free, non-viable organisms and host cell mitochondria, were resuspended in PBS, transferred to microfuge tubes and re-pelleted at 21,000 x g for 30 min. Supernatants were removed completely and the pellets were processed immediately or stored at −20°C. For every 10⁸ host cells used at 90-100% infectivity, enough organisms were recovered to yield on average 1–1.5 μg high-quality DNA using either the Gentra Puregene Yeast/Bact. kit (Qiagen Inc., Valencia, CA) or the QIAGEN Blood & Cell Culture DNA mini kit following the manufacturer’s protocols.

For the Dog2 and Ap variant 1 strains, organisms were cultured and isolated from ISE6 tick cells as described 49 . Host cell-free bacteria were prepared from two cultures in 25 cm² flasks, collected by centrifugation for 10 min at 11,000 xg at 4°C, and lysed in Gentra Puregene lysis buffer (Qiagen) at 80°C for 5 min. Since these DNA samples also contained a considerable amount of small (<500 bp) DNA species naturally associated with the ISE6 host cell line, the A. phagocytophilum gDNA was further purified by electroelution from agarose gels, followed by phenol/chloroform extraction and EtOH precipitation using conventional protocols.

ApDog1 was initially selected for complete genome sequencing to compare with the published HZ strain. When a draft genome was assembled for ApDog1 it was largely syntenic with HZ except for the virB6 locus, indicating a possible error in the sequence of one or both of the strains. Accordingly, the ApDog1 draft genome sequence was verified by Optical Mapping. In preparation for Optical Mapping (performed by OpGen Inc., Gaithersburg, MD), host cell-free organisms were embedded in 0.5% low-melting point agarose plugs and subsequently lysed, allowing access to the intact, ~1.48 Mb circular A. phagocytophilum chromosome. A procedure recommended by OpGen was followed. All solutions were made fresh using OpGen suggested reagents. Intact ApDog1 organisms were purified as above, except that the pellet of free organisms obtained following centrifugation of the filtrate was resupended and washed in cell suspension buffer [200 mM NaCl, 100 mM EDTA-Na₂ (pH 8.0), 10 mM Tris (pH 7.2)]. Plugs were made immediately on completion of the isolation procedure. Briefly, following the final centrifugation of the purified organisms, the pellet was resuspended in cell suspension buffer using 40–50 μl for every 10⁸ host cells used at >95% infectivity. The sample was passed 2x through a 31 G needle (3/10 ml capacity Insulin Syringe with fused 8 mm long needle, BD #328438; Becton, Dickinson & Co., Franklin Lakes, NJ) to ensure homogeneity of the thick suspension, and an equal volume of 1% low melting point SeaPlaque GTG agarose [(Lonza #50111) dissolved in DEPC-treated water (Invitrogen #750023; Carlsbad, CA) and held at 55°C] was immediately added. Following mixing, 100 μl aliquots were dispensed into plug molds (Bio-Rad #170-3713; Hercules, CA) and allowed to set for 1 hr at 4°C prior to transfer into a 50 ml tube containing 5–10 ml, 50°C NDSK solution [filter sterile NDS solution (1% N-lauroylsarcosine (Sigma #L5000; St. Louis, MO) in 0.5 M EDTA-Na₂ (pH 9.5), supplemented with final 2 mg/ml proteinase K (Pierce #17916; Rockford, IL) immediately prior to use]. The tube was incubated upright at 50°C with mild shaking (40 rpm) for 8–24 hrs, until the plugs turned clear and colorless. Plugs were gently washed 3x in 5 ml 0.5 M EDTA-Na₂ (pH 9.5), then transferred to a fresh tube and stored in EDTA at 4°C. Optical Mapping data generated from the BamHI-digested ApDog1 chromosome was analyzed using the OpGen MapSolver software.

Isolated DNA was provided to the Interdisciplinary Center for Biotechnology Research (ICBR) core facilities, University of Florida for library construction and pyrosequencing on the Roche/454 Genome Sequencer according to standard manufacturer protocols. Regular read libraries were generated for all strains. Additionally, 3 kb paired end libraries were made for ApHZ, ApDog1 and ApMRK. Genome coverage range was 31.3x to 72.1x. For each strain, the SFF format flow files, returned by ICBR for bioinformatics analysis, were first combined and converted to .fasta and .qual files (or the two combined in .fastq format) using Roche/454 Genome Sequencer FLX System software. Genome drafts were assembled using the CLC Genomics Workbench software suite (version 4.0-4.9) by mapping reads initially against the fully annotated, Sanger sequenced ApHZ genome (GenBank CP000235), then against the completed ApDog1 genome. Default parameters were used: length fraction, 0.5; similarity, 0.8; and for paired end reads, minimum distance, 1500/maximum distance, 4500. To obtain the vir loci, the resulting consensus sequence and underlying aligned reads were inspected for conflicts and mismatched paired ends suggesting the presence of insertions and/or deletions not mirrored in the consensus. These were manually corrected. Gaps were also manually closed where possible. Briefly, overlapping reads covering at least 2 kb of sequence on both sides of a gap and extending into it were individually extracted from the alignment. A new consensus for each side was obtained by assembling the reads against each other, and 250 N’s were added to its ends. These were individually used as the reference sequence against which all the 454 reads were re-mapped to pull out novel reads extending into the unknown region. The process was repeated multiple times, allowing for the incremental filling of the gap. PCR, followed by sequencing was performed when sequences extrapolated in this fashion spanned complex tandem repeat regions such as repeat regions 1 and 3 (R1 and R3 in Figure 3A) of the virB6-4 gene, or when gap closure could not be completed due to such structures, as was the case with the extremely long virB6-4 R4 (Figure 3A) region.

Amino acid sequences were aligned with MAFFT 50 and displayed with CHROMA 51 . Taxonomic relationships used a neighbor-joining tree and the ITT substitution model 52 and were displayed using Archaeopteryx (http://www.phylosoft.org/archaeopteryx). Hydrophobicity analyses were conducted using the method of Hopp and Woods 53 54 at web.expasy.org and transmembrane segments were predicted with TMpred at http://www.ch.embnet.org/software/TMPRED_form.html.

Due to difficulties in amplifying tandem repeat-containing DNA, all PCR reactions spanning the virB6-4 gene repeat regions were performed in the presence of 1.5-1.7 M Betaine (Sigma). The 8.36 kb PCR product spanning R3 and R4 in the ApHZ strain (Figure 3A, 3C, and Additional file 2: Figure S2A) was amplified using the iProof High-Fidelity DNA Polymerase system with GC buffer (Bio-Rad). Reactions totaled 50 μl with 5 ng purified A. phagocytophilum gDNA, 1.0 U polymerase, 1.5 mM MgCl₂, 200 μM each dNTP, and 250 nM each primer (AB1393: 5^′-CGGGATCTAAGACAGATGATGATTC-3^′, forward; AB1466: 5^′-CTCATCCTGATGCGTCTCCTTAG-3^′, reverse; Figure 3A). 35 cycles of 30 sec denaturing at 98°C, 20 sec annealing at 67°C, and 5 min extension at 72°C were performed. PCR products spanning R4 in ApJM and ApDog1 (both ~10.3 kb; Figure 3C) were derived using Takara’s PrimeSTAR GXL DNA Polymerase system (Clontech Laboratories, Mountain View, CA). Reactions contained 5 ng gDNA, 1.25 U polymerase, 1.0 mM MgCl₂, 200 μM each dNTP, and 200 nM each primer (AB1395: 5^′-CACCAGAGGATGCAGCATTAG-3^′, forward; AB1466, reverse; Figure 3A) in total 50 μl. Following the manufacturer’s recommendations, 2-step PCR was performed with 30 cycles of 10 sec denaturing at 98°C and 10 min annealing/extension at 68°C. PCR products were analyzed on 0.5% agarose gels alongside the 1 kb Plus (Invitrogen) and the GeneRulerHighRange (Fermentas, Inc., Glen Burnie, MD) DNA ladders. In order to TA-clone the amplicons, A-overhangs were added to the ends using 0.5-1.0 units AmpliTaq DNA polymerase (Applied Biosystems, Foster City, CA) in a 10–15 min reaction at 72°C. Products purified from agarose gels (before or after A-overhang addition) were cloned into the pCR-XL-TOPO vector (Invitrogen) and transformed into E. coli Stbl2 (Invitrogen), which is more permissive to repeat-laden foreign DNA. Recombinants containing the correct size insert were end sequenced to verify their identity.

In preparation for sequencing with the long-read length Pacific Biosciences (PacBio) next-generation sequencing RS instrument, constructs were linearized with restriction enzymes which cut the vector only, but on opposite sides of the insert within the Multiple Cloning Site. For ApHZ, equimolar amounts of the TA clone were cut with either HindIII or EcoRV. Following pooling and EtOH precipitation, the linearized DNA mix was submitted to ICBR/UF for SMRTbell library construction and sequencing. Libraries were constructed using a commercial strobe library preparation kit (#001-326-530; Pacific Biosciences, Menlo Park, CA) following standard manufacturer protocols. To further increase the likelihood of full coverage, the strobe-sequencing run was performed using two different conditions: I) 45 min light period (continuous collection time); and II) (5 min light period, 10 min dark period), followed by (45 min light period, 10 min dark period). The ApJM and ApDog1 constructs were double-digested with HindIII/XbaI to excise the ~10.3 kb inserts. Following separation on 0.5% agarose gels, the inserts were recovered from agarose slices by electroelution and further purified and concentrated by passage over QIAquick spin columns following the PCR Purification kit protocol (Qiagen). SMRTbell libraries were made as above then sequenced using a single 75 min movie time run.

Due to the repetitive nature of the cloned gene fragments, combined with the relatively high error-rate of the PacBio system, all attempts to assemble the reads de novo failed to yield a sequence of the expected size. Therefore, for each construct, reads >3 kb were selected from the multi-fasta files using the Galaxy suite 55 , and imported into the CLC Genomics Workbench for assembly and further analysis. These were assembled at low stringency initially against a consensus sequence representing an entire (vector and insert sequence) linear construct to which sufficient N’s were added based on the estimated gap-size. Starting with reads initiating outside the repeat region, the longest of the assembled reads were visually inspected for the presence of virB6 4 R4 repeat signature-sequences (Additional file 2: Figure S2) and their sequence manually corrected where necessary. The extended sequences were used to replace N’s in the consensus and the process repeated several times until sufficient reads with >2 kb sequence overlap were recovered spanning the entire insert region. For verification, the completed sequence for each strain was used as the reference to re-map all the respective >3 kb PacBio reads and the Roche/454 reads at higher stringency.

The sequences of vir loci are complete for strains ApDog1 and ApJM. The sequence of the repetitive virB6 4 locus was incomplete (ApDog2) or not determined for the other strains except ApHz. We provide a revised sequence of virB6 4 for the previously sequenced 15 ApHZ strain.

ApDog1:JX415845 - JX415868

B2-1 B2-2, B2-3, B2-4, B2-5, B2-6, B2-7, B2-8, B2-9, B3, B4-1, B4-t1, B4-2, B6-1, B6-2, B6-3, B6-4, B8-1, B8-2, B9-1, B9-2, B10, B11, D4

ApJM:JX415869 - JX415892

B2-1, B2-2, B2-3, B2-4, B2-5, B2-6, B2-7, B2-8, B2-9, B3, B4-1, B4-t1, B4-2, B6-1, B6-2, B6-3, B6-4, B8-1, B8-2, B9-1, B9-2, B10, B11, D4

ApDog2:JX415893 - JX415915 (virB6-4 submitted separately as gapped)

B2-1, B2-2, B2-3, B2-4, B2-5, B2-6, B2-7, B2-8, B2-9, B3, B4-1, B4-t1, B4-2, B6-1, B6-2, B6-3, B8-1, B8-2, B9-1, B9-2, B10, B11, D4

ApNorLambV2:JX415916 - JX415938

B2-1, B2-2, B2-3, B2-4, B2-5, B2-6, B2-7, B2-8, B2-9, B3, B4-1, B4-t1, B4-2, B6-1, B6-2, B6-3, B8-1, B8-2, B9-1, B9-2, B10, B11, D4

ApNorLambV1:JX415939 - JX415966

B2-1, B2-2, B2-3, B2-4, B2-5, B2-6, B2-7, B2-8, B2-novel1, B2- novel2, B2-novel3, B2-novel4, B2-novel5, B2-novel6, B3, B4-1, B4- t1, B4-2, B6-1, B6-2, B6-3, B8-1, B8-2, B9-1, B9-2, B10, B11, D4

ApHZvirB6-4:JX415967

ApVar1:JX415968 - JX415996

B2-1, B2-2, B2-3, B2-4, B2-5, B2-6, B2-7, B2-8, B2-novel1, B2- novel2, B2-novel3, B2-novel4, B2-novel5, B2-novel6, B2-novel7, B3, B4-1, B4-t1, B4-2, B6-1, B6-2, B6-3, B8-1, B8-2, B9-1, B9-2, B10, B11, D4

ApMRK:JX415997 - JX416019

B2-1, B2-2, B2-3, B2-4, B2-5, B2-6, B2-7, B2-8, B2-9, B3, B4-1, B4-t1, B4-2, B6-1, B6-2, B6-3, B8-1, B8-2, B9-1, B9-2, B10, B11, D4

ApDog2virB6-4Gapped:JX416020.