Pyrosequencing indicated that the MAP strain S397 has a circular chromosome with at least 4,814,922 bp, a G + C content of 69.31% and contains 4,700 predicted open reading frames (ORFs). The majority of these genes (44.5%) were predicted 23 to encode cytoplasmic proteins (Additional file 1: Table S1) involved in various cellular functions and a minority of extracellular proteins (< 1%). The number of annotated genes in S397 was more than the bovine K-10 strain (Table 1) due to the different annotation methods used on each genome 19. However, like MAP K-10, the S397 genome contains one rRNA operon and 46 tRNA genes representing all 20 amino acids. A detailed comparison between MAP strains K-10 and S397 as well as the human, MAH 104 is shown in Table 1. The de novo assembly of the compiled S397 genome had an average sequencing depth of 24 × in 184 scaffolds (Additional file 2: Table S2). When aligned to the K-10 sequence, over 110 of these scaffolds are separated by a sequence gap of less than 500 bp suggesting the small size of most gaps. Furthermore, when gaps of 3.5 kb or less were ignored, we were able to assemble the whole genome into 3 scaffolds. The two largest sequence gaps are between contig00150c and contig00149c, which is estimated at 30.19 kb and the contig00082-contig00041c gap, which is estimated at 18.87 kb. Additional file 3: Table S3 gives an overview of the ordered scaffolds.

Analysis of the two additional genomes sequenced in this study (JTC1074 and JTC7565) revealed more than 99% identity to the S397 genome sequence (Table 2). A de novo assembly of these genomes sequenced using Illumina platform produced an average sequence depth of 60 ×. As expected, no significant differences were found between the common features of the 3 sequenced sheep isolate genomes. In fact, there were no gene differences; hence all three genomes were identically annotated. Similar to other sequenced mycobacterial genomes, dnaA was assigned the first locus tag (MAPs_00010).

The IS elements usually play a role in the genomic diversity among strains of mycobacteria 24 and could act as a good target for molecular diagnostics 25. Similar to K-10, the S397 genome has all the well-studied insertion sequences (e.g. IS900, IS1311 and IS_map02). IS900 is generally considered a MAP specific element that was originally discovered in 1989 2627. A total of 17 copies of IS900 were found in the S397 genome, which is identical to the K-10 strain. Another element, IS_map02, is a MAP specific insertion sequence that was discovered by sequencing the K-10 genome. A total of 6 copies of IS_map02 are present in both S397 and K-10. Likewise, IS1311 is present 7 times in each genome. No IS elements were found to be unique to one or the other genome.

Sequence analysis alone was not sufficient to decipher the synteny of the genome. Previously, we used an optical mapping protocol to confirm the organization of the MAP K-10 genome 20. A similar strategy was used to analyze the genome of S397. The raw optical map dataset comprised 2,950 single molecule maps with a total mass of 784.5 Mb, and an average molecule size of 333.6 Kb (Figure 1). After assembly, the compiled optical map contained 905 single molecule optical maps (301.9 Mb; total mass), which covers the genome 58 ×. After a G + C content adjustment by a factor of 0.95, the estimated size of MAP S397 optical map is 4.95 Mb, which is slightly higher than the sequence data suggested. However, if the estimated sequence gaps are added in, the estimated sizes are very similar.

To our surprise, there were 7 inversions that are larger than 22 kb when the S397 genome was compared to the sequenced genome of K-10 compiled by Wynne 2028. The total size of these inversions spanned 2.4 Mb of the S397 genome. Individual sizes of those inversions range from 22 to 1,174 kb. As shown in Figure 2B, homologous segments between MAP K-10 and S397 are represented by color boxes and to each segment a number was assigned. Detail information of each segment is shown in Table 3. Thirteen out of the 14 segments have at least one IS element on the flanking regions (Figure 2).

Similar to our analysis of inversions discovered in the K-10 strain, we used a PCR-based approach to examine two of the inversion breakpoints in the S397 genome (Figure 3), which are the right end of segment ID #1 and the left end of segment ID#2 (Table 3). As expected, our PCR analysis confirmed the inversion predicted in the genome of K-10 and S397 strains. Because these inversions were readily identified from the optical map and sequence alignment data, we did not attempt to confirm all of the inverted fragments by PCR. Despite these inversions, there is strong synteny between these genomes, underscoring their close relatedness. Both genomes share a number of large-scale clusters of homology where gene order is highly conserved (Additional file 4: Table S4).

Further comparative sequence analysis identified several regions that are present in MAP S397 and MAH 104, but not in MAP K-10 (Additional file 5: Table S5). The largest of these is a 9-kb gene cluster encompassing 13 ORFs (MAPs_15940-MAPs_16060). This region was partially identified by representation difference analysis and termed PIG-RDA20 for pigmented strain representational difference analysis-20, as detailed before 7. It was also mapped to the MAH 104 genome by Dohmann and coworkers 7 and was subsequently described by Semret and coworkers as large sequence polymorphism (LSP), LSP^A4-II 29. This region contains a copy of the IS1311 insertion sequence and within the MAH 104 genome is flanked by an additional copy of IS1311. Another previously described LSP included 9 ORFs (MAPs_46190-MAPs_46270) and totals 6.6 kb. This region was partially identified as the PIG-RDA10 sequence and was mapped to a 16 kb segment of the MAH 104 genome 7. The full sequence was later identified as LSP^A18 29, which is equivalent to MAV island 24 3. An interesting feature of LSP^A18 is that it begins and ends with a transcriptional regulator. Eight other LSPs containing 4 or more ORFs not present in K-10 were also observed (Table 4). Overall, a total of 70 ORFs were present in MAP S397 but absent in the MAP K-10 genome (Additional file 5: Table S5).

Several new or only partially described LSPs common to MAP S397 and MAH 104 strains were also identified. A good example here is the novel LSP found in MAP sheep and MAH 104 genomes is comprised of 14 ORFs (MAPs_17580 - MAPs_17710), predicted to encode proteins involved in the biosynthesis of glycopeptidolipids 30. This region in MAP S397 revealed the presence of four additional ORFs (hyp, hlpA, dhgA and mtfC) with homology to glycopeptidolipid biosynthesis genes immediately downstream. The additional 4 ORFs were also not present in the MAH 104 sequence. Finally, a putative transcriptional regular labeled as MAPs_44910 is present in MAP S397. The protein encoded by this ORF has homology to the GntR-family of transcriptional regulators, which are widely distributed across bacterial species and regulate a variety of cellular processes 3132.

A second subset of sequence polymorphism was represented by 32 ORFs that were present in the MAP K-10 genome but absent from the genome of MAP S397 (Additional file 6: Figure S1). Several of these deletions have already been described earlier. The deletion encompassing MAP1485c-MAP1491 was previously identified by Marsh and coworkers as S strain deletion #1 in an Australian MAP sheep isolate 33 and by Semret and coworkers as LSP^A20 29. An additional larger deletion in the MAP S397 included the cluster of ORFs between MAP1728c and MAP1744. This deletion was partially identified by Marsh and coworkers as RDA3 34, and later fully described as S deletion #2 33.

A novel deletion comprising the ORFs MAP1432-MAP1438c (partial) was identified in the current study as absent from MAP S397. This deletion, termed sΔ-1, was originally discovered by comparative genomic analysis and subsequently confirmed by PCR analysis. This gene cluster is predicted to encode four energy metabolism enzymes as well as a lipase (MAP1438c). MAP1432 encodes a hypothetical protein with homology to the REP13E12, a family of repetitive elements that were originally described in M. tuberculosis and have been shown to be targets of phage integration 35. There is a homolog to MAP1434 that is present in S397 (MAPs_13210). The region around MAPs_13210 is not near the end of a contig and is nearly identical to an inverted stretch in K-10, thus leading to the conclusion that MAPs_13210 is only a homolog of MAP1434, but that the gene itself is not present in the S397 genome.

Interestingly, MAP2656 was initially identified as absent via microarray analysis 5 but sequencing of MAP S397 identified a homologue with 100% identity (MAPs_10401 & MAPs_10402). Likewise, MAP2325 was identified as being absent from Australian sheep isolates of MAP 33. This ORF was not identified as missing from MAP S397 as sequencing confirmed the presence of an ORF (MAPs_34380) with 100% identity to MAP2325. These discrepancies may represent a geographic difference between MAP isolates recovered from sheep in Australia and the United States or it may be an error from the microarray experiment. These were the only observed differences between the microarray and sequence data. Overall, genomic alignments indicated the presence of a significant number of insertions and deletions between ovine and bovine strains of MAP that are suggested to be associated with their respective host.

Genomic insertions and deletions have been previously used to determine evolutionary relationships among MAC strains 36. With the genome sequence of these ovine isolates of MAP, we can now add comprehensive SNP and inversion data to strengthen evolutionary hypotheses. Earlier genotyping of the MAP S397 utilizing SNP of recF, gyrA and gyrB genes indicated that this strain belong to the MAP type III, a sublineage of the MAP-S cluster of isolates 37. To examine the evolutionary history of MAP, we analyzed the genome sequence of S397 compared to other clinical isolates circulating in sheep as well as the standard cattle strain, K-10. Our first level of analysis included the alignment of the S397 genome to that of the JTC1074 and JTC7565. This alignment resulted in identical genome organization of all three ovine isolates, as expected. Additionally, we examined the relationship among S397 (ovine origin) with both K-10 (bovine origin) and MAH 104 (human origin). Such analysis identified several events of inversions and potential insertions/deletions between genomes belonging to the ovine isolates and other isolates of bovine and human origins (Figure 4). The optical map of S397 confirmed these inversions as well. Moreover, when the draft genome sequence of M. intracellulare was added to the comparison, the whole contig00148 (accession number GenBank: ABIN01000141) aligns to the region spanning the right breakpoint (Figure 4) of MAH 104 and MAP ovine strains, an indication of a conserved genome synteny among M. intracellulare, MAH and MAP sheep strains, but distinct from MAP bovine strains.

In the second level of analysis on the nucleotide level, a core of 42 single nucleotide polymorphisms (SNPs) were present in both JTC isolates compared to S397. In addition, a very small number of unique SNPs in JTC1074 (N = 22) and JTC7565 (N = 11) were not present in any other genome in this study. Collectively, this small level of polymorphism indicates the clonal nature of ovine isolates, which contrasts sharply with the 4,438 SNPs between the ovine S397 and the bovine K-10 strains (Figure 5A). Additionally, when analyzing genome-wide SNPs, it appears that MAP S397 and K-10 split off recently from the hominissuis progenitor strain (Figure 5B). A similar result is obtained when SNPs are restricted to coding sequences (Figure 5C).

Isolates were cultured in Middlebrook 7H9 broth (BD Biosciences, San Jose, CA) media supplemented with 10% OADC (2% glucose, 5% bovine serum albumin factor V, and 0.85% NaCl), 0.05% Tween 80 and 2 μg/ml of Mycobactin J at 37°C 45. The MAP ovine S397 strain was obtained from a Suffolk breed in Iowa. It was isolated from the distal ileum at necropsy in 2004. The other 2 sheep isolates of MAP (JTC1074 and JTC7565) were isolated from the intestine of infected sheep in Texas and obtained from the Johne's Testing Center at the University of Wisconsin-Madison. All isolates were genotyped using the IS1311 restriction endonuclease, which yielded the 2-band pattern typical of ovine strains 6.

Genomic DNA was extracted as described in detail previously 346. For the S397 strain, the DNA (1-5 μg) was sequenced using Roche 454 pyrosequencing (GS20 and FLX) at the National Animal Disease Center. A whole-genomic shotgun sequencing library was prepared according to Roche protocols. The library was used with the appropriate emulsion based PCR kits to produce sufficient beads for sequencing using the Roche Standard Chemistry GS-LR 70 sequencing kit. For the JTC1074 and JTC7565, the purified genomic DNA (~5 μg) of each strain was sent to Genomic Resource Center at the University of Maryland for Illumina whole genome sequencing (Multiplexing Sample Preparation oligonucleotide Kit) as outline before 47. The adapters and indexing oligonucleotides were purchased from Illumina (5 Paired End Cluster Generation Kits-v4). The CLC Genomic Workbench software (version 4.0.3) was used to perform reference and de novo assembly on all sequenced genomes.

The S397 sequence was annotated using the Integrated Microbial Genomes Expert Review (IMG-ER) pipeline 48. The sequences of the JTC isolates were annotated based on S397. Genes were each designated by the locus tag "MAPs" to distinguish it as a MAP sheep strain gene. This locus tag is followed by a five digit unique identifier, which incrementally increases by ten (i.e. MAPs_45660... MAPs_45670... MAPs_45680...). With this numbering configuration, additional genes can easily be added as they are discovered or when remaining gaps are closed.

The genome data for MAP K-10 (accession no. GenBank: NC_002944.2) and M. avium subsp. hominissuis (MAH) strain 104 (GenBank: NC_008595.1) were used in alignments in the Artemis and Artemis Comparison Tool (ACT) programs or Mauve 2.3.1 49. BLASTP analysis was used for similarity searches and protein sequence analysis. In addition, Mauve algorithm was used to align two or more genomes 50. For detecting single nucleotide polymorphisms (SNPs) among sheep isolates, the CLC Genomic Workbench was used. The coverage range setting for each strain was at 10-55 reads, and the frequency of the mutation was at least in 50% of the reads.

Shotgun optical mapping, as previously described 205152535455, was used to construct a physical restriction map for the S397 genome. Genomic DNA, in agarose inserts 56, was electroeluted into a solution containing a lambda DNA sizing standard (30 pg/μl), and then were mounted on cleaned, derivitized glass surfaces using a microfluidic device 57 followed by polymerization of a thin layer of polyacrylamide (3.3% containing 0.02% Triton X-100). Mounted DNA was digested with 20 units of BsiWI (NEB, Ipswich, MA) for 1 to 2 hrs at 37°C. Fluorochrome-stained DNA fragments were imaged by fluorescence microscopy with a 63 × objective lens (Carl Zeiss, Thornwood, NY) and a high-resolution digital camera (Princeton Instruments, Trenton, NJ). Images were acquired and processed using "ChannelCollect" and "Pathfinder" -custom software 57 that converts captured images into map data sets. Bayesian inference and an efficient dynamic programming algorithm were also being used to fine-tune the parameters including standard deviation, digestion rate, false cut, and false match probability etc. 545859. The final circular optical map contig was built using an iterative assembly process 60 including rounds of pair-wise alignment (single molecule maps vs. seed maps; provisional assemblies) and assembly 5254. Due to the high G + C content of MAP, which skews fragment sizing by integrated fluorescence intensity measurement, the final maps were globally scaled (0.95) to correct this problem 2061. A laboratory software implementation of an optical map alignment algorithm 62 was used to align between optical fragments generated from MAP S397 and the in silico restriction maps of MAP K-10, which provided a whole-genome rearrangement comparison between the two genomes. This restriction framework was used to generate a temporary rearranged genome as the reference sequence to guide the assembly of MAP S397 de novo contigs with the function "move contigs" in Mauve 2.3.1 49.

PCR reactions were performed in 25-μl reaction mixture containing 1 M betaine (Sigma-Aldrich, St. Louis, MO), 50 mM potassium glutamate (Sigma-Aldrich), 10 mM Tris-HCl pH 8.8, 0.1% Triton X-100, 2 mM magnesium chloride, 0.2 mM dNTPs, 0.5 μM each primer, 0.5 U of GoTaq^® Flexi DNA Polymerase (Promega, Madison, WI) and 25 ng of genomic DNA. The amplification thermocycle started with an initial step of 94°C for 5 minutes followed by 5 cycles of 94°C for 30 s, 62°C for 30 s with 1°C decrease for each cycle and 72°C for 3.5 min, and followed by 30 cycles of 94°C for 30 s, 57°C for 30 s and 72°C for 3.5 min. PCR primers used for examining the breakpoints included control F: AAGCATCACCTGCATGAGC, control R: CGGGAATTTATCCGTTTCAG, F1: GGGATCGATCTTGACCACAT, R1: GTGCCTGGACTCGATTTTGT, F2: AAGAGGTCGGAGGTTCGAGT and R2: CGGTGAGAGATTTCGTCACA. Primers used to demonstrate the S397 sΔ-1 deletion included F18: CGTCTTCCCCGTCGTCGTTC, B24: CGATGAGAGTCCGTGCGTGG, F15: CGGCGGGCGGTCAGGGTTTG, B17: GCAGGTTGGGGTTCGGCTTG, F7: GGTGGTCGGCGTCCTCGTAG, B9: CGTCGTCACAGCGAAAACGG, F3: CCACCCGCCTCACACCACTC, B4: AGGACGCCGACCACCAAACG. Conditions for the amplifications are essentially as described immediately above except that Advantage GC Genomic LA PCR Polymerase kit (Clontech) was used for each reaction.

This Whole Genome Shotgun project has been deposited at DDBJ/EMBL/GenBank under the accession AFIF00000000.