The MHC regions of two individuals were characterized. Cedric, whose parents came from DFTD-free northwestern Tasmania (father from Arthur River and mother from Woolnorth; Figure 1), produced an antibody response when injected with irradiated DFTD cells (Alex Kreiss, University of Tasmania, personal communication) and at the beginning of the project was thought to be a "resistant" animal because he did not develop DFTD following a disease challenge. Since then Cedric has succumbed to DFTD, but remains one of very few animals to be able to mount an antibody response to DFTD. The other individual, named Spirit, was originally from Bangor (Figure 1) and was euthanized in 2008 due to multiple DFTD lesions and metastases to the lungs.

Four genomic regions of approximately 960 kb in length were assembled and annotated based on ten fully sequenced BAC clones (seven from Cedric and three from Spirit) (Figure 2). 34 genes and pseudogenes were predicted, comprising five Class I genes, four Class II genes, four antigen-processing genes, eight Class III genes and thirteen other genes and pseudogenes. GenBank accession numbers of the BAC clones and coordinates of the predicted genes are shown in Table 1. The annotated genes were mostly named after their orthologues in the human with the exceptions of MHC Class I and II genes, which were given species-specific names based on the nomenclature proposed by Klein and colleagues 20 and their evolutionary relationships with previously characterized marsupial genes 21 22 .

Two of the analysed regions contain Class I genes (Region 1 and 2) and the other two contain Class II (Region 3 and 4) (Figure 2). All four regions were physically mapped to the same area on devil Chromosome 4q, indicating the genomic location of the devil MHC (see Figure 3 for Region 1 and 3, and Additional file 1: Figure S1 for Region 2 and 4). Based on comparisons to the MHC of the grey short-tailed opossum (Monodelphis domestica; herein referred to as "opossum") and the tammar wallaby (Macropus eugenii), Region 3 is likely located between Region 1 and Region 4, which encompasses a suite of well-conserved Class III genes (Figure 4). In Region 2, Class I Saha-UD is flanked by five non-MHC genes USP49, TOMM6, PRICKLE4, FRS3 and PGC, which are found ~12 MB proximal from the MHC in human and ~16 Mb distal from the MHC in opossum; however, according to FISH mapping result, these genes are closer to the MHC in the devil than in opossum (Additional file 1: Figure S1).

A total of five Class I genes were identified and named Saha-UA, UB, UC, UD and UK. Except Saha-UD, they are all closely linked (Region 1 in Figure 2) and are localised with PSMB8, PSMB9, TAP1 and TAP2 genes, which are involved in antigen processing and transportation. Saha-UA, UB, UC and UD have evolved from recent gene duplications in the devil lineage (shown in Figure 5) and are not orthologous to Class I genes in the opossum or the tammar wallaby. Saha-UK is orthologous to Modo-UK in the opossum and Maeu-UK in the tammar wallaby (Figure 5), both of which have been suggested to be nonclassical genes 22 23 .

The genomic data allowed us to assign MHC alleles to loci using phylogenetics (Figure 5, Additional file 2: Figure S2). Table 2 describes the alleles sequenced on each of the two haplotypes, and by deduction the alleles found on the un-sequenced haplotypes of Cedric and Spirit. We were able to resolve that previously isolated devil Class I sequence variants SahaI*27 and 28, 35, 46, and 32 belong to locus Saha-UC, UA, UB and UD, respectively. Meanwhile, five novel alleles, named SahaI*86 - 90, were found and assigned to genes. Alleles from Saha-UD, which were previously called group 2 alleles 19 , are divergent from other genes and form a discrete phylogenetic clade with 94% bootstrap support. Saha-UA, UB and UC are closely related, with UA and UB alleles interspersed within the same clade. At nucleotide level, the sequence identity between these three genes is extremely high (> 97.7%), with Saha-UA and UB sharing up to 99% nucleotide identity in their introns (Table 3).

The sequence 200 bp upstream of the translation start sites was searched for gene regulatory elements, as promoters of MHC Class I genes in mammals are mostly contained within this range 24 . In Saha-UA, UB and UC, this region is highly similar, with only one single nucleotide variation in the TATA sites. Putative sites of an enhancer A, an interferon stimulated response element (ISRE), the S-X-Y motifs, CAAT and TATA boxes were identified (Figure 6). Saha-UD was not included in the alignment in Figure 6 due to low sequence identity. Compared to Saha-UA, UB and UC, regulatory elements including enhancers, ISRE, S and X motifs are not conserved at this locus, which may indicate different expression or function of this gene.

The MHC region containing the Class Ia genes in Cedric and Spirit (Region 1 in Figure 2) contains three indels. Spirit's haplotype has an intact Saha-UA gene, whereas Cedric's has a deletion (~1646 bp in size) that results in the loss of a large portion (from exon 2 to intron 5) of Saha-UA and renders it a pseudogene. Furthermore, Cedric's haplotype also lacks two other long segments that are present in Spirit's, one (~12230 bp) lying between the pseudogene and Saha-UB, and the other (~16970 bp) between Saha-UB and Saha-UC. Further investigation on this genomic region revealed 12 putative Long INterspersed Element (LINE) segments interspersing the MHC genes (Figure 2), nine of which closely flank (within 2 kb) the indels and Saha-UA, UB and UC. Two of the putative LINEs are located inside the indels and only found in Spirit's haplotype.

A PCR test was developed to detect the presence or absence of the deletion in Saha-UA in devils. Haplotypes without the deletion generate a single band of 1917 bp in size, which contains amplicons from Saha-UA, UB and UC genes. A second band (271 bp) is generated when a haplotype contains the deletion in Saha-UA, representing an amplicon from only the pseudogene, which has arisen due to the deletion (Figure 7).

The frequency of the Saha-UA deletion was assessed in samples from seven geographic areas across Tasmania (see pie charts in Figure 1). 72 individuals were investigated: 25 from DFTD-affected areas in eastern Tasmania (five from each of St Marys, Narawntapu, Bronte Park, Buckland and Fentonbury); 12 from West Pencil Pine, an area on the disease front; and 35 from the DFTD-free northwestern coast region. The deletion is most prevalent in the northwest (found in 54.8% of tested individuals), followed by West Pencil Pine (41.7%), Bronte Park (40%), Narawntapu (40%) and St Marys (20%). None of individuals from Fentonbury and Buckland contained the deletion.

Four Class II genes were identified, all belonging to the marsupial Class II DA gene family. Previously isolated DAB transcripts were aligned against the three β chain paralogues and 100% match was found between transcripts SahaDAB*01 [GenBank:EF591102], SahaDAB*03 [GenBank:EF591104] and SahaDAB*05 [GenBank:EF591105] and genes Saha-DAB1, DAB2 and DAB3, respectively. Sequence comparison of these three genes is shown in Table 4. Saha-DAB1 and DAB2 share high nucleotide identity in both exons (96.9%) and introns (95.7%). They are both similar to Saha-DAB3 in exon sequences, but differ from it significantly in the introns due to multiple deletions/insertions of nucleotide fragments. In the 5' untranslated region, putative sites of S-X-Y motifs are identified in both α and β chain genes, though TATA and CAAT elements are not found in these genes (Figure 8).

Based on the genomic characterization of the MHC genes, as well as expression and diversity studies, we are now in the position to assign previously identified MHC alleles to loci and make informed deductions about the likely role of these genes in immune response.

Four Class II loci are described in this study, Saha-DAA, DAB1, DAB2 and DAB3. Transcripts from the three β chain genes have previously been described 18 . These loci likely encode functional antigen-presenting molecules based on homology to other members of the DA gene family, which have been studied in a variety of marsupial species [reviewed in 25 ]. No other Class II gene families have been identified in the devil, indicating that either the devil only has one functional Class II gene family or the other genes are too divergent from known marsupial Class II genes to be detected by the probes.

Amongst the five characterized Class I loci, we propose that Saha-UA, UB and UC are functional Class Ia genes on the basis of three facts. First, their transcripts have been amplified from all tissue types examined so far, including blood, spleen, skin, liver, kidney and DFTD cells 9 18 . Second, genetic variation, though the level is not high, is present at these loci, particularly in the putative peptide-binding regions (Figure 9). Third, promoter elements in the 5' untranslated region are well conserved between these genes and the Class Ia genes in the opossum and the tammar wallaby 22 , sharing 91.8% and 83.6% nucleotide identity respectively. Whether the single nucleotide variation in TATA box affects the expression level of Saha-UA remains uncertain. The number of Class Ia loci in the devil is comparable to those in the opossum (one) and the tammar wallaby (three).

Saha-UK has been found transcribed in the blood and spleen and is likely a Class Ib gene based on orthology with the Class Ib Modo-UK in the opossum and Maeu-UK in the tammar wallaby. It has been suggested that the high conservation of this gene over extended evolutionary periods indicates that it may serve a critical, marsupial-specific function 22 .

Saha-UD, which has been found transcribed in the blood, spleen and DFTD cells, shows features of a Class Ib gene. This locus has significantly lower levels of polymorphism than Saha-UA, UB and UC, and its 5' regulatory elements are divergent. Within the α1 domain, Saha-UD alleles share extremely high sequence identity (> 97.7%) and only one of the 15 putative peptide-binding amino acid residues is polymorphic (Figure 10). Elucidation of the functional role of Saha-UD in the devil remains to be determined.

Genomic characterization allowed us to assign previously characterized "group 1" alleles to Class Ia loci Saha-UA, UB and UC, and "group 2" alleles to Saha-UD ( 19 , Figure 5). The previous description of individuals containing only "group 1" or "group 2" alleles 19 was likely due to experimental artefact caused by nucleotide mismatches (two in Saha-UA, UB and UC; four in Saha-UD) within the PCR primer hybridisation sites. Here, we propose that Class Ia loci Saha-UB and UC are likely present in all devils, whereas Saha-UA is lost in certain haplotypes. We suggest that in future studies new primers should be designed for Saha-UD and the three Class Ia genes separately to ensure high primer efficiency.

A genomic CNV is a duplication or deletion of a genomic segment larger than 1 kb in size 26 . In addition to single nucleotide polymorphisms, CNVs represent a major class of genetic variation and are widespread in the human genome 27 28 . Large-scale genome-wide disease association studies have revealed a number of CNVs in the HLA that are associated with Crohn's disease, rheumatoid arthritis and type 1 diabetes 26 29 .

When we first proposed the hypothesis of "Class I gene copy number variation" in the devil, we were referring to the concept of CNV in a narrow sense, namely variation in the number of Class I loci 19 . With the new findings from this study, it is necessary to expand the concept to take into account other CNVs of non-coding sequences lying within the devil MHC.

Three genomic copy number variants that intersperse with the MHC Class Ia genes were identified between Cedric's and Spirit's haplotypes. The one within Saha-UA is convincingly a deletion, while the other two are difficult to classify as either duplications or deletions. The loss of functional Class I genes due to CNVs is not specific to the devil. A case has been reported for a human haplotype where a 4 kb deletion turns HLA-A into a pseudogene 30 . The mechanisms underlying the generation of these CNVs in the devil MHC are not clear, though a possible explanation may be implied from the presence of the putative LINE fragments in the affected regions encompassing the three CNVs and Saha-UA, UB and UC genes (Figure 2). These retrotransposons may play a role in causing the CNVs as evidence has been found in primate genomes that LINEs are centrally involved in the generation of CNVs and can mediate deletions up to 18 kb in size 31 32 .

The higher frequency of the Saha-UA deletion in the northwest may indicate that the eastern and western populations have been exposed to different selective pressures. It is tempting to continue to speculate that the deletion of Saha-UA, which was found in Cedric and occurs in high frequency in the northwest of Tasmania, may provide animals with some level of resilience to DFTD. This is consistent with the observation that the spread of DFTD is slowing down as the disease front reaches genetically disparate populations in the northwest and that the incidence of disease in these populations is much lower than in similar populations in the east 33 . This gene deletion may be advantageous for the following reasons. First, the loss of Saha-UA may increase the MHC antigenic dissimilarity between host and tumour, as a Saha-UA allele, SahaI*35, has been found to be transcribed in the tumour 19 . Second, the CNVs may alter the expression level of the adjacent MHC genes by affecting regulatory elements such as promoters and enhancers as well as inhibitory elements. In fact, it has been found that CNVs can even influence the expression of genes that are up to 1 Mb away 34 . Here we have only looked at two devil MHC haplotypes and have only scratched the surface of the genetic variation that could be present amongst devil genomes. The role of this variation in the varied susceptibility/resistance of devils to DFTD needs to be further investigated.

Previous comparative studies revealed that the MHC of the opossum and the tammar wallaby exhibit distinct features in gene content and organisation 21 35 . Sequencing of Tasmanian devil MHC regions has provided us a better understanding of the marsupial MHC.

In the tammar wallaby, all Class Ia genes (UA, UB and UC) are un-linked to the MHC region 22 , whereas in the opossum the only confirmed Class Ia gene (UA1) is located within the MHC and close to antigen-processing genes ( 21 , Figure 4). The organisation of Class Ia genes in the devil is similar to that of the opossum, with UA, UB and UC adjacent to TAP1, TAP2, PSMB9 and PSMB8. This indicates that the dispersal of Class I genes across the genome in the tammar wallaby is not a common characteristic of the Australasian marsupial and has occurred after the divergence of devil and tammar wallaby's common ancestor at ~66 million years before present 36 .

Four Class II gene families have been characterized in the opossum - DA, DB, DC and DM 21 . DA, DB and DM have also been identified in the tammar wallaby. The Class II genes have undergone large-scale expansion in the tammar wallaby, resulting in up to 10 DAB loci 35 . However, such an expansion of Class II genes is not seen in the devil, with devils having a single DAA and three DAB genes. On the contrary, the devil genome may have undergone gene deletions that have lead to loss of the DB and DC gene families.

In this study, we have focused on characterizing MHC Class Ia and II genes in the devil. Future work will involve characterization of the functional roles of the putative Class Ib genes, as well as genes involved in the antigen processing machinery.

Two BAC libraries were produced. The first one, designated VMRC-49, was constructed from whole blood genomic DNA of Cedric (three-year-old male). The second library, designated VMRC-50, was constructed from genomic DNA extracted from the liver of Spirit (two-year-old male). The genomes of both animals were recently sequenced 37 . The genome assembly was not able to provide an accurate picture of the MHC, which contains multiple closely related genes that have arisen from recent gene duplications. Therefore, a BAC clone sequencing project was required.

The BAC libraries were produced in the Genome Resource Center at the Benaroya Research Institute at Virginia Mason, Seattle, USA. Detailed procedures of library construction have been described previously 38 39 . The quality of the DNA was checked by running a pulsed field gel electrophoresis (PFGE) on a CHEF-DR^® III system (BioRad). The DNA was partially digested in an EcoRI/EcoRI-methylase competition reaction and size fractionated by analytical PFGE on a BioRad CHEF Mapper^® XA system. DNA fragments from the appropriate size fraction were ligated into the CopyControl™ pCC1BAC™ vector from Epicentre Technologies and transformed into Invitrogen ElectroMAX™ DH10B™ T1 Phage-Resistant E. coli cells. Transformants were arrayed into 384-well LB/chloramphenicol/glycerin microtiter plates (Genetix) using colony-picking robots (Norgren Systems) and subsequently gridded onto 22 × 22 cm high-density nylon filters with a Total Array System (BioRobotics Ltd.). The average insert size was estimated to be ~140 kb. A total of 672 microtiter plates, which contain 258,048 BAC clones, and 14 high-density filters were produced for library VMRC-49. Assuming the size of the devil genome is similar to that of the opossum genome, which is around 3.6 Gb, this library will represent 10x coverage of the devil genome. For library VMRC-50, 432 plates containing 165,888 BAC clones and nine high-density filters were generated, estimated to represent 6.5x whole genome coverage.

BAC clones containing MHC Class I or II genes were physically mapped to a male devil karyotype following the protocol described previously by Alsop and colleagues 43 . Approximately 1 mg of BAC DNA was used to produce probes that were labelled by nick translation with either SpectrumOrange dUTP or SpectrumGreen dUTP (Abbott Molecular Inc.). Labelled probes were hybridised overnight to devil chromosomes, which were denatured for 1 min 40 sec. Slides were washed once at 60°C in 0.4x SSC with 0.3% Tween20 for 2 min and then once in 2x SSC with 0.1% Tween20 for 30 sec at room temperature. Chromosomes were counterstained in DAPI and mounted in VECTASHIELD^® Mounting Medium from Vector Laboratories Inc. A Zeiss Axioplan2 epifluorescence microscope was used to visualize the fluorescent signals. Images of DAPI stained metaphase chromosomes and fluorescent signals were captured using a SPOT RT Monochrome CCD charge-coupled device camera (Diagnostic Instruments Inc.) and merged using IP Lab imaging software (Scanalytics Inc.).

To test the Class I gene number variation hypothesis, it was necessary to ensure that all Class I loci in both Cedric and Spirit were characterized. To do this, we isolated and sequenced all Class I alleles from both individuals and assigned them to genes. We amplified exon 2 of the Class I genes from genomic DNA using protocols developed by Siddle and colleagues 9 with modifications in the reverse primer sequence (5'- CTCGCTCTGGTTGTAGTAGCC 3'). Two independent PCRs were performed for each individual. The PCR amplicons were gel-purified and cloned in a pGEM-T Easy Vector (Promega)/JM109 High Efficiency Competent Cells (Promega) cloning system. 32 positive clones were picked for each individual and plasmids were extracted using QIAGEN DirectPrep 96 MiniPrep Kit on a QIAvac Multiwell vacuum manifold (QIAGEN). Plasmids were sequenced with T7 primer at the Australian Genome Research Facility, Sydney, Australia. Sequences were quality-checked using Sequencher 4.1.4 (Gene Codes) and aligned with previous identified devil Class I alleles 19 in BioEdit 7.0.9 44 . To minimize errors yielded during PCR, cloning and sequencing, new sequence variants were determined to be real alleles only if they were found in more than one PCR amplification. Alleles were assigned to genomic loci based on nucleotide sequence similarity. Evolutionary relationships of devil Class I sequences were analysed by constructing a Neighbour-Joining phylogenetic tree with MEGA5 45 .

Two pairs of new PCR primers were designed using program Oligo 6.7 (Molecular Biology Insights). One pair (forward 5'- TTTGCAAGCTTCCATGTCTCT -3', reverse 5'-CACTTGTGCTTGGAGTTCAGA -3') amplified partial exon 2 to exon 3 of the Class I gene Saha-UK from devil cDNA. The other pair was designed to detect a deletion in Class I gene Saha-UA, with the forward primer 5'- TGTCCCCCCCTCCGTCTCAG -3'binding to the end of intron 1 and the reverse 5'- CAGGAGAGGAGACCACACTAAGAT -3'binding to the end of intron 5. In about 5% of samples, this set of primers was found to work less efficiently, which could be due to nucleotide variations in primer binding sites. PCR conditions for both primer sets were the same as the one provided above. PCR products were sequenced to confirm correct amplification sites.