In silico searching revealed a total of 23 cytokine sequences, all of which are described in the opossum for the first time and 5 of which are novel for any marsupial species (see Table 1). A number of critical cytokine receptors are also identified, as are the sequences for the hallmark T cell cluster of differentiation markers, CD4 and CD8.

The majority of genes reported in this study were identified using sensitive peptide BLAST searches (Table 2). The most divergent genes, interleukins 2, 4 and 13, were identified using synteny searches. Properties of the putative proteins identified in this study, predicted structures and comparison with human sequences are summarised in Tables 1 and 2. Sequence data of the predicted proteins are available online 33.

Interleukins 2, 4 and 21 and their common gamma chain receptor were identified using both BLAST and syntenic strategies. IL-21 was identified by a sensitive TBLASTN search (e-value = 2e-18) on Chromosome 5:7034081–7057815. The predicted protein is of similar size and contains the same number of exons as human IL-21 [see Additional File 1]. The signal peptide was predicted to be encoded within the first 21 amino acids (score = 7.6, p = 0.06), with N-linked glycosylation sites predicted at positions 46 and 106 and O-linked glycosylation of threonine predicted at position 55. Instability motifs (ATTTA) were not found in the 3' UTR of the sequence before the poly(A)+ signal.

Opossum IL-2 was found by searching genomic sequence flanking IL-21, which is adjacent to, and has significant structural homology with IL-2 in humans [see Additional File 2]. This strategy was adopted since the alignment of human IL-2 against the opossum genome using TBLASTN resulted in no hits. A 395 kb region adjacent to IL-21 was extracted and 15 genes within this region were predicted with GENSCAN. The predicted gene most similar to IL-2 was identified using BLASTP. The sequence was extracted and GenomeScan was used with an IL-2 orthologue to obtain a more accurate prediction. Opossum IL-2 was located on Chromosome 5:7191593–7196834 (Fig 1) and contains several conserved residues essential for biological activities, including two cysteine residues that provide structural stability 34 and the amino acids leucine and aspartic acid within helix A, which are crucial for binding of the ligand to IL-2Rβ in humans 35. Also well conserved is a glutamine residue in the D helix, which is directly involved in the binding of the IL-2Rγ chain 36. Similar to the human sequence, the putative peptide is 142 amino acids in length and contains 4 exons. A signal peptide that contains a potential O-linked glycosylation site (position 13 – Thr) is predicted from positions 1–22 (score = 9.9, p = 0.03). A potential N-linked glycosylation site, not found in humans or mice, but present in several eutherians including the cat and dog, is found at position 101. Four mRNA instability motifs (ATTTA) are present upstream of the poly(A)+ signal.

Opossum IL-2Rγ was identified using TBLASTN (e-value = 8e-119) (Table 1). It shares 61% amino acid similarity with the human sequence [see Additional File 3].

IL-5 was identified on chromosome 1:307529660–307531352. It shares 53.0% identity to human IL-5, and 86.7% identity to the tammar wallaby IL-5 13 [see Additional File 4].

Synteny searches located the sequence for IL-4 [see Additional File 5]. RAD50 (GenBank accession no: AAB07119) and kinesin-like protein KIF3A (GenBank accession no: NP_008985) are situated adjacent to IL-4 and IL-13 in humans. The area between these proteins in opossum was extracted and GENSCAN predictions were searched with BLASTP and FASTP for suitable matches. IL-4 was identified using FASTP and was located on Chromosome 1 (307752915–307754456). The predicted peptide is 138 amino acids in length (Fig 2). It has low levels of identity to human IL-4 (30.8%). Two putative N-linked glycosylation sites were identified. SPScan was unable to predict a putative signal sequence although two instability motifs (ATTTA) were recognised in the 3' UTR region. Despite the variation in sequence between the predicted opossum and human IL-4 protein sequences, disulfide bonds that join helix B to the CD loop and that are important for biological activity are conserved 37.

IL-4 and IL-13 were identified simultaneously using a syntenic approach since they sit adjacent in the human genome [see Additional File 5]. Opossum IL-13 (Chromosome 1:307682382–307686155) is found 74.30 kb upstream from opossum IL-4 and does not contain any glycosylation sites. Alignment with mammal and chicken protein sequences (Fig 3) revealed a truncation of 32 amino acids from the 5' end of the peptide in opossum IL-13. This is probably due to incorrect gene prediction, a fact supported by the absence of signal peptide and any instability motifs.

Opossum IL-6 was identified using a sensitive TBLASTN search (e-value = 0.08). Opossum IL-6 is located on Chromosome 8:296810942–296824133 and the PROSITE IL-6 family motif (C-x(9)-C-x(6)-G-L-x(2)- [F,Y]-x(3)-L) is conserved [see Additional File 6]. The signal peptide is predicted from positions 1–28 (score = 8.1, p = 0.20) and no instability motifs (ATTTA) are found in the 3' UTR. Opossum IL-6 has maintained significant structural similarities to human and other mammalian IL-6 genes despite its comparatively low sequence identity. The number and position of cysteine residues in opossum IL-6 are identical to those found in eutherian and chicken sequences. An arginine molecule in helix D that is involved in IL-6β binding 38 is also conserved.

Opossum IL-12 alpha chain (chr7:260,616,009–260,626,803) was identified using a TBLASTN search and is predicted to be 58% similar to its human orthologue [see Additional File 7]. Cysteine residues are conserved between the marsupial, eutherians and chicken sequences.

IL-10 family members were identified in two clusters. Chromosome 2 contained IL-10 (113139397–113144942; [see Additional File 8]), IL-19 (113283404–113294773; [see Additional File 9]), IL-20 (113319666–113324608; [see Additional File 10]), IL-24 (113362216–113377467; [see Additional File 11]) with identical head-to-tail transcriptional orientation and organisation to their human orthologues. Chromosome 8 contained IL-26 (23485674–23494985; [see Additional File 12]) and IL-22 (23457582–23460076; [see Additional File 13]). The complete IL-22 open reading frame was not identified since the 3' end (approximately 33 amino acids and 2 exons) fell in an unsequenced gap. However, conservation of a predicted N-linked glycosylation site at N54 between putative opossum IL-22 and human IL-22 (a site crucial for IL-22 modulation during the inflammatory response) suggests that this partial sequence is opossum IL-22. Both chicken and the amphibia contain IL-10 family members, although only one IL-19-like ancestral gene replaces IL-19, IL-20 and IL-24 in the chicken 39. Orthology of the IL-10 family cytokines with their eutherian counterparts was confirmed by phylogenetic analysis [see Additional File 14]. All putative IL-10 family members clustered closely with their eutherian orthologs.

CD4 [see Additional File 15] and CD8 [see Additional File 16] were identified by TBLASTN search and found on chromosome 8 (104157682–104183462) and chromosome 1 (716671734–716675645) respectively. Their number of amino acids and potential glycosylation sites are noted in Table 2. Neither we, nor Ensembl, were able to successfully predict the terminal exons of these two genes.

Where possible, gene predictions were verified by alignment with known marsupial cDNA sequences, and compared to Ensembl gene predictions and UCSC similarity features. For instance, a known cDNA sequence is available for Trichosurus vulpecula (possum) IL-10 cDNA (Genbank ref: AF026277). Our predicted opossum IL-10 protein shared 76% amino acid identity with possum IL-10, and exon-intron boundaries match. However, despite our use of robust methodologies, we are still not confident with prediction of the most divergent immune genes sequences. Some doubt exists with our predications for IL-4 and IL-13 and the terminal exons of IL-22, CD4 and CD8. Characterisation of their cDNA, together with laboratory-based assays will ultimately confirm the reliability of the predictions reported here.

Draft sequencing of the genome of a female opossum (Monodelphis domestica) has recently been completed by the Broad Institute 49. Analysis was performed on assembly MonDom4 (January 2006).

To optimise the chances of identifying previously undiscovered sequences, our search strategy relied on a preliminary database screen for sequence conservation, together with positional analyses of the gene sequence relative to other genes within the genome (synteny). Finally, the putative sequence was analysed for the presence of biologically significant sites associated with both structure and function in their eutherian homologues.

Sequence similarity searching (TBLASTN) was performed with known eutherian sequences. Positive hits from the BLAST search with good potential were extracted for further structural analysis. When ambiguities existed between alignments from BLAST results, each of the multiple hits were extracted and inspected. Assessment methods for ambiguous hits included tests for reciprocal-best-hit where the aligned sequence was blasted against SWISS-PROT and TrEMBL protein databases to confirm preliminary findings. Proteins discovered in BLAST searches were used to mine additional homologues. To do this, parameters were optimised for sensitive searching. In order to increase our ability to detect highly divergent sequences, the BLOSUM 45 similarity matrix 50 was used. Additionally, application of soft-masking and the lowering of the neighbourhood word threshold score to 9 increased the chance of detecting homologous sequences that otherwise might have been overlooked using default parameters.

If the protein of interest was not detected by the initial BLAST search, other methods were employed. Similarity searches were performed with genes found in close syntenic regions in the human genome. Syntenic regions were extracted from the opossum database, and passed into GENSCAN 24. The predicted peptide sequences were analysed by performing similarity searches against the SWISS-PROT and TrEMBL databases using BLASTP and FASTP 51. In order to improve the accuracy of identified cytokine sequences, the sequence was re-extracted from the opossum database and the putative protein was re-evaluated using GenomeScan 29. Combined results from GenomeScan and GENSCAN were compared with documented structural features of the cytokine.

For sequences that were not detected using the above methods, a hidden Markov model (HMM) was built and calibrated using the HMMER 2.3.2 package 52. The model was built as a multiple local alignment profile with the Krogh/Mitchison substitution weight matrix 53 and used to search the six-frame translation of the opossum genome.

Ancestral sequences were included in the HMM. These were calculated by programs in the Phylip package 54. PRODIST was used to compute a distance matrix under default settings. After this, the program NEIGHBOUR was used to create a neighbour joining (NJ) tree from the matrix. The tree was rooted with a teleost species. Following this, ProML was set to produce ancestral sequences at each of the NJ tree nodes.

Once the gene of interest was located, exon/intron boundaries were identified using the gene prediction programs GENSCAN 24 and GenomeScan 29. Our experience suggests that some caution is advisable in the interpretation of data from existing gene prediction software; excessively long predicted genes ('thready' gene predictions) due to mis-identification of first exons and merging of adjacent genes, and unlikely predictions of splice sites (based on comparison with orthologous sequences) were the most common problems we observed. Mindful of these limitations, our gene predictions were compared with known gene structures. The presence of signal peptides was predicted by SPScan (Accelrys GCG) and estimation of glycosylation sites were made with NetOGlyc 3.1 55 and NetNGlyc 1.0 56. Finally, sequences were submitted to the PROSITE database 57 for detection of protein family motifs that would confirm gene identify.

Sequences from the opossum and other species were aligned using ClustalW 58. Accession numbers of sequences used in analyses are shown in figure legends. Sequence labels in the alignments are abbreviated by the first letter from the genus with the first two letters from the species name followed by the gene name. In figures, residues with functional importance are highlighted.

Neighbour-joining (NJ) trees were constructed using the Jones-Taylor-Thornton substitution model 59 and 500 bootstrap replicates in MEGA 3.1 60. The tree, constructed from amino acid sequences, was rooted using chicken sequences.

Sequence identity and similarity calculations were carried out using GAP (Accelrys GCG), with the Needleman-Wunsch alignment 61, except for IFN-α genes which were calculated in GenDoc 62 using the BLOSUM 35 similarity matrix 50 for comparisons of human and opossum IFN-α genes and BLOSUM 80 matrix 50 for comparisons among opossum sequences. GCG, GENSCAN, BLASTP and FastA programs were accessed through the Australian National Genomic Information Service (ANGIS) 63.