Most of the IFN gene candidates were identified through the National Center for Biotechnology and Information (NCBI)'s bovine genome resource by using the basic local alignment search tools (BLAST) http://www.ncbi.nlm.nih.gov/genome/seq/BlastGen/BlastGen.cgi?taxid=991339. Additional searches were performed through NCBI by using the appropriate genome resource http://www.ncbi.nlm.nih.gov/mapview/40 for other species, which are discussed later in this section, and by using the basic nucleotide BLAST suite http://www.ncbi.nlm.nih.gov/blast/Blast.cgi?CMD=Web&PAGE_TYPE=BlastHome4142. Several combinations of BLAST algorithms and databases within NCBI were utilized for this work and are described below 43.

1. MegaBLAST was designed to compare highly related nucleotide sequences and works best when the target sequence has a 95% identity or higher to the query sequence.

2. Cross-species megaBLAST, also referred to as discontiguous BLAST, is a derivative of megBLAST that ignores certain bases, thereby allowing mismatches. It was designed to compare nucleotide sequences from one species to nucleotide sequences in another species.

3. BLASTN also compares nucleotide query sequences to a nucleotide database. This algorithm is slower than megaBLAST, but it can identify shorter sequence matches than megaBLAST. It was not specifically designed for cross-species comparisons.

4. TBLASTN was designed to compare a protein sequence with a nucleotide database dynamically translated in all reading frames.

1. The "genome (reference)" database represents the most current publicly available assembly of a genome. The most current assembly of the bovine genome at the time this work was completed was assembly 3.1. The most current assembly for other species examined in this work are placed in parenthesis here – human (36.2), mouse (37.1), horse (1.1), and dog (2.1).

2. The "WGS contigs" database contains the contigs, or overlapping unassembled sequences, that forms the basis for the assembled genome. Both pig and cat do not have an assembled genome available at this time and only the "WGS contigs" database could be searched for genomic information for these species.

3. The "traces-WGS" database contains the trace data for whole genome shotgun sequence (WGS) bacterial artificial chromosome (BAC) end sequencing. This database contains single pass sequencing reads that are not trimmed based on quality or vector contamination.

4. The "nucleotide collection (nr/nt)" database contains all Genbank, RefSeq, EMBL (Europe's primary nucleotide database), DNA Database of Japan (DDJB), and many Protein Databank (PDB) sequences. The "nucleotide (nr/nt)" database is subdivided into "human nucleotide (nr/nt)," "mouse nucleotide (nr/nt)," and "others nucleotide (nr/nt)" databases. The "others nucleotide (nr/nt)" database does not contain any mouse or human sequences.

Bovine IFNB, IFNA, IFNW, and IFNT cDNA sequences (Table 1) from GenBank were used to perform a megaBLAST search in the bovine "genome (resource)" database. Human IFNE, murine IFNZ, porcine IFND, human IFNK, and human IFNL/IL28-29 sequences (Table 1) were queried with cross-species megaBLAST in the bovine "genome (reference)" database because no bovine homologues for the latter group of genes have been reported. The translated sequence for each of the non-bovine IFN cDNA were also queried with TBLASTN in the bovine "genome (reference)" database because the TBLASTN algorithm can often identify homologues that are not detected through other searches. IFNL/IL28-29 nucleotide and amino acid sequence were also queried in the bovine "WGS contig" database by using a cross-species megaBLAST and TBLASTN search, respectively, to verify no sequence with high identity to IFNL/IL28-29 in the bovine genome was missed. Specific genes were analyzed in the bovine "traces-WGS" database to verify frameshift mutations or nucleotide variations from the query sequences.

Alignments for the genomic IFN ORFs were created through ClustalW in BioEdit version 7.09, with individual genes denoted by their GLEAN numbers. Phylogenetic trees were constructed in MEGA4 through the Neighbor-joining (NJ) method with bootstrapping test (1000 replicates). The tree was rooted to IFNK 5 and a second tree was created with the assumption of a non-uniform rate of change between sites (gamma = 1).

The localization and identity of all repetitive elements were determined by using the RepeatMasker program http://www.repeatmasker.org/52, which uses the RepBase library of repeat elements 53. Sub-locus 1, corresponding to 20000–711500 bp in scaffold Chr8.25, and sub-locus 2, corresponding to 2000–446000 bp in scaffold Chr8.34, sequences were first selected through Apollo and imported into a word processing program, Microsoft Word. All gaps within the scaffolds, which are represented by an "N" in the bovine assembly, were removed manually. IFN sub-loci sequences were then analyzed in RepeatMasker version 3.1.9 run in default mode with blastp version 2.0MP-WashU http://blast.wustl.edu/54 to determine the percentage of repetitive elements. Bos taurus was set as the assumed species within the program parameters. Simple repeats and low complexity regions were not masked, which means they were not excluded as start sites for a BLAST match, and the matrix was optimized for 42% GC content based on sub-loci optimization pre-runs.

Evidence for the presence of all previously known Type I IFN subfamilies except IFNZ was found on chromosome 8 of the bovine genome assembly 3.1 (Table 2). IFNZ has only been reported in mouse 2755, and so its absence in Bos taurus was anticipated. Both IFNK and IFNE are present as single genes with intact ORF and are assumed to be functional, providing the first evidence that either of these subfamilies is present in ruminants. Bovine IFNK and IFNE have 81.2% and 84.7% nucleotide identity respectively when compared to their human orthologues, values that are similar to their degree of conservation between human, cat, dog, and pig (A. Walker, unpublished data). The IFNW family is greatly expanded compared to other species that have so far been examined. There are 24 potential IFNW and at least 8 pseudogenes. The query IFNW [Genbank:M11002] 34 exactly matches two IFNW annotated from the genome database (GLEAN 09983 [Bovine Genome Database temporary ID:2733] and GLEAN 10004 [Bovine Genome Database temporary ID: 2525]). The remaining IFNW range in sequence identity from 86% to 96% relative to the query sequence.

The IFNA and IFNB are also present in multiple copies, with 13 and 6 genes, respectively, although neither family is as large as the IFNW. An apparent IFNB pseudogene, deemed nonfunctional due to a frameshift deletion, also exists.

IFND are only represented as three pseudogenes, a not unexpected finding, as a functional gene has only been reported previously for the pig 56.

Three apparently functional IFNT are found within the locus. Surprisingly none of these provide an exact match for any of the many cDNA and gene sequences that have previously been reported. Previous mRNA sequencing of the IFNT family had indicated that at least 18 bovine IFNT might exist 57. Only three IFNT are present in the bovine genome assembly 3.1, however. One particular, well established sub-family, the IFNT2 grouping 57, is not represented at all in the assembly. Additional analysis revealed 45 acceptable matches to IFNT in the WGS contig database. Since the bovine genome at this time has 7.1 X coverage, the number of IFNT matches divided by this coverage value suggests the possibility of around six IFNT. One explanation is that these "extra" genes have been lost in the assembly process, but even this higher value is still significantly lower than the 10 to 18 IFNT previously believed to exist. Some of the latter are most likely alleles.

Most interestingly, we detected a novel Type I IFN, which, as we shall discuss later, consists of three potentially functional genes and one pseudogene, none of which provides a close sequence match with any previously described Type I IFN. For convenience, and until an appropriate nomenclature is approved, this new family will be termed IFNX.

A weak sequence identity to IFNL was found on chromosome 13, specifically located on scaffold Chr13.80 [Genbank:NW_001493172] from 635,850 to 636,120 bp. This sequence appears not to encode a functional gene in either the 3.1 assembly or the WGS contig database. These data suggest that the Type III IFN family exists only as a relic and is no longer a functional component in bovine pathogen defenses.

The Type I IFN locus is organized similarly in mouse and human, and possibly also in pig, i.e. there is relatively conserved synteny across rodents, primates, and swine. Two mammalian IFN genes of ancient origin, IFNB and IFNE, define the outer limits of the locus, with all the other genes, except IFNK, distributed between these two markers. The genes are predominantly (but not exclusively) localized on one strand and transcribed in the same direction as the IFNB and IFNE 27. The relative arrangements of the murine and human versus the bovine IFN locus are illustrated in Fig. 1. The bovine Type I IFN locus is clearly organized differently than that of the other two species. Instead of a single stretch of DNA defining the locus, cattle have two sub-loci (1 & 2) encompassing 701 kb and 441 kb, respectively, separated by a gap estimated to be approximately 11 megabases (Mb) in assembly 3.1 (but ~750 kb in assembly 4.0). The gene density based on the both predicted open reading frames (Gene Sequence map) and bovine EST and mRNA alignments with the assembled sequence (BT UniG map) is lower in the region between the two IFN sub-loci than much of the rest of chromosome 8, but many genes are present and actively transcribed.

The following explanation of the locus organization has been based on the chromosome map assigned during the assembly process and reported by NCBI (Fig. 2). An IFNB defines the distal end (relative to the start of the chromosome map) of sub-locus 2 (Fig. 1 &2). IFNE, while present in the bovine, is located towards the distal end of sub-locus 1. The majority of the genes in both sub-loci are transcribed in the same direction as the distally placed IFNB, except one cluster of IFNW and IFNA and the solitary IFNE, which are transcribed in the opposite direction. IFNK is present in a single "copy" nearer to the start of the chromosome map and well separated (6.044 Mb) from the closest sub-locus (sub-locus 1) (Fig. 1). The bovine IFNK location is very similar to that in the human where a single IFNK is located 6.5 Mb from the Type I IFN locus 22 (Fig. 1).

There are three clusters of IFNA/IFNW. Two of them are on sub-locus 1, one at the proximal end, the second placed about half way along (Fig. 2). A gene set in the first IFNA/IFNW cluster 1 is a palindrome to one in the second cluster. The corresponding gene pairs have complete nucleotide identity within their coding regions, suggesting that the duplication or gene conversion event that led to their formation occurred quite recently. The third cluster of IFNA/IFNW is at the distal end of sub-locus 2, but lacks the duplicated group of four genes in IFNA/IFNW clusters 1 and 2.

Only one non-IFN gene is detectable within sub-loci 1 and 2, an intronless kelch-like 9 (KLHL9) located 33.5 kb proximal to IFNA/IFNW cluster 2 in sub-locus 1 (Fig. 2). The orthologous KLHL9 gene can be found in the Type I IFN locus of the mouse approximately 25 kb from the nearest functional IFN (IFNA8) 27 and 29 kb from the nearest IFN (IFNA6) in human. The fact KLHL9 has resisted duplication despite residing close to genes undergoing multiple duplications is noteworthy and possibly indicates that multiple copies of this gene are not well tolerated.

The presence of KLHL9, which appears to be under different evolutionary constraints than IFNW and IFNA, close to IFNA/IFNW cluster 2 suggested the cluster as whole might be under different evolutionary control than cluster 1 or 3. Evolutionary divergence rates do not indicate that this is the case, however (Table 3). While cluster 1 does have a slightly higher divergence rate than cluster 2, all three clusters are relatively constant in their rate of change. Inter-cluster divergence is actually very low in all three of the IFNA/IFNW clusters.

The IFNT and two IFNW pseudogenes are neighbors at the distal end of sub-locus 1, suggesting that this cluster of genes originated from an IFNW that had become isolated from other IFNW before the divergence of the IFNT. Its unique position outside the IFNA/IFNW clusters and close to the edge of the sub-locus may have permitted the rapid expansion and evolution of the IFNT family without the restraints placed on the clustered IFNW.

All non-ruminant species examined to date, including mouse, human, cats, dogs, rabbits, and pigs, contain only one IFNB 16. In cattle, this family has clearly expanded and extends from the distal end of IFNA/IFNW cluster 3 to the end of sub-locus 2. Interspersed within these multiple IFNB are members of the previously unidentified IFN family, IFNX. Again, it is tempting to hypothesize that the IFNX and expanded IFNB family were able to emerge due to their location on the edge of the sub-locus 2, as suggested for the IFNT in sub-locus 1.

Repetitive elements have been implicated in gene duplication by creating regions predisposed to homologous recombination 5859 and also in organizing the assembly of enhanceasomes, as recently described for transcription of IFNB 60. Over one third of the bovine Type I IFN locus consists of interspersed repeats, rather less than the 43% assessed for the murine Type I IFN locus 27. Repeats are more enriched in sub-locus 1 largely because of the presence of a greater number of long interspersed nucleotide elements (LINE) and long terminal repeats (LTR) (Table 4). LINE2 elements are absent in both sub-loci. Short interspersed nucleotide elements (SINE)s are present in similar proportions, approximately 15%, for both sub-loci. The arrangement of these elements provides no obvious insight into whether the sequences are involved in gene duplication or in controlling transcription.

As mentioned previously, a gene set located on the distal end of IFNA/IFNW cluster 1 is a palindrome to a gene set on the distal end of cluster 2 (Fig. 3A). Two IFNA and two IFNW are present in each gene set and these are designated as A(1) and A(2) for the two IFNA and W(1) and W(2) for the two IFNW in cluster 1. A(1) and A(2) are located at the proximal and distal ends of the gene set in cluster 1, respectively. W(1) and W(2) are located from proximal to distal between A(1) and A(2). The genes with identical ORFs in cluster 2 are designated as A(1)^/, A(2)^/, W(1)^/, and W(2)^/ (Fig. 3B). The ORF and the first 350 bp of the 3^/ UTR for the two IFNA and IFNW gene pairs are identical, and an approximately 550 bp promoter for both IFNA gene pairs and the W(2)/W(2)^/ pair are also identical. The promoters of the W(1)/W(1)^/pair, while closely similar (99%), are not identical, however. All nucleotide differences between the W(1)/W(1)^/ gene pair are within a region between 300 bp and 400 bp upstream of the transcriptional start site.

Two different evolutionary processes, either gene duplication or gene conversion, could possibly explain the existence of the IFNA/IFNW palindromic gene sets. Gene duplication involves the formation of a new gene copy. Gene conversion, on the other hand, does not generate new gene copies, but instead homogenizes existing genes. Both gene duplication and gene conversion have been specifically implicated in the evolution of the IFNA in human, chimpanzee, dog, rhesus monkey, rat, and mouse 306162. Gene conversion, specifically, was predicted by two different statistical programs, GARD and GENECONV, in humans, chimpanzee, rhesus monkey, and mice. Furthermore, despite IFNA genes aligning in conserved positions on a locus map for chimpanzees and humans, the subfamily separated into species-specific clades on phylogenetic analysis 61, strongly indicating gene conversion has occurred in the IFNA subfamily in these two species. Although gene duplications cannot be unambiguously distinguished from gene conversions 63, the latter seldom involve sequence longer than 1 kb in mammals, with 3 kb considered the maximum length 63. Therefore, when the sequence tract involved is "too large" for gene conversion, gene duplication is usually implicated 64. The palindomic gene set involves at least a 27 kb tract, far exceeding this size limit and reducing the likelihood of a conversion event. Therefore, a segmental duplication event, which is a specific type of gene duplication that involves a large segment of a locus, combined with an inversion is the best explanation for the palindrome 65.

Comparison of the rate of non-synonymous nucleotide change relative to the rate of synonymous change can provide information about the type of selection operating on the members of a multigene families 66. If neutral selection is occurring, then all nucleotides in a sequence are equally likely to change. Consequently the rate of synonymous nucleotide changes (dS) will be equal to the rate of non-synonymous changes (dN) and dS:dN will equal 1. Rapid change in the amino acid sequence is the desired endpoint for positive selective pressure. Hence, in this scenario, dN will exceed dS, and dN:dS will be greater than 1. Conversely, if strong selection against amino acid change is present (purifying selection), dN will be less than dS and dN:dS will be less than 1. Virtually all pairwise comparisons within IFNA, whatever the species 6167, and IFNT 57 have shown the overall value for dN not to be significantly higher than dS. Indeed, dN values have been generally calculated to be lower than dS, consistent with the conclusion that there has not been strong positive selection for amino acid change within the coding regions of these subfamilies of IFN.

The dN:dS for all multigene bovine IFN subfamilies, including only IFN annotated during this work, is illustrated in Fig. 4. The dN and dS values for every gene pair in a Type I IFN subfamily were calculated and plotted against each other. No multigene IFN subfamily (IFNA, IFNB, IFNW, or IFNT) in bovine has a dN significantly exceeding dS. In fact, bovine IFNA and IFNW provide strong evidence for purifying selection. The significance of purifying selection within IFNW and IFNA subfamilies was also verified through a codon based Z-test (p < 0.001), which determines selective pressure for a gene pair or group of genes based on the difference between dN and dS through a one-tailed t-test (Table 5). Four IFNB pairs out of 15 pairwise comparisons examined provide some evidence for purifying selection on the basis of the pairwise codon based Z-test (Table 6), but such selective pressures are not evident for the family as a whole (p = 0.171). On the other hand, the same test employed for neutral selection provided no evidence for a lack of selective pressure operating on any members of the IFNB subfamily (p = 0.343), indicating the six genes comprising the bovine IFNB subfamily may be too few for a meaningful analysis. Since only three IFNT were identified, statistical analysis of this subfamily from the genomic data was not possible.

The classic model of gene duplication states that after a duplication event one gene continues to perform the ancestral function while the second either rapidly evolves to fill a new niche or becomes inactive 656869. As a consequence gene duplication is usually followed by a period in which there is an acquisition of non-synonymous nucleotide changes in one of the two genes, leading to a divergence in amino acid sequence. This temporary relaxation of purifying selection, in which dN:dS approaches 1, permits the gene to become fine-tuned to its new role or, more commonly, results in pseudogenization. Such a sequence of events does not appear to have occurred during the large scale expansion of the IFNW family where there is strong evidence for purifying selection operating on the coding regions of the genes (Fig. 4 & Table 5). One potential explanation is that sub-functionalization is occurring through alterations in the manner in which these genes are subjected to transcriptional control rather than in the structure of the proteins themselves. Evidence for positive selective pressure in the regulatory regions of IFNW would strongly support this hypothesis, but a detailed promoter analysis is beyond the scope of this work.

Based on the current assembly, the IFNX subfamily consists of three potential genes and one probable pseudogene. The origin of these genes is currently unclear, but they appear to constitute a unique IFN subfamily, whose closest relatives are the IFNA based on a Maximum Composite Likelihood divergence analysis, with IFNX sharing over 65% nucleotide identity to IFNA and only about 45% nucleotide identity to IFNB. They are clearly defined on phylogenetic trees based on their ORF as a distinct clade that is most closely related to IFNA; however, phylogenetic analyses are conflicting with regard to when IFNX first appeared. Phylogenetic trees calculated on the assumption that all nucleotide sites within the coding sequence change at the same rate indicate IFNA separated from IFNX prior to or corresponding with the radiation of eutherian mammals, i.e. more than 150 MYA 35 (Fig. 5A). Conversely, phylogenetic trees that take into account substitution rate variation from site to site indicate that bovine IFNX and bovine IFNA emerged from a common ancestor after the radiation of the major mammalian orders (Fig. 5B). In other words, this model recognizes that certain amino acids and, by corollary certain nucleotides, are more highly conserved than others across subfamilies. This second model most closely matches the amino acid differences among IFN subfamilies, since amino acids critical to preserving the tertiary structures of IFNB, IFNA, IFNT, and IFNL are much more highly conserved than ones in less critical regions of the proteins 5570. Preliminary examination of IFNX indicate some of the more highly conserved amino acids for other Type I IFN, such as cysteine residues 1 and 99 (discussed below), are also conserved in IFNX. In addition IFNX and the IFNX pseudogene are absent in human, mouse, porcine, feline, and canine genomic databases, yet some remnants of their presence might be expected if this family emerged early in the evolution of mammals and prior to the main radiation events.

Two of the three potential IFNX genes contain the four conserved cysteine residues required for the disulfide bonds (1–99; 29–139) encountered in IFNA and those Type I subfamilies that emerged from IFNA, namely IFNW, IFND, and IFNT. The third potential gene, GLEAN 24316 [Bovine Genome Database temporary ID:2755], contains an early termination codon at codon 125, which eliminates the second disulfide bond (29–139) (Fig. 6), raising the possibility that it is a second pseudogene. The "trace-WGS" database was visually examined to verify the presence of this early stop codon. Five BACs, all of which were in the minus orientation, contained sequences that exactly matched the complementary sequence to Glean 24316, i.e. all contained the early termination codon.

Previous studies of IFNA have identified three regions that are strongly associated with IFN-receptor interaction and are termed interferon receptor recognition peptides (IRRP)1–3 707172. IRRP1 (27–35) and IRRP2 (78–105) control the initial binding of IFN to the Type I receptor and are highly conserved among IFNA. IRRP3 (123–140) modulates the downstream signaling pathways, so that amino acid changes in this region can explain some differences in biological activity among different IFNA. The protein products of the three IFNX do not possess identical IRRP1 and IRRP2 motifs as IFNA, but these two regions are highly conserved within the subfamily, emphasizing, first, the possible importance of this motif and second that the IFNX family is unique and distinct from IFNA. IRRP3 was absent in GLEAN 24316 again suggesting that it may be a pseudogene. The two remaining IFNX members differed in their IRRP3 sequences, a not unexpected finding as changes in this region may provide subtle differences in biologic activity between the two family members. None of the IFNX genes contain the N-glycosylation sequence (N-X-S/T) common in other Type I IFNs that could alter IFN-receptor interaction.

No evidence for IFNX expression could be found in any EST databases, although, genes with high identity to IFNX exist in the equine genomic database. The conservation of this gene family in species that diverged at least 80 million years ago suggests that the family may have an important function in ungulates. However, the apparent absence of IFNX genes in pigs, also an ungulate, is puzzling. Possibly, IFNX has a specific function in herbivores that is not required in omnivores, most likely in immune defense against particular viruses or other pathogenic organisms affecting such species.

The identification of a novel Type I IFN gene, the IFNX, is an unexpected and possibly important finding. The proteins encoded by this family of genes differ sufficiently in primary sequence from related Type I IFN to justify a separate designation from the related IFNA and IFNB. The presence of a distinct cluster of IFNX within the Type I IFN locus, the phylogenetic position of IFNX as a separate clade within the IFN tree, and the conservation of critical amino acid residues, are totally consistent with classifying the IFNX as a distinct Type I IFN subfamily. Whether IFNX are responsive to a viral challenge and able to interact with the Type I IFN receptor and elicit a typical Type I response in their target cells has yet to be verified. Substantial work will be necessary to characterize this subfamily fully, but its place as a separate clade within the Type I IFN would appear to be assured.