Identification and characterisation of a novel immune-type receptor (NITR) gene cluster in the European sea bass, Dicentrarchus labrax, reveals recurrent gene expansion and diversification by positive selection
- First Online:
- Cite this article as:
- Ferraresso, S., Kuhl, H., Milan, M. et al. Immunogenetics (2009) 61: 773. doi:10.1007/s00251-009-0398-3
- 458 Views
In the last decade, a new gene family encoding non-rearranging receptors, called novel immune-type receptors (NITRs), has been discovered in teleost fish. NITRs belong to the immunoglobulin superfamily and represent an extraordinarily divergent and rapidly evolving gene complex. Genomic analysis of a region spanning 270 kb led to the discovery of a NITR gene cluster in the European sea bass (Dicentrarchus labrax). In total, 27 NITR genes and three putative pseudogenes, organised in a tandemly arrayed cluster, were identified. Sea bass NITR genes maintain the three major genomic organisations that appear to be essentially conserved among fish species along with new features presumably involving processes of intron loss, exon deletion and acquisition of new exons. Comparative and evolutionary analyses suggest that these receptors have evolved following a “birth-and-death” model of gene evolution in which duplication events together with lineage-specific gain and loss of individual members contributed to the rapid diversification of individual gene families. In this study, we demonstrate that species-specific gene expansions provide the raw material for diversifying, positive Darwinian selection favouring the evolution of a highly diverse array of molecules.
KeywordsNovel immune-type receptor (NITR)EvolutionGenomic structuremRNA expressionTeleost
Increasing attention is being directed to the role of innate processes within the immune system, the variety and diversity of innate immune receptors, their interaction with other surface molecules and the mechanisms employed for transduction of intracellular signals. Several clusters of innate immune receptors have been identified in mammals. In primates, particular attention has been given to receptors belonging to the immunoglobulin superfamily; the killer Ig-like receptors (KIR; Martinez-Borra and Khakoo 2008), the Ig-like transcripts (ILT; Samaridis and Colonna 1997; leukocyte inhibitory receptors and monocyte inhibitory receptors), the leukocyte-associated inhibitory receptors (Meyaard et al. 1997) and the Fc receptors. Interestingly, all these genes, which are structurally and functionally related, are encoded in the same region of the human chromosome 19q13.4, termed the leukocyte receptor complex (LRC; Torkar et al. 1998), and also in non-human primates where they show a tandem arrangement (Canavez et al. 2001). KIR receptors, members of the natural killer complex (NKC), have been proposed to interact with MHC class I molecules, enabling NK cells to discern target cells exposing “missing self” signals (Yokoyama and Plougastel 2003). These receptors are type I transmembrane proteins characterised by the presence of a variable number of extracellular Ig domains of either the C2-type (I) or the novel V-type (V). The Ig domains are followed by a stem region, a transmembrane region (TM) and a cytoplasmic tail (Cyt). Depending on the cytoplasmic tail, they can be subdivided into inhibitory and activating receptors. In general, inhibitory KIRs possess long cytoplasmic tails containing one or two immunoreceptor tyrosine-based inhibitory motifs (ITIM), whereas short-tailed KIRs generate activating signals. In the last decade, two gene families encoding non-rearranging receptors with a similar molecular structure to that of KIRs have been discovered in teleost fish, the novel immune-type receptors (NITRs) and the novel immunoglobulin-like transcripts (NILTs). So far, two and four NILTs have been isolated from the common carp (Stet et al. 2005) and the rainbow trout (Kock and Fischer 2008; Østergaard et al. 2008), respectively. NILTs possess one or two V-type Ig domains followed by a transmembrane domain and a cytoplasmic tail containing one immunoreceptor tyrosine-based activation motif (ITAM) and/or a variable number of ITIM domains.
NITR genes have been identified in several fish species. Twenty-six individual genes, all located on a single genomic region spanning ?113 kb, were first characterised in the Southern pufferfish (Strong et al. 1999), then followed the isolation and description of four NITRs in the rainbow trout (Yoder et al. 2002), 18 individual genes in the channel catfish (Hawke et al. 2001; Evenhuis et al. 2007) and one transcript in the Japanese flounder (Piyaviriyakul et al. 2007). NITRs are best characterised in the zebrafish in which 12 families were found on chromosome (chr.) 14 and two on chr. 7 (Yoder et al. 2001, 2004) and in medaka with the identification of 24 families located in three gene clusters, respectively, on chromosomes 10, 18 and 21 (Desai et al. 2008). Transcripts encoding a NITR protein have also been found in the Atlantic salmon, Atlantic cod, Atlantic halibut, lake whitefish and stickleback (Yoder 2009). NITRs appear to represent a large and highly diversified gene complex, with several forms identified according to: (1) number of extracellular Ig domains (V-type and I-type), (2) number of exons encoding the ectodomains, (3) presence and location of joining (J) or J-like sequences, (4) presence of ITIM motifs in the cytoplasmic tail and (5) presence of charged residues in the transmembrane region. NITRs have been proposed to be functional homologs of mammalian KIRs, as both NITRs and KIRs appear to stimulate the same signalling pathways (Yoder 2009). In zebrafish, activating NITRs have been shown to redirect NK cell-mediated cytolysis by interaction with the fish ortholog of DAP12 (Yoder et al. 2004), whilst an inhibitory activity has been proposed for ITIM and ITIM-like domains, exerted through recruitment of the tyrosine phosphatases SHP-1 and SHP-2 and/or the inositol phosphatase SHIP-1, which transduces negative signals (Bléry et al. 2000; Yoder et al. 2007; Piyaviriyakul et al. 2007). Moreover, the crystal structure of NITR Ig domains provides evidence that binding specificity for the allogeneic determinant resides in the V region CDR1-analogous loop and demonstrates their potential interaction with a polymorphic set of ligands (Cannon et al. 2008), like NK cell receptors.
Even though NITR genes are structurally similar to certain members of NKC, their origin remains to be elucidated. Comparative genomic analysis between the NITR gene clusters and mammalian LRC showed only a weakly conserved synteny, and it cannot be excluded that the presence of NITRs may be restricted to teleost fish (Yoder 2009). Data collected so far demonstrate that NITRs represent an extraordinarily divergent and rapidly evolving gene family. Different NITR profiles and a remarkable diversity of gene content were found across teleost species, suggesting that independent, lineage-specific tandem gene duplications and family expansions are the most frequent mode for the NITR gene complex evolution (Litman et al. 2001; Yoder 2009). The functional significance of such lineage-specific gene expansions remains to be further investigated.
In this paper, we describe the identification and characterisation of 27 NITRs in the European sea bass (Dicentrarchus labrax), all encoded in a single genomic region spanning 270 kb. Classified into at least 18 families, sea bass NITRs represent a highly diversified gene complex generated by repeated tandem duplication events followed by rapid diversification. The resolution of this gene cluster together with the previous genomic characterisation of NITRs in zebrafish, medaka and Southern pufferfish allowed an exhaustive evolutionary analysis as well as the examination of shared synteny.
The European sea bass is a marine teleost fish of great importance for fisheries and aquaculture, being an excellent food fish with high commercial value. The great importance of D. labrax for marine aquaculture has fuelled an increasing number of studies in many different areas such as immunology, endocrinology, bone morphology and muscle physiology, whilst genomic resources for this species have been constantly improving, leading to a whole genome sequencing project which is now well advanced (R. Reinhardt, unpublished data). In particular, the D. labrax immune system is one of the best characterised among teleost species (Scapigliati et al. 2002; Randelli et al. 2009), providing a nice model for immunological studies, which is complementary to the best fish models such as the channel catfish and zebrafish. These two are freshwater fish belonging to Ostariophysi, a less derived group, whilst D. labrax is a marine fish representative of the highly derived Perciform order, which is the most species-rich vertebrate group with over 20,000 species (Nelson 2006).
Materials and methods
In silico identification of sea bass NITRs
A first bacterial artificial chromosome (BAC) putatively containing sea bass NITRs was found through Blast analysis of sequenced BAC-ends of a sea bass BAC library (Kuhl et al., in preparation). After complete shotgun sequencing of the first BAC (bassbac-18k1), a second BAC (bassbac-79e10) was sequenced, which potentially contained a further fragment of the candidate genomic region based on Blast analysis.
The DNA of identified BAC clones was isolated by alkaline lysis, and subsequently remaining Escherichia coli DNA was removed by ATP-dependent exonuclease digestion. Purified BAC-DNA was sheared by ultrasonic sound, and fragment sizes of 1–4 kb were selected for end-polishing with T4 DNA polymerase and DNA polymerase I (Klenow). Fragments were ligated with T4 DNA ligase into the SmaI restriction site of puC19 sequencing vector. Competent E. coli DH10B cells were transformed by electroporation. For each BAC, a library with approximately tenfold coverage was constructed, and plasmid DNA was purified for sequencing with ABI BigDye v3.1 Terminator chemistry on ABI3730xl (Applied Biosystems Inc., Foster City, CA, USA) capillary sequencers. Raw sequences were processed by PHRED (Ewing and Green 1998), and removal of vector backbone or low-quality sequence was done by LUCY (Chou and Holmes 2001). The remaining sequences were screened for E. coli or BAC vector contamination by megablast. The BAC inserts were assembled using PHRAP (available from Phil Green, University of Washington; www.phrap.org).
In order to identify NITR genes in the sea bass genomic contig, all the publicly available NITR protein sequences were used as queries for Blast analyses (Wolfsberg and Madden 2001) using the tblastn option. The APOLLO software (Lewis et al. 2002) was used to visualise the organisation of the analysed genomic region and to manually annotate NITR genes structures based on all the available supporting evidence. Signal peptide/leader sequences were predicted using SignalP 3.0 server (http://www.cbs.dtu.dk/services/SignalP/), whilst TMHMM v2.0 server (http://www.cbs.dtu.dk/services/TMHMM/) was used to search for transmembrane domains.
Families of sea bass NITRs were defined using the MatGat programme (MATrix global alignment tool; Campanella et al. 2003) based on the criterion of 70% shared identity within the peptide sequence of the V domain.
Tissue collection and RNA extraction
Four juvenile sea bass individuals (17–20 g) were collected from the experimental aquaria of the Istituto Zooprofilattico Sperimentale delle Venezie (Padova, Italy) and sacrificed using an excess of anaesthetic. Eight different tissues/organs (gill, spleen, liver, intestine, skeletal muscle, skin, head kidney and whole blood) were collected from each animal and stored in RNALater® (Ambion, Austin TX, USA) at 4°C for 24 h followed by long-term storage at ?20°C. Total RNA was extracted using an RNAeasy Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer’s specifications. The quality of the RNA was checked by gel electrophoresis on a 1% agarose gel containing SYBR Safe™ DNA gel stain ×10,000 (Invitrogen™, Carlsbad, CA, USA).
Amplification of NITR cDNA segments and full-length transcripts
In order to validate exon–intron boundaries and transmembrane predictions and to search for short sequence motifs, cDNA amplification and sequencing of all predicted NITR genes in the cluster was carried out using RNA extracted from D. labrax head kidney. One microgram of total RNA was reverse-transcribed to cDNA using Superscript II (Invitrogen™). Primer pairs for all the sea bass NITR genes were designed based on the in silico predicted gene sequence. Whenever possible, forward and reverse primers included the whole putative coding region of the gene. One microlitre of diluted (1:10) cDNA was used as template in PCR. Cycling conditions were: initial incubation at 94°C for 2 min followed by 45 cycles at 94°C for 45 s, 60°C for 30 s and 72°C for 45 s. A final extension step at 72°C for 5 min was added at the end of the last cycle. For each NITR gene, annealing temperature was set according to the predicted melting temperature of primers. Both 3? and 5? rapid amplification of cDNA ends reactions were also carried out for five transcripts.
Evolutionary analyses were performed to determine patterns of divergence of the NITR genes in D. labrax as well as to define putative orthology between NITR genes in different teleost species. All published protein sequences of NITR V domains were aligned using ClustalW (Thompson et al. 1994), applying default settings, whereas CodonAlign v2.0 was used to carry out the nucleotide alignment of V and I coding regions. NITR sequences described from medaka, zebrafish, Southern pufferfish, channel catfish, rainbow trout, Japanese flounder and green-spotted pufferfish were included in the alignment. It should be noted that sea bass NITRs, as well as other teleost ones, are named accordingly to the order that they were identified in each species. Thus, NITRs with the same name in different fish species are generally not orthologous.
Phylogenetic trees were inferred using the maximum likelihood (ML) method (Felsenstein 2004). For amino acid sequences, the most appropriate model of protein evolution was selected using the Akaike information criterion as implemented in ProtTest v1.4 (Abascal et al. 2005). The ML analyses were performed with the Phyml 2.4.4 programme (Guindon and Gascuel 2003) by applying the JTT+I+G+F evolutionary model; non-parametric bootstrap resampling (500 replicates) was also performed to evaluate the robustness of tree topology. For nucleotide sequences, ML analysis was performed by applying the GTR+I+G model. The TREEFINDER programme (Jobb et al. 2004) was then used in order to evaluate alternative phylogenetic hypotheses.
Tests for positive selection were carried out by comparing the number of non-synonymous substitutions per non-synonymous site (dN) with the number of synonymous substitutions per synonymous site (dS) in a ML framework. A ratio of dN/dS (?) greater than 1 is usually taken as evidence of selection following different methodological approaches. First, a series of nested models as implemented in the software PAML ver. 4 (Yang 2007) were evaluated, comparing “neutral” models, which restrict ? to values lower or equal to 1 (models M1, M7 and M8a) with models allowing for positive selection (? greater than 1 for a class of sites; models M2 and M8). Comparison of nested models was carried out using a likelihood ratio test (LRT). All analyses were performed twice starting from different initial values of ? to check for convergence in ML parameter estimates. Adopted models for codon frequencies were F3×4 and F61. Virtually identical results were obtained using both codon frequency models; thus, only data from the simplest one (F3×4) are reported. Amino acids under selection for model M8 were identified using a Bayes empirical Bayes approach (BEB) (Yang et al. 2005). BEB probabilities higher than 0.95 were considered significant. Two additional likelihood methods, incorporating variation in the rate of synonymous substitution, random effects likelihood (REL) and fixed effects likelihood (FEL; Kosakovsky Pond and Frost 2005) were implemented at the DATAMONKEY web server (Pond and Frost 2005). Finally, a codon model (MEC), which takes into account the physicochemical properties of amino acids when estimating non-synonymous substitution rates (Doron-Faigenboim and Pupko 2006), was implemented at the web server Selecton (Stern et al. 2007).
Comparative genomic analysis
Three proteins flanking the sea bass NITR gene cluster, RBM4 (RNA binding motif 4), TRAF2-binding protein (TNF receptor-associated factor 2) and FXR2 (fragile X mental retardation gene 2), were used to investigate the syntenic relationship between D. labrax and the other teleost species. Putative orthologs of these sea bass genes were found and their location identified with TBLASTN searches against the Danio rerio, Gasterosteus aculeatus and Oryzias latipes genomes. Latest genome releases of the Ensembl web site (http://www.ensembl.org/index.html) were used to achieve this task.
Comparative modelling of NITR V domain
Theoretical models of the Dl28 family V domains were generated with Modeller 9v4 (http://www.salilab.org/modeller/), with symmetry constraints applied to build symmetry-related domain dimers. The comparative modelling approach implemented in Modeller was based on the satisfaction of spatial restraints obtained from the alignment of the target sequence to one or more model structures (Sali and Blundell 1993). The recently solved Ictalurus punctatus NITR10 and NITR11 (Cannon et al. 2008) were selected as templates according to the highest sequence similarity. Firstly, sea bass sequences were used to perform a FASTA search on the Protein Data Bank (http://www.rcsb.org/) for each NITR domain. Next, the ClustalW server (http://www.ebi.ac.uk/Tools/clustalw2/index.html) was used to align V domains for each target/template combination. The model-default routine of Modeller was then used to generate five protein models for each sea bass sequence. Best models were selected based on their stereochemical properties.
Expression of NITR mRNAs in D. labrax
NITR mRNA expression, in the eight tissues listed above, was examined by real-time polymerase chain reaction (RT-PCR) for four individual fishes. Total RNA extraction and RT-PCR were performed as described earlier. To examine expression of NITRs, ?-actin was amplified as positive control. Amplicons were then resolved by gel electrophoresis in 1.8% agarose containing SYBR Safe™ DNA gel stain ×10,000 (Invitrogen). Expression data (presence–absence) in individual tissue samples were transformed into binary data, and non-parametric Spearman rank correlation tests were carried out for all possible pairwise comparisons of NITR genes.
Sea bass NITRs identification and mapping
Shotgun sequencing of the two BAC clones yielded 2,021 reads for bassbac-18k1 (min Q20 length 300 bp, average length 649 bp) and 1,979 reads for bassbac-79e10 (min Q20 length 300 bp, average length 675 bp). Assembly of the data resulted in two contigs (136,804 and 10,268 bp with an average sequencing coverage of 8.9-fold) for bassbac-18k1 that could be ordered in a scaffold according to the BAC-ends sequences. For bassbac-79e10, the assembly yielded four contigs (3,184, 29,024, 100,505 and 4,881 bp with an average sequencing coverage of 9.7-fold) that were ordered into a scaffold by paired end information of plasmid reads and BAC end sequences. Both BACs share an overlapping region of ?13.5 kb and were manually aligned to build a single scaffold of 271,448 bp (including gaps). Average G+C content of the sequenced region was 40.5%.
Sea bass NITRs share conserved features
The NITR gene cluster is composed of multiple families
Phylogenetic analyses suggest NITRs conform to the “birth-and-death” model of evolution
NITRs experienced positive selection
Sea bass NITR amino acid sites under positive selection identified using different methods
Maximum likelihood method
Proportionally, a smaller number of amino acid positions are inferred to have experienced adaptive selection in the remaining three gene expansions. Following either a 50% or a 75% MC rule, 29 and 4 positively selected positions are found in the sea bass (ESM, Supplementary file 5), 22 and 3 in the Southern pufferfish and six and two in medaka (ESM, Supplementary file 6), respectively. The most supported sites are observed in CDRs as well as other protein domains (cyt2, I).
Sea bass NITRs share conserved synteny
Expression analysis of sea bass NITRs
A preliminary survey on sea bass NITR gene expression was carried out using end-point RT-PCR on tissue samples taken from four different fish. Little or no expression was observed for Dl2, Dl4 and Dl16, whilst variation across individuals and tissues were noticed for the other NITR genes. Not surprisingly, NITRs are primarily detected in tissues that play a relevant role in immune responses. All NITRs are expressed in the head kidney and, with a few exceptions, in gills, gut and spleen. Weak (and variable across individuals) expression was observed in whole blood, liver and skeletal muscle (Table S1 in ESM, Supplementary file 7).
Non-parametric correlation tests on presence–absence data (Table S2 in ESM, Supplementary file 7) shows that for several gene pairs, mRNA expression is positively correlated, suggesting that in most cases, transcriptional regulation is partially conserved after duplication. However, correlation of expression patterns does not reflect well the phylogenetic/evolutionary relatedness of different NITR genes. The group of NITR genes Dl10–14 provides an interesting example on this issue. Based on the sea bass NITR phylogenetic tree (Fig. 4) and their genomic organisation (Fig. 1), the genes appear to have evolved from a single ancestral copy first through a gene duplication, giving rise to the ancestor of Dl10/Dl11and Dl12/13/14, respectively. A second round of duplication gave rise to Dl10 and Dl11, as well as to Dl14 and Dl12/Dl13. A final duplication event resulted in the emergence of Dl12 and Dl13. Comparison of NITR10 and NITR11 expression reveals 84% identity (27/32) across tissue/sample combinations, with a Spearman rho equal to 0.74 (p?<?0.00001), in agreement with the small phylogenetic distance between these two genes. On the other hand, between NITR12 and NITR13, only 56% identity (Spearman rho 0.24, p?=?0.17) is observed despite the two sequences being closely related. In this case, NITR13 gene expression seems to be under different selective constraints because the comparison between NITR12 and NITR14 shows a better overlap (68%) and a higher, marginally significant correlation (Spearman rho?=?0.33, p?=?0.06).
NITR genes have been proposed to share functional features with mammalian natural killer receptors (NKRs), which can function to either inhibit or activate NK cell cytotoxicity and/or cytokine release (Yoder 2009). In common with other immune genes, both NKRs and NITRs appear to follow a “birth-and-death” model of evolution (Nei and Rooney 2005). This model assumes that, when two species are compared for a specific multi-gene family, most genes originate from a set of paralogous sequences present in the common ancestor through tandem duplications (“birth” process) after the divergence of the two lineages. On the other hand, several copies are lost through pseudogenization–deletion (“death” process; ESM, Supplementary file 8). In contrast to what happens in concerted evolution, a variable degree of sequence divergence is expected among family members as well as a certain proportion of pseudogenes depending on the selective forces acting on the specific gene family (Nei and Rooney 2005). In agreement with model predictions, three pseudogenes (10% of all isolated genes) are found in the sea bass, similar to what has been observed in other fish species, and three main gene groups are found, each one with a distinctive domain composition and genomic organisation (Fig. 4). Sequence identity >70% was considered a suitable criterion to define a gene family and allowed sea bass NITRs to be further divided into a large number of families (18), as already described for zebrafish, catfish and medaka NITRs (Yoder et al. 2001, 2004; Hawke et al. 2001; Desai et al. 2008). Diversification of sea bass NITR genes appear to have occurred through repeated tandem duplications, since most genes are organised in smaller, co-linear genomic subclusters which largely reflect their phylogenetic position (Figs. 1 and 4). The observed association between phylogenetic and physical clustering of members of different families suggests unequal crossing over as the major mechanism for their amplification. For instance, the sea bass Dl8 family includes three different genes (Dl8, Dl9 and Dl16) which are likely derived from a recent duplication event. Their encoded proteins are >85% identical, in particular Dl9 and Dl16 encode proteins with 95% identity, with intronic sequences showing >90% identity. Similar evidence comes from the Dl28 family (Dl28–30) where exon sequences are 85–91% identical, whilst all introns but one could be aligned.
When set in a larger evolutionary perspective, the different groups (I–III) of sea bass NITRs again appear to conform to the “birth-and-death” model predictions. The origin of group I can be traced back to a minimal set of two genes in the common ancestor of acanthopterygian fish (medaka, pufferfish, sea bass). One of these two genes has been retained in single copy in T. nigroviridis, S. nephelus and D. labrax (grouped under node L in Fig. 5), whereas it has been lost in O. latipes. The other ancestral gene underwent lineage-specific, independent expansions (node N in Fig. 5 and, in details, ESM, Supplementary file 3). Group II is the least numerous in the sea bass, and it appears that homologous genes in other teleosts (node H in Fig. 5) have experienced a limited number of duplication events as well. Group III is characterised by lack of the I domain-encoding exon, which is replaced by a novel exon in four genes (Dl15, Dl28–30). Whilst V-only NITRs have been reported in several other species (channel catfish, zebrafish, medaka), the acquisition of an additional exon has been observed so far only in I. punctatus NITR5-11. As discussed earlier, this acquisition is likely to have occurred independently three times. Two mechanisms have been proposed for the evolution of novel exons. The first one, “exonization” of an intronic region, requires the evolution of a splice site surrounding the recruited genomic sequence (Sorek 2007). The second one, “exon shuffling”, promotes the gain of an exon through crossing over between non-homologous genes, causing the insertion of an exon from a “donor” gene into a preexisting intron of a recipient gene (Patthy 2008). Under the exonization model, a sequence with some similarity with the novel exon should be retained in the corresponding introns of related genes that did not acquire an additional exonic sequence. However, genomic alignment of intron 2 of Dl10–14 shows no such evidence. Exon shuffling is therefore the most likely mechanism for novel exon acquisition, as already reported for KIR genes (Rajalingam et al. 2004).
The function of the peptide encoded in the extra exon is not clear at present. No functional domain or motif has been identified. Homology modelling (Fig. 3) maps it to the “stalk” region reported for the channel catfish NITR11 (Hawke et al. 2001; Cannon et al. 2008), and further analysis using TMHMM v2.0 confirms its extracellular localisation, suggesting it might encode a connecting peptide. Connecting peptides are commonly found in other immunoglobulin-like molecules such as NILTs in teleosts (Stet et al. 2005; Østergaard et al. 2008) as well as mammalian KIRs and ILTs (Torkar et al. 1998; Martin et al. 2000). In NILT genes, it has been suggested that this region delimits a potentially accessible pocket playing a crucial role for ligand binding. Whether the repeated acquisition of the stalk-encoding exon has any functional meaning in fish NITRs remains to be further elucidated.
Convergent evolution towards similar genomic organisations (number of exons, intron–exon structure) seems to be a more general feature of fish NITR genes. At present, it is not clear which was the original organisation in the ancestral NITR. However, if it is postulated that a structure with six exons and two Ig domains was the ancestral state for the genomic structure of NITR genes, then loss of I, TM and cytoplasmic domains, as well as of the intron spacing V-encoding and I-encoding exons appears to have occurred several times independently. Modular organisation and evolution through domain shuffling provide additional evidence for the similarity between NITRs and NKRs, particularly KIRs where two to three Ig (C) domains are alternatively present and long or short forms, either including or excluding a long cytoplasmic tail, are found. Both activating and inhibitory alleles of KIR 3DL1/S1 locus have been found to be under positive, directional selection in some human populations based on a codon model ML analysis (Norman et al. 2007). Selected amino acid positions were located in the two Ig domains (D1 and D2) that bind MHC-I proteins as well as on the third Ig domain (D0), which is supposed to enhance binding capacity. In the present study, using a similar approach, four clusters of NITR genes, in three different species, were demonstrated to have experienced positive, divergent selection. This is the first evidence that the extraordinary diversity of NITR genes reported in several fish lineages might be, at least in part, the consequence of adaptive selection favouring the evolution of a highly diverse array of molecules. Maximum likelihood methods for detecting positive selection have some limitations (Nozawa et al. 2009), yet they represent the most rigorous approach available to date to assess which evolutionary forces (negative–positive–(nearly) neutral) have been acting on a group of sequences (Anisimova and Kosiol 2009). Lineage-specific gene expansions have already been suggested to be a peculiar, recurrent feature of fish NITRs (Yoder 2009). Evidence presented here further supports the important role of this phenomenon, indicating that repeated gene duplications within NITR families likely provide the raw material for positive, diversifying selection to work on. This observation is in perfect agreement with genome-wide surveys conducted to assess the role of positive selection in gene evolution, which have already pointed out that positively selected genes are significantly enriched for roles in defence/immunity (e.g. Bustamante et al. 2005; Kosiol et al. 2008). Whilst these studies have generally excluded recent gene families, Han et al. (2009) have reported that recently duplicated genes show more frequently evidence of positive selection compared to single-copy genes.
The application of probabilistic codon models can also predict which sites are most likely under positive selection in a specific protein (Anisimova and Kosiol 2009). Although not conclusive and requiring further experimental studies, the consensus among different methods suggests that positively selected codons are often located in putative ligand-binding regions. Whilst ligands for NITRs remain to be identified, a recent study (Cannon et al. 2008) clearly demonstrated that a single amino acid substitution in CDR1 of channel catfish NITRs is sufficient to acquire/abrogate recognition of an allogeneic B cell target. Also, the close similarity in protein structure between NITR and TCR/Ig-LC suggests that CDR1-3 might have a similar role of ligand recognition in NITRs. Thus, it is not unexpected to find positively selected sites in these regions which might be under diversifying selection as already shown for recombining receptors of adaptive immunity. Evidence of positive Darwinian selection in the extracellular V domain of NITR genes also fits well with the hypothesis that fish NITRs are functionally homologous to mammalian KIRs, which in primates mediate self-recognition of NK cells under the missing self model. Inhibitory, self-recognising KIRs are supposed to be under diversifying selection to keep pace with the enormous diversity of their ligands, encoded by hundreds of different alleles at the MHC-I loci. Although evolutionary strategies are different across teleosts, ranging from a single classical MHC-I locus with highly polymorphic alleles in the rainbow trout (Phillips et al. 2003) to tens–hundreds of loci in the Nile tilapia (Sato et al. 2006) and the European sea bass (L. Bargelloni, unpublished observations), fish show high intraspecific diversity of MHC-I proteins as observed in mammals. Finally, it should be noticed that codon models likely underestimate coding sequence variation between NITR sequences since insertions and/or deletions cannot be incorporated in such models. For instance, sea bass NITR28 shows a four amino acid insertion at the end of CDR2 where the most significant positively selected site was also found (Table 1), with a visible effect on predicted protein structure (Fig. 3).
In addition to the putative ligand recognition region, there is evidence that a few sites in the cytoplasmic portion of the protein might be positively selected. The regulatory role of the cytoplasmic tail appears to be conserved compared to NKRs (Yoder et al. 2001, 2007; Wei et al. 2007). Activating or inhibitory signals are transduced through binding of intracellular partner proteins (e.g. DAP12, SHPs). Thus, it might be possible that amino acid changes in the cytoplasmic tail modify the specificity and/or affinity of these protein–protein interactions, modulating signal transduction between different NITRs.
Comparative genomics of fish NITRs yields additional evidence in favour of a “birth-and-death” model of evolution. Two major NITR clusters were found in zebrafish on chromosomes 7 and 14 (Yoder et al. 2001, 2008), medaka chromosomes 18 and 10 (Desai et al. 2008) and stickleback groups VII and IV (Desai et al. 2008; this study), which are onhologs (chromosome pairs deriving from a single ancestral chromosome after a whole genome duplication; Nakatani et al. 2007; Negrisolo et al., submitted). As already pointed out (Yoder 2009), this evidence suggests the presence of a substantial set of NITR genes in the common ancestor of all teleost fish before the occurrence of the teleost-specific whole genome duplication event (Nakatani et al. 2007). Once duplicated, the original cluster was retained in two copies, although individual NITR genes have been alternatively lost, retained or expanded mostly independently in each fish lineage examined so far. For instance, a closer examination of NITRs grouped under node H (Fig. 5) reveals that sea bass group II are more closely related to zebrafish NITR 6a,b and 8 (chr. 7) than with medaka Ol 6a,b,c and Ol18-22 (chr. 22) despite the fact that medaka and seabass are phylogenetically closer than sea bass with zebrafish. These zebrafish and sea bass genes, however, are located on orthologous clusters, whereas the medaka ones are positioned on the corresponding onholog. This indicates that orthologs of sea bass group II genes have been retained in zebrafish, whilst only their paralogs are found in medaka. All these evidence, together with the shared synteny found between sea bass NITR cluster and stickleback/medaka/zebrafish chromosomes IV/10/14 (Fig. 6), suggest that in sea bass as well might exist at least one more NITR cluster not yet identified.
In agreement with observations in other species (Piyaviriyakul et al. 2007; Yoder et al. 2002), NITR genes are predominantly expressed in tissues playing a relevant role in the immunological response, whilst gene-specific variation was noticed in other tissues. Weak or null expression observed on blood samples might be partially explained by the fact that whole blood was used rather than pheripheral blood leucocytes due to the limited amount of available biological material.
The results of expression analysis presented here are preliminary and await further quantitative studies with qRT-PCR, possibly on cell line-specific RNAs rather than whole tissue samples. Nevertheless, moderate correlation between gene pairs in presence–absence expression data suggests that in most cases, transcriptional regulation is partially conserved after duplication. However, expression pattern does not reflect well the phylogenetic/evolutionary relatedness of different NITR genes, thereby revealing differential selective constraints on gene expression and providing further room for evolutionary diversification of NITR genes.
In summary, 27 NITR genes and three pseudogenes were identified and analysed in the European sea bass, providing significant insights into the evolution and function of NITR genes in bony fish. Based on the results reported here, in addition to those collected so far in other species, the following conclusions can be drawn: (1) starting from two gene clusters in the common ancestor of all teleosts, NITRs have evolved following a “birth-and-death” model of evolution, with lineage-specific gain and loss of individual family members, (2) recurrent events of domain shuffling and exon/intron acquisition/deletion contributed to the remarkable diversity of domain and genomic organisation of fish NITRs, (3) repeated duplications have generated largely species-specific gene expansions, which have been demonstrated to be under positive Darwinian selection in at least four independent cases, (5) directional selection seems to significantly affect putative ligand binding regions, confirming the predicted function of NITR proteins and (5) differential expression might be a further dimension of NITR diversity.
We thank Fabio Borghesan for providing sea bass individuals. We thank also Enrico Negrisolo for very helpful discussion and advice on phylogenetic analyses. This work was partially supported by EU project Aquagenome–genomics in fish and shellfish: From research to aquaculture.