Transspecies dimorphic allelic lineages of the proteasome subunit β-type 8 gene (PSMB8) in the teleost genus Oryzias

Abstract

The proteasome subunit β-type 8 (PSMB8) gene in the jawed vertebrate MHC genomic region encodes a catalytic subunit of the immunoproteasome involved in the generation of peptides to be presented by the MHC class I molecules. A teleost, the medaka (Oryzias latipes), has highly diverged dimorphic allelic lineages of the PSMB8 gene with only about 80% amino acid identity, termed “PSMB8d” and “PSMB8N,” which have been retained by most wild populations analyzed. To elucidate the evolutionary origin of these two allelic lineages, seven species of the genus Oryzias were analyzed for their PSMB8 allelic sequences using a large number of individuals from wild populations. All the PSMB8 alleles of these species were classified into one of these two allelic lineages based on their nucleotide sequences of exons and introns, indicating that the Oryzias PSMB8 gene has a truly dichotomous allelic lineage. Retention of both allelic lineages was confirmed except for one species. The PSMB8d lineage showed a higher frequency than the PSMB8N lineage in all seven species. The two allelic lineages showed curious substitutions at the 31st and 53rd residues of the mature peptide, probably involved in formation of the S1 pocket, suggesting that these allelic lineages show a functional difference in cleavage specificity. These results indicate that the PSMB8 dimorphism was established before speciation within the genus Oryzias and has been maintained for more than 30–60 million years under a strict and asymmetric balancing selection through several speciation events.

Transspecies polymorphism (TSP) is the passage of allelic lineages from ancestral to descendent species (1), and the underlying selective mechanism is referred to as balancing selection caused by “overdominant selection,” “frequency-dependent selection,” or “selection that varies in time and space” (2). The vertebrate MHC class I and II genes with a large number of alleles and wide allelic differences provide a classical example of TSP (3, 4), most probably arising by overdominant selection (5). TSP of the MHC genes sometimes persists for very long periods. Certain HLA-DRB alleles are reported to have persisted for 50–60 million years (6). Another well-characterized TSP has been reported for the rabbit Ig heavy-chain variable-region genes; its persistence has been estimated at 50 million years based on molecular phylogenetic analysis (7, 8). Except for these genes encoding molecules showing high degree of binding specificity, information for long-term balancing selection resulting in TSP is limited (9–11). Moreover, all TSP reported thus far involve only a few species, and the possible presence of a long-lasting TSP surviving through many speciation events is still to be clarified by systematic phylogenetic analysis.

The MHC class I molecules deliver the peptides derived from cytosolic proteins to the cell surface for recognition by cytotoxic T cells. The proteins are degraded proteolytically into short peptides by proteasomes whose catalytic core, or 20S proteasome, is a large complex composed of four stacks of two outer α-rings and two inner β-rings containing seven α and seven β subunits, respectively (12, 13). Of these subunits, only three β subunits—PSMB5, PSMB6, and PSMB7—have proteolytic activity with chymotrypsin-like, caspase-like, and trypsin-like specificity, respectively (14). Immunoproteasomes are formed by replacing PSMB5, PSMB6, and PSMB7 with the IFN-γ–inducible β subunits, PSMB8, PSMB9, and PSMB10, respectively (15, 16). These subunit substitutions enhance the chymotrypsin-like activity of the immunoproteasomes, which are responsible for generation of peptides with a hydrophobic residue at the C terminus suitable for binding to MHC class I molecules. In particular, PSMB8, with its chymotrypsin-like activity, is a critical component for supplying MHC class I-binding peptides, because PSMB8-knockout mice show reduced expression of MHC class I molecules on the cell surface (17).

The Japanese population of medaka, Oryzias latipes, is divided into two subpopulations, the Northern Population (NP) and the Southern Population (SP), which diverged 5–18 Mya (18, 19). Using BAC clones, we determined the complete nucleotide sequences of the approximately 400-kb MHC class I region of two inbred strains, HNI (derived from the NP) and Hd-rR (derived from the SP) (20, 21). Although the order and transcriptional orientation of the 22 identified genes were conserved perfectly between these two strains, the nucleotide sequences showed an extremely high degree of divergence at the ~100 kb segment in the middle of the medaka MHC class I region harboring the two MHC class IA genes, Orla-UAA and -UBA, and the two immunoproteasome β subunit genes, PSMB8 and PSMB10. The sequence divergence was especially conspicuous with the PSMB8 gene, which showed only 81.5% deduced amino acid sequence identity between these two inbred strains. The average nucleotide divergence between these two inbred strains was estimated to be 3.4% based on the whole-genome sequences (22), and the two sides of this highly diverged segment of the MHC class I region showed similar levels of nucleotide divergence from this average (21, 23). Extensive analysis of allelic polymorphism of the PSMB8 gene using wild individuals from nine localities representing both the NP and SP clarified that all PSMB8 alleles were clearly classified into the dichotomous allelic lineages, PSMB8N or PSMB8d (24). Both PSMB8d and PSMB8N lineages were retained in NP and SP although the allelic frequency of the PSMB8d lineage (73–100%) was much higher than that of the PSMB8N lineage (0–27%) in all the populations we analyzed. These findings suggest that the dichotomous allelic lineages were established before the divergence between NP and SP and have been maintained in each group for 5–18 million years (18, 19).

The genus Oryzias contains about 20 species distributed in southeastern Asia. These species most likely were generated by allopatric speciation, and the few hybridization experiments among them performed thus far showed the presence of reproductive isolation resulting in abortive embryonic development in the case of O. latipes x O. javanicus (25) or sterile F1 males in the case of O. latipes x O. celebensis (26) and O. latipes x O. curvinotus (27). The Oryzias species are divided into three species groups, the latipes, javanicus, and celebensis groups, based on mitochondrial DNA sequences (18, 19), corresponding to the biarmed, monoarmed, and fused chromosome groups of karyological grouping. The divergence time among these three species groups was estimated as 29–32 Mya, based on the molecular clock assumption using the mitochondrial 12S and 16S rRNA sequences (18), or as 58–65 Mya, based on a Bayesian relaxed molecular clock analysis of whole-mitogenome sequences (19). On the other hand, the formation of the Makassar Strait, the possible vicariant event between the celebensis group and the other two groups, occurred in the Eocene (34–56 Mya) (28). All these results suggest that the speciation within genus Oryzias started 30–60 Mya.

In this study, to elucidate the origin and evolution of these two allelic lineages of the PSMB8 gene found in O. latipes, we analyzed the genetic polymorphism of PSMB8 in this genus using wild populations of seven species from the latipes (O. curvinotus), javanicus (O. javanicus, O. minutillus, and O. dancena), and celebensis (O. celebensis, O. marmoratus, and O. matanensis) groups.

Results

Identification of the PSMB8 Alleles in Wild Populations of Seven Oryzias Species.

To investigate the allelic diversity of the PSMB8 gene in wild populations, seven Oryzias species [O. minutillus (n = 244 individuals), O. javanicus (n = 178), O. curvinotus (n = 69), O. celebensis (n = 190), O. matanensis (n = 106), O. marmoratus (n = 106), and O. dancena (n = 150)] were analyzed (Fig. 1). First, the PSMB8 gene fragment from exons 2–3 was amplified by genomic PCR. Each individual gave either single or double bands upon agarose gel electrophoresis, probably representing the homozygous or heterozygous state, respectively. Two to eight bands with distinct sizes were detected in each species. All bands amplified from all individuals were sequenced directly, and the lineage was determined based on the 31st amino acid residue of the mature peptide encoded by exon 3 (24). One homozygous individual for each distinctive band was selected for another round of PCR amplification from exon 1 to exon 6 (the last exon), and nucleotide sequences were determined by direct sequencing. The entire coding sequences for the mature peptide were elucidated by determining the 3′-most 69 bp not covered by this PCR amplification using 3′ RACE.

Fig. 1.

Map showing collection sites of wild populations of Oryzias species and allelic frequencies of the PSMB8d and PSMB8N lineages in wild populations. Dots indicate the collection site of each wild population of Oryzias species. Allelic frequencies of the PSMB8d and PSMB8N lineages in each wild population are shown in black and white, respectively, in the circle diagrams. Actual frequencies of the d allele (number of d alleles/number of total alleles) for each species are O. curvinotus, 0.862 (119/138); O. celebensis, 0.600 (228/380); O. matanensis, 0.764 (162/212); O. marmoratus, 0.816 (173/212); O. dancena, 1.00 (300/300); O. minutillus, 1.00 (488/488); and O. javanicus, 0.778 (277/356).

The deduced complete amino acid sequences of the mature peptide of the PSMB8d and PSMB8N allelic lineages identified from these eight Oryzias species, including O. latipes, were aligned, together with dimorphic PSMB8 alleles of Xenopus (29), two paralogous PSMB8 genes of sharks (30, 31), and human PSMB8 (Fig. S1). These sequences were aligned perfectly without any insertion/deletion except for a few residues at the C terminus. Fig. 2 shows amino acid substitutions found among the PSMB8 alleles of eight Oryzias species in the mature peptide region. Based on the 31st amino acid residue of the mature peptide and degree of amino acid sequence identity to the HNI (PSMB8N) and Hd-rR (PSMB8d) alleles of O. latipes (Fig. 2), these Oryzias PSMB8 alleles clearly were classified into either the PSMB8d or the PSMB8N lineages. Both lineages were identified from wild populations of O. javanicus, O. curvinotus, O. celebensis, O. matanensis, and O. marmoratus. In contrast, only the PSMB8d lineage was detected from the wild populations of O. minutillus and O. dancena. However, O. dancena has the PSMB8N lineage at the species level, because a PSMB8N lineage gene coding for tyrosine at the 31st position (Tyr31) has been isolated from a BAC library constructed using individuals from the closed colony kept at the University of Tokyo (32). This O. dancena PSMB8N sequence was added to Fig. 2 and Fig. S1 and was used in the following analyses. There are two sublineages, PSMB8d(V) and PSMB8d(A), in thePSMB8d lineage of medaka, having valine and alanine, respectively, at the 31st amino acid position, respectively (24). These two sublineages were present in all the wild populations of Oryzias species we analyzed. Names for the PSMB8 alleles of the Oryzias species were designated as follows: the species name; lineage name d or N; sublineage name V or A in parenthesis in the case of the d lineage, and the number if there are multiple alleles in that lineage or sublineage of each species. In the coding region, the PSMB8d(V) alleles of these species showed 92.7–98.5% nucleotide identity and 96.6–99.6% amino acid identity to the Hd-rR allele [O. latid(V) in Fig. 2 and Fig. S1], and the PSMB8d(A) alleles of these species showed 93.0–95.6% nucleotide identity and 94.0–97.5% amino acid identity to the Hd-rR allele. The PSMB8N alleles of these species showed 90.1–94.8% nucleotide identity and 93.6–96.1% amino acid identity to the HNI allele (O. latiN in Fig. 2 and Fig. S1). Because the nucleotide and amino acid identities between the coding regions of the Hd-rR and HNI PSMB8 genes were only 80.3% and 81.5%, respectively (21), there was no ambiguity in allocating the analyzed sequences into the PSMB8d or PSMB8N lineage.

Fig. 2.

Comparison of the amino acid sequences of the mature peptides of PSMB8. PSMB8 mature peptides of Oryzias species and two allelic lineages of O. latipes (NCBI accession nos. AB183488 and BA000027) were aligned with ClustalX 2.0. All Oryzias PSMB8 sequences were determined in this study, except for two O. latipes PSMB8N and PSMB8d(V) sequences and the O. dancena PSMB8N sequence (NCBI accession no. FJ481084). Of the 204 positions of the mature peptides, only the 48 positions where amino acid substitutions were observed are shown. Dots indicate identity with the residues in the uppermost sequence.

Phylogenetic Analysis of the Oryzias PSMB8 Alleles.

To clarify the phylogenetic relationship of the PSMB8 alleles from these Oryzias species, phylogenetic analysis was performed using the nucleotide sequences of the mature peptide region. When the phylogenetic trees were constructed using the neighbor-joining (NJ) method (33) with the PSMB8 sequences of spotted green pufferfish, human, and mouse as an outgroup, Oryzias PSMB8 alleles were divided into the PSMB8d and PSMB8N clades supported by 100% bootstrap percentage (Fig. 3). Both clades contained the PSMB8 alleles from the latipes, javanicus, and celebensis groups, indicating that the divergence between these two PSMB8 lineages occurred before separation of the Oryzias species. The PSMB8d(V) and PSMB8d(A) sublineages did not form their respective subclades and intermingled with each other. On the other hand, the PSMB8N clade was divided into two subclades supported by high bootstrap percentages of 96% and 93%. However, one subclade contained only the javanicus and celebensis groups, and the other contained only the latipes and celebensis groups. When the amino acid sequences of the mature peptides were used to construct a NJ tree, the dichotomous PSMB8d and PSMB8N clades were reproduced with high bootstrap percentages of 100% and 99%, respectively (Fig. S2). However, the topology within the PSMB8d clade showed a significant divergence from that of the NJ tree based on nucleotide sequences. It is especially noteworthy that the PSMB8d(A) sublineage formed a clade, although with a low bootstrap percentage of 65%.

Fig. 3.

Phylogenetic tree of Oryzias PSMB8 alleles. The nucleotide sequences of mature peptides of 612 residues were aligned by ClustalX 2.0(37), and the phylogenetic tree was constructed by the NJ method (33). The numbers on each branch represent bootstrap probabilities (>50%) based on 1,000 bootstrap trials. The sequences of TeniPSMB8 (Tetraodon nigroviridis, NCBI accession no. CR697191), HosaPSMB8 (Homo sapiens, NCBI CR541661), and MumuPSMB8 (Mus musculus, NCBI BC013785) were used as an outgroup.

Polymorphism of Deduced Amino Acid Sequences Between Two PSMB8 Lineages.

The Oryzias PSMB8d and PSMB8N lineages were clearly discriminated by 10 diagnostic substitutions: R⁹K, Q⁵³K, C¹³⁰S, V¹³⁹L, C¹⁵⁷R, A¹⁶⁸S, S¹⁷³V, M¹⁷⁹I, E¹⁹⁰D, and R²⁰⁰K (Fig. 2). In addition, these two lineages show substitutions in five other positions, although two different amino acids are found in one of the lineages: N²⁹C/G, D/N³⁰E, V/A³¹Y, R¹⁸⁰Q/K, and E²⁰¹Q/K (Fig. 2). Among these diagnostic substitutions, V/A³¹Y and Q⁵³K are especially interesting because these residues are likely to be involved in the formation of the S1 pocket that determines cleaving specificity (13). The other residues involved in S1 pocket shaping—A²⁰, I³⁵, M⁴⁵, and A⁴⁹—were conserved between these two lineages. Xenopus dimorphic alleles and two shark paralogous copies of the PSMB8 gene also show a similar substitution at the 31st position (Fig. S1) (29, 30). On the other hand, it is not clear whether there is functional diversification between PSMB8d(V) and PSMB8d(A), because sublineage-specific substitutions were recognized only at two positions, V³¹A and R⁶⁷K (Fig. 2). The valine/alanine substitution at the 31st residue is a relatively conservative one, and no direct functional importance can be assigned to the 67th position. The two sublineages of the PSMB8N lineage show seven sublineage-specific substitutions, A²²T, N²⁴S, G²⁹C, N¹³³S, V¹⁶¹A, M¹⁷⁶L, and L¹⁹⁶I (Fig.2). However, the functional significance of these substitutions, if any, is still to be clarified.

Comparison of Nucleotide Sequences of Exons and Introns Among the PSMB8 Alleles.

The sizes of exons 2–6 were conserved completely in all the analyzed PSMB8 genes of these eight Oryzias species. In contrast, intronic sequences showed interspecific as well as intraspecific variation in length. The nucleotide sequences of these PSMB8 alleles were compared with those of the PSMB8d(V) and PSMB8N of O. latipes by dot-plot analysis (Fig. S3). One PSMB8d(V) allele, one PSMB8d(A) allele, and one PSMB8N allele were selected from O. curvinotus, representing the latipes group, from O. celebensis, representing the celebensis group, and from O. javanicus, representing the javanicus group. The only exception was the PSMB8N of O. javanicus, for which PCR amplification was unsuccessful, probably because of the presence of long intron(s). As shown in Fig. S3, diagonal lines were detected in the dot plots for all exons in any pair of comparisons. In contrast, diagonal lines were detected only in certain pairs of comparisons for the intronic regions. In the comparison with PSMB8N of O. latipes, only the PSMB8N alleles of O. curvinotus and O. celebensis showed diagonal lines in these regions. On the other hand, in the comparison with the PSMB8d(V) of O. latipes, only the PSMB8d(V) alleles showed clear diagonal lines in intronic regions, and the PSMB8d(A) alleles showed only short intronic lines. These results indicate that the intronic sequences of the PSMB8d and PSMB8N lineages have almost no sequence similarity and that there has been considerable intronic sequence diversification even between the PSMB8d(V) and PSMB8d(A) sublineages. On the other hand, intronic sequences have been more or less conserved within the PSMB8N and PSMB8d lineages, excluding the remote possibility that the observed sequence similarity in the protein-coding region among the PSMB8 sequences of various species belonging to the same lineage has been caused by convergent evolution rather than by common ancestry.

Frequency of Two PSMB8 Allelic Lineages in Wild Oryzias Populations.

The allelic frequencies of the PSMB8d and PSMB8N lineages in wild populations of each species are shown by circle diagrams in Fig. 1. The allelic frequency of the PSMB8d lineage (black) was higher than that of the PSMB8N lineage (white) in all analyzed species, ranging from 0.6 to 1.0. The results of the genotyping of each individual of these Oryzias species are summarized in Table S1. The observed numbers do not deviate significantly from the values expected from allelic frequency. The PSMB8N lineage was not identified in wild populations of O. minutillus and O. dancena. However, because the frequency of the PSMB8N lineage was less than 0.01 in certain populations of O. latipes (24), it still is unclear whether the PSMB8N lineage is missing or is present at an extremely low frequency in these populations.

Discussion

We found that dichotomous allelic lineages of the single-copy PSMB8 gene (20, 21), PSMB8d and PSMB8N, are present in most of the Oryzias species we analyzed. Lineage-specific sequences were identified not only in the protein-coding regions (Fig. 2) but also in introns (Fig. S3), indicating that these lineages reflect real lineages established by ancient diversification rather than false lineages formed by convergent evolution. Because these two allelic lineages were found among Oryzias species belonging to all three species groups—O. latipes and O. curvinotus of the latipes group, O. dancena and O. javanicus of the javanicus group, and O. celebensis, O. marmoratus, and O. matanensis of the celebensis group—they should have been established before the start of speciation in the genus Oryzias. Thus, it is likely that these dimorphic lineages of the Oryzias PSMB8 gene survived through the Oryzias speciation process and were transferred from species to species.

Long-term balancing selection maintaining multiple alleles through multiple speciation events has been reported with the MHC and other host-defense genes in vertebrates (3, 4, 7–10). The longest persistence time estimated thus far based on molecular data is 50–60 million years for the human HLA-DRB and rabbit Ig heavy-chain variable-region genes (6–8). The phylogenetic analysis of the Oryzias PSMB8 genes (Fig. 3) indicated that the PSMB8d and PSMB8N lineages diverged long before the speciation of the Oryzias species, which is believed to have started 30–60 Mya (18, 19). To estimate roughly the time of divergence between these two PSMB8 lineages, we constructed a linearized tree (Fig. S4) based on the NJ tree of Fig. S2. To make the estimation of the persistence time comparable with that of the rabbit Ig heavy-chain variable-region genes (8), the divergence time of 100 Mya between human and mouse (34) was used as a standard. The rate of amino acid substitution for the PSMB8 gene was estimated to be 0.42 × 10⁻⁹ · site⁻¹ · y⁻¹, and the divergence time between the two allelic lineages of the PSMB8 gene was calculated to be 178 Mya, much more ancient than the origin of the rabbit Ig heavy-chain variable-region gene polymorphism. Based on this rate of amino acid substitution, the divergence time between spotted green pufferfish and Oryzias was calculated to be about 240 Mya. This value does not show great discrepancy with the recent estimation of 191 Mya based on mitochondria DNA sequences (35). This result suggests that age estimation based on this linearized tree is not unrealistic. However, the evolutionary rate of the PSMB8N lineage apparently is faster than that of the PSMB8d lineage (Fig.3 and Fig. S2), limiting the reliability of age estimation based on this linearized tree. Thus, we conclude that the two Oryzias PSMB8 lineages diverged well before the start of Oryzias speciation 30–60 Mya.

Like the Oryzias PSMB8 gene, the Xenopus PSMB8 gene showed dichotomous allelic lineages that were transferred from ancestral to descendant species for more than 80 million years (36). However, transspecies dimorphism of the Xenopus PSMB8 gene was inferred indirectly from Southern blotting analysis; our current analysis of the Oryzias PSMB8 genes provides conclusive evidence for transspecies dimorphism of PSMB8 based on the nucleotide and amino acid sequences. Interestingly, phylogenetic analysis has indicated that the origin of the PSMB8 dimorphism of Xenopus and Oryzias species was independent (21), suggesting that the PSMB8 dimorphism was not maintained throughout the evolution of jawed vertebrates. However, once established, the dimorphism was transferred from species to species for more than 30–60 or 80 million years. In addition to the dichotomous, highly diverged allelic lineages in Xenopus and Oryzias, two paralogous PSMB8 genes with a similar level of sequence diversity have been reported in sharks (30). Again, the phylogenetic tree analysis of these two types of the PSMB8 genes in these species did not show any orthologous relationship, indicating that they were generated by independent evolutionary events (21). However, the two diverged PSMB8 types of each species, whether alleles or paralogous genes, show curious similarities in amino acid substitution pattern at certain positions. Thus, the amino acid residues at the 31st position of the mature peptide are either alanine or phenylalanine in the shark and Xenopus and either valine/alanine or tryptophan in Oryzias. Because this position in bovine PSMB5 is involved in the S1 pocket formation (13) and is occupied by phenylalanine or tyrosine with bulky and neutral side chains or alanine or valine with smaller side chains, it is conceivable that the two types of the PSMB8 gene in these animals show similar differences in cleaving specificity. Because there is no orthologous relationship, the similar amino acid substitutions at the 31st position in the two types of PSMB8 gene of each animal lineage might have been formed by convergent evolution, suggesting the presence of a strong selective pressure to induce functional diversification of the PSMB8 gene within each species group.

In all the wild populations we analyzed here, PSMB8d was the major allelic lineage with a gene frequency of 0.6–1.0. This predominance also is true for wild populations of O. latipes in Japan and Korea, where the frequency of the PSMB8N lineage was extremely low (<0.01) in some populations (24). Thus, it still is not clear whether the PSMB8N lineage is absent from the wild populations of O. minutillus and O. dancena we analyzed or is present at an extremely low frequency. Recently, the nucleotide sequence of the MHC class I region of O. dancena was determined, and a BAC clone encompassing the PSMB8N lineage was identified (32), indicating that two allelic lineages were maintained in at least some wild populations of O. dancena. Conservation of this biased dimorphism of the PSMB8 gene in Oryzias species for at least 30–60 million years suggests the presence of asymmetric balancing selection. One possible explanation for this asymmetric balancing selection may be that the PSMB8d lineage is more efficient in processing most pathogen-derived proteins but that there are certain pathogen-derived proteins for which PSMB8N is more effective than PSMB8d. Overdominant selection with the differential relative fitness for PSMB8d and PSMB8N could explain this asymmetric balancing selection. However, the molecular basis for the differential relative fitness is still to be clarified through biochemical analysis addressed at different cleaving specificities of these two PSMB8 lineages.

Materials and Methods

Fishes.

The wild individuals of Oryzias species were 244 specimens of O. minutillus from Patum Thani and Chai Nat, Thailand; 178 specimens of O. javanicus from Singapore; 69 specimens of O. curvinotus from Hong Kong, China; 190 specimens of O. celebensis from Malino and Pattunnuang, Indonesia; 106 specimens of O. matanensis from Lake Matano, Indonesia; 106 specimens of O. marmoratus from Lake Towuti, Indonesia; and 150 specimens of O. dancena from Linggi, Malaysia. All specimens were fixed in100% ethanol after being collected from the field and were kept at 4 °C until DNA and/or RNA extraction.

Genomic DNA Extraction.

Genomic DNA was extracted from the caudal fin using the Puregene Genomic DNA Purification Kit (Gentra Systems) according to the manufacturer's instructions and was finally dissolved in 30–50 μL TE (10 mM Tris, 1 mM EDTA).

PCR Amplification and Sequencing of the Oryzias PSMB8 Alleles.

The PSMB8 gene was amplified using genomic DNA as template with two primer sets. First, the PSMB8 fragment from the second to third coding exons was amplified with a pair of primers that were designed on the conserved PSMB8 sequences among two allelic lineages of O. latipes (NCBI accession nos. AB183488 and BA000027), fugu (Takifugu ruburipes, NCBI accession no. CAC13117, and Tetraodon nigroviridis, NCBI accession no. CAG11683), and zebrafish (Danio rerio, NCBI accession no. BC066288). The forward primer (Oryzias PSMB8.E2F) was 5′-CATGGAGTCATHGTNGCNGTNGA-3′ at the second coding exon, and the reverse primer (Oryzias PSMB8.E3R) was 5′-AGTCTGCNGCRCTNCCNGACAT-3′ at the third coding exon of PSMB8. The PCR condition was denaturation at 98 °C for 30 s, 35 cycles of denaturation at 98 °C for 10 s, annealing and elongation at 56 °C for 2 min, and final elongation at 72 °C for 3 min with LA-Taq (Takara Bio Inc.). Sequencing reaction was performed with BigDye Terminator v3.1 Sequencing Standard kit (Applied Biosystems), and nucleotide sequences were determined by a 3100/3130xl Genetic Analyzer (Applied Biosystems).

Determination of the Full Coding Sequence of the PSMB8 Mature Peptide.

To elucidate the entire coding sequence of the mature peptide, a region from exon 1 to exon 6 (the last exon) was amplified by the previously reported primers and PCR conditions (24). The PCR products were sequenced directly using homozygous individuals or after cloning into pCR 2.1-TOPO vector or pCR-XL-TOPO vector (Invitrogen).

Because the 3′-most 69 bp of the coding sequence are not covered by this PCR amplification, 3′ RACE was performed according to the SMART RACE cDNA Amplification Kit protocol (Clontech Laboratories). Total RNA was isolated from 100% ethanol-fixed internal organs of fish using ISOGEN (Nippon Gene) according to the manufacture's instructions and was reverse transcribed into cDNA by SuperScript II (Invitrogen). The cDNA was used for PCR amplification of PSMB8N and PSMB8d alleles using the primers specific for each type.

Sequence Alignment and Phylogenetic Analysis.

The nucleotide sequences of PSMB8 mature peptides were aligned using ClustalX 2.0 (37). Based on the alignments, the phylogenetic trees were constructed using the NJ method (33) with the PSMB8 sequences of spotted green pufferfish, human, and mouse as an outgroup, and bootstrap possibilities were determined with 1,000 bootstrap replications. The evolutionary distances were computed using the Kimura two-parameter method (38). These evolutionary analyses were conducted in MEGA4 (39).

Sequence Comparison Among PSMB8 Alleles.

Nucleotide sequences of alleles were compared using the dot plot generated by PipMaker (http://bio.cse.psu.edu) (40). In the dot plot, gap-free segments with more than 50% identity between two sequences were plotted. Percent identities of protein-coding regions were computed using the homology search program of GENETYX-MAC version 11.1.0 (GENETYX Corp.).